1 SPDX-License-Identifier: GPL-2.0
2 aufs4.x-rcN kbuild patch
4 diff --git a/fs/Kconfig b/fs/Kconfig
5 index 7aee6d6..ec92031 100644
8 @@ -248,6 +248,7 @@ source "fs/pstore/Kconfig"
9 source "fs/sysv/Kconfig"
10 source "fs/ufs/Kconfig"
11 source "fs/exofs/Kconfig"
12 +source "fs/aufs/Kconfig"
14 endif # MISC_FILESYSTEMS
16 diff --git a/fs/Makefile b/fs/Makefile
17 index ef772f1..51779e68c 100644
20 @@ -129,3 +129,4 @@ obj-y += exofs/ # Multiple modules
21 obj-$(CONFIG_CEPH_FS) += ceph/
22 obj-$(CONFIG_PSTORE) += pstore/
23 obj-$(CONFIG_EFIVAR_FS) += efivarfs/
24 +obj-$(CONFIG_AUFS_FS) += aufs/
25 SPDX-License-Identifier: GPL-2.0
26 aufs4.x-rcN base patch
28 diff --git a/MAINTAINERS b/MAINTAINERS
29 index 82ad0ea..7d8b461 100644
32 @@ -2478,6 +2478,19 @@ F: include/linux/audit.h
33 F: include/uapi/linux/audit.h
36 +AUFS (advanced multi layered unification filesystem) FILESYSTEM
37 +M: "J. R. Okajima" <hooanon05g@gmail.com>
38 +L: linux-unionfs@vger.kernel.org
39 +L: aufs-users@lists.sourceforge.net (members only)
40 +W: http://aufs.sourceforge.net
41 +T: git://github.com/sfjro/aufs4-linux.git
43 +F: Documentation/filesystems/aufs/
44 +F: Documentation/ABI/testing/debugfs-aufs
45 +F: Documentation/ABI/testing/sysfs-aufs
47 +F: include/uapi/linux/aufs_type.h
49 AUXILIARY DISPLAY DRIVERS
50 M: Miguel Ojeda Sandonis <miguel.ojeda.sandonis@gmail.com>
51 W: http://miguelojeda.es/auxdisplay.htm
52 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
53 index bc8e615..e51a59d 100644
54 --- a/drivers/block/loop.c
55 +++ b/drivers/block/loop.c
56 @@ -691,6 +691,24 @@ static inline int is_loop_device(struct file *file)
57 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
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 5c7df1d..019f14b 100644
85 @@ -1197,7 +1197,7 @@ enum d_walk_ret {
87 * The @enter() and @finish() 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 *),
92 void (*finish)(void *))
94 diff --git a/fs/fcntl.c b/fs/fcntl.c
95 index 0522e28..74c255d 100644
100 #define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
102 -static int setfl(int fd, struct file * filp, unsigned long arg)
103 +int setfl(int fd, struct file * filp, unsigned long arg)
105 struct inode * inode = file_inode(filp);
107 @@ -63,6 +63,8 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
109 if (filp->f_op->check_flags)
110 error = filp->f_op->check_flags(arg);
111 + if (!error && filp->f_op->setfl)
112 + error = filp->f_op->setfl(filp, arg);
116 diff --git a/fs/inode.c b/fs/inode.c
117 index 03102d6..517883c 100644
120 @@ -1655,7 +1655,7 @@ EXPORT_SYMBOL(generic_update_time);
121 * This does the actual work of updating an inodes time or version. Must have
122 * had called mnt_want_write() before calling this.
124 -static int update_time(struct inode *inode, struct timespec *time, int flags)
125 +int update_time(struct inode *inode, struct timespec *time, int flags)
127 int (*update_time)(struct inode *, struct timespec *, int);
129 diff --git a/fs/namespace.c b/fs/namespace.c
130 index e158ec6..312bdbd8 100644
133 @@ -846,6 +846,12 @@ static inline int check_mnt(struct mount *mnt)
134 return mnt->mnt_ns == current->nsproxy->mnt_ns;
137 +/* for aufs, CONFIG_AUFS_BR_FUSE */
138 +int is_current_mnt_ns(struct vfsmount *mnt)
140 + return check_mnt(real_mount(mnt));
144 * vfsmount lock must be held for write
146 diff --git a/fs/read_write.c b/fs/read_write.c
147 index f8547b8..0a5c47b 100644
148 --- a/fs/read_write.c
149 +++ b/fs/read_write.c
150 @@ -484,6 +484,28 @@ ssize_t __vfs_write(struct file *file, const char __user *p, size_t count,
154 +vfs_readf_t vfs_readf(struct file *file)
156 + const struct file_operations *fop = file->f_op;
160 + if (fop->read_iter)
161 + return new_sync_read;
162 + return ERR_PTR(-ENOSYS);
165 +vfs_writef_t vfs_writef(struct file *file)
167 + const struct file_operations *fop = file->f_op;
171 + if (fop->write_iter)
172 + return new_sync_write;
173 + return ERR_PTR(-ENOSYS);
176 ssize_t __kernel_write(struct file *file, const void *buf, size_t count, loff_t *pos)
179 diff --git a/fs/splice.c b/fs/splice.c
180 index 39e2dc0..c5fb195 100644
183 @@ -837,8 +837,8 @@ EXPORT_SYMBOL(generic_splice_sendpage);
185 * Attempt to initiate a splice from pipe to file.
187 -static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
188 - loff_t *ppos, size_t len, unsigned int flags)
189 +long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
190 + loff_t *ppos, size_t len, unsigned int flags)
192 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
193 loff_t *, size_t, unsigned int);
194 @@ -854,9 +854,9 @@ static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
196 * Attempt to initiate a splice from a file to a pipe.
198 -static long do_splice_to(struct file *in, loff_t *ppos,
199 - struct pipe_inode_info *pipe, size_t len,
200 - unsigned int flags)
201 +long do_splice_to(struct file *in, loff_t *ppos,
202 + struct pipe_inode_info *pipe, size_t len,
203 + unsigned int flags)
205 ssize_t (*splice_read)(struct file *, loff_t *,
206 struct pipe_inode_info *, size_t, unsigned int);
207 diff --git a/fs/sync.c b/fs/sync.c
208 index 6e0a2cb..a6891ee 100644
212 * wait == 1 case since in that case write_inode() functions do
213 * sync_dirty_buffer() and thus effectively write one block at a time.
215 -static int __sync_filesystem(struct super_block *sb, int wait)
216 +int __sync_filesystem(struct super_block *sb, int wait)
220 diff --git a/include/linux/file.h b/include/linux/file.h
221 index 279720d..76e38ea 100644
222 --- a/include/linux/file.h
223 +++ b/include/linux/file.h
224 @@ -20,6 +20,7 @@ struct dentry;
226 extern struct file *alloc_file(const struct path *, fmode_t mode,
227 const struct file_operations *fop);
228 +extern struct file *get_empty_filp(void);
230 static inline void fput_light(struct file *file, int fput_needed)
232 diff --git a/include/linux/fs.h b/include/linux/fs.h
233 index 511fbaa..96e05b3 100644
234 --- a/include/linux/fs.h
235 +++ b/include/linux/fs.h
236 @@ -1265,6 +1265,7 @@ extern void fasync_free(struct fasync_struct *);
237 /* can be called from interrupts */
238 extern void kill_fasync(struct fasync_struct **, int, int);
240 +extern int setfl(int fd, struct file * filp, unsigned long arg);
241 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
242 extern int f_setown(struct file *filp, unsigned long arg, int force);
243 extern void f_delown(struct file *filp);
244 @@ -1712,6 +1713,7 @@ struct file_operations {
245 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
246 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
247 int (*check_flags)(int);
248 + int (*setfl)(struct file *, unsigned long);
249 int (*flock) (struct file *, int, struct file_lock *);
250 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
251 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
252 @@ -1782,6 +1784,12 @@ ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
253 struct iovec *fast_pointer,
254 struct iovec **ret_pointer);
256 +typedef ssize_t (*vfs_readf_t)(struct file *, char __user *, size_t, loff_t *);
257 +typedef ssize_t (*vfs_writef_t)(struct file *, const char __user *, size_t,
259 +vfs_readf_t vfs_readf(struct file *file);
260 +vfs_writef_t vfs_writef(struct file *file);
262 extern ssize_t __vfs_read(struct file *, char __user *, size_t, loff_t *);
263 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
264 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
265 @@ -2201,6 +2209,7 @@ extern int current_umask(void);
266 extern void ihold(struct inode * inode);
267 extern void iput(struct inode *);
268 extern int generic_update_time(struct inode *, struct timespec *, int);
269 +extern int update_time(struct inode *, struct timespec *, int);
272 extern struct kobject *fs_kobj;
273 @@ -2481,6 +2490,7 @@ static inline bool sb_is_blkdev_sb(struct super_block *sb)
277 +extern int __sync_filesystem(struct super_block *, int);
278 extern int sync_filesystem(struct super_block *);
279 extern const struct file_operations def_blk_fops;
280 extern const struct file_operations def_chr_fops;
281 diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
282 index a842551..453e941 100644
283 --- a/include/linux/lockdep.h
284 +++ b/include/linux/lockdep.h
285 @@ -406,6 +406,8 @@ static inline int lockdep_match_key(struct lockdep_map *lock,
286 return lock->key == key;
289 +struct lock_class *lockdep_hlock_class(struct held_lock *hlock);
294 @@ -535,6 +537,7 @@ struct lockdep_map { };
296 #define lockdep_depth(tsk) (0)
298 +#define lockdep_is_held(lock) (1)
299 #define lockdep_is_held_type(l, r) (1)
301 #define lockdep_assert_held(l) do { (void)(l); } while (0)
302 diff --git a/include/linux/mnt_namespace.h b/include/linux/mnt_namespace.h
303 index 3594208..24f5fd1 100644
304 --- a/include/linux/mnt_namespace.h
305 +++ b/include/linux/mnt_namespace.h
307 struct mnt_namespace;
309 struct user_namespace;
312 extern struct mnt_namespace *copy_mnt_ns(unsigned long, struct mnt_namespace *,
313 struct user_namespace *, struct fs_struct *);
314 extern void put_mnt_ns(struct mnt_namespace *ns);
316 +extern int is_current_mnt_ns(struct vfsmount *mnt);
318 extern const struct file_operations proc_mounts_operations;
319 extern const struct file_operations proc_mountinfo_operations;
320 extern const struct file_operations proc_mountstats_operations;
321 diff --git a/include/linux/splice.h b/include/linux/splice.h
322 index 74b4911..19789fb 100644
323 --- a/include/linux/splice.h
324 +++ b/include/linux/splice.h
325 @@ -87,4 +87,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
327 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
328 extern const struct pipe_buf_operations default_pipe_buf_ops;
330 +extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
331 + loff_t *ppos, size_t len, unsigned int flags);
332 +extern long do_splice_to(struct file *in, loff_t *ppos,
333 + struct pipe_inode_info *pipe, size_t len,
334 + unsigned int flags);
336 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
337 index 670d8d7..2cd0282 100644
338 --- a/kernel/locking/lockdep.c
339 +++ b/kernel/locking/lockdep.c
340 @@ -156,7 +156,7 @@ static struct lock_list list_entries[MAX_LOCKDEP_ENTRIES];
341 unsigned long nr_lock_classes;
342 static struct lock_class lock_classes[MAX_LOCKDEP_KEYS];
344 -static inline struct lock_class *hlock_class(struct held_lock *hlock)
345 +inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
347 if (!hlock->class_idx) {
349 @@ -167,6 +167,7 @@ static inline struct lock_class *hlock_class(struct held_lock *hlock)
351 return lock_classes + hlock->class_idx - 1;
353 +#define hlock_class(hlock) lockdep_hlock_class(hlock)
355 #ifdef CONFIG_LOCK_STAT
356 static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
357 SPDX-License-Identifier: GPL-2.0
358 aufs4.x-rcN mmap patch
360 diff --git a/fs/proc/base.c b/fs/proc/base.c
361 index 60316b5..ce5314e 100644
364 @@ -1987,7 +1987,7 @@ static int map_files_get_link(struct dentry *dentry, struct path *path)
365 down_read(&mm->mmap_sem);
366 vma = find_exact_vma(mm, vm_start, vm_end);
367 if (vma && vma->vm_file) {
368 - *path = vma->vm_file->f_path;
369 + *path = vma_pr_or_file(vma)->f_path;
373 diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
374 index 7563437..7c0dc0f 100644
375 --- a/fs/proc/nommu.c
376 +++ b/fs/proc/nommu.c
377 @@ -45,7 +45,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
378 file = region->vm_file;
381 - struct inode *inode = file_inode(region->vm_file);
382 + struct inode *inode;
384 + file = vmr_pr_or_file(region);
385 + inode = file_inode(file);
386 dev = inode->i_sb->s_dev;
389 diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
390 index 339e4c1..1138098 100644
391 --- a/fs/proc/task_mmu.c
392 +++ b/fs/proc/task_mmu.c
393 @@ -306,7 +306,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid)
394 const char *name = NULL;
397 - struct inode *inode = file_inode(vma->vm_file);
398 + struct inode *inode;
400 + file = vma_pr_or_file(vma);
401 + inode = file_inode(file);
402 dev = inode->i_sb->s_dev;
404 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
405 @@ -1736,7 +1739,7 @@ static int show_numa_map(struct seq_file *m, void *v, int is_pid)
406 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
407 struct vm_area_struct *vma = v;
408 struct numa_maps *md = &numa_priv->md;
409 - struct file *file = vma->vm_file;
410 + struct file *file = vma_pr_or_file(vma);
411 struct mm_struct *mm = vma->vm_mm;
412 struct mm_walk walk = {
413 .hugetlb_entry = gather_hugetlb_stats,
414 diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
415 index 5b62f57..dfb4a3b 100644
416 --- a/fs/proc/task_nommu.c
417 +++ b/fs/proc/task_nommu.c
418 @@ -156,7 +156,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma,
422 - struct inode *inode = file_inode(vma->vm_file);
423 + struct inode *inode;
425 + file = vma_pr_or_file(vma);
426 + inode = file_inode(file);
427 dev = inode->i_sb->s_dev;
429 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
430 diff --git a/include/linux/mm.h b/include/linux/mm.h
431 index ea818ff..fbd4799 100644
432 --- a/include/linux/mm.h
433 +++ b/include/linux/mm.h
434 @@ -1362,6 +1362,28 @@ static inline int fixup_user_fault(struct task_struct *tsk,
435 unmap_mapping_range(mapping, holebegin, holelen, 0);
438 +extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
439 +extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
441 +extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
442 +extern void vma_do_fput(struct vm_area_struct *, const char[], int);
444 +#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
446 +#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
448 +#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
449 +#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
452 +extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
453 +extern void vmr_do_fput(struct vm_region *, const char[], int);
455 +#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
457 +#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
458 +#endif /* !CONFIG_MMU */
460 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
461 void *buf, int len, unsigned int gup_flags);
462 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
463 diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
464 index cfd0ac4..135e11c 100644
465 --- a/include/linux/mm_types.h
466 +++ b/include/linux/mm_types.h
467 @@ -255,6 +255,7 @@ struct vm_region {
468 unsigned long vm_top; /* region allocated to here */
469 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
470 struct file *vm_file; /* the backing file or NULL */
471 + struct file *vm_prfile; /* the virtual backing file or NULL */
473 int vm_usage; /* region usage count (access under nommu_region_sem) */
474 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
475 @@ -329,6 +330,7 @@ struct vm_area_struct {
476 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
478 struct file * vm_file; /* File we map to (can be NULL). */
479 + struct file *vm_prfile; /* shadow of vm_file */
480 void * vm_private_data; /* was vm_pte (shared mem) */
482 atomic_long_t swap_readahead_info;
483 diff --git a/kernel/fork.c b/kernel/fork.c
484 index 432eadf..8b2ba5b 100644
487 @@ -676,7 +676,7 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
488 struct inode *inode = file_inode(file);
489 struct address_space *mapping = file->f_mapping;
493 if (tmp->vm_flags & VM_DENYWRITE)
494 atomic_dec(&inode->i_writecount);
495 i_mmap_lock_write(mapping);
496 diff --git a/mm/Makefile b/mm/Makefile
497 index e669f02..9c36567 100644
500 @@ -39,7 +39,7 @@ obj-y := filemap.o mempool.o oom_kill.o \
501 mm_init.o mmu_context.o percpu.o slab_common.o \
502 compaction.o vmacache.o swap_slots.o \
503 interval_tree.o list_lru.o workingset.o \
505 + prfile.o debug.o $(mmu-y)
509 diff --git a/mm/filemap.c b/mm/filemap.c
510 index ee83baa..7677d13 100644
513 @@ -2704,7 +2704,7 @@ int filemap_page_mkwrite(struct vm_fault *vmf)
514 int ret = VM_FAULT_LOCKED;
516 sb_start_pagefault(inode->i_sb);
517 - file_update_time(vmf->vma->vm_file);
518 + vma_file_update_time(vmf->vma);
520 if (page->mapping != inode->i_mapping) {
522 diff --git a/mm/mmap.c b/mm/mmap.c
523 index a4d5468..cb06cbd 100644
526 @@ -171,7 +171,7 @@ static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
527 if (vma->vm_ops && vma->vm_ops->close)
528 vma->vm_ops->close(vma);
530 - fput(vma->vm_file);
532 mpol_put(vma_policy(vma));
533 kmem_cache_free(vm_area_cachep, vma);
535 @@ -896,7 +896,7 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
538 uprobe_munmap(next, next->vm_start, next->vm_end);
543 anon_vma_merge(vma, next);
544 @@ -1761,8 +1761,8 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
552 /* Undo any partial mapping done by a device driver. */
553 unmap_region(mm, vma, prev, vma->vm_start, vma->vm_end);
554 @@ -2586,7 +2586,7 @@ int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
558 - get_file(new->vm_file);
561 if (new->vm_ops && new->vm_ops->open)
562 new->vm_ops->open(new);
563 @@ -2605,7 +2605,7 @@ int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
564 if (new->vm_ops && new->vm_ops->close)
565 new->vm_ops->close(new);
567 - fput(new->vm_file);
569 unlink_anon_vmas(new);
571 mpol_put(vma_policy(new));
572 @@ -2767,7 +2767,7 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
573 struct vm_area_struct *vma;
574 unsigned long populate = 0;
575 unsigned long ret = -EINVAL;
577 + struct file *file, *prfile;
579 pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. See Documentation/vm/remap_file_pages.txt.\n",
580 current->comm, current->pid);
581 @@ -2842,10 +2842,27 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
585 - file = get_file(vma->vm_file);
587 + file = vma->vm_file;
588 + prfile = vma->vm_prfile;
589 ret = do_mmap_pgoff(vma->vm_file, start, size,
590 prot, flags, pgoff, &populate, NULL);
591 + if (!IS_ERR_VALUE(ret) && file && prfile) {
592 + struct vm_area_struct *new_vma;
594 + new_vma = find_vma(mm, ret);
595 + if (!new_vma->vm_prfile)
596 + new_vma->vm_prfile = prfile;
597 + if (new_vma != vma)
601 + * two fput()s instead of vma_fput(vma),
602 + * coz vma may not be available anymore.
608 up_write(&mm->mmap_sem);
610 @@ -3153,7 +3170,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
611 if (anon_vma_clone(new_vma, vma))
612 goto out_free_mempol;
613 if (new_vma->vm_file)
614 - get_file(new_vma->vm_file);
615 + vma_get_file(new_vma);
616 if (new_vma->vm_ops && new_vma->vm_ops->open)
617 new_vma->vm_ops->open(new_vma);
618 vma_link(mm, new_vma, prev, rb_link, rb_parent);
619 diff --git a/mm/nommu.c b/mm/nommu.c
620 index 17c00d9..4bcdf94 100644
623 @@ -641,7 +641,7 @@ static void __put_nommu_region(struct vm_region *region)
624 up_write(&nommu_region_sem);
627 - fput(region->vm_file);
630 /* IO memory and memory shared directly out of the pagecache
631 * from ramfs/tmpfs mustn't be released here */
632 @@ -799,7 +799,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
633 if (vma->vm_ops && vma->vm_ops->close)
634 vma->vm_ops->close(vma);
636 - fput(vma->vm_file);
638 put_nommu_region(vma->vm_region);
639 kmem_cache_free(vm_area_cachep, vma);
641 @@ -1321,7 +1321,7 @@ unsigned long do_mmap(struct file *file,
642 goto error_just_free;
645 - fput(region->vm_file);
647 kmem_cache_free(vm_region_jar, region);
650 @@ -1396,10 +1396,10 @@ unsigned long do_mmap(struct file *file,
651 up_write(&nommu_region_sem);
654 - fput(region->vm_file);
656 kmem_cache_free(vm_region_jar, region);
658 - fput(vma->vm_file);
660 kmem_cache_free(vm_area_cachep, vma);
663 diff --git a/mm/prfile.c b/mm/prfile.c
665 index 0000000..3f56669
670 + * SPDX-License-Identifier: GPL-2.0
671 + * Mainly for aufs which mmap(2) different file and wants to print different
672 + * path in /proc/PID/maps.
673 + * Call these functions via macros defined in linux/mm.h.
675 + * See Documentation/filesystems/aufs/design/06mmap.txt
677 + * Copyright (c) 2014-2017 Junjro R. Okajima
678 + * Copyright (c) 2014 Ian Campbell
681 +#include <linux/mm.h>
682 +#include <linux/file.h>
683 +#include <linux/fs.h>
685 +/* #define PRFILE_TRACE */
686 +static inline void prfile_trace(struct file *f, struct file *pr,
687 + const char func[], int line, const char func2[])
691 + pr_info("%s:%d: %s, %pD2\n", func, line, func2, f);
695 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
698 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
700 + prfile_trace(f, pr, func, line, __func__);
701 + file_update_time(f);
703 + file_update_time(pr);
706 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
709 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
711 + prfile_trace(f, pr, func, line, __func__);
712 + return (f && pr) ? pr : f;
715 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
717 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
719 + prfile_trace(f, pr, func, line, __func__);
725 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
727 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
729 + prfile_trace(f, pr, func, line, __func__);
736 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
739 + struct file *f = region->vm_file, *pr = region->vm_prfile;
741 + prfile_trace(f, pr, func, line, __func__);
742 + return (f && pr) ? pr : f;
745 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
747 + struct file *f = region->vm_file, *pr = region->vm_prfile;
749 + prfile_trace(f, pr, func, line, __func__);
754 +#endif /* !CONFIG_MMU */
755 SPDX-License-Identifier: GPL-2.0
756 aufs4.x-rcN standalone patch
758 diff --git a/fs/dcache.c b/fs/dcache.c
759 index 019f14b..10c1a6d 100644
762 @@ -1305,6 +1305,7 @@ void d_walk(struct dentry *parent, void *data,
766 +EXPORT_SYMBOL_GPL(d_walk);
769 struct vfsmount *mnt;
770 @@ -2892,6 +2893,7 @@ void d_exchange(struct dentry *dentry1, struct dentry *dentry2)
772 write_sequnlock(&rename_lock);
774 +EXPORT_SYMBOL_GPL(d_exchange);
777 * d_ancestor - search for an ancestor
778 diff --git a/fs/exec.c b/fs/exec.c
779 index 6be2aa0..1e003f9 100644
782 @@ -109,6 +109,7 @@ bool path_noexec(const struct path *path)
783 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
784 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
786 +EXPORT_SYMBOL_GPL(path_noexec);
790 diff --git a/fs/fcntl.c b/fs/fcntl.c
791 index 74c255d..ec53ee1 100644
794 @@ -85,6 +85,7 @@ int setfl(int fd, struct file * filp, unsigned long arg)
798 +EXPORT_SYMBOL_GPL(setfl);
800 static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
802 diff --git a/fs/file_table.c b/fs/file_table.c
803 index 2dc9f38..7bf57df 100644
804 --- a/fs/file_table.c
805 +++ b/fs/file_table.c
806 @@ -148,6 +148,7 @@ struct file *get_empty_filp(void)
808 return ERR_PTR(-ENFILE);
810 +EXPORT_SYMBOL_GPL(get_empty_filp);
813 * alloc_file - allocate and initialize a 'struct file'
814 @@ -258,6 +259,7 @@ void flush_delayed_fput(void)
818 +EXPORT_SYMBOL_GPL(flush_delayed_fput);
820 static DECLARE_DELAYED_WORK(delayed_fput_work, delayed_fput);
822 @@ -300,6 +302,7 @@ void __fput_sync(struct file *file)
826 +EXPORT_SYMBOL_GPL(__fput_sync);
828 void put_filp(struct file *file)
830 @@ -308,6 +311,7 @@ void put_filp(struct file *file)
834 +EXPORT_SYMBOL_GPL(put_filp);
836 void __init files_init(void)
838 diff --git a/fs/inode.c b/fs/inode.c
839 index 517883c..5cece5e 100644
842 @@ -1664,6 +1664,7 @@ int update_time(struct inode *inode, struct timespec *time, int flags)
844 return update_time(inode, time, flags);
846 +EXPORT_SYMBOL_GPL(update_time);
849 * touch_atime - update the access time
850 diff --git a/fs/namespace.c b/fs/namespace.c
851 index 312bdbd8..a5baeb5 100644
854 @@ -517,6 +517,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
855 mnt_dec_writers(real_mount(mnt));
858 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
861 * mnt_drop_write - give up write access to a mount
862 @@ -851,6 +852,7 @@ int is_current_mnt_ns(struct vfsmount *mnt)
864 return check_mnt(real_mount(mnt));
866 +EXPORT_SYMBOL_GPL(is_current_mnt_ns);
869 * vfsmount lock must be held for write
870 @@ -1887,6 +1889,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
874 +EXPORT_SYMBOL_GPL(iterate_mounts);
876 static void cleanup_group_ids(struct mount *mnt, struct mount *end)
878 diff --git a/fs/notify/group.c b/fs/notify/group.c
879 index b7a4b6a..5a69d60 100644
880 --- a/fs/notify/group.c
881 +++ b/fs/notify/group.c
883 #include <linux/srcu.h>
884 #include <linux/rculist.h>
885 #include <linux/wait.h>
886 +#include <linux/module.h>
888 #include <linux/fsnotify_backend.h>
889 #include "fsnotify.h"
890 @@ -109,6 +110,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
892 refcount_inc(&group->refcnt);
894 +EXPORT_SYMBOL_GPL(fsnotify_get_group);
897 * Drop a reference to a group. Free it if it's through.
898 @@ -118,6 +120,7 @@ void fsnotify_put_group(struct fsnotify_group *group)
899 if (refcount_dec_and_test(&group->refcnt))
900 fsnotify_final_destroy_group(group);
902 +EXPORT_SYMBOL_GPL(fsnotify_put_group);
905 * Create a new fsnotify_group and hold a reference for the group returned.
906 @@ -147,6 +150,7 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
910 +EXPORT_SYMBOL_GPL(fsnotify_alloc_group);
912 int fsnotify_fasync(int fd, struct file *file, int on)
914 diff --git a/fs/notify/mark.c b/fs/notify/mark.c
915 index e9191b4..1f8ccfa 100644
916 --- a/fs/notify/mark.c
917 +++ b/fs/notify/mark.c
918 @@ -108,6 +108,7 @@ void fsnotify_get_mark(struct fsnotify_mark *mark)
919 WARN_ON_ONCE(!refcount_read(&mark->refcnt));
920 refcount_inc(&mark->refcnt);
922 +EXPORT_SYMBOL_GPL(fsnotify_put_mark);
924 static void __fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
926 @@ -392,6 +393,7 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark,
927 mutex_unlock(&group->mark_mutex);
928 fsnotify_free_mark(mark);
930 +EXPORT_SYMBOL_GPL(fsnotify_destroy_mark);
933 * Sorting function for lists of fsnotify marks.
934 @@ -606,6 +608,7 @@ int fsnotify_add_mark_locked(struct fsnotify_mark *mark, struct inode *inode,
935 fsnotify_put_mark(mark);
938 +EXPORT_SYMBOL_GPL(fsnotify_add_mark);
940 int fsnotify_add_mark(struct fsnotify_mark *mark, struct inode *inode,
941 struct vfsmount *mnt, int allow_dups)
942 @@ -741,6 +744,7 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
943 fsnotify_get_group(group);
946 +EXPORT_SYMBOL_GPL(fsnotify_init_mark);
949 * Destroy all marks in destroy_list, waits for SRCU period to finish before
950 diff --git a/fs/open.c b/fs/open.c
951 index 7ea1184..6e2e241 100644
954 @@ -64,6 +64,7 @@ int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
955 inode_unlock(dentry->d_inode);
958 +EXPORT_SYMBOL_GPL(do_truncate);
960 long vfs_truncate(const struct path *path, loff_t length)
962 @@ -691,6 +692,7 @@ int open_check_o_direct(struct file *f)
966 +EXPORT_SYMBOL_GPL(open_check_o_direct);
968 static int do_dentry_open(struct file *f,
970 diff --git a/fs/read_write.c b/fs/read_write.c
971 index 0a5c47b..d423a5f 100644
972 --- a/fs/read_write.c
973 +++ b/fs/read_write.c
974 @@ -454,6 +454,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
978 +EXPORT_SYMBOL_GPL(vfs_read);
980 static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
982 @@ -494,6 +495,7 @@ vfs_readf_t vfs_readf(struct file *file)
983 return new_sync_read;
984 return ERR_PTR(-ENOSYS);
986 +EXPORT_SYMBOL_GPL(vfs_readf);
988 vfs_writef_t vfs_writef(struct file *file)
990 @@ -505,6 +507,7 @@ vfs_writef_t vfs_writef(struct file *file)
991 return new_sync_write;
992 return ERR_PTR(-ENOSYS);
994 +EXPORT_SYMBOL_GPL(vfs_writef);
996 ssize_t __kernel_write(struct file *file, const void *buf, size_t count, loff_t *pos)
998 @@ -574,6 +577,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
1002 +EXPORT_SYMBOL_GPL(vfs_write);
1004 static inline loff_t file_pos_read(struct file *file)
1006 diff --git a/fs/splice.c b/fs/splice.c
1007 index c5fb195..ce01a74 100644
1010 @@ -850,6 +850,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
1012 return splice_write(pipe, out, ppos, len, flags);
1014 +EXPORT_SYMBOL_GPL(do_splice_from);
1017 * Attempt to initiate a splice from a file to a pipe.
1018 @@ -879,6 +880,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
1020 return splice_read(in, ppos, pipe, len, flags);
1022 +EXPORT_SYMBOL_GPL(do_splice_to);
1025 * splice_direct_to_actor - splices data directly between two non-pipes
1026 diff --git a/fs/sync.c b/fs/sync.c
1027 index a6891ee..47a78bd 100644
1030 @@ -39,6 +39,7 @@ int __sync_filesystem(struct super_block *sb, int wait)
1031 sb->s_op->sync_fs(sb, wait);
1032 return __sync_blockdev(sb->s_bdev, wait);
1034 +EXPORT_SYMBOL_GPL(__sync_filesystem);
1037 * Write out and wait upon all dirty data associated with this
1038 diff --git a/fs/xattr.c b/fs/xattr.c
1039 index 61cd28b..35570cd 100644
1042 @@ -297,6 +297,7 @@ vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
1043 *xattr_value = value;
1046 +EXPORT_SYMBOL_GPL(vfs_getxattr_alloc);
1049 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
1050 diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
1051 index 2cd0282..af59768 100644
1052 --- a/kernel/locking/lockdep.c
1053 +++ b/kernel/locking/lockdep.c
1054 @@ -167,6 +167,7 @@ inline struct lock_class *lockdep_hlock_class(struct held_lock *hlock)
1056 return lock_classes + hlock->class_idx - 1;
1058 +EXPORT_SYMBOL_GPL(lockdep_hlock_class);
1059 #define hlock_class(hlock) lockdep_hlock_class(hlock)
1061 #ifdef CONFIG_LOCK_STAT
1062 diff --git a/kernel/task_work.c b/kernel/task_work.c
1063 index 0fef395..83fb1ec 100644
1064 --- a/kernel/task_work.c
1065 +++ b/kernel/task_work.c
1066 @@ -116,3 +116,4 @@ void task_work_run(void)
1070 +EXPORT_SYMBOL_GPL(task_work_run);
1071 diff --git a/security/commoncap.c b/security/commoncap.c
1072 index 4f8e093..f1e0544 100644
1073 --- a/security/commoncap.c
1074 +++ b/security/commoncap.c
1075 @@ -1333,12 +1333,14 @@ int cap_mmap_addr(unsigned long addr)
1079 +EXPORT_SYMBOL_GPL(cap_mmap_addr);
1081 int cap_mmap_file(struct file *file, unsigned long reqprot,
1082 unsigned long prot, unsigned long flags)
1086 +EXPORT_SYMBOL_GPL(cap_mmap_file);
1088 #ifdef CONFIG_SECURITY
1090 diff --git a/security/device_cgroup.c b/security/device_cgroup.c
1091 index c65b39b..e363d22 100644
1092 --- a/security/device_cgroup.c
1093 +++ b/security/device_cgroup.c
1095 #include <linux/device_cgroup.h>
1096 #include <linux/cgroup.h>
1097 #include <linux/ctype.h>
1098 +#include <linux/export.h>
1099 #include <linux/list.h>
1100 #include <linux/uaccess.h>
1101 #include <linux/seq_file.h>
1102 @@ -824,3 +825,4 @@ int __devcgroup_check_permission(short type, u32 major, u32 minor,
1106 +EXPORT_SYMBOL_GPL(__devcgroup_check_permission);
1107 diff --git a/security/security.c b/security/security.c
1108 index 1cd8526..f2e4736 100644
1109 --- a/security/security.c
1110 +++ b/security/security.c
1111 @@ -531,6 +531,7 @@ int security_path_rmdir(const struct path *dir, struct dentry *dentry)
1113 return call_int_hook(path_rmdir, 0, dir, dentry);
1115 +EXPORT_SYMBOL_GPL(security_path_rmdir);
1117 int security_path_unlink(const struct path *dir, struct dentry *dentry)
1119 @@ -547,6 +548,7 @@ int security_path_symlink(const struct path *dir, struct dentry *dentry,
1121 return call_int_hook(path_symlink, 0, dir, dentry, old_name);
1123 +EXPORT_SYMBOL_GPL(security_path_symlink);
1125 int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
1126 struct dentry *new_dentry)
1127 @@ -555,6 +557,7 @@ int security_path_link(struct dentry *old_dentry, const struct path *new_dir,
1129 return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
1131 +EXPORT_SYMBOL_GPL(security_path_link);
1133 int security_path_rename(const struct path *old_dir, struct dentry *old_dentry,
1134 const struct path *new_dir, struct dentry *new_dentry,
1135 @@ -582,6 +585,7 @@ int security_path_truncate(const struct path *path)
1137 return call_int_hook(path_truncate, 0, path);
1139 +EXPORT_SYMBOL_GPL(security_path_truncate);
1141 int security_path_chmod(const struct path *path, umode_t mode)
1143 @@ -589,6 +593,7 @@ int security_path_chmod(const struct path *path, umode_t mode)
1145 return call_int_hook(path_chmod, 0, path, mode);
1147 +EXPORT_SYMBOL_GPL(security_path_chmod);
1149 int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
1151 @@ -596,6 +601,7 @@ int security_path_chown(const struct path *path, kuid_t uid, kgid_t gid)
1153 return call_int_hook(path_chown, 0, path, uid, gid);
1155 +EXPORT_SYMBOL_GPL(security_path_chown);
1157 int security_path_chroot(const struct path *path)
1159 @@ -681,6 +687,7 @@ int security_inode_readlink(struct dentry *dentry)
1161 return call_int_hook(inode_readlink, 0, dentry);
1163 +EXPORT_SYMBOL_GPL(security_inode_readlink);
1165 int security_inode_follow_link(struct dentry *dentry, struct inode *inode,
1167 @@ -696,6 +703,7 @@ int security_inode_permission(struct inode *inode, int mask)
1169 return call_int_hook(inode_permission, 0, inode, mask);
1171 +EXPORT_SYMBOL_GPL(security_inode_permission);
1173 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
1175 @@ -867,6 +875,7 @@ int security_file_permission(struct file *file, int mask)
1177 return fsnotify_perm(file, mask);
1179 +EXPORT_SYMBOL_GPL(security_file_permission);
1181 int security_file_alloc(struct file *file)
1183 @@ -926,6 +935,7 @@ int security_mmap_file(struct file *file, unsigned long prot,
1185 return ima_file_mmap(file, prot);
1187 +EXPORT_SYMBOL_GPL(security_mmap_file);
1189 int security_mmap_addr(unsigned long addr)
1191 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
1192 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
1193 +++ linux/Documentation/ABI/testing/debugfs-aufs 2016-01-13 20:11:11.663093444 +0100
1195 +What: /debug/aufs/si_<id>/
1197 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1199 + Under /debug/aufs, a directory named si_<id> is created
1200 + per aufs mount, where <id> is a unique id generated
1203 +What: /debug/aufs/si_<id>/plink
1205 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1207 + It has three lines and shows the information about the
1208 + pseudo-link. The first line is a single number
1209 + representing a number of buckets. The second line is a
1210 + number of pseudo-links per buckets (separated by a
1211 + blank). The last line is a single number representing a
1212 + total number of psedo-links.
1213 + When the aufs mount option 'noplink' is specified, it
1214 + will show "1\n0\n0\n".
1216 +What: /debug/aufs/si_<id>/xib
1218 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1220 + It shows the consumed blocks by xib (External Inode Number
1221 + Bitmap), its block size and file size.
1222 + When the aufs mount option 'noxino' is specified, it
1223 + will be empty. About XINO files, see the aufs manual.
1225 +What: /debug/aufs/si_<id>/xino0, xino1 ... xinoN
1227 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1229 + It shows the consumed blocks by xino (External Inode Number
1230 + Translation Table), its link count, block size and file
1232 + When the aufs mount option 'noxino' is specified, it
1233 + will be empty. About XINO files, see the aufs manual.
1235 +What: /debug/aufs/si_<id>/xigen
1237 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1239 + It shows the consumed blocks by xigen (External Inode
1240 + Generation Table), its block size and file size.
1241 + If CONFIG_AUFS_EXPORT is disabled, this entry will not
1243 + When the aufs mount option 'noxino' is specified, it
1244 + will be empty. About XINO files, see the aufs manual.
1245 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
1246 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
1247 +++ linux/Documentation/ABI/testing/sysfs-aufs 2016-01-13 20:11:11.663093444 +0100
1249 +What: /sys/fs/aufs/si_<id>/
1251 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1253 + Under /sys/fs/aufs, a directory named si_<id> is created
1254 + per aufs mount, where <id> is a unique id generated
1257 +What: /sys/fs/aufs/si_<id>/br0, br1 ... brN
1259 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1261 + It shows the abolute path of a member directory (which
1262 + is called branch) in aufs, and its permission.
1264 +What: /sys/fs/aufs/si_<id>/brid0, brid1 ... bridN
1266 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1268 + It shows the id of a member directory (which is called
1271 +What: /sys/fs/aufs/si_<id>/xi_path
1273 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1275 + It shows the abolute path of XINO (External Inode Number
1276 + Bitmap, Translation Table and Generation Table) file
1277 + even if it is the default path.
1278 + When the aufs mount option 'noxino' is specified, it
1279 + will be empty. About XINO files, see the aufs manual.
1280 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1281 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
1282 +++ linux/Documentation/filesystems/aufs/design/01intro.txt 2018-01-29 07:56:20.046658203 +0100
1285 +# Copyright (C) 2005-2017 Junjiro R. Okajima
1287 +# This program is free software; you can redistribute it and/or modify
1288 +# it under the terms of the GNU General Public License as published by
1289 +# the Free Software Foundation; either version 2 of the License, or
1290 +# (at your option) any later version.
1292 +# This program is distributed in the hope that it will be useful,
1293 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1294 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1295 +# GNU General Public License for more details.
1297 +# You should have received a copy of the GNU General Public License
1298 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1301 +----------------------------------------
1303 +aufs [ei ju: ef es] | /ey-yoo-ef-es/ | [a u f s]
1304 +1. abbrev. for "advanced multi-layered unification filesystem".
1305 +2. abbrev. for "another unionfs".
1306 +3. abbrev. for "auf das" in German which means "on the" in English.
1307 + Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1308 + But "Filesystem aufs Filesystem" is hard to understand.
1309 +4. abbrev. for "African Urban Fashion Show".
1311 +AUFS is a filesystem with features:
1312 +- multi layered stackable unification filesystem, the member directory
1313 + is called as a branch.
1314 +- branch permission and attribute, 'readonly', 'real-readonly',
1315 + 'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1317 +- internal "file copy-on-write".
1318 +- logical deletion, whiteout.
1319 +- dynamic branch manipulation, adding, deleting and changing permission.
1320 +- allow bypassing aufs, user's direct branch access.
1321 +- external inode number translation table and bitmap which maintains the
1322 + persistent aufs inode number.
1323 +- seekable directory, including NFS readdir.
1324 +- file mapping, mmap and sharing pages.
1325 +- pseudo-link, hardlink over branches.
1326 +- loopback mounted filesystem as a branch.
1327 +- several policies to select one among multiple writable branches.
1328 +- revert a single systemcall when an error occurs in aufs.
1332 +Multi Layered Stackable Unification Filesystem
1333 +----------------------------------------------------------------------
1334 +Most people already knows what it is.
1335 +It is a filesystem which unifies several directories and provides a
1336 +merged single directory. When users access a file, the access will be
1337 +passed/re-directed/converted (sorry, I am not sure which English word is
1338 +correct) to the real file on the member filesystem. The member
1339 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1340 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1341 +readonly branch is handled by creating 'whiteout' on the upper writable
1344 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1345 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1346 +different approaches to implement the merged-view.
1347 +The former tries putting it into VFS, and the latter implements as a
1348 +separate filesystem.
1349 +(If I misunderstand about these implementations, please let me know and
1350 +I shall correct it. Because it is a long time ago when I read their
1351 +source files last time).
1353 +UnionMount's approach will be able to small, but may be hard to share
1354 +branches between several UnionMount since the whiteout in it is
1355 +implemented in the inode on branch filesystem and always
1356 +shared. According to Bharata's post, readdir does not seems to be
1358 +There are several missing features known in this implementations such as
1359 +- for users, the inode number may change silently. eg. copy-up.
1360 +- link(2) may break by copy-up.
1361 +- read(2) may get an obsoleted filedata (fstat(2) too).
1362 +- fcntl(F_SETLK) may be broken by copy-up.
1363 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1366 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1367 +merged into mainline. This is another implementation of UnionMount as a
1368 +separated filesystem. All the limitations and known problems which
1369 +UnionMount are equally inherited to "overlay" filesystem.
1371 +Unionfs has a longer history. When I started implementing a stackable
1372 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1373 +inode, dentry and file objects and they have an array pointing lower
1374 +same kind objects. After contributing many patches for Unionfs, I
1375 +re-started my project AUFS (Jun 2006).
1377 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1378 +implemented my own ideas, approaches and enhancements and it became
1379 +totally different one.
1381 +Comparing DM snapshot and fs based implementation
1382 +- the number of bytes to be copied between devices is much smaller.
1383 +- the type of filesystem must be one and only.
1384 +- the fs must be writable, no readonly fs, even for the lower original
1385 + device. so the compression fs will not be usable. but if we use
1386 + loopback mount, we may address this issue.
1388 + mount /cdrom/squashfs.img /sq
1389 + losetup /sq/ext2.img
1390 + losetup /somewhere/cow
1391 + dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1392 +- it will be difficult (or needs more operations) to extract the
1393 + difference between the original device and COW.
1394 +- DM snapshot-merge may help a lot when users try merging. in the
1395 + fs-layer union, users will use rsync(1).
1397 +You may want to read my old paper "Filesystems in LiveCD"
1398 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1401 +Several characters/aspects/persona of aufs
1402 +----------------------------------------------------------------------
1404 +Aufs has several characters, aspects or persona.
1405 +1. a filesystem, callee of VFS helper
1406 +2. sub-VFS, caller of VFS helper for branches
1407 +3. a virtual filesystem which maintains persistent inode number
1408 +4. reader/writer of files on branches such like an application
1410 +1. Callee of VFS Helper
1411 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1412 +unlink(2) from an application reaches sys_unlink() kernel function and
1413 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1414 +calls filesystem specific unlink operation. Actually aufs implements the
1415 +unlink operation but it behaves like a redirector.
1417 +2. Caller of VFS Helper for Branches
1418 +aufs_unlink() passes the unlink request to the branch filesystem as if
1419 +it were called from VFS. So the called unlink operation of the branch
1420 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1421 +every necessary pre/post operation for the branch filesystem.
1422 +- acquire the lock for the parent dir on a branch
1423 +- lookup in a branch
1424 +- revalidate dentry on a branch
1425 +- mnt_want_write() for a branch
1426 +- vfs_unlink() for a branch
1427 +- mnt_drop_write() for a branch
1428 +- release the lock on a branch
1430 +3. Persistent Inode Number
1431 +One of the most important issue for a filesystem is to maintain inode
1432 +numbers. This is particularly important to support exporting a
1433 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1434 +backend block device for its own. But some storage is necessary to
1435 +keep and maintain the inode numbers. It may be a large space and may not
1436 +suit to keep in memory. Aufs rents some space from its first writable
1437 +branch filesystem (by default) and creates file(s) on it. These files
1438 +are created by aufs internally and removed soon (currently) keeping
1440 +Note: Because these files are removed, they are totally gone after
1441 + unmounting aufs. It means the inode numbers are not persistent
1442 + across unmount or reboot. I have a plan to make them really
1443 + persistent which will be important for aufs on NFS server.
1445 +4. Read/Write Files Internally (copy-on-write)
1446 +Because a branch can be readonly, when you write a file on it, aufs will
1447 +"copy-up" it to the upper writable branch internally. And then write the
1448 +originally requested thing to the file. Generally kernel doesn't
1449 +open/read/write file actively. In aufs, even a single write may cause a
1450 +internal "file copy". This behaviour is very similar to cp(1) command.
1452 +Some people may think it is better to pass such work to user space
1453 +helper, instead of doing in kernel space. Actually I am still thinking
1454 +about it. But currently I have implemented it in kernel space.
1455 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1456 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1457 +++ linux/Documentation/filesystems/aufs/design/02struct.txt 2018-01-29 07:56:20.049991637 +0100
1460 +# Copyright (C) 2005-2017 Junjiro R. Okajima
1462 +# This program is free software; you can redistribute it and/or modify
1463 +# it under the terms of the GNU General Public License as published by
1464 +# the Free Software Foundation; either version 2 of the License, or
1465 +# (at your option) any later version.
1467 +# This program is distributed in the hope that it will be useful,
1468 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1469 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1470 +# GNU General Public License for more details.
1472 +# You should have received a copy of the GNU General Public License
1473 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1475 +Basic Aufs Internal Structure
1477 +Superblock/Inode/Dentry/File Objects
1478 +----------------------------------------------------------------------
1479 +As like an ordinary filesystem, aufs has its own
1480 +superblock/inode/dentry/file objects. All these objects have a
1481 +dynamically allocated array and store the same kind of pointers to the
1482 +lower filesystem, branch.
1483 +For example, when you build a union with one readwrite branch and one
1484 +readonly, mounted /au, /rw and /ro respectively.
1486 +- /ro/fileA exists but /rw/fileA
1488 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1489 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1490 +- [0] = NULL (because /rw/fileA doesn't exist)
1493 +This style of an array is essentially same to the aufs
1494 +superblock/inode/dentry/file objects.
1496 +Because aufs supports manipulating branches, ie. add/delete/change
1497 +branches dynamically, these objects has its own generation. When
1498 +branches are changed, the generation in aufs superblock is
1499 +incremented. And a generation in other object are compared when it is
1500 +accessed. When a generation in other objects are obsoleted, aufs
1501 +refreshes the internal array.
1505 +----------------------------------------------------------------------
1506 +Additionally aufs superblock has some data for policies to select one
1507 +among multiple writable branches, XIB files, pseudo-links and kobject.
1508 +See below in detail.
1509 +About the policies which supports copy-down a directory, see
1510 +wbr_policy.txt too.
1513 +Branch and XINO(External Inode Number Translation Table)
1514 +----------------------------------------------------------------------
1515 +Every branch has its own xino (external inode number translation table)
1516 +file. The xino file is created and unlinked by aufs internally. When two
1517 +members of a union exist on the same filesystem, they share the single
1519 +The struct of a xino file is simple, just a sequence of aufs inode
1520 +numbers which is indexed by the lower inode number.
1521 +In the above sample, assume the inode number of /ro/fileA is i111 and
1522 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1523 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1525 +When the inode numbers are not contiguous, the xino file will be sparse
1526 +which has a hole in it and doesn't consume as much disk space as it
1527 +might appear. If your branch filesystem consumes disk space for such
1528 +holes, then you should specify 'xino=' option at mounting aufs.
1530 +Aufs has a mount option to free the disk blocks for such holes in XINO
1531 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1532 +meet a problem of disk shortage due to XINO files, then you should try
1533 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1534 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1535 +the holes in XINO files.
1537 +Also a writable branch has three kinds of "whiteout bases". All these
1538 +are existed when the branch is joined to aufs, and their names are
1539 +whiteout-ed doubly, so that users will never see their names in aufs
1541 +1. a regular file which will be hardlinked to all whiteouts.
1542 +2. a directory to store a pseudo-link.
1543 +3. a directory to store an "orphan"-ed file temporary.
1546 + When you remove a file on a readonly branch, aufs handles it as a
1547 + logical deletion and creates a whiteout on the upper writable branch
1548 + as a hardlink of this file in order not to consume inode on the
1551 + See below, Pseudo-link.
1553 + When "fileC" exists on the lower readonly branch only and it is
1554 + opened and removed with its parent dir, and then user writes
1555 + something into it, then aufs copies-up fileC to this
1556 + directory. Because there is no other dir to store fileC. After
1557 + creating a file under this dir, the file is unlinked.
1559 +Because aufs supports manipulating branches, ie. add/delete/change
1560 +dynamically, a branch has its own id. When the branch order changes,
1561 +aufs finds the new index by searching the branch id.
1565 +----------------------------------------------------------------------
1566 +Assume "fileA" exists on the lower readonly branch only and it is
1567 +hardlinked to "fileB" on the branch. When you write something to fileA,
1568 +aufs copies-up it to the upper writable branch. Additionally aufs
1569 +creates a hardlink under the Pseudo-link Directory of the writable
1570 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1571 +simple list. If fileB is read after unlinking fileA, aufs returns
1572 +filedata from the pseudo-link instead of the lower readonly
1573 +branch. Because the pseudo-link is based upon the inode, to keep the
1574 +inode number by xino (see above) is essentially necessary.
1576 +All the hardlinks under the Pseudo-link Directory of the writable branch
1577 +should be restored in a proper location later. Aufs provides a utility
1578 +to do this. The userspace helpers executed at remounting and unmounting
1580 +During this utility is running, it puts aufs into the pseudo-link
1581 +maintenance mode. In this mode, only the process which began the
1582 +maintenance mode (and its child processes) is allowed to operate in
1583 +aufs. Some other processes which are not related to the pseudo-link will
1584 +be allowed to run too, but the rest have to return an error or wait
1585 +until the maintenance mode ends. If a process already acquires an inode
1586 +mutex (in VFS), it has to return an error.
1589 +XIB(external inode number bitmap)
1590 +----------------------------------------------------------------------
1591 +Addition to the xino file per a branch, aufs has an external inode number
1592 +bitmap in a superblock object. It is also an internal file such like a
1594 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1596 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1598 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1599 +reduce the number of consumed disk blocks for these files.
1602 +Virtual or Vertical Dir, and Readdir in Userspace
1603 +----------------------------------------------------------------------
1604 +In order to support multiple layers (branches), aufs readdir operation
1605 +constructs a virtual dir block on memory. For readdir, aufs calls
1606 +vfs_readdir() internally for each dir on branches, merges their entries
1607 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1608 +object. So the file object has its entry list until it is closed. The
1609 +entry list will be updated when the file position is zero and becomes
1610 +obsoleted. This decision is made in aufs automatically.
1612 +The dynamically allocated memory block for the name of entries has a
1613 +unit of 512 bytes (by default) and stores the names contiguously (no
1614 +padding). Another block for each entry is handled by kmem_cache too.
1615 +During building dir blocks, aufs creates hash list and judging whether
1616 +the entry is whiteouted by its upper branch or already listed.
1617 +The merged result is cached in the corresponding inode object and
1618 +maintained by a customizable life-time option.
1620 +Some people may call it can be a security hole or invite DoS attack
1621 +since the opened and once readdir-ed dir (file object) holds its entry
1622 +list and becomes a pressure for system memory. But I'd say it is similar
1623 +to files under /proc or /sys. The virtual files in them also holds a
1624 +memory page (generally) while they are opened. When an idea to reduce
1625 +memory for them is introduced, it will be applied to aufs too.
1626 +For those who really hate this situation, I've developed readdir(3)
1627 +library which operates this merging in userspace. You just need to set
1628 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1629 +kernel space for readdir(3).
1633 +----------------------------------------------------------------------
1634 +Aufs sometimes requires privilege access to a branch. For instance,
1635 +in copy-up/down operation. When a user process is going to make changes
1636 +to a file which exists in the lower readonly branch only, and the mode
1637 +of one of ancestor directories may not be writable by a user
1638 +process. Here aufs copy-up the file with its ancestors and they may
1639 +require privilege to set its owner/group/mode/etc.
1640 +This is a typical case of a application character of aufs (see
1643 +Aufs uses workqueue synchronously for this case. It creates its own
1644 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1645 +passes the request to call mkdir or write (for example), and wait for
1646 +its completion. This approach solves a problem of a signal handler
1648 +If aufs didn't adopt the workqueue and changed the privilege of the
1649 +process, then the process may receive the unexpected SIGXFSZ or other
1652 +Also aufs uses the system global workqueue ("events" kernel thread) too
1653 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1654 +whiteout base and etc. This is unrelated to a privilege.
1655 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1656 +superblock at the beginning, at the same time waits for the completion
1657 +of all queued asynchronous tasks.
1661 +----------------------------------------------------------------------
1662 +The whiteout in aufs is very similar to Unionfs's. That is represented
1663 +by its filename. UnionMount takes an approach of a file mode, but I am
1664 +afraid several utilities (find(1) or something) will have to support it.
1666 +Basically the whiteout represents "logical deletion" which stops aufs to
1667 +lookup further, but also it represents "dir is opaque" which also stop
1670 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1671 +In order to make several functions in a single systemcall to be
1672 +revertible, aufs adopts an approach to rename a directory to a temporary
1673 +unique whiteouted name.
1674 +For example, in rename(2) dir where the target dir already existed, aufs
1675 +renames the target dir to a temporary unique whiteouted name before the
1676 +actual rename on a branch, and then handles other actions (make it opaque,
1677 +update the attributes, etc). If an error happens in these actions, aufs
1678 +simply renames the whiteouted name back and returns an error. If all are
1679 +succeeded, aufs registers a function to remove the whiteouted unique
1680 +temporary name completely and asynchronously to the system global
1685 +----------------------------------------------------------------------
1686 +It is a well-known feature or concept.
1687 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1688 +internally and makes change to the new file on the upper writable branch.
1689 +When the trigger systemcall does not update the timestamps of the parent
1690 +dir, aufs reverts it after copy-up.
1693 +Move-down (aufs3.9 and later)
1694 +----------------------------------------------------------------------
1695 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1696 +the lower readonly branch to the upper writable branch when a user
1697 +changes something about the file.
1698 +"Move-down" is an opposite action of copy-up. Basically this action is
1699 +ran manually instead of automatically and internally.
1700 +For desgin and implementation, aufs has to consider these issues.
1701 +- whiteout for the file may exist on the lower branch.
1702 +- ancestor directories may not exist on the lower branch.
1703 +- diropq for the ancestor directories may exist on the upper branch.
1704 +- free space on the lower branch will reduce.
1705 +- another access to the file may happen during moving-down, including
1706 + UDBA (see "Revalidate Dentry and UDBA").
1707 +- the file should not be hard-linked nor pseudo-linked. they should be
1708 + handled by auplink utility later.
1710 +Sometimes users want to move-down a file from the upper writable branch
1711 +to the lower readonly or writable branch. For instance,
1712 +- the free space of the upper writable branch is going to run out.
1713 +- create a new intermediate branch between the upper and lower branch.
1716 +For this purpose, use "aumvdown" command in aufs-util.git.
1717 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1718 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt 1970-01-01 01:00:00.000000000 +0100
1719 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt 2018-01-29 07:56:20.049991637 +0100
1722 +# Copyright (C) 2015-2017 Junjiro R. Okajima
1724 +# This program is free software; you can redistribute it and/or modify
1725 +# it under the terms of the GNU General Public License as published by
1726 +# the Free Software Foundation; either version 2 of the License, or
1727 +# (at your option) any later version.
1729 +# This program is distributed in the hope that it will be useful,
1730 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1731 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1732 +# GNU General Public License for more details.
1734 +# You should have received a copy of the GNU General Public License
1735 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1737 +Support for a branch who has its ->atomic_open()
1738 +----------------------------------------------------------------------
1739 +The filesystems who implement its ->atomic_open() are not majority. For
1740 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1741 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1742 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1743 +sure whether all filesystems who have ->atomic_open() behave like this,
1744 +but NFSv4 surely returns the error.
1746 +In order to support ->atomic_open() for aufs, there are a few
1749 +A. Introduce aufs_atomic_open()
1750 + - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1752 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1753 + an aufs user Pip Cet's approach
1754 + - calls aufs_create(), VFS finish_open() and notify_change().
1755 + - pass fake-mode to finish_open(), and then correct the mode by
1757 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1758 + - no aufs_atomic_open().
1759 + - aufs_lookup() registers the TID to an aufs internal object.
1760 + - aufs_create() does nothing when the matching TID is registered, but
1761 + registers the mode.
1762 + - aufs_open() calls branch fs's ->atomic_open() when the matching
1763 + TID is registered.
1764 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1766 + - no aufs_atomic_open().
1767 + - aufs_create() registers the TID to an internal object. this info
1768 + represents "this process created this file just now."
1769 + - when aufs gets EACCES from branch fs's ->open(), then confirm the
1770 + registered TID and re-try open() with superuser's credential.
1772 +Pros and cons for each approach.
1775 + - straightforward but highly depends upon VFS internal.
1776 + - the atomic behavaiour is kept.
1777 + - some of parameters such as nameidata are hard to reproduce for
1781 + - easy to implement.
1782 + - the atomic behavaiour is lost.
1784 + - the atomic behavaiour is kept.
1785 + - dirty and tricky.
1786 + - VFS checks whether the file is created correctly after calling
1787 + ->create(), which means this approach doesn't work.
1789 + - easy to implement.
1790 + - the atomic behavaiour is lost.
1791 + - to open a file with superuser's credential and give it to a user
1792 + process is a bad idea, since the file object keeps the credential
1793 + in it. It may affect LSM or something. This approach doesn't work
1796 +The approach A is ideal, but it hard to implement. So here is a
1797 +variation of A, which is to be implemented.
1799 +A-1. Introduce aufs_atomic_open()
1800 + - calls branch fs ->atomic_open() if exists. otherwise calls
1801 + vfs_create() and finish_open().
1802 + - the demerit is that the several checks after branch fs
1803 + ->atomic_open() are lost. in the ordinary case, the checks are
1804 + done by VFS:do_last(), lookup_open() and atomic_open(). some can
1805 + be implemented in aufs, but not all I am afraid.
1806 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1807 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1808 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2018-01-29 07:56:20.049991637 +0100
1811 +# Copyright (C) 2005-2017 Junjiro R. Okajima
1813 +# This program is free software; you can redistribute it and/or modify
1814 +# it under the terms of the GNU General Public License as published by
1815 +# the Free Software Foundation; either version 2 of the License, or
1816 +# (at your option) any later version.
1818 +# This program is distributed in the hope that it will be useful,
1819 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1820 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1821 +# GNU General Public License for more details.
1823 +# You should have received a copy of the GNU General Public License
1824 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1827 +----------------------------------------------------------------------
1828 +Since aufs has a character of sub-VFS (see Introduction), it operates
1829 +lookup for branches as VFS does. It may be a heavy work. But almost all
1830 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1831 +directly connected to its parent. Digging down the directory hierarchy
1832 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1835 +When a branch is a remote filesystem, aufs basically relies upon its
1836 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1838 +For d_revalidate, aufs implements three levels of revalidate tests. See
1839 +"Revalidate Dentry and UDBA" in detail.
1842 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1843 +----------------------------------------------------------------------
1844 +Let's try case study.
1845 +- aufs has two branches, upper readwrite and lower readonly.
1847 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1848 +- user invoked "chmod a+rx /au/dirA"
1849 +- the internal copy-up is activated and "/rw/dirA" is created and its
1850 + permission bits are set to world readable.
1851 +- then "/au/dirA" becomes world readable?
1853 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1854 +or it may be a natively readonly filesystem. If aufs respects the lower
1855 +branch, it should not respond readdir request from other users. But user
1856 +allowed it by chmod. Should really aufs rejects showing the entries
1859 +To be honest, I don't have a good solution for this case. So aufs
1860 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1862 +When dirperm1 is specified, aufs checks only the highest one for the
1863 +directory permission, and shows the entries. Otherwise, as usual, checks
1864 +every dir existing on all branches and rejects the request.
1866 +As a side effect, dirperm1 option improves the performance of aufs
1867 +because the number of permission check is reduced when the number of
1871 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1872 +----------------------------------------------------------------------
1873 +Generally VFS helpers re-validate a dentry as a part of lookup.
1874 +0. digging down the directory hierarchy.
1875 +1. lock the parent dir by its i_mutex.
1876 +2. lookup the final (child) entry.
1878 +4. call the actual operation (create, unlink, etc.)
1879 +5. unlock the parent dir
1881 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1882 +called. Actually aufs implements it and checks the dentry on a branch is
1884 +But it is not enough. Because aufs has to release the lock for the
1885 +parent dir on a branch at the end of ->lookup() (step 2) and
1886 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1888 +If the file on a branch is changed directly, eg. bypassing aufs, after
1889 +aufs released the lock, then the subsequent operation may cause
1890 +something unpleasant result.
1892 +This situation is a result of VFS architecture, ->lookup() and
1893 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1894 +design from VFS's point of view. It is just not suitable for sub-VFS
1897 +Aufs supports such case by three level of revalidation which is
1898 +selectable by user.
1899 +1. Simple Revalidate
1900 + Addition to the native flow in VFS's, confirm the child-parent
1901 + relationship on the branch just after locking the parent dir on the
1902 + branch in the "actual operation" (step 4). When this validation
1903 + fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1904 + checks the validation of the dentry on branches.
1905 +2. Monitor Changes Internally by Inotify/Fsnotify
1906 + Addition to above, in the "actual operation" (step 4) aufs re-lookup
1907 + the dentry on the branch, and returns EBUSY if it finds different
1909 + Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1910 + during it is in cache. When the event is notified, aufs registers a
1911 + function to kernel 'events' thread by schedule_work(). And the
1912 + function sets some special status to the cached aufs dentry and inode
1913 + private data. If they are not cached, then aufs has nothing to
1914 + do. When the same file is accessed through aufs (step 0-3) later,
1915 + aufs will detect the status and refresh all necessary data.
1916 + In this mode, aufs has to ignore the event which is fired by aufs
1918 +3. No Extra Validation
1919 + This is the simplest test and doesn't add any additional revalidation
1920 + test, and skip the revalidation in step 4. It is useful and improves
1921 + aufs performance when system surely hide the aufs branches from user,
1922 + by over-mounting something (or another method).
1923 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1924 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1925 +++ linux/Documentation/filesystems/aufs/design/04branch.txt 2018-01-29 07:56:20.049991637 +0100
1928 +# Copyright (C) 2005-2017 Junjiro R. Okajima
1930 +# This program is free software; you can redistribute it and/or modify
1931 +# it under the terms of the GNU General Public License as published by
1932 +# the Free Software Foundation; either version 2 of the License, or
1933 +# (at your option) any later version.
1935 +# This program is distributed in the hope that it will be useful,
1936 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1937 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1938 +# GNU General Public License for more details.
1940 +# You should have received a copy of the GNU General Public License
1941 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1943 +Branch Manipulation
1945 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1946 +and changing its permission/attribute, there are a lot of works to do.
1950 +----------------------------------------------------------------------
1951 +o Confirm the adding dir exists outside of aufs, including loopback
1952 + mount, and its various attributes.
1953 +o Initialize the xino file and whiteout bases if necessary.
1956 +o Check the owner/group/mode of the directory
1957 + When the owner/group/mode of the adding directory differs from the
1958 + existing branch, aufs issues a warning because it may impose a
1960 + For example, when a upper writable branch has a world writable empty
1961 + top directory, a malicious user can create any files on the writable
1962 + branch directly, like copy-up and modify manually. If something like
1963 + /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1964 + writable branch, and the writable branch is world-writable, then a
1965 + malicious guy may create /etc/passwd on the writable branch directly
1966 + and the infected file will be valid in aufs.
1967 + I am afraid it can be a security issue, but aufs can do nothing except
1968 + producing a warning.
1972 +----------------------------------------------------------------------
1973 +o Confirm the deleting branch is not busy
1974 + To be general, there is one merit to adopt "remount" interface to
1975 + manipulate branches. It is to discard caches. At deleting a branch,
1976 + aufs checks the still cached (and connected) dentries and inodes. If
1977 + there are any, then they are all in-use. An inode without its
1978 + corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1980 + For the cached one, aufs checks whether the same named entry exists on
1982 + If the cached one is a directory, because aufs provides a merged view
1983 + to users, as long as one dir is left on any branch aufs can show the
1984 + dir to users. In this case, the branch can be removed from aufs.
1985 + Otherwise aufs rejects deleting the branch.
1987 + If any file on the deleting branch is opened by aufs, then aufs
1991 +Modify the Permission of a Branch
1992 +----------------------------------------------------------------------
1993 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1996 +o rw --> ro: Confirm the modifying branch is not busy
1997 + Aufs rejects the request if any of these conditions are true.
1998 + - a file on the branch is mmap-ed.
1999 + - a regular file on the branch is opened for write and there is no
2000 + same named entry on the upper branch.
2001 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
2002 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
2003 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2018-01-29 07:56:20.049991637 +0100
2006 +# Copyright (C) 2005-2017 Junjiro R. Okajima
2008 +# This program is free software; you can redistribute it and/or modify
2009 +# it under the terms of the GNU General Public License as published by
2010 +# the Free Software Foundation; either version 2 of the License, or
2011 +# (at your option) any later version.
2013 +# This program is distributed in the hope that it will be useful,
2014 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2015 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2016 +# GNU General Public License for more details.
2018 +# You should have received a copy of the GNU General Public License
2019 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2021 +Policies to Select One among Multiple Writable Branches
2022 +----------------------------------------------------------------------
2023 +When the number of writable branch is more than one, aufs has to decide
2024 +the target branch for file creation or copy-up. By default, the highest
2025 +writable branch which has the parent (or ancestor) dir of the target
2026 +file is chosen (top-down-parent policy).
2027 +By user's request, aufs implements some other policies to select the
2028 +writable branch, for file creation several policies, round-robin,
2029 +most-free-space, and other policies. For copy-up, top-down-parent,
2030 +bottom-up-parent, bottom-up and others.
2032 +As expected, the round-robin policy selects the branch in circular. When
2033 +you have two writable branches and creates 10 new files, 5 files will be
2034 +created for each branch. mkdir(2) systemcall is an exception. When you
2035 +create 10 new directories, all will be created on the same branch.
2036 +And the most-free-space policy selects the one which has most free
2037 +space among the writable branches. The amount of free space will be
2038 +checked by aufs internally, and users can specify its time interval.
2040 +The policies for copy-up is more simple,
2041 +top-down-parent is equivalent to the same named on in create policy,
2042 +bottom-up-parent selects the writable branch where the parent dir
2043 +exists and the nearest upper one from the copyup-source,
2044 +bottom-up selects the nearest upper writable branch from the
2045 +copyup-source, regardless the existence of the parent dir.
2047 +There are some rules or exceptions to apply these policies.
2048 +- If there is a readonly branch above the policy-selected branch and
2049 + the parent dir is marked as opaque (a variation of whiteout), or the
2050 + target (creating) file is whiteout-ed on the upper readonly branch,
2051 + then the result of the policy is ignored and the target file will be
2052 + created on the nearest upper writable branch than the readonly branch.
2053 +- If there is a writable branch above the policy-selected branch and
2054 + the parent dir is marked as opaque or the target file is whiteouted
2055 + on the branch, then the result of the policy is ignored and the target
2056 + file will be created on the highest one among the upper writable
2057 + branches who has diropq or whiteout. In case of whiteout, aufs removes
2059 +- link(2) and rename(2) systemcalls are exceptions in every policy.
2060 + They try selecting the branch where the source exists as possible
2061 + since copyup a large file will take long time. If it can't be,
2062 + ie. the branch where the source exists is readonly, then they will
2063 + follow the copyup policy.
2064 +- There is an exception for rename(2) when the target exists.
2065 + If the rename target exists, aufs compares the index of the branches
2066 + where the source and the target exists and selects the higher
2067 + one. If the selected branch is readonly, then aufs follows the
2069 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot linux/Documentation/filesystems/aufs/design/06dirren.dot
2070 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.dot 1970-01-01 01:00:00.000000000 +0100
2071 +++ linux/Documentation/filesystems/aufs/design/06dirren.dot 2018-01-29 07:56:20.049991637 +0100
2074 +// to view this graph, run dot(1) command in GRAPHVIZ.
2078 +whinfo [label="detailed info file\n(lower_brid_root-hinum, h_inum, namelen, old name)"];
2082 +aufs_rename -> whinfo [label="store/remove"];
2085 +inode_list [label="h_inum list in branch\ncache"];
2088 +whinode [label="h_inum list file"];
2091 +brmgmt [label="br_add/del/mod/umount"];
2093 +brmgmt -> inode_list [label="create/remove"];
2094 +brmgmt -> whinode [label="load/store"];
2096 +inode_list -> whinode [style=dashed,dir=both];
2098 +aufs_rename -> inode_list [label="add/del"];
2100 +aufs_lookup -> inode_list [label="search"];
2102 +aufs_lookup -> whinfo [label="load/remove"];
2104 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt linux/Documentation/filesystems/aufs/design/06dirren.txt
2105 --- /usr/share/empty/Documentation/filesystems/aufs/design/06dirren.txt 1970-01-01 01:00:00.000000000 +0100
2106 +++ linux/Documentation/filesystems/aufs/design/06dirren.txt 2018-01-29 07:56:20.049991637 +0100
2109 +# Copyright (C) 2017 Junjiro R. Okajima
2111 +# This program is free software; you can redistribute it and/or modify
2112 +# it under the terms of the GNU General Public License as published by
2113 +# the Free Software Foundation; either version 2 of the License, or
2114 +# (at your option) any later version.
2116 +# This program is distributed in the hope that it will be useful,
2117 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2118 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2119 +# GNU General Public License for more details.
2121 +# You should have received a copy of the GNU General Public License
2122 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2124 +Special handling for renaming a directory (DIRREN)
2125 +----------------------------------------------------------------------
2126 +First, let's assume we have a simple usecase.
2130 +- /ro/dirA and /ro/dirA/file exist too
2131 +- there is no dirB on both branches
2132 +- a user issues rename("dirA", "dirB")
2134 +Now, what should aufs behave against this rename(2)?
2135 +There are a few possible cases.
2138 + since dirA exists on a readonly branch which cannot be renamed.
2140 + it is possible to copy-up dirA (only the dir itself), but the child
2141 + entries ("file" in this case) should not be. it must be a bad
2142 + approach to copy-up recursively.
2143 +C. returns a success.
2144 + even the branch /ro is readonly, aufs tries renaming it. Obviously it
2145 + is a violation of aufs' policy.
2146 +D. construct an extra information which indicates that /ro/dirA should
2147 + be handled as the name of dirB.
2148 + overlayfs has a similar feature called REDIRECT.
2150 +Until now, aufs implements the case B only which returns EXDEV, and
2151 +expects the userspace application behaves like mv(1) which tries
2152 +issueing rename(2) recursively.
2154 +A new aufs feature called DIRREN is introduced which implements the case
2155 +D. There are several "extra information" added.
2157 +1. detailed info per renamed directory
2158 + path: /rw/dirB/$AUFS_WH_DR_INFO_PFX.<lower branch-id>
2159 +2. the inode-number list of directories on a branch
2160 + path: /rw/dirB/$AUFS_WH_DR_BRHINO
2162 +The filename of "detailed info per directory" represents the lower
2163 +branch, and its format is
2164 +- a type of the branch id
2166 + + uuid (not implemented yet)
2169 +- the inode-number of the branch root dir
2171 +And it contains these info in a single regular file.
2173 +- branch's inode-number of the logically renamed dir
2174 +- the name of the before-renamed dir
2176 +The "detailed info per directory" file is created in aufs rename(2), and
2177 +loaded in any lookup.
2178 +The info is considered in lookup for the matching case only. Here
2179 +"matching" means that the root of branch (in the info filename) is same
2180 +to the current looking-up branch. After looking-up the before-renamed
2181 +name, the inode-number is compared. And the matched dentry is used.
2183 +The "inode-number list of directories" is a regular file which contains
2184 +simply the inode-numbers on the branch. The file is created or updated
2185 +in removing the branch, and loaded in adding the branch. Its lifetime is
2186 +equal to the branch.
2187 +The list is refered in lookup, and when the current target inode is
2188 +found in the list, the aufs tries loading the "detailed info per
2189 +directory" and get the changed and valid name of the dir.
2191 +Theoretically these "extra informaiton" may be able to be put into XATTR
2192 +in the dir inode. But aufs doesn't choose this way because
2193 +1. XATTR may not be supported by the branch (or its configuration)
2194 +2. XATTR may have its size limit.
2195 +3. XATTR may be less easy to convert than a regular file, when the
2196 + format of the info is changed in the future.
2197 +At the same time, I agree that the regular file approach is much slower
2198 +than XATTR approach. So, in the future, aufs may take the XATTR or other
2201 +This DIRREN feature is enabled by aufs configuration, and is activated
2202 +by a new mount option.
2204 +For the more complicated case, there is a work with UDBA option, which
2205 +is to dected the direct access to the branches (by-passing aufs) and to
2206 +maintain the cashes in aufs. Since a single cached aufs dentry may
2207 +contains two names, before- and after-rename, the name comparision in
2208 +UDBA handler may not work correctly. In this case, the behaviour will be
2209 +equivalen to udba=reval case.
2210 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
2211 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt 1970-01-01 01:00:00.000000000 +0100
2212 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt 2018-01-29 07:56:20.049991637 +0100
2215 +# Copyright (C) 2011-2017 Junjiro R. Okajima
2217 +# This program is free software; you can redistribute it and/or modify
2218 +# it under the terms of the GNU General Public License as published by
2219 +# the Free Software Foundation; either version 2 of the License, or
2220 +# (at your option) any later version.
2222 +# This program is distributed in the hope that it will be useful,
2223 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2224 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2225 +# GNU General Public License for more details.
2227 +# You should have received a copy of the GNU General Public License
2228 +# along with this program; if not, write to the Free Software
2229 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2232 +File-based Hierarchical Storage Management (FHSM)
2233 +----------------------------------------------------------------------
2234 +Hierarchical Storage Management (or HSM) is a well-known feature in the
2235 +storage world. Aufs provides this feature as file-based with multiple
2236 +writable branches, based upon the principle of "Colder, the Lower".
2237 +Here the word "colder" means that the less used files, and "lower" means
2238 +that the position in the order of the stacked branches vertically.
2239 +These multiple writable branches are prioritized, ie. the topmost one
2240 +should be the fastest drive and be used heavily.
2242 +o Characters in aufs FHSM story
2243 +- aufs itself and a new branch attribute.
2244 +- a new ioctl interface to move-down and to establish a connection with
2245 + the daemon ("move-down" is a converse of "copy-up").
2246 +- userspace tool and daemon.
2248 +The userspace daemon establishes a connection with aufs and waits for
2249 +the notification. The notified information is very similar to struct
2250 +statfs containing the number of consumed blocks and inodes.
2251 +When the consumed blocks/inodes of a branch exceeds the user-specified
2252 +upper watermark, the daemon activates its move-down process until the
2253 +consumed blocks/inodes reaches the user-specified lower watermark.
2255 +The actual move-down is done by aufs based upon the request from
2256 +user-space since we need to maintain the inode number and the internal
2257 +pointer arrays in aufs.
2259 +Currently aufs FHSM handles the regular files only. Additionally they
2260 +must not be hard-linked nor pseudo-linked.
2263 +o Cowork of aufs and the user-space daemon
2264 + During the userspace daemon established the connection, aufs sends a
2265 + small notification to it whenever aufs writes something into the
2266 + writable branch. But it may cost high since aufs issues statfs(2)
2267 + internally. So user can specify a new option to cache the
2268 + info. Actually the notification is controlled by these factors.
2269 + + the specified cache time.
2270 + + classified as "force" by aufs internally.
2271 + Until the specified time expires, aufs doesn't send the info
2272 + except the forced cases. When aufs decide forcing, the info is always
2273 + notified to userspace.
2274 + For example, the number of free inodes is generally large enough and
2275 + the shortage of it happens rarely. So aufs doesn't force the
2276 + notification when creating a new file, directory and others. This is
2277 + the typical case which aufs doesn't force.
2278 + When aufs writes the actual filedata and the files consumes any of new
2279 + blocks, the aufs forces notifying.
2282 +o Interfaces in aufs
2283 +- New branch attribute.
2285 + Specifies that the branch is managed by FHSM feature. In other word,
2286 + participant in the FHSM.
2287 + When nofhsm is set to the branch, it will not be the source/target
2288 + branch of the move-down operation. This attribute is set
2289 + independently from coo and moo attributes, and if you want full
2290 + FHSM, you should specify them as well.
2291 +- New mount option.
2293 + Specifies a second to suppress many less important info to be
2296 + + AUFS_CTL_FHSM_FD
2297 + create a new file descriptor which userspace can read the notification
2298 + (a subset of struct statfs) from aufs.
2299 +- Module parameter 'brs'
2300 + It has to be set to 1. Otherwise the new mount option 'fhsm' will not
2302 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
2303 + When there are two or more branches with fhsm attributes,
2304 + /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
2305 + terminates it. As a result of remounting and branch-manipulation, the
2306 + number of branches with fhsm attribute can be one. In this case,
2307 + /sbin/mount.aufs will terminate the user-space daemon.
2310 +Finally the operation is done as these steps in kernel-space.
2312 + + no one else is using the file.
2313 + + the file is not hard-linked.
2314 + + the file is not pseudo-linked.
2315 + + the file is a regular file.
2316 + + the parent dir is not opaqued.
2317 +- find the target writable branch.
2318 +- make sure the file is not whiteout-ed by the upper (than the target)
2320 +- make the parent dir on the target branch.
2321 +- mutex lock the inode on the branch.
2322 +- unlink the whiteout on the target branch (if exists).
2323 +- lookup and create the whiteout-ed temporary name on the target branch.
2324 +- copy the file as the whiteout-ed temporary name on the target branch.
2325 +- rename the whiteout-ed temporary name to the original name.
2326 +- unlink the file on the source branch.
2327 +- maintain the internal pointer array and the external inode number
2329 +- maintain the timestamps and other attributes of the parent dir and the
2332 +And of course, in every step, an error may happen. So the operation
2333 +should restore the original file state after an error happens.
2334 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
2335 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
2336 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2018-01-29 07:56:20.049991637 +0100
2339 +# Copyright (C) 2005-2017 Junjiro R. Okajima
2341 +# This program is free software; you can redistribute it and/or modify
2342 +# it under the terms of the GNU General Public License as published by
2343 +# the Free Software Foundation; either version 2 of the License, or
2344 +# (at your option) any later version.
2346 +# This program is distributed in the hope that it will be useful,
2347 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2348 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2349 +# GNU General Public License for more details.
2351 +# You should have received a copy of the GNU General Public License
2352 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2354 +mmap(2) -- File Memory Mapping
2355 +----------------------------------------------------------------------
2356 +In aufs, the file-mapped pages are handled by a branch fs directly, no
2357 +interaction with aufs. It means aufs_mmap() calls the branch fs's
2359 +This approach is simple and good, but there is one problem.
2360 +Under /proc, several entries show the mmapped files by its path (with
2361 +device and inode number), and the printed path will be the path on the
2362 +branch fs's instead of virtual aufs's.
2363 +This is not a problem in most cases, but some utilities lsof(1) (and its
2364 +user) may expect the path on aufs.
2366 +To address this issue, aufs adds a new member called vm_prfile in struct
2367 +vm_area_struct (and struct vm_region). The original vm_file points to
2368 +the file on the branch fs in order to handle everything correctly as
2369 +usual. The new vm_prfile points to a virtual file in aufs, and the
2370 +show-functions in procfs refers to vm_prfile if it is set.
2371 +Also we need to maintain several other places where touching vm_file
2373 +- fork()/clone() copies vma and the reference count of vm_file is
2375 +- merging vma maintains the ref count too.
2377 +This is not a good approach. It just fakes the printed path. But it
2378 +leaves all behaviour around f_mapping unchanged. This is surely an
2380 +Actually aufs had adopted another complicated approach which calls
2381 +generic_file_mmap() and handles struct vm_operations_struct. In this
2382 +approach, aufs met a hard problem and I could not solve it without
2383 +switching the approach.
2385 +There may be one more another approach which is
2386 +- bind-mount the branch-root onto the aufs-root internally
2387 +- grab the new vfsmount (ie. struct mount)
2388 +- lazy-umount the branch-root internally
2389 +- in open(2) the aufs-file, open the branch-file with the hidden
2390 + vfsmount (instead of the original branch's vfsmount)
2391 +- ideally this "bind-mount and lazy-umount" should be done atomically,
2392 + but it may be possible from userspace by the mount helper.
2394 +Adding the internal hidden vfsmount and using it in opening a file, the
2395 +file path under /proc will be printed correctly. This approach looks
2396 +smarter, but is not possible I am afraid.
2397 +- aufs-root may be bind-mount later. when it happens, another hidden
2398 + vfsmount will be required.
2399 +- it is hard to get the chance to bind-mount and lazy-umount
2400 + + in kernel-space, FS can have vfsmount in open(2) via
2401 + file->f_path, and aufs can know its vfsmount. But several locks are
2402 + already acquired, and if aufs tries to bind-mount and lazy-umount
2403 + here, then it may cause a deadlock.
2404 + + in user-space, bind-mount doesn't invoke the mount helper.
2405 +- since /proc shows dev and ino, aufs has to give vma these info. it
2406 + means a new member vm_prinode will be necessary. this is essentially
2407 + equivalent to vm_prfile described above.
2409 +I have to give up this "looks-smater" approach.
2410 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2411 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt 1970-01-01 01:00:00.000000000 +0100
2412 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt 2018-01-29 07:56:20.049991637 +0100
2415 +# Copyright (C) 2014-2017 Junjiro R. Okajima
2417 +# This program is free software; you can redistribute it and/or modify
2418 +# it under the terms of the GNU General Public License as published by
2419 +# the Free Software Foundation; either version 2 of the License, or
2420 +# (at your option) any later version.
2422 +# This program is distributed in the hope that it will be useful,
2423 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2424 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2425 +# GNU General Public License for more details.
2427 +# You should have received a copy of the GNU General Public License
2428 +# along with this program; if not, write to the Free Software
2429 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2432 +Listing XATTR/EA and getting the value
2433 +----------------------------------------------------------------------
2434 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2435 +shows the values from the topmost existing file. This behaviour is good
2436 +for the non-dir entries since the bahaviour exactly matches the shown
2437 +information. But for the directories, aufs considers all the same named
2438 +entries on the lower branches. Which means, if one of the lower entry
2439 +rejects readdir call, then aufs returns an error even if the topmost
2440 +entry allows it. This behaviour is necessary to respect the branch fs's
2441 +security, but can make users confused since the user-visible standard
2442 +attributes don't match the behaviour.
2443 +To address this issue, aufs has a mount option called dirperm1 which
2444 +checks the permission for the topmost entry only, and ignores the lower
2445 +entry's permission.
2447 +A similar issue can happen around XATTR.
2448 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2449 +always set. Otherwise these very unpleasant situation would happen.
2450 +- listxattr(2) may return the duplicated entries.
2451 +- users may not be able to remove or reset the XATTR forever,
2454 +XATTR/EA support in the internal (copy,move)-(up,down)
2455 +----------------------------------------------------------------------
2456 +Generally the extended attributes of inode are categorized as these.
2457 +- "security" for LSM and capability.
2458 +- "system" for posix ACL, 'acl' mount option is required for the branch
2460 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2461 +- "user" for userspace, 'user_xattr' mount option is required for the
2462 + branch fs generally.
2464 +Moreover there are some other categories. Aufs handles these rather
2465 +unpopular categories as the ordinary ones, ie. there is no special
2466 +condition nor exception.
2468 +In copy-up, the support for XATTR on the dst branch may differ from the
2469 +src branch. In this case, the copy-up operation will get an error and
2470 +the original user operation which triggered the copy-up will fail. It
2471 +can happen that even all copy-up will fail.
2472 +When both of src and dst branches support XATTR and if an error occurs
2473 +during copying XATTR, then the copy-up should fail obviously. That is a
2474 +good reason and aufs should return an error to userspace. But when only
2475 +the src branch support that XATTR, aufs should not return an error.
2476 +For example, the src branch supports ACL but the dst branch doesn't
2477 +because the dst branch may natively un-support it or temporary
2478 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2479 +may NOT return an error even if the XATTR is not supported. It is
2480 +totally up to the branch fs.
2482 +Anyway when the aufs internal copy-up gets an error from the dst branch
2483 +fs, then aufs tries removing the just copied entry and returns the error
2484 +to the userspace. The worst case of this situation will be all copy-up
2487 +For the copy-up operation, there two basic approaches.
2488 +- copy the specified XATTR only (by category above), and return the
2489 + error unconditionally if it happens.
2490 +- copy all XATTR, and ignore the error on the specified category only.
2492 +In order to support XATTR and to implement the correct behaviour, aufs
2493 +chooses the latter approach and introduces some new branch attributes,
2494 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2495 +They correspond to the XATTR namespaces (see above). Additionally, to be
2496 +convenient, "icex" is also provided which means all "icex*" attributes
2497 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2499 +The meaning of these attributes is to ignore the error from setting
2500 +XATTR on that branch.
2501 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2502 +error from the dst branch according to the specified attributes.
2504 +Some XATTR may have its default value. The default value may come from
2505 +the parent dir or the environment. If the default value is set at the
2506 +file creating-time, it will be overwritten by copy-up.
2507 +Some contradiction may happen I am afraid.
2508 +Do we need another attribute to stop copying XATTR? I am unsure. For
2509 +now, aufs implements the branch attributes to ignore the error.
2510 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2511 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2512 +++ linux/Documentation/filesystems/aufs/design/07export.txt 2018-01-29 07:56:20.049991637 +0100
2515 +# Copyright (C) 2005-2017 Junjiro R. Okajima
2517 +# This program is free software; you can redistribute it and/or modify
2518 +# it under the terms of the GNU General Public License as published by
2519 +# the Free Software Foundation; either version 2 of the License, or
2520 +# (at your option) any later version.
2522 +# This program is distributed in the hope that it will be useful,
2523 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2524 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2525 +# GNU General Public License for more details.
2527 +# You should have received a copy of the GNU General Public License
2528 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2530 +Export Aufs via NFS
2531 +----------------------------------------------------------------------
2532 +Here is an approach.
2533 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2535 +- iget_locked(): initialize aufs inode generation for a new inode, and
2536 + store it in xigen file.
2537 +- destroy_inode(): increment aufs inode generation and store it in xigen
2538 + file. it is necessary even if it is not unlinked, because any data of
2539 + inode may be changed by UDBA.
2540 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2541 + build file handle by
2542 + + branch id (4 bytes)
2543 + + superblock generation (4 bytes)
2544 + + inode number (4 or 8 bytes)
2545 + + parent dir inode number (4 or 8 bytes)
2546 + + inode generation (4 bytes))
2547 + + return value of exportfs_encode_fh() for the parent on a branch (4
2549 + + file handle for a branch (by exportfs_encode_fh())
2551 + + find the index of a branch from its id in handle, and check it is
2552 + still exist in aufs.
2553 + + 1st level: get the inode number from handle and search it in cache.
2554 + + 2nd level: if not found in cache, get the parent inode number from
2555 + the handle and search it in cache. and then open the found parent
2556 + dir, find the matching inode number by vfs_readdir() and get its
2557 + name, and call lookup_one_len() for the target dentry.
2558 + + 3rd level: if the parent dir is not cached, call
2559 + exportfs_decode_fh() for a branch and get the parent on a branch,
2560 + build a pathname of it, convert it a pathname in aufs, call
2561 + path_lookup(). now aufs gets a parent dir dentry, then handle it as
2563 + + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2564 + for every branch, but not itself. to get this, (currently) aufs
2565 + searches in current->nsproxy->mnt_ns list. it may not be a good
2566 + idea, but I didn't get other approach.
2567 + + test the generation of the gotten inode.
2568 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2569 + convert it into ESTALE for NFSD.
2570 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2571 + lookup_one_len(), vfs_getattr(), encode_fh() and others.
2572 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2573 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
2574 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2018-01-29 07:56:20.049991637 +0100
2577 +# Copyright (C) 2005-2017 Junjiro R. Okajima
2579 +# This program is free software; you can redistribute it and/or modify
2580 +# it under the terms of the GNU General Public License as published by
2581 +# the Free Software Foundation; either version 2 of the License, or
2582 +# (at your option) any later version.
2584 +# This program is distributed in the hope that it will be useful,
2585 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2586 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2587 +# GNU General Public License for more details.
2589 +# You should have received a copy of the GNU General Public License
2590 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2592 +Show Whiteout Mode (shwh)
2593 +----------------------------------------------------------------------
2594 +Generally aufs hides the name of whiteouts. But in some cases, to show
2595 +them is very useful for users. For instance, creating a new middle layer
2596 +(branch) by merging existing layers.
2598 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2599 +When you have three branches,
2600 +- Bottom: 'system', squashfs (underlying base system), read-only
2601 +- Middle: 'mods', squashfs, read-only
2602 +- Top: 'overlay', ram (tmpfs), read-write
2604 +The top layer is loaded at boot time and saved at shutdown, to preserve
2605 +the changes made to the system during the session.
2606 +When larger changes have been made, or smaller changes have accumulated,
2607 +the size of the saved top layer data grows. At this point, it would be
2608 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2609 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2610 +restoring save and load speed.
2612 +This merging is simplified by the use of another aufs mount, of just the
2613 +two overlay branches using the 'shwh' option.
2614 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2615 + aufs /livesys/merge_union
2617 +A merged view of these two branches is then available at
2618 +/livesys/merge_union, and the new feature is that the whiteouts are
2620 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2621 +writing to all branches. Also the default mode for all branches is 'ro'.
2622 +It is now possible to save the combined contents of the two overlay
2623 +branches to a new squashfs, e.g.:
2624 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2626 +This new squashfs archive can be stored on the boot device and the
2627 +initramfs will use it to replace the old one at the next boot.
2628 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2629 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
2630 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2018-01-29 07:56:20.049991637 +0100
2633 +# Copyright (C) 2010-2017 Junjiro R. Okajima
2635 +# This program is free software; you can redistribute it and/or modify
2636 +# it under the terms of the GNU General Public License as published by
2637 +# the Free Software Foundation; either version 2 of the License, or
2638 +# (at your option) any later version.
2640 +# This program is distributed in the hope that it will be useful,
2641 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2642 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2643 +# GNU General Public License for more details.
2645 +# You should have received a copy of the GNU General Public License
2646 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2648 +Dynamically customizable FS operations
2649 +----------------------------------------------------------------------
2650 +Generally FS operations (struct inode_operations, struct
2651 +address_space_operations, struct file_operations, etc.) are defined as
2652 +"static const", but it never means that FS have only one set of
2653 +operation. Some FS have multiple sets of them. For instance, ext2 has
2654 +three sets, one for XIP, for NOBH, and for normal.
2655 +Since aufs overrides and redirects these operations, sometimes aufs has
2656 +to change its behaviour according to the branch FS type. More importantly
2657 +VFS acts differently if a function (member in the struct) is set or
2658 +not. It means aufs should have several sets of operations and select one
2659 +among them according to the branch FS definition.
2661 +In order to solve this problem and not to affect the behaviour of VFS,
2662 +aufs defines these operations dynamically. For instance, aufs defines
2663 +dummy direct_IO function for struct address_space_operations, but it may
2664 +not be set to the address_space_operations actually. When the branch FS
2665 +doesn't have it, aufs doesn't set it to its address_space_operations
2666 +while the function definition itself is still alive. So the behaviour
2667 +itself will not change, and it will return an error when direct_IO is
2670 +The lifetime of these dynamically generated operation object is
2671 +maintained by aufs branch object. When the branch is removed from aufs,
2672 +the reference counter of the object is decremented. When it reaches
2673 +zero, the dynamically generated operation object will be freed.
2675 +This approach is designed to support AIO (io_submit), Direct I/O and
2677 +Currently this approach is applied to address_space_operations for
2678 +regular files only.
2679 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2680 --- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
2681 +++ linux/Documentation/filesystems/aufs/README 2018-01-29 07:56:20.046658203 +0100
2684 +Aufs4 -- advanced multi layered unification filesystem version 4.x
2690 +----------------------------------------
2691 +In the early days, aufs was entirely re-designed and re-implemented
2692 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2693 +improvements and implementations, it becomes totally different from
2694 +Unionfs while keeping the basic features.
2695 +Recently, Unionfs Version 2.x series begin taking some of the same
2696 +approaches to aufs1's.
2697 +Unionfs is being developed by Professor Erez Zadok at Stony Brook
2698 +University and his team.
2700 +Aufs4 supports linux-4.0 and later, and for linux-3.x series try aufs3.
2701 +If you want older kernel version support, try aufs2-2.6.git or
2702 +aufs2-standalone.git repository, aufs1 from CVS on SourceForge.
2704 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2705 + According to Christoph Hellwig, linux rejects all union-type
2706 + filesystems but UnionMount.
2707 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2709 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2710 + UnionMount, and he pointed out an issue around a directory mutex
2711 + lock and aufs addressed it. But it is still unsure whether aufs will
2712 + be merged (or any other union solution).
2713 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2717 +----------------------------------------
2718 +- unite several directories into a single virtual filesystem. The member
2719 + directory is called as a branch.
2720 +- you can specify the permission flags to the branch, which are 'readonly',
2721 + 'readwrite' and 'whiteout-able.'
2722 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2723 + readonly branch are modifiable logically.
2724 +- dynamic branch manipulation, add, del.
2727 +Also there are many enhancements in aufs, such as:
2728 +- test only the highest one for the directory permission (dirperm1)
2729 +- copyup on open (coo=)
2730 +- 'move' policy for copy-up between two writable branches, after
2731 + checking free space.
2733 +- readdir(3) in userspace.
2734 +- keep inode number by external inode number table
2735 +- keep the timestamps of file/dir in internal copyup operation
2736 +- seekable directory, supporting NFS readdir.
2737 +- whiteout is hardlinked in order to reduce the consumption of inodes
2739 +- do not copyup, nor create a whiteout when it is unnecessary
2740 +- revert a single systemcall when an error occurs in aufs
2741 +- remount interface instead of ioctl
2742 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2743 +- loopback mounted filesystem as a branch
2744 +- kernel thread for removing the dir who has a plenty of whiteouts
2745 +- support copyup sparse file (a file which has a 'hole' in it)
2746 +- default permission flags for branches
2747 +- selectable permission flags for ro branch, whether whiteout can
2750 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2751 +- support multiple writable branches, some policies to select one
2752 + among multiple writable branches.
2753 +- a new semantics for link(2) and rename(2) to support multiple
2754 + writable branches.
2755 +- no glibc changes are required.
2756 +- pseudo hardlink (hardlink over branches)
2757 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2758 + including NFS or remote filesystem branch.
2759 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2762 +Currently these features are dropped temporary from aufs4.
2763 +See design/08plan.txt in detail.
2764 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2766 +- statistics of aufs thread (/sys/fs/aufs/stat)
2768 +Features or just an idea in the future (see also design/*.txt),
2769 +- reorder the branch index without del/re-add.
2770 +- permanent xino files for NFSD
2771 +- an option for refreshing the opened files after add/del branches
2772 +- light version, without branch manipulation. (unnecessary?)
2773 +- copyup in userspace
2774 +- inotify in userspace
2779 +----------------------------------------
2780 +There are three GIT trees for aufs4, aufs4-linux.git,
2781 +aufs4-standalone.git, and aufs-util.git. Note that there is no "4" in
2783 +While the aufs-util is always necessary, you need either of aufs4-linux
2784 +or aufs4-standalone.
2786 +The aufs4-linux tree includes the whole linux mainline GIT tree,
2787 +git://git.kernel.org/.../torvalds/linux.git.
2788 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2789 +build aufs4 as an external kernel module.
2790 +Several extra patches are not included in this tree. Only
2791 +aufs4-standalone tree contains them. They are described in the later
2792 +section "Configuration and Compilation."
2794 +On the other hand, the aufs4-standalone tree has only aufs source files
2795 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2796 +But you need to apply all aufs patches manually.
2798 +You will find GIT branches whose name is in form of "aufs4.x" where "x"
2799 +represents the linux kernel version, "linux-4.x". For instance,
2800 +"aufs4.0" is for linux-4.0. For latest "linux-4.x-rcN", use
2801 +"aufs4.x-rcN" branch.
2804 +$ git clone --reference /your/linux/git/tree \
2805 + git://github.com/sfjro/aufs4-linux.git aufs4-linux.git
2806 +- if you don't have linux GIT tree, then remove "--reference ..."
2807 +$ cd aufs4-linux.git
2808 +$ git checkout origin/aufs4.0
2810 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2811 +leave the patch-work to GIT.
2812 +$ cd /your/linux/git/tree
2813 +$ git remote add aufs4 git://github.com/sfjro/aufs4-linux.git
2815 +$ git checkout -b my4.0 v4.0
2816 +$ (add your local change...)
2817 +$ git pull aufs4 aufs4.0
2818 +- now you have v4.0 + your_changes + aufs4.0 in you my4.0 branch.
2819 +- you may need to solve some conflicts between your_changes and
2820 + aufs4.0. in this case, git-rerere is recommended so that you can
2821 + solve the similar conflicts automatically when you upgrade to 4.1 or
2822 + later in the future.
2824 +o aufs4-standalone tree
2825 +$ git clone git://github.com/sfjro/aufs4-standalone.git aufs4-standalone.git
2826 +$ cd aufs4-standalone.git
2827 +$ git checkout origin/aufs4.0
2830 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2831 +- note that the public aufs-util.git is on SourceForge instead of
2834 +$ git checkout origin/aufs4.0
2836 +Note: The 4.x-rcN branch is to be used with `rc' kernel versions ONLY.
2837 +The minor version number, 'x' in '4.x', of aufs may not always
2838 +follow the minor version number of the kernel.
2839 +Because changes in the kernel that cause the use of a new
2840 +minor version number do not always require changes to aufs-util.
2842 +Since aufs-util has its own minor version number, you may not be
2843 +able to find a GIT branch in aufs-util for your kernel's
2844 +exact minor version number.
2845 +In this case, you should git-checkout the branch for the
2846 +nearest lower number.
2848 +For (an unreleased) example:
2849 +If you are using "linux-4.10" and the "aufs4.10" branch
2850 +does not exist in aufs-util repository, then "aufs4.9", "aufs4.8"
2851 +or something numerically smaller is the branch for your kernel.
2853 +Also you can view all branches by
2857 +3. Configuration and Compilation
2858 +----------------------------------------
2859 +Make sure you have git-checkout'ed the correct branch.
2861 +For aufs4-linux tree,
2862 +- enable CONFIG_AUFS_FS.
2863 +- set other aufs configurations if necessary.
2865 +For aufs4-standalone tree,
2866 +There are several ways to build.
2869 +- apply ./aufs4-kbuild.patch to your kernel source files.
2870 +- apply ./aufs4-base.patch too.
2871 +- apply ./aufs4-mmap.patch too.
2872 +- apply ./aufs4-standalone.patch too, if you have a plan to set
2873 + CONFIG_AUFS_FS=m. otherwise you don't need ./aufs4-standalone.patch.
2874 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2875 + kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2876 +- enable CONFIG_AUFS_FS, you can select either
2878 +- and build your kernel as usual.
2879 +- install the built kernel.
2880 + Note: Since linux-3.9, every filesystem module requires an alias
2881 + "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2882 + modules.aliases file if you set CONFIG_AUFS_FS=m.
2883 +- install the header files too by "make headers_install" to the
2884 + directory where you specify. By default, it is $PWD/usr.
2885 + "make help" shows a brief note for headers_install.
2886 +- and reboot your system.
2889 +- module only (CONFIG_AUFS_FS=m).
2890 +- apply ./aufs4-base.patch to your kernel source files.
2891 +- apply ./aufs4-mmap.patch too.
2892 +- apply ./aufs4-standalone.patch too.
2893 +- build your kernel, don't forget "make headers_install", and reboot.
2894 +- edit ./config.mk and set other aufs configurations if necessary.
2895 + Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2896 + every aufs configurations.
2897 +- build the module by simple "make".
2898 + Note: Since linux-3.9, every filesystem module requires an alias
2899 + "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2900 + modules.aliases file.
2901 +- you can specify ${KDIR} make variable which points to your kernel
2903 +- install the files
2904 + + run "make install" to install the aufs module, or copy the built
2905 + $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2906 + + run "make install_headers" (instead of headers_install) to install
2907 + the modified aufs header file (you can specify DESTDIR which is
2908 + available in aufs standalone version's Makefile only), or copy
2909 + $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2910 + you like manually. By default, the target directory is $PWD/usr.
2911 +- no need to apply aufs4-kbuild.patch, nor copying source files to your
2912 + kernel source tree.
2914 +Note: The header file aufs_type.h is necessary to build aufs-util
2915 + as well as "make headers_install" in the kernel source tree.
2916 + headers_install is subject to be forgotten, but it is essentially
2917 + necessary, not only for building aufs-util.
2918 + You may not meet problems without headers_install in some older
2922 +- read README in aufs-util, build and install it
2923 +- note that your distribution may contain an obsoleted version of
2924 + aufs_type.h in /usr/include/linux or something. When you build aufs
2925 + utilities, make sure that your compiler refers the correct aufs header
2926 + file which is built by "make headers_install."
2927 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2928 + then run "make install_ulib" too. And refer to the aufs manual in
2931 +There several other patches in aufs4-standalone.git. They are all
2932 +optional. When you meet some problems, they will help you.
2933 +- aufs4-loopback.patch
2934 + Supports a nested loopback mount in a branch-fs. This patch is
2935 + unnecessary until aufs produces a message like "you may want to try
2936 + another patch for loopback file".
2938 + Modifies a system global kernel internal function get_next_ino() in
2939 + order to stop assigning 0 for an inode-number. Not directly related to
2940 + aufs, but recommended generally.
2942 + Keeps the tmpfs inode number as the lowest value. Effective to reduce
2943 + the size of aufs XINO files for tmpfs branch. Also it prevents the
2944 + duplication of inode number, which is important for backup tools and
2945 + other utilities. When you find aufs XINO files for tmpfs branch
2946 + growing too much, try this patch.
2947 +- lockdep-debug.patch
2948 + Because aufs is not only an ordinary filesystem (callee of VFS), but
2949 + also a caller of VFS functions for branch filesystems, subclassing of
2950 + the internal locks for LOCKDEP is necessary. LOCKDEP is a debugging
2951 + feature of linux kernel. If you enable CONFIG_LOCKDEP, then you will
2952 + need to apply this debug patch to expand several constant values.
2953 + If don't know what LOCKDEP, then you don't have apply this patch.
2957 +----------------------------------------
2958 +At first, make sure aufs-util are installed, and please read the aufs
2959 +manual, aufs.5 in aufs-util.git tree.
2963 +$ mkdir /tmp/rw /tmp/aufs
2964 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2966 +Here is another example. The result is equivalent.
2967 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2969 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2970 +# mount -o remount,append:${HOME} /tmp/aufs
2972 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2973 +you modify a file under /tmp/aufs, the one on your home directory is
2974 +not affected, instead the same named file will be newly created under
2975 +/tmp/rw. And all of your modification to a file will be applied to
2976 +the one under /tmp/rw. This is called the file based Copy on Write
2978 +Aufs mount options are described in aufs.5.
2979 +If you run chroot or something and make your aufs as a root directory,
2980 +then you need to customize the shutdown script. See the aufs manual in
2983 +Additionally, there are some sample usages of aufs which are a
2984 +diskless system with network booting, and LiveCD over NFS.
2985 +See sample dir in CVS tree on SourceForge.
2989 +----------------------------------------
2990 +When you have any problems or strange behaviour in aufs, please let me
2992 +- /proc/mounts (instead of the output of mount(8))
2993 +- /sys/module/aufs/*
2994 +- /sys/fs/aufs/* (if you have them)
2995 +- /debug/aufs/* (if you have them)
2996 +- linux kernel version
2997 + if your kernel is not plain, for example modified by distributor,
2998 + the url where i can download its source is necessary too.
2999 +- aufs version which was printed at loading the module or booting the
3000 + system, instead of the date you downloaded.
3001 +- configuration (define/undefine CONFIG_AUFS_xxx)
3002 +- kernel configuration or /proc/config.gz (if you have it)
3003 +- behaviour which you think to be incorrect
3004 +- actual operation, reproducible one is better
3005 +- mailto: aufs-users at lists.sourceforge.net
3007 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
3008 +and Feature Requests) on SourceForge. Please join and write to
3012 +6. Acknowledgements
3013 +----------------------------------------
3014 +Thanks to everyone who have tried and are using aufs, whoever
3015 +have reported a bug or any feedback.
3017 +Especially donators:
3018 +Tomas Matejicek(slax.org) made a donation (much more than once).
3019 + Since Apr 2010, Tomas M (the author of Slax and Linux Live
3020 + scripts) is making "doubling" donations.
3021 + Unfortunately I cannot list all of the donators, but I really
3023 + It ends Aug 2010, but the ordinary donation URL is still available.
3024 + <http://sourceforge.net/donate/index.php?group_id=167503>
3025 +Dai Itasaka made a donation (2007/8).
3026 +Chuck Smith made a donation (2008/4, 10 and 12).
3027 +Henk Schoneveld made a donation (2008/9).
3028 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
3029 +Francois Dupoux made a donation (2008/11).
3030 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
3031 + aufs2 GIT tree (2009/2).
3032 +William Grant made a donation (2009/3).
3033 +Patrick Lane made a donation (2009/4).
3034 +The Mail Archive (mail-archive.com) made donations (2009/5).
3035 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
3036 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
3037 +Pavel Pronskiy made a donation (2011/2).
3038 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
3039 + Networks (Ed Wildgoose) made a donation for hardware (2011/3).
3040 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
3042 +Sam Liddicott made a donation (2011/9).
3043 +Era Scarecrow made a donation (2013/4).
3044 +Bor Ratajc made a donation (2013/4).
3045 +Alessandro Gorreta made a donation (2013/4).
3046 +POIRETTE Marc made a donation (2013/4).
3047 +Alessandro Gorreta made a donation (2013/4).
3048 +lauri kasvandik made a donation (2013/5).
3049 +"pemasu from Finland" made a donation (2013/7).
3050 +The Parted Magic Project made a donation (2013/9 and 11).
3051 +Pavel Barta made a donation (2013/10).
3052 +Nikolay Pertsev made a donation (2014/5).
3053 +James B made a donation (2014/7 and 2015/7).
3054 +Stefano Di Biase made a donation (2014/8).
3055 +Daniel Epellei made a donation (2015/1).
3056 +OmegaPhil made a donation (2016/1).
3057 +Tomasz Szewczyk made a donation (2016/4).
3058 +James Burry made a donation (2016/12).
3060 +Thank you very much.
3061 +Donations are always, including future donations, very important and
3062 +helpful for me to keep on developing aufs.
3066 +----------------------------------------
3067 +If you are an experienced user, no explanation is needed. Aufs is
3068 +just a linux filesystem.
3073 +# Local variables: ;
3076 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
3077 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
3078 +++ linux/fs/aufs/aufs.h 2018-01-29 07:56:20.049991637 +0100
3081 + * Copyright (C) 2005-2017 Junjiro R. Okajima
3083 + * This program, aufs is free software; you can redistribute it and/or modify
3084 + * it under the terms of the GNU General Public License as published by
3085 + * the Free Software Foundation; either version 2 of the License, or
3086 + * (at your option) any later version.
3088 + * This program is distributed in the hope that it will be useful,
3089 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
3090 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
3091 + * GNU General Public License for more details.
3093 + * You should have received a copy of the GNU General Public License
3094 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
3098 + * all header files
3106 +#define AuStub(type, name, body, ...) \
3107 + static inline type name(__VA_ARGS__) { body; }
3109 +#define AuStubVoid(name, ...) \
3110 + AuStub(void, name, , __VA_ARGS__)
3111 +#define AuStubInt0(name, ...) \
3112 + AuStub(int, name, return 0, __VA_ARGS__)
3116 +#include "branch.h"
3119 +#include "dbgaufs.h"
3120 +#include "dentry.h"
3122 +#include "dirren.h"
3125 +#include "fstype.h"
3129 +#include "module.h"
3133 +#include "sysaufs.h"
3138 +#endif /* __KERNEL__ */
3139 +#endif /* __AUFS_H__ */
3140 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
3141 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
3142 +++ linux/fs/aufs/branch.c 2018-01-29 07:56:20.049991637 +0100
3145 + * Copyright (C) 2005-2017 Junjiro R. Okajima
3147 + * This program, aufs is free software; you can redistribute it and/or modify
3148 + * it under the terms of the GNU General Public License as published by
3149 + * the Free Software Foundation; either version 2 of the License, or
3150 + * (at your option) any later version.
3152 + * This program is distributed in the hope that it will be useful,
3153 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
3154 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
3155 + * GNU General Public License for more details.
3157 + * You should have received a copy of the GNU General Public License
3158 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
3162 + * branch management
3165 +#include <linux/compat.h>
3166 +#include <linux/statfs.h>
3170 + * free a single branch
3172 +static void au_br_do_free(struct au_branch *br)
3175 + struct au_wbr *wbr;
3176 + struct au_dykey **key;
3178 + au_hnotify_fin_br(br);
3179 + /* always, regardless the mount option */
3180 + au_dr_hino_free(&br->br_dirren);
3182 + if (br->br_xino.xi_file)
3183 + fput(br->br_xino.xi_file);
3184 + for (i = br->br_xino.xi_nondir.total - 1; i >= 0; i--)
3185 + AuDebugOn(br->br_xino.xi_nondir.array[i]);
3186 + kfree(br->br_xino.xi_nondir.array);
3188 + AuDebugOn(au_br_count(br));
3189 + au_br_count_fin(br);
3193 + for (i = 0; i < AuBrWh_Last; i++)
3194 + dput(wbr->wbr_wh[i]);
3195 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
3196 + AuRwDestroy(&wbr->wbr_wh_rwsem);
3199 + if (br->br_fhsm) {
3200 + au_br_fhsm_fin(br->br_fhsm);
3201 + kfree(br->br_fhsm);
3204 + key = br->br_dykey;
3205 + for (i = 0; i < AuBrDynOp; i++, key++)
3211 + /* recursive lock, s_umount of branch's */
3213 + path_put(&br->br_path);
3220 + * frees all branches
3222 +void au_br_free(struct au_sbinfo *sbinfo)
3224 + aufs_bindex_t bmax;
3225 + struct au_branch **br;
3227 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3229 + bmax = sbinfo->si_bbot + 1;
3230 + br = sbinfo->si_branch;
3232 + au_br_do_free(*br++);
3236 + * find the index of a branch which is specified by @br_id.
3238 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
3240 + aufs_bindex_t bindex, bbot;
3242 + bbot = au_sbbot(sb);
3243 + for (bindex = 0; bindex <= bbot; bindex++)
3244 + if (au_sbr_id(sb, bindex) == br_id)
3249 +/* ---------------------------------------------------------------------- */
3255 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
3256 + struct dentry *h_root)
3258 + if (unlikely(h_adding == h_root
3259 + || au_test_loopback_overlap(sb, h_adding)))
3261 + if (h_adding->d_sb != h_root->d_sb)
3263 + return au_test_subdir(h_adding, h_root)
3264 + || au_test_subdir(h_root, h_adding);
3268 + * returns a newly allocated branch. @new_nbranch is a number of branches
3269 + * after adding a branch.
3271 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
3274 + struct au_branch *add_branch;
3275 + struct dentry *root;
3276 + struct inode *inode;
3280 + add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
3281 + if (unlikely(!add_branch))
3283 + add_branch->br_xino.xi_nondir.total = 8; /* initial size */
3284 + add_branch->br_xino.xi_nondir.array
3285 + = kcalloc(add_branch->br_xino.xi_nondir.total, sizeof(ino_t),
3287 + if (unlikely(!add_branch->br_xino.xi_nondir.array))
3290 + err = au_hnotify_init_br(add_branch, perm);
3291 + if (unlikely(err))
3292 + goto out_xinondir;
3294 + if (au_br_writable(perm)) {
3295 + /* may be freed separately at changing the branch permission */
3296 + add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
3298 + if (unlikely(!add_branch->br_wbr))
3302 + if (au_br_fhsm(perm)) {
3303 + err = au_fhsm_br_alloc(add_branch);
3304 + if (unlikely(err))
3308 + root = sb->s_root;
3309 + err = au_sbr_realloc(au_sbi(sb), new_nbranch, /*may_shrink*/0);
3311 + err = au_di_realloc(au_di(root), new_nbranch, /*may_shrink*/0);
3313 + inode = d_inode(root);
3314 + err = au_hinode_realloc(au_ii(inode), new_nbranch,
3318 + return add_branch; /* success */
3321 + kfree(add_branch->br_wbr);
3323 + au_hnotify_fin_br(add_branch);
3325 + kfree(add_branch->br_xino.xi_nondir.array);
3327 + kfree(add_branch);
3329 + return ERR_PTR(err);
3333 + * test if the branch permission is legal or not.
3335 +static int test_br(struct inode *inode, int brperm, char *path)
3339 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
3344 + pr_err("write permission for readonly mount or inode, %s\n", path);
3352 + * 0: success, the caller will add it
3353 + * plus: success, it is already unified, the caller should ignore it
3356 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
3359 + aufs_bindex_t bbot, bindex;
3360 + struct dentry *root, *h_dentry;
3361 + struct inode *inode, *h_inode;
3363 + root = sb->s_root;
3364 + bbot = au_sbbot(sb);
3365 + if (unlikely(bbot >= 0
3366 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
3370 + pr_err("%s duplicated\n", add->pathname);
3375 + err = -ENOSPC; /* -E2BIG; */
3376 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
3377 + || AUFS_BRANCH_MAX - 1 <= bbot)) {
3378 + pr_err("number of branches exceeded %s\n", add->pathname);
3383 + if (unlikely(add->bindex < 0 || bbot + 1 < add->bindex)) {
3384 + pr_err("bad index %d\n", add->bindex);
3388 + inode = d_inode(add->path.dentry);
3390 + if (unlikely(!inode->i_nlink)) {
3391 + pr_err("no existence %s\n", add->pathname);
3396 + if (unlikely(inode->i_sb == sb)) {
3397 + pr_err("%s must be outside\n", add->pathname);
3401 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
3402 + pr_err("unsupported filesystem, %s (%s)\n",
3403 + add->pathname, au_sbtype(inode->i_sb));
3407 + if (unlikely(inode->i_sb->s_stack_depth)) {
3408 + pr_err("already stacked, %s (%s)\n",
3409 + add->pathname, au_sbtype(inode->i_sb));
3413 + err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3414 + if (unlikely(err))
3418 + return 0; /* success */
3421 + for (bindex = 0; bindex <= bbot; bindex++)
3422 + if (unlikely(test_overlap(sb, add->path.dentry,
3423 + au_h_dptr(root, bindex)))) {
3424 + pr_err("%s is overlapped\n", add->pathname);
3429 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3430 + h_dentry = au_h_dptr(root, 0);
3431 + h_inode = d_inode(h_dentry);
3432 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3433 + || !uid_eq(h_inode->i_uid, inode->i_uid)
3434 + || !gid_eq(h_inode->i_gid, inode->i_gid))
3435 + pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3437 + i_uid_read(inode), i_gid_read(inode),
3438 + (inode->i_mode & S_IALLUGO),
3439 + i_uid_read(h_inode), i_gid_read(h_inode),
3440 + (h_inode->i_mode & S_IALLUGO));
3448 + * initialize or clean the whiteouts for an adding branch
3450 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3453 + int err, old_perm;
3454 + aufs_bindex_t bindex;
3455 + struct inode *h_inode;
3456 + struct au_wbr *wbr;
3457 + struct au_hinode *hdir;
3458 + struct dentry *h_dentry;
3460 + err = vfsub_mnt_want_write(au_br_mnt(br));
3461 + if (unlikely(err))
3465 + old_perm = br->br_perm;
3466 + br->br_perm = new_perm;
3469 + bindex = au_br_index(sb, br->br_id);
3470 + if (0 <= bindex) {
3471 + hdir = au_hi(d_inode(sb->s_root), bindex);
3472 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
3474 + h_dentry = au_br_dentry(br);
3475 + h_inode = d_inode(h_dentry);
3476 + inode_lock_nested(h_inode, AuLsc_I_PARENT);
3479 + err = au_wh_init(br, sb);
3481 + wbr_wh_write_lock(wbr);
3482 + err = au_wh_init(br, sb);
3483 + wbr_wh_write_unlock(wbr);
3486 + au_hn_inode_unlock(hdir);
3488 + inode_unlock(h_inode);
3489 + vfsub_mnt_drop_write(au_br_mnt(br));
3490 + br->br_perm = old_perm;
3492 + if (!err && wbr && !au_br_writable(new_perm)) {
3494 + br->br_wbr = NULL;
3501 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3505 + struct kstatfs kst;
3506 + struct au_wbr *wbr;
3509 + au_rw_init(&wbr->wbr_wh_rwsem);
3510 + atomic_set(&wbr->wbr_wh_running, 0);
3513 + * a limit for rmdir/rename a dir
3514 + * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3516 + err = vfs_statfs(&br->br_path, &kst);
3517 + if (unlikely(err))
3520 + if (kst.f_namelen >= NAME_MAX)
3521 + err = au_br_init_wh(sb, br, perm);
3523 + pr_err("%pd(%s), unsupported namelen %ld\n",
3525 + au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3531 +/* initialize a new branch */
3532 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3533 + struct au_opt_add *add)
3536 + struct inode *h_inode;
3539 + spin_lock_init(&br->br_xino.xi_nondir.spin);
3540 + init_waitqueue_head(&br->br_xino.xi_nondir.wqh);
3541 + br->br_perm = add->perm;
3542 + br->br_path = add->path; /* set first, path_get() later */
3543 + spin_lock_init(&br->br_dykey_lock);
3544 + au_br_count_init(br);
3545 + atomic_set(&br->br_xino_running, 0);
3546 + br->br_id = au_new_br_id(sb);
3547 + AuDebugOn(br->br_id < 0);
3549 + /* always, regardless the given option */
3550 + err = au_dr_br_init(sb, br, &add->path);
3551 + if (unlikely(err))
3554 + if (au_br_writable(add->perm)) {
3555 + err = au_wbr_init(br, sb, add->perm);
3556 + if (unlikely(err))
3560 + if (au_opt_test(au_mntflags(sb), XINO)) {
3561 + h_inode = d_inode(add->path.dentry);
3562 + err = au_xino_br(sb, br, h_inode->i_ino,
3563 + au_sbr(sb, 0)->br_xino.xi_file, /*do_test*/1);
3564 + if (unlikely(err)) {
3565 + AuDebugOn(br->br_xino.xi_file);
3570 + sysaufs_br_init(br);
3571 + path_get(&br->br_path);
3572 + goto out; /* success */
3575 + memset(&br->br_path, 0, sizeof(br->br_path));
3580 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3581 + struct au_branch *br, aufs_bindex_t bbot,
3582 + aufs_bindex_t amount)
3584 + struct au_branch **brp;
3586 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3588 + brp = sbinfo->si_branch + bindex;
3589 + memmove(brp + 1, brp, sizeof(*brp) * amount);
3591 + sbinfo->si_bbot++;
3592 + if (unlikely(bbot < 0))
3593 + sbinfo->si_bbot = 0;
3596 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3597 + aufs_bindex_t bbot, aufs_bindex_t amount)
3599 + struct au_hdentry *hdp;
3601 + AuRwMustWriteLock(&dinfo->di_rwsem);
3603 + hdp = au_hdentry(dinfo, bindex);
3604 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3605 + au_h_dentry_init(hdp);
3607 + if (unlikely(bbot < 0))
3608 + dinfo->di_btop = 0;
3611 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3612 + aufs_bindex_t bbot, aufs_bindex_t amount)
3614 + struct au_hinode *hip;
3616 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3618 + hip = au_hinode(iinfo, bindex);
3619 + memmove(hip + 1, hip, sizeof(*hip) * amount);
3620 + au_hinode_init(hip);
3622 + if (unlikely(bbot < 0))
3623 + iinfo->ii_btop = 0;
3626 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3627 + aufs_bindex_t bindex)
3629 + struct dentry *root, *h_dentry;
3630 + struct inode *root_inode, *h_inode;
3631 + aufs_bindex_t bbot, amount;
3633 + root = sb->s_root;
3634 + root_inode = d_inode(root);
3635 + bbot = au_sbbot(sb);
3636 + amount = bbot + 1 - bindex;
3637 + h_dentry = au_br_dentry(br);
3638 + au_sbilist_lock();
3639 + au_br_do_add_brp(au_sbi(sb), bindex, br, bbot, amount);
3640 + au_br_do_add_hdp(au_di(root), bindex, bbot, amount);
3641 + au_br_do_add_hip(au_ii(root_inode), bindex, bbot, amount);
3642 + au_set_h_dptr(root, bindex, dget(h_dentry));
3643 + h_inode = d_inode(h_dentry);
3644 + au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3645 + au_sbilist_unlock();
3648 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3651 + aufs_bindex_t bbot, add_bindex;
3652 + struct dentry *root, *h_dentry;
3653 + struct inode *root_inode;
3654 + struct au_branch *add_branch;
3656 + root = sb->s_root;
3657 + root_inode = d_inode(root);
3658 + IMustLock(root_inode);
3659 + IiMustWriteLock(root_inode);
3660 + err = test_add(sb, add, remount);
3661 + if (unlikely(err < 0))
3665 + goto out; /* success */
3668 + bbot = au_sbbot(sb);
3669 + add_branch = au_br_alloc(sb, bbot + 2, add->perm);
3670 + err = PTR_ERR(add_branch);
3671 + if (IS_ERR(add_branch))
3674 + err = au_br_init(add_branch, sb, add);
3675 + if (unlikely(err)) {
3676 + au_br_do_free(add_branch);
3680 + add_bindex = add->bindex;
3682 + au_br_do_add(sb, add_branch, add_bindex);
3684 + sysaufs_brs_del(sb, add_bindex);
3685 + au_br_do_add(sb, add_branch, add_bindex);
3686 + sysaufs_brs_add(sb, add_bindex);
3689 + h_dentry = add->path.dentry;
3690 + if (!add_bindex) {
3691 + au_cpup_attr_all(root_inode, /*force*/1);
3692 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3694 + au_add_nlink(root_inode, d_inode(h_dentry));
3697 + * this test/set prevents aufs from handling unnecesary notify events
3698 + * of xino files, in case of re-adding a writable branch which was
3699 + * once detached from aufs.
3701 + if (au_xino_brid(sb) < 0
3702 + && au_br_writable(add_branch->br_perm)
3703 + && !au_test_fs_bad_xino(h_dentry->d_sb)
3704 + && add_branch->br_xino.xi_file
3705 + && add_branch->br_xino.xi_file->f_path.dentry->d_parent == h_dentry)
3706 + au_xino_brid_set(sb, add_branch->br_id);
3712 +/* ---------------------------------------------------------------------- */
3714 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3715 + unsigned long long max __maybe_unused,
3718 + unsigned long long n;
3719 + struct file **p, *f;
3720 + struct hlist_bl_head *files;
3721 + struct hlist_bl_node *pos;
3722 + struct au_finfo *finfo;
3726 + files = &au_sbi(sb)->si_files;
3727 + hlist_bl_lock(files);
3728 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
3729 + f = finfo->fi_file;
3731 + && !special_file(file_inode(f)->i_mode)) {
3735 + AuDebugOn(n > max);
3738 + hlist_bl_unlock(files);
3743 +static struct file **au_farray_alloc(struct super_block *sb,
3744 + unsigned long long *max)
3746 + *max = au_nfiles(sb);
3747 + return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3750 +static void au_farray_free(struct file **a, unsigned long long max)
3752 + unsigned long long ull;
3754 + for (ull = 0; ull < max; ull++)
3760 +/* ---------------------------------------------------------------------- */
3766 +/* to show the line number, do not make it inlined function */
3767 +#define AuVerbose(do_info, fmt, ...) do { \
3769 + pr_info(fmt, ##__VA_ARGS__); \
3772 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t btop,
3773 + aufs_bindex_t bbot)
3775 + return (inode && !S_ISDIR(inode->i_mode)) || btop == bbot;
3778 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t btop,
3779 + aufs_bindex_t bbot)
3781 + return au_test_ibusy(d_inode(dentry), btop, bbot);
3785 + * test if the branch is deletable or not.
3787 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3788 + unsigned int sigen, const unsigned int verbose)
3790 + int err, i, j, ndentry;
3791 + aufs_bindex_t btop, bbot;
3792 + struct au_dcsub_pages dpages;
3793 + struct au_dpage *dpage;
3796 + err = au_dpages_init(&dpages, GFP_NOFS);
3797 + if (unlikely(err))
3799 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
3800 + if (unlikely(err))
3803 + for (i = 0; !err && i < dpages.ndpage; i++) {
3804 + dpage = dpages.dpages + i;
3805 + ndentry = dpage->ndentry;
3806 + for (j = 0; !err && j < ndentry; j++) {
3807 + d = dpage->dentries[j];
3808 + AuDebugOn(au_dcount(d) <= 0);
3809 + if (!au_digen_test(d, sigen)) {
3810 + di_read_lock_child(d, AuLock_IR);
3811 + if (unlikely(au_dbrange_test(d))) {
3812 + di_read_unlock(d, AuLock_IR);
3816 + di_write_lock_child(d);
3817 + if (unlikely(au_dbrange_test(d))) {
3818 + di_write_unlock(d);
3821 + err = au_reval_dpath(d, sigen);
3823 + di_downgrade_lock(d, AuLock_IR);
3825 + di_write_unlock(d);
3830 + /* AuDbgDentry(d); */
3831 + btop = au_dbtop(d);
3832 + bbot = au_dbbot(d);
3833 + if (btop <= bindex
3835 + && au_h_dptr(d, bindex)
3836 + && au_test_dbusy(d, btop, bbot)) {
3838 + AuVerbose(verbose, "busy %pd\n", d);
3841 + di_read_unlock(d, AuLock_IR);
3846 + au_dpages_free(&dpages);
3851 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3852 + unsigned int sigen, const unsigned int verbose)
3855 + unsigned long long max, ull;
3856 + struct inode *i, **array;
3857 + aufs_bindex_t btop, bbot;
3859 + array = au_iarray_alloc(sb, &max);
3860 + err = PTR_ERR(array);
3861 + if (IS_ERR(array))
3865 + AuDbg("b%d\n", bindex);
3866 + for (ull = 0; !err && ull < max; ull++) {
3870 + if (i->i_ino == AUFS_ROOT_INO)
3873 + /* AuDbgInode(i); */
3874 + if (au_iigen(i, NULL) == sigen)
3875 + ii_read_lock_child(i);
3877 + ii_write_lock_child(i);
3878 + err = au_refresh_hinode_self(i);
3881 + ii_downgrade_lock(i);
3883 + ii_write_unlock(i);
3888 + btop = au_ibtop(i);
3889 + bbot = au_ibbot(i);
3890 + if (btop <= bindex
3892 + && au_h_iptr(i, bindex)
3893 + && au_test_ibusy(i, btop, bbot)) {
3895 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3898 + ii_read_unlock(i);
3900 + au_iarray_free(array, max);
3906 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3907 + const unsigned int verbose)
3910 + unsigned int sigen;
3912 + sigen = au_sigen(root->d_sb);
3913 + DiMustNoWaiters(root);
3914 + IiMustNoWaiters(d_inode(root));
3915 + di_write_unlock(root);
3916 + err = test_dentry_busy(root, bindex, sigen, verbose);
3918 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3919 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3924 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3925 + struct file **to_free, int *idx)
3928 + unsigned char matched, root;
3929 + aufs_bindex_t bindex, bbot;
3930 + struct au_fidir *fidir;
3931 + struct au_hfile *hfile;
3934 + root = IS_ROOT(file->f_path.dentry);
3937 + to_free[*idx] = file;
3943 + fidir = au_fi(file)->fi_hdir;
3944 + AuDebugOn(!fidir);
3945 + bbot = au_fbbot_dir(file);
3946 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++) {
3947 + hfile = fidir->fd_hfile + bindex;
3948 + if (!hfile->hf_file)
3951 + if (hfile->hf_br->br_id == br_id) {
3963 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3964 + struct file **to_free, int opened)
3967 + unsigned long long ull, max;
3968 + aufs_bindex_t btop;
3969 + struct file *file, **array;
3970 + struct dentry *root;
3971 + struct au_hfile *hfile;
3973 + array = au_farray_alloc(sb, &max);
3974 + err = PTR_ERR(array);
3975 + if (IS_ERR(array))
3980 + root = sb->s_root;
3981 + di_write_unlock(root);
3982 + for (ull = 0; ull < max; ull++) {
3983 + file = array[ull];
3984 + if (unlikely(!file))
3987 + /* AuDbg("%pD\n", file); */
3988 + fi_read_lock(file);
3989 + btop = au_fbtop(file);
3990 + if (!d_is_dir(file->f_path.dentry)) {
3991 + hfile = &au_fi(file)->fi_htop;
3992 + if (hfile->hf_br->br_id == br_id)
3995 + err = test_dir_busy(file, br_id, to_free, &idx);
3996 + fi_read_unlock(file);
3997 + if (unlikely(err))
4000 + di_write_lock_child(root);
4001 + au_farray_free(array, max);
4002 + AuDebugOn(idx > opened);
4008 +static void br_del_file(struct file **to_free, unsigned long long opened,
4009 + aufs_bindex_t br_id)
4011 + unsigned long long ull;
4012 + aufs_bindex_t bindex, btop, bbot, bfound;
4013 + struct file *file;
4014 + struct au_fidir *fidir;
4015 + struct au_hfile *hfile;
4017 + for (ull = 0; ull < opened; ull++) {
4018 + file = to_free[ull];
4019 + if (unlikely(!file))
4022 + /* AuDbg("%pD\n", file); */
4023 + AuDebugOn(!d_is_dir(file->f_path.dentry));
4025 + fidir = au_fi(file)->fi_hdir;
4026 + AuDebugOn(!fidir);
4027 + fi_write_lock(file);
4028 + btop = au_fbtop(file);
4029 + bbot = au_fbbot_dir(file);
4030 + for (bindex = btop; bindex <= bbot; bindex++) {
4031 + hfile = fidir->fd_hfile + bindex;
4032 + if (!hfile->hf_file)
4035 + if (hfile->hf_br->br_id == br_id) {
4040 + AuDebugOn(bfound < 0);
4041 + au_set_h_fptr(file, bfound, NULL);
4042 + if (bfound == btop) {
4043 + for (btop++; btop <= bbot; btop++)
4044 + if (au_hf_dir(file, btop)) {
4045 + au_set_fbtop(file, btop);
4049 + fi_write_unlock(file);
4053 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
4054 + const aufs_bindex_t bindex,
4055 + const aufs_bindex_t bbot)
4057 + struct au_branch **brp, **p;
4059 + AuRwMustWriteLock(&sbinfo->si_rwsem);
4061 + brp = sbinfo->si_branch + bindex;
4062 + if (bindex < bbot)
4063 + memmove(brp, brp + 1, sizeof(*brp) * (bbot - bindex));
4064 + sbinfo->si_branch[0 + bbot] = NULL;
4065 + sbinfo->si_bbot--;
4067 + p = au_krealloc(sbinfo->si_branch, sizeof(*p) * bbot, AuGFP_SBILIST,
4070 + sbinfo->si_branch = p;
4071 + /* harmless error */
4074 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
4075 + const aufs_bindex_t bbot)
4077 + struct au_hdentry *hdp, *p;
4079 + AuRwMustWriteLock(&dinfo->di_rwsem);
4081 + hdp = au_hdentry(dinfo, bindex);
4082 + if (bindex < bbot)
4083 + memmove(hdp, hdp + 1, sizeof(*hdp) * (bbot - bindex));
4084 + /* au_h_dentry_init(au_hdentry(dinfo, bbot); */
4087 + p = au_krealloc(dinfo->di_hdentry, sizeof(*p) * bbot, AuGFP_SBILIST,
4090 + dinfo->di_hdentry = p;
4091 + /* harmless error */
4094 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
4095 + const aufs_bindex_t bbot)
4097 + struct au_hinode *hip, *p;
4099 + AuRwMustWriteLock(&iinfo->ii_rwsem);
4101 + hip = au_hinode(iinfo, bindex);
4102 + if (bindex < bbot)
4103 + memmove(hip, hip + 1, sizeof(*hip) * (bbot - bindex));
4104 + /* au_hinode_init(au_hinode(iinfo, bbot)); */
4107 + p = au_krealloc(iinfo->ii_hinode, sizeof(*p) * bbot, AuGFP_SBILIST,
4110 + iinfo->ii_hinode = p;
4111 + /* harmless error */
4114 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
4115 + struct au_branch *br)
4117 + aufs_bindex_t bbot;
4118 + struct au_sbinfo *sbinfo;
4119 + struct dentry *root, *h_root;
4120 + struct inode *inode, *h_inode;
4121 + struct au_hinode *hinode;
4123 + SiMustWriteLock(sb);
4125 + root = sb->s_root;
4126 + inode = d_inode(root);
4127 + sbinfo = au_sbi(sb);
4128 + bbot = sbinfo->si_bbot;
4130 + h_root = au_h_dptr(root, bindex);
4131 + hinode = au_hi(inode, bindex);
4132 + h_inode = au_igrab(hinode->hi_inode);
4135 + au_sbilist_lock();
4136 + au_br_do_del_brp(sbinfo, bindex, bbot);
4137 + au_br_do_del_hdp(au_di(root), bindex, bbot);
4138 + au_br_do_del_hip(au_ii(inode), bindex, bbot);
4139 + au_sbilist_unlock();
4141 + /* ignore an error */
4142 + au_dr_br_fin(sb, br); /* always, regardless the mount option */
4146 + au_br_do_free(br);
4149 +static unsigned long long empty_cb(struct super_block *sb, void *array,
4150 + unsigned long long max, void *arg)
4155 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
4158 + unsigned long long opened;
4159 + unsigned int mnt_flags;
4160 + aufs_bindex_t bindex, bbot, br_id;
4161 + unsigned char do_wh, verbose;
4162 + struct au_branch *br;
4163 + struct au_wbr *wbr;
4164 + struct dentry *root;
4165 + struct file **to_free;
4170 + root = sb->s_root;
4171 + bindex = au_find_dbindex(root, del->h_path.dentry);
4174 + goto out; /* success */
4176 + pr_err("%s no such branch\n", del->pathname);
4179 + AuDbg("bindex b%d\n", bindex);
4182 + mnt_flags = au_mntflags(sb);
4183 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4184 + bbot = au_sbbot(sb);
4185 + if (unlikely(!bbot)) {
4186 + AuVerbose(verbose, "no more branches left\n");
4189 + br = au_sbr(sb, bindex);
4190 + AuDebugOn(!path_equal(&br->br_path, &del->h_path));
4192 + br_id = br->br_id;
4193 + opened = au_br_count(br);
4194 + if (unlikely(opened)) {
4195 + to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
4196 + err = PTR_ERR(to_free);
4197 + if (IS_ERR(to_free))
4200 + err = test_file_busy(sb, br_id, to_free, opened);
4201 + if (unlikely(err)) {
4202 + AuVerbose(verbose, "%llu file(s) opened\n", opened);
4208 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
4210 + /* instead of WbrWhMustWriteLock(wbr) */
4211 + SiMustWriteLock(sb);
4212 + for (i = 0; i < AuBrWh_Last; i++) {
4213 + dput(wbr->wbr_wh[i]);
4214 + wbr->wbr_wh[i] = NULL;
4218 + err = test_children_busy(root, bindex, verbose);
4219 + if (unlikely(err)) {
4228 + * now we confirmed the branch is deletable.
4229 + * let's free the remaining opened dirs on the branch.
4231 + di_write_unlock(root);
4232 + br_del_file(to_free, opened, br_id);
4233 + di_write_lock_child(root);
4237 + au_br_do_del(sb, bindex, br);
4239 + sysaufs_brs_del(sb, bindex);
4240 + au_br_do_del(sb, bindex, br);
4241 + sysaufs_brs_add(sb, bindex);
4245 + au_cpup_attr_all(d_inode(root), /*force*/1);
4246 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
4248 + au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
4249 + if (au_opt_test(mnt_flags, PLINK))
4250 + au_plink_half_refresh(sb, br_id);
4252 + if (au_xino_brid(sb) == br_id)
4253 + au_xino_brid_set(sb, -1);
4254 + goto out; /* success */
4258 + rerr = au_br_init_wh(sb, br, br->br_perm);
4260 + pr_warn("failed re-creating base whiteout, %s. (%d)\n",
4261 + del->pathname, rerr);
4264 + au_farray_free(to_free, opened);
4268 +/* ---------------------------------------------------------------------- */
4270 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
4273 + aufs_bindex_t btop, bbot;
4274 + struct aufs_ibusy ibusy;
4275 + struct inode *inode, *h_inode;
4278 + if (unlikely(!capable(CAP_SYS_ADMIN)))
4281 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
4283 + err = !access_ok(VERIFY_WRITE, &arg->h_ino, sizeof(arg->h_ino));
4284 + if (unlikely(err)) {
4291 + si_read_lock(sb, AuLock_FLUSH);
4292 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbbot(sb)))
4296 + ibusy.h_ino = 0; /* invalid */
4297 + inode = ilookup(sb, ibusy.ino);
4299 + || inode->i_ino == AUFS_ROOT_INO
4300 + || au_is_bad_inode(inode))
4303 + ii_read_lock_child(inode);
4304 + btop = au_ibtop(inode);
4305 + bbot = au_ibbot(inode);
4306 + if (btop <= ibusy.bindex && ibusy.bindex <= bbot) {
4307 + h_inode = au_h_iptr(inode, ibusy.bindex);
4308 + if (h_inode && au_test_ibusy(inode, btop, bbot))
4309 + ibusy.h_ino = h_inode->i_ino;
4311 + ii_read_unlock(inode);
4315 + si_read_unlock(sb);
4317 + err = __put_user(ibusy.h_ino, &arg->h_ino);
4318 + if (unlikely(err)) {
4327 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
4329 + return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
4332 +#ifdef CONFIG_COMPAT
4333 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
4335 + return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
4339 +/* ---------------------------------------------------------------------- */
4342 + * change a branch permission
4345 +static void au_warn_ima(void)
4348 + /* since it doesn't support mark_files_ro() */
4349 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
4353 +static int do_need_sigen_inc(int a, int b)
4355 + return au_br_whable(a) && !au_br_whable(b);
4358 +static int need_sigen_inc(int old, int new)
4360 + return do_need_sigen_inc(old, new)
4361 + || do_need_sigen_inc(new, old);
4364 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
4367 + unsigned int mnt_flags;
4368 + unsigned long long ull, max;
4369 + aufs_bindex_t br_id;
4370 + unsigned char verbose, writer;
4371 + struct file *file, *hf, **array;
4372 + struct au_hfile *hfile;
4374 + mnt_flags = au_mntflags(sb);
4375 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4377 + array = au_farray_alloc(sb, &max);
4378 + err = PTR_ERR(array);
4379 + if (IS_ERR(array))
4383 + br_id = au_sbr_id(sb, bindex);
4384 + for (ull = 0; ull < max; ull++) {
4385 + file = array[ull];
4386 + if (unlikely(!file))
4389 + /* AuDbg("%pD\n", file); */
4390 + fi_read_lock(file);
4391 + if (unlikely(au_test_mmapped(file))) {
4393 + AuVerbose(verbose, "mmapped %pD\n", file);
4395 + FiMustNoWaiters(file);
4396 + fi_read_unlock(file);
4400 + hfile = &au_fi(file)->fi_htop;
4401 + hf = hfile->hf_file;
4402 + if (!d_is_reg(file->f_path.dentry)
4403 + || !(file->f_mode & FMODE_WRITE)
4404 + || hfile->hf_br->br_id != br_id
4405 + || !(hf->f_mode & FMODE_WRITE))
4406 + array[ull] = NULL;
4412 + FiMustNoWaiters(file);
4413 + fi_read_unlock(file);
4421 + for (ull = 0; ull < max; ull++) {
4422 + file = array[ull];
4426 + /* todo: already flushed? */
4428 + * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4429 + * approach which resets f_mode and calls mnt_drop_write() and
4430 + * file_release_write() for each file, because the branch
4431 + * attribute in aufs world is totally different from the native
4434 + /* fi_read_lock(file); */
4435 + hfile = &au_fi(file)->fi_htop;
4436 + hf = hfile->hf_file;
4437 + /* fi_read_unlock(file); */
4438 + spin_lock(&hf->f_lock);
4439 + writer = !!(hf->f_mode & FMODE_WRITER);
4440 + hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4441 + spin_unlock(&hf->f_lock);
4443 + put_write_access(file_inode(hf));
4444 + __mnt_drop_write(hf->f_path.mnt);
4449 + au_farray_free(array, max);
4455 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4459 + aufs_bindex_t bindex;
4460 + struct dentry *root;
4461 + struct au_branch *br;
4462 + struct au_br_fhsm *bf;
4464 + root = sb->s_root;
4465 + bindex = au_find_dbindex(root, mod->h_root);
4468 + return 0; /* success */
4470 + pr_err("%s no such branch\n", mod->path);
4473 + AuDbg("bindex b%d\n", bindex);
4475 + err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4476 + if (unlikely(err))
4479 + br = au_sbr(sb, bindex);
4480 + AuDebugOn(mod->h_root != au_br_dentry(br));
4481 + if (br->br_perm == mod->perm)
4482 + return 0; /* success */
4484 + /* pre-allocate for non-fhsm --> fhsm */
4486 + if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4487 + err = au_fhsm_br_alloc(br);
4488 + if (unlikely(err))
4491 + br->br_fhsm = NULL;
4494 + if (au_br_writable(br->br_perm)) {
4495 + /* remove whiteout base */
4496 + err = au_br_init_wh(sb, br, mod->perm);
4497 + if (unlikely(err))
4500 + if (!au_br_writable(mod->perm)) {
4501 + /* rw --> ro, file might be mmapped */
4502 + DiMustNoWaiters(root);
4503 + IiMustNoWaiters(d_inode(root));
4504 + di_write_unlock(root);
4505 + err = au_br_mod_files_ro(sb, bindex);
4506 + /* aufs_write_lock() calls ..._child() */
4507 + di_write_lock_child(root);
4509 + if (unlikely(err)) {
4511 + br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4514 + rerr = au_wbr_init(br, sb, br->br_perm);
4515 + if (unlikely(rerr)) {
4516 + AuIOErr("nested error %d (%d)\n",
4518 + br->br_perm = mod->perm;
4522 + } else if (au_br_writable(mod->perm)) {
4525 + br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4527 + err = au_wbr_init(br, sb, mod->perm);
4528 + if (unlikely(err)) {
4529 + kfree(br->br_wbr);
4530 + br->br_wbr = NULL;
4534 + if (unlikely(err))
4537 + if (au_br_fhsm(br->br_perm)) {
4538 + if (!au_br_fhsm(mod->perm)) {
4539 + /* fhsm --> non-fhsm */
4540 + au_br_fhsm_fin(br->br_fhsm);
4541 + kfree(br->br_fhsm);
4542 + br->br_fhsm = NULL;
4544 + } else if (au_br_fhsm(mod->perm))
4545 + /* non-fhsm --> fhsm */
4548 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4549 + br->br_perm = mod->perm;
4550 + goto out; /* success */
4559 +/* ---------------------------------------------------------------------- */
4561 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4564 + struct kstatfs kstfs;
4566 + err = vfs_statfs(&br->br_path, &kstfs);
4568 + stfs->f_blocks = kstfs.f_blocks;
4569 + stfs->f_bavail = kstfs.f_bavail;
4570 + stfs->f_files = kstfs.f_files;
4571 + stfs->f_ffree = kstfs.f_ffree;
4576 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4577 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
4578 +++ linux/fs/aufs/branch.h 2018-01-29 07:56:20.049991637 +0100
4581 + * Copyright (C) 2005-2017 Junjiro R. Okajima
4583 + * This program, aufs is free software; you can redistribute it and/or modify
4584 + * it under the terms of the GNU General Public License as published by
4585 + * the Free Software Foundation; either version 2 of the License, or
4586 + * (at your option) any later version.
4588 + * This program is distributed in the hope that it will be useful,
4589 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4590 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4591 + * GNU General Public License for more details.
4593 + * You should have received a copy of the GNU General Public License
4594 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4598 + * branch filesystems and xino for them
4601 +#ifndef __AUFS_BRANCH_H__
4602 +#define __AUFS_BRANCH_H__
4606 +#include <linux/mount.h>
4607 +#include "dirren.h"
4612 +/* ---------------------------------------------------------------------- */
4615 +struct au_xino_file {
4616 + struct file *xi_file;
4621 + /* reserved for future use */
4622 + /* unsigned long *bitmap; */
4623 + wait_queue_head_t wqh;
4626 + /* todo: make xino files an array to support huge inode number */
4628 +#ifdef CONFIG_DEBUG_FS
4629 + struct dentry *xi_dbgaufs;
4633 +/* File-based Hierarchical Storage Management */
4634 +struct au_br_fhsm {
4635 +#ifdef CONFIG_AUFS_FHSM
4636 + struct mutex bf_lock;
4637 + unsigned long bf_jiffy;
4638 + struct aufs_stfs bf_stfs;
4643 +/* members for writable branch only */
4644 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4646 + struct au_rwsem wbr_wh_rwsem;
4647 + struct dentry *wbr_wh[AuBrWh_Last];
4648 + atomic_t wbr_wh_running;
4649 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4650 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4651 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4654 + unsigned long long wbr_bytes;
4657 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4658 +#define AuBrDynOp (AuDyLast * 4)
4660 +#ifdef CONFIG_AUFS_HFSNOTIFY
4661 +/* support for asynchronous destruction */
4662 +struct au_br_hfsnotify {
4663 + struct fsnotify_group *hfsn_group;
4667 +/* sysfs entries */
4668 +struct au_brsysfs {
4670 + struct attribute attr;
4679 +/* protected by superblock rwsem */
4681 + struct au_xino_file br_xino;
4683 + aufs_bindex_t br_id;
4686 + struct path br_path;
4687 + spinlock_t br_dykey_lock;
4688 + struct au_dykey *br_dykey[AuBrDynOp];
4689 + struct percpu_counter br_count;
4691 + struct au_wbr *br_wbr;
4692 + struct au_br_fhsm *br_fhsm;
4694 + /* xino truncation */
4695 + atomic_t br_xino_running;
4697 +#ifdef CONFIG_AUFS_HFSNOTIFY
4698 + struct au_br_hfsnotify *br_hfsn;
4701 +#ifdef CONFIG_SYSFS
4702 + /* entries under sysfs per mount-point */
4703 + struct au_brsysfs br_sysfs[AuBrSysfs_Last];
4706 + struct au_dr_br br_dirren;
4709 +/* ---------------------------------------------------------------------- */
4711 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4713 + return br->br_path.mnt;
4716 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4718 + return br->br_path.dentry;
4721 +static inline struct super_block *au_br_sb(struct au_branch *br)
4723 + return au_br_mnt(br)->mnt_sb;
4726 +static inline void au_br_get(struct au_branch *br)
4728 + percpu_counter_inc(&br->br_count);
4731 +static inline void au_br_put(struct au_branch *br)
4733 + percpu_counter_dec(&br->br_count);
4736 +static inline s64 au_br_count(struct au_branch *br)
4738 + return percpu_counter_sum(&br->br_count);
4741 +static inline void au_br_count_init(struct au_branch *br)
4743 + percpu_counter_init(&br->br_count, 0, GFP_NOFS);
4746 +static inline void au_br_count_fin(struct au_branch *br)
4748 + percpu_counter_destroy(&br->br_count);
4751 +static inline int au_br_rdonly(struct au_branch *br)
4753 + return (sb_rdonly(au_br_sb(br))
4754 + || !au_br_writable(br->br_perm))
4758 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4760 +#ifdef CONFIG_AUFS_HNOTIFY
4761 + return !(brperm & AuBrPerm_RR);
4767 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4769 + int err, exec_flag;
4772 + exec_flag = oflag & __FMODE_EXEC;
4773 + if (unlikely(exec_flag && path_noexec(&br->br_path)))
4779 +/* ---------------------------------------------------------------------- */
4783 +void au_br_free(struct au_sbinfo *sinfo);
4784 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4786 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4788 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4789 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4790 +#ifdef CONFIG_COMPAT
4791 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4794 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4797 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4800 +static const loff_t au_loff_max = LLONG_MAX;
4802 +int au_xib_trunc(struct super_block *sb);
4803 +ssize_t xino_fread(vfs_readf_t func, struct file *file, void *buf, size_t size,
4805 +ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
4806 + size_t size, loff_t *pos);
4807 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src);
4808 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent);
4809 +ino_t au_xino_new_ino(struct super_block *sb);
4810 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4811 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4813 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4815 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4816 + struct file *base_file, int do_test);
4817 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex);
4819 +struct au_opt_xino;
4820 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount);
4821 +void au_xino_clr(struct super_block *sb);
4822 +struct file *au_xino_def(struct super_block *sb);
4823 +int au_xino_path(struct seq_file *seq, struct file *file);
4825 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
4826 + ino_t h_ino, int idx);
4827 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4830 +/* ---------------------------------------------------------------------- */
4832 +/* Superblock to branch */
4834 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4836 + return au_sbr(sb, bindex)->br_id;
4840 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4842 + return au_br_mnt(au_sbr(sb, bindex));
4846 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4848 + return au_br_sb(au_sbr(sb, bindex));
4851 +static inline void au_sbr_get(struct super_block *sb, aufs_bindex_t bindex)
4853 + au_br_get(au_sbr(sb, bindex));
4856 +static inline void au_sbr_put(struct super_block *sb, aufs_bindex_t bindex)
4858 + au_br_put(au_sbr(sb, bindex));
4861 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4863 + return au_sbr(sb, bindex)->br_perm;
4866 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4868 + return au_br_whable(au_sbr_perm(sb, bindex));
4871 +/* ---------------------------------------------------------------------- */
4873 +#define wbr_wh_read_lock(wbr) au_rw_read_lock(&(wbr)->wbr_wh_rwsem)
4874 +#define wbr_wh_write_lock(wbr) au_rw_write_lock(&(wbr)->wbr_wh_rwsem)
4875 +#define wbr_wh_read_trylock(wbr) au_rw_read_trylock(&(wbr)->wbr_wh_rwsem)
4876 +#define wbr_wh_write_trylock(wbr) au_rw_write_trylock(&(wbr)->wbr_wh_rwsem)
4878 +#define wbr_wh_read_trylock_nested(wbr) \
4879 + au_rw_read_trylock_nested(&(wbr)->wbr_wh_rwsem)
4880 +#define wbr_wh_write_trylock_nested(wbr) \
4881 + au_rw_write_trylock_nested(&(wbr)->wbr_wh_rwsem)
4884 +#define wbr_wh_read_unlock(wbr) au_rw_read_unlock(&(wbr)->wbr_wh_rwsem)
4885 +#define wbr_wh_write_unlock(wbr) au_rw_write_unlock(&(wbr)->wbr_wh_rwsem)
4886 +#define wbr_wh_downgrade_lock(wbr) au_rw_dgrade_lock(&(wbr)->wbr_wh_rwsem)
4888 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&(wbr)->wbr_wh_rwsem)
4889 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&(wbr)->wbr_wh_rwsem)
4890 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&(wbr)->wbr_wh_rwsem)
4892 +/* ---------------------------------------------------------------------- */
4894 +#ifdef CONFIG_AUFS_FHSM
4895 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4897 + mutex_init(&brfhsm->bf_lock);
4898 + brfhsm->bf_jiffy = 0;
4899 + brfhsm->bf_readable = 0;
4902 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4904 + mutex_destroy(&brfhsm->bf_lock);
4907 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4908 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4911 +#endif /* __KERNEL__ */
4912 +#endif /* __AUFS_BRANCH_H__ */
4913 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4914 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
4915 +++ linux/fs/aufs/conf.mk 2018-01-29 07:56:20.049991637 +0100
4917 +# SPDX-License-Identifier: GPL-2.0
4919 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4923 +AuConfStr += ${1}=${${1}}
4927 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4929 + HNOTIFY HFSNOTIFY \
4941 +$(foreach i, ${AuConfAll}, \
4942 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
4944 +AuConfName = ${obj}/conf.str
4945 +${AuConfName}.tmp: FORCE
4946 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4947 +${AuConfName}: ${AuConfName}.tmp
4948 + @diff -q $< $@ > /dev/null 2>&1 || { \
4949 + echo ' GEN ' $@; \
4953 +clean-files += ${AuConfName} ${AuConfName}.tmp
4954 +${obj}/sysfs.o: ${AuConfName}
4956 +-include ${srctree}/${src}/conf_priv.mk
4957 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4958 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
4959 +++ linux/fs/aufs/cpup.c 2018-01-29 07:56:20.053325069 +0100
4962 + * Copyright (C) 2005-2017 Junjiro R. Okajima
4964 + * This program, aufs is free software; you can redistribute it and/or modify
4965 + * it under the terms of the GNU General Public License as published by
4966 + * the Free Software Foundation; either version 2 of the License, or
4967 + * (at your option) any later version.
4969 + * This program is distributed in the hope that it will be useful,
4970 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4971 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4972 + * GNU General Public License for more details.
4974 + * You should have received a copy of the GNU General Public License
4975 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4979 + * copy-up functions, see wbr_policy.c for copy-down
4982 +#include <linux/fs_stack.h>
4983 +#include <linux/mm.h>
4984 +#include <linux/task_work.h>
4987 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4989 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4990 + | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4992 + BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4994 + dst->i_flags |= iflags & ~mask;
4995 + if (au_test_fs_notime(dst->i_sb))
4996 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
4999 +void au_cpup_attr_timesizes(struct inode *inode)
5001 + struct inode *h_inode;
5003 + h_inode = au_h_iptr(inode, au_ibtop(inode));
5004 + fsstack_copy_attr_times(inode, h_inode);
5005 + fsstack_copy_inode_size(inode, h_inode);
5008 +void au_cpup_attr_nlink(struct inode *inode, int force)
5010 + struct inode *h_inode;
5011 + struct super_block *sb;
5012 + aufs_bindex_t bindex, bbot;
5015 + bindex = au_ibtop(inode);
5016 + h_inode = au_h_iptr(inode, bindex);
5018 + && !S_ISDIR(h_inode->i_mode)
5019 + && au_opt_test(au_mntflags(sb), PLINK)
5020 + && au_plink_test(inode))
5024 + * 0 can happen in revalidating.
5025 + * h_inode->i_mutex may not be held here, but it is harmless since once
5026 + * i_nlink reaches 0, it will never become positive except O_TMPFILE
5028 + * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
5029 + * the incorrect link count.
5031 + set_nlink(inode, h_inode->i_nlink);
5034 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
5035 + * it may includes whplink directory.
5037 + if (S_ISDIR(h_inode->i_mode)) {
5038 + bbot = au_ibbot(inode);
5039 + for (bindex++; bindex <= bbot; bindex++) {
5040 + h_inode = au_h_iptr(inode, bindex);
5042 + au_add_nlink(inode, h_inode);
5047 +void au_cpup_attr_changeable(struct inode *inode)
5049 + struct inode *h_inode;
5051 + h_inode = au_h_iptr(inode, au_ibtop(inode));
5052 + inode->i_mode = h_inode->i_mode;
5053 + inode->i_uid = h_inode->i_uid;
5054 + inode->i_gid = h_inode->i_gid;
5055 + au_cpup_attr_timesizes(inode);
5056 + au_cpup_attr_flags(inode, h_inode->i_flags);
5059 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
5061 + struct au_iinfo *iinfo = au_ii(inode);
5063 + IiMustWriteLock(inode);
5065 + iinfo->ii_higen = h_inode->i_generation;
5066 + iinfo->ii_hsb1 = h_inode->i_sb;
5069 +void au_cpup_attr_all(struct inode *inode, int force)
5071 + struct inode *h_inode;
5073 + h_inode = au_h_iptr(inode, au_ibtop(inode));
5074 + au_cpup_attr_changeable(inode);
5075 + if (inode->i_nlink > 0)
5076 + au_cpup_attr_nlink(inode, force);
5077 + inode->i_rdev = h_inode->i_rdev;
5078 + inode->i_blkbits = h_inode->i_blkbits;
5079 + au_cpup_igen(inode, h_inode);
5082 +/* ---------------------------------------------------------------------- */
5084 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
5086 +/* keep the timestamps of the parent dir when cpup */
5087 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
5088 + struct path *h_path)
5090 + struct inode *h_inode;
5092 + dt->dt_dentry = dentry;
5093 + dt->dt_h_path = *h_path;
5094 + h_inode = d_inode(h_path->dentry);
5095 + dt->dt_atime = h_inode->i_atime;
5096 + dt->dt_mtime = h_inode->i_mtime;
5100 +void au_dtime_revert(struct au_dtime *dt)
5102 + struct iattr attr;
5105 + attr.ia_atime = dt->dt_atime;
5106 + attr.ia_mtime = dt->dt_mtime;
5107 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
5108 + | ATTR_ATIME | ATTR_ATIME_SET;
5110 + /* no delegation since this is a directory */
5111 + err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
5112 + if (unlikely(err))
5113 + pr_warn("restoring timestamps failed(%d). ignored\n", err);
5116 +/* ---------------------------------------------------------------------- */
5118 +/* internal use only */
5119 +struct au_cpup_reg_attr {
5122 + unsigned int iflags; /* inode->i_flags */
5125 +static noinline_for_stack
5126 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct dentry *h_src,
5127 + struct au_cpup_reg_attr *h_src_attr)
5129 + int err, sbits, icex;
5130 + unsigned int mnt_flags;
5131 + unsigned char verbose;
5133 + struct path h_path;
5134 + struct inode *h_isrc, *h_idst;
5135 + struct kstat *h_st;
5136 + struct au_branch *br;
5138 + h_path.dentry = au_h_dptr(dst, bindex);
5139 + h_idst = d_inode(h_path.dentry);
5140 + br = au_sbr(dst->d_sb, bindex);
5141 + h_path.mnt = au_br_mnt(br);
5142 + h_isrc = d_inode(h_src);
5143 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
5144 + | ATTR_ATIME | ATTR_MTIME
5145 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
5146 + if (h_src_attr && h_src_attr->valid) {
5147 + h_st = &h_src_attr->st;
5148 + ia.ia_uid = h_st->uid;
5149 + ia.ia_gid = h_st->gid;
5150 + ia.ia_atime = h_st->atime;
5151 + ia.ia_mtime = h_st->mtime;
5152 + if (h_idst->i_mode != h_st->mode
5153 + && !S_ISLNK(h_idst->i_mode)) {
5154 + ia.ia_valid |= ATTR_MODE;
5155 + ia.ia_mode = h_st->mode;
5157 + sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
5158 + au_cpup_attr_flags(h_idst, h_src_attr->iflags);
5160 + ia.ia_uid = h_isrc->i_uid;
5161 + ia.ia_gid = h_isrc->i_gid;
5162 + ia.ia_atime = h_isrc->i_atime;
5163 + ia.ia_mtime = h_isrc->i_mtime;
5164 + if (h_idst->i_mode != h_isrc->i_mode
5165 + && !S_ISLNK(h_idst->i_mode)) {
5166 + ia.ia_valid |= ATTR_MODE;
5167 + ia.ia_mode = h_isrc->i_mode;
5169 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
5170 + au_cpup_attr_flags(h_idst, h_isrc->i_flags);
5172 + /* no delegation since it is just created */
5173 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
5175 + /* is this nfs only? */
5176 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
5177 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
5178 + ia.ia_mode = h_isrc->i_mode;
5179 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
5182 + icex = br->br_perm & AuBrAttr_ICEX;
5184 + mnt_flags = au_mntflags(dst->d_sb);
5185 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
5186 + err = au_cpup_xattr(h_path.dentry, h_src, icex, verbose);
5192 +/* ---------------------------------------------------------------------- */
5194 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
5195 + char *buf, unsigned long blksize)
5198 + size_t sz, rbytes, wbytes;
5199 + unsigned char all_zero;
5201 + struct inode *h_inode;
5202 + /* reduce stack usage */
5205 + zp = page_address(ZERO_PAGE(0));
5206 + if (unlikely(!zp))
5207 + return -ENOMEM; /* possible? */
5212 + AuDbg("len %lld\n", len);
5214 + if (len < blksize)
5218 + /* todo: signal_pending? */
5219 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
5220 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
5223 + if (unlikely(err < 0))
5227 + if (len >= rbytes && rbytes == blksize)
5228 + all_zero = !memcmp(buf, zp, rbytes);
5235 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
5237 + /* todo: signal_pending? */
5238 + if (unlikely(err == -EAGAIN || err == -EINTR))
5240 + if (unlikely(err < 0))
5245 + if (unlikely(err < 0))
5251 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
5253 + if (unlikely(res < 0))
5260 + /* the last block may be a hole */
5261 + if (!err && all_zero) {
5262 + AuLabel(last hole);
5265 + if (au_test_nfs(dst->f_path.dentry->d_sb)) {
5266 + /* nfs requires this step to make last hole */
5267 + /* is this only nfs? */
5269 + /* todo: signal_pending? */
5270 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
5271 + } while (err == -EAGAIN || err == -EINTR);
5278 + ia->ia_size = dst->f_pos;
5279 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
5280 + ia->ia_file = dst;
5281 + h_inode = file_inode(dst);
5282 + inode_lock_nested(h_inode, AuLsc_I_CHILD2);
5283 + /* no delegation since it is just created */
5284 + err = vfsub_notify_change(&dst->f_path, ia,
5285 + /*delegated*/NULL);
5286 + inode_unlock(h_inode);
5293 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
5296 + unsigned long blksize;
5297 + unsigned char do_kfree;
5301 + blksize = dst->f_path.dentry->d_sb->s_blocksize;
5302 + if (!blksize || PAGE_SIZE < blksize)
5303 + blksize = PAGE_SIZE;
5304 + AuDbg("blksize %lu\n", blksize);
5305 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
5307 + buf = kmalloc(blksize, GFP_NOFS);
5309 + buf = (void *)__get_free_page(GFP_NOFS);
5310 + if (unlikely(!buf))
5313 + if (len > (1 << 22))
5314 + AuDbg("copying a large file %lld\n", (long long)len);
5318 + err = au_do_copy_file(dst, src, len, buf, blksize);
5322 + free_page((unsigned long)buf);
5328 +static int au_do_copy(struct file *dst, struct file *src, loff_t len)
5331 + struct super_block *h_src_sb;
5332 + struct inode *h_src_inode;
5334 + h_src_inode = file_inode(src);
5335 + h_src_sb = h_src_inode->i_sb;
5337 + /* XFS acquires inode_lock */
5338 + if (!au_test_xfs(h_src_sb))
5339 + err = au_copy_file(dst, src, len);
5341 + inode_unlock_shared(h_src_inode);
5342 + err = au_copy_file(dst, src, len);
5343 + vfsub_inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5349 +static int au_clone_or_copy(struct file *dst, struct file *src, loff_t len)
5352 + struct super_block *h_src_sb;
5353 + struct inode *h_src_inode;
5355 + h_src_inode = file_inode(src);
5356 + h_src_sb = h_src_inode->i_sb;
5357 + if (h_src_sb != file_inode(dst)->i_sb
5358 + || !dst->f_op->clone_file_range) {
5359 + err = au_do_copy(dst, src, len);
5363 + if (!au_test_nfs(h_src_sb)) {
5364 + inode_unlock_shared(h_src_inode);
5365 + err = vfsub_clone_file_range(src, dst, len);
5366 + vfsub_inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5368 + err = vfsub_clone_file_range(src, dst, len);
5369 + /* older XFS has a condition in cloning */
5370 + if (unlikely(err != -EOPNOTSUPP))
5373 + /* the backend fs on NFS may not support cloning */
5374 + err = au_do_copy(dst, src, len);
5382 + * to support a sparse file which is opened with O_APPEND,
5383 + * we need to close the file.
5385 +static int au_cp_regular(struct au_cp_generic *cpg)
5388 + enum { SRC, DST };
5390 + aufs_bindex_t bindex;
5391 + unsigned int flags;
5392 + struct dentry *dentry;
5394 + struct file *file;
5398 + .bindex = cpg->bsrc,
5399 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
5403 + .bindex = cpg->bdst,
5404 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
5405 + .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
5406 + .label = &&out_src
5409 + struct super_block *sb, *h_src_sb;
5410 + struct inode *h_src_inode;
5411 + struct task_struct *tsk = current;
5413 + /* bsrc branch can be ro/rw. */
5414 + sb = cpg->dentry->d_sb;
5416 + for (i = 0; i < 2; i++, f++) {
5417 + f->dentry = au_h_dptr(cpg->dentry, f->bindex);
5418 + f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
5419 + /*file*/NULL, f->force_wr);
5420 + err = PTR_ERR(f->file);
5421 + if (IS_ERR(f->file))
5425 + /* try stopping to update while we copyup */
5426 + h_src_inode = d_inode(file[SRC].dentry);
5427 + h_src_sb = h_src_inode->i_sb;
5428 + if (!au_test_nfs(h_src_sb))
5429 + IMustLock(h_src_inode);
5430 + err = au_clone_or_copy(file[DST].file, file[SRC].file, cpg->len);
5432 + /* i wonder if we had O_NO_DELAY_FPUT flag */
5433 + if (tsk->flags & PF_KTHREAD)
5434 + __fput_sync(file[DST].file);
5436 + /* it happend actually */
5437 + fput(file[DST].file);
5440 + * we have to call both since we don't know which place the file
5444 + flush_delayed_fput();
5446 + au_sbr_put(sb, file[DST].bindex);
5449 + fput(file[SRC].file);
5450 + au_sbr_put(sb, file[SRC].bindex);
5455 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
5456 + struct au_cpup_reg_attr *h_src_attr)
5460 + struct path h_path;
5461 + struct inode *h_src_inode, *h_dst_inode;
5464 + h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
5465 + l = i_size_read(h_src_inode);
5466 + if (cpg->len == -1 || l < cpg->len)
5469 + /* try stopping to update while we are referencing */
5470 + vfsub_inode_lock_shared_nested(h_src_inode, AuLsc_I_CHILD);
5471 + au_pin_hdir_unlock(cpg->pin);
5473 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5474 + h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
5475 + h_src_attr->iflags = h_src_inode->i_flags;
5476 + if (!au_test_nfs(h_src_inode->i_sb))
5477 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5479 + inode_unlock_shared(h_src_inode);
5480 + err = vfsub_getattr(&h_path, &h_src_attr->st);
5481 + vfsub_inode_lock_shared_nested(h_src_inode,
5484 + if (unlikely(err)) {
5485 + inode_unlock_shared(h_src_inode);
5488 + h_src_attr->valid = 1;
5489 + if (!au_test_nfs(h_src_inode->i_sb)) {
5490 + err = au_cp_regular(cpg);
5491 + inode_unlock_shared(h_src_inode);
5493 + inode_unlock_shared(h_src_inode);
5494 + err = au_cp_regular(cpg);
5496 + rerr = au_pin_hdir_relock(cpg->pin);
5500 + if (!err && (h_src_inode->i_state & I_LINKABLE)) {
5501 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5502 + h_dst_inode = d_inode(h_path.dentry);
5503 + spin_lock(&h_dst_inode->i_lock);
5504 + h_dst_inode->i_state |= I_LINKABLE;
5505 + spin_unlock(&h_dst_inode->i_lock);
5512 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
5513 + struct inode *h_dir)
5516 + mm_segment_t old_fs;
5523 + sym.k = (void *)__get_free_page(GFP_NOFS);
5524 + if (unlikely(!sym.k))
5527 + /* unnecessary to support mmap_sem since symlink is not mmap-able */
5528 + old_fs = get_fs();
5529 + set_fs(KERNEL_DS);
5530 + symlen = vfs_readlink(h_src, sym.u, PATH_MAX);
5535 + sym.k[symlen] = 0;
5536 + err = vfsub_symlink(h_dir, h_path, sym.k);
5538 + free_page((unsigned long)sym.k);
5545 + * regardless 'acl' option, reset all ACL.
5546 + * All ACL will be copied up later from the original entry on the lower branch.
5548 +static int au_reset_acl(struct inode *h_dir, struct path *h_path, umode_t mode)
5551 + struct dentry *h_dentry;
5552 + struct inode *h_inode;
5554 + h_dentry = h_path->dentry;
5555 + h_inode = d_inode(h_dentry);
5556 + /* forget_all_cached_acls(h_inode)); */
5557 + err = vfsub_removexattr(h_dentry, XATTR_NAME_POSIX_ACL_ACCESS);
5559 + if (err == -EOPNOTSUPP)
5562 + err = vfsub_acl_chmod(h_inode, mode);
5568 +static int au_do_cpup_dir(struct au_cp_generic *cpg, struct dentry *dst_parent,
5569 + struct inode *h_dir, struct path *h_path)
5572 + struct inode *dir, *inode;
5574 + err = vfsub_removexattr(h_path->dentry, XATTR_NAME_POSIX_ACL_DEFAULT);
5576 + if (err == -EOPNOTSUPP)
5578 + if (unlikely(err))
5582 + * strange behaviour from the users view,
5583 + * particularry setattr case
5585 + dir = d_inode(dst_parent);
5586 + if (au_ibtop(dir) == cpg->bdst)
5587 + au_cpup_attr_nlink(dir, /*force*/1);
5588 + inode = d_inode(cpg->dentry);
5589 + au_cpup_attr_nlink(inode, /*force*/1);
5595 +static noinline_for_stack
5596 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5597 + struct au_cpup_reg_attr *h_src_attr)
5601 + unsigned int mnt_flags;
5602 + unsigned char isdir, isreg, force;
5603 + const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5604 + struct au_dtime dt;
5605 + struct path h_path;
5606 + struct dentry *h_src, *h_dst, *h_parent;
5607 + struct inode *h_inode, *h_dir;
5608 + struct super_block *sb;
5610 + /* bsrc branch can be ro/rw. */
5611 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5612 + h_inode = d_inode(h_src);
5613 + AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5615 + /* try stopping to be referenced while we are creating */
5616 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5617 + if (au_ftest_cpup(cpg->flags, RENAME))
5618 + AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5619 + AUFS_WH_PFX_LEN));
5620 + h_parent = h_dst->d_parent; /* dir inode is locked */
5621 + h_dir = d_inode(h_parent);
5623 + AuDebugOn(h_parent != h_dst->d_parent);
5625 + sb = cpg->dentry->d_sb;
5626 + h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5628 + h_path.dentry = h_parent;
5629 + au_dtime_store(&dt, dst_parent, &h_path);
5631 + h_path.dentry = h_dst;
5635 + mode = h_inode->i_mode;
5636 + switch (mode & S_IFMT) {
5639 + err = vfsub_create(h_dir, &h_path, S_IRUSR | S_IWUSR,
5640 + /*want_excl*/true);
5642 + err = au_do_cpup_regular(cpg, h_src_attr);
5646 + err = vfsub_mkdir(h_dir, &h_path, mode);
5648 + err = au_do_cpup_dir(cpg, dst_parent, h_dir, &h_path);
5651 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5655 + AuDebugOn(!capable(CAP_MKNOD));
5659 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5662 + AuIOErr("Unknown inode type 0%o\n", mode);
5666 + err = au_reset_acl(h_dir, &h_path, mode);
5668 + mnt_flags = au_mntflags(sb);
5669 + if (!au_opt_test(mnt_flags, UDBA_NONE)
5671 + && au_opt_test(mnt_flags, XINO)
5672 + && (h_inode->i_nlink == 1
5673 + || (h_inode->i_state & I_LINKABLE))
5674 + /* todo: unnecessary? */
5675 + /* && d_inode(cpg->dentry)->i_nlink == 1 */
5676 + && cpg->bdst < cpg->bsrc
5677 + && !au_ftest_cpup(cpg->flags, KEEPLINO))
5678 + au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5679 + /* ignore this error */
5684 + force = !!cpg->len;
5685 + if (cpg->len == -1)
5686 + force = !!i_size_read(h_inode);
5688 + au_fhsm_wrote(sb, cpg->bdst, force);
5692 + au_dtime_revert(&dt);
5696 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5699 + struct dentry *dentry, *h_dentry, *h_parent, *parent;
5700 + struct inode *h_dir;
5701 + aufs_bindex_t bdst;
5703 + dentry = cpg->dentry;
5705 + h_dentry = au_h_dptr(dentry, bdst);
5706 + if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5708 + au_set_h_dptr(dentry, bdst, NULL);
5709 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
5711 + h_path->dentry = dget(au_h_dptr(dentry, bdst));
5712 + au_set_h_dptr(dentry, bdst, h_dentry);
5715 + parent = dget_parent(dentry);
5716 + h_parent = au_h_dptr(parent, bdst);
5718 + h_path->dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
5719 + if (IS_ERR(h_path->dentry))
5720 + err = PTR_ERR(h_path->dentry);
5722 + if (unlikely(err))
5725 + h_parent = h_dentry->d_parent; /* dir inode is locked */
5726 + h_dir = d_inode(h_parent);
5728 + AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5729 + /* no delegation since it is just created */
5730 + err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL,
5732 + dput(h_path->dentry);
5739 + * copyup the @dentry from @bsrc to @bdst.
5740 + * the caller must set the both of lower dentries.
5741 + * @len is for truncating when it is -1 copyup the entire file.
5742 + * in link/rename cases, @dst_parent may be different from the real one.
5743 + * basic->bsrc can be larger than basic->bdst.
5744 + * aufs doesn't touch the credential so
5745 + * security_inode_copy_up{,_xattr}() are unnecrssary.
5747 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5750 + aufs_bindex_t old_ibtop;
5751 + unsigned char isdir, plink;
5752 + struct dentry *h_src, *h_dst, *h_parent;
5753 + struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5754 + struct super_block *sb;
5755 + struct au_branch *br;
5756 + /* to reuduce stack size */
5758 + struct au_dtime dt;
5759 + struct path h_path;
5760 + struct au_cpup_reg_attr h_src_attr;
5764 + a = kmalloc(sizeof(*a), GFP_NOFS);
5767 + a->h_src_attr.valid = 0;
5769 + sb = cpg->dentry->d_sb;
5770 + br = au_sbr(sb, cpg->bdst);
5771 + a->h_path.mnt = au_br_mnt(br);
5772 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5773 + h_parent = h_dst->d_parent; /* dir inode is locked */
5774 + h_dir = d_inode(h_parent);
5777 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5778 + inode = d_inode(cpg->dentry);
5781 + dst_parent = dget_parent(cpg->dentry);
5785 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
5786 + dst_inode = au_h_iptr(inode, cpg->bdst);
5788 + if (unlikely(!plink)) {
5790 + AuIOErr("hi%lu(i%lu) exists on b%d "
5791 + "but plink is disabled\n",
5792 + dst_inode->i_ino, inode->i_ino, cpg->bdst);
5796 + if (dst_inode->i_nlink) {
5797 + const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5799 + h_src = au_plink_lkup(inode, cpg->bdst);
5800 + err = PTR_ERR(h_src);
5801 + if (IS_ERR(h_src))
5803 + if (unlikely(d_is_negative(h_src))) {
5805 + AuIOErr("i%lu exists on b%d "
5806 + "but not pseudo-linked\n",
5807 + inode->i_ino, cpg->bdst);
5813 + a->h_path.dentry = h_parent;
5814 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5817 + a->h_path.dentry = h_dst;
5819 + err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5820 + if (!err && au_ftest_cpup(cpg->flags, RENAME))
5821 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5823 + au_dtime_revert(&a->dt);
5824 + if (unlikely(err == -EWOULDBLOCK)) {
5825 + pr_warn("cannot retry for NFSv4 delegation"
5826 + " for an internal link\n");
5832 + /* todo: cpup_wh_file? */
5834 + au_update_ibrange(inode, /*do_put_zero*/1);
5837 + isdir = S_ISDIR(inode->i_mode);
5838 + old_ibtop = au_ibtop(inode);
5839 + err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5840 + if (unlikely(err))
5842 + dst_inode = d_inode(h_dst);
5843 + inode_lock_nested(dst_inode, AuLsc_I_CHILD2);
5844 + /* todo: necessary? */
5845 + /* au_pin_hdir_unlock(cpg->pin); */
5847 + err = cpup_iattr(cpg->dentry, cpg->bdst, h_src, &a->h_src_attr);
5848 + if (unlikely(err)) {
5849 + /* todo: necessary? */
5850 + /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5851 + inode_unlock(dst_inode);
5855 + if (cpg->bdst < old_ibtop) {
5856 + if (S_ISREG(inode->i_mode)) {
5857 + err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5858 + if (unlikely(err)) {
5859 + /* ignore an error */
5860 + /* au_pin_hdir_relock(cpg->pin); */
5861 + inode_unlock(dst_inode);
5865 + au_set_ibtop(inode, cpg->bdst);
5867 + au_set_ibbot(inode, cpg->bdst);
5868 + au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5869 + au_hi_flags(inode, isdir));
5871 + /* todo: necessary? */
5872 + /* err = au_pin_hdir_relock(cpg->pin); */
5873 + inode_unlock(dst_inode);
5874 + if (unlikely(err))
5877 + src_inode = d_inode(h_src);
5879 + && (src_inode->i_nlink > 1
5880 + || src_inode->i_state & I_LINKABLE)
5882 + au_plink_append(inode, cpg->bdst, h_dst);
5884 + if (au_ftest_cpup(cpg->flags, RENAME)) {
5885 + a->h_path.dentry = h_dst;
5886 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5889 + goto out_parent; /* success */
5893 + a->h_path.dentry = h_parent;
5894 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5895 + a->h_path.dentry = h_dst;
5897 + if (d_is_positive(h_dst)) {
5899 + /* no delegation since it is just created */
5900 + rerr = vfsub_unlink(h_dir, &a->h_path,
5901 + /*delegated*/NULL, /*force*/0);
5903 + rerr = vfsub_rmdir(h_dir, &a->h_path);
5905 + au_dtime_revert(&a->dt);
5907 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5917 +#if 0 /* reserved */
5918 +struct au_cpup_single_args {
5920 + struct au_cp_generic *cpg;
5921 + struct dentry *dst_parent;
5924 +static void au_call_cpup_single(void *args)
5926 + struct au_cpup_single_args *a = args;
5928 + au_pin_hdir_acquire_nest(a->cpg->pin);
5929 + *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5930 + au_pin_hdir_release(a->cpg->pin);
5935 + * prevent SIGXFSZ in copy-up.
5936 + * testing CAP_MKNOD is for generic fs,
5937 + * but CAP_FSETID is for xfs only, currently.
5939 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5942 + struct super_block *sb;
5943 + struct inode *h_dir;
5946 + sb = au_pinned_parent(pin)->d_sb;
5947 + if (!au_wkq_test()
5948 + && (!au_sbi(sb)->si_plink_maint_pid
5949 + || au_plink_maint(sb, AuLock_NOPLM))) {
5950 + switch (mode & S_IFMT) {
5952 + /* no condition about RLIMIT_FSIZE and the file size */
5957 + do_sio = !capable(CAP_MKNOD);
5961 + do_sio = ((mode & (S_ISUID | S_ISGID))
5962 + && !capable(CAP_FSETID));
5963 + /* this workaround may be removed in the future */
5965 + h_dir = au_pinned_h_dir(pin);
5966 + do_sio = h_dir->i_mode & S_ISVTX;
5973 +#if 0 /* reserved */
5974 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5977 + struct dentry *h_dentry;
5979 + h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5980 + if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5981 + err = au_cpup_single(cpg, dst_parent);
5983 + struct au_cpup_single_args args = {
5986 + .dst_parent = dst_parent
5988 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5989 + if (unlikely(wkq_err))
5998 + * copyup the @dentry from the first active lower branch to @bdst,
5999 + * using au_cpup_single().
6001 +static int au_cpup_simple(struct au_cp_generic *cpg)
6004 + unsigned int flags_orig;
6005 + struct dentry *dentry;
6007 + AuDebugOn(cpg->bsrc < 0);
6009 + dentry = cpg->dentry;
6010 + DiMustWriteLock(dentry);
6012 + err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
6014 + flags_orig = cpg->flags;
6015 + au_fset_cpup(cpg->flags, RENAME);
6016 + err = au_cpup_single(cpg, NULL);
6017 + cpg->flags = flags_orig;
6019 + return 0; /* success */
6022 + au_set_h_dptr(dentry, cpg->bdst, NULL);
6023 + au_set_dbtop(dentry, cpg->bsrc);
6029 +struct au_cpup_simple_args {
6031 + struct au_cp_generic *cpg;
6034 +static void au_call_cpup_simple(void *args)
6036 + struct au_cpup_simple_args *a = args;
6038 + au_pin_hdir_acquire_nest(a->cpg->pin);
6039 + *a->errp = au_cpup_simple(a->cpg);
6040 + au_pin_hdir_release(a->cpg->pin);
6043 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
6046 + struct dentry *dentry, *parent;
6047 + struct file *h_file;
6048 + struct inode *h_dir;
6050 + dentry = cpg->dentry;
6052 + if (au_ftest_cpup(cpg->flags, HOPEN)) {
6053 + AuDebugOn(cpg->bsrc < 0);
6054 + h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
6055 + err = PTR_ERR(h_file);
6056 + if (IS_ERR(h_file))
6060 + parent = dget_parent(dentry);
6061 + h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
6062 + if (!au_test_h_perm_sio(h_dir, MAY_EXEC | MAY_WRITE)
6063 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6064 + err = au_cpup_simple(cpg);
6066 + struct au_cpup_simple_args args = {
6070 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
6071 + if (unlikely(wkq_err))
6077 + au_h_open_post(dentry, cpg->bsrc, h_file);
6083 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
6085 + aufs_bindex_t bsrc, bbot;
6086 + struct dentry *dentry, *h_dentry;
6088 + if (cpg->bsrc < 0) {
6089 + dentry = cpg->dentry;
6090 + bbot = au_dbbot(dentry);
6091 + for (bsrc = cpg->bdst + 1; bsrc <= bbot; bsrc++) {
6092 + h_dentry = au_h_dptr(dentry, bsrc);
6094 + AuDebugOn(d_is_negative(h_dentry));
6098 + AuDebugOn(bsrc > bbot);
6101 + AuDebugOn(cpg->bsrc <= cpg->bdst);
6102 + return au_do_sio_cpup_simple(cpg);
6105 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
6107 + AuDebugOn(cpg->bdst <= cpg->bsrc);
6108 + return au_do_sio_cpup_simple(cpg);
6111 +/* ---------------------------------------------------------------------- */
6114 + * copyup the deleted file for writing.
6116 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
6117 + struct file *file)
6120 + unsigned int flags_orig;
6121 + aufs_bindex_t bsrc_orig;
6122 + struct au_dinfo *dinfo;
6124 + struct au_hdentry *hd;
6125 + struct dentry *h_dentry;
6128 + dinfo = au_di(cpg->dentry);
6129 + AuRwMustWriteLock(&dinfo->di_rwsem);
6131 + bsrc_orig = cpg->bsrc;
6132 + cpg->bsrc = dinfo->di_btop;
6133 + hdst.hd = au_hdentry(dinfo, cpg->bdst);
6134 + hdst.h_dentry = hdst.hd->hd_dentry;
6135 + hdst.hd->hd_dentry = wh_dentry;
6136 + dinfo->di_btop = cpg->bdst;
6138 + hsrc.h_dentry = NULL;
6140 + hsrc.hd = au_hdentry(dinfo, cpg->bsrc);
6141 + hsrc.h_dentry = hsrc.hd->hd_dentry;
6142 + hsrc.hd->hd_dentry = au_hf_top(file)->f_path.dentry;
6144 + flags_orig = cpg->flags;
6145 + cpg->flags = !AuCpup_DTIME;
6146 + err = au_cpup_single(cpg, /*h_parent*/NULL);
6147 + cpg->flags = flags_orig;
6150 + err = au_reopen_nondir(file);
6151 + hsrc.hd->hd_dentry = hsrc.h_dentry;
6153 + hdst.hd->hd_dentry = hdst.h_dentry;
6154 + dinfo->di_btop = cpg->bsrc;
6155 + cpg->bsrc = bsrc_orig;
6160 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6163 + aufs_bindex_t bdst;
6164 + struct au_dtime dt;
6165 + struct dentry *dentry, *parent, *h_parent, *wh_dentry;
6166 + struct au_branch *br;
6167 + struct path h_path;
6169 + dentry = cpg->dentry;
6171 + br = au_sbr(dentry->d_sb, bdst);
6172 + parent = dget_parent(dentry);
6173 + h_parent = au_h_dptr(parent, bdst);
6174 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
6175 + err = PTR_ERR(wh_dentry);
6176 + if (IS_ERR(wh_dentry))
6179 + h_path.dentry = h_parent;
6180 + h_path.mnt = au_br_mnt(br);
6181 + au_dtime_store(&dt, parent, &h_path);
6182 + err = au_do_cpup_wh(cpg, wh_dentry, file);
6183 + if (unlikely(err))
6187 + h_path.dentry = wh_dentry;
6188 + if (!d_is_dir(wh_dentry)) {
6189 + /* no delegation since it is just created */
6190 + err = vfsub_unlink(d_inode(h_parent), &h_path,
6191 + /*delegated*/NULL, /*force*/0);
6193 + err = vfsub_rmdir(d_inode(h_parent), &h_path);
6194 + if (unlikely(err)) {
6195 + AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
6199 + au_dtime_revert(&dt);
6200 + au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
6209 +struct au_cpup_wh_args {
6211 + struct au_cp_generic *cpg;
6212 + struct file *file;
6215 +static void au_call_cpup_wh(void *args)
6217 + struct au_cpup_wh_args *a = args;
6219 + au_pin_hdir_acquire_nest(a->cpg->pin);
6220 + *a->errp = au_cpup_wh(a->cpg, a->file);
6221 + au_pin_hdir_release(a->cpg->pin);
6224 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
6227 + aufs_bindex_t bdst;
6228 + struct dentry *dentry, *parent, *h_orph, *h_parent;
6229 + struct inode *dir, *h_dir, *h_tmpdir;
6230 + struct au_wbr *wbr;
6231 + struct au_pin wh_pin, *pin_orig;
6233 + dentry = cpg->dentry;
6235 + parent = dget_parent(dentry);
6236 + dir = d_inode(parent);
6239 + h_dir = au_igrab(au_h_iptr(dir, bdst));
6242 + if (!h_dir->i_nlink) {
6243 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
6244 + h_orph = wbr->wbr_orph;
6246 + h_parent = dget(au_h_dptr(parent, bdst));
6247 + au_set_h_dptr(parent, bdst, dget(h_orph));
6248 + h_tmpdir = d_inode(h_orph);
6249 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
6251 + inode_lock_nested(h_tmpdir, AuLsc_I_PARENT3);
6252 + /* todo: au_h_open_pre()? */
6254 + pin_orig = cpg->pin;
6255 + au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
6256 + AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
6257 + cpg->pin = &wh_pin;
6260 + if (!au_test_h_perm_sio(h_tmpdir, MAY_EXEC | MAY_WRITE)
6261 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
6262 + err = au_cpup_wh(cpg, file);
6264 + struct au_cpup_wh_args args = {
6269 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
6270 + if (unlikely(wkq_err))
6275 + inode_unlock(h_tmpdir);
6276 + /* todo: au_h_open_post()? */
6277 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
6278 + au_set_h_dptr(parent, bdst, h_parent);
6279 + AuDebugOn(!pin_orig);
6280 + cpg->pin = pin_orig;
6288 +/* ---------------------------------------------------------------------- */
6291 + * generic routine for both of copy-up and copy-down.
6293 +/* cf. revalidate function in file.c */
6294 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6295 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6296 + struct au_pin *pin,
6297 + struct dentry *h_parent, void *arg),
6301 + struct au_pin pin;
6302 + struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
6305 + parent = dget_parent(dentry);
6306 + if (IS_ROOT(parent))
6309 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
6310 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
6312 + /* do not use au_dpage */
6313 + real_parent = parent;
6316 + parent = dget_parent(dentry);
6317 + h_parent = au_h_dptr(parent, bdst);
6319 + goto out; /* success */
6321 + /* find top dir which is necessary to cpup */
6325 + parent = dget_parent(d);
6326 + di_read_lock_parent3(parent, !AuLock_IR);
6327 + h_parent = au_h_dptr(parent, bdst);
6328 + di_read_unlock(parent, !AuLock_IR);
6329 + } while (!h_parent);
6331 + if (d != real_parent)
6332 + di_write_lock_child3(d);
6334 + /* somebody else might create while we were sleeping */
6335 + h_dentry = au_h_dptr(d, bdst);
6336 + if (!h_dentry || d_is_negative(h_dentry)) {
6338 + au_update_dbtop(d);
6340 + au_pin_set_dentry(&pin, d);
6341 + err = au_do_pin(&pin);
6343 + err = cp(d, bdst, &pin, h_parent, arg);
6348 + if (d != real_parent)
6349 + di_write_unlock(d);
6350 + if (unlikely(err))
6359 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
6360 + struct au_pin *pin,
6361 + struct dentry *h_parent __maybe_unused,
6362 + void *arg __maybe_unused)
6364 + struct au_cp_generic cpg = {
6370 + .flags = AuCpup_DTIME
6372 + return au_sio_cpup_simple(&cpg);
6375 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6377 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
6380 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
6383 + struct dentry *parent;
6384 + struct inode *dir;
6386 + parent = dget_parent(dentry);
6387 + dir = d_inode(parent);
6389 + if (au_h_iptr(dir, bdst))
6392 + di_read_unlock(parent, AuLock_IR);
6393 + di_write_lock_parent(parent);
6394 + /* someone else might change our inode while we were sleeping */
6395 + if (!au_h_iptr(dir, bdst))
6396 + err = au_cpup_dirs(dentry, bdst);
6397 + di_downgrade_lock(parent, AuLock_IR);
6403 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
6404 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
6405 +++ linux/fs/aufs/cpup.h 2018-01-29 07:56:20.053325069 +0100
6408 + * Copyright (C) 2005-2017 Junjiro R. Okajima
6410 + * This program, aufs is free software; you can redistribute it and/or modify
6411 + * it under the terms of the GNU General Public License as published by
6412 + * the Free Software Foundation; either version 2 of the License, or
6413 + * (at your option) any later version.
6415 + * This program is distributed in the hope that it will be useful,
6416 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6417 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6418 + * GNU General Public License for more details.
6420 + * You should have received a copy of the GNU General Public License
6421 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6425 + * copy-up/down functions
6428 +#ifndef __AUFS_CPUP_H__
6429 +#define __AUFS_CPUP_H__
6433 +#include <linux/path.h>
6439 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
6440 +void au_cpup_attr_timesizes(struct inode *inode);
6441 +void au_cpup_attr_nlink(struct inode *inode, int force);
6442 +void au_cpup_attr_changeable(struct inode *inode);
6443 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
6444 +void au_cpup_attr_all(struct inode *inode, int force);
6446 +/* ---------------------------------------------------------------------- */
6448 +struct au_cp_generic {
6449 + struct dentry *dentry;
6450 + aufs_bindex_t bdst, bsrc;
6452 + struct au_pin *pin;
6453 + unsigned int flags;
6457 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
6458 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
6460 +#define AuCpup_RENAME (1 << 2) /* rename after cpup */
6461 +#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
6463 +#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
6465 +#define AuCpup_RWDST (1 << 5) /* force write target even if
6466 + the branch is marked as RO */
6468 +#ifndef CONFIG_AUFS_BR_HFSPLUS
6469 +#undef AuCpup_HOPEN
6470 +#define AuCpup_HOPEN 0
6473 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
6474 +#define au_fset_cpup(flags, name) \
6475 + do { (flags) |= AuCpup_##name; } while (0)
6476 +#define au_fclr_cpup(flags, name) \
6477 + do { (flags) &= ~AuCpup_##name; } while (0)
6479 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
6480 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
6481 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
6482 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
6484 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
6485 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
6486 + struct au_pin *pin,
6487 + struct dentry *h_parent, void *arg),
6489 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6490 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
6492 +/* ---------------------------------------------------------------------- */
6494 +/* keep timestamps when copyup */
6496 + struct dentry *dt_dentry;
6497 + struct path dt_h_path;
6498 + struct timespec dt_atime, dt_mtime;
6500 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
6501 + struct path *h_path);
6502 +void au_dtime_revert(struct au_dtime *dt);
6504 +#endif /* __KERNEL__ */
6505 +#endif /* __AUFS_CPUP_H__ */
6506 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
6507 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
6508 +++ linux/fs/aufs/dbgaufs.c 2018-01-29 07:56:20.053325069 +0100
6511 + * Copyright (C) 2005-2017 Junjiro R. Okajima
6513 + * This program, aufs is free software; you can redistribute it and/or modify
6514 + * it under the terms of the GNU General Public License as published by
6515 + * the Free Software Foundation; either version 2 of the License, or
6516 + * (at your option) any later version.
6518 + * This program is distributed in the hope that it will be useful,
6519 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6520 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6521 + * GNU General Public License for more details.
6523 + * You should have received a copy of the GNU General Public License
6524 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6528 + * debugfs interface
6531 +#include <linux/debugfs.h>
6534 +#ifndef CONFIG_SYSFS
6535 +#error DEBUG_FS depends upon SYSFS
6538 +static struct dentry *dbgaufs;
6539 +static const mode_t dbgaufs_mode = S_IRUSR | S_IRGRP | S_IROTH;
6541 +/* 20 is max digits length of ulong 64 */
6542 +struct dbgaufs_arg {
6548 + * common function for all XINO files
6550 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
6551 + struct file *file)
6553 + kfree(file->private_data);
6557 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt)
6561 + struct dbgaufs_arg *p;
6564 + p = kmalloc(sizeof(*p), GFP_NOFS);
6570 + file->private_data = p;
6574 + err = vfsub_getattr(&xf->f_path, &st);
6578 + (p->a, sizeof(p->a), "%ld, %llux%u %lld\n",
6579 + (long)file_count(xf), st.blocks, st.blksize,
6580 + (long long)st.size);
6582 + p->n = snprintf(p->a, sizeof(p->a), "%llux%u %lld\n",
6583 + st.blocks, st.blksize,
6584 + (long long)st.size);
6585 + AuDebugOn(p->n >= sizeof(p->a));
6587 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
6596 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
6597 + size_t count, loff_t *ppos)
6599 + struct dbgaufs_arg *p;
6601 + p = file->private_data;
6602 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6605 +/* ---------------------------------------------------------------------- */
6607 +struct dbgaufs_plink_arg {
6612 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6613 + struct file *file)
6615 + free_page((unsigned long)file->private_data);
6619 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6621 + int err, i, limit;
6622 + unsigned long n, sum;
6623 + struct dbgaufs_plink_arg *p;
6624 + struct au_sbinfo *sbinfo;
6625 + struct super_block *sb;
6626 + struct hlist_bl_head *hbl;
6629 + p = (void *)get_zeroed_page(GFP_NOFS);
6634 + sbinfo = inode->i_private;
6635 + sb = sbinfo->si_sb;
6636 + si_noflush_read_lock(sb);
6637 + if (au_opt_test(au_mntflags(sb), PLINK)) {
6638 + limit = PAGE_SIZE - sizeof(p->n);
6640 + /* the number of buckets */
6641 + n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6646 + for (i = 0, hbl = sbinfo->si_plink; i < AuPlink_NHASH;
6648 + n = au_hbl_count(hbl);
6651 + n = snprintf(p->a + p->n, limit, "%lu ", n);
6654 + if (unlikely(limit <= 0))
6657 + p->a[p->n - 1] = '\n';
6659 + /* the sum of plinks */
6660 + n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6663 + if (unlikely(limit <= 0))
6666 +#define str "1\n0\n0\n"
6667 + p->n = sizeof(str) - 1;
6668 + strcpy(p->a, str);
6671 + si_read_unlock(sb);
6674 + file->private_data = p;
6675 + goto out; /* success */
6678 + free_page((unsigned long)p);
6683 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6684 + size_t count, loff_t *ppos)
6686 + struct dbgaufs_plink_arg *p;
6688 + p = file->private_data;
6689 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6692 +static const struct file_operations dbgaufs_plink_fop = {
6693 + .owner = THIS_MODULE,
6694 + .open = dbgaufs_plink_open,
6695 + .release = dbgaufs_plink_release,
6696 + .read = dbgaufs_plink_read
6699 +/* ---------------------------------------------------------------------- */
6701 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6704 + struct au_sbinfo *sbinfo;
6705 + struct super_block *sb;
6707 + sbinfo = inode->i_private;
6708 + sb = sbinfo->si_sb;
6709 + si_noflush_read_lock(sb);
6710 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0);
6711 + si_read_unlock(sb);
6715 +static const struct file_operations dbgaufs_xib_fop = {
6716 + .owner = THIS_MODULE,
6717 + .open = dbgaufs_xib_open,
6718 + .release = dbgaufs_xi_release,
6719 + .read = dbgaufs_xi_read
6722 +/* ---------------------------------------------------------------------- */
6724 +#define DbgaufsXi_PREFIX "xi"
6726 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6730 + struct au_sbinfo *sbinfo;
6731 + struct super_block *sb;
6733 + struct qstr *name;
6737 + name = &file->f_path.dentry->d_name;
6738 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6739 + || memcmp(name->name, DbgaufsXi_PREFIX,
6740 + sizeof(DbgaufsXi_PREFIX) - 1)))
6742 + err = kstrtol(name->name + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6743 + if (unlikely(err))
6746 + sbinfo = inode->i_private;
6747 + sb = sbinfo->si_sb;
6748 + si_noflush_read_lock(sb);
6749 + if (l <= au_sbbot(sb)) {
6750 + xf = au_sbr(sb, (aufs_bindex_t)l)->br_xino.xi_file;
6751 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1);
6754 + si_read_unlock(sb);
6760 +static const struct file_operations dbgaufs_xino_fop = {
6761 + .owner = THIS_MODULE,
6762 + .open = dbgaufs_xino_open,
6763 + .release = dbgaufs_xi_release,
6764 + .read = dbgaufs_xi_read
6767 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6769 + aufs_bindex_t bbot;
6770 + struct au_branch *br;
6771 + struct au_xino_file *xi;
6773 + if (!au_sbi(sb)->si_dbgaufs)
6776 + bbot = au_sbbot(sb);
6777 + for (; bindex <= bbot; bindex++) {
6778 + br = au_sbr(sb, bindex);
6779 + xi = &br->br_xino;
6780 + /* debugfs acquires the parent i_mutex */
6782 + debugfs_remove(xi->xi_dbgaufs);
6784 + xi->xi_dbgaufs = NULL;
6788 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
6790 + struct au_sbinfo *sbinfo;
6791 + struct dentry *parent;
6792 + struct au_branch *br;
6793 + struct au_xino_file *xi;
6794 + aufs_bindex_t bbot;
6795 + char name[sizeof(DbgaufsXi_PREFIX) + 5]; /* "xi" bindex NULL */
6797 + sbinfo = au_sbi(sb);
6798 + parent = sbinfo->si_dbgaufs;
6802 + bbot = au_sbbot(sb);
6803 + for (; bindex <= bbot; bindex++) {
6804 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6805 + br = au_sbr(sb, bindex);
6806 + xi = &br->br_xino;
6807 + AuDebugOn(xi->xi_dbgaufs);
6808 + /* debugfs acquires the parent i_mutex */
6810 + xi->xi_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6811 + sbinfo, &dbgaufs_xino_fop);
6813 + /* ignore an error */
6814 + if (unlikely(!xi->xi_dbgaufs))
6815 + AuWarn1("failed %s under debugfs\n", name);
6819 +/* ---------------------------------------------------------------------- */
6821 +#ifdef CONFIG_AUFS_EXPORT
6822 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6825 + struct au_sbinfo *sbinfo;
6826 + struct super_block *sb;
6828 + sbinfo = inode->i_private;
6829 + sb = sbinfo->si_sb;
6830 + si_noflush_read_lock(sb);
6831 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0);
6832 + si_read_unlock(sb);
6836 +static const struct file_operations dbgaufs_xigen_fop = {
6837 + .owner = THIS_MODULE,
6838 + .open = dbgaufs_xigen_open,
6839 + .release = dbgaufs_xi_release,
6840 + .read = dbgaufs_xi_read
6843 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6848 + * This function is a dynamic '__init' function actually,
6849 + * so the tiny check for si_rwsem is unnecessary.
6851 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6854 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
6855 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6856 + &dbgaufs_xigen_fop);
6857 + if (sbinfo->si_dbgaufs_xigen)
6863 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6867 +#endif /* CONFIG_AUFS_EXPORT */
6869 +/* ---------------------------------------------------------------------- */
6871 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6874 + * This function is a dynamic '__fin' function actually,
6875 + * so the tiny check for si_rwsem is unnecessary.
6877 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6879 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
6880 + sbinfo->si_dbgaufs = NULL;
6881 + kobject_put(&sbinfo->si_kobj);
6884 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6887 + char name[SysaufsSiNameLen];
6890 + * This function is a dynamic '__init' function actually,
6891 + * so the tiny check for si_rwsem is unnecessary.
6893 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6897 + AuErr1("/debug/aufs is uninitialized\n");
6902 + sysaufs_name(sbinfo, name);
6903 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6904 + if (unlikely(!sbinfo->si_dbgaufs))
6906 + kobject_get(&sbinfo->si_kobj);
6908 + sbinfo->si_dbgaufs_xib = debugfs_create_file
6909 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6910 + &dbgaufs_xib_fop);
6911 + if (unlikely(!sbinfo->si_dbgaufs_xib))
6914 + sbinfo->si_dbgaufs_plink = debugfs_create_file
6915 + ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6916 + &dbgaufs_plink_fop);
6917 + if (unlikely(!sbinfo->si_dbgaufs_plink))
6920 + err = dbgaufs_xigen_init(sbinfo);
6922 + goto out; /* success */
6925 + dbgaufs_si_fin(sbinfo);
6930 +/* ---------------------------------------------------------------------- */
6932 +void dbgaufs_fin(void)
6934 + debugfs_remove(dbgaufs);
6937 +int __init dbgaufs_init(void)
6942 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6947 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6948 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
6949 +++ linux/fs/aufs/dbgaufs.h 2018-01-29 07:56:20.053325069 +0100
6952 + * Copyright (C) 2005-2017 Junjiro R. Okajima
6954 + * This program, aufs is free software; you can redistribute it and/or modify
6955 + * it under the terms of the GNU General Public License as published by
6956 + * the Free Software Foundation; either version 2 of the License, or
6957 + * (at your option) any later version.
6959 + * This program is distributed in the hope that it will be useful,
6960 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6961 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6962 + * GNU General Public License for more details.
6964 + * You should have received a copy of the GNU General Public License
6965 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6969 + * debugfs interface
6972 +#ifndef __DBGAUFS_H__
6973 +#define __DBGAUFS_H__
6977 +struct super_block;
6980 +#ifdef CONFIG_DEBUG_FS
6982 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6983 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
6984 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6985 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6986 +void dbgaufs_fin(void);
6987 +int __init dbgaufs_init(void);
6989 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6990 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
6991 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6992 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6993 +AuStubVoid(dbgaufs_fin, void)
6994 +AuStubInt0(__init dbgaufs_init, void)
6995 +#endif /* CONFIG_DEBUG_FS */
6997 +#endif /* __KERNEL__ */
6998 +#endif /* __DBGAUFS_H__ */
6999 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
7000 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
7001 +++ linux/fs/aufs/dcsub.c 2018-01-29 07:56:20.053325069 +0100
7004 + * Copyright (C) 2005-2017 Junjiro R. Okajima
7006 + * This program, aufs is free software; you can redistribute it and/or modify
7007 + * it under the terms of the GNU General Public License as published by
7008 + * the Free Software Foundation; either version 2 of the License, or
7009 + * (at your option) any later version.
7011 + * This program is distributed in the hope that it will be useful,
7012 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7013 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7014 + * GNU General Public License for more details.
7016 + * You should have received a copy of the GNU General Public License
7017 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7021 + * sub-routines for dentry cache
7026 +static void au_dpage_free(struct au_dpage *dpage)
7029 + struct dentry **p;
7031 + p = dpage->dentries;
7032 + for (i = 0; i < dpage->ndentry; i++)
7034 + free_page((unsigned long)dpage->dentries);
7037 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
7043 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
7044 + if (unlikely(!dpages->dpages))
7047 + p = (void *)__get_free_page(gfp);
7051 + dpages->dpages[0].ndentry = 0;
7052 + dpages->dpages[0].dentries = p;
7053 + dpages->ndpage = 1;
7054 + return 0; /* success */
7057 + kfree(dpages->dpages);
7062 +void au_dpages_free(struct au_dcsub_pages *dpages)
7065 + struct au_dpage *p;
7067 + p = dpages->dpages;
7068 + for (i = 0; i < dpages->ndpage; i++)
7069 + au_dpage_free(p++);
7070 + kfree(dpages->dpages);
7073 +static int au_dpages_append(struct au_dcsub_pages *dpages,
7074 + struct dentry *dentry, gfp_t gfp)
7077 + struct au_dpage *dpage;
7080 + dpage = dpages->dpages + dpages->ndpage - 1;
7081 + sz = PAGE_SIZE / sizeof(dentry);
7082 + if (unlikely(dpage->ndentry >= sz)) {
7083 + AuLabel(new dpage);
7085 + sz = dpages->ndpage * sizeof(*dpages->dpages);
7086 + p = au_kzrealloc(dpages->dpages, sz,
7087 + sz + sizeof(*dpages->dpages), gfp,
7092 + dpages->dpages = p;
7093 + dpage = dpages->dpages + dpages->ndpage;
7094 + p = (void *)__get_free_page(gfp);
7098 + dpage->ndentry = 0;
7099 + dpage->dentries = p;
7103 + AuDebugOn(au_dcount(dentry) <= 0);
7104 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
7105 + return 0; /* success */
7111 +/* todo: BAD approach */
7112 +/* copied from linux/fs/dcache.c */
7120 +extern void d_walk(struct dentry *parent, void *data,
7121 + enum d_walk_ret (*enter)(void *, struct dentry *),
7122 + void (*finish)(void *));
7124 +struct ac_dpages_arg {
7126 + struct au_dcsub_pages *dpages;
7127 + struct super_block *sb;
7128 + au_dpages_test test;
7132 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
7134 + enum d_walk_ret ret;
7135 + struct ac_dpages_arg *arg = _arg;
7137 + ret = D_WALK_CONTINUE;
7138 + if (dentry->d_sb == arg->sb
7139 + && !IS_ROOT(dentry)
7140 + && au_dcount(dentry) > 0
7142 + && (!arg->test || arg->test(dentry, arg->arg))) {
7143 + arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
7144 + if (unlikely(arg->err))
7145 + ret = D_WALK_QUIT;
7151 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7152 + au_dpages_test test, void *arg)
7154 + struct ac_dpages_arg args = {
7162 + d_walk(root, &args, au_call_dpages_append, NULL);
7167 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7168 + int do_include, au_dpages_test test, void *arg)
7173 + write_seqlock(&rename_lock);
7174 + spin_lock(&dentry->d_lock);
7176 + && au_dcount(dentry) > 0
7177 + && (!test || test(dentry, arg)))
7178 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7179 + spin_unlock(&dentry->d_lock);
7180 + if (unlikely(err))
7184 + * RCU for vfsmount is unnecessary since this is a traverse in a single
7187 + while (!IS_ROOT(dentry)) {
7188 + dentry = dentry->d_parent; /* rename_lock is locked */
7189 + spin_lock(&dentry->d_lock);
7190 + if (au_dcount(dentry) > 0
7191 + && (!test || test(dentry, arg)))
7192 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
7193 + spin_unlock(&dentry->d_lock);
7194 + if (unlikely(err))
7199 + write_sequnlock(&rename_lock);
7203 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
7205 + return au_di(dentry) && dentry->d_sb == arg;
7208 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7209 + struct dentry *dentry, int do_include)
7211 + return au_dcsub_pages_rev(dpages, dentry, do_include,
7212 + au_dcsub_dpages_aufs, dentry->d_sb);
7215 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
7217 + struct path path[2] = {
7226 + return path_is_under(path + 0, path + 1);
7228 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
7229 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
7230 +++ linux/fs/aufs/dcsub.h 2018-01-29 07:56:20.053325069 +0100
7233 + * Copyright (C) 2005-2017 Junjiro R. Okajima
7235 + * This program, aufs is free software; you can redistribute it and/or modify
7236 + * it under the terms of the GNU General Public License as published by
7237 + * the Free Software Foundation; either version 2 of the License, or
7238 + * (at your option) any later version.
7240 + * This program is distributed in the hope that it will be useful,
7241 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7242 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7243 + * GNU General Public License for more details.
7245 + * You should have received a copy of the GNU General Public License
7246 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7250 + * sub-routines for dentry cache
7253 +#ifndef __AUFS_DCSUB_H__
7254 +#define __AUFS_DCSUB_H__
7258 +#include <linux/dcache.h>
7259 +#include <linux/fs.h>
7263 + struct dentry **dentries;
7266 +struct au_dcsub_pages {
7268 + struct au_dpage *dpages;
7271 +/* ---------------------------------------------------------------------- */
7274 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
7275 +void au_dpages_free(struct au_dcsub_pages *dpages);
7276 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
7277 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
7278 + au_dpages_test test, void *arg);
7279 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
7280 + int do_include, au_dpages_test test, void *arg);
7281 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
7282 + struct dentry *dentry, int do_include);
7283 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
7285 +/* ---------------------------------------------------------------------- */
7288 + * todo: in linux-3.13, several similar (but faster) helpers are added to
7289 + * include/linux/dcache.h. Try them (in the future).
7292 +static inline int au_d_hashed_positive(struct dentry *d)
7295 + struct inode *inode = d_inode(d);
7298 + if (unlikely(d_unhashed(d)
7299 + || d_is_negative(d)
7300 + || !inode->i_nlink))
7305 +static inline int au_d_linkable(struct dentry *d)
7308 + struct inode *inode = d_inode(d);
7310 + err = au_d_hashed_positive(d);
7312 + && d_is_positive(d)
7313 + && (inode->i_state & I_LINKABLE))
7318 +static inline int au_d_alive(struct dentry *d)
7321 + struct inode *inode;
7325 + err = au_d_hashed_positive(d);
7327 + inode = d_inode(d);
7328 + if (unlikely(d_unlinked(d)
7329 + || d_is_negative(d)
7330 + || !inode->i_nlink))
7336 +static inline int au_alive_dir(struct dentry *d)
7340 + err = au_d_alive(d);
7341 + if (unlikely(err || IS_DEADDIR(d_inode(d))))
7346 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
7348 + return a->len == b->len
7349 + && !memcmp(a->name, b->name, a->len);
7354 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
7356 + * the type of d_lockref.count became int, but the inlined function d_count()
7357 + * still returns unsigned int.
7358 + * I don't know why. Maybe it is for every d_count() users?
7359 + * Anyway au_dcount() lives on.
7361 +static inline int au_dcount(struct dentry *d)
7363 + return (int)d_count(d);
7366 +#endif /* __KERNEL__ */
7367 +#endif /* __AUFS_DCSUB_H__ */
7368 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
7369 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
7370 +++ linux/fs/aufs/debug.c 2018-01-29 07:56:20.053325069 +0100
7373 + * Copyright (C) 2005-2017 Junjiro R. Okajima
7375 + * This program, aufs is free software; you can redistribute it and/or modify
7376 + * it under the terms of the GNU General Public License as published by
7377 + * the Free Software Foundation; either version 2 of the License, or
7378 + * (at your option) any later version.
7380 + * This program is distributed in the hope that it will be useful,
7381 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7382 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7383 + * GNU General Public License for more details.
7385 + * You should have received a copy of the GNU General Public License
7386 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7390 + * debug print functions
7395 +/* Returns 0, or -errno. arg is in kp->arg. */
7396 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
7400 + err = kstrtoint(val, 0, &n);
7410 +/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
7411 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
7416 + return sprintf(buffer, "%d", atomic_read(a));
7419 +static struct kernel_param_ops param_ops_atomic_t = {
7420 + .set = param_atomic_t_set,
7421 + .get = param_atomic_t_get
7422 + /* void (*free)(void *arg) */
7425 +atomic_t aufs_debug = ATOMIC_INIT(0);
7426 +MODULE_PARM_DESC(debug, "debug print");
7427 +module_param_named(debug, aufs_debug, atomic_t, S_IRUGO | S_IWUSR | S_IWGRP);
7429 +DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
7430 +char *au_plevel = KERN_DEBUG;
7431 +#define dpri(fmt, ...) do { \
7433 + && strcmp(au_plevel, KERN_DEBUG)) \
7434 + || au_debug_test()) \
7435 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
7438 +/* ---------------------------------------------------------------------- */
7440 +void au_dpri_whlist(struct au_nhash *whlist)
7442 + unsigned long ul, n;
7443 + struct hlist_head *head;
7444 + struct au_vdir_wh *pos;
7446 + n = whlist->nh_num;
7447 + head = whlist->nh_head;
7448 + for (ul = 0; ul < n; ul++) {
7449 + hlist_for_each_entry(pos, head, wh_hash)
7450 + dpri("b%d, %.*s, %d\n",
7452 + pos->wh_str.len, pos->wh_str.name,
7458 +void au_dpri_vdir(struct au_vdir *vdir)
7461 + union au_vdir_deblk_p p;
7464 + if (!vdir || IS_ERR(vdir)) {
7465 + dpri("err %ld\n", PTR_ERR(vdir));
7469 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %lu\n",
7470 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
7471 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
7472 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
7473 + p.deblk = vdir->vd_deblk[ul];
7475 + dpri("[%lu]: %p\n", ul, o);
7479 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
7480 + struct dentry *wh)
7485 + if (!inode || IS_ERR(inode)) {
7486 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
7490 + /* the type of i_blocks depends upon CONFIG_LBDAF */
7491 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
7492 + && sizeof(inode->i_blocks) != sizeof(u64));
7494 + n = (void *)wh->d_name.name;
7495 + l = wh->d_name.len;
7498 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
7499 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
7501 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
7502 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
7503 + i_size_read(inode), (unsigned long long)inode->i_blocks,
7504 + hn, (long long)timespec_to_ns(&inode->i_ctime) & 0x0ffff,
7505 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
7506 + inode->i_state, inode->i_flags, inode->i_version,
7507 + inode->i_generation,
7508 + l ? ", wh " : "", l, n);
7512 +void au_dpri_inode(struct inode *inode)
7514 + struct au_iinfo *iinfo;
7515 + struct au_hinode *hi;
7516 + aufs_bindex_t bindex;
7519 + err = do_pri_inode(-1, inode, -1, NULL);
7520 + if (err || !au_test_aufs(inode->i_sb) || au_is_bad_inode(inode))
7523 + iinfo = au_ii(inode);
7524 + dpri("i-1: btop %d, bbot %d, gen %d\n",
7525 + iinfo->ii_btop, iinfo->ii_bbot, au_iigen(inode, NULL));
7526 + if (iinfo->ii_btop < 0)
7529 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot; bindex++) {
7530 + hi = au_hinode(iinfo, bindex);
7532 + do_pri_inode(bindex, hi->hi_inode, hn, hi->hi_whdentry);
7536 +void au_dpri_dalias(struct inode *inode)
7540 + spin_lock(&inode->i_lock);
7541 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
7542 + au_dpri_dentry(d);
7543 + spin_unlock(&inode->i_lock);
7546 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
7548 + struct dentry *wh = NULL;
7550 + struct inode *inode;
7551 + struct au_iinfo *iinfo;
7552 + struct au_hinode *hi;
7554 + if (!dentry || IS_ERR(dentry)) {
7555 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
7558 + /* do not call dget_parent() here */
7559 + /* note: access d_xxx without d_lock */
7560 + dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
7561 + bindex, dentry, dentry,
7562 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
7563 + au_dcount(dentry), dentry->d_flags,
7564 + d_unhashed(dentry) ? "un" : "");
7567 + if (d_is_positive(dentry))
7568 + inode = d_inode(dentry);
7570 + && au_test_aufs(dentry->d_sb)
7572 + && !au_is_bad_inode(inode)) {
7573 + iinfo = au_ii(inode);
7574 + hi = au_hinode(iinfo, bindex);
7576 + wh = hi->hi_whdentry;
7578 + do_pri_inode(bindex, inode, hn, wh);
7582 +void au_dpri_dentry(struct dentry *dentry)
7584 + struct au_dinfo *dinfo;
7585 + aufs_bindex_t bindex;
7588 + err = do_pri_dentry(-1, dentry);
7589 + if (err || !au_test_aufs(dentry->d_sb))
7592 + dinfo = au_di(dentry);
7595 + dpri("d-1: btop %d, bbot %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
7596 + dinfo->di_btop, dinfo->di_bbot,
7597 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
7598 + dinfo->di_tmpfile);
7599 + if (dinfo->di_btop < 0)
7601 + for (bindex = dinfo->di_btop; bindex <= dinfo->di_bbot; bindex++)
7602 + do_pri_dentry(bindex, au_hdentry(dinfo, bindex)->hd_dentry);
7605 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7609 + if (!file || IS_ERR(file)) {
7610 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7615 + && !IS_ERR_OR_NULL(file->f_path.dentry)
7616 + && au_test_aufs(file->f_path.dentry->d_sb)
7618 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7619 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7620 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7621 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
7622 + file->f_version, file->f_pos, a);
7623 + if (!IS_ERR_OR_NULL(file->f_path.dentry))
7624 + do_pri_dentry(bindex, file->f_path.dentry);
7628 +void au_dpri_file(struct file *file)
7630 + struct au_finfo *finfo;
7631 + struct au_fidir *fidir;
7632 + struct au_hfile *hfile;
7633 + aufs_bindex_t bindex;
7636 + err = do_pri_file(-1, file);
7638 + || IS_ERR_OR_NULL(file->f_path.dentry)
7639 + || !au_test_aufs(file->f_path.dentry->d_sb))
7642 + finfo = au_fi(file);
7645 + if (finfo->fi_btop < 0)
7647 + fidir = finfo->fi_hdir;
7649 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7651 + for (bindex = finfo->fi_btop;
7652 + bindex >= 0 && bindex <= fidir->fd_bbot;
7654 + hfile = fidir->fd_hfile + bindex;
7655 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7659 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7661 + struct vfsmount *mnt;
7662 + struct super_block *sb;
7664 + if (!br || IS_ERR(br))
7666 + mnt = au_br_mnt(br);
7667 + if (!mnt || IS_ERR(mnt))
7670 + if (!sb || IS_ERR(sb))
7673 + dpri("s%d: {perm 0x%x, id %d, cnt %lld, wbr %p}, "
7674 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7676 + bindex, br->br_perm, br->br_id, au_br_count(br),
7677 + br->br_wbr, au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7678 + sb->s_flags, sb->s_count,
7679 + atomic_read(&sb->s_active), !!br->br_xino.xi_file);
7683 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7687 +void au_dpri_sb(struct super_block *sb)
7689 + struct au_sbinfo *sbinfo;
7690 + aufs_bindex_t bindex;
7692 + /* to reuduce stack size */
7694 + struct vfsmount mnt;
7695 + struct au_branch fake;
7698 + /* this function can be called from magic sysrq */
7699 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
7700 + if (unlikely(!a)) {
7701 + dpri("no memory\n");
7705 + a->mnt.mnt_sb = sb;
7706 + a->fake.br_path.mnt = &a->mnt;
7707 + au_br_count_init(&a->fake);
7708 + err = do_pri_br(-1, &a->fake);
7709 + au_br_count_fin(&a->fake);
7711 + dpri("dev 0x%x\n", sb->s_dev);
7712 + if (err || !au_test_aufs(sb))
7715 + sbinfo = au_sbi(sb);
7718 + dpri("nw %d, gen %u, kobj %d\n",
7719 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7720 + kref_read(&sbinfo->si_kobj.kref));
7721 + for (bindex = 0; bindex <= sbinfo->si_bbot; bindex++)
7722 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7725 +/* ---------------------------------------------------------------------- */
7727 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7729 + struct inode *h_inode, *inode = d_inode(dentry);
7730 + struct dentry *h_dentry;
7731 + aufs_bindex_t bindex, bbot, bi;
7733 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7736 + bbot = au_dbbot(dentry);
7737 + bi = au_ibbot(inode);
7740 + bindex = au_dbtop(dentry);
7741 + bi = au_ibtop(inode);
7745 + for (; bindex <= bbot; bindex++) {
7746 + h_dentry = au_h_dptr(dentry, bindex);
7749 + h_inode = au_h_iptr(inode, bindex);
7750 + if (unlikely(h_inode != d_inode(h_dentry))) {
7752 + AuDbg("b%d, %s:%d\n", bindex, func, line);
7753 + AuDbgDentry(dentry);
7754 + AuDbgInode(inode);
7761 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7764 + struct au_dcsub_pages dpages;
7765 + struct au_dpage *dpage;
7766 + struct dentry **dentries;
7768 + err = au_dpages_init(&dpages, GFP_NOFS);
7770 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7772 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7773 + dpage = dpages.dpages + i;
7774 + dentries = dpage->dentries;
7775 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7776 + AuDebugOn(au_digen_test(dentries[j], sigen));
7778 + au_dpages_free(&dpages);
7781 +void au_dbg_verify_kthread(void)
7783 + if (au_wkq_test()) {
7786 + * It may be recursive, but udba=notify between two aufs mounts,
7787 + * where a single ro branch is shared, is not a problem.
7793 +/* ---------------------------------------------------------------------- */
7795 +int __init au_debug_init(void)
7797 + aufs_bindex_t bindex;
7798 + struct au_vdir_destr destr;
7801 + AuDebugOn(bindex >= 0);
7804 + AuDebugOn(destr.len < NAME_MAX);
7806 +#ifdef CONFIG_4KSTACKS
7807 + pr_warn("CONFIG_4KSTACKS is defined.\n");
7812 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7813 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
7814 +++ linux/fs/aufs/debug.h 2018-01-29 07:56:20.053325069 +0100
7817 + * Copyright (C) 2005-2017 Junjiro R. Okajima
7819 + * This program, aufs is free software; you can redistribute it and/or modify
7820 + * it under the terms of the GNU General Public License as published by
7821 + * the Free Software Foundation; either version 2 of the License, or
7822 + * (at your option) any later version.
7824 + * This program is distributed in the hope that it will be useful,
7825 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7826 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7827 + * GNU General Public License for more details.
7829 + * You should have received a copy of the GNU General Public License
7830 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7834 + * debug print functions
7837 +#ifndef __AUFS_DEBUG_H__
7838 +#define __AUFS_DEBUG_H__
7842 +#include <linux/atomic.h>
7843 +#include <linux/module.h>
7844 +#include <linux/kallsyms.h>
7845 +#include <linux/sysrq.h>
7847 +#ifdef CONFIG_AUFS_DEBUG
7848 +#define AuDebugOn(a) BUG_ON(a)
7850 +/* module parameter */
7851 +extern atomic_t aufs_debug;
7852 +static inline void au_debug_on(void)
7854 + atomic_inc(&aufs_debug);
7856 +static inline void au_debug_off(void)
7858 + atomic_dec_if_positive(&aufs_debug);
7861 +static inline int au_debug_test(void)
7863 + return atomic_read(&aufs_debug) > 0;
7866 +#define AuDebugOn(a) do {} while (0)
7867 +AuStubVoid(au_debug_on, void)
7868 +AuStubVoid(au_debug_off, void)
7869 +AuStubInt0(au_debug_test, void)
7870 +#endif /* CONFIG_AUFS_DEBUG */
7872 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7874 +/* ---------------------------------------------------------------------- */
7878 +#define AuDbg(fmt, ...) do { \
7879 + if (au_debug_test()) \
7880 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7882 +#define AuLabel(l) AuDbg(#l "\n")
7883 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7884 +#define AuWarn1(fmt, ...) do { \
7885 + static unsigned char _c; \
7887 + pr_warn(fmt, ##__VA_ARGS__); \
7890 +#define AuErr1(fmt, ...) do { \
7891 + static unsigned char _c; \
7893 + pr_err(fmt, ##__VA_ARGS__); \
7896 +#define AuIOErr1(fmt, ...) do { \
7897 + static unsigned char _c; \
7899 + AuIOErr(fmt, ##__VA_ARGS__); \
7902 +#define AuUnsupportMsg "This operation is not supported." \
7903 + " Please report this application to aufs-users ML."
7904 +#define AuUnsupport(fmt, ...) do { \
7905 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7909 +#define AuTraceErr(e) do { \
7910 + if (unlikely((e) < 0)) \
7911 + AuDbg("err %d\n", (int)(e)); \
7914 +#define AuTraceErrPtr(p) do { \
7916 + AuDbg("err %ld\n", PTR_ERR(p)); \
7919 +/* dirty macros for debug print, use with "%.*s" and caution */
7920 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
7922 +/* ---------------------------------------------------------------------- */
7925 +#ifdef CONFIG_AUFS_DEBUG
7926 +extern struct mutex au_dbg_mtx;
7927 +extern char *au_plevel;
7929 +void au_dpri_whlist(struct au_nhash *whlist);
7931 +void au_dpri_vdir(struct au_vdir *vdir);
7933 +void au_dpri_inode(struct inode *inode);
7934 +void au_dpri_dalias(struct inode *inode);
7935 +void au_dpri_dentry(struct dentry *dentry);
7937 +void au_dpri_file(struct file *filp);
7938 +struct super_block;
7939 +void au_dpri_sb(struct super_block *sb);
7941 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7942 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7943 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7944 +void au_dbg_verify_kthread(void);
7946 +int __init au_debug_init(void);
7948 +#define AuDbgWhlist(w) do { \
7949 + mutex_lock(&au_dbg_mtx); \
7951 + au_dpri_whlist(w); \
7952 + mutex_unlock(&au_dbg_mtx); \
7955 +#define AuDbgVdir(v) do { \
7956 + mutex_lock(&au_dbg_mtx); \
7958 + au_dpri_vdir(v); \
7959 + mutex_unlock(&au_dbg_mtx); \
7962 +#define AuDbgInode(i) do { \
7963 + mutex_lock(&au_dbg_mtx); \
7965 + au_dpri_inode(i); \
7966 + mutex_unlock(&au_dbg_mtx); \
7969 +#define AuDbgDAlias(i) do { \
7970 + mutex_lock(&au_dbg_mtx); \
7972 + au_dpri_dalias(i); \
7973 + mutex_unlock(&au_dbg_mtx); \
7976 +#define AuDbgDentry(d) do { \
7977 + mutex_lock(&au_dbg_mtx); \
7979 + au_dpri_dentry(d); \
7980 + mutex_unlock(&au_dbg_mtx); \
7983 +#define AuDbgFile(f) do { \
7984 + mutex_lock(&au_dbg_mtx); \
7986 + au_dpri_file(f); \
7987 + mutex_unlock(&au_dbg_mtx); \
7990 +#define AuDbgSb(sb) do { \
7991 + mutex_lock(&au_dbg_mtx); \
7992 + AuDbg(#sb "\n"); \
7994 + mutex_unlock(&au_dbg_mtx); \
7997 +#define AuDbgSym(addr) do { \
7998 + char sym[KSYM_SYMBOL_LEN]; \
7999 + sprint_symbol(sym, (unsigned long)addr); \
8000 + AuDbg("%s\n", sym); \
8003 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
8004 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
8005 +AuStubVoid(au_dbg_verify_kthread, void)
8006 +AuStubInt0(__init au_debug_init, void)
8008 +#define AuDbgWhlist(w) do {} while (0)
8009 +#define AuDbgVdir(v) do {} while (0)
8010 +#define AuDbgInode(i) do {} while (0)
8011 +#define AuDbgDAlias(i) do {} while (0)
8012 +#define AuDbgDentry(d) do {} while (0)
8013 +#define AuDbgFile(f) do {} while (0)
8014 +#define AuDbgSb(sb) do {} while (0)
8015 +#define AuDbgSym(addr) do {} while (0)
8016 +#endif /* CONFIG_AUFS_DEBUG */
8018 +/* ---------------------------------------------------------------------- */
8020 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
8021 +int __init au_sysrq_init(void);
8022 +void au_sysrq_fin(void);
8024 +#ifdef CONFIG_HW_CONSOLE
8025 +#define au_dbg_blocked() do { \
8027 + handle_sysrq('w'); \
8030 +AuStubVoid(au_dbg_blocked, void)
8034 +AuStubInt0(__init au_sysrq_init, void)
8035 +AuStubVoid(au_sysrq_fin, void)
8036 +AuStubVoid(au_dbg_blocked, void)
8037 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
8039 +#endif /* __KERNEL__ */
8040 +#endif /* __AUFS_DEBUG_H__ */
8041 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
8042 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
8043 +++ linux/fs/aufs/dentry.c 2018-01-29 07:56:20.053325069 +0100
8046 + * Copyright (C) 2005-2017 Junjiro R. Okajima
8048 + * This program, aufs is free software; you can redistribute it and/or modify
8049 + * it under the terms of the GNU General Public License as published by
8050 + * the Free Software Foundation; either version 2 of the License, or
8051 + * (at your option) any later version.
8053 + * This program is distributed in the hope that it will be useful,
8054 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8055 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8056 + * GNU General Public License for more details.
8058 + * You should have received a copy of the GNU General Public License
8059 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
8063 + * lookup and dentry operations
8066 +#include <linux/namei.h>
8070 + * returns positive/negative dentry, NULL or an error.
8071 + * NULL means whiteout-ed or not-found.
8073 +static struct dentry*
8074 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
8075 + aufs_bindex_t bindex, struct au_do_lookup_args *args)
8077 + struct dentry *h_dentry;
8078 + struct inode *h_inode;
8079 + struct au_branch *br;
8080 + int wh_found, opq;
8081 + unsigned char wh_able;
8082 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
8083 + const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
8087 + br = au_sbr(dentry->d_sb, bindex);
8088 + wh_able = !!au_br_whable(br->br_perm);
8090 + wh_found = au_wh_test(h_parent, &args->whname, ignore_perm);
8091 + h_dentry = ERR_PTR(wh_found);
8094 + if (unlikely(wh_found < 0))
8097 + /* We found a whiteout */
8098 + /* au_set_dbbot(dentry, bindex); */
8099 + au_set_dbwh(dentry, bindex);
8101 + return NULL; /* success */
8105 + h_dentry = vfsub_lkup_one(args->name, h_parent);
8107 + h_dentry = au_sio_lkup_one(args->name, h_parent);
8108 + if (IS_ERR(h_dentry)) {
8109 + if (PTR_ERR(h_dentry) == -ENAMETOOLONG
8115 + h_inode = d_inode(h_dentry);
8116 + if (d_is_negative(h_dentry)) {
8119 + } else if (wh_found
8120 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
8122 + else if (au_ftest_lkup(args->flags, DIRREN)
8124 + && !au_dr_lkup_h_ino(args, bindex, h_inode->i_ino)) {
8125 + AuDbg("b%d %pd ignored hi%llu\n", bindex, h_dentry,
8126 + (unsigned long long)h_inode->i_ino);
8130 + if (au_dbbot(dentry) <= bindex)
8131 + au_set_dbbot(dentry, bindex);
8132 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
8133 + au_set_dbtop(dentry, bindex);
8134 + au_set_h_dptr(dentry, bindex, h_dentry);
8136 + if (!d_is_dir(h_dentry)
8138 + || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
8139 + goto out; /* success */
8141 + vfsub_inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
8142 + opq = au_diropq_test(h_dentry);
8143 + inode_unlock_shared(h_inode);
8145 + au_set_dbdiropq(dentry, bindex);
8146 + else if (unlikely(opq < 0)) {
8147 + au_set_h_dptr(dentry, bindex, NULL);
8148 + h_dentry = ERR_PTR(opq);
8159 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
8161 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
8162 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
8168 + * returns the number of lower positive dentries,
8169 + * otherwise an error.
8170 + * can be called at unlinking with @type is zero.
8172 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
8173 + unsigned int flags)
8175 + int npositive, err;
8176 + aufs_bindex_t bindex, btail, bdiropq;
8177 + unsigned char isdir, dirperm1, dirren;
8178 + struct au_do_lookup_args args = {
8180 + .name = &dentry->d_name
8182 + struct dentry *parent;
8183 + struct super_block *sb;
8185 + sb = dentry->d_sb;
8186 + err = au_test_shwh(sb, args.name);
8187 + if (unlikely(err))
8190 + err = au_wh_name_alloc(&args.whname, args.name);
8191 + if (unlikely(err))
8194 + isdir = !!d_is_dir(dentry);
8195 + dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
8196 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
8198 + au_fset_lkup(args.flags, DIRREN);
8201 + parent = dget_parent(dentry);
8202 + btail = au_dbtaildir(parent);
8203 + for (bindex = btop; bindex <= btail; bindex++) {
8204 + struct dentry *h_parent, *h_dentry;
8205 + struct inode *h_inode, *h_dir;
8206 + struct au_branch *br;
8208 + h_dentry = au_h_dptr(dentry, bindex);
8210 + if (d_is_positive(h_dentry))
8214 + h_parent = au_h_dptr(parent, bindex);
8215 + if (!h_parent || !d_is_dir(h_parent))
8219 + /* if the inum matches, then use the prepared name */
8220 + err = au_dr_lkup_name(&args, bindex);
8221 + if (unlikely(err))
8225 + h_dir = d_inode(h_parent);
8226 + vfsub_inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
8227 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &args);
8228 + inode_unlock_shared(h_dir);
8229 + err = PTR_ERR(h_dentry);
8230 + if (IS_ERR(h_dentry))
8233 + au_fclr_lkup(args.flags, ALLOW_NEG);
8235 + au_fset_lkup(args.flags, IGNORE_PERM);
8237 + if (au_dbwh(dentry) == bindex)
8241 + if (d_is_negative(h_dentry))
8243 + h_inode = d_inode(h_dentry);
8246 + args.type = h_inode->i_mode & S_IFMT;
8247 + if (args.type != S_IFDIR)
8250 + /* the type of lower may be different */
8251 + bdiropq = au_dbdiropq(dentry);
8252 + if (bdiropq >= 0 && bdiropq <= bindex)
8255 + br = au_sbr(sb, bindex);
8257 + && au_dr_hino_test_add(&br->br_dirren, h_inode->i_ino,
8258 + /*add_ent*/NULL)) {
8259 + /* prepare next name to lookup */
8260 + err = au_dr_lkup(&args, dentry, bindex);
8261 + if (unlikely(err))
8267 + AuLabel(positive);
8268 + au_update_dbtop(dentry);
8271 + if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
8272 + && au_dbtop(dentry) < 0)) {
8274 + AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
8280 + kfree(args.whname.name);
8282 + au_dr_lkup_fin(&args);
8287 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent)
8289 + struct dentry *dentry;
8292 + if (!au_test_h_perm_sio(d_inode(parent), MAY_EXEC))
8293 + dentry = vfsub_lkup_one(name, parent);
8295 + struct vfsub_lkup_one_args args = {
8301 + wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
8302 + if (unlikely(wkq_err))
8303 + dentry = ERR_PTR(wkq_err);
8310 + * lookup @dentry on @bindex which should be negative.
8312 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
8315 + struct dentry *parent, *h_parent, *h_dentry;
8316 + struct au_branch *br;
8318 + parent = dget_parent(dentry);
8319 + h_parent = au_h_dptr(parent, bindex);
8320 + br = au_sbr(dentry->d_sb, bindex);
8322 + h_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
8324 + h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
8325 + err = PTR_ERR(h_dentry);
8326 + if (IS_ERR(h_dentry))
8328 + if (unlikely(d_is_positive(h_dentry))) {
8330 + AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
8336 + if (bindex < au_dbtop(dentry))
8337 + au_set_dbtop(dentry, bindex);
8338 + if (au_dbbot(dentry) < bindex)
8339 + au_set_dbbot(dentry, bindex);
8340 + au_set_h_dptr(dentry, bindex, h_dentry);
8347 +/* ---------------------------------------------------------------------- */
8349 +/* subset of struct inode */
8351 + unsigned long i_ino;
8352 + /* unsigned int i_nlink; */
8358 + blkcnt_t i_blocks;
8363 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
8365 + ia->i_ino = h_inode->i_ino;
8366 + /* ia->i_nlink = h_inode->i_nlink; */
8367 + ia->i_uid = h_inode->i_uid;
8368 + ia->i_gid = h_inode->i_gid;
8369 + ia->i_version = h_inode->i_version;
8371 + ia->i_size = h_inode->i_size;
8372 + ia->i_blocks = h_inode->i_blocks;
8374 + ia->i_mode = (h_inode->i_mode & S_IFMT);
8377 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
8379 + return ia->i_ino != h_inode->i_ino
8380 + /* || ia->i_nlink != h_inode->i_nlink */
8381 + || !uid_eq(ia->i_uid, h_inode->i_uid)
8382 + || !gid_eq(ia->i_gid, h_inode->i_gid)
8383 + || ia->i_version != h_inode->i_version
8385 + || ia->i_size != h_inode->i_size
8386 + || ia->i_blocks != h_inode->i_blocks
8388 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
8391 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
8392 + struct au_branch *br)
8395 + struct au_iattr ia;
8396 + struct inode *h_inode;
8397 + struct dentry *h_d;
8398 + struct super_block *h_sb;
8401 + memset(&ia, -1, sizeof(ia));
8402 + h_sb = h_dentry->d_sb;
8404 + if (d_is_positive(h_dentry)) {
8405 + h_inode = d_inode(h_dentry);
8406 + au_iattr_save(&ia, h_inode);
8407 + } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
8408 + /* nfs d_revalidate may return 0 for negative dentry */
8409 + /* fuse d_revalidate always return 0 for negative dentry */
8412 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
8413 + h_d = vfsub_lkup_one(&h_dentry->d_name, h_parent);
8414 + err = PTR_ERR(h_d);
8419 + if (unlikely(h_d != h_dentry
8420 + || d_inode(h_d) != h_inode
8421 + || (h_inode && au_iattr_test(&ia, h_inode))))
8422 + err = au_busy_or_stale();
8430 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8431 + struct dentry *h_parent, struct au_branch *br)
8436 + if (udba == AuOpt_UDBA_REVAL
8437 + && !au_test_fs_remote(h_dentry->d_sb)) {
8439 + err = (d_inode(h_dentry->d_parent) != h_dir);
8440 + } else if (udba != AuOpt_UDBA_NONE)
8441 + err = au_h_verify_dentry(h_dentry, h_parent, br);
8446 +/* ---------------------------------------------------------------------- */
8448 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
8451 + aufs_bindex_t new_bindex, bindex, bbot, bwh, bdiropq;
8452 + struct au_hdentry tmp, *p, *q;
8453 + struct au_dinfo *dinfo;
8454 + struct super_block *sb;
8456 + DiMustWriteLock(dentry);
8458 + sb = dentry->d_sb;
8459 + dinfo = au_di(dentry);
8460 + bbot = dinfo->di_bbot;
8461 + bwh = dinfo->di_bwh;
8462 + bdiropq = dinfo->di_bdiropq;
8463 + bindex = dinfo->di_btop;
8464 + p = au_hdentry(dinfo, bindex);
8465 + for (; bindex <= bbot; bindex++, p++) {
8466 + if (!p->hd_dentry)
8469 + new_bindex = au_br_index(sb, p->hd_id);
8470 + if (new_bindex == bindex)
8473 + if (dinfo->di_bwh == bindex)
8475 + if (dinfo->di_bdiropq == bindex)
8476 + bdiropq = new_bindex;
8477 + if (new_bindex < 0) {
8479 + p->hd_dentry = NULL;
8483 + /* swap two lower dentries, and loop again */
8484 + q = au_hdentry(dinfo, new_bindex);
8488 + if (tmp.hd_dentry) {
8494 + dinfo->di_bwh = -1;
8495 + if (bwh >= 0 && bwh <= au_sbbot(sb) && au_sbr_whable(sb, bwh))
8496 + dinfo->di_bwh = bwh;
8498 + dinfo->di_bdiropq = -1;
8500 + && bdiropq <= au_sbbot(sb)
8501 + && au_sbr_whable(sb, bdiropq))
8502 + dinfo->di_bdiropq = bdiropq;
8505 + dinfo->di_btop = -1;
8506 + dinfo->di_bbot = -1;
8507 + bbot = au_dbbot(parent);
8509 + p = au_hdentry(dinfo, bindex);
8510 + for (; bindex <= bbot; bindex++, p++)
8511 + if (p->hd_dentry) {
8512 + dinfo->di_btop = bindex;
8516 + if (dinfo->di_btop >= 0) {
8518 + p = au_hdentry(dinfo, bindex);
8519 + for (; bindex >= 0; bindex--, p--)
8520 + if (p->hd_dentry) {
8521 + dinfo->di_bbot = bindex;
8530 +static void au_do_hide(struct dentry *dentry)
8532 + struct inode *inode;
8534 + if (d_really_is_positive(dentry)) {
8535 + inode = d_inode(dentry);
8536 + if (!d_is_dir(dentry)) {
8537 + if (inode->i_nlink && !d_unhashed(dentry))
8538 + drop_nlink(inode);
8540 + clear_nlink(inode);
8541 + /* stop next lookup */
8542 + inode->i_flags |= S_DEAD;
8544 + smp_mb(); /* necessary? */
8549 +static int au_hide_children(struct dentry *parent)
8551 + int err, i, j, ndentry;
8552 + struct au_dcsub_pages dpages;
8553 + struct au_dpage *dpage;
8554 + struct dentry *dentry;
8556 + err = au_dpages_init(&dpages, GFP_NOFS);
8557 + if (unlikely(err))
8559 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
8560 + if (unlikely(err))
8563 + /* in reverse order */
8564 + for (i = dpages.ndpage - 1; i >= 0; i--) {
8565 + dpage = dpages.dpages + i;
8566 + ndentry = dpage->ndentry;
8567 + for (j = ndentry - 1; j >= 0; j--) {
8568 + dentry = dpage->dentries[j];
8569 + if (dentry != parent)
8570 + au_do_hide(dentry);
8575 + au_dpages_free(&dpages);
8580 +static void au_hide(struct dentry *dentry)
8584 + AuDbgDentry(dentry);
8585 + if (d_is_dir(dentry)) {
8586 + /* shrink_dcache_parent(dentry); */
8587 + err = au_hide_children(dentry);
8588 + if (unlikely(err))
8589 + AuIOErr("%pd, failed hiding children, ignored %d\n",
8592 + au_do_hide(dentry);
8596 + * By adding a dirty branch, a cached dentry may be affected in various ways.
8598 + * a dirty branch is added
8599 + * - on the top of layers
8600 + * - in the middle of layers
8601 + * - to the bottom of layers
8603 + * on the added branch there exists
8606 + * - a same named entry
8608 + * * negative --> positive
8609 + * * positive --> positive
8610 + * - type is unchanged
8611 + * - type is changed
8613 + * * negative --> negative
8614 + * * positive --> negative (rejected by au_br_del() for non-dir case)
8617 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8618 + struct au_dinfo *tmp)
8621 + aufs_bindex_t bindex, bbot;
8623 + struct dentry *dentry;
8624 + struct inode *inode;
8626 + } orig_h, tmp_h = {
8629 + struct au_hdentry *hd;
8630 + struct inode *inode, *h_inode;
8631 + struct dentry *h_dentry;
8634 + AuDebugOn(dinfo->di_btop < 0);
8636 + orig_h.dentry = au_hdentry(dinfo, dinfo->di_btop)->hd_dentry;
8637 + orig_h.inode = NULL;
8638 + if (d_is_positive(orig_h.dentry)) {
8639 + orig_h.inode = d_inode(orig_h.dentry);
8640 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8642 + if (tmp->di_btop >= 0) {
8643 + tmp_h.dentry = au_hdentry(tmp, tmp->di_btop)->hd_dentry;
8644 + if (d_is_positive(tmp_h.dentry)) {
8645 + tmp_h.inode = d_inode(tmp_h.dentry);
8646 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8651 + if (d_really_is_positive(dentry))
8652 + inode = d_inode(dentry);
8653 + if (!orig_h.inode) {
8654 + AuDbg("nagative originally\n");
8660 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8661 + AuDebugOn(dinfo->di_bdiropq != -1);
8663 + if (!tmp_h.inode) {
8664 + AuDbg("negative --> negative\n");
8665 + /* should have only one negative lower */
8666 + if (tmp->di_btop >= 0
8667 + && tmp->di_btop < dinfo->di_btop) {
8668 + AuDebugOn(tmp->di_btop != tmp->di_bbot);
8669 + AuDebugOn(dinfo->di_btop != dinfo->di_bbot);
8670 + au_set_h_dptr(dentry, dinfo->di_btop, NULL);
8671 + au_di_cp(dinfo, tmp);
8672 + hd = au_hdentry(tmp, tmp->di_btop);
8673 + au_set_h_dptr(dentry, tmp->di_btop,
8674 + dget(hd->hd_dentry));
8676 + au_dbg_verify_dinode(dentry);
8678 + AuDbg("negative --> positive\n");
8680 + * similar to the behaviour of creating with bypassing
8682 + * unhash it in order to force an error in the
8683 + * succeeding create operation.
8684 + * we should not set S_DEAD here.
8687 + /* au_di_swap(tmp, dinfo); */
8688 + au_dbg_verify_dinode(dentry);
8691 + AuDbg("positive originally\n");
8692 + /* inode may be NULL */
8693 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8694 + if (!tmp_h.inode) {
8695 + AuDbg("positive --> negative\n");
8696 + /* or bypassing aufs */
8698 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_btop)
8699 + dinfo->di_bwh = tmp->di_bwh;
8701 + err = au_refresh_hinode_self(inode);
8702 + au_dbg_verify_dinode(dentry);
8703 + } else if (orig_h.mode == tmp_h.mode) {
8704 + AuDbg("positive --> positive, same type\n");
8705 + if (!S_ISDIR(orig_h.mode)
8706 + && dinfo->di_btop > tmp->di_btop) {
8708 + * similar to the behaviour of removing and
8713 + err = au_refresh_hinode_self(inode);
8714 + au_dbg_verify_dinode(dentry);
8716 + /* fill empty slots */
8717 + if (dinfo->di_btop > tmp->di_btop)
8718 + dinfo->di_btop = tmp->di_btop;
8719 + if (dinfo->di_bbot < tmp->di_bbot)
8720 + dinfo->di_bbot = tmp->di_bbot;
8721 + dinfo->di_bwh = tmp->di_bwh;
8722 + dinfo->di_bdiropq = tmp->di_bdiropq;
8723 + bbot = dinfo->di_bbot;
8724 + bindex = tmp->di_btop;
8725 + hd = au_hdentry(tmp, bindex);
8726 + for (; bindex <= bbot; bindex++, hd++) {
8727 + if (au_h_dptr(dentry, bindex))
8729 + h_dentry = hd->hd_dentry;
8732 + AuDebugOn(d_is_negative(h_dentry));
8733 + h_inode = d_inode(h_dentry);
8734 + AuDebugOn(orig_h.mode
8735 + != (h_inode->i_mode
8737 + au_set_h_dptr(dentry, bindex,
8741 + err = au_refresh_hinode(inode, dentry);
8742 + au_dbg_verify_dinode(dentry);
8745 + AuDbg("positive --> positive, different type\n");
8746 + /* similar to the behaviour of removing and creating */
8749 + err = au_refresh_hinode_self(inode);
8750 + au_dbg_verify_dinode(dentry);
8758 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8760 + const struct dentry_operations *dop
8761 + = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8762 + static const unsigned int mask
8763 + = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8765 + BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8767 + if (dentry->d_op == dop)
8770 + AuDbg("%pd\n", dentry);
8771 + spin_lock(&dentry->d_lock);
8772 + if (dop == &aufs_dop)
8773 + dentry->d_flags |= mask;
8775 + dentry->d_flags &= ~mask;
8776 + dentry->d_op = dop;
8777 + spin_unlock(&dentry->d_lock);
8780 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8782 + int err, ebrange, nbr;
8783 + unsigned int sigen;
8784 + struct au_dinfo *dinfo, *tmp;
8785 + struct super_block *sb;
8786 + struct inode *inode;
8788 + DiMustWriteLock(dentry);
8789 + AuDebugOn(IS_ROOT(dentry));
8790 + AuDebugOn(d_really_is_negative(parent));
8792 + sb = dentry->d_sb;
8793 + sigen = au_sigen(sb);
8794 + err = au_digen_test(parent, sigen);
8795 + if (unlikely(err))
8798 + nbr = au_sbbot(sb) + 1;
8799 + dinfo = au_di(dentry);
8800 + err = au_di_realloc(dinfo, nbr, /*may_shrink*/0);
8801 + if (unlikely(err))
8803 + ebrange = au_dbrange_test(dentry);
8805 + ebrange = au_do_refresh_hdentry(dentry, parent);
8807 + if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8808 + AuDebugOn(au_dbtop(dentry) < 0 && au_dbbot(dentry) >= 0);
8809 + if (d_really_is_positive(dentry)) {
8810 + inode = d_inode(dentry);
8811 + err = au_refresh_hinode_self(inode);
8813 + au_dbg_verify_dinode(dentry);
8815 + goto out_dgen; /* success */
8819 + /* temporary dinfo */
8820 + AuDbgDentry(dentry);
8822 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8823 + if (unlikely(!tmp))
8825 + au_di_swap(tmp, dinfo);
8826 + /* returns the number of positive dentries */
8828 + * if current working dir is removed, it returns an error.
8829 + * but the dentry is legal.
8831 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
8832 + AuDbgDentry(dentry);
8833 + au_di_swap(tmp, dinfo);
8834 + if (err == -ENOENT)
8837 + /* compare/refresh by dinfo */
8838 + AuDbgDentry(dentry);
8839 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8840 + au_dbg_verify_dinode(dentry);
8843 + au_di_realloc(dinfo, nbr, /*may_shrink*/1); /* harmless if err */
8844 + au_rw_write_unlock(&tmp->di_rwsem);
8846 + if (unlikely(err))
8850 + au_update_digen(dentry);
8852 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8853 + AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8854 + AuDbgDentry(dentry);
8860 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8861 + struct dentry *dentry, aufs_bindex_t bindex)
8866 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8869 + AuDbg("b%d\n", bindex);
8871 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8872 + * due to whiteout and branch permission.
8874 + flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8875 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
8876 + /* it may return tri-state */
8877 + valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8879 + if (unlikely(valid < 0))
8889 +/* todo: remove this */
8890 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8891 + unsigned int flags, int do_udba, int dirren)
8894 + umode_t mode, h_mode;
8895 + aufs_bindex_t bindex, btail, btop, ibs, ibe;
8896 + unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8897 + struct inode *h_inode, *h_cached_inode;
8898 + struct dentry *h_dentry;
8899 + struct qstr *name, *h_name;
8906 + unhashed = !!d_unhashed(dentry);
8907 + is_root = !!IS_ROOT(dentry);
8908 + name = &dentry->d_name;
8909 + tmpfile = au_di(dentry)->di_tmpfile;
8912 + * Theoretically, REVAL test should be unnecessary in case of
8914 + * But {fs,i}notify doesn't fire some necessary events,
8915 + * IN_ATTRIB for atime/nlink/pageio
8916 + * Let's do REVAL test too.
8918 + if (do_udba && inode) {
8919 + mode = (inode->i_mode & S_IFMT);
8920 + plus = (inode->i_nlink > 0);
8921 + ibs = au_ibtop(inode);
8922 + ibe = au_ibbot(inode);
8925 + btop = au_dbtop(dentry);
8927 + if (inode && S_ISDIR(inode->i_mode))
8928 + btail = au_dbtaildir(dentry);
8929 + for (bindex = btop; bindex <= btail; bindex++) {
8930 + h_dentry = au_h_dptr(dentry, bindex);
8934 + AuDbg("b%d, %pd\n", bindex, h_dentry);
8935 + h_nfs = !!au_test_nfs(h_dentry->d_sb);
8936 + spin_lock(&h_dentry->d_lock);
8937 + h_name = &h_dentry->d_name;
8938 + if (unlikely(do_udba
8941 + && (unhashed != !!d_unhashed(h_dentry)
8942 + || (!tmpfile && !dirren
8943 + && !au_qstreq(name, h_name))
8946 + && !(flags & LOOKUP_OPEN)
8947 + && (h_dentry->d_flags
8948 + & DCACHE_NFSFS_RENAMED)))
8952 + h_unhashed = d_unhashed(h_dentry);
8953 + spin_unlock(&h_dentry->d_lock);
8954 + AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8955 + unhashed, h_unhashed, dentry, h_dentry);
8958 + spin_unlock(&h_dentry->d_lock);
8960 + err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8961 + if (unlikely(err))
8962 + /* do not goto err, to keep the errno */
8965 + /* todo: plink too? */
8970 + if (unlikely(!!inode != d_is_positive(h_dentry)))
8974 + if (d_is_positive(h_dentry))
8975 + h_inode = d_inode(h_dentry);
8978 + h_cached_inode = h_inode;
8980 + h_mode = (h_inode->i_mode & S_IFMT);
8981 + h_plus = (h_inode->i_nlink > 0);
8983 + if (inode && ibs <= bindex && bindex <= ibe)
8984 + h_cached_inode = au_h_iptr(inode, bindex);
8987 + if (unlikely(plus != h_plus && !tmpfile))
8990 + if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8992 + && !IS_ROOT(h_dentry)
8993 + && unhashed != d_unhashed(h_dentry)))
8996 + if (unlikely(mode != h_mode
8997 + || h_cached_inode != h_inode))
9010 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
9011 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
9014 + struct dentry *parent;
9016 + if (!au_digen_test(dentry, sigen))
9019 + parent = dget_parent(dentry);
9020 + di_read_lock_parent(parent, AuLock_IR);
9021 + AuDebugOn(au_digen_test(parent, sigen));
9022 + au_dbg_verify_gen(parent, sigen);
9023 + err = au_refresh_dentry(dentry, parent);
9024 + di_read_unlock(parent, AuLock_IR);
9030 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
9033 + struct dentry *d, *parent;
9035 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
9036 + return simple_reval_dpath(dentry, sigen);
9038 + /* slow loop, keep it simple and stupid */
9039 + /* cf: au_cpup_dirs() */
9042 + while (au_digen_test(dentry, sigen)) {
9046 + parent = dget_parent(d);
9047 + if (!au_digen_test(parent, sigen))
9053 + di_write_lock_child2(d);
9055 + /* someone might update our dentry while we were sleeping */
9056 + if (au_digen_test(d, sigen)) {
9058 + * todo: consolidate with simple_reval_dpath(),
9059 + * do_refresh() and au_reval_for_attr().
9061 + di_read_lock_parent(parent, AuLock_IR);
9062 + err = au_refresh_dentry(d, parent);
9063 + di_read_unlock(parent, AuLock_IR);
9067 + di_write_unlock(d);
9069 + if (unlikely(err))
9077 + * if valid returns 1, otherwise 0.
9079 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
9082 + unsigned int sigen;
9083 + unsigned char do_udba, dirren;
9084 + struct super_block *sb;
9085 + struct inode *inode;
9087 + /* todo: support rcu-walk? */
9088 + if (flags & LOOKUP_RCU)
9092 + if (unlikely(!au_di(dentry)))
9096 + sb = dentry->d_sb;
9099 + * i_mutex of parent dir may be held,
9100 + * but we should not return 'invalid' due to busy.
9102 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
9103 + if (unlikely(err)) {
9109 + if (d_really_is_positive(dentry))
9110 + inode = d_inode(dentry);
9111 + if (unlikely(inode && au_is_bad_inode(inode))) {
9116 + if (unlikely(au_dbrange_test(dentry))) {
9122 + sigen = au_sigen(sb);
9123 + if (au_digen_test(dentry, sigen)) {
9124 + AuDebugOn(IS_ROOT(dentry));
9125 + err = au_reval_dpath(dentry, sigen);
9126 + if (unlikely(err)) {
9131 + di_downgrade_lock(dentry, AuLock_IR);
9134 + if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
9136 + && !(inode->i_state && I_LINKABLE)
9137 + && (IS_DEADDIR(inode) || !inode->i_nlink)) {
9142 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
9143 + if (do_udba && inode) {
9144 + aufs_bindex_t btop = au_ibtop(inode);
9145 + struct inode *h_inode;
9148 + h_inode = au_h_iptr(inode, btop);
9149 + if (h_inode && au_test_higen(inode, h_inode)) {
9156 + dirren = !!au_opt_test(au_mntflags(sb), DIRREN);
9157 + err = h_d_revalidate(dentry, inode, flags, do_udba, dirren);
9158 + if (unlikely(!err && do_udba && au_dbtop(dentry) < 0)) {
9160 + AuDbg("both of real entry and whiteout found, %p, err %d\n",
9166 + di_downgrade_lock(dentry, AuLock_IR);
9168 + aufs_read_unlock(dentry, AuLock_IR);
9173 + AuDbg("%pd invalid, %d\n", dentry, valid);
9179 +static void aufs_d_release(struct dentry *dentry)
9181 + if (au_di(dentry)) {
9182 + au_di_fin(dentry);
9183 + au_hn_di_reinit(dentry);
9187 +const struct dentry_operations aufs_dop = {
9188 + .d_revalidate = aufs_d_revalidate,
9189 + .d_weak_revalidate = aufs_d_revalidate,
9190 + .d_release = aufs_d_release
9193 +/* aufs_dop without d_revalidate */
9194 +const struct dentry_operations aufs_dop_noreval = {
9195 + .d_release = aufs_d_release
9197 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
9198 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
9199 +++ linux/fs/aufs/dentry.h 2018-01-29 07:56:20.053325069 +0100
9202 + * Copyright (C) 2005-2017 Junjiro R. Okajima
9204 + * This program, aufs is free software; you can redistribute it and/or modify
9205 + * it under the terms of the GNU General Public License as published by
9206 + * the Free Software Foundation; either version 2 of the License, or
9207 + * (at your option) any later version.
9209 + * This program is distributed in the hope that it will be useful,
9210 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9211 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9212 + * GNU General Public License for more details.
9214 + * You should have received a copy of the GNU General Public License
9215 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9219 + * lookup and dentry operations
9222 +#ifndef __AUFS_DENTRY_H__
9223 +#define __AUFS_DENTRY_H__
9227 +#include <linux/dcache.h>
9228 +#include "dirren.h"
9231 +struct au_hdentry {
9232 + struct dentry *hd_dentry;
9233 + aufs_bindex_t hd_id;
9237 + atomic_t di_generation;
9239 + struct au_rwsem di_rwsem;
9240 + aufs_bindex_t di_btop, di_bbot, di_bwh, di_bdiropq;
9241 + unsigned char di_tmpfile; /* to allow the different name */
9242 + struct au_hdentry *di_hdentry;
9243 +} ____cacheline_aligned_in_smp;
9245 +/* ---------------------------------------------------------------------- */
9247 +/* flags for au_lkup_dentry() */
9248 +#define AuLkup_ALLOW_NEG 1
9249 +#define AuLkup_IGNORE_PERM (1 << 1)
9250 +#define AuLkup_DIRREN (1 << 2)
9251 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
9252 +#define au_fset_lkup(flags, name) \
9253 + do { (flags) |= AuLkup_##name; } while (0)
9254 +#define au_fclr_lkup(flags, name) \
9255 + do { (flags) &= ~AuLkup_##name; } while (0)
9257 +#ifndef CONFIG_AUFS_DIRREN
9258 +#undef AuLkup_DIRREN
9259 +#define AuLkup_DIRREN 0
9262 +struct au_do_lookup_args {
9263 + unsigned int flags;
9265 + struct qstr whname, *name;
9266 + struct au_dr_lookup dirren;
9269 +/* ---------------------------------------------------------------------- */
9272 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
9274 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent);
9275 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
9276 + struct dentry *h_parent, struct au_branch *br);
9278 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t btop,
9279 + unsigned int flags);
9280 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
9281 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
9282 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
9283 +void au_refresh_dop(struct dentry *dentry, int force_reval);
9286 +void au_di_init_once(void *_di);
9287 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
9288 +void au_di_free(struct au_dinfo *dinfo);
9289 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
9290 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
9291 +int au_di_init(struct dentry *dentry);
9292 +void au_di_fin(struct dentry *dentry);
9293 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink);
9295 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
9296 +void di_read_unlock(struct dentry *d, int flags);
9297 +void di_downgrade_lock(struct dentry *d, int flags);
9298 +void di_write_lock(struct dentry *d, unsigned int lsc);
9299 +void di_write_unlock(struct dentry *d);
9300 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
9301 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
9302 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
9304 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
9305 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
9306 +aufs_bindex_t au_dbtail(struct dentry *dentry);
9307 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
9309 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9310 + struct dentry *h_dentry);
9311 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
9312 +int au_dbrange_test(struct dentry *dentry);
9313 +void au_update_digen(struct dentry *dentry);
9314 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
9315 +void au_update_dbtop(struct dentry *dentry);
9316 +void au_update_dbbot(struct dentry *dentry);
9317 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
9319 +/* ---------------------------------------------------------------------- */
9321 +static inline struct au_dinfo *au_di(struct dentry *dentry)
9323 + return dentry->d_fsdata;
9326 +/* ---------------------------------------------------------------------- */
9328 +/* lock subclass for dinfo */
9330 + AuLsc_DI_CHILD, /* child first */
9331 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
9332 + AuLsc_DI_CHILD3, /* copyup dirs */
9336 + AuLsc_DI_TMP /* temp for replacing dinfo */
9340 + * di_read_lock_child, di_write_lock_child,
9341 + * di_read_lock_child2, di_write_lock_child2,
9342 + * di_read_lock_child3, di_write_lock_child3,
9343 + * di_read_lock_parent, di_write_lock_parent,
9344 + * di_read_lock_parent2, di_write_lock_parent2,
9345 + * di_read_lock_parent3, di_write_lock_parent3,
9347 +#define AuReadLockFunc(name, lsc) \
9348 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
9349 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
9351 +#define AuWriteLockFunc(name, lsc) \
9352 +static inline void di_write_lock_##name(struct dentry *d) \
9353 +{ di_write_lock(d, AuLsc_DI_##lsc); }
9355 +#define AuRWLockFuncs(name, lsc) \
9356 + AuReadLockFunc(name, lsc) \
9357 + AuWriteLockFunc(name, lsc)
9359 +AuRWLockFuncs(child, CHILD);
9360 +AuRWLockFuncs(child2, CHILD2);
9361 +AuRWLockFuncs(child3, CHILD3);
9362 +AuRWLockFuncs(parent, PARENT);
9363 +AuRWLockFuncs(parent2, PARENT2);
9364 +AuRWLockFuncs(parent3, PARENT3);
9366 +#undef AuReadLockFunc
9367 +#undef AuWriteLockFunc
9368 +#undef AuRWLockFuncs
9370 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
9371 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
9372 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
9374 +/* ---------------------------------------------------------------------- */
9376 +/* todo: memory barrier? */
9377 +static inline unsigned int au_digen(struct dentry *d)
9379 + return atomic_read(&au_di(d)->di_generation);
9382 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
9384 + hdentry->hd_dentry = NULL;
9387 +static inline struct au_hdentry *au_hdentry(struct au_dinfo *di,
9388 + aufs_bindex_t bindex)
9390 + return di->di_hdentry + bindex;
9393 +static inline void au_hdput(struct au_hdentry *hd)
9396 + dput(hd->hd_dentry);
9399 +static inline aufs_bindex_t au_dbtop(struct dentry *dentry)
9401 + DiMustAnyLock(dentry);
9402 + return au_di(dentry)->di_btop;
9405 +static inline aufs_bindex_t au_dbbot(struct dentry *dentry)
9407 + DiMustAnyLock(dentry);
9408 + return au_di(dentry)->di_bbot;
9411 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
9413 + DiMustAnyLock(dentry);
9414 + return au_di(dentry)->di_bwh;
9417 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
9419 + DiMustAnyLock(dentry);
9420 + return au_di(dentry)->di_bdiropq;
9423 +/* todo: hard/soft set? */
9424 +static inline void au_set_dbtop(struct dentry *dentry, aufs_bindex_t bindex)
9426 + DiMustWriteLock(dentry);
9427 + au_di(dentry)->di_btop = bindex;
9430 +static inline void au_set_dbbot(struct dentry *dentry, aufs_bindex_t bindex)
9432 + DiMustWriteLock(dentry);
9433 + au_di(dentry)->di_bbot = bindex;
9436 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
9438 + DiMustWriteLock(dentry);
9439 + /* dbwh can be outside of btop - bbot range */
9440 + au_di(dentry)->di_bwh = bindex;
9443 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
9445 + DiMustWriteLock(dentry);
9446 + au_di(dentry)->di_bdiropq = bindex;
9449 +/* ---------------------------------------------------------------------- */
9451 +#ifdef CONFIG_AUFS_HNOTIFY
9452 +static inline void au_digen_dec(struct dentry *d)
9454 + atomic_dec(&au_di(d)->di_generation);
9457 +static inline void au_hn_di_reinit(struct dentry *dentry)
9459 + dentry->d_fsdata = NULL;
9462 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
9463 +#endif /* CONFIG_AUFS_HNOTIFY */
9465 +#endif /* __KERNEL__ */
9466 +#endif /* __AUFS_DENTRY_H__ */
9467 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
9468 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
9469 +++ linux/fs/aufs/dinfo.c 2018-01-29 07:56:20.053325069 +0100
9472 + * Copyright (C) 2005-2017 Junjiro R. Okajima
9474 + * This program, aufs is free software; you can redistribute it and/or modify
9475 + * it under the terms of the GNU General Public License as published by
9476 + * the Free Software Foundation; either version 2 of the License, or
9477 + * (at your option) any later version.
9479 + * This program is distributed in the hope that it will be useful,
9480 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9481 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9482 + * GNU General Public License for more details.
9484 + * You should have received a copy of the GNU General Public License
9485 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9489 + * dentry private data
9494 +void au_di_init_once(void *_dinfo)
9496 + struct au_dinfo *dinfo = _dinfo;
9498 + au_rw_init(&dinfo->di_rwsem);
9501 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
9503 + struct au_dinfo *dinfo;
9506 + dinfo = au_cache_alloc_dinfo();
9507 + if (unlikely(!dinfo))
9510 + nbr = au_sbbot(sb) + 1;
9513 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
9514 + if (dinfo->di_hdentry) {
9515 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
9516 + dinfo->di_btop = -1;
9517 + dinfo->di_bbot = -1;
9518 + dinfo->di_bwh = -1;
9519 + dinfo->di_bdiropq = -1;
9520 + dinfo->di_tmpfile = 0;
9521 + for (i = 0; i < nbr; i++)
9522 + dinfo->di_hdentry[i].hd_id = -1;
9526 + au_cache_free_dinfo(dinfo);
9533 +void au_di_free(struct au_dinfo *dinfo)
9535 + struct au_hdentry *p;
9536 + aufs_bindex_t bbot, bindex;
9538 + /* dentry may not be revalidated */
9539 + bindex = dinfo->di_btop;
9540 + if (bindex >= 0) {
9541 + bbot = dinfo->di_bbot;
9542 + p = au_hdentry(dinfo, bindex);
9543 + while (bindex++ <= bbot)
9546 + kfree(dinfo->di_hdentry);
9547 + au_cache_free_dinfo(dinfo);
9550 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
9552 + struct au_hdentry *p;
9555 + AuRwMustWriteLock(&a->di_rwsem);
9556 + AuRwMustWriteLock(&b->di_rwsem);
9558 +#define DiSwap(v, name) \
9560 + v = a->di_##name; \
9561 + a->di_##name = b->di_##name; \
9562 + b->di_##name = v; \
9565 + DiSwap(p, hdentry);
9569 + DiSwap(bi, bdiropq);
9575 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
9577 + AuRwMustWriteLock(&dst->di_rwsem);
9578 + AuRwMustWriteLock(&src->di_rwsem);
9580 + dst->di_btop = src->di_btop;
9581 + dst->di_bbot = src->di_bbot;
9582 + dst->di_bwh = src->di_bwh;
9583 + dst->di_bdiropq = src->di_bdiropq;
9587 +int au_di_init(struct dentry *dentry)
9590 + struct super_block *sb;
9591 + struct au_dinfo *dinfo;
9594 + sb = dentry->d_sb;
9595 + dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
9597 + atomic_set(&dinfo->di_generation, au_sigen(sb));
9598 + /* smp_mb(); */ /* atomic_set */
9599 + dentry->d_fsdata = dinfo;
9606 +void au_di_fin(struct dentry *dentry)
9608 + struct au_dinfo *dinfo;
9610 + dinfo = au_di(dentry);
9611 + AuRwDestroy(&dinfo->di_rwsem);
9612 + au_di_free(dinfo);
9615 +int au_di_realloc(struct au_dinfo *dinfo, int nbr, int may_shrink)
9618 + struct au_hdentry *hdp;
9620 + AuRwMustWriteLock(&dinfo->di_rwsem);
9623 + sz = sizeof(*hdp) * (dinfo->di_bbot + 1);
9625 + sz = sizeof(*hdp);
9626 + hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS,
9629 + dinfo->di_hdentry = hdp;
9636 +/* ---------------------------------------------------------------------- */
9638 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
9641 + case AuLsc_DI_CHILD:
9642 + ii_write_lock_child(inode);
9644 + case AuLsc_DI_CHILD2:
9645 + ii_write_lock_child2(inode);
9647 + case AuLsc_DI_CHILD3:
9648 + ii_write_lock_child3(inode);
9650 + case AuLsc_DI_PARENT:
9651 + ii_write_lock_parent(inode);
9653 + case AuLsc_DI_PARENT2:
9654 + ii_write_lock_parent2(inode);
9656 + case AuLsc_DI_PARENT3:
9657 + ii_write_lock_parent3(inode);
9664 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9667 + case AuLsc_DI_CHILD:
9668 + ii_read_lock_child(inode);
9670 + case AuLsc_DI_CHILD2:
9671 + ii_read_lock_child2(inode);
9673 + case AuLsc_DI_CHILD3:
9674 + ii_read_lock_child3(inode);
9676 + case AuLsc_DI_PARENT:
9677 + ii_read_lock_parent(inode);
9679 + case AuLsc_DI_PARENT2:
9680 + ii_read_lock_parent2(inode);
9682 + case AuLsc_DI_PARENT3:
9683 + ii_read_lock_parent3(inode);
9690 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9692 + struct inode *inode;
9694 + au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9695 + if (d_really_is_positive(d)) {
9696 + inode = d_inode(d);
9697 + if (au_ftest_lock(flags, IW))
9698 + do_ii_write_lock(inode, lsc);
9699 + else if (au_ftest_lock(flags, IR))
9700 + do_ii_read_lock(inode, lsc);
9704 +void di_read_unlock(struct dentry *d, int flags)
9706 + struct inode *inode;
9708 + if (d_really_is_positive(d)) {
9709 + inode = d_inode(d);
9710 + if (au_ftest_lock(flags, IW)) {
9711 + au_dbg_verify_dinode(d);
9712 + ii_write_unlock(inode);
9713 + } else if (au_ftest_lock(flags, IR)) {
9714 + au_dbg_verify_dinode(d);
9715 + ii_read_unlock(inode);
9718 + au_rw_read_unlock(&au_di(d)->di_rwsem);
9721 +void di_downgrade_lock(struct dentry *d, int flags)
9723 + if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9724 + ii_downgrade_lock(d_inode(d));
9725 + au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9728 +void di_write_lock(struct dentry *d, unsigned int lsc)
9730 + au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9731 + if (d_really_is_positive(d))
9732 + do_ii_write_lock(d_inode(d), lsc);
9735 +void di_write_unlock(struct dentry *d)
9737 + au_dbg_verify_dinode(d);
9738 + if (d_really_is_positive(d))
9739 + ii_write_unlock(d_inode(d));
9740 + au_rw_write_unlock(&au_di(d)->di_rwsem);
9743 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9745 + AuDebugOn(d1 == d2
9746 + || d_inode(d1) == d_inode(d2)
9747 + || d1->d_sb != d2->d_sb);
9749 + if ((isdir && au_test_subdir(d1, d2))
9751 + di_write_lock_child(d1);
9752 + di_write_lock_child2(d2);
9754 + di_write_lock_child(d2);
9755 + di_write_lock_child2(d1);
9759 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9761 + AuDebugOn(d1 == d2
9762 + || d_inode(d1) == d_inode(d2)
9763 + || d1->d_sb != d2->d_sb);
9765 + if ((isdir && au_test_subdir(d1, d2))
9767 + di_write_lock_parent(d1);
9768 + di_write_lock_parent2(d2);
9770 + di_write_lock_parent(d2);
9771 + di_write_lock_parent2(d1);
9775 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9777 + di_write_unlock(d1);
9778 + if (d_inode(d1) == d_inode(d2))
9779 + au_rw_write_unlock(&au_di(d2)->di_rwsem);
9781 + di_write_unlock(d2);
9784 +/* ---------------------------------------------------------------------- */
9786 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9790 + DiMustAnyLock(dentry);
9792 + if (au_dbtop(dentry) < 0 || bindex < au_dbtop(dentry))
9794 + AuDebugOn(bindex < 0);
9795 + d = au_hdentry(au_di(dentry), bindex)->hd_dentry;
9796 + AuDebugOn(d && au_dcount(d) <= 0);
9801 + * extended version of au_h_dptr().
9802 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9805 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9807 + struct dentry *h_dentry;
9808 + struct inode *inode, *h_inode;
9810 + AuDebugOn(d_really_is_negative(dentry));
9813 + if (au_dbtop(dentry) <= bindex
9814 + && bindex <= au_dbbot(dentry))
9815 + h_dentry = au_h_dptr(dentry, bindex);
9816 + if (h_dentry && !au_d_linkable(h_dentry)) {
9818 + goto out; /* success */
9821 + inode = d_inode(dentry);
9822 + AuDebugOn(bindex < au_ibtop(inode));
9823 + AuDebugOn(au_ibbot(inode) < bindex);
9824 + h_inode = au_h_iptr(inode, bindex);
9825 + h_dentry = d_find_alias(h_inode);
9827 + if (!IS_ERR(h_dentry)) {
9828 + if (!au_d_linkable(h_dentry))
9829 + goto out; /* success */
9835 + if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9836 + h_dentry = au_plink_lkup(inode, bindex);
9837 + AuDebugOn(!h_dentry);
9838 + if (!IS_ERR(h_dentry)) {
9839 + if (!au_d_hashed_positive(h_dentry))
9840 + goto out; /* success */
9847 + AuDbgDentry(h_dentry);
9851 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9853 + aufs_bindex_t bbot, bwh;
9855 + bbot = au_dbbot(dentry);
9857 + bwh = au_dbwh(dentry);
9860 + if (0 < bwh && bwh < bbot)
9866 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9868 + aufs_bindex_t bbot, bopq;
9870 + bbot = au_dbtail(dentry);
9872 + bopq = au_dbdiropq(dentry);
9873 + if (0 <= bopq && bopq < bbot)
9879 +/* ---------------------------------------------------------------------- */
9881 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9882 + struct dentry *h_dentry)
9884 + struct au_dinfo *dinfo;
9885 + struct au_hdentry *hd;
9886 + struct au_branch *br;
9888 + DiMustWriteLock(dentry);
9890 + dinfo = au_di(dentry);
9891 + hd = au_hdentry(dinfo, bindex);
9893 + hd->hd_dentry = h_dentry;
9895 + br = au_sbr(dentry->d_sb, bindex);
9896 + hd->hd_id = br->br_id;
9900 +int au_dbrange_test(struct dentry *dentry)
9903 + aufs_bindex_t btop, bbot;
9906 + btop = au_dbtop(dentry);
9907 + bbot = au_dbbot(dentry);
9909 + AuDebugOn(bbot < 0 && btop > bbot);
9912 + AuDebugOn(bbot >= 0);
9918 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9923 + if (unlikely(au_digen(dentry) != sigen
9924 + || au_iigen_test(d_inode(dentry), sigen)))
9930 +void au_update_digen(struct dentry *dentry)
9932 + atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9933 + /* smp_mb(); */ /* atomic_set */
9936 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9938 + struct au_dinfo *dinfo;
9939 + struct dentry *h_d;
9940 + struct au_hdentry *hdp;
9941 + aufs_bindex_t bindex, bbot;
9943 + DiMustWriteLock(dentry);
9945 + dinfo = au_di(dentry);
9946 + if (!dinfo || dinfo->di_btop < 0)
9949 + if (do_put_zero) {
9950 + bbot = dinfo->di_bbot;
9951 + bindex = dinfo->di_btop;
9952 + hdp = au_hdentry(dinfo, bindex);
9953 + for (; bindex <= bbot; bindex++, hdp++) {
9954 + h_d = hdp->hd_dentry;
9955 + if (h_d && d_is_negative(h_d))
9956 + au_set_h_dptr(dentry, bindex, NULL);
9960 + dinfo->di_btop = 0;
9961 + hdp = au_hdentry(dinfo, dinfo->di_btop);
9962 + for (; dinfo->di_btop <= dinfo->di_bbot; dinfo->di_btop++, hdp++)
9963 + if (hdp->hd_dentry)
9965 + if (dinfo->di_btop > dinfo->di_bbot) {
9966 + dinfo->di_btop = -1;
9967 + dinfo->di_bbot = -1;
9971 + hdp = au_hdentry(dinfo, dinfo->di_bbot);
9972 + for (; dinfo->di_bbot >= 0; dinfo->di_bbot--, hdp--)
9973 + if (hdp->hd_dentry)
9975 + AuDebugOn(dinfo->di_btop > dinfo->di_bbot || dinfo->di_bbot < 0);
9978 +void au_update_dbtop(struct dentry *dentry)
9980 + aufs_bindex_t bindex, bbot;
9981 + struct dentry *h_dentry;
9983 + bbot = au_dbbot(dentry);
9984 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
9985 + h_dentry = au_h_dptr(dentry, bindex);
9988 + if (d_is_positive(h_dentry)) {
9989 + au_set_dbtop(dentry, bindex);
9992 + au_set_h_dptr(dentry, bindex, NULL);
9996 +void au_update_dbbot(struct dentry *dentry)
9998 + aufs_bindex_t bindex, btop;
9999 + struct dentry *h_dentry;
10001 + btop = au_dbtop(dentry);
10002 + for (bindex = au_dbbot(dentry); bindex >= btop; bindex--) {
10003 + h_dentry = au_h_dptr(dentry, bindex);
10006 + if (d_is_positive(h_dentry)) {
10007 + au_set_dbbot(dentry, bindex);
10010 + au_set_h_dptr(dentry, bindex, NULL);
10014 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
10016 + aufs_bindex_t bindex, bbot;
10018 + bbot = au_dbbot(dentry);
10019 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++)
10020 + if (au_h_dptr(dentry, bindex) == h_dentry)
10024 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
10025 --- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
10026 +++ linux/fs/aufs/dir.c 2018-01-29 07:56:20.053325069 +0100
10029 + * Copyright (C) 2005-2017 Junjiro R. Okajima
10031 + * This program, aufs is free software; you can redistribute it and/or modify
10032 + * it under the terms of the GNU General Public License as published by
10033 + * the Free Software Foundation; either version 2 of the License, or
10034 + * (at your option) any later version.
10036 + * This program is distributed in the hope that it will be useful,
10037 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10038 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10039 + * GNU General Public License for more details.
10041 + * You should have received a copy of the GNU General Public License
10042 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10046 + * directory operations
10049 +#include <linux/fs_stack.h>
10052 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
10054 + unsigned int nlink;
10056 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10058 + nlink = dir->i_nlink;
10059 + nlink += h_dir->i_nlink - 2;
10060 + if (h_dir->i_nlink < 2)
10062 + smp_mb(); /* for i_nlink */
10063 + /* 0 can happen in revaliding */
10064 + set_nlink(dir, nlink);
10067 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
10069 + unsigned int nlink;
10071 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
10073 + nlink = dir->i_nlink;
10074 + nlink -= h_dir->i_nlink - 2;
10075 + if (h_dir->i_nlink < 2)
10077 + smp_mb(); /* for i_nlink */
10078 + /* nlink == 0 means the branch-fs is broken */
10079 + set_nlink(dir, nlink);
10082 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
10085 + aufs_bindex_t bindex, bbot;
10086 + struct file *h_file;
10087 + struct dentry *h_dentry;
10091 + AuDebugOn(!d_is_dir(file->f_path.dentry));
10093 + bbot = au_fbbot_dir(file);
10094 + for (bindex = au_fbtop(file);
10095 + bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10097 + h_file = au_hf_dir(file, bindex);
10098 + if (h_file && file_inode(h_file))
10099 + sz += vfsub_f_size_read(h_file);
10102 + AuDebugOn(!dentry);
10103 + AuDebugOn(!d_is_dir(dentry));
10105 + bbot = au_dbtaildir(dentry);
10106 + for (bindex = au_dbtop(dentry);
10107 + bindex <= bbot && sz < KMALLOC_MAX_SIZE;
10109 + h_dentry = au_h_dptr(dentry, bindex);
10110 + if (h_dentry && d_is_positive(h_dentry))
10111 + sz += i_size_read(d_inode(h_dentry));
10114 + if (sz < KMALLOC_MAX_SIZE)
10115 + sz = roundup_pow_of_two(sz);
10116 + if (sz > KMALLOC_MAX_SIZE)
10117 + sz = KMALLOC_MAX_SIZE;
10118 + else if (sz < NAME_MAX) {
10119 + BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
10120 + sz = AUFS_RDBLK_DEF;
10125 +struct au_dir_ts_arg {
10126 + struct dentry *dentry;
10127 + aufs_bindex_t brid;
10130 +static void au_do_dir_ts(void *arg)
10132 + struct au_dir_ts_arg *a = arg;
10133 + struct au_dtime dt;
10134 + struct path h_path;
10135 + struct inode *dir, *h_dir;
10136 + struct super_block *sb;
10137 + struct au_branch *br;
10138 + struct au_hinode *hdir;
10140 + aufs_bindex_t btop, bindex;
10142 + sb = a->dentry->d_sb;
10143 + if (d_really_is_negative(a->dentry))
10145 + /* no dir->i_mutex lock */
10146 + aufs_read_lock(a->dentry, AuLock_DW); /* noflush */
10148 + dir = d_inode(a->dentry);
10149 + btop = au_ibtop(dir);
10150 + bindex = au_br_index(sb, a->brid);
10151 + if (bindex < btop)
10154 + br = au_sbr(sb, bindex);
10155 + h_path.dentry = au_h_dptr(a->dentry, bindex);
10156 + if (!h_path.dentry)
10158 + h_path.mnt = au_br_mnt(br);
10159 + au_dtime_store(&dt, a->dentry, &h_path);
10161 + br = au_sbr(sb, btop);
10162 + if (!au_br_writable(br->br_perm))
10164 + h_path.dentry = au_h_dptr(a->dentry, btop);
10165 + h_path.mnt = au_br_mnt(br);
10166 + err = vfsub_mnt_want_write(h_path.mnt);
10169 + hdir = au_hi(dir, btop);
10170 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
10171 + h_dir = au_h_iptr(dir, btop);
10172 + if (h_dir->i_nlink
10173 + && timespec_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
10174 + dt.dt_h_path = h_path;
10175 + au_dtime_revert(&dt);
10177 + au_hn_inode_unlock(hdir);
10178 + vfsub_mnt_drop_write(h_path.mnt);
10179 + au_cpup_attr_timesizes(dir);
10182 + aufs_read_unlock(a->dentry, AuLock_DW);
10185 + au_nwt_done(&au_sbi(sb)->si_nowait);
10189 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
10191 + int perm, wkq_err;
10192 + aufs_bindex_t btop;
10193 + struct au_dir_ts_arg *arg;
10194 + struct dentry *dentry;
10195 + struct super_block *sb;
10199 + dentry = d_find_any_alias(dir);
10200 + AuDebugOn(!dentry);
10201 + sb = dentry->d_sb;
10202 + btop = au_ibtop(dir);
10203 + if (btop == bindex) {
10204 + au_cpup_attr_timesizes(dir);
10208 + perm = au_sbr_perm(sb, btop);
10209 + if (!au_br_writable(perm))
10212 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
10216 + arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
10217 + arg->brid = au_sbr_id(sb, bindex);
10218 + wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
10219 + if (unlikely(wkq_err)) {
10220 + pr_err("wkq %d\n", wkq_err);
10229 +/* ---------------------------------------------------------------------- */
10231 +static int reopen_dir(struct file *file)
10234 + unsigned int flags;
10235 + aufs_bindex_t bindex, btail, btop;
10236 + struct dentry *dentry, *h_dentry;
10237 + struct file *h_file;
10239 + /* open all lower dirs */
10240 + dentry = file->f_path.dentry;
10241 + btop = au_dbtop(dentry);
10242 + for (bindex = au_fbtop(file); bindex < btop; bindex++)
10243 + au_set_h_fptr(file, bindex, NULL);
10244 + au_set_fbtop(file, btop);
10246 + btail = au_dbtaildir(dentry);
10247 + for (bindex = au_fbbot_dir(file); btail < bindex; bindex--)
10248 + au_set_h_fptr(file, bindex, NULL);
10249 + au_set_fbbot_dir(file, btail);
10251 + flags = vfsub_file_flags(file);
10252 + for (bindex = btop; bindex <= btail; bindex++) {
10253 + h_dentry = au_h_dptr(dentry, bindex);
10256 + h_file = au_hf_dir(file, bindex);
10260 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10261 + err = PTR_ERR(h_file);
10262 + if (IS_ERR(h_file))
10263 + goto out; /* close all? */
10264 + au_set_h_fptr(file, bindex, h_file);
10266 + au_update_figen(file);
10267 + /* todo: necessary? */
10268 + /* file->f_ra = h_file->f_ra; */
10275 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
10278 + aufs_bindex_t bindex, btail;
10279 + struct dentry *dentry, *h_dentry;
10280 + struct vfsmount *mnt;
10282 + FiMustWriteLock(file);
10283 + AuDebugOn(h_file);
10286 + mnt = file->f_path.mnt;
10287 + dentry = file->f_path.dentry;
10288 + file->f_version = d_inode(dentry)->i_version;
10289 + bindex = au_dbtop(dentry);
10290 + au_set_fbtop(file, bindex);
10291 + btail = au_dbtaildir(dentry);
10292 + au_set_fbbot_dir(file, btail);
10293 + for (; !err && bindex <= btail; bindex++) {
10294 + h_dentry = au_h_dptr(dentry, bindex);
10298 + err = vfsub_test_mntns(mnt, h_dentry->d_sb);
10299 + if (unlikely(err))
10301 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
10302 + if (IS_ERR(h_file)) {
10303 + err = PTR_ERR(h_file);
10306 + au_set_h_fptr(file, bindex, h_file);
10308 + au_update_figen(file);
10309 + /* todo: necessary? */
10310 + /* file->f_ra = h_file->f_ra; */
10312 + return 0; /* success */
10315 + for (bindex = au_fbtop(file); bindex <= btail; bindex++)
10316 + au_set_h_fptr(file, bindex, NULL);
10317 + au_set_fbtop(file, -1);
10318 + au_set_fbbot_dir(file, -1);
10323 +static int aufs_open_dir(struct inode *inode __maybe_unused,
10324 + struct file *file)
10327 + struct super_block *sb;
10328 + struct au_fidir *fidir;
10331 + sb = file->f_path.dentry->d_sb;
10332 + si_read_lock(sb, AuLock_FLUSH);
10333 + fidir = au_fidir_alloc(sb);
10335 + struct au_do_open_args args = {
10336 + .open = do_open_dir,
10339 + err = au_do_open(file, &args);
10340 + if (unlikely(err))
10343 + si_read_unlock(sb);
10347 +static int aufs_release_dir(struct inode *inode __maybe_unused,
10348 + struct file *file)
10350 + struct au_vdir *vdir_cache;
10351 + struct au_finfo *finfo;
10352 + struct au_fidir *fidir;
10353 + struct au_hfile *hf;
10354 + aufs_bindex_t bindex, bbot;
10356 + finfo = au_fi(file);
10357 + fidir = finfo->fi_hdir;
10359 + au_hbl_del(&finfo->fi_hlist,
10360 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
10361 + vdir_cache = fidir->fd_vdir_cache; /* lock-free */
10363 + au_vdir_free(vdir_cache);
10365 + bindex = finfo->fi_btop;
10366 + if (bindex >= 0) {
10367 + hf = fidir->fd_hfile + bindex;
10369 + * calls fput() instead of filp_close(),
10370 + * since no dnotify or lock for the lower file.
10372 + bbot = fidir->fd_bbot;
10373 + for (; bindex <= bbot; bindex++, hf++)
10375 + au_hfput(hf, /*execed*/0);
10378 + finfo->fi_hdir = NULL;
10380 + au_finfo_fin(file);
10384 +/* ---------------------------------------------------------------------- */
10386 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
10389 + aufs_bindex_t bindex, bbot;
10390 + struct file *h_file;
10393 + bbot = au_fbbot_dir(file);
10394 + for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10395 + h_file = au_hf_dir(file, bindex);
10397 + err = vfsub_flush(h_file, id);
10402 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
10404 + return au_do_flush(file, id, au_do_flush_dir);
10407 +/* ---------------------------------------------------------------------- */
10409 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
10412 + aufs_bindex_t bbot, bindex;
10413 + struct inode *inode;
10414 + struct super_block *sb;
10417 + sb = dentry->d_sb;
10418 + inode = d_inode(dentry);
10419 + IMustLock(inode);
10420 + bbot = au_dbbot(dentry);
10421 + for (bindex = au_dbtop(dentry); !err && bindex <= bbot; bindex++) {
10422 + struct path h_path;
10424 + if (au_test_ro(sb, bindex, inode))
10426 + h_path.dentry = au_h_dptr(dentry, bindex);
10427 + if (!h_path.dentry)
10430 + h_path.mnt = au_sbr_mnt(sb, bindex);
10431 + err = vfsub_fsync(NULL, &h_path, datasync);
10437 +static int au_do_fsync_dir(struct file *file, int datasync)
10440 + aufs_bindex_t bbot, bindex;
10441 + struct file *h_file;
10442 + struct super_block *sb;
10443 + struct inode *inode;
10445 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10446 + if (unlikely(err))
10449 + inode = file_inode(file);
10450 + sb = inode->i_sb;
10451 + bbot = au_fbbot_dir(file);
10452 + for (bindex = au_fbtop(file); !err && bindex <= bbot; bindex++) {
10453 + h_file = au_hf_dir(file, bindex);
10454 + if (!h_file || au_test_ro(sb, bindex, inode))
10457 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
10465 + * @file may be NULL
10467 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
10471 + struct dentry *dentry;
10472 + struct inode *inode;
10473 + struct super_block *sb;
10476 + dentry = file->f_path.dentry;
10477 + inode = d_inode(dentry);
10478 + inode_lock(inode);
10479 + sb = dentry->d_sb;
10480 + si_noflush_read_lock(sb);
10482 + err = au_do_fsync_dir(file, datasync);
10484 + di_write_lock_child(dentry);
10485 + err = au_do_fsync_dir_no_file(dentry, datasync);
10487 + au_cpup_attr_timesizes(inode);
10488 + di_write_unlock(dentry);
10490 + fi_write_unlock(file);
10492 + si_read_unlock(sb);
10493 + inode_unlock(inode);
10497 +/* ---------------------------------------------------------------------- */
10499 +static int aufs_iterate_shared(struct file *file, struct dir_context *ctx)
10502 + struct dentry *dentry;
10503 + struct inode *inode, *h_inode;
10504 + struct super_block *sb;
10506 + AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
10508 + dentry = file->f_path.dentry;
10509 + inode = d_inode(dentry);
10510 + IMustLock(inode);
10512 + sb = dentry->d_sb;
10513 + si_read_lock(sb, AuLock_FLUSH);
10514 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1, /*fi_lsc*/0);
10515 + if (unlikely(err))
10517 + err = au_alive_dir(dentry);
10519 + err = au_vdir_init(file);
10520 + di_downgrade_lock(dentry, AuLock_IR);
10521 + if (unlikely(err))
10524 + h_inode = au_h_iptr(inode, au_ibtop(inode));
10525 + if (!au_test_nfsd()) {
10526 + err = au_vdir_fill_de(file, ctx);
10527 + fsstack_copy_attr_atime(inode, h_inode);
10530 + * nfsd filldir may call lookup_one_len(), vfs_getattr(),
10531 + * encode_fh() and others.
10533 + atomic_inc(&h_inode->i_count);
10534 + di_read_unlock(dentry, AuLock_IR);
10535 + si_read_unlock(sb);
10536 + err = au_vdir_fill_de(file, ctx);
10537 + fsstack_copy_attr_atime(inode, h_inode);
10538 + fi_write_unlock(file);
10546 + di_read_unlock(dentry, AuLock_IR);
10547 + fi_write_unlock(file);
10549 + si_read_unlock(sb);
10553 +/* ---------------------------------------------------------------------- */
10555 +#define AuTestEmpty_WHONLY 1
10556 +#define AuTestEmpty_CALLED (1 << 1)
10557 +#define AuTestEmpty_SHWH (1 << 2)
10558 +#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
10559 +#define au_fset_testempty(flags, name) \
10560 + do { (flags) |= AuTestEmpty_##name; } while (0)
10561 +#define au_fclr_testempty(flags, name) \
10562 + do { (flags) &= ~AuTestEmpty_##name; } while (0)
10564 +#ifndef CONFIG_AUFS_SHWH
10565 +#undef AuTestEmpty_SHWH
10566 +#define AuTestEmpty_SHWH 0
10569 +struct test_empty_arg {
10570 + struct dir_context ctx;
10571 + struct au_nhash *whlist;
10572 + unsigned int flags;
10574 + aufs_bindex_t bindex;
10577 +static int test_empty_cb(struct dir_context *ctx, const char *__name,
10578 + int namelen, loff_t offset __maybe_unused, u64 ino,
10579 + unsigned int d_type)
10581 + struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
10583 + char *name = (void *)__name;
10586 + au_fset_testempty(arg->flags, CALLED);
10588 + if (name[0] == '.'
10589 + && (namelen == 1 || (name[1] == '.' && namelen == 2)))
10590 + goto out; /* success */
10592 + if (namelen <= AUFS_WH_PFX_LEN
10593 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
10594 + if (au_ftest_testempty(arg->flags, WHONLY)
10595 + && !au_nhash_test_known_wh(arg->whlist, name, namelen))
10596 + arg->err = -ENOTEMPTY;
10600 + name += AUFS_WH_PFX_LEN;
10601 + namelen -= AUFS_WH_PFX_LEN;
10602 + if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
10603 + arg->err = au_nhash_append_wh
10604 + (arg->whlist, name, namelen, ino, d_type, arg->bindex,
10605 + au_ftest_testempty(arg->flags, SHWH));
10609 + AuTraceErr(arg->err);
10613 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10616 + struct file *h_file;
10618 + h_file = au_h_open(dentry, arg->bindex,
10619 + O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
10620 + /*file*/NULL, /*force_wr*/0);
10621 + err = PTR_ERR(h_file);
10622 + if (IS_ERR(h_file))
10626 + if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
10627 + && !file_inode(h_file)->i_nlink)
10632 + au_fclr_testempty(arg->flags, CALLED);
10634 + err = vfsub_iterate_dir(h_file, &arg->ctx);
10637 + } while (!err && au_ftest_testempty(arg->flags, CALLED));
10641 + au_sbr_put(dentry->d_sb, arg->bindex);
10646 +struct do_test_empty_args {
10648 + struct dentry *dentry;
10649 + struct test_empty_arg *arg;
10652 +static void call_do_test_empty(void *args)
10654 + struct do_test_empty_args *a = args;
10655 + *a->errp = do_test_empty(a->dentry, a->arg);
10658 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
10660 + int err, wkq_err;
10661 + struct dentry *h_dentry;
10662 + struct inode *h_inode;
10664 + h_dentry = au_h_dptr(dentry, arg->bindex);
10665 + h_inode = d_inode(h_dentry);
10666 + /* todo: i_mode changes anytime? */
10667 + vfsub_inode_lock_shared_nested(h_inode, AuLsc_I_CHILD);
10668 + err = au_test_h_perm_sio(h_inode, MAY_EXEC | MAY_READ);
10669 + inode_unlock_shared(h_inode);
10671 + err = do_test_empty(dentry, arg);
10673 + struct do_test_empty_args args = {
10675 + .dentry = dentry,
10678 + unsigned int flags = arg->flags;
10680 + wkq_err = au_wkq_wait(call_do_test_empty, &args);
10681 + if (unlikely(wkq_err))
10683 + arg->flags = flags;
10689 +int au_test_empty_lower(struct dentry *dentry)
10692 + unsigned int rdhash;
10693 + aufs_bindex_t bindex, btop, btail;
10694 + struct au_nhash whlist;
10695 + struct test_empty_arg arg = {
10697 + .actor = test_empty_cb
10700 + int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10702 + SiMustAnyLock(dentry->d_sb);
10704 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10706 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10707 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10708 + if (unlikely(err))
10712 + arg.whlist = &whlist;
10713 + btop = au_dbtop(dentry);
10714 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10715 + au_fset_testempty(arg.flags, SHWH);
10716 + test_empty = do_test_empty;
10717 + if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10718 + test_empty = sio_test_empty;
10719 + arg.bindex = btop;
10720 + err = test_empty(dentry, &arg);
10721 + if (unlikely(err))
10724 + au_fset_testempty(arg.flags, WHONLY);
10725 + btail = au_dbtaildir(dentry);
10726 + for (bindex = btop + 1; !err && bindex <= btail; bindex++) {
10727 + struct dentry *h_dentry;
10729 + h_dentry = au_h_dptr(dentry, bindex);
10730 + if (h_dentry && d_is_positive(h_dentry)) {
10731 + arg.bindex = bindex;
10732 + err = test_empty(dentry, &arg);
10737 + au_nhash_wh_free(&whlist);
10742 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10745 + struct test_empty_arg arg = {
10747 + .actor = test_empty_cb
10750 + aufs_bindex_t bindex, btail;
10753 + arg.whlist = whlist;
10754 + arg.flags = AuTestEmpty_WHONLY;
10755 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10756 + au_fset_testempty(arg.flags, SHWH);
10757 + btail = au_dbtaildir(dentry);
10758 + for (bindex = au_dbtop(dentry); !err && bindex <= btail; bindex++) {
10759 + struct dentry *h_dentry;
10761 + h_dentry = au_h_dptr(dentry, bindex);
10762 + if (h_dentry && d_is_positive(h_dentry)) {
10763 + arg.bindex = bindex;
10764 + err = sio_test_empty(dentry, &arg);
10771 +/* ---------------------------------------------------------------------- */
10773 +const struct file_operations aufs_dir_fop = {
10774 + .owner = THIS_MODULE,
10775 + .llseek = default_llseek,
10776 + .read = generic_read_dir,
10777 + .iterate_shared = aufs_iterate_shared,
10778 + .unlocked_ioctl = aufs_ioctl_dir,
10779 +#ifdef CONFIG_COMPAT
10780 + .compat_ioctl = aufs_compat_ioctl_dir,
10782 + .open = aufs_open_dir,
10783 + .release = aufs_release_dir,
10784 + .flush = aufs_flush_dir,
10785 + .fsync = aufs_fsync_dir
10787 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10788 --- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
10789 +++ linux/fs/aufs/dir.h 2018-01-29 07:56:20.053325069 +0100
10792 + * Copyright (C) 2005-2017 Junjiro R. Okajima
10794 + * This program, aufs is free software; you can redistribute it and/or modify
10795 + * it under the terms of the GNU General Public License as published by
10796 + * the Free Software Foundation; either version 2 of the License, or
10797 + * (at your option) any later version.
10799 + * This program is distributed in the hope that it will be useful,
10800 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10801 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10802 + * GNU General Public License for more details.
10804 + * You should have received a copy of the GNU General Public License
10805 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10809 + * directory operations
10812 +#ifndef __AUFS_DIR_H__
10813 +#define __AUFS_DIR_H__
10817 +#include <linux/fs.h>
10819 +/* ---------------------------------------------------------------------- */
10821 +/* need to be faster and smaller */
10824 + unsigned int nh_num;
10825 + struct hlist_head *nh_head;
10828 +struct au_vdir_destr {
10829 + unsigned char len;
10830 + unsigned char name[0];
10833 +struct au_vdir_dehstr {
10834 + struct hlist_node hash;
10835 + struct au_vdir_destr *str;
10836 +} ____cacheline_aligned_in_smp;
10838 +struct au_vdir_de {
10840 + unsigned char de_type;
10841 + /* caution: packed */
10842 + struct au_vdir_destr de_str;
10845 +struct au_vdir_wh {
10846 + struct hlist_node wh_hash;
10847 +#ifdef CONFIG_AUFS_SHWH
10849 + aufs_bindex_t wh_bindex;
10850 + unsigned char wh_type;
10852 + aufs_bindex_t wh_bindex;
10854 + /* caution: packed */
10855 + struct au_vdir_destr wh_str;
10858 +union au_vdir_deblk_p {
10859 + unsigned char *deblk;
10860 + struct au_vdir_de *de;
10864 + unsigned char **vd_deblk;
10865 + unsigned long vd_nblk;
10867 + unsigned long ul;
10868 + union au_vdir_deblk_p p;
10871 + unsigned long vd_version;
10872 + unsigned int vd_deblk_sz;
10873 + unsigned long vd_jiffy;
10874 +} ____cacheline_aligned_in_smp;
10876 +/* ---------------------------------------------------------------------- */
10879 +extern const struct file_operations aufs_dir_fop;
10880 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10881 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10882 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10883 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10884 +int au_test_empty_lower(struct dentry *dentry);
10885 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10888 +unsigned int au_rdhash_est(loff_t sz);
10889 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10890 +void au_nhash_wh_free(struct au_nhash *whlist);
10891 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10893 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10894 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10895 + unsigned int d_type, aufs_bindex_t bindex,
10896 + unsigned char shwh);
10897 +void au_vdir_free(struct au_vdir *vdir);
10898 +int au_vdir_init(struct file *file);
10899 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10902 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10904 +#ifdef CONFIG_AUFS_RDU
10906 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10907 +#ifdef CONFIG_COMPAT
10908 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10909 + unsigned long arg);
10912 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10913 + unsigned int cmd, unsigned long arg)
10914 +#ifdef CONFIG_COMPAT
10915 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10916 + unsigned int cmd, unsigned long arg)
10920 +#endif /* __KERNEL__ */
10921 +#endif /* __AUFS_DIR_H__ */
10922 diff -urN /usr/share/empty/fs/aufs/dirren.c linux/fs/aufs/dirren.c
10923 --- /usr/share/empty/fs/aufs/dirren.c 1970-01-01 01:00:00.000000000 +0100
10924 +++ linux/fs/aufs/dirren.c 2018-01-29 07:56:20.053325069 +0100
10927 + * Copyright (C) 2017 Junjiro R. Okajima
10929 + * This program, aufs is free software; you can redistribute it and/or modify
10930 + * it under the terms of the GNU General Public License as published by
10931 + * the Free Software Foundation; either version 2 of the License, or
10932 + * (at your option) any later version.
10934 + * This program is distributed in the hope that it will be useful,
10935 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10936 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10937 + * GNU General Public License for more details.
10939 + * You should have received a copy of the GNU General Public License
10940 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10944 + * special handling in renaming a directoy
10945 + * in order to support looking-up the before-renamed name on the lower readonly
10949 +#include <linux/byteorder/generic.h>
10952 +static void au_dr_hino_del(struct au_dr_br *dr, struct au_dr_hino *ent)
10956 + idx = au_dr_ihash(ent->dr_h_ino);
10957 + au_hbl_del(&ent->dr_hnode, dr->dr_h_ino + idx);
10960 +static int au_dr_hino_test_empty(struct au_dr_br *dr)
10963 + struct hlist_bl_head *hbl;
10966 + for (i = 0; ret && i < AuDirren_NHASH; i++) {
10967 + hbl = dr->dr_h_ino + i;
10968 + hlist_bl_lock(hbl);
10969 + ret &= hlist_bl_empty(hbl);
10970 + hlist_bl_unlock(hbl);
10976 +static struct au_dr_hino *au_dr_hino_find(struct au_dr_br *dr, ino_t ino)
10978 + struct au_dr_hino *found, *ent;
10979 + struct hlist_bl_head *hbl;
10980 + struct hlist_bl_node *pos;
10984 + idx = au_dr_ihash(ino);
10985 + hbl = dr->dr_h_ino + idx;
10986 + hlist_bl_lock(hbl);
10987 + hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
10988 + if (ent->dr_h_ino == ino) {
10992 + hlist_bl_unlock(hbl);
10997 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t ino,
10998 + struct au_dr_hino *add_ent)
11001 + struct hlist_bl_head *hbl;
11002 + struct hlist_bl_node *pos;
11003 + struct au_dr_hino *ent;
11006 + idx = au_dr_ihash(ino);
11007 + hbl = dr->dr_h_ino + idx;
11010 + struct hlist_bl_node *tmp;
11012 + hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
11013 + AuDbg("hi%llu\n", (unsigned long long)ent->dr_h_ino);
11016 + hlist_bl_lock(hbl);
11017 + hlist_bl_for_each_entry(ent, pos, hbl, dr_hnode)
11018 + if (ent->dr_h_ino == ino) {
11022 + if (!found && add_ent)
11023 + hlist_bl_add_head(&add_ent->dr_hnode, hbl);
11024 + hlist_bl_unlock(hbl);
11026 + if (!found && add_ent)
11027 + AuDbg("i%llu added\n", (unsigned long long)add_ent->dr_h_ino);
11032 +void au_dr_hino_free(struct au_dr_br *dr)
11035 + struct hlist_bl_head *hbl;
11036 + struct hlist_bl_node *pos, *tmp;
11037 + struct au_dr_hino *ent;
11039 + /* SiMustWriteLock(sb); */
11041 + for (i = 0; i < AuDirren_NHASH; i++) {
11042 + hbl = dr->dr_h_ino + i;
11043 + /* no spinlock since sbinfo must be write-locked */
11044 + hlist_bl_for_each_entry_safe(ent, pos, tmp, hbl, dr_hnode)
11046 + INIT_HLIST_BL_HEAD(hbl);
11050 +/* returns the number of inodes or an error */
11051 +static int au_dr_hino_store(struct super_block *sb, struct au_branch *br,
11052 + struct file *hinofile)
11056 + loff_t pos, oldsize;
11058 + struct inode *hinoinode;
11059 + struct hlist_bl_head *hbl;
11060 + struct hlist_bl_node *n1, *n2;
11061 + struct au_dr_hino *ent;
11063 + SiMustWriteLock(sb);
11064 + AuDebugOn(!au_br_writable(br->br_perm));
11066 + hinoinode = file_inode(hinofile);
11067 + oldsize = i_size_read(hinoinode);
11071 + hbl = br->br_dirren.dr_h_ino;
11072 + for (i = 0; !err && i < AuDirren_NHASH; i++, hbl++) {
11073 + /* no bit-lock since sbinfo must be write-locked */
11074 + hlist_bl_for_each_entry_safe(ent, n1, n2, hbl, dr_hnode) {
11075 + AuDbg("hi%llu, %pD2\n",
11076 + (unsigned long long)ent->dr_h_ino, hinofile);
11077 + u64 = cpu_to_be64(ent->dr_h_ino);
11078 + ssz = vfsub_write_k(hinofile, &u64, sizeof(u64), &pos);
11079 + if (ssz == sizeof(u64))
11082 + /* write error */
11083 + pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11090 + /* regardless the error */
11091 + if (pos < oldsize) {
11092 + err = vfsub_trunc(&hinofile->f_path, pos, /*attr*/0, hinofile);
11100 +static int au_dr_hino_load(struct au_dr_br *dr, struct file *hinofile)
11107 + struct au_dr_hino *ent;
11108 + struct inode *hinoinode;
11109 + struct hlist_bl_head *hbl;
11113 + hbl = dr->dr_h_ino;
11114 + hinoinode = file_inode(hinofile);
11115 + sz = i_size_read(hinoinode);
11116 + AuDebugOn(sz % sizeof(u64));
11117 + n = sz / sizeof(u64);
11119 + ssz = vfsub_read_k(hinofile, &u64, sizeof(u64), &pos);
11120 + if (unlikely(ssz != sizeof(u64))) {
11121 + pr_err("ssz %zd, %pD2\n", ssz, hinofile);
11128 + ent = kmalloc(sizeof(*ent), GFP_NOFS);
11134 + ent->dr_h_ino = be64_to_cpu((__force __be64)u64);
11135 + AuDbg("hi%llu, %pD2\n",
11136 + (unsigned long long)ent->dr_h_ino, hinofile);
11137 + hidx = au_dr_ihash(ent->dr_h_ino);
11138 + au_hbl_add(&ent->dr_hnode, hbl + hidx);
11140 + goto out; /* success */
11143 + au_dr_hino_free(dr);
11150 + * @bindex/@br is a switch to distinguish whether suspending hnotify or not.
11151 + * @path is a switch to distinguish load and store.
11153 +static int au_dr_hino(struct super_block *sb, aufs_bindex_t bindex,
11154 + struct au_branch *br, const struct path *path)
11157 + unsigned char load, suspend;
11158 + struct file *hinofile;
11159 + struct au_hinode *hdir;
11160 + struct inode *dir, *delegated;
11161 + struct path hinopath;
11162 + struct qstr hinoname = QSTR_INIT(AUFS_WH_DR_BRHINO,
11163 + sizeof(AUFS_WH_DR_BRHINO) - 1);
11165 + AuDebugOn(bindex < 0 && !br);
11166 + AuDebugOn(bindex >= 0 && br);
11171 + br = au_sbr(sb, bindex);
11174 + path = &br->br_path;
11175 + AuDebugOn(!au_br_writable(br->br_perm));
11176 + if (unlikely(!au_br_writable(br->br_perm)))
11182 + dir = d_inode(sb->s_root);
11183 + hdir = au_hinode(au_ii(dir), bindex);
11184 + dir = hdir->hi_inode;
11185 + au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11187 + dir = d_inode(path->dentry);
11188 + inode_lock_nested(dir, AuLsc_I_CHILD);
11190 + hinopath.dentry = vfsub_lkup_one(&hinoname, path->dentry);
11191 + err = PTR_ERR(hinopath.dentry);
11192 + if (IS_ERR(hinopath.dentry))
11196 + flags = O_RDONLY;
11198 + if (d_is_negative(hinopath.dentry))
11199 + goto out_dput; /* success */
11201 + if (au_dr_hino_test_empty(&br->br_dirren)) {
11202 + if (d_is_positive(hinopath.dentry)) {
11203 + delegated = NULL;
11204 + err = vfsub_unlink(dir, &hinopath, &delegated,
11207 + if (unlikely(err))
11208 + pr_err("ignored err %d, %pd2\n",
11209 + err, hinopath.dentry);
11210 + if (unlikely(err == -EWOULDBLOCK))
11215 + } else if (!d_is_positive(hinopath.dentry)) {
11216 + err = vfsub_create(dir, &hinopath, 0600,
11217 + /*want_excl*/false);
11219 + if (unlikely(err))
11222 + flags = O_WRONLY;
11224 + hinopath.mnt = path->mnt;
11225 + hinofile = vfsub_dentry_open(&hinopath, flags);
11227 + au_hn_inode_unlock(hdir);
11229 + inode_unlock(dir);
11230 + dput(hinopath.dentry);
11231 + AuTraceErrPtr(hinofile);
11232 + if (IS_ERR(hinofile)) {
11233 + err = PTR_ERR(hinofile);
11238 + err = au_dr_hino_load(&br->br_dirren, hinofile);
11240 + err = au_dr_hino_store(sb, br, hinofile);
11245 + dput(hinopath.dentry);
11248 + au_hn_inode_unlock(hdir);
11250 + inode_unlock(dir);
11256 +/* ---------------------------------------------------------------------- */
11258 +static int au_dr_brid_init(struct au_dr_brid *brid, const struct path *path)
11261 + struct kstatfs kstfs;
11263 + struct dentry *dentry;
11264 + struct super_block *sb;
11266 + err = vfs_statfs((void *)path, &kstfs);
11268 + if (unlikely(err))
11271 + /* todo: support for UUID */
11273 + if (kstfs.f_fsid.val[0] || kstfs.f_fsid.val[1]) {
11274 + brid->type = AuBrid_FSID;
11275 + brid->fsid = kstfs.f_fsid;
11277 + dentry = path->dentry;
11278 + sb = dentry->d_sb;
11281 + brid->type = AuBrid_DEV;
11290 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
11291 + const struct path *path)
11294 + struct au_dr_br *dr;
11295 + struct hlist_bl_head *hbl;
11297 + dr = &br->br_dirren;
11298 + hbl = dr->dr_h_ino;
11299 + for (i = 0; i < AuDirren_NHASH; i++, hbl++)
11300 + INIT_HLIST_BL_HEAD(hbl);
11302 + err = au_dr_brid_init(&dr->dr_brid, path);
11303 + if (unlikely(err))
11306 + if (au_opt_test(au_mntflags(sb), DIRREN))
11307 + err = au_dr_hino(sb, /*bindex*/-1, br, path);
11314 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br)
11319 + if (au_br_writable(br->br_perm))
11320 + err = au_dr_hino(sb, /*bindex*/-1, br, /*path*/NULL);
11322 + au_dr_hino_free(&br->br_dirren);
11327 +/* ---------------------------------------------------------------------- */
11329 +static int au_brid_str(struct au_dr_brid *brid, struct inode *h_inode,
11330 + char *buf, size_t sz)
11333 + unsigned int major, minor;
11337 + err = snprintf(p, sz, "%d_", brid->type);
11338 + AuDebugOn(err > sz);
11341 + switch (brid->type) {
11342 + case AuBrid_Unset:
11344 + case AuBrid_UUID:
11345 + err = snprintf(p, sz, "%pU", brid->uuid.b);
11347 + case AuBrid_FSID:
11348 + err = snprintf(p, sz, "%08x-%08x",
11349 + brid->fsid.val[0], brid->fsid.val[1]);
11352 + major = MAJOR(brid->dev);
11353 + minor = MINOR(brid->dev);
11354 + if (major <= 0xff && minor <= 0xff)
11355 + err = snprintf(p, sz, "%02x%02x", major, minor);
11357 + err = snprintf(p, sz, "%03x:%05x", major, minor);
11360 + AuDebugOn(err > sz);
11363 + err = snprintf(p, sz, "_%llu", (unsigned long long)h_inode->i_ino);
11364 + AuDebugOn(err > sz);
11371 +static int au_drinfo_name(struct au_branch *br, char *name, int len)
11374 + struct dentry *br_dentry;
11375 + struct inode *br_inode;
11377 + br_dentry = au_br_dentry(br);
11378 + br_inode = d_inode(br_dentry);
11379 + rlen = au_brid_str(&br->br_dirren.dr_brid, br_inode, name, len);
11380 + AuDebugOn(rlen >= AUFS_DIRREN_ENV_VAL_SZ);
11381 + AuDebugOn(rlen > len);
11386 +/* ---------------------------------------------------------------------- */
11389 + * from the given @h_dentry, construct drinfo at @*fdata.
11390 + * when the size of @*fdata is not enough, reallocate and return new @fdata and
11393 +static int au_drinfo_construct(struct au_drinfo_fdata **fdata,
11394 + struct dentry *h_dentry,
11395 + unsigned char *allocated)
11398 + struct au_drinfo_fdata *f, *p;
11399 + struct au_drinfo *drinfo;
11400 + struct inode *h_inode;
11401 + struct qstr *qname;
11405 + h_inode = d_inode(h_dentry);
11406 + qname = &h_dentry->d_name;
11407 + drinfo = &f->drinfo;
11408 + drinfo->ino = (__force uint64_t)cpu_to_be64(h_inode->i_ino);
11409 + drinfo->oldnamelen = qname->len;
11410 + if (*allocated < sizeof(*f) + qname->len) {
11411 + v = roundup_pow_of_two(*allocated + qname->len);
11412 + p = au_krealloc(f, v, GFP_NOFS, /*may_shrink*/0);
11413 + if (unlikely(!p)) {
11421 + drinfo = &f->drinfo;
11423 + memcpy(drinfo->oldname, qname->name, qname->len);
11424 + AuDbg("i%llu, %.*s\n",
11425 + be64_to_cpu((__force __be64)drinfo->ino), drinfo->oldnamelen,
11426 + drinfo->oldname);
11433 +/* callers have to free the return value */
11434 +static struct au_drinfo *au_drinfo_read_k(struct file *file, ino_t h_ino)
11436 + struct au_drinfo *ret, *drinfo;
11437 + struct au_drinfo_fdata fdata;
11442 + ret = ERR_PTR(-EIO);
11444 + ssz = vfsub_read_k(file, &fdata, sizeof(fdata), &pos);
11445 + if (unlikely(ssz != sizeof(fdata))) {
11446 + AuIOErr("ssz %zd, %u, %pD2\n",
11447 + ssz, (unsigned int)sizeof(fdata), file);
11451 + fdata.magic = ntohl((__force __be32)fdata.magic);
11452 + switch (fdata.magic) {
11453 + case AUFS_DRINFO_MAGIC_V1:
11456 + AuIOErr("magic-num 0x%x, 0x%x, %pD2\n",
11457 + fdata.magic, AUFS_DRINFO_MAGIC_V1, file);
11461 + drinfo = &fdata.drinfo;
11462 + len = drinfo->oldnamelen;
11464 + AuIOErr("broken drinfo %pD2\n", file);
11469 + drinfo->ino = be64_to_cpu((__force __be64)drinfo->ino);
11470 + if (unlikely(h_ino && drinfo->ino != h_ino)) {
11471 + AuDbg("ignored i%llu, i%llu, %pD2\n",
11472 + (unsigned long long)drinfo->ino,
11473 + (unsigned long long)h_ino, file);
11474 + goto out; /* success */
11477 + ret = kmalloc(sizeof(*ret) + len, GFP_NOFS);
11478 + if (unlikely(!ret)) {
11479 + ret = ERR_PTR(-ENOMEM);
11480 + AuTraceErrPtr(ret);
11485 + ssz = vfsub_read_k(file, (void *)ret->oldname, len, &pos);
11486 + if (unlikely(ssz != len)) {
11488 + ret = ERR_PTR(-EIO);
11489 + AuIOErr("ssz %zd, %u, %pD2\n", ssz, len, file);
11493 + AuDbg("oldname %.*s\n", ret->oldnamelen, ret->oldname);
11499 +/* ---------------------------------------------------------------------- */
11501 +/* in order to be revertible */
11502 +struct au_drinfo_rev_elm {
11504 + struct dentry *info_dentry;
11505 + struct au_drinfo *info_last;
11508 +struct au_drinfo_rev {
11509 + unsigned char already;
11510 + aufs_bindex_t nelm;
11511 + struct au_drinfo_rev_elm elm[0];
11514 +/* todo: isn't it too large? */
11515 +struct au_drinfo_store {
11516 + struct path h_ppath;
11517 + struct dentry *h_dentry;
11518 + struct au_drinfo_fdata *fdata;
11519 + char *infoname; /* inside of whname, just after PFX */
11520 + char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ];
11521 + aufs_bindex_t btgt, btail;
11522 + unsigned char no_sio,
11523 + allocated, /* current size of *fdata */
11524 + infonamelen, /* room size for p */
11525 + whnamelen, /* length of the genarated name */
11526 + renameback; /* renamed back */
11529 +/* on rename(2) error, the caller should revert it using @elm */
11530 +static int au_drinfo_do_store(struct au_drinfo_store *w,
11531 + struct au_drinfo_rev_elm *elm)
11536 + struct path infopath = {
11537 + .mnt = w->h_ppath.mnt
11539 + struct inode *h_dir, *h_inode, *delegated;
11540 + struct file *infofile;
11541 + struct qstr *qname;
11544 + && memcmp(elm, page_address(ZERO_PAGE(0)), sizeof(*elm)));
11546 + infopath.dentry = vfsub_lookup_one_len(w->whname, w->h_ppath.dentry,
11548 + AuTraceErrPtr(infopath.dentry);
11549 + if (IS_ERR(infopath.dentry)) {
11550 + err = PTR_ERR(infopath.dentry);
11555 + h_dir = d_inode(w->h_ppath.dentry);
11556 + if (elm && d_is_negative(infopath.dentry)) {
11557 + err = vfsub_create(h_dir, &infopath, 0600, /*want_excl*/true);
11559 + if (unlikely(err))
11561 + elm->created = 1;
11562 + elm->info_dentry = dget(infopath.dentry);
11565 + infofile = vfsub_dentry_open(&infopath, O_RDWR);
11566 + AuTraceErrPtr(infofile);
11567 + if (IS_ERR(infofile)) {
11568 + err = PTR_ERR(infofile);
11572 + h_inode = d_inode(infopath.dentry);
11573 + if (elm && i_size_read(h_inode)) {
11574 + h_inode = d_inode(w->h_dentry);
11575 + elm->info_last = au_drinfo_read_k(infofile, h_inode->i_ino);
11576 + AuTraceErrPtr(elm->info_last);
11577 + if (IS_ERR(elm->info_last)) {
11578 + err = PTR_ERR(elm->info_last);
11579 + elm->info_last = NULL;
11580 + AuDebugOn(elm->info_dentry);
11585 + if (elm && w->renameback) {
11586 + delegated = NULL;
11587 + err = vfsub_unlink(h_dir, &infopath, &delegated, /*force*/0);
11589 + if (unlikely(err == -EWOULDBLOCK))
11595 + qname = &w->h_dentry->d_name;
11596 + len = sizeof(*w->fdata) + qname->len;
11598 + len = sizeof(*w->fdata) + w->fdata->drinfo.oldnamelen;
11599 + ssz = vfsub_write_k(infofile, w->fdata, len, &pos);
11600 + if (ssz == len) {
11601 + AuDbg("hi%llu, %.*s\n", w->fdata->drinfo.ino,
11602 + w->fdata->drinfo.oldnamelen, w->fdata->drinfo.oldname);
11603 + goto out_fput; /* success */
11608 + /* the caller should revert it using @elm */
11614 + dput(infopath.dentry);
11620 +struct au_call_drinfo_do_store_args {
11622 + struct au_drinfo_store *w;
11623 + struct au_drinfo_rev_elm *elm;
11626 +static void au_call_drinfo_do_store(void *args)
11628 + struct au_call_drinfo_do_store_args *a = args;
11630 + *a->errp = au_drinfo_do_store(a->w, a->elm);
11633 +static int au_drinfo_store_sio(struct au_drinfo_store *w,
11634 + struct au_drinfo_rev_elm *elm)
11636 + int err, wkq_err;
11639 + err = au_drinfo_do_store(w, elm);
11641 + struct au_call_drinfo_do_store_args a = {
11646 + wkq_err = au_wkq_wait(au_call_drinfo_do_store, &a);
11647 + if (unlikely(wkq_err))
11655 +static int au_drinfo_store_work_init(struct au_drinfo_store *w,
11656 + aufs_bindex_t btgt)
11660 + memset(w, 0, sizeof(*w));
11661 + w->allocated = roundup_pow_of_two(sizeof(*w->fdata) + 40);
11662 + strcpy(w->whname, AUFS_WH_DR_INFO_PFX);
11663 + w->infoname = w->whname + sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11664 + w->infonamelen = sizeof(w->whname) - sizeof(AUFS_WH_DR_INFO_PFX);
11666 + w->no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
11669 + w->fdata = kcalloc(1, w->allocated, GFP_NOFS);
11670 + if (unlikely(!w->fdata)) {
11674 + w->fdata->magic = (__force uint32_t)htonl(AUFS_DRINFO_MAGIC_V1);
11681 +static void au_drinfo_store_work_fin(struct au_drinfo_store *w)
11686 +static void au_drinfo_store_rev(struct au_drinfo_rev *rev,
11687 + struct au_drinfo_store *w)
11689 + struct au_drinfo_rev_elm *elm;
11690 + struct inode *h_dir, *delegated;
11692 + struct path infopath = {
11693 + .mnt = w->h_ppath.mnt
11696 + h_dir = d_inode(w->h_ppath.dentry);
11697 + IMustLock(h_dir);
11701 + for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11702 + AuDebugOn(elm->created && elm->info_last);
11703 + if (elm->created) {
11705 + delegated = NULL;
11706 + infopath.dentry = elm->info_dentry;
11707 + err = vfsub_unlink(h_dir, &infopath, &delegated,
11710 + if (unlikely(err == -EWOULDBLOCK))
11712 + dput(elm->info_dentry);
11713 + } else if (elm->info_last) {
11715 + w->fdata->drinfo = *elm->info_last;
11716 + memcpy(w->fdata->drinfo.oldname,
11717 + elm->info_last->oldname,
11718 + elm->info_last->oldnamelen);
11719 + err = au_drinfo_store_sio(w, /*elm*/NULL);
11720 + kfree(elm->info_last);
11722 + if (unlikely(err))
11723 + AuIOErr("%d, %s\n", err, w->whname);
11724 + /* go on even if err */
11728 +/* caller has to call au_dr_rename_fin() later */
11729 +static int au_drinfo_store(struct dentry *dentry, aufs_bindex_t btgt,
11730 + struct qstr *dst_name, void *_rev)
11732 + int err, sz, nelm;
11733 + aufs_bindex_t bindex, btail;
11734 + struct au_drinfo_store work;
11735 + struct au_drinfo_rev *rev, **p;
11736 + struct au_drinfo_rev_elm *elm;
11737 + struct super_block *sb;
11738 + struct au_branch *br;
11739 + struct au_hinode *hdir;
11741 + err = au_drinfo_store_work_init(&work, btgt);
11743 + if (unlikely(err))
11747 + btail = au_dbtaildir(dentry);
11748 + nelm = btail - btgt;
11749 + sz = sizeof(*rev) + sizeof(*elm) * nelm;
11750 + rev = kcalloc(1, sz, GFP_NOFS);
11751 + if (unlikely(!rev)) {
11755 + rev->nelm = nelm;
11761 + sb = dentry->d_sb;
11762 + work.h_ppath.dentry = au_h_dptr(dentry, btgt);
11763 + work.h_ppath.mnt = au_sbr_mnt(sb, btgt);
11764 + hdir = au_hi(d_inode(dentry), btgt);
11765 + au_hn_inode_lock_nested(hdir, AuLsc_I_CHILD);
11766 + for (bindex = btgt + 1; bindex <= btail; bindex++, elm++) {
11767 + work.h_dentry = au_h_dptr(dentry, bindex);
11768 + if (!work.h_dentry)
11771 + err = au_drinfo_construct(&work.fdata, work.h_dentry,
11772 + &work.allocated);
11774 + if (unlikely(err))
11777 + work.renameback = au_qstreq(&work.h_dentry->d_name, dst_name);
11778 + br = au_sbr(sb, bindex);
11779 + work.whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
11780 + work.whnamelen += au_drinfo_name(br, work.infoname,
11781 + work.infonamelen);
11782 + AuDbg("whname %.*s, i%llu, %.*s\n",
11783 + work.whnamelen, work.whname,
11784 + be64_to_cpu((__force __be64)work.fdata->drinfo.ino),
11785 + work.fdata->drinfo.oldnamelen,
11786 + work.fdata->drinfo.oldname);
11788 + err = au_drinfo_store_sio(&work, elm);
11790 + if (unlikely(err))
11793 + if (unlikely(err)) {
11794 + /* revert all drinfo */
11795 + au_drinfo_store_rev(rev, &work);
11799 + au_hn_inode_unlock(hdir);
11802 + au_drinfo_store_work_fin(&work);
11807 +/* ---------------------------------------------------------------------- */
11809 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
11810 + struct qstr *dst_name, void *_rev)
11812 + int err, already;
11814 + struct super_block *sb;
11815 + struct au_branch *br;
11816 + struct au_dr_br *dr;
11817 + struct dentry *h_dentry;
11818 + struct inode *h_inode;
11819 + struct au_dr_hino *ent;
11820 + struct au_drinfo_rev *rev, **p;
11822 + AuDbg("bindex %d\n", bindex);
11825 + ent = kmalloc(sizeof(*ent), GFP_NOFS);
11826 + if (unlikely(!ent))
11830 + br = au_sbr(sb, bindex);
11831 + dr = &br->br_dirren;
11832 + h_dentry = au_h_dptr(src, bindex);
11833 + h_inode = d_inode(h_dentry);
11834 + ino = h_inode->i_ino;
11835 + ent->dr_h_ino = ino;
11836 + already = au_dr_hino_test_add(dr, ino, ent);
11837 + AuDbg("b%d, hi%llu, already %d\n",
11838 + bindex, (unsigned long long)ino, already);
11840 + err = au_drinfo_store(src, bindex, dst_name, _rev);
11845 + rev->already = already;
11846 + goto out; /* success */
11851 + au_dr_hino_del(dr, ent);
11859 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11861 + struct au_drinfo_rev *rev;
11862 + struct au_drinfo_rev_elm *elm;
11867 + for (nelm = rev->nelm; nelm > 0; nelm--, elm++) {
11868 + dput(elm->info_dentry);
11869 + kfree(elm->info_last);
11874 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t btgt, void *_rev)
11877 + struct au_drinfo_store work;
11878 + struct au_drinfo_rev *rev = _rev;
11879 + struct super_block *sb;
11880 + struct au_branch *br;
11881 + struct inode *h_inode;
11882 + struct au_dr_br *dr;
11883 + struct au_dr_hino *ent;
11885 + err = au_drinfo_store_work_init(&work, btgt);
11886 + if (unlikely(err))
11890 + br = au_sbr(sb, btgt);
11891 + work.h_ppath.dentry = au_h_dptr(src, btgt);
11892 + work.h_ppath.mnt = au_br_mnt(br);
11893 + au_drinfo_store_rev(rev, &work);
11894 + au_drinfo_store_work_fin(&work);
11895 + if (rev->already)
11898 + dr = &br->br_dirren;
11899 + h_inode = d_inode(work.h_ppath.dentry);
11900 + ent = au_dr_hino_find(dr, h_inode->i_ino);
11902 + au_dr_hino_del(dr, ent);
11907 + if (unlikely(err))
11908 + pr_err("failed to remove dirren info\n");
11911 +/* ---------------------------------------------------------------------- */
11913 +static struct au_drinfo *au_drinfo_do_load(struct path *h_ppath,
11914 + char *whname, int whnamelen,
11915 + struct dentry **info_dentry)
11917 + struct au_drinfo *drinfo;
11919 + struct inode *h_dir;
11920 + struct path infopath;
11923 + AuDbg("%pd/%.*s\n", h_ppath->dentry, whnamelen, whname);
11925 + *info_dentry = NULL;
11928 + h_dir = d_inode(h_ppath->dentry);
11929 + vfsub_inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
11930 + infopath.dentry = vfsub_lookup_one_len(whname, h_ppath->dentry,
11932 + if (IS_ERR(infopath.dentry)) {
11933 + drinfo = (void *)infopath.dentry;
11937 + if (d_is_negative(infopath.dentry))
11938 + goto out_dput; /* success */
11940 + infopath.mnt = h_ppath->mnt;
11941 + f = vfsub_dentry_open(&infopath, O_RDONLY);
11942 + inode_unlock_shared(h_dir);
11945 + drinfo = (void *)f;
11949 + drinfo = au_drinfo_read_k(f, /*h_ino*/0);
11950 + if (IS_ERR_OR_NULL(drinfo))
11953 + AuDbg("oldname %.*s\n", drinfo->oldnamelen, drinfo->oldname);
11954 + *info_dentry = dget(infopath.dentry); /* keep it alive */
11959 + dput(infopath.dentry);
11962 + inode_unlock_shared(h_dir);
11963 + AuTraceErrPtr(drinfo);
11967 +struct au_drinfo_do_load_args {
11968 + struct au_drinfo **drinfop;
11969 + struct path *h_ppath;
11972 + struct dentry **info_dentry;
11975 +static void au_call_drinfo_do_load(void *args)
11977 + struct au_drinfo_do_load_args *a = args;
11979 + *a->drinfop = au_drinfo_do_load(a->h_ppath, a->whname, a->whnamelen,
11983 +struct au_drinfo_load {
11984 + struct path h_ppath;
11985 + struct qstr *qname;
11986 + unsigned char no_sio;
11988 + aufs_bindex_t ninfo;
11989 + struct au_drinfo **drinfo;
11992 +static int au_drinfo_load(struct au_drinfo_load *w, aufs_bindex_t bindex,
11993 + struct au_branch *br)
11995 + int err, wkq_err, whnamelen, e;
11996 + char whname[sizeof(AUFS_WH_DR_INFO_PFX) + AUFS_DIRREN_ENV_VAL_SZ]
11997 + = AUFS_WH_DR_INFO_PFX;
11998 + struct au_drinfo *drinfo;
11999 + struct qstr oldname;
12000 + struct inode *h_dir, *delegated;
12001 + struct dentry *info_dentry;
12002 + struct path infopath;
12004 + whnamelen = sizeof(AUFS_WH_DR_INFO_PFX) - 1;
12005 + whnamelen += au_drinfo_name(br, whname + whnamelen,
12006 + sizeof(whname) - whnamelen);
12008 + drinfo = au_drinfo_do_load(&w->h_ppath, whname, whnamelen,
12011 + struct au_drinfo_do_load_args args = {
12012 + .drinfop = &drinfo,
12013 + .h_ppath = &w->h_ppath,
12014 + .whname = whname,
12015 + .whnamelen = whnamelen,
12016 + .info_dentry = &info_dentry
12018 + wkq_err = au_wkq_wait(au_call_drinfo_do_load, &args);
12019 + if (unlikely(wkq_err))
12020 + drinfo = ERR_PTR(wkq_err);
12022 + err = PTR_ERR(drinfo);
12023 + if (IS_ERR_OR_NULL(drinfo))
12027 + oldname.len = drinfo->oldnamelen;
12028 + oldname.name = drinfo->oldname;
12029 + if (au_qstreq(w->qname, &oldname)) {
12030 + /* the name is renamed back */
12034 + infopath.dentry = info_dentry;
12035 + infopath.mnt = w->h_ppath.mnt;
12036 + h_dir = d_inode(w->h_ppath.dentry);
12037 + delegated = NULL;
12038 + inode_lock_nested(h_dir, AuLsc_I_PARENT);
12039 + e = vfsub_unlink(h_dir, &infopath, &delegated, !w->no_sio);
12040 + inode_unlock(h_dir);
12042 + AuIOErr("ignored %d, %pd2\n", e, &infopath.dentry);
12043 + if (unlikely(e == -EWOULDBLOCK))
12046 + kfree(w->drinfo[bindex]);
12047 + w->drinfo[bindex] = drinfo;
12048 + dput(info_dentry);
12055 +/* ---------------------------------------------------------------------- */
12057 +static void au_dr_lkup_free(struct au_drinfo **drinfo, int n)
12059 + struct au_drinfo **p = drinfo;
12062 + kfree(*drinfo++);
12066 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12067 + aufs_bindex_t btgt)
12070 + struct au_drinfo_load w;
12071 + aufs_bindex_t bindex, bbot;
12072 + struct au_branch *br;
12073 + struct inode *h_dir;
12074 + struct au_dr_hino *ent;
12075 + struct super_block *sb;
12077 + AuDbg("%.*s, name %.*s, whname %.*s, b%d\n",
12078 + AuLNPair(&dentry->d_name), AuLNPair(&lkup->dirren.dr_name),
12079 + AuLNPair(&lkup->whname), btgt);
12081 + sb = dentry->d_sb;
12082 + bbot = au_sbbot(sb);
12083 + w.ninfo = bbot + 1;
12084 + if (!lkup->dirren.drinfo) {
12085 + lkup->dirren.drinfo = kcalloc(w.ninfo,
12086 + sizeof(*lkup->dirren.drinfo),
12088 + if (unlikely(!lkup->dirren.drinfo)) {
12092 + lkup->dirren.ninfo = w.ninfo;
12094 + w.drinfo = lkup->dirren.drinfo;
12095 + w.no_sio = !!uid_eq(current_fsuid(), GLOBAL_ROOT_UID);
12096 + w.h_ppath.dentry = au_h_dptr(dentry, btgt);
12097 + AuDebugOn(!w.h_ppath.dentry);
12098 + w.h_ppath.mnt = au_sbr_mnt(sb, btgt);
12099 + w.qname = &dentry->d_name;
12102 + for (bindex = btgt + 1; bindex <= bbot; bindex++) {
12103 + br = au_sbr(sb, bindex);
12104 + err = au_drinfo_load(&w, bindex, br);
12105 + if (unlikely(err))
12107 + if (w.drinfo[bindex])
12111 + br = au_sbr(sb, btgt);
12112 + h_dir = d_inode(w.h_ppath.dentry);
12113 + ent = au_dr_hino_find(&br->br_dirren, h_dir->i_ino);
12115 + au_dr_hino_del(&br->br_dirren, ent);
12118 + goto out; /* success */
12121 + au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12122 + lkup->dirren.ninfo = 0;
12123 + lkup->dirren.drinfo = NULL;
12129 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup)
12131 + au_dr_lkup_free(lkup->dirren.drinfo, lkup->dirren.ninfo);
12134 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt)
12137 + struct au_drinfo *drinfo;
12140 + if (!lkup->dirren.drinfo)
12142 + AuDebugOn(lkup->dirren.ninfo < btgt + 1);
12143 + drinfo = lkup->dirren.drinfo[btgt + 1];
12147 + kfree(lkup->whname.name);
12148 + lkup->whname.name = NULL;
12149 + lkup->dirren.dr_name.len = drinfo->oldnamelen;
12150 + lkup->dirren.dr_name.name = drinfo->oldname;
12151 + lkup->name = &lkup->dirren.dr_name;
12152 + err = au_wh_name_alloc(&lkup->whname, lkup->name);
12154 + AuDbg("name %.*s, whname %.*s, b%d\n",
12155 + AuLNPair(lkup->name), AuLNPair(&lkup->whname),
12163 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12167 + struct au_drinfo *drinfo;
12170 + if (!lkup->dirren.drinfo)
12172 + AuDebugOn(lkup->dirren.ninfo < bindex + 1);
12173 + drinfo = lkup->dirren.drinfo[bindex + 1];
12177 + match = (drinfo->ino == h_ino);
12178 + AuDbg("match %d\n", match);
12184 +/* ---------------------------------------------------------------------- */
12186 +int au_dr_opt_set(struct super_block *sb)
12189 + aufs_bindex_t bindex, bbot;
12190 + struct au_branch *br;
12193 + bbot = au_sbbot(sb);
12194 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
12195 + br = au_sbr(sb, bindex);
12196 + err = au_dr_hino(sb, bindex, /*br*/NULL, &br->br_path);
12202 +int au_dr_opt_flush(struct super_block *sb)
12205 + aufs_bindex_t bindex, bbot;
12206 + struct au_branch *br;
12209 + bbot = au_sbbot(sb);
12210 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
12211 + br = au_sbr(sb, bindex);
12212 + if (au_br_writable(br->br_perm))
12213 + err = au_dr_hino(sb, bindex, /*br*/NULL, /*path*/NULL);
12219 +int au_dr_opt_clr(struct super_block *sb, int no_flush)
12222 + aufs_bindex_t bindex, bbot;
12223 + struct au_branch *br;
12227 + err = au_dr_opt_flush(sb);
12228 + if (unlikely(err))
12232 + bbot = au_sbbot(sb);
12233 + for (bindex = 0; bindex <= bbot; bindex++) {
12234 + br = au_sbr(sb, bindex);
12235 + au_dr_hino_free(&br->br_dirren);
12241 diff -urN /usr/share/empty/fs/aufs/dirren.h linux/fs/aufs/dirren.h
12242 --- /usr/share/empty/fs/aufs/dirren.h 1970-01-01 01:00:00.000000000 +0100
12243 +++ linux/fs/aufs/dirren.h 2018-01-29 07:56:20.053325069 +0100
12246 + * Copyright (C) 2017 Junjiro R. Okajima
12248 + * This program, aufs is free software; you can redistribute it and/or modify
12249 + * it under the terms of the GNU General Public License as published by
12250 + * the Free Software Foundation; either version 2 of the License, or
12251 + * (at your option) any later version.
12253 + * This program is distributed in the hope that it will be useful,
12254 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12255 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12256 + * GNU General Public License for more details.
12258 + * You should have received a copy of the GNU General Public License
12259 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12263 + * renamed dir info
12266 +#ifndef __AUFS_DIRREN_H__
12267 +#define __AUFS_DIRREN_H__
12271 +#include <linux/dcache.h>
12272 +#include <linux/statfs.h>
12273 +#include <linux/uuid.h>
12276 +#define AuDirren_NHASH 100
12278 +#ifdef CONFIG_AUFS_DIRREN
12279 +enum au_brid_type {
12286 +struct au_dr_brid {
12287 + enum au_brid_type type;
12289 + uuid_t uuid; /* unimplemented yet */
12295 +/* 20 is the max digits length of ulong 64 */
12296 +/* brid-type "_" uuid "_" inum */
12297 +#define AUFS_DIRREN_FNAME_SZ (1 + 1 + UUID_STRING_LEN + 20)
12298 +#define AUFS_DIRREN_ENV_VAL_SZ (AUFS_DIRREN_FNAME_SZ + 1 + 20)
12300 +struct au_dr_hino {
12301 + struct hlist_bl_node dr_hnode;
12306 + struct hlist_bl_head dr_h_ino[AuDirren_NHASH];
12307 + struct au_dr_brid dr_brid;
12310 +struct au_dr_lookup {
12311 + /* dr_name is pointed by struct au_do_lookup_args.name */
12312 + struct qstr dr_name; /* subset of dr_info */
12313 + aufs_bindex_t ninfo;
12314 + struct au_drinfo **drinfo;
12317 +struct au_dr_hino;
12319 +struct au_dr_br { };
12320 +struct au_dr_lookup { };
12323 +/* ---------------------------------------------------------------------- */
12326 +struct au_do_lookup_args;
12328 +#ifdef CONFIG_AUFS_DIRREN
12329 +int au_dr_hino_test_add(struct au_dr_br *dr, ino_t h_ino,
12330 + struct au_dr_hino *add_ent);
12331 +void au_dr_hino_free(struct au_dr_br *dr);
12332 +int au_dr_br_init(struct super_block *sb, struct au_branch *br,
12333 + const struct path *path);
12334 +int au_dr_br_fin(struct super_block *sb, struct au_branch *br);
12335 +int au_dr_rename(struct dentry *src, aufs_bindex_t bindex,
12336 + struct qstr *dst_name, void *_rev);
12337 +void au_dr_rename_fin(struct dentry *src, aufs_bindex_t btgt, void *rev);
12338 +void au_dr_rename_rev(struct dentry *src, aufs_bindex_t bindex, void *rev);
12339 +int au_dr_lkup(struct au_do_lookup_args *lkup, struct dentry *dentry,
12340 + aufs_bindex_t bindex);
12341 +int au_dr_lkup_name(struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12342 +int au_dr_lkup_h_ino(struct au_do_lookup_args *lkup, aufs_bindex_t bindex,
12344 +void au_dr_lkup_fin(struct au_do_lookup_args *lkup);
12345 +int au_dr_opt_set(struct super_block *sb);
12346 +int au_dr_opt_flush(struct super_block *sb);
12347 +int au_dr_opt_clr(struct super_block *sb, int no_flush);
12349 +AuStubInt0(au_dr_hino_test_add, struct au_dr_br *dr, ino_t h_ino,
12350 + struct au_dr_hino *add_ent);
12351 +AuStubVoid(au_dr_hino_free, struct au_dr_br *dr);
12352 +AuStubInt0(au_dr_br_init, struct super_block *sb, struct au_branch *br,
12353 + const struct path *path);
12354 +AuStubInt0(au_dr_br_fin, struct super_block *sb, struct au_branch *br);
12355 +AuStubInt0(au_dr_rename, struct dentry *src, aufs_bindex_t bindex,
12356 + struct qstr *dst_name, void *_rev);
12357 +AuStubVoid(au_dr_rename_fin, struct dentry *src, aufs_bindex_t btgt, void *rev);
12358 +AuStubVoid(au_dr_rename_rev, struct dentry *src, aufs_bindex_t bindex,
12360 +AuStubInt0(au_dr_lkup, struct au_do_lookup_args *lkup, struct dentry *dentry,
12361 + aufs_bindex_t bindex);
12362 +AuStubInt0(au_dr_lkup_name, struct au_do_lookup_args *lkup, aufs_bindex_t btgt);
12363 +AuStubInt0(au_dr_lkup_h_ino, struct au_do_lookup_args *lkup,
12364 + aufs_bindex_t bindex, ino_t h_ino);
12365 +AuStubVoid(au_dr_lkup_fin, struct au_do_lookup_args *lkup);
12366 +AuStubInt0(au_dr_opt_set, struct super_block *sb);
12367 +AuStubInt0(au_dr_opt_flush, struct super_block *sb);
12368 +AuStubInt0(au_dr_opt_clr, struct super_block *sb, int no_flush);
12371 +/* ---------------------------------------------------------------------- */
12373 +#ifdef CONFIG_AUFS_DIRREN
12374 +static inline int au_dr_ihash(ino_t h_ino)
12376 + return h_ino % AuDirren_NHASH;
12379 +AuStubInt0(au_dr_ihash, ino_t h_ino);
12382 +#endif /* __KERNEL__ */
12383 +#endif /* __AUFS_DIRREN_H__ */
12384 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
12385 --- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
12386 +++ linux/fs/aufs/dynop.c 2018-01-29 07:56:20.053325069 +0100
12389 + * Copyright (C) 2010-2017 Junjiro R. Okajima
12391 + * This program, aufs is free software; you can redistribute it and/or modify
12392 + * it under the terms of the GNU General Public License as published by
12393 + * the Free Software Foundation; either version 2 of the License, or
12394 + * (at your option) any later version.
12396 + * This program is distributed in the hope that it will be useful,
12397 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12398 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12399 + * GNU General Public License for more details.
12401 + * You should have received a copy of the GNU General Public License
12402 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12406 + * dynamically customizable operations for regular files
12411 +#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
12414 + * How large will these lists be?
12415 + * Usually just a few elements, 20-30 at most for each, I guess.
12417 +static struct hlist_bl_head dynop[AuDyLast];
12419 +static struct au_dykey *dy_gfind_get(struct hlist_bl_head *hbl,
12420 + const void *h_op)
12422 + struct au_dykey *key, *tmp;
12423 + struct hlist_bl_node *pos;
12426 + hlist_bl_lock(hbl);
12427 + hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12428 + if (tmp->dk_op.dy_hop == h_op) {
12430 + kref_get(&key->dk_kref);
12433 + hlist_bl_unlock(hbl);
12438 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
12440 + struct au_dykey **k, *found;
12441 + const void *h_op = key->dk_op.dy_hop;
12445 + k = br->br_dykey;
12446 + for (i = 0; i < AuBrDynOp; i++)
12448 + if (k[i]->dk_op.dy_hop == h_op) {
12455 + spin_lock(&br->br_dykey_lock);
12456 + for (; i < AuBrDynOp; i++)
12458 + if (k[i]->dk_op.dy_hop == h_op) {
12466 + spin_unlock(&br->br_dykey_lock);
12467 + BUG_ON(i == AuBrDynOp); /* expand the array */
12473 +/* kref_get() if @key is already added */
12474 +static struct au_dykey *dy_gadd(struct hlist_bl_head *hbl, struct au_dykey *key)
12476 + struct au_dykey *tmp, *found;
12477 + struct hlist_bl_node *pos;
12478 + const void *h_op = key->dk_op.dy_hop;
12481 + hlist_bl_lock(hbl);
12482 + hlist_bl_for_each_entry(tmp, pos, hbl, dk_hnode)
12483 + if (tmp->dk_op.dy_hop == h_op) {
12484 + kref_get(&tmp->dk_kref);
12489 + hlist_bl_add_head(&key->dk_hnode, hbl);
12490 + hlist_bl_unlock(hbl);
12497 +static void dy_free_rcu(struct rcu_head *rcu)
12499 + struct au_dykey *key;
12501 + key = container_of(rcu, struct au_dykey, dk_rcu);
12506 +static void dy_free(struct kref *kref)
12508 + struct au_dykey *key;
12509 + struct hlist_bl_head *hbl;
12511 + key = container_of(kref, struct au_dykey, dk_kref);
12512 + hbl = dynop + key->dk_op.dy_type;
12513 + au_hbl_del(&key->dk_hnode, hbl);
12514 + call_rcu(&key->dk_rcu, dy_free_rcu);
12517 +void au_dy_put(struct au_dykey *key)
12519 + kref_put(&key->dk_kref, dy_free);
12522 +/* ---------------------------------------------------------------------- */
12524 +#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
12526 +#ifdef CONFIG_AUFS_DEBUG
12527 +#define DyDbgDeclare(cnt) unsigned int cnt = 0
12528 +#define DyDbgInc(cnt) do { cnt++; } while (0)
12530 +#define DyDbgDeclare(cnt) do {} while (0)
12531 +#define DyDbgInc(cnt) do {} while (0)
12534 +#define DySet(func, dst, src, h_op, h_sb) do { \
12536 + if (h_op->func) { \
12538 + dst.func = src.func; \
12540 + AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
12544 +#define DySetForce(func, dst, src) do { \
12545 + AuDebugOn(!src.func); \
12547 + dst.func = src.func; \
12550 +#define DySetAop(func) \
12551 + DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
12552 +#define DySetAopForce(func) \
12553 + DySetForce(func, dyaop->da_op, aufs_aop)
12555 +static void dy_aop(struct au_dykey *key, const void *h_op,
12556 + struct super_block *h_sb __maybe_unused)
12558 + struct au_dyaop *dyaop = (void *)key;
12559 + const struct address_space_operations *h_aop = h_op;
12560 + DyDbgDeclare(cnt);
12562 + AuDbg("%s\n", au_sbtype(h_sb));
12564 + DySetAop(writepage);
12565 + DySetAopForce(readpage); /* force */
12566 + DySetAop(writepages);
12567 + DySetAop(set_page_dirty);
12568 + DySetAop(readpages);
12569 + DySetAop(write_begin);
12570 + DySetAop(write_end);
12572 + DySetAop(invalidatepage);
12573 + DySetAop(releasepage);
12574 + DySetAop(freepage);
12575 + /* this one will be changed according to an aufs mount option */
12576 + DySetAop(direct_IO);
12577 + DySetAop(migratepage);
12578 + DySetAop(isolate_page);
12579 + DySetAop(putback_page);
12580 + DySetAop(launder_page);
12581 + DySetAop(is_partially_uptodate);
12582 + DySetAop(is_dirty_writeback);
12583 + DySetAop(error_remove_page);
12584 + DySetAop(swap_activate);
12585 + DySetAop(swap_deactivate);
12587 + DyDbgSize(cnt, *h_aop);
12590 +/* ---------------------------------------------------------------------- */
12592 +static void dy_bug(struct kref *kref)
12597 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
12599 + struct au_dykey *key, *old;
12600 + struct hlist_bl_head *hbl;
12603 + void (*set)(struct au_dykey *key, const void *h_op,
12604 + struct super_block *h_sb __maybe_unused);
12606 + static const struct op a[] = {
12608 + .sz = sizeof(struct au_dyaop),
12612 + const struct op *p;
12614 + hbl = dynop + op->dy_type;
12615 + key = dy_gfind_get(hbl, op->dy_hop);
12617 + goto out_add; /* success */
12619 + p = a + op->dy_type;
12620 + key = kzalloc(p->sz, GFP_NOFS);
12621 + if (unlikely(!key)) {
12622 + key = ERR_PTR(-ENOMEM);
12626 + key->dk_op.dy_hop = op->dy_hop;
12627 + kref_init(&key->dk_kref);
12628 + p->set(key, op->dy_hop, au_br_sb(br));
12629 + old = dy_gadd(hbl, key);
12636 + old = dy_bradd(br, key);
12638 + /* its ref-count should never be zero here */
12639 + kref_put(&key->dk_kref, dy_bug);
12644 +/* ---------------------------------------------------------------------- */
12646 + * Aufs prohibits O_DIRECT by defaut even if the branch supports it.
12647 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
12648 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
12649 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
12650 + * See the aufs manual in detail.
12652 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
12655 + dyaop->da_op.direct_IO = NULL;
12657 + dyaop->da_op.direct_IO = aufs_aop.direct_IO;
12660 +static struct au_dyaop *dy_aget(struct au_branch *br,
12661 + const struct address_space_operations *h_aop,
12664 + struct au_dyaop *dyaop;
12665 + struct au_dynop op;
12667 + op.dy_type = AuDy_AOP;
12668 + op.dy_haop = h_aop;
12669 + dyaop = (void *)dy_get(&op, br);
12670 + if (IS_ERR(dyaop))
12672 + dy_adx(dyaop, do_dx);
12678 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12679 + struct inode *h_inode)
12682 + struct super_block *sb;
12683 + struct au_branch *br;
12684 + struct au_dyaop *dyaop;
12686 + AuDebugOn(!S_ISREG(h_inode->i_mode));
12687 + IiMustWriteLock(inode);
12689 + sb = inode->i_sb;
12690 + br = au_sbr(sb, bindex);
12691 + do_dx = !!au_opt_test(au_mntflags(sb), DIO);
12692 + dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
12693 + err = PTR_ERR(dyaop);
12694 + if (IS_ERR(dyaop))
12695 + /* unnecessary to call dy_fput() */
12699 + inode->i_mapping->a_ops = &dyaop->da_op;
12706 + * Is it safe to replace a_ops during the inode/file is in operation?
12707 + * Yes, I hope so.
12709 +int au_dy_irefresh(struct inode *inode)
12712 + aufs_bindex_t btop;
12713 + struct inode *h_inode;
12716 + if (S_ISREG(inode->i_mode)) {
12717 + btop = au_ibtop(inode);
12718 + h_inode = au_h_iptr(inode, btop);
12719 + err = au_dy_iaop(inode, btop, h_inode);
12724 +void au_dy_arefresh(int do_dx)
12726 + struct hlist_bl_head *hbl;
12727 + struct hlist_bl_node *pos;
12728 + struct au_dykey *key;
12730 + hbl = dynop + AuDy_AOP;
12731 + hlist_bl_lock(hbl);
12732 + hlist_bl_for_each_entry(key, pos, hbl, dk_hnode)
12733 + dy_adx((void *)key, do_dx);
12734 + hlist_bl_unlock(hbl);
12737 +/* ---------------------------------------------------------------------- */
12739 +void __init au_dy_init(void)
12743 + /* make sure that 'struct au_dykey *' can be any type */
12744 + BUILD_BUG_ON(offsetof(struct au_dyaop, da_key));
12746 + for (i = 0; i < AuDyLast; i++)
12747 + INIT_HLIST_BL_HEAD(dynop + i);
12750 +void au_dy_fin(void)
12754 + for (i = 0; i < AuDyLast; i++)
12755 + WARN_ON(!hlist_bl_empty(dynop + i));
12757 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
12758 --- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
12759 +++ linux/fs/aufs/dynop.h 2018-01-29 07:56:20.053325069 +0100
12762 + * Copyright (C) 2010-2017 Junjiro R. Okajima
12764 + * This program, aufs is free software; you can redistribute it and/or modify
12765 + * it under the terms of the GNU General Public License as published by
12766 + * the Free Software Foundation; either version 2 of the License, or
12767 + * (at your option) any later version.
12769 + * This program is distributed in the hope that it will be useful,
12770 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12771 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12772 + * GNU General Public License for more details.
12774 + * You should have received a copy of the GNU General Public License
12775 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12779 + * dynamically customizable operations (for regular files only)
12782 +#ifndef __AUFS_DYNOP_H__
12783 +#define __AUFS_DYNOP_H__
12787 +#include <linux/fs.h>
12788 +#include <linux/kref.h>
12790 +enum {AuDy_AOP, AuDyLast};
12795 + const void *dy_hop;
12796 + const struct address_space_operations *dy_haop;
12802 + struct hlist_bl_node dk_hnode;
12803 + struct rcu_head dk_rcu;
12805 + struct au_dynop dk_op;
12808 + * during I am in the branch local array, kref is gotten. when the
12809 + * branch is removed, kref is put.
12811 + struct kref dk_kref;
12814 +/* stop unioning since their sizes are very different from each other */
12816 + struct au_dykey da_key;
12817 + struct address_space_operations da_op; /* not const */
12820 +/* ---------------------------------------------------------------------- */
12824 +void au_dy_put(struct au_dykey *key);
12825 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
12826 + struct inode *h_inode);
12827 +int au_dy_irefresh(struct inode *inode);
12828 +void au_dy_arefresh(int do_dio);
12830 +void __init au_dy_init(void);
12831 +void au_dy_fin(void);
12833 +#endif /* __KERNEL__ */
12834 +#endif /* __AUFS_DYNOP_H__ */
12835 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
12836 --- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
12837 +++ linux/fs/aufs/export.c 2018-01-29 07:56:20.053325069 +0100
12840 + * Copyright (C) 2005-2017 Junjiro R. Okajima
12842 + * This program, aufs is free software; you can redistribute it and/or modify
12843 + * it under the terms of the GNU General Public License as published by
12844 + * the Free Software Foundation; either version 2 of the License, or
12845 + * (at your option) any later version.
12847 + * This program is distributed in the hope that it will be useful,
12848 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12849 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12850 + * GNU General Public License for more details.
12852 + * You should have received a copy of the GNU General Public License
12853 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12860 +#include <linux/exportfs.h>
12861 +#include <linux/fs_struct.h>
12862 +#include <linux/namei.h>
12863 +#include <linux/nsproxy.h>
12864 +#include <linux/random.h>
12865 +#include <linux/writeback.h>
12869 +#ifdef CONFIG_AUFS_INO_T_64
12877 +static ino_t decode_ino(__u32 *a)
12881 + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
12883 +#ifdef CONFIG_AUFS_INO_T_64
12889 +static void encode_ino(__u32 *a, ino_t ino)
12895 +#ifdef CONFIG_AUFS_INO_T_64
12900 +/* NFS file handle */
12904 +#ifdef CONFIG_AUFS_INO_T_64
12905 + /* support 64bit inode number */
12918 + Fh_ino = Fh_ino1,
12919 + Fh_dir_ino = Fh_dir_ino1
12922 +static int au_test_anon(struct dentry *dentry)
12924 + /* note: read d_flags without d_lock */
12925 + return !!(dentry->d_flags & DCACHE_DISCONNECTED);
12928 +int au_test_nfsd(void)
12931 + struct task_struct *tsk = current;
12932 + char comm[sizeof(tsk->comm)];
12935 + if (tsk->flags & PF_KTHREAD) {
12936 + get_task_comm(comm, tsk);
12937 + ret = !strcmp(comm, "nfsd");
12943 +/* ---------------------------------------------------------------------- */
12944 +/* inode generation external table */
12946 +void au_xigen_inc(struct inode *inode)
12951 + struct super_block *sb;
12952 + struct au_sbinfo *sbinfo;
12954 + sb = inode->i_sb;
12955 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
12957 + sbinfo = au_sbi(sb);
12958 + pos = inode->i_ino;
12959 + pos *= sizeof(igen);
12960 + igen = inode->i_generation + 1;
12961 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xigen, &igen,
12962 + sizeof(igen), &pos);
12963 + if (sz == sizeof(igen))
12964 + return; /* success */
12966 + if (unlikely(sz >= 0))
12967 + AuIOErr("xigen error (%zd)\n", sz);
12970 +int au_xigen_new(struct inode *inode)
12975 + struct super_block *sb;
12976 + struct au_sbinfo *sbinfo;
12977 + struct file *file;
12980 + /* todo: dirty, at mount time */
12981 + if (inode->i_ino == AUFS_ROOT_INO)
12983 + sb = inode->i_sb;
12984 + SiMustAnyLock(sb);
12985 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
12989 + pos = inode->i_ino;
12990 + if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
12991 + AuIOErr1("too large i%lld\n", pos);
12994 + pos *= sizeof(inode->i_generation);
12997 + sbinfo = au_sbi(sb);
12998 + file = sbinfo->si_xigen;
13001 + if (vfsub_f_size_read(file)
13002 + < pos + sizeof(inode->i_generation)) {
13003 + inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
13004 + sz = xino_fwrite(sbinfo->si_xwrite, file, &inode->i_generation,
13005 + sizeof(inode->i_generation), &pos);
13007 + sz = xino_fread(sbinfo->si_xread, file, &inode->i_generation,
13008 + sizeof(inode->i_generation), &pos);
13009 + if (sz == sizeof(inode->i_generation))
13010 + goto out; /* success */
13013 + if (unlikely(sz >= 0)) {
13015 + AuIOErr("xigen error (%zd)\n", sz);
13022 +int au_xigen_set(struct super_block *sb, struct file *base)
13025 + struct au_sbinfo *sbinfo;
13026 + struct file *file;
13028 + SiMustWriteLock(sb);
13030 + sbinfo = au_sbi(sb);
13031 + file = au_xino_create2(base, sbinfo->si_xigen);
13032 + err = PTR_ERR(file);
13033 + if (IS_ERR(file))
13036 + if (sbinfo->si_xigen)
13037 + fput(sbinfo->si_xigen);
13038 + sbinfo->si_xigen = file;
13044 +void au_xigen_clr(struct super_block *sb)
13046 + struct au_sbinfo *sbinfo;
13048 + SiMustWriteLock(sb);
13050 + sbinfo = au_sbi(sb);
13051 + if (sbinfo->si_xigen) {
13052 + fput(sbinfo->si_xigen);
13053 + sbinfo->si_xigen = NULL;
13057 +/* ---------------------------------------------------------------------- */
13059 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
13062 + struct dentry *dentry, *d;
13063 + struct inode *inode;
13064 + unsigned int sigen;
13067 + inode = ilookup(sb, ino);
13071 + dentry = ERR_PTR(-ESTALE);
13072 + sigen = au_sigen(sb);
13073 + if (unlikely(au_is_bad_inode(inode)
13074 + || IS_DEADDIR(inode)
13075 + || sigen != au_iigen(inode, NULL)))
13079 + if (!dir_ino || S_ISDIR(inode->i_mode))
13080 + dentry = d_find_alias(inode);
13082 + spin_lock(&inode->i_lock);
13083 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
13084 + spin_lock(&d->d_lock);
13085 + if (!au_test_anon(d)
13086 + && d_inode(d->d_parent)->i_ino == dir_ino) {
13087 + dentry = dget_dlock(d);
13088 + spin_unlock(&d->d_lock);
13091 + spin_unlock(&d->d_lock);
13093 + spin_unlock(&inode->i_lock);
13095 + if (unlikely(dentry && au_digen_test(dentry, sigen))) {
13096 + /* need to refresh */
13104 + AuTraceErrPtr(dentry);
13108 +/* ---------------------------------------------------------------------- */
13110 +/* todo: dirty? */
13111 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
13113 +struct au_compare_mnt_args {
13115 + struct super_block *sb;
13118 + struct vfsmount *mnt;
13121 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
13123 + struct au_compare_mnt_args *a = arg;
13125 + if (mnt->mnt_sb != a->sb)
13127 + a->mnt = mntget(mnt);
13131 +static struct vfsmount *au_mnt_get(struct super_block *sb)
13134 + struct path root;
13135 + struct au_compare_mnt_args args = {
13139 + get_fs_root(current->fs, &root);
13141 + err = iterate_mounts(au_compare_mnt, &args, root.mnt);
13142 + rcu_read_unlock();
13145 + AuDebugOn(!args.mnt);
13149 +struct au_nfsd_si_lock {
13150 + unsigned int sigen;
13151 + aufs_bindex_t bindex, br_id;
13152 + unsigned char force_lock;
13155 +static int si_nfsd_read_lock(struct super_block *sb,
13156 + struct au_nfsd_si_lock *nsi_lock)
13159 + aufs_bindex_t bindex;
13161 + si_read_lock(sb, AuLock_FLUSH);
13163 + /* branch id may be wrapped around */
13165 + bindex = au_br_index(sb, nsi_lock->br_id);
13166 + if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
13167 + goto out; /* success */
13171 + if (!nsi_lock->force_lock)
13172 + si_read_unlock(sb);
13175 + nsi_lock->bindex = bindex;
13179 +struct find_name_by_ino {
13180 + struct dir_context ctx;
13181 + int called, found;
13188 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
13189 + loff_t offset, u64 ino, unsigned int d_type)
13191 + struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
13195 + if (a->ino != ino)
13198 + memcpy(a->name, name, namelen);
13199 + a->namelen = namelen;
13204 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
13205 + struct au_nfsd_si_lock *nsi_lock)
13207 + struct dentry *dentry, *parent;
13208 + struct file *file;
13209 + struct inode *dir;
13210 + struct find_name_by_ino arg = {
13212 + .actor = find_name_by_ino
13217 + parent = path->dentry;
13219 + si_read_unlock(parent->d_sb);
13220 + file = vfsub_dentry_open(path, au_dir_roflags);
13221 + dentry = (void *)file;
13222 + if (IS_ERR(file))
13225 + dentry = ERR_PTR(-ENOMEM);
13226 + arg.name = (void *)__get_free_page(GFP_NOFS);
13227 + if (unlikely(!arg.name))
13234 + err = vfsub_iterate_dir(file, &arg.ctx);
13235 + } while (!err && !arg.found && arg.called);
13236 + dentry = ERR_PTR(err);
13237 + if (unlikely(err))
13239 + /* instead of ENOENT */
13240 + dentry = ERR_PTR(-ESTALE);
13244 + /* do not call vfsub_lkup_one() */
13245 + dir = d_inode(parent);
13246 + dentry = vfsub_lookup_one_len_unlocked(arg.name, parent, arg.namelen);
13247 + AuTraceErrPtr(dentry);
13248 + if (IS_ERR(dentry))
13250 + AuDebugOn(au_test_anon(dentry));
13251 + if (unlikely(d_really_is_negative(dentry))) {
13253 + dentry = ERR_PTR(-ENOENT);
13257 + free_page((unsigned long)arg.name);
13261 + if (unlikely(nsi_lock
13262 + && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
13263 + if (!IS_ERR(dentry)) {
13265 + dentry = ERR_PTR(-ESTALE);
13267 + AuTraceErrPtr(dentry);
13271 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
13273 + struct au_nfsd_si_lock *nsi_lock)
13275 + struct dentry *dentry;
13276 + struct path path;
13278 + if (dir_ino != AUFS_ROOT_INO) {
13279 + path.dentry = decode_by_ino(sb, dir_ino, 0);
13280 + dentry = path.dentry;
13281 + if (!path.dentry || IS_ERR(path.dentry))
13283 + AuDebugOn(au_test_anon(path.dentry));
13285 + path.dentry = dget(sb->s_root);
13287 + path.mnt = au_mnt_get(sb);
13288 + dentry = au_lkup_by_ino(&path, ino, nsi_lock);
13292 + AuTraceErrPtr(dentry);
13296 +/* ---------------------------------------------------------------------- */
13298 +static int h_acceptable(void *expv, struct dentry *dentry)
13303 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
13304 + char *buf, int len, struct super_block *sb)
13308 + struct path path;
13310 + p = d_path(h_rootpath, buf, len);
13315 + path.mnt = h_rootpath->mnt;
13316 + path.dentry = h_parent;
13317 + p = d_path(&path, buf, len);
13323 + path.mnt = au_mnt_get(sb);
13324 + path.dentry = sb->s_root;
13325 + p = d_path(&path, buf, len - strlen(p));
13326 + mntput(path.mnt);
13330 + p[strlen(p)] = '/';
13333 + AuTraceErrPtr(p);
13338 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
13339 + int fh_len, struct au_nfsd_si_lock *nsi_lock)
13341 + struct dentry *dentry, *h_parent, *root;
13342 + struct super_block *h_sb;
13343 + char *pathname, *p;
13344 + struct vfsmount *h_mnt;
13345 + struct au_branch *br;
13347 + struct path path;
13349 + br = au_sbr(sb, nsi_lock->bindex);
13350 + h_mnt = au_br_mnt(br);
13351 + h_sb = h_mnt->mnt_sb;
13352 + /* todo: call lower fh_to_dentry()? fh_to_parent()? */
13354 + h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
13355 + fh_len - Fh_tail, fh[Fh_h_type],
13356 + h_acceptable, /*context*/NULL);
13358 + dentry = h_parent;
13359 + if (unlikely(!h_parent || IS_ERR(h_parent))) {
13360 + AuWarn1("%s decode_fh failed, %ld\n",
13361 + au_sbtype(h_sb), PTR_ERR(h_parent));
13365 + if (unlikely(au_test_anon(h_parent))) {
13366 + AuWarn1("%s decode_fh returned a disconnected dentry\n",
13367 + au_sbtype(h_sb));
13368 + goto out_h_parent;
13371 + dentry = ERR_PTR(-ENOMEM);
13372 + pathname = (void *)__get_free_page(GFP_NOFS);
13373 + if (unlikely(!pathname))
13374 + goto out_h_parent;
13376 + root = sb->s_root;
13377 + path.mnt = h_mnt;
13378 + di_read_lock_parent(root, !AuLock_IR);
13379 + path.dentry = au_h_dptr(root, nsi_lock->bindex);
13380 + di_read_unlock(root, !AuLock_IR);
13381 + p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
13382 + dentry = (void *)p;
13384 + goto out_pathname;
13386 + si_read_unlock(sb);
13387 + err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
13388 + dentry = ERR_PTR(err);
13389 + if (unlikely(err))
13392 + dentry = ERR_PTR(-ENOENT);
13393 + AuDebugOn(au_test_anon(path.dentry));
13394 + if (unlikely(d_really_is_negative(path.dentry)))
13397 + if (ino != d_inode(path.dentry)->i_ino)
13398 + dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
13400 + dentry = dget(path.dentry);
13405 + if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
13406 + if (!IS_ERR(dentry)) {
13408 + dentry = ERR_PTR(-ESTALE);
13411 + free_page((unsigned long)pathname);
13415 + AuTraceErrPtr(dentry);
13419 +/* ---------------------------------------------------------------------- */
13421 +static struct dentry *
13422 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
13425 + struct dentry *dentry;
13426 + __u32 *fh = fid->raw;
13427 + struct au_branch *br;
13428 + ino_t ino, dir_ino;
13429 + struct au_nfsd_si_lock nsi_lock = {
13433 + dentry = ERR_PTR(-ESTALE);
13434 + /* it should never happen, but the file handle is unreliable */
13435 + if (unlikely(fh_len < Fh_tail))
13437 + nsi_lock.sigen = fh[Fh_sigen];
13438 + nsi_lock.br_id = fh[Fh_br_id];
13440 + /* branch id may be wrapped around */
13442 + if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
13444 + nsi_lock.force_lock = 1;
13446 + /* is this inode still cached? */
13447 + ino = decode_ino(fh + Fh_ino);
13448 + /* it should never happen */
13449 + if (unlikely(ino == AUFS_ROOT_INO))
13452 + dir_ino = decode_ino(fh + Fh_dir_ino);
13453 + dentry = decode_by_ino(sb, ino, dir_ino);
13454 + if (IS_ERR(dentry))
13459 + /* is the parent dir cached? */
13460 + br = au_sbr(sb, nsi_lock.bindex);
13462 + dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
13463 + if (IS_ERR(dentry))
13468 + /* lookup path */
13469 + dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
13470 + if (IS_ERR(dentry))
13472 + if (unlikely(!dentry))
13473 + /* todo?: make it ESTALE */
13477 + if (!au_digen_test(dentry, au_sigen(sb))
13478 + && d_inode(dentry)->i_generation == fh[Fh_igen])
13479 + goto out_unlock; /* success */
13482 + dentry = ERR_PTR(-ESTALE);
13486 + si_read_unlock(sb);
13488 + AuTraceErrPtr(dentry);
13492 +#if 0 /* reserved for future use */
13493 +/* support subtreecheck option */
13494 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
13495 + int fh_len, int fh_type)
13497 + struct dentry *parent;
13498 + __u32 *fh = fid->raw;
13501 + dir_ino = decode_ino(fh + Fh_dir_ino);
13502 + parent = decode_by_ino(sb, dir_ino, 0);
13503 + if (IS_ERR(parent))
13506 + parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
13507 + dir_ino, fh, fh_len);
13510 + AuTraceErrPtr(parent);
13515 +/* ---------------------------------------------------------------------- */
13517 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
13518 + struct inode *dir)
13521 + aufs_bindex_t bindex;
13522 + struct super_block *sb, *h_sb;
13523 + struct dentry *dentry, *parent, *h_parent;
13524 + struct inode *h_dir;
13525 + struct au_branch *br;
13528 + if (unlikely(*max_len <= Fh_tail)) {
13529 + AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
13533 + err = FILEID_ROOT;
13534 + if (inode->i_ino == AUFS_ROOT_INO) {
13535 + AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
13540 + sb = inode->i_sb;
13541 + err = si_read_lock(sb, AuLock_FLUSH);
13542 + if (unlikely(err))
13545 +#ifdef CONFIG_AUFS_DEBUG
13546 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
13547 + AuWarn1("NFS-exporting requires xino\n");
13551 + ii_read_lock_child(inode);
13552 + bindex = au_ibtop(inode);
13554 + dentry = d_find_any_alias(inode);
13555 + if (unlikely(!dentry))
13557 + AuDebugOn(au_test_anon(dentry));
13558 + parent = dget_parent(dentry);
13560 + if (unlikely(!parent))
13562 + if (d_really_is_positive(parent))
13563 + dir = d_inode(parent);
13566 + ii_read_lock_parent(dir);
13567 + h_dir = au_h_iptr(dir, bindex);
13568 + ii_read_unlock(dir);
13569 + if (unlikely(!h_dir))
13571 + h_parent = d_find_any_alias(h_dir);
13572 + if (unlikely(!h_parent))
13573 + goto out_hparent;
13576 + br = au_sbr(sb, bindex);
13577 + h_sb = au_br_sb(br);
13578 + if (unlikely(!h_sb->s_export_op)) {
13579 + AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
13580 + goto out_hparent;
13583 + fh[Fh_br_id] = br->br_id;
13584 + fh[Fh_sigen] = au_sigen(sb);
13585 + encode_ino(fh + Fh_ino, inode->i_ino);
13586 + encode_ino(fh + Fh_dir_ino, dir->i_ino);
13587 + fh[Fh_igen] = inode->i_generation;
13589 + *max_len -= Fh_tail;
13590 + fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
13592 + /*connectable or subtreecheck*/0);
13593 + err = fh[Fh_h_type];
13594 + *max_len += Fh_tail;
13595 + /* todo: macros? */
13596 + if (err != FILEID_INVALID)
13599 + AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
13606 + ii_read_unlock(inode);
13607 + si_read_unlock(sb);
13609 + if (unlikely(err < 0))
13610 + err = FILEID_INVALID;
13614 +/* ---------------------------------------------------------------------- */
13616 +static int aufs_commit_metadata(struct inode *inode)
13619 + aufs_bindex_t bindex;
13620 + struct super_block *sb;
13621 + struct inode *h_inode;
13622 + int (*f)(struct inode *inode);
13624 + sb = inode->i_sb;
13625 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13626 + ii_write_lock_child(inode);
13627 + bindex = au_ibtop(inode);
13628 + AuDebugOn(bindex < 0);
13629 + h_inode = au_h_iptr(inode, bindex);
13631 + f = h_inode->i_sb->s_export_op->commit_metadata;
13633 + err = f(h_inode);
13635 + struct writeback_control wbc = {
13636 + .sync_mode = WB_SYNC_ALL,
13637 + .nr_to_write = 0 /* metadata only */
13640 + err = sync_inode(h_inode, &wbc);
13643 + au_cpup_attr_timesizes(inode);
13644 + ii_write_unlock(inode);
13645 + si_read_unlock(sb);
13649 +/* ---------------------------------------------------------------------- */
13651 +static struct export_operations aufs_export_op = {
13652 + .fh_to_dentry = aufs_fh_to_dentry,
13653 + /* .fh_to_parent = aufs_fh_to_parent, */
13654 + .encode_fh = aufs_encode_fh,
13655 + .commit_metadata = aufs_commit_metadata
13658 +void au_export_init(struct super_block *sb)
13660 + struct au_sbinfo *sbinfo;
13663 + BUILD_BUG_ON_MSG(IS_BUILTIN(CONFIG_AUFS_FS)
13664 + && IS_MODULE(CONFIG_EXPORTFS),
13665 + AUFS_NAME ": unsupported configuration "
13666 + "CONFIG_EXPORTFS=m and CONFIG_AUFS_FS=y");
13668 + sb->s_export_op = &aufs_export_op;
13669 + sbinfo = au_sbi(sb);
13670 + sbinfo->si_xigen = NULL;
13671 + get_random_bytes(&u, sizeof(u));
13672 + BUILD_BUG_ON(sizeof(u) != sizeof(int));
13673 + atomic_set(&sbinfo->si_xigen_next, u);
13675 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
13676 --- /usr/share/empty/fs/aufs/fhsm.c 1970-01-01 01:00:00.000000000 +0100
13677 +++ linux/fs/aufs/fhsm.c 2018-01-29 07:56:20.053325069 +0100
13680 + * Copyright (C) 2011-2017 Junjiro R. Okajima
13682 + * This program, aufs is free software; you can redistribute it and/or modify
13683 + * it under the terms of the GNU General Public License as published by
13684 + * the Free Software Foundation; either version 2 of the License, or
13685 + * (at your option) any later version.
13687 + * This program is distributed in the hope that it will be useful,
13688 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13689 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13690 + * GNU General Public License for more details.
13692 + * You should have received a copy of the GNU General Public License
13693 + * along with this program; if not, write to the Free Software
13694 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
13698 + * File-based Hierarchy Storage Management
13701 +#include <linux/anon_inodes.h>
13702 +#include <linux/poll.h>
13703 +#include <linux/seq_file.h>
13704 +#include <linux/statfs.h>
13707 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
13709 + struct au_sbinfo *sbinfo;
13710 + struct au_fhsm *fhsm;
13712 + SiMustAnyLock(sb);
13714 + sbinfo = au_sbi(sb);
13715 + fhsm = &sbinfo->si_fhsm;
13716 + AuDebugOn(!fhsm);
13717 + return fhsm->fhsm_bottom;
13720 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
13722 + struct au_sbinfo *sbinfo;
13723 + struct au_fhsm *fhsm;
13725 + SiMustWriteLock(sb);
13727 + sbinfo = au_sbi(sb);
13728 + fhsm = &sbinfo->si_fhsm;
13729 + AuDebugOn(!fhsm);
13730 + fhsm->fhsm_bottom = bindex;
13733 +/* ---------------------------------------------------------------------- */
13735 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
13737 + struct au_br_fhsm *bf;
13739 + bf = br->br_fhsm;
13740 + MtxMustLock(&bf->bf_lock);
13742 + return !bf->bf_readable
13743 + || time_after(jiffies,
13744 + bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
13747 +/* ---------------------------------------------------------------------- */
13749 +static void au_fhsm_notify(struct super_block *sb, int val)
13751 + struct au_sbinfo *sbinfo;
13752 + struct au_fhsm *fhsm;
13754 + SiMustAnyLock(sb);
13756 + sbinfo = au_sbi(sb);
13757 + fhsm = &sbinfo->si_fhsm;
13758 + if (au_fhsm_pid(fhsm)
13759 + && atomic_read(&fhsm->fhsm_readable) != -1) {
13760 + atomic_set(&fhsm->fhsm_readable, val);
13762 + wake_up(&fhsm->fhsm_wqh);
13766 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
13767 + struct aufs_stfs *rstfs, int do_lock, int do_notify)
13770 + struct au_branch *br;
13771 + struct au_br_fhsm *bf;
13773 + br = au_sbr(sb, bindex);
13774 + AuDebugOn(au_br_rdonly(br));
13775 + bf = br->br_fhsm;
13779 + mutex_lock(&bf->bf_lock);
13781 + MtxMustLock(&bf->bf_lock);
13783 + /* sb->s_root for NFS is unreliable */
13784 + err = au_br_stfs(br, &bf->bf_stfs);
13785 + if (unlikely(err)) {
13786 + AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
13790 + bf->bf_jiffy = jiffies;
13791 + bf->bf_readable = 1;
13793 + au_fhsm_notify(sb, /*val*/1);
13795 + *rstfs = bf->bf_stfs;
13799 + mutex_unlock(&bf->bf_lock);
13800 + au_fhsm_notify(sb, /*val*/1);
13805 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
13808 + struct au_sbinfo *sbinfo;
13809 + struct au_fhsm *fhsm;
13810 + struct au_branch *br;
13811 + struct au_br_fhsm *bf;
13813 + AuDbg("b%d, force %d\n", bindex, force);
13814 + SiMustAnyLock(sb);
13816 + sbinfo = au_sbi(sb);
13817 + fhsm = &sbinfo->si_fhsm;
13818 + if (!au_ftest_si(sbinfo, FHSM)
13819 + || fhsm->fhsm_bottom == bindex)
13822 + br = au_sbr(sb, bindex);
13823 + bf = br->br_fhsm;
13825 + mutex_lock(&bf->bf_lock);
13827 + || au_fhsm_pid(fhsm)
13828 + || au_fhsm_test_jiffy(sbinfo, br))
13829 + err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
13831 + mutex_unlock(&bf->bf_lock);
13834 +void au_fhsm_wrote_all(struct super_block *sb, int force)
13836 + aufs_bindex_t bindex, bbot;
13837 + struct au_branch *br;
13839 + /* exclude the bottom */
13840 + bbot = au_fhsm_bottom(sb);
13841 + for (bindex = 0; bindex < bbot; bindex++) {
13842 + br = au_sbr(sb, bindex);
13843 + if (au_br_fhsm(br->br_perm))
13844 + au_fhsm_wrote(sb, bindex, force);
13848 +/* ---------------------------------------------------------------------- */
13850 +static unsigned int au_fhsm_poll(struct file *file,
13851 + struct poll_table_struct *wait)
13853 + unsigned int mask;
13854 + struct au_sbinfo *sbinfo;
13855 + struct au_fhsm *fhsm;
13858 + sbinfo = file->private_data;
13859 + fhsm = &sbinfo->si_fhsm;
13860 + poll_wait(file, &fhsm->fhsm_wqh, wait);
13861 + if (atomic_read(&fhsm->fhsm_readable))
13862 + mask = POLLIN /* | POLLRDNORM */;
13864 + AuTraceErr((int)mask);
13868 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
13869 + struct aufs_stfs *stfs, __s16 brid)
13873 + err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
13875 + err = __put_user(brid, &stbr->brid);
13876 + if (unlikely(err))
13882 +static ssize_t au_fhsm_do_read(struct super_block *sb,
13883 + struct aufs_stbr __user *stbr, size_t count)
13887 + aufs_bindex_t bindex, bbot;
13888 + struct au_branch *br;
13889 + struct au_br_fhsm *bf;
13891 + /* except the bottom branch */
13894 + bbot = au_fhsm_bottom(sb);
13895 + for (bindex = 0; !err && bindex < bbot; bindex++) {
13896 + br = au_sbr(sb, bindex);
13897 + if (!au_br_fhsm(br->br_perm))
13900 + bf = br->br_fhsm;
13901 + mutex_lock(&bf->bf_lock);
13902 + if (bf->bf_readable) {
13904 + if (count >= sizeof(*stbr))
13905 + err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
13908 + bf->bf_readable = 0;
13909 + count -= sizeof(*stbr);
13913 + mutex_unlock(&bf->bf_lock);
13916 + err = sizeof(*stbr) * nstbr;
13921 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
13926 + aufs_bindex_t nfhsm, bindex, bbot;
13927 + struct au_sbinfo *sbinfo;
13928 + struct au_fhsm *fhsm;
13929 + struct au_branch *br;
13930 + struct super_block *sb;
13933 + sbinfo = file->private_data;
13934 + fhsm = &sbinfo->si_fhsm;
13936 + spin_lock_irq(&fhsm->fhsm_wqh.lock);
13937 + if (!atomic_read(&fhsm->fhsm_readable)) {
13938 + if (vfsub_file_flags(file) & O_NONBLOCK)
13941 + err = wait_event_interruptible_locked_irq
13943 + atomic_read(&fhsm->fhsm_readable));
13945 + spin_unlock_irq(&fhsm->fhsm_wqh.lock);
13946 + if (unlikely(err))
13949 + /* sb may already be dead */
13950 + au_rw_read_lock(&sbinfo->si_rwsem);
13951 + readable = atomic_read(&fhsm->fhsm_readable);
13952 + if (readable > 0) {
13953 + sb = sbinfo->si_sb;
13955 + /* exclude the bottom branch */
13957 + bbot = au_fhsm_bottom(sb);
13958 + for (bindex = 0; bindex < bbot; bindex++) {
13959 + br = au_sbr(sb, bindex);
13960 + if (au_br_fhsm(br->br_perm))
13964 + if (nfhsm * sizeof(struct aufs_stbr) <= count) {
13965 + atomic_set(&fhsm->fhsm_readable, 0);
13966 + err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
13970 + au_rw_read_unlock(&sbinfo->si_rwsem);
13978 +static int au_fhsm_release(struct inode *inode, struct file *file)
13980 + struct au_sbinfo *sbinfo;
13981 + struct au_fhsm *fhsm;
13983 + /* sb may already be dead */
13984 + sbinfo = file->private_data;
13985 + fhsm = &sbinfo->si_fhsm;
13986 + spin_lock(&fhsm->fhsm_spin);
13987 + fhsm->fhsm_pid = 0;
13988 + spin_unlock(&fhsm->fhsm_spin);
13989 + kobject_put(&sbinfo->si_kobj);
13994 +static const struct file_operations au_fhsm_fops = {
13995 + .owner = THIS_MODULE,
13996 + .llseek = noop_llseek,
13997 + .read = au_fhsm_read,
13998 + .poll = au_fhsm_poll,
13999 + .release = au_fhsm_release
14002 +int au_fhsm_fd(struct super_block *sb, int oflags)
14005 + struct au_sbinfo *sbinfo;
14006 + struct au_fhsm *fhsm;
14009 + if (unlikely(!capable(CAP_SYS_ADMIN)))
14013 + if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
14017 + sbinfo = au_sbi(sb);
14018 + fhsm = &sbinfo->si_fhsm;
14019 + spin_lock(&fhsm->fhsm_spin);
14020 + if (!fhsm->fhsm_pid)
14021 + fhsm->fhsm_pid = current->pid;
14024 + spin_unlock(&fhsm->fhsm_spin);
14025 + if (unlikely(err))
14028 + oflags |= O_RDONLY;
14029 + /* oflags |= FMODE_NONOTIFY; */
14030 + fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
14032 + if (unlikely(fd < 0))
14035 + /* succeed reglardless 'fhsm' status */
14036 + kobject_get(&sbinfo->si_kobj);
14037 + si_noflush_read_lock(sb);
14038 + if (au_ftest_si(sbinfo, FHSM))
14039 + au_fhsm_wrote_all(sb, /*force*/0);
14040 + si_read_unlock(sb);
14041 + goto out; /* success */
14044 + spin_lock(&fhsm->fhsm_spin);
14045 + fhsm->fhsm_pid = 0;
14046 + spin_unlock(&fhsm->fhsm_spin);
14052 +/* ---------------------------------------------------------------------- */
14054 +int au_fhsm_br_alloc(struct au_branch *br)
14059 + br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
14061 + au_br_fhsm_init(br->br_fhsm);
14068 +/* ---------------------------------------------------------------------- */
14070 +void au_fhsm_fin(struct super_block *sb)
14072 + au_fhsm_notify(sb, /*val*/-1);
14075 +void au_fhsm_init(struct au_sbinfo *sbinfo)
14077 + struct au_fhsm *fhsm;
14079 + fhsm = &sbinfo->si_fhsm;
14080 + spin_lock_init(&fhsm->fhsm_spin);
14081 + init_waitqueue_head(&fhsm->fhsm_wqh);
14082 + atomic_set(&fhsm->fhsm_readable, 0);
14083 + fhsm->fhsm_expire
14084 + = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
14085 + fhsm->fhsm_bottom = -1;
14088 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
14090 + sbinfo->si_fhsm.fhsm_expire
14091 + = msecs_to_jiffies(sec * MSEC_PER_SEC);
14094 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
14098 + if (!au_ftest_si(sbinfo, FHSM))
14101 + u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
14102 + if (u != AUFS_FHSM_CACHE_DEF_SEC)
14103 + seq_printf(seq, ",fhsm_sec=%u", u);
14105 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
14106 --- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
14107 +++ linux/fs/aufs/file.c 2018-01-29 07:56:20.056658502 +0100
14110 + * Copyright (C) 2005-2017 Junjiro R. Okajima
14112 + * This program, aufs is free software; you can redistribute it and/or modify
14113 + * it under the terms of the GNU General Public License as published by
14114 + * the Free Software Foundation; either version 2 of the License, or
14115 + * (at your option) any later version.
14117 + * This program is distributed in the hope that it will be useful,
14118 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14119 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14120 + * GNU General Public License for more details.
14122 + * You should have received a copy of the GNU General Public License
14123 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14127 + * handling file/dir, and address_space operation
14130 +#ifdef CONFIG_AUFS_DEBUG
14131 +#include <linux/migrate.h>
14133 +#include <linux/pagemap.h>
14136 +/* drop flags for writing */
14137 +unsigned int au_file_roflags(unsigned int flags)
14139 + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
14140 + flags |= O_RDONLY | O_NOATIME;
14144 +/* common functions to regular file and dir */
14145 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
14146 + struct file *file, int force_wr)
14148 + struct file *h_file;
14149 + struct dentry *h_dentry;
14150 + struct inode *h_inode;
14151 + struct super_block *sb;
14152 + struct au_branch *br;
14153 + struct path h_path;
14156 + /* a race condition can happen between open and unlink/rmdir */
14157 + h_file = ERR_PTR(-ENOENT);
14158 + h_dentry = au_h_dptr(dentry, bindex);
14159 + if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
14161 + h_inode = d_inode(h_dentry);
14162 + spin_lock(&h_dentry->d_lock);
14163 + err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
14164 + /* || !d_inode(dentry)->i_nlink */
14166 + spin_unlock(&h_dentry->d_lock);
14167 + if (unlikely(err))
14170 + sb = dentry->d_sb;
14171 + br = au_sbr(sb, bindex);
14172 + err = au_br_test_oflag(flags, br);
14173 + h_file = ERR_PTR(err);
14174 + if (unlikely(err))
14177 + /* drop flags for writing */
14178 + if (au_test_ro(sb, bindex, d_inode(dentry))) {
14179 + if (force_wr && !(flags & O_WRONLY))
14181 + flags = au_file_roflags(flags);
14183 + h_file = ERR_PTR(-EROFS);
14184 + flags = au_file_roflags(flags);
14185 + if (unlikely(vfsub_native_ro(h_inode)
14186 + || IS_APPEND(h_inode)))
14188 + flags &= ~O_ACCMODE;
14189 + flags |= O_WRONLY;
14192 + flags &= ~O_CREAT;
14194 + h_path.dentry = h_dentry;
14195 + h_path.mnt = au_br_mnt(br);
14196 + h_file = vfsub_dentry_open(&h_path, flags);
14197 + if (IS_ERR(h_file))
14200 + if (flags & __FMODE_EXEC) {
14201 + err = deny_write_access(h_file);
14202 + if (unlikely(err)) {
14204 + h_file = ERR_PTR(err);
14208 + fsnotify_open(h_file);
14209 + goto out; /* success */
14217 +static int au_cmoo(struct dentry *dentry)
14219 + int err, cmoo, matched;
14220 + unsigned int udba;
14221 + struct path h_path;
14222 + struct au_pin pin;
14223 + struct au_cp_generic cpg = {
14224 + .dentry = dentry,
14229 + .flags = AuCpup_DTIME | AuCpup_HOPEN
14231 + struct inode *delegated;
14232 + struct super_block *sb;
14233 + struct au_sbinfo *sbinfo;
14234 + struct au_fhsm *fhsm;
14236 + struct au_branch *br;
14237 + struct dentry *parent;
14238 + struct au_hinode *hdir;
14240 + DiMustWriteLock(dentry);
14241 + IiMustWriteLock(d_inode(dentry));
14244 + if (IS_ROOT(dentry))
14246 + cpg.bsrc = au_dbtop(dentry);
14250 + sb = dentry->d_sb;
14251 + sbinfo = au_sbi(sb);
14252 + fhsm = &sbinfo->si_fhsm;
14253 + pid = au_fhsm_pid(fhsm);
14256 + && (current->pid == pid
14257 + || rcu_dereference(current->real_parent)->pid == pid));
14258 + rcu_read_unlock();
14262 + br = au_sbr(sb, cpg.bsrc);
14263 + cmoo = au_br_cmoo(br->br_perm);
14266 + if (!d_is_reg(dentry))
14267 + cmoo &= AuBrAttr_COO_ALL;
14271 + parent = dget_parent(dentry);
14272 + di_write_lock_parent(parent);
14273 + err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
14275 + if (unlikely(err < 0)) {
14276 + err = 0; /* there is no upper writable branch */
14279 + AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
14281 + /* do not respect the coo attrib for the target branch */
14282 + err = au_cpup_dirs(dentry, cpg.bdst);
14283 + if (unlikely(err))
14286 + di_downgrade_lock(parent, AuLock_IR);
14287 + udba = au_opt_udba(sb);
14288 + err = au_pin(&pin, dentry, cpg.bdst, udba,
14289 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14290 + if (unlikely(err))
14293 + err = au_sio_cpup_simple(&cpg);
14295 + if (unlikely(err))
14297 + if (!(cmoo & AuBrWAttr_MOO))
14298 + goto out_parent; /* success */
14300 + err = au_pin(&pin, dentry, cpg.bsrc, udba,
14301 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14302 + if (unlikely(err))
14305 + h_path.mnt = au_br_mnt(br);
14306 + h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
14307 + hdir = au_hi(d_inode(parent), cpg.bsrc);
14308 + delegated = NULL;
14309 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
14311 + /* todo: keep h_dentry or not? */
14312 + if (unlikely(err == -EWOULDBLOCK)) {
14313 + pr_warn("cannot retry for NFSv4 delegation"
14314 + " for an internal unlink\n");
14317 + if (unlikely(err)) {
14318 + pr_err("unlink %pd after coo failed (%d), ignored\n",
14322 + goto out_parent; /* success */
14325 + di_downgrade_lock(parent, AuLock_IR);
14327 + di_read_unlock(parent, AuLock_IR);
14334 +int au_do_open(struct file *file, struct au_do_open_args *args)
14336 + int err, aopen = args->aopen;
14337 + struct dentry *dentry;
14338 + struct au_finfo *finfo;
14341 + err = au_finfo_init(file, args->fidir);
14344 + err = au_finfo_init(file, args->fidir);
14347 + if (unlikely(err))
14350 + dentry = file->f_path.dentry;
14351 + AuDebugOn(IS_ERR_OR_NULL(dentry));
14352 + di_write_lock_child(dentry);
14353 + err = au_cmoo(dentry);
14354 + di_downgrade_lock(dentry, AuLock_IR);
14357 + err = args->open(file, vfsub_file_flags(file), NULL);
14360 + err = args->open(file, vfsub_file_flags(file), NULL);
14364 + di_read_unlock(dentry, AuLock_IR);
14366 + finfo = au_fi(file);
14368 + finfo->fi_file = file;
14369 + au_hbl_add(&finfo->fi_hlist,
14370 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
14373 + fi_write_unlock(file);
14376 + fi_write_unlock(file);
14379 + if (unlikely(err)) {
14380 + finfo->fi_hdir = NULL;
14381 + au_finfo_fin(file);
14389 +int au_reopen_nondir(struct file *file)
14392 + aufs_bindex_t btop;
14393 + struct dentry *dentry;
14394 + struct file *h_file, *h_file_tmp;
14396 + dentry = file->f_path.dentry;
14397 + btop = au_dbtop(dentry);
14398 + h_file_tmp = NULL;
14399 + if (au_fbtop(file) == btop) {
14400 + h_file = au_hf_top(file);
14401 + if (file->f_mode == h_file->f_mode)
14402 + return 0; /* success */
14403 + h_file_tmp = h_file;
14404 + get_file(h_file_tmp);
14405 + au_set_h_fptr(file, btop, NULL);
14407 + AuDebugOn(au_fi(file)->fi_hdir);
14410 + * file exists on both of rw and ro
14411 + * open --> dbtop and fbtop are both 0
14412 + * prepend a branch as rw, "rw" become ro
14414 + * delete the top branch, "rw" becomes rw again
14415 + * --> dbtop is 1, fbtop is still 0
14416 + * write --> fbtop is 0 but dbtop is 1
14418 + /* AuDebugOn(au_fbtop(file) < btop); */
14420 + h_file = au_h_open(dentry, btop, vfsub_file_flags(file) & ~O_TRUNC,
14421 + file, /*force_wr*/0);
14422 + err = PTR_ERR(h_file);
14423 + if (IS_ERR(h_file)) {
14424 + if (h_file_tmp) {
14425 + au_sbr_get(dentry->d_sb, btop);
14426 + au_set_h_fptr(file, btop, h_file_tmp);
14427 + h_file_tmp = NULL;
14429 + goto out; /* todo: close all? */
14433 + au_set_fbtop(file, btop);
14434 + au_set_h_fptr(file, btop, h_file);
14435 + au_update_figen(file);
14436 + /* todo: necessary? */
14437 + /* file->f_ra = h_file->f_ra; */
14441 + fput(h_file_tmp);
14445 +/* ---------------------------------------------------------------------- */
14447 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
14448 + struct dentry *hi_wh)
14451 + aufs_bindex_t btop;
14452 + struct au_dinfo *dinfo;
14453 + struct dentry *h_dentry;
14454 + struct au_hdentry *hdp;
14456 + dinfo = au_di(file->f_path.dentry);
14457 + AuRwMustWriteLock(&dinfo->di_rwsem);
14459 + btop = dinfo->di_btop;
14460 + dinfo->di_btop = btgt;
14461 + hdp = au_hdentry(dinfo, btgt);
14462 + h_dentry = hdp->hd_dentry;
14463 + hdp->hd_dentry = hi_wh;
14464 + err = au_reopen_nondir(file);
14465 + hdp->hd_dentry = h_dentry;
14466 + dinfo->di_btop = btop;
14471 +static int au_ready_to_write_wh(struct file *file, loff_t len,
14472 + aufs_bindex_t bcpup, struct au_pin *pin)
14475 + struct inode *inode, *h_inode;
14476 + struct dentry *h_dentry, *hi_wh;
14477 + struct au_cp_generic cpg = {
14478 + .dentry = file->f_path.dentry,
14485 + au_update_dbtop(cpg.dentry);
14486 + inode = d_inode(cpg.dentry);
14488 + if (au_dbtop(cpg.dentry) <= bcpup
14489 + && au_dbbot(cpg.dentry) >= bcpup) {
14490 + h_dentry = au_h_dptr(cpg.dentry, bcpup);
14491 + if (h_dentry && d_is_positive(h_dentry))
14492 + h_inode = d_inode(h_dentry);
14494 + hi_wh = au_hi_wh(inode, bcpup);
14495 + if (!hi_wh && !h_inode)
14496 + err = au_sio_cpup_wh(&cpg, file);
14498 + /* already copied-up after unlink */
14499 + err = au_reopen_wh(file, bcpup, hi_wh);
14502 + && (inode->i_nlink > 1
14503 + || (inode->i_state & I_LINKABLE))
14504 + && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
14505 + au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
14511 + * prepare the @file for writing.
14513 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
14516 + aufs_bindex_t dbtop;
14517 + struct dentry *parent;
14518 + struct inode *inode;
14519 + struct super_block *sb;
14520 + struct file *h_file;
14521 + struct au_cp_generic cpg = {
14522 + .dentry = file->f_path.dentry,
14527 + .flags = AuCpup_DTIME
14530 + sb = cpg.dentry->d_sb;
14531 + inode = d_inode(cpg.dentry);
14532 + cpg.bsrc = au_fbtop(file);
14533 + err = au_test_ro(sb, cpg.bsrc, inode);
14534 + if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
14535 + err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
14540 + /* need to cpup or reopen */
14541 + parent = dget_parent(cpg.dentry);
14542 + di_write_lock_parent(parent);
14543 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14545 + if (unlikely(err < 0))
14549 + if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
14550 + err = au_cpup_dirs(cpg.dentry, cpg.bdst);
14551 + if (unlikely(err))
14555 + err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14556 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14557 + if (unlikely(err))
14560 + dbtop = au_dbtop(cpg.dentry);
14561 + if (dbtop <= cpg.bdst)
14562 + cpg.bsrc = cpg.bdst;
14564 + if (dbtop <= cpg.bdst /* just reopen */
14565 + || !d_unhashed(cpg.dentry) /* copyup and reopen */
14567 + h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
14568 + if (IS_ERR(h_file))
14569 + err = PTR_ERR(h_file);
14571 + di_downgrade_lock(parent, AuLock_IR);
14572 + if (dbtop > cpg.bdst)
14573 + err = au_sio_cpup_simple(&cpg);
14575 + err = au_reopen_nondir(file);
14576 + au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
14578 + } else { /* copyup as wh and reopen */
14580 + * since writable hfsplus branch is not supported,
14581 + * h_open_pre/post() are unnecessary.
14583 + err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
14584 + di_downgrade_lock(parent, AuLock_IR);
14588 + au_pin_set_parent_lflag(pin, /*lflag*/0);
14589 + goto out_dput; /* success */
14595 + di_downgrade_lock(parent, AuLock_IR);
14597 + di_read_unlock(parent, AuLock_IR);
14604 +/* ---------------------------------------------------------------------- */
14606 +int au_do_flush(struct file *file, fl_owner_t id,
14607 + int (*flush)(struct file *file, fl_owner_t id))
14610 + struct super_block *sb;
14611 + struct inode *inode;
14613 + inode = file_inode(file);
14614 + sb = inode->i_sb;
14615 + si_noflush_read_lock(sb);
14616 + fi_read_lock(file);
14617 + ii_read_lock_child(inode);
14619 + err = flush(file, id);
14620 + au_cpup_attr_timesizes(inode);
14622 + ii_read_unlock(inode);
14623 + fi_read_unlock(file);
14624 + si_read_unlock(sb);
14628 +/* ---------------------------------------------------------------------- */
14630 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
14633 + struct au_pin pin;
14634 + struct au_finfo *finfo;
14635 + struct dentry *parent, *hi_wh;
14636 + struct inode *inode;
14637 + struct super_block *sb;
14638 + struct au_cp_generic cpg = {
14639 + .dentry = file->f_path.dentry,
14644 + .flags = AuCpup_DTIME
14647 + FiMustWriteLock(file);
14650 + finfo = au_fi(file);
14651 + sb = cpg.dentry->d_sb;
14652 + inode = d_inode(cpg.dentry);
14653 + cpg.bdst = au_ibtop(inode);
14654 + if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
14657 + parent = dget_parent(cpg.dentry);
14658 + if (au_test_ro(sb, cpg.bdst, inode)) {
14659 + di_read_lock_parent(parent, !AuLock_IR);
14660 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
14662 + di_read_unlock(parent, !AuLock_IR);
14663 + if (unlikely(err < 0))
14668 + di_read_lock_parent(parent, AuLock_IR);
14669 + hi_wh = au_hi_wh(inode, cpg.bdst);
14670 + if (!S_ISDIR(inode->i_mode)
14671 + && au_opt_test(au_mntflags(sb), PLINK)
14672 + && au_plink_test(inode)
14673 + && !d_unhashed(cpg.dentry)
14674 + && cpg.bdst < au_dbtop(cpg.dentry)) {
14675 + err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
14676 + if (unlikely(err))
14679 + /* always superio. */
14680 + err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
14681 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
14683 + err = au_sio_cpup_simple(&cpg);
14686 + } else if (hi_wh) {
14687 + /* already copied-up after unlink */
14688 + err = au_reopen_wh(file, cpg.bdst, hi_wh);
14689 + *need_reopen = 0;
14693 + di_read_unlock(parent, AuLock_IR);
14700 +static void au_do_refresh_dir(struct file *file)
14702 + aufs_bindex_t bindex, bbot, new_bindex, brid;
14703 + struct au_hfile *p, tmp, *q;
14704 + struct au_finfo *finfo;
14705 + struct super_block *sb;
14706 + struct au_fidir *fidir;
14708 + FiMustWriteLock(file);
14710 + sb = file->f_path.dentry->d_sb;
14711 + finfo = au_fi(file);
14712 + fidir = finfo->fi_hdir;
14713 + AuDebugOn(!fidir);
14714 + p = fidir->fd_hfile + finfo->fi_btop;
14715 + brid = p->hf_br->br_id;
14716 + bbot = fidir->fd_bbot;
14717 + for (bindex = finfo->fi_btop; bindex <= bbot; bindex++, p++) {
14721 + new_bindex = au_br_index(sb, p->hf_br->br_id);
14722 + if (new_bindex == bindex)
14724 + if (new_bindex < 0) {
14725 + au_set_h_fptr(file, bindex, NULL);
14729 + /* swap two lower inode, and loop again */
14730 + q = fidir->fd_hfile + new_bindex;
14734 + if (tmp.hf_file) {
14740 + p = fidir->fd_hfile;
14741 + if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
14742 + bbot = au_sbbot(sb);
14743 + for (finfo->fi_btop = 0; finfo->fi_btop <= bbot;
14744 + finfo->fi_btop++, p++)
14745 + if (p->hf_file) {
14746 + if (file_inode(p->hf_file))
14748 + au_hfput(p, /*execed*/0);
14751 + bbot = au_br_index(sb, brid);
14752 + for (finfo->fi_btop = 0; finfo->fi_btop < bbot;
14753 + finfo->fi_btop++, p++)
14755 + au_hfput(p, /*execed*/0);
14756 + bbot = au_sbbot(sb);
14759 + p = fidir->fd_hfile + bbot;
14760 + for (fidir->fd_bbot = bbot; fidir->fd_bbot >= finfo->fi_btop;
14761 + fidir->fd_bbot--, p--)
14762 + if (p->hf_file) {
14763 + if (file_inode(p->hf_file))
14765 + au_hfput(p, /*execed*/0);
14767 + AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
14771 + * after branch manipulating, refresh the file.
14773 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
14775 + int err, need_reopen, nbr;
14776 + aufs_bindex_t bbot, bindex;
14777 + struct dentry *dentry;
14778 + struct super_block *sb;
14779 + struct au_finfo *finfo;
14780 + struct au_hfile *hfile;
14782 + dentry = file->f_path.dentry;
14783 + sb = dentry->d_sb;
14784 + nbr = au_sbbot(sb) + 1;
14785 + finfo = au_fi(file);
14786 + if (!finfo->fi_hdir) {
14787 + hfile = &finfo->fi_htop;
14788 + AuDebugOn(!hfile->hf_file);
14789 + bindex = au_br_index(sb, hfile->hf_br->br_id);
14790 + AuDebugOn(bindex < 0);
14791 + if (bindex != finfo->fi_btop)
14792 + au_set_fbtop(file, bindex);
14794 + err = au_fidir_realloc(finfo, nbr, /*may_shrink*/0);
14795 + if (unlikely(err))
14797 + au_do_refresh_dir(file);
14802 + if (!au_test_mmapped(file))
14803 + err = au_file_refresh_by_inode(file, &need_reopen);
14804 + if (finfo->fi_hdir)
14805 + /* harmless if err */
14806 + au_fidir_realloc(finfo, nbr, /*may_shrink*/1);
14807 + if (!err && need_reopen && !d_unlinked(dentry))
14808 + err = reopen(file);
14810 + au_update_figen(file);
14811 + goto out; /* success */
14814 + /* error, close all lower files */
14815 + if (finfo->fi_hdir) {
14816 + bbot = au_fbbot_dir(file);
14817 + for (bindex = au_fbtop(file); bindex <= bbot; bindex++)
14818 + au_set_h_fptr(file, bindex, NULL);
14825 +/* common function to regular file and dir */
14826 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
14827 + int wlock, unsigned int fi_lsc)
14830 + unsigned int sigen, figen;
14831 + aufs_bindex_t btop;
14832 + unsigned char pseudo_link;
14833 + struct dentry *dentry;
14834 + struct inode *inode;
14837 + dentry = file->f_path.dentry;
14838 + inode = d_inode(dentry);
14839 + sigen = au_sigen(dentry->d_sb);
14840 + fi_write_lock_nested(file, fi_lsc);
14841 + figen = au_figen(file);
14843 + di_write_lock_child(dentry);
14845 + di_write_lock_child2(dentry);
14846 + btop = au_dbtop(dentry);
14847 + pseudo_link = (btop != au_ibtop(inode));
14848 + if (sigen == figen && !pseudo_link && au_fbtop(file) == btop) {
14850 + di_downgrade_lock(dentry, AuLock_IR);
14851 + fi_downgrade_lock(file);
14853 + goto out; /* success */
14856 + AuDbg("sigen %d, figen %d\n", sigen, figen);
14857 + if (au_digen_test(dentry, sigen)) {
14858 + err = au_reval_dpath(dentry, sigen);
14859 + AuDebugOn(!err && au_digen_test(dentry, sigen));
14863 + err = refresh_file(file, reopen);
14866 + di_downgrade_lock(dentry, AuLock_IR);
14867 + fi_downgrade_lock(file);
14870 + di_write_unlock(dentry);
14871 + fi_write_unlock(file);
14878 +/* ---------------------------------------------------------------------- */
14880 +/* cf. aufs_nopage() */
14881 +/* for madvise(2) */
14882 +static int aufs_readpage(struct file *file __maybe_unused, struct page *page)
14884 + unlock_page(page);
14888 +/* it will never be called, but necessary to support O_DIRECT */
14889 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
14890 +{ BUG(); return 0; }
14892 +/* they will never be called. */
14893 +#ifdef CONFIG_AUFS_DEBUG
14894 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
14895 + loff_t pos, unsigned len, unsigned flags,
14896 + struct page **pagep, void **fsdata)
14897 +{ AuUnsupport(); return 0; }
14898 +static int aufs_write_end(struct file *file, struct address_space *mapping,
14899 + loff_t pos, unsigned len, unsigned copied,
14900 + struct page *page, void *fsdata)
14901 +{ AuUnsupport(); return 0; }
14902 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
14903 +{ AuUnsupport(); return 0; }
14905 +static int aufs_set_page_dirty(struct page *page)
14906 +{ AuUnsupport(); return 0; }
14907 +static void aufs_invalidatepage(struct page *page, unsigned int offset,
14908 + unsigned int length)
14909 +{ AuUnsupport(); }
14910 +static int aufs_releasepage(struct page *page, gfp_t gfp)
14911 +{ AuUnsupport(); return 0; }
14912 +#if 0 /* called by memory compaction regardless file */
14913 +static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
14914 + struct page *page, enum migrate_mode mode)
14915 +{ AuUnsupport(); return 0; }
14917 +static bool aufs_isolate_page(struct page *page, isolate_mode_t mode)
14918 +{ AuUnsupport(); return true; }
14919 +static void aufs_putback_page(struct page *page)
14920 +{ AuUnsupport(); }
14921 +static int aufs_launder_page(struct page *page)
14922 +{ AuUnsupport(); return 0; }
14923 +static int aufs_is_partially_uptodate(struct page *page,
14924 + unsigned long from,
14925 + unsigned long count)
14926 +{ AuUnsupport(); return 0; }
14927 +static void aufs_is_dirty_writeback(struct page *page, bool *dirty,
14929 +{ AuUnsupport(); }
14930 +static int aufs_error_remove_page(struct address_space *mapping,
14931 + struct page *page)
14932 +{ AuUnsupport(); return 0; }
14933 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
14935 +{ AuUnsupport(); return 0; }
14936 +static void aufs_swap_deactivate(struct file *file)
14937 +{ AuUnsupport(); }
14938 +#endif /* CONFIG_AUFS_DEBUG */
14940 +const struct address_space_operations aufs_aop = {
14941 + .readpage = aufs_readpage,
14942 + .direct_IO = aufs_direct_IO,
14943 +#ifdef CONFIG_AUFS_DEBUG
14944 + .writepage = aufs_writepage,
14945 + /* no writepages, because of writepage */
14946 + .set_page_dirty = aufs_set_page_dirty,
14947 + /* no readpages, because of readpage */
14948 + .write_begin = aufs_write_begin,
14949 + .write_end = aufs_write_end,
14950 + /* no bmap, no block device */
14951 + .invalidatepage = aufs_invalidatepage,
14952 + .releasepage = aufs_releasepage,
14953 + /* is fallback_migrate_page ok? */
14954 + /* .migratepage = aufs_migratepage, */
14955 + .isolate_page = aufs_isolate_page,
14956 + .putback_page = aufs_putback_page,
14957 + .launder_page = aufs_launder_page,
14958 + .is_partially_uptodate = aufs_is_partially_uptodate,
14959 + .is_dirty_writeback = aufs_is_dirty_writeback,
14960 + .error_remove_page = aufs_error_remove_page,
14961 + .swap_activate = aufs_swap_activate,
14962 + .swap_deactivate = aufs_swap_deactivate
14963 +#endif /* CONFIG_AUFS_DEBUG */
14965 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
14966 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
14967 +++ linux/fs/aufs/file.h 2018-01-29 07:56:20.056658502 +0100
14970 + * Copyright (C) 2005-2017 Junjiro R. Okajima
14972 + * This program, aufs is free software; you can redistribute it and/or modify
14973 + * it under the terms of the GNU General Public License as published by
14974 + * the Free Software Foundation; either version 2 of the License, or
14975 + * (at your option) any later version.
14977 + * This program is distributed in the hope that it will be useful,
14978 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14979 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14980 + * GNU General Public License for more details.
14982 + * You should have received a copy of the GNU General Public License
14983 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14987 + * file operations
14990 +#ifndef __AUFS_FILE_H__
14991 +#define __AUFS_FILE_H__
14995 +#include <linux/file.h>
14996 +#include <linux/fs.h>
14997 +#include <linux/mm_types.h>
14998 +#include <linux/poll.h>
14999 +#include "rwsem.h"
15003 + struct file *hf_file;
15004 + struct au_branch *hf_br;
15009 + aufs_bindex_t fd_bbot;
15010 + aufs_bindex_t fd_nent;
15011 + struct au_vdir *fd_vdir_cache;
15012 + struct au_hfile fd_hfile[];
15015 +static inline int au_fidir_sz(int nent)
15017 + AuDebugOn(nent < 0);
15018 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
15022 + atomic_t fi_generation;
15024 + struct au_rwsem fi_rwsem;
15025 + aufs_bindex_t fi_btop;
15027 + /* do not union them */
15028 + struct { /* for non-dir */
15029 + struct au_hfile fi_htop;
15030 + atomic_t fi_mmapped;
15032 + struct au_fidir *fi_hdir; /* for dir only */
15034 + struct hlist_bl_node fi_hlist;
15035 + struct file *fi_file; /* very ugly */
15036 +} ____cacheline_aligned_in_smp;
15038 +/* ---------------------------------------------------------------------- */
15041 +extern const struct address_space_operations aufs_aop;
15042 +unsigned int au_file_roflags(unsigned int flags);
15043 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
15044 + struct file *file, int force_wr);
15045 +struct au_do_open_args {
15047 + int (*open)(struct file *file, int flags,
15048 + struct file *h_file);
15049 + struct au_fidir *fidir;
15050 + struct file *h_file;
15052 +int au_do_open(struct file *file, struct au_do_open_args *args);
15053 +int au_reopen_nondir(struct file *file);
15055 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
15056 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
15057 + int wlock, unsigned int fi_lsc);
15058 +int au_do_flush(struct file *file, fl_owner_t id,
15059 + int (*flush)(struct file *file, fl_owner_t id));
15062 +#ifdef CONFIG_AUFS_POLL
15063 +unsigned int aufs_poll(struct file *file, poll_table *wait);
15066 +#ifdef CONFIG_AUFS_BR_HFSPLUS
15068 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
15070 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
15071 + struct file *h_file);
15073 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
15074 + aufs_bindex_t bindex, int force_wr)
15075 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
15076 + struct file *h_file);
15080 +extern const struct file_operations aufs_file_fop;
15081 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
15082 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
15083 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc);
15086 +void au_hfput(struct au_hfile *hf, int execed);
15087 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
15088 + struct file *h_file);
15090 +void au_update_figen(struct file *file);
15091 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
15092 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink);
15094 +void au_fi_init_once(void *_fi);
15095 +void au_finfo_fin(struct file *file);
15096 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
15099 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
15100 +#ifdef CONFIG_COMPAT
15101 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15102 + unsigned long arg);
15103 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15104 + unsigned long arg);
15107 +/* ---------------------------------------------------------------------- */
15109 +static inline struct au_finfo *au_fi(struct file *file)
15111 + return file->private_data;
15114 +/* ---------------------------------------------------------------------- */
15116 +#define fi_read_lock(f) au_rw_read_lock(&au_fi(f)->fi_rwsem)
15117 +#define fi_write_lock(f) au_rw_write_lock(&au_fi(f)->fi_rwsem)
15118 +#define fi_read_trylock(f) au_rw_read_trylock(&au_fi(f)->fi_rwsem)
15119 +#define fi_write_trylock(f) au_rw_write_trylock(&au_fi(f)->fi_rwsem)
15121 +#define fi_read_trylock_nested(f) \
15122 + au_rw_read_trylock_nested(&au_fi(f)->fi_rwsem)
15123 +#define fi_write_trylock_nested(f) \
15124 + au_rw_write_trylock_nested(&au_fi(f)->fi_rwsem)
15127 +#define fi_read_unlock(f) au_rw_read_unlock(&au_fi(f)->fi_rwsem)
15128 +#define fi_write_unlock(f) au_rw_write_unlock(&au_fi(f)->fi_rwsem)
15129 +#define fi_downgrade_lock(f) au_rw_dgrade_lock(&au_fi(f)->fi_rwsem)
15131 +/* lock subclass for finfo */
15137 +static inline void fi_read_lock_nested(struct file *f, unsigned int lsc)
15139 + au_rw_read_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15142 +static inline void fi_write_lock_nested(struct file *f, unsigned int lsc)
15144 + au_rw_write_lock_nested(&au_fi(f)->fi_rwsem, lsc);
15148 + * fi_read_lock_1, fi_write_lock_1,
15149 + * fi_read_lock_2, fi_write_lock_2
15151 +#define AuReadLockFunc(name) \
15152 +static inline void fi_read_lock_##name(struct file *f) \
15153 +{ fi_read_lock_nested(f, AuLsc_FI_##name); }
15155 +#define AuWriteLockFunc(name) \
15156 +static inline void fi_write_lock_##name(struct file *f) \
15157 +{ fi_write_lock_nested(f, AuLsc_FI_##name); }
15159 +#define AuRWLockFuncs(name) \
15160 + AuReadLockFunc(name) \
15161 + AuWriteLockFunc(name)
15166 +#undef AuReadLockFunc
15167 +#undef AuWriteLockFunc
15168 +#undef AuRWLockFuncs
15170 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
15171 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
15172 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
15174 +/* ---------------------------------------------------------------------- */
15176 +/* todo: hard/soft set? */
15177 +static inline aufs_bindex_t au_fbtop(struct file *file)
15179 + FiMustAnyLock(file);
15180 + return au_fi(file)->fi_btop;
15183 +static inline aufs_bindex_t au_fbbot_dir(struct file *file)
15185 + FiMustAnyLock(file);
15186 + AuDebugOn(!au_fi(file)->fi_hdir);
15187 + return au_fi(file)->fi_hdir->fd_bbot;
15190 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
15192 + FiMustAnyLock(file);
15193 + AuDebugOn(!au_fi(file)->fi_hdir);
15194 + return au_fi(file)->fi_hdir->fd_vdir_cache;
15197 +static inline void au_set_fbtop(struct file *file, aufs_bindex_t bindex)
15199 + FiMustWriteLock(file);
15200 + au_fi(file)->fi_btop = bindex;
15203 +static inline void au_set_fbbot_dir(struct file *file, aufs_bindex_t bindex)
15205 + FiMustWriteLock(file);
15206 + AuDebugOn(!au_fi(file)->fi_hdir);
15207 + au_fi(file)->fi_hdir->fd_bbot = bindex;
15210 +static inline void au_set_fvdir_cache(struct file *file,
15211 + struct au_vdir *vdir_cache)
15213 + FiMustWriteLock(file);
15214 + AuDebugOn(!au_fi(file)->fi_hdir);
15215 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
15218 +static inline struct file *au_hf_top(struct file *file)
15220 + FiMustAnyLock(file);
15221 + AuDebugOn(au_fi(file)->fi_hdir);
15222 + return au_fi(file)->fi_htop.hf_file;
15225 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
15227 + FiMustAnyLock(file);
15228 + AuDebugOn(!au_fi(file)->fi_hdir);
15229 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
15232 +/* todo: memory barrier? */
15233 +static inline unsigned int au_figen(struct file *f)
15235 + return atomic_read(&au_fi(f)->fi_generation);
15238 +static inline void au_set_mmapped(struct file *f)
15240 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
15242 + pr_warn("fi_mmapped wrapped around\n");
15243 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
15247 +static inline void au_unset_mmapped(struct file *f)
15249 + atomic_dec(&au_fi(f)->fi_mmapped);
15252 +static inline int au_test_mmapped(struct file *f)
15254 + return atomic_read(&au_fi(f)->fi_mmapped);
15257 +/* customize vma->vm_file */
15259 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
15260 + struct file *file)
15264 + f = vma->vm_file;
15266 + vma->vm_file = file;
15271 +#define AuDbgVmRegion(file, vma) do {} while (0)
15273 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15274 + struct file *file)
15276 + au_do_vm_file_reset(vma, file);
15279 +#define AuDbgVmRegion(file, vma) \
15280 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
15282 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
15283 + struct file *file)
15287 + au_do_vm_file_reset(vma, file);
15288 + f = vma->vm_region->vm_file;
15290 + vma->vm_region->vm_file = file;
15293 +#endif /* CONFIG_MMU */
15295 +/* handle vma->vm_prfile */
15296 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
15297 + struct file *file)
15300 + vma->vm_prfile = file;
15301 +#ifndef CONFIG_MMU
15303 + vma->vm_region->vm_prfile = file;
15307 +#endif /* __KERNEL__ */
15308 +#endif /* __AUFS_FILE_H__ */
15309 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
15310 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
15311 +++ linux/fs/aufs/finfo.c 2018-01-29 07:56:20.056658502 +0100
15314 + * Copyright (C) 2005-2017 Junjiro R. Okajima
15316 + * This program, aufs is free software; you can redistribute it and/or modify
15317 + * it under the terms of the GNU General Public License as published by
15318 + * the Free Software Foundation; either version 2 of the License, or
15319 + * (at your option) any later version.
15321 + * This program is distributed in the hope that it will be useful,
15322 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15323 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15324 + * GNU General Public License for more details.
15326 + * You should have received a copy of the GNU General Public License
15327 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15331 + * file private data
15336 +void au_hfput(struct au_hfile *hf, int execed)
15339 + allow_write_access(hf->hf_file);
15340 + fput(hf->hf_file);
15341 + hf->hf_file = NULL;
15342 + au_br_put(hf->hf_br);
15343 + hf->hf_br = NULL;
15346 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
15348 + struct au_finfo *finfo = au_fi(file);
15349 + struct au_hfile *hf;
15350 + struct au_fidir *fidir;
15352 + fidir = finfo->fi_hdir;
15354 + AuDebugOn(finfo->fi_btop != bindex);
15355 + hf = &finfo->fi_htop;
15357 + hf = fidir->fd_hfile + bindex;
15359 + if (hf && hf->hf_file)
15360 + au_hfput(hf, vfsub_file_execed(file));
15362 + FiMustWriteLock(file);
15363 + AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
15364 + hf->hf_file = val;
15365 + hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
15369 +void au_update_figen(struct file *file)
15371 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
15372 + /* smp_mb(); */ /* atomic_set */
15375 +/* ---------------------------------------------------------------------- */
15377 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
15379 + struct au_fidir *fidir;
15382 + nbr = au_sbbot(sb) + 1;
15384 + nbr = 2; /* initial allocate for 2 branches */
15385 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
15387 + fidir->fd_bbot = -1;
15388 + fidir->fd_nent = nbr;
15394 +int au_fidir_realloc(struct au_finfo *finfo, int nbr, int may_shrink)
15397 + struct au_fidir *fidir, *p;
15399 + AuRwMustWriteLock(&finfo->fi_rwsem);
15400 + fidir = finfo->fi_hdir;
15401 + AuDebugOn(!fidir);
15404 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
15405 + GFP_NOFS, may_shrink);
15407 + p->fd_nent = nbr;
15408 + finfo->fi_hdir = p;
15415 +/* ---------------------------------------------------------------------- */
15417 +void au_finfo_fin(struct file *file)
15419 + struct au_finfo *finfo;
15421 + au_nfiles_dec(file->f_path.dentry->d_sb);
15423 + finfo = au_fi(file);
15424 + AuDebugOn(finfo->fi_hdir);
15425 + AuRwDestroy(&finfo->fi_rwsem);
15426 + au_cache_free_finfo(finfo);
15429 +void au_fi_init_once(void *_finfo)
15431 + struct au_finfo *finfo = _finfo;
15433 + au_rw_init(&finfo->fi_rwsem);
15436 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
15439 + struct au_finfo *finfo;
15440 + struct dentry *dentry;
15443 + dentry = file->f_path.dentry;
15444 + finfo = au_cache_alloc_finfo();
15445 + if (unlikely(!finfo))
15449 + au_nfiles_inc(dentry->d_sb);
15450 + au_rw_write_lock(&finfo->fi_rwsem);
15451 + finfo->fi_btop = -1;
15452 + finfo->fi_hdir = fidir;
15453 + atomic_set(&finfo->fi_generation, au_digen(dentry));
15454 + /* smp_mb(); */ /* atomic_set */
15456 + file->private_data = finfo;
15461 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
15462 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
15463 +++ linux/fs/aufs/f_op.c 2018-01-29 07:56:20.053325069 +0100
15466 + * Copyright (C) 2005-2017 Junjiro R. Okajima
15468 + * This program, aufs is free software; you can redistribute it and/or modify
15469 + * it under the terms of the GNU General Public License as published by
15470 + * the Free Software Foundation; either version 2 of the License, or
15471 + * (at your option) any later version.
15473 + * This program is distributed in the hope that it will be useful,
15474 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15475 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15476 + * GNU General Public License for more details.
15478 + * You should have received a copy of the GNU General Public License
15479 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15483 + * file and vm operations
15486 +#include <linux/aio.h>
15487 +#include <linux/fs_stack.h>
15488 +#include <linux/mman.h>
15489 +#include <linux/security.h>
15492 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
15495 + aufs_bindex_t bindex;
15496 + struct dentry *dentry, *h_dentry;
15497 + struct au_finfo *finfo;
15498 + struct inode *h_inode;
15500 + FiMustWriteLock(file);
15503 + dentry = file->f_path.dentry;
15504 + AuDebugOn(IS_ERR_OR_NULL(dentry));
15505 + finfo = au_fi(file);
15506 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
15507 + atomic_set(&finfo->fi_mmapped, 0);
15508 + bindex = au_dbtop(dentry);
15510 + h_dentry = au_h_dptr(dentry, bindex);
15511 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15512 + if (unlikely(err))
15514 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
15516 + h_dentry = h_file->f_path.dentry;
15517 + err = vfsub_test_mntns(file->f_path.mnt, h_dentry->d_sb);
15518 + if (unlikely(err))
15520 + get_file(h_file);
15522 + if (IS_ERR(h_file))
15523 + err = PTR_ERR(h_file);
15525 + if ((flags & __O_TMPFILE)
15526 + && !(flags & O_EXCL)) {
15527 + h_inode = file_inode(h_file);
15528 + spin_lock(&h_inode->i_lock);
15529 + h_inode->i_state |= I_LINKABLE;
15530 + spin_unlock(&h_inode->i_lock);
15532 + au_set_fbtop(file, bindex);
15533 + au_set_h_fptr(file, bindex, h_file);
15534 + au_update_figen(file);
15535 + /* todo: necessary? */
15536 + /* file->f_ra = h_file->f_ra; */
15543 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
15544 + struct file *file)
15547 + struct super_block *sb;
15548 + struct au_do_open_args args = {
15549 + .open = au_do_open_nondir
15552 + AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
15553 + file, vfsub_file_flags(file), file->f_mode);
15555 + sb = file->f_path.dentry->d_sb;
15556 + si_read_lock(sb, AuLock_FLUSH);
15557 + err = au_do_open(file, &args);
15558 + si_read_unlock(sb);
15562 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
15564 + struct au_finfo *finfo;
15565 + aufs_bindex_t bindex;
15567 + finfo = au_fi(file);
15568 + au_hbl_del(&finfo->fi_hlist,
15569 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
15570 + bindex = finfo->fi_btop;
15572 + au_set_h_fptr(file, bindex, NULL);
15574 + au_finfo_fin(file);
15578 +/* ---------------------------------------------------------------------- */
15580 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
15583 + struct file *h_file;
15586 + h_file = au_hf_top(file);
15588 + err = vfsub_flush(h_file, id);
15592 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
15594 + return au_do_flush(file, id, au_do_flush_nondir);
15597 +/* ---------------------------------------------------------------------- */
15599 + * read and write functions acquire [fdi]_rwsem once, but release before
15600 + * mmap_sem. This is because to stop a race condition between mmap(2).
15601 + * Releasing these aufs-rwsem should be safe, no branch-mamagement (by keeping
15602 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
15603 + * read functions after [fdi]_rwsem are released, but it should be harmless.
15606 +/* Callers should call au_read_post() or fput() in the end */
15607 +struct file *au_read_pre(struct file *file, int keep_fi, unsigned int lsc)
15609 + struct file *h_file;
15612 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0, lsc);
15614 + di_read_unlock(file->f_path.dentry, AuLock_IR);
15615 + h_file = au_hf_top(file);
15616 + get_file(h_file);
15618 + fi_read_unlock(file);
15620 + h_file = ERR_PTR(err);
15625 +static void au_read_post(struct inode *inode, struct file *h_file)
15627 + /* update without lock, I don't think it a problem */
15628 + fsstack_copy_attr_atime(inode, file_inode(h_file));
15632 +struct au_write_pre {
15634 + unsigned int lsc;
15638 + aufs_bindex_t btop;
15642 + * return with iinfo is write-locked
15643 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
15646 +static struct file *au_write_pre(struct file *file, int do_ready,
15647 + struct au_write_pre *wpre)
15649 + struct file *h_file;
15650 + struct dentry *dentry;
15652 + unsigned int lsc;
15653 + struct au_pin pin;
15658 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1, lsc);
15659 + h_file = ERR_PTR(err);
15660 + if (unlikely(err))
15663 + dentry = file->f_path.dentry;
15665 + err = au_ready_to_write(file, -1, &pin);
15666 + if (unlikely(err)) {
15667 + h_file = ERR_PTR(err);
15668 + di_write_unlock(dentry);
15673 + di_downgrade_lock(dentry, /*flags*/0);
15675 + wpre->btop = au_fbtop(file);
15676 + h_file = au_hf_top(file);
15677 + get_file(h_file);
15679 + wpre->blks = file_inode(h_file)->i_blocks;
15682 + di_read_unlock(dentry, /*flags*/0);
15685 + fi_write_unlock(file);
15690 +static void au_write_post(struct inode *inode, struct file *h_file,
15691 + struct au_write_pre *wpre, ssize_t written)
15693 + struct inode *h_inode;
15695 + au_cpup_attr_timesizes(inode);
15696 + AuDebugOn(au_ibtop(inode) != wpre->btop);
15697 + h_inode = file_inode(h_file);
15698 + inode->i_mode = h_inode->i_mode;
15699 + ii_write_unlock(inode);
15700 + /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
15702 + au_fhsm_wrote(inode->i_sb, wpre->btop,
15703 + /*force*/h_inode->i_blocks > wpre->blks);
15707 +static ssize_t aufs_read(struct file *file, char __user *buf, size_t count,
15711 + struct inode *inode;
15712 + struct file *h_file;
15713 + struct super_block *sb;
15715 + inode = file_inode(file);
15716 + sb = inode->i_sb;
15717 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15719 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15720 + err = PTR_ERR(h_file);
15721 + if (IS_ERR(h_file))
15724 + /* filedata may be obsoleted by concurrent copyup, but no problem */
15725 + err = vfsub_read_u(h_file, buf, count, ppos);
15726 + /* todo: necessary? */
15727 + /* file->f_ra = h_file->f_ra; */
15728 + au_read_post(inode, h_file);
15731 + si_read_unlock(sb);
15736 + * todo: very ugly
15737 + * it locks both of i_mutex and si_rwsem for read in safe.
15738 + * if the plink maintenance mode continues forever (that is the problem),
15739 + * may loop forever.
15741 +static void au_mtx_and_read_lock(struct inode *inode)
15744 + struct super_block *sb = inode->i_sb;
15747 + inode_lock(inode);
15748 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
15751 + inode_unlock(inode);
15752 + si_read_lock(sb, AuLock_NOPLMW);
15753 + si_read_unlock(sb);
15757 +static ssize_t aufs_write(struct file *file, const char __user *ubuf,
15758 + size_t count, loff_t *ppos)
15761 + struct au_write_pre wpre;
15762 + struct inode *inode;
15763 + struct file *h_file;
15764 + char __user *buf = (char __user *)ubuf;
15766 + inode = file_inode(file);
15767 + au_mtx_and_read_lock(inode);
15770 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15771 + err = PTR_ERR(h_file);
15772 + if (IS_ERR(h_file))
15775 + err = vfsub_write_u(h_file, buf, count, ppos);
15776 + au_write_post(inode, h_file, &wpre, err);
15779 + si_read_unlock(inode->i_sb);
15780 + inode_unlock(inode);
15784 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
15785 + struct iov_iter *iov_iter)
15788 + struct file *file;
15789 + ssize_t (*iter)(struct kiocb *, struct iov_iter *);
15791 + err = security_file_permission(h_file, rw);
15792 + if (unlikely(err))
15797 + if (rw == MAY_READ)
15798 + iter = h_file->f_op->read_iter;
15799 + else if (rw == MAY_WRITE)
15800 + iter = h_file->f_op->write_iter;
15802 + file = kio->ki_filp;
15803 + kio->ki_filp = h_file;
15806 + err = iter(kio, iov_iter);
15809 + /* currently there is no such fs */
15811 + kio->ki_filp = file;
15817 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15820 + struct file *file, *h_file;
15821 + struct inode *inode;
15822 + struct super_block *sb;
15824 + file = kio->ki_filp;
15825 + inode = file_inode(file);
15826 + sb = inode->i_sb;
15827 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15829 + h_file = au_read_pre(file, /*keep_fi*/1, /*lsc*/0);
15830 + err = PTR_ERR(h_file);
15831 + if (IS_ERR(h_file))
15834 + if (au_test_loopback_kthread()) {
15835 + au_warn_loopback(h_file->f_path.dentry->d_sb);
15836 + if (file->f_mapping != h_file->f_mapping) {
15837 + file->f_mapping = h_file->f_mapping;
15838 + smp_mb(); /* unnecessary? */
15841 + fi_read_unlock(file);
15843 + err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
15844 + /* todo: necessary? */
15845 + /* file->f_ra = h_file->f_ra; */
15846 + au_read_post(inode, h_file);
15849 + si_read_unlock(sb);
15853 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
15856 + struct au_write_pre wpre;
15857 + struct inode *inode;
15858 + struct file *file, *h_file;
15860 + file = kio->ki_filp;
15861 + inode = file_inode(file);
15862 + au_mtx_and_read_lock(inode);
15865 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15866 + err = PTR_ERR(h_file);
15867 + if (IS_ERR(h_file))
15870 + err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
15871 + au_write_post(inode, h_file, &wpre, err);
15874 + si_read_unlock(inode->i_sb);
15875 + inode_unlock(inode);
15879 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
15880 + struct pipe_inode_info *pipe, size_t len,
15881 + unsigned int flags)
15884 + struct file *h_file;
15885 + struct inode *inode;
15886 + struct super_block *sb;
15888 + inode = file_inode(file);
15889 + sb = inode->i_sb;
15890 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
15892 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
15893 + err = PTR_ERR(h_file);
15894 + if (IS_ERR(h_file))
15897 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
15898 + /* todo: necessasry? */
15899 + /* file->f_ra = h_file->f_ra; */
15900 + au_read_post(inode, h_file);
15903 + si_read_unlock(sb);
15908 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
15909 + size_t len, unsigned int flags)
15912 + struct au_write_pre wpre;
15913 + struct inode *inode;
15914 + struct file *h_file;
15916 + inode = file_inode(file);
15917 + au_mtx_and_read_lock(inode);
15920 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15921 + err = PTR_ERR(h_file);
15922 + if (IS_ERR(h_file))
15925 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
15926 + au_write_post(inode, h_file, &wpre, err);
15929 + si_read_unlock(inode->i_sb);
15930 + inode_unlock(inode);
15934 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
15938 + struct au_write_pre wpre;
15939 + struct inode *inode;
15940 + struct file *h_file;
15942 + inode = file_inode(file);
15943 + au_mtx_and_read_lock(inode);
15946 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
15947 + err = PTR_ERR(h_file);
15948 + if (IS_ERR(h_file))
15952 + err = vfs_fallocate(h_file, mode, offset, len);
15954 + au_write_post(inode, h_file, &wpre, /*written*/1);
15957 + si_read_unlock(inode->i_sb);
15958 + inode_unlock(inode);
15962 +static ssize_t aufs_copy_file_range(struct file *src, loff_t src_pos,
15963 + struct file *dst, loff_t dst_pos,
15964 + size_t len, unsigned int flags)
15967 + struct au_write_pre wpre;
15968 + enum { SRC, DST };
15970 + struct inode *inode;
15971 + struct file *h_file;
15972 + struct super_block *h_sb;
15974 +#define a_src a[SRC]
15975 +#define a_dst a[DST]
15978 + a_src.inode = file_inode(src);
15979 + if (unlikely(!S_ISREG(a_src.inode->i_mode)))
15981 + a_dst.inode = file_inode(dst);
15982 + if (unlikely(!S_ISREG(a_dst.inode->i_mode)))
15985 + au_mtx_and_read_lock(a_dst.inode);
15987 + * in order to match the order in di_write_lock2_{child,parent}(),
15988 + * use f_path.dentry for this comparision.
15990 + if (src->f_path.dentry < dst->f_path.dentry) {
15991 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_1);
15992 + err = PTR_ERR(a_src.h_file);
15993 + if (IS_ERR(a_src.h_file))
15996 + wpre.lsc = AuLsc_FI_2;
15997 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
15998 + err = PTR_ERR(a_dst.h_file);
15999 + if (IS_ERR(a_dst.h_file)) {
16000 + au_read_post(a_src.inode, a_src.h_file);
16004 + wpre.lsc = AuLsc_FI_1;
16005 + a_dst.h_file = au_write_pre(dst, /*do_ready*/1, &wpre);
16006 + err = PTR_ERR(a_dst.h_file);
16007 + if (IS_ERR(a_dst.h_file))
16010 + a_src.h_file = au_read_pre(src, /*keep_fi*/1, AuLsc_FI_2);
16011 + err = PTR_ERR(a_src.h_file);
16012 + if (IS_ERR(a_src.h_file)) {
16013 + au_write_post(a_dst.inode, a_dst.h_file, &wpre,
16020 + a_src.h_sb = file_inode(a_src.h_file)->i_sb;
16021 + a_dst.h_sb = file_inode(a_dst.h_file)->i_sb;
16022 + if (unlikely(a_src.h_sb != a_dst.h_sb)) {
16028 + err = vfsub_copy_file_range(a_src.h_file, src_pos, a_dst.h_file,
16029 + dst_pos, len, flags);
16032 + au_write_post(a_dst.inode, a_dst.h_file, &wpre, err);
16033 + fi_read_unlock(src);
16034 + au_read_post(a_src.inode, a_src.h_file);
16036 + si_read_unlock(a_dst.inode->i_sb);
16037 + inode_unlock(a_dst.inode);
16044 +/* ---------------------------------------------------------------------- */
16047 + * The locking order around current->mmap_sem.
16048 + * - in most and regular cases
16049 + * file I/O syscall -- aufs_read() or something
16050 + * -- si_rwsem for read -- mmap_sem
16051 + * (Note that [fdi]i_rwsem are released before mmap_sem).
16053 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
16054 + * This AB-BA order is definitly bad, but is not a problem since "si_rwsem for
16055 + * read" allows muliple processes to acquire it and [fdi]i_rwsem are not held in
16056 + * file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
16057 + * It means that when aufs acquires si_rwsem for write, the process should never
16058 + * acquire mmap_sem.
16060 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
16061 + * problem either since any directory is not able to be mmap-ed.
16062 + * The similar scenario is applied to aufs_readlink() too.
16065 +#if 0 /* stop calling security_file_mmap() */
16066 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
16067 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
16069 +static unsigned long au_arch_prot_conv(unsigned long flags)
16071 + /* currently ppc64 only */
16072 +#ifdef CONFIG_PPC64
16073 + /* cf. linux/arch/powerpc/include/asm/mman.h */
16074 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
16075 + return AuConv_VM_PROT(flags, SAO);
16077 + AuDebugOn(arch_calc_vm_prot_bits(-1));
16082 +static unsigned long au_prot_conv(unsigned long flags)
16084 + return AuConv_VM_PROT(flags, READ)
16085 + | AuConv_VM_PROT(flags, WRITE)
16086 + | AuConv_VM_PROT(flags, EXEC)
16087 + | au_arch_prot_conv(flags);
16090 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
16091 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
16093 +static unsigned long au_flag_conv(unsigned long flags)
16095 + return AuConv_VM_MAP(flags, GROWSDOWN)
16096 + | AuConv_VM_MAP(flags, DENYWRITE)
16097 + | AuConv_VM_MAP(flags, LOCKED);
16101 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
16104 + const unsigned char wlock
16105 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
16106 + struct super_block *sb;
16107 + struct file *h_file;
16108 + struct inode *inode;
16110 + AuDbgVmRegion(file, vma);
16112 + inode = file_inode(file);
16113 + sb = inode->i_sb;
16115 + si_read_lock(sb, AuLock_NOPLMW);
16117 + h_file = au_write_pre(file, wlock, /*wpre*/NULL);
16119 + err = PTR_ERR(h_file);
16120 + if (IS_ERR(h_file))
16124 + au_set_mmapped(file);
16125 + au_vm_file_reset(vma, h_file);
16127 + * we cannot call security_mmap_file() here since it may acquire
16128 + * mmap_sem or i_mutex.
16130 + * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
16131 + * au_flag_conv(vma->vm_flags));
16134 + err = call_mmap(h_file, vma);
16136 + au_vm_prfile_set(vma, file);
16137 + fsstack_copy_attr_atime(inode, file_inode(h_file));
16138 + goto out_fput; /* success */
16140 + au_unset_mmapped(file);
16141 + au_vm_file_reset(vma, file);
16145 + ii_write_unlock(inode);
16150 + si_read_unlock(sb);
16156 +/* ---------------------------------------------------------------------- */
16158 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
16162 + struct au_write_pre wpre;
16163 + struct inode *inode;
16164 + struct file *h_file;
16166 + err = 0; /* -EBADF; */ /* posix? */
16167 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
16170 + inode = file_inode(file);
16171 + au_mtx_and_read_lock(inode);
16174 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
16175 + err = PTR_ERR(h_file);
16176 + if (IS_ERR(h_file))
16179 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
16180 + au_write_post(inode, h_file, &wpre, /*written*/0);
16183 + si_read_unlock(inode->i_sb);
16184 + inode_unlock(inode);
16189 +static int aufs_fasync(int fd, struct file *file, int flag)
16192 + struct file *h_file;
16193 + struct super_block *sb;
16195 + sb = file->f_path.dentry->d_sb;
16196 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16198 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16199 + err = PTR_ERR(h_file);
16200 + if (IS_ERR(h_file))
16203 + if (h_file->f_op->fasync)
16204 + err = h_file->f_op->fasync(fd, h_file, flag);
16205 + fput(h_file); /* instead of au_read_post() */
16208 + si_read_unlock(sb);
16212 +static int aufs_setfl(struct file *file, unsigned long arg)
16215 + struct file *h_file;
16216 + struct super_block *sb;
16218 + sb = file->f_path.dentry->d_sb;
16219 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
16221 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
16222 + err = PTR_ERR(h_file);
16223 + if (IS_ERR(h_file))
16226 + /* stop calling h_file->fasync */
16227 + arg |= vfsub_file_flags(file) & FASYNC;
16228 + err = setfl(/*unused fd*/-1, h_file, arg);
16229 + fput(h_file); /* instead of au_read_post() */
16232 + si_read_unlock(sb);
16236 +/* ---------------------------------------------------------------------- */
16238 +/* no one supports this operation, currently */
16240 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
16241 + size_t len, loff_t *pos, int more)
16246 +/* ---------------------------------------------------------------------- */
16248 +const struct file_operations aufs_file_fop = {
16249 + .owner = THIS_MODULE,
16251 + .llseek = default_llseek,
16253 + .read = aufs_read,
16254 + .write = aufs_write,
16255 + .read_iter = aufs_read_iter,
16256 + .write_iter = aufs_write_iter,
16258 +#ifdef CONFIG_AUFS_POLL
16259 + .poll = aufs_poll,
16261 + .unlocked_ioctl = aufs_ioctl_nondir,
16262 +#ifdef CONFIG_COMPAT
16263 + .compat_ioctl = aufs_compat_ioctl_nondir,
16265 + .mmap = aufs_mmap,
16266 + .open = aufs_open_nondir,
16267 + .flush = aufs_flush_nondir,
16268 + .release = aufs_release_nondir,
16269 + .fsync = aufs_fsync_nondir,
16270 + .fasync = aufs_fasync,
16271 + /* .sendpage = aufs_sendpage, */
16272 + .setfl = aufs_setfl,
16273 + .splice_write = aufs_splice_write,
16274 + .splice_read = aufs_splice_read,
16276 + .aio_splice_write = aufs_aio_splice_write,
16277 + .aio_splice_read = aufs_aio_splice_read,
16279 + .fallocate = aufs_fallocate,
16280 + .copy_file_range = aufs_copy_file_range
16282 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
16283 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
16284 +++ linux/fs/aufs/fstype.h 2018-01-29 07:56:20.056658502 +0100
16287 + * Copyright (C) 2005-2017 Junjiro R. Okajima
16289 + * This program, aufs is free software; you can redistribute it and/or modify
16290 + * it under the terms of the GNU General Public License as published by
16291 + * the Free Software Foundation; either version 2 of the License, or
16292 + * (at your option) any later version.
16294 + * This program is distributed in the hope that it will be useful,
16295 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16296 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16297 + * GNU General Public License for more details.
16299 + * You should have received a copy of the GNU General Public License
16300 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16304 + * judging filesystem type
16307 +#ifndef __AUFS_FSTYPE_H__
16308 +#define __AUFS_FSTYPE_H__
16312 +#include <linux/fs.h>
16313 +#include <linux/magic.h>
16314 +#include <linux/nfs_fs.h>
16315 +#include <linux/romfs_fs.h>
16317 +static inline int au_test_aufs(struct super_block *sb)
16319 + return sb->s_magic == AUFS_SUPER_MAGIC;
16322 +static inline const char *au_sbtype(struct super_block *sb)
16324 + return sb->s_type->name;
16327 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
16329 +#if IS_ENABLED(CONFIG_ISO9660_FS)
16330 + return sb->s_magic == ISOFS_SUPER_MAGIC;
16336 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
16338 +#if IS_ENABLED(CONFIG_ROMFS_FS)
16339 + return sb->s_magic == ROMFS_MAGIC;
16345 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
16347 +#if IS_ENABLED(CONFIG_CRAMFS)
16348 + return sb->s_magic == CRAMFS_MAGIC;
16353 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
16355 +#if IS_ENABLED(CONFIG_NFS_FS)
16356 + return sb->s_magic == NFS_SUPER_MAGIC;
16362 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
16364 +#if IS_ENABLED(CONFIG_FUSE_FS)
16365 + return sb->s_magic == FUSE_SUPER_MAGIC;
16371 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
16373 +#if IS_ENABLED(CONFIG_XFS_FS)
16374 + return sb->s_magic == XFS_SB_MAGIC;
16380 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
16382 +#ifdef CONFIG_TMPFS
16383 + return sb->s_magic == TMPFS_MAGIC;
16389 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
16391 +#if IS_ENABLED(CONFIG_ECRYPT_FS)
16392 + return !strcmp(au_sbtype(sb), "ecryptfs");
16398 +static inline int au_test_ramfs(struct super_block *sb)
16400 + return sb->s_magic == RAMFS_MAGIC;
16403 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
16405 +#if IS_ENABLED(CONFIG_UBIFS_FS)
16406 + return sb->s_magic == UBIFS_SUPER_MAGIC;
16412 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
16414 +#ifdef CONFIG_PROC_FS
16415 + return sb->s_magic == PROC_SUPER_MAGIC;
16421 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
16423 +#ifdef CONFIG_SYSFS
16424 + return sb->s_magic == SYSFS_MAGIC;
16430 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
16432 +#if IS_ENABLED(CONFIG_CONFIGFS_FS)
16433 + return sb->s_magic == CONFIGFS_MAGIC;
16439 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
16441 +#if IS_ENABLED(CONFIG_MINIX_FS)
16442 + return sb->s_magic == MINIX3_SUPER_MAGIC
16443 + || sb->s_magic == MINIX2_SUPER_MAGIC
16444 + || sb->s_magic == MINIX2_SUPER_MAGIC2
16445 + || sb->s_magic == MINIX_SUPER_MAGIC
16446 + || sb->s_magic == MINIX_SUPER_MAGIC2;
16452 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
16454 +#if IS_ENABLED(CONFIG_FAT_FS)
16455 + return sb->s_magic == MSDOS_SUPER_MAGIC;
16461 +static inline int au_test_msdos(struct super_block *sb)
16463 + return au_test_fat(sb);
16466 +static inline int au_test_vfat(struct super_block *sb)
16468 + return au_test_fat(sb);
16471 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
16473 +#ifdef CONFIG_SECURITYFS
16474 + return sb->s_magic == SECURITYFS_MAGIC;
16480 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
16482 +#if IS_ENABLED(CONFIG_SQUASHFS)
16483 + return sb->s_magic == SQUASHFS_MAGIC;
16489 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
16491 +#if IS_ENABLED(CONFIG_BTRFS_FS)
16492 + return sb->s_magic == BTRFS_SUPER_MAGIC;
16498 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
16500 +#if IS_ENABLED(CONFIG_XENFS)
16501 + return sb->s_magic == XENFS_SUPER_MAGIC;
16507 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
16509 +#ifdef CONFIG_DEBUG_FS
16510 + return sb->s_magic == DEBUGFS_MAGIC;
16516 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
16518 +#if IS_ENABLED(CONFIG_NILFS)
16519 + return sb->s_magic == NILFS_SUPER_MAGIC;
16525 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
16527 +#if IS_ENABLED(CONFIG_HFSPLUS_FS)
16528 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
16534 +/* ---------------------------------------------------------------------- */
16536 + * they can't be an aufs branch.
16538 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
16541 +#ifndef CONFIG_AUFS_BR_RAMFS
16542 + au_test_ramfs(sb) ||
16544 + au_test_procfs(sb)
16545 + || au_test_sysfs(sb)
16546 + || au_test_configfs(sb)
16547 + || au_test_debugfs(sb)
16548 + || au_test_securityfs(sb)
16549 + || au_test_xenfs(sb)
16550 + || au_test_ecryptfs(sb)
16551 + /* || !strcmp(au_sbtype(sb), "unionfs") */
16552 + || au_test_aufs(sb); /* will be supported in next version */
16555 +static inline int au_test_fs_remote(struct super_block *sb)
16557 + return !au_test_tmpfs(sb)
16558 +#ifdef CONFIG_AUFS_BR_RAMFS
16559 + && !au_test_ramfs(sb)
16561 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
16564 +/* ---------------------------------------------------------------------- */
16567 + * Note: these functions (below) are created after reading ->getattr() in all
16568 + * filesystems under linux/fs. it means we have to do so in every update...
16572 + * some filesystems require getattr to refresh the inode attributes before
16574 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
16575 + * and leave the work for d_revalidate()
16577 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
16579 + return au_test_nfs(sb)
16580 + || au_test_fuse(sb)
16581 + /* || au_test_btrfs(sb) */ /* untested */
16586 + * filesystems which don't maintain i_size or i_blocks.
16588 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
16590 + return au_test_xfs(sb)
16591 + || au_test_btrfs(sb)
16592 + || au_test_ubifs(sb)
16593 + || au_test_hfsplus(sb) /* maintained, but incorrect */
16594 + /* || au_test_minix(sb) */ /* untested */
16599 + * filesystems which don't store the correct value in some of their inode
16602 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
16604 + return au_test_fs_bad_iattr_size(sb)
16605 + || au_test_fat(sb)
16606 + || au_test_msdos(sb)
16607 + || au_test_vfat(sb);
16610 +/* they don't check i_nlink in link(2) */
16611 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
16613 + return au_test_tmpfs(sb)
16614 +#ifdef CONFIG_AUFS_BR_RAMFS
16615 + || au_test_ramfs(sb)
16617 + || au_test_ubifs(sb)
16618 + || au_test_hfsplus(sb);
16622 + * filesystems which sets S_NOATIME and S_NOCMTIME.
16624 +static inline int au_test_fs_notime(struct super_block *sb)
16626 + return au_test_nfs(sb)
16627 + || au_test_fuse(sb)
16628 + || au_test_ubifs(sb)
16632 +/* temporary support for i#1 in cramfs */
16633 +static inline int au_test_fs_unique_ino(struct inode *inode)
16635 + if (au_test_cramfs(inode->i_sb))
16636 + return inode->i_ino != 1;
16640 +/* ---------------------------------------------------------------------- */
16643 + * the filesystem where the xino files placed must support i/o after unlink and
16644 + * maintain i_size and i_blocks.
16646 +static inline int au_test_fs_bad_xino(struct super_block *sb)
16648 + return au_test_fs_remote(sb)
16649 + || au_test_fs_bad_iattr_size(sb)
16650 + /* don't want unnecessary work for xino */
16651 + || au_test_aufs(sb)
16652 + || au_test_ecryptfs(sb)
16653 + || au_test_nilfs(sb);
16656 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
16658 + return au_test_tmpfs(sb)
16659 + || au_test_ramfs(sb);
16663 + * test if the @sb is real-readonly.
16665 +static inline int au_test_fs_rr(struct super_block *sb)
16667 + return au_test_squashfs(sb)
16668 + || au_test_iso9660(sb)
16669 + || au_test_cramfs(sb)
16670 + || au_test_romfs(sb);
16674 + * test if the @inode is nfs with 'noacl' option
16675 + * NFS always sets SB_POSIXACL regardless its mount option 'noacl.'
16677 +static inline int au_test_nfs_noacl(struct inode *inode)
16679 + return au_test_nfs(inode->i_sb)
16680 + /* && IS_POSIXACL(inode) */
16681 + && !nfs_server_capable(inode, NFS_CAP_ACLS);
16684 +#endif /* __KERNEL__ */
16685 +#endif /* __AUFS_FSTYPE_H__ */
16686 diff -urN /usr/share/empty/fs/aufs/hbl.h linux/fs/aufs/hbl.h
16687 --- /usr/share/empty/fs/aufs/hbl.h 1970-01-01 01:00:00.000000000 +0100
16688 +++ linux/fs/aufs/hbl.h 2018-01-29 07:56:20.056658502 +0100
16691 + * Copyright (C) 2017 Junjiro R. Okajima
16693 + * This program, aufs is free software; you can redistribute it and/or modify
16694 + * it under the terms of the GNU General Public License as published by
16695 + * the Free Software Foundation; either version 2 of the License, or
16696 + * (at your option) any later version.
16698 + * This program is distributed in the hope that it will be useful,
16699 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16700 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16701 + * GNU General Public License for more details.
16703 + * You should have received a copy of the GNU General Public License
16704 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16708 + * helpers for hlist_bl.h
16711 +#ifndef __AUFS_HBL_H__
16712 +#define __AUFS_HBL_H__
16716 +#include <linux/list_bl.h>
16718 +static inline void au_hbl_add(struct hlist_bl_node *node,
16719 + struct hlist_bl_head *hbl)
16721 + hlist_bl_lock(hbl);
16722 + hlist_bl_add_head(node, hbl);
16723 + hlist_bl_unlock(hbl);
16726 +static inline void au_hbl_del(struct hlist_bl_node *node,
16727 + struct hlist_bl_head *hbl)
16729 + hlist_bl_lock(hbl);
16730 + hlist_bl_del(node);
16731 + hlist_bl_unlock(hbl);
16734 +#define au_hbl_for_each(pos, head) \
16735 + for (pos = hlist_bl_first(head); \
16739 +static inline unsigned long au_hbl_count(struct hlist_bl_head *hbl)
16741 + unsigned long cnt;
16742 + struct hlist_bl_node *pos;
16745 + hlist_bl_lock(hbl);
16746 + au_hbl_for_each(pos, hbl)
16748 + hlist_bl_unlock(hbl);
16752 +#endif /* __KERNEL__ */
16753 +#endif /* __AUFS_HBL_H__ */
16754 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
16755 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
16756 +++ linux/fs/aufs/hfsnotify.c 2018-01-29 07:56:20.056658502 +0100
16759 + * Copyright (C) 2005-2017 Junjiro R. Okajima
16761 + * This program, aufs is free software; you can redistribute it and/or modify
16762 + * it under the terms of the GNU General Public License as published by
16763 + * the Free Software Foundation; either version 2 of the License, or
16764 + * (at your option) any later version.
16766 + * This program is distributed in the hope that it will be useful,
16767 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16768 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16769 + * GNU General Public License for more details.
16771 + * You should have received a copy of the GNU General Public License
16772 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16776 + * fsnotify for the lower directories
16781 +/* FS_IN_IGNORED is unnecessary */
16782 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
16783 + | FS_CREATE | FS_EVENT_ON_CHILD);
16784 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
16785 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
16787 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
16789 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
16791 + /* AuDbg("here\n"); */
16792 + au_cache_free_hnotify(hn);
16793 + smp_mb__before_atomic(); /* for atomic64_dec */
16794 + if (atomic64_dec_and_test(&au_hfsn_ifree))
16795 + wake_up(&au_hfsn_wq);
16798 +static int au_hfsn_alloc(struct au_hinode *hinode)
16801 + struct au_hnotify *hn;
16802 + struct super_block *sb;
16803 + struct au_branch *br;
16804 + struct fsnotify_mark *mark;
16805 + aufs_bindex_t bindex;
16807 + hn = hinode->hi_notify;
16808 + sb = hn->hn_aufs_inode->i_sb;
16809 + bindex = au_br_index(sb, hinode->hi_id);
16810 + br = au_sbr(sb, bindex);
16811 + AuDebugOn(!br->br_hfsn);
16813 + mark = &hn->hn_mark;
16814 + fsnotify_init_mark(mark, br->br_hfsn->hfsn_group);
16815 + mark->mask = AuHfsnMask;
16817 + * by udba rename or rmdir, aufs assign a new inode to the known
16818 + * h_inode, so specify 1 to allow dups.
16821 + err = fsnotify_add_mark(mark, hinode->hi_inode, /*mnt*/NULL,
16822 + /*allow_dups*/1);
16828 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
16830 + struct fsnotify_mark *mark;
16831 + unsigned long long ull;
16832 + struct fsnotify_group *group;
16834 + ull = atomic64_inc_return(&au_hfsn_ifree);
16837 + mark = &hn->hn_mark;
16838 + spin_lock(&mark->lock);
16839 + group = mark->group;
16840 + fsnotify_get_group(group);
16841 + spin_unlock(&mark->lock);
16843 + fsnotify_destroy_mark(mark, group);
16844 + fsnotify_put_mark(mark);
16845 + fsnotify_put_group(group);
16848 + /* free hn by myself */
16852 +/* ---------------------------------------------------------------------- */
16854 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
16856 + struct fsnotify_mark *mark;
16858 + mark = &hinode->hi_notify->hn_mark;
16859 + spin_lock(&mark->lock);
16861 + AuDebugOn(mark->mask & AuHfsnMask);
16862 + mark->mask |= AuHfsnMask;
16864 + AuDebugOn(!(mark->mask & AuHfsnMask));
16865 + mark->mask &= ~AuHfsnMask;
16867 + spin_unlock(&mark->lock);
16868 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
16871 +/* ---------------------------------------------------------------------- */
16873 +/* #define AuDbgHnotify */
16874 +#ifdef AuDbgHnotify
16875 +static char *au_hfsn_name(u32 mask)
16877 +#ifdef CONFIG_AUFS_DEBUG
16878 +#define test_ret(flag) \
16880 + if (mask & flag) \
16883 + test_ret(FS_ACCESS);
16884 + test_ret(FS_MODIFY);
16885 + test_ret(FS_ATTRIB);
16886 + test_ret(FS_CLOSE_WRITE);
16887 + test_ret(FS_CLOSE_NOWRITE);
16888 + test_ret(FS_OPEN);
16889 + test_ret(FS_MOVED_FROM);
16890 + test_ret(FS_MOVED_TO);
16891 + test_ret(FS_CREATE);
16892 + test_ret(FS_DELETE);
16893 + test_ret(FS_DELETE_SELF);
16894 + test_ret(FS_MOVE_SELF);
16895 + test_ret(FS_UNMOUNT);
16896 + test_ret(FS_Q_OVERFLOW);
16897 + test_ret(FS_IN_IGNORED);
16898 + test_ret(FS_ISDIR);
16899 + test_ret(FS_IN_ONESHOT);
16900 + test_ret(FS_EVENT_ON_CHILD);
16909 +/* ---------------------------------------------------------------------- */
16911 +static void au_hfsn_free_group(struct fsnotify_group *group)
16913 + struct au_br_hfsnotify *hfsn = group->private;
16915 + /* AuDbg("here\n"); */
16919 +static int au_hfsn_handle_event(struct fsnotify_group *group,
16920 + struct inode *inode,
16921 + struct fsnotify_mark *inode_mark,
16922 + struct fsnotify_mark *vfsmount_mark,
16923 + u32 mask, const void *data, int data_type,
16924 + const unsigned char *file_name, u32 cookie,
16925 + struct fsnotify_iter_info *iter_info)
16928 + struct au_hnotify *hnotify;
16929 + struct inode *h_dir, *h_inode;
16930 + struct qstr h_child_qstr = QSTR_INIT(file_name, strlen(file_name));
16932 + AuDebugOn(data_type != FSNOTIFY_EVENT_INODE);
16935 + /* if FS_UNMOUNT happens, there must be another bug */
16936 + AuDebugOn(mask & FS_UNMOUNT);
16937 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
16942 +#ifdef AuDbgHnotify
16944 + if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
16945 + || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
16946 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
16947 + h_dir->i_ino, mask, au_hfsn_name(mask),
16948 + AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
16949 + /* WARN_ON(1); */
16954 + AuDebugOn(!inode_mark);
16955 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
16956 + err = au_hnotify(h_dir, hnotify, mask, &h_child_qstr, h_inode);
16962 +static struct fsnotify_ops au_hfsn_ops = {
16963 + .handle_event = au_hfsn_handle_event,
16964 + .free_group_priv = au_hfsn_free_group,
16965 + .free_mark = au_hfsn_free_mark
16968 +/* ---------------------------------------------------------------------- */
16970 +static void au_hfsn_fin_br(struct au_branch *br)
16972 + struct au_br_hfsnotify *hfsn;
16974 + hfsn = br->br_hfsn;
16977 + fsnotify_put_group(hfsn->hfsn_group);
16982 +static int au_hfsn_init_br(struct au_branch *br, int perm)
16985 + struct fsnotify_group *group;
16986 + struct au_br_hfsnotify *hfsn;
16989 + br->br_hfsn = NULL;
16990 + if (!au_br_hnotifyable(perm))
16994 + hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
16995 + if (unlikely(!hfsn))
16999 + group = fsnotify_alloc_group(&au_hfsn_ops);
17000 + if (IS_ERR(group)) {
17001 + err = PTR_ERR(group);
17002 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
17006 + group->private = hfsn;
17007 + hfsn->hfsn_group = group;
17008 + br->br_hfsn = hfsn;
17009 + goto out; /* success */
17017 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
17022 + if (!br->br_hfsn)
17023 + err = au_hfsn_init_br(br, perm);
17028 +/* ---------------------------------------------------------------------- */
17030 +static void au_hfsn_fin(void)
17032 + AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
17033 + wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
17036 +const struct au_hnotify_op au_hnotify_op = {
17037 + .ctl = au_hfsn_ctl,
17038 + .alloc = au_hfsn_alloc,
17039 + .free = au_hfsn_free,
17041 + .fin = au_hfsn_fin,
17043 + .reset_br = au_hfsn_reset_br,
17044 + .fin_br = au_hfsn_fin_br,
17045 + .init_br = au_hfsn_init_br
17047 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
17048 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
17049 +++ linux/fs/aufs/hfsplus.c 2018-01-29 07:56:20.056658502 +0100
17052 + * Copyright (C) 2010-2017 Junjiro R. Okajima
17054 + * This program, aufs is free software; you can redistribute it and/or modify
17055 + * it under the terms of the GNU General Public License as published by
17056 + * the Free Software Foundation; either version 2 of the License, or
17057 + * (at your option) any later version.
17059 + * This program is distributed in the hope that it will be useful,
17060 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17061 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17062 + * GNU General Public License for more details.
17064 + * You should have received a copy of the GNU General Public License
17065 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17069 + * special support for filesystems which aqucires an inode mutex
17070 + * at final closing a file, eg, hfsplus.
17072 + * This trick is very simple and stupid, just to open the file before really
17073 + * neceeary open to tell hfsplus that this is not the final closing.
17074 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
17075 + * and au_h_open_post() after releasing it.
17080 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
17083 + struct file *h_file;
17084 + struct dentry *h_dentry;
17086 + h_dentry = au_h_dptr(dentry, bindex);
17087 + AuDebugOn(!h_dentry);
17088 + AuDebugOn(d_is_negative(h_dentry));
17091 + if (au_test_hfsplus(h_dentry->d_sb)
17092 + && d_is_reg(h_dentry))
17093 + h_file = au_h_open(dentry, bindex,
17094 + O_RDONLY | O_NOATIME | O_LARGEFILE,
17095 + /*file*/NULL, force_wr);
17099 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
17100 + struct file *h_file)
17104 + au_sbr_put(dentry->d_sb, bindex);
17107 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
17108 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
17109 +++ linux/fs/aufs/hnotify.c 2018-01-29 07:56:20.056658502 +0100
17112 + * Copyright (C) 2005-2017 Junjiro R. Okajima
17114 + * This program, aufs is free software; you can redistribute it and/or modify
17115 + * it under the terms of the GNU General Public License as published by
17116 + * the Free Software Foundation; either version 2 of the License, or
17117 + * (at your option) any later version.
17119 + * This program is distributed in the hope that it will be useful,
17120 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17121 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17122 + * GNU General Public License for more details.
17124 + * You should have received a copy of the GNU General Public License
17125 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17129 + * abstraction to notify the direct changes on lower directories
17134 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
17137 + struct au_hnotify *hn;
17140 + hn = au_cache_alloc_hnotify();
17142 + hn->hn_aufs_inode = inode;
17143 + hinode->hi_notify = hn;
17144 + err = au_hnotify_op.alloc(hinode);
17146 + if (unlikely(err)) {
17147 + hinode->hi_notify = NULL;
17148 + au_cache_free_hnotify(hn);
17150 + * The upper dir was removed by udba, but the same named
17151 + * dir left. In this case, aufs assignes a new inode
17152 + * number and set the monitor again.
17153 + * For the lower dir, the old monitnor is still left.
17155 + if (err == -EEXIST)
17164 +void au_hn_free(struct au_hinode *hinode)
17166 + struct au_hnotify *hn;
17168 + hn = hinode->hi_notify;
17170 + hinode->hi_notify = NULL;
17171 + if (au_hnotify_op.free(hinode, hn))
17172 + au_cache_free_hnotify(hn);
17176 +/* ---------------------------------------------------------------------- */
17178 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
17180 + if (hinode->hi_notify)
17181 + au_hnotify_op.ctl(hinode, do_set);
17184 +void au_hn_reset(struct inode *inode, unsigned int flags)
17186 + aufs_bindex_t bindex, bbot;
17187 + struct inode *hi;
17188 + struct dentry *iwhdentry;
17190 + bbot = au_ibbot(inode);
17191 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
17192 + hi = au_h_iptr(inode, bindex);
17196 + /* inode_lock_nested(hi, AuLsc_I_CHILD); */
17197 + iwhdentry = au_hi_wh(inode, bindex);
17201 + au_set_h_iptr(inode, bindex, NULL, 0);
17202 + au_set_h_iptr(inode, bindex, au_igrab(hi),
17203 + flags & ~AuHi_XINO);
17206 + /* inode_unlock(hi); */
17210 +/* ---------------------------------------------------------------------- */
17212 +static int hn_xino(struct inode *inode, struct inode *h_inode)
17215 + aufs_bindex_t bindex, bbot, bfound, btop;
17216 + struct inode *h_i;
17219 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
17220 + pr_warn("branch root dir was changed\n");
17225 + bbot = au_ibbot(inode);
17226 + btop = au_ibtop(inode);
17227 +#if 0 /* reserved for future use */
17228 + if (bindex == bbot) {
17229 + /* keep this ino in rename case */
17233 + for (bindex = btop; bindex <= bbot; bindex++)
17234 + if (au_h_iptr(inode, bindex) == h_inode) {
17241 + for (bindex = btop; bindex <= bbot; bindex++) {
17242 + h_i = au_h_iptr(inode, bindex);
17246 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
17247 + /* ignore this error */
17248 + /* bad action? */
17251 + /* children inode number will be broken */
17258 +static int hn_gen_tree(struct dentry *dentry)
17260 + int err, i, j, ndentry;
17261 + struct au_dcsub_pages dpages;
17262 + struct au_dpage *dpage;
17263 + struct dentry **dentries;
17265 + err = au_dpages_init(&dpages, GFP_NOFS);
17266 + if (unlikely(err))
17268 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
17269 + if (unlikely(err))
17272 + for (i = 0; i < dpages.ndpage; i++) {
17273 + dpage = dpages.dpages + i;
17274 + dentries = dpage->dentries;
17275 + ndentry = dpage->ndentry;
17276 + for (j = 0; j < ndentry; j++) {
17277 + struct dentry *d;
17284 + if (d_really_is_positive(d))
17285 + /* todo: reset children xino?
17286 + cached children only? */
17287 + au_iigen_dec(d_inode(d));
17292 + au_dpages_free(&dpages);
17295 + /* discard children */
17296 + dentry_unhash(dentry);
17304 + * return 0 if processed.
17306 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
17307 + const unsigned int isdir)
17310 + struct dentry *d;
17311 + struct qstr *dname;
17314 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
17315 + pr_warn("branch root dir was changed\n");
17321 + AuDebugOn(!name);
17322 + au_iigen_dec(inode);
17323 + spin_lock(&inode->i_lock);
17324 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
17325 + spin_lock(&d->d_lock);
17326 + dname = &d->d_name;
17327 + if (dname->len != nlen
17328 + && memcmp(dname->name, name, nlen)) {
17329 + spin_unlock(&d->d_lock);
17334 + spin_unlock(&d->d_lock);
17337 + spin_unlock(&inode->i_lock);
17339 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
17340 + d = d_find_any_alias(inode);
17342 + au_iigen_dec(inode);
17346 + spin_lock(&d->d_lock);
17347 + dname = &d->d_name;
17348 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
17349 + spin_unlock(&d->d_lock);
17350 + err = hn_gen_tree(d);
17351 + spin_lock(&d->d_lock);
17353 + spin_unlock(&d->d_lock);
17362 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
17366 + if (IS_ROOT(dentry)) {
17367 + pr_warn("branch root dir was changed\n");
17373 + au_digen_dec(dentry);
17374 + if (d_really_is_positive(dentry))
17375 + au_iigen_dec(d_inode(dentry));
17377 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
17378 + if (d_really_is_positive(dentry))
17379 + err = hn_gen_tree(dentry);
17386 +/* ---------------------------------------------------------------------- */
17388 +/* hnotify job flags */
17389 +#define AuHnJob_XINO0 1
17390 +#define AuHnJob_GEN (1 << 1)
17391 +#define AuHnJob_DIRENT (1 << 2)
17392 +#define AuHnJob_ISDIR (1 << 3)
17393 +#define AuHnJob_TRYXINO0 (1 << 4)
17394 +#define AuHnJob_MNTPNT (1 << 5)
17395 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
17396 +#define au_fset_hnjob(flags, name) \
17397 + do { (flags) |= AuHnJob_##name; } while (0)
17398 +#define au_fclr_hnjob(flags, name) \
17399 + do { (flags) &= ~AuHnJob_##name; } while (0)
17407 +struct au_hnotify_args {
17408 + struct inode *h_dir, *dir, *h_child_inode;
17410 + unsigned int flags[AuHnLast];
17411 + unsigned int h_child_nlen;
17412 + char h_child_name[];
17415 +struct hn_job_args {
17416 + unsigned int flags;
17417 + struct inode *inode, *h_inode, *dir, *h_dir;
17418 + struct dentry *dentry;
17423 +static int hn_job(struct hn_job_args *a)
17425 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
17429 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
17430 + hn_xino(a->inode, a->h_inode); /* ignore this error */
17432 + if (au_ftest_hnjob(a->flags, TRYXINO0)
17435 + vfsub_inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
17436 + if (!a->h_inode->i_nlink
17437 + && !(a->h_inode->i_state & I_LINKABLE))
17438 + hn_xino(a->inode, a->h_inode); /* ignore this error */
17439 + inode_unlock_shared(a->h_inode);
17442 + /* make the generation obsolete */
17443 + if (au_ftest_hnjob(a->flags, GEN)) {
17446 + e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
17448 + if (e && a->dentry)
17449 + hn_gen_by_name(a->dentry, isdir);
17450 + /* ignore this error */
17453 + /* make dir entries obsolete */
17454 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
17455 + struct au_vdir *vdir;
17457 + vdir = au_ivdir(a->inode);
17459 + vdir->vd_jiffy = 0;
17460 + /* IMustLock(a->inode); */
17461 + /* a->inode->i_version++; */
17464 + /* can do nothing but warn */
17465 + if (au_ftest_hnjob(a->flags, MNTPNT)
17467 + && d_mountpoint(a->dentry))
17468 + pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
17473 +/* ---------------------------------------------------------------------- */
17475 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
17476 + struct inode *dir)
17478 + struct dentry *dentry, *d, *parent;
17479 + struct qstr *dname;
17481 + parent = d_find_any_alias(dir);
17486 + spin_lock(&parent->d_lock);
17487 + list_for_each_entry(d, &parent->d_subdirs, d_child) {
17488 + /* AuDbg("%pd\n", d); */
17489 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
17490 + dname = &d->d_name;
17491 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
17492 + goto cont_unlock;
17496 + goto cont_unlock;
17497 + if (au_dcount(d) > 0) {
17498 + dentry = dget_dlock(d);
17499 + spin_unlock(&d->d_lock);
17504 + spin_unlock(&d->d_lock);
17506 + spin_unlock(&parent->d_lock);
17510 + di_write_lock_child(dentry);
17515 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
17516 + aufs_bindex_t bindex, ino_t h_ino)
17518 + struct inode *inode;
17523 + err = au_xino_read(sb, bindex, h_ino, &ino);
17525 + inode = ilookup(sb, ino);
17529 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
17530 + pr_warn("wrong root branch\n");
17536 + ii_write_lock_child(inode);
17542 +static void au_hn_bh(void *_args)
17544 + struct au_hnotify_args *a = _args;
17545 + struct super_block *sb;
17546 + aufs_bindex_t bindex, bbot, bfound;
17547 + unsigned char xino, try_iput;
17549 + struct inode *inode;
17551 + struct hn_job_args args;
17552 + struct dentry *dentry;
17553 + struct au_sbinfo *sbinfo;
17555 + AuDebugOn(!_args);
17556 + AuDebugOn(!a->h_dir);
17557 + AuDebugOn(!a->dir);
17558 + AuDebugOn(!a->mask);
17559 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
17560 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
17561 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
17566 + * do not lock a->dir->i_mutex here
17567 + * because of d_revalidate() may cause a deadlock.
17569 + sb = a->dir->i_sb;
17571 + sbinfo = au_sbi(sb);
17572 + AuDebugOn(!sbinfo);
17573 + si_write_lock(sb, AuLock_NOPLMW);
17575 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
17576 + switch (a->mask & FS_EVENTS_POSS_ON_CHILD) {
17577 + case FS_MOVED_FROM:
17578 + case FS_MOVED_TO:
17579 + AuWarn1("DIRREN with UDBA may not work correctly "
17580 + "for the direct rename(2)\n");
17583 + ii_read_lock_parent(a->dir);
17585 + bbot = au_ibbot(a->dir);
17586 + for (bindex = au_ibtop(a->dir); bindex <= bbot; bindex++)
17587 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
17591 + ii_read_unlock(a->dir);
17592 + if (unlikely(bfound < 0))
17595 + xino = !!au_opt_test(au_mntflags(sb), XINO);
17597 + if (a->h_child_inode)
17598 + h_ino = a->h_child_inode->i_ino;
17600 + if (a->h_child_nlen
17601 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
17602 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
17603 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
17606 + if (dentry && d_really_is_positive(dentry))
17607 + inode = d_inode(dentry);
17608 + if (xino && !inode && h_ino
17609 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
17610 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
17611 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
17612 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
17616 + args.flags = a->flags[AuHn_CHILD];
17617 + args.dentry = dentry;
17618 + args.inode = inode;
17619 + args.h_inode = a->h_child_inode;
17620 + args.dir = a->dir;
17621 + args.h_dir = a->h_dir;
17622 + args.h_name = a->h_child_name;
17623 + args.h_nlen = a->h_child_nlen;
17624 + err = hn_job(&args);
17626 + if (au_di(dentry))
17627 + di_write_unlock(dentry);
17630 + if (inode && try_iput) {
17631 + ii_write_unlock(inode);
17635 + ii_write_lock_parent(a->dir);
17636 + args.flags = a->flags[AuHn_PARENT];
17637 + args.dentry = NULL;
17638 + args.inode = a->dir;
17639 + args.h_inode = a->h_dir;
17641 + args.h_dir = NULL;
17642 + args.h_name = NULL;
17644 + err = hn_job(&args);
17645 + ii_write_unlock(a->dir);
17648 + iput(a->h_child_inode);
17651 + si_write_unlock(sb);
17652 + au_nwt_done(&sbinfo->si_nowait);
17656 +/* ---------------------------------------------------------------------- */
17658 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
17659 + struct qstr *h_child_qstr, struct inode *h_child_inode)
17662 + unsigned int flags[AuHnLast], f;
17663 + unsigned char isdir, isroot, wh;
17664 + struct inode *dir;
17665 + struct au_hnotify_args *args;
17666 + char *p, *h_child_name;
17669 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
17670 + dir = igrab(hnotify->hn_aufs_inode);
17674 + isroot = (dir->i_ino == AUFS_ROOT_INO);
17676 + h_child_name = (void *)h_child_qstr->name;
17677 + len = h_child_qstr->len;
17678 + if (h_child_name) {
17679 + if (len > AUFS_WH_PFX_LEN
17680 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
17681 + h_child_name += AUFS_WH_PFX_LEN;
17682 + len -= AUFS_WH_PFX_LEN;
17688 + if (h_child_inode)
17689 + isdir = !!S_ISDIR(h_child_inode->i_mode);
17690 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
17691 + flags[AuHn_CHILD] = 0;
17693 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
17694 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
17695 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
17696 + switch (mask & FS_EVENTS_POSS_ON_CHILD) {
17697 + case FS_MOVED_FROM:
17698 + case FS_MOVED_TO:
17699 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
17700 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
17703 + AuDebugOn(!h_child_name);
17708 + * aufs never be able to get this child inode.
17709 + * revalidation should be in d_revalidate()
17710 + * by checking i_nlink, i_generation or d_unhashed().
17712 + AuDebugOn(!h_child_name);
17713 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
17714 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
17722 + h_child_inode = NULL;
17725 + /* iput() and kfree() will be called in au_hnotify() */
17726 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
17727 + if (unlikely(!args)) {
17728 + AuErr1("no memory\n");
17732 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
17733 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
17734 + args->mask = mask;
17736 + args->h_dir = igrab(h_dir);
17737 + if (h_child_inode)
17738 + h_child_inode = igrab(h_child_inode); /* can be NULL */
17739 + args->h_child_inode = h_child_inode;
17740 + args->h_child_nlen = len;
17742 + p = (void *)args;
17743 + p += sizeof(*args);
17744 + memcpy(p, h_child_name, len);
17748 + /* NFS fires the event for silly-renamed one from kworker */
17750 + if (!dir->i_nlink
17751 + || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
17753 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
17754 + if (unlikely(err)) {
17755 + pr_err("wkq %d\n", err);
17756 + iput(args->h_child_inode);
17757 + iput(args->h_dir);
17766 +/* ---------------------------------------------------------------------- */
17768 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
17772 + AuDebugOn(!(udba & AuOptMask_UDBA));
17775 + if (au_hnotify_op.reset_br)
17776 + err = au_hnotify_op.reset_br(udba, br, perm);
17781 +int au_hnotify_init_br(struct au_branch *br, int perm)
17786 + if (au_hnotify_op.init_br)
17787 + err = au_hnotify_op.init_br(br, perm);
17792 +void au_hnotify_fin_br(struct au_branch *br)
17794 + if (au_hnotify_op.fin_br)
17795 + au_hnotify_op.fin_br(br);
17798 +static void au_hn_destroy_cache(void)
17800 + kmem_cache_destroy(au_cache[AuCache_HNOTIFY]);
17801 + au_cache[AuCache_HNOTIFY] = NULL;
17804 +int __init au_hnotify_init(void)
17809 + au_cache[AuCache_HNOTIFY] = AuCache(au_hnotify);
17810 + if (au_cache[AuCache_HNOTIFY]) {
17812 + if (au_hnotify_op.init)
17813 + err = au_hnotify_op.init();
17814 + if (unlikely(err))
17815 + au_hn_destroy_cache();
17821 +void au_hnotify_fin(void)
17823 + if (au_hnotify_op.fin)
17824 + au_hnotify_op.fin();
17826 + /* cf. au_cache_fin() */
17827 + if (au_cache[AuCache_HNOTIFY])
17828 + au_hn_destroy_cache();
17830 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
17831 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
17832 +++ linux/fs/aufs/iinfo.c 2018-01-29 07:56:20.056658502 +0100
17835 + * Copyright (C) 2005-2017 Junjiro R. Okajima
17837 + * This program, aufs is free software; you can redistribute it and/or modify
17838 + * it under the terms of the GNU General Public License as published by
17839 + * the Free Software Foundation; either version 2 of the License, or
17840 + * (at your option) any later version.
17842 + * This program is distributed in the hope that it will be useful,
17843 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17844 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17845 + * GNU General Public License for more details.
17847 + * You should have received a copy of the GNU General Public License
17848 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17852 + * inode private data
17857 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
17859 + struct inode *h_inode;
17860 + struct au_hinode *hinode;
17862 + IiMustAnyLock(inode);
17864 + hinode = au_hinode(au_ii(inode), bindex);
17865 + h_inode = hinode->hi_inode;
17866 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
17870 +/* todo: hard/soft set? */
17871 +void au_hiput(struct au_hinode *hinode)
17873 + au_hn_free(hinode);
17874 + dput(hinode->hi_whdentry);
17875 + iput(hinode->hi_inode);
17878 +unsigned int au_hi_flags(struct inode *inode, int isdir)
17880 + unsigned int flags;
17881 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
17884 + if (au_opt_test(mnt_flags, XINO))
17885 + au_fset_hi(flags, XINO);
17886 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
17887 + au_fset_hi(flags, HNOTIFY);
17891 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
17892 + struct inode *h_inode, unsigned int flags)
17894 + struct au_hinode *hinode;
17895 + struct inode *hi;
17896 + struct au_iinfo *iinfo = au_ii(inode);
17898 + IiMustWriteLock(inode);
17900 + hinode = au_hinode(iinfo, bindex);
17901 + hi = hinode->hi_inode;
17902 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
17905 + au_hiput(hinode);
17906 + hinode->hi_inode = h_inode;
17909 + struct super_block *sb = inode->i_sb;
17910 + struct au_branch *br;
17912 + AuDebugOn(inode->i_mode
17913 + && (h_inode->i_mode & S_IFMT)
17914 + != (inode->i_mode & S_IFMT));
17915 + if (bindex == iinfo->ii_btop)
17916 + au_cpup_igen(inode, h_inode);
17917 + br = au_sbr(sb, bindex);
17918 + hinode->hi_id = br->br_id;
17919 + if (au_ftest_hi(flags, XINO)) {
17920 + err = au_xino_write(sb, bindex, h_inode->i_ino,
17922 + if (unlikely(err))
17923 + AuIOErr1("failed au_xino_write() %d\n", err);
17926 + if (au_ftest_hi(flags, HNOTIFY)
17927 + && au_br_hnotifyable(br->br_perm)) {
17928 + err = au_hn_alloc(hinode, inode);
17929 + if (unlikely(err))
17930 + AuIOErr1("au_hn_alloc() %d\n", err);
17935 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
17936 + struct dentry *h_wh)
17938 + struct au_hinode *hinode;
17940 + IiMustWriteLock(inode);
17942 + hinode = au_hinode(au_ii(inode), bindex);
17943 + AuDebugOn(hinode->hi_whdentry);
17944 + hinode->hi_whdentry = h_wh;
17947 +void au_update_iigen(struct inode *inode, int half)
17949 + struct au_iinfo *iinfo;
17950 + struct au_iigen *iigen;
17951 + unsigned int sigen;
17953 + sigen = au_sigen(inode->i_sb);
17954 + iinfo = au_ii(inode);
17955 + iigen = &iinfo->ii_generation;
17956 + spin_lock(&iigen->ig_spin);
17957 + iigen->ig_generation = sigen;
17959 + au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
17961 + au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
17962 + spin_unlock(&iigen->ig_spin);
17965 +/* it may be called at remount time, too */
17966 +void au_update_ibrange(struct inode *inode, int do_put_zero)
17968 + struct au_iinfo *iinfo;
17969 + aufs_bindex_t bindex, bbot;
17971 + AuDebugOn(au_is_bad_inode(inode));
17972 + IiMustWriteLock(inode);
17974 + iinfo = au_ii(inode);
17975 + if (do_put_zero && iinfo->ii_btop >= 0) {
17976 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
17978 + struct inode *h_i;
17980 + h_i = au_hinode(iinfo, bindex)->hi_inode;
17983 + && !(h_i->i_state & I_LINKABLE))
17984 + au_set_h_iptr(inode, bindex, NULL, 0);
17988 + iinfo->ii_btop = -1;
17989 + iinfo->ii_bbot = -1;
17990 + bbot = au_sbbot(inode->i_sb);
17991 + for (bindex = 0; bindex <= bbot; bindex++)
17992 + if (au_hinode(iinfo, bindex)->hi_inode) {
17993 + iinfo->ii_btop = bindex;
17996 + if (iinfo->ii_btop >= 0)
17997 + for (bindex = bbot; bindex >= iinfo->ii_btop; bindex--)
17998 + if (au_hinode(iinfo, bindex)->hi_inode) {
17999 + iinfo->ii_bbot = bindex;
18002 + AuDebugOn(iinfo->ii_btop > iinfo->ii_bbot);
18005 +/* ---------------------------------------------------------------------- */
18007 +void au_icntnr_init_once(void *_c)
18009 + struct au_icntnr *c = _c;
18010 + struct au_iinfo *iinfo = &c->iinfo;
18012 + spin_lock_init(&iinfo->ii_generation.ig_spin);
18013 + au_rw_init(&iinfo->ii_rwsem);
18014 + inode_init_once(&c->vfs_inode);
18017 +void au_hinode_init(struct au_hinode *hinode)
18019 + hinode->hi_inode = NULL;
18020 + hinode->hi_id = -1;
18021 + au_hn_init(hinode);
18022 + hinode->hi_whdentry = NULL;
18025 +int au_iinfo_init(struct inode *inode)
18027 + struct au_iinfo *iinfo;
18028 + struct super_block *sb;
18029 + struct au_hinode *hi;
18032 + sb = inode->i_sb;
18033 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
18034 + nbr = au_sbbot(sb) + 1;
18035 + if (unlikely(nbr <= 0))
18037 + hi = kmalloc_array(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
18039 + au_ninodes_inc(sb);
18041 + iinfo->ii_hinode = hi;
18042 + for (i = 0; i < nbr; i++, hi++)
18043 + au_hinode_init(hi);
18045 + iinfo->ii_generation.ig_generation = au_sigen(sb);
18046 + iinfo->ii_btop = -1;
18047 + iinfo->ii_bbot = -1;
18048 + iinfo->ii_vdir = NULL;
18054 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink)
18057 + struct au_hinode *hip;
18059 + AuRwMustWriteLock(&iinfo->ii_rwsem);
18062 + hip = au_krealloc(iinfo->ii_hinode, sizeof(*hip) * nbr, GFP_NOFS,
18065 + iinfo->ii_hinode = hip;
18066 + i = iinfo->ii_bbot + 1;
18068 + for (; i < nbr; i++, hip++)
18069 + au_hinode_init(hip);
18076 +void au_iinfo_fin(struct inode *inode)
18078 + struct au_iinfo *iinfo;
18079 + struct au_hinode *hi;
18080 + struct super_block *sb;
18081 + aufs_bindex_t bindex, bbot;
18082 + const unsigned char unlinked = !inode->i_nlink;
18084 + AuDebugOn(au_is_bad_inode(inode));
18086 + sb = inode->i_sb;
18087 + au_ninodes_dec(sb);
18088 + if (si_pid_test(sb))
18089 + au_xino_delete_inode(inode, unlinked);
18092 + * it is safe to hide the dependency between sbinfo and
18096 + si_noflush_read_lock(sb);
18097 + au_xino_delete_inode(inode, unlinked);
18098 + si_read_unlock(sb);
18102 + iinfo = au_ii(inode);
18103 + if (iinfo->ii_vdir)
18104 + au_vdir_free(iinfo->ii_vdir);
18106 + bindex = iinfo->ii_btop;
18107 + if (bindex >= 0) {
18108 + hi = au_hinode(iinfo, bindex);
18109 + bbot = iinfo->ii_bbot;
18110 + while (bindex++ <= bbot) {
18111 + if (hi->hi_inode)
18116 + kfree(iinfo->ii_hinode);
18117 + AuRwDestroy(&iinfo->ii_rwsem);
18119 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
18120 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
18121 +++ linux/fs/aufs/inode.c 2018-01-29 07:56:20.056658502 +0100
18124 + * Copyright (C) 2005-2017 Junjiro R. Okajima
18126 + * This program, aufs is free software; you can redistribute it and/or modify
18127 + * it under the terms of the GNU General Public License as published by
18128 + * the Free Software Foundation; either version 2 of the License, or
18129 + * (at your option) any later version.
18131 + * This program is distributed in the hope that it will be useful,
18132 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18133 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18134 + * GNU General Public License for more details.
18136 + * You should have received a copy of the GNU General Public License
18137 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18141 + * inode functions
18146 +struct inode *au_igrab(struct inode *inode)
18149 + AuDebugOn(!atomic_read(&inode->i_count));
18155 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
18157 + au_cpup_attr_all(inode, /*force*/0);
18158 + au_update_iigen(inode, /*half*/1);
18160 + inode->i_version++;
18163 +static int au_ii_refresh(struct inode *inode, int *update)
18167 + aufs_bindex_t bindex, new_bindex;
18168 + struct super_block *sb;
18169 + struct au_iinfo *iinfo;
18170 + struct au_hinode *p, *q, tmp;
18172 + AuDebugOn(au_is_bad_inode(inode));
18173 + IiMustWriteLock(inode);
18176 + sb = inode->i_sb;
18177 + nbr = au_sbbot(sb) + 1;
18178 + type = inode->i_mode & S_IFMT;
18179 + iinfo = au_ii(inode);
18180 + err = au_hinode_realloc(iinfo, nbr, /*may_shrink*/0);
18181 + if (unlikely(err))
18184 + AuDebugOn(iinfo->ii_btop < 0);
18185 + p = au_hinode(iinfo, iinfo->ii_btop);
18186 + for (bindex = iinfo->ii_btop; bindex <= iinfo->ii_bbot;
18188 + if (!p->hi_inode)
18191 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
18192 + new_bindex = au_br_index(sb, p->hi_id);
18193 + if (new_bindex == bindex)
18196 + if (new_bindex < 0) {
18199 + p->hi_inode = NULL;
18203 + if (new_bindex < iinfo->ii_btop)
18204 + iinfo->ii_btop = new_bindex;
18205 + if (iinfo->ii_bbot < new_bindex)
18206 + iinfo->ii_bbot = new_bindex;
18207 + /* swap two lower inode, and loop again */
18208 + q = au_hinode(iinfo, new_bindex);
18212 + if (tmp.hi_inode) {
18217 + au_update_ibrange(inode, /*do_put_zero*/0);
18218 + au_hinode_realloc(iinfo, nbr, /*may_shrink*/1); /* harmless if err */
18219 + e = au_dy_irefresh(inode);
18220 + if (unlikely(e && !err))
18228 +void au_refresh_iop(struct inode *inode, int force_getattr)
18231 + struct au_sbinfo *sbi = au_sbi(inode->i_sb);
18232 + const struct inode_operations *iop
18233 + = force_getattr ? aufs_iop : sbi->si_iop_array;
18235 + if (inode->i_op == iop)
18238 + switch (inode->i_mode & S_IFMT) {
18240 + type = AuIop_DIR;
18243 + type = AuIop_SYMLINK;
18246 + type = AuIop_OTHER;
18250 + inode->i_op = iop + type;
18251 + /* unnecessary smp_wmb() */
18254 +int au_refresh_hinode_self(struct inode *inode)
18258 + err = au_ii_refresh(inode, &update);
18260 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
18266 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
18268 + int err, e, update;
18269 + unsigned int flags;
18271 + aufs_bindex_t bindex, bbot;
18272 + unsigned char isdir;
18273 + struct au_hinode *p;
18274 + struct au_iinfo *iinfo;
18276 + err = au_ii_refresh(inode, &update);
18277 + if (unlikely(err))
18281 + iinfo = au_ii(inode);
18282 + p = au_hinode(iinfo, iinfo->ii_btop);
18283 + mode = (inode->i_mode & S_IFMT);
18284 + isdir = S_ISDIR(mode);
18285 + flags = au_hi_flags(inode, isdir);
18286 + bbot = au_dbbot(dentry);
18287 + for (bindex = au_dbtop(dentry); bindex <= bbot; bindex++) {
18288 + struct inode *h_i, *h_inode;
18289 + struct dentry *h_d;
18291 + h_d = au_h_dptr(dentry, bindex);
18292 + if (!h_d || d_is_negative(h_d))
18295 + h_inode = d_inode(h_d);
18296 + AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
18297 + if (iinfo->ii_btop <= bindex && bindex <= iinfo->ii_bbot) {
18298 + h_i = au_h_iptr(inode, bindex);
18300 + if (h_i == h_inode)
18306 + if (bindex < iinfo->ii_btop)
18307 + iinfo->ii_btop = bindex;
18308 + if (iinfo->ii_bbot < bindex)
18309 + iinfo->ii_bbot = bindex;
18310 + au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
18313 + au_update_ibrange(inode, /*do_put_zero*/0);
18314 + e = au_dy_irefresh(inode);
18315 + if (unlikely(e && !err))
18318 + au_refresh_hinode_attr(inode, update && isdir);
18325 +static int set_inode(struct inode *inode, struct dentry *dentry)
18328 + unsigned int flags;
18330 + aufs_bindex_t bindex, btop, btail;
18331 + unsigned char isdir;
18332 + struct dentry *h_dentry;
18333 + struct inode *h_inode;
18334 + struct au_iinfo *iinfo;
18335 + struct inode_operations *iop;
18337 + IiMustWriteLock(inode);
18341 + iop = au_sbi(inode->i_sb)->si_iop_array;
18342 + btop = au_dbtop(dentry);
18343 + h_dentry = au_h_dptr(dentry, btop);
18344 + h_inode = d_inode(h_dentry);
18345 + mode = h_inode->i_mode;
18346 + switch (mode & S_IFMT) {
18348 + btail = au_dbtail(dentry);
18349 + inode->i_op = iop + AuIop_OTHER;
18350 + inode->i_fop = &aufs_file_fop;
18351 + err = au_dy_iaop(inode, btop, h_inode);
18352 + if (unlikely(err))
18357 + btail = au_dbtaildir(dentry);
18358 + inode->i_op = iop + AuIop_DIR;
18359 + inode->i_fop = &aufs_dir_fop;
18362 + btail = au_dbtail(dentry);
18363 + inode->i_op = iop + AuIop_SYMLINK;
18369 + btail = au_dbtail(dentry);
18370 + inode->i_op = iop + AuIop_OTHER;
18371 + init_special_inode(inode, mode, h_inode->i_rdev);
18374 + AuIOErr("Unknown file type 0%o\n", mode);
18379 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
18380 + flags = au_hi_flags(inode, isdir);
18381 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
18382 + && au_ftest_hi(flags, HNOTIFY)
18383 + && dentry->d_name.len > AUFS_WH_PFX_LEN
18384 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
18385 + au_fclr_hi(flags, HNOTIFY);
18386 + iinfo = au_ii(inode);
18387 + iinfo->ii_btop = btop;
18388 + iinfo->ii_bbot = btail;
18389 + for (bindex = btop; bindex <= btail; bindex++) {
18390 + h_dentry = au_h_dptr(dentry, bindex);
18392 + au_set_h_iptr(inode, bindex,
18393 + au_igrab(d_inode(h_dentry)), flags);
18395 + au_cpup_attr_all(inode, /*force*/1);
18397 + * to force calling aufs_get_acl() every time,
18398 + * do not call cache_no_acl() for aufs inode.
18406 + * successful returns with iinfo write_locked
18408 + * zero: success, matched
18409 + * plus: no error, but unmatched
18411 +static int reval_inode(struct inode *inode, struct dentry *dentry)
18414 + unsigned int gen, igflags;
18415 + aufs_bindex_t bindex, bbot;
18416 + struct inode *h_inode, *h_dinode;
18417 + struct dentry *h_dentry;
18420 + * before this function, if aufs got any iinfo lock, it must be only
18421 + * one, the parent dir.
18422 + * it can happen by UDBA and the obsoleted inode number.
18425 + if (unlikely(inode->i_ino == parent_ino(dentry)))
18429 + ii_write_lock_new_child(inode);
18430 + h_dentry = au_h_dptr(dentry, au_dbtop(dentry));
18431 + h_dinode = d_inode(h_dentry);
18432 + bbot = au_ibbot(inode);
18433 + for (bindex = au_ibtop(inode); bindex <= bbot; bindex++) {
18434 + h_inode = au_h_iptr(inode, bindex);
18435 + if (!h_inode || h_inode != h_dinode)
18439 + gen = au_iigen(inode, &igflags);
18440 + if (gen == au_digen(dentry)
18441 + && !au_ig_ftest(igflags, HALF_REFRESHED))
18444 + /* fully refresh inode using dentry */
18445 + err = au_refresh_hinode(inode, dentry);
18447 + au_update_iigen(inode, /*half*/0);
18451 + if (unlikely(err))
18452 + ii_write_unlock(inode);
18457 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
18458 + unsigned int d_type, ino_t *ino)
18461 + const int isnondir = d_type != DT_DIR;
18463 + /* prevent hardlinked inode number from race condition */
18465 + err = au_xinondir_enter(sb, bindex, h_ino, &idx);
18466 + if (unlikely(err))
18470 + err = au_xino_read(sb, bindex, h_ino, ino);
18471 + if (unlikely(err))
18472 + goto out_xinondir;
18476 + *ino = au_xino_new_ino(sb);
18477 + if (unlikely(!*ino))
18478 + goto out_xinondir;
18479 + err = au_xino_write(sb, bindex, h_ino, *ino);
18480 + if (unlikely(err))
18481 + goto out_xinondir;
18485 + if (isnondir && idx >= 0)
18486 + au_xinondir_leave(sb, bindex, h_ino, idx);
18491 +/* successful returns with iinfo write_locked */
18492 +/* todo: return with unlocked? */
18493 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
18495 + struct inode *inode, *h_inode;
18496 + struct dentry *h_dentry;
18497 + struct super_block *sb;
18498 + ino_t h_ino, ino;
18499 + int err, idx, hlinked;
18500 + aufs_bindex_t btop;
18502 + sb = dentry->d_sb;
18503 + btop = au_dbtop(dentry);
18504 + h_dentry = au_h_dptr(dentry, btop);
18505 + h_inode = d_inode(h_dentry);
18506 + h_ino = h_inode->i_ino;
18507 + hlinked = !d_is_dir(h_dentry) && h_inode->i_nlink > 1;
18511 + * stop 'race'-ing between hardlinks under different
18515 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
18516 + inode = ERR_PTR(err);
18517 + if (unlikely(err))
18521 + err = au_xino_read(sb, btop, h_ino, &ino);
18522 + inode = ERR_PTR(err);
18523 + if (unlikely(err))
18524 + goto out_xinondir;
18527 + ino = au_xino_new_ino(sb);
18528 + if (unlikely(!ino)) {
18529 + inode = ERR_PTR(-EIO);
18530 + goto out_xinondir;
18534 + AuDbg("i%lu\n", (unsigned long)ino);
18535 + inode = au_iget_locked(sb, ino);
18536 + err = PTR_ERR(inode);
18537 + if (IS_ERR(inode))
18538 + goto out_xinondir;
18540 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
18541 + if (inode->i_state & I_NEW) {
18542 + ii_write_lock_new_child(inode);
18543 + err = set_inode(inode, dentry);
18545 + unlock_new_inode(inode);
18546 + goto out_xinondir; /* success */
18550 + * iget_failed() calls iput(), but we need to call
18551 + * ii_write_unlock() after iget_failed(). so dirty hack for
18554 + atomic_inc(&inode->i_count);
18555 + iget_failed(inode);
18556 + ii_write_unlock(inode);
18557 + au_xino_write(sb, btop, h_ino, /*ino*/0);
18558 + /* ignore this error */
18560 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
18562 + * horrible race condition between lookup, readdir and copyup
18563 + * (or something).
18565 + if (hlinked && idx >= 0)
18566 + au_xinondir_leave(sb, btop, h_ino, idx);
18567 + err = reval_inode(inode, dentry);
18568 + if (unlikely(err < 0)) {
18573 + goto out; /* success */
18574 + else if (hlinked && idx >= 0) {
18575 + err = au_xinondir_enter(sb, btop, h_ino, &idx);
18576 + if (unlikely(err)) {
18578 + inode = ERR_PTR(err);
18584 + if (unlikely(au_test_fs_unique_ino(h_inode)))
18585 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
18586 + " b%d, %s, %pd, hi%lu, i%lu.\n",
18587 + btop, au_sbtype(h_dentry->d_sb), dentry,
18588 + (unsigned long)h_ino, (unsigned long)ino);
18590 + err = au_xino_write(sb, btop, h_ino, /*ino*/0);
18593 + if (hlinked && idx >= 0)
18594 + au_xinondir_leave(sb, btop, h_ino, idx);
18600 + inode = ERR_PTR(err);
18602 + if (hlinked && idx >= 0)
18603 + au_xinondir_leave(sb, btop, h_ino, idx);
18608 +/* ---------------------------------------------------------------------- */
18610 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
18611 + struct inode *inode)
18614 + struct inode *hi;
18616 + err = au_br_rdonly(au_sbr(sb, bindex));
18618 + /* pseudo-link after flushed may happen out of bounds */
18621 + && au_ibtop(inode) <= bindex
18622 + && bindex <= au_ibbot(inode)) {
18624 + * permission check is unnecessary since vfsub routine
18625 + * will be called later
18627 + hi = au_h_iptr(inode, bindex);
18629 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
18635 +int au_test_h_perm(struct inode *h_inode, int mask)
18637 + if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
18639 + return inode_permission(h_inode, mask);
18642 +int au_test_h_perm_sio(struct inode *h_inode, int mask)
18644 + if (au_test_nfs(h_inode->i_sb)
18645 + && (mask & MAY_WRITE)
18646 + && S_ISDIR(h_inode->i_mode))
18647 + mask |= MAY_READ; /* force permission check */
18648 + return au_test_h_perm(h_inode, mask);
18650 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
18651 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
18652 +++ linux/fs/aufs/inode.h 2018-01-29 07:56:20.056658502 +0100
18655 + * Copyright (C) 2005-2017 Junjiro R. Okajima
18657 + * This program, aufs is free software; you can redistribute it and/or modify
18658 + * it under the terms of the GNU General Public License as published by
18659 + * the Free Software Foundation; either version 2 of the License, or
18660 + * (at your option) any later version.
18662 + * This program is distributed in the hope that it will be useful,
18663 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18664 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18665 + * GNU General Public License for more details.
18667 + * You should have received a copy of the GNU General Public License
18668 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18672 + * inode operations
18675 +#ifndef __AUFS_INODE_H__
18676 +#define __AUFS_INODE_H__
18680 +#include <linux/fsnotify.h>
18681 +#include "rwsem.h"
18685 +struct au_hnotify {
18686 +#ifdef CONFIG_AUFS_HNOTIFY
18687 +#ifdef CONFIG_AUFS_HFSNOTIFY
18688 + /* never use fsnotify_add_vfsmount_mark() */
18689 + struct fsnotify_mark hn_mark;
18691 + struct inode *hn_aufs_inode; /* no get/put */
18693 +} ____cacheline_aligned_in_smp;
18695 +struct au_hinode {
18696 + struct inode *hi_inode;
18697 + aufs_bindex_t hi_id;
18698 +#ifdef CONFIG_AUFS_HNOTIFY
18699 + struct au_hnotify *hi_notify;
18702 + /* reference to the copied-up whiteout with get/put */
18703 + struct dentry *hi_whdentry;
18707 +#define AuIG_HALF_REFRESHED 1
18708 +#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
18709 +#define au_ig_fset(flags, name) \
18710 + do { (flags) |= AuIG_##name; } while (0)
18711 +#define au_ig_fclr(flags, name) \
18712 + do { (flags) &= ~AuIG_##name; } while (0)
18715 + spinlock_t ig_spin;
18716 + __u32 ig_generation, ig_flags;
18721 + struct au_iigen ii_generation;
18722 + struct super_block *ii_hsb1; /* no get/put */
18724 + struct au_rwsem ii_rwsem;
18725 + aufs_bindex_t ii_btop, ii_bbot;
18727 + struct au_hinode *ii_hinode;
18728 + struct au_vdir *ii_vdir;
18731 +struct au_icntnr {
18732 + struct au_iinfo iinfo;
18733 + struct inode vfs_inode;
18734 + struct hlist_bl_node plink;
18735 +} ____cacheline_aligned_in_smp;
18737 +/* au_pin flags */
18738 +#define AuPin_DI_LOCKED 1
18739 +#define AuPin_MNT_WRITE (1 << 1)
18740 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
18741 +#define au_fset_pin(flags, name) \
18742 + do { (flags) |= AuPin_##name; } while (0)
18743 +#define au_fclr_pin(flags, name) \
18744 + do { (flags) &= ~AuPin_##name; } while (0)
18748 + struct dentry *dentry;
18749 + unsigned int udba;
18750 + unsigned char lsc_di, lsc_hi, flags;
18751 + aufs_bindex_t bindex;
18754 + struct dentry *parent;
18755 + struct au_hinode *hdir;
18756 + struct vfsmount *h_mnt;
18758 + /* temporary unlock/relock for copyup */
18759 + struct dentry *h_dentry, *h_parent;
18760 + struct au_branch *br;
18761 + struct task_struct *task;
18764 +void au_pin_hdir_unlock(struct au_pin *p);
18765 +int au_pin_hdir_lock(struct au_pin *p);
18766 +int au_pin_hdir_relock(struct au_pin *p);
18767 +void au_pin_hdir_acquire_nest(struct au_pin *p);
18768 +void au_pin_hdir_release(struct au_pin *p);
18770 +/* ---------------------------------------------------------------------- */
18772 +static inline struct au_iinfo *au_ii(struct inode *inode)
18774 + BUG_ON(is_bad_inode(inode));
18775 + return &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
18778 +/* ---------------------------------------------------------------------- */
18781 +struct inode *au_igrab(struct inode *inode);
18782 +void au_refresh_iop(struct inode *inode, int force_getattr);
18783 +int au_refresh_hinode_self(struct inode *inode);
18784 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
18785 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
18786 + unsigned int d_type, ino_t *ino);
18787 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
18788 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
18789 + struct inode *inode);
18790 +int au_test_h_perm(struct inode *h_inode, int mask);
18791 +int au_test_h_perm_sio(struct inode *h_inode, int mask);
18793 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
18794 + ino_t h_ino, unsigned int d_type, ino_t *ino)
18796 +#ifdef CONFIG_AUFS_SHWH
18797 + return au_ino(sb, bindex, h_ino, d_type, ino);
18810 +extern struct inode_operations aufs_iop[AuIop_Last],
18811 + aufs_iop_nogetattr[AuIop_Last];
18813 +/* au_wr_dir flags */
18814 +#define AuWrDir_ADD_ENTRY 1
18815 +#define AuWrDir_ISDIR (1 << 1)
18816 +#define AuWrDir_TMPFILE (1 << 2)
18817 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
18818 +#define au_fset_wrdir(flags, name) \
18819 + do { (flags) |= AuWrDir_##name; } while (0)
18820 +#define au_fclr_wrdir(flags, name) \
18821 + do { (flags) &= ~AuWrDir_##name; } while (0)
18823 +struct au_wr_dir_args {
18824 + aufs_bindex_t force_btgt;
18825 + unsigned char flags;
18827 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
18828 + struct au_wr_dir_args *args);
18830 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
18831 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
18832 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
18833 + unsigned int udba, unsigned char flags);
18834 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
18835 + unsigned int udba, unsigned char flags) __must_check;
18836 +int au_do_pin(struct au_pin *pin) __must_check;
18837 +void au_unpin(struct au_pin *pin);
18838 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
18840 +#define AuIcpup_DID_CPUP 1
18841 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
18842 +#define au_fset_icpup(flags, name) \
18843 + do { (flags) |= AuIcpup_##name; } while (0)
18844 +#define au_fclr_icpup(flags, name) \
18845 + do { (flags) &= ~AuIcpup_##name; } while (0)
18847 +struct au_icpup_args {
18848 + unsigned char flags;
18849 + unsigned char pin_flags;
18850 + aufs_bindex_t btgt;
18851 + unsigned int udba;
18852 + struct au_pin pin;
18853 + struct path h_path;
18854 + struct inode *h_inode;
18857 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
18858 + struct au_icpup_args *a);
18860 +int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path,
18864 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
18865 + struct dentry *h_parent, int isdir);
18866 +int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
18868 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname);
18869 +int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
18871 +struct vfsub_aopen_args;
18872 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
18873 + struct vfsub_aopen_args *args);
18874 +int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode);
18875 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
18876 + struct dentry *dentry);
18877 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
18880 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
18881 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
18882 + struct dentry *h_parent, int isdir);
18883 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
18884 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
18887 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
18888 +int aufs_rename(struct inode *src_dir, struct dentry *src_dentry,
18889 + struct inode *dir, struct dentry *dentry,
18890 + unsigned int flags);
18893 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
18894 +void au_hiput(struct au_hinode *hinode);
18895 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
18896 + struct dentry *h_wh);
18897 +unsigned int au_hi_flags(struct inode *inode, int isdir);
18899 +/* hinode flags */
18900 +#define AuHi_XINO 1
18901 +#define AuHi_HNOTIFY (1 << 1)
18902 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
18903 +#define au_fset_hi(flags, name) \
18904 + do { (flags) |= AuHi_##name; } while (0)
18905 +#define au_fclr_hi(flags, name) \
18906 + do { (flags) &= ~AuHi_##name; } while (0)
18908 +#ifndef CONFIG_AUFS_HNOTIFY
18909 +#undef AuHi_HNOTIFY
18910 +#define AuHi_HNOTIFY 0
18913 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
18914 + struct inode *h_inode, unsigned int flags);
18916 +void au_update_iigen(struct inode *inode, int half);
18917 +void au_update_ibrange(struct inode *inode, int do_put_zero);
18919 +void au_icntnr_init_once(void *_c);
18920 +void au_hinode_init(struct au_hinode *hinode);
18921 +int au_iinfo_init(struct inode *inode);
18922 +void au_iinfo_fin(struct inode *inode);
18923 +int au_hinode_realloc(struct au_iinfo *iinfo, int nbr, int may_shrink);
18925 +#ifdef CONFIG_PROC_FS
18927 +int au_plink_maint(struct super_block *sb, int flags);
18929 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
18930 +int au_plink_maint_enter(struct super_block *sb);
18931 +#ifdef CONFIG_AUFS_DEBUG
18932 +void au_plink_list(struct super_block *sb);
18934 +AuStubVoid(au_plink_list, struct super_block *sb)
18936 +int au_plink_test(struct inode *inode);
18937 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
18938 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
18939 + struct dentry *h_dentry);
18940 +void au_plink_put(struct super_block *sb, int verbose);
18941 +void au_plink_clean(struct super_block *sb, int verbose);
18942 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
18944 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
18945 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
18946 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
18947 +AuStubVoid(au_plink_list, struct super_block *sb);
18948 +AuStubInt0(au_plink_test, struct inode *inode);
18949 +AuStub(struct dentry *, au_plink_lkup, return NULL,
18950 + struct inode *inode, aufs_bindex_t bindex);
18951 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
18952 + struct dentry *h_dentry);
18953 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
18954 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
18955 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
18956 +#endif /* CONFIG_PROC_FS */
18958 +#ifdef CONFIG_AUFS_XATTR
18960 +int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
18961 + unsigned int verbose);
18962 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
18963 +void au_xattr_init(struct super_block *sb);
18965 +AuStubInt0(au_cpup_xattr, struct dentry *h_dst, struct dentry *h_src,
18966 + int ignore_flags, unsigned int verbose);
18967 +AuStubVoid(au_xattr_init, struct super_block *sb);
18970 +#ifdef CONFIG_FS_POSIX_ACL
18971 +struct posix_acl *aufs_get_acl(struct inode *inode, int type);
18972 +int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
18975 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
18981 +struct au_sxattr {
18985 + const char *name;
18986 + const void *value;
18991 + struct posix_acl *acl;
18996 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
18997 + struct au_sxattr *arg);
19000 +/* ---------------------------------------------------------------------- */
19002 +/* lock subclass for iinfo */
19004 + AuLsc_II_CHILD, /* child first */
19005 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
19006 + AuLsc_II_CHILD3, /* copyup dirs */
19007 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
19008 + AuLsc_II_PARENT2,
19009 + AuLsc_II_PARENT3, /* copyup dirs */
19010 + AuLsc_II_NEW_CHILD
19014 + * ii_read_lock_child, ii_write_lock_child,
19015 + * ii_read_lock_child2, ii_write_lock_child2,
19016 + * ii_read_lock_child3, ii_write_lock_child3,
19017 + * ii_read_lock_parent, ii_write_lock_parent,
19018 + * ii_read_lock_parent2, ii_write_lock_parent2,
19019 + * ii_read_lock_parent3, ii_write_lock_parent3,
19020 + * ii_read_lock_new_child, ii_write_lock_new_child,
19022 +#define AuReadLockFunc(name, lsc) \
19023 +static inline void ii_read_lock_##name(struct inode *i) \
19025 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
19028 +#define AuWriteLockFunc(name, lsc) \
19029 +static inline void ii_write_lock_##name(struct inode *i) \
19031 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
19034 +#define AuRWLockFuncs(name, lsc) \
19035 + AuReadLockFunc(name, lsc) \
19036 + AuWriteLockFunc(name, lsc)
19038 +AuRWLockFuncs(child, CHILD);
19039 +AuRWLockFuncs(child2, CHILD2);
19040 +AuRWLockFuncs(child3, CHILD3);
19041 +AuRWLockFuncs(parent, PARENT);
19042 +AuRWLockFuncs(parent2, PARENT2);
19043 +AuRWLockFuncs(parent3, PARENT3);
19044 +AuRWLockFuncs(new_child, NEW_CHILD);
19046 +#undef AuReadLockFunc
19047 +#undef AuWriteLockFunc
19048 +#undef AuRWLockFuncs
19050 +#define ii_read_unlock(i) au_rw_read_unlock(&au_ii(i)->ii_rwsem)
19051 +#define ii_write_unlock(i) au_rw_write_unlock(&au_ii(i)->ii_rwsem)
19052 +#define ii_downgrade_lock(i) au_rw_dgrade_lock(&au_ii(i)->ii_rwsem)
19054 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
19055 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
19056 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
19058 +/* ---------------------------------------------------------------------- */
19060 +static inline void au_icntnr_init(struct au_icntnr *c)
19062 +#ifdef CONFIG_AUFS_DEBUG
19063 + c->vfs_inode.i_mode = 0;
19067 +static inline unsigned int au_iigen(struct inode *inode, unsigned int *igflags)
19069 + unsigned int gen;
19070 + struct au_iinfo *iinfo;
19071 + struct au_iigen *iigen;
19073 + iinfo = au_ii(inode);
19074 + iigen = &iinfo->ii_generation;
19075 + spin_lock(&iigen->ig_spin);
19077 + *igflags = iigen->ig_flags;
19078 + gen = iigen->ig_generation;
19079 + spin_unlock(&iigen->ig_spin);
19084 +/* tiny test for inode number */
19085 +/* tmpfs generation is too rough */
19086 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
19088 + struct au_iinfo *iinfo;
19090 + iinfo = au_ii(inode);
19091 + AuRwMustAnyLock(&iinfo->ii_rwsem);
19092 + return !(iinfo->ii_hsb1 == h_inode->i_sb
19093 + && iinfo->ii_higen == h_inode->i_generation);
19096 +static inline void au_iigen_dec(struct inode *inode)
19098 + struct au_iinfo *iinfo;
19099 + struct au_iigen *iigen;
19101 + iinfo = au_ii(inode);
19102 + iigen = &iinfo->ii_generation;
19103 + spin_lock(&iigen->ig_spin);
19104 + iigen->ig_generation--;
19105 + spin_unlock(&iigen->ig_spin);
19108 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
19113 + if (unlikely(inode && au_iigen(inode, NULL) != sigen))
19119 +/* ---------------------------------------------------------------------- */
19121 +static inline struct au_hinode *au_hinode(struct au_iinfo *iinfo,
19122 + aufs_bindex_t bindex)
19124 + return iinfo->ii_hinode + bindex;
19127 +static inline int au_is_bad_inode(struct inode *inode)
19129 + return !!(is_bad_inode(inode) || !au_hinode(au_ii(inode), 0));
19132 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
19133 + aufs_bindex_t bindex)
19135 + IiMustAnyLock(inode);
19136 + return au_hinode(au_ii(inode), bindex)->hi_id;
19139 +static inline aufs_bindex_t au_ibtop(struct inode *inode)
19141 + IiMustAnyLock(inode);
19142 + return au_ii(inode)->ii_btop;
19145 +static inline aufs_bindex_t au_ibbot(struct inode *inode)
19147 + IiMustAnyLock(inode);
19148 + return au_ii(inode)->ii_bbot;
19151 +static inline struct au_vdir *au_ivdir(struct inode *inode)
19153 + IiMustAnyLock(inode);
19154 + return au_ii(inode)->ii_vdir;
19157 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
19159 + IiMustAnyLock(inode);
19160 + return au_hinode(au_ii(inode), bindex)->hi_whdentry;
19163 +static inline void au_set_ibtop(struct inode *inode, aufs_bindex_t bindex)
19165 + IiMustWriteLock(inode);
19166 + au_ii(inode)->ii_btop = bindex;
19169 +static inline void au_set_ibbot(struct inode *inode, aufs_bindex_t bindex)
19171 + IiMustWriteLock(inode);
19172 + au_ii(inode)->ii_bbot = bindex;
19175 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
19177 + IiMustWriteLock(inode);
19178 + au_ii(inode)->ii_vdir = vdir;
19181 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
19183 + IiMustAnyLock(inode);
19184 + return au_hinode(au_ii(inode), bindex);
19187 +/* ---------------------------------------------------------------------- */
19189 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
19192 + return pin->parent;
19196 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
19198 + if (pin && pin->hdir)
19199 + return pin->hdir->hi_inode;
19203 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
19206 + return pin->hdir;
19210 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
19213 + pin->dentry = dentry;
19216 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
19217 + unsigned char lflag)
19221 + au_fset_pin(pin->flags, DI_LOCKED);
19223 + au_fclr_pin(pin->flags, DI_LOCKED);
19227 +#if 0 /* reserved */
19228 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
19231 + dput(pin->parent);
19232 + pin->parent = dget(parent);
19237 +/* ---------------------------------------------------------------------- */
19240 +#ifdef CONFIG_AUFS_HNOTIFY
19241 +struct au_hnotify_op {
19242 + void (*ctl)(struct au_hinode *hinode, int do_set);
19243 + int (*alloc)(struct au_hinode *hinode);
19246 + * if it returns true, the the caller should free hinode->hi_notify,
19247 + * otherwise ->free() frees it.
19249 + int (*free)(struct au_hinode *hinode,
19250 + struct au_hnotify *hn) __must_check;
19252 + void (*fin)(void);
19253 + int (*init)(void);
19255 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
19256 + void (*fin_br)(struct au_branch *br);
19257 + int (*init_br)(struct au_branch *br, int perm);
19261 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
19262 +void au_hn_free(struct au_hinode *hinode);
19263 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
19264 +void au_hn_reset(struct inode *inode, unsigned int flags);
19265 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
19266 + struct qstr *h_child_qstr, struct inode *h_child_inode);
19267 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
19268 +int au_hnotify_init_br(struct au_branch *br, int perm);
19269 +void au_hnotify_fin_br(struct au_branch *br);
19270 +int __init au_hnotify_init(void);
19271 +void au_hnotify_fin(void);
19274 +extern const struct au_hnotify_op au_hnotify_op;
19277 +void au_hn_init(struct au_hinode *hinode)
19279 + hinode->hi_notify = NULL;
19282 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
19284 + return hinode->hi_notify;
19288 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
19289 + struct au_hinode *hinode __maybe_unused,
19290 + struct inode *inode __maybe_unused)
19291 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
19292 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
19293 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
19294 + int do_set __maybe_unused)
19295 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
19296 + unsigned int flags __maybe_unused)
19297 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
19298 + struct au_branch *br __maybe_unused,
19299 + int perm __maybe_unused)
19300 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
19301 + int perm __maybe_unused)
19302 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
19303 +AuStubInt0(__init au_hnotify_init, void)
19304 +AuStubVoid(au_hnotify_fin, void)
19305 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
19306 +#endif /* CONFIG_AUFS_HNOTIFY */
19308 +static inline void au_hn_suspend(struct au_hinode *hdir)
19310 + au_hn_ctl(hdir, /*do_set*/0);
19313 +static inline void au_hn_resume(struct au_hinode *hdir)
19315 + au_hn_ctl(hdir, /*do_set*/1);
19318 +static inline void au_hn_inode_lock(struct au_hinode *hdir)
19320 + inode_lock(hdir->hi_inode);
19321 + au_hn_suspend(hdir);
19324 +static inline void au_hn_inode_lock_nested(struct au_hinode *hdir,
19325 + unsigned int sc __maybe_unused)
19327 + inode_lock_nested(hdir->hi_inode, sc);
19328 + au_hn_suspend(hdir);
19331 +#if 0 /* unused */
19332 +#include "vfsub.h"
19333 +static inline void au_hn_inode_lock_shared_nested(struct au_hinode *hdir,
19336 + vfsub_inode_lock_shared_nested(hdir->hi_inode, sc);
19337 + au_hn_suspend(hdir);
19341 +static inline void au_hn_inode_unlock(struct au_hinode *hdir)
19343 + au_hn_resume(hdir);
19344 + inode_unlock(hdir->hi_inode);
19347 +#endif /* __KERNEL__ */
19348 +#endif /* __AUFS_INODE_H__ */
19349 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
19350 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
19351 +++ linux/fs/aufs/ioctl.c 2018-01-29 07:56:20.056658502 +0100
19354 + * Copyright (C) 2005-2017 Junjiro R. Okajima
19356 + * This program, aufs is free software; you can redistribute it and/or modify
19357 + * it under the terms of the GNU General Public License as published by
19358 + * the Free Software Foundation; either version 2 of the License, or
19359 + * (at your option) any later version.
19361 + * This program is distributed in the hope that it will be useful,
19362 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19363 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19364 + * GNU General Public License for more details.
19366 + * You should have received a copy of the GNU General Public License
19367 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19372 + * plink-management and readdir in userspace.
19373 + * assist the pathconf(3) wrapper library.
19375 + * File-based Hierarchical Storage Management.
19378 +#include <linux/compat.h>
19379 +#include <linux/file.h>
19382 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
19385 + aufs_bindex_t wbi, bindex, bbot;
19386 + struct file *h_file;
19387 + struct super_block *sb;
19388 + struct dentry *root;
19389 + struct au_branch *br;
19390 + struct aufs_wbr_fd wbrfd = {
19391 + .oflags = au_dir_roflags,
19394 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
19395 + | O_NOATIME | O_CLOEXEC;
19397 + AuDebugOn(wbrfd.oflags & ~valid);
19400 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
19401 + if (unlikely(err)) {
19407 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
19408 + wbrfd.oflags |= au_dir_roflags;
19409 + AuDbg("0%o\n", wbrfd.oflags);
19410 + if (unlikely(wbrfd.oflags & ~valid))
19414 + fd = get_unused_fd_flags(0);
19416 + if (unlikely(fd < 0))
19419 + h_file = ERR_PTR(-EINVAL);
19422 + sb = path->dentry->d_sb;
19423 + root = sb->s_root;
19424 + aufs_read_lock(root, AuLock_IR);
19425 + bbot = au_sbbot(sb);
19426 + if (wbrfd.brid >= 0) {
19427 + wbi = au_br_index(sb, wbrfd.brid);
19428 + if (unlikely(wbi < 0 || wbi > bbot))
19432 + h_file = ERR_PTR(-ENOENT);
19433 + br = au_sbr(sb, wbi);
19434 + if (!au_br_writable(br->br_perm)) {
19438 + bindex = wbi + 1;
19440 + for (; bindex <= bbot; bindex++) {
19441 + br = au_sbr(sb, bindex);
19442 + if (au_br_writable(br->br_perm)) {
19444 + br = au_sbr(sb, wbi);
19449 + AuDbg("wbi %d\n", wbi);
19451 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
19455 + aufs_read_unlock(root, AuLock_IR);
19456 + err = PTR_ERR(h_file);
19457 + if (IS_ERR(h_file))
19460 + au_br_put(br); /* cf. au_h_open() */
19461 + fd_install(fd, h_file);
19463 + goto out; /* success */
19466 + put_unused_fd(fd);
19472 +/* ---------------------------------------------------------------------- */
19474 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
19477 + struct dentry *dentry;
19480 + case AUFS_CTL_RDU:
19481 + case AUFS_CTL_RDU_INO:
19482 + err = au_rdu_ioctl(file, cmd, arg);
19485 + case AUFS_CTL_WBR_FD:
19486 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
19489 + case AUFS_CTL_IBUSY:
19490 + err = au_ibusy_ioctl(file, arg);
19493 + case AUFS_CTL_BRINFO:
19494 + err = au_brinfo_ioctl(file, arg);
19497 + case AUFS_CTL_FHSM_FD:
19498 + dentry = file->f_path.dentry;
19499 + if (IS_ROOT(dentry))
19500 + err = au_fhsm_fd(dentry->d_sb, arg);
19506 + /* do not call the lower */
19507 + AuDbg("0x%x\n", cmd);
19515 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
19520 + case AUFS_CTL_MVDOWN:
19521 + err = au_mvdown(file->f_path.dentry, (void __user *)arg);
19524 + case AUFS_CTL_WBR_FD:
19525 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
19529 + /* do not call the lower */
19530 + AuDbg("0x%x\n", cmd);
19538 +#ifdef CONFIG_COMPAT
19539 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
19540 + unsigned long arg)
19545 + case AUFS_CTL_RDU:
19546 + case AUFS_CTL_RDU_INO:
19547 + err = au_rdu_compat_ioctl(file, cmd, arg);
19550 + case AUFS_CTL_IBUSY:
19551 + err = au_ibusy_compat_ioctl(file, arg);
19554 + case AUFS_CTL_BRINFO:
19555 + err = au_brinfo_compat_ioctl(file, arg);
19559 + err = aufs_ioctl_dir(file, cmd, arg);
19566 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
19567 + unsigned long arg)
19569 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
19572 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
19573 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
19574 +++ linux/fs/aufs/i_op_add.c 2018-01-29 07:56:20.056658502 +0100
19577 + * Copyright (C) 2005-2017 Junjiro R. Okajima
19579 + * This program, aufs is free software; you can redistribute it and/or modify
19580 + * it under the terms of the GNU General Public License as published by
19581 + * the Free Software Foundation; either version 2 of the License, or
19582 + * (at your option) any later version.
19584 + * This program is distributed in the hope that it will be useful,
19585 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19586 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19587 + * GNU General Public License for more details.
19589 + * You should have received a copy of the GNU General Public License
19590 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19594 + * inode operations (add entry)
19600 + * final procedure of adding a new entry, except link(2).
19601 + * remove whiteout, instantiate, copyup the parent dir's times and size
19602 + * and update version.
19603 + * if it failed, re-create the removed whiteout.
19605 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
19606 + struct dentry *wh_dentry, struct dentry *dentry)
19609 + aufs_bindex_t bwh;
19610 + struct path h_path;
19611 + struct super_block *sb;
19612 + struct inode *inode, *h_dir;
19613 + struct dentry *wh;
19618 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
19619 + IMustLock(h_dir);
19620 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
19621 + bwh = au_dbwh(dentry);
19622 + h_path.dentry = wh_dentry;
19623 + h_path.mnt = au_sbr_mnt(sb, bindex);
19624 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
19626 + if (unlikely(err))
19630 + inode = au_new_inode(dentry, /*must_new*/1);
19631 + if (!IS_ERR(inode)) {
19632 + d_instantiate(dentry, inode);
19633 + dir = d_inode(dentry->d_parent); /* dir inode is locked */
19635 + au_dir_ts(dir, bindex);
19636 + dir->i_version++;
19637 + au_fhsm_wrote(sb, bindex, /*force*/0);
19638 + return 0; /* success */
19641 + err = PTR_ERR(inode);
19646 + /* dir inode is locked */
19647 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
19648 + rerr = PTR_ERR(wh);
19649 + if (IS_ERR(wh)) {
19650 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
19651 + dentry, err, rerr);
19660 +static int au_d_may_add(struct dentry *dentry)
19665 + if (unlikely(d_unhashed(dentry)))
19667 + if (unlikely(d_really_is_positive(dentry)))
19673 + * simple tests for the adding inode operations.
19674 + * following the checks in vfs, plus the parent-child relationship.
19676 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
19677 + struct dentry *h_parent, int isdir)
19681 + struct dentry *h_dentry;
19682 + struct inode *h_inode;
19684 + err = -ENAMETOOLONG;
19685 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
19688 + h_dentry = au_h_dptr(dentry, bindex);
19689 + if (d_really_is_negative(dentry)) {
19691 + if (unlikely(d_is_positive(h_dentry)))
19694 + /* rename(2) case */
19696 + if (unlikely(d_is_negative(h_dentry)))
19698 + h_inode = d_inode(h_dentry);
19699 + if (unlikely(!h_inode->i_nlink))
19702 + h_mode = h_inode->i_mode;
19705 + if (unlikely(S_ISDIR(h_mode)))
19707 + } else if (unlikely(!S_ISDIR(h_mode))) {
19714 + /* expected parent dir is locked */
19715 + if (unlikely(h_parent != h_dentry->d_parent))
19724 + * initial procedure of adding a new entry.
19725 + * prepare writable branch and the parent dir, lock it,
19726 + * and lookup whiteout for the new entry.
19728 +static struct dentry*
19729 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
19730 + struct dentry *src_dentry, struct au_pin *pin,
19731 + struct au_wr_dir_args *wr_dir_args)
19733 + struct dentry *wh_dentry, *h_parent;
19734 + struct super_block *sb;
19735 + struct au_branch *br;
19737 + unsigned int udba;
19738 + aufs_bindex_t bcpup;
19740 + AuDbg("%pd\n", dentry);
19742 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
19744 + wh_dentry = ERR_PTR(err);
19745 + if (unlikely(err < 0))
19748 + sb = dentry->d_sb;
19749 + udba = au_opt_udba(sb);
19750 + err = au_pin(pin, dentry, bcpup, udba,
19751 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
19752 + wh_dentry = ERR_PTR(err);
19753 + if (unlikely(err))
19756 + h_parent = au_pinned_h_parent(pin);
19757 + if (udba != AuOpt_UDBA_NONE
19758 + && au_dbtop(dentry) == bcpup)
19759 + err = au_may_add(dentry, bcpup, h_parent,
19760 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
19761 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
19762 + err = -ENAMETOOLONG;
19763 + wh_dentry = ERR_PTR(err);
19764 + if (unlikely(err))
19767 + br = au_sbr(sb, bcpup);
19769 + struct path tmp = {
19770 + .dentry = h_parent,
19771 + .mnt = au_br_mnt(br)
19773 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
19776 + wh_dentry = NULL;
19777 + if (bcpup != au_dbwh(dentry))
19778 + goto out; /* success */
19781 + * ENAMETOOLONG here means that if we allowed create such name, then it
19782 + * would not be able to removed in the future. So we don't allow such
19783 + * name here and we don't handle ENAMETOOLONG differently here.
19785 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
19788 + if (IS_ERR(wh_dentry))
19791 + return wh_dentry;
19794 +/* ---------------------------------------------------------------------- */
19796 +enum { Mknod, Symlink, Creat };
19797 +struct simple_arg {
19804 + struct vfsub_aopen_args *aopen;
19807 + const char *symname;
19816 +static int add_simple(struct inode *dir, struct dentry *dentry,
19817 + struct simple_arg *arg)
19820 + aufs_bindex_t btop;
19821 + unsigned char created;
19822 + const unsigned char try_aopen
19823 + = (arg->type == Creat && arg->u.c.try_aopen);
19824 + struct dentry *wh_dentry, *parent;
19825 + struct inode *h_dir;
19826 + struct super_block *sb;
19827 + struct au_branch *br;
19828 + /* to reuduce stack size */
19830 + struct au_dtime dt;
19831 + struct au_pin pin;
19832 + struct path h_path;
19833 + struct au_wr_dir_args wr_dir_args;
19836 + AuDbg("%pd\n", dentry);
19840 + a = kmalloc(sizeof(*a), GFP_NOFS);
19841 + if (unlikely(!a))
19843 + a->wr_dir_args.force_btgt = -1;
19844 + a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
19846 + parent = dentry->d_parent; /* dir inode is locked */
19847 + if (!try_aopen) {
19848 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
19849 + if (unlikely(err))
19852 + err = au_d_may_add(dentry);
19853 + if (unlikely(err))
19856 + di_write_lock_parent(parent);
19857 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
19858 + &a->pin, &a->wr_dir_args);
19859 + err = PTR_ERR(wh_dentry);
19860 + if (IS_ERR(wh_dentry))
19863 + btop = au_dbtop(dentry);
19864 + sb = dentry->d_sb;
19865 + br = au_sbr(sb, btop);
19866 + a->h_path.dentry = au_h_dptr(dentry, btop);
19867 + a->h_path.mnt = au_br_mnt(br);
19868 + h_dir = au_pinned_h_dir(&a->pin);
19869 + switch (arg->type) {
19872 + if (!try_aopen || !h_dir->i_op->atomic_open)
19873 + err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
19874 + arg->u.c.want_excl);
19876 + err = vfsub_atomic_open(h_dir, a->h_path.dentry,
19877 + arg->u.c.aopen, br);
19880 + err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
19883 + err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
19891 + err = epilog(dir, btop, wh_dentry, dentry);
19894 + if (unlikely(created && err && d_is_positive(a->h_path.dentry))) {
19895 + /* no delegation since it is just created */
19896 + rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
19899 + AuIOErr("%pd revert failure(%d, %d)\n",
19900 + dentry, err, rerr);
19903 + au_dtime_revert(&a->dt);
19906 + if (!err && try_aopen && !h_dir->i_op->atomic_open)
19907 + *arg->u.c.aopen->opened |= FILE_CREATED;
19909 + au_unpin(&a->pin);
19914 + di_write_unlock(parent);
19916 + if (unlikely(err)) {
19917 + au_update_dbtop(dentry);
19921 + aufs_read_unlock(dentry, AuLock_DW);
19928 +int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
19931 + struct simple_arg arg = {
19938 + return add_simple(dir, dentry, &arg);
19941 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
19943 + struct simple_arg arg = {
19945 + .u.s.symname = symname
19947 + return add_simple(dir, dentry, &arg);
19950 +int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
19953 + struct simple_arg arg = {
19957 + .want_excl = want_excl
19960 + return add_simple(dir, dentry, &arg);
19963 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
19964 + struct vfsub_aopen_args *aopen_args)
19966 + struct simple_arg arg = {
19969 + .mode = aopen_args->create_mode,
19970 + .want_excl = aopen_args->open_flag & O_EXCL,
19971 + .try_aopen = true,
19972 + .aopen = aopen_args
19975 + return add_simple(dir, dentry, &arg);
19978 +int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
19981 + aufs_bindex_t bindex;
19982 + struct super_block *sb;
19983 + struct dentry *parent, *h_parent, *h_dentry;
19984 + struct inode *h_dir, *inode;
19985 + struct vfsmount *h_mnt;
19986 + struct au_wr_dir_args wr_dir_args = {
19987 + .force_btgt = -1,
19988 + .flags = AuWrDir_TMPFILE
19991 + /* copy-up may happen */
19995 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19996 + if (unlikely(err))
19999 + err = au_di_init(dentry);
20000 + if (unlikely(err))
20004 + parent = d_find_any_alias(dir);
20005 + AuDebugOn(!parent);
20006 + di_write_lock_parent(parent);
20007 + if (unlikely(d_inode(parent) != dir))
20010 + err = au_digen_test(parent, au_sigen(sb));
20011 + if (unlikely(err))
20014 + bindex = au_dbtop(parent);
20015 + au_set_dbtop(dentry, bindex);
20016 + au_set_dbbot(dentry, bindex);
20017 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
20019 + if (unlikely(err < 0))
20022 + err = -EOPNOTSUPP;
20023 + h_dir = au_h_iptr(dir, bindex);
20024 + if (unlikely(!h_dir->i_op->tmpfile))
20027 + h_mnt = au_sbr_mnt(sb, bindex);
20028 + err = vfsub_mnt_want_write(h_mnt);
20029 + if (unlikely(err))
20032 + h_parent = au_h_dptr(parent, bindex);
20033 + h_dentry = vfs_tmpfile(h_parent, mode, /*open_flag*/0);
20034 + if (IS_ERR(h_dentry)) {
20035 + err = PTR_ERR(h_dentry);
20039 + au_set_dbtop(dentry, bindex);
20040 + au_set_dbbot(dentry, bindex);
20041 + au_set_h_dptr(dentry, bindex, dget(h_dentry));
20042 + inode = au_new_inode(dentry, /*must_new*/1);
20043 + if (IS_ERR(inode)) {
20044 + err = PTR_ERR(inode);
20045 + au_set_h_dptr(dentry, bindex, NULL);
20046 + au_set_dbtop(dentry, -1);
20047 + au_set_dbbot(dentry, -1);
20049 + if (!inode->i_nlink)
20050 + set_nlink(inode, 1);
20051 + d_tmpfile(dentry, inode);
20052 + au_di(dentry)->di_tmpfile = 1;
20054 + /* update without i_mutex */
20055 + if (au_ibtop(dir) == au_dbtop(dentry))
20056 + au_cpup_attr_timesizes(dir);
20061 + vfsub_mnt_drop_write(h_mnt);
20063 + di_write_unlock(parent);
20065 + di_write_unlock(dentry);
20066 + if (unlikely(err)) {
20067 + au_di_fin(dentry);
20068 + dentry->d_fsdata = NULL;
20071 + si_read_unlock(sb);
20073 + inode_unlock(dir);
20077 +/* ---------------------------------------------------------------------- */
20079 +struct au_link_args {
20080 + aufs_bindex_t bdst, bsrc;
20081 + struct au_pin pin;
20082 + struct path h_path;
20083 + struct dentry *src_parent, *parent;
20086 +static int au_cpup_before_link(struct dentry *src_dentry,
20087 + struct au_link_args *a)
20090 + struct dentry *h_src_dentry;
20091 + struct au_cp_generic cpg = {
20092 + .dentry = src_dentry,
20097 + .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
20100 + di_read_lock_parent(a->src_parent, AuLock_IR);
20101 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
20102 + if (unlikely(err))
20105 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
20106 + err = au_pin(&a->pin, src_dentry, a->bdst,
20107 + au_opt_udba(src_dentry->d_sb),
20108 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20109 + if (unlikely(err))
20112 + err = au_sio_cpup_simple(&cpg);
20113 + au_unpin(&a->pin);
20116 + di_read_unlock(a->src_parent, AuLock_IR);
20120 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
20121 + struct au_link_args *a)
20124 + unsigned char plink;
20125 + aufs_bindex_t bbot;
20126 + struct dentry *h_src_dentry;
20127 + struct inode *h_inode, *inode, *delegated;
20128 + struct super_block *sb;
20129 + struct file *h_file;
20133 + sb = src_dentry->d_sb;
20134 + inode = d_inode(src_dentry);
20135 + if (au_ibtop(inode) <= a->bdst)
20136 + h_inode = au_h_iptr(inode, a->bdst);
20137 + if (!h_inode || !h_inode->i_nlink) {
20138 + /* copyup src_dentry as the name of dentry. */
20139 + bbot = au_dbbot(dentry);
20140 + if (bbot < a->bsrc)
20141 + au_set_dbbot(dentry, a->bsrc);
20142 + au_set_h_dptr(dentry, a->bsrc,
20143 + dget(au_h_dptr(src_dentry, a->bsrc)));
20144 + dget(a->h_path.dentry);
20145 + au_set_h_dptr(dentry, a->bdst, NULL);
20146 + AuDbg("temporary d_inode...\n");
20147 + spin_lock(&dentry->d_lock);
20148 + dentry->d_inode = d_inode(src_dentry); /* tmp */
20149 + spin_unlock(&dentry->d_lock);
20150 + h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
20151 + if (IS_ERR(h_file))
20152 + err = PTR_ERR(h_file);
20154 + struct au_cp_generic cpg = {
20155 + .dentry = dentry,
20160 + .flags = AuCpup_KEEPLINO
20162 + err = au_sio_cpup_simple(&cpg);
20163 + au_h_open_post(dentry, a->bsrc, h_file);
20165 + dput(a->h_path.dentry);
20166 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
20168 + au_set_h_dptr(dentry, a->bdst,
20169 + a->h_path.dentry);
20171 + spin_lock(&dentry->d_lock);
20172 + dentry->d_inode = NULL; /* restore */
20173 + spin_unlock(&dentry->d_lock);
20174 + AuDbg("temporary d_inode...done\n");
20175 + au_set_h_dptr(dentry, a->bsrc, NULL);
20176 + au_set_dbbot(dentry, bbot);
20178 + /* the inode of src_dentry already exists on a.bdst branch */
20179 + h_src_dentry = d_find_alias(h_inode);
20180 + if (!h_src_dentry && au_plink_test(inode)) {
20182 + h_src_dentry = au_plink_lkup(inode, a->bdst);
20183 + err = PTR_ERR(h_src_dentry);
20184 + if (IS_ERR(h_src_dentry))
20187 + if (unlikely(d_is_negative(h_src_dentry))) {
20188 + dput(h_src_dentry);
20189 + h_src_dentry = NULL;
20193 + if (h_src_dentry) {
20194 + delegated = NULL;
20195 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
20196 + &a->h_path, &delegated);
20197 + if (unlikely(err == -EWOULDBLOCK)) {
20198 + pr_warn("cannot retry for NFSv4 delegation"
20199 + " for an internal link\n");
20202 + dput(h_src_dentry);
20204 + AuIOErr("no dentry found for hi%lu on b%d\n",
20205 + h_inode->i_ino, a->bdst);
20210 + if (!err && !plink)
20211 + au_plink_append(inode, a->bdst, a->h_path.dentry);
20218 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
20219 + struct dentry *dentry)
20222 + struct au_dtime dt;
20223 + struct au_link_args *a;
20224 + struct dentry *wh_dentry, *h_src_dentry;
20225 + struct inode *inode, *delegated;
20226 + struct super_block *sb;
20227 + struct au_wr_dir_args wr_dir_args = {
20228 + /* .force_btgt = -1, */
20229 + .flags = AuWrDir_ADD_ENTRY
20233 + inode = d_inode(src_dentry);
20234 + IMustLock(inode);
20237 + a = kzalloc(sizeof(*a), GFP_NOFS);
20238 + if (unlikely(!a))
20241 + a->parent = dentry->d_parent; /* dir inode is locked */
20242 + err = aufs_read_and_write_lock2(dentry, src_dentry,
20243 + AuLock_NOPLM | AuLock_GEN);
20244 + if (unlikely(err))
20246 + err = au_d_linkable(src_dentry);
20247 + if (unlikely(err))
20249 + err = au_d_may_add(dentry);
20250 + if (unlikely(err))
20253 + a->src_parent = dget_parent(src_dentry);
20254 + wr_dir_args.force_btgt = au_ibtop(inode);
20256 + di_write_lock_parent(a->parent);
20257 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
20258 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
20260 + err = PTR_ERR(wh_dentry);
20261 + if (IS_ERR(wh_dentry))
20265 + sb = dentry->d_sb;
20266 + a->bdst = au_dbtop(dentry);
20267 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
20268 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
20269 + a->bsrc = au_ibtop(inode);
20270 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
20271 + if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
20272 + h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
20273 + if (!h_src_dentry) {
20274 + a->bsrc = au_dbtop(src_dentry);
20275 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
20276 + AuDebugOn(!h_src_dentry);
20277 + } else if (IS_ERR(h_src_dentry)) {
20278 + err = PTR_ERR(h_src_dentry);
20283 + * aufs doesn't touch the credential so
20284 + * security_dentry_create_files_as() is unnecrssary.
20286 + if (au_opt_test(au_mntflags(sb), PLINK)) {
20287 + if (a->bdst < a->bsrc
20288 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
20289 + err = au_cpup_or_link(src_dentry, dentry, a);
20291 + delegated = NULL;
20292 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
20293 + &a->h_path, &delegated);
20294 + if (unlikely(err == -EWOULDBLOCK)) {
20295 + pr_warn("cannot retry for NFSv4 delegation"
20296 + " for an internal link\n");
20300 + dput(h_src_dentry);
20303 + * copyup src_dentry to the branch we process,
20304 + * and then link(2) to it.
20306 + dput(h_src_dentry);
20307 + if (a->bdst < a->bsrc
20308 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
20309 + au_unpin(&a->pin);
20310 + di_write_unlock(a->parent);
20311 + err = au_cpup_before_link(src_dentry, a);
20312 + di_write_lock_parent(a->parent);
20314 + err = au_pin(&a->pin, dentry, a->bdst,
20316 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
20317 + if (unlikely(err))
20321 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
20323 + if (h_src_dentry && d_is_positive(h_src_dentry)) {
20324 + delegated = NULL;
20325 + err = vfsub_link(h_src_dentry,
20326 + au_pinned_h_dir(&a->pin),
20327 + &a->h_path, &delegated);
20328 + if (unlikely(err == -EWOULDBLOCK)) {
20329 + pr_warn("cannot retry"
20330 + " for NFSv4 delegation"
20331 + " for an internal link\n");
20337 + if (unlikely(err))
20341 + a->h_path.dentry = wh_dentry;
20342 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
20344 + if (unlikely(err))
20348 + au_dir_ts(dir, a->bdst);
20349 + dir->i_version++;
20350 + inc_nlink(inode);
20351 + inode->i_ctime = dir->i_ctime;
20352 + d_instantiate(dentry, au_igrab(inode));
20353 + if (d_unhashed(a->h_path.dentry))
20354 + /* some filesystem calls d_drop() */
20356 + /* some filesystems consume an inode even hardlink */
20357 + au_fhsm_wrote(sb, a->bdst, /*force*/0);
20358 + goto out_unpin; /* success */
20361 + /* no delegation since it is just created */
20362 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
20363 + /*delegated*/NULL, /*force*/0);
20364 + if (unlikely(rerr)) {
20365 + AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
20368 + au_dtime_revert(&dt);
20370 + au_unpin(&a->pin);
20374 + di_write_unlock(a->parent);
20375 + dput(a->src_parent);
20377 + if (unlikely(err)) {
20378 + au_update_dbtop(dentry);
20381 + aufs_read_and_write_unlock2(dentry, src_dentry);
20389 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
20392 + aufs_bindex_t bindex;
20393 + unsigned char diropq;
20394 + struct path h_path;
20395 + struct dentry *wh_dentry, *parent, *opq_dentry;
20396 + struct inode *h_inode;
20397 + struct super_block *sb;
20399 + struct au_pin pin;
20400 + struct au_dtime dt;
20401 + } *a; /* reduce the stack usage */
20402 + struct au_wr_dir_args wr_dir_args = {
20403 + .force_btgt = -1,
20404 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
20410 + a = kmalloc(sizeof(*a), GFP_NOFS);
20411 + if (unlikely(!a))
20414 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
20415 + if (unlikely(err))
20417 + err = au_d_may_add(dentry);
20418 + if (unlikely(err))
20421 + parent = dentry->d_parent; /* dir inode is locked */
20422 + di_write_lock_parent(parent);
20423 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
20424 + &a->pin, &wr_dir_args);
20425 + err = PTR_ERR(wh_dentry);
20426 + if (IS_ERR(wh_dentry))
20429 + sb = dentry->d_sb;
20430 + bindex = au_dbtop(dentry);
20431 + h_path.dentry = au_h_dptr(dentry, bindex);
20432 + h_path.mnt = au_sbr_mnt(sb, bindex);
20433 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
20434 + if (unlikely(err))
20437 + /* make the dir opaque */
20439 + h_inode = d_inode(h_path.dentry);
20441 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
20442 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
20443 + opq_dentry = au_diropq_create(dentry, bindex);
20444 + inode_unlock(h_inode);
20445 + err = PTR_ERR(opq_dentry);
20446 + if (IS_ERR(opq_dentry))
20448 + dput(opq_dentry);
20452 + err = epilog(dir, bindex, wh_dentry, dentry);
20455 + goto out_unpin; /* success */
20460 + AuLabel(revert opq);
20461 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
20462 + rerr = au_diropq_remove(dentry, bindex);
20463 + inode_unlock(h_inode);
20465 + AuIOErr("%pd reverting diropq failed(%d, %d)\n",
20466 + dentry, err, rerr);
20472 + AuLabel(revert dir);
20473 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
20475 + AuIOErr("%pd reverting dir failed(%d, %d)\n",
20476 + dentry, err, rerr);
20479 + au_dtime_revert(&a->dt);
20481 + au_unpin(&a->pin);
20484 + di_write_unlock(parent);
20486 + if (unlikely(err)) {
20487 + au_update_dbtop(dentry);
20490 + aufs_read_unlock(dentry, AuLock_DW);
20496 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
20497 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
20498 +++ linux/fs/aufs/i_op.c 2018-01-29 07:56:20.056658502 +0100
20501 + * Copyright (C) 2005-2017 Junjiro R. Okajima
20503 + * This program, aufs is free software; you can redistribute it and/or modify
20504 + * it under the terms of the GNU General Public License as published by
20505 + * the Free Software Foundation; either version 2 of the License, or
20506 + * (at your option) any later version.
20508 + * This program is distributed in the hope that it will be useful,
20509 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20510 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20511 + * GNU General Public License for more details.
20513 + * You should have received a copy of the GNU General Public License
20514 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20518 + * inode operations (except add/del/rename)
20521 +#include <linux/device_cgroup.h>
20522 +#include <linux/fs_stack.h>
20523 +#include <linux/namei.h>
20524 +#include <linux/security.h>
20527 +static int h_permission(struct inode *h_inode, int mask,
20528 + struct path *h_path, int brperm)
20531 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
20534 + if (write_mask && IS_IMMUTABLE(h_inode))
20538 + if (((mask & MAY_EXEC)
20539 + && S_ISREG(h_inode->i_mode)
20540 + && (path_noexec(h_path)
20541 + || !(h_inode->i_mode & S_IXUGO))))
20545 + * - skip the lower fs test in the case of write to ro branch.
20546 + * - nfs dir permission write check is optimized, but a policy for
20547 + * link/rename requires a real check.
20548 + * - nfs always sets SB_POSIXACL regardless its mount option 'noacl.'
20549 + * in this case, generic_permission() returns -EOPNOTSUPP.
20551 + if ((write_mask && !au_br_writable(brperm))
20552 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
20553 + && write_mask && !(mask & MAY_READ))
20554 + || !h_inode->i_op->permission) {
20555 + /* AuLabel(generic_permission); */
20556 + /* AuDbg("get_acl %pf\n", h_inode->i_op->get_acl); */
20557 + err = generic_permission(h_inode, mask);
20558 + if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
20559 + err = h_inode->i_op->permission(h_inode, mask);
20562 + /* AuLabel(h_inode->permission); */
20563 + err = h_inode->i_op->permission(h_inode, mask);
20568 + err = devcgroup_inode_permission(h_inode, mask);
20570 + err = security_inode_permission(h_inode, mask);
20574 + /* todo: do we need to call ima_path_check()? */
20575 + struct path h_path = {
20579 + err = ima_path_check(&h_path,
20580 + mask & (MAY_READ | MAY_WRITE | MAY_EXEC),
20581 + IMA_COUNT_LEAVE);
20589 +static int aufs_permission(struct inode *inode, int mask)
20592 + aufs_bindex_t bindex, bbot;
20593 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
20594 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
20595 + struct inode *h_inode;
20596 + struct super_block *sb;
20597 + struct au_branch *br;
20599 + /* todo: support rcu-walk? */
20600 + if (mask & MAY_NOT_BLOCK)
20603 + sb = inode->i_sb;
20604 + si_read_lock(sb, AuLock_FLUSH);
20605 + ii_read_lock_child(inode);
20607 + err = au_iigen_test(inode, au_sigen(sb));
20608 + if (unlikely(err))
20614 + || au_opt_test(au_mntflags(sb), DIRPERM1)) {
20615 + err = au_busy_or_stale();
20616 + h_inode = au_h_iptr(inode, au_ibtop(inode));
20617 + if (unlikely(!h_inode
20618 + || (h_inode->i_mode & S_IFMT)
20619 + != (inode->i_mode & S_IFMT)))
20623 + bindex = au_ibtop(inode);
20624 + br = au_sbr(sb, bindex);
20625 + err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
20628 + && !special_file(h_inode->i_mode)) {
20629 + /* test whether the upper writable branch exists */
20631 + for (; bindex >= 0; bindex--)
20632 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
20640 + /* non-write to dir */
20642 + bbot = au_ibbot(inode);
20643 + for (bindex = au_ibtop(inode); !err && bindex <= bbot; bindex++) {
20644 + h_inode = au_h_iptr(inode, bindex);
20646 + err = au_busy_or_stale();
20647 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
20650 + br = au_sbr(sb, bindex);
20651 + err = h_permission(h_inode, mask, &br->br_path,
20657 + ii_read_unlock(inode);
20658 + si_read_unlock(sb);
20662 +/* ---------------------------------------------------------------------- */
20664 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
20665 + unsigned int flags)
20667 + struct dentry *ret, *parent;
20668 + struct inode *inode;
20669 + struct super_block *sb;
20670 + int err, npositive;
20674 + /* todo: support rcu-walk? */
20675 + ret = ERR_PTR(-ECHILD);
20676 + if (flags & LOOKUP_RCU)
20679 + ret = ERR_PTR(-ENAMETOOLONG);
20680 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
20684 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
20685 + ret = ERR_PTR(err);
20686 + if (unlikely(err))
20689 + err = au_di_init(dentry);
20690 + ret = ERR_PTR(err);
20691 + if (unlikely(err))
20695 + npositive = 0; /* suppress a warning */
20696 + parent = dentry->d_parent; /* dir inode is locked */
20697 + di_read_lock_parent(parent, AuLock_IR);
20698 + err = au_alive_dir(parent);
20700 + err = au_digen_test(parent, au_sigen(sb));
20702 + /* regardless LOOKUP_CREATE, always ALLOW_NEG */
20703 + npositive = au_lkup_dentry(dentry, au_dbtop(parent),
20704 + AuLkup_ALLOW_NEG);
20707 + di_read_unlock(parent, AuLock_IR);
20708 + ret = ERR_PTR(err);
20709 + if (unlikely(err < 0))
20713 + inode = au_new_inode(dentry, /*must_new*/0);
20714 + if (IS_ERR(inode)) {
20715 + ret = (void *)inode;
20722 + atomic_inc(&inode->i_count);
20723 + ret = d_splice_alias(inode, dentry);
20725 + if (unlikely(d_need_lookup(dentry))) {
20726 + spin_lock(&dentry->d_lock);
20727 + dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
20728 + spin_unlock(&dentry->d_lock);
20732 + if (!IS_ERR(ret)) {
20734 + if (ret && ret != dentry)
20735 + ii_write_unlock(inode);
20737 + ii_write_unlock(inode);
20744 + di_write_unlock(dentry);
20746 + si_read_unlock(sb);
20751 +/* ---------------------------------------------------------------------- */
20753 +struct aopen_node {
20754 + struct hlist_bl_node hblist;
20755 + struct file *file, *h_file;
20758 +static int au_do_aopen(struct inode *inode, struct file *file)
20760 + struct hlist_bl_head *aopen;
20761 + struct hlist_bl_node *pos;
20762 + struct aopen_node *node;
20763 + struct au_do_open_args args = {
20765 + .open = au_do_open_nondir
20768 + aopen = &au_sbi(inode->i_sb)->si_aopen;
20769 + hlist_bl_lock(aopen);
20770 + hlist_bl_for_each_entry(node, pos, aopen, hblist)
20771 + if (node->file == file) {
20772 + args.h_file = node->h_file;
20775 + hlist_bl_unlock(aopen);
20776 + /* AuDebugOn(!args.h_file); */
20778 + return au_do_open(file, &args);
20781 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
20782 + struct file *file, unsigned int open_flag,
20783 + umode_t create_mode, int *opened)
20785 + int err, unlocked, h_opened = *opened;
20786 + unsigned int lkup_flags;
20787 + struct dentry *parent, *d;
20788 + struct hlist_bl_head *aopen;
20789 + struct vfsub_aopen_args args = {
20790 + .open_flag = open_flag,
20791 + .create_mode = create_mode,
20792 + .opened = &h_opened
20794 + struct aopen_node aopen_node = {
20799 + AuDbg("open_flag 0%o\n", open_flag);
20800 + AuDbgDentry(dentry);
20803 + if (!au_di(dentry)) {
20804 + lkup_flags = LOOKUP_OPEN;
20805 + if (open_flag & O_CREAT)
20806 + lkup_flags |= LOOKUP_CREATE;
20807 + d = aufs_lookup(dir, dentry, lkup_flags);
20809 + err = PTR_ERR(d);
20814 + * obsoleted dentry found.
20815 + * another error will be returned later.
20821 + AuDbgDentry(dentry);
20824 + if (d_is_positive(dentry)
20825 + || d_unhashed(dentry)
20826 + || d_unlinked(dentry)
20827 + || !(open_flag & O_CREAT))
20828 + goto out_no_open;
20831 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
20832 + if (unlikely(err))
20835 + parent = dentry->d_parent; /* dir is locked */
20836 + di_write_lock_parent(parent);
20837 + err = au_lkup_dentry(dentry, /*btop*/0, AuLkup_ALLOW_NEG);
20838 + if (unlikely(err))
20841 + AuDbgDentry(dentry);
20842 + if (d_is_positive(dentry))
20845 + args.file = get_empty_filp();
20846 + err = PTR_ERR(args.file);
20847 + if (IS_ERR(args.file))
20850 + args.file->f_flags = file->f_flags;
20851 + err = au_aopen_or_create(dir, dentry, &args);
20853 + AuDbgFile(args.file);
20854 + if (unlikely(err < 0)) {
20855 + if (h_opened & FILE_OPENED)
20858 + put_filp(args.file);
20861 + di_write_unlock(parent);
20862 + di_write_unlock(dentry);
20865 + /* some filesystems don't set FILE_CREATED while succeeded? */
20866 + *opened |= FILE_CREATED;
20867 + if (h_opened & FILE_OPENED)
20868 + aopen_node.h_file = args.file;
20870 + put_filp(args.file);
20871 + args.file = NULL;
20873 + aopen = &au_sbi(dir->i_sb)->si_aopen;
20874 + au_hbl_add(&aopen_node.hblist, aopen);
20875 + err = finish_open(file, dentry, au_do_aopen, opened);
20876 + au_hbl_del(&aopen_node.hblist, aopen);
20879 + if (aopen_node.h_file)
20880 + fput(aopen_node.h_file);
20884 + si_read_unlock(dentry->d_sb);
20886 + di_write_unlock(parent);
20887 + aufs_read_unlock(dentry, AuLock_DW);
20889 + AuDbgDentry(dentry);
20890 + if (unlikely(err < 0))
20893 + if (err >= 0 && !(*opened & FILE_CREATED)) {
20894 + AuLabel(out_no_open);
20896 + err = finish_no_open(file, dentry);
20899 + AuDbg("%pd%s%s\n", dentry,
20900 + (*opened & FILE_CREATED) ? " created" : "",
20901 + (*opened & FILE_OPENED) ? " opened" : "");
20907 +/* ---------------------------------------------------------------------- */
20909 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
20910 + const unsigned char add_entry, aufs_bindex_t bcpup,
20911 + aufs_bindex_t btop)
20914 + struct dentry *h_parent;
20915 + struct inode *h_dir;
20918 + IMustLock(d_inode(parent));
20920 + di_write_lock_parent(parent);
20923 + if (!au_h_dptr(parent, bcpup)) {
20924 + if (btop > bcpup)
20925 + err = au_cpup_dirs(dentry, bcpup);
20926 + else if (btop < bcpup)
20927 + err = au_cpdown_dirs(dentry, bcpup);
20931 + if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
20932 + h_parent = au_h_dptr(parent, bcpup);
20933 + h_dir = d_inode(h_parent);
20934 + vfsub_inode_lock_shared_nested(h_dir, AuLsc_I_PARENT);
20935 + err = au_lkup_neg(dentry, bcpup, /*wh*/0);
20936 + /* todo: no unlock here */
20937 + inode_unlock_shared(h_dir);
20939 + AuDbg("bcpup %d\n", bcpup);
20941 + if (d_really_is_negative(dentry))
20942 + au_set_h_dptr(dentry, btop, NULL);
20943 + au_update_dbrange(dentry, /*do_put_zero*/0);
20948 + di_write_unlock(parent);
20950 + err = bcpup; /* success */
20957 + * decide the branch and the parent dir where we will create a new entry.
20958 + * returns new bindex or an error.
20959 + * copyup the parent dir if needed.
20961 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
20962 + struct au_wr_dir_args *args)
20965 + unsigned int flags;
20966 + aufs_bindex_t bcpup, btop, src_btop;
20967 + const unsigned char add_entry
20968 + = au_ftest_wrdir(args->flags, ADD_ENTRY)
20969 + | au_ftest_wrdir(args->flags, TMPFILE);
20970 + struct super_block *sb;
20971 + struct dentry *parent;
20972 + struct au_sbinfo *sbinfo;
20974 + sb = dentry->d_sb;
20975 + sbinfo = au_sbi(sb);
20976 + parent = dget_parent(dentry);
20977 + btop = au_dbtop(dentry);
20979 + if (args->force_btgt < 0) {
20980 + if (src_dentry) {
20981 + src_btop = au_dbtop(src_dentry);
20982 + if (src_btop < btop)
20983 + bcpup = src_btop;
20984 + } else if (add_entry) {
20986 + if (au_ftest_wrdir(args->flags, ISDIR))
20987 + au_fset_wbr(flags, DIR);
20988 + err = AuWbrCreate(sbinfo, dentry, flags);
20992 + if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
20994 + err = AuWbrCopyup(sbinfo, dentry);
20996 + if (!IS_ROOT(dentry)) {
20997 + di_read_lock_parent(parent, !AuLock_IR);
20998 + err = AuWbrCopyup(sbinfo, dentry);
20999 + di_read_unlock(parent, !AuLock_IR);
21001 + err = AuWbrCopyup(sbinfo, dentry);
21004 + if (unlikely(err < 0))
21008 + bcpup = args->force_btgt;
21009 + AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
21012 + AuDbg("btop %d, bcpup %d\n", btop, bcpup);
21014 + if (bcpup == btop)
21015 + goto out; /* success */
21017 + /* copyup the new parent into the branch we process */
21018 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, btop);
21020 + if (d_really_is_negative(dentry)) {
21021 + au_set_h_dptr(dentry, btop, NULL);
21022 + au_set_dbtop(dentry, bcpup);
21023 + au_set_dbbot(dentry, bcpup);
21025 + AuDebugOn(add_entry
21026 + && !au_ftest_wrdir(args->flags, TMPFILE)
21027 + && !au_h_dptr(dentry, bcpup));
21035 +/* ---------------------------------------------------------------------- */
21037 +void au_pin_hdir_unlock(struct au_pin *p)
21040 + au_hn_inode_unlock(p->hdir);
21043 +int au_pin_hdir_lock(struct au_pin *p)
21051 + /* even if an error happens later, keep this lock */
21052 + au_hn_inode_lock_nested(p->hdir, p->lsc_hi);
21055 + if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
21060 + err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
21061 + p->h_parent, p->br);
21067 +int au_pin_hdir_relock(struct au_pin *p)
21070 + struct inode *h_i;
21071 + struct dentry *h_d[] = {
21076 + err = au_pin_hdir_lock(p);
21077 + if (unlikely(err))
21080 + for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
21083 + if (d_is_positive(h_d[i])) {
21084 + h_i = d_inode(h_d[i]);
21085 + err = !h_i->i_nlink;
21093 +static void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
21095 +#if !defined(CONFIG_RWSEM_GENERIC_SPINLOCK) && defined(CONFIG_RWSEM_SPIN_ON_OWNER)
21096 + p->hdir->hi_inode->i_rwsem.owner = task;
21100 +void au_pin_hdir_acquire_nest(struct au_pin *p)
21103 + rwsem_acquire_nest(&p->hdir->hi_inode->i_rwsem.dep_map,
21104 + p->lsc_hi, 0, NULL, _RET_IP_);
21105 + au_pin_hdir_set_owner(p, current);
21109 +void au_pin_hdir_release(struct au_pin *p)
21112 + au_pin_hdir_set_owner(p, p->task);
21113 + rwsem_release(&p->hdir->hi_inode->i_rwsem.dep_map, 1, _RET_IP_);
21117 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
21119 + if (pin && pin->parent)
21120 + return au_h_dptr(pin->parent, pin->bindex);
21124 +void au_unpin(struct au_pin *p)
21127 + au_pin_hdir_unlock(p);
21128 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
21129 + vfsub_mnt_drop_write(p->h_mnt);
21133 + if (!au_ftest_pin(p->flags, DI_LOCKED))
21134 + di_read_unlock(p->parent, AuLock_IR);
21135 + iput(p->hdir->hi_inode);
21137 + p->parent = NULL;
21140 + /* do not clear p->task */
21143 +int au_do_pin(struct au_pin *p)
21146 + struct super_block *sb;
21147 + struct inode *h_dir;
21150 + sb = p->dentry->d_sb;
21151 + p->br = au_sbr(sb, p->bindex);
21152 + if (IS_ROOT(p->dentry)) {
21153 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
21154 + p->h_mnt = au_br_mnt(p->br);
21155 + err = vfsub_mnt_want_write(p->h_mnt);
21156 + if (unlikely(err)) {
21157 + au_fclr_pin(p->flags, MNT_WRITE);
21164 + p->h_dentry = NULL;
21165 + if (p->bindex <= au_dbbot(p->dentry))
21166 + p->h_dentry = au_h_dptr(p->dentry, p->bindex);
21168 + p->parent = dget_parent(p->dentry);
21169 + if (!au_ftest_pin(p->flags, DI_LOCKED))
21170 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
21173 + p->h_parent = au_h_dptr(p->parent, p->bindex);
21174 + p->hdir = au_hi(d_inode(p->parent), p->bindex);
21176 + h_dir = p->hdir->hi_inode;
21180 + * if DI_LOCKED is not set, then p->parent may be different
21181 + * and h_parent can be NULL.
21183 + if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
21185 + if (!au_ftest_pin(p->flags, DI_LOCKED))
21186 + di_read_unlock(p->parent, AuLock_IR);
21188 + p->parent = NULL;
21192 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
21193 + p->h_mnt = au_br_mnt(p->br);
21194 + err = vfsub_mnt_want_write(p->h_mnt);
21195 + if (unlikely(err)) {
21196 + au_fclr_pin(p->flags, MNT_WRITE);
21197 + if (!au_ftest_pin(p->flags, DI_LOCKED))
21198 + di_read_unlock(p->parent, AuLock_IR);
21200 + p->parent = NULL;
21206 + err = au_pin_hdir_lock(p);
21208 + goto out; /* success */
21213 + pr_err("err %d\n", err);
21214 + err = au_busy_or_stale();
21219 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
21220 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
21221 + unsigned int udba, unsigned char flags)
21223 + p->dentry = dentry;
21225 + p->lsc_di = lsc_di;
21226 + p->lsc_hi = lsc_hi;
21227 + p->flags = flags;
21228 + p->bindex = bindex;
21230 + p->parent = NULL;
21234 + p->h_dentry = NULL;
21235 + p->h_parent = NULL;
21237 + p->task = current;
21240 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
21241 + unsigned int udba, unsigned char flags)
21243 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
21245 + return au_do_pin(pin);
21248 +/* ---------------------------------------------------------------------- */
21251 + * ->setattr() and ->getattr() are called in various cases.
21252 + * chmod, stat: dentry is revalidated.
21253 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
21255 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
21257 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
21258 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
21261 + struct dentry *parent;
21264 + if (au_digen_test(dentry, sigen)) {
21265 + parent = dget_parent(dentry);
21266 + di_read_lock_parent(parent, AuLock_IR);
21267 + err = au_refresh_dentry(dentry, parent);
21268 + di_read_unlock(parent, AuLock_IR);
21276 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
21277 + struct au_icpup_args *a)
21281 + aufs_bindex_t btop, ibtop;
21282 + struct dentry *hi_wh, *parent;
21283 + struct inode *inode;
21284 + struct au_wr_dir_args wr_dir_args = {
21285 + .force_btgt = -1,
21289 + if (d_is_dir(dentry))
21290 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
21291 + /* plink or hi_wh() case */
21292 + btop = au_dbtop(dentry);
21293 + inode = d_inode(dentry);
21294 + ibtop = au_ibtop(inode);
21295 + if (btop != ibtop && !au_test_ro(inode->i_sb, ibtop, inode))
21296 + wr_dir_args.force_btgt = ibtop;
21297 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
21298 + if (unlikely(err < 0))
21302 + au_fset_icpup(a->flags, DID_CPUP);
21305 + a->pin_flags = AuPin_MNT_WRITE;
21307 + if (!IS_ROOT(dentry)) {
21308 + au_fset_pin(a->pin_flags, DI_LOCKED);
21309 + parent = dget_parent(dentry);
21310 + di_write_lock_parent(parent);
21313 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
21314 + if (unlikely(err))
21318 + a->h_path.dentry = au_h_dptr(dentry, btop);
21319 + a->h_inode = d_inode(a->h_path.dentry);
21320 + if (ia && (ia->ia_valid & ATTR_SIZE)) {
21321 + vfsub_inode_lock_shared_nested(a->h_inode, AuLsc_I_CHILD);
21322 + if (ia->ia_size < i_size_read(a->h_inode))
21323 + sz = ia->ia_size;
21324 + inode_unlock_shared(a->h_inode);
21328 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
21329 + hi_wh = au_hi_wh(inode, a->btgt);
21331 + struct au_cp_generic cpg = {
21332 + .dentry = dentry,
21338 + err = au_sio_cpup_wh(&cpg, /*file*/NULL);
21339 + if (unlikely(err))
21341 + hi_wh = au_hi_wh(inode, a->btgt);
21342 + /* todo: revalidate hi_wh? */
21347 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
21348 + di_downgrade_lock(parent, AuLock_IR);
21352 + if (!au_ftest_icpup(a->flags, DID_CPUP))
21353 + goto out; /* success */
21355 + if (!d_unhashed(dentry)) {
21356 + struct au_cp_generic cpg = {
21357 + .dentry = dentry,
21362 + .flags = AuCpup_DTIME | AuCpup_HOPEN
21364 + err = au_sio_cpup_simple(&cpg);
21366 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
21367 + } else if (!hi_wh)
21368 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
21370 + a->h_path.dentry = hi_wh; /* do not dget here */
21373 + a->h_inode = d_inode(a->h_path.dentry);
21375 + goto out; /* success */
21376 + au_unpin(&a->pin);
21379 + di_write_unlock(parent);
21384 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
21388 +static int aufs_setattr(struct dentry *dentry, struct iattr *ia)
21391 + struct inode *inode, *delegated;
21392 + struct super_block *sb;
21393 + struct file *file;
21394 + struct au_icpup_args *a;
21396 + inode = d_inode(dentry);
21397 + IMustLock(inode);
21399 + err = setattr_prepare(dentry, ia);
21400 + if (unlikely(err))
21404 + a = kzalloc(sizeof(*a), GFP_NOFS);
21405 + if (unlikely(!a))
21408 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
21409 + ia->ia_valid &= ~ATTR_MODE;
21412 + sb = dentry->d_sb;
21413 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21414 + if (unlikely(err))
21417 + if (ia->ia_valid & ATTR_FILE) {
21418 + /* currently ftruncate(2) only */
21419 + AuDebugOn(!d_is_reg(dentry));
21420 + file = ia->ia_file;
21421 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1,
21423 + if (unlikely(err))
21425 + ia->ia_file = au_hf_top(file);
21426 + a->udba = AuOpt_UDBA_NONE;
21428 + /* fchmod() doesn't pass ia_file */
21429 + a->udba = au_opt_udba(sb);
21430 + di_write_lock_child(dentry);
21431 + /* no d_unlinked(), to set UDBA_NONE for root */
21432 + if (d_unhashed(dentry))
21433 + a->udba = AuOpt_UDBA_NONE;
21434 + if (a->udba != AuOpt_UDBA_NONE) {
21435 + AuDebugOn(IS_ROOT(dentry));
21436 + err = au_reval_for_attr(dentry, au_sigen(sb));
21437 + if (unlikely(err))
21442 + err = au_pin_and_icpup(dentry, ia, a);
21443 + if (unlikely(err < 0))
21445 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
21446 + ia->ia_file = NULL;
21447 + ia->ia_valid &= ~ATTR_FILE;
21450 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
21451 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
21452 + == (ATTR_MODE | ATTR_CTIME)) {
21453 + err = security_path_chmod(&a->h_path, ia->ia_mode);
21454 + if (unlikely(err))
21456 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
21457 + && (ia->ia_valid & ATTR_CTIME)) {
21458 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
21459 + if (unlikely(err))
21463 + if (ia->ia_valid & ATTR_SIZE) {
21466 + if (ia->ia_size < i_size_read(inode))
21468 + truncate_setsize(inode, ia->ia_size);
21471 + if (ia->ia_valid & ATTR_FILE)
21473 + inode_unlock(a->h_inode);
21474 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
21475 + inode_lock_nested(a->h_inode, AuLsc_I_CHILD);
21477 + delegated = NULL;
21479 + err = vfsub_notify_change(&a->h_path, ia, &delegated);
21481 + err = break_deleg_wait(&delegated);
21489 + * regardless aufs 'acl' option setting.
21490 + * why don't all acl-aware fs call this func from their ->setattr()?
21492 + if (!err && (ia->ia_valid & ATTR_MODE))
21493 + err = vfsub_acl_chmod(a->h_inode, ia->ia_mode);
21495 + au_cpup_attr_changeable(inode);
21498 + inode_unlock(a->h_inode);
21499 + au_unpin(&a->pin);
21500 + if (unlikely(err))
21501 + au_update_dbtop(dentry);
21503 + di_write_unlock(dentry);
21505 + fi_write_unlock(file);
21506 + ia->ia_file = file;
21507 + ia->ia_valid |= ATTR_FILE;
21510 + si_read_unlock(sb);
21518 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
21519 +static int au_h_path_to_set_attr(struct dentry *dentry,
21520 + struct au_icpup_args *a, struct path *h_path)
21523 + struct super_block *sb;
21525 + sb = dentry->d_sb;
21526 + a->udba = au_opt_udba(sb);
21527 + /* no d_unlinked(), to set UDBA_NONE for root */
21528 + if (d_unhashed(dentry))
21529 + a->udba = AuOpt_UDBA_NONE;
21530 + if (a->udba != AuOpt_UDBA_NONE) {
21531 + AuDebugOn(IS_ROOT(dentry));
21532 + err = au_reval_for_attr(dentry, au_sigen(sb));
21533 + if (unlikely(err))
21536 + err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
21537 + if (unlikely(err < 0))
21540 + h_path->dentry = a->h_path.dentry;
21541 + h_path->mnt = au_sbr_mnt(sb, a->btgt);
21547 +ssize_t au_sxattr(struct dentry *dentry, struct inode *inode,
21548 + struct au_sxattr *arg)
21551 + struct path h_path;
21552 + struct super_block *sb;
21553 + struct au_icpup_args *a;
21554 + struct inode *h_inode;
21556 + IMustLock(inode);
21559 + a = kzalloc(sizeof(*a), GFP_NOFS);
21560 + if (unlikely(!a))
21563 + sb = dentry->d_sb;
21564 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21565 + if (unlikely(err))
21568 + h_path.dentry = NULL; /* silence gcc */
21569 + di_write_lock_child(dentry);
21570 + err = au_h_path_to_set_attr(dentry, a, &h_path);
21571 + if (unlikely(err))
21574 + inode_unlock(a->h_inode);
21575 + switch (arg->type) {
21576 + case AU_XATTR_SET:
21577 + AuDebugOn(d_is_negative(h_path.dentry));
21578 + err = vfsub_setxattr(h_path.dentry,
21579 + arg->u.set.name, arg->u.set.value,
21580 + arg->u.set.size, arg->u.set.flags);
21583 + err = -EOPNOTSUPP;
21584 + h_inode = d_inode(h_path.dentry);
21585 + if (h_inode->i_op->set_acl)
21586 + /* this will call posix_acl_update_mode */
21587 + err = h_inode->i_op->set_acl(h_inode,
21588 + arg->u.acl_set.acl,
21589 + arg->u.acl_set.type);
21593 + au_cpup_attr_timesizes(inode);
21595 + au_unpin(&a->pin);
21596 + if (unlikely(err))
21597 + au_update_dbtop(dentry);
21600 + di_write_unlock(dentry);
21601 + si_read_unlock(sb);
21610 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
21611 + unsigned int nlink)
21615 + inode->i_mode = st->mode;
21616 + /* don't i_[ug]id_write() here */
21617 + inode->i_uid = st->uid;
21618 + inode->i_gid = st->gid;
21619 + inode->i_atime = st->atime;
21620 + inode->i_mtime = st->mtime;
21621 + inode->i_ctime = st->ctime;
21623 + au_cpup_attr_nlink(inode, /*force*/0);
21624 + if (S_ISDIR(inode->i_mode)) {
21625 + n = inode->i_nlink;
21628 + smp_mb(); /* for i_nlink */
21629 + /* 0 can happen */
21630 + set_nlink(inode, n);
21633 + spin_lock(&inode->i_lock);
21634 + inode->i_blocks = st->blocks;
21635 + i_size_write(inode, st->size);
21636 + spin_unlock(&inode->i_lock);
21640 + * common routine for aufs_getattr() and au_getxattr().
21641 + * returns zero or negative (an error).
21642 + * @dentry will be read-locked in success.
21644 +int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path,
21648 + unsigned int mnt_flags, sigen;
21649 + unsigned char udba_none;
21650 + aufs_bindex_t bindex;
21651 + struct super_block *sb, *h_sb;
21652 + struct inode *inode;
21654 + h_path->mnt = NULL;
21655 + h_path->dentry = NULL;
21658 + sb = dentry->d_sb;
21659 + mnt_flags = au_mntflags(sb);
21660 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
21662 + if (unlikely(locked))
21663 + goto body; /* skip locking dinfo */
21665 + /* support fstat(2) */
21666 + if (!d_unlinked(dentry) && !udba_none) {
21667 + sigen = au_sigen(sb);
21668 + err = au_digen_test(dentry, sigen);
21670 + di_read_lock_child(dentry, AuLock_IR);
21671 + err = au_dbrange_test(dentry);
21672 + if (unlikely(err)) {
21673 + di_read_unlock(dentry, AuLock_IR);
21677 + AuDebugOn(IS_ROOT(dentry));
21678 + di_write_lock_child(dentry);
21679 + err = au_dbrange_test(dentry);
21681 + err = au_reval_for_attr(dentry, sigen);
21683 + di_downgrade_lock(dentry, AuLock_IR);
21685 + di_write_unlock(dentry);
21690 + di_read_lock_child(dentry, AuLock_IR);
21693 + inode = d_inode(dentry);
21694 + bindex = au_ibtop(inode);
21695 + h_path->mnt = au_sbr_mnt(sb, bindex);
21696 + h_sb = h_path->mnt->mnt_sb;
21698 + && !au_test_fs_bad_iattr(h_sb)
21700 + goto out; /* success */
21702 + if (au_dbtop(dentry) == bindex)
21703 + h_path->dentry = au_h_dptr(dentry, bindex);
21704 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
21705 + h_path->dentry = au_plink_lkup(inode, bindex);
21706 + if (IS_ERR(h_path->dentry))
21707 + /* pretending success */
21708 + h_path->dentry = NULL;
21710 + dput(h_path->dentry);
21717 +static int aufs_getattr(const struct path *path, struct kstat *st,
21718 + u32 request, unsigned int query)
21721 + unsigned char positive;
21722 + struct path h_path;
21723 + struct dentry *dentry;
21724 + struct inode *inode;
21725 + struct super_block *sb;
21727 + dentry = path->dentry;
21728 + inode = d_inode(dentry);
21729 + sb = dentry->d_sb;
21730 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
21731 + if (unlikely(err))
21733 + err = au_h_path_getattr(dentry, /*force*/0, &h_path, /*locked*/0);
21734 + if (unlikely(err))
21736 + if (unlikely(!h_path.dentry))
21737 + /* illegally overlapped or something */
21738 + goto out_fill; /* pretending success */
21740 + positive = d_is_positive(h_path.dentry);
21742 + /* no vfsub version */
21743 + err = vfs_getattr(&h_path, st, request, query);
21746 + au_refresh_iattr(inode, st,
21747 + d_inode(h_path.dentry)->i_nlink);
21748 + goto out_fill; /* success */
21754 + generic_fillattr(inode, st);
21756 + di_read_unlock(dentry, AuLock_IR);
21758 + si_read_unlock(sb);
21764 +/* ---------------------------------------------------------------------- */
21766 +static const char *aufs_get_link(struct dentry *dentry, struct inode *inode,
21767 + struct delayed_call *done)
21770 + struct dentry *h_dentry;
21771 + struct inode *h_inode;
21773 + aufs_bindex_t bindex;
21775 + ret = NULL; /* suppress a warning */
21780 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
21781 + if (unlikely(err))
21784 + err = au_d_hashed_positive(dentry);
21785 + if (unlikely(err))
21789 + inode = d_inode(dentry);
21790 + bindex = au_ibtop(inode);
21791 + h_inode = au_h_iptr(inode, bindex);
21792 + if (unlikely(!h_inode->i_op->get_link))
21797 + if (au_dbtop(dentry) <= bindex) {
21798 + h_dentry = au_h_dptr(dentry, bindex);
21803 + h_dentry = d_find_any_alias(h_inode);
21804 + if (IS_ERR(h_dentry)) {
21805 + err = PTR_ERR(h_dentry);
21809 + if (unlikely(!h_dentry))
21813 + AuDbg("%pf\n", h_inode->i_op->get_link);
21814 + AuDbgDentry(h_dentry);
21815 + ret = vfs_get_link(h_dentry, done);
21818 + err = PTR_ERR(ret);
21821 + aufs_read_unlock(dentry, AuLock_IR);
21823 + if (unlikely(err))
21824 + ret = ERR_PTR(err);
21825 + AuTraceErrPtr(ret);
21829 +/* ---------------------------------------------------------------------- */
21831 +static int au_is_special(struct inode *inode)
21833 + return (inode->i_mode & (S_IFBLK | S_IFCHR | S_IFIFO | S_IFSOCK));
21836 +static int aufs_update_time(struct inode *inode, struct timespec *ts, int flags)
21839 + aufs_bindex_t bindex;
21840 + struct super_block *sb;
21841 + struct inode *h_inode;
21842 + struct vfsmount *h_mnt;
21844 + sb = inode->i_sb;
21845 + WARN_ONCE((flags & S_ATIME) && !IS_NOATIME(inode),
21846 + "unexpected s_flags 0x%lx", sb->s_flags);
21848 + /* mmap_sem might be acquired already, cf. aufs_mmap() */
21850 + si_read_lock(sb, AuLock_FLUSH);
21851 + ii_write_lock_child(inode);
21854 + bindex = au_ibtop(inode);
21855 + h_inode = au_h_iptr(inode, bindex);
21856 + if (!au_test_ro(sb, bindex, inode)) {
21857 + h_mnt = au_sbr_mnt(sb, bindex);
21858 + err = vfsub_mnt_want_write(h_mnt);
21860 + err = vfsub_update_time(h_inode, ts, flags);
21861 + vfsub_mnt_drop_write(h_mnt);
21863 + } else if (au_is_special(h_inode)) {
21865 + * Never copy-up here.
21866 + * These special files may already be opened and used for
21867 + * communicating. If we copied it up, then the communication
21868 + * would be corrupted.
21870 + AuWarn1("timestamps for i%lu are ignored "
21871 + "since it is on readonly branch (hi%lu).\n",
21872 + inode->i_ino, h_inode->i_ino);
21873 + } else if (flags & ~S_ATIME) {
21875 + AuIOErr1("unexpected flags 0x%x\n", flags);
21880 + au_cpup_attr_timesizes(inode);
21881 + ii_write_unlock(inode);
21882 + si_read_unlock(sb);
21885 + if (!err && (flags & S_VERSION))
21886 + inode_inc_iversion(inode);
21891 +/* ---------------------------------------------------------------------- */
21893 +/* no getattr version will be set by module.c:aufs_init() */
21894 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
21896 + [AuIop_SYMLINK] = {
21897 + .permission = aufs_permission,
21898 +#ifdef CONFIG_FS_POSIX_ACL
21899 + .get_acl = aufs_get_acl,
21900 + .set_acl = aufs_set_acl, /* unsupport for symlink? */
21903 + .setattr = aufs_setattr,
21904 + .getattr = aufs_getattr,
21906 +#ifdef CONFIG_AUFS_XATTR
21907 + .listxattr = aufs_listxattr,
21910 + .get_link = aufs_get_link,
21912 + /* .update_time = aufs_update_time */
21915 + .create = aufs_create,
21916 + .lookup = aufs_lookup,
21917 + .link = aufs_link,
21918 + .unlink = aufs_unlink,
21919 + .symlink = aufs_symlink,
21920 + .mkdir = aufs_mkdir,
21921 + .rmdir = aufs_rmdir,
21922 + .mknod = aufs_mknod,
21923 + .rename = aufs_rename,
21925 + .permission = aufs_permission,
21926 +#ifdef CONFIG_FS_POSIX_ACL
21927 + .get_acl = aufs_get_acl,
21928 + .set_acl = aufs_set_acl,
21931 + .setattr = aufs_setattr,
21932 + .getattr = aufs_getattr,
21934 +#ifdef CONFIG_AUFS_XATTR
21935 + .listxattr = aufs_listxattr,
21938 + .update_time = aufs_update_time,
21939 + .atomic_open = aufs_atomic_open,
21940 + .tmpfile = aufs_tmpfile
21942 + [AuIop_OTHER] = {
21943 + .permission = aufs_permission,
21944 +#ifdef CONFIG_FS_POSIX_ACL
21945 + .get_acl = aufs_get_acl,
21946 + .set_acl = aufs_set_acl,
21949 + .setattr = aufs_setattr,
21950 + .getattr = aufs_getattr,
21952 +#ifdef CONFIG_AUFS_XATTR
21953 + .listxattr = aufs_listxattr,
21956 + .update_time = aufs_update_time
21959 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
21960 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
21961 +++ linux/fs/aufs/i_op_del.c 2018-01-29 07:56:20.056658502 +0100
21964 + * Copyright (C) 2005-2017 Junjiro R. Okajima
21966 + * This program, aufs is free software; you can redistribute it and/or modify
21967 + * it under the terms of the GNU General Public License as published by
21968 + * the Free Software Foundation; either version 2 of the License, or
21969 + * (at your option) any later version.
21971 + * This program is distributed in the hope that it will be useful,
21972 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21973 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21974 + * GNU General Public License for more details.
21976 + * You should have received a copy of the GNU General Public License
21977 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21981 + * inode operations (del entry)
21987 + * decide if a new whiteout for @dentry is necessary or not.
21988 + * when it is necessary, prepare the parent dir for the upper branch whose
21989 + * branch index is @bcpup for creation. the actual creation of the whiteout will
21990 + * be done by caller.
21992 + * 0: wh is unnecessary
21993 + * plus: wh is necessary
21996 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
21998 + int need_wh, err;
21999 + aufs_bindex_t btop;
22000 + struct super_block *sb;
22002 + sb = dentry->d_sb;
22003 + btop = au_dbtop(dentry);
22004 + if (*bcpup < 0) {
22006 + if (au_test_ro(sb, btop, d_inode(dentry))) {
22007 + err = AuWbrCopyup(au_sbi(sb), dentry);
22009 + if (unlikely(err < 0))
22013 + AuDebugOn(btop < *bcpup
22014 + || au_test_ro(sb, *bcpup, d_inode(dentry)));
22015 + AuDbg("bcpup %d, btop %d\n", *bcpup, btop);
22017 + if (*bcpup != btop) {
22018 + err = au_cpup_dirs(dentry, *bcpup);
22019 + if (unlikely(err))
22023 + struct au_dinfo *dinfo, *tmp;
22025 + need_wh = -ENOMEM;
22026 + dinfo = au_di(dentry);
22027 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
22029 + au_di_cp(tmp, dinfo);
22030 + au_di_swap(tmp, dinfo);
22031 + /* returns the number of positive dentries */
22032 + need_wh = au_lkup_dentry(dentry, btop + 1,
22033 + /* AuLkup_IGNORE_PERM */ 0);
22034 + au_di_swap(tmp, dinfo);
22035 + au_rw_write_unlock(&tmp->di_rwsem);
22039 + AuDbg("need_wh %d\n", need_wh);
22047 + * simple tests for the del-entry operations.
22048 + * following the checks in vfs, plus the parent-child relationship.
22050 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
22051 + struct dentry *h_parent, int isdir)
22055 + struct dentry *h_dentry, *h_latest;
22056 + struct inode *h_inode;
22058 + h_dentry = au_h_dptr(dentry, bindex);
22059 + if (d_really_is_positive(dentry)) {
22061 + if (unlikely(d_is_negative(h_dentry)))
22063 + h_inode = d_inode(h_dentry);
22064 + if (unlikely(!h_inode->i_nlink))
22067 + h_mode = h_inode->i_mode;
22070 + if (unlikely(S_ISDIR(h_mode)))
22072 + } else if (unlikely(!S_ISDIR(h_mode))) {
22077 + /* rename(2) case */
22079 + if (unlikely(d_is_positive(h_dentry)))
22084 + /* expected parent dir is locked */
22085 + if (unlikely(h_parent != h_dentry->d_parent))
22090 + * rmdir a dir may break the consistency on some filesystem.
22091 + * let's try heavy test.
22094 + if (unlikely(!au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1)
22095 + && au_test_h_perm(d_inode(h_parent),
22096 + MAY_EXEC | MAY_WRITE)))
22099 + h_latest = au_sio_lkup_one(&dentry->d_name, h_parent);
22101 + if (IS_ERR(h_latest))
22103 + if (h_latest == h_dentry)
22112 + * decide the branch where we operate for @dentry. the branch index will be set
22113 + * @rbcpup. after diciding it, 'pin' it and store the timestamps of the parent
22114 + * dir for reverting.
22115 + * when a new whiteout is necessary, create it.
22117 +static struct dentry*
22118 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
22119 + struct au_dtime *dt, struct au_pin *pin)
22121 + struct dentry *wh_dentry;
22122 + struct super_block *sb;
22123 + struct path h_path;
22124 + int err, need_wh;
22125 + unsigned int udba;
22126 + aufs_bindex_t bcpup;
22128 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
22129 + wh_dentry = ERR_PTR(need_wh);
22130 + if (unlikely(need_wh < 0))
22133 + sb = dentry->d_sb;
22134 + udba = au_opt_udba(sb);
22136 + err = au_pin(pin, dentry, bcpup, udba,
22137 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
22138 + wh_dentry = ERR_PTR(err);
22139 + if (unlikely(err))
22142 + h_path.dentry = au_pinned_h_parent(pin);
22143 + if (udba != AuOpt_UDBA_NONE
22144 + && au_dbtop(dentry) == bcpup) {
22145 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
22146 + wh_dentry = ERR_PTR(err);
22147 + if (unlikely(err))
22151 + h_path.mnt = au_sbr_mnt(sb, bcpup);
22152 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
22153 + wh_dentry = NULL;
22155 + goto out; /* success, no need to create whiteout */
22157 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
22158 + if (IS_ERR(wh_dentry))
22161 + /* returns with the parent is locked and wh_dentry is dget-ed */
22162 + goto out; /* success */
22167 + return wh_dentry;
22171 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
22172 + * in order to be revertible and save time for removing many child whiteouts
22174 + * returns 1 when there are too many child whiteout and caller should remove
22175 + * them asynchronously. returns 0 when the number of children is enough small to
22176 + * remove now or the branch fs is a remote fs.
22177 + * otherwise return an error.
22179 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
22180 + struct au_nhash *whlist, struct inode *dir)
22182 + int rmdir_later, err, dirwh;
22183 + struct dentry *h_dentry;
22184 + struct super_block *sb;
22185 + struct inode *inode;
22187 + sb = dentry->d_sb;
22188 + SiMustAnyLock(sb);
22189 + h_dentry = au_h_dptr(dentry, bindex);
22190 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
22191 + if (unlikely(err))
22194 + /* stop monitoring */
22195 + inode = d_inode(dentry);
22196 + au_hn_free(au_hi(inode, bindex));
22198 + if (!au_test_fs_remote(h_dentry->d_sb)) {
22199 + dirwh = au_sbi(sb)->si_dirwh;
22200 + rmdir_later = (dirwh <= 1);
22201 + if (!rmdir_later)
22202 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
22205 + return rmdir_later;
22208 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
22209 + if (unlikely(err)) {
22210 + AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
22211 + h_dentry, bindex, err);
22221 + * final procedure for deleting a entry.
22222 + * maintain dentry and iattr.
22224 +static void epilog(struct inode *dir, struct dentry *dentry,
22225 + aufs_bindex_t bindex)
22227 + struct inode *inode;
22229 + inode = d_inode(dentry);
22231 + inode->i_ctime = dir->i_ctime;
22233 + au_dir_ts(dir, bindex);
22234 + dir->i_version++;
22238 + * when an error happened, remove the created whiteout and revert everything.
22240 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
22241 + aufs_bindex_t bwh, struct dentry *wh_dentry,
22242 + struct dentry *dentry, struct au_dtime *dt)
22245 + struct path h_path = {
22246 + .dentry = wh_dentry,
22247 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
22250 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
22252 + au_set_dbwh(dentry, bwh);
22253 + au_dtime_revert(dt);
22257 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
22261 +/* ---------------------------------------------------------------------- */
22263 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
22266 + aufs_bindex_t bwh, bindex, btop;
22267 + struct inode *inode, *h_dir, *delegated;
22268 + struct dentry *parent, *wh_dentry;
22269 + /* to reuduce stack size */
22271 + struct au_dtime dt;
22272 + struct au_pin pin;
22273 + struct path h_path;
22279 + a = kmalloc(sizeof(*a), GFP_NOFS);
22280 + if (unlikely(!a))
22283 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
22284 + if (unlikely(err))
22286 + err = au_d_hashed_positive(dentry);
22287 + if (unlikely(err))
22289 + inode = d_inode(dentry);
22290 + IMustLock(inode);
22292 + if (unlikely(d_is_dir(dentry)))
22293 + goto out_unlock; /* possible? */
22295 + btop = au_dbtop(dentry);
22296 + bwh = au_dbwh(dentry);
22298 + parent = dentry->d_parent; /* dir inode is locked */
22299 + di_write_lock_parent(parent);
22300 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
22302 + err = PTR_ERR(wh_dentry);
22303 + if (IS_ERR(wh_dentry))
22306 + a->h_path.mnt = au_sbr_mnt(dentry->d_sb, btop);
22307 + a->h_path.dentry = au_h_dptr(dentry, btop);
22308 + dget(a->h_path.dentry);
22309 + if (bindex == btop) {
22310 + h_dir = au_pinned_h_dir(&a->pin);
22311 + delegated = NULL;
22312 + err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
22313 + if (unlikely(err == -EWOULDBLOCK)) {
22314 + pr_warn("cannot retry for NFSv4 delegation"
22315 + " for an internal unlink\n");
22319 + /* dir inode is locked */
22320 + h_dir = d_inode(wh_dentry->d_parent);
22321 + IMustLock(h_dir);
22326 + vfsub_drop_nlink(inode);
22327 + epilog(dir, dentry, bindex);
22329 + /* update target timestamps */
22330 + if (bindex == btop) {
22331 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
22333 + inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
22335 + /* todo: this timestamp may be reverted later */
22336 + inode->i_ctime = h_dir->i_ctime;
22337 + goto out_unpin; /* success */
22344 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
22351 + au_unpin(&a->pin);
22353 + dput(a->h_path.dentry);
22355 + di_write_unlock(parent);
22357 + aufs_read_unlock(dentry, AuLock_DW);
22364 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
22366 + int err, rmdir_later;
22367 + aufs_bindex_t bwh, bindex, btop;
22368 + struct inode *inode;
22369 + struct dentry *parent, *wh_dentry, *h_dentry;
22370 + struct au_whtmp_rmdir *args;
22371 + /* to reuduce stack size */
22373 + struct au_dtime dt;
22374 + struct au_pin pin;
22380 + a = kmalloc(sizeof(*a), GFP_NOFS);
22381 + if (unlikely(!a))
22384 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
22385 + if (unlikely(err))
22387 + err = au_alive_dir(dentry);
22388 + if (unlikely(err))
22390 + inode = d_inode(dentry);
22391 + IMustLock(inode);
22393 + if (unlikely(!d_is_dir(dentry)))
22394 + goto out_unlock; /* possible? */
22397 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
22398 + if (unlikely(!args))
22401 + parent = dentry->d_parent; /* dir inode is locked */
22402 + di_write_lock_parent(parent);
22403 + err = au_test_empty(dentry, &args->whlist);
22404 + if (unlikely(err))
22407 + btop = au_dbtop(dentry);
22408 + bwh = au_dbwh(dentry);
22410 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
22412 + err = PTR_ERR(wh_dentry);
22413 + if (IS_ERR(wh_dentry))
22416 + h_dentry = au_h_dptr(dentry, btop);
22419 + if (bindex == btop) {
22420 + err = renwh_and_rmdir(dentry, btop, &args->whlist, dir);
22422 + rmdir_later = err;
22426 + /* stop monitoring */
22427 + au_hn_free(au_hi(inode, btop));
22429 + /* dir inode is locked */
22430 + IMustLock(d_inode(wh_dentry->d_parent));
22435 + vfsub_dead_dir(inode);
22436 + au_set_dbdiropq(dentry, -1);
22437 + epilog(dir, dentry, bindex);
22439 + if (rmdir_later) {
22440 + au_whtmp_kick_rmdir(dir, btop, h_dentry, args);
22444 + goto out_unpin; /* success */
22452 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
22459 + au_unpin(&a->pin);
22463 + di_write_unlock(parent);
22465 + au_whtmp_rmdir_free(args);
22467 + aufs_read_unlock(dentry, AuLock_DW);
22474 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
22475 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
22476 +++ linux/fs/aufs/i_op_ren.c 2018-01-29 07:56:20.056658502 +0100
22479 + * Copyright (C) 2005-2017 Junjiro R. Okajima
22481 + * This program, aufs is free software; you can redistribute it and/or modify
22482 + * it under the terms of the GNU General Public License as published by
22483 + * the Free Software Foundation; either version 2 of the License, or
22484 + * (at your option) any later version.
22486 + * This program is distributed in the hope that it will be useful,
22487 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22488 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22489 + * GNU General Public License for more details.
22491 + * You should have received a copy of the GNU General Public License
22492 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
22496 + * inode operation (rename entry)
22497 + * todo: this is crazy monster
22502 +enum { AuSRC, AuDST, AuSrcDst };
22503 +enum { AuPARENT, AuCHILD, AuParentChild };
22505 +#define AuRen_ISDIR_SRC 1
22506 +#define AuRen_ISDIR_DST (1 << 1)
22507 +#define AuRen_ISSAMEDIR (1 << 2)
22508 +#define AuRen_WHSRC (1 << 3)
22509 +#define AuRen_WHDST (1 << 4)
22510 +#define AuRen_MNT_WRITE (1 << 5)
22511 +#define AuRen_DT_DSTDIR (1 << 6)
22512 +#define AuRen_DIROPQ_SRC (1 << 7)
22513 +#define AuRen_DIROPQ_DST (1 << 8)
22514 +#define AuRen_DIRREN (1 << 9)
22515 +#define AuRen_DROPPED_SRC (1 << 10)
22516 +#define AuRen_DROPPED_DST (1 << 11)
22517 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
22518 +#define au_fset_ren(flags, name) \
22519 + do { (flags) |= AuRen_##name; } while (0)
22520 +#define au_fclr_ren(flags, name) \
22521 + do { (flags) &= ~AuRen_##name; } while (0)
22523 +#ifndef CONFIG_AUFS_DIRREN
22524 +#undef AuRen_DIRREN
22525 +#define AuRen_DIRREN 0
22528 +struct au_ren_args {
22530 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
22532 + struct inode *dir, *inode;
22533 + struct au_hinode *hdir, *hinode;
22534 + struct au_dtime dt[AuParentChild];
22535 + aufs_bindex_t btop, bdiropq;
22538 +#define src_dentry sd[AuSRC].dentry
22539 +#define src_dir sd[AuSRC].dir
22540 +#define src_inode sd[AuSRC].inode
22541 +#define src_h_dentry sd[AuSRC].h_dentry
22542 +#define src_parent sd[AuSRC].parent
22543 +#define src_h_parent sd[AuSRC].h_parent
22544 +#define src_wh_dentry sd[AuSRC].wh_dentry
22545 +#define src_hdir sd[AuSRC].hdir
22546 +#define src_hinode sd[AuSRC].hinode
22547 +#define src_h_dir sd[AuSRC].hdir->hi_inode
22548 +#define src_dt sd[AuSRC].dt
22549 +#define src_btop sd[AuSRC].btop
22550 +#define src_bdiropq sd[AuSRC].bdiropq
22552 +#define dst_dentry sd[AuDST].dentry
22553 +#define dst_dir sd[AuDST].dir
22554 +#define dst_inode sd[AuDST].inode
22555 +#define dst_h_dentry sd[AuDST].h_dentry
22556 +#define dst_parent sd[AuDST].parent
22557 +#define dst_h_parent sd[AuDST].h_parent
22558 +#define dst_wh_dentry sd[AuDST].wh_dentry
22559 +#define dst_hdir sd[AuDST].hdir
22560 +#define dst_hinode sd[AuDST].hinode
22561 +#define dst_h_dir sd[AuDST].hdir->hi_inode
22562 +#define dst_dt sd[AuDST].dt
22563 +#define dst_btop sd[AuDST].btop
22564 +#define dst_bdiropq sd[AuDST].bdiropq
22566 + struct dentry *h_trap;
22567 + struct au_branch *br;
22568 + struct path h_path;
22569 + struct au_nhash whlist;
22570 + aufs_bindex_t btgt, src_bwh;
22573 + unsigned short auren_flags;
22574 + unsigned char flags; /* syscall parameter */
22575 + unsigned char exchange;
22578 + struct au_whtmp_rmdir *thargs;
22579 + struct dentry *h_dst;
22580 + struct au_hinode *h_root;
22583 +/* ---------------------------------------------------------------------- */
22586 + * functions for reverting.
22587 + * when an error happened in a single rename systemcall, we should revert
22588 + * everything as if nothing happened.
22589 + * we don't need to revert the copied-up/down the parent dir since they are
22593 +#define RevertFailure(fmt, ...) do { \
22594 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
22595 + ##__VA_ARGS__, err, rerr); \
22599 +static void au_ren_do_rev_diropq(int err, struct au_ren_args *a, int idx)
22602 + struct dentry *d;
22603 +#define src_or_dst(member) a->sd[idx].member
22605 + d = src_or_dst(dentry); /* {src,dst}_dentry */
22606 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
22607 + rerr = au_diropq_remove(d, a->btgt);
22608 + au_hn_inode_unlock(src_or_dst(hinode));
22609 + au_set_dbdiropq(d, src_or_dst(bdiropq));
22611 + RevertFailure("remove diropq %pd", d);
22613 +#undef src_or_dst_
22616 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
22618 + if (au_ftest_ren(a->auren_flags, DIROPQ_SRC))
22619 + au_ren_do_rev_diropq(err, a, AuSRC);
22620 + if (au_ftest_ren(a->auren_flags, DIROPQ_DST))
22621 + au_ren_do_rev_diropq(err, a, AuDST);
22624 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
22627 + struct inode *delegated;
22629 + a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name,
22630 + a->src_h_parent);
22631 + rerr = PTR_ERR(a->h_path.dentry);
22632 + if (IS_ERR(a->h_path.dentry)) {
22633 + RevertFailure("lkup one %pd", a->src_dentry);
22637 + delegated = NULL;
22638 + rerr = vfsub_rename(a->dst_h_dir,
22639 + au_h_dptr(a->src_dentry, a->btgt),
22640 + a->src_h_dir, &a->h_path, &delegated, a->flags);
22641 + if (unlikely(rerr == -EWOULDBLOCK)) {
22642 + pr_warn("cannot retry for NFSv4 delegation"
22643 + " for an internal rename\n");
22646 + d_drop(a->h_path.dentry);
22647 + dput(a->h_path.dentry);
22648 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
22650 + RevertFailure("rename %pd", a->src_dentry);
22653 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
22656 + struct inode *delegated;
22658 + a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name,
22659 + a->dst_h_parent);
22660 + rerr = PTR_ERR(a->h_path.dentry);
22661 + if (IS_ERR(a->h_path.dentry)) {
22662 + RevertFailure("lkup one %pd", a->dst_dentry);
22665 + if (d_is_positive(a->h_path.dentry)) {
22666 + d_drop(a->h_path.dentry);
22667 + dput(a->h_path.dentry);
22671 + delegated = NULL;
22672 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
22673 + &delegated, a->flags);
22674 + if (unlikely(rerr == -EWOULDBLOCK)) {
22675 + pr_warn("cannot retry for NFSv4 delegation"
22676 + " for an internal rename\n");
22679 + d_drop(a->h_path.dentry);
22680 + dput(a->h_path.dentry);
22682 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
22684 + RevertFailure("rename %pd", a->h_dst);
22687 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
22691 + a->h_path.dentry = a->src_wh_dentry;
22692 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
22693 + au_set_dbwh(a->src_dentry, a->src_bwh);
22695 + RevertFailure("unlink %pd", a->src_wh_dentry);
22697 +#undef RevertFailure
22699 +/* ---------------------------------------------------------------------- */
22702 + * when we have to copyup the renaming entry, do it with the rename-target name
22703 + * in order to minimize the cost (the later actual rename is unnecessary).
22704 + * otherwise rename it on the target branch.
22706 +static int au_ren_or_cpup(struct au_ren_args *a)
22709 + struct dentry *d;
22710 + struct inode *delegated;
22712 + d = a->src_dentry;
22713 + if (au_dbtop(d) == a->btgt) {
22714 + a->h_path.dentry = a->dst_h_dentry;
22715 + AuDebugOn(au_dbtop(d) != a->btgt);
22716 + delegated = NULL;
22717 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
22718 + a->dst_h_dir, &a->h_path, &delegated,
22720 + if (unlikely(err == -EWOULDBLOCK)) {
22721 + pr_warn("cannot retry for NFSv4 delegation"
22722 + " for an internal rename\n");
22728 + if (!err && a->h_dst)
22729 + /* it will be set to dinfo later */
22735 +/* cf. aufs_rmdir() */
22736 +static int au_ren_del_whtmp(struct au_ren_args *a)
22739 + struct inode *dir;
22741 + dir = a->dst_dir;
22742 + SiMustAnyLock(dir->i_sb);
22743 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
22744 + au_sbi(dir->i_sb)->si_dirwh)
22745 + || au_test_fs_remote(a->h_dst->d_sb)) {
22746 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
22747 + if (unlikely(err))
22748 + pr_warn("failed removing whtmp dir %pd (%d), "
22749 + "ignored.\n", a->h_dst, err);
22751 + au_nhash_wh_free(&a->thargs->whlist);
22752 + a->thargs->whlist = a->whlist;
22753 + a->whlist.nh_num = 0;
22754 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
22756 + a->thargs = NULL;
22762 +/* make it 'opaque' dir. */
22763 +static int au_ren_do_diropq(struct au_ren_args *a, int idx)
22766 + struct dentry *d, *diropq;
22767 +#define src_or_dst(member) a->sd[idx].member
22770 + d = src_or_dst(dentry); /* {src,dst}_dentry */
22771 + src_or_dst(bdiropq) = au_dbdiropq(d);
22772 + src_or_dst(hinode) = au_hi(src_or_dst(inode), a->btgt);
22773 + au_hn_inode_lock_nested(src_or_dst(hinode), AuLsc_I_CHILD);
22774 + diropq = au_diropq_create(d, a->btgt);
22775 + au_hn_inode_unlock(src_or_dst(hinode));
22776 + if (IS_ERR(diropq))
22777 + err = PTR_ERR(diropq);
22781 +#undef src_or_dst_
22785 +static int au_ren_diropq(struct au_ren_args *a)
22788 + unsigned char always;
22789 + struct dentry *d;
22792 + d = a->dst_dentry; /* already renamed on the branch */
22793 + always = !!au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ);
22794 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
22795 + && !au_ftest_ren(a->auren_flags, DIRREN)
22796 + && a->btgt != au_dbdiropq(a->src_dentry)
22797 + && (a->dst_wh_dentry
22798 + || a->btgt <= au_dbdiropq(d)
22799 + /* hide the lower to keep xino */
22800 + /* the lowers may not be a dir, but we hide them anyway */
22801 + || a->btgt < au_dbbot(d)
22804 + err = au_ren_do_diropq(a, AuSRC);
22805 + if (unlikely(err))
22807 + au_fset_ren(a->auren_flags, DIROPQ_SRC);
22809 + if (!a->exchange)
22810 + goto out; /* success */
22812 + d = a->src_dentry; /* already renamed on the branch */
22813 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
22814 + && a->btgt != au_dbdiropq(a->dst_dentry)
22815 + && (a->btgt < au_dbdiropq(d)
22816 + || a->btgt < au_dbbot(d)
22818 + AuDbgDentry(a->src_dentry);
22819 + AuDbgDentry(a->dst_dentry);
22820 + err = au_ren_do_diropq(a, AuDST);
22821 + if (unlikely(err))
22822 + goto out_rev_src;
22823 + au_fset_ren(a->auren_flags, DIROPQ_DST);
22825 + goto out; /* success */
22828 + AuDbg("err %d, reverting src\n", err);
22829 + au_ren_rev_diropq(err, a);
22834 +static int do_rename(struct au_ren_args *a)
22837 + struct dentry *d, *h_d;
22839 + if (!a->exchange) {
22840 + /* prepare workqueue args for asynchronous rmdir */
22841 + h_d = a->dst_h_dentry;
22842 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)
22843 + /* && !au_ftest_ren(a->auren_flags, DIRREN) */
22844 + && d_is_positive(h_d)) {
22846 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb,
22848 + if (unlikely(!a->thargs))
22850 + a->h_dst = dget(h_d);
22853 + /* create whiteout for src_dentry */
22854 + if (au_ftest_ren(a->auren_flags, WHSRC)) {
22855 + a->src_bwh = au_dbwh(a->src_dentry);
22856 + AuDebugOn(a->src_bwh >= 0);
22857 + a->src_wh_dentry = au_wh_create(a->src_dentry, a->btgt,
22858 + a->src_h_parent);
22859 + err = PTR_ERR(a->src_wh_dentry);
22860 + if (IS_ERR(a->src_wh_dentry))
22864 + /* lookup whiteout for dentry */
22865 + if (au_ftest_ren(a->auren_flags, WHDST)) {
22866 + h_d = au_wh_lkup(a->dst_h_parent,
22867 + &a->dst_dentry->d_name, a->br);
22868 + err = PTR_ERR(h_d);
22871 + if (d_is_negative(h_d))
22874 + a->dst_wh_dentry = h_d;
22877 + /* rename dentry to tmpwh */
22879 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
22880 + if (unlikely(err))
22883 + d = a->dst_dentry;
22884 + au_set_h_dptr(d, a->btgt, NULL);
22885 + err = au_lkup_neg(d, a->btgt, /*wh*/0);
22886 + if (unlikely(err))
22888 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
22892 + BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_btop != a->btgt);
22894 + BUG_ON(!au_ftest_ren(a->auren_flags, DIRREN)
22895 + && d_is_positive(a->dst_h_dentry)
22896 + && a->src_btop != a->btgt);
22899 + /* rename by vfs_rename or cpup */
22900 + err = au_ren_or_cpup(a);
22901 + if (unlikely(err))
22902 + /* leave the copied-up one */
22905 + /* make dir opaque */
22906 + err = au_ren_diropq(a);
22907 + if (unlikely(err))
22910 + /* update target timestamps */
22911 + if (a->exchange) {
22912 + AuDebugOn(au_dbtop(a->dst_dentry) != a->btgt);
22913 + a->h_path.dentry = au_h_dptr(a->dst_dentry, a->btgt);
22914 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
22915 + a->dst_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
22917 + AuDebugOn(au_dbtop(a->src_dentry) != a->btgt);
22918 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
22919 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
22920 + a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
22922 + if (!a->exchange) {
22923 + /* remove whiteout for dentry */
22924 + if (a->dst_wh_dentry) {
22925 + a->h_path.dentry = a->dst_wh_dentry;
22926 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
22928 + if (unlikely(err))
22932 + /* remove whtmp */
22934 + au_ren_del_whtmp(a); /* ignore this error */
22936 + au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
22939 + goto out_success;
22942 + au_ren_rev_diropq(err, a);
22944 + au_ren_rev_rename(err, a);
22948 + au_ren_rev_whtmp(err, a);
22950 + dput(a->dst_wh_dentry);
22951 + a->dst_wh_dentry = NULL;
22953 + if (a->src_wh_dentry)
22954 + au_ren_rev_whsrc(err, a);
22956 + dput(a->src_wh_dentry);
22957 + dput(a->dst_wh_dentry);
22961 + au_whtmp_rmdir_free(a->thargs);
22962 + a->thargs = NULL;
22968 +/* ---------------------------------------------------------------------- */
22971 + * test if @dentry dir can be rename destination or not.
22972 + * success means, it is a logically empty dir.
22974 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
22976 + return au_test_empty(dentry, whlist);
22980 + * test if @a->src_dentry dir can be rename source or not.
22981 + * if it can, return 0.
22983 + * - it is a logically empty dir.
22984 + * - or, it exists on writable branch and has no children including whiteouts
22985 + * on the lower branch unless DIRREN is on.
22987 +static int may_rename_srcdir(struct au_ren_args *a)
22990 + unsigned int rdhash;
22991 + aufs_bindex_t btop, btgt;
22992 + struct dentry *dentry;
22993 + struct super_block *sb;
22994 + struct au_sbinfo *sbinfo;
22996 + dentry = a->src_dentry;
22997 + sb = dentry->d_sb;
22998 + sbinfo = au_sbi(sb);
22999 + if (au_opt_test(sbinfo->si_mntflags, DIRREN))
23000 + au_fset_ren(a->auren_flags, DIRREN);
23003 + btop = au_dbtop(dentry);
23004 + if (btop != btgt) {
23005 + struct au_nhash whlist;
23007 + SiMustAnyLock(sb);
23008 + rdhash = sbinfo->si_rdhash;
23010 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
23012 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
23013 + if (unlikely(err))
23015 + err = au_test_empty(dentry, &whlist);
23016 + au_nhash_wh_free(&whlist);
23020 + if (btop == au_dbtaildir(dentry))
23021 + return 0; /* success */
23023 + err = au_test_empty_lower(dentry);
23026 + if (err == -ENOTEMPTY) {
23027 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
23030 + AuWarn1("renaming dir who has child(ren) on multiple "
23031 + "branches, is not supported\n");
23038 +/* side effect: sets whlist and h_dentry */
23039 +static int au_ren_may_dir(struct au_ren_args *a)
23042 + unsigned int rdhash;
23043 + struct dentry *d;
23045 + d = a->dst_dentry;
23046 + SiMustAnyLock(d->d_sb);
23049 + if (au_ftest_ren(a->auren_flags, ISDIR_DST) && a->dst_inode) {
23050 + rdhash = au_sbi(d->d_sb)->si_rdhash;
23052 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
23053 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
23054 + if (unlikely(err))
23057 + if (!a->exchange) {
23058 + au_set_dbtop(d, a->dst_btop);
23059 + err = may_rename_dstdir(d, &a->whlist);
23060 + au_set_dbtop(d, a->btgt);
23062 + err = may_rename_srcdir(a);
23064 + a->dst_h_dentry = au_h_dptr(d, au_dbtop(d));
23065 + if (unlikely(err))
23068 + d = a->src_dentry;
23069 + a->src_h_dentry = au_h_dptr(d, au_dbtop(d));
23070 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
23071 + err = may_rename_srcdir(a);
23072 + if (unlikely(err)) {
23073 + au_nhash_wh_free(&a->whlist);
23074 + a->whlist.nh_num = 0;
23081 +/* ---------------------------------------------------------------------- */
23084 + * simple tests for rename.
23085 + * following the checks in vfs, plus the parent-child relationship.
23087 +static int au_may_ren(struct au_ren_args *a)
23090 + struct inode *h_inode;
23092 + if (a->src_btop == a->btgt) {
23093 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
23094 + au_ftest_ren(a->auren_flags, ISDIR_SRC));
23095 + if (unlikely(err))
23098 + if (unlikely(a->src_h_dentry == a->h_trap))
23103 + if (a->dst_btop != a->btgt)
23106 + err = -ENOTEMPTY;
23107 + if (unlikely(a->dst_h_dentry == a->h_trap))
23111 + isdir = !!au_ftest_ren(a->auren_flags, ISDIR_DST);
23112 + if (d_really_is_negative(a->dst_dentry)) {
23113 + if (d_is_negative(a->dst_h_dentry))
23114 + err = au_may_add(a->dst_dentry, a->btgt,
23115 + a->dst_h_parent, isdir);
23117 + if (unlikely(d_is_negative(a->dst_h_dentry)))
23119 + h_inode = d_inode(a->dst_h_dentry);
23120 + if (h_inode->i_nlink)
23121 + err = au_may_del(a->dst_dentry, a->btgt,
23122 + a->dst_h_parent, isdir);
23126 + if (unlikely(err == -ENOENT || err == -EEXIST))
23132 +/* ---------------------------------------------------------------------- */
23137 + * - src_dir and dir by lock_rename()
23138 + * - inode if exitsts
23141 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
23143 + * + di_write_lock2_child()
23144 + * + di_write_lock_child()
23145 + * + ii_write_lock_child()
23146 + * + di_write_lock_child2()
23147 + * + ii_write_lock_child2()
23148 + * + src_parent and parent
23149 + * + di_write_lock_parent()
23150 + * + ii_write_lock_parent()
23151 + * + di_write_lock_parent2()
23152 + * + ii_write_lock_parent2()
23153 + * + lower src_dir and dir by vfsub_lock_rename()
23154 + * + verify the every relationships between child and parent. if any
23155 + * of them failed, unlock all and return -EBUSY.
23157 +static void au_ren_unlock(struct au_ren_args *a)
23159 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
23160 + a->dst_h_parent, a->dst_hdir);
23161 + if (au_ftest_ren(a->auren_flags, DIRREN)
23163 + au_hn_inode_unlock(a->h_root);
23164 + if (au_ftest_ren(a->auren_flags, MNT_WRITE))
23165 + vfsub_mnt_drop_write(au_br_mnt(a->br));
23168 +static int au_ren_lock(struct au_ren_args *a)
23171 + unsigned int udba;
23174 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
23175 + a->src_hdir = au_hi(a->src_dir, a->btgt);
23176 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
23177 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
23179 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
23180 + if (unlikely(err))
23182 + au_fset_ren(a->auren_flags, MNT_WRITE);
23183 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
23184 + struct dentry *root;
23185 + struct inode *dir;
23188 + * sbinfo is already locked, so this ii_read_lock is
23189 + * unnecessary. but our debugging feature checks it.
23191 + root = a->src_inode->i_sb->s_root;
23192 + if (root != a->src_parent && root != a->dst_parent) {
23193 + dir = d_inode(root);
23194 + ii_read_lock_parent3(dir);
23195 + a->h_root = au_hi(dir, a->btgt);
23196 + ii_read_unlock(dir);
23197 + au_hn_inode_lock_nested(a->h_root, AuLsc_I_PARENT3);
23200 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
23201 + a->dst_h_parent, a->dst_hdir);
23202 + udba = au_opt_udba(a->src_dentry->d_sb);
23203 + if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
23204 + || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
23205 + err = au_busy_or_stale();
23206 + if (!err && au_dbtop(a->src_dentry) == a->btgt)
23207 + err = au_h_verify(a->src_h_dentry, udba,
23208 + d_inode(a->src_h_parent), a->src_h_parent,
23210 + if (!err && au_dbtop(a->dst_dentry) == a->btgt)
23211 + err = au_h_verify(a->dst_h_dentry, udba,
23212 + d_inode(a->dst_h_parent), a->dst_h_parent,
23215 + goto out; /* success */
23217 + err = au_busy_or_stale();
23218 + au_ren_unlock(a);
23224 +/* ---------------------------------------------------------------------- */
23226 +static void au_ren_refresh_dir(struct au_ren_args *a)
23228 + struct inode *dir;
23230 + dir = a->dst_dir;
23231 + dir->i_version++;
23232 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)) {
23233 + /* is this updating defined in POSIX? */
23234 + au_cpup_attr_timesizes(a->src_inode);
23235 + au_cpup_attr_nlink(dir, /*force*/1);
23237 + au_dir_ts(dir, a->btgt);
23239 + if (a->exchange) {
23240 + dir = a->src_dir;
23241 + dir->i_version++;
23242 + if (au_ftest_ren(a->auren_flags, ISDIR_DST)) {
23243 + /* is this updating defined in POSIX? */
23244 + au_cpup_attr_timesizes(a->dst_inode);
23245 + au_cpup_attr_nlink(dir, /*force*/1);
23247 + au_dir_ts(dir, a->btgt);
23250 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
23253 + dir = a->src_dir;
23254 + dir->i_version++;
23255 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC))
23256 + au_cpup_attr_nlink(dir, /*force*/1);
23257 + au_dir_ts(dir, a->btgt);
23260 +static void au_ren_refresh(struct au_ren_args *a)
23262 + aufs_bindex_t bbot, bindex;
23263 + struct dentry *d, *h_d;
23264 + struct inode *i, *h_i;
23265 + struct super_block *sb;
23267 + d = a->dst_dentry;
23270 + /* already dget-ed by au_ren_or_cpup() */
23271 + au_set_h_dptr(d, a->btgt, a->h_dst);
23273 + i = a->dst_inode;
23275 + if (!a->exchange) {
23276 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST))
23277 + vfsub_drop_nlink(i);
23279 + vfsub_dead_dir(i);
23280 + au_cpup_attr_timesizes(i);
23282 + au_update_dbrange(d, /*do_put_zero*/1);
23284 + au_cpup_attr_nlink(i, /*force*/1);
23287 + for (bindex = au_dbtop(d); bindex < bbot; bindex++)
23288 + au_set_h_dptr(d, bindex, NULL);
23289 + bbot = au_dbbot(d);
23290 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++)
23291 + au_set_h_dptr(d, bindex, NULL);
23292 + au_update_dbrange(d, /*do_put_zero*/0);
23296 + || au_ftest_ren(a->auren_flags, DIRREN)) {
23297 + d_drop(a->src_dentry);
23298 + if (au_ftest_ren(a->auren_flags, DIRREN))
23299 + au_set_dbwh(a->src_dentry, -1);
23303 + d = a->src_dentry;
23304 + au_set_dbwh(d, -1);
23305 + bbot = au_dbbot(d);
23306 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
23307 + h_d = au_h_dptr(d, bindex);
23309 + au_set_h_dptr(d, bindex, NULL);
23311 + au_set_dbbot(d, a->btgt);
23314 + i = a->src_inode;
23315 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
23316 + return; /* success */
23318 + bbot = au_ibbot(i);
23319 + for (bindex = a->btgt + 1; bindex <= bbot; bindex++) {
23320 + h_i = au_h_iptr(i, bindex);
23322 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
23323 + /* ignore this error */
23324 + au_set_h_iptr(i, bindex, NULL, 0);
23327 + au_set_ibbot(i, a->btgt);
23330 +/* ---------------------------------------------------------------------- */
23332 +/* mainly for link(2) and rename(2) */
23333 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
23335 + aufs_bindex_t bdiropq, bwh;
23336 + struct dentry *parent;
23337 + struct au_branch *br;
23339 + parent = dentry->d_parent;
23340 + IMustLock(d_inode(parent)); /* dir is locked */
23342 + bdiropq = au_dbdiropq(parent);
23343 + bwh = au_dbwh(dentry);
23344 + br = au_sbr(dentry->d_sb, btgt);
23345 + if (au_br_rdonly(br)
23346 + || (0 <= bdiropq && bdiropq < btgt)
23347 + || (0 <= bwh && bwh < btgt))
23350 + AuDbg("btgt %d\n", btgt);
23354 +/* sets src_btop, dst_btop and btgt */
23355 +static int au_ren_wbr(struct au_ren_args *a)
23358 + struct au_wr_dir_args wr_dir_args = {
23359 + /* .force_btgt = -1, */
23360 + .flags = AuWrDir_ADD_ENTRY
23363 + a->src_btop = au_dbtop(a->src_dentry);
23364 + a->dst_btop = au_dbtop(a->dst_dentry);
23365 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC)
23366 + || au_ftest_ren(a->auren_flags, ISDIR_DST))
23367 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
23368 + wr_dir_args.force_btgt = a->src_btop;
23369 + if (a->dst_inode && a->dst_btop < a->src_btop)
23370 + wr_dir_args.force_btgt = a->dst_btop;
23371 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
23372 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
23375 + au_update_dbtop(a->dst_dentry);
23380 +static void au_ren_dt(struct au_ren_args *a)
23382 + a->h_path.dentry = a->src_h_parent;
23383 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
23384 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR)) {
23385 + a->h_path.dentry = a->dst_h_parent;
23386 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
23389 + au_fclr_ren(a->auren_flags, DT_DSTDIR);
23390 + if (!au_ftest_ren(a->auren_flags, ISDIR_SRC)
23394 + a->h_path.dentry = a->src_h_dentry;
23395 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
23396 + if (d_is_positive(a->dst_h_dentry)) {
23397 + au_fset_ren(a->auren_flags, DT_DSTDIR);
23398 + a->h_path.dentry = a->dst_h_dentry;
23399 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
23403 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
23405 + struct dentry *h_d;
23406 + struct inode *h_inode;
23408 + au_dtime_revert(a->src_dt + AuPARENT);
23409 + if (!au_ftest_ren(a->auren_flags, ISSAMEDIR))
23410 + au_dtime_revert(a->dst_dt + AuPARENT);
23412 + if (au_ftest_ren(a->auren_flags, ISDIR_SRC) && err != -EIO) {
23413 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
23414 + h_inode = d_inode(h_d);
23415 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
23416 + au_dtime_revert(a->src_dt + AuCHILD);
23417 + inode_unlock(h_inode);
23419 + if (au_ftest_ren(a->auren_flags, DT_DSTDIR)) {
23420 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
23421 + h_inode = d_inode(h_d);
23422 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
23423 + au_dtime_revert(a->dst_dt + AuCHILD);
23424 + inode_unlock(h_inode);
23429 +/* ---------------------------------------------------------------------- */
23431 +int aufs_rename(struct inode *_src_dir, struct dentry *_src_dentry,
23432 + struct inode *_dst_dir, struct dentry *_dst_dentry,
23433 + unsigned int _flags)
23435 + int err, lock_flags;
23437 + /* reduce stack space */
23438 + struct au_ren_args *a;
23439 + struct au_pin pin;
23441 + AuDbg("%pd, %pd, 0x%x\n", _src_dentry, _dst_dentry, _flags);
23442 + IMustLock(_src_dir);
23443 + IMustLock(_dst_dir);
23446 + if (unlikely(_flags & RENAME_WHITEOUT))
23450 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
23451 + a = kzalloc(sizeof(*a), GFP_NOFS);
23452 + if (unlikely(!a))
23455 + a->flags = _flags;
23456 + a->exchange = _flags & RENAME_EXCHANGE;
23457 + a->src_dir = _src_dir;
23458 + a->src_dentry = _src_dentry;
23459 + a->src_inode = NULL;
23460 + if (d_really_is_positive(a->src_dentry))
23461 + a->src_inode = d_inode(a->src_dentry);
23462 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
23463 + a->dst_dir = _dst_dir;
23464 + a->dst_dentry = _dst_dentry;
23465 + a->dst_inode = NULL;
23466 + if (d_really_is_positive(a->dst_dentry))
23467 + a->dst_inode = d_inode(a->dst_dentry);
23468 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
23469 + if (a->dst_inode) {
23471 + * if EXCHANGE && src is non-dir && dst is dir,
23472 + * dst is not locked.
23474 + /* IMustLock(a->dst_inode); */
23475 + au_igrab(a->dst_inode);
23479 + lock_flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
23480 + if (d_is_dir(a->src_dentry)) {
23481 + au_fset_ren(a->auren_flags, ISDIR_SRC);
23482 + if (unlikely(!a->exchange
23483 + && d_really_is_positive(a->dst_dentry)
23484 + && !d_is_dir(a->dst_dentry)))
23486 + lock_flags |= AuLock_DIRS;
23488 + if (a->dst_inode && d_is_dir(a->dst_dentry)) {
23489 + au_fset_ren(a->auren_flags, ISDIR_DST);
23490 + if (unlikely(!a->exchange
23491 + && d_really_is_positive(a->src_dentry)
23492 + && !d_is_dir(a->src_dentry)))
23494 + lock_flags |= AuLock_DIRS;
23496 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
23498 + if (unlikely(err))
23501 + err = au_d_hashed_positive(a->src_dentry);
23502 + if (unlikely(err))
23505 + if (a->dst_inode) {
23507 + * If it is a dir, VFS unhash it before this
23508 + * function. It means we cannot rely upon d_unhashed().
23510 + if (unlikely(!a->dst_inode->i_nlink))
23512 + if (!au_ftest_ren(a->auren_flags, ISDIR_DST)) {
23513 + err = au_d_hashed_positive(a->dst_dentry);
23514 + if (unlikely(err && !a->exchange))
23516 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
23518 + } else if (unlikely(d_unhashed(a->dst_dentry)))
23522 + * is it possible?
23523 + * yes, it happened (in linux-3.3-rcN) but I don't know why.
23524 + * there may exist a problem somewhere else.
23527 + if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
23530 + au_fset_ren(a->auren_flags, ISSAMEDIR); /* temporary */
23531 + di_write_lock_parent(a->dst_parent);
23533 + /* which branch we process */
23534 + err = au_ren_wbr(a);
23535 + if (unlikely(err < 0))
23537 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
23538 + a->h_path.mnt = au_br_mnt(a->br);
23540 + /* are they available to be renamed */
23541 + err = au_ren_may_dir(a);
23542 + if (unlikely(err))
23543 + goto out_children;
23545 + /* prepare the writable parent dir on the same branch */
23546 + if (a->dst_btop == a->btgt) {
23547 + au_fset_ren(a->auren_flags, WHDST);
23549 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
23550 + if (unlikely(err))
23551 + goto out_children;
23555 + if (!a->exchange) {
23556 + if (a->src_dir != a->dst_dir) {
23558 + * this temporary unlock is safe,
23559 + * because both dir->i_mutex are locked.
23561 + di_write_unlock(a->dst_parent);
23562 + di_write_lock_parent(a->src_parent);
23563 + err = au_wr_dir_need_wh(a->src_dentry,
23564 + au_ftest_ren(a->auren_flags,
23567 + di_write_unlock(a->src_parent);
23568 + di_write_lock2_parent(a->src_parent, a->dst_parent,
23570 + au_fclr_ren(a->auren_flags, ISSAMEDIR);
23572 + err = au_wr_dir_need_wh(a->src_dentry,
23573 + au_ftest_ren(a->auren_flags,
23577 + if (unlikely(err < 0))
23578 + goto out_children;
23580 + au_fset_ren(a->auren_flags, WHSRC);
23583 + if (a->src_btop != a->btgt) {
23584 + err = au_pin(&pin, a->src_dentry, a->btgt,
23585 + au_opt_udba(a->src_dentry->d_sb),
23586 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23588 + struct au_cp_generic cpg = {
23589 + .dentry = a->src_dentry,
23591 + .bsrc = a->src_btop,
23594 + .flags = AuCpup_DTIME | AuCpup_HOPEN
23596 + AuDebugOn(au_dbtop(a->src_dentry) != a->src_btop);
23597 + err = au_sio_cpup_simple(&cpg);
23600 + if (unlikely(err))
23601 + goto out_children;
23602 + a->src_btop = a->btgt;
23603 + a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
23604 + if (!a->exchange)
23605 + au_fset_ren(a->auren_flags, WHSRC);
23609 + if (a->exchange && a->dst_inode
23610 + && a->dst_btop != a->btgt) {
23611 + err = au_pin(&pin, a->dst_dentry, a->btgt,
23612 + au_opt_udba(a->dst_dentry->d_sb),
23613 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
23615 + struct au_cp_generic cpg = {
23616 + .dentry = a->dst_dentry,
23618 + .bsrc = a->dst_btop,
23621 + .flags = AuCpup_DTIME | AuCpup_HOPEN
23623 + err = au_sio_cpup_simple(&cpg);
23626 + if (unlikely(err))
23627 + goto out_children;
23628 + a->dst_btop = a->btgt;
23629 + a->dst_h_dentry = au_h_dptr(a->dst_dentry, a->btgt);
23632 + /* lock them all */
23633 + err = au_ren_lock(a);
23634 + if (unlikely(err))
23635 + /* leave the copied-up one */
23636 + goto out_children;
23638 + if (!a->exchange) {
23639 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
23640 + err = au_may_ren(a);
23641 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
23642 + err = -ENAMETOOLONG;
23643 + if (unlikely(err))
23647 + /* store timestamps to be revertible */
23650 + /* store dirren info */
23651 + if (au_ftest_ren(a->auren_flags, DIRREN)) {
23652 + err = au_dr_rename(a->src_dentry, a->btgt,
23653 + &a->dst_dentry->d_name, &rev);
23655 + if (unlikely(err))
23660 + err = do_rename(a);
23661 + if (unlikely(err))
23664 + if (au_ftest_ren(a->auren_flags, DIRREN))
23665 + au_dr_rename_fin(a->src_dentry, a->btgt, rev);
23667 + /* update dir attributes */
23668 + au_ren_refresh_dir(a);
23670 + /* dput/iput all lower dentries */
23671 + au_ren_refresh(a);
23673 + goto out_hdir; /* success */
23676 + if (au_ftest_ren(a->auren_flags, DIRREN))
23677 + au_dr_rename_rev(a->src_dentry, a->btgt, rev);
23679 + au_ren_rev_dt(err, a);
23681 + au_ren_unlock(a);
23683 + au_nhash_wh_free(&a->whlist);
23684 + if (err && a->dst_inode && a->dst_btop != a->btgt) {
23685 + AuDbg("btop %d, btgt %d\n", a->dst_btop, a->btgt);
23686 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
23687 + au_set_dbtop(a->dst_dentry, a->dst_btop);
23691 + if (d_unhashed(a->src_dentry))
23692 + au_fset_ren(a->auren_flags, DROPPED_SRC);
23693 + if (d_unhashed(a->dst_dentry))
23694 + au_fset_ren(a->auren_flags, DROPPED_DST);
23695 + if (!a->exchange)
23696 + d_move(a->src_dentry, a->dst_dentry);
23698 + d_exchange(a->src_dentry, a->dst_dentry);
23699 + if (au_ftest_ren(a->auren_flags, DROPPED_DST))
23700 + d_drop(a->dst_dentry);
23702 + if (au_ftest_ren(a->auren_flags, DROPPED_SRC))
23703 + d_drop(a->src_dentry);
23705 + au_update_dbtop(a->dst_dentry);
23706 + if (!a->dst_inode)
23707 + d_drop(a->dst_dentry);
23709 + if (au_ftest_ren(a->auren_flags, ISSAMEDIR))
23710 + di_write_unlock(a->dst_parent);
23712 + di_write_unlock2(a->src_parent, a->dst_parent);
23714 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
23716 + iput(a->dst_inode);
23718 + au_whtmp_rmdir_free(a->thargs);
23724 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
23725 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
23726 +++ linux/fs/aufs/Kconfig 2018-01-29 07:56:20.049991637 +0100
23728 +# SPDX-License-Identifier: GPL-2.0
23730 + tristate "Aufs (Advanced multi layered unification filesystem) support"
23732 + Aufs is a stackable unification filesystem such as Unionfs,
23733 + which unifies several directories and provides a merged single
23735 + In the early days, aufs was entirely re-designed and
23736 + re-implemented Unionfs Version 1.x series. Introducing many
23737 + original ideas, approaches and improvements, it becomes totally
23738 + different from Unionfs while keeping the basic features.
23742 + prompt "Maximum number of branches"
23743 + default AUFS_BRANCH_MAX_127
23745 + Specifies the maximum number of branches (or member directories)
23746 + in a single aufs. The larger value consumes more system
23747 + resources and has a minor impact to performance.
23748 +config AUFS_BRANCH_MAX_127
23751 + Specifies the maximum number of branches (or member directories)
23752 + in a single aufs. The larger value consumes more system
23753 + resources and has a minor impact to performance.
23754 +config AUFS_BRANCH_MAX_511
23757 + Specifies the maximum number of branches (or member directories)
23758 + in a single aufs. The larger value consumes more system
23759 + resources and has a minor impact to performance.
23760 +config AUFS_BRANCH_MAX_1023
23763 + Specifies the maximum number of branches (or member directories)
23764 + in a single aufs. The larger value consumes more system
23765 + resources and has a minor impact to performance.
23766 +config AUFS_BRANCH_MAX_32767
23769 + Specifies the maximum number of branches (or member directories)
23770 + in a single aufs. The larger value consumes more system
23771 + resources and has a minor impact to performance.
23774 +config AUFS_SBILIST
23776 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
23779 + Automatic configuration for internal use.
23780 + When aufs supports Magic SysRq or /proc, enabled automatically.
23782 +config AUFS_HNOTIFY
23783 + bool "Detect direct branch access (bypassing aufs)"
23785 + If you want to modify files on branches directly, eg. bypassing aufs,
23786 + and want aufs to detect the changes of them fully, then enable this
23787 + option and use 'udba=notify' mount option.
23788 + Currently there is only one available configuration, "fsnotify".
23789 + It will have a negative impact to the performance.
23790 + See detail in aufs.5.
23793 + prompt "method" if AUFS_HNOTIFY
23794 + default AUFS_HFSNOTIFY
23795 +config AUFS_HFSNOTIFY
23800 +config AUFS_EXPORT
23801 + bool "NFS-exportable aufs"
23802 + depends on EXPORTFS
23804 + If you want to export your mounted aufs via NFS, then enable this
23805 + option. There are several requirements for this configuration.
23806 + See detail in aufs.5.
23808 +config AUFS_INO_T_64
23810 + depends on AUFS_EXPORT
23811 + depends on 64BIT && !(ALPHA || S390)
23814 + Automatic configuration for internal use.
23815 + /* typedef unsigned long/int __kernel_ino_t */
23816 + /* alpha and s390x are int */
23819 + bool "support for XATTR/EA (including Security Labels)"
23821 + If your branch fs supports XATTR/EA and you want to make them
23822 + available in aufs too, then enable this opsion and specify the
23823 + branch attributes for EA.
23824 + See detail in aufs.5.
23827 + bool "File-based Hierarchical Storage Management"
23829 + Hierarchical Storage Management (or HSM) is a well-known feature
23830 + in the storage world. Aufs provides this feature as file-based.
23831 + with multiple branches.
23832 + These multiple branches are prioritized, ie. the topmost one
23833 + should be the fastest drive and be used heavily.
23836 + bool "Readdir in userspace"
23838 + Aufs has two methods to provide a merged view for a directory,
23839 + by a user-space library and by kernel-space natively. The latter
23840 + is always enabled but sometimes large and slow.
23841 + If you enable this option, install the library in aufs2-util
23842 + package, and set some environment variables for your readdir(3),
23843 + then the work will be handled in user-space which generally
23844 + shows better performance in most cases.
23845 + See detail in aufs.5.
23847 +config AUFS_DIRREN
23848 + bool "Workaround for rename(2)-ing a directory"
23850 + By default, aufs returns EXDEV error in renameing a dir who has
23851 + his child on the lower branch, since it is a bad idea to issue
23852 + rename(2) internally for every lower branch. But user may not
23853 + accept this behaviour. So here is a workaround to allow such
23854 + rename(2) and store some extra infromation on the writable
23855 + branch. Obviously this costs high (and I don't like it).
23856 + To use this feature, you need to enable this configuration AND
23857 + to specify the mount option `dirren.'
23858 + See details in aufs.5 and the design documents.
23861 + bool "Show whiteouts"
23863 + If you want to make the whiteouts in aufs visible, then enable
23864 + this option and specify 'shwh' mount option. Although it may
23865 + sounds like philosophy or something, but in technically it
23866 + simply shows the name of whiteout with keeping its behaviour.
23868 +config AUFS_BR_RAMFS
23869 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
23871 + If you want to use ramfs as an aufs branch fs, then enable this
23872 + option. Generally tmpfs is recommended.
23873 + Aufs prohibited them to be a branch fs by default, because
23874 + initramfs becomes unusable after switch_root or something
23875 + generally. If you sets initramfs as an aufs branch and boot your
23876 + system by switch_root, you will meet a problem easily since the
23877 + files in initramfs may be inaccessible.
23878 + Unless you are going to use ramfs as an aufs branch fs without
23879 + switch_root or something, leave it N.
23881 +config AUFS_BR_FUSE
23882 + bool "Fuse fs as an aufs branch"
23883 + depends on FUSE_FS
23886 + If you want to use fuse-based userspace filesystem as an aufs
23887 + branch fs, then enable this option.
23888 + It implements the internal poll(2) operation which is
23889 + implemented by fuse only (curretnly).
23894 + Automatic configuration for internal use.
23896 +config AUFS_BR_HFSPLUS
23897 + bool "Hfsplus as an aufs branch"
23898 + depends on HFSPLUS_FS
23901 + If you want to use hfsplus fs as an aufs branch fs, then enable
23902 + this option. This option introduces a small overhead at
23903 + copying-up a file on hfsplus.
23905 +config AUFS_BDEV_LOOP
23907 + depends on BLK_DEV_LOOP
23910 + Automatic configuration for internal use.
23911 + Convert =[ym] into =y.
23914 + bool "Debug aufs"
23916 + Enable this to compile aufs internal debug code.
23917 + It will have a negative impact to the performance.
23919 +config AUFS_MAGIC_SYSRQ
23921 + depends on AUFS_DEBUG && MAGIC_SYSRQ
23924 + Automatic configuration for internal use.
23925 + When aufs supports Magic SysRq, enabled automatically.
23927 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
23928 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
23929 +++ linux/fs/aufs/loop.c 2018-01-29 07:56:20.056658502 +0100
23932 + * Copyright (C) 2005-2017 Junjiro R. Okajima
23934 + * This program, aufs is free software; you can redistribute it and/or modify
23935 + * it under the terms of the GNU General Public License as published by
23936 + * the Free Software Foundation; either version 2 of the License, or
23937 + * (at your option) any later version.
23939 + * This program is distributed in the hope that it will be useful,
23940 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23941 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23942 + * GNU General Public License for more details.
23944 + * You should have received a copy of the GNU General Public License
23945 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
23949 + * support for loopback block device as a branch
23954 +/* added into drivers/block/loop.c */
23955 +static struct file *(*backing_file_func)(struct super_block *sb);
23958 + * test if two lower dentries have overlapping branches.
23960 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
23962 + struct super_block *h_sb;
23963 + struct file *backing_file;
23965 + if (unlikely(!backing_file_func)) {
23966 + /* don't load "loop" module here */
23967 + backing_file_func = symbol_get(loop_backing_file);
23968 + if (unlikely(!backing_file_func))
23969 + /* "loop" module is not loaded */
23973 + h_sb = h_adding->d_sb;
23974 + backing_file = backing_file_func(h_sb);
23975 + if (!backing_file)
23978 + h_adding = backing_file->f_path.dentry;
23980 + * h_adding can be local NFS.
23981 + * in this case aufs cannot detect the loop.
23983 + if (unlikely(h_adding->d_sb == sb))
23985 + return !!au_test_subdir(h_adding, sb->s_root);
23988 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
23989 +int au_test_loopback_kthread(void)
23992 + struct task_struct *tsk = current;
23993 + char c, comm[sizeof(tsk->comm)];
23996 + if (tsk->flags & PF_KTHREAD) {
23997 + get_task_comm(comm, tsk);
23999 + ret = ('0' <= c && c <= '9'
24000 + && !strncmp(comm, "loop", 4));
24006 +/* ---------------------------------------------------------------------- */
24008 +#define au_warn_loopback_step 16
24009 +static int au_warn_loopback_nelem = au_warn_loopback_step;
24010 +static unsigned long *au_warn_loopback_array;
24012 +void au_warn_loopback(struct super_block *h_sb)
24014 + int i, new_nelem;
24015 + unsigned long *a, magic;
24016 + static DEFINE_SPINLOCK(spin);
24018 + magic = h_sb->s_magic;
24019 + spin_lock(&spin);
24020 + a = au_warn_loopback_array;
24021 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
24022 + if (a[i] == magic) {
24023 + spin_unlock(&spin);
24027 + /* h_sb is new to us, print it */
24028 + if (i < au_warn_loopback_nelem) {
24033 + /* expand the array */
24034 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
24035 + a = au_kzrealloc(au_warn_loopback_array,
24036 + au_warn_loopback_nelem * sizeof(unsigned long),
24037 + new_nelem * sizeof(unsigned long), GFP_ATOMIC,
24038 + /*may_shrink*/0);
24040 + au_warn_loopback_nelem = new_nelem;
24041 + au_warn_loopback_array = a;
24046 + spin_unlock(&spin);
24047 + AuWarn1("realloc failed, ignored\n");
24051 + spin_unlock(&spin);
24052 + pr_warn("you may want to try another patch for loopback file "
24053 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
24056 +int au_loopback_init(void)
24059 + struct super_block *sb __maybe_unused;
24061 + BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(unsigned long));
24064 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
24065 + sizeof(unsigned long), GFP_NOFS);
24066 + if (unlikely(!au_warn_loopback_array))
24072 +void au_loopback_fin(void)
24074 + if (backing_file_func)
24075 + symbol_put(loop_backing_file);
24076 + kfree(au_warn_loopback_array);
24078 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
24079 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
24080 +++ linux/fs/aufs/loop.h 2018-01-29 07:56:20.059991935 +0100
24083 + * Copyright (C) 2005-2017 Junjiro R. Okajima
24085 + * This program, aufs is free software; you can redistribute it and/or modify
24086 + * it under the terms of the GNU General Public License as published by
24087 + * the Free Software Foundation; either version 2 of the License, or
24088 + * (at your option) any later version.
24090 + * This program is distributed in the hope that it will be useful,
24091 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24092 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24093 + * GNU General Public License for more details.
24095 + * You should have received a copy of the GNU General Public License
24096 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24100 + * support for loopback mount as a branch
24103 +#ifndef __AUFS_LOOP_H__
24104 +#define __AUFS_LOOP_H__
24109 +struct super_block;
24111 +#ifdef CONFIG_AUFS_BDEV_LOOP
24112 +/* drivers/block/loop.c */
24113 +struct file *loop_backing_file(struct super_block *sb);
24116 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
24117 +int au_test_loopback_kthread(void);
24118 +void au_warn_loopback(struct super_block *h_sb);
24120 +int au_loopback_init(void);
24121 +void au_loopback_fin(void);
24123 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
24124 + struct dentry *h_adding)
24125 +AuStubInt0(au_test_loopback_kthread, void)
24126 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
24128 +AuStubInt0(au_loopback_init, void)
24129 +AuStubVoid(au_loopback_fin, void)
24130 +#endif /* BLK_DEV_LOOP */
24132 +#endif /* __KERNEL__ */
24133 +#endif /* __AUFS_LOOP_H__ */
24134 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
24135 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
24136 +++ linux/fs/aufs/magic.mk 2018-01-29 07:56:20.059991935 +0100
24138 +# SPDX-License-Identifier: GPL-2.0
24140 +# defined in ${srctree}/fs/fuse/inode.c
24142 +ifdef CONFIG_FUSE_FS
24143 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
24146 +# defined in ${srctree}/fs/xfs/xfs_sb.h
24148 +ifdef CONFIG_XFS_FS
24149 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
24152 +# defined in ${srctree}/fs/configfs/mount.c
24154 +ifdef CONFIG_CONFIGFS_FS
24155 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
24158 +# defined in ${srctree}/fs/ubifs/ubifs.h
24160 +ifdef CONFIG_UBIFS_FS
24161 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
24164 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
24166 +ifdef CONFIG_HFSPLUS_FS
24167 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
24169 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
24170 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
24171 +++ linux/fs/aufs/Makefile 2018-01-29 07:56:20.049991637 +0100
24173 +# SPDX-License-Identifier: GPL-2.0
24175 +include ${src}/magic.mk
24176 +ifeq (${CONFIG_AUFS_FS},m)
24177 +include ${src}/conf.mk
24179 +-include ${src}/priv_def.mk
24181 +# cf. include/linux/kernel.h
24183 +ccflags-y += -DDEBUG
24184 +# sparse requires the full pathname
24186 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
24188 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
24191 +obj-$(CONFIG_AUFS_FS) += aufs.o
24192 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o \
24193 + wkq.o vfsub.o dcsub.o \
24194 + cpup.o whout.o wbr_policy.o \
24195 + dinfo.o dentry.o \
24197 + finfo.o file.o f_op.o \
24199 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
24203 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
24204 +aufs-$(CONFIG_SYSFS) += sysfs.o
24205 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
24206 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
24207 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
24208 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
24209 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
24210 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
24211 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
24212 +aufs-$(CONFIG_AUFS_DIRREN) += dirren.o
24213 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
24214 +aufs-$(CONFIG_AUFS_POLL) += poll.o
24215 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
24216 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
24217 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
24218 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
24219 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
24220 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
24221 +++ linux/fs/aufs/module.c 2018-01-29 07:56:20.059991935 +0100
24224 + * Copyright (C) 2005-2017 Junjiro R. Okajima
24226 + * This program, aufs is free software; you can redistribute it and/or modify
24227 + * it under the terms of the GNU General Public License as published by
24228 + * the Free Software Foundation; either version 2 of the License, or
24229 + * (at your option) any later version.
24231 + * This program is distributed in the hope that it will be useful,
24232 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24233 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24234 + * GNU General Public License for more details.
24236 + * You should have received a copy of the GNU General Public License
24237 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24241 + * module global variables and operations
24244 +#include <linux/module.h>
24245 +#include <linux/seq_file.h>
24248 +/* shrinkable realloc */
24249 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink)
24257 +#if 0 /* unused */
24264 + AuDebugOn(!new_sz);
24267 + diff = au_kmidx_sub(sz, new_sz);
24273 + /* expand or SLOB */
24274 + p = krealloc(p, new_sz, gfp);
24275 + else if (new_sz < sz && may_shrink) {
24279 + q = kmalloc(new_sz, gfp);
24282 + memcpy(q, p, new_sz);
24294 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
24297 + p = au_krealloc(p, new_sz, gfp, may_shrink);
24298 + if (p && new_sz > nused)
24299 + memset(p + nused, 0, new_sz - nused);
24303 +/* ---------------------------------------------------------------------- */
24307 +struct kmem_cache *au_cache[AuCache_Last];
24309 +static void au_cache_fin(void)
24314 + * Make sure all delayed rcu free inodes are flushed before we
24319 + /* excluding AuCache_HNOTIFY */
24320 + BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
24321 + for (i = 0; i < AuCache_HNOTIFY; i++) {
24322 + kmem_cache_destroy(au_cache[i]);
24323 + au_cache[i] = NULL;
24327 +static int __init au_cache_init(void)
24329 + au_cache[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
24330 + if (au_cache[AuCache_DINFO])
24331 + /* SLAB_DESTROY_BY_RCU */
24332 + au_cache[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
24333 + au_icntnr_init_once);
24334 + if (au_cache[AuCache_ICNTNR])
24335 + au_cache[AuCache_FINFO] = AuCacheCtor(au_finfo,
24336 + au_fi_init_once);
24337 + if (au_cache[AuCache_FINFO])
24338 + au_cache[AuCache_VDIR] = AuCache(au_vdir);
24339 + if (au_cache[AuCache_VDIR])
24340 + au_cache[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
24341 + if (au_cache[AuCache_DEHSTR])
24348 +/* ---------------------------------------------------------------------- */
24350 +int au_dir_roflags;
24352 +#ifdef CONFIG_AUFS_SBILIST
24354 + * iterate_supers_type() doesn't protect us from
24355 + * remounting (branch management)
24357 +struct hlist_bl_head au_sbilist;
24361 + * functions for module interface.
24363 +MODULE_LICENSE("GPL");
24364 +/* MODULE_LICENSE("GPL v2"); */
24365 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
24366 +MODULE_DESCRIPTION(AUFS_NAME
24367 + " -- Advanced multi layered unification filesystem");
24368 +MODULE_VERSION(AUFS_VERSION);
24369 +MODULE_ALIAS_FS(AUFS_NAME);
24371 +/* this module parameter has no meaning when SYSFS is disabled */
24372 +int sysaufs_brs = 1;
24373 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
24374 +module_param_named(brs, sysaufs_brs, int, S_IRUGO);
24376 +/* this module parameter has no meaning when USER_NS is disabled */
24378 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
24379 +module_param_named(allow_userns, au_userns, bool, S_IRUGO);
24381 +/* ---------------------------------------------------------------------- */
24383 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
24385 +int au_seq_path(struct seq_file *seq, struct path *path)
24389 + err = seq_path(seq, path, au_esc_chars);
24398 +/* ---------------------------------------------------------------------- */
24400 +static int __init aufs_init(void)
24405 + p = au_esc_chars;
24406 + for (i = 1; i <= ' '; i++)
24412 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
24414 + memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
24415 + for (i = 0; i < AuIop_Last; i++)
24416 + aufs_iop_nogetattr[i].getattr = NULL;
24418 + memset(au_cache, 0, sizeof(au_cache)); /* including hnotify */
24420 + au_sbilist_init();
24421 + sysaufs_brs_init();
24424 + err = sysaufs_init();
24425 + if (unlikely(err))
24427 + err = au_procfs_init();
24428 + if (unlikely(err))
24429 + goto out_sysaufs;
24430 + err = au_wkq_init();
24431 + if (unlikely(err))
24433 + err = au_loopback_init();
24434 + if (unlikely(err))
24436 + err = au_hnotify_init();
24437 + if (unlikely(err))
24438 + goto out_loopback;
24439 + err = au_sysrq_init();
24440 + if (unlikely(err))
24442 + err = au_cache_init();
24443 + if (unlikely(err))
24446 + aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
24447 + err = register_filesystem(&aufs_fs_type);
24448 + if (unlikely(err))
24451 + /* since we define pr_fmt, call printk directly */
24452 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
24453 + goto out; /* success */
24460 + au_hnotify_fin();
24462 + au_loopback_fin();
24474 +static void __exit aufs_exit(void)
24476 + unregister_filesystem(&aufs_fs_type);
24479 + au_hnotify_fin();
24480 + au_loopback_fin();
24487 +module_init(aufs_init);
24488 +module_exit(aufs_exit);
24489 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
24490 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
24491 +++ linux/fs/aufs/module.h 2018-01-29 07:56:20.059991935 +0100
24494 + * Copyright (C) 2005-2017 Junjiro R. Okajima
24496 + * This program, aufs is free software; you can redistribute it and/or modify
24497 + * it under the terms of the GNU General Public License as published by
24498 + * the Free Software Foundation; either version 2 of the License, or
24499 + * (at your option) any later version.
24501 + * This program is distributed in the hope that it will be useful,
24502 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24503 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24504 + * GNU General Public License for more details.
24506 + * You should have received a copy of the GNU General Public License
24507 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24511 + * module initialization and module-global
24514 +#ifndef __AUFS_MODULE_H__
24515 +#define __AUFS_MODULE_H__
24519 +#include <linux/slab.h>
24524 +/* module parameters */
24525 +extern int sysaufs_brs;
24526 +extern bool au_userns;
24528 +/* ---------------------------------------------------------------------- */
24530 +extern int au_dir_roflags;
24532 +void *au_krealloc(void *p, unsigned int new_sz, gfp_t gfp, int may_shrink);
24533 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp,
24536 +static inline int au_kmidx_sub(size_t sz, size_t new_sz)
24538 +#ifndef CONFIG_SLOB
24539 + return kmalloc_index(sz) - kmalloc_index(new_sz);
24541 + return -1; /* SLOB is untested */
24545 +int au_seq_path(struct seq_file *seq, struct path *path);
24547 +#ifdef CONFIG_PROC_FS
24549 +int __init au_procfs_init(void);
24550 +void au_procfs_fin(void);
24552 +AuStubInt0(au_procfs_init, void);
24553 +AuStubVoid(au_procfs_fin, void);
24556 +/* ---------------------------------------------------------------------- */
24565 + AuCache_HNOTIFY, /* must be last */
24569 +extern struct kmem_cache *au_cache[AuCache_Last];
24571 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
24572 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
24573 +#define AuCacheCtor(type, ctor) \
24574 + kmem_cache_create(#type, sizeof(struct type), \
24575 + __alignof__(struct type), AuCacheFlags, ctor)
24577 +#define AuCacheFuncs(name, index) \
24578 +static inline struct au_##name *au_cache_alloc_##name(void) \
24579 +{ return kmem_cache_alloc(au_cache[AuCache_##index], GFP_NOFS); } \
24580 +static inline void au_cache_free_##name(struct au_##name *p) \
24581 +{ kmem_cache_free(au_cache[AuCache_##index], p); }
24583 +AuCacheFuncs(dinfo, DINFO);
24584 +AuCacheFuncs(icntnr, ICNTNR);
24585 +AuCacheFuncs(finfo, FINFO);
24586 +AuCacheFuncs(vdir, VDIR);
24587 +AuCacheFuncs(vdir_dehstr, DEHSTR);
24588 +#ifdef CONFIG_AUFS_HNOTIFY
24589 +AuCacheFuncs(hnotify, HNOTIFY);
24592 +#endif /* __KERNEL__ */
24593 +#endif /* __AUFS_MODULE_H__ */
24594 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
24595 --- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
24596 +++ linux/fs/aufs/mvdown.c 2018-01-29 07:56:20.059991935 +0100
24599 + * Copyright (C) 2011-2017 Junjiro R. Okajima
24601 + * This program, aufs is free software; you can redistribute it and/or modify
24602 + * it under the terms of the GNU General Public License as published by
24603 + * the Free Software Foundation; either version 2 of the License, or
24604 + * (at your option) any later version.
24606 + * This program is distributed in the hope that it will be useful,
24607 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24608 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24609 + * GNU General Public License for more details.
24611 + * You should have received a copy of the GNU General Public License
24612 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24616 + * move-down, opposite of copy-up
24621 +struct au_mvd_args {
24623 + struct super_block *h_sb;
24624 + struct dentry *h_parent;
24625 + struct au_hinode *hdir;
24626 + struct inode *h_dir, *h_inode;
24627 + struct au_pin pin;
24628 + } info[AUFS_MVDOWN_NARRAY];
24630 + struct aufs_mvdown mvdown;
24631 + struct dentry *dentry, *parent;
24632 + struct inode *inode, *dir;
24633 + struct super_block *sb;
24634 + aufs_bindex_t bopq, bwh, bfound;
24635 + unsigned char rename_lock;
24638 +#define mvd_errno mvdown.au_errno
24639 +#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
24640 +#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
24641 +#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
24642 +#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
24644 +#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
24645 +#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
24646 +#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
24647 +#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
24648 +#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
24649 +#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
24651 +#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
24652 +#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
24653 +#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
24654 +#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
24655 +#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
24656 +#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
24658 +#define AU_MVD_PR(flag, ...) do { \
24660 + pr_err(__VA_ARGS__); \
24663 +static int find_lower_writable(struct au_mvd_args *a)
24665 + struct super_block *sb;
24666 + aufs_bindex_t bindex, bbot;
24667 + struct au_branch *br;
24670 + bindex = a->mvd_bsrc;
24671 + bbot = au_sbbot(sb);
24672 + if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
24673 + for (bindex++; bindex <= bbot; bindex++) {
24674 + br = au_sbr(sb, bindex);
24675 + if (au_br_fhsm(br->br_perm)
24676 + && !sb_rdonly(au_br_sb(br)))
24679 + else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
24680 + for (bindex++; bindex <= bbot; bindex++) {
24681 + br = au_sbr(sb, bindex);
24682 + if (!au_br_rdonly(br))
24686 + for (bindex++; bindex <= bbot; bindex++) {
24687 + br = au_sbr(sb, bindex);
24688 + if (!sb_rdonly(au_br_sb(br))) {
24689 + if (au_br_rdonly(br))
24691 + |= AUFS_MVDOWN_ROLOWER_R;
24699 +/* make the parent dir on bdst */
24700 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
24705 + a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
24706 + a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
24707 + a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
24708 + a->mvd_h_dst_parent = NULL;
24709 + if (au_dbbot(a->parent) >= a->mvd_bdst)
24710 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
24711 + if (!a->mvd_h_dst_parent) {
24712 + err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
24713 + if (unlikely(err)) {
24714 + AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
24717 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
24725 +/* lock them all */
24726 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
24729 + struct dentry *h_trap;
24731 + a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
24732 + a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
24733 + err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
24734 + au_opt_udba(a->sb),
24735 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
24737 + if (unlikely(err)) {
24738 + AU_MVD_PR(dmsg, "pin_dst failed\n");
24742 + if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
24743 + a->rename_lock = 0;
24744 + au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
24745 + AuLsc_DI_PARENT, AuLsc_I_PARENT3,
24746 + au_opt_udba(a->sb),
24747 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
24748 + err = au_do_pin(&a->mvd_pin_src);
24750 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
24751 + if (unlikely(err)) {
24752 + AU_MVD_PR(dmsg, "pin_src failed\n");
24755 + goto out; /* success */
24758 + a->rename_lock = 1;
24759 + au_pin_hdir_unlock(&a->mvd_pin_dst);
24760 + err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
24761 + au_opt_udba(a->sb),
24762 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
24764 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
24765 + if (unlikely(err)) {
24766 + AU_MVD_PR(dmsg, "pin_src failed\n");
24767 + au_pin_hdir_lock(&a->mvd_pin_dst);
24770 + au_pin_hdir_unlock(&a->mvd_pin_src);
24771 + h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
24772 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
24774 + err = (h_trap != a->mvd_h_src_parent);
24776 + err = (h_trap != a->mvd_h_dst_parent);
24778 + BUG_ON(err); /* it should never happen */
24779 + if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
24782 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
24783 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
24784 + au_pin_hdir_lock(&a->mvd_pin_src);
24785 + au_unpin(&a->mvd_pin_src);
24786 + au_pin_hdir_lock(&a->mvd_pin_dst);
24789 + goto out; /* success */
24792 + au_unpin(&a->mvd_pin_dst);
24798 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
24800 + if (!a->rename_lock)
24801 + au_unpin(&a->mvd_pin_src);
24803 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
24804 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
24805 + au_pin_hdir_lock(&a->mvd_pin_src);
24806 + au_unpin(&a->mvd_pin_src);
24807 + au_pin_hdir_lock(&a->mvd_pin_dst);
24809 + au_unpin(&a->mvd_pin_dst);
24812 +/* copy-down the file */
24813 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
24816 + struct au_cp_generic cpg = {
24817 + .dentry = a->dentry,
24818 + .bdst = a->mvd_bdst,
24819 + .bsrc = a->mvd_bsrc,
24821 + .pin = &a->mvd_pin_dst,
24822 + .flags = AuCpup_DTIME | AuCpup_HOPEN
24825 + AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
24826 + if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
24827 + au_fset_cpup(cpg.flags, OVERWRITE);
24828 + if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
24829 + au_fset_cpup(cpg.flags, RWDST);
24830 + err = au_sio_cpdown_simple(&cpg);
24831 + if (unlikely(err))
24832 + AU_MVD_PR(dmsg, "cpdown failed\n");
24839 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
24842 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
24845 + struct path h_path;
24846 + struct au_branch *br;
24847 + struct inode *delegated;
24849 + br = au_sbr(a->sb, a->mvd_bdst);
24850 + h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
24851 + err = PTR_ERR(h_path.dentry);
24852 + if (IS_ERR(h_path.dentry)) {
24853 + AU_MVD_PR(dmsg, "wh_lkup failed\n");
24858 + if (d_is_positive(h_path.dentry)) {
24859 + h_path.mnt = au_br_mnt(br);
24860 + delegated = NULL;
24861 + err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
24862 + &delegated, /*force*/0);
24863 + if (unlikely(err == -EWOULDBLOCK)) {
24864 + pr_warn("cannot retry for NFSv4 delegation"
24865 + " for an internal unlink\n");
24868 + if (unlikely(err))
24869 + AU_MVD_PR(dmsg, "wh_unlink failed\n");
24871 + dput(h_path.dentry);
24879 + * unlink the topmost h_dentry
24881 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
24884 + struct path h_path;
24885 + struct inode *delegated;
24887 + h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
24888 + h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
24889 + delegated = NULL;
24890 + err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
24891 + if (unlikely(err == -EWOULDBLOCK)) {
24892 + pr_warn("cannot retry for NFSv4 delegation"
24893 + " for an internal unlink\n");
24896 + if (unlikely(err))
24897 + AU_MVD_PR(dmsg, "unlink failed\n");
24903 +/* Since mvdown succeeded, we ignore an error of this function */
24904 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
24907 + struct au_branch *br;
24909 + a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
24910 + br = au_sbr(a->sb, a->mvd_bsrc);
24911 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
24913 + br = au_sbr(a->sb, a->mvd_bdst);
24914 + a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
24915 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
24918 + a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
24920 + AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
24924 + * copy-down the file and unlink the bsrc file.
24925 + * - unlink the bdst whout if exist
24926 + * - copy-down the file (with whtmp name and rename)
24927 + * - unlink the bsrc file
24929 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
24933 + err = au_do_mkdir(dmsg, a);
24935 + err = au_do_lock(dmsg, a);
24936 + if (unlikely(err))
24940 + * do not revert the activities we made on bdst since they should be
24941 + * harmless in aufs.
24944 + err = au_do_cpdown(dmsg, a);
24946 + err = au_do_unlink_wh(dmsg, a);
24947 + if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
24948 + err = au_do_unlink(dmsg, a);
24949 + if (unlikely(err))
24952 + AuDbg("%pd2, 0x%x, %d --> %d\n",
24953 + a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
24954 + if (find_lower_writable(a) < 0)
24955 + a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
24957 + if (a->mvdown.flags & AUFS_MVDOWN_STFS)
24958 + au_do_stfs(dmsg, a);
24960 + /* maintain internal array */
24961 + if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
24962 + au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
24963 + au_set_dbtop(a->dentry, a->mvd_bdst);
24964 + au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
24965 + au_set_ibtop(a->inode, a->mvd_bdst);
24967 + /* hide the lower */
24968 + au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
24969 + au_set_dbbot(a->dentry, a->mvd_bsrc);
24970 + au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
24971 + au_set_ibbot(a->inode, a->mvd_bsrc);
24973 + if (au_dbbot(a->dentry) < a->mvd_bdst)
24974 + au_set_dbbot(a->dentry, a->mvd_bdst);
24975 + if (au_ibbot(a->inode) < a->mvd_bdst)
24976 + au_set_ibbot(a->inode, a->mvd_bdst);
24979 + au_do_unlock(dmsg, a);
24985 +/* ---------------------------------------------------------------------- */
24987 +/* make sure the file is idle */
24988 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
24990 + int err, plinked;
24993 + plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
24994 + if (au_dbtop(a->dentry) == a->mvd_bsrc
24995 + && au_dcount(a->dentry) == 1
24996 + && atomic_read(&a->inode->i_count) == 1
24997 + /* && a->mvd_h_src_inode->i_nlink == 1 */
24998 + && (!plinked || !au_plink_test(a->inode))
24999 + && a->inode->i_nlink == 1)
25004 + "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
25005 + a->mvd_bsrc, au_dbtop(a->dentry), au_dcount(a->dentry),
25006 + atomic_read(&a->inode->i_count), a->inode->i_nlink,
25007 + a->mvd_h_src_inode->i_nlink,
25008 + plinked, plinked ? au_plink_test(a->inode) : 0);
25015 +/* make sure the parent dir is fine */
25016 +static int au_mvd_args_parent(const unsigned char dmsg,
25017 + struct au_mvd_args *a)
25020 + aufs_bindex_t bindex;
25023 + if (unlikely(au_alive_dir(a->parent))) {
25025 + AU_MVD_PR(dmsg, "parent dir is dead\n");
25029 + a->bopq = au_dbdiropq(a->parent);
25030 + bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
25031 + AuDbg("b%d\n", bindex);
25032 + if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
25033 + || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
25035 + a->mvd_errno = EAU_MVDOWN_OPAQUE;
25036 + AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
25037 + a->bopq, a->mvd_bdst);
25045 +static int au_mvd_args_intermediate(const unsigned char dmsg,
25046 + struct au_mvd_args *a)
25049 + struct au_dinfo *dinfo, *tmp;
25051 + /* lookup the next lower positive entry */
25053 + tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
25054 + if (unlikely(!tmp))
25059 + dinfo = au_di(a->dentry);
25060 + au_di_cp(tmp, dinfo);
25061 + au_di_swap(tmp, dinfo);
25063 + /* returns the number of positive dentries */
25064 + err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1,
25065 + /* AuLkup_IGNORE_PERM */ 0);
25067 + a->bwh = au_dbwh(a->dentry);
25068 + else if (err > 0)
25069 + a->bfound = au_dbtop(a->dentry);
25071 + au_di_swap(tmp, dinfo);
25072 + au_rw_write_unlock(&tmp->di_rwsem);
25074 + if (unlikely(err < 0))
25075 + AU_MVD_PR(dmsg, "failed look-up lower\n");
25078 + * here, we have these cases.
25080 + * no positive dentry under bsrc. there are more sub-cases.
25082 + * there no whiteout, we can safely move-down.
25085 + * bsrc < bwh && bwh < bdst
25086 + * there is a whiteout on RO branch. cannot proceed.
25088 + * there is a whiteout on the RW target branch. it should
25091 + * there is a whiteout somewhere unrelated branch.
25092 + * -1 < bfound && bfound <= bsrc
25095 + * found, but it is on RO branch between bsrc and bdst. cannot
25098 + * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
25101 + * found, after we create the file on bdst, it will be hidden.
25104 + AuDebugOn(a->bfound == -1
25106 + && a->bwh <= a->mvd_bsrc);
25107 + AuDebugOn(-1 < a->bfound
25108 + && a->bfound <= a->mvd_bsrc);
25111 + if (a->bfound == -1
25112 + && a->mvd_bsrc < a->bwh
25114 + && a->bwh < a->mvd_bdst) {
25115 + a->mvd_errno = EAU_MVDOWN_WHITEOUT;
25116 + AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
25117 + a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
25119 + } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
25120 + a->mvd_errno = EAU_MVDOWN_UPPER;
25121 + AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
25122 + a->mvd_bdst, a->bfound);
25126 + err = 0; /* success */
25133 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
25138 + if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
25139 + && a->bfound == a->mvd_bdst)
25145 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
25148 + struct au_branch *br;
25151 + if (unlikely(S_ISDIR(a->inode->i_mode)))
25155 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
25156 + a->mvd_bsrc = au_ibtop(a->inode);
25158 + a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
25159 + if (unlikely(a->mvd_bsrc < 0
25160 + || (a->mvd_bsrc < au_dbtop(a->dentry)
25161 + || au_dbbot(a->dentry) < a->mvd_bsrc
25162 + || !au_h_dptr(a->dentry, a->mvd_bsrc))
25163 + || (a->mvd_bsrc < au_ibtop(a->inode)
25164 + || au_ibbot(a->inode) < a->mvd_bsrc
25165 + || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
25166 + a->mvd_errno = EAU_MVDOWN_NOUPPER;
25167 + AU_MVD_PR(dmsg, "no upper\n");
25171 + if (unlikely(a->mvd_bsrc == au_sbbot(a->sb))) {
25172 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
25173 + AU_MVD_PR(dmsg, "on the bottom\n");
25176 + a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
25177 + br = au_sbr(a->sb, a->mvd_bsrc);
25178 + err = au_br_rdonly(br);
25179 + if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
25180 + if (unlikely(err))
25182 + } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
25183 + || IS_APPEND(a->mvd_h_src_inode))) {
25185 + a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
25191 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
25192 + a->mvd_bdst = find_lower_writable(a);
25193 + if (unlikely(a->mvd_bdst < 0)) {
25194 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
25195 + AU_MVD_PR(dmsg, "no writable lower branch\n");
25199 + a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
25200 + if (unlikely(a->mvd_bdst < 0
25201 + || au_sbbot(a->sb) < a->mvd_bdst)) {
25202 + a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
25203 + AU_MVD_PR(dmsg, "no lower brid\n");
25208 + err = au_mvd_args_busy(dmsg, a);
25210 + err = au_mvd_args_parent(dmsg, a);
25212 + err = au_mvd_args_intermediate(dmsg, a);
25214 + err = au_mvd_args_exist(dmsg, a);
25216 + AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
25223 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
25226 + unsigned char dmsg;
25227 + struct au_mvd_args *args;
25228 + struct inode *inode;
25230 + inode = d_inode(dentry);
25232 + if (unlikely(!capable(CAP_SYS_ADMIN)))
25236 + args = kmalloc(sizeof(*args), GFP_NOFS);
25237 + if (unlikely(!args))
25240 + err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
25242 + err = !access_ok(VERIFY_WRITE, uarg, sizeof(*uarg));
25243 + if (unlikely(err)) {
25248 + AuDbg("flags 0x%x\n", args->mvdown.flags);
25249 + args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
25250 + args->mvdown.au_errno = 0;
25251 + args->dentry = dentry;
25252 + args->inode = inode;
25253 + args->sb = dentry->d_sb;
25256 + dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
25257 + args->parent = dget_parent(dentry);
25258 + args->dir = d_inode(args->parent);
25259 + inode_lock_nested(args->dir, I_MUTEX_PARENT);
25260 + dput(args->parent);
25261 + if (unlikely(args->parent != dentry->d_parent)) {
25262 + AU_MVD_PR(dmsg, "parent dir is moved\n");
25266 + inode_lock_nested(inode, I_MUTEX_CHILD);
25267 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
25268 + if (unlikely(err))
25271 + di_write_lock_parent(args->parent);
25272 + err = au_mvd_args(dmsg, args);
25273 + if (unlikely(err))
25276 + err = au_do_mvdown(dmsg, args);
25277 + if (unlikely(err))
25280 + au_cpup_attr_timesizes(args->dir);
25281 + au_cpup_attr_timesizes(inode);
25282 + if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
25283 + au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
25284 + /* au_digen_dec(dentry); */
25287 + di_write_unlock(args->parent);
25288 + aufs_read_unlock(dentry, AuLock_DW);
25290 + inode_unlock(inode);
25292 + inode_unlock(args->dir);
25294 + e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
25302 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
25303 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
25304 +++ linux/fs/aufs/opts.c 2018-01-29 07:56:20.059991935 +0100
25307 + * Copyright (C) 2005-2017 Junjiro R. Okajima
25309 + * This program, aufs is free software; you can redistribute it and/or modify
25310 + * it under the terms of the GNU General Public License as published by
25311 + * the Free Software Foundation; either version 2 of the License, or
25312 + * (at your option) any later version.
25314 + * This program is distributed in the hope that it will be useful,
25315 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25316 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25317 + * GNU General Public License for more details.
25319 + * You should have received a copy of the GNU General Public License
25320 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25324 + * mount options/flags
25327 +#include <linux/namei.h>
25328 +#include <linux/types.h> /* a distribution requires */
25329 +#include <linux/parser.h>
25332 +/* ---------------------------------------------------------------------- */
25336 + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
25337 + Opt_idel, Opt_imod,
25338 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
25339 + Opt_rdblk_def, Opt_rdhash_def,
25340 + Opt_xino, Opt_noxino,
25341 + Opt_trunc_xino, Opt_trunc_xino_v, Opt_notrunc_xino,
25342 + Opt_trunc_xino_path, Opt_itrunc_xino,
25343 + Opt_trunc_xib, Opt_notrunc_xib,
25344 + Opt_shwh, Opt_noshwh,
25345 + Opt_plink, Opt_noplink, Opt_list_plink,
25347 + Opt_dio, Opt_nodio,
25348 + Opt_diropq_a, Opt_diropq_w,
25349 + Opt_warn_perm, Opt_nowarn_perm,
25350 + Opt_wbr_copyup, Opt_wbr_create,
25352 + Opt_verbose, Opt_noverbose,
25353 + Opt_sum, Opt_nosum, Opt_wsum,
25354 + Opt_dirperm1, Opt_nodirperm1,
25355 + Opt_dirren, Opt_nodirren,
25356 + Opt_acl, Opt_noacl,
25357 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
25360 +static match_table_t options = {
25361 + {Opt_br, "br=%s"},
25362 + {Opt_br, "br:%s"},
25364 + {Opt_add, "add=%d:%s"},
25365 + {Opt_add, "add:%d:%s"},
25366 + {Opt_add, "ins=%d:%s"},
25367 + {Opt_add, "ins:%d:%s"},
25368 + {Opt_append, "append=%s"},
25369 + {Opt_append, "append:%s"},
25370 + {Opt_prepend, "prepend=%s"},
25371 + {Opt_prepend, "prepend:%s"},
25373 + {Opt_del, "del=%s"},
25374 + {Opt_del, "del:%s"},
25375 + /* {Opt_idel, "idel:%d"}, */
25376 + {Opt_mod, "mod=%s"},
25377 + {Opt_mod, "mod:%s"},
25378 + /* {Opt_imod, "imod:%d:%s"}, */
25380 + {Opt_dirwh, "dirwh=%d"},
25382 + {Opt_xino, "xino=%s"},
25383 + {Opt_noxino, "noxino"},
25384 + {Opt_trunc_xino, "trunc_xino"},
25385 + {Opt_trunc_xino_v, "trunc_xino_v=%d:%d"},
25386 + {Opt_notrunc_xino, "notrunc_xino"},
25387 + {Opt_trunc_xino_path, "trunc_xino=%s"},
25388 + {Opt_itrunc_xino, "itrunc_xino=%d"},
25389 + /* {Opt_zxino, "zxino=%s"}, */
25390 + {Opt_trunc_xib, "trunc_xib"},
25391 + {Opt_notrunc_xib, "notrunc_xib"},
25393 +#ifdef CONFIG_PROC_FS
25394 + {Opt_plink, "plink"},
25396 + {Opt_ignore_silent, "plink"},
25399 + {Opt_noplink, "noplink"},
25401 +#ifdef CONFIG_AUFS_DEBUG
25402 + {Opt_list_plink, "list_plink"},
25405 + {Opt_udba, "udba=%s"},
25407 + {Opt_dio, "dio"},
25408 + {Opt_nodio, "nodio"},
25410 +#ifdef CONFIG_AUFS_DIRREN
25411 + {Opt_dirren, "dirren"},
25412 + {Opt_nodirren, "nodirren"},
25414 + {Opt_ignore, "dirren"},
25415 + {Opt_ignore_silent, "nodirren"},
25418 +#ifdef CONFIG_AUFS_FHSM
25419 + {Opt_fhsm_sec, "fhsm_sec=%d"},
25421 + {Opt_ignore, "fhsm_sec=%d"},
25424 + {Opt_diropq_a, "diropq=always"},
25425 + {Opt_diropq_a, "diropq=a"},
25426 + {Opt_diropq_w, "diropq=whiteouted"},
25427 + {Opt_diropq_w, "diropq=w"},
25429 + {Opt_warn_perm, "warn_perm"},
25430 + {Opt_nowarn_perm, "nowarn_perm"},
25432 + /* keep them temporary */
25433 + {Opt_ignore_silent, "nodlgt"},
25434 + {Opt_ignore, "clean_plink"},
25436 +#ifdef CONFIG_AUFS_SHWH
25437 + {Opt_shwh, "shwh"},
25439 + {Opt_noshwh, "noshwh"},
25441 + {Opt_dirperm1, "dirperm1"},
25442 + {Opt_nodirperm1, "nodirperm1"},
25444 + {Opt_verbose, "verbose"},
25445 + {Opt_verbose, "v"},
25446 + {Opt_noverbose, "noverbose"},
25447 + {Opt_noverbose, "quiet"},
25448 + {Opt_noverbose, "q"},
25449 + {Opt_noverbose, "silent"},
25451 + {Opt_sum, "sum"},
25452 + {Opt_nosum, "nosum"},
25453 + {Opt_wsum, "wsum"},
25455 + {Opt_rdcache, "rdcache=%d"},
25456 + {Opt_rdblk, "rdblk=%d"},
25457 + {Opt_rdblk_def, "rdblk=def"},
25458 + {Opt_rdhash, "rdhash=%d"},
25459 + {Opt_rdhash_def, "rdhash=def"},
25461 + {Opt_wbr_create, "create=%s"},
25462 + {Opt_wbr_create, "create_policy=%s"},
25463 + {Opt_wbr_copyup, "cpup=%s"},
25464 + {Opt_wbr_copyup, "copyup=%s"},
25465 + {Opt_wbr_copyup, "copyup_policy=%s"},
25467 + /* generic VFS flag */
25468 +#ifdef CONFIG_FS_POSIX_ACL
25469 + {Opt_acl, "acl"},
25470 + {Opt_noacl, "noacl"},
25472 + {Opt_ignore, "acl"},
25473 + {Opt_ignore_silent, "noacl"},
25476 + /* internal use for the scripts */
25477 + {Opt_ignore_silent, "si=%s"},
25479 + {Opt_br, "dirs=%s"},
25480 + {Opt_ignore, "debug=%d"},
25481 + {Opt_ignore, "delete=whiteout"},
25482 + {Opt_ignore, "delete=all"},
25483 + {Opt_ignore, "imap=%s"},
25485 + /* temporary workaround, due to old mount(8)? */
25486 + {Opt_ignore_silent, "relatime"},
25491 +/* ---------------------------------------------------------------------- */
25493 +static const char *au_parser_pattern(int val, match_table_t tbl)
25495 + struct match_token *p;
25498 + while (p->pattern) {
25499 + if (p->token == val)
25500 + return p->pattern;
25507 +static const char *au_optstr(int *val, match_table_t tbl)
25509 + struct match_token *p;
25516 + while (p->pattern) {
25518 + && (v & p->token) == p->token) {
25519 + *val &= ~p->token;
25520 + return p->pattern;
25529 +/* ---------------------------------------------------------------------- */
25531 +static match_table_t brperm = {
25532 + {AuBrPerm_RO, AUFS_BRPERM_RO},
25533 + {AuBrPerm_RR, AUFS_BRPERM_RR},
25534 + {AuBrPerm_RW, AUFS_BRPERM_RW},
25538 +static match_table_t brattr = {
25540 + {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
25541 + {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
25542 + /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
25543 + {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
25544 +#ifdef CONFIG_AUFS_FHSM
25545 + {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
25547 +#ifdef CONFIG_AUFS_XATTR
25548 + {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
25549 + {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
25550 + {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
25551 + {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
25552 + {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
25553 + {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
25556 + /* ro/rr branch */
25557 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
25560 + {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
25561 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
25566 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
25573 + p = strchr(str, '+');
25576 + v = match_token(str, table, args);
25578 + if (v & AuBrAttr_CMOO_Mask)
25579 + attr &= ~AuBrAttr_CMOO_Mask;
25584 + pr_warn("ignored branch attribute %s\n", str);
25594 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
25602 + p = au_optstr(&perm, brattr);
25605 + memcpy(q, p, sz + 1);
25611 + p = au_optstr(&perm, brattr);
25615 + memcpy(q, p, sz + 1);
25621 + return q - str->a;
25624 +static int noinline_for_stack br_perm_val(char *perm)
25626 + int val, bad, sz;
25628 + substring_t args[MAX_OPT_ARGS];
25629 + au_br_perm_str_t attr;
25631 + p = strchr(perm, '+');
25634 + val = match_token(perm, brperm, args);
25638 + pr_warn("ignored branch permission %s\n", perm);
25639 + val = AuBrPerm_RO;
25645 + val |= br_attr_val(p + 1, brattr, args);
25648 + switch (val & AuBrPerm_Mask) {
25649 + case AuBrPerm_RO:
25650 + case AuBrPerm_RR:
25651 + bad = val & AuBrWAttr_Mask;
25652 + val &= ~AuBrWAttr_Mask;
25654 + case AuBrPerm_RW:
25655 + bad = val & AuBrRAttr_Mask;
25656 + val &= ~AuBrRAttr_Mask;
25661 + * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
25662 + * does not treat it as an error, just warning.
25663 + * this is a tiny guard for the user operation.
25665 + if (val & AuBrAttr_UNPIN) {
25666 + bad |= AuBrAttr_UNPIN;
25667 + val &= ~AuBrAttr_UNPIN;
25670 + if (unlikely(bad)) {
25671 + sz = au_do_optstr_br_attr(&attr, bad);
25673 + pr_warn("ignored branch attribute %s\n", attr.a);
25680 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
25682 + au_br_perm_str_t attr;
25688 + p = au_optstr(&perm, brperm);
25689 + AuDebugOn(!p || !*p);
25691 + memcpy(q, p, sz + 1);
25694 + sz = au_do_optstr_br_attr(&attr, perm);
25697 + memcpy(q, attr.a, sz + 1);
25700 + AuDebugOn(strlen(str->a) >= sizeof(str->a));
25703 +/* ---------------------------------------------------------------------- */
25705 +static match_table_t udbalevel = {
25706 + {AuOpt_UDBA_REVAL, "reval"},
25707 + {AuOpt_UDBA_NONE, "none"},
25708 +#ifdef CONFIG_AUFS_HNOTIFY
25709 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
25710 +#ifdef CONFIG_AUFS_HFSNOTIFY
25711 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
25717 +static int noinline_for_stack udba_val(char *str)
25719 + substring_t args[MAX_OPT_ARGS];
25721 + return match_token(str, udbalevel, args);
25724 +const char *au_optstr_udba(int udba)
25726 + return au_parser_pattern(udba, udbalevel);
25729 +/* ---------------------------------------------------------------------- */
25731 +static match_table_t au_wbr_create_policy = {
25732 + {AuWbrCreate_TDP, "tdp"},
25733 + {AuWbrCreate_TDP, "top-down-parent"},
25734 + {AuWbrCreate_RR, "rr"},
25735 + {AuWbrCreate_RR, "round-robin"},
25736 + {AuWbrCreate_MFS, "mfs"},
25737 + {AuWbrCreate_MFS, "most-free-space"},
25738 + {AuWbrCreate_MFSV, "mfs:%d"},
25739 + {AuWbrCreate_MFSV, "most-free-space:%d"},
25741 + /* top-down regardless the parent, and then mfs */
25742 + {AuWbrCreate_TDMFS, "tdmfs:%d"},
25743 + {AuWbrCreate_TDMFSV, "tdmfs:%d:%d"},
25745 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
25746 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
25747 + {AuWbrCreate_PMFS, "pmfs"},
25748 + {AuWbrCreate_PMFSV, "pmfs:%d"},
25749 + {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
25750 + {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
25755 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
25756 + struct au_opt_wbr_create *create)
25759 + unsigned long long ull;
25762 + if (!match_u64(arg, &ull))
25763 + create->mfsrr_watermark = ull;
25765 + pr_err("bad integer in %s\n", str);
25772 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
25773 + struct au_opt_wbr_create *create)
25778 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
25779 + create->mfs_second = n;
25781 + pr_err("bad integer in %s\n", str);
25788 +static int noinline_for_stack
25789 +au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
25792 + substring_t args[MAX_OPT_ARGS];
25794 + err = match_token(str, au_wbr_create_policy, args);
25795 + create->wbr_create = err;
25797 + case AuWbrCreate_MFSRRV:
25798 + case AuWbrCreate_TDMFSV:
25799 + case AuWbrCreate_PMFSRRV:
25800 + e = au_wbr_mfs_wmark(&args[0], str, create);
25802 + e = au_wbr_mfs_sec(&args[1], str, create);
25806 + case AuWbrCreate_MFSRR:
25807 + case AuWbrCreate_TDMFS:
25808 + case AuWbrCreate_PMFSRR:
25809 + e = au_wbr_mfs_wmark(&args[0], str, create);
25810 + if (unlikely(e)) {
25815 + case AuWbrCreate_MFS:
25816 + case AuWbrCreate_PMFS:
25817 + create->mfs_second = AUFS_MFS_DEF_SEC;
25819 + case AuWbrCreate_MFSV:
25820 + case AuWbrCreate_PMFSV:
25821 + e = au_wbr_mfs_sec(&args[0], str, create);
25830 +const char *au_optstr_wbr_create(int wbr_create)
25832 + return au_parser_pattern(wbr_create, au_wbr_create_policy);
25835 +static match_table_t au_wbr_copyup_policy = {
25836 + {AuWbrCopyup_TDP, "tdp"},
25837 + {AuWbrCopyup_TDP, "top-down-parent"},
25838 + {AuWbrCopyup_BUP, "bup"},
25839 + {AuWbrCopyup_BUP, "bottom-up-parent"},
25840 + {AuWbrCopyup_BU, "bu"},
25841 + {AuWbrCopyup_BU, "bottom-up"},
25845 +static int noinline_for_stack au_wbr_copyup_val(char *str)
25847 + substring_t args[MAX_OPT_ARGS];
25849 + return match_token(str, au_wbr_copyup_policy, args);
25852 +const char *au_optstr_wbr_copyup(int wbr_copyup)
25854 + return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
25857 +/* ---------------------------------------------------------------------- */
25859 +static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
25861 +static void dump_opts(struct au_opts *opts)
25863 +#ifdef CONFIG_AUFS_DEBUG
25864 + /* reduce stack space */
25866 + struct au_opt_add *add;
25867 + struct au_opt_del *del;
25868 + struct au_opt_mod *mod;
25869 + struct au_opt_xino *xino;
25870 + struct au_opt_xino_itrunc *xino_itrunc;
25871 + struct au_opt_wbr_create *create;
25873 + struct au_opt *opt;
25876 + while (opt->type != Opt_tail) {
25877 + switch (opt->type) {
25879 + u.add = &opt->add;
25880 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
25881 + u.add->bindex, u.add->pathname, u.add->perm,
25882 + u.add->path.dentry);
25886 + u.del = &opt->del;
25887 + AuDbg("del {%s, %p}\n",
25888 + u.del->pathname, u.del->h_path.dentry);
25892 + u.mod = &opt->mod;
25893 + AuDbg("mod {%s, 0x%x, %p}\n",
25894 + u.mod->path, u.mod->perm, u.mod->h_root);
25897 + u.add = &opt->add;
25898 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
25899 + u.add->bindex, u.add->pathname, u.add->perm,
25900 + u.add->path.dentry);
25902 + case Opt_prepend:
25903 + u.add = &opt->add;
25904 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
25905 + u.add->bindex, u.add->pathname, u.add->perm,
25906 + u.add->path.dentry);
25909 + AuDbg("dirwh %d\n", opt->dirwh);
25911 + case Opt_rdcache:
25912 + AuDbg("rdcache %d\n", opt->rdcache);
25915 + AuDbg("rdblk %u\n", opt->rdblk);
25917 + case Opt_rdblk_def:
25918 + AuDbg("rdblk_def\n");
25921 + AuDbg("rdhash %u\n", opt->rdhash);
25923 + case Opt_rdhash_def:
25924 + AuDbg("rdhash_def\n");
25927 + u.xino = &opt->xino;
25928 + AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
25930 + case Opt_trunc_xino:
25931 + AuLabel(trunc_xino);
25933 + case Opt_notrunc_xino:
25934 + AuLabel(notrunc_xino);
25936 + case Opt_trunc_xino_path:
25937 + case Opt_itrunc_xino:
25938 + u.xino_itrunc = &opt->xino_itrunc;
25939 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
25944 + case Opt_trunc_xib:
25945 + AuLabel(trunc_xib);
25947 + case Opt_notrunc_xib:
25948 + AuLabel(notrunc_xib);
25956 + case Opt_dirperm1:
25957 + AuLabel(dirperm1);
25959 + case Opt_nodirperm1:
25960 + AuLabel(nodirperm1);
25965 + case Opt_noplink:
25966 + AuLabel(noplink);
25968 + case Opt_list_plink:
25969 + AuLabel(list_plink);
25972 + AuDbg("udba %d, %s\n",
25973 + opt->udba, au_optstr_udba(opt->udba));
25981 + case Opt_diropq_a:
25982 + AuLabel(diropq_a);
25984 + case Opt_diropq_w:
25985 + AuLabel(diropq_w);
25987 + case Opt_warn_perm:
25988 + AuLabel(warn_perm);
25990 + case Opt_nowarn_perm:
25991 + AuLabel(nowarn_perm);
25993 + case Opt_verbose:
25994 + AuLabel(verbose);
25996 + case Opt_noverbose:
25997 + AuLabel(noverbose);
26008 + case Opt_wbr_create:
26009 + u.create = &opt->wbr_create;
26010 + AuDbg("create %d, %s\n", u.create->wbr_create,
26011 + au_optstr_wbr_create(u.create->wbr_create));
26012 + switch (u.create->wbr_create) {
26013 + case AuWbrCreate_MFSV:
26014 + case AuWbrCreate_PMFSV:
26015 + AuDbg("%d sec\n", u.create->mfs_second);
26017 + case AuWbrCreate_MFSRR:
26018 + case AuWbrCreate_TDMFS:
26019 + AuDbg("%llu watermark\n",
26020 + u.create->mfsrr_watermark);
26022 + case AuWbrCreate_MFSRRV:
26023 + case AuWbrCreate_TDMFSV:
26024 + case AuWbrCreate_PMFSRRV:
26025 + AuDbg("%llu watermark, %d sec\n",
26026 + u.create->mfsrr_watermark,
26027 + u.create->mfs_second);
26031 + case Opt_wbr_copyup:
26032 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
26033 + au_optstr_wbr_copyup(opt->wbr_copyup));
26035 + case Opt_fhsm_sec:
26036 + AuDbg("fhsm_sec %u\n", opt->fhsm_second);
26041 + case Opt_nodirren:
26042 + AuLabel(nodirren);
26058 +void au_opts_free(struct au_opts *opts)
26060 + struct au_opt *opt;
26063 + while (opt->type != Opt_tail) {
26064 + switch (opt->type) {
26067 + case Opt_prepend:
26068 + path_put(&opt->add.path);
26072 + path_put(&opt->del.h_path);
26076 + dput(opt->mod.h_root);
26079 + fput(opt->xino.file);
26086 +static int opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
26087 + aufs_bindex_t bindex)
26090 + struct au_opt_add *add = &opt->add;
26093 + add->bindex = bindex;
26094 + add->perm = AuBrPerm_RO;
26095 + add->pathname = opt_str;
26096 + p = strchr(opt_str, '=');
26100 + add->perm = br_perm_val(p);
26103 + err = vfsub_kern_path(add->pathname, lkup_dirflags, &add->path);
26106 + add->perm = AuBrPerm_RO;
26107 + if (au_test_fs_rr(add->path.dentry->d_sb))
26108 + add->perm = AuBrPerm_RR;
26109 + else if (!bindex && !(sb_flags & SB_RDONLY))
26110 + add->perm = AuBrPerm_RW;
26112 + opt->type = Opt_add;
26115 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
26122 +static int au_opts_parse_del(struct au_opt_del *del, substring_t args[])
26126 + del->pathname = args[0].from;
26127 + AuDbg("del path %s\n", del->pathname);
26129 + err = vfsub_kern_path(del->pathname, lkup_dirflags, &del->h_path);
26130 + if (unlikely(err))
26131 + pr_err("lookup failed %s (%d)\n", del->pathname, err);
26136 +#if 0 /* reserved for future use */
26137 +static int au_opts_parse_idel(struct super_block *sb, aufs_bindex_t bindex,
26138 + struct au_opt_del *del, substring_t args[])
26141 + struct dentry *root;
26144 + root = sb->s_root;
26145 + aufs_read_lock(root, AuLock_FLUSH);
26146 + if (bindex < 0 || au_sbbot(sb) < bindex) {
26147 + pr_err("out of bounds, %d\n", bindex);
26152 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
26153 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
26156 + aufs_read_unlock(root, !AuLock_IR);
26161 +static int noinline_for_stack
26162 +au_opts_parse_mod(struct au_opt_mod *mod, substring_t args[])
26165 + struct path path;
26169 + mod->path = args[0].from;
26170 + p = strchr(mod->path, '=');
26171 + if (unlikely(!p)) {
26172 + pr_err("no permssion %s\n", args[0].from);
26177 + err = vfsub_kern_path(mod->path, lkup_dirflags, &path);
26178 + if (unlikely(err)) {
26179 + pr_err("lookup failed %s (%d)\n", mod->path, err);
26183 + mod->perm = br_perm_val(p);
26184 + AuDbg("mod path %s, perm 0x%x, %s\n", mod->path, mod->perm, p);
26185 + mod->h_root = dget(path.dentry);
26192 +#if 0 /* reserved for future use */
26193 +static int au_opts_parse_imod(struct super_block *sb, aufs_bindex_t bindex,
26194 + struct au_opt_mod *mod, substring_t args[])
26197 + struct dentry *root;
26200 + root = sb->s_root;
26201 + aufs_read_lock(root, AuLock_FLUSH);
26202 + if (bindex < 0 || au_sbbot(sb) < bindex) {
26203 + pr_err("out of bounds, %d\n", bindex);
26208 + mod->perm = br_perm_val(args[1].from);
26209 + AuDbg("mod path %s, perm 0x%x, %s\n",
26210 + mod->path, mod->perm, args[1].from);
26211 + mod->h_root = dget(au_h_dptr(root, bindex));
26214 + aufs_read_unlock(root, !AuLock_IR);
26219 +static int au_opts_parse_xino(struct super_block *sb, struct au_opt_xino *xino,
26220 + substring_t args[])
26223 + struct file *file;
26225 + file = au_xino_create(sb, args[0].from, /*silent*/0);
26226 + err = PTR_ERR(file);
26227 + if (IS_ERR(file))
26231 + if (unlikely(file->f_path.dentry->d_sb == sb)) {
26233 + pr_err("%s must be outside\n", args[0].from);
26238 + xino->file = file;
26239 + xino->path = args[0].from;
26245 +static int noinline_for_stack
26246 +au_opts_parse_xino_itrunc_path(struct super_block *sb,
26247 + struct au_opt_xino_itrunc *xino_itrunc,
26248 + substring_t args[])
26251 + aufs_bindex_t bbot, bindex;
26252 + struct path path;
26253 + struct dentry *root;
26255 + err = vfsub_kern_path(args[0].from, lkup_dirflags, &path);
26256 + if (unlikely(err)) {
26257 + pr_err("lookup failed %s (%d)\n", args[0].from, err);
26261 + xino_itrunc->bindex = -1;
26262 + root = sb->s_root;
26263 + aufs_read_lock(root, AuLock_FLUSH);
26264 + bbot = au_sbbot(sb);
26265 + for (bindex = 0; bindex <= bbot; bindex++) {
26266 + if (au_h_dptr(root, bindex) == path.dentry) {
26267 + xino_itrunc->bindex = bindex;
26271 + aufs_read_unlock(root, !AuLock_IR);
26274 + if (unlikely(xino_itrunc->bindex < 0)) {
26275 + pr_err("no such branch %s\n", args[0].from);
26283 +/* called without aufs lock */
26284 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts)
26286 + int err, n, token;
26287 + aufs_bindex_t bindex;
26288 + unsigned char skipped;
26289 + struct dentry *root;
26290 + struct au_opt *opt, *opt_tail;
26292 + /* reduce the stack space */
26294 + struct au_opt_xino_itrunc *xino_itrunc;
26295 + struct au_opt_wbr_create *create;
26298 + substring_t args[MAX_OPT_ARGS];
26302 + a = kmalloc(sizeof(*a), GFP_NOFS);
26303 + if (unlikely(!a))
26306 + root = sb->s_root;
26310 + opt_tail = opt + opts->max_opt - 1;
26311 + opt->type = Opt_tail;
26312 + while (!err && (opt_str = strsep(&str, ",")) && *opt_str) {
26315 + token = match_token(opt_str, options, a->args);
26319 + while (!err && (opt_str = strsep(&a->args[0].from, ":"))
26321 + err = opt_add(opt, opt_str, opts->sb_flags,
26323 + if (unlikely(!err && ++opt > opt_tail)) {
26327 + opt->type = Opt_tail;
26332 + if (unlikely(match_int(&a->args[0], &n))) {
26333 + pr_err("bad integer in %s\n", opt_str);
26337 + err = opt_add(opt, a->args[1].from, opts->sb_flags,
26340 + opt->type = token;
26343 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
26344 + /*dummy bindex*/1);
26346 + opt->type = token;
26348 + case Opt_prepend:
26349 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
26352 + opt->type = token;
26355 + err = au_opts_parse_del(&opt->del, a->args);
26357 + opt->type = token;
26359 +#if 0 /* reserved for future use */
26361 + del->pathname = "(indexed)";
26362 + if (unlikely(match_int(&args[0], &n))) {
26363 + pr_err("bad integer in %s\n", opt_str);
26366 + err = au_opts_parse_idel(sb, n, &opt->del, a->args);
26368 + opt->type = token;
26372 + err = au_opts_parse_mod(&opt->mod, a->args);
26374 + opt->type = token;
26376 +#ifdef IMOD /* reserved for future use */
26378 + u.mod->path = "(indexed)";
26379 + if (unlikely(match_int(&a->args[0], &n))) {
26380 + pr_err("bad integer in %s\n", opt_str);
26383 + err = au_opts_parse_imod(sb, n, &opt->mod, a->args);
26385 + opt->type = token;
26389 + err = au_opts_parse_xino(sb, &opt->xino, a->args);
26391 + opt->type = token;
26394 + case Opt_trunc_xino_path:
26395 + err = au_opts_parse_xino_itrunc_path
26396 + (sb, &opt->xino_itrunc, a->args);
26398 + opt->type = token;
26401 + case Opt_itrunc_xino:
26402 + u.xino_itrunc = &opt->xino_itrunc;
26403 + if (unlikely(match_int(&a->args[0], &n))) {
26404 + pr_err("bad integer in %s\n", opt_str);
26407 + u.xino_itrunc->bindex = n;
26408 + aufs_read_lock(root, AuLock_FLUSH);
26409 + if (n < 0 || au_sbbot(sb) < n) {
26410 + pr_err("out of bounds, %d\n", n);
26411 + aufs_read_unlock(root, !AuLock_IR);
26414 + aufs_read_unlock(root, !AuLock_IR);
26416 + opt->type = token;
26420 + if (unlikely(match_int(&a->args[0], &opt->dirwh)))
26423 + opt->type = token;
26426 + case Opt_rdcache:
26427 + if (unlikely(match_int(&a->args[0], &n))) {
26428 + pr_err("bad integer in %s\n", opt_str);
26431 + if (unlikely(n > AUFS_RDCACHE_MAX)) {
26432 + pr_err("rdcache must be smaller than %d\n",
26433 + AUFS_RDCACHE_MAX);
26436 + opt->rdcache = n;
26438 + opt->type = token;
26441 + if (unlikely(match_int(&a->args[0], &n)
26443 + || n > KMALLOC_MAX_SIZE)) {
26444 + pr_err("bad integer in %s\n", opt_str);
26447 + if (unlikely(n && n < NAME_MAX)) {
26448 + pr_err("rdblk must be larger than %d\n",
26454 + opt->type = token;
26457 + if (unlikely(match_int(&a->args[0], &n)
26459 + || n * sizeof(struct hlist_head)
26460 + > KMALLOC_MAX_SIZE)) {
26461 + pr_err("bad integer in %s\n", opt_str);
26466 + opt->type = token;
26469 + case Opt_trunc_xino:
26470 + case Opt_notrunc_xino:
26472 + case Opt_trunc_xib:
26473 + case Opt_notrunc_xib:
26476 + case Opt_dirperm1:
26477 + case Opt_nodirperm1:
26479 + case Opt_noplink:
26480 + case Opt_list_plink:
26483 + case Opt_diropq_a:
26484 + case Opt_diropq_w:
26485 + case Opt_warn_perm:
26486 + case Opt_nowarn_perm:
26487 + case Opt_verbose:
26488 + case Opt_noverbose:
26492 + case Opt_rdblk_def:
26493 + case Opt_rdhash_def:
26495 + case Opt_nodirren:
26499 + opt->type = token;
26503 + opt->udba = udba_val(a->args[0].from);
26504 + if (opt->udba >= 0) {
26506 + opt->type = token;
26508 + pr_err("wrong value, %s\n", opt_str);
26511 + case Opt_wbr_create:
26512 + u.create = &opt->wbr_create;
26513 + u.create->wbr_create
26514 + = au_wbr_create_val(a->args[0].from, u.create);
26515 + if (u.create->wbr_create >= 0) {
26517 + opt->type = token;
26519 + pr_err("wrong value, %s\n", opt_str);
26521 + case Opt_wbr_copyup:
26522 + opt->wbr_copyup = au_wbr_copyup_val(a->args[0].from);
26523 + if (opt->wbr_copyup >= 0) {
26525 + opt->type = token;
26527 + pr_err("wrong value, %s\n", opt_str);
26530 + case Opt_fhsm_sec:
26531 + if (unlikely(match_int(&a->args[0], &n)
26533 + pr_err("bad integer in %s\n", opt_str);
26536 + if (sysaufs_brs) {
26537 + opt->fhsm_second = n;
26538 + opt->type = token;
26540 + pr_warn("ignored %s\n", opt_str);
26545 + pr_warn("ignored %s\n", opt_str);
26547 + case Opt_ignore_silent:
26552 + pr_err("unknown option %s\n", opt_str);
26556 + if (!err && !skipped) {
26557 + if (unlikely(++opt > opt_tail)) {
26560 + opt->type = Opt_tail;
26563 + opt->type = Opt_tail;
26569 + if (unlikely(err))
26570 + au_opts_free(opts);
26576 +static int au_opt_wbr_create(struct super_block *sb,
26577 + struct au_opt_wbr_create *create)
26580 + struct au_sbinfo *sbinfo;
26582 + SiMustWriteLock(sb);
26584 + err = 1; /* handled */
26585 + sbinfo = au_sbi(sb);
26586 + if (sbinfo->si_wbr_create_ops->fin) {
26587 + err = sbinfo->si_wbr_create_ops->fin(sb);
26592 + sbinfo->si_wbr_create = create->wbr_create;
26593 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
26594 + switch (create->wbr_create) {
26595 + case AuWbrCreate_MFSRRV:
26596 + case AuWbrCreate_MFSRR:
26597 + case AuWbrCreate_TDMFS:
26598 + case AuWbrCreate_TDMFSV:
26599 + case AuWbrCreate_PMFSRR:
26600 + case AuWbrCreate_PMFSRRV:
26601 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
26603 + case AuWbrCreate_MFS:
26604 + case AuWbrCreate_MFSV:
26605 + case AuWbrCreate_PMFS:
26606 + case AuWbrCreate_PMFSV:
26607 + sbinfo->si_wbr_mfs.mfs_expire
26608 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
26612 + if (sbinfo->si_wbr_create_ops->init)
26613 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
26620 + * plus: processed without an error
26621 + * zero: unprocessed
26623 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
26624 + struct au_opts *opts)
26627 + struct au_sbinfo *sbinfo;
26629 + SiMustWriteLock(sb);
26631 + err = 1; /* handled */
26632 + sbinfo = au_sbi(sb);
26633 + switch (opt->type) {
26635 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
26636 + sbinfo->si_mntflags |= opt->udba;
26637 + opts->given_udba |= opt->udba;
26641 + au_opt_set(sbinfo->si_mntflags, PLINK);
26643 + case Opt_noplink:
26644 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
26645 + au_plink_put(sb, /*verbose*/1);
26646 + au_opt_clr(sbinfo->si_mntflags, PLINK);
26648 + case Opt_list_plink:
26649 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
26650 + au_plink_list(sb);
26654 + au_opt_set(sbinfo->si_mntflags, DIO);
26655 + au_fset_opts(opts->flags, REFRESH_DYAOP);
26658 + au_opt_clr(sbinfo->si_mntflags, DIO);
26659 + au_fset_opts(opts->flags, REFRESH_DYAOP);
26662 + case Opt_fhsm_sec:
26663 + au_fhsm_set(sbinfo, opt->fhsm_second);
26666 + case Opt_diropq_a:
26667 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
26669 + case Opt_diropq_w:
26670 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
26673 + case Opt_warn_perm:
26674 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
26676 + case Opt_nowarn_perm:
26677 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
26680 + case Opt_verbose:
26681 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
26683 + case Opt_noverbose:
26684 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
26688 + au_opt_set(sbinfo->si_mntflags, SUM);
26691 + au_opt_clr(sbinfo->si_mntflags, SUM);
26692 + au_opt_set(sbinfo->si_mntflags, SUM_W);
26694 + au_opt_clr(sbinfo->si_mntflags, SUM);
26695 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
26698 + case Opt_wbr_create:
26699 + err = au_opt_wbr_create(sb, &opt->wbr_create);
26701 + case Opt_wbr_copyup:
26702 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
26703 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
26707 + sbinfo->si_dirwh = opt->dirwh;
26710 + case Opt_rdcache:
26711 + sbinfo->si_rdcache
26712 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
26715 + sbinfo->si_rdblk = opt->rdblk;
26717 + case Opt_rdblk_def:
26718 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
26721 + sbinfo->si_rdhash = opt->rdhash;
26723 + case Opt_rdhash_def:
26724 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
26728 + au_opt_set(sbinfo->si_mntflags, SHWH);
26731 + au_opt_clr(sbinfo->si_mntflags, SHWH);
26734 + case Opt_dirperm1:
26735 + au_opt_set(sbinfo->si_mntflags, DIRPERM1);
26737 + case Opt_nodirperm1:
26738 + au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
26741 + case Opt_trunc_xino:
26742 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
26744 + case Opt_notrunc_xino:
26745 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
26748 + case Opt_trunc_xino_path:
26749 + case Opt_itrunc_xino:
26750 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex);
26755 + case Opt_trunc_xib:
26756 + au_fset_opts(opts->flags, TRUNC_XIB);
26758 + case Opt_notrunc_xib:
26759 + au_fclr_opts(opts->flags, TRUNC_XIB);
26764 + if (!au_opt_test(sbinfo->si_mntflags, DIRREN)) {
26765 + err = au_dr_opt_set(sb);
26770 + au_opt_set(sbinfo->si_mntflags, DIRREN);
26772 + case Opt_nodirren:
26774 + if (au_opt_test(sbinfo->si_mntflags, DIRREN)) {
26775 + err = au_dr_opt_clr(sb, au_ftest_opts(opts->flags,
26781 + au_opt_clr(sbinfo->si_mntflags, DIRREN);
26785 + sb->s_flags |= SB_POSIXACL;
26788 + sb->s_flags &= ~SB_POSIXACL;
26800 + * returns tri-state.
26801 + * plus: processed without an error
26802 + * zero: unprocessed
26805 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
26806 + struct au_opts *opts)
26808 + int err, do_refresh;
26811 + switch (opt->type) {
26813 + opt->add.bindex = au_sbbot(sb) + 1;
26814 + if (opt->add.bindex < 0)
26815 + opt->add.bindex = 0;
26817 + case Opt_prepend:
26818 + opt->add.bindex = 0;
26819 + add: /* indented label */
26821 + err = au_br_add(sb, &opt->add,
26822 + au_ftest_opts(opts->flags, REMOUNT));
26825 + au_fset_opts(opts->flags, REFRESH);
26831 + err = au_br_del(sb, &opt->del,
26832 + au_ftest_opts(opts->flags, REMOUNT));
26835 + au_fset_opts(opts->flags, TRUNC_XIB);
26836 + au_fset_opts(opts->flags, REFRESH);
26842 + err = au_br_mod(sb, &opt->mod,
26843 + au_ftest_opts(opts->flags, REMOUNT),
26848 + au_fset_opts(opts->flags, REFRESH);
26855 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
26856 + struct au_opt_xino **opt_xino,
26857 + struct au_opts *opts)
26860 + aufs_bindex_t bbot, bindex;
26861 + struct dentry *root, *parent, *h_root;
26864 + switch (opt->type) {
26866 + err = au_xino_set(sb, &opt->xino,
26867 + !!au_ftest_opts(opts->flags, REMOUNT));
26868 + if (unlikely(err))
26871 + *opt_xino = &opt->xino;
26872 + au_xino_brid_set(sb, -1);
26874 + /* safe d_parent access */
26875 + parent = opt->xino.file->f_path.dentry->d_parent;
26876 + root = sb->s_root;
26877 + bbot = au_sbbot(sb);
26878 + for (bindex = 0; bindex <= bbot; bindex++) {
26879 + h_root = au_h_dptr(root, bindex);
26880 + if (h_root == parent) {
26881 + au_xino_brid_set(sb, au_sbr_id(sb, bindex));
26889 + au_xino_brid_set(sb, -1);
26890 + *opt_xino = (void *)-1;
26897 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
26898 + unsigned int pending)
26901 + aufs_bindex_t bindex, bbot;
26902 + unsigned char do_plink, skip, do_free, can_no_dreval;
26903 + struct au_branch *br;
26904 + struct au_wbr *wbr;
26905 + struct dentry *root, *dentry;
26906 + struct inode *dir, *h_dir;
26907 + struct au_sbinfo *sbinfo;
26908 + struct au_hinode *hdir;
26910 + SiMustAnyLock(sb);
26912 + sbinfo = au_sbi(sb);
26913 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
26915 + if (!(sb_flags & SB_RDONLY)) {
26916 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
26917 + pr_warn("first branch should be rw\n");
26918 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
26919 + pr_warn_once("shwh should be used with ro\n");
26922 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
26923 + && !au_opt_test(sbinfo->si_mntflags, XINO))
26924 + pr_warn_once("udba=*notify requires xino\n");
26926 + if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
26927 + pr_warn_once("dirperm1 breaks the protection"
26928 + " by the permission bits on the lower branch\n");
26932 + root = sb->s_root;
26933 + dir = d_inode(root);
26934 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
26935 + can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
26937 + bbot = au_sbbot(sb);
26938 + for (bindex = 0; !err && bindex <= bbot; bindex++) {
26940 + h_dir = au_h_iptr(dir, bindex);
26941 + br = au_sbr(sb, bindex);
26943 + if ((br->br_perm & AuBrAttr_ICEX)
26944 + && !h_dir->i_op->listxattr)
26945 + br->br_perm &= ~AuBrAttr_ICEX;
26947 + if ((br->br_perm & AuBrAttr_ICEX_SEC)
26948 + && (au_br_sb(br)->s_flags & SB_NOSEC))
26949 + br->br_perm &= ~AuBrAttr_ICEX_SEC;
26953 + wbr = br->br_wbr;
26955 + wbr_wh_read_lock(wbr);
26957 + if (!au_br_writable(br->br_perm)) {
26960 + || (!wbr->wbr_whbase
26961 + && !wbr->wbr_plink
26962 + && !wbr->wbr_orph));
26963 + } else if (!au_br_wh_linkable(br->br_perm)) {
26964 + /* skip = (!br->br_whbase && !br->br_orph); */
26965 + skip = (!wbr || !wbr->wbr_whbase);
26966 + if (skip && wbr) {
26968 + skip = !!wbr->wbr_plink;
26970 + skip = !wbr->wbr_plink;
26973 + /* skip = (br->br_whbase && br->br_ohph); */
26974 + skip = (wbr && wbr->wbr_whbase);
26977 + skip = !!wbr->wbr_plink;
26979 + skip = !wbr->wbr_plink;
26983 + wbr_wh_read_unlock(wbr);
26985 + if (can_no_dreval) {
26986 + dentry = br->br_path.dentry;
26987 + spin_lock(&dentry->d_lock);
26988 + if (dentry->d_flags &
26989 + (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
26990 + can_no_dreval = 0;
26991 + spin_unlock(&dentry->d_lock);
26994 + if (au_br_fhsm(br->br_perm)) {
26996 + AuDebugOn(!br->br_fhsm);
27002 + hdir = au_hi(dir, bindex);
27003 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
27005 + wbr_wh_write_lock(wbr);
27006 + err = au_wh_init(br, sb);
27008 + wbr_wh_write_unlock(wbr);
27009 + au_hn_inode_unlock(hdir);
27011 + if (!err && do_free) {
27013 + br->br_wbr = NULL;
27017 + if (can_no_dreval)
27018 + au_fset_si(sbinfo, NO_DREVAL);
27020 + au_fclr_si(sbinfo, NO_DREVAL);
27023 + au_fset_si(sbinfo, FHSM);
27024 + for (bindex = bbot; bindex >= 0; bindex--) {
27025 + br = au_sbr(sb, bindex);
27026 + if (au_br_fhsm(br->br_perm)) {
27027 + au_fhsm_set_bottom(sb, bindex);
27032 + au_fclr_si(sbinfo, FHSM);
27033 + au_fhsm_set_bottom(sb, -1);
27039 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
27042 + unsigned int tmp;
27043 + aufs_bindex_t bindex, bbot;
27044 + struct au_opt *opt;
27045 + struct au_opt_xino *opt_xino, xino;
27046 + struct au_sbinfo *sbinfo;
27047 + struct au_branch *br;
27048 + struct inode *dir;
27050 + SiMustWriteLock(sb);
27055 + while (err >= 0 && opt->type != Opt_tail)
27056 + err = au_opt_simple(sb, opt++, opts);
27059 + else if (unlikely(err < 0))
27062 + /* disable xino and udba temporary */
27063 + sbinfo = au_sbi(sb);
27064 + tmp = sbinfo->si_mntflags;
27065 + au_opt_clr(sbinfo->si_mntflags, XINO);
27066 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
27069 + while (err >= 0 && opt->type != Opt_tail)
27070 + err = au_opt_br(sb, opt++, opts);
27073 + else if (unlikely(err < 0))
27076 + bbot = au_sbbot(sb);
27077 + if (unlikely(bbot < 0)) {
27079 + pr_err("no branches\n");
27083 + if (au_opt_test(tmp, XINO))
27084 + au_opt_set(sbinfo->si_mntflags, XINO);
27086 + while (!err && opt->type != Opt_tail)
27087 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
27088 + if (unlikely(err))
27091 + err = au_opts_verify(sb, sb->s_flags, tmp);
27092 + if (unlikely(err))
27095 + /* restore xino */
27096 + if (au_opt_test(tmp, XINO) && !opt_xino) {
27097 + xino.file = au_xino_def(sb);
27098 + err = PTR_ERR(xino.file);
27099 + if (IS_ERR(xino.file))
27102 + err = au_xino_set(sb, &xino, /*remount*/0);
27104 + if (unlikely(err))
27108 + /* restore udba */
27109 + tmp &= AuOptMask_UDBA;
27110 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
27111 + sbinfo->si_mntflags |= tmp;
27112 + bbot = au_sbbot(sb);
27113 + for (bindex = 0; bindex <= bbot; bindex++) {
27114 + br = au_sbr(sb, bindex);
27115 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
27116 + if (unlikely(err))
27117 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
27119 + /* go on even if err */
27121 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
27122 + dir = d_inode(sb->s_root);
27123 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
27130 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
27133 + unsigned char no_dreval;
27134 + struct inode *dir;
27135 + struct au_opt_xino *opt_xino;
27136 + struct au_opt *opt;
27137 + struct au_sbinfo *sbinfo;
27139 + SiMustWriteLock(sb);
27141 + err = au_dr_opt_flush(sb);
27142 + if (unlikely(err))
27144 + au_fset_opts(opts->flags, DR_FLUSHED);
27146 + dir = d_inode(sb->s_root);
27147 + sbinfo = au_sbi(sb);
27150 + while (err >= 0 && opt->type != Opt_tail) {
27151 + err = au_opt_simple(sb, opt, opts);
27153 + err = au_opt_br(sb, opt, opts);
27155 + err = au_opt_xino(sb, opt, &opt_xino, opts);
27161 + /* go on even err */
27163 + no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
27164 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
27165 + if (unlikely(rerr && !err))
27168 + if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
27169 + au_fset_opts(opts->flags, REFRESH_IDOP);
27171 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
27172 + rerr = au_xib_trunc(sb);
27173 + if (unlikely(rerr && !err))
27177 + /* will be handled by the caller */
27178 + if (!au_ftest_opts(opts->flags, REFRESH)
27179 + && (opts->given_udba
27180 + || au_opt_test(sbinfo->si_mntflags, XINO)
27181 + || au_ftest_opts(opts->flags, REFRESH_IDOP)
27183 + au_fset_opts(opts->flags, REFRESH);
27185 + AuDbg("status 0x%x\n", opts->flags);
27191 +/* ---------------------------------------------------------------------- */
27193 +unsigned int au_opt_udba(struct super_block *sb)
27195 + return au_mntflags(sb) & AuOptMask_UDBA;
27197 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
27198 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
27199 +++ linux/fs/aufs/opts.h 2018-01-29 07:56:20.059991935 +0100
27202 + * Copyright (C) 2005-2017 Junjiro R. Okajima
27204 + * This program, aufs is free software; you can redistribute it and/or modify
27205 + * it under the terms of the GNU General Public License as published by
27206 + * the Free Software Foundation; either version 2 of the License, or
27207 + * (at your option) any later version.
27209 + * This program is distributed in the hope that it will be useful,
27210 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27211 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27212 + * GNU General Public License for more details.
27214 + * You should have received a copy of the GNU General Public License
27215 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27219 + * mount options/flags
27222 +#ifndef __AUFS_OPTS_H__
27223 +#define __AUFS_OPTS_H__
27227 +#include <linux/path.h>
27231 +/* ---------------------------------------------------------------------- */
27234 +#define AuOpt_XINO 1 /* external inode number bitmap
27235 + and translation table */
27236 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
27237 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
27238 +#define AuOpt_UDBA_REVAL (1 << 3)
27239 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
27240 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
27241 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
27242 +#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
27244 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
27245 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
27246 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
27247 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
27248 +#define AuOpt_VERBOSE (1 << 13) /* busy inode when del-branch */
27249 +#define AuOpt_DIO (1 << 14) /* direct io */
27250 +#define AuOpt_DIRREN (1 << 15) /* directory rename */
27252 +#ifndef CONFIG_AUFS_HNOTIFY
27253 +#undef AuOpt_UDBA_HNOTIFY
27254 +#define AuOpt_UDBA_HNOTIFY 0
27256 +#ifndef CONFIG_AUFS_DIRREN
27257 +#undef AuOpt_DIRREN
27258 +#define AuOpt_DIRREN 0
27260 +#ifndef CONFIG_AUFS_SHWH
27262 +#define AuOpt_SHWH 0
27265 +#define AuOpt_Def (AuOpt_XINO \
27266 + | AuOpt_UDBA_REVAL \
27268 + /* | AuOpt_DIRPERM1 */ \
27269 + | AuOpt_WARN_PERM)
27270 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
27271 + | AuOpt_UDBA_REVAL \
27272 + | AuOpt_UDBA_HNOTIFY)
27274 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
27275 +#define au_opt_set(flags, name) do { \
27276 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
27277 + ((flags) |= AuOpt_##name); \
27279 +#define au_opt_set_udba(flags, name) do { \
27280 + (flags) &= ~AuOptMask_UDBA; \
27281 + ((flags) |= AuOpt_##name); \
27283 +#define au_opt_clr(flags, name) do { \
27284 + ((flags) &= ~AuOpt_##name); \
27287 +static inline unsigned int au_opts_plink(unsigned int mntflags)
27289 +#ifdef CONFIG_PROC_FS
27292 + return mntflags & ~AuOpt_PLINK;
27296 +/* ---------------------------------------------------------------------- */
27298 +/* policies to select one among multiple writable branches */
27300 + AuWbrCreate_TDP, /* top down parent */
27301 + AuWbrCreate_RR, /* round robin */
27302 + AuWbrCreate_MFS, /* most free space */
27303 + AuWbrCreate_MFSV, /* mfs with seconds */
27304 + AuWbrCreate_MFSRR, /* mfs then rr */
27305 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
27306 + AuWbrCreate_TDMFS, /* top down regardless parent and mfs */
27307 + AuWbrCreate_TDMFSV, /* top down regardless parent and mfs */
27308 + AuWbrCreate_PMFS, /* parent and mfs */
27309 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
27310 + AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
27311 + AuWbrCreate_PMFSRRV, /* plus seconds */
27313 + AuWbrCreate_Def = AuWbrCreate_TDP
27317 + AuWbrCopyup_TDP, /* top down parent */
27318 + AuWbrCopyup_BUP, /* bottom up parent */
27319 + AuWbrCopyup_BU, /* bottom up */
27321 + AuWbrCopyup_Def = AuWbrCopyup_TDP
27324 +/* ---------------------------------------------------------------------- */
27326 +struct au_opt_add {
27327 + aufs_bindex_t bindex;
27330 + struct path path;
27333 +struct au_opt_del {
27335 + struct path h_path;
27338 +struct au_opt_mod {
27341 + struct dentry *h_root;
27344 +struct au_opt_xino {
27346 + struct file *file;
27349 +struct au_opt_xino_itrunc {
27350 + aufs_bindex_t bindex;
27353 +struct au_opt_wbr_create {
27356 + unsigned long long mfsrr_watermark;
27362 + struct au_opt_xino xino;
27363 + struct au_opt_xino_itrunc xino_itrunc;
27364 + struct au_opt_add add;
27365 + struct au_opt_del del;
27366 + struct au_opt_mod mod;
27369 + unsigned int rdblk;
27370 + unsigned int rdhash;
27372 + struct au_opt_wbr_create wbr_create;
27374 + unsigned int fhsm_second;
27379 +#define AuOpts_REMOUNT 1
27380 +#define AuOpts_REFRESH (1 << 1)
27381 +#define AuOpts_TRUNC_XIB (1 << 2)
27382 +#define AuOpts_REFRESH_DYAOP (1 << 3)
27383 +#define AuOpts_REFRESH_IDOP (1 << 4)
27384 +#define AuOpts_DR_FLUSHED (1 << 5)
27385 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
27386 +#define au_fset_opts(flags, name) \
27387 + do { (flags) |= AuOpts_##name; } while (0)
27388 +#define au_fclr_opts(flags, name) \
27389 + do { (flags) &= ~AuOpts_##name; } while (0)
27391 +#ifndef CONFIG_AUFS_DIRREN
27392 +#undef AuOpts_DR_FLUSHED
27393 +#define AuOpts_DR_FLUSHED 0
27397 + struct au_opt *opt;
27400 + unsigned int given_udba;
27401 + unsigned int flags;
27402 + unsigned long sb_flags;
27405 +/* ---------------------------------------------------------------------- */
27408 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
27409 +const char *au_optstr_udba(int udba);
27410 +const char *au_optstr_wbr_copyup(int wbr_copyup);
27411 +const char *au_optstr_wbr_create(int wbr_create);
27413 +void au_opts_free(struct au_opts *opts);
27414 +struct super_block;
27415 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts);
27416 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
27417 + unsigned int pending);
27418 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
27419 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
27421 +unsigned int au_opt_udba(struct super_block *sb);
27423 +#endif /* __KERNEL__ */
27424 +#endif /* __AUFS_OPTS_H__ */
27425 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
27426 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
27427 +++ linux/fs/aufs/plink.c 2018-01-29 07:56:20.059991935 +0100
27430 + * Copyright (C) 2005-2017 Junjiro R. Okajima
27432 + * This program, aufs is free software; you can redistribute it and/or modify
27433 + * it under the terms of the GNU General Public License as published by
27434 + * the Free Software Foundation; either version 2 of the License, or
27435 + * (at your option) any later version.
27437 + * This program is distributed in the hope that it will be useful,
27438 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27439 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27440 + * GNU General Public License for more details.
27442 + * You should have received a copy of the GNU General Public License
27443 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27453 + * the pseudo-link maintenance mode.
27454 + * during a user process maintains the pseudo-links,
27455 + * prohibit adding a new plink and branch manipulation.
27459 + * For entry functions which will handle plink, and i_mutex is already held
27461 + * They cannot wait and should return an error at once.
27462 + * Callers has to check the error.
27464 + * For entry functions which will handle plink, but i_mutex is not held
27466 + * They can wait the plink maintenance mode to finish.
27468 + * They behave like F_SETLK and F_SETLKW.
27469 + * If the caller never handle plink, then both flags are unnecessary.
27472 +int au_plink_maint(struct super_block *sb, int flags)
27476 + struct task_struct *parent, *prev;
27477 + struct au_sbinfo *sbi;
27479 + SiMustAnyLock(sb);
27482 + if (!au_opt_test(au_mntflags(sb), PLINK))
27485 + sbi = au_sbi(sb);
27486 + pid = sbi->si_plink_maint_pid;
27487 + if (!pid || pid == current->pid)
27490 + /* todo: it highly depends upon /sbin/mount.aufs */
27492 + parent = current;
27496 + parent = rcu_dereference(parent->real_parent);
27497 + if (parent == prev)
27499 + ppid = task_pid_vnr(parent);
27500 + if (pid == ppid) {
27501 + rcu_read_unlock();
27506 + rcu_read_unlock();
27508 + if (au_ftest_lock(flags, NOPLMW)) {
27509 + /* if there is no i_mutex lock in VFS, we don't need to wait */
27510 + /* AuDebugOn(!lockdep_depth(current)); */
27511 + while (sbi->si_plink_maint_pid) {
27512 + si_read_unlock(sb);
27513 + /* gave up wake_up_bit() */
27514 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
27516 + if (au_ftest_lock(flags, FLUSH))
27517 + au_nwt_flush(&sbi->si_nowait);
27518 + si_noflush_read_lock(sb);
27520 + } else if (au_ftest_lock(flags, NOPLM)) {
27521 + AuDbg("ppid %d, pid %d\n", ppid, pid);
27529 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
27531 + spin_lock(&sbinfo->si_plink_maint_lock);
27532 + sbinfo->si_plink_maint_pid = 0;
27533 + spin_unlock(&sbinfo->si_plink_maint_lock);
27534 + wake_up_all(&sbinfo->si_plink_wq);
27537 +int au_plink_maint_enter(struct super_block *sb)
27540 + struct au_sbinfo *sbinfo;
27543 + sbinfo = au_sbi(sb);
27544 + /* make sure i am the only one in this fs */
27545 + si_write_lock(sb, AuLock_FLUSH);
27546 + if (au_opt_test(au_mntflags(sb), PLINK)) {
27547 + spin_lock(&sbinfo->si_plink_maint_lock);
27548 + if (!sbinfo->si_plink_maint_pid)
27549 + sbinfo->si_plink_maint_pid = current->pid;
27552 + spin_unlock(&sbinfo->si_plink_maint_lock);
27554 + si_write_unlock(sb);
27559 +/* ---------------------------------------------------------------------- */
27561 +#ifdef CONFIG_AUFS_DEBUG
27562 +void au_plink_list(struct super_block *sb)
27565 + struct au_sbinfo *sbinfo;
27566 + struct hlist_bl_head *hbl;
27567 + struct hlist_bl_node *pos;
27568 + struct au_icntnr *icntnr;
27570 + SiMustAnyLock(sb);
27572 + sbinfo = au_sbi(sb);
27573 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
27574 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
27576 + for (i = 0; i < AuPlink_NHASH; i++) {
27577 + hbl = sbinfo->si_plink + i;
27578 + hlist_bl_lock(hbl);
27579 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
27580 + AuDbg("%lu\n", icntnr->vfs_inode.i_ino);
27581 + hlist_bl_unlock(hbl);
27586 +/* is the inode pseudo-linked? */
27587 +int au_plink_test(struct inode *inode)
27590 + struct au_sbinfo *sbinfo;
27591 + struct hlist_bl_head *hbl;
27592 + struct hlist_bl_node *pos;
27593 + struct au_icntnr *icntnr;
27595 + sbinfo = au_sbi(inode->i_sb);
27596 + AuRwMustAnyLock(&sbinfo->si_rwsem);
27597 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
27598 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
27601 + i = au_plink_hash(inode->i_ino);
27602 + hbl = sbinfo->si_plink + i;
27603 + hlist_bl_lock(hbl);
27604 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink)
27605 + if (&icntnr->vfs_inode == inode) {
27609 + hlist_bl_unlock(hbl);
27613 +/* ---------------------------------------------------------------------- */
27616 + * generate a name for plink.
27617 + * the file will be stored under AUFS_WH_PLINKDIR.
27619 +/* 20 is max digits length of ulong 64 */
27620 +#define PLINK_NAME_LEN ((20 + 1) * 2)
27622 +static int plink_name(char *name, int len, struct inode *inode,
27623 + aufs_bindex_t bindex)
27626 + struct inode *h_inode;
27628 + h_inode = au_h_iptr(inode, bindex);
27629 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
27633 +struct au_do_plink_lkup_args {
27634 + struct dentry **errp;
27635 + struct qstr *tgtname;
27636 + struct dentry *h_parent;
27637 + struct au_branch *br;
27640 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
27641 + struct dentry *h_parent,
27642 + struct au_branch *br)
27644 + struct dentry *h_dentry;
27645 + struct inode *h_inode;
27647 + h_inode = d_inode(h_parent);
27648 + vfsub_inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2);
27649 + h_dentry = vfsub_lkup_one(tgtname, h_parent);
27650 + inode_unlock_shared(h_inode);
27654 +static void au_call_do_plink_lkup(void *args)
27656 + struct au_do_plink_lkup_args *a = args;
27657 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br);
27660 +/* lookup the plink-ed @inode under the branch at @bindex */
27661 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
27663 + struct dentry *h_dentry, *h_parent;
27664 + struct au_branch *br;
27666 + char a[PLINK_NAME_LEN];
27667 + struct qstr tgtname = QSTR_INIT(a, 0);
27669 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
27671 + br = au_sbr(inode->i_sb, bindex);
27672 + h_parent = br->br_wbr->wbr_plink;
27673 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
27675 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
27676 + struct au_do_plink_lkup_args args = {
27677 + .errp = &h_dentry,
27678 + .tgtname = &tgtname,
27679 + .h_parent = h_parent,
27683 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
27684 + if (unlikely(wkq_err))
27685 + h_dentry = ERR_PTR(wkq_err);
27687 + h_dentry = au_do_plink_lkup(&tgtname, h_parent, br);
27692 +/* create a pseudo-link */
27693 +static int do_whplink(struct qstr *tgt, struct dentry *h_parent,
27694 + struct dentry *h_dentry, struct au_branch *br)
27697 + struct path h_path = {
27698 + .mnt = au_br_mnt(br)
27700 + struct inode *h_dir, *delegated;
27702 + h_dir = d_inode(h_parent);
27703 + inode_lock_nested(h_dir, AuLsc_I_CHILD2);
27705 + h_path.dentry = vfsub_lkup_one(tgt, h_parent);
27706 + err = PTR_ERR(h_path.dentry);
27707 + if (IS_ERR(h_path.dentry))
27711 + /* wh.plink dir is not monitored */
27712 + /* todo: is it really safe? */
27713 + if (d_is_positive(h_path.dentry)
27714 + && d_inode(h_path.dentry) != d_inode(h_dentry)) {
27715 + delegated = NULL;
27716 + err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
27717 + if (unlikely(err == -EWOULDBLOCK)) {
27718 + pr_warn("cannot retry for NFSv4 delegation"
27719 + " for an internal unlink\n");
27722 + dput(h_path.dentry);
27723 + h_path.dentry = NULL;
27727 + if (!err && d_is_negative(h_path.dentry)) {
27728 + delegated = NULL;
27729 + err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
27730 + if (unlikely(err == -EWOULDBLOCK)) {
27731 + pr_warn("cannot retry for NFSv4 delegation"
27732 + " for an internal link\n");
27736 + dput(h_path.dentry);
27739 + inode_unlock(h_dir);
27743 +struct do_whplink_args {
27745 + struct qstr *tgt;
27746 + struct dentry *h_parent;
27747 + struct dentry *h_dentry;
27748 + struct au_branch *br;
27751 +static void call_do_whplink(void *args)
27753 + struct do_whplink_args *a = args;
27754 + *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br);
27757 +static int whplink(struct dentry *h_dentry, struct inode *inode,
27758 + aufs_bindex_t bindex, struct au_branch *br)
27760 + int err, wkq_err;
27761 + struct au_wbr *wbr;
27762 + struct dentry *h_parent;
27763 + char a[PLINK_NAME_LEN];
27764 + struct qstr tgtname = QSTR_INIT(a, 0);
27766 + wbr = au_sbr(inode->i_sb, bindex)->br_wbr;
27767 + h_parent = wbr->wbr_plink;
27768 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
27770 + /* always superio. */
27771 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
27772 + struct do_whplink_args args = {
27775 + .h_parent = h_parent,
27776 + .h_dentry = h_dentry,
27779 + wkq_err = au_wkq_wait(call_do_whplink, &args);
27780 + if (unlikely(wkq_err))
27783 + err = do_whplink(&tgtname, h_parent, h_dentry, br);
27789 + * create a new pseudo-link for @h_dentry on @bindex.
27790 + * the linked inode is held in aufs @inode.
27792 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
27793 + struct dentry *h_dentry)
27795 + struct super_block *sb;
27796 + struct au_sbinfo *sbinfo;
27797 + struct hlist_bl_head *hbl;
27798 + struct hlist_bl_node *pos;
27799 + struct au_icntnr *icntnr;
27800 + int found, err, cnt, i;
27802 + sb = inode->i_sb;
27803 + sbinfo = au_sbi(sb);
27804 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
27805 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
27807 + found = au_plink_test(inode);
27811 + i = au_plink_hash(inode->i_ino);
27812 + hbl = sbinfo->si_plink + i;
27815 + hlist_bl_lock(hbl);
27816 + hlist_bl_for_each_entry(icntnr, pos, hbl, plink) {
27817 + if (&icntnr->vfs_inode == inode) {
27823 + icntnr = container_of(inode, struct au_icntnr, vfs_inode);
27824 + hlist_bl_add_head(&icntnr->plink, hbl);
27826 + hlist_bl_unlock(hbl);
27828 + cnt = au_hbl_count(hbl);
27829 +#define msg "unexpectedly unblanced or too many pseudo-links"
27830 + if (cnt > AUFS_PLINK_WARN)
27831 + AuWarn1(msg ", %d\n", cnt);
27833 + err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex));
27834 + if (unlikely(err)) {
27835 + pr_warn("err %d, damaged pseudo link.\n", err);
27836 + au_hbl_del(&icntnr->plink, hbl);
27837 + iput(&icntnr->vfs_inode);
27840 + iput(&icntnr->vfs_inode);
27843 +/* free all plinks */
27844 +void au_plink_put(struct super_block *sb, int verbose)
27847 + struct au_sbinfo *sbinfo;
27848 + struct hlist_bl_head *hbl;
27849 + struct hlist_bl_node *pos, *tmp;
27850 + struct au_icntnr *icntnr;
27852 + SiMustWriteLock(sb);
27854 + sbinfo = au_sbi(sb);
27855 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
27856 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
27858 + /* no spin_lock since sbinfo is write-locked */
27860 + for (i = 0; i < AuPlink_NHASH; i++) {
27861 + hbl = sbinfo->si_plink + i;
27862 + if (!warned && verbose && !hlist_bl_empty(hbl)) {
27863 + pr_warn("pseudo-link is not flushed");
27866 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink)
27867 + iput(&icntnr->vfs_inode);
27868 + INIT_HLIST_BL_HEAD(hbl);
27872 +void au_plink_clean(struct super_block *sb, int verbose)
27874 + struct dentry *root;
27876 + root = sb->s_root;
27877 + aufs_write_lock(root);
27878 + if (au_opt_test(au_mntflags(sb), PLINK))
27879 + au_plink_put(sb, verbose);
27880 + aufs_write_unlock(root);
27883 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
27886 + aufs_bindex_t btop, bbot, bindex;
27889 + btop = au_ibtop(inode);
27890 + bbot = au_ibbot(inode);
27892 + for (bindex = btop; bindex <= bbot; bindex++) {
27893 + if (!au_h_iptr(inode, bindex)
27894 + || au_ii_br_id(inode, bindex) != br_id)
27896 + au_set_h_iptr(inode, bindex, NULL, 0);
27901 + for (bindex = btop; bindex <= bbot; bindex++)
27902 + if (au_h_iptr(inode, bindex)) {
27912 +/* free the plinks on a branch specified by @br_id */
27913 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
27915 + struct au_sbinfo *sbinfo;
27916 + struct hlist_bl_head *hbl;
27917 + struct hlist_bl_node *pos, *tmp;
27918 + struct au_icntnr *icntnr;
27919 + struct inode *inode;
27922 + SiMustWriteLock(sb);
27924 + sbinfo = au_sbi(sb);
27925 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
27926 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
27928 + /* no bit_lock since sbinfo is write-locked */
27929 + for (i = 0; i < AuPlink_NHASH; i++) {
27930 + hbl = sbinfo->si_plink + i;
27931 + hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) {
27932 + inode = au_igrab(&icntnr->vfs_inode);
27933 + ii_write_lock_child(inode);
27934 + do_put = au_plink_do_half_refresh(inode, br_id);
27936 + hlist_bl_del(&icntnr->plink);
27939 + ii_write_unlock(inode);
27944 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
27945 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
27946 +++ linux/fs/aufs/poll.c 2018-01-29 07:56:20.059991935 +0100
27949 + * Copyright (C) 2005-2017 Junjiro R. Okajima
27951 + * This program, aufs is free software; you can redistribute it and/or modify
27952 + * it under the terms of the GNU General Public License as published by
27953 + * the Free Software Foundation; either version 2 of the License, or
27954 + * (at your option) any later version.
27956 + * This program is distributed in the hope that it will be useful,
27957 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27958 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27959 + * GNU General Public License for more details.
27961 + * You should have received a copy of the GNU General Public License
27962 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27967 + * There is only one filesystem which implements ->poll operation, currently.
27972 +unsigned int aufs_poll(struct file *file, poll_table *wait)
27974 + unsigned int mask;
27976 + struct file *h_file;
27977 + struct super_block *sb;
27979 + /* We should pretend an error happened. */
27980 + mask = POLLERR /* | POLLIN | POLLOUT */;
27981 + sb = file->f_path.dentry->d_sb;
27982 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
27984 + h_file = au_read_pre(file, /*keep_fi*/0, /*lsc*/0);
27985 + err = PTR_ERR(h_file);
27986 + if (IS_ERR(h_file))
27989 + /* it is not an error if h_file has no operation */
27990 + mask = DEFAULT_POLLMASK;
27991 + if (h_file->f_op->poll)
27992 + mask = h_file->f_op->poll(h_file, wait);
27993 + fput(h_file); /* instead of au_read_post() */
27996 + si_read_unlock(sb);
27997 + AuTraceErr((int)mask);
28000 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
28001 --- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
28002 +++ linux/fs/aufs/posix_acl.c 2018-01-29 07:56:20.059991935 +0100
28005 + * Copyright (C) 2014-2017 Junjiro R. Okajima
28007 + * This program, aufs is free software; you can redistribute it and/or modify
28008 + * it under the terms of the GNU General Public License as published by
28009 + * the Free Software Foundation; either version 2 of the License, or
28010 + * (at your option) any later version.
28012 + * This program is distributed in the hope that it will be useful,
28013 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28014 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28015 + * GNU General Public License for more details.
28017 + * You should have received a copy of the GNU General Public License
28018 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28022 + * posix acl operations
28025 +#include <linux/fs.h>
28028 +struct posix_acl *aufs_get_acl(struct inode *inode, int type)
28030 + struct posix_acl *acl;
28032 + aufs_bindex_t bindex;
28033 + struct inode *h_inode;
28034 + struct super_block *sb;
28037 + sb = inode->i_sb;
28038 + si_read_lock(sb, AuLock_FLUSH);
28039 + ii_read_lock_child(inode);
28040 + if (!(sb->s_flags & SB_POSIXACL))
28043 + bindex = au_ibtop(inode);
28044 + h_inode = au_h_iptr(inode, bindex);
28045 + if (unlikely(!h_inode
28046 + || ((h_inode->i_mode & S_IFMT)
28047 + != (inode->i_mode & S_IFMT)))) {
28048 + err = au_busy_or_stale();
28049 + acl = ERR_PTR(err);
28053 + /* always topmost only */
28054 + acl = get_acl(h_inode, type);
28055 + if (!IS_ERR_OR_NULL(acl))
28056 + set_cached_acl(inode, type, acl);
28059 + ii_read_unlock(inode);
28060 + si_read_unlock(sb);
28062 + AuTraceErrPtr(acl);
28066 +int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
28070 + struct dentry *dentry;
28071 + struct au_sxattr arg = {
28072 + .type = AU_ACL_SET,
28079 + IMustLock(inode);
28081 + if (inode->i_ino == AUFS_ROOT_INO)
28082 + dentry = dget(inode->i_sb->s_root);
28084 + dentry = d_find_alias(inode);
28086 + dentry = d_find_any_alias(inode);
28088 + pr_warn("cannot handle this inode, "
28089 + "please report to aufs-users ML\n");
28095 + ssz = au_sxattr(dentry, inode, &arg);
28100 + set_cached_acl(inode, type, acl);
28106 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
28107 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
28108 +++ linux/fs/aufs/procfs.c 2018-01-29 07:56:20.059991935 +0100
28111 + * Copyright (C) 2010-2017 Junjiro R. Okajima
28113 + * This program, aufs is free software; you can redistribute it and/or modify
28114 + * it under the terms of the GNU General Public License as published by
28115 + * the Free Software Foundation; either version 2 of the License, or
28116 + * (at your option) any later version.
28118 + * This program is distributed in the hope that it will be useful,
28119 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28120 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28121 + * GNU General Public License for more details.
28123 + * You should have received a copy of the GNU General Public License
28124 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28128 + * procfs interfaces
28131 +#include <linux/proc_fs.h>
28134 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
28136 + struct au_sbinfo *sbinfo;
28138 + sbinfo = file->private_data;
28140 + au_plink_maint_leave(sbinfo);
28141 + kobject_put(&sbinfo->si_kobj);
28147 +static void au_procfs_plm_write_clean(struct file *file)
28149 + struct au_sbinfo *sbinfo;
28151 + sbinfo = file->private_data;
28153 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
28156 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
28159 + struct super_block *sb;
28160 + struct au_sbinfo *sbinfo;
28161 + struct hlist_bl_node *pos;
28164 + if (unlikely(file->private_data))
28168 + /* don't use au_sbilist_lock() here */
28169 + hlist_bl_lock(&au_sbilist);
28170 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
28171 + if (id == sysaufs_si_id(sbinfo)) {
28172 + kobject_get(&sbinfo->si_kobj);
28173 + sb = sbinfo->si_sb;
28176 + hlist_bl_unlock(&au_sbilist);
28179 + if (unlikely(!sb))
28182 + err = au_plink_maint_enter(sb);
28184 + /* keep kobject_get() */
28185 + file->private_data = sbinfo;
28187 + kobject_put(&sbinfo->si_kobj);
28193 + * Accept a valid "si=xxxx" only.
28194 + * Once it is accepted successfully, accept "clean" too.
28196 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
28197 + size_t count, loff_t *ppos)
28200 + unsigned long id;
28201 + /* last newline is allowed */
28202 + char buf[3 + sizeof(unsigned long) * 2 + 1];
28205 + if (unlikely(!capable(CAP_SYS_ADMIN)))
28209 + if (unlikely(count > sizeof(buf)))
28212 + err = copy_from_user(buf, ubuf, count);
28213 + if (unlikely(err)) {
28220 + if (!strcmp("clean", buf)) {
28221 + au_procfs_plm_write_clean(file);
28222 + goto out_success;
28223 + } else if (unlikely(strncmp("si=", buf, 3)))
28226 + err = kstrtoul(buf + 3, 16, &id);
28227 + if (unlikely(err))
28230 + err = au_procfs_plm_write_si(file, id);
28231 + if (unlikely(err))
28235 + err = count; /* success */
28240 +static const struct file_operations au_procfs_plm_fop = {
28241 + .write = au_procfs_plm_write,
28242 + .release = au_procfs_plm_release,
28243 + .owner = THIS_MODULE
28246 +/* ---------------------------------------------------------------------- */
28248 +static struct proc_dir_entry *au_procfs_dir;
28250 +void au_procfs_fin(void)
28252 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
28253 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
28256 +int __init au_procfs_init(void)
28259 + struct proc_dir_entry *entry;
28262 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
28263 + if (unlikely(!au_procfs_dir))
28266 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | S_IWUSR,
28267 + au_procfs_dir, &au_procfs_plm_fop);
28268 + if (unlikely(!entry))
28272 + goto out; /* success */
28276 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
28280 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
28281 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
28282 +++ linux/fs/aufs/rdu.c 2018-01-29 07:56:20.059991935 +0100
28285 + * Copyright (C) 2005-2017 Junjiro R. Okajima
28287 + * This program, aufs is free software; you can redistribute it and/or modify
28288 + * it under the terms of the GNU General Public License as published by
28289 + * the Free Software Foundation; either version 2 of the License, or
28290 + * (at your option) any later version.
28292 + * This program is distributed in the hope that it will be useful,
28293 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28294 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28295 + * GNU General Public License for more details.
28297 + * You should have received a copy of the GNU General Public License
28298 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28302 + * readdir in userspace.
28305 +#include <linux/compat.h>
28306 +#include <linux/fs_stack.h>
28307 +#include <linux/security.h>
28310 +/* bits for struct aufs_rdu.flags */
28311 +#define AuRdu_CALLED 1
28312 +#define AuRdu_CONT (1 << 1)
28313 +#define AuRdu_FULL (1 << 2)
28314 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
28315 +#define au_fset_rdu(flags, name) \
28316 + do { (flags) |= AuRdu_##name; } while (0)
28317 +#define au_fclr_rdu(flags, name) \
28318 + do { (flags) &= ~AuRdu_##name; } while (0)
28320 +struct au_rdu_arg {
28321 + struct dir_context ctx;
28322 + struct aufs_rdu *rdu;
28323 + union au_rdu_ent_ul ent;
28324 + unsigned long end;
28326 + struct super_block *sb;
28330 +static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
28331 + loff_t offset, u64 h_ino, unsigned int d_type)
28334 + struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
28335 + struct aufs_rdu *rdu = arg->rdu;
28336 + struct au_rdu_ent ent;
28340 + au_fset_rdu(rdu->cookie.flags, CALLED);
28341 + len = au_rdu_len(nlen);
28342 + if (arg->ent.ul + len < arg->end) {
28344 + ent.bindex = rdu->cookie.bindex;
28345 + ent.type = d_type;
28347 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
28348 + ent.type = DT_UNKNOWN;
28350 + /* unnecessary to support mmap_sem since this is a dir */
28352 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
28354 + if (copy_to_user(arg->ent.e->name, name, nlen))
28356 + /* the terminating NULL */
28357 + if (__put_user(0, arg->ent.e->name + nlen))
28360 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
28361 + arg->ent.ul += len;
28365 + au_fset_rdu(rdu->cookie.flags, FULL);
28367 + rdu->tail = arg->ent;
28371 + /* AuTraceErr(err); */
28375 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
28379 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
28381 + /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
28382 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
28384 + if (unlikely(offset != cookie->h_pos))
28390 + au_fclr_rdu(cookie->flags, CALLED);
28392 + err = vfsub_iterate_dir(h_file, &arg->ctx);
28396 + && au_ftest_rdu(cookie->flags, CALLED)
28397 + && !au_ftest_rdu(cookie->flags, FULL));
28398 + cookie->h_pos = h_file->f_pos;
28405 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
28408 + aufs_bindex_t bbot;
28409 + struct au_rdu_arg arg = {
28411 + .actor = au_rdu_fill
28414 + struct dentry *dentry;
28415 + struct inode *inode;
28416 + struct file *h_file;
28417 + struct au_rdu_cookie *cookie = &rdu->cookie;
28419 + err = !access_ok(VERIFY_WRITE, rdu->ent.e, rdu->sz);
28420 + if (unlikely(err)) {
28426 + rdu->tail = rdu->ent;
28429 + arg.ent = rdu->ent;
28430 + arg.end = arg.ent.ul;
28431 + arg.end += rdu->sz;
28434 + if (unlikely(!file->f_op->iterate && !file->f_op->iterate_shared))
28437 + err = security_file_permission(file, MAY_READ);
28439 + if (unlikely(err))
28442 + dentry = file->f_path.dentry;
28443 + inode = d_inode(dentry);
28444 + inode_lock_shared(inode);
28446 + arg.sb = inode->i_sb;
28447 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
28448 + if (unlikely(err))
28450 + err = au_alive_dir(dentry);
28451 + if (unlikely(err))
28453 + /* todo: reval? */
28454 + fi_read_lock(file);
28457 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
28458 + && cookie->generation != au_figen(file)))
28463 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
28465 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
28467 + bbot = au_fbtop(file);
28468 + if (cookie->bindex < bbot)
28469 + cookie->bindex = bbot;
28470 + bbot = au_fbbot_dir(file);
28471 + /* AuDbg("b%d, b%d\n", cookie->bindex, bbot); */
28472 + for (; !err && cookie->bindex <= bbot;
28473 + cookie->bindex++, cookie->h_pos = 0) {
28474 + h_file = au_hf_dir(file, cookie->bindex);
28478 + au_fclr_rdu(cookie->flags, FULL);
28479 + err = au_rdu_do(h_file, &arg);
28481 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
28484 + AuDbg("rent %llu\n", rdu->rent);
28486 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
28487 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
28488 + au_fset_rdu(cookie->flags, CONT);
28489 + cookie->generation = au_figen(file);
28492 + ii_read_lock_child(inode);
28493 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibtop(inode)));
28494 + ii_read_unlock(inode);
28497 + fi_read_unlock(file);
28499 + si_read_unlock(arg.sb);
28501 + inode_unlock_shared(inode);
28507 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
28511 + unsigned long long nent;
28512 + union au_rdu_ent_ul *u;
28513 + struct au_rdu_ent ent;
28514 + struct super_block *sb;
28517 + nent = rdu->nent;
28519 + sb = file->f_path.dentry->d_sb;
28520 + si_read_lock(sb, AuLock_FLUSH);
28521 + while (nent-- > 0) {
28522 + /* unnecessary to support mmap_sem since this is a dir */
28523 + err = copy_from_user(&ent, u->e, sizeof(ent));
28525 + err = !access_ok(VERIFY_WRITE, &u->e->ino, sizeof(ino));
28526 + if (unlikely(err)) {
28532 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
28534 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
28536 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
28538 + if (unlikely(err)) {
28543 + err = __put_user(ino, &u->e->ino);
28544 + if (unlikely(err)) {
28549 + u->ul += au_rdu_len(ent.nlen);
28551 + si_read_unlock(sb);
28556 +/* ---------------------------------------------------------------------- */
28558 +static int au_rdu_verify(struct aufs_rdu *rdu)
28560 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
28561 + "%llu, b%d, 0x%x, g%u}\n",
28562 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
28564 + rdu->rent, rdu->shwh, rdu->full,
28565 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
28566 + rdu->cookie.generation);
28568 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
28572 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
28576 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
28579 + struct aufs_rdu rdu;
28580 + void __user *p = (void __user *)arg;
28582 + err = copy_from_user(&rdu, p, sizeof(rdu));
28583 + if (unlikely(err)) {
28588 + err = au_rdu_verify(&rdu);
28589 + if (unlikely(err))
28593 + case AUFS_CTL_RDU:
28594 + err = au_rdu(file, &rdu);
28595 + if (unlikely(err))
28598 + e = copy_to_user(p, &rdu, sizeof(rdu));
28599 + if (unlikely(e)) {
28604 + case AUFS_CTL_RDU_INO:
28605 + err = au_rdu_ino(file, &rdu);
28609 + /* err = -ENOTTY; */
28618 +#ifdef CONFIG_COMPAT
28619 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
28622 + struct aufs_rdu rdu;
28623 + void __user *p = compat_ptr(arg);
28625 + /* todo: get_user()? */
28626 + err = copy_from_user(&rdu, p, sizeof(rdu));
28627 + if (unlikely(err)) {
28632 + rdu.ent.e = compat_ptr(rdu.ent.ul);
28633 + err = au_rdu_verify(&rdu);
28634 + if (unlikely(err))
28638 + case AUFS_CTL_RDU:
28639 + err = au_rdu(file, &rdu);
28640 + if (unlikely(err))
28643 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
28644 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
28645 + e = copy_to_user(p, &rdu, sizeof(rdu));
28646 + if (unlikely(e)) {
28651 + case AUFS_CTL_RDU_INO:
28652 + err = au_rdu_ino(file, &rdu);
28656 + /* err = -ENOTTY; */
28665 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
28666 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
28667 +++ linux/fs/aufs/rwsem.h 2018-01-29 07:56:20.059991935 +0100
28670 + * Copyright (C) 2005-2017 Junjiro R. Okajima
28672 + * This program, aufs is free software; you can redistribute it and/or modify
28673 + * it under the terms of the GNU General Public License as published by
28674 + * the Free Software Foundation; either version 2 of the License, or
28675 + * (at your option) any later version.
28677 + * This program is distributed in the hope that it will be useful,
28678 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28679 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28680 + * GNU General Public License for more details.
28682 + * You should have received a copy of the GNU General Public License
28683 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28687 + * simple read-write semaphore wrappers
28690 +#ifndef __AUFS_RWSEM_H__
28691 +#define __AUFS_RWSEM_H__
28695 +#include "debug.h"
28697 +/* in the futre, the name 'au_rwsem' will be totally gone */
28698 +#define au_rwsem rw_semaphore
28700 +/* to debug easier, do not make them inlined functions */
28701 +#define AuRwMustNoWaiters(rw) AuDebugOn(rwsem_is_contended(rw))
28702 +/* rwsem_is_locked() is unusable */
28703 +#define AuRwMustReadLock(rw) AuDebugOn(!lockdep_recursing(current) \
28705 + && !lockdep_is_held_type(rw, 1))
28706 +#define AuRwMustWriteLock(rw) AuDebugOn(!lockdep_recursing(current) \
28708 + && !lockdep_is_held_type(rw, 0))
28709 +#define AuRwMustAnyLock(rw) AuDebugOn(!lockdep_recursing(current) \
28711 + && !lockdep_is_held(rw))
28712 +#define AuRwDestroy(rw) AuDebugOn(!lockdep_recursing(current) \
28714 + && lockdep_is_held(rw))
28716 +#define au_rw_init(rw) init_rwsem(rw)
28718 +#define au_rw_init_wlock(rw) do { \
28719 + au_rw_init(rw); \
28720 + down_write(rw); \
28723 +#define au_rw_init_wlock_nested(rw, lsc) do { \
28724 + au_rw_init(rw); \
28725 + down_write_nested(rw, lsc); \
28728 +#define au_rw_read_lock(rw) down_read(rw)
28729 +#define au_rw_read_lock_nested(rw, lsc) down_read_nested(rw, lsc)
28730 +#define au_rw_read_unlock(rw) up_read(rw)
28731 +#define au_rw_dgrade_lock(rw) downgrade_write(rw)
28732 +#define au_rw_write_lock(rw) down_write(rw)
28733 +#define au_rw_write_lock_nested(rw, lsc) down_write_nested(rw, lsc)
28734 +#define au_rw_write_unlock(rw) up_write(rw)
28735 +/* why is not _nested version defined? */
28736 +#define au_rw_read_trylock(rw) down_read_trylock(rw)
28737 +#define au_rw_write_trylock(rw) down_write_trylock(rw)
28739 +#endif /* __KERNEL__ */
28740 +#endif /* __AUFS_RWSEM_H__ */
28741 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
28742 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
28743 +++ linux/fs/aufs/sbinfo.c 2018-01-29 07:56:20.059991935 +0100
28746 + * Copyright (C) 2005-2017 Junjiro R. Okajima
28748 + * This program, aufs is free software; you can redistribute it and/or modify
28749 + * it under the terms of the GNU General Public License as published by
28750 + * the Free Software Foundation; either version 2 of the License, or
28751 + * (at your option) any later version.
28753 + * This program is distributed in the hope that it will be useful,
28754 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28755 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28756 + * GNU General Public License for more details.
28758 + * You should have received a copy of the GNU General Public License
28759 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28763 + * superblock private data
28769 + * they are necessary regardless sysfs is disabled.
28771 +void au_si_free(struct kobject *kobj)
28774 + struct au_sbinfo *sbinfo;
28775 + char *locked __maybe_unused; /* debug only */
28777 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
28778 + for (i = 0; i < AuPlink_NHASH; i++)
28779 + AuDebugOn(!hlist_bl_empty(sbinfo->si_plink + i));
28780 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
28782 + AuDebugOn(percpu_counter_sum(&sbinfo->si_ninodes));
28783 + percpu_counter_destroy(&sbinfo->si_ninodes);
28784 + AuDebugOn(percpu_counter_sum(&sbinfo->si_nfiles));
28785 + percpu_counter_destroy(&sbinfo->si_nfiles);
28787 + au_rw_write_lock(&sbinfo->si_rwsem);
28788 + au_br_free(sbinfo);
28789 + au_rw_write_unlock(&sbinfo->si_rwsem);
28791 + kfree(sbinfo->si_branch);
28792 + mutex_destroy(&sbinfo->si_xib_mtx);
28793 + AuRwDestroy(&sbinfo->si_rwsem);
28798 +int au_si_alloc(struct super_block *sb)
28801 + struct au_sbinfo *sbinfo;
28804 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
28805 + if (unlikely(!sbinfo))
28808 + /* will be reallocated separately */
28809 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
28810 + if (unlikely(!sbinfo->si_branch))
28813 + err = sysaufs_si_init(sbinfo);
28814 + if (unlikely(err))
28817 + au_nwt_init(&sbinfo->si_nowait);
28818 + au_rw_init_wlock(&sbinfo->si_rwsem);
28820 + percpu_counter_init(&sbinfo->si_ninodes, 0, GFP_NOFS);
28821 + percpu_counter_init(&sbinfo->si_nfiles, 0, GFP_NOFS);
28823 + sbinfo->si_bbot = -1;
28824 + sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
28826 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
28827 + sbinfo->si_wbr_create = AuWbrCreate_Def;
28828 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
28829 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
28831 + au_fhsm_init(sbinfo);
28833 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
28835 + sbinfo->si_xino_jiffy = jiffies;
28836 + sbinfo->si_xino_expire
28837 + = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
28838 + mutex_init(&sbinfo->si_xib_mtx);
28839 + sbinfo->si_xino_brid = -1;
28840 + /* leave si_xib_last_pindex and si_xib_next_bit */
28842 + INIT_HLIST_BL_HEAD(&sbinfo->si_aopen);
28844 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
28845 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
28846 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
28847 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
28849 + for (i = 0; i < AuPlink_NHASH; i++)
28850 + INIT_HLIST_BL_HEAD(sbinfo->si_plink + i);
28851 + init_waitqueue_head(&sbinfo->si_plink_wq);
28852 + spin_lock_init(&sbinfo->si_plink_maint_lock);
28854 + INIT_HLIST_BL_HEAD(&sbinfo->si_files);
28856 + /* with getattr by default */
28857 + sbinfo->si_iop_array = aufs_iop;
28859 + /* leave other members for sysaufs and si_mnt. */
28860 + sbinfo->si_sb = sb;
28861 + sb->s_fs_info = sbinfo;
28863 + return 0; /* success */
28866 + kfree(sbinfo->si_branch);
28873 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink)
28876 + struct au_branch **brp;
28878 + AuRwMustWriteLock(&sbinfo->si_rwsem);
28881 + sz = sizeof(*brp) * (sbinfo->si_bbot + 1);
28882 + if (unlikely(!sz))
28883 + sz = sizeof(*brp);
28884 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS,
28887 + sbinfo->si_branch = brp;
28894 +/* ---------------------------------------------------------------------- */
28896 +unsigned int au_sigen_inc(struct super_block *sb)
28898 + unsigned int gen;
28899 + struct inode *inode;
28901 + SiMustWriteLock(sb);
28903 + gen = ++au_sbi(sb)->si_generation;
28904 + au_update_digen(sb->s_root);
28905 + inode = d_inode(sb->s_root);
28906 + au_update_iigen(inode, /*half*/0);
28907 + inode->i_version++;
28911 +aufs_bindex_t au_new_br_id(struct super_block *sb)
28913 + aufs_bindex_t br_id;
28915 + struct au_sbinfo *sbinfo;
28917 + SiMustWriteLock(sb);
28919 + sbinfo = au_sbi(sb);
28920 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
28921 + br_id = ++sbinfo->si_last_br_id;
28922 + AuDebugOn(br_id < 0);
28923 + if (br_id && au_br_index(sb, br_id) < 0)
28930 +/* ---------------------------------------------------------------------- */
28932 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
28933 +int si_read_lock(struct super_block *sb, int flags)
28938 + if (au_ftest_lock(flags, FLUSH))
28939 + au_nwt_flush(&au_sbi(sb)->si_nowait);
28941 + si_noflush_read_lock(sb);
28942 + err = au_plink_maint(sb, flags);
28943 + if (unlikely(err))
28944 + si_read_unlock(sb);
28949 +int si_write_lock(struct super_block *sb, int flags)
28953 + if (au_ftest_lock(flags, FLUSH))
28954 + au_nwt_flush(&au_sbi(sb)->si_nowait);
28956 + si_noflush_write_lock(sb);
28957 + err = au_plink_maint(sb, flags);
28958 + if (unlikely(err))
28959 + si_write_unlock(sb);
28964 +/* dentry and super_block lock. call at entry point */
28965 +int aufs_read_lock(struct dentry *dentry, int flags)
28968 + struct super_block *sb;
28970 + sb = dentry->d_sb;
28971 + err = si_read_lock(sb, flags);
28972 + if (unlikely(err))
28975 + if (au_ftest_lock(flags, DW))
28976 + di_write_lock_child(dentry);
28978 + di_read_lock_child(dentry, flags);
28980 + if (au_ftest_lock(flags, GEN)) {
28981 + err = au_digen_test(dentry, au_sigen(sb));
28982 + if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
28983 + AuDebugOn(!err && au_dbrange_test(dentry));
28985 + err = au_dbrange_test(dentry);
28986 + if (unlikely(err))
28987 + aufs_read_unlock(dentry, flags);
28994 +void aufs_read_unlock(struct dentry *dentry, int flags)
28996 + if (au_ftest_lock(flags, DW))
28997 + di_write_unlock(dentry);
28999 + di_read_unlock(dentry, flags);
29000 + si_read_unlock(dentry->d_sb);
29003 +void aufs_write_lock(struct dentry *dentry)
29005 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
29006 + di_write_lock_child(dentry);
29009 +void aufs_write_unlock(struct dentry *dentry)
29011 + di_write_unlock(dentry);
29012 + si_write_unlock(dentry->d_sb);
29015 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
29018 + unsigned int sigen;
29019 + struct super_block *sb;
29022 + err = si_read_lock(sb, flags);
29023 + if (unlikely(err))
29026 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
29028 + if (au_ftest_lock(flags, GEN)) {
29029 + sigen = au_sigen(sb);
29030 + err = au_digen_test(d1, sigen);
29031 + AuDebugOn(!err && au_dbrange_test(d1));
29033 + err = au_digen_test(d2, sigen);
29034 + AuDebugOn(!err && au_dbrange_test(d2));
29036 + if (unlikely(err))
29037 + aufs_read_and_write_unlock2(d1, d2);
29044 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
29046 + di_write_unlock2(d1, d2);
29047 + si_read_unlock(d1->d_sb);
29049 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
29050 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
29051 +++ linux/fs/aufs/super.c 2018-01-29 07:56:20.059991935 +0100
29054 + * Copyright (C) 2005-2017 Junjiro R. Okajima
29056 + * This program, aufs is free software; you can redistribute it and/or modify
29057 + * it under the terms of the GNU General Public License as published by
29058 + * the Free Software Foundation; either version 2 of the License, or
29059 + * (at your option) any later version.
29061 + * This program is distributed in the hope that it will be useful,
29062 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29063 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29064 + * GNU General Public License for more details.
29066 + * You should have received a copy of the GNU General Public License
29067 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29071 + * mount and super_block operations
29074 +#include <linux/mm.h>
29075 +#include <linux/seq_file.h>
29076 +#include <linux/statfs.h>
29077 +#include <linux/vmalloc.h>
29081 + * super_operations
29083 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
29085 + struct au_icntnr *c;
29087 + c = au_cache_alloc_icntnr();
29089 + au_icntnr_init(c);
29090 + c->vfs_inode.i_version = 1; /* sigen(sb); */
29091 + c->iinfo.ii_hinode = NULL;
29092 + return &c->vfs_inode;
29097 +static void aufs_destroy_inode_cb(struct rcu_head *head)
29099 + struct inode *inode = container_of(head, struct inode, i_rcu);
29101 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
29104 +static void aufs_destroy_inode(struct inode *inode)
29106 + if (!au_is_bad_inode(inode))
29107 + au_iinfo_fin(inode);
29108 + call_rcu(&inode->i_rcu, aufs_destroy_inode_cb);
29111 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
29113 + struct inode *inode;
29116 + inode = iget_locked(sb, ino);
29117 + if (unlikely(!inode)) {
29118 + inode = ERR_PTR(-ENOMEM);
29121 + if (!(inode->i_state & I_NEW))
29124 + err = au_xigen_new(inode);
29126 + err = au_iinfo_init(inode);
29128 + inode->i_version++;
29130 + iget_failed(inode);
29131 + inode = ERR_PTR(err);
29135 + /* never return NULL */
29136 + AuDebugOn(!inode);
29137 + AuTraceErrPtr(inode);
29141 +/* lock free root dinfo */
29142 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
29145 + aufs_bindex_t bindex, bbot;
29146 + struct path path;
29147 + struct au_hdentry *hdp;
29148 + struct au_branch *br;
29149 + au_br_perm_str_t perm;
29152 + bbot = au_sbbot(sb);
29154 + hdp = au_hdentry(au_di(sb->s_root), bindex);
29155 + for (; !err && bindex <= bbot; bindex++, hdp++) {
29156 + br = au_sbr(sb, bindex);
29157 + path.mnt = au_br_mnt(br);
29158 + path.dentry = hdp->hd_dentry;
29159 + err = au_seq_path(seq, &path);
29161 + au_optstr_br_perm(&perm, br->br_perm);
29162 + seq_printf(seq, "=%s", perm.a);
29163 + if (bindex != bbot)
29164 + seq_putc(seq, ':');
29167 + if (unlikely(err || seq_has_overflowed(seq)))
29173 +static void au_gen_fmt(char *fmt, int len __maybe_unused, const char *pat,
29174 + const char *append)
29179 + while (*pat != ':')
29182 + strcpy(p, append);
29183 + AuDebugOn(strlen(fmt) >= len);
29186 +static void au_show_wbr_create(struct seq_file *m, int v,
29187 + struct au_sbinfo *sbinfo)
29191 + struct au_wbr_mfs *mfs;
29193 + AuRwMustAnyLock(&sbinfo->si_rwsem);
29195 + seq_puts(m, ",create=");
29196 + pat = au_optstr_wbr_create(v);
29197 + mfs = &sbinfo->si_wbr_mfs;
29199 + case AuWbrCreate_TDP:
29200 + case AuWbrCreate_RR:
29201 + case AuWbrCreate_MFS:
29202 + case AuWbrCreate_PMFS:
29203 + seq_puts(m, pat);
29205 + case AuWbrCreate_MFSRR:
29206 + case AuWbrCreate_TDMFS:
29207 + case AuWbrCreate_PMFSRR:
29208 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu");
29209 + seq_printf(m, fmt, mfs->mfsrr_watermark);
29211 + case AuWbrCreate_MFSV:
29212 + case AuWbrCreate_PMFSV:
29213 + au_gen_fmt(fmt, sizeof(fmt), pat, "%lu");
29214 + seq_printf(m, fmt,
29215 + jiffies_to_msecs(mfs->mfs_expire)
29218 + case AuWbrCreate_MFSRRV:
29219 + case AuWbrCreate_TDMFSV:
29220 + case AuWbrCreate_PMFSRRV:
29221 + au_gen_fmt(fmt, sizeof(fmt), pat, "%llu:%lu");
29222 + seq_printf(m, fmt, mfs->mfsrr_watermark,
29223 + jiffies_to_msecs(mfs->mfs_expire) / MSEC_PER_SEC);
29230 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
29232 +#ifdef CONFIG_SYSFS
29236 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
29237 + aufs_bindex_t bindex, brid;
29238 + struct qstr *name;
29240 + struct dentry *d, *h_root;
29242 + AuRwMustAnyLock(&sbinfo->si_rwsem);
29245 + f = au_sbi(sb)->si_xib;
29249 + /* stop printing the default xino path on the first writable branch */
29251 + brid = au_xino_brid(sb);
29253 + bindex = au_br_index(sb, brid);
29254 + h_root = au_hdentry(au_di(sb->s_root), bindex)->hd_dentry;
29256 + d = f->f_path.dentry;
29257 + name = &d->d_name;
29258 + /* safe ->d_parent because the file is unlinked */
29259 + if (d->d_parent == h_root
29260 + && name->len == len
29261 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
29264 + seq_puts(seq, ",xino=");
29265 + err = au_xino_path(seq, f);
29272 +/* seq_file will re-call me in case of too long string */
29273 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
29276 + unsigned int mnt_flags, v;
29277 + struct super_block *sb;
29278 + struct au_sbinfo *sbinfo;
29280 +#define AuBool(name, str) do { \
29281 + v = au_opt_test(mnt_flags, name); \
29282 + if (v != au_opt_test(AuOpt_Def, name)) \
29283 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
29286 +#define AuStr(name, str) do { \
29287 + v = mnt_flags & AuOptMask_##name; \
29288 + if (v != (AuOpt_Def & AuOptMask_##name)) \
29289 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
29292 +#define AuUInt(name, str, val) do { \
29293 + if (val != AUFS_##name##_DEF) \
29294 + seq_printf(m, "," #str "=%u", val); \
29297 + sb = dentry->d_sb;
29298 + if (sb->s_flags & SB_POSIXACL)
29299 + seq_puts(m, ",acl");
29301 + /* lock free root dinfo */
29302 + si_noflush_read_lock(sb);
29303 + sbinfo = au_sbi(sb);
29304 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
29306 + mnt_flags = au_mntflags(sb);
29307 + if (au_opt_test(mnt_flags, XINO)) {
29308 + err = au_show_xino(m, sb);
29309 + if (unlikely(err))
29312 + seq_puts(m, ",noxino");
29314 + AuBool(TRUNC_XINO, trunc_xino);
29315 + AuStr(UDBA, udba);
29316 + AuBool(SHWH, shwh);
29317 + AuBool(PLINK, plink);
29318 + AuBool(DIO, dio);
29319 + AuBool(DIRPERM1, dirperm1);
29321 + v = sbinfo->si_wbr_create;
29322 + if (v != AuWbrCreate_Def)
29323 + au_show_wbr_create(m, v, sbinfo);
29325 + v = sbinfo->si_wbr_copyup;
29326 + if (v != AuWbrCopyup_Def)
29327 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
29329 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
29330 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
29331 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
29333 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
29335 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
29336 + AuUInt(RDCACHE, rdcache, v);
29338 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
29339 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
29341 + au_fhsm_show(m, sbinfo);
29343 + AuBool(DIRREN, dirren);
29344 + AuBool(SUM, sum);
29345 + /* AuBool(SUM_W, wsum); */
29346 + AuBool(WARN_PERM, warn_perm);
29347 + AuBool(VERBOSE, verbose);
29350 + /* be sure to print "br:" last */
29351 + if (!sysaufs_brs) {
29352 + seq_puts(m, ",br:");
29353 + au_show_brs(m, sb);
29355 + si_read_unlock(sb);
29363 +/* ---------------------------------------------------------------------- */
29365 +/* sum mode which returns the summation for statfs(2) */
29367 +static u64 au_add_till_max(u64 a, u64 b)
29375 + return ULLONG_MAX;
29378 +static u64 au_mul_till_max(u64 a, long mul)
29386 + return ULLONG_MAX;
29389 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
29392 + long bsize, factor;
29393 + u64 blocks, bfree, bavail, files, ffree;
29394 + aufs_bindex_t bbot, bindex, i;
29395 + unsigned char shared;
29396 + struct path h_path;
29397 + struct super_block *h_sb;
29400 + bsize = LONG_MAX;
29406 + bbot = au_sbbot(sb);
29407 + for (bindex = 0; bindex <= bbot; bindex++) {
29408 + h_path.mnt = au_sbr_mnt(sb, bindex);
29409 + h_sb = h_path.mnt->mnt_sb;
29411 + for (i = 0; !shared && i < bindex; i++)
29412 + shared = (au_sbr_sb(sb, i) == h_sb);
29416 + /* sb->s_root for NFS is unreliable */
29417 + h_path.dentry = h_path.mnt->mnt_root;
29418 + err = vfs_statfs(&h_path, buf);
29419 + if (unlikely(err))
29422 + if (bsize > buf->f_bsize) {
29424 + * we will reduce bsize, so we have to expand blocks
29425 + * etc. to match them again
29427 + factor = (bsize / buf->f_bsize);
29428 + blocks = au_mul_till_max(blocks, factor);
29429 + bfree = au_mul_till_max(bfree, factor);
29430 + bavail = au_mul_till_max(bavail, factor);
29431 + bsize = buf->f_bsize;
29434 + factor = (buf->f_bsize / bsize);
29435 + blocks = au_add_till_max(blocks,
29436 + au_mul_till_max(buf->f_blocks, factor));
29437 + bfree = au_add_till_max(bfree,
29438 + au_mul_till_max(buf->f_bfree, factor));
29439 + bavail = au_add_till_max(bavail,
29440 + au_mul_till_max(buf->f_bavail, factor));
29441 + files = au_add_till_max(files, buf->f_files);
29442 + ffree = au_add_till_max(ffree, buf->f_ffree);
29445 + buf->f_bsize = bsize;
29446 + buf->f_blocks = blocks;
29447 + buf->f_bfree = bfree;
29448 + buf->f_bavail = bavail;
29449 + buf->f_files = files;
29450 + buf->f_ffree = ffree;
29451 + buf->f_frsize = 0;
29457 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
29460 + struct path h_path;
29461 + struct super_block *sb;
29463 + /* lock free root dinfo */
29464 + sb = dentry->d_sb;
29465 + si_noflush_read_lock(sb);
29466 + if (!au_opt_test(au_mntflags(sb), SUM)) {
29467 + /* sb->s_root for NFS is unreliable */
29468 + h_path.mnt = au_sbr_mnt(sb, 0);
29469 + h_path.dentry = h_path.mnt->mnt_root;
29470 + err = vfs_statfs(&h_path, buf);
29472 + err = au_statfs_sum(sb, buf);
29473 + si_read_unlock(sb);
29476 + buf->f_type = AUFS_SUPER_MAGIC;
29477 + buf->f_namelen = AUFS_MAX_NAMELEN;
29478 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
29480 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
29485 +/* ---------------------------------------------------------------------- */
29487 +static int aufs_sync_fs(struct super_block *sb, int wait)
29490 + aufs_bindex_t bbot, bindex;
29491 + struct au_branch *br;
29492 + struct super_block *h_sb;
29495 + si_noflush_read_lock(sb);
29496 + bbot = au_sbbot(sb);
29497 + for (bindex = 0; bindex <= bbot; bindex++) {
29498 + br = au_sbr(sb, bindex);
29499 + if (!au_br_writable(br->br_perm))
29502 + h_sb = au_sbr_sb(sb, bindex);
29503 + e = vfsub_sync_filesystem(h_sb, wait);
29504 + if (unlikely(e && !err))
29506 + /* go on even if an error happens */
29508 + si_read_unlock(sb);
29513 +/* ---------------------------------------------------------------------- */
29515 +/* final actions when unmounting a file system */
29516 +static void aufs_put_super(struct super_block *sb)
29518 + struct au_sbinfo *sbinfo;
29520 + sbinfo = au_sbi(sb);
29524 + dbgaufs_si_fin(sbinfo);
29525 + kobject_put(&sbinfo->si_kobj);
29528 +/* ---------------------------------------------------------------------- */
29530 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
29531 + struct super_block *sb, void *arg)
29534 + unsigned long long n, sz;
29541 + if (*hint > ULLONG_MAX / sizeof(array)) {
29542 + array = ERR_PTR(-EMFILE);
29543 + pr_err("hint %llu\n", *hint);
29547 + sz = sizeof(array) * *hint;
29548 + array = kzalloc(sz, GFP_NOFS);
29549 + if (unlikely(!array))
29550 + array = vzalloc(sz);
29551 + if (unlikely(!array)) {
29552 + array = ERR_PTR(-ENOMEM);
29556 + n = cb(sb, array, *hint, arg);
29557 + AuDebugOn(n > *hint);
29564 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
29565 + unsigned long long max __maybe_unused,
29568 + unsigned long long n;
29569 + struct inode **p, *inode;
29570 + struct list_head *head;
29575 + spin_lock(&sb->s_inode_list_lock);
29576 + list_for_each_entry(inode, head, i_sb_list) {
29577 + if (!au_is_bad_inode(inode)
29578 + && au_ii(inode)->ii_btop >= 0) {
29579 + spin_lock(&inode->i_lock);
29580 + if (atomic_read(&inode->i_count)) {
29584 + AuDebugOn(n > max);
29586 + spin_unlock(&inode->i_lock);
29589 + spin_unlock(&sb->s_inode_list_lock);
29594 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
29596 + *max = au_ninodes(sb);
29597 + return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
29600 +void au_iarray_free(struct inode **a, unsigned long long max)
29602 + unsigned long long ull;
29604 + for (ull = 0; ull < max; ull++)
29609 +/* ---------------------------------------------------------------------- */
29612 + * refresh dentry and inode at remount time.
29614 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
29615 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
29616 + struct dentry *parent)
29620 + di_write_lock_child(dentry);
29621 + di_read_lock_parent(parent, AuLock_IR);
29622 + err = au_refresh_dentry(dentry, parent);
29623 + if (!err && dir_flags)
29624 + au_hn_reset(d_inode(dentry), dir_flags);
29625 + di_read_unlock(parent, AuLock_IR);
29626 + di_write_unlock(dentry);
29631 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
29632 + struct au_sbinfo *sbinfo,
29633 + const unsigned int dir_flags, unsigned int do_idop)
29636 + struct dentry *parent;
29639 + parent = dget_parent(dentry);
29640 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
29641 + if (d_really_is_positive(dentry)) {
29642 + if (!d_is_dir(dentry))
29643 + err = au_do_refresh(dentry, /*dir_flags*/0,
29646 + err = au_do_refresh(dentry, dir_flags, parent);
29647 + if (unlikely(err))
29648 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
29651 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
29652 + AuDbgDentry(dentry);
29658 + au_refresh_dop(dentry, /*force_reval*/0);
29660 + au_refresh_dop(dentry, /*force_reval*/1);
29666 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
29668 + int err, i, j, ndentry, e;
29669 + unsigned int sigen;
29670 + struct au_dcsub_pages dpages;
29671 + struct au_dpage *dpage;
29672 + struct dentry **dentries, *d;
29673 + struct au_sbinfo *sbinfo;
29674 + struct dentry *root = sb->s_root;
29675 + const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
29678 + au_refresh_dop(root, /*force_reval*/0);
29680 + err = au_dpages_init(&dpages, GFP_NOFS);
29681 + if (unlikely(err))
29683 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
29684 + if (unlikely(err))
29687 + sigen = au_sigen(sb);
29688 + sbinfo = au_sbi(sb);
29689 + for (i = 0; i < dpages.ndpage; i++) {
29690 + dpage = dpages.dpages + i;
29691 + dentries = dpage->dentries;
29692 + ndentry = dpage->ndentry;
29693 + for (j = 0; j < ndentry; j++) {
29695 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
29697 + if (unlikely(e && !err))
29699 + /* go on even err */
29704 + au_dpages_free(&dpages);
29709 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
29712 + unsigned int sigen;
29713 + unsigned long long max, ull;
29714 + struct inode *inode, **array;
29716 + array = au_iarray_alloc(sb, &max);
29717 + err = PTR_ERR(array);
29718 + if (IS_ERR(array))
29722 + sigen = au_sigen(sb);
29723 + for (ull = 0; ull < max; ull++) {
29724 + inode = array[ull];
29725 + if (unlikely(!inode))
29729 + ii_write_lock_child(inode);
29730 + if (au_iigen(inode, NULL) != sigen) {
29731 + e = au_refresh_hinode_self(inode);
29732 + if (unlikely(e)) {
29733 + au_refresh_iop(inode, /*force_getattr*/1);
29734 + pr_err("error %d, i%lu\n", e, inode->i_ino);
29737 + /* go on even if err */
29740 + if (!e && do_idop)
29741 + au_refresh_iop(inode, /*force_getattr*/0);
29742 + ii_write_unlock(inode);
29745 + au_iarray_free(array, max);
29751 +static void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
29754 + unsigned int udba;
29755 + aufs_bindex_t bindex, bbot;
29756 + struct dentry *root;
29757 + struct inode *inode;
29758 + struct au_branch *br;
29759 + struct au_sbinfo *sbi;
29761 + au_sigen_inc(sb);
29762 + sbi = au_sbi(sb);
29763 + au_fclr_si(sbi, FAILED_REFRESH_DIR);
29765 + root = sb->s_root;
29766 + DiMustNoWaiters(root);
29767 + inode = d_inode(root);
29768 + IiMustNoWaiters(inode);
29770 + udba = au_opt_udba(sb);
29771 + bbot = au_sbbot(sb);
29772 + for (bindex = 0; bindex <= bbot; bindex++) {
29773 + br = au_sbr(sb, bindex);
29774 + err = au_hnotify_reset_br(udba, br, br->br_perm);
29775 + if (unlikely(err))
29776 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
29778 + /* go on even if err */
29780 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
29783 + if (au_ftest_si(sbi, NO_DREVAL)) {
29784 + AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
29785 + sb->s_d_op = &aufs_dop_noreval;
29786 + AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
29787 + sbi->si_iop_array = aufs_iop_nogetattr;
29789 + AuDebugOn(sb->s_d_op == &aufs_dop);
29790 + sb->s_d_op = &aufs_dop;
29791 + AuDebugOn(sbi->si_iop_array == aufs_iop);
29792 + sbi->si_iop_array = aufs_iop;
29794 + pr_info("reset to %pf and %pf\n",
29795 + sb->s_d_op, sbi->si_iop_array);
29798 + di_write_unlock(root);
29799 + err = au_refresh_d(sb, do_idop);
29800 + e = au_refresh_i(sb, do_idop);
29801 + if (unlikely(e && !err))
29803 + /* aufs_write_lock() calls ..._child() */
29804 + di_write_lock_child(root);
29806 + au_cpup_attr_all(inode, /*force*/1);
29808 + if (unlikely(err))
29809 + AuIOErr("refresh failed, ignored, %d\n", err);
29812 +/* stop extra interpretation of errno in mount(8), and strange error messages */
29813 +static int cvt_err(int err)
29827 +static int aufs_remount_fs(struct super_block *sb, int *flags, char *data)
29830 + unsigned int mntflags;
29831 + struct au_opts opts = {
29834 + struct dentry *root;
29835 + struct inode *inode;
29836 + struct au_sbinfo *sbinfo;
29839 + root = sb->s_root;
29840 + if (!data || !*data) {
29841 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
29843 + di_write_lock_child(root);
29844 + err = au_opts_verify(sb, *flags, /*pending*/0);
29845 + aufs_write_unlock(root);
29851 + opts.opt = (void *)__get_free_page(GFP_NOFS);
29852 + if (unlikely(!opts.opt))
29854 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
29855 + opts.flags = AuOpts_REMOUNT;
29856 + opts.sb_flags = *flags;
29858 + /* parse it before aufs lock */
29859 + err = au_opts_parse(sb, data, &opts);
29860 + if (unlikely(err))
29863 + sbinfo = au_sbi(sb);
29864 + inode = d_inode(root);
29865 + inode_lock(inode);
29866 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
29867 + if (unlikely(err))
29869 + di_write_lock_child(root);
29871 + /* au_opts_remount() may return an error */
29872 + err = au_opts_remount(sb, &opts);
29873 + au_opts_free(&opts);
29875 + if (au_ftest_opts(opts.flags, REFRESH))
29876 + au_remount_refresh(sb, au_ftest_opts(opts.flags, REFRESH_IDOP));
29878 + if (au_ftest_opts(opts.flags, REFRESH_DYAOP)) {
29879 + mntflags = au_mntflags(sb);
29880 + do_dx = !!au_opt_test(mntflags, DIO);
29881 + au_dy_arefresh(do_dx);
29884 + au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
29885 + aufs_write_unlock(root);
29888 + inode_unlock(inode);
29890 + free_page((unsigned long)opts.opt);
29892 + err = cvt_err(err);
29897 +static const struct super_operations aufs_sop = {
29898 + .alloc_inode = aufs_alloc_inode,
29899 + .destroy_inode = aufs_destroy_inode,
29900 + /* always deleting, no clearing */
29901 + .drop_inode = generic_delete_inode,
29902 + .show_options = aufs_show_options,
29903 + .statfs = aufs_statfs,
29904 + .put_super = aufs_put_super,
29905 + .sync_fs = aufs_sync_fs,
29906 + .remount_fs = aufs_remount_fs
29909 +/* ---------------------------------------------------------------------- */
29911 +static int alloc_root(struct super_block *sb)
29914 + struct inode *inode;
29915 + struct dentry *root;
29918 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
29919 + err = PTR_ERR(inode);
29920 + if (IS_ERR(inode))
29923 + inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
29924 + inode->i_fop = &aufs_dir_fop;
29925 + inode->i_mode = S_IFDIR;
29926 + set_nlink(inode, 2);
29927 + unlock_new_inode(inode);
29929 + root = d_make_root(inode);
29930 + if (unlikely(!root))
29932 + err = PTR_ERR(root);
29933 + if (IS_ERR(root))
29936 + err = au_di_init(root);
29938 + sb->s_root = root;
29939 + return 0; /* success */
29947 +static int aufs_fill_super(struct super_block *sb, void *raw_data,
29948 + int silent __maybe_unused)
29951 + struct au_opts opts = {
29954 + struct au_sbinfo *sbinfo;
29955 + struct dentry *root;
29956 + struct inode *inode;
29957 + char *arg = raw_data;
29959 + if (unlikely(!arg || !*arg)) {
29961 + pr_err("no arg\n");
29966 + opts.opt = (void *)__get_free_page(GFP_NOFS);
29967 + if (unlikely(!opts.opt))
29969 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
29970 + opts.sb_flags = sb->s_flags;
29972 + err = au_si_alloc(sb);
29973 + if (unlikely(err))
29975 + sbinfo = au_sbi(sb);
29977 + /* all timestamps always follow the ones on the branch */
29978 + sb->s_flags |= SB_NOATIME | SB_NODIRATIME;
29979 + sb->s_op = &aufs_sop;
29980 + sb->s_d_op = &aufs_dop;
29981 + sb->s_magic = AUFS_SUPER_MAGIC;
29982 + sb->s_maxbytes = 0;
29983 + sb->s_stack_depth = 1;
29984 + au_export_init(sb);
29985 + au_xattr_init(sb);
29987 + err = alloc_root(sb);
29988 + if (unlikely(err)) {
29989 + si_write_unlock(sb);
29992 + root = sb->s_root;
29993 + inode = d_inode(root);
29996 + * actually we can parse options regardless aufs lock here.
29997 + * but at remount time, parsing must be done before aufs lock.
29998 + * so we follow the same rule.
30000 + ii_write_lock_parent(inode);
30001 + aufs_write_unlock(root);
30002 + err = au_opts_parse(sb, arg, &opts);
30003 + if (unlikely(err))
30006 + /* lock vfs_inode first, then aufs. */
30007 + inode_lock(inode);
30008 + aufs_write_lock(root);
30009 + err = au_opts_mount(sb, &opts);
30010 + au_opts_free(&opts);
30011 + if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
30012 + sb->s_d_op = &aufs_dop_noreval;
30013 + pr_info("%pf\n", sb->s_d_op);
30014 + au_refresh_dop(root, /*force_reval*/0);
30015 + sbinfo->si_iop_array = aufs_iop_nogetattr;
30016 + au_refresh_iop(inode, /*force_getattr*/0);
30018 + aufs_write_unlock(root);
30019 + inode_unlock(inode);
30021 + goto out_opts; /* success */
30025 + sb->s_root = NULL;
30027 + dbgaufs_si_fin(sbinfo);
30028 + kobject_put(&sbinfo->si_kobj);
30029 + sb->s_fs_info = NULL;
30031 + free_page((unsigned long)opts.opt);
30034 + err = cvt_err(err);
30039 +/* ---------------------------------------------------------------------- */
30041 +static struct dentry *aufs_mount(struct file_system_type *fs_type, int flags,
30042 + const char *dev_name __maybe_unused,
30045 + struct dentry *root;
30046 + struct super_block *sb;
30048 + /* all timestamps always follow the ones on the branch */
30049 + /* mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; */
30050 + root = mount_nodev(fs_type, flags, raw_data, aufs_fill_super);
30051 + if (IS_ERR(root))
30055 + si_write_lock(sb, !AuLock_FLUSH);
30056 + sysaufs_brs_add(sb, 0);
30057 + si_write_unlock(sb);
30058 + au_sbilist_add(sb);
30064 +static void aufs_kill_sb(struct super_block *sb)
30066 + struct au_sbinfo *sbinfo;
30068 + sbinfo = au_sbi(sb);
30070 + au_sbilist_del(sb);
30071 + aufs_write_lock(sb->s_root);
30073 + if (sbinfo->si_wbr_create_ops->fin)
30074 + sbinfo->si_wbr_create_ops->fin(sb);
30075 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
30076 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
30077 + au_remount_refresh(sb, /*do_idop*/0);
30079 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
30080 + au_plink_put(sb, /*verbose*/1);
30082 + au_dr_opt_flush(sb);
30083 + sbinfo->si_sb = NULL;
30084 + aufs_write_unlock(sb->s_root);
30085 + au_nwt_flush(&sbinfo->si_nowait);
30087 + kill_anon_super(sb);
30090 +struct file_system_type aufs_fs_type = {
30091 + .name = AUFS_FSTYPE,
30092 + /* a race between rename and others */
30093 + .fs_flags = FS_RENAME_DOES_D_MOVE,
30094 + .mount = aufs_mount,
30095 + .kill_sb = aufs_kill_sb,
30096 + /* no need to __module_get() and module_put(). */
30097 + .owner = THIS_MODULE,
30099 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
30100 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
30101 +++ linux/fs/aufs/super.h 2018-01-29 07:56:20.059991935 +0100
30104 + * Copyright (C) 2005-2017 Junjiro R. Okajima
30106 + * This program, aufs is free software; you can redistribute it and/or modify
30107 + * it under the terms of the GNU General Public License as published by
30108 + * the Free Software Foundation; either version 2 of the License, or
30109 + * (at your option) any later version.
30111 + * This program is distributed in the hope that it will be useful,
30112 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30113 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30114 + * GNU General Public License for more details.
30116 + * You should have received a copy of the GNU General Public License
30117 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30121 + * super_block operations
30124 +#ifndef __AUFS_SUPER_H__
30125 +#define __AUFS_SUPER_H__
30129 +#include <linux/fs.h>
30130 +#include <linux/kobject.h>
30132 +#include "rwsem.h"
30135 +/* policies to select one among multiple writable branches */
30136 +struct au_wbr_copyup_operations {
30137 + int (*copyup)(struct dentry *dentry);
30140 +#define AuWbr_DIR 1 /* target is a dir */
30141 +#define AuWbr_PARENT (1 << 1) /* always require a parent */
30143 +#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
30144 +#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
30145 +#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
30147 +struct au_wbr_create_operations {
30148 + int (*create)(struct dentry *dentry, unsigned int flags);
30149 + int (*init)(struct super_block *sb);
30150 + int (*fin)(struct super_block *sb);
30153 +struct au_wbr_mfs {
30154 + struct mutex mfs_lock; /* protect this structure */
30155 + unsigned long mfs_jiffy;
30156 + unsigned long mfs_expire;
30157 + aufs_bindex_t mfs_bindex;
30159 + unsigned long long mfsrr_bytes;
30160 + unsigned long long mfsrr_watermark;
30163 +#define AuPlink_NHASH 100
30164 +static inline int au_plink_hash(ino_t ino)
30166 + return ino % AuPlink_NHASH;
30169 +/* File-based Hierarchical Storage Management */
30171 +#ifdef CONFIG_AUFS_FHSM
30172 + /* allow only one process who can receive the notification */
30173 + spinlock_t fhsm_spin;
30175 + wait_queue_head_t fhsm_wqh;
30176 + atomic_t fhsm_readable;
30178 + /* these are protected by si_rwsem */
30179 + unsigned long fhsm_expire;
30180 + aufs_bindex_t fhsm_bottom;
30185 +struct au_sbinfo {
30186 + /* nowait tasks in the system-wide workqueue */
30187 + struct au_nowait_tasks si_nowait;
30190 + * tried sb->s_umount, but failed due to the dependecy between i_mutex.
30191 + * rwsem for au_sbinfo is necessary.
30193 + struct au_rwsem si_rwsem;
30196 + * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
30199 + struct percpu_counter si_ninodes, si_nfiles;
30201 + /* branch management */
30202 + unsigned int si_generation;
30204 + /* see AuSi_ flags */
30205 + unsigned char au_si_status;
30207 + aufs_bindex_t si_bbot;
30209 + /* dirty trick to keep br_id plus */
30210 + unsigned int si_last_br_id :
30211 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
30212 + struct au_branch **si_branch;
30214 + /* policy to select a writable branch */
30215 + unsigned char si_wbr_copyup;
30216 + unsigned char si_wbr_create;
30217 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
30218 + struct au_wbr_create_operations *si_wbr_create_ops;
30220 + /* round robin */
30221 + atomic_t si_wbr_rr_next;
30223 + /* most free space */
30224 + struct au_wbr_mfs si_wbr_mfs;
30226 + /* File-based Hierarchical Storage Management */
30227 + struct au_fhsm si_fhsm;
30229 + /* mount flags */
30230 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
30231 + unsigned int si_mntflags;
30233 + /* external inode number (bitmap and translation table) */
30234 + vfs_readf_t si_xread;
30235 + vfs_writef_t si_xwrite;
30236 + struct file *si_xib;
30237 + struct mutex si_xib_mtx; /* protect xib members */
30238 + unsigned long *si_xib_buf;
30239 + unsigned long si_xib_last_pindex;
30240 + int si_xib_next_bit;
30241 + aufs_bindex_t si_xino_brid;
30242 + unsigned long si_xino_jiffy;
30243 + unsigned long si_xino_expire;
30244 + /* reserved for future use */
30245 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
30247 +#ifdef CONFIG_AUFS_EXPORT
30248 + /* i_generation */
30249 + struct file *si_xigen;
30250 + atomic_t si_xigen_next;
30253 + /* dirty trick to suppoer atomic_open */
30254 + struct hlist_bl_head si_aopen;
30256 + /* vdir parameters */
30257 + unsigned long si_rdcache; /* max cache time in jiffies */
30258 + unsigned int si_rdblk; /* deblk size */
30259 + unsigned int si_rdhash; /* hash size */
30262 + * If the number of whiteouts are larger than si_dirwh, leave all of
30263 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
30264 + * future fsck.aufs or kernel thread will remove them later.
30265 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
30267 + unsigned int si_dirwh;
30269 + /* pseudo_link list */
30270 + struct hlist_bl_head si_plink[AuPlink_NHASH];
30271 + wait_queue_head_t si_plink_wq;
30272 + spinlock_t si_plink_maint_lock;
30273 + pid_t si_plink_maint_pid;
30276 + struct hlist_bl_head si_files;
30278 + /* with/without getattr, brother of sb->s_d_op */
30279 + struct inode_operations *si_iop_array;
30282 + * sysfs and lifetime management.
30283 + * this is not a small structure and it may be a waste of memory in case
30284 + * of sysfs is disabled, particulary when many aufs-es are mounted.
30285 + * but using sysfs is majority.
30287 + struct kobject si_kobj;
30288 +#ifdef CONFIG_DEBUG_FS
30289 + struct dentry *si_dbgaufs;
30290 + struct dentry *si_dbgaufs_plink;
30291 + struct dentry *si_dbgaufs_xib;
30292 +#ifdef CONFIG_AUFS_EXPORT
30293 + struct dentry *si_dbgaufs_xigen;
30297 +#ifdef CONFIG_AUFS_SBILIST
30298 + struct hlist_bl_node si_list;
30301 + /* dirty, necessary for unmounting, sysfs and sysrq */
30302 + struct super_block *si_sb;
30305 +/* sbinfo status flags */
30307 + * set true when refresh_dirs() failed at remount time.
30308 + * then try refreshing dirs at access time again.
30309 + * if it is false, refreshing dirs at access time is unnecesary
30311 +#define AuSi_FAILED_REFRESH_DIR 1
30312 +#define AuSi_FHSM (1 << 1) /* fhsm is active now */
30313 +#define AuSi_NO_DREVAL (1 << 2) /* disable all d_revalidate */
30315 +#ifndef CONFIG_AUFS_FHSM
30317 +#define AuSi_FHSM 0
30320 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
30321 + unsigned int flag)
30323 + AuRwMustAnyLock(&sbi->si_rwsem);
30324 + return sbi->au_si_status & flag;
30326 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
30327 +#define au_fset_si(sbinfo, name) do { \
30328 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30329 + (sbinfo)->au_si_status |= AuSi_##name; \
30331 +#define au_fclr_si(sbinfo, name) do { \
30332 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
30333 + (sbinfo)->au_si_status &= ~AuSi_##name; \
30336 +/* ---------------------------------------------------------------------- */
30338 +/* policy to select one among writable branches */
30339 +#define AuWbrCopyup(sbinfo, ...) \
30340 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
30341 +#define AuWbrCreate(sbinfo, ...) \
30342 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
30344 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
30345 +#define AuLock_DW 1 /* write-lock dentry */
30346 +#define AuLock_IR (1 << 1) /* read-lock inode */
30347 +#define AuLock_IW (1 << 2) /* write-lock inode */
30348 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
30349 +#define AuLock_DIRS (1 << 4) /* target is a pair of dirs */
30350 + /* except RENAME_EXCHANGE */
30351 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
30352 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
30353 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
30354 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
30355 +#define au_fset_lock(flags, name) \
30356 + do { (flags) |= AuLock_##name; } while (0)
30357 +#define au_fclr_lock(flags, name) \
30358 + do { (flags) &= ~AuLock_##name; } while (0)
30360 +/* ---------------------------------------------------------------------- */
30363 +extern struct file_system_type aufs_fs_type;
30364 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
30365 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
30366 + unsigned long long max, void *arg);
30367 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
30368 + struct super_block *sb, void *arg);
30369 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
30370 +void au_iarray_free(struct inode **a, unsigned long long max);
30373 +void au_si_free(struct kobject *kobj);
30374 +int au_si_alloc(struct super_block *sb);
30375 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr, int may_shrink);
30377 +unsigned int au_sigen_inc(struct super_block *sb);
30378 +aufs_bindex_t au_new_br_id(struct super_block *sb);
30380 +int si_read_lock(struct super_block *sb, int flags);
30381 +int si_write_lock(struct super_block *sb, int flags);
30382 +int aufs_read_lock(struct dentry *dentry, int flags);
30383 +void aufs_read_unlock(struct dentry *dentry, int flags);
30384 +void aufs_write_lock(struct dentry *dentry);
30385 +void aufs_write_unlock(struct dentry *dentry);
30386 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
30387 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
30389 +/* wbr_policy.c */
30390 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
30391 +extern struct au_wbr_create_operations au_wbr_create_ops[];
30392 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
30393 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
30394 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop);
30397 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
30399 +#ifdef CONFIG_AUFS_FHSM
30402 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
30406 + spin_lock(&fhsm->fhsm_spin);
30407 + pid = fhsm->fhsm_pid;
30408 + spin_unlock(&fhsm->fhsm_spin);
30413 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
30414 +void au_fhsm_wrote_all(struct super_block *sb, int force);
30415 +int au_fhsm_fd(struct super_block *sb, int oflags);
30416 +int au_fhsm_br_alloc(struct au_branch *br);
30417 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
30418 +void au_fhsm_fin(struct super_block *sb);
30419 +void au_fhsm_init(struct au_sbinfo *sbinfo);
30420 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
30421 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
30423 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
30425 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
30426 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
30427 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
30428 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
30429 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
30430 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
30431 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
30432 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
30433 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
30436 +/* ---------------------------------------------------------------------- */
30438 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
30440 + return sb->s_fs_info;
30443 +/* ---------------------------------------------------------------------- */
30445 +#ifdef CONFIG_AUFS_EXPORT
30446 +int au_test_nfsd(void);
30447 +void au_export_init(struct super_block *sb);
30448 +void au_xigen_inc(struct inode *inode);
30449 +int au_xigen_new(struct inode *inode);
30450 +int au_xigen_set(struct super_block *sb, struct file *base);
30451 +void au_xigen_clr(struct super_block *sb);
30453 +static inline int au_busy_or_stale(void)
30455 + if (!au_test_nfsd())
30460 +AuStubInt0(au_test_nfsd, void)
30461 +AuStubVoid(au_export_init, struct super_block *sb)
30462 +AuStubVoid(au_xigen_inc, struct inode *inode)
30463 +AuStubInt0(au_xigen_new, struct inode *inode)
30464 +AuStubInt0(au_xigen_set, struct super_block *sb, struct file *base)
30465 +AuStubVoid(au_xigen_clr, struct super_block *sb)
30466 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
30467 +#endif /* CONFIG_AUFS_EXPORT */
30469 +/* ---------------------------------------------------------------------- */
30471 +#ifdef CONFIG_AUFS_SBILIST
30473 +extern struct hlist_bl_head au_sbilist;
30475 +static inline void au_sbilist_init(void)
30477 + INIT_HLIST_BL_HEAD(&au_sbilist);
30480 +static inline void au_sbilist_add(struct super_block *sb)
30482 + au_hbl_add(&au_sbi(sb)->si_list, &au_sbilist);
30485 +static inline void au_sbilist_del(struct super_block *sb)
30487 + au_hbl_del(&au_sbi(sb)->si_list, &au_sbilist);
30490 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
30491 +static inline void au_sbilist_lock(void)
30493 + hlist_bl_lock(&au_sbilist);
30496 +static inline void au_sbilist_unlock(void)
30498 + hlist_bl_unlock(&au_sbilist);
30500 +#define AuGFP_SBILIST GFP_ATOMIC
30502 +AuStubVoid(au_sbilist_lock, void)
30503 +AuStubVoid(au_sbilist_unlock, void)
30504 +#define AuGFP_SBILIST GFP_NOFS
30505 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
30507 +AuStubVoid(au_sbilist_init, void)
30508 +AuStubVoid(au_sbilist_add, struct super_block *sb)
30509 +AuStubVoid(au_sbilist_del, struct super_block *sb)
30510 +AuStubVoid(au_sbilist_lock, void)
30511 +AuStubVoid(au_sbilist_unlock, void)
30512 +#define AuGFP_SBILIST GFP_NOFS
30515 +/* ---------------------------------------------------------------------- */
30517 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
30520 + * This function is a dynamic '__init' function actually,
30521 + * so the tiny check for si_rwsem is unnecessary.
30523 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
30524 +#ifdef CONFIG_DEBUG_FS
30525 + sbinfo->si_dbgaufs = NULL;
30526 + sbinfo->si_dbgaufs_plink = NULL;
30527 + sbinfo->si_dbgaufs_xib = NULL;
30528 +#ifdef CONFIG_AUFS_EXPORT
30529 + sbinfo->si_dbgaufs_xigen = NULL;
30534 +/* ---------------------------------------------------------------------- */
30536 +/* current->atomic_flags */
30537 +/* this value should never corrupt the ones defined in linux/sched.h */
30538 +#define PFA_AUFS 7
30540 +TASK_PFA_TEST(AUFS, test_aufs) /* task_test_aufs */
30541 +TASK_PFA_SET(AUFS, aufs) /* task_set_aufs */
30542 +TASK_PFA_CLEAR(AUFS, aufs) /* task_clear_aufs */
30544 +static inline int si_pid_test(struct super_block *sb)
30546 + return !!task_test_aufs(current);
30549 +static inline void si_pid_clr(struct super_block *sb)
30551 + AuDebugOn(!task_test_aufs(current));
30552 + task_clear_aufs(current);
30555 +static inline void si_pid_set(struct super_block *sb)
30557 + AuDebugOn(task_test_aufs(current));
30558 + task_set_aufs(current);
30561 +/* ---------------------------------------------------------------------- */
30563 +/* lock superblock. mainly for entry point functions */
30564 +#define __si_read_lock(sb) au_rw_read_lock(&au_sbi(sb)->si_rwsem)
30565 +#define __si_write_lock(sb) au_rw_write_lock(&au_sbi(sb)->si_rwsem)
30566 +#define __si_read_trylock(sb) au_rw_read_trylock(&au_sbi(sb)->si_rwsem)
30567 +#define __si_write_trylock(sb) au_rw_write_trylock(&au_sbi(sb)->si_rwsem)
30569 +#define __si_read_trylock_nested(sb) \
30570 + au_rw_read_trylock_nested(&au_sbi(sb)->si_rwsem)
30571 +#define __si_write_trylock_nested(sb) \
30572 + au_rw_write_trylock_nested(&au_sbi(sb)->si_rwsem)
30575 +#define __si_read_unlock(sb) au_rw_read_unlock(&au_sbi(sb)->si_rwsem)
30576 +#define __si_write_unlock(sb) au_rw_write_unlock(&au_sbi(sb)->si_rwsem)
30577 +#define __si_downgrade_lock(sb) au_rw_dgrade_lock(&au_sbi(sb)->si_rwsem)
30579 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
30580 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
30581 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
30583 +static inline void si_noflush_read_lock(struct super_block *sb)
30585 + __si_read_lock(sb);
30589 +static inline int si_noflush_read_trylock(struct super_block *sb)
30593 + locked = __si_read_trylock(sb);
30599 +static inline void si_noflush_write_lock(struct super_block *sb)
30601 + __si_write_lock(sb);
30605 +static inline int si_noflush_write_trylock(struct super_block *sb)
30609 + locked = __si_write_trylock(sb);
30615 +#if 0 /* reserved */
30616 +static inline int si_read_trylock(struct super_block *sb, int flags)
30618 + if (au_ftest_lock(flags, FLUSH))
30619 + au_nwt_flush(&au_sbi(sb)->si_nowait);
30620 + return si_noflush_read_trylock(sb);
30624 +static inline void si_read_unlock(struct super_block *sb)
30627 + __si_read_unlock(sb);
30630 +#if 0 /* reserved */
30631 +static inline int si_write_trylock(struct super_block *sb, int flags)
30633 + if (au_ftest_lock(flags, FLUSH))
30634 + au_nwt_flush(&au_sbi(sb)->si_nowait);
30635 + return si_noflush_write_trylock(sb);
30639 +static inline void si_write_unlock(struct super_block *sb)
30642 + __si_write_unlock(sb);
30645 +#if 0 /* reserved */
30646 +static inline void si_downgrade_lock(struct super_block *sb)
30648 + __si_downgrade_lock(sb);
30652 +/* ---------------------------------------------------------------------- */
30654 +static inline aufs_bindex_t au_sbbot(struct super_block *sb)
30656 + SiMustAnyLock(sb);
30657 + return au_sbi(sb)->si_bbot;
30660 +static inline unsigned int au_mntflags(struct super_block *sb)
30662 + SiMustAnyLock(sb);
30663 + return au_sbi(sb)->si_mntflags;
30666 +static inline unsigned int au_sigen(struct super_block *sb)
30668 + SiMustAnyLock(sb);
30669 + return au_sbi(sb)->si_generation;
30672 +static inline unsigned long long au_ninodes(struct super_block *sb)
30674 + s64 n = percpu_counter_sum(&au_sbi(sb)->si_ninodes);
30680 +static inline void au_ninodes_inc(struct super_block *sb)
30682 + percpu_counter_inc(&au_sbi(sb)->si_ninodes);
30685 +static inline void au_ninodes_dec(struct super_block *sb)
30687 + percpu_counter_dec(&au_sbi(sb)->si_ninodes);
30690 +static inline unsigned long long au_nfiles(struct super_block *sb)
30692 + s64 n = percpu_counter_sum(&au_sbi(sb)->si_nfiles);
30698 +static inline void au_nfiles_inc(struct super_block *sb)
30700 + percpu_counter_inc(&au_sbi(sb)->si_nfiles);
30703 +static inline void au_nfiles_dec(struct super_block *sb)
30705 + percpu_counter_dec(&au_sbi(sb)->si_nfiles);
30708 +static inline struct au_branch *au_sbr(struct super_block *sb,
30709 + aufs_bindex_t bindex)
30711 + SiMustAnyLock(sb);
30712 + return au_sbi(sb)->si_branch[0 + bindex];
30715 +static inline void au_xino_brid_set(struct super_block *sb, aufs_bindex_t brid)
30717 + SiMustWriteLock(sb);
30718 + au_sbi(sb)->si_xino_brid = brid;
30721 +static inline aufs_bindex_t au_xino_brid(struct super_block *sb)
30723 + SiMustAnyLock(sb);
30724 + return au_sbi(sb)->si_xino_brid;
30727 +#endif /* __KERNEL__ */
30728 +#endif /* __AUFS_SUPER_H__ */
30729 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
30730 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
30731 +++ linux/fs/aufs/sysaufs.c 2018-01-29 07:56:20.059991935 +0100
30734 + * Copyright (C) 2005-2017 Junjiro R. Okajima
30736 + * This program, aufs is free software; you can redistribute it and/or modify
30737 + * it under the terms of the GNU General Public License as published by
30738 + * the Free Software Foundation; either version 2 of the License, or
30739 + * (at your option) any later version.
30741 + * This program is distributed in the hope that it will be useful,
30742 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30743 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30744 + * GNU General Public License for more details.
30746 + * You should have received a copy of the GNU General Public License
30747 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30751 + * sysfs interface and lifetime management
30752 + * they are necessary regardless sysfs is disabled.
30755 +#include <linux/random.h>
30758 +unsigned long sysaufs_si_mask;
30759 +struct kset *sysaufs_kset;
30761 +#define AuSiAttr(_name) { \
30762 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
30763 + .show = sysaufs_si_##_name, \
30766 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
30767 +struct attribute *sysaufs_si_attrs[] = {
30768 + &sysaufs_si_attr_xi_path.attr,
30772 +static const struct sysfs_ops au_sbi_ops = {
30773 + .show = sysaufs_si_show
30776 +static struct kobj_type au_sbi_ktype = {
30777 + .release = au_si_free,
30778 + .sysfs_ops = &au_sbi_ops,
30779 + .default_attrs = sysaufs_si_attrs
30782 +/* ---------------------------------------------------------------------- */
30784 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
30788 + sbinfo->si_kobj.kset = sysaufs_kset;
30789 + /* cf. sysaufs_name() */
30790 + err = kobject_init_and_add
30791 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
30792 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
30794 + dbgaufs_si_null(sbinfo);
30796 + err = dbgaufs_si_init(sbinfo);
30797 + if (unlikely(err))
30798 + kobject_put(&sbinfo->si_kobj);
30803 +void sysaufs_fin(void)
30806 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
30807 + kset_unregister(sysaufs_kset);
30810 +int __init sysaufs_init(void)
30815 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
30816 + } while (!sysaufs_si_mask);
30819 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
30820 + if (unlikely(!sysaufs_kset))
30822 + err = PTR_ERR(sysaufs_kset);
30823 + if (IS_ERR(sysaufs_kset))
30825 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
30826 + if (unlikely(err)) {
30827 + kset_unregister(sysaufs_kset);
30831 + err = dbgaufs_init();
30832 + if (unlikely(err))
30837 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
30838 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
30839 +++ linux/fs/aufs/sysaufs.h 2018-01-29 07:56:20.059991935 +0100
30842 + * Copyright (C) 2005-2017 Junjiro R. Okajima
30844 + * This program, aufs is free software; you can redistribute it and/or modify
30845 + * it under the terms of the GNU General Public License as published by
30846 + * the Free Software Foundation; either version 2 of the License, or
30847 + * (at your option) any later version.
30849 + * This program is distributed in the hope that it will be useful,
30850 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30851 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30852 + * GNU General Public License for more details.
30854 + * You should have received a copy of the GNU General Public License
30855 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30859 + * sysfs interface and mount lifetime management
30862 +#ifndef __SYSAUFS_H__
30863 +#define __SYSAUFS_H__
30867 +#include <linux/sysfs.h>
30868 +#include "module.h"
30870 +struct super_block;
30873 +struct sysaufs_si_attr {
30874 + struct attribute attr;
30875 + int (*show)(struct seq_file *seq, struct super_block *sb);
30878 +/* ---------------------------------------------------------------------- */
30881 +extern unsigned long sysaufs_si_mask;
30882 +extern struct kset *sysaufs_kset;
30883 +extern struct attribute *sysaufs_si_attrs[];
30884 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
30885 +int __init sysaufs_init(void);
30886 +void sysaufs_fin(void);
30888 +/* ---------------------------------------------------------------------- */
30890 +/* some people doesn't like to show a pointer in kernel */
30891 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
30893 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
30896 +#define SysaufsSiNamePrefix "si_"
30897 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
30898 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
30900 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
30901 + sysaufs_si_id(sbinfo));
30905 +#ifdef CONFIG_SYSFS
30907 +extern struct attribute_group *sysaufs_attr_group;
30909 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
30910 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
30912 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
30913 +#ifdef CONFIG_COMPAT
30914 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
30917 +void sysaufs_br_init(struct au_branch *br);
30918 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
30919 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
30921 +#define sysaufs_brs_init() do {} while (0)
30924 +#define sysaufs_attr_group NULL
30926 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
30927 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
30928 + struct attribute *attr, char *buf)
30929 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
30930 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
30931 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
30933 +static inline void sysaufs_brs_init(void)
30938 +#endif /* CONFIG_SYSFS */
30940 +#endif /* __KERNEL__ */
30941 +#endif /* __SYSAUFS_H__ */
30942 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
30943 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
30944 +++ linux/fs/aufs/sysfs.c 2018-01-29 07:56:20.059991935 +0100
30947 + * Copyright (C) 2005-2017 Junjiro R. Okajima
30949 + * This program, aufs is free software; you can redistribute it and/or modify
30950 + * it under the terms of the GNU General Public License as published by
30951 + * the Free Software Foundation; either version 2 of the License, or
30952 + * (at your option) any later version.
30954 + * This program is distributed in the hope that it will be useful,
30955 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30956 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30957 + * GNU General Public License for more details.
30959 + * You should have received a copy of the GNU General Public License
30960 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30964 + * sysfs interface
30967 +#include <linux/compat.h>
30968 +#include <linux/seq_file.h>
30971 +#ifdef CONFIG_AUFS_FS_MODULE
30972 +/* this entry violates the "one line per file" policy of sysfs */
30973 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
30977 + static char *conf =
30978 +/* this file is generated at compiling */
30979 +#include "conf.str"
30982 + err = snprintf(buf, PAGE_SIZE, conf);
30983 + if (unlikely(err >= PAGE_SIZE))
30988 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
30991 +static struct attribute *au_attr[] = {
30992 +#ifdef CONFIG_AUFS_FS_MODULE
30993 + &au_config_attr.attr,
30995 + NULL, /* need to NULL terminate the list of attributes */
30998 +static struct attribute_group sysaufs_attr_group_body = {
31002 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
31004 +/* ---------------------------------------------------------------------- */
31006 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
31010 + SiMustAnyLock(sb);
31013 + if (au_opt_test(au_mntflags(sb), XINO)) {
31014 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
31015 + seq_putc(seq, '\n');
31021 + * the lifetime of branch is independent from the entry under sysfs.
31022 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
31025 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
31026 + aufs_bindex_t bindex, int idx)
31029 + struct path path;
31030 + struct dentry *root;
31031 + struct au_branch *br;
31032 + au_br_perm_str_t perm;
31034 + AuDbg("b%d\n", bindex);
31037 + root = sb->s_root;
31038 + di_read_lock_parent(root, !AuLock_IR);
31039 + br = au_sbr(sb, bindex);
31042 + case AuBrSysfs_BR:
31043 + path.mnt = au_br_mnt(br);
31044 + path.dentry = au_h_dptr(root, bindex);
31045 + err = au_seq_path(seq, &path);
31047 + au_optstr_br_perm(&perm, br->br_perm);
31048 + seq_printf(seq, "=%s\n", perm.a);
31051 + case AuBrSysfs_BRID:
31052 + seq_printf(seq, "%d\n", br->br_id);
31055 + di_read_unlock(root, !AuLock_IR);
31056 + if (unlikely(err || seq_has_overflowed(seq)))
31062 +/* ---------------------------------------------------------------------- */
31064 +static struct seq_file *au_seq(char *p, ssize_t len)
31066 + struct seq_file *seq;
31068 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
31070 + /* mutex_init(&seq.lock); */
31073 + return seq; /* success */
31076 + seq = ERR_PTR(-ENOMEM);
31080 +#define SysaufsBr_PREFIX "br"
31081 +#define SysaufsBrid_PREFIX "brid"
31083 +/* todo: file size may exceed PAGE_SIZE */
31084 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
31090 + aufs_bindex_t bbot;
31091 + struct au_sbinfo *sbinfo;
31092 + struct super_block *sb;
31093 + struct seq_file *seq;
31095 + struct attribute **cattr;
31097 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
31098 + sb = sbinfo->si_sb;
31101 + * prevent a race condition between sysfs and aufs.
31102 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
31103 + * prohibits maintaining the sysfs entries.
31104 + * hew we acquire read lock after sysfs_get_active_two().
31105 + * on the other hand, the remount process may maintain the sysfs/aufs
31106 + * entries after acquiring write lock.
31107 + * it can cause a deadlock.
31108 + * simply we gave up processing read here.
31111 + if (unlikely(!si_noflush_read_trylock(sb)))
31114 + seq = au_seq(buf, PAGE_SIZE);
31115 + err = PTR_ERR(seq);
31119 + name = (void *)attr->name;
31120 + cattr = sysaufs_si_attrs;
31122 + if (!strcmp(name, (*cattr)->name)) {
31123 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
31130 + if (!strncmp(name, SysaufsBrid_PREFIX,
31131 + sizeof(SysaufsBrid_PREFIX) - 1)) {
31132 + idx = AuBrSysfs_BRID;
31133 + name += sizeof(SysaufsBrid_PREFIX) - 1;
31134 + } else if (!strncmp(name, SysaufsBr_PREFIX,
31135 + sizeof(SysaufsBr_PREFIX) - 1)) {
31136 + idx = AuBrSysfs_BR;
31137 + name += sizeof(SysaufsBr_PREFIX) - 1;
31141 + err = kstrtol(name, 10, &l);
31143 + bbot = au_sbbot(sb);
31145 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
31152 + err = seq->count;
31153 + /* sysfs limit */
31154 + if (unlikely(err == PAGE_SIZE))
31159 + si_read_unlock(sb);
31164 +/* ---------------------------------------------------------------------- */
31166 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
31170 + aufs_bindex_t bindex, bbot;
31173 + struct seq_file *seq;
31174 + struct au_branch *br;
31176 + si_read_lock(sb, AuLock_FLUSH);
31177 + bbot = au_sbbot(sb);
31183 + buf = (void *)__get_free_page(GFP_NOFS);
31184 + if (unlikely(!buf))
31187 + seq = au_seq(buf, PAGE_SIZE);
31188 + err = PTR_ERR(seq);
31192 + sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
31193 + for (bindex = 0; bindex <= bbot; bindex++, arg++) {
31194 + err = !access_ok(VERIFY_WRITE, arg, sizeof(*arg));
31195 + if (unlikely(err))
31198 + br = au_sbr(sb, bindex);
31199 + brid = br->br_id;
31200 + BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
31201 + err = __put_user(brid, &arg->id);
31202 + if (unlikely(err))
31205 + BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
31206 + err = __put_user(br->br_perm, &arg->perm);
31207 + if (unlikely(err))
31210 + err = au_seq_path(seq, &br->br_path);
31211 + if (unlikely(err))
31213 + seq_putc(seq, '\0');
31214 + if (!seq_has_overflowed(seq)) {
31215 + err = copy_to_user(arg->path, seq->buf, seq->count);
31217 + if (unlikely(err))
31224 + if (unlikely(err))
31230 + free_page((unsigned long)buf);
31232 + si_read_unlock(sb);
31236 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
31238 + return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
31241 +#ifdef CONFIG_COMPAT
31242 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
31244 + return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
31248 +/* ---------------------------------------------------------------------- */
31250 +void sysaufs_br_init(struct au_branch *br)
31253 + struct au_brsysfs *br_sysfs;
31254 + struct attribute *attr;
31256 + br_sysfs = br->br_sysfs;
31257 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31258 + attr = &br_sysfs->attr;
31259 + sysfs_attr_init(attr);
31260 + attr->name = br_sysfs->name;
31261 + attr->mode = S_IRUGO;
31266 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
31268 + struct au_branch *br;
31269 + struct kobject *kobj;
31270 + struct au_brsysfs *br_sysfs;
31272 + aufs_bindex_t bbot;
31274 + dbgaufs_brs_del(sb, bindex);
31276 + if (!sysaufs_brs)
31279 + kobj = &au_sbi(sb)->si_kobj;
31280 + bbot = au_sbbot(sb);
31281 + for (; bindex <= bbot; bindex++) {
31282 + br = au_sbr(sb, bindex);
31283 + br_sysfs = br->br_sysfs;
31284 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31285 + sysfs_remove_file(kobj, &br_sysfs->attr);
31291 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
31294 + aufs_bindex_t bbot;
31295 + struct kobject *kobj;
31296 + struct au_branch *br;
31297 + struct au_brsysfs *br_sysfs;
31299 + dbgaufs_brs_add(sb, bindex);
31301 + if (!sysaufs_brs)
31304 + kobj = &au_sbi(sb)->si_kobj;
31305 + bbot = au_sbbot(sb);
31306 + for (; bindex <= bbot; bindex++) {
31307 + br = au_sbr(sb, bindex);
31308 + br_sysfs = br->br_sysfs;
31309 + snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
31310 + SysaufsBr_PREFIX "%d", bindex);
31311 + snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
31312 + SysaufsBrid_PREFIX "%d", bindex);
31313 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
31314 + err = sysfs_create_file(kobj, &br_sysfs->attr);
31315 + if (unlikely(err))
31316 + pr_warn("failed %s under sysfs(%d)\n",
31317 + br_sysfs->name, err);
31322 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
31323 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
31324 +++ linux/fs/aufs/sysrq.c 2018-01-29 07:56:20.059991935 +0100
31327 + * Copyright (C) 2005-2017 Junjiro R. Okajima
31329 + * This program, aufs is free software; you can redistribute it and/or modify
31330 + * it under the terms of the GNU General Public License as published by
31331 + * the Free Software Foundation; either version 2 of the License, or
31332 + * (at your option) any later version.
31334 + * This program is distributed in the hope that it will be useful,
31335 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31336 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31337 + * GNU General Public License for more details.
31339 + * You should have received a copy of the GNU General Public License
31340 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31344 + * magic sysrq hanlder
31347 +/* #include <linux/sysrq.h> */
31348 +#include <linux/writeback.h>
31351 +/* ---------------------------------------------------------------------- */
31353 +static void sysrq_sb(struct super_block *sb)
31356 + struct au_sbinfo *sbinfo;
31357 + struct file *file;
31358 + struct hlist_bl_head *files;
31359 + struct hlist_bl_node *pos;
31360 + struct au_finfo *finfo;
31362 + plevel = au_plevel;
31363 + au_plevel = KERN_WARNING;
31365 + /* since we define pr_fmt, call printk directly */
31366 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
31368 + sbinfo = au_sbi(sb);
31369 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
31370 + pr("superblock\n");
31374 + pr("root dentry\n");
31375 + au_dpri_dentry(sb->s_root);
31376 + pr("root inode\n");
31377 + au_dpri_inode(d_inode(sb->s_root));
31382 + int err, i, j, ndentry;
31383 + struct au_dcsub_pages dpages;
31384 + struct au_dpage *dpage;
31386 + err = au_dpages_init(&dpages, GFP_ATOMIC);
31387 + if (unlikely(err))
31389 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
31391 + for (i = 0; i < dpages.ndpage; i++) {
31392 + dpage = dpages.dpages + i;
31393 + ndentry = dpage->ndentry;
31394 + for (j = 0; j < ndentry; j++)
31395 + au_dpri_dentry(dpage->dentries[j]);
31397 + au_dpages_free(&dpages);
31405 + pr("isolated inode\n");
31406 + spin_lock(&sb->s_inode_list_lock);
31407 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
31408 + spin_lock(&i->i_lock);
31409 + if (1 || hlist_empty(&i->i_dentry))
31410 + au_dpri_inode(i);
31411 + spin_unlock(&i->i_lock);
31413 + spin_unlock(&sb->s_inode_list_lock);
31417 + files = &au_sbi(sb)->si_files;
31418 + hlist_bl_lock(files);
31419 + hlist_bl_for_each_entry(finfo, pos, files, fi_hlist) {
31422 + file = finfo->fi_file;
31423 + mode = file_inode(file)->i_mode;
31424 + if (!special_file(mode))
31425 + au_dpri_file(file);
31427 + hlist_bl_unlock(files);
31431 + au_plevel = plevel;
31434 +/* ---------------------------------------------------------------------- */
31436 +/* module parameter */
31437 +static char *aufs_sysrq_key = "a";
31438 +module_param_named(sysrq, aufs_sysrq_key, charp, S_IRUGO);
31439 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
31441 +static void au_sysrq(int key __maybe_unused)
31443 + struct au_sbinfo *sbinfo;
31444 + struct hlist_bl_node *pos;
31447 + au_sbilist_lock();
31448 + hlist_bl_for_each_entry(sbinfo, pos, &au_sbilist, si_list)
31449 + sysrq_sb(sbinfo->si_sb);
31450 + au_sbilist_unlock();
31454 +static struct sysrq_key_op au_sysrq_op = {
31455 + .handler = au_sysrq,
31456 + .help_msg = "Aufs",
31457 + .action_msg = "Aufs",
31458 + .enable_mask = SYSRQ_ENABLE_DUMP
31461 +/* ---------------------------------------------------------------------- */
31463 +int __init au_sysrq_init(void)
31469 + key = *aufs_sysrq_key;
31470 + if ('a' <= key && key <= 'z')
31471 + err = register_sysrq_key(key, &au_sysrq_op);
31472 + if (unlikely(err))
31473 + pr_err("err %d, sysrq=%c\n", err, key);
31477 +void au_sysrq_fin(void)
31481 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
31482 + if (unlikely(err))
31483 + pr_err("err %d (ignored)\n", err);
31485 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
31486 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
31487 +++ linux/fs/aufs/vdir.c 2018-01-29 07:56:20.063325368 +0100
31490 + * Copyright (C) 2005-2017 Junjiro R. Okajima
31492 + * This program, aufs is free software; you can redistribute it and/or modify
31493 + * it under the terms of the GNU General Public License as published by
31494 + * the Free Software Foundation; either version 2 of the License, or
31495 + * (at your option) any later version.
31497 + * This program is distributed in the hope that it will be useful,
31498 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31499 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31500 + * GNU General Public License for more details.
31502 + * You should have received a copy of the GNU General Public License
31503 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31507 + * virtual or vertical directory
31512 +static unsigned int calc_size(int nlen)
31514 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
31517 +static int set_deblk_end(union au_vdir_deblk_p *p,
31518 + union au_vdir_deblk_p *deblk_end)
31520 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
31521 + p->de->de_str.len = 0;
31525 + return -1; /* error */
31528 +/* returns true or false */
31529 +static int is_deblk_end(union au_vdir_deblk_p *p,
31530 + union au_vdir_deblk_p *deblk_end)
31532 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
31533 + return !p->de->de_str.len;
31537 +static unsigned char *last_deblk(struct au_vdir *vdir)
31539 + return vdir->vd_deblk[vdir->vd_nblk - 1];
31542 +/* ---------------------------------------------------------------------- */
31544 +/* estimate the appropriate size for name hash table */
31545 +unsigned int au_rdhash_est(loff_t sz)
31553 + if (sz < AUFS_RDHASH_DEF)
31554 + n = AUFS_RDHASH_DEF;
31555 + /* pr_info("n %u\n", n); */
31560 + * the allocated memory has to be freed by
31561 + * au_nhash_wh_free() or au_nhash_de_free().
31563 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
31565 + struct hlist_head *head;
31569 + sz = sizeof(*nhash->nh_head) * num_hash;
31570 + head = kmalloc(sz, gfp);
31572 + nhash->nh_num = num_hash;
31573 + nhash->nh_head = head;
31574 + for (u = 0; u < num_hash; u++)
31575 + INIT_HLIST_HEAD(head++);
31576 + return 0; /* success */
31582 +static void nhash_count(struct hlist_head *head)
31586 + struct hlist_node *pos;
31589 + hlist_for_each(pos, head)
31591 + pr_info("%lu\n", n);
31595 +static void au_nhash_wh_do_free(struct hlist_head *head)
31597 + struct au_vdir_wh *pos;
31598 + struct hlist_node *node;
31600 + hlist_for_each_entry_safe(pos, node, head, wh_hash)
31604 +static void au_nhash_de_do_free(struct hlist_head *head)
31606 + struct au_vdir_dehstr *pos;
31607 + struct hlist_node *node;
31609 + hlist_for_each_entry_safe(pos, node, head, hash)
31610 + au_cache_free_vdir_dehstr(pos);
31613 +static void au_nhash_do_free(struct au_nhash *nhash,
31614 + void (*free)(struct hlist_head *head))
31617 + struct hlist_head *head;
31619 + n = nhash->nh_num;
31623 + head = nhash->nh_head;
31624 + while (n-- > 0) {
31625 + nhash_count(head);
31628 + kfree(nhash->nh_head);
31631 +void au_nhash_wh_free(struct au_nhash *whlist)
31633 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
31636 +static void au_nhash_de_free(struct au_nhash *delist)
31638 + au_nhash_do_free(delist, au_nhash_de_do_free);
31641 +/* ---------------------------------------------------------------------- */
31643 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
31647 + unsigned int u, n;
31648 + struct hlist_head *head;
31649 + struct au_vdir_wh *pos;
31652 + n = whlist->nh_num;
31653 + head = whlist->nh_head;
31654 + for (u = 0; u < n; u++, head++)
31655 + hlist_for_each_entry(pos, head, wh_hash)
31656 + if (pos->wh_bindex == btgt && ++num > limit)
31661 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
31662 + unsigned char *name,
31663 + unsigned int len)
31666 + /* const unsigned int magic_bit = 12; */
31668 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
31675 + /* v = hash_long(v, magic_bit); */
31676 + v %= nhash->nh_num;
31677 + return nhash->nh_head + v;
31680 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
31683 + return str->len == nlen && !memcmp(str->name, name, nlen);
31686 +/* returns found or not */
31687 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
31689 + struct hlist_head *head;
31690 + struct au_vdir_wh *pos;
31691 + struct au_vdir_destr *str;
31693 + head = au_name_hash(whlist, name, nlen);
31694 + hlist_for_each_entry(pos, head, wh_hash) {
31695 + str = &pos->wh_str;
31696 + AuDbg("%.*s\n", str->len, str->name);
31697 + if (au_nhash_test_name(str, name, nlen))
31703 +/* returns found(true) or not */
31704 +static int test_known(struct au_nhash *delist, char *name, int nlen)
31706 + struct hlist_head *head;
31707 + struct au_vdir_dehstr *pos;
31708 + struct au_vdir_destr *str;
31710 + head = au_name_hash(delist, name, nlen);
31711 + hlist_for_each_entry(pos, head, hash) {
31713 + AuDbg("%.*s\n", str->len, str->name);
31714 + if (au_nhash_test_name(str, name, nlen))
31720 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
31721 + unsigned char d_type)
31723 +#ifdef CONFIG_AUFS_SHWH
31724 + wh->wh_ino = ino;
31725 + wh->wh_type = d_type;
31729 +/* ---------------------------------------------------------------------- */
31731 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
31732 + unsigned int d_type, aufs_bindex_t bindex,
31733 + unsigned char shwh)
31736 + struct au_vdir_destr *str;
31737 + struct au_vdir_wh *wh;
31739 + AuDbg("%.*s\n", nlen, name);
31740 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
31743 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
31744 + if (unlikely(!wh))
31748 + wh->wh_bindex = bindex;
31750 + au_shwh_init_wh(wh, ino, d_type);
31751 + str = &wh->wh_str;
31753 + memcpy(str->name, name, nlen);
31754 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
31761 +static int append_deblk(struct au_vdir *vdir)
31764 + unsigned long ul;
31765 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
31766 + union au_vdir_deblk_p p, deblk_end;
31767 + unsigned char **o;
31770 + o = au_krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
31771 + GFP_NOFS, /*may_shrink*/0);
31772 + if (unlikely(!o))
31775 + vdir->vd_deblk = o;
31776 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
31778 + ul = vdir->vd_nblk++;
31779 + vdir->vd_deblk[ul] = p.deblk;
31780 + vdir->vd_last.ul = ul;
31781 + vdir->vd_last.p.deblk = p.deblk;
31782 + deblk_end.deblk = p.deblk + deblk_sz;
31783 + err = set_deblk_end(&p, &deblk_end);
31790 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
31791 + unsigned int d_type, struct au_nhash *delist)
31795 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
31796 + union au_vdir_deblk_p p, *room, deblk_end;
31797 + struct au_vdir_dehstr *dehstr;
31799 + p.deblk = last_deblk(vdir);
31800 + deblk_end.deblk = p.deblk + deblk_sz;
31801 + room = &vdir->vd_last.p;
31802 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
31803 + || !is_deblk_end(room, &deblk_end));
31805 + sz = calc_size(nlen);
31806 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
31807 + err = append_deblk(vdir);
31808 + if (unlikely(err))
31811 + p.deblk = last_deblk(vdir);
31812 + deblk_end.deblk = p.deblk + deblk_sz;
31814 + AuDebugOn(room->deblk != p.deblk);
31818 + dehstr = au_cache_alloc_vdir_dehstr();
31819 + if (unlikely(!dehstr))
31822 + dehstr->str = &room->de->de_str;
31823 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
31824 + room->de->de_ino = ino;
31825 + room->de->de_type = d_type;
31826 + room->de->de_str.len = nlen;
31827 + memcpy(room->de->de_str.name, name, nlen);
31830 + room->deblk += sz;
31831 + if (unlikely(set_deblk_end(room, &deblk_end)))
31832 + err = append_deblk(vdir);
31839 +/* ---------------------------------------------------------------------- */
31841 +void au_vdir_free(struct au_vdir *vdir)
31843 + unsigned char **deblk;
31845 + deblk = vdir->vd_deblk;
31846 + while (vdir->vd_nblk--)
31848 + kfree(vdir->vd_deblk);
31849 + au_cache_free_vdir(vdir);
31852 +static struct au_vdir *alloc_vdir(struct file *file)
31854 + struct au_vdir *vdir;
31855 + struct super_block *sb;
31858 + sb = file->f_path.dentry->d_sb;
31859 + SiMustAnyLock(sb);
31862 + vdir = au_cache_alloc_vdir();
31863 + if (unlikely(!vdir))
31866 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
31867 + if (unlikely(!vdir->vd_deblk))
31870 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
31871 + if (!vdir->vd_deblk_sz) {
31872 + /* estimate the appropriate size for deblk */
31873 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
31874 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
31876 + vdir->vd_nblk = 0;
31877 + vdir->vd_version = 0;
31878 + vdir->vd_jiffy = 0;
31879 + err = append_deblk(vdir);
31881 + return vdir; /* success */
31883 + kfree(vdir->vd_deblk);
31886 + au_cache_free_vdir(vdir);
31888 + vdir = ERR_PTR(err);
31892 +static int reinit_vdir(struct au_vdir *vdir)
31895 + union au_vdir_deblk_p p, deblk_end;
31897 + while (vdir->vd_nblk > 1) {
31898 + kfree(vdir->vd_deblk[vdir->vd_nblk - 1]);
31899 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
31902 + p.deblk = vdir->vd_deblk[0];
31903 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
31904 + err = set_deblk_end(&p, &deblk_end);
31905 + /* keep vd_dblk_sz */
31906 + vdir->vd_last.ul = 0;
31907 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
31908 + vdir->vd_version = 0;
31909 + vdir->vd_jiffy = 0;
31914 +/* ---------------------------------------------------------------------- */
31916 +#define AuFillVdir_CALLED 1
31917 +#define AuFillVdir_WHABLE (1 << 1)
31918 +#define AuFillVdir_SHWH (1 << 2)
31919 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
31920 +#define au_fset_fillvdir(flags, name) \
31921 + do { (flags) |= AuFillVdir_##name; } while (0)
31922 +#define au_fclr_fillvdir(flags, name) \
31923 + do { (flags) &= ~AuFillVdir_##name; } while (0)
31925 +#ifndef CONFIG_AUFS_SHWH
31926 +#undef AuFillVdir_SHWH
31927 +#define AuFillVdir_SHWH 0
31930 +struct fillvdir_arg {
31931 + struct dir_context ctx;
31932 + struct file *file;
31933 + struct au_vdir *vdir;
31934 + struct au_nhash delist;
31935 + struct au_nhash whlist;
31936 + aufs_bindex_t bindex;
31937 + unsigned int flags;
31941 +static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
31942 + loff_t offset __maybe_unused, u64 h_ino,
31943 + unsigned int d_type)
31945 + struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
31946 + char *name = (void *)__name;
31947 + struct super_block *sb;
31949 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
31952 + sb = arg->file->f_path.dentry->d_sb;
31953 + au_fset_fillvdir(arg->flags, CALLED);
31955 + if (nlen <= AUFS_WH_PFX_LEN
31956 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
31957 + if (test_known(&arg->delist, name, nlen)
31958 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
31959 + goto out; /* already exists or whiteouted */
31961 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
31963 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
31964 + d_type = DT_UNKNOWN;
31965 + arg->err = append_de(arg->vdir, name, nlen, ino,
31966 + d_type, &arg->delist);
31968 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
31969 + name += AUFS_WH_PFX_LEN;
31970 + nlen -= AUFS_WH_PFX_LEN;
31971 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
31972 + goto out; /* already whiteouted */
31975 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
31978 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
31979 + d_type = DT_UNKNOWN;
31980 + arg->err = au_nhash_append_wh
31981 + (&arg->whlist, name, nlen, ino, d_type,
31982 + arg->bindex, shwh);
31988 + arg->vdir->vd_jiffy = jiffies;
31990 + AuTraceErr(arg->err);
31994 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
31995 + struct au_nhash *whlist, struct au_nhash *delist)
31997 +#ifdef CONFIG_AUFS_SHWH
31999 + unsigned int nh, u;
32000 + struct hlist_head *head;
32001 + struct au_vdir_wh *pos;
32002 + struct hlist_node *n;
32004 + struct au_vdir_destr *destr;
32006 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
32009 + o = p = (void *)__get_free_page(GFP_NOFS);
32010 + if (unlikely(!p))
32014 + nh = whlist->nh_num;
32015 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32016 + p += AUFS_WH_PFX_LEN;
32017 + for (u = 0; u < nh; u++) {
32018 + head = whlist->nh_head + u;
32019 + hlist_for_each_entry_safe(pos, n, head, wh_hash) {
32020 + destr = &pos->wh_str;
32021 + memcpy(p, destr->name, destr->len);
32022 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
32023 + pos->wh_ino, pos->wh_type, delist);
32024 + if (unlikely(err))
32029 + free_page((unsigned long)o);
32039 +static int au_do_read_vdir(struct fillvdir_arg *arg)
32042 + unsigned int rdhash;
32044 + aufs_bindex_t bbot, bindex, btop;
32045 + unsigned char shwh;
32046 + struct file *hf, *file;
32047 + struct super_block *sb;
32049 + file = arg->file;
32050 + sb = file->f_path.dentry->d_sb;
32051 + SiMustAnyLock(sb);
32053 + rdhash = au_sbi(sb)->si_rdhash;
32055 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
32056 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
32057 + if (unlikely(err))
32059 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
32060 + if (unlikely(err))
32066 + if (au_opt_test(au_mntflags(sb), SHWH)) {
32068 + au_fset_fillvdir(arg->flags, SHWH);
32070 + btop = au_fbtop(file);
32071 + bbot = au_fbbot_dir(file);
32072 + for (bindex = btop; !err && bindex <= bbot; bindex++) {
32073 + hf = au_hf_dir(file, bindex);
32077 + offset = vfsub_llseek(hf, 0, SEEK_SET);
32079 + if (unlikely(offset))
32082 + arg->bindex = bindex;
32083 + au_fclr_fillvdir(arg->flags, WHABLE);
32085 + || (bindex != bbot
32086 + && au_br_whable(au_sbr_perm(sb, bindex))))
32087 + au_fset_fillvdir(arg->flags, WHABLE);
32090 + au_fclr_fillvdir(arg->flags, CALLED);
32092 + err = vfsub_iterate_dir(hf, &arg->ctx);
32095 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
32098 + * dir_relax() may be good for concurrency, but aufs should not
32099 + * use it since it will cause a lockdep problem.
32103 + if (!err && shwh)
32104 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
32106 + au_nhash_wh_free(&arg->whlist);
32109 + au_nhash_de_free(&arg->delist);
32114 +static int read_vdir(struct file *file, int may_read)
32117 + unsigned long expire;
32118 + unsigned char do_read;
32119 + struct fillvdir_arg arg = {
32121 + .actor = fillvdir
32124 + struct inode *inode;
32125 + struct au_vdir *vdir, *allocated;
32128 + inode = file_inode(file);
32129 + IMustLock(inode);
32130 + IiMustWriteLock(inode);
32131 + SiMustAnyLock(inode->i_sb);
32133 + allocated = NULL;
32135 + expire = au_sbi(inode->i_sb)->si_rdcache;
32136 + vdir = au_ivdir(inode);
32139 + vdir = alloc_vdir(file);
32140 + err = PTR_ERR(vdir);
32141 + if (IS_ERR(vdir))
32144 + allocated = vdir;
32145 + } else if (may_read
32146 + && (inode->i_version != vdir->vd_version
32147 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
32149 + err = reinit_vdir(vdir);
32150 + if (unlikely(err))
32155 + return 0; /* success */
32159 + err = au_do_read_vdir(&arg);
32161 + /* file->f_pos = 0; */ /* todo: ctx->pos? */
32162 + vdir->vd_version = inode->i_version;
32163 + vdir->vd_last.ul = 0;
32164 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
32166 + au_set_ivdir(inode, allocated);
32167 + } else if (allocated)
32168 + au_vdir_free(allocated);
32174 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
32177 + unsigned long ul, n;
32178 + const unsigned int deblk_sz = src->vd_deblk_sz;
32180 + AuDebugOn(tgt->vd_nblk != 1);
32183 + if (tgt->vd_nblk < src->vd_nblk) {
32184 + unsigned char **p;
32186 + p = au_krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
32187 + GFP_NOFS, /*may_shrink*/0);
32188 + if (unlikely(!p))
32190 + tgt->vd_deblk = p;
32193 + if (tgt->vd_deblk_sz != deblk_sz) {
32194 + unsigned char *p;
32196 + tgt->vd_deblk_sz = deblk_sz;
32197 + p = au_krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS,
32198 + /*may_shrink*/1);
32199 + if (unlikely(!p))
32201 + tgt->vd_deblk[0] = p;
32203 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
32204 + tgt->vd_version = src->vd_version;
32205 + tgt->vd_jiffy = src->vd_jiffy;
32207 + n = src->vd_nblk;
32208 + for (ul = 1; ul < n; ul++) {
32209 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
32211 + if (unlikely(!tgt->vd_deblk[ul]))
32215 + tgt->vd_nblk = n;
32216 + tgt->vd_last.ul = tgt->vd_last.ul;
32217 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
32218 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
32219 + - src->vd_deblk[src->vd_last.ul];
32221 + return 0; /* success */
32224 + rerr = reinit_vdir(tgt);
32229 +int au_vdir_init(struct file *file)
32232 + struct inode *inode;
32233 + struct au_vdir *vdir_cache, *allocated;
32235 + /* test file->f_pos here instead of ctx->pos */
32236 + err = read_vdir(file, !file->f_pos);
32237 + if (unlikely(err))
32240 + allocated = NULL;
32241 + vdir_cache = au_fvdir_cache(file);
32242 + if (!vdir_cache) {
32243 + vdir_cache = alloc_vdir(file);
32244 + err = PTR_ERR(vdir_cache);
32245 + if (IS_ERR(vdir_cache))
32247 + allocated = vdir_cache;
32248 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
32249 + /* test file->f_pos here instead of ctx->pos */
32250 + err = reinit_vdir(vdir_cache);
32251 + if (unlikely(err))
32254 + return 0; /* success */
32256 + inode = file_inode(file);
32257 + err = copy_vdir(vdir_cache, au_ivdir(inode));
32259 + file->f_version = inode->i_version;
32261 + au_set_fvdir_cache(file, allocated);
32262 + } else if (allocated)
32263 + au_vdir_free(allocated);
32269 +static loff_t calc_offset(struct au_vdir *vdir)
32272 + union au_vdir_deblk_p p;
32274 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
32275 + offset = vdir->vd_last.p.deblk - p.deblk;
32276 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
32280 +/* returns true or false */
32281 +static int seek_vdir(struct file *file, struct dir_context *ctx)
32284 + unsigned int deblk_sz;
32285 + unsigned long ul, n;
32287 + union au_vdir_deblk_p p, deblk_end;
32288 + struct au_vdir *vdir_cache;
32291 + vdir_cache = au_fvdir_cache(file);
32292 + offset = calc_offset(vdir_cache);
32293 + AuDbg("offset %lld\n", offset);
32294 + if (ctx->pos == offset)
32297 + vdir_cache->vd_last.ul = 0;
32298 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
32303 + deblk_sz = vdir_cache->vd_deblk_sz;
32304 + ul = div64_u64(ctx->pos, deblk_sz);
32305 + AuDbg("ul %lu\n", ul);
32306 + if (ul >= vdir_cache->vd_nblk)
32309 + n = vdir_cache->vd_nblk;
32310 + for (; ul < n; ul++) {
32311 + p.deblk = vdir_cache->vd_deblk[ul];
32312 + deblk_end.deblk = p.deblk + deblk_sz;
32314 + offset *= deblk_sz;
32315 + while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
32318 + l = calc_size(p.de->de_str.len);
32322 + if (!is_deblk_end(&p, &deblk_end)) {
32324 + vdir_cache->vd_last.ul = ul;
32325 + vdir_cache->vd_last.p = p;
32332 + AuTraceErr(!valid);
32336 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
32338 + unsigned int l, deblk_sz;
32339 + union au_vdir_deblk_p deblk_end;
32340 + struct au_vdir *vdir_cache;
32341 + struct au_vdir_de *de;
32343 + vdir_cache = au_fvdir_cache(file);
32344 + if (!seek_vdir(file, ctx))
32347 + deblk_sz = vdir_cache->vd_deblk_sz;
32349 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32350 + deblk_end.deblk += deblk_sz;
32351 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
32352 + de = vdir_cache->vd_last.p.de;
32353 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
32354 + de->de_str.len, de->de_str.name, ctx->pos,
32355 + (unsigned long)de->de_ino, de->de_type);
32356 + if (unlikely(!dir_emit(ctx, de->de_str.name,
32357 + de->de_str.len, de->de_ino,
32359 + /* todo: ignore the error caused by udba? */
32360 + /* return err; */
32364 + l = calc_size(de->de_str.len);
32365 + vdir_cache->vd_last.p.deblk += l;
32368 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
32369 + vdir_cache->vd_last.ul++;
32370 + vdir_cache->vd_last.p.deblk
32371 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
32372 + ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
32381 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
32382 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
32383 +++ linux/fs/aufs/vfsub.c 2018-01-29 07:56:20.063325368 +0100
32386 + * Copyright (C) 2005-2017 Junjiro R. Okajima
32388 + * This program, aufs is free software; you can redistribute it and/or modify
32389 + * it under the terms of the GNU General Public License as published by
32390 + * the Free Software Foundation; either version 2 of the License, or
32391 + * (at your option) any later version.
32393 + * This program is distributed in the hope that it will be useful,
32394 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32395 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32396 + * GNU General Public License for more details.
32398 + * You should have received a copy of the GNU General Public License
32399 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32403 + * sub-routines for VFS
32406 +#include <linux/mnt_namespace.h>
32407 +#include <linux/namei.h>
32408 +#include <linux/nsproxy.h>
32409 +#include <linux/security.h>
32410 +#include <linux/splice.h>
32413 +#ifdef CONFIG_AUFS_BR_FUSE
32414 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb)
32416 + if (!au_test_fuse(h_sb) || !au_userns)
32419 + return is_current_mnt_ns(mnt) ? 0 : -EACCES;
32423 +int vfsub_sync_filesystem(struct super_block *h_sb, int wait)
32428 + down_read(&h_sb->s_umount);
32429 + err = __sync_filesystem(h_sb, wait);
32430 + up_read(&h_sb->s_umount);
32436 +/* ---------------------------------------------------------------------- */
32438 +int vfsub_update_h_iattr(struct path *h_path, int *did)
32442 + struct super_block *h_sb;
32444 + /* for remote fs, leave work for its getattr or d_revalidate */
32445 + /* for bad i_attr fs, handle them in aufs_getattr() */
32446 + /* still some fs may acquire i_mutex. we need to skip them */
32450 + h_sb = h_path->dentry->d_sb;
32451 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
32453 + err = vfsub_getattr(h_path, &st);
32458 +/* ---------------------------------------------------------------------- */
32460 +struct file *vfsub_dentry_open(struct path *path, int flags)
32462 + struct file *file;
32464 + file = dentry_open(path, flags /* | __FMODE_NONOTIFY */,
32466 + if (!IS_ERR_OR_NULL(file)
32467 + && (file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
32468 + i_readcount_inc(d_inode(path->dentry));
32473 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
32475 + struct file *file;
32478 + file = filp_open(path,
32479 + oflags /* | __FMODE_NONOTIFY */,
32482 + if (IS_ERR(file))
32484 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
32491 + * Ideally this function should call VFS:do_last() in order to keep all its
32492 + * checkings. But it is very hard for aufs to regenerate several VFS internal
32493 + * structure such as nameidata. This is a second (or third) best approach.
32494 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
32496 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
32497 + struct vfsub_aopen_args *args, struct au_branch *br)
32500 + struct file *file = args->file;
32501 + /* copied from linux/fs/namei.c:atomic_open() */
32502 + struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
32505 + AuDebugOn(!dir->i_op->atomic_open);
32507 + err = au_br_test_oflag(args->open_flag, br);
32508 + if (unlikely(err))
32511 + args->file->f_path.dentry = DENTRY_NOT_SET;
32512 + args->file->f_path.mnt = au_br_mnt(br);
32513 + err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
32514 + args->create_mode, args->opened);
32516 + /* some filesystems don't set FILE_CREATED while succeeded? */
32517 + if (*args->opened & FILE_CREATED)
32518 + fsnotify_create(dir, dentry);
32524 + /* todo: call VFS:may_open() here */
32525 + err = open_check_o_direct(file);
32526 + /* todo: ima_file_check() too? */
32527 + if (!err && (args->open_flag & __FMODE_EXEC))
32528 + err = deny_write_access(file);
32529 + if (unlikely(err))
32530 + /* note that the file is created and still opened */
32535 + fsnotify_open(file);
32541 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
32545 + err = kern_path(name, flags, path);
32546 + if (!err && d_is_positive(path->dentry))
32547 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
32551 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
32552 + struct dentry *parent, int len)
32554 + struct path path = {
32558 + path.dentry = lookup_one_len_unlocked(name, parent, len);
32559 + if (IS_ERR(path.dentry))
32561 + if (d_is_positive(path.dentry))
32562 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
32565 + AuTraceErrPtr(path.dentry);
32566 + return path.dentry;
32569 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
32572 + struct path path = {
32576 + /* VFS checks it too, but by WARN_ON_ONCE() */
32577 + IMustLock(d_inode(parent));
32579 + path.dentry = lookup_one_len(name, parent, len);
32580 + if (IS_ERR(path.dentry))
32582 + if (d_is_positive(path.dentry))
32583 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
32586 + AuTraceErrPtr(path.dentry);
32587 + return path.dentry;
32590 +void vfsub_call_lkup_one(void *args)
32592 + struct vfsub_lkup_one_args *a = args;
32593 + *a->errp = vfsub_lkup_one(a->name, a->parent);
32596 +/* ---------------------------------------------------------------------- */
32598 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
32599 + struct dentry *d2, struct au_hinode *hdir2)
32601 + struct dentry *d;
32604 + d = lock_rename(d1, d2);
32606 + au_hn_suspend(hdir1);
32607 + if (hdir1 != hdir2)
32608 + au_hn_suspend(hdir2);
32613 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
32614 + struct dentry *d2, struct au_hinode *hdir2)
32616 + au_hn_resume(hdir1);
32617 + if (hdir1 != hdir2)
32618 + au_hn_resume(hdir2);
32620 + unlock_rename(d1, d2);
32624 +/* ---------------------------------------------------------------------- */
32626 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
32629 + struct dentry *d;
32633 + d = path->dentry;
32634 + path->dentry = d->d_parent;
32635 + err = security_path_mknod(path, d, mode, 0);
32636 + path->dentry = d;
32637 + if (unlikely(err))
32641 + err = vfs_create(dir, path->dentry, mode, want_excl);
32644 + struct path tmp = *path;
32647 + vfsub_update_h_iattr(&tmp, &did);
32649 + tmp.dentry = path->dentry->d_parent;
32650 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32659 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
32662 + struct dentry *d;
32666 + d = path->dentry;
32667 + path->dentry = d->d_parent;
32668 + err = security_path_symlink(path, d, symname);
32669 + path->dentry = d;
32670 + if (unlikely(err))
32674 + err = vfs_symlink(dir, path->dentry, symname);
32677 + struct path tmp = *path;
32680 + vfsub_update_h_iattr(&tmp, &did);
32682 + tmp.dentry = path->dentry->d_parent;
32683 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32692 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
32695 + struct dentry *d;
32699 + d = path->dentry;
32700 + path->dentry = d->d_parent;
32701 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
32702 + path->dentry = d;
32703 + if (unlikely(err))
32707 + err = vfs_mknod(dir, path->dentry, mode, dev);
32710 + struct path tmp = *path;
32713 + vfsub_update_h_iattr(&tmp, &did);
32715 + tmp.dentry = path->dentry->d_parent;
32716 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32725 +static int au_test_nlink(struct inode *inode)
32727 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
32729 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
32730 + || inode->i_nlink < link_max)
32735 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
32736 + struct inode **delegated_inode)
32739 + struct dentry *d;
32743 + err = au_test_nlink(d_inode(src_dentry));
32744 + if (unlikely(err))
32747 + /* we don't call may_linkat() */
32748 + d = path->dentry;
32749 + path->dentry = d->d_parent;
32750 + err = security_path_link(src_dentry, path, d);
32751 + path->dentry = d;
32752 + if (unlikely(err))
32756 + err = vfs_link(src_dentry, dir, path->dentry, delegated_inode);
32759 + struct path tmp = *path;
32762 + /* fuse has different memory inode for the same inumber */
32763 + vfsub_update_h_iattr(&tmp, &did);
32765 + tmp.dentry = path->dentry->d_parent;
32766 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32767 + tmp.dentry = src_dentry;
32768 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32777 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
32778 + struct inode *dir, struct path *path,
32779 + struct inode **delegated_inode, unsigned int flags)
32782 + struct path tmp = {
32785 + struct dentry *d;
32788 + IMustLock(src_dir);
32790 + d = path->dentry;
32791 + path->dentry = d->d_parent;
32792 + tmp.dentry = src_dentry->d_parent;
32793 + err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
32794 + path->dentry = d;
32795 + if (unlikely(err))
32799 + err = vfs_rename(src_dir, src_dentry, dir, path->dentry,
32800 + delegated_inode, flags);
32805 + tmp.dentry = d->d_parent;
32806 + vfsub_update_h_iattr(&tmp, &did);
32808 + tmp.dentry = src_dentry;
32809 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32810 + tmp.dentry = src_dentry->d_parent;
32811 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32820 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
32823 + struct dentry *d;
32827 + d = path->dentry;
32828 + path->dentry = d->d_parent;
32829 + err = security_path_mkdir(path, d, mode);
32830 + path->dentry = d;
32831 + if (unlikely(err))
32835 + err = vfs_mkdir(dir, path->dentry, mode);
32838 + struct path tmp = *path;
32841 + vfsub_update_h_iattr(&tmp, &did);
32843 + tmp.dentry = path->dentry->d_parent;
32844 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
32853 +int vfsub_rmdir(struct inode *dir, struct path *path)
32856 + struct dentry *d;
32860 + d = path->dentry;
32861 + path->dentry = d->d_parent;
32862 + err = security_path_rmdir(path, d);
32863 + path->dentry = d;
32864 + if (unlikely(err))
32868 + err = vfs_rmdir(dir, path->dentry);
32871 + struct path tmp = {
32872 + .dentry = path->dentry->d_parent,
32876 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
32883 +/* ---------------------------------------------------------------------- */
32885 +/* todo: support mmap_sem? */
32886 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
32892 + err = vfs_read(file, ubuf, count, ppos);
32895 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
32899 +/* todo: kernel_read()? */
32900 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
32904 + mm_segment_t oldfs;
32911 + oldfs = get_fs();
32912 + set_fs(KERNEL_DS);
32913 + err = vfsub_read_u(file, buf.u, count, ppos);
32918 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
32924 + err = vfs_write(file, ubuf, count, ppos);
32927 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
32931 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
32934 + mm_segment_t oldfs;
32937 + const char __user *u;
32941 + oldfs = get_fs();
32942 + set_fs(KERNEL_DS);
32943 + err = vfsub_write_u(file, buf.u, count, ppos);
32948 +int vfsub_flush(struct file *file, fl_owner_t id)
32953 + if (file->f_op->flush) {
32954 + if (!au_test_nfs(file->f_path.dentry->d_sb))
32955 + err = file->f_op->flush(file, id);
32958 + err = file->f_op->flush(file, id);
32962 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
32968 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
32972 + AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
32975 + err = iterate_dir(file, ctx);
32978 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
32983 +long vfsub_splice_to(struct file *in, loff_t *ppos,
32984 + struct pipe_inode_info *pipe, size_t len,
32985 + unsigned int flags)
32990 + err = do_splice_to(in, ppos, pipe, len, flags);
32992 + file_accessed(in);
32994 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
32998 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
32999 + loff_t *ppos, size_t len, unsigned int flags)
33004 + err = do_splice_from(pipe, out, ppos, len, flags);
33007 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
33011 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
33015 + /* file can be NULL */
33017 + err = vfs_fsync(file, datasync);
33021 + AuDebugOn(!file);
33022 + path = &file->f_path;
33024 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
33029 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
33030 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33031 + struct file *h_file)
33034 + struct inode *h_inode;
33035 + struct super_block *h_sb;
33038 + err = vfsub_truncate(h_path, length);
33042 + h_inode = d_inode(h_path->dentry);
33043 + h_sb = h_inode->i_sb;
33045 + sb_start_write(h_sb);
33047 + err = locks_verify_truncate(h_inode, h_file, length);
33049 + err = security_path_truncate(h_path);
33052 + err = do_truncate(h_path->dentry, length, attr, h_file);
33056 + sb_end_write(h_sb);
33063 +/* ---------------------------------------------------------------------- */
33065 +struct au_vfsub_mkdir_args {
33067 + struct inode *dir;
33068 + struct path *path;
33072 +static void au_call_vfsub_mkdir(void *args)
33074 + struct au_vfsub_mkdir_args *a = args;
33075 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
33078 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
33080 + int err, do_sio, wkq_err;
33082 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
33085 + err = vfsub_mkdir(dir, path, mode);
33088 + struct au_vfsub_mkdir_args args = {
33094 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
33095 + if (unlikely(wkq_err))
33102 +struct au_vfsub_rmdir_args {
33104 + struct inode *dir;
33105 + struct path *path;
33108 +static void au_call_vfsub_rmdir(void *args)
33110 + struct au_vfsub_rmdir_args *a = args;
33111 + *a->errp = vfsub_rmdir(a->dir, a->path);
33114 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
33116 + int err, do_sio, wkq_err;
33118 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
33121 + err = vfsub_rmdir(dir, path);
33124 + struct au_vfsub_rmdir_args args = {
33129 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
33130 + if (unlikely(wkq_err))
33137 +/* ---------------------------------------------------------------------- */
33139 +struct notify_change_args {
33141 + struct path *path;
33142 + struct iattr *ia;
33143 + struct inode **delegated_inode;
33146 +static void call_notify_change(void *args)
33148 + struct notify_change_args *a = args;
33149 + struct inode *h_inode;
33151 + h_inode = d_inode(a->path->dentry);
33152 + IMustLock(h_inode);
33154 + *a->errp = -EPERM;
33155 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
33157 + *a->errp = notify_change(a->path->dentry, a->ia,
33158 + a->delegated_inode);
33161 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
33163 + AuTraceErr(*a->errp);
33166 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33167 + struct inode **delegated_inode)
33170 + struct notify_change_args args = {
33174 + .delegated_inode = delegated_inode
33177 + call_notify_change(&args);
33182 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33183 + struct inode **delegated_inode)
33185 + int err, wkq_err;
33186 + struct notify_change_args args = {
33190 + .delegated_inode = delegated_inode
33193 + wkq_err = au_wkq_wait(call_notify_change, &args);
33194 + if (unlikely(wkq_err))
33200 +/* ---------------------------------------------------------------------- */
33202 +struct unlink_args {
33204 + struct inode *dir;
33205 + struct path *path;
33206 + struct inode **delegated_inode;
33209 +static void call_unlink(void *args)
33211 + struct unlink_args *a = args;
33212 + struct dentry *d = a->path->dentry;
33213 + struct inode *h_inode;
33214 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
33215 + && au_dcount(d) == 1);
33217 + IMustLock(a->dir);
33219 + a->path->dentry = d->d_parent;
33220 + *a->errp = security_path_unlink(a->path, d);
33221 + a->path->dentry = d;
33222 + if (unlikely(*a->errp))
33225 + if (!stop_sillyrename)
33228 + if (d_is_positive(d)) {
33229 + h_inode = d_inode(d);
33234 + *a->errp = vfs_unlink(a->dir, d, a->delegated_inode);
33237 + struct path tmp = {
33238 + .dentry = d->d_parent,
33239 + .mnt = a->path->mnt
33241 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
33244 + if (!stop_sillyrename)
33249 + AuTraceErr(*a->errp);
33253 + * @dir: must be locked.
33254 + * @dentry: target dentry.
33256 +int vfsub_unlink(struct inode *dir, struct path *path,
33257 + struct inode **delegated_inode, int force)
33260 + struct unlink_args args = {
33264 + .delegated_inode = delegated_inode
33268 + call_unlink(&args);
33272 + wkq_err = au_wkq_wait(call_unlink, &args);
33273 + if (unlikely(wkq_err))
33279 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
33280 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
33281 +++ linux/fs/aufs/vfsub.h 2018-01-29 07:56:20.063325368 +0100
33284 + * Copyright (C) 2005-2017 Junjiro R. Okajima
33286 + * This program, aufs is free software; you can redistribute it and/or modify
33287 + * it under the terms of the GNU General Public License as published by
33288 + * the Free Software Foundation; either version 2 of the License, or
33289 + * (at your option) any later version.
33291 + * This program is distributed in the hope that it will be useful,
33292 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33293 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33294 + * GNU General Public License for more details.
33296 + * You should have received a copy of the GNU General Public License
33297 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33301 + * sub-routines for VFS
33304 +#ifndef __AUFS_VFSUB_H__
33305 +#define __AUFS_VFSUB_H__
33309 +#include <linux/fs.h>
33310 +#include <linux/mount.h>
33311 +#include <linux/posix_acl.h>
33312 +#include <linux/xattr.h>
33313 +#include "debug.h"
33315 +/* copied from linux/fs/internal.h */
33316 +/* todo: BAD approach!! */
33317 +extern void __mnt_drop_write(struct vfsmount *);
33318 +extern int open_check_o_direct(struct file *f);
33320 +/* ---------------------------------------------------------------------- */
33322 +/* lock subclass for lower inode */
33323 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
33324 +/* reduce? gave up. */
33326 + AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
33327 + AuLsc_I_PARENT, /* lower inode, parent first */
33328 + AuLsc_I_PARENT2, /* copyup dirs */
33329 + AuLsc_I_PARENT3, /* copyup wh */
33335 +/* to debug easier, do not make them inlined functions */
33336 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
33337 +#define IMustLock(i) AuDebugOn(!inode_is_locked(i))
33339 +/* why VFS doesn't define it? */
33341 +void vfsub_inode_lock_shared_nested(struct inode *inode, unsigned int sc)
33343 + down_read_nested(&inode->i_rwsem, sc);
33346 +/* ---------------------------------------------------------------------- */
33348 +static inline void vfsub_drop_nlink(struct inode *inode)
33350 + AuDebugOn(!inode->i_nlink);
33351 + drop_nlink(inode);
33354 +static inline void vfsub_dead_dir(struct inode *inode)
33356 + AuDebugOn(!S_ISDIR(inode->i_mode));
33357 + inode->i_flags |= S_DEAD;
33358 + clear_nlink(inode);
33361 +static inline int vfsub_native_ro(struct inode *inode)
33363 + return sb_rdonly(inode->i_sb)
33364 + || IS_RDONLY(inode)
33365 + /* || IS_APPEND(inode) */
33366 + || IS_IMMUTABLE(inode);
33369 +#ifdef CONFIG_AUFS_BR_FUSE
33370 +int vfsub_test_mntns(struct vfsmount *mnt, struct super_block *h_sb);
33372 +AuStubInt0(vfsub_test_mntns, struct vfsmount *mnt, struct super_block *h_sb);
33375 +int vfsub_sync_filesystem(struct super_block *h_sb, int wait);
33377 +/* ---------------------------------------------------------------------- */
33379 +int vfsub_update_h_iattr(struct path *h_path, int *did);
33380 +struct file *vfsub_dentry_open(struct path *path, int flags);
33381 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
33382 +struct vfsub_aopen_args {
33383 + struct file *file;
33384 + unsigned int open_flag;
33385 + umode_t create_mode;
33389 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
33390 + struct vfsub_aopen_args *args, struct au_branch *br);
33391 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
33393 +struct dentry *vfsub_lookup_one_len_unlocked(const char *name,
33394 + struct dentry *parent, int len);
33395 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
33398 +struct vfsub_lkup_one_args {
33399 + struct dentry **errp;
33400 + struct qstr *name;
33401 + struct dentry *parent;
33404 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
33405 + struct dentry *parent)
33407 + return vfsub_lookup_one_len(name->name, parent, name->len);
33410 +void vfsub_call_lkup_one(void *args);
33412 +/* ---------------------------------------------------------------------- */
33414 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
33419 + err = mnt_want_write(mnt);
33424 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
33427 + mnt_drop_write(mnt);
33431 +#if 0 /* reserved */
33432 +static inline void vfsub_mnt_drop_write_file(struct file *file)
33435 + mnt_drop_write_file(file);
33440 +/* ---------------------------------------------------------------------- */
33443 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
33444 + struct dentry *d2, struct au_hinode *hdir2);
33445 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
33446 + struct dentry *d2, struct au_hinode *hdir2);
33448 +int vfsub_create(struct inode *dir, struct path *path, int mode,
33450 +int vfsub_symlink(struct inode *dir, struct path *path,
33451 + const char *symname);
33452 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
33453 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
33454 + struct path *path, struct inode **delegated_inode);
33455 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
33456 + struct inode *hdir, struct path *path,
33457 + struct inode **delegated_inode, unsigned int flags);
33458 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
33459 +int vfsub_rmdir(struct inode *dir, struct path *path);
33461 +/* ---------------------------------------------------------------------- */
33463 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
33465 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
33467 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
33469 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
33471 +int vfsub_flush(struct file *file, fl_owner_t id);
33472 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
33474 +static inline loff_t vfsub_f_size_read(struct file *file)
33476 + return i_size_read(file_inode(file));
33479 +static inline unsigned int vfsub_file_flags(struct file *file)
33481 + unsigned int flags;
33483 + spin_lock(&file->f_lock);
33484 + flags = file->f_flags;
33485 + spin_unlock(&file->f_lock);
33490 +static inline int vfsub_file_execed(struct file *file)
33492 + /* todo: direct access f_flags */
33493 + return !!(vfsub_file_flags(file) & __FMODE_EXEC);
33496 +#if 0 /* reserved */
33497 +static inline void vfsub_file_accessed(struct file *h_file)
33499 + file_accessed(h_file);
33500 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
33504 +#if 0 /* reserved */
33505 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
33506 + struct dentry *h_dentry)
33508 + struct path h_path = {
33509 + .dentry = h_dentry,
33512 + touch_atime(&h_path);
33513 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
33517 +static inline int vfsub_update_time(struct inode *h_inode, struct timespec *ts,
33520 + return update_time(h_inode, ts, flags);
33521 + /* no vfsub_update_h_iattr() since we don't have struct path */
33524 +#ifdef CONFIG_FS_POSIX_ACL
33525 +static inline int vfsub_acl_chmod(struct inode *h_inode, umode_t h_mode)
33529 + err = posix_acl_chmod(h_inode, h_mode);
33530 + if (err == -EOPNOTSUPP)
33535 +AuStubInt0(vfsub_acl_chmod, struct inode *h_inode, umode_t h_mode);
33538 +long vfsub_splice_to(struct file *in, loff_t *ppos,
33539 + struct pipe_inode_info *pipe, size_t len,
33540 + unsigned int flags);
33541 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
33542 + loff_t *ppos, size_t len, unsigned int flags);
33544 +static inline long vfsub_truncate(struct path *path, loff_t length)
33549 + err = vfs_truncate(path, length);
33554 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
33555 + struct file *h_file);
33556 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
33559 + * re-use branch fs's ioctl(FICLONE) while aufs itself doesn't support such
33562 +static inline int vfsub_clone_file_range(struct file *src, struct file *dst,
33568 + err = vfs_clone_file_range(src, 0, dst, 0, len);
33574 +/* copy_file_range(2) is a systemcall */
33575 +static inline ssize_t vfsub_copy_file_range(struct file *src, loff_t src_pos,
33576 + struct file *dst, loff_t dst_pos,
33577 + size_t len, unsigned int flags)
33582 + ssz = vfs_copy_file_range(src, src_pos, dst, dst_pos, len, flags);
33588 +/* ---------------------------------------------------------------------- */
33590 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
33595 + err = vfs_llseek(file, offset, origin);
33600 +/* ---------------------------------------------------------------------- */
33602 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
33603 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
33604 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
33605 + struct inode **delegated_inode);
33606 +int vfsub_notify_change(struct path *path, struct iattr *ia,
33607 + struct inode **delegated_inode);
33608 +int vfsub_unlink(struct inode *dir, struct path *path,
33609 + struct inode **delegated_inode, int force);
33611 +static inline int vfsub_getattr(const struct path *path, struct kstat *st)
33613 + return vfs_getattr(path, st, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
33616 +/* ---------------------------------------------------------------------- */
33618 +static inline int vfsub_setxattr(struct dentry *dentry, const char *name,
33619 + const void *value, size_t size, int flags)
33624 + err = vfs_setxattr(dentry, name, value, size, flags);
33630 +static inline int vfsub_removexattr(struct dentry *dentry, const char *name)
33635 + err = vfs_removexattr(dentry, name);
33641 +#endif /* __KERNEL__ */
33642 +#endif /* __AUFS_VFSUB_H__ */
33643 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
33644 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
33645 +++ linux/fs/aufs/wbr_policy.c 2018-01-29 07:56:20.063325368 +0100
33648 + * Copyright (C) 2005-2017 Junjiro R. Okajima
33650 + * This program, aufs is free software; you can redistribute it and/or modify
33651 + * it under the terms of the GNU General Public License as published by
33652 + * the Free Software Foundation; either version 2 of the License, or
33653 + * (at your option) any later version.
33655 + * This program is distributed in the hope that it will be useful,
33656 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33657 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33658 + * GNU General Public License for more details.
33660 + * You should have received a copy of the GNU General Public License
33661 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33665 + * policies for selecting one among multiple writable branches
33668 +#include <linux/statfs.h>
33671 +/* subset of cpup_attr() */
33672 +static noinline_for_stack
33673 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
33677 + struct inode *h_isrc;
33679 + h_isrc = d_inode(h_src);
33680 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
33681 + ia.ia_mode = h_isrc->i_mode;
33682 + ia.ia_uid = h_isrc->i_uid;
33683 + ia.ia_gid = h_isrc->i_gid;
33684 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
33685 + au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
33686 + /* no delegation since it is just created */
33687 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
33689 + /* is this nfs only? */
33690 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
33691 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
33692 + ia.ia_mode = h_isrc->i_mode;
33693 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
33699 +#define AuCpdown_PARENT_OPQ 1
33700 +#define AuCpdown_WHED (1 << 1)
33701 +#define AuCpdown_MADE_DIR (1 << 2)
33702 +#define AuCpdown_DIROPQ (1 << 3)
33703 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
33704 +#define au_fset_cpdown(flags, name) \
33705 + do { (flags) |= AuCpdown_##name; } while (0)
33706 +#define au_fclr_cpdown(flags, name) \
33707 + do { (flags) &= ~AuCpdown_##name; } while (0)
33709 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
33710 + unsigned int *flags)
33713 + struct dentry *opq_dentry;
33715 + opq_dentry = au_diropq_create(dentry, bdst);
33716 + err = PTR_ERR(opq_dentry);
33717 + if (IS_ERR(opq_dentry))
33719 + dput(opq_dentry);
33720 + au_fset_cpdown(*flags, DIROPQ);
33726 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
33727 + struct inode *dir, aufs_bindex_t bdst)
33730 + struct path h_path;
33731 + struct au_branch *br;
33733 + br = au_sbr(dentry->d_sb, bdst);
33734 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
33735 + err = PTR_ERR(h_path.dentry);
33736 + if (IS_ERR(h_path.dentry))
33740 + if (d_is_positive(h_path.dentry)) {
33741 + h_path.mnt = au_br_mnt(br);
33742 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
33745 + dput(h_path.dentry);
33751 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
33752 + struct au_pin *pin,
33753 + struct dentry *h_parent, void *arg)
33756 + aufs_bindex_t bopq, btop;
33757 + struct path h_path;
33758 + struct dentry *parent;
33759 + struct inode *h_dir, *h_inode, *inode, *dir;
33760 + unsigned int *flags = arg;
33762 + btop = au_dbtop(dentry);
33763 + /* dentry is di-locked */
33764 + parent = dget_parent(dentry);
33765 + dir = d_inode(parent);
33766 + h_dir = d_inode(h_parent);
33767 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
33768 + IMustLock(h_dir);
33770 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
33771 + if (unlikely(err < 0))
33773 + h_path.dentry = au_h_dptr(dentry, bdst);
33774 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
33775 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path,
33776 + S_IRWXU | S_IRUGO | S_IXUGO);
33777 + if (unlikely(err))
33779 + au_fset_cpdown(*flags, MADE_DIR);
33781 + bopq = au_dbdiropq(dentry);
33782 + au_fclr_cpdown(*flags, WHED);
33783 + au_fclr_cpdown(*flags, DIROPQ);
33784 + if (au_dbwh(dentry) == bdst)
33785 + au_fset_cpdown(*flags, WHED);
33786 + if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
33787 + au_fset_cpdown(*flags, PARENT_OPQ);
33788 + h_inode = d_inode(h_path.dentry);
33789 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
33790 + if (au_ftest_cpdown(*flags, WHED)) {
33791 + err = au_cpdown_dir_opq(dentry, bdst, flags);
33792 + if (unlikely(err)) {
33793 + inode_unlock(h_inode);
33798 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, btop));
33799 + inode_unlock(h_inode);
33800 + if (unlikely(err))
33803 + if (au_ftest_cpdown(*flags, WHED)) {
33804 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
33805 + if (unlikely(err))
33809 + inode = d_inode(dentry);
33810 + if (au_ibbot(inode) < bdst)
33811 + au_set_ibbot(inode, bdst);
33812 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
33813 + au_hi_flags(inode, /*isdir*/1));
33814 + au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
33815 + goto out; /* success */
33819 + if (au_ftest_cpdown(*flags, DIROPQ)) {
33820 + inode_lock_nested(h_inode, AuLsc_I_CHILD);
33821 + rerr = au_diropq_remove(dentry, bdst);
33822 + inode_unlock(h_inode);
33823 + if (unlikely(rerr)) {
33824 + AuIOErr("failed removing diropq for %pd b%d (%d)\n",
33825 + dentry, bdst, rerr);
33831 + if (au_ftest_cpdown(*flags, MADE_DIR)) {
33832 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
33833 + if (unlikely(rerr)) {
33834 + AuIOErr("failed removing %pd b%d (%d)\n",
33835 + dentry, bdst, rerr);
33840 + au_set_h_dptr(dentry, bdst, NULL);
33841 + if (au_dbbot(dentry) == bdst)
33842 + au_update_dbbot(dentry);
33848 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
33851 + unsigned int flags;
33854 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
33859 +/* ---------------------------------------------------------------------- */
33861 +/* policies for create */
33863 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
33865 + int err, i, j, ndentry;
33866 + aufs_bindex_t bopq;
33867 + struct au_dcsub_pages dpages;
33868 + struct au_dpage *dpage;
33869 + struct dentry **dentries, *parent, *d;
33871 + err = au_dpages_init(&dpages, GFP_NOFS);
33872 + if (unlikely(err))
33874 + parent = dget_parent(dentry);
33875 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
33876 + if (unlikely(err))
33880 + for (i = 0; i < dpages.ndpage; i++) {
33881 + dpage = dpages.dpages + i;
33882 + dentries = dpage->dentries;
33883 + ndentry = dpage->ndentry;
33884 + for (j = 0; j < ndentry; j++) {
33886 + di_read_lock_parent2(d, !AuLock_IR);
33887 + bopq = au_dbdiropq(d);
33888 + di_read_unlock(d, !AuLock_IR);
33889 + if (bopq >= 0 && bopq < err)
33896 + au_dpages_free(&dpages);
33901 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
33903 + for (; bindex >= 0; bindex--)
33904 + if (!au_br_rdonly(au_sbr(sb, bindex)))
33909 +/* top down parent */
33910 +static int au_wbr_create_tdp(struct dentry *dentry,
33911 + unsigned int flags __maybe_unused)
33914 + aufs_bindex_t btop, bindex;
33915 + struct super_block *sb;
33916 + struct dentry *parent, *h_parent;
33918 + sb = dentry->d_sb;
33919 + btop = au_dbtop(dentry);
33921 + if (!au_br_rdonly(au_sbr(sb, btop)))
33925 + parent = dget_parent(dentry);
33926 + for (bindex = au_dbtop(parent); bindex < btop; bindex++) {
33927 + h_parent = au_h_dptr(parent, bindex);
33928 + if (!h_parent || d_is_negative(h_parent))
33931 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
33938 + /* bottom up here */
33939 + if (unlikely(err < 0)) {
33940 + err = au_wbr_bu(sb, btop - 1);
33942 + err = au_wbr_nonopq(dentry, err);
33946 + AuDbg("b%d\n", err);
33950 +/* ---------------------------------------------------------------------- */
33952 +/* an exception for the policy other than tdp */
33953 +static int au_wbr_create_exp(struct dentry *dentry)
33956 + aufs_bindex_t bwh, bdiropq;
33957 + struct dentry *parent;
33960 + bwh = au_dbwh(dentry);
33961 + parent = dget_parent(dentry);
33962 + bdiropq = au_dbdiropq(parent);
33964 + if (bdiropq >= 0)
33965 + err = min(bdiropq, bwh);
33968 + AuDbg("%d\n", err);
33969 + } else if (bdiropq >= 0) {
33971 + AuDbg("%d\n", err);
33976 + err = au_wbr_nonopq(dentry, err);
33978 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
33981 + AuDbg("%d\n", err);
33985 +/* ---------------------------------------------------------------------- */
33988 +static int au_wbr_create_init_rr(struct super_block *sb)
33992 + err = au_wbr_bu(sb, au_sbbot(sb));
33993 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
33996 + AuDbg("b%d\n", err);
34000 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
34004 + aufs_bindex_t bindex, bbot;
34005 + struct super_block *sb;
34008 + err = au_wbr_create_exp(dentry);
34012 + sb = dentry->d_sb;
34013 + next = &au_sbi(sb)->si_wbr_rr_next;
34014 + bbot = au_sbbot(sb);
34016 + for (bindex = 0; bindex <= bbot; bindex++) {
34017 + if (!au_ftest_wbr(flags, DIR)) {
34018 + err = atomic_dec_return(next) + 1;
34019 + /* modulo for 0 is meaningless */
34020 + if (unlikely(!err))
34021 + err = atomic_dec_return(next) + 1;
34023 + err = atomic_read(next);
34024 + AuDbg("%d\n", err);
34027 + AuDbg("%d\n", err);
34028 + if (!au_br_rdonly(au_sbr(sb, err)))
34034 + err = au_wbr_nonopq(dentry, err);
34037 + AuDbg("%d\n", err);
34041 +/* ---------------------------------------------------------------------- */
34043 +/* most free space */
34044 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
34046 + struct super_block *sb;
34047 + struct au_branch *br;
34048 + struct au_wbr_mfs *mfs;
34049 + struct dentry *h_parent;
34050 + aufs_bindex_t bindex, bbot;
34052 + unsigned long long b, bavail;
34053 + struct path h_path;
34054 + /* reduce the stack usage */
34055 + struct kstatfs *st;
34057 + st = kmalloc(sizeof(*st), GFP_NOFS);
34058 + if (unlikely(!st)) {
34059 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
34064 + sb = dentry->d_sb;
34065 + mfs = &au_sbi(sb)->si_wbr_mfs;
34066 + MtxMustLock(&mfs->mfs_lock);
34067 + mfs->mfs_bindex = -EROFS;
34068 + mfs->mfsrr_bytes = 0;
34071 + bbot = au_sbbot(sb);
34073 + bindex = au_dbtop(parent);
34074 + bbot = au_dbtaildir(parent);
34077 + for (; bindex <= bbot; bindex++) {
34079 + h_parent = au_h_dptr(parent, bindex);
34080 + if (!h_parent || d_is_negative(h_parent))
34083 + br = au_sbr(sb, bindex);
34084 + if (au_br_rdonly(br))
34087 + /* sb->s_root for NFS is unreliable */
34088 + h_path.mnt = au_br_mnt(br);
34089 + h_path.dentry = h_path.mnt->mnt_root;
34090 + err = vfs_statfs(&h_path, st);
34091 + if (unlikely(err)) {
34092 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
34096 + /* when the available size is equal, select the lower one */
34097 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
34098 + || sizeof(b) < sizeof(st->f_bsize));
34099 + b = st->f_bavail * st->f_bsize;
34100 + br->br_wbr->wbr_bytes = b;
34101 + if (b >= bavail) {
34103 + mfs->mfs_bindex = bindex;
34104 + mfs->mfs_jiffy = jiffies;
34108 + mfs->mfsrr_bytes = bavail;
34109 + AuDbg("b%d\n", mfs->mfs_bindex);
34113 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
34116 + struct dentry *parent;
34117 + struct super_block *sb;
34118 + struct au_wbr_mfs *mfs;
34120 + err = au_wbr_create_exp(dentry);
34124 + sb = dentry->d_sb;
34126 + if (au_ftest_wbr(flags, PARENT))
34127 + parent = dget_parent(dentry);
34128 + mfs = &au_sbi(sb)->si_wbr_mfs;
34129 + mutex_lock(&mfs->mfs_lock);
34130 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34131 + || mfs->mfs_bindex < 0
34132 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
34133 + au_mfs(dentry, parent);
34134 + mutex_unlock(&mfs->mfs_lock);
34135 + err = mfs->mfs_bindex;
34139 + err = au_wbr_nonopq(dentry, err);
34142 + AuDbg("b%d\n", err);
34146 +static int au_wbr_create_init_mfs(struct super_block *sb)
34148 + struct au_wbr_mfs *mfs;
34150 + mfs = &au_sbi(sb)->si_wbr_mfs;
34151 + mutex_init(&mfs->mfs_lock);
34152 + mfs->mfs_jiffy = 0;
34153 + mfs->mfs_bindex = -EROFS;
34158 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
34160 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
34164 +/* ---------------------------------------------------------------------- */
34166 +/* top down regardless parent, and then mfs */
34167 +static int au_wbr_create_tdmfs(struct dentry *dentry,
34168 + unsigned int flags __maybe_unused)
34171 + aufs_bindex_t bwh, btail, bindex, bfound, bmfs;
34172 + unsigned long long watermark;
34173 + struct super_block *sb;
34174 + struct au_wbr_mfs *mfs;
34175 + struct au_branch *br;
34176 + struct dentry *parent;
34178 + sb = dentry->d_sb;
34179 + mfs = &au_sbi(sb)->si_wbr_mfs;
34180 + mutex_lock(&mfs->mfs_lock);
34181 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
34182 + || mfs->mfs_bindex < 0)
34183 + au_mfs(dentry, /*parent*/NULL);
34184 + watermark = mfs->mfsrr_watermark;
34185 + bmfs = mfs->mfs_bindex;
34186 + mutex_unlock(&mfs->mfs_lock);
34188 + /* another style of au_wbr_create_exp() */
34189 + bwh = au_dbwh(dentry);
34190 + parent = dget_parent(dentry);
34191 + btail = au_dbtaildir(parent);
34192 + if (bwh >= 0 && bwh < btail)
34195 + err = au_wbr_nonopq(dentry, btail);
34196 + if (unlikely(err < 0))
34200 + for (bindex = 0; bindex <= btail; bindex++) {
34201 + br = au_sbr(sb, bindex);
34202 + if (au_br_rdonly(br))
34204 + if (br->br_wbr->wbr_bytes > watermark) {
34215 + AuDbg("b%d\n", err);
34219 +/* ---------------------------------------------------------------------- */
34221 +/* most free space and then round robin */
34222 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
34225 + struct au_wbr_mfs *mfs;
34227 + err = au_wbr_create_mfs(dentry, flags);
34229 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
34230 + mutex_lock(&mfs->mfs_lock);
34231 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
34232 + err = au_wbr_create_rr(dentry, flags);
34233 + mutex_unlock(&mfs->mfs_lock);
34236 + AuDbg("b%d\n", err);
34240 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
34244 + au_wbr_create_init_mfs(sb); /* ignore */
34245 + err = au_wbr_create_init_rr(sb);
34250 +/* ---------------------------------------------------------------------- */
34252 +/* top down parent and most free space */
34253 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
34256 + unsigned long long b;
34257 + aufs_bindex_t bindex, btop, bbot;
34258 + struct super_block *sb;
34259 + struct dentry *parent, *h_parent;
34260 + struct au_branch *br;
34262 + err = au_wbr_create_tdp(dentry, flags);
34263 + if (unlikely(err < 0))
34265 + parent = dget_parent(dentry);
34266 + btop = au_dbtop(parent);
34267 + bbot = au_dbtaildir(parent);
34268 + if (btop == bbot)
34269 + goto out_parent; /* success */
34271 + e2 = au_wbr_create_mfs(dentry, flags);
34273 + goto out_parent; /* success */
34275 + /* when the available size is equal, select upper one */
34276 + sb = dentry->d_sb;
34277 + br = au_sbr(sb, err);
34278 + b = br->br_wbr->wbr_bytes;
34279 + AuDbg("b%d, %llu\n", err, b);
34281 + for (bindex = btop; bindex <= bbot; bindex++) {
34282 + h_parent = au_h_dptr(parent, bindex);
34283 + if (!h_parent || d_is_negative(h_parent))
34286 + br = au_sbr(sb, bindex);
34287 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
34288 + b = br->br_wbr->wbr_bytes;
34290 + AuDbg("b%d, %llu\n", err, b);
34295 + err = au_wbr_nonopq(dentry, err);
34300 + AuDbg("b%d\n", err);
34304 +/* ---------------------------------------------------------------------- */
34307 + * - top down parent
34308 + * - most free space with parent
34309 + * - most free space round-robin regardless parent
34311 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
34314 + unsigned long long watermark;
34315 + struct super_block *sb;
34316 + struct au_branch *br;
34317 + struct au_wbr_mfs *mfs;
34319 + err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
34320 + if (unlikely(err < 0))
34323 + sb = dentry->d_sb;
34324 + br = au_sbr(sb, err);
34325 + mfs = &au_sbi(sb)->si_wbr_mfs;
34326 + mutex_lock(&mfs->mfs_lock);
34327 + watermark = mfs->mfsrr_watermark;
34328 + mutex_unlock(&mfs->mfs_lock);
34329 + if (br->br_wbr->wbr_bytes < watermark)
34330 + /* regardless the parent dir */
34331 + err = au_wbr_create_mfsrr(dentry, flags);
34334 + AuDbg("b%d\n", err);
34338 +/* ---------------------------------------------------------------------- */
34340 +/* policies for copyup */
34342 +/* top down parent */
34343 +static int au_wbr_copyup_tdp(struct dentry *dentry)
34345 + return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
34348 +/* bottom up parent */
34349 +static int au_wbr_copyup_bup(struct dentry *dentry)
34352 + aufs_bindex_t bindex, btop;
34353 + struct dentry *parent, *h_parent;
34354 + struct super_block *sb;
34357 + sb = dentry->d_sb;
34358 + parent = dget_parent(dentry);
34359 + btop = au_dbtop(parent);
34360 + for (bindex = au_dbtop(dentry); bindex >= btop; bindex--) {
34361 + h_parent = au_h_dptr(parent, bindex);
34362 + if (!h_parent || d_is_negative(h_parent))
34365 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
34372 + /* bottom up here */
34373 + if (unlikely(err < 0))
34374 + err = au_wbr_bu(sb, btop - 1);
34376 + AuDbg("b%d\n", err);
34381 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t btop)
34385 + err = au_wbr_bu(dentry->d_sb, btop);
34386 + AuDbg("b%d\n", err);
34388 + err = au_wbr_nonopq(dentry, err);
34390 + AuDbg("b%d\n", err);
34394 +static int au_wbr_copyup_bu(struct dentry *dentry)
34397 + aufs_bindex_t btop;
34399 + btop = au_dbtop(dentry);
34400 + err = au_wbr_do_copyup_bu(dentry, btop);
34404 +/* ---------------------------------------------------------------------- */
34406 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
34407 + [AuWbrCopyup_TDP] = {
34408 + .copyup = au_wbr_copyup_tdp
34410 + [AuWbrCopyup_BUP] = {
34411 + .copyup = au_wbr_copyup_bup
34413 + [AuWbrCopyup_BU] = {
34414 + .copyup = au_wbr_copyup_bu
34418 +struct au_wbr_create_operations au_wbr_create_ops[] = {
34419 + [AuWbrCreate_TDP] = {
34420 + .create = au_wbr_create_tdp
34422 + [AuWbrCreate_RR] = {
34423 + .create = au_wbr_create_rr,
34424 + .init = au_wbr_create_init_rr
34426 + [AuWbrCreate_MFS] = {
34427 + .create = au_wbr_create_mfs,
34428 + .init = au_wbr_create_init_mfs,
34429 + .fin = au_wbr_create_fin_mfs
34431 + [AuWbrCreate_MFSV] = {
34432 + .create = au_wbr_create_mfs,
34433 + .init = au_wbr_create_init_mfs,
34434 + .fin = au_wbr_create_fin_mfs
34436 + [AuWbrCreate_MFSRR] = {
34437 + .create = au_wbr_create_mfsrr,
34438 + .init = au_wbr_create_init_mfsrr,
34439 + .fin = au_wbr_create_fin_mfs
34441 + [AuWbrCreate_MFSRRV] = {
34442 + .create = au_wbr_create_mfsrr,
34443 + .init = au_wbr_create_init_mfsrr,
34444 + .fin = au_wbr_create_fin_mfs
34446 + [AuWbrCreate_TDMFS] = {
34447 + .create = au_wbr_create_tdmfs,
34448 + .init = au_wbr_create_init_mfs,
34449 + .fin = au_wbr_create_fin_mfs
34451 + [AuWbrCreate_TDMFSV] = {
34452 + .create = au_wbr_create_tdmfs,
34453 + .init = au_wbr_create_init_mfs,
34454 + .fin = au_wbr_create_fin_mfs
34456 + [AuWbrCreate_PMFS] = {
34457 + .create = au_wbr_create_pmfs,
34458 + .init = au_wbr_create_init_mfs,
34459 + .fin = au_wbr_create_fin_mfs
34461 + [AuWbrCreate_PMFSV] = {
34462 + .create = au_wbr_create_pmfs,
34463 + .init = au_wbr_create_init_mfs,
34464 + .fin = au_wbr_create_fin_mfs
34466 + [AuWbrCreate_PMFSRR] = {
34467 + .create = au_wbr_create_pmfsrr,
34468 + .init = au_wbr_create_init_mfsrr,
34469 + .fin = au_wbr_create_fin_mfs
34471 + [AuWbrCreate_PMFSRRV] = {
34472 + .create = au_wbr_create_pmfsrr,
34473 + .init = au_wbr_create_init_mfsrr,
34474 + .fin = au_wbr_create_fin_mfs
34477 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
34478 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
34479 +++ linux/fs/aufs/whout.c 2018-01-29 07:56:20.063325368 +0100
34482 + * Copyright (C) 2005-2017 Junjiro R. Okajima
34484 + * This program, aufs is free software; you can redistribute it and/or modify
34485 + * it under the terms of the GNU General Public License as published by
34486 + * the Free Software Foundation; either version 2 of the License, or
34487 + * (at your option) any later version.
34489 + * This program is distributed in the hope that it will be useful,
34490 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
34491 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
34492 + * GNU General Public License for more details.
34494 + * You should have received a copy of the GNU General Public License
34495 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
34499 + * whiteout for logical deletion and opaque directory
34504 +#define WH_MASK S_IRUGO
34507 + * If a directory contains this file, then it is opaque. We start with the
34508 + * .wh. flag so that it is blocked by lookup.
34510 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
34511 + sizeof(AUFS_WH_DIROPQ) - 1);
34514 + * generate whiteout name, which is NOT terminated by NULL.
34515 + * @name: original d_name.name
34516 + * @len: original d_name.len
34517 + * @wh: whiteout qstr
34518 + * returns zero when succeeds, otherwise error.
34519 + * succeeded value as wh->name should be freed by kfree().
34521 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
34525 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
34526 + return -ENAMETOOLONG;
34528 + wh->len = name->len + AUFS_WH_PFX_LEN;
34529 + p = kmalloc(wh->len, GFP_NOFS);
34532 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
34533 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
34540 +/* ---------------------------------------------------------------------- */
34543 + * test if the @wh_name exists under @h_parent.
34544 + * @try_sio specifies the necessary of super-io.
34546 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio)
34549 + struct dentry *wh_dentry;
34552 + wh_dentry = vfsub_lkup_one(wh_name, h_parent);
34554 + wh_dentry = au_sio_lkup_one(wh_name, h_parent);
34555 + err = PTR_ERR(wh_dentry);
34556 + if (IS_ERR(wh_dentry)) {
34557 + if (err == -ENAMETOOLONG)
34563 + if (d_is_negative(wh_dentry))
34564 + goto out_wh; /* success */
34567 + if (d_is_reg(wh_dentry))
34568 + goto out_wh; /* success */
34571 + AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
34572 + wh_dentry, d_inode(wh_dentry)->i_mode);
34581 + * test if the @h_dentry sets opaque or not.
34583 +int au_diropq_test(struct dentry *h_dentry)
34586 + struct inode *h_dir;
34588 + h_dir = d_inode(h_dentry);
34589 + err = au_wh_test(h_dentry, &diropq_name,
34590 + au_test_h_perm_sio(h_dir, MAY_EXEC));
34595 + * returns a negative dentry whose name is unique and temporary.
34597 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
34598 + struct qstr *prefix)
34600 + struct dentry *dentry;
34602 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
34604 + /* strict atomic_t is unnecessary here */
34605 + static unsigned short cnt;
34608 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
34611 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
34612 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
34613 + dentry = ERR_PTR(-ENAMETOOLONG);
34614 + if (unlikely(qs.len > NAME_MAX))
34616 + dentry = ERR_PTR(-ENOMEM);
34617 + name = kmalloc(qs.len + 1, GFP_NOFS);
34618 + if (unlikely(!name))
34622 + /* doubly whiteout-ed */
34623 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
34624 + p = name + AUFS_WH_PFX_LEN * 2;
34625 + memcpy(p, prefix->name, prefix->len);
34626 + p += prefix->len;
34628 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
34631 + for (i = 0; i < 3; i++) {
34632 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
34633 + dentry = au_sio_lkup_one(&qs, h_parent);
34634 + if (IS_ERR(dentry) || d_is_negative(dentry))
34638 + /* pr_warn("could not get random name\n"); */
34639 + dentry = ERR_PTR(-EEXIST);
34640 + AuDbg("%.*s\n", AuLNPair(&qs));
34644 + if (name != defname)
34647 + AuTraceErrPtr(dentry);
34652 + * rename the @h_dentry on @br to the whiteouted temporary name.
34654 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
34657 + struct path h_path = {
34658 + .mnt = au_br_mnt(br)
34660 + struct inode *h_dir, *delegated;
34661 + struct dentry *h_parent;
34663 + h_parent = h_dentry->d_parent; /* dir inode is locked */
34664 + h_dir = d_inode(h_parent);
34665 + IMustLock(h_dir);
34667 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
34668 + err = PTR_ERR(h_path.dentry);
34669 + if (IS_ERR(h_path.dentry))
34672 + /* under the same dir, no need to lock_rename() */
34673 + delegated = NULL;
34674 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated,
34677 + if (unlikely(err == -EWOULDBLOCK)) {
34678 + pr_warn("cannot retry for NFSv4 delegation"
34679 + " for an internal rename\n");
34682 + dput(h_path.dentry);
34689 +/* ---------------------------------------------------------------------- */
34691 + * functions for removing a whiteout
34694 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
34697 + struct inode *delegated;
34700 + * forces superio when the dir has a sticky bit.
34701 + * this may be a violation of unix fs semantics.
34703 + force = (h_dir->i_mode & S_ISVTX)
34704 + && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
34705 + delegated = NULL;
34706 + err = vfsub_unlink(h_dir, h_path, &delegated, force);
34707 + if (unlikely(err == -EWOULDBLOCK)) {
34708 + pr_warn("cannot retry for NFSv4 delegation"
34709 + " for an internal unlink\n");
34715 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
34716 + struct dentry *dentry)
34720 + err = do_unlink_wh(h_dir, h_path);
34721 + if (!err && dentry)
34722 + au_set_dbwh(dentry, -1);
34727 +static int unlink_wh_name(struct dentry *h_parent, struct qstr *wh,
34728 + struct au_branch *br)
34731 + struct path h_path = {
34732 + .mnt = au_br_mnt(br)
34736 + h_path.dentry = vfsub_lkup_one(wh, h_parent);
34737 + if (IS_ERR(h_path.dentry))
34738 + err = PTR_ERR(h_path.dentry);
34740 + if (d_is_reg(h_path.dentry))
34741 + err = do_unlink_wh(d_inode(h_parent), &h_path);
34742 + dput(h_path.dentry);
34748 +/* ---------------------------------------------------------------------- */
34750 + * initialize/clean whiteout for a branch
34753 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
34757 + struct inode *delegated;
34759 + if (d_is_negative(whpath->dentry))
34763 + err = vfsub_rmdir(h_dir, whpath);
34765 + delegated = NULL;
34766 + err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
34767 + if (unlikely(err == -EWOULDBLOCK)) {
34768 + pr_warn("cannot retry for NFSv4 delegation"
34769 + " for an internal unlink\n");
34773 + if (unlikely(err))
34774 + pr_warn("failed removing %pd (%d), ignored.\n",
34775 + whpath->dentry, err);
34778 +static int test_linkable(struct dentry *h_root)
34780 + struct inode *h_dir = d_inode(h_root);
34782 + if (h_dir->i_op->link)
34785 + pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
34786 + h_root, au_sbtype(h_root->d_sb));
34790 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
34791 +static int au_whdir(struct inode *h_dir, struct path *path)
34796 + if (d_is_negative(path->dentry)) {
34797 + int mode = S_IRWXU;
34799 + if (au_test_nfs(path->dentry->d_sb))
34801 + err = vfsub_mkdir(h_dir, path, mode);
34802 + } else if (d_is_dir(path->dentry))
34805 + pr_err("unknown %pd exists\n", path->dentry);
34810 +struct au_wh_base {
34811 + const struct qstr *name;
34812 + struct dentry *dentry;
34815 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
34816 + struct path *h_path)
34818 + h_path->dentry = base[AuBrWh_BASE].dentry;
34819 + au_wh_clean(h_dir, h_path, /*isdir*/0);
34820 + h_path->dentry = base[AuBrWh_PLINK].dentry;
34821 + au_wh_clean(h_dir, h_path, /*isdir*/1);
34822 + h_path->dentry = base[AuBrWh_ORPH].dentry;
34823 + au_wh_clean(h_dir, h_path, /*isdir*/1);
34827 + * returns tri-state,
34828 + * minus: error, caller should print the message
34830 + * plus: error, caller should NOT print the message
34832 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
34833 + int do_plink, struct au_wh_base base[],
34834 + struct path *h_path)
34837 + struct inode *h_dir;
34839 + h_dir = d_inode(h_root);
34840 + h_path->dentry = base[AuBrWh_BASE].dentry;
34841 + au_wh_clean(h_dir, h_path, /*isdir*/0);
34842 + h_path->dentry = base[AuBrWh_PLINK].dentry;
34844 + err = test_linkable(h_root);
34845 + if (unlikely(err)) {
34850 + err = au_whdir(h_dir, h_path);
34851 + if (unlikely(err))
34853 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
34855 + au_wh_clean(h_dir, h_path, /*isdir*/1);
34856 + h_path->dentry = base[AuBrWh_ORPH].dentry;
34857 + err = au_whdir(h_dir, h_path);
34858 + if (unlikely(err))
34860 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
34867 + * for the moment, aufs supports the branch filesystem which does not support
34868 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
34869 + * copyup failed. finally, such filesystem will not be used as the writable
34872 + * returns tri-state, see above.
34874 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
34875 + int do_plink, struct au_wh_base base[],
34876 + struct path *h_path)
34879 + struct inode *h_dir;
34881 + WbrWhMustWriteLock(wbr);
34883 + err = test_linkable(h_root);
34884 + if (unlikely(err)) {
34890 + * todo: should this create be done in /sbin/mount.aufs helper?
34893 + h_dir = d_inode(h_root);
34894 + if (d_is_negative(base[AuBrWh_BASE].dentry)) {
34895 + h_path->dentry = base[AuBrWh_BASE].dentry;
34896 + err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
34897 + } else if (d_is_reg(base[AuBrWh_BASE].dentry))
34900 + pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
34901 + if (unlikely(err))
34904 + h_path->dentry = base[AuBrWh_PLINK].dentry;
34906 + err = au_whdir(h_dir, h_path);
34907 + if (unlikely(err))
34909 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
34911 + au_wh_clean(h_dir, h_path, /*isdir*/1);
34912 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
34914 + h_path->dentry = base[AuBrWh_ORPH].dentry;
34915 + err = au_whdir(h_dir, h_path);
34916 + if (unlikely(err))
34918 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
34925 + * initialize the whiteout base file/dir for @br.
34927 +int au_wh_init(struct au_branch *br, struct super_block *sb)
34930 + const unsigned char do_plink
34931 + = !!au_opt_test(au_mntflags(sb), PLINK);
34932 + struct inode *h_dir;
34933 + struct path path = br->br_path;
34934 + struct dentry *h_root = path.dentry;
34935 + struct au_wbr *wbr = br->br_wbr;
34936 + static const struct qstr base_name[] = {
34937 + [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
34938 + sizeof(AUFS_BASE_NAME) - 1),
34939 + [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
34940 + sizeof(AUFS_PLINKDIR_NAME) - 1),
34941 + [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
34942 + sizeof(AUFS_ORPHDIR_NAME) - 1)
34944 + struct au_wh_base base[] = {
34945 + [AuBrWh_BASE] = {
34946 + .name = base_name + AuBrWh_BASE,
34949 + [AuBrWh_PLINK] = {
34950 + .name = base_name + AuBrWh_PLINK,
34953 + [AuBrWh_ORPH] = {
34954 + .name = base_name + AuBrWh_ORPH,
34960 + WbrWhMustWriteLock(wbr);
34962 + for (i = 0; i < AuBrWh_Last; i++) {
34963 + /* doubly whiteouted */
34964 + struct dentry *d;
34966 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
34967 + err = PTR_ERR(d);
34971 + base[i].dentry = d;
34973 + && wbr->wbr_wh[i]
34974 + && wbr->wbr_wh[i] != base[i].dentry);
34978 + for (i = 0; i < AuBrWh_Last; i++) {
34979 + dput(wbr->wbr_wh[i]);
34980 + wbr->wbr_wh[i] = NULL;
34984 + if (!au_br_writable(br->br_perm)) {
34985 + h_dir = d_inode(h_root);
34986 + au_wh_init_ro(h_dir, base, &path);
34987 + } else if (!au_br_wh_linkable(br->br_perm)) {
34988 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
34994 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
35000 + goto out; /* success */
35003 + pr_err("an error(%d) on the writable branch %pd(%s)\n",
35004 + err, h_root, au_sbtype(h_root->d_sb));
35006 + for (i = 0; i < AuBrWh_Last; i++)
35007 + dput(base[i].dentry);
35011 +/* ---------------------------------------------------------------------- */
35013 + * whiteouts are all hard-linked usually.
35014 + * when its link count reaches a ceiling, we create a new whiteout base
35015 + * asynchronously.
35018 +struct reinit_br_wh {
35019 + struct super_block *sb;
35020 + struct au_branch *br;
35023 +static void reinit_br_wh(void *arg)
35026 + aufs_bindex_t bindex;
35027 + struct path h_path;
35028 + struct reinit_br_wh *a = arg;
35029 + struct au_wbr *wbr;
35030 + struct inode *dir, *delegated;
35031 + struct dentry *h_root;
35032 + struct au_hinode *hdir;
35035 + wbr = a->br->br_wbr;
35036 + /* big aufs lock */
35037 + si_noflush_write_lock(a->sb);
35038 + if (!au_br_writable(a->br->br_perm))
35040 + bindex = au_br_index(a->sb, a->br->br_id);
35041 + if (unlikely(bindex < 0))
35044 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
35045 + dir = d_inode(a->sb->s_root);
35046 + hdir = au_hi(dir, bindex);
35047 + h_root = au_h_dptr(a->sb->s_root, bindex);
35048 + AuDebugOn(h_root != au_br_dentry(a->br));
35050 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35051 + wbr_wh_write_lock(wbr);
35052 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
35055 + h_path.dentry = wbr->wbr_whbase;
35056 + h_path.mnt = au_br_mnt(a->br);
35057 + delegated = NULL;
35058 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
35060 + if (unlikely(err == -EWOULDBLOCK)) {
35061 + pr_warn("cannot retry for NFSv4 delegation"
35062 + " for an internal unlink\n");
35066 + pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
35069 + dput(wbr->wbr_whbase);
35070 + wbr->wbr_whbase = NULL;
35072 + err = au_wh_init(a->br, a->sb);
35073 + wbr_wh_write_unlock(wbr);
35074 + au_hn_inode_unlock(hdir);
35075 + di_read_unlock(a->sb->s_root, AuLock_IR);
35077 + au_fhsm_wrote(a->sb, bindex, /*force*/0);
35081 + atomic_dec(&wbr->wbr_wh_running);
35082 + au_br_put(a->br);
35083 + si_write_unlock(a->sb);
35084 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
35086 + if (unlikely(err))
35087 + AuIOErr("err %d\n", err);
35090 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
35092 + int do_dec, wkq_err;
35093 + struct reinit_br_wh *arg;
35096 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
35099 + /* ignore ENOMEM */
35100 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
35103 + * dec(wh_running), kfree(arg) and dec(br_count)
35104 + * in reinit function
35109 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
35110 + if (unlikely(wkq_err)) {
35111 + atomic_dec(&br->br_wbr->wbr_wh_running);
35120 + atomic_dec(&br->br_wbr->wbr_wh_running);
35123 +/* ---------------------------------------------------------------------- */
35126 + * create the whiteout @wh.
35128 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
35129 + struct dentry *wh)
35132 + struct path h_path = {
35135 + struct au_branch *br;
35136 + struct au_wbr *wbr;
35137 + struct dentry *h_parent;
35138 + struct inode *h_dir, *delegated;
35140 + h_parent = wh->d_parent; /* dir inode is locked */
35141 + h_dir = d_inode(h_parent);
35142 + IMustLock(h_dir);
35144 + br = au_sbr(sb, bindex);
35145 + h_path.mnt = au_br_mnt(br);
35146 + wbr = br->br_wbr;
35147 + wbr_wh_read_lock(wbr);
35148 + if (wbr->wbr_whbase) {
35149 + delegated = NULL;
35150 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
35151 + if (unlikely(err == -EWOULDBLOCK)) {
35152 + pr_warn("cannot retry for NFSv4 delegation"
35153 + " for an internal link\n");
35156 + if (!err || err != -EMLINK)
35159 + /* link count full. re-initialize br_whbase. */
35160 + kick_reinit_br_wh(sb, br);
35163 + /* return this error in this context */
35164 + err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
35166 + au_fhsm_wrote(sb, bindex, /*force*/0);
35169 + wbr_wh_read_unlock(wbr);
35173 +/* ---------------------------------------------------------------------- */
35176 + * create or remove the diropq.
35178 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
35179 + unsigned int flags)
35181 + struct dentry *opq_dentry, *h_dentry;
35182 + struct super_block *sb;
35183 + struct au_branch *br;
35186 + sb = dentry->d_sb;
35187 + br = au_sbr(sb, bindex);
35188 + h_dentry = au_h_dptr(dentry, bindex);
35189 + opq_dentry = vfsub_lkup_one(&diropq_name, h_dentry);
35190 + if (IS_ERR(opq_dentry))
35193 + if (au_ftest_diropq(flags, CREATE)) {
35194 + err = link_or_create_wh(sb, bindex, opq_dentry);
35196 + au_set_dbdiropq(dentry, bindex);
35197 + goto out; /* success */
35200 + struct path tmp = {
35201 + .dentry = opq_dentry,
35202 + .mnt = au_br_mnt(br)
35204 + err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &tmp);
35206 + au_set_dbdiropq(dentry, -1);
35208 + dput(opq_dentry);
35209 + opq_dentry = ERR_PTR(err);
35212 + return opq_dentry;
35215 +struct do_diropq_args {
35216 + struct dentry **errp;
35217 + struct dentry *dentry;
35218 + aufs_bindex_t bindex;
35219 + unsigned int flags;
35222 +static void call_do_diropq(void *args)
35224 + struct do_diropq_args *a = args;
35225 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
35228 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35229 + unsigned int flags)
35231 + struct dentry *diropq, *h_dentry;
35233 + h_dentry = au_h_dptr(dentry, bindex);
35234 + if (!au_test_h_perm_sio(d_inode(h_dentry), MAY_EXEC | MAY_WRITE))
35235 + diropq = do_diropq(dentry, bindex, flags);
35238 + struct do_diropq_args args = {
35240 + .dentry = dentry,
35241 + .bindex = bindex,
35245 + wkq_err = au_wkq_wait(call_do_diropq, &args);
35246 + if (unlikely(wkq_err))
35247 + diropq = ERR_PTR(wkq_err);
35253 +/* ---------------------------------------------------------------------- */
35256 + * lookup whiteout dentry.
35257 + * @h_parent: lower parent dentry which must exist and be locked
35258 + * @base_name: name of dentry which will be whiteouted
35259 + * returns dentry for whiteout.
35261 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35262 + struct au_branch *br)
35265 + struct qstr wh_name;
35266 + struct dentry *wh_dentry;
35268 + err = au_wh_name_alloc(&wh_name, base_name);
35269 + wh_dentry = ERR_PTR(err);
35271 + wh_dentry = vfsub_lkup_one(&wh_name, h_parent);
35272 + kfree(wh_name.name);
35274 + return wh_dentry;
35278 + * link/create a whiteout for @dentry on @bindex.
35280 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35281 + struct dentry *h_parent)
35283 + struct dentry *wh_dentry;
35284 + struct super_block *sb;
35287 + sb = dentry->d_sb;
35288 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
35289 + if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
35290 + err = link_or_create_wh(sb, bindex, wh_dentry);
35292 + au_set_dbwh(dentry, bindex);
35293 + au_fhsm_wrote(sb, bindex, /*force*/0);
35296 + wh_dentry = ERR_PTR(err);
35300 + return wh_dentry;
35303 +/* ---------------------------------------------------------------------- */
35305 +/* Delete all whiteouts in this directory on branch bindex. */
35306 +static int del_wh_children(struct dentry *h_dentry, struct au_nhash *whlist,
35307 + aufs_bindex_t bindex, struct au_branch *br)
35310 + unsigned long ul, n;
35311 + struct qstr wh_name;
35313 + struct hlist_head *head;
35314 + struct au_vdir_wh *pos;
35315 + struct au_vdir_destr *str;
35318 + p = (void *)__get_free_page(GFP_NOFS);
35319 + wh_name.name = p;
35320 + if (unlikely(!wh_name.name))
35324 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
35325 + p += AUFS_WH_PFX_LEN;
35326 + n = whlist->nh_num;
35327 + head = whlist->nh_head;
35328 + for (ul = 0; !err && ul < n; ul++, head++) {
35329 + hlist_for_each_entry(pos, head, wh_hash) {
35330 + if (pos->wh_bindex != bindex)
35333 + str = &pos->wh_str;
35334 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
35335 + memcpy(p, str->name, str->len);
35336 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
35337 + err = unlink_wh_name(h_dentry, &wh_name, br);
35342 + AuIOErr("whiteout name too long %.*s\n",
35343 + str->len, str->name);
35348 + free_page((unsigned long)wh_name.name);
35354 +struct del_wh_children_args {
35356 + struct dentry *h_dentry;
35357 + struct au_nhash *whlist;
35358 + aufs_bindex_t bindex;
35359 + struct au_branch *br;
35362 +static void call_del_wh_children(void *args)
35364 + struct del_wh_children_args *a = args;
35365 + *a->errp = del_wh_children(a->h_dentry, a->whlist, a->bindex, a->br);
35368 +/* ---------------------------------------------------------------------- */
35370 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
35372 + struct au_whtmp_rmdir *whtmp;
35374 + unsigned int rdhash;
35376 + SiMustAnyLock(sb);
35378 + whtmp = kzalloc(sizeof(*whtmp), gfp);
35379 + if (unlikely(!whtmp)) {
35380 + whtmp = ERR_PTR(-ENOMEM);
35384 + /* no estimation for dir size */
35385 + rdhash = au_sbi(sb)->si_rdhash;
35387 + rdhash = AUFS_RDHASH_DEF;
35388 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
35389 + if (unlikely(err)) {
35391 + whtmp = ERR_PTR(err);
35398 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
35401 + au_br_put(whtmp->br);
35402 + dput(whtmp->wh_dentry);
35403 + iput(whtmp->dir);
35404 + au_nhash_wh_free(&whtmp->whlist);
35409 + * rmdir the whiteouted temporary named dir @h_dentry.
35410 + * @whlist: whiteouted children.
35412 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
35413 + struct dentry *wh_dentry, struct au_nhash *whlist)
35416 + unsigned int h_nlink;
35417 + struct path h_tmp;
35418 + struct inode *wh_inode, *h_dir;
35419 + struct au_branch *br;
35421 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
35422 + IMustLock(h_dir);
35424 + br = au_sbr(dir->i_sb, bindex);
35425 + wh_inode = d_inode(wh_dentry);
35426 + inode_lock_nested(wh_inode, AuLsc_I_CHILD);
35429 + * someone else might change some whiteouts while we were sleeping.
35430 + * it means this whlist may have an obsoleted entry.
35432 + if (!au_test_h_perm_sio(wh_inode, MAY_EXEC | MAY_WRITE))
35433 + err = del_wh_children(wh_dentry, whlist, bindex, br);
35436 + struct del_wh_children_args args = {
35438 + .h_dentry = wh_dentry,
35439 + .whlist = whlist,
35440 + .bindex = bindex,
35444 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
35445 + if (unlikely(wkq_err))
35448 + inode_unlock(wh_inode);
35451 + h_tmp.dentry = wh_dentry;
35452 + h_tmp.mnt = au_br_mnt(br);
35453 + h_nlink = h_dir->i_nlink;
35454 + err = vfsub_rmdir(h_dir, &h_tmp);
35455 + /* some fs doesn't change the parent nlink in some cases */
35456 + h_nlink -= h_dir->i_nlink;
35460 + if (au_ibtop(dir) == bindex) {
35461 + /* todo: dir->i_mutex is necessary */
35462 + au_cpup_attr_timesizes(dir);
35464 + vfsub_drop_nlink(dir);
35466 + return 0; /* success */
35469 + pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
35473 +static void call_rmdir_whtmp(void *args)
35476 + aufs_bindex_t bindex;
35477 + struct au_whtmp_rmdir *a = args;
35478 + struct super_block *sb;
35479 + struct dentry *h_parent;
35480 + struct inode *h_dir;
35481 + struct au_hinode *hdir;
35483 + /* rmdir by nfsd may cause deadlock with this i_mutex */
35484 + /* inode_lock(a->dir); */
35486 + sb = a->dir->i_sb;
35487 + si_read_lock(sb, !AuLock_FLUSH);
35488 + if (!au_br_writable(a->br->br_perm))
35490 + bindex = au_br_index(sb, a->br->br_id);
35491 + if (unlikely(bindex < 0))
35495 + ii_write_lock_parent(a->dir);
35496 + h_parent = dget_parent(a->wh_dentry);
35497 + h_dir = d_inode(h_parent);
35498 + hdir = au_hi(a->dir, bindex);
35499 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
35500 + if (unlikely(err))
35502 + au_hn_inode_lock_nested(hdir, AuLsc_I_PARENT);
35503 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
35506 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
35507 + au_hn_inode_unlock(hdir);
35508 + vfsub_mnt_drop_write(au_br_mnt(a->br));
35512 + ii_write_unlock(a->dir);
35514 + /* inode_unlock(a->dir); */
35515 + au_whtmp_rmdir_free(a);
35516 + si_read_unlock(sb);
35517 + au_nwt_done(&au_sbi(sb)->si_nowait);
35518 + if (unlikely(err))
35519 + AuIOErr("err %d\n", err);
35522 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
35523 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
35526 + struct super_block *sb;
35530 + /* all post-process will be done in do_rmdir_whtmp(). */
35532 + args->dir = au_igrab(dir);
35533 + args->br = au_sbr(sb, bindex);
35534 + au_br_get(args->br);
35535 + args->wh_dentry = dget(wh_dentry);
35536 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
35537 + if (unlikely(wkq_err)) {
35538 + pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
35539 + au_whtmp_rmdir_free(args);
35542 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
35543 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
35544 +++ linux/fs/aufs/whout.h 2018-01-29 07:56:20.063325368 +0100
35547 + * Copyright (C) 2005-2017 Junjiro R. Okajima
35549 + * This program, aufs is free software; you can redistribute it and/or modify
35550 + * it under the terms of the GNU General Public License as published by
35551 + * the Free Software Foundation; either version 2 of the License, or
35552 + * (at your option) any later version.
35554 + * This program is distributed in the hope that it will be useful,
35555 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35556 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35557 + * GNU General Public License for more details.
35559 + * You should have received a copy of the GNU General Public License
35560 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35564 + * whiteout for logical deletion and opaque directory
35567 +#ifndef __AUFS_WHOUT_H__
35568 +#define __AUFS_WHOUT_H__
35575 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
35576 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio);
35577 +int au_diropq_test(struct dentry *h_dentry);
35579 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
35580 + struct qstr *prefix);
35581 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
35582 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
35583 + struct dentry *dentry);
35584 +int au_wh_init(struct au_branch *br, struct super_block *sb);
35586 +/* diropq flags */
35587 +#define AuDiropq_CREATE 1
35588 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
35589 +#define au_fset_diropq(flags, name) \
35590 + do { (flags) |= AuDiropq_##name; } while (0)
35591 +#define au_fclr_diropq(flags, name) \
35592 + do { (flags) &= ~AuDiropq_##name; } while (0)
35594 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
35595 + unsigned int flags);
35596 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
35597 + struct au_branch *br);
35598 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
35599 + struct dentry *h_parent);
35601 +/* real rmdir for the whiteout-ed dir */
35602 +struct au_whtmp_rmdir {
35603 + struct inode *dir;
35604 + struct au_branch *br;
35605 + struct dentry *wh_dentry;
35606 + struct au_nhash whlist;
35609 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
35610 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
35611 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
35612 + struct dentry *wh_dentry, struct au_nhash *whlist);
35613 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
35614 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
35616 +/* ---------------------------------------------------------------------- */
35618 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
35619 + aufs_bindex_t bindex)
35621 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
35624 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
35626 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
35629 +#endif /* __KERNEL__ */
35630 +#endif /* __AUFS_WHOUT_H__ */
35631 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
35632 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
35633 +++ linux/fs/aufs/wkq.c 2018-01-29 07:56:20.063325368 +0100
35636 + * Copyright (C) 2005-2017 Junjiro R. Okajima
35638 + * This program, aufs is free software; you can redistribute it and/or modify
35639 + * it under the terms of the GNU General Public License as published by
35640 + * the Free Software Foundation; either version 2 of the License, or
35641 + * (at your option) any later version.
35643 + * This program is distributed in the hope that it will be useful,
35644 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35645 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35646 + * GNU General Public License for more details.
35648 + * You should have received a copy of the GNU General Public License
35649 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35653 + * workqueue for asynchronous/super-io operations
35654 + * todo: try new dredential scheme
35657 +#include <linux/module.h>
35660 +/* internal workqueue named AUFS_WKQ_NAME */
35662 +static struct workqueue_struct *au_wkq;
35664 +struct au_wkinfo {
35665 + struct work_struct wk;
35666 + struct kobject *kobj;
35668 + unsigned int flags; /* see wkq.h */
35670 + au_wkq_func_t func;
35673 +#ifdef CONFIG_LOCKDEP
35675 + struct held_lock **hlock;
35678 + struct completion *comp;
35681 +/* ---------------------------------------------------------------------- */
35683 + * Aufs passes some operations to the workqueue such as the internal copyup.
35684 + * This scheme looks rather unnatural for LOCKDEP debugging feature, since the
35685 + * job run by workqueue depends upon the locks acquired in the other task.
35686 + * Delegating a small operation to the workqueue, aufs passes its lockdep
35687 + * information too. And the job in the workqueue restores the info in order to
35688 + * pretend as if it acquired those locks. This is just to make LOCKDEP work
35689 + * correctly and expectedly.
35692 +#ifndef CONFIG_LOCKDEP
35693 +AuStubInt0(au_wkq_lockdep_alloc, struct au_wkinfo *wkinfo);
35694 +AuStubVoid(au_wkq_lockdep_free, struct au_wkinfo *wkinfo);
35695 +AuStubVoid(au_wkq_lockdep_pre, struct au_wkinfo *wkinfo);
35696 +AuStubVoid(au_wkq_lockdep_post, struct au_wkinfo *wkinfo);
35697 +AuStubVoid(au_wkq_lockdep_init, struct au_wkinfo *wkinfo);
35699 +static void au_wkq_lockdep_init(struct au_wkinfo *wkinfo)
35701 + wkinfo->hlock = NULL;
35702 + wkinfo->dont_check = 0;
35709 +static int au_wkq_lockdep_test(struct lock_class_key *key, const char *name)
35711 + static DEFINE_SPINLOCK(spin);
35714 + struct lock_class_key *key;
35716 + { .name = "&sbinfo->si_rwsem" },
35717 + { .name = "&finfo->fi_rwsem" },
35718 + { .name = "&dinfo->di_rwsem" },
35719 + { .name = "&iinfo->ii_rwsem" }
35724 + /* lockless read from 'set.' see below */
35725 + if (set == ARRAY_SIZE(a)) {
35726 + for (i = 0; i < ARRAY_SIZE(a); i++)
35727 + if (a[i].key == key)
35732 + spin_lock(&spin);
35734 + for (i = 0; i < ARRAY_SIZE(a); i++)
35735 + if (a[i].key == key) {
35736 + spin_unlock(&spin);
35739 + for (i = 0; i < ARRAY_SIZE(a); i++) {
35741 + if (unlikely(a[i].key == key)) { /* rare but possible */
35742 + spin_unlock(&spin);
35747 + if (strstr(a[i].name, name)) {
35749 + * the order of these three lines is important for the
35750 + * lockless read above.
35753 + spin_unlock(&spin);
35755 + /* AuDbg("%d, %s\n", set, name); */
35759 + spin_unlock(&spin);
35768 +static int au_wkq_lockdep_alloc(struct au_wkinfo *wkinfo)
35771 + struct task_struct *curr;
35772 + struct held_lock **hl, *held_locks, *p;
35776 + wkinfo->dont_check = lockdep_recursing(curr);
35777 + if (wkinfo->dont_check)
35779 + n = curr->lockdep_depth;
35784 + wkinfo->hlock = kmalloc_array(n + 1, sizeof(*wkinfo->hlock), GFP_NOFS);
35785 + if (unlikely(!wkinfo->hlock))
35790 + if (0 && au_debug_test()) /* left for debugging */
35791 + lockdep_print_held_locks(curr);
35793 + held_locks = curr->held_locks;
35794 + hl = wkinfo->hlock;
35796 + p = held_locks++;
35797 + if (au_wkq_lockdep_test(p->instance->key, p->instance->name))
35806 +static void au_wkq_lockdep_free(struct au_wkinfo *wkinfo)
35808 + kfree(wkinfo->hlock);
35811 +static void au_wkq_lockdep_pre(struct au_wkinfo *wkinfo)
35813 + struct held_lock *p, **hl = wkinfo->hlock;
35816 + if (wkinfo->dont_check)
35820 + while ((p = *hl++)) { /* assignment */
35821 + subclass = lockdep_hlock_class(p)->subclass;
35822 + /* AuDbg("%s, %d\n", p->instance->name, subclass); */
35824 + rwsem_acquire_read(p->instance, subclass, 0,
35825 + /*p->acquire_ip*/_RET_IP_);
35827 + rwsem_acquire(p->instance, subclass, 0,
35828 + /*p->acquire_ip*/_RET_IP_);
35832 +static void au_wkq_lockdep_post(struct au_wkinfo *wkinfo)
35834 + struct held_lock *p, **hl = wkinfo->hlock;
35836 + if (wkinfo->dont_check)
35840 + while ((p = *hl++)) /* assignment */
35841 + rwsem_release(p->instance, 0, /*p->acquire_ip*/_RET_IP_);
35845 +static void wkq_func(struct work_struct *wk)
35847 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
35849 + AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
35850 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
35852 + au_wkq_lockdep_pre(wkinfo);
35853 + wkinfo->func(wkinfo->args);
35854 + au_wkq_lockdep_post(wkinfo);
35855 + if (au_ftest_wkq(wkinfo->flags, WAIT))
35856 + complete(wkinfo->comp);
35858 + kobject_put(wkinfo->kobj);
35859 + module_put(THIS_MODULE); /* todo: ?? */
35865 + * Since struct completion is large, try allocating it dynamically.
35867 +#if 1 /* defined(CONFIG_4KSTACKS) || defined(AuTest4KSTACKS) */
35868 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
35870 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
35872 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
35874 + init_completion(*comp);
35875 + wkinfo->comp = *comp;
35881 +static void au_wkq_comp_free(struct completion *comp)
35889 +#define AuWkqCompDeclare(name) \
35890 + DECLARE_COMPLETION_ONSTACK(_ ## name); \
35891 + struct completion *comp = &_ ## name
35893 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
35895 + wkinfo->comp = *comp;
35899 +static void au_wkq_comp_free(struct completion *comp __maybe_unused)
35903 +#endif /* 4KSTACKS */
35905 +static void au_wkq_run(struct au_wkinfo *wkinfo)
35907 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
35908 + if (au_wkq_test()) {
35909 + AuWarn1("wkq from wkq, unless silly-rename on NFS,"
35910 + " due to a dead dir by UDBA?\n");
35911 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
35914 + au_dbg_verify_kthread();
35916 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
35917 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
35918 + queue_work(au_wkq, &wkinfo->wk);
35920 + INIT_WORK(&wkinfo->wk, wkq_func);
35921 + schedule_work(&wkinfo->wk);
35926 + * Be careful. It is easy to make deadlock happen.
35927 + * processA: lock, wkq and wait
35928 + * processB: wkq and wait, lock in wkq
35931 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
35934 + AuWkqCompDeclare(comp);
35935 + struct au_wkinfo wkinfo = {
35941 + err = au_wkq_comp_alloc(&wkinfo, &comp);
35942 + if (unlikely(err))
35944 + err = au_wkq_lockdep_alloc(&wkinfo);
35945 + if (unlikely(err))
35948 + au_wkq_run(&wkinfo);
35949 + /* no timeout, no interrupt */
35950 + wait_for_completion(wkinfo.comp);
35952 + au_wkq_lockdep_free(&wkinfo);
35955 + au_wkq_comp_free(comp);
35957 + destroy_work_on_stack(&wkinfo.wk);
35962 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
35963 + * problem in a concurrent umounting.
35965 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
35966 + unsigned int flags)
35969 + struct au_wkinfo *wkinfo;
35971 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
35974 + * wkq_func() must free this wkinfo.
35975 + * it highly depends upon the implementation of workqueue.
35978 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
35980 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
35981 + wkinfo->flags = flags & ~AuWkq_WAIT;
35982 + wkinfo->func = func;
35983 + wkinfo->args = args;
35984 + wkinfo->comp = NULL;
35985 + au_wkq_lockdep_init(wkinfo);
35986 + kobject_get(wkinfo->kobj);
35987 + __module_get(THIS_MODULE); /* todo: ?? */
35989 + au_wkq_run(wkinfo);
35992 + au_nwt_done(&au_sbi(sb)->si_nowait);
35998 +/* ---------------------------------------------------------------------- */
36000 +void au_nwt_init(struct au_nowait_tasks *nwt)
36002 + atomic_set(&nwt->nw_len, 0);
36003 + /* smp_mb(); */ /* atomic_set */
36004 + init_waitqueue_head(&nwt->nw_wq);
36007 +void au_wkq_fin(void)
36009 + destroy_workqueue(au_wkq);
36012 +int __init au_wkq_init(void)
36017 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
36018 + if (IS_ERR(au_wkq))
36019 + err = PTR_ERR(au_wkq);
36020 + else if (!au_wkq)
36025 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
36026 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
36027 +++ linux/fs/aufs/wkq.h 2018-01-29 07:56:20.063325368 +0100
36030 + * Copyright (C) 2005-2017 Junjiro R. Okajima
36032 + * This program, aufs is free software; you can redistribute it and/or modify
36033 + * it under the terms of the GNU General Public License as published by
36034 + * the Free Software Foundation; either version 2 of the License, or
36035 + * (at your option) any later version.
36037 + * This program is distributed in the hope that it will be useful,
36038 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36039 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36040 + * GNU General Public License for more details.
36042 + * You should have received a copy of the GNU General Public License
36043 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36047 + * workqueue for asynchronous/super-io operations
36048 + * todo: try new credentials management scheme
36051 +#ifndef __AUFS_WKQ_H__
36052 +#define __AUFS_WKQ_H__
36056 +#include <linux/wait.h>
36058 +struct super_block;
36060 +/* ---------------------------------------------------------------------- */
36063 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
36065 +struct au_nowait_tasks {
36067 + wait_queue_head_t nw_wq;
36070 +/* ---------------------------------------------------------------------- */
36072 +typedef void (*au_wkq_func_t)(void *args);
36075 +#define AuWkq_WAIT 1
36076 +#define AuWkq_NEST (1 << 1)
36077 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
36078 +#define au_fset_wkq(flags, name) \
36079 + do { (flags) |= AuWkq_##name; } while (0)
36080 +#define au_fclr_wkq(flags, name) \
36081 + do { (flags) &= ~AuWkq_##name; } while (0)
36083 +#ifndef CONFIG_AUFS_HNOTIFY
36085 +#define AuWkq_NEST 0
36089 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
36090 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
36091 + unsigned int flags);
36092 +void au_nwt_init(struct au_nowait_tasks *nwt);
36093 +int __init au_wkq_init(void);
36094 +void au_wkq_fin(void);
36096 +/* ---------------------------------------------------------------------- */
36098 +static inline int au_wkq_test(void)
36100 + return current->flags & PF_WQ_WORKER;
36103 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
36105 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
36108 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
36110 + if (atomic_dec_and_test(&nwt->nw_len))
36111 + wake_up_all(&nwt->nw_wq);
36114 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
36116 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
36120 +#endif /* __KERNEL__ */
36121 +#endif /* __AUFS_WKQ_H__ */
36122 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
36123 --- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
36124 +++ linux/fs/aufs/xattr.c 2018-01-29 07:56:20.063325368 +0100
36127 + * Copyright (C) 2014-2017 Junjiro R. Okajima
36129 + * This program, aufs is free software; you can redistribute it and/or modify
36130 + * it under the terms of the GNU General Public License as published by
36131 + * the Free Software Foundation; either version 2 of the License, or
36132 + * (at your option) any later version.
36134 + * This program is distributed in the hope that it will be useful,
36135 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36136 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36137 + * GNU General Public License for more details.
36139 + * You should have received a copy of the GNU General Public License
36140 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36144 + * handling xattr functions
36147 +#include <linux/fs.h>
36148 +#include <linux/posix_acl_xattr.h>
36149 +#include <linux/xattr.h>
36152 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
36154 + if (!ignore_flags)
36162 + if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
36167 +#define cmp(brattr, prefix) do { \
36168 + if (!strncmp(name, XATTR_##prefix##_PREFIX, \
36169 + XATTR_##prefix##_PREFIX_LEN)) { \
36170 + if (ignore_flags & AuBrAttr_ICEX_##brattr) \
36176 + cmp(SEC, SECURITY);
36177 + cmp(SYS, SYSTEM);
36178 + cmp(TR, TRUSTED);
36182 + if (ignore_flags & AuBrAttr_ICEX_OTH)
36189 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
36191 +static int au_do_cpup_xattr(struct dentry *h_dst, struct dentry *h_src,
36192 + char *name, char **buf, unsigned int ignore_flags,
36193 + unsigned int verbose)
36197 + struct inode *h_idst;
36199 + ssz = vfs_getxattr_alloc(h_src, name, buf, 0, GFP_NOFS);
36201 + if (unlikely(err <= 0)) {
36202 + if (err == -ENODATA
36203 + || (err == -EOPNOTSUPP
36204 + && ((ignore_flags & au_xattr_out_of_list)
36205 + || (au_test_nfs_noacl(d_inode(h_src))
36206 + && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
36208 + XATTR_NAME_POSIX_ACL_DEFAULT))))
36211 + if (err && (verbose || au_debug_test()))
36212 + pr_err("%s, err %d\n", name, err);
36216 + /* unlock it temporary */
36217 + h_idst = d_inode(h_dst);
36218 + inode_unlock(h_idst);
36219 + err = vfsub_setxattr(h_dst, name, *buf, ssz, /*flags*/0);
36220 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36221 + if (unlikely(err)) {
36222 + if (verbose || au_debug_test())
36223 + pr_err("%s, err %d\n", name, err);
36224 + err = au_xattr_ignore(err, name, ignore_flags);
36231 +int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
36232 + unsigned int verbose)
36234 + int err, unlocked, acl_access, acl_default;
36236 + struct inode *h_isrc, *h_idst;
36237 + char *value, *p, *o, *e;
36239 + /* try stopping to update the source inode while we are referencing */
36240 + /* there should not be the parent-child relationship between them */
36241 + h_isrc = d_inode(h_src);
36242 + h_idst = d_inode(h_dst);
36243 + inode_unlock(h_idst);
36244 + vfsub_inode_lock_shared_nested(h_isrc, AuLsc_I_CHILD);
36245 + inode_lock_nested(h_idst, AuLsc_I_CHILD2);
36248 + /* some filesystems don't list POSIX ACL, for example tmpfs */
36249 + ssz = vfs_listxattr(h_src, NULL, 0);
36251 + if (unlikely(err < 0)) {
36253 + if (err == -ENODATA
36254 + || err == -EOPNOTSUPP)
36255 + err = 0; /* ignore */
36264 + p = kmalloc(ssz, GFP_NOFS);
36266 + if (unlikely(!p))
36268 + err = vfs_listxattr(h_src, p, ssz);
36270 + inode_unlock_shared(h_isrc);
36272 + AuDbg("err %d, ssz %zd\n", err, ssz);
36273 + if (unlikely(err < 0))
36281 + while (!err && p < e) {
36282 + acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
36283 + sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
36284 + acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
36285 + sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
36287 + err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
36289 + p += strlen(p) + 1;
36292 + ignore_flags |= au_xattr_out_of_list;
36293 + if (!err && !acl_access) {
36294 + err = au_do_cpup_xattr(h_dst, h_src,
36295 + XATTR_NAME_POSIX_ACL_ACCESS, &value,
36296 + ignore_flags, verbose);
36299 + if (!err && !acl_default) {
36300 + err = au_do_cpup_xattr(h_dst, h_src,
36301 + XATTR_NAME_POSIX_ACL_DEFAULT, &value,
36302 + ignore_flags, verbose);
36312 + inode_unlock_shared(h_isrc);
36317 +/* ---------------------------------------------------------------------- */
36319 +static int au_smack_reentering(struct super_block *sb)
36321 +#if IS_ENABLED(CONFIG_SECURITY_SMACK)
36323 + * as a part of lookup, smack_d_instantiate() is called, and it calls
36324 + * i_op->getxattr(). ouch.
36326 + return si_pid_test(sb);
36337 +struct au_lgxattr {
36345 + const char *name;
36352 +static ssize_t au_lgxattr(struct dentry *dentry, struct au_lgxattr *arg)
36356 + struct path h_path;
36357 + struct super_block *sb;
36359 + sb = dentry->d_sb;
36360 + reenter = au_smack_reentering(sb);
36362 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
36363 + if (unlikely(err))
36366 + err = au_h_path_getattr(dentry, /*force*/1, &h_path, reenter);
36367 + if (unlikely(err))
36369 + if (unlikely(!h_path.dentry))
36370 + /* illegally overlapped or something */
36371 + goto out_di; /* pretending success */
36373 + /* always topmost entry only */
36374 + switch (arg->type) {
36375 + case AU_XATTR_LIST:
36376 + err = vfs_listxattr(h_path.dentry,
36377 + arg->u.list.list, arg->u.list.size);
36379 + case AU_XATTR_GET:
36380 + AuDebugOn(d_is_negative(h_path.dentry));
36381 + err = vfs_getxattr(h_path.dentry,
36382 + arg->u.get.name, arg->u.get.value,
36383 + arg->u.get.size);
36389 + di_read_unlock(dentry, AuLock_IR);
36392 + si_read_unlock(sb);
36398 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
36400 + struct au_lgxattr arg = {
36401 + .type = AU_XATTR_LIST,
36408 + return au_lgxattr(dentry, &arg);
36411 +static ssize_t au_getxattr(struct dentry *dentry,
36412 + struct inode *inode __maybe_unused,
36413 + const char *name, void *value, size_t size)
36415 + struct au_lgxattr arg = {
36416 + .type = AU_XATTR_GET,
36424 + return au_lgxattr(dentry, &arg);
36427 +static int au_setxattr(struct dentry *dentry, struct inode *inode,
36428 + const char *name, const void *value, size_t size,
36431 + struct au_sxattr arg = {
36432 + .type = AU_XATTR_SET,
36441 + return au_sxattr(dentry, inode, &arg);
36444 +/* ---------------------------------------------------------------------- */
36446 +static int au_xattr_get(const struct xattr_handler *handler,
36447 + struct dentry *dentry, struct inode *inode,
36448 + const char *name, void *buffer, size_t size)
36450 + return au_getxattr(dentry, inode, name, buffer, size);
36453 +static int au_xattr_set(const struct xattr_handler *handler,
36454 + struct dentry *dentry, struct inode *inode,
36455 + const char *name, const void *value, size_t size,
36458 + return au_setxattr(dentry, inode, name, value, size, flags);
36461 +static const struct xattr_handler au_xattr_handler = {
36464 + .get = au_xattr_get,
36465 + .set = au_xattr_set
36468 +static const struct xattr_handler *au_xattr_handlers[] = {
36469 +#ifdef CONFIG_FS_POSIX_ACL
36470 + &posix_acl_access_xattr_handler,
36471 + &posix_acl_default_xattr_handler,
36473 + &au_xattr_handler, /* must be last */
36477 +void au_xattr_init(struct super_block *sb)
36479 + sb->s_xattr = au_xattr_handlers;
36481 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
36482 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
36483 +++ linux/fs/aufs/xino.c 2018-01-29 07:56:20.063325368 +0100
36486 + * Copyright (C) 2005-2017 Junjiro R. Okajima
36488 + * This program, aufs is free software; you can redistribute it and/or modify
36489 + * it under the terms of the GNU General Public License as published by
36490 + * the Free Software Foundation; either version 2 of the License, or
36491 + * (at your option) any later version.
36493 + * This program is distributed in the hope that it will be useful,
36494 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
36495 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36496 + * GNU General Public License for more details.
36498 + * You should have received a copy of the GNU General Public License
36499 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
36503 + * external inode number translation table and bitmap
36506 +#include <linux/seq_file.h>
36507 +#include <linux/statfs.h>
36510 +/* todo: unnecessary to support mmap_sem since kernel-space? */
36511 +ssize_t xino_fread(vfs_readf_t func, struct file *file, void *kbuf, size_t size,
36515 + mm_segment_t oldfs;
36522 + oldfs = get_fs();
36523 + set_fs(KERNEL_DS);
36525 + /* todo: signal_pending? */
36526 + err = func(file, buf.u, size, pos);
36527 + } while (err == -EAGAIN || err == -EINTR);
36530 +#if 0 /* reserved for future use */
36532 + fsnotify_access(file->f_path.dentry);
36538 +/* ---------------------------------------------------------------------- */
36540 +static ssize_t xino_fwrite_wkq(vfs_writef_t func, struct file *file, void *buf,
36541 + size_t size, loff_t *pos);
36543 +static ssize_t do_xino_fwrite(vfs_writef_t func, struct file *file, void *kbuf,
36544 + size_t size, loff_t *pos)
36547 + mm_segment_t oldfs;
36550 + const char __user *u;
36553 + const int prevent_endless = 10;
36557 + oldfs = get_fs();
36558 + set_fs(KERNEL_DS);
36560 + err = func(file, buf.u, size, pos);
36561 + if (err == -EINTR
36562 + && !au_wkq_test()
36563 + && fatal_signal_pending(current)) {
36565 + err = xino_fwrite_wkq(func, file, kbuf, size, pos);
36566 + BUG_ON(err == -EINTR);
36567 + oldfs = get_fs();
36568 + set_fs(KERNEL_DS);
36570 + } while (i++ < prevent_endless
36571 + && (err == -EAGAIN || err == -EINTR));
36574 +#if 0 /* reserved for future use */
36576 + fsnotify_modify(file->f_path.dentry);
36582 +struct do_xino_fwrite_args {
36584 + vfs_writef_t func;
36585 + struct file *file;
36591 +static void call_do_xino_fwrite(void *args)
36593 + struct do_xino_fwrite_args *a = args;
36594 + *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos);
36597 +static ssize_t xino_fwrite_wkq(vfs_writef_t func, struct file *file, void *buf,
36598 + size_t size, loff_t *pos)
36602 + struct do_xino_fwrite_args args = {
36612 + * it breaks RLIMIT_FSIZE and normal user's limit,
36613 + * users should care about quota and real 'filesystem full.'
36615 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
36616 + if (unlikely(wkq_err))
36622 +ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
36623 + size_t size, loff_t *pos)
36627 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
36629 + err = do_xino_fwrite(func, file, buf, size, pos);
36633 + err = xino_fwrite_wkq(func, file, buf, size, pos);
36640 +/* ---------------------------------------------------------------------- */
36643 + * create a new xinofile at the same place/path as @base_file.
36645 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src)
36647 + struct file *file;
36648 + struct dentry *base, *parent;
36649 + struct inode *dir, *delegated;
36650 + struct qstr *name;
36651 + struct path path;
36654 + base = base_file->f_path.dentry;
36655 + parent = base->d_parent; /* dir inode is locked */
36656 + dir = d_inode(parent);
36659 + file = ERR_PTR(-EINVAL);
36660 + name = &base->d_name;
36661 + path.dentry = vfsub_lookup_one_len(name->name, parent, name->len);
36662 + if (IS_ERR(path.dentry)) {
36663 + file = (void *)path.dentry;
36664 + pr_err("%pd lookup err %ld\n",
36665 + base, PTR_ERR(path.dentry));
36669 + /* no need to mnt_want_write() since we call dentry_open() later */
36670 + err = vfs_create(dir, path.dentry, S_IRUGO | S_IWUGO, NULL);
36671 + if (unlikely(err)) {
36672 + file = ERR_PTR(err);
36673 + pr_err("%pd create err %d\n", base, err);
36677 + path.mnt = base_file->f_path.mnt;
36678 + file = vfsub_dentry_open(&path,
36679 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
36680 + /* | __FMODE_NONOTIFY */);
36681 + if (IS_ERR(file)) {
36682 + pr_err("%pd open err %ld\n", base, PTR_ERR(file));
36686 + delegated = NULL;
36687 + err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
36688 + if (unlikely(err == -EWOULDBLOCK)) {
36689 + pr_warn("cannot retry for NFSv4 delegation"
36690 + " for an internal unlink\n");
36693 + if (unlikely(err)) {
36694 + pr_err("%pd unlink err %d\n", base, err);
36699 + /* no one can touch copy_src xino */
36700 + err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
36701 + if (unlikely(err)) {
36702 + pr_err("%pd copy err %d\n", base, err);
36706 + goto out_dput; /* success */
36710 + file = ERR_PTR(err);
36712 + dput(path.dentry);
36717 +struct au_xino_lock_dir {
36718 + struct au_hinode *hdir;
36719 + struct dentry *parent;
36720 + struct inode *dir;
36723 +static void au_xino_lock_dir(struct super_block *sb, struct file *xino,
36724 + struct au_xino_lock_dir *ldir)
36726 + aufs_bindex_t brid, bindex;
36728 + ldir->hdir = NULL;
36730 + brid = au_xino_brid(sb);
36732 + bindex = au_br_index(sb, brid);
36733 + if (bindex >= 0) {
36734 + ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
36735 + au_hn_inode_lock_nested(ldir->hdir, AuLsc_I_PARENT);
36737 + ldir->parent = dget_parent(xino->f_path.dentry);
36738 + ldir->dir = d_inode(ldir->parent);
36739 + inode_lock_nested(ldir->dir, AuLsc_I_PARENT);
36743 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
36746 + au_hn_inode_unlock(ldir->hdir);
36748 + inode_unlock(ldir->dir);
36749 + dput(ldir->parent);
36753 +/* ---------------------------------------------------------------------- */
36755 +/* trucate xino files asynchronously */
36757 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex)
36760 + unsigned long jiffy;
36762 + aufs_bindex_t bi, bbot;
36763 + struct kstatfs *st;
36764 + struct au_branch *br;
36765 + struct file *new_xino, *file;
36766 + struct super_block *h_sb;
36767 + struct au_xino_lock_dir ldir;
36770 + st = kmalloc(sizeof(*st), GFP_NOFS);
36771 + if (unlikely(!st))
36775 + bbot = au_sbbot(sb);
36776 + if (unlikely(bindex < 0 || bbot < bindex))
36778 + br = au_sbr(sb, bindex);
36779 + file = br->br_xino.xi_file;
36783 + err = vfs_statfs(&file->f_path, st);
36784 + if (unlikely(err))
36785 + AuErr1("statfs err %d, ignored\n", err);
36787 + blocks = file_inode(file)->i_blocks;
36788 + pr_info("begin truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
36789 + bindex, (u64)blocks, st->f_bfree, st->f_blocks);
36791 + au_xino_lock_dir(sb, file, &ldir);
36792 + /* mnt_want_write() is unnecessary here */
36793 + new_xino = au_xino_create2(file, file);
36794 + au_xino_unlock_dir(&ldir);
36795 + err = PTR_ERR(new_xino);
36796 + if (IS_ERR(new_xino)) {
36797 + pr_err("err %d, ignored\n", err);
36802 + br->br_xino.xi_file = new_xino;
36804 + h_sb = au_br_sb(br);
36805 + for (bi = 0; bi <= bbot; bi++) {
36806 + if (unlikely(bi == bindex))
36808 + br = au_sbr(sb, bi);
36809 + if (au_br_sb(br) != h_sb)
36812 + fput(br->br_xino.xi_file);
36813 + br->br_xino.xi_file = new_xino;
36814 + get_file(new_xino);
36817 + err = vfs_statfs(&new_xino->f_path, st);
36819 + pr_info("end truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
36820 + bindex, (u64)file_inode(new_xino)->i_blocks,
36821 + st->f_bfree, st->f_blocks);
36822 + if (file_inode(new_xino)->i_blocks < blocks)
36823 + au_sbi(sb)->si_xino_jiffy = jiffy;
36825 + AuErr1("statfs err %d, ignored\n", err);
36833 +struct xino_do_trunc_args {
36834 + struct super_block *sb;
36835 + struct au_branch *br;
36838 +static void xino_do_trunc(void *_args)
36840 + struct xino_do_trunc_args *args = _args;
36841 + struct super_block *sb;
36842 + struct au_branch *br;
36843 + struct inode *dir;
36845 + aufs_bindex_t bindex;
36849 + dir = d_inode(sb->s_root);
36852 + si_noflush_write_lock(sb);
36853 + ii_read_lock_parent(dir);
36854 + bindex = au_br_index(sb, br->br_id);
36855 + err = au_xino_trunc(sb, bindex);
36856 + ii_read_unlock(dir);
36857 + if (unlikely(err))
36858 + pr_warn("err b%d, (%d)\n", bindex, err);
36859 + atomic_dec(&br->br_xino_running);
36861 + si_write_unlock(sb);
36862 + au_nwt_done(&au_sbi(sb)->si_nowait);
36866 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
36869 + struct kstatfs st;
36870 + struct au_sbinfo *sbinfo;
36872 + /* todo: si_xino_expire and the ratio should be customizable */
36873 + sbinfo = au_sbi(sb);
36874 + if (time_before(jiffies,
36875 + sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
36878 + /* truncation border */
36879 + err = vfs_statfs(&br->br_xino.xi_file->f_path, &st);
36880 + if (unlikely(err)) {
36881 + AuErr1("statfs err %d, ignored\n", err);
36884 + if (div64_u64(st.f_bfree * 100, st.f_blocks) >= AUFS_XINO_DEF_TRUNC)
36890 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
36892 + struct xino_do_trunc_args *args;
36895 + if (!xino_trunc_test(sb, br))
36898 + if (atomic_inc_return(&br->br_xino_running) > 1)
36901 + /* lock and kfree() will be called in trunc_xino() */
36902 + args = kmalloc(sizeof(*args), GFP_NOFS);
36903 + if (unlikely(!args)) {
36904 + AuErr1("no memory\n");
36911 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
36913 + return; /* success */
36915 + pr_err("wkq %d\n", wkq_err);
36920 + atomic_dec(&br->br_xino_running);
36923 +/* ---------------------------------------------------------------------- */
36925 +static int au_xino_do_write(vfs_writef_t write, struct file *file,
36926 + ino_t h_ino, ino_t ino)
36932 + if (unlikely(au_loff_max / sizeof(ino) - 1 < pos)) {
36933 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
36936 + pos *= sizeof(ino);
36937 + sz = xino_fwrite(write, file, &ino, sizeof(ino), &pos);
36938 + if (sz == sizeof(ino))
36939 + return 0; /* success */
36941 + AuIOErr("write failed (%zd)\n", sz);
36946 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
36947 + * at the position of @h_ino.
36948 + * even if @ino is zero, it is written to the xinofile and means no entry.
36949 + * if the size of the xino file on a specific filesystem exceeds the watermark,
36950 + * try truncating it.
36952 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
36956 + unsigned int mnt_flags;
36957 + struct au_branch *br;
36959 + BUILD_BUG_ON(sizeof(long long) != sizeof(au_loff_max)
36960 + || ((loff_t)-1) > 0);
36961 + SiMustAnyLock(sb);
36963 + mnt_flags = au_mntflags(sb);
36964 + if (!au_opt_test(mnt_flags, XINO))
36967 + br = au_sbr(sb, bindex);
36968 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
36971 + if (au_opt_test(mnt_flags, TRUNC_XINO)
36972 + && au_test_fs_trunc_xino(au_br_sb(br)))
36973 + xino_try_trunc(sb, br);
36974 + return 0; /* success */
36977 + AuIOErr("write failed (%d)\n", err);
36981 +/* ---------------------------------------------------------------------- */
36983 +/* aufs inode number bitmap */
36985 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
36986 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
36990 + AuDebugOn(bit < 0 || page_bits <= bit);
36991 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
36995 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
36997 + AuDebugOn(ino < AUFS_FIRST_INO);
36998 + ino -= AUFS_FIRST_INO;
36999 + *pindex = ino / page_bits;
37000 + *bit = ino % page_bits;
37003 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
37008 + struct au_sbinfo *sbinfo;
37009 + struct file *xib;
37010 + unsigned long *p;
37012 + sbinfo = au_sbi(sb);
37013 + MtxMustLock(&sbinfo->si_xib_mtx);
37014 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
37015 + || !au_opt_test(sbinfo->si_mntflags, XINO));
37017 + if (pindex == sbinfo->si_xib_last_pindex)
37020 + xib = sbinfo->si_xib;
37021 + p = sbinfo->si_xib_buf;
37022 + pos = sbinfo->si_xib_last_pindex;
37023 + pos *= PAGE_SIZE;
37024 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
37025 + if (unlikely(sz != PAGE_SIZE))
37029 + pos *= PAGE_SIZE;
37030 + if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
37031 + sz = xino_fread(sbinfo->si_xread, xib, p, PAGE_SIZE, &pos);
37033 + memset(p, 0, PAGE_SIZE);
37034 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
37036 + if (sz == PAGE_SIZE) {
37037 + sbinfo->si_xib_last_pindex = pindex;
37038 + return 0; /* success */
37042 + AuIOErr1("write failed (%zd)\n", sz);
37049 +/* ---------------------------------------------------------------------- */
37051 +static void au_xib_clear_bit(struct inode *inode)
37054 + unsigned long pindex;
37055 + struct super_block *sb;
37056 + struct au_sbinfo *sbinfo;
37058 + AuDebugOn(inode->i_nlink);
37060 + sb = inode->i_sb;
37061 + xib_calc_bit(inode->i_ino, &pindex, &bit);
37062 + AuDebugOn(page_bits <= bit);
37063 + sbinfo = au_sbi(sb);
37064 + mutex_lock(&sbinfo->si_xib_mtx);
37065 + err = xib_pindex(sb, pindex);
37067 + clear_bit(bit, sbinfo->si_xib_buf);
37068 + sbinfo->si_xib_next_bit = bit;
37070 + mutex_unlock(&sbinfo->si_xib_mtx);
37073 +/* for s_op->delete_inode() */
37074 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
37077 + unsigned int mnt_flags;
37078 + aufs_bindex_t bindex, bbot, bi;
37079 + unsigned char try_trunc;
37080 + struct au_iinfo *iinfo;
37081 + struct super_block *sb;
37082 + struct au_hinode *hi;
37083 + struct inode *h_inode;
37084 + struct au_branch *br;
37085 + vfs_writef_t xwrite;
37087 + AuDebugOn(au_is_bad_inode(inode));
37089 + sb = inode->i_sb;
37090 + mnt_flags = au_mntflags(sb);
37091 + if (!au_opt_test(mnt_flags, XINO)
37092 + || inode->i_ino == AUFS_ROOT_INO)
37096 + au_xigen_inc(inode);
37097 + au_xib_clear_bit(inode);
37100 + iinfo = au_ii(inode);
37101 + bindex = iinfo->ii_btop;
37105 + xwrite = au_sbi(sb)->si_xwrite;
37106 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
37107 + hi = au_hinode(iinfo, bindex);
37108 + bbot = iinfo->ii_bbot;
37109 + for (; bindex <= bbot; bindex++, hi++) {
37110 + h_inode = hi->hi_inode;
37112 + || (!unlinked && h_inode->i_nlink))
37115 + /* inode may not be revalidated */
37116 + bi = au_br_index(sb, hi->hi_id);
37120 + br = au_sbr(sb, bi);
37121 + err = au_xino_do_write(xwrite, br->br_xino.xi_file,
37122 + h_inode->i_ino, /*ino*/0);
37123 + if (!err && try_trunc
37124 + && au_test_fs_trunc_xino(au_br_sb(br)))
37125 + xino_try_trunc(sb, br);
37129 +/* get an unused inode number from bitmap */
37130 +ino_t au_xino_new_ino(struct super_block *sb)
37133 + unsigned long *p, pindex, ul, pend;
37134 + struct au_sbinfo *sbinfo;
37135 + struct file *file;
37136 + int free_bit, err;
37138 + if (!au_opt_test(au_mntflags(sb), XINO))
37139 + return iunique(sb, AUFS_FIRST_INO);
37141 + sbinfo = au_sbi(sb);
37142 + mutex_lock(&sbinfo->si_xib_mtx);
37143 + p = sbinfo->si_xib_buf;
37144 + free_bit = sbinfo->si_xib_next_bit;
37145 + if (free_bit < page_bits && !test_bit(free_bit, p))
37146 + goto out; /* success */
37147 + free_bit = find_first_zero_bit(p, page_bits);
37148 + if (free_bit < page_bits)
37149 + goto out; /* success */
37151 + pindex = sbinfo->si_xib_last_pindex;
37152 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
37153 + err = xib_pindex(sb, ul);
37154 + if (unlikely(err))
37156 + free_bit = find_first_zero_bit(p, page_bits);
37157 + if (free_bit < page_bits)
37158 + goto out; /* success */
37161 + file = sbinfo->si_xib;
37162 + pend = vfsub_f_size_read(file) / PAGE_SIZE;
37163 + for (ul = pindex + 1; ul <= pend; ul++) {
37164 + err = xib_pindex(sb, ul);
37165 + if (unlikely(err))
37167 + free_bit = find_first_zero_bit(p, page_bits);
37168 + if (free_bit < page_bits)
37169 + goto out; /* success */
37174 + set_bit(free_bit, p);
37175 + sbinfo->si_xib_next_bit = free_bit + 1;
37176 + pindex = sbinfo->si_xib_last_pindex;
37177 + mutex_unlock(&sbinfo->si_xib_mtx);
37178 + ino = xib_calc_ino(pindex, free_bit);
37179 + AuDbg("i%lu\n", (unsigned long)ino);
37182 + mutex_unlock(&sbinfo->si_xib_mtx);
37188 + * read @ino from xinofile for the specified branch{@sb, @bindex}
37189 + * at the position of @h_ino.
37190 + * if @ino does not exist and @do_new is true, get new one.
37192 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37198 + struct file *file;
37199 + struct au_sbinfo *sbinfo;
37202 + if (!au_opt_test(au_mntflags(sb), XINO))
37203 + return 0; /* no xino */
37206 + sbinfo = au_sbi(sb);
37208 + if (unlikely(au_loff_max / sizeof(*ino) - 1 < pos)) {
37209 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
37212 + pos *= sizeof(*ino);
37214 + file = au_sbr(sb, bindex)->br_xino.xi_file;
37215 + if (vfsub_f_size_read(file) < pos + sizeof(*ino))
37216 + return 0; /* no ino */
37218 + sz = xino_fread(sbinfo->si_xread, file, ino, sizeof(*ino), &pos);
37219 + if (sz == sizeof(*ino))
37220 + return 0; /* success */
37223 + if (unlikely(sz >= 0)) {
37225 + AuIOErr("xino read error (%zd)\n", sz);
37231 +/* ---------------------------------------------------------------------- */
37233 +/* create and set a new xino file */
37235 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent)
37237 + struct file *file;
37238 + struct dentry *h_parent, *d;
37239 + struct inode *h_dir, *inode;
37243 + * at mount-time, and the xino file is the default path,
37244 + * hnotify is disabled so we have no notify events to ignore.
37245 + * when a user specified the xino, we cannot get au_hdir to be ignored.
37247 + file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
37248 + /* | __FMODE_NONOTIFY */,
37249 + S_IRUGO | S_IWUGO);
37250 + if (IS_ERR(file)) {
37252 + pr_err("open %s(%ld)\n", fname, PTR_ERR(file));
37256 + /* keep file count */
37258 + inode = file_inode(file);
37259 + h_parent = dget_parent(file->f_path.dentry);
37260 + h_dir = d_inode(h_parent);
37261 + inode_lock_nested(h_dir, AuLsc_I_PARENT);
37262 + /* mnt_want_write() is unnecessary here */
37263 + /* no delegation since it is just created */
37264 + if (inode->i_nlink)
37265 + err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
37267 + inode_unlock(h_dir);
37269 + if (unlikely(err)) {
37271 + pr_err("unlink %s(%d)\n", fname, err);
37276 + d = file->f_path.dentry;
37277 + if (unlikely(sb == d->d_sb)) {
37279 + pr_err("%s must be outside\n", fname);
37282 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
37284 + pr_err("xino doesn't support %s(%s)\n",
37285 + fname, au_sbtype(d->d_sb));
37288 + return file; /* success */
37292 + file = ERR_PTR(err);
37297 + * find another branch who is on the same filesystem of the specified
37298 + * branch{@btgt}. search until @bbot.
37300 +static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
37301 + aufs_bindex_t bbot)
37303 + aufs_bindex_t bindex;
37304 + struct super_block *tgt_sb = au_sbr_sb(sb, btgt);
37306 + for (bindex = 0; bindex < btgt; bindex++)
37307 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
37309 + for (bindex++; bindex <= bbot; bindex++)
37310 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
37315 +/* ---------------------------------------------------------------------- */
37318 + * initialize the xinofile for the specified branch @br
37319 + * at the place/path where @base_file indicates.
37320 + * test whether another branch is on the same filesystem or not,
37321 + * if @do_test is true.
37323 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
37324 + struct file *base_file, int do_test)
37328 + aufs_bindex_t bbot, bindex;
37329 + struct au_branch *shared_br, *b;
37330 + struct file *file;
37331 + struct super_block *tgt_sb;
37333 + shared_br = NULL;
37334 + bbot = au_sbbot(sb);
37336 + tgt_sb = au_br_sb(br);
37337 + for (bindex = 0; bindex <= bbot; bindex++) {
37338 + b = au_sbr(sb, bindex);
37339 + if (tgt_sb == au_br_sb(b)) {
37346 + if (!shared_br || !shared_br->br_xino.xi_file) {
37347 + struct au_xino_lock_dir ldir;
37349 + au_xino_lock_dir(sb, base_file, &ldir);
37350 + /* mnt_want_write() is unnecessary here */
37351 + file = au_xino_create2(base_file, NULL);
37352 + au_xino_unlock_dir(&ldir);
37353 + err = PTR_ERR(file);
37354 + if (IS_ERR(file))
37356 + br->br_xino.xi_file = file;
37358 + br->br_xino.xi_file = shared_br->br_xino.xi_file;
37359 + get_file(br->br_xino.xi_file);
37362 + ino = AUFS_ROOT_INO;
37363 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
37365 + if (unlikely(err)) {
37366 + fput(br->br_xino.xi_file);
37367 + br->br_xino.xi_file = NULL;
37374 +/* ---------------------------------------------------------------------- */
37376 +/* trucate a xino bitmap file */
37379 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
37383 + unsigned long pindex;
37384 + loff_t pos, pend;
37385 + struct au_sbinfo *sbinfo;
37386 + vfs_readf_t func;
37388 + unsigned long *p;
37391 + sbinfo = au_sbi(sb);
37392 + MtxMustLock(&sbinfo->si_xib_mtx);
37393 + p = sbinfo->si_xib_buf;
37394 + func = sbinfo->si_xread;
37395 + pend = vfsub_f_size_read(file);
37397 + while (pos < pend) {
37398 + sz = xino_fread(func, file, page, PAGE_SIZE, &pos);
37400 + if (unlikely(sz <= 0))
37404 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
37405 + if (unlikely(*ino < AUFS_FIRST_INO))
37408 + xib_calc_bit(*ino, &pindex, &bit);
37409 + AuDebugOn(page_bits <= bit);
37410 + err = xib_pindex(sb, pindex);
37422 +static int xib_restore(struct super_block *sb)
37425 + aufs_bindex_t bindex, bbot;
37429 + page = (void *)__get_free_page(GFP_NOFS);
37430 + if (unlikely(!page))
37434 + bbot = au_sbbot(sb);
37435 + for (bindex = 0; !err && bindex <= bbot; bindex++)
37436 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0)
37437 + err = do_xib_restore
37438 + (sb, au_sbr(sb, bindex)->br_xino.xi_file, page);
37440 + AuDbg("b%d\n", bindex);
37441 + free_page((unsigned long)page);
37447 +int au_xib_trunc(struct super_block *sb)
37452 + struct au_xino_lock_dir ldir;
37453 + struct au_sbinfo *sbinfo;
37454 + unsigned long *p;
37455 + struct file *file;
37457 + SiMustWriteLock(sb);
37460 + sbinfo = au_sbi(sb);
37461 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
37464 + file = sbinfo->si_xib;
37465 + if (vfsub_f_size_read(file) <= PAGE_SIZE)
37468 + au_xino_lock_dir(sb, file, &ldir);
37469 + /* mnt_want_write() is unnecessary here */
37470 + file = au_xino_create2(sbinfo->si_xib, NULL);
37471 + au_xino_unlock_dir(&ldir);
37472 + err = PTR_ERR(file);
37473 + if (IS_ERR(file))
37475 + fput(sbinfo->si_xib);
37476 + sbinfo->si_xib = file;
37478 + p = sbinfo->si_xib_buf;
37479 + memset(p, 0, PAGE_SIZE);
37481 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xib, p, PAGE_SIZE, &pos);
37482 + if (unlikely(sz != PAGE_SIZE)) {
37484 + AuIOErr("err %d\n", err);
37490 + mutex_lock(&sbinfo->si_xib_mtx);
37491 + /* mnt_want_write() is unnecessary here */
37492 + err = xib_restore(sb);
37493 + mutex_unlock(&sbinfo->si_xib_mtx);
37499 +/* ---------------------------------------------------------------------- */
37502 + * xino mount option handlers
37506 +static void xino_clear_xib(struct super_block *sb)
37508 + struct au_sbinfo *sbinfo;
37510 + SiMustWriteLock(sb);
37512 + sbinfo = au_sbi(sb);
37513 + sbinfo->si_xread = NULL;
37514 + sbinfo->si_xwrite = NULL;
37515 + if (sbinfo->si_xib)
37516 + fput(sbinfo->si_xib);
37517 + sbinfo->si_xib = NULL;
37518 + if (sbinfo->si_xib_buf)
37519 + free_page((unsigned long)sbinfo->si_xib_buf);
37520 + sbinfo->si_xib_buf = NULL;
37523 +static int au_xino_set_xib(struct super_block *sb, struct file *base)
37527 + struct au_sbinfo *sbinfo;
37528 + struct file *file;
37530 + SiMustWriteLock(sb);
37532 + sbinfo = au_sbi(sb);
37533 + file = au_xino_create2(base, sbinfo->si_xib);
37534 + err = PTR_ERR(file);
37535 + if (IS_ERR(file))
37537 + if (sbinfo->si_xib)
37538 + fput(sbinfo->si_xib);
37539 + sbinfo->si_xib = file;
37540 + sbinfo->si_xread = vfs_readf(file);
37541 + sbinfo->si_xwrite = vfs_writef(file);
37544 + if (!sbinfo->si_xib_buf)
37545 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
37546 + if (unlikely(!sbinfo->si_xib_buf))
37549 + sbinfo->si_xib_last_pindex = 0;
37550 + sbinfo->si_xib_next_bit = 0;
37551 + if (vfsub_f_size_read(file) < PAGE_SIZE) {
37553 + err = xino_fwrite(sbinfo->si_xwrite, file, sbinfo->si_xib_buf,
37554 + PAGE_SIZE, &pos);
37555 + if (unlikely(err != PAGE_SIZE))
37559 + goto out; /* success */
37562 + if (sbinfo->si_xib_buf)
37563 + free_page((unsigned long)sbinfo->si_xib_buf);
37564 + sbinfo->si_xib_buf = NULL;
37568 + fput(sbinfo->si_xib);
37569 + sbinfo->si_xib = NULL;
37570 + sbinfo->si_xread = NULL;
37571 + sbinfo->si_xwrite = NULL;
37576 +/* xino for each branch */
37577 +static void xino_clear_br(struct super_block *sb)
37579 + aufs_bindex_t bindex, bbot;
37580 + struct au_branch *br;
37582 + bbot = au_sbbot(sb);
37583 + for (bindex = 0; bindex <= bbot; bindex++) {
37584 + br = au_sbr(sb, bindex);
37585 + if (!br || !br->br_xino.xi_file)
37588 + fput(br->br_xino.xi_file);
37589 + br->br_xino.xi_file = NULL;
37593 +static int au_xino_set_br(struct super_block *sb, struct file *base)
37597 + aufs_bindex_t bindex, bbot, bshared;
37599 + struct file *old, *new;
37601 + struct au_branch *br;
37602 + struct inode *inode;
37603 + vfs_writef_t writef;
37605 + SiMustWriteLock(sb);
37608 + bbot = au_sbbot(sb);
37609 + fpair = kcalloc(bbot + 1, sizeof(*fpair), GFP_NOFS);
37610 + if (unlikely(!fpair))
37613 + inode = d_inode(sb->s_root);
37614 + ino = AUFS_ROOT_INO;
37615 + writef = au_sbi(sb)->si_xwrite;
37616 + for (bindex = 0, p = fpair; bindex <= bbot; bindex++, p++) {
37617 + bshared = is_sb_shared(sb, bindex, bindex - 1);
37618 + if (bshared >= 0) {
37619 + /* shared xino */
37620 + *p = fpair[bshared];
37621 + get_file(p->new);
37626 + br = au_sbr(sb, bindex);
37627 + p->old = br->br_xino.xi_file;
37628 + p->new = au_xino_create2(base, br->br_xino.xi_file);
37629 + err = PTR_ERR(p->new);
37630 + if (IS_ERR(p->new)) {
37636 + err = au_xino_do_write(writef, p->new,
37637 + au_h_iptr(inode, bindex)->i_ino, ino);
37638 + if (unlikely(err))
37642 + for (bindex = 0, p = fpair; bindex <= bbot; bindex++, p++) {
37643 + br = au_sbr(sb, bindex);
37644 + if (br->br_xino.xi_file)
37645 + fput(br->br_xino.xi_file);
37646 + get_file(p->new);
37647 + br->br_xino.xi_file = p->new;
37651 + for (bindex = 0, p = fpair; bindex <= bbot; bindex++, p++)
37661 +void au_xino_clr(struct super_block *sb)
37663 + struct au_sbinfo *sbinfo;
37665 + au_xigen_clr(sb);
37666 + xino_clear_xib(sb);
37667 + xino_clear_br(sb);
37668 + sbinfo = au_sbi(sb);
37669 + /* lvalue, do not call au_mntflags() */
37670 + au_opt_clr(sbinfo->si_mntflags, XINO);
37673 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount)
37676 + struct dentry *parent, *cur_parent;
37677 + struct qstr *dname, *cur_name;
37678 + struct file *cur_xino;
37679 + struct inode *dir;
37680 + struct au_sbinfo *sbinfo;
37682 + SiMustWriteLock(sb);
37685 + sbinfo = au_sbi(sb);
37686 + parent = dget_parent(xino->file->f_path.dentry);
37689 + dname = &xino->file->f_path.dentry->d_name;
37690 + cur_xino = sbinfo->si_xib;
37692 + cur_parent = dget_parent(cur_xino->f_path.dentry);
37693 + cur_name = &cur_xino->f_path.dentry->d_name;
37694 + skip = (cur_parent == parent
37695 + && au_qstreq(dname, cur_name));
37696 + dput(cur_parent);
37702 + au_opt_set(sbinfo->si_mntflags, XINO);
37703 + dir = d_inode(parent);
37704 + inode_lock_nested(dir, AuLsc_I_PARENT);
37705 + /* mnt_want_write() is unnecessary here */
37706 + err = au_xino_set_xib(sb, xino->file);
37708 + err = au_xigen_set(sb, xino->file);
37710 + err = au_xino_set_br(sb, xino->file);
37711 + inode_unlock(dir);
37713 + goto out; /* success */
37716 + AuIOErr("failed creating xino(%d).\n", err);
37717 + au_xigen_clr(sb);
37718 + xino_clear_xib(sb);
37725 +/* ---------------------------------------------------------------------- */
37728 + * create a xinofile at the default place/path.
37730 +struct file *au_xino_def(struct super_block *sb)
37732 + struct file *file;
37734 + struct au_branch *br;
37735 + struct super_block *h_sb;
37736 + struct path path;
37737 + aufs_bindex_t bbot, bindex, bwr;
37740 + bbot = au_sbbot(sb);
37742 + for (bindex = 0; bindex <= bbot; bindex++) {
37743 + br = au_sbr(sb, bindex);
37744 + if (au_br_writable(br->br_perm)
37745 + && !au_test_fs_bad_xino(au_br_sb(br))) {
37752 + file = ERR_PTR(-ENOMEM);
37753 + page = (void *)__get_free_page(GFP_NOFS);
37754 + if (unlikely(!page))
37756 + path.mnt = au_br_mnt(br);
37757 + path.dentry = au_h_dptr(sb->s_root, bwr);
37758 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
37759 + file = (void *)p;
37760 + if (!IS_ERR(p)) {
37761 + strcat(p, "/" AUFS_XINO_FNAME);
37762 + AuDbg("%s\n", p);
37763 + file = au_xino_create(sb, p, /*silent*/0);
37764 + if (!IS_ERR(file))
37765 + au_xino_brid_set(sb, br->br_id);
37767 + free_page((unsigned long)page);
37769 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0);
37770 + if (IS_ERR(file))
37772 + h_sb = file->f_path.dentry->d_sb;
37773 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
37774 + pr_err("xino doesn't support %s(%s)\n",
37775 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
37777 + file = ERR_PTR(-EINVAL);
37779 + if (!IS_ERR(file))
37780 + au_xino_brid_set(sb, -1);
37787 +/* ---------------------------------------------------------------------- */
37789 +int au_xino_path(struct seq_file *seq, struct file *file)
37793 + err = au_seq_path(seq, &file->f_path);
37794 + if (unlikely(err))
37797 +#define Deleted "\\040(deleted)"
37798 + seq->count -= sizeof(Deleted) - 1;
37799 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
37800 + sizeof(Deleted) - 1));
37807 +/* ---------------------------------------------------------------------- */
37809 +void au_xinondir_leave(struct super_block *sb, aufs_bindex_t bindex,
37810 + ino_t h_ino, int idx)
37812 + struct au_xino_file *xino;
37814 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
37815 + xino = &au_sbr(sb, bindex)->br_xino;
37816 + AuDebugOn(idx < 0 || xino->xi_nondir.total <= idx);
37818 + spin_lock(&xino->xi_nondir.spin);
37819 + AuDebugOn(xino->xi_nondir.array[idx] != h_ino);
37820 + xino->xi_nondir.array[idx] = 0;
37821 + spin_unlock(&xino->xi_nondir.spin);
37822 + wake_up_all(&xino->xi_nondir.wqh);
37825 +static int au_xinondir_find(struct au_xino_file *xino, ino_t h_ino)
37827 + int found, total, i;
37830 + total = xino->xi_nondir.total;
37831 + for (i = 0; i < total; i++) {
37832 + if (xino->xi_nondir.array[i] != h_ino)
37841 +static int au_xinondir_expand(struct au_xino_file *xino)
37846 + BUILD_BUG_ON(KMALLOC_MAX_SIZE > INT_MAX);
37849 + sz = xino->xi_nondir.total * sizeof(ino_t);
37850 + if (unlikely(sz > KMALLOC_MAX_SIZE / 2))
37852 + p = au_kzrealloc(xino->xi_nondir.array, sz, sz << 1, GFP_ATOMIC,
37853 + /*may_shrink*/0);
37855 + xino->xi_nondir.array = p;
37856 + xino->xi_nondir.total <<= 1;
37857 + AuDbg("xi_nondir.total %d\n", xino->xi_nondir.total);
37865 +int au_xinondir_enter(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
37868 + int err, found, empty;
37869 + struct au_xino_file *xino;
37873 + if (!au_opt_test(au_mntflags(sb), XINO))
37874 + goto out; /* no xino */
37876 + xino = &au_sbr(sb, bindex)->br_xino;
37879 + spin_lock(&xino->xi_nondir.spin);
37880 + found = au_xinondir_find(xino, h_ino);
37881 + if (found == -1) {
37882 + empty = au_xinondir_find(xino, /*h_ino*/0);
37883 + if (empty == -1) {
37884 + empty = xino->xi_nondir.total;
37885 + err = au_xinondir_expand(xino);
37886 + if (unlikely(err))
37889 + xino->xi_nondir.array[empty] = h_ino;
37892 + spin_unlock(&xino->xi_nondir.spin);
37893 + wait_event(xino->xi_nondir.wqh,
37894 + xino->xi_nondir.array[found] != h_ino);
37899 + spin_unlock(&xino->xi_nondir.spin);
37903 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
37904 --- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
37905 +++ linux/include/uapi/linux/aufs_type.h 2018-01-29 07:56:20.063325368 +0100
37908 + * Copyright (C) 2005-2017 Junjiro R. Okajima
37910 + * This program, aufs is free software; you can redistribute it and/or modify
37911 + * it under the terms of the GNU General Public License as published by
37912 + * the Free Software Foundation; either version 2 of the License, or
37913 + * (at your option) any later version.
37915 + * This program is distributed in the hope that it will be useful,
37916 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
37917 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
37918 + * GNU General Public License for more details.
37920 + * You should have received a copy of the GNU General Public License
37921 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
37924 +#ifndef __AUFS_TYPE_H__
37925 +#define __AUFS_TYPE_H__
37927 +#define AUFS_NAME "aufs"
37931 + * define it before including all other headers.
37932 + * sched.h may use pr_* macros before defining "current", so define the
37933 + * no-current version first, and re-define later.
37935 +#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
37936 +#include <linux/sched.h>
37938 +#define pr_fmt(fmt) \
37939 + AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
37940 + (int)sizeof(current->comm), current->comm, current->pid
37942 +#include <stdint.h>
37943 +#include <sys/types.h>
37944 +#endif /* __KERNEL__ */
37946 +#include <linux/limits.h>
37948 +#define AUFS_VERSION "4.x-rcN-20171218"
37950 +/* todo? move this to linux-2.6.19/include/magic.h */
37951 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
37953 +/* ---------------------------------------------------------------------- */
37955 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
37956 +typedef int8_t aufs_bindex_t;
37957 +#define AUFS_BRANCH_MAX 127
37959 +typedef int16_t aufs_bindex_t;
37960 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
37961 +#define AUFS_BRANCH_MAX 511
37962 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
37963 +#define AUFS_BRANCH_MAX 1023
37964 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
37965 +#define AUFS_BRANCH_MAX 32767
37970 +#ifndef AUFS_BRANCH_MAX
37971 +#error unknown CONFIG_AUFS_BRANCH_MAX value
37973 +#endif /* __KERNEL__ */
37975 +/* ---------------------------------------------------------------------- */
37977 +#define AUFS_FSTYPE AUFS_NAME
37979 +#define AUFS_ROOT_INO 2
37980 +#define AUFS_FIRST_INO 11
37982 +#define AUFS_WH_PFX ".wh."
37983 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
37984 +#define AUFS_WH_TMP_LEN 4
37985 +/* a limit for rmdir/rename a dir and copyup */
37986 +#define AUFS_MAX_NAMELEN (NAME_MAX \
37987 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
37989 + - AUFS_WH_TMP_LEN) /* hex */
37990 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
37991 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
37992 +#define AUFS_XINO_DEF_SEC 30 /* seconds */
37993 +#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
37994 +#define AUFS_DIRWH_DEF 3
37995 +#define AUFS_RDCACHE_DEF 10 /* seconds */
37996 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
37997 +#define AUFS_RDBLK_DEF 512 /* bytes */
37998 +#define AUFS_RDHASH_DEF 32
37999 +#define AUFS_WKQ_NAME AUFS_NAME "d"
38000 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
38001 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
38002 +#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
38003 +#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
38005 +/* pseudo-link maintenace under /proc */
38006 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
38007 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
38008 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
38010 +/* dirren, renamed dir */
38011 +#define AUFS_DR_INFO_PFX AUFS_WH_PFX ".dr."
38012 +#define AUFS_DR_BRHINO_NAME AUFS_WH_PFX "hino"
38013 +/* whiteouted doubly */
38014 +#define AUFS_WH_DR_INFO_PFX AUFS_WH_PFX AUFS_DR_INFO_PFX
38015 +#define AUFS_WH_DR_BRHINO AUFS_WH_PFX AUFS_DR_BRHINO_NAME
38017 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
38018 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
38020 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
38021 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
38022 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
38024 +/* doubly whiteouted */
38025 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
38026 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
38027 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
38029 +/* branch permissions and attributes */
38030 +#define AUFS_BRPERM_RW "rw"
38031 +#define AUFS_BRPERM_RO "ro"
38032 +#define AUFS_BRPERM_RR "rr"
38033 +#define AUFS_BRATTR_COO_REG "coo_reg"
38034 +#define AUFS_BRATTR_COO_ALL "coo_all"
38035 +#define AUFS_BRATTR_FHSM "fhsm"
38036 +#define AUFS_BRATTR_UNPIN "unpin"
38037 +#define AUFS_BRATTR_ICEX "icex"
38038 +#define AUFS_BRATTR_ICEX_SEC "icexsec"
38039 +#define AUFS_BRATTR_ICEX_SYS "icexsys"
38040 +#define AUFS_BRATTR_ICEX_TR "icextr"
38041 +#define AUFS_BRATTR_ICEX_USR "icexusr"
38042 +#define AUFS_BRATTR_ICEX_OTH "icexoth"
38043 +#define AUFS_BRRATTR_WH "wh"
38044 +#define AUFS_BRWATTR_NLWH "nolwh"
38045 +#define AUFS_BRWATTR_MOO "moo"
38047 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
38048 +#define AuBrPerm_RO (1 << 1) /* readonly */
38049 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
38050 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
38052 +#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
38053 +#define AuBrAttr_COO_ALL (1 << 4)
38054 +#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
38056 +#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
38057 +#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
38058 + branch. meaningless since
38059 + linux-3.18-rc1 */
38061 +/* ignore error in copying XATTR */
38062 +#define AuBrAttr_ICEX_SEC (1 << 7)
38063 +#define AuBrAttr_ICEX_SYS (1 << 8)
38064 +#define AuBrAttr_ICEX_TR (1 << 9)
38065 +#define AuBrAttr_ICEX_USR (1 << 10)
38066 +#define AuBrAttr_ICEX_OTH (1 << 11)
38067 +#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
38068 + | AuBrAttr_ICEX_SYS \
38069 + | AuBrAttr_ICEX_TR \
38070 + | AuBrAttr_ICEX_USR \
38071 + | AuBrAttr_ICEX_OTH)
38073 +#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
38074 +#define AuBrRAttr_Mask AuBrRAttr_WH
38076 +#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
38077 +#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
38078 +#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
38080 +#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
38082 +/* #warning test userspace */
38084 +#ifndef CONFIG_AUFS_FHSM
38085 +#undef AuBrAttr_FHSM
38086 +#define AuBrAttr_FHSM 0
38088 +#ifndef CONFIG_AUFS_XATTR
38089 +#undef AuBrAttr_ICEX
38090 +#define AuBrAttr_ICEX 0
38091 +#undef AuBrAttr_ICEX_SEC
38092 +#define AuBrAttr_ICEX_SEC 0
38093 +#undef AuBrAttr_ICEX_SYS
38094 +#define AuBrAttr_ICEX_SYS 0
38095 +#undef AuBrAttr_ICEX_TR
38096 +#define AuBrAttr_ICEX_TR 0
38097 +#undef AuBrAttr_ICEX_USR
38098 +#define AuBrAttr_ICEX_USR 0
38099 +#undef AuBrAttr_ICEX_OTH
38100 +#define AuBrAttr_ICEX_OTH 0
38104 +/* the longest combination */
38105 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
38106 +#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
38107 + "+" AUFS_BRATTR_COO_REG \
38108 + "+" AUFS_BRATTR_FHSM \
38109 + "+" AUFS_BRATTR_UNPIN \
38110 + "+" AUFS_BRATTR_ICEX_SEC \
38111 + "+" AUFS_BRATTR_ICEX_SYS \
38112 + "+" AUFS_BRATTR_ICEX_USR \
38113 + "+" AUFS_BRATTR_ICEX_OTH \
38114 + "+" AUFS_BRWATTR_NLWH)
38117 + char a[AuBrPermStrSz];
38118 +} au_br_perm_str_t;
38120 +static inline int au_br_writable(int brperm)
38122 + return brperm & AuBrPerm_RW;
38125 +static inline int au_br_whable(int brperm)
38127 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
38130 +static inline int au_br_wh_linkable(int brperm)
38132 + return !(brperm & AuBrWAttr_NoLinkWH);
38135 +static inline int au_br_cmoo(int brperm)
38137 + return brperm & AuBrAttr_CMOO_Mask;
38140 +static inline int au_br_fhsm(int brperm)
38142 + return brperm & AuBrAttr_FHSM;
38145 +/* ---------------------------------------------------------------------- */
38149 + /* readdir in userspace */
38153 + AuCtl_WBR_FD, /* pathconf wrapper */
38154 + AuCtl_IBUSY, /* busy inode */
38155 + AuCtl_MVDOWN, /* move-down */
38156 + AuCtl_BR, /* info about branches */
38157 + AuCtl_FHSM_FD /* connection for fhsm */
38160 +/* borrowed from linux/include/linux/kernel.h */
38162 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
38163 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
38166 +/* borrowed from linux/include/linux/compiler-gcc3.h */
38168 +#define __aligned(x) __attribute__((aligned(x)))
38173 +#define __packed __attribute__((packed))
38177 +struct au_rdu_cookie {
38182 + uint32_t generation;
38185 +struct au_rdu_ent {
38194 +static inline int au_rdu_len(int nlen)
38196 + /* include the terminating NULL */
38197 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
38198 + sizeof(uint64_t));
38201 +union au_rdu_ent_ul {
38202 + struct au_rdu_ent __user *e;
38214 + uint64_t sz; /* AuCtl_RDU */
38215 + uint64_t nent; /* AuCtl_RDU_INO */
38217 + union au_rdu_ent_ul ent;
38218 + uint16_t verify[AufsCtlRduV_End];
38220 + /* input/output */
38224 + union au_rdu_ent_ul tail;
38225 + /* number of entries which were added in a single call */
38230 + struct au_rdu_cookie cookie;
38233 +/* ---------------------------------------------------------------------- */
38235 +/* dirren. the branch is identified by the filename who contains this */
38236 +struct au_drinfo {
38239 + uint8_t oldnamelen;
38240 + uint64_t _padding;
38242 + uint8_t oldname[0];
38245 +struct au_drinfo_fdata {
38247 + struct au_drinfo drinfo;
38250 +#define AUFS_DRINFO_MAGIC_V1 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x01)
38252 +#define AUFS_DRINFO_MAGIC_V2 ('a' << 24 | 'd' << 16 | 'r' << 8 | 0x02)
38254 +/* ---------------------------------------------------------------------- */
38256 +struct aufs_wbr_fd {
38261 +/* ---------------------------------------------------------------------- */
38263 +struct aufs_ibusy {
38264 + uint64_t ino, h_ino;
38268 +/* ---------------------------------------------------------------------- */
38270 +/* error code for move-down */
38271 +/* the actual message strings are implemented in aufs-util.git */
38273 + EAU_MVDOWN_OPAQUE = 1,
38274 + EAU_MVDOWN_WHITEOUT,
38275 + EAU_MVDOWN_UPPER,
38276 + EAU_MVDOWN_BOTTOM,
38277 + EAU_MVDOWN_NOUPPER,
38278 + EAU_MVDOWN_NOLOWERBR,
38282 +/* flags for move-down */
38283 +#define AUFS_MVDOWN_DMSG 1
38284 +#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
38285 +#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
38286 +#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
38287 +#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
38288 +#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
38289 +#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
38290 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
38291 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
38292 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
38293 +#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
38294 +#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
38295 +#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
38297 +/* index for move-down */
38299 + AUFS_MVDOWN_UPPER,
38300 + AUFS_MVDOWN_LOWER,
38301 + AUFS_MVDOWN_NARRAY
38305 + * additional info of move-down
38306 + * number of free blocks and inodes.
38307 + * subset of struct kstatfs, but smaller and always 64bit.
38309 +struct aufs_stfs {
38310 + uint64_t f_blocks;
38311 + uint64_t f_bavail;
38312 + uint64_t f_files;
38313 + uint64_t f_ffree;
38316 +struct aufs_stbr {
38317 + int16_t brid; /* optional input */
38318 + int16_t bindex; /* output */
38319 + struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
38322 +struct aufs_mvdown {
38323 + uint32_t flags; /* input/output */
38324 + struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
38325 + int8_t au_errno; /* output */
38328 +/* ---------------------------------------------------------------------- */
38330 +union aufs_brinfo {
38331 + /* PATH_MAX may differ between kernel-space and user-space */
38332 + char _spacer[4096];
38340 +/* ---------------------------------------------------------------------- */
38342 +#define AuCtlType 'A'
38343 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
38344 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
38345 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
38346 + struct aufs_wbr_fd)
38347 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
38348 +#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
38349 + struct aufs_mvdown)
38350 +#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
38351 +#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
38353 +#endif /* __AUFS_TYPE_H__ */
38354 SPDX-License-Identifier: GPL-2.0
38355 aufs4.x-rcN loopback patch
38357 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
38358 index e51a59d..b4f7d3a 100644
38359 --- a/drivers/block/loop.c
38360 +++ b/drivers/block/loop.c
38361 @@ -600,6 +600,15 @@ static inline void loop_update_dio(struct loop_device *lo)
38365 +static struct file *loop_real_file(struct file *file)
38367 + struct file *f = NULL;
38369 + if (file->f_path.dentry->d_sb->s_op->real_loop)
38370 + f = file->f_path.dentry->d_sb->s_op->real_loop(file);
38374 static void loop_reread_partitions(struct loop_device *lo,
38375 struct block_device *bdev)
38377 @@ -634,6 +643,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
38380 struct file *file, *old_file;
38381 + struct file *f, *virt_file = NULL, *old_virt_file;
38382 struct inode *inode;
38385 @@ -650,9 +660,16 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
38389 + f = loop_real_file(file);
38391 + virt_file = file;
38396 inode = file->f_mapping->host;
38397 old_file = lo->lo_backing_file;
38398 + old_virt_file = lo->lo_backing_virt_file;
38402 @@ -667,6 +684,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
38403 blk_mq_freeze_queue(lo->lo_queue);
38404 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
38405 lo->lo_backing_file = file;
38406 + lo->lo_backing_virt_file = virt_file;
38407 lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
38408 mapping_set_gfp_mask(file->f_mapping,
38409 lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
38410 @@ -674,12 +692,16 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
38411 blk_mq_unfreeze_queue(lo->lo_queue);
38414 + if (old_virt_file)
38415 + fput(old_virt_file);
38416 if (lo->lo_flags & LO_FLAGS_PARTSCAN)
38417 loop_reread_partitions(lo, bdev);
38427 @@ -873,7 +895,7 @@ static int loop_prepare_queue(struct loop_device *lo)
38428 static int loop_set_fd(struct loop_device *lo, fmode_t mode,
38429 struct block_device *bdev, unsigned int arg)
38431 - struct file *file, *f;
38432 + struct file *file, *f, *virt_file = NULL;
38433 struct inode *inode;
38434 struct address_space *mapping;
38436 @@ -887,6 +909,12 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
38440 + f = loop_real_file(file);
38442 + virt_file = file;
38448 if (lo->lo_state != Lo_unbound)
38449 @@ -935,6 +963,7 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
38450 lo->lo_device = bdev;
38451 lo->lo_flags = lo_flags;
38452 lo->lo_backing_file = file;
38453 + lo->lo_backing_virt_file = virt_file;
38454 lo->transfer = NULL;
38456 lo->lo_sizelimit = 0;
38457 @@ -968,6 +997,8 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
38464 /* This is safe: open() is still holding a reference. */
38465 module_put(THIS_MODULE);
38466 @@ -1014,6 +1045,7 @@ loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
38467 static int loop_clr_fd(struct loop_device *lo)
38469 struct file *filp = lo->lo_backing_file;
38470 + struct file *virt_filp = lo->lo_backing_virt_file;
38471 gfp_t gfp = lo->old_gfp_mask;
38472 struct block_device *bdev = lo->lo_device;
38474 @@ -1045,6 +1077,7 @@ static int loop_clr_fd(struct loop_device *lo)
38475 spin_lock_irq(&lo->lo_lock);
38476 lo->lo_state = Lo_rundown;
38477 lo->lo_backing_file = NULL;
38478 + lo->lo_backing_virt_file = NULL;
38479 spin_unlock_irq(&lo->lo_lock);
38481 loop_release_xfer(lo);
38482 @@ -1092,6 +1125,8 @@ static int loop_clr_fd(struct loop_device *lo)
38483 * bd_mutex which is usually taken before lo_ctl_mutex.
38491 diff --git a/drivers/block/loop.h b/drivers/block/loop.h
38492 index 0f45416..101f193 100644
38493 --- a/drivers/block/loop.h
38494 +++ b/drivers/block/loop.h
38495 @@ -46,7 +46,7 @@ struct loop_device {
38496 int (*ioctl)(struct loop_device *, int cmd,
38497 unsigned long arg);
38499 - struct file * lo_backing_file;
38500 + struct file * lo_backing_file, *lo_backing_virt_file;
38501 struct block_device *lo_device;
38504 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
38505 index 0d4ea929..af293c2 100644
38506 --- a/fs/aufs/f_op.c
38507 +++ b/fs/aufs/f_op.c
38508 @@ -358,7 +358,7 @@ static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
38509 if (IS_ERR(h_file))
38512 - if (au_test_loopback_kthread()) {
38513 + if (0 && au_test_loopback_kthread()) {
38514 au_warn_loopback(h_file->f_path.dentry->d_sb);
38515 if (file->f_mapping != h_file->f_mapping) {
38516 file->f_mapping = h_file->f_mapping;
38517 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
38518 index 3b217c2..0f5ab22 100644
38519 --- a/fs/aufs/loop.c
38520 +++ b/fs/aufs/loop.c
38521 @@ -133,3 +133,19 @@ void au_loopback_fin(void)
38522 symbol_put(loop_backing_file);
38523 kfree(au_warn_loopback_array);
38526 +/* ---------------------------------------------------------------------- */
38528 +/* support the loopback block device insude aufs */
38530 +struct file *aufs_real_loop(struct file *file)
38534 + BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
38535 + fi_read_lock(file);
38536 + f = au_hf_top(file);
38537 + fi_read_unlock(file);
38541 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
38542 index 7d7bf34..ba7c188 100644
38543 --- a/fs/aufs/loop.h
38544 +++ b/fs/aufs/loop.h
38545 @@ -26,7 +26,11 @@ void au_warn_loopback(struct super_block *h_sb);
38547 int au_loopback_init(void);
38548 void au_loopback_fin(void);
38550 +struct file *aufs_real_loop(struct file *file);
38552 +AuStub(struct file *, loop_backing_file, return NULL)
38554 AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
38555 struct dentry *h_adding)
38556 AuStubInt0(au_test_loopback_kthread, void)
38557 @@ -34,6 +38,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
38559 AuStubInt0(au_loopback_init, void)
38560 AuStubVoid(au_loopback_fin, void)
38562 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
38563 #endif /* BLK_DEV_LOOP */
38565 #endif /* __KERNEL__ */
38566 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
38567 index 357bf04..c6614df 100644
38568 --- a/fs/aufs/super.c
38569 +++ b/fs/aufs/super.c
38570 @@ -839,7 +839,10 @@ static const struct super_operations aufs_sop = {
38571 .statfs = aufs_statfs,
38572 .put_super = aufs_put_super,
38573 .sync_fs = aufs_sync_fs,
38574 - .remount_fs = aufs_remount_fs
38575 + .remount_fs = aufs_remount_fs,
38576 +#ifdef CONFIG_AUFS_BDEV_LOOP
38577 + .real_loop = aufs_real_loop
38581 /* ---------------------------------------------------------------------- */
38582 diff --git a/include/linux/fs.h b/include/linux/fs.h
38583 index 96e05b3..ba5e627 100644
38584 --- a/include/linux/fs.h
38585 +++ b/include/linux/fs.h
38586 @@ -1840,6 +1840,10 @@ struct super_operations {
38587 struct shrink_control *);
38588 long (*free_cached_objects)(struct super_block *,
38589 struct shrink_control *);
38590 +#if defined(CONFIG_BLK_DEV_LOOP) || defined(CONFIG_BLK_DEV_LOOP_MODULE)
38592 + struct file *(*real_loop)(struct file *);