1 aufs4.x-rcN kbuild patch
3 diff --git a/fs/Kconfig b/fs/Kconfig
4 index 6ce72d8..4aa31ea 100644
7 @@ -221,6 +221,7 @@ source "fs/pstore/Kconfig"
8 source "fs/sysv/Kconfig"
9 source "fs/ufs/Kconfig"
10 source "fs/exofs/Kconfig"
11 +source "fs/aufs/Kconfig"
13 endif # MISC_FILESYSTEMS
15 diff --git a/fs/Makefile b/fs/Makefile
16 index 79f5225..a7c7f16 100644
19 @@ -126,3 +126,4 @@ obj-y += exofs/ # Multiple modules
20 obj-$(CONFIG_CEPH_FS) += ceph/
21 obj-$(CONFIG_PSTORE) += pstore/
22 obj-$(CONFIG_EFIVAR_FS) += efivarfs/
23 +obj-$(CONFIG_AUFS_FS) += aufs/
24 diff --git a/include/uapi/linux/Kbuild b/include/uapi/linux/Kbuild
25 index c2e5d6c..d736c11 100644
26 --- a/include/uapi/linux/Kbuild
27 +++ b/include/uapi/linux/Kbuild
28 @@ -59,6 +59,7 @@ header-y += atmsvc.h
30 header-y += atm_zatm.h
32 +header-y += aufs_type.h
33 header-y += auto_fs4.h
36 aufs4.x-rcN base patch
38 diff --git a/MAINTAINERS b/MAINTAINERS
39 index 233f834..c250892 100644
42 @@ -2029,6 +2029,19 @@ F: include/linux/audit.h
43 F: include/uapi/linux/audit.h
46 +AUFS (advanced multi layered unification filesystem) FILESYSTEM
47 +M: "J. R. Okajima" <hooanon05g@gmail.com>
48 +L: linux-unionfs@vger.kernel.org
49 +L: aufs-users@lists.sourceforge.net (members only)
50 +W: http://aufs.sourceforge.net
51 +T: git://github.com/sfjro/aufs4-linux.git
53 +F: Documentation/filesystems/aufs/
54 +F: Documentation/ABI/testing/debugfs-aufs
55 +F: Documentation/ABI/testing/sysfs-aufs
57 +F: include/uapi/linux/aufs_type.h
59 AUXILIARY DISPLAY DRIVERS
60 M: Miguel Ojeda Sandonis <miguel.ojeda.sandonis@gmail.com>
61 W: http://miguelojeda.es/auxdisplay.htm
62 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
63 index 423f4ca..abfdd2b 100644
64 --- a/drivers/block/loop.c
65 +++ b/drivers/block/loop.c
66 @@ -706,6 +706,24 @@ static inline int is_loop_device(struct file *file)
67 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
72 + * no get/put for file.
74 +struct file *loop_backing_file(struct super_block *sb)
77 + struct loop_device *l;
80 + if (MAJOR(sb->s_dev) == LOOP_MAJOR) {
81 + l = sb->s_bdev->bd_disk->private_data;
82 + ret = l->lo_backing_file;
86 +EXPORT_SYMBOL(loop_backing_file);
88 /* loop sysfs attributes */
90 static ssize_t loop_attr_show(struct device *dev, char *page,
91 diff --git a/fs/dcache.c b/fs/dcache.c
92 index 5c33aeb..8aa7f26 100644
95 @@ -1167,7 +1167,7 @@ enum d_walk_ret {
97 * The @enter() and @finish() callbacks are called with d_lock held.
99 -static void d_walk(struct dentry *parent, void *data,
100 +void d_walk(struct dentry *parent, void *data,
101 enum d_walk_ret (*enter)(void *, struct dentry *),
102 void (*finish)(void *))
104 diff --git a/fs/read_write.c b/fs/read_write.c
105 index 819ef3f..fd0414e 100644
106 --- a/fs/read_write.c
107 +++ b/fs/read_write.c
108 @@ -494,6 +494,28 @@ ssize_t __vfs_write(struct file *file, const char __user *p, size_t count,
110 EXPORT_SYMBOL(__vfs_write);
112 +vfs_readf_t vfs_readf(struct file *file)
114 + const struct file_operations *fop = file->f_op;
118 + if (fop->read_iter)
119 + return new_sync_read;
120 + return ERR_PTR(-ENOSYS);
123 +vfs_writef_t vfs_writef(struct file *file)
125 + const struct file_operations *fop = file->f_op;
129 + if (fop->write_iter)
130 + return new_sync_write;
131 + return ERR_PTR(-ENOSYS);
134 ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
137 diff --git a/fs/splice.c b/fs/splice.c
138 index 4cf700d..30a091d 100644
141 @@ -1110,8 +1110,8 @@ EXPORT_SYMBOL(generic_splice_sendpage);
143 * Attempt to initiate a splice from pipe to file.
145 -static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
146 - loff_t *ppos, size_t len, unsigned int flags)
147 +long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
148 + loff_t *ppos, size_t len, unsigned int flags)
150 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
151 loff_t *, size_t, unsigned int);
152 @@ -1127,9 +1127,9 @@ static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
154 * Attempt to initiate a splice from a file to a pipe.
156 -static long do_splice_to(struct file *in, loff_t *ppos,
157 - struct pipe_inode_info *pipe, size_t len,
158 - unsigned int flags)
159 +long do_splice_to(struct file *in, loff_t *ppos,
160 + struct pipe_inode_info *pipe, size_t len,
161 + unsigned int flags)
163 ssize_t (*splice_read)(struct file *, loff_t *,
164 struct pipe_inode_info *, size_t, unsigned int);
165 diff --git a/include/linux/file.h b/include/linux/file.h
166 index f87d308..9a290b3 100644
167 --- a/include/linux/file.h
168 +++ b/include/linux/file.h
169 @@ -19,6 +19,7 @@ struct dentry;
171 extern struct file *alloc_file(struct path *, fmode_t mode,
172 const struct file_operations *fop);
173 +extern struct file *get_empty_filp(void);
175 static inline void fput_light(struct file *file, int fput_needed)
177 diff --git a/include/linux/fs.h b/include/linux/fs.h
178 index 3aa5142..8d48506 100644
179 --- a/include/linux/fs.h
180 +++ b/include/linux/fs.h
181 @@ -1672,6 +1672,12 @@ ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
182 struct iovec *fast_pointer,
183 struct iovec **ret_pointer);
185 +typedef ssize_t (*vfs_readf_t)(struct file *, char __user *, size_t, loff_t *);
186 +typedef ssize_t (*vfs_writef_t)(struct file *, const char __user *, size_t,
188 +vfs_readf_t vfs_readf(struct file *file);
189 +vfs_writef_t vfs_writef(struct file *file);
191 extern ssize_t __vfs_read(struct file *, char __user *, size_t, loff_t *);
192 extern ssize_t __vfs_write(struct file *, const char __user *, size_t, loff_t *);
193 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
194 diff --git a/include/linux/splice.h b/include/linux/splice.h
195 index da2751d..2e0fca6 100644
196 --- a/include/linux/splice.h
197 +++ b/include/linux/splice.h
198 @@ -83,4 +83,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
199 extern void spd_release_page(struct splice_pipe_desc *, unsigned int);
201 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
203 +extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
204 + loff_t *ppos, size_t len, unsigned int flags);
205 +extern long do_splice_to(struct file *in, loff_t *ppos,
206 + struct pipe_inode_info *pipe, size_t len,
207 + unsigned int flags);
209 aufs4.x-rcN mmap patch
211 diff --git a/fs/proc/base.c b/fs/proc/base.c
212 index 4bd5d31..aa41f2a 100644
215 @@ -1921,7 +1921,7 @@ static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
216 down_read(&mm->mmap_sem);
217 vma = find_exact_vma(mm, vm_start, vm_end);
218 if (vma && vma->vm_file) {
219 - *path = vma->vm_file->f_path;
220 + *path = vma_pr_or_file(vma)->f_path;
224 diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
225 index f8595e8..cb8eda0 100644
226 --- a/fs/proc/nommu.c
227 +++ b/fs/proc/nommu.c
228 @@ -45,7 +45,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
229 file = region->vm_file;
232 - struct inode *inode = file_inode(region->vm_file);
233 + struct inode *inode;
235 + file = vmr_pr_or_file(region);
236 + inode = file_inode(file);
237 dev = inode->i_sb->s_dev;
240 diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
241 index 187b3b5..e03793e 100644
242 --- a/fs/proc/task_mmu.c
243 +++ b/fs/proc/task_mmu.c
244 @@ -281,7 +281,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid)
245 const char *name = NULL;
248 - struct inode *inode = file_inode(vma->vm_file);
249 + struct inode *inode;
251 + file = vma_pr_or_file(vma);
252 + inode = file_inode(file);
253 dev = inode->i_sb->s_dev;
255 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
256 @@ -1505,7 +1508,7 @@ static int show_numa_map(struct seq_file *m, void *v, int is_pid)
257 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
258 struct vm_area_struct *vma = v;
259 struct numa_maps *md = &numa_priv->md;
260 - struct file *file = vma->vm_file;
261 + struct file *file = vma_pr_or_file(vma);
262 struct mm_struct *mm = vma->vm_mm;
263 struct mm_walk walk = {
264 .hugetlb_entry = gather_hugetlb_stats,
265 diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
266 index e0d64c9..7aa92db 100644
267 --- a/fs/proc/task_nommu.c
268 +++ b/fs/proc/task_nommu.c
269 @@ -160,7 +160,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma,
273 - struct inode *inode = file_inode(vma->vm_file);
274 + struct inode *inode;
276 + file = vma_pr_or_file(vma);
277 + inode = file_inode(file);
278 dev = inode->i_sb->s_dev;
280 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
281 diff --git a/include/linux/mm.h b/include/linux/mm.h
282 index 00bad77..cc616b0 100644
283 --- a/include/linux/mm.h
284 +++ b/include/linux/mm.h
285 @@ -1183,6 +1183,28 @@ static inline int fixup_user_fault(struct task_struct *tsk,
289 +extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
290 +extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
292 +extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
293 +extern void vma_do_fput(struct vm_area_struct *, const char[], int);
295 +#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
297 +#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
299 +#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
300 +#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
303 +extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
304 +extern void vmr_do_fput(struct vm_region *, const char[], int);
306 +#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
308 +#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
309 +#endif /* !CONFIG_MMU */
311 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
312 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
313 void *buf, int len, int write);
314 diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
315 index f8d1492..c3a3760 100644
316 --- a/include/linux/mm_types.h
317 +++ b/include/linux/mm_types.h
318 @@ -272,6 +272,7 @@ struct vm_region {
319 unsigned long vm_top; /* region allocated to here */
320 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
321 struct file *vm_file; /* the backing file or NULL */
322 + struct file *vm_prfile; /* the virtual backing file or NULL */
324 int vm_usage; /* region usage count (access under nommu_region_sem) */
325 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
326 @@ -346,6 +347,7 @@ struct vm_area_struct {
327 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
328 units, *not* PAGE_CACHE_SIZE */
329 struct file * vm_file; /* File we map to (can be NULL). */
330 + struct file *vm_prfile; /* shadow of vm_file */
331 void * vm_private_data; /* was vm_pte (shared mem) */
334 diff --git a/kernel/fork.c b/kernel/fork.c
335 index fce002e..2230a0f 100644
338 @@ -464,7 +464,7 @@ static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
339 struct inode *inode = file_inode(file);
340 struct address_space *mapping = file->f_mapping;
344 if (tmp->vm_flags & VM_DENYWRITE)
345 atomic_dec(&inode->i_writecount);
346 i_mmap_lock_write(mapping);
347 diff --git a/mm/Makefile b/mm/Makefile
348 index 2ed4319..e3a53f5 100644
351 @@ -21,7 +21,7 @@ obj-y := filemap.o mempool.o oom_kill.o \
352 mm_init.o mmu_context.o percpu.o slab_common.o \
353 compaction.o vmacache.o \
354 interval_tree.o list_lru.o workingset.o \
356 + prfile.o debug.o $(mmu-y)
360 diff --git a/mm/filemap.c b/mm/filemap.c
361 index 1bb0076..8eaece8 100644
364 @@ -2128,7 +2128,7 @@ int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
365 int ret = VM_FAULT_LOCKED;
367 sb_start_pagefault(inode->i_sb);
368 - file_update_time(vma->vm_file);
369 + vma_file_update_time(vma);
371 if (page->mapping != inode->i_mapping) {
373 diff --git a/mm/memory.c b/mm/memory.c
374 index c387430..d434404 100644
377 @@ -2035,7 +2035,7 @@ static inline int wp_page_reuse(struct mm_struct *mm,
381 - file_update_time(vma->vm_file);
382 + vma_file_update_time(vma);
385 return VM_FAULT_WRITE;
386 diff --git a/mm/mmap.c b/mm/mmap.c
387 index 2ce04a6..f555c0a 100644
390 @@ -275,7 +275,7 @@ static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
391 if (vma->vm_ops && vma->vm_ops->close)
392 vma->vm_ops->close(vma);
394 - fput(vma->vm_file);
396 mpol_put(vma_policy(vma));
397 kmem_cache_free(vm_area_cachep, vma);
399 @@ -887,7 +887,7 @@ again: remove_next = 1 + (end > next->vm_end);
402 uprobe_munmap(next, next->vm_start, next->vm_end);
407 anon_vma_merge(vma, next);
408 @@ -1681,8 +1681,8 @@ out:
416 /* Undo any partial mapping done by a device driver. */
417 unmap_region(mm, vma, prev, vma->vm_start, vma->vm_end);
418 @@ -2488,7 +2488,7 @@ static int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
422 - get_file(new->vm_file);
425 if (new->vm_ops && new->vm_ops->open)
426 new->vm_ops->open(new);
427 @@ -2507,7 +2507,7 @@ static int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
428 if (new->vm_ops && new->vm_ops->close)
429 new->vm_ops->close(new);
431 - fput(new->vm_file);
433 unlink_anon_vmas(new);
435 mpol_put(vma_policy(new));
436 @@ -2649,7 +2649,6 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
437 struct vm_area_struct *vma;
438 unsigned long populate = 0;
439 unsigned long ret = -EINVAL;
442 pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. "
443 "See Documentation/vm/remap_file_pages.txt.\n",
444 @@ -2693,10 +2692,10 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
445 munlock_vma_pages_range(vma, start, start + size);
448 - file = get_file(vma->vm_file);
450 ret = do_mmap_pgoff(vma->vm_file, start, size,
451 prot, flags, pgoff, &populate);
455 up_write(&mm->mmap_sem);
457 @@ -2966,7 +2965,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
458 if (anon_vma_clone(new_vma, vma))
459 goto out_free_mempol;
460 if (new_vma->vm_file)
461 - get_file(new_vma->vm_file);
462 + vma_get_file(new_vma);
463 if (new_vma->vm_ops && new_vma->vm_ops->open)
464 new_vma->vm_ops->open(new_vma);
465 vma_link(mm, new_vma, prev, rb_link, rb_parent);
466 diff --git a/mm/nommu.c b/mm/nommu.c
467 index 92be862..29179f7 100644
470 @@ -671,7 +671,7 @@ static void __put_nommu_region(struct vm_region *region)
471 up_write(&nommu_region_sem);
474 - fput(region->vm_file);
477 /* IO memory and memory shared directly out of the pagecache
478 * from ramfs/tmpfs mustn't be released here */
479 @@ -829,7 +829,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
480 if (vma->vm_ops && vma->vm_ops->close)
481 vma->vm_ops->close(vma);
483 - fput(vma->vm_file);
485 put_nommu_region(vma->vm_region);
486 kmem_cache_free(vm_area_cachep, vma);
488 @@ -1355,7 +1355,7 @@ unsigned long do_mmap(struct file *file,
489 goto error_just_free;
492 - fput(region->vm_file);
494 kmem_cache_free(vm_region_jar, region);
497 @@ -1430,10 +1430,10 @@ error_just_free:
498 up_write(&nommu_region_sem);
501 - fput(region->vm_file);
503 kmem_cache_free(vm_region_jar, region);
505 - fput(vma->vm_file);
507 kmem_cache_free(vm_area_cachep, vma);
510 diff --git a/mm/prfile.c b/mm/prfile.c
512 index 0000000..b323b8a
517 + * Mainly for aufs which mmap(2) diffrent file and wants to print different path
518 + * in /proc/PID/maps.
519 + * Call these functions via macros defined in linux/mm.h.
521 + * See Documentation/filesystems/aufs/design/06mmap.txt
523 + * Copyright (c) 2014 Junjro R. Okajima
524 + * Copyright (c) 2014 Ian Campbell
527 +#include <linux/mm.h>
528 +#include <linux/file.h>
529 +#include <linux/fs.h>
531 +/* #define PRFILE_TRACE */
532 +static inline void prfile_trace(struct file *f, struct file *pr,
533 + const char func[], int line, const char func2[])
537 + pr_info("%s:%d: %s, %s\n", func, line, func2,
538 + f ? (char *)f->f_path.dentry->d_name.name : "(null)");
542 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
545 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
547 + prfile_trace(f, pr, func, line, __func__);
548 + file_update_time(f);
550 + file_update_time(pr);
553 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
556 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
558 + prfile_trace(f, pr, func, line, __func__);
559 + return (f && pr) ? pr : f;
562 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
564 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
566 + prfile_trace(f, pr, func, line, __func__);
572 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
574 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
576 + prfile_trace(f, pr, func, line, __func__);
583 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
586 + struct file *f = region->vm_file, *pr = region->vm_prfile;
588 + prfile_trace(f, pr, func, line, __func__);
589 + return (f && pr) ? pr : f;
592 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
594 + struct file *f = region->vm_file, *pr = region->vm_prfile;
596 + prfile_trace(f, pr, func, line, __func__);
601 +#endif /* !CONFIG_MMU */
602 aufs4.x-rcN standalone patch
604 diff --git a/fs/dcache.c b/fs/dcache.c
605 index 8aa7f26..f997345 100644
608 @@ -1272,6 +1272,7 @@ rename_retry:
612 +EXPORT_SYMBOL(d_walk);
615 * Search for at least 1 mount point in the dentry's subdirs.
616 diff --git a/fs/exec.c b/fs/exec.c
617 index b06623a..b9206c5 100644
620 @@ -103,6 +103,7 @@ bool path_noexec(const struct path *path)
621 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
622 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
624 +EXPORT_SYMBOL(path_noexec);
628 diff --git a/fs/file_table.c b/fs/file_table.c
629 index ad17e05..df66450 100644
630 --- a/fs/file_table.c
631 +++ b/fs/file_table.c
632 @@ -147,6 +147,7 @@ over:
634 return ERR_PTR(-ENFILE);
636 +EXPORT_SYMBOL(get_empty_filp);
639 * alloc_file - allocate and initialize a 'struct file'
640 @@ -308,6 +309,7 @@ void put_filp(struct file *file)
644 +EXPORT_SYMBOL(put_filp);
646 void __init files_init(void)
648 diff --git a/fs/namespace.c b/fs/namespace.c
649 index 0570729..ec560d8 100644
652 @@ -463,6 +463,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
653 mnt_dec_writers(real_mount(mnt));
656 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
659 * mnt_drop_write - give up write access to a mount
660 @@ -1803,6 +1804,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
664 +EXPORT_SYMBOL(iterate_mounts);
666 static void cleanup_group_ids(struct mount *mnt, struct mount *end)
668 diff --git a/fs/notify/group.c b/fs/notify/group.c
669 index d16b62c..06ca6bc 100644
670 --- a/fs/notify/group.c
671 +++ b/fs/notify/group.c
673 #include <linux/srcu.h>
674 #include <linux/rculist.h>
675 #include <linux/wait.h>
676 +#include <linux/module.h>
678 #include <linux/fsnotify_backend.h>
679 #include "fsnotify.h"
680 @@ -72,6 +73,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
682 atomic_inc(&group->refcnt);
684 +EXPORT_SYMBOL(fsnotify_get_group);
687 * Drop a reference to a group. Free it if it's through.
688 @@ -81,6 +83,7 @@ void fsnotify_put_group(struct fsnotify_group *group)
689 if (atomic_dec_and_test(&group->refcnt))
690 fsnotify_final_destroy_group(group);
692 +EXPORT_SYMBOL(fsnotify_put_group);
695 * Create a new fsnotify_group and hold a reference for the group returned.
696 @@ -109,6 +112,7 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
700 +EXPORT_SYMBOL(fsnotify_alloc_group);
702 int fsnotify_fasync(int fd, struct file *file, int on)
704 diff --git a/fs/notify/mark.c b/fs/notify/mark.c
705 index fc0df44..325b5c6 100644
706 --- a/fs/notify/mark.c
707 +++ b/fs/notify/mark.c
708 @@ -109,6 +109,7 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
709 mark->free_mark(mark);
712 +EXPORT_SYMBOL(fsnotify_put_mark);
714 /* Calculate mask of events for a list of marks */
715 u32 fsnotify_recalc_mask(struct hlist_head *head)
716 @@ -208,6 +209,7 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark,
717 mutex_unlock(&group->mark_mutex);
718 fsnotify_free_mark(mark);
720 +EXPORT_SYMBOL(fsnotify_destroy_mark);
722 void fsnotify_destroy_marks(struct hlist_head *head, spinlock_t *lock)
724 @@ -392,6 +394,7 @@ err:
728 +EXPORT_SYMBOL(fsnotify_add_mark);
730 int fsnotify_add_mark(struct fsnotify_mark *mark, struct fsnotify_group *group,
731 struct inode *inode, struct vfsmount *mnt, int allow_dups)
732 @@ -492,6 +495,7 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
733 atomic_set(&mark->refcnt, 1);
734 mark->free_mark = free_mark;
736 +EXPORT_SYMBOL(fsnotify_init_mark);
738 static int fsnotify_mark_destroy(void *ignored)
740 diff --git a/fs/open.c b/fs/open.c
741 index b6f1e96..4ab0d4e 100644
744 @@ -64,6 +64,7 @@ int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
745 mutex_unlock(&dentry->d_inode->i_mutex);
748 +EXPORT_SYMBOL(do_truncate);
750 long vfs_truncate(struct path *path, loff_t length)
752 @@ -678,6 +679,7 @@ int open_check_o_direct(struct file *f)
756 +EXPORT_SYMBOL(open_check_o_direct);
758 static int do_dentry_open(struct file *f,
760 diff --git a/fs/read_write.c b/fs/read_write.c
761 index fd0414e..8ace6ec 100644
762 --- a/fs/read_write.c
763 +++ b/fs/read_write.c
764 @@ -504,6 +504,7 @@ vfs_readf_t vfs_readf(struct file *file)
765 return new_sync_read;
766 return ERR_PTR(-ENOSYS);
768 +EXPORT_SYMBOL(vfs_readf);
770 vfs_writef_t vfs_writef(struct file *file)
772 @@ -515,6 +516,7 @@ vfs_writef_t vfs_writef(struct file *file)
773 return new_sync_write;
774 return ERR_PTR(-ENOSYS);
776 +EXPORT_SYMBOL(vfs_writef);
778 ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
780 diff --git a/fs/splice.c b/fs/splice.c
781 index 30a091d..c37c311 100644
784 @@ -1123,6 +1123,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
786 return splice_write(pipe, out, ppos, len, flags);
788 +EXPORT_SYMBOL(do_splice_from);
791 * Attempt to initiate a splice from a file to a pipe.
792 @@ -1149,6 +1150,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
794 return splice_read(in, ppos, pipe, len, flags);
796 +EXPORT_SYMBOL(do_splice_to);
799 * splice_direct_to_actor - splices data directly between two non-pipes
800 diff --git a/fs/xattr.c b/fs/xattr.c
801 index 9b932b9..44c457a 100644
804 @@ -207,6 +207,7 @@ vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
805 *xattr_value = value;
808 +EXPORT_SYMBOL(vfs_getxattr_alloc);
810 /* Compare an extended attribute value with the given value */
811 int vfs_xattr_cmp(struct dentry *dentry, const char *xattr_name,
812 diff --git a/security/commoncap.c b/security/commoncap.c
813 index 1832cf7..987ff5f 100644
814 --- a/security/commoncap.c
815 +++ b/security/commoncap.c
816 @@ -1053,12 +1053,14 @@ int cap_mmap_addr(unsigned long addr)
820 +EXPORT_SYMBOL(cap_mmap_addr);
822 int cap_mmap_file(struct file *file, unsigned long reqprot,
823 unsigned long prot, unsigned long flags)
827 +EXPORT_SYMBOL(cap_mmap_file);
829 #ifdef CONFIG_SECURITY
831 diff --git a/security/device_cgroup.c b/security/device_cgroup.c
832 index 03c1652..b00aa76 100644
833 --- a/security/device_cgroup.c
834 +++ b/security/device_cgroup.c
836 #include <linux/device_cgroup.h>
837 #include <linux/cgroup.h>
838 #include <linux/ctype.h>
839 +#include <linux/export.h>
840 #include <linux/list.h>
841 #include <linux/uaccess.h>
842 #include <linux/seq_file.h>
843 @@ -849,6 +850,7 @@ int __devcgroup_inode_permission(struct inode *inode, int mask)
844 return __devcgroup_check_permission(type, imajor(inode), iminor(inode),
847 +EXPORT_SYMBOL(__devcgroup_inode_permission);
849 int devcgroup_inode_mknod(int mode, dev_t dev)
851 diff --git a/security/security.c b/security/security.c
852 index 46f405c..54488b0 100644
853 --- a/security/security.c
854 +++ b/security/security.c
855 @@ -433,6 +433,7 @@ int security_path_rmdir(struct path *dir, struct dentry *dentry)
857 return call_int_hook(path_rmdir, 0, dir, dentry);
859 +EXPORT_SYMBOL(security_path_rmdir);
861 int security_path_unlink(struct path *dir, struct dentry *dentry)
863 @@ -449,6 +450,7 @@ int security_path_symlink(struct path *dir, struct dentry *dentry,
865 return call_int_hook(path_symlink, 0, dir, dentry, old_name);
867 +EXPORT_SYMBOL(security_path_symlink);
869 int security_path_link(struct dentry *old_dentry, struct path *new_dir,
870 struct dentry *new_dentry)
871 @@ -457,6 +459,7 @@ int security_path_link(struct dentry *old_dentry, struct path *new_dir,
873 return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
875 +EXPORT_SYMBOL(security_path_link);
877 int security_path_rename(struct path *old_dir, struct dentry *old_dentry,
878 struct path *new_dir, struct dentry *new_dentry,
879 @@ -484,6 +487,7 @@ int security_path_truncate(struct path *path)
881 return call_int_hook(path_truncate, 0, path);
883 +EXPORT_SYMBOL(security_path_truncate);
885 int security_path_chmod(struct path *path, umode_t mode)
887 @@ -491,6 +495,7 @@ int security_path_chmod(struct path *path, umode_t mode)
889 return call_int_hook(path_chmod, 0, path, mode);
891 +EXPORT_SYMBOL(security_path_chmod);
893 int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
895 @@ -498,6 +503,7 @@ int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
897 return call_int_hook(path_chown, 0, path, uid, gid);
899 +EXPORT_SYMBOL(security_path_chown);
901 int security_path_chroot(struct path *path)
903 @@ -583,6 +589,7 @@ int security_inode_readlink(struct dentry *dentry)
905 return call_int_hook(inode_readlink, 0, dentry);
907 +EXPORT_SYMBOL(security_inode_readlink);
909 int security_inode_follow_link(struct dentry *dentry, struct inode *inode,
911 @@ -598,6 +605,7 @@ int security_inode_permission(struct inode *inode, int mask)
913 return call_int_hook(inode_permission, 0, inode, mask);
915 +EXPORT_SYMBOL(security_inode_permission);
917 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
919 @@ -736,6 +744,7 @@ int security_file_permission(struct file *file, int mask)
921 return fsnotify_perm(file, mask);
923 +EXPORT_SYMBOL(security_file_permission);
925 int security_file_alloc(struct file *file)
927 @@ -795,6 +804,7 @@ int security_mmap_file(struct file *file, unsigned long prot,
929 return ima_file_mmap(file, prot);
931 +EXPORT_SYMBOL(security_mmap_file);
933 int security_mmap_addr(unsigned long addr)
935 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
936 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
937 +++ linux/Documentation/ABI/testing/debugfs-aufs 2016-01-13 20:11:11.663093444 +0100
939 +What: /debug/aufs/si_<id>/
941 +Contact: J. R. Okajima <hooanon05g@gmail.com>
943 + Under /debug/aufs, a directory named si_<id> is created
944 + per aufs mount, where <id> is a unique id generated
947 +What: /debug/aufs/si_<id>/plink
949 +Contact: J. R. Okajima <hooanon05g@gmail.com>
951 + It has three lines and shows the information about the
952 + pseudo-link. The first line is a single number
953 + representing a number of buckets. The second line is a
954 + number of pseudo-links per buckets (separated by a
955 + blank). The last line is a single number representing a
956 + total number of psedo-links.
957 + When the aufs mount option 'noplink' is specified, it
958 + will show "1\n0\n0\n".
960 +What: /debug/aufs/si_<id>/xib
962 +Contact: J. R. Okajima <hooanon05g@gmail.com>
964 + It shows the consumed blocks by xib (External Inode Number
965 + Bitmap), its block size and file size.
966 + When the aufs mount option 'noxino' is specified, it
967 + will be empty. About XINO files, see the aufs manual.
969 +What: /debug/aufs/si_<id>/xino0, xino1 ... xinoN
971 +Contact: J. R. Okajima <hooanon05g@gmail.com>
973 + It shows the consumed blocks by xino (External Inode Number
974 + Translation Table), its link count, block size and file
976 + When the aufs mount option 'noxino' is specified, it
977 + will be empty. About XINO files, see the aufs manual.
979 +What: /debug/aufs/si_<id>/xigen
981 +Contact: J. R. Okajima <hooanon05g@gmail.com>
983 + It shows the consumed blocks by xigen (External Inode
984 + Generation Table), its block size and file size.
985 + If CONFIG_AUFS_EXPORT is disabled, this entry will not
987 + When the aufs mount option 'noxino' is specified, it
988 + will be empty. About XINO files, see the aufs manual.
989 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
990 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
991 +++ linux/Documentation/ABI/testing/sysfs-aufs 2016-01-13 20:11:11.663093444 +0100
993 +What: /sys/fs/aufs/si_<id>/
995 +Contact: J. R. Okajima <hooanon05g@gmail.com>
997 + Under /sys/fs/aufs, a directory named si_<id> is created
998 + per aufs mount, where <id> is a unique id generated
1001 +What: /sys/fs/aufs/si_<id>/br0, br1 ... brN
1003 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1005 + It shows the abolute path of a member directory (which
1006 + is called branch) in aufs, and its permission.
1008 +What: /sys/fs/aufs/si_<id>/brid0, brid1 ... bridN
1010 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1012 + It shows the id of a member directory (which is called
1015 +What: /sys/fs/aufs/si_<id>/xi_path
1017 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1019 + It shows the abolute path of XINO (External Inode Number
1020 + Bitmap, Translation Table and Generation Table) file
1021 + even if it is the default path.
1022 + When the aufs mount option 'noxino' is specified, it
1023 + will be empty. About XINO files, see the aufs manual.
1024 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1025 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
1026 +++ linux/Documentation/filesystems/aufs/design/01intro.txt 2016-01-13 20:11:11.663093444 +0100
1029 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1031 +# This program is free software; you can redistribute it and/or modify
1032 +# it under the terms of the GNU General Public License as published by
1033 +# the Free Software Foundation; either version 2 of the License, or
1034 +# (at your option) any later version.
1036 +# This program is distributed in the hope that it will be useful,
1037 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1038 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1039 +# GNU General Public License for more details.
1041 +# You should have received a copy of the GNU General Public License
1042 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1045 +----------------------------------------
1047 +aufs [ei ju: ef es] | [a u f s]
1048 +1. abbrev. for "advanced multi-layered unification filesystem".
1049 +2. abbrev. for "another unionfs".
1050 +3. abbrev. for "auf das" in German which means "on the" in English.
1051 + Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1052 + But "Filesystem aufs Filesystem" is hard to understand.
1054 +AUFS is a filesystem with features:
1055 +- multi layered stackable unification filesystem, the member directory
1056 + is called as a branch.
1057 +- branch permission and attribute, 'readonly', 'real-readonly',
1058 + 'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1060 +- internal "file copy-on-write".
1061 +- logical deletion, whiteout.
1062 +- dynamic branch manipulation, adding, deleting and changing permission.
1063 +- allow bypassing aufs, user's direct branch access.
1064 +- external inode number translation table and bitmap which maintains the
1065 + persistent aufs inode number.
1066 +- seekable directory, including NFS readdir.
1067 +- file mapping, mmap and sharing pages.
1068 +- pseudo-link, hardlink over branches.
1069 +- loopback mounted filesystem as a branch.
1070 +- several policies to select one among multiple writable branches.
1071 +- revert a single systemcall when an error occurs in aufs.
1075 +Multi Layered Stackable Unification Filesystem
1076 +----------------------------------------------------------------------
1077 +Most people already knows what it is.
1078 +It is a filesystem which unifies several directories and provides a
1079 +merged single directory. When users access a file, the access will be
1080 +passed/re-directed/converted (sorry, I am not sure which English word is
1081 +correct) to the real file on the member filesystem. The member
1082 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1083 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1084 +readonly branch is handled by creating 'whiteout' on the upper writable
1087 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1088 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1089 +different approaches to implement the merged-view.
1090 +The former tries putting it into VFS, and the latter implements as a
1091 +separate filesystem.
1092 +(If I misunderstand about these implementations, please let me know and
1093 +I shall correct it. Because it is a long time ago when I read their
1094 +source files last time).
1096 +UnionMount's approach will be able to small, but may be hard to share
1097 +branches between several UnionMount since the whiteout in it is
1098 +implemented in the inode on branch filesystem and always
1099 +shared. According to Bharata's post, readdir does not seems to be
1101 +There are several missing features known in this implementations such as
1102 +- for users, the inode number may change silently. eg. copy-up.
1103 +- link(2) may break by copy-up.
1104 +- read(2) may get an obsoleted filedata (fstat(2) too).
1105 +- fcntl(F_SETLK) may be broken by copy-up.
1106 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1109 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1110 +merged into mainline. This is another implementation of UnionMount as a
1111 +separated filesystem. All the limitations and known problems which
1112 +UnionMount are equally inherited to "overlay" filesystem.
1114 +Unionfs has a longer history. When I started implementing a stackable
1115 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1116 +inode, dentry and file objects and they have an array pointing lower
1117 +same kind objects. After contributing many patches for Unionfs, I
1118 +re-started my project AUFS (Jun 2006).
1120 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1121 +implemented my own ideas, approaches and enhancements and it became
1122 +totally different one.
1124 +Comparing DM snapshot and fs based implementation
1125 +- the number of bytes to be copied between devices is much smaller.
1126 +- the type of filesystem must be one and only.
1127 +- the fs must be writable, no readonly fs, even for the lower original
1128 + device. so the compression fs will not be usable. but if we use
1129 + loopback mount, we may address this issue.
1131 + mount /cdrom/squashfs.img /sq
1132 + losetup /sq/ext2.img
1133 + losetup /somewhere/cow
1134 + dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1135 +- it will be difficult (or needs more operations) to extract the
1136 + difference between the original device and COW.
1137 +- DM snapshot-merge may help a lot when users try merging. in the
1138 + fs-layer union, users will use rsync(1).
1140 +You may want to read my old paper "Filesystems in LiveCD"
1141 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1144 +Several characters/aspects/persona of aufs
1145 +----------------------------------------------------------------------
1147 +Aufs has several characters, aspects or persona.
1148 +1. a filesystem, callee of VFS helper
1149 +2. sub-VFS, caller of VFS helper for branches
1150 +3. a virtual filesystem which maintains persistent inode number
1151 +4. reader/writer of files on branches such like an application
1153 +1. Callee of VFS Helper
1154 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1155 +unlink(2) from an application reaches sys_unlink() kernel function and
1156 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1157 +calls filesystem specific unlink operation. Actually aufs implements the
1158 +unlink operation but it behaves like a redirector.
1160 +2. Caller of VFS Helper for Branches
1161 +aufs_unlink() passes the unlink request to the branch filesystem as if
1162 +it were called from VFS. So the called unlink operation of the branch
1163 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1164 +every necessary pre/post operation for the branch filesystem.
1165 +- acquire the lock for the parent dir on a branch
1166 +- lookup in a branch
1167 +- revalidate dentry on a branch
1168 +- mnt_want_write() for a branch
1169 +- vfs_unlink() for a branch
1170 +- mnt_drop_write() for a branch
1171 +- release the lock on a branch
1173 +3. Persistent Inode Number
1174 +One of the most important issue for a filesystem is to maintain inode
1175 +numbers. This is particularly important to support exporting a
1176 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1177 +backend block device for its own. But some storage is necessary to
1178 +keep and maintain the inode numbers. It may be a large space and may not
1179 +suit to keep in memory. Aufs rents some space from its first writable
1180 +branch filesystem (by default) and creates file(s) on it. These files
1181 +are created by aufs internally and removed soon (currently) keeping
1183 +Note: Because these files are removed, they are totally gone after
1184 + unmounting aufs. It means the inode numbers are not persistent
1185 + across unmount or reboot. I have a plan to make them really
1186 + persistent which will be important for aufs on NFS server.
1188 +4. Read/Write Files Internally (copy-on-write)
1189 +Because a branch can be readonly, when you write a file on it, aufs will
1190 +"copy-up" it to the upper writable branch internally. And then write the
1191 +originally requested thing to the file. Generally kernel doesn't
1192 +open/read/write file actively. In aufs, even a single write may cause a
1193 +internal "file copy". This behaviour is very similar to cp(1) command.
1195 +Some people may think it is better to pass such work to user space
1196 +helper, instead of doing in kernel space. Actually I am still thinking
1197 +about it. But currently I have implemented it in kernel space.
1198 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1199 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1200 +++ linux/Documentation/filesystems/aufs/design/02struct.txt 2016-01-13 20:11:11.663093444 +0100
1203 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1205 +# This program is free software; you can redistribute it and/or modify
1206 +# it under the terms of the GNU General Public License as published by
1207 +# the Free Software Foundation; either version 2 of the License, or
1208 +# (at your option) any later version.
1210 +# This program is distributed in the hope that it will be useful,
1211 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1212 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1213 +# GNU General Public License for more details.
1215 +# You should have received a copy of the GNU General Public License
1216 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1218 +Basic Aufs Internal Structure
1220 +Superblock/Inode/Dentry/File Objects
1221 +----------------------------------------------------------------------
1222 +As like an ordinary filesystem, aufs has its own
1223 +superblock/inode/dentry/file objects. All these objects have a
1224 +dynamically allocated array and store the same kind of pointers to the
1225 +lower filesystem, branch.
1226 +For example, when you build a union with one readwrite branch and one
1227 +readonly, mounted /au, /rw and /ro respectively.
1229 +- /ro/fileA exists but /rw/fileA
1231 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1232 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1233 +- [0] = NULL (because /rw/fileA doesn't exist)
1236 +This style of an array is essentially same to the aufs
1237 +superblock/inode/dentry/file objects.
1239 +Because aufs supports manipulating branches, ie. add/delete/change
1240 +branches dynamically, these objects has its own generation. When
1241 +branches are changed, the generation in aufs superblock is
1242 +incremented. And a generation in other object are compared when it is
1243 +accessed. When a generation in other objects are obsoleted, aufs
1244 +refreshes the internal array.
1248 +----------------------------------------------------------------------
1249 +Additionally aufs superblock has some data for policies to select one
1250 +among multiple writable branches, XIB files, pseudo-links and kobject.
1251 +See below in detail.
1252 +About the policies which supports copy-down a directory, see
1253 +wbr_policy.txt too.
1256 +Branch and XINO(External Inode Number Translation Table)
1257 +----------------------------------------------------------------------
1258 +Every branch has its own xino (external inode number translation table)
1259 +file. The xino file is created and unlinked by aufs internally. When two
1260 +members of a union exist on the same filesystem, they share the single
1262 +The struct of a xino file is simple, just a sequence of aufs inode
1263 +numbers which is indexed by the lower inode number.
1264 +In the above sample, assume the inode number of /ro/fileA is i111 and
1265 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1266 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1268 +When the inode numbers are not contiguous, the xino file will be sparse
1269 +which has a hole in it and doesn't consume as much disk space as it
1270 +might appear. If your branch filesystem consumes disk space for such
1271 +holes, then you should specify 'xino=' option at mounting aufs.
1273 +Aufs has a mount option to free the disk blocks for such holes in XINO
1274 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1275 +meet a problem of disk shortage due to XINO files, then you should try
1276 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1277 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1278 +the holes in XINO files.
1280 +Also a writable branch has three kinds of "whiteout bases". All these
1281 +are existed when the branch is joined to aufs, and their names are
1282 +whiteout-ed doubly, so that users will never see their names in aufs
1284 +1. a regular file which will be hardlinked to all whiteouts.
1285 +2. a directory to store a pseudo-link.
1286 +3. a directory to store an "orphan"-ed file temporary.
1289 + When you remove a file on a readonly branch, aufs handles it as a
1290 + logical deletion and creates a whiteout on the upper writable branch
1291 + as a hardlink of this file in order not to consume inode on the
1294 + See below, Pseudo-link.
1296 + When "fileC" exists on the lower readonly branch only and it is
1297 + opened and removed with its parent dir, and then user writes
1298 + something into it, then aufs copies-up fileC to this
1299 + directory. Because there is no other dir to store fileC. After
1300 + creating a file under this dir, the file is unlinked.
1302 +Because aufs supports manipulating branches, ie. add/delete/change
1303 +dynamically, a branch has its own id. When the branch order changes,
1304 +aufs finds the new index by searching the branch id.
1308 +----------------------------------------------------------------------
1309 +Assume "fileA" exists on the lower readonly branch only and it is
1310 +hardlinked to "fileB" on the branch. When you write something to fileA,
1311 +aufs copies-up it to the upper writable branch. Additionally aufs
1312 +creates a hardlink under the Pseudo-link Directory of the writable
1313 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1314 +simple list. If fileB is read after unlinking fileA, aufs returns
1315 +filedata from the pseudo-link instead of the lower readonly
1316 +branch. Because the pseudo-link is based upon the inode, to keep the
1317 +inode number by xino (see above) is essentially necessary.
1319 +All the hardlinks under the Pseudo-link Directory of the writable branch
1320 +should be restored in a proper location later. Aufs provides a utility
1321 +to do this. The userspace helpers executed at remounting and unmounting
1323 +During this utility is running, it puts aufs into the pseudo-link
1324 +maintenance mode. In this mode, only the process which began the
1325 +maintenance mode (and its child processes) is allowed to operate in
1326 +aufs. Some other processes which are not related to the pseudo-link will
1327 +be allowed to run too, but the rest have to return an error or wait
1328 +until the maintenance mode ends. If a process already acquires an inode
1329 +mutex (in VFS), it has to return an error.
1332 +XIB(external inode number bitmap)
1333 +----------------------------------------------------------------------
1334 +Addition to the xino file per a branch, aufs has an external inode number
1335 +bitmap in a superblock object. It is also an internal file such like a
1337 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1339 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1341 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1342 +reduce the number of consumed disk blocks for these files.
1345 +Virtual or Vertical Dir, and Readdir in Userspace
1346 +----------------------------------------------------------------------
1347 +In order to support multiple layers (branches), aufs readdir operation
1348 +constructs a virtual dir block on memory. For readdir, aufs calls
1349 +vfs_readdir() internally for each dir on branches, merges their entries
1350 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1351 +object. So the file object has its entry list until it is closed. The
1352 +entry list will be updated when the file position is zero and becomes
1353 +obsoleted. This decision is made in aufs automatically.
1355 +The dynamically allocated memory block for the name of entries has a
1356 +unit of 512 bytes (by default) and stores the names contiguously (no
1357 +padding). Another block for each entry is handled by kmem_cache too.
1358 +During building dir blocks, aufs creates hash list and judging whether
1359 +the entry is whiteouted by its upper branch or already listed.
1360 +The merged result is cached in the corresponding inode object and
1361 +maintained by a customizable life-time option.
1363 +Some people may call it can be a security hole or invite DoS attack
1364 +since the opened and once readdir-ed dir (file object) holds its entry
1365 +list and becomes a pressure for system memory. But I'd say it is similar
1366 +to files under /proc or /sys. The virtual files in them also holds a
1367 +memory page (generally) while they are opened. When an idea to reduce
1368 +memory for them is introduced, it will be applied to aufs too.
1369 +For those who really hate this situation, I've developed readdir(3)
1370 +library which operates this merging in userspace. You just need to set
1371 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1372 +kernel space for readdir(3).
1376 +----------------------------------------------------------------------
1377 +Aufs sometimes requires privilege access to a branch. For instance,
1378 +in copy-up/down operation. When a user process is going to make changes
1379 +to a file which exists in the lower readonly branch only, and the mode
1380 +of one of ancestor directories may not be writable by a user
1381 +process. Here aufs copy-up the file with its ancestors and they may
1382 +require privilege to set its owner/group/mode/etc.
1383 +This is a typical case of a application character of aufs (see
1386 +Aufs uses workqueue synchronously for this case. It creates its own
1387 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1388 +passes the request to call mkdir or write (for example), and wait for
1389 +its completion. This approach solves a problem of a signal handler
1391 +If aufs didn't adopt the workqueue and changed the privilege of the
1392 +process, then the process may receive the unexpected SIGXFSZ or other
1395 +Also aufs uses the system global workqueue ("events" kernel thread) too
1396 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1397 +whiteout base and etc. This is unrelated to a privilege.
1398 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1399 +superblock at the beginning, at the same time waits for the completion
1400 +of all queued asynchronous tasks.
1404 +----------------------------------------------------------------------
1405 +The whiteout in aufs is very similar to Unionfs's. That is represented
1406 +by its filename. UnionMount takes an approach of a file mode, but I am
1407 +afraid several utilities (find(1) or something) will have to support it.
1409 +Basically the whiteout represents "logical deletion" which stops aufs to
1410 +lookup further, but also it represents "dir is opaque" which also stop
1413 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1414 +In order to make several functions in a single systemcall to be
1415 +revertible, aufs adopts an approach to rename a directory to a temporary
1416 +unique whiteouted name.
1417 +For example, in rename(2) dir where the target dir already existed, aufs
1418 +renames the target dir to a temporary unique whiteouted name before the
1419 +actual rename on a branch, and then handles other actions (make it opaque,
1420 +update the attributes, etc). If an error happens in these actions, aufs
1421 +simply renames the whiteouted name back and returns an error. If all are
1422 +succeeded, aufs registers a function to remove the whiteouted unique
1423 +temporary name completely and asynchronously to the system global
1428 +----------------------------------------------------------------------
1429 +It is a well-known feature or concept.
1430 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1431 +internally and makes change to the new file on the upper writable branch.
1432 +When the trigger systemcall does not update the timestamps of the parent
1433 +dir, aufs reverts it after copy-up.
1436 +Move-down (aufs3.9 and later)
1437 +----------------------------------------------------------------------
1438 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1439 +the lower readonly branch to the upper writable branch when a user
1440 +changes something about the file.
1441 +"Move-down" is an opposite action of copy-up. Basically this action is
1442 +ran manually instead of automatically and internally.
1443 +For desgin and implementation, aufs has to consider these issues.
1444 +- whiteout for the file may exist on the lower branch.
1445 +- ancestor directories may not exist on the lower branch.
1446 +- diropq for the ancestor directories may exist on the upper branch.
1447 +- free space on the lower branch will reduce.
1448 +- another access to the file may happen during moving-down, including
1449 + UDBA (see "Revalidate Dentry and UDBA").
1450 +- the file should not be hard-linked nor pseudo-linked. they should be
1451 + handled by auplink utility later.
1453 +Sometimes users want to move-down a file from the upper writable branch
1454 +to the lower readonly or writable branch. For instance,
1455 +- the free space of the upper writable branch is going to run out.
1456 +- create a new intermediate branch between the upper and lower branch.
1459 +For this purpose, use "aumvdown" command in aufs-util.git.
1460 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1461 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt 1970-01-01 01:00:00.000000000 +0100
1462 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt 2016-01-13 20:11:11.663093444 +0100
1465 +# Copyright (C) 2015 Junjiro R. Okajima
1467 +# This program is free software; you can redistribute it and/or modify
1468 +# it under the terms of the GNU General Public License as published by
1469 +# the Free Software Foundation; either version 2 of the License, or
1470 +# (at your option) any later version.
1472 +# This program is distributed in the hope that it will be useful,
1473 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1474 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1475 +# GNU General Public License for more details.
1477 +# You should have received a copy of the GNU General Public License
1478 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1480 +Support for a branch who has its ->atomic_open()
1481 +----------------------------------------------------------------------
1482 +The filesystems who implement its ->atomic_open() are not majority. For
1483 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1484 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1485 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1486 +sure whether all filesystems who have ->atomic_open() behave like this,
1487 +but NFSv4 surely returns the error.
1489 +In order to support ->atomic_open() for aufs, there are a few
1492 +A. Introduce aufs_atomic_open()
1493 + - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1495 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1496 + an aufs user Pip Cet's approach
1497 + - calls aufs_create(), VFS finish_open() and notify_change().
1498 + - pass fake-mode to finish_open(), and then correct the mode by
1500 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1501 + - no aufs_atomic_open().
1502 + - aufs_lookup() registers the TID to an aufs internal object.
1503 + - aufs_create() does nothing when the matching TID is registered, but
1504 + registers the mode.
1505 + - aufs_open() calls branch fs's ->atomic_open() when the matching
1506 + TID is registered.
1507 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1509 + - no aufs_atomic_open().
1510 + - aufs_create() registers the TID to an internal object. this info
1511 + represents "this process created this file just now."
1512 + - when aufs gets EACCES from branch fs's ->open(), then confirm the
1513 + registered TID and re-try open() with superuser's credential.
1515 +Pros and cons for each approach.
1518 + - straightforward but highly depends upon VFS internal.
1519 + - the atomic behavaiour is kept.
1520 + - some of parameters such as nameidata are hard to reproduce for
1524 + - easy to implement.
1525 + - the atomic behavaiour is lost.
1527 + - the atomic behavaiour is kept.
1528 + - dirty and tricky.
1529 + - VFS checks whether the file is created correctly after calling
1530 + ->create(), which means this approach doesn't work.
1532 + - easy to implement.
1533 + - the atomic behavaiour is lost.
1534 + - to open a file with superuser's credential and give it to a user
1535 + process is a bad idea, since the file object keeps the credential
1536 + in it. It may affect LSM or something. This approach doesn't work
1539 +The approach A is ideal, but it hard to implement. So here is a
1540 +variation of A, which is to be implemented.
1542 +A-1. Introduce aufs_atomic_open()
1543 + - calls branch fs ->atomic_open() if exists. otherwise calls
1544 + vfs_create() and finish_open().
1545 + - the demerit is that the several checks after branch fs
1546 + ->atomic_open() are lost. in the ordinary case, the checks are
1547 + done by VFS:do_last(), lookup_open() and atomic_open(). some can
1548 + be implemented in aufs, but not all I am afraid.
1549 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1550 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1551 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2016-01-13 20:11:11.663093444 +0100
1554 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1556 +# This program is free software; you can redistribute it and/or modify
1557 +# it under the terms of the GNU General Public License as published by
1558 +# the Free Software Foundation; either version 2 of the License, or
1559 +# (at your option) any later version.
1561 +# This program is distributed in the hope that it will be useful,
1562 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1563 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1564 +# GNU General Public License for more details.
1566 +# You should have received a copy of the GNU General Public License
1567 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1570 +----------------------------------------------------------------------
1571 +Since aufs has a character of sub-VFS (see Introduction), it operates
1572 +lookup for branches as VFS does. It may be a heavy work. But almost all
1573 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1574 +directly connected to its parent. Digging down the directory hierarchy
1575 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1578 +When a branch is a remote filesystem, aufs basically relies upon its
1579 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1581 +For d_revalidate, aufs implements three levels of revalidate tests. See
1582 +"Revalidate Dentry and UDBA" in detail.
1585 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1586 +----------------------------------------------------------------------
1587 +Let's try case study.
1588 +- aufs has two branches, upper readwrite and lower readonly.
1590 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1591 +- user invoked "chmod a+rx /au/dirA"
1592 +- the internal copy-up is activated and "/rw/dirA" is created and its
1593 + permission bits are set to world readable.
1594 +- then "/au/dirA" becomes world readable?
1596 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1597 +or it may be a natively readonly filesystem. If aufs respects the lower
1598 +branch, it should not respond readdir request from other users. But user
1599 +allowed it by chmod. Should really aufs rejects showing the entries
1602 +To be honest, I don't have a good solution for this case. So aufs
1603 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1605 +When dirperm1 is specified, aufs checks only the highest one for the
1606 +directory permission, and shows the entries. Otherwise, as usual, checks
1607 +every dir existing on all branches and rejects the request.
1609 +As a side effect, dirperm1 option improves the performance of aufs
1610 +because the number of permission check is reduced when the number of
1614 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1615 +----------------------------------------------------------------------
1616 +Generally VFS helpers re-validate a dentry as a part of lookup.
1617 +0. digging down the directory hierarchy.
1618 +1. lock the parent dir by its i_mutex.
1619 +2. lookup the final (child) entry.
1621 +4. call the actual operation (create, unlink, etc.)
1622 +5. unlock the parent dir
1624 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1625 +called. Actually aufs implements it and checks the dentry on a branch is
1627 +But it is not enough. Because aufs has to release the lock for the
1628 +parent dir on a branch at the end of ->lookup() (step 2) and
1629 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1631 +If the file on a branch is changed directly, eg. bypassing aufs, after
1632 +aufs released the lock, then the subsequent operation may cause
1633 +something unpleasant result.
1635 +This situation is a result of VFS architecture, ->lookup() and
1636 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1637 +design from VFS's point of view. It is just not suitable for sub-VFS
1640 +Aufs supports such case by three level of revalidation which is
1641 +selectable by user.
1642 +1. Simple Revalidate
1643 + Addition to the native flow in VFS's, confirm the child-parent
1644 + relationship on the branch just after locking the parent dir on the
1645 + branch in the "actual operation" (step 4). When this validation
1646 + fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1647 + checks the validation of the dentry on branches.
1648 +2. Monitor Changes Internally by Inotify/Fsnotify
1649 + Addition to above, in the "actual operation" (step 4) aufs re-lookup
1650 + the dentry on the branch, and returns EBUSY if it finds different
1652 + Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1653 + during it is in cache. When the event is notified, aufs registers a
1654 + function to kernel 'events' thread by schedule_work(). And the
1655 + function sets some special status to the cached aufs dentry and inode
1656 + private data. If they are not cached, then aufs has nothing to
1657 + do. When the same file is accessed through aufs (step 0-3) later,
1658 + aufs will detect the status and refresh all necessary data.
1659 + In this mode, aufs has to ignore the event which is fired by aufs
1661 +3. No Extra Validation
1662 + This is the simplest test and doesn't add any additional revalidation
1663 + test, and skip the revalidation in step 4. It is useful and improves
1664 + aufs performance when system surely hide the aufs branches from user,
1665 + by over-mounting something (or another method).
1666 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1667 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1668 +++ linux/Documentation/filesystems/aufs/design/04branch.txt 2016-01-13 20:11:11.663093444 +0100
1671 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1673 +# This program is free software; you can redistribute it and/or modify
1674 +# it under the terms of the GNU General Public License as published by
1675 +# the Free Software Foundation; either version 2 of the License, or
1676 +# (at your option) any later version.
1678 +# This program is distributed in the hope that it will be useful,
1679 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1680 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1681 +# GNU General Public License for more details.
1683 +# You should have received a copy of the GNU General Public License
1684 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1686 +Branch Manipulation
1688 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1689 +and changing its permission/attribute, there are a lot of works to do.
1693 +----------------------------------------------------------------------
1694 +o Confirm the adding dir exists outside of aufs, including loopback
1695 + mount, and its various attributes.
1696 +o Initialize the xino file and whiteout bases if necessary.
1699 +o Check the owner/group/mode of the directory
1700 + When the owner/group/mode of the adding directory differs from the
1701 + existing branch, aufs issues a warning because it may impose a
1703 + For example, when a upper writable branch has a world writable empty
1704 + top directory, a malicious user can create any files on the writable
1705 + branch directly, like copy-up and modify manually. If something like
1706 + /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1707 + writable branch, and the writable branch is world-writable, then a
1708 + malicious guy may create /etc/passwd on the writable branch directly
1709 + and the infected file will be valid in aufs.
1710 + I am afraid it can be a security issue, but aufs can do nothing except
1711 + producing a warning.
1715 +----------------------------------------------------------------------
1716 +o Confirm the deleting branch is not busy
1717 + To be general, there is one merit to adopt "remount" interface to
1718 + manipulate branches. It is to discard caches. At deleting a branch,
1719 + aufs checks the still cached (and connected) dentries and inodes. If
1720 + there are any, then they are all in-use. An inode without its
1721 + corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1723 + For the cached one, aufs checks whether the same named entry exists on
1725 + If the cached one is a directory, because aufs provides a merged view
1726 + to users, as long as one dir is left on any branch aufs can show the
1727 + dir to users. In this case, the branch can be removed from aufs.
1728 + Otherwise aufs rejects deleting the branch.
1730 + If any file on the deleting branch is opened by aufs, then aufs
1734 +Modify the Permission of a Branch
1735 +----------------------------------------------------------------------
1736 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1739 +o rw --> ro: Confirm the modifying branch is not busy
1740 + Aufs rejects the request if any of these conditions are true.
1741 + - a file on the branch is mmap-ed.
1742 + - a regular file on the branch is opened for write and there is no
1743 + same named entry on the upper branch.
1744 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1745 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
1746 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2016-01-13 20:11:11.663093444 +0100
1749 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1751 +# This program is free software; you can redistribute it and/or modify
1752 +# it under the terms of the GNU General Public License as published by
1753 +# the Free Software Foundation; either version 2 of the License, or
1754 +# (at your option) any later version.
1756 +# This program is distributed in the hope that it will be useful,
1757 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1758 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1759 +# GNU General Public License for more details.
1761 +# You should have received a copy of the GNU General Public License
1762 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1764 +Policies to Select One among Multiple Writable Branches
1765 +----------------------------------------------------------------------
1766 +When the number of writable branch is more than one, aufs has to decide
1767 +the target branch for file creation or copy-up. By default, the highest
1768 +writable branch which has the parent (or ancestor) dir of the target
1769 +file is chosen (top-down-parent policy).
1770 +By user's request, aufs implements some other policies to select the
1771 +writable branch, for file creation several policies, round-robin,
1772 +most-free-space, and other policies. For copy-up, top-down-parent,
1773 +bottom-up-parent, bottom-up and others.
1775 +As expected, the round-robin policy selects the branch in circular. When
1776 +you have two writable branches and creates 10 new files, 5 files will be
1777 +created for each branch. mkdir(2) systemcall is an exception. When you
1778 +create 10 new directories, all will be created on the same branch.
1779 +And the most-free-space policy selects the one which has most free
1780 +space among the writable branches. The amount of free space will be
1781 +checked by aufs internally, and users can specify its time interval.
1783 +The policies for copy-up is more simple,
1784 +top-down-parent is equivalent to the same named on in create policy,
1785 +bottom-up-parent selects the writable branch where the parent dir
1786 +exists and the nearest upper one from the copyup-source,
1787 +bottom-up selects the nearest upper writable branch from the
1788 +copyup-source, regardless the existence of the parent dir.
1790 +There are some rules or exceptions to apply these policies.
1791 +- If there is a readonly branch above the policy-selected branch and
1792 + the parent dir is marked as opaque (a variation of whiteout), or the
1793 + target (creating) file is whiteout-ed on the upper readonly branch,
1794 + then the result of the policy is ignored and the target file will be
1795 + created on the nearest upper writable branch than the readonly branch.
1796 +- If there is a writable branch above the policy-selected branch and
1797 + the parent dir is marked as opaque or the target file is whiteouted
1798 + on the branch, then the result of the policy is ignored and the target
1799 + file will be created on the highest one among the upper writable
1800 + branches who has diropq or whiteout. In case of whiteout, aufs removes
1802 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1803 + They try selecting the branch where the source exists as possible
1804 + since copyup a large file will take long time. If it can't be,
1805 + ie. the branch where the source exists is readonly, then they will
1806 + follow the copyup policy.
1807 +- There is an exception for rename(2) when the target exists.
1808 + If the rename target exists, aufs compares the index of the branches
1809 + where the source and the target exists and selects the higher
1810 + one. If the selected branch is readonly, then aufs follows the
1812 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1813 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt 1970-01-01 01:00:00.000000000 +0100
1814 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt 2016-01-13 20:11:11.663093444 +0100
1817 +# Copyright (C) 2011-2015 Junjiro R. Okajima
1819 +# This program is free software; you can redistribute it and/or modify
1820 +# it under the terms of the GNU General Public License as published by
1821 +# the Free Software Foundation; either version 2 of the License, or
1822 +# (at your option) any later version.
1824 +# This program is distributed in the hope that it will be useful,
1825 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1826 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1827 +# GNU General Public License for more details.
1829 +# You should have received a copy of the GNU General Public License
1830 +# along with this program; if not, write to the Free Software
1831 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1834 +File-based Hierarchical Storage Management (FHSM)
1835 +----------------------------------------------------------------------
1836 +Hierarchical Storage Management (or HSM) is a well-known feature in the
1837 +storage world. Aufs provides this feature as file-based with multiple
1838 +writable branches, based upon the principle of "Colder, the Lower".
1839 +Here the word "colder" means that the less used files, and "lower" means
1840 +that the position in the order of the stacked branches vertically.
1841 +These multiple writable branches are prioritized, ie. the topmost one
1842 +should be the fastest drive and be used heavily.
1844 +o Characters in aufs FHSM story
1845 +- aufs itself and a new branch attribute.
1846 +- a new ioctl interface to move-down and to establish a connection with
1847 + the daemon ("move-down" is a converse of "copy-up").
1848 +- userspace tool and daemon.
1850 +The userspace daemon establishes a connection with aufs and waits for
1851 +the notification. The notified information is very similar to struct
1852 +statfs containing the number of consumed blocks and inodes.
1853 +When the consumed blocks/inodes of a branch exceeds the user-specified
1854 +upper watermark, the daemon activates its move-down process until the
1855 +consumed blocks/inodes reaches the user-specified lower watermark.
1857 +The actual move-down is done by aufs based upon the request from
1858 +user-space since we need to maintain the inode number and the internal
1859 +pointer arrays in aufs.
1861 +Currently aufs FHSM handles the regular files only. Additionally they
1862 +must not be hard-linked nor pseudo-linked.
1865 +o Cowork of aufs and the user-space daemon
1866 + During the userspace daemon established the connection, aufs sends a
1867 + small notification to it whenever aufs writes something into the
1868 + writable branch. But it may cost high since aufs issues statfs(2)
1869 + internally. So user can specify a new option to cache the
1870 + info. Actually the notification is controlled by these factors.
1871 + + the specified cache time.
1872 + + classified as "force" by aufs internally.
1873 + Until the specified time expires, aufs doesn't send the info
1874 + except the forced cases. When aufs decide forcing, the info is always
1875 + notified to userspace.
1876 + For example, the number of free inodes is generally large enough and
1877 + the shortage of it happens rarely. So aufs doesn't force the
1878 + notification when creating a new file, directory and others. This is
1879 + the typical case which aufs doesn't force.
1880 + When aufs writes the actual filedata and the files consumes any of new
1881 + blocks, the aufs forces notifying.
1884 +o Interfaces in aufs
1885 +- New branch attribute.
1887 + Specifies that the branch is managed by FHSM feature. In other word,
1888 + participant in the FHSM.
1889 + When nofhsm is set to the branch, it will not be the source/target
1890 + branch of the move-down operation. This attribute is set
1891 + independently from coo and moo attributes, and if you want full
1892 + FHSM, you should specify them as well.
1893 +- New mount option.
1895 + Specifies a second to suppress many less important info to be
1898 + + AUFS_CTL_FHSM_FD
1899 + create a new file descriptor which userspace can read the notification
1900 + (a subset of struct statfs) from aufs.
1901 +- Module parameter 'brs'
1902 + It has to be set to 1. Otherwise the new mount option 'fhsm' will not
1904 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
1905 + When there are two or more branches with fhsm attributes,
1906 + /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
1907 + terminates it. As a result of remounting and branch-manipulation, the
1908 + number of branches with fhsm attribute can be one. In this case,
1909 + /sbin/mount.aufs will terminate the user-space daemon.
1912 +Finally the operation is done as these steps in kernel-space.
1914 + + no one else is using the file.
1915 + + the file is not hard-linked.
1916 + + the file is not pseudo-linked.
1917 + + the file is a regular file.
1918 + + the parent dir is not opaqued.
1919 +- find the target writable branch.
1920 +- make sure the file is not whiteout-ed by the upper (than the target)
1922 +- make the parent dir on the target branch.
1923 +- mutex lock the inode on the branch.
1924 +- unlink the whiteout on the target branch (if exists).
1925 +- lookup and create the whiteout-ed temporary name on the target branch.
1926 +- copy the file as the whiteout-ed temporary name on the target branch.
1927 +- rename the whiteout-ed temporary name to the original name.
1928 +- unlink the file on the source branch.
1929 +- maintain the internal pointer array and the external inode number
1931 +- maintain the timestamps and other attributes of the parent dir and the
1934 +And of course, in every step, an error may happen. So the operation
1935 +should restore the original file state after an error happens.
1936 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
1937 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
1938 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2016-01-13 20:11:11.663093444 +0100
1941 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1943 +# This program is free software; you can redistribute it and/or modify
1944 +# it under the terms of the GNU General Public License as published by
1945 +# the Free Software Foundation; either version 2 of the License, or
1946 +# (at your option) any later version.
1948 +# This program is distributed in the hope that it will be useful,
1949 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1950 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1951 +# GNU General Public License for more details.
1953 +# You should have received a copy of the GNU General Public License
1954 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1956 +mmap(2) -- File Memory Mapping
1957 +----------------------------------------------------------------------
1958 +In aufs, the file-mapped pages are handled by a branch fs directly, no
1959 +interaction with aufs. It means aufs_mmap() calls the branch fs's
1961 +This approach is simple and good, but there is one problem.
1962 +Under /proc, several entries show the mmapped files by its path (with
1963 +device and inode number), and the printed path will be the path on the
1964 +branch fs's instead of virtual aufs's.
1965 +This is not a problem in most cases, but some utilities lsof(1) (and its
1966 +user) may expect the path on aufs.
1968 +To address this issue, aufs adds a new member called vm_prfile in struct
1969 +vm_area_struct (and struct vm_region). The original vm_file points to
1970 +the file on the branch fs in order to handle everything correctly as
1971 +usual. The new vm_prfile points to a virtual file in aufs, and the
1972 +show-functions in procfs refers to vm_prfile if it is set.
1973 +Also we need to maintain several other places where touching vm_file
1975 +- fork()/clone() copies vma and the reference count of vm_file is
1977 +- merging vma maintains the ref count too.
1979 +This is not a good approach. It just fakes the printed path. But it
1980 +leaves all behaviour around f_mapping unchanged. This is surely an
1982 +Actually aufs had adopted another complicated approach which calls
1983 +generic_file_mmap() and handles struct vm_operations_struct. In this
1984 +approach, aufs met a hard problem and I could not solve it without
1985 +switching the approach.
1987 +There may be one more another approach which is
1988 +- bind-mount the branch-root onto the aufs-root internally
1989 +- grab the new vfsmount (ie. struct mount)
1990 +- lazy-umount the branch-root internally
1991 +- in open(2) the aufs-file, open the branch-file with the hidden
1992 + vfsmount (instead of the original branch's vfsmount)
1993 +- ideally this "bind-mount and lazy-umount" should be done atomically,
1994 + but it may be possible from userspace by the mount helper.
1996 +Adding the internal hidden vfsmount and using it in opening a file, the
1997 +file path under /proc will be printed correctly. This approach looks
1998 +smarter, but is not possible I am afraid.
1999 +- aufs-root may be bind-mount later. when it happens, another hidden
2000 + vfsmount will be required.
2001 +- it is hard to get the chance to bind-mount and lazy-umount
2002 + + in kernel-space, FS can have vfsmount in open(2) via
2003 + file->f_path, and aufs can know its vfsmount. But several locks are
2004 + already acquired, and if aufs tries to bind-mount and lazy-umount
2005 + here, then it may cause a deadlock.
2006 + + in user-space, bind-mount doesn't invoke the mount helper.
2007 +- since /proc shows dev and ino, aufs has to give vma these info. it
2008 + means a new member vm_prinode will be necessary. this is essentially
2009 + equivalent to vm_prfile described above.
2011 +I have to give up this "looks-smater" approach.
2012 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2013 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt 1970-01-01 01:00:00.000000000 +0100
2014 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt 2016-01-13 20:11:11.663093444 +0100
2017 +# Copyright (C) 2014-2015 Junjiro R. Okajima
2019 +# This program is free software; you can redistribute it and/or modify
2020 +# it under the terms of the GNU General Public License as published by
2021 +# the Free Software Foundation; either version 2 of the License, or
2022 +# (at your option) any later version.
2024 +# This program is distributed in the hope that it will be useful,
2025 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2026 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2027 +# GNU General Public License for more details.
2029 +# You should have received a copy of the GNU General Public License
2030 +# along with this program; if not, write to the Free Software
2031 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2034 +Listing XATTR/EA and getting the value
2035 +----------------------------------------------------------------------
2036 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2037 +shows the values from the topmost existing file. This behaviour is good
2038 +for the non-dir entries since the bahaviour exactly matches the shown
2039 +information. But for the directories, aufs considers all the same named
2040 +entries on the lower branches. Which means, if one of the lower entry
2041 +rejects readdir call, then aufs returns an error even if the topmost
2042 +entry allows it. This behaviour is necessary to respect the branch fs's
2043 +security, but can make users confused since the user-visible standard
2044 +attributes don't match the behaviour.
2045 +To address this issue, aufs has a mount option called dirperm1 which
2046 +checks the permission for the topmost entry only, and ignores the lower
2047 +entry's permission.
2049 +A similar issue can happen around XATTR.
2050 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2051 +always set. Otherwise these very unpleasant situation would happen.
2052 +- listxattr(2) may return the duplicated entries.
2053 +- users may not be able to remove or reset the XATTR forever,
2056 +XATTR/EA support in the internal (copy,move)-(up,down)
2057 +----------------------------------------------------------------------
2058 +Generally the extended attributes of inode are categorized as these.
2059 +- "security" for LSM and capability.
2060 +- "system" for posix ACL, 'acl' mount option is required for the branch
2062 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2063 +- "user" for userspace, 'user_xattr' mount option is required for the
2064 + branch fs generally.
2066 +Moreover there are some other categories. Aufs handles these rather
2067 +unpopular categories as the ordinary ones, ie. there is no special
2068 +condition nor exception.
2070 +In copy-up, the support for XATTR on the dst branch may differ from the
2071 +src branch. In this case, the copy-up operation will get an error and
2072 +the original user operation which triggered the copy-up will fail. It
2073 +can happen that even all copy-up will fail.
2074 +When both of src and dst branches support XATTR and if an error occurs
2075 +during copying XATTR, then the copy-up should fail obviously. That is a
2076 +good reason and aufs should return an error to userspace. But when only
2077 +the src branch support that XATTR, aufs should not return an error.
2078 +For example, the src branch supports ACL but the dst branch doesn't
2079 +because the dst branch may natively un-support it or temporary
2080 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2081 +may NOT return an error even if the XATTR is not supported. It is
2082 +totally up to the branch fs.
2084 +Anyway when the aufs internal copy-up gets an error from the dst branch
2085 +fs, then aufs tries removing the just copied entry and returns the error
2086 +to the userspace. The worst case of this situation will be all copy-up
2089 +For the copy-up operation, there two basic approaches.
2090 +- copy the specified XATTR only (by category above), and return the
2091 + error unconditionally if it happens.
2092 +- copy all XATTR, and ignore the error on the specified category only.
2094 +In order to support XATTR and to implement the correct behaviour, aufs
2095 +chooses the latter approach and introduces some new branch attributes,
2096 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2097 +They correspond to the XATTR namespaces (see above). Additionally, to be
2098 +convenient, "icex" is also provided which means all "icex*" attributes
2099 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2101 +The meaning of these attributes is to ignore the error from setting
2102 +XATTR on that branch.
2103 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2104 +error from the dst branch according to the specified attributes.
2106 +Some XATTR may have its default value. The default value may come from
2107 +the parent dir or the environment. If the default value is set at the
2108 +file creating-time, it will be overwritten by copy-up.
2109 +Some contradiction may happen I am afraid.
2110 +Do we need another attribute to stop copying XATTR? I am unsure. For
2111 +now, aufs implements the branch attributes to ignore the error.
2112 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2113 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2114 +++ linux/Documentation/filesystems/aufs/design/07export.txt 2016-01-13 20:11:11.663093444 +0100
2117 +# Copyright (C) 2005-2015 Junjiro R. Okajima
2119 +# This program is free software; you can redistribute it and/or modify
2120 +# it under the terms of the GNU General Public License as published by
2121 +# the Free Software Foundation; either version 2 of the License, or
2122 +# (at your option) any later version.
2124 +# This program is distributed in the hope that it will be useful,
2125 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2126 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2127 +# GNU General Public License for more details.
2129 +# You should have received a copy of the GNU General Public License
2130 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2132 +Export Aufs via NFS
2133 +----------------------------------------------------------------------
2134 +Here is an approach.
2135 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2137 +- iget_locked(): initialize aufs inode generation for a new inode, and
2138 + store it in xigen file.
2139 +- destroy_inode(): increment aufs inode generation and store it in xigen
2140 + file. it is necessary even if it is not unlinked, because any data of
2141 + inode may be changed by UDBA.
2142 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2143 + build file handle by
2144 + + branch id (4 bytes)
2145 + + superblock generation (4 bytes)
2146 + + inode number (4 or 8 bytes)
2147 + + parent dir inode number (4 or 8 bytes)
2148 + + inode generation (4 bytes))
2149 + + return value of exportfs_encode_fh() for the parent on a branch (4
2151 + + file handle for a branch (by exportfs_encode_fh())
2153 + + find the index of a branch from its id in handle, and check it is
2154 + still exist in aufs.
2155 + + 1st level: get the inode number from handle and search it in cache.
2156 + + 2nd level: if not found in cache, get the parent inode number from
2157 + the handle and search it in cache. and then open the found parent
2158 + dir, find the matching inode number by vfs_readdir() and get its
2159 + name, and call lookup_one_len() for the target dentry.
2160 + + 3rd level: if the parent dir is not cached, call
2161 + exportfs_decode_fh() for a branch and get the parent on a branch,
2162 + build a pathname of it, convert it a pathname in aufs, call
2163 + path_lookup(). now aufs gets a parent dir dentry, then handle it as
2165 + + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2166 + for every branch, but not itself. to get this, (currently) aufs
2167 + searches in current->nsproxy->mnt_ns list. it may not be a good
2168 + idea, but I didn't get other approach.
2169 + + test the generation of the gotten inode.
2170 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2171 + convert it into ESTALE for NFSD.
2172 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2173 + lookup_one_len(), vfs_getattr(), encode_fh() and others.
2174 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2175 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
2176 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2016-01-13 20:11:11.663093444 +0100
2179 +# Copyright (C) 2005-2015 Junjiro R. Okajima
2181 +# This program is free software; you can redistribute it and/or modify
2182 +# it under the terms of the GNU General Public License as published by
2183 +# the Free Software Foundation; either version 2 of the License, or
2184 +# (at your option) any later version.
2186 +# This program is distributed in the hope that it will be useful,
2187 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2188 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2189 +# GNU General Public License for more details.
2191 +# You should have received a copy of the GNU General Public License
2192 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2194 +Show Whiteout Mode (shwh)
2195 +----------------------------------------------------------------------
2196 +Generally aufs hides the name of whiteouts. But in some cases, to show
2197 +them is very useful for users. For instance, creating a new middle layer
2198 +(branch) by merging existing layers.
2200 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2201 +When you have three branches,
2202 +- Bottom: 'system', squashfs (underlying base system), read-only
2203 +- Middle: 'mods', squashfs, read-only
2204 +- Top: 'overlay', ram (tmpfs), read-write
2206 +The top layer is loaded at boot time and saved at shutdown, to preserve
2207 +the changes made to the system during the session.
2208 +When larger changes have been made, or smaller changes have accumulated,
2209 +the size of the saved top layer data grows. At this point, it would be
2210 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2211 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2212 +restoring save and load speed.
2214 +This merging is simplified by the use of another aufs mount, of just the
2215 +two overlay branches using the 'shwh' option.
2216 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2217 + aufs /livesys/merge_union
2219 +A merged view of these two branches is then available at
2220 +/livesys/merge_union, and the new feature is that the whiteouts are
2222 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2223 +writing to all branches. Also the default mode for all branches is 'ro'.
2224 +It is now possible to save the combined contents of the two overlay
2225 +branches to a new squashfs, e.g.:
2226 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2228 +This new squashfs archive can be stored on the boot device and the
2229 +initramfs will use it to replace the old one at the next boot.
2230 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2231 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
2232 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2016-01-13 20:11:11.663093444 +0100
2235 +# Copyright (C) 2010-2015 Junjiro R. Okajima
2237 +# This program is free software; you can redistribute it and/or modify
2238 +# it under the terms of the GNU General Public License as published by
2239 +# the Free Software Foundation; either version 2 of the License, or
2240 +# (at your option) any later version.
2242 +# This program is distributed in the hope that it will be useful,
2243 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2244 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2245 +# GNU General Public License for more details.
2247 +# You should have received a copy of the GNU General Public License
2248 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2250 +Dynamically customizable FS operations
2251 +----------------------------------------------------------------------
2252 +Generally FS operations (struct inode_operations, struct
2253 +address_space_operations, struct file_operations, etc.) are defined as
2254 +"static const", but it never means that FS have only one set of
2255 +operation. Some FS have multiple sets of them. For instance, ext2 has
2256 +three sets, one for XIP, for NOBH, and for normal.
2257 +Since aufs overrides and redirects these operations, sometimes aufs has
2258 +to change its behaviour according to the branch FS type. More importantly
2259 +VFS acts differently if a function (member in the struct) is set or
2260 +not. It means aufs should have several sets of operations and select one
2261 +among them according to the branch FS definition.
2263 +In order to solve this problem and not to affect the behaviour of VFS,
2264 +aufs defines these operations dynamically. For instance, aufs defines
2265 +dummy direct_IO function for struct address_space_operations, but it may
2266 +not be set to the address_space_operations actually. When the branch FS
2267 +doesn't have it, aufs doesn't set it to its address_space_operations
2268 +while the function definition itself is still alive. So the behaviour
2269 +itself will not change, and it will return an error when direct_IO is
2272 +The lifetime of these dynamically generated operation object is
2273 +maintained by aufs branch object. When the branch is removed from aufs,
2274 +the reference counter of the object is decremented. When it reaches
2275 +zero, the dynamically generated operation object will be freed.
2277 +This approach is designed to support AIO (io_submit), Direct I/O and
2279 +Currently this approach is applied to address_space_operations for
2280 +regular files only.
2281 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2282 --- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
2283 +++ linux/Documentation/filesystems/aufs/README 2016-01-13 20:11:11.663093444 +0100
2286 +Aufs4 -- advanced multi layered unification filesystem version 4.x
2292 +----------------------------------------
2293 +In the early days, aufs was entirely re-designed and re-implemented
2294 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2295 +improvements and implementations, it becomes totally different from
2296 +Unionfs while keeping the basic features.
2297 +Recently, Unionfs Version 2.x series begin taking some of the same
2298 +approaches to aufs1's.
2299 +Unionfs is being developed by Professor Erez Zadok at Stony Brook
2300 +University and his team.
2302 +Aufs4 supports linux-4.0 and later, and for linux-3.x series try aufs3.
2303 +If you want older kernel version support, try aufs2-2.6.git or
2304 +aufs2-standalone.git repository, aufs1 from CVS on SourceForge.
2306 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2307 + According to Christoph Hellwig, linux rejects all union-type
2308 + filesystems but UnionMount.
2309 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2311 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2312 + UnionMount, and he pointed out an issue around a directory mutex
2313 + lock and aufs addressed it. But it is still unsure whether aufs will
2314 + be merged (or any other union solution).
2315 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2319 +----------------------------------------
2320 +- unite several directories into a single virtual filesystem. The member
2321 + directory is called as a branch.
2322 +- you can specify the permission flags to the branch, which are 'readonly',
2323 + 'readwrite' and 'whiteout-able.'
2324 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2325 + readonly branch are modifiable logically.
2326 +- dynamic branch manipulation, add, del.
2329 +Also there are many enhancements in aufs, such as:
2330 +- test only the highest one for the directory permission (dirperm1)
2331 +- copyup on open (coo=)
2332 +- 'move' policy for copy-up between two writable branches, after
2333 + checking free space.
2335 +- readdir(3) in userspace.
2336 +- keep inode number by external inode number table
2337 +- keep the timestamps of file/dir in internal copyup operation
2338 +- seekable directory, supporting NFS readdir.
2339 +- whiteout is hardlinked in order to reduce the consumption of inodes
2341 +- do not copyup, nor create a whiteout when it is unnecessary
2342 +- revert a single systemcall when an error occurs in aufs
2343 +- remount interface instead of ioctl
2344 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2345 +- loopback mounted filesystem as a branch
2346 +- kernel thread for removing the dir who has a plenty of whiteouts
2347 +- support copyup sparse file (a file which has a 'hole' in it)
2348 +- default permission flags for branches
2349 +- selectable permission flags for ro branch, whether whiteout can
2352 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2353 +- support multiple writable branches, some policies to select one
2354 + among multiple writable branches.
2355 +- a new semantics for link(2) and rename(2) to support multiple
2356 + writable branches.
2357 +- no glibc changes are required.
2358 +- pseudo hardlink (hardlink over branches)
2359 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2360 + including NFS or remote filesystem branch.
2361 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2364 +Currently these features are dropped temporary from aufs4.
2365 +See design/08plan.txt in detail.
2366 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2368 +- statistics of aufs thread (/sys/fs/aufs/stat)
2370 +Features or just an idea in the future (see also design/*.txt),
2371 +- reorder the branch index without del/re-add.
2372 +- permanent xino files for NFSD
2373 +- an option for refreshing the opened files after add/del branches
2374 +- light version, without branch manipulation. (unnecessary?)
2375 +- copyup in userspace
2376 +- inotify in userspace
2381 +----------------------------------------
2382 +There are three GIT trees for aufs4, aufs4-linux.git,
2383 +aufs4-standalone.git, and aufs-util.git. Note that there is no "4" in
2385 +While the aufs-util is always necessary, you need either of aufs4-linux
2386 +or aufs4-standalone.
2388 +The aufs4-linux tree includes the whole linux mainline GIT tree,
2389 +git://git.kernel.org/.../torvalds/linux.git.
2390 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2391 +build aufs4 as an external kernel module.
2392 +Several extra patches are not included in this tree. Only
2393 +aufs4-standalone tree contains them. They are described in the later
2394 +section "Configuration and Compilation."
2396 +On the other hand, the aufs4-standalone tree has only aufs source files
2397 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2398 +But you need to apply all aufs patches manually.
2400 +You will find GIT branches whose name is in form of "aufs4.x" where "x"
2401 +represents the linux kernel version, "linux-4.x". For instance,
2402 +"aufs4.0" is for linux-4.0. For latest "linux-4.x-rcN", use
2403 +"aufs4.x-rcN" branch.
2406 +$ git clone --reference /your/linux/git/tree \
2407 + git://github.com/sfjro/aufs4-linux.git aufs4-linux.git
2408 +- if you don't have linux GIT tree, then remove "--reference ..."
2409 +$ cd aufs4-linux.git
2410 +$ git checkout origin/aufs4.0
2412 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2413 +leave the patch-work to GIT.
2414 +$ cd /your/linux/git/tree
2415 +$ git remote add aufs4 git://github.com/sfjro/aufs4-linux.git
2417 +$ git checkout -b my4.0 v4.0
2418 +$ (add your local change...)
2419 +$ git pull aufs4 aufs4.0
2420 +- now you have v4.0 + your_changes + aufs4.0 in you my4.0 branch.
2421 +- you may need to solve some conflicts between your_changes and
2422 + aufs4.0. in this case, git-rerere is recommended so that you can
2423 + solve the similar conflicts automatically when you upgrade to 4.1 or
2424 + later in the future.
2426 +o aufs4-standalone tree
2427 +$ git clone git://github.com/sfjro/aufs4-standalone.git aufs4-standalone.git
2428 +$ cd aufs4-standalone.git
2429 +$ git checkout origin/aufs4.0
2432 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2433 +- note that the public aufs-util.git is on SourceForge instead of
2436 +$ git checkout origin/aufs4.0
2438 +Note: The 4.x-rcN branch is to be used with `rc' kernel versions ONLY.
2439 +The minor version number, 'x' in '4.x', of aufs may not always
2440 +follow the minor version number of the kernel.
2441 +Because changes in the kernel that cause the use of a new
2442 +minor version number do not always require changes to aufs-util.
2444 +Since aufs-util has its own minor version number, you may not be
2445 +able to find a GIT branch in aufs-util for your kernel's
2446 +exact minor version number.
2447 +In this case, you should git-checkout the branch for the
2448 +nearest lower number.
2450 +For (an unreleased) example:
2451 +If you are using "linux-4.10" and the "aufs4.10" branch
2452 +does not exist in aufs-util repository, then "aufs4.9", "aufs4.8"
2453 +or something numerically smaller is the branch for your kernel.
2455 +Also you can view all branches by
2459 +3. Configuration and Compilation
2460 +----------------------------------------
2461 +Make sure you have git-checkout'ed the correct branch.
2463 +For aufs4-linux tree,
2464 +- enable CONFIG_AUFS_FS.
2465 +- set other aufs configurations if necessary.
2467 +For aufs4-standalone tree,
2468 +There are several ways to build.
2471 +- apply ./aufs4-kbuild.patch to your kernel source files.
2472 +- apply ./aufs4-base.patch too.
2473 +- apply ./aufs4-mmap.patch too.
2474 +- apply ./aufs4-standalone.patch too, if you have a plan to set
2475 + CONFIG_AUFS_FS=m. otherwise you don't need ./aufs4-standalone.patch.
2476 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2477 + kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2478 +- enable CONFIG_AUFS_FS, you can select either
2480 +- and build your kernel as usual.
2481 +- install the built kernel.
2482 + Note: Since linux-3.9, every filesystem module requires an alias
2483 + "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2484 + modules.aliases file if you set CONFIG_AUFS_FS=m.
2485 +- install the header files too by "make headers_install" to the
2486 + directory where you specify. By default, it is $PWD/usr.
2487 + "make help" shows a brief note for headers_install.
2488 +- and reboot your system.
2491 +- module only (CONFIG_AUFS_FS=m).
2492 +- apply ./aufs4-base.patch to your kernel source files.
2493 +- apply ./aufs4-mmap.patch too.
2494 +- apply ./aufs4-standalone.patch too.
2495 +- build your kernel, don't forget "make headers_install", and reboot.
2496 +- edit ./config.mk and set other aufs configurations if necessary.
2497 + Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2498 + every aufs configurations.
2499 +- build the module by simple "make".
2500 + Note: Since linux-3.9, every filesystem module requires an alias
2501 + "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2502 + modules.aliases file.
2503 +- you can specify ${KDIR} make variable which points to your kernel
2505 +- install the files
2506 + + run "make install" to install the aufs module, or copy the built
2507 + $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2508 + + run "make install_headers" (instead of headers_install) to install
2509 + the modified aufs header file (you can specify DESTDIR which is
2510 + available in aufs standalone version's Makefile only), or copy
2511 + $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2512 + you like manually. By default, the target directory is $PWD/usr.
2513 +- no need to apply aufs4-kbuild.patch, nor copying source files to your
2514 + kernel source tree.
2516 +Note: The header file aufs_type.h is necessary to build aufs-util
2517 + as well as "make headers_install" in the kernel source tree.
2518 + headers_install is subject to be forgotten, but it is essentially
2519 + necessary, not only for building aufs-util.
2520 + You may not meet problems without headers_install in some older
2524 +- read README in aufs-util, build and install it
2525 +- note that your distribution may contain an obsoleted version of
2526 + aufs_type.h in /usr/include/linux or something. When you build aufs
2527 + utilities, make sure that your compiler refers the correct aufs header
2528 + file which is built by "make headers_install."
2529 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2530 + then run "make install_ulib" too. And refer to the aufs manual in
2533 +There several other patches in aufs4-standalone.git. They are all
2534 +optional. When you meet some problems, they will help you.
2535 +- aufs4-loopback.patch
2536 + Supports a nested loopback mount in a branch-fs. This patch is
2537 + unnecessary until aufs produces a message like "you may want to try
2538 + another patch for loopback file".
2540 + Modifies a system global kernel internal function get_next_ino() in
2541 + order to stop assigning 0 for an inode-number. Not directly related to
2542 + aufs, but recommended generally.
2544 + Keeps the tmpfs inode number as the lowest value. Effective to reduce
2545 + the size of aufs XINO files for tmpfs branch. Also it prevents the
2546 + duplication of inode number, which is important for backup tools and
2547 + other utilities. When you find aufs XINO files for tmpfs branch
2548 + growing too much, try this patch.
2549 +- lockdep-debug.patch
2550 + Because aufs is not only an ordinary filesystem (callee of VFS), but
2551 + also a caller of VFS functions for branch filesystems, subclassing of
2552 + the internal locks for LOCKDEP is necessary. LOCKDEP is a debugging
2553 + feature of linux kernel. If you enable CONFIG_LOCKDEP, then you will
2554 + need to apply this debug patch to expand several constant values.
2555 + If don't know what LOCKDEP, then you don't have apply this patch.
2559 +----------------------------------------
2560 +At first, make sure aufs-util are installed, and please read the aufs
2561 +manual, aufs.5 in aufs-util.git tree.
2565 +$ mkdir /tmp/rw /tmp/aufs
2566 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2568 +Here is another example. The result is equivalent.
2569 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2571 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2572 +# mount -o remount,append:${HOME} /tmp/aufs
2574 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2575 +you modify a file under /tmp/aufs, the one on your home directory is
2576 +not affected, instead the same named file will be newly created under
2577 +/tmp/rw. And all of your modification to a file will be applied to
2578 +the one under /tmp/rw. This is called the file based Copy on Write
2580 +Aufs mount options are described in aufs.5.
2581 +If you run chroot or something and make your aufs as a root directory,
2582 +then you need to customize the shutdown script. See the aufs manual in
2585 +Additionally, there are some sample usages of aufs which are a
2586 +diskless system with network booting, and LiveCD over NFS.
2587 +See sample dir in CVS tree on SourceForge.
2591 +----------------------------------------
2592 +When you have any problems or strange behaviour in aufs, please let me
2594 +- /proc/mounts (instead of the output of mount(8))
2595 +- /sys/module/aufs/*
2596 +- /sys/fs/aufs/* (if you have them)
2597 +- /debug/aufs/* (if you have them)
2598 +- linux kernel version
2599 + if your kernel is not plain, for example modified by distributor,
2600 + the url where i can download its source is necessary too.
2601 +- aufs version which was printed at loading the module or booting the
2602 + system, instead of the date you downloaded.
2603 +- configuration (define/undefine CONFIG_AUFS_xxx)
2604 +- kernel configuration or /proc/config.gz (if you have it)
2605 +- behaviour which you think to be incorrect
2606 +- actual operation, reproducible one is better
2607 +- mailto: aufs-users at lists.sourceforge.net
2609 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2610 +and Feature Requests) on SourceForge. Please join and write to
2614 +6. Acknowledgements
2615 +----------------------------------------
2616 +Thanks to everyone who have tried and are using aufs, whoever
2617 +have reported a bug or any feedback.
2619 +Especially donators:
2620 +Tomas Matejicek(slax.org) made a donation (much more than once).
2621 + Since Apr 2010, Tomas M (the author of Slax and Linux Live
2622 + scripts) is making "doubling" donations.
2623 + Unfortunately I cannot list all of the donators, but I really
2625 + It ends Aug 2010, but the ordinary donation URL is still available.
2626 + <http://sourceforge.net/donate/index.php?group_id=167503>
2627 +Dai Itasaka made a donation (2007/8).
2628 +Chuck Smith made a donation (2008/4, 10 and 12).
2629 +Henk Schoneveld made a donation (2008/9).
2630 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2631 +Francois Dupoux made a donation (2008/11).
2632 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2633 + aufs2 GIT tree (2009/2).
2634 +William Grant made a donation (2009/3).
2635 +Patrick Lane made a donation (2009/4).
2636 +The Mail Archive (mail-archive.com) made donations (2009/5).
2637 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2638 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2639 +Pavel Pronskiy made a donation (2011/2).
2640 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2641 + Networks (Ed Wildgoose) made a donation for hardware (2011/3).
2642 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2644 +Sam Liddicott made a donation (2011/9).
2645 +Era Scarecrow made a donation (2013/4).
2646 +Bor Ratajc made a donation (2013/4).
2647 +Alessandro Gorreta made a donation (2013/4).
2648 +POIRETTE Marc made a donation (2013/4).
2649 +Alessandro Gorreta made a donation (2013/4).
2650 +lauri kasvandik made a donation (2013/5).
2651 +"pemasu from Finland" made a donation (2013/7).
2652 +The Parted Magic Project made a donation (2013/9 and 11).
2653 +Pavel Barta made a donation (2013/10).
2654 +Nikolay Pertsev made a donation (2014/5).
2655 +James B made a donation (2014/7 and 2015/7).
2656 +Stefano Di Biase made a donation (2014/8).
2657 +Daniel Epellei made a donation (2015/1).
2659 +Thank you very much.
2660 +Donations are always, including future donations, very important and
2661 +helpful for me to keep on developing aufs.
2665 +----------------------------------------
2666 +If you are an experienced user, no explanation is needed. Aufs is
2667 +just a linux filesystem.
2672 +# Local variables: ;
2675 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2676 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
2677 +++ linux/fs/aufs/aufs.h 2016-01-13 20:11:11.666426853 +0100
2680 + * Copyright (C) 2005-2015 Junjiro R. Okajima
2682 + * This program, aufs is free software; you can redistribute it and/or modify
2683 + * it under the terms of the GNU General Public License as published by
2684 + * the Free Software Foundation; either version 2 of the License, or
2685 + * (at your option) any later version.
2687 + * This program is distributed in the hope that it will be useful,
2688 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2689 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2690 + * GNU General Public License for more details.
2692 + * You should have received a copy of the GNU General Public License
2693 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2697 + * all header files
2705 +#define AuStub(type, name, body, ...) \
2706 + static inline type name(__VA_ARGS__) { body; }
2708 +#define AuStubVoid(name, ...) \
2709 + AuStub(void, name, , __VA_ARGS__)
2710 +#define AuStubInt0(name, ...) \
2711 + AuStub(int, name, return 0, __VA_ARGS__)
2715 +#include "branch.h"
2718 +#include "dbgaufs.h"
2719 +#include "dentry.h"
2723 +#include "fstype.h"
2726 +#include "module.h"
2731 +#include "sysaufs.h"
2736 +#endif /* __KERNEL__ */
2737 +#endif /* __AUFS_H__ */
2738 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2739 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
2740 +++ linux/fs/aufs/branch.c 2016-01-13 20:11:11.666426853 +0100
2743 + * Copyright (C) 2005-2015 Junjiro R. Okajima
2745 + * This program, aufs is free software; you can redistribute it and/or modify
2746 + * it under the terms of the GNU General Public License as published by
2747 + * the Free Software Foundation; either version 2 of the License, or
2748 + * (at your option) any later version.
2750 + * This program is distributed in the hope that it will be useful,
2751 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2752 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2753 + * GNU General Public License for more details.
2755 + * You should have received a copy of the GNU General Public License
2756 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2760 + * branch management
2763 +#include <linux/compat.h>
2764 +#include <linux/statfs.h>
2768 + * free a single branch
2770 +static void au_br_do_free(struct au_branch *br)
2773 + struct au_wbr *wbr;
2774 + struct au_dykey **key;
2776 + au_hnotify_fin_br(br);
2778 + if (br->br_xino.xi_file)
2779 + fput(br->br_xino.xi_file);
2780 + mutex_destroy(&br->br_xino.xi_nondir_mtx);
2782 + AuDebugOn(atomic_read(&br->br_count));
2786 + for (i = 0; i < AuBrWh_Last; i++)
2787 + dput(wbr->wbr_wh[i]);
2788 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2789 + AuRwDestroy(&wbr->wbr_wh_rwsem);
2792 + if (br->br_fhsm) {
2793 + au_br_fhsm_fin(br->br_fhsm);
2794 + kfree(br->br_fhsm);
2797 + key = br->br_dykey;
2798 + for (i = 0; i < AuBrDynOp; i++, key++)
2804 + /* recursive lock, s_umount of branch's */
2806 + path_put(&br->br_path);
2813 + * frees all branches
2815 +void au_br_free(struct au_sbinfo *sbinfo)
2817 + aufs_bindex_t bmax;
2818 + struct au_branch **br;
2820 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2822 + bmax = sbinfo->si_bend + 1;
2823 + br = sbinfo->si_branch;
2825 + au_br_do_free(*br++);
2829 + * find the index of a branch which is specified by @br_id.
2831 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2833 + aufs_bindex_t bindex, bend;
2835 + bend = au_sbend(sb);
2836 + for (bindex = 0; bindex <= bend; bindex++)
2837 + if (au_sbr_id(sb, bindex) == br_id)
2842 +/* ---------------------------------------------------------------------- */
2848 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
2849 + struct dentry *h_root)
2851 + if (unlikely(h_adding == h_root
2852 + || au_test_loopback_overlap(sb, h_adding)))
2854 + if (h_adding->d_sb != h_root->d_sb)
2856 + return au_test_subdir(h_adding, h_root)
2857 + || au_test_subdir(h_root, h_adding);
2861 + * returns a newly allocated branch. @new_nbranch is a number of branches
2862 + * after adding a branch.
2864 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
2867 + struct au_branch *add_branch;
2868 + struct dentry *root;
2869 + struct inode *inode;
2873 + root = sb->s_root;
2874 + add_branch = kzalloc(sizeof(*add_branch), GFP_NOFS);
2875 + if (unlikely(!add_branch))
2878 + err = au_hnotify_init_br(add_branch, perm);
2879 + if (unlikely(err))
2882 + if (au_br_writable(perm)) {
2883 + /* may be freed separately at changing the branch permission */
2884 + add_branch->br_wbr = kzalloc(sizeof(*add_branch->br_wbr),
2886 + if (unlikely(!add_branch->br_wbr))
2890 + if (au_br_fhsm(perm)) {
2891 + err = au_fhsm_br_alloc(add_branch);
2892 + if (unlikely(err))
2896 + err = au_sbr_realloc(au_sbi(sb), new_nbranch);
2898 + err = au_di_realloc(au_di(root), new_nbranch);
2900 + inode = d_inode(root);
2901 + err = au_ii_realloc(au_ii(inode), new_nbranch);
2904 + return add_branch; /* success */
2907 + kfree(add_branch->br_wbr);
2909 + au_hnotify_fin_br(add_branch);
2911 + kfree(add_branch);
2913 + return ERR_PTR(err);
2917 + * test if the branch permission is legal or not.
2919 +static int test_br(struct inode *inode, int brperm, char *path)
2923 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
2928 + pr_err("write permission for readonly mount or inode, %s\n", path);
2936 + * 0: success, the caller will add it
2937 + * plus: success, it is already unified, the caller should ignore it
2940 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
2943 + aufs_bindex_t bend, bindex;
2944 + struct dentry *root, *h_dentry;
2945 + struct inode *inode, *h_inode;
2947 + root = sb->s_root;
2948 + bend = au_sbend(sb);
2949 + if (unlikely(bend >= 0
2950 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
2954 + pr_err("%s duplicated\n", add->pathname);
2959 + err = -ENOSPC; /* -E2BIG; */
2960 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
2961 + || AUFS_BRANCH_MAX - 1 <= bend)) {
2962 + pr_err("number of branches exceeded %s\n", add->pathname);
2967 + if (unlikely(add->bindex < 0 || bend + 1 < add->bindex)) {
2968 + pr_err("bad index %d\n", add->bindex);
2972 + inode = d_inode(add->path.dentry);
2974 + if (unlikely(!inode->i_nlink)) {
2975 + pr_err("no existence %s\n", add->pathname);
2980 + if (unlikely(inode->i_sb == sb)) {
2981 + pr_err("%s must be outside\n", add->pathname);
2985 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
2986 + pr_err("unsupported filesystem, %s (%s)\n",
2987 + add->pathname, au_sbtype(inode->i_sb));
2991 + if (unlikely(inode->i_sb->s_stack_depth)) {
2992 + pr_err("already stacked, %s (%s)\n",
2993 + add->pathname, au_sbtype(inode->i_sb));
2997 + err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
2998 + if (unlikely(err))
3002 + return 0; /* success */
3005 + for (bindex = 0; bindex <= bend; bindex++)
3006 + if (unlikely(test_overlap(sb, add->path.dentry,
3007 + au_h_dptr(root, bindex)))) {
3008 + pr_err("%s is overlapped\n", add->pathname);
3013 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3014 + h_dentry = au_h_dptr(root, 0);
3015 + h_inode = d_inode(h_dentry);
3016 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3017 + || !uid_eq(h_inode->i_uid, inode->i_uid)
3018 + || !gid_eq(h_inode->i_gid, inode->i_gid))
3019 + pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3021 + i_uid_read(inode), i_gid_read(inode),
3022 + (inode->i_mode & S_IALLUGO),
3023 + i_uid_read(h_inode), i_gid_read(h_inode),
3024 + (h_inode->i_mode & S_IALLUGO));
3032 + * initialize or clean the whiteouts for an adding branch
3034 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3037 + int err, old_perm;
3038 + aufs_bindex_t bindex;
3039 + struct mutex *h_mtx;
3040 + struct au_wbr *wbr;
3041 + struct au_hinode *hdir;
3042 + struct dentry *h_dentry;
3044 + err = vfsub_mnt_want_write(au_br_mnt(br));
3045 + if (unlikely(err))
3049 + old_perm = br->br_perm;
3050 + br->br_perm = new_perm;
3053 + bindex = au_br_index(sb, br->br_id);
3054 + if (0 <= bindex) {
3055 + hdir = au_hi(d_inode(sb->s_root), bindex);
3056 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
3058 + h_dentry = au_br_dentry(br);
3059 + h_mtx = &d_inode(h_dentry)->i_mutex;
3060 + mutex_lock_nested(h_mtx, AuLsc_I_PARENT);
3063 + err = au_wh_init(br, sb);
3065 + wbr_wh_write_lock(wbr);
3066 + err = au_wh_init(br, sb);
3067 + wbr_wh_write_unlock(wbr);
3070 + au_hn_imtx_unlock(hdir);
3072 + mutex_unlock(h_mtx);
3073 + vfsub_mnt_drop_write(au_br_mnt(br));
3074 + br->br_perm = old_perm;
3076 + if (!err && wbr && !au_br_writable(new_perm)) {
3078 + br->br_wbr = NULL;
3085 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3089 + struct kstatfs kst;
3090 + struct au_wbr *wbr;
3093 + au_rw_init(&wbr->wbr_wh_rwsem);
3094 + atomic_set(&wbr->wbr_wh_running, 0);
3097 + * a limit for rmdir/rename a dir
3098 + * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3100 + err = vfs_statfs(&br->br_path, &kst);
3101 + if (unlikely(err))
3104 + if (kst.f_namelen >= NAME_MAX)
3105 + err = au_br_init_wh(sb, br, perm);
3107 + pr_err("%pd(%s), unsupported namelen %ld\n",
3109 + au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3115 +/* initialize a new branch */
3116 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3117 + struct au_opt_add *add)
3120 + struct inode *h_inode;
3123 + mutex_init(&br->br_xino.xi_nondir_mtx);
3124 + br->br_perm = add->perm;
3125 + br->br_path = add->path; /* set first, path_get() later */
3126 + spin_lock_init(&br->br_dykey_lock);
3127 + atomic_set(&br->br_count, 0);
3128 + atomic_set(&br->br_xino_running, 0);
3129 + br->br_id = au_new_br_id(sb);
3130 + AuDebugOn(br->br_id < 0);
3132 + if (au_br_writable(add->perm)) {
3133 + err = au_wbr_init(br, sb, add->perm);
3134 + if (unlikely(err))
3138 + if (au_opt_test(au_mntflags(sb), XINO)) {
3139 + h_inode = d_inode(add->path.dentry);
3140 + err = au_xino_br(sb, br, h_inode->i_ino,
3141 + au_sbr(sb, 0)->br_xino.xi_file, /*do_test*/1);
3142 + if (unlikely(err)) {
3143 + AuDebugOn(br->br_xino.xi_file);
3148 + sysaufs_br_init(br);
3149 + path_get(&br->br_path);
3150 + goto out; /* success */
3153 + memset(&br->br_path, 0, sizeof(br->br_path));
3158 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3159 + struct au_branch *br, aufs_bindex_t bend,
3160 + aufs_bindex_t amount)
3162 + struct au_branch **brp;
3164 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3166 + brp = sbinfo->si_branch + bindex;
3167 + memmove(brp + 1, brp, sizeof(*brp) * amount);
3169 + sbinfo->si_bend++;
3170 + if (unlikely(bend < 0))
3171 + sbinfo->si_bend = 0;
3174 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3175 + aufs_bindex_t bend, aufs_bindex_t amount)
3177 + struct au_hdentry *hdp;
3179 + AuRwMustWriteLock(&dinfo->di_rwsem);
3181 + hdp = dinfo->di_hdentry + bindex;
3182 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3183 + au_h_dentry_init(hdp);
3185 + if (unlikely(bend < 0))
3186 + dinfo->di_bstart = 0;
3189 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3190 + aufs_bindex_t bend, aufs_bindex_t amount)
3192 + struct au_hinode *hip;
3194 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3196 + hip = iinfo->ii_hinode + bindex;
3197 + memmove(hip + 1, hip, sizeof(*hip) * amount);
3198 + hip->hi_inode = NULL;
3201 + if (unlikely(bend < 0))
3202 + iinfo->ii_bstart = 0;
3205 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3206 + aufs_bindex_t bindex)
3208 + struct dentry *root, *h_dentry;
3209 + struct inode *root_inode, *h_inode;
3210 + aufs_bindex_t bend, amount;
3212 + root = sb->s_root;
3213 + root_inode = d_inode(root);
3214 + bend = au_sbend(sb);
3215 + amount = bend + 1 - bindex;
3216 + h_dentry = au_br_dentry(br);
3217 + au_sbilist_lock();
3218 + au_br_do_add_brp(au_sbi(sb), bindex, br, bend, amount);
3219 + au_br_do_add_hdp(au_di(root), bindex, bend, amount);
3220 + au_br_do_add_hip(au_ii(root_inode), bindex, bend, amount);
3221 + au_set_h_dptr(root, bindex, dget(h_dentry));
3222 + h_inode = d_inode(h_dentry);
3223 + au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3224 + au_sbilist_unlock();
3227 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3230 + aufs_bindex_t bend, add_bindex;
3231 + struct dentry *root, *h_dentry;
3232 + struct inode *root_inode;
3233 + struct au_branch *add_branch;
3235 + root = sb->s_root;
3236 + root_inode = d_inode(root);
3237 + IMustLock(root_inode);
3238 + err = test_add(sb, add, remount);
3239 + if (unlikely(err < 0))
3243 + goto out; /* success */
3246 + bend = au_sbend(sb);
3247 + add_branch = au_br_alloc(sb, bend + 2, add->perm);
3248 + err = PTR_ERR(add_branch);
3249 + if (IS_ERR(add_branch))
3252 + err = au_br_init(add_branch, sb, add);
3253 + if (unlikely(err)) {
3254 + au_br_do_free(add_branch);
3258 + add_bindex = add->bindex;
3260 + au_br_do_add(sb, add_branch, add_bindex);
3262 + sysaufs_brs_del(sb, add_bindex);
3263 + au_br_do_add(sb, add_branch, add_bindex);
3264 + sysaufs_brs_add(sb, add_bindex);
3267 + h_dentry = add->path.dentry;
3268 + if (!add_bindex) {
3269 + au_cpup_attr_all(root_inode, /*force*/1);
3270 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3272 + au_add_nlink(root_inode, d_inode(h_dentry));
3275 + * this test/set prevents aufs from handling unnecesary notify events
3276 + * of xino files, in case of re-adding a writable branch which was
3277 + * once detached from aufs.
3279 + if (au_xino_brid(sb) < 0
3280 + && au_br_writable(add_branch->br_perm)
3281 + && !au_test_fs_bad_xino(h_dentry->d_sb)
3282 + && add_branch->br_xino.xi_file
3283 + && add_branch->br_xino.xi_file->f_path.dentry->d_parent == h_dentry)
3284 + au_xino_brid_set(sb, add_branch->br_id);
3290 +/* ---------------------------------------------------------------------- */
3292 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3293 + unsigned long long max __maybe_unused,
3296 + unsigned long long n;
3297 + struct file **p, *f;
3298 + struct au_sphlhead *files;
3299 + struct au_finfo *finfo;
3303 + files = &au_sbi(sb)->si_files;
3304 + spin_lock(&files->spin);
3305 + hlist_for_each_entry(finfo, &files->head, fi_hlist) {
3306 + f = finfo->fi_file;
3308 + && !special_file(file_inode(f)->i_mode)) {
3312 + AuDebugOn(n > max);
3315 + spin_unlock(&files->spin);
3320 +static struct file **au_farray_alloc(struct super_block *sb,
3321 + unsigned long long *max)
3323 + *max = atomic_long_read(&au_sbi(sb)->si_nfiles);
3324 + return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3327 +static void au_farray_free(struct file **a, unsigned long long max)
3329 + unsigned long long ull;
3331 + for (ull = 0; ull < max; ull++)
3337 +/* ---------------------------------------------------------------------- */
3343 +/* to show the line number, do not make it inlined function */
3344 +#define AuVerbose(do_info, fmt, ...) do { \
3346 + pr_info(fmt, ##__VA_ARGS__); \
3349 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t bstart,
3350 + aufs_bindex_t bend)
3352 + return (inode && !S_ISDIR(inode->i_mode)) || bstart == bend;
3355 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t bstart,
3356 + aufs_bindex_t bend)
3358 + return au_test_ibusy(d_inode(dentry), bstart, bend);
3362 + * test if the branch is deletable or not.
3364 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3365 + unsigned int sigen, const unsigned int verbose)
3367 + int err, i, j, ndentry;
3368 + aufs_bindex_t bstart, bend;
3369 + struct au_dcsub_pages dpages;
3370 + struct au_dpage *dpage;
3373 + err = au_dpages_init(&dpages, GFP_NOFS);
3374 + if (unlikely(err))
3376 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
3377 + if (unlikely(err))
3380 + for (i = 0; !err && i < dpages.ndpage; i++) {
3381 + dpage = dpages.dpages + i;
3382 + ndentry = dpage->ndentry;
3383 + for (j = 0; !err && j < ndentry; j++) {
3384 + d = dpage->dentries[j];
3385 + AuDebugOn(au_dcount(d) <= 0);
3386 + if (!au_digen_test(d, sigen)) {
3387 + di_read_lock_child(d, AuLock_IR);
3388 + if (unlikely(au_dbrange_test(d))) {
3389 + di_read_unlock(d, AuLock_IR);
3393 + di_write_lock_child(d);
3394 + if (unlikely(au_dbrange_test(d))) {
3395 + di_write_unlock(d);
3398 + err = au_reval_dpath(d, sigen);
3400 + di_downgrade_lock(d, AuLock_IR);
3402 + di_write_unlock(d);
3407 + /* AuDbgDentry(d); */
3408 + bstart = au_dbstart(d);
3409 + bend = au_dbend(d);
3410 + if (bstart <= bindex
3412 + && au_h_dptr(d, bindex)
3413 + && au_test_dbusy(d, bstart, bend)) {
3415 + AuVerbose(verbose, "busy %pd\n", d);
3418 + di_read_unlock(d, AuLock_IR);
3423 + au_dpages_free(&dpages);
3428 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3429 + unsigned int sigen, const unsigned int verbose)
3432 + unsigned long long max, ull;
3433 + struct inode *i, **array;
3434 + aufs_bindex_t bstart, bend;
3436 + array = au_iarray_alloc(sb, &max);
3437 + err = PTR_ERR(array);
3438 + if (IS_ERR(array))
3442 + AuDbg("b%d\n", bindex);
3443 + for (ull = 0; !err && ull < max; ull++) {
3447 + if (i->i_ino == AUFS_ROOT_INO)
3450 + /* AuDbgInode(i); */
3451 + if (au_iigen(i, NULL) == sigen)
3452 + ii_read_lock_child(i);
3454 + ii_write_lock_child(i);
3455 + err = au_refresh_hinode_self(i);
3458 + ii_downgrade_lock(i);
3460 + ii_write_unlock(i);
3465 + bstart = au_ibstart(i);
3466 + bend = au_ibend(i);
3467 + if (bstart <= bindex
3469 + && au_h_iptr(i, bindex)
3470 + && au_test_ibusy(i, bstart, bend)) {
3472 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3475 + ii_read_unlock(i);
3477 + au_iarray_free(array, max);
3483 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3484 + const unsigned int verbose)
3487 + unsigned int sigen;
3489 + sigen = au_sigen(root->d_sb);
3490 + DiMustNoWaiters(root);
3491 + IiMustNoWaiters(d_inode(root));
3492 + di_write_unlock(root);
3493 + err = test_dentry_busy(root, bindex, sigen, verbose);
3495 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3496 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3501 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3502 + struct file **to_free, int *idx)
3505 + unsigned char matched, root;
3506 + aufs_bindex_t bindex, bend;
3507 + struct au_fidir *fidir;
3508 + struct au_hfile *hfile;
3511 + root = IS_ROOT(file->f_path.dentry);
3514 + to_free[*idx] = file;
3520 + fidir = au_fi(file)->fi_hdir;
3521 + AuDebugOn(!fidir);
3522 + bend = au_fbend_dir(file);
3523 + for (bindex = au_fbstart(file); bindex <= bend; bindex++) {
3524 + hfile = fidir->fd_hfile + bindex;
3525 + if (!hfile->hf_file)
3528 + if (hfile->hf_br->br_id == br_id) {
3540 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3541 + struct file **to_free, int opened)
3544 + unsigned long long ull, max;
3545 + aufs_bindex_t bstart;
3546 + struct file *file, **array;
3547 + struct dentry *root;
3548 + struct au_hfile *hfile;
3550 + array = au_farray_alloc(sb, &max);
3551 + err = PTR_ERR(array);
3552 + if (IS_ERR(array))
3557 + root = sb->s_root;
3558 + di_write_unlock(root);
3559 + for (ull = 0; ull < max; ull++) {
3560 + file = array[ull];
3561 + if (unlikely(!file))
3564 + /* AuDbg("%pD\n", file); */
3565 + fi_read_lock(file);
3566 + bstart = au_fbstart(file);
3567 + if (!d_is_dir(file->f_path.dentry)) {
3568 + hfile = &au_fi(file)->fi_htop;
3569 + if (hfile->hf_br->br_id == br_id)
3572 + err = test_dir_busy(file, br_id, to_free, &idx);
3573 + fi_read_unlock(file);
3574 + if (unlikely(err))
3577 + di_write_lock_child(root);
3578 + au_farray_free(array, max);
3579 + AuDebugOn(idx > opened);
3585 +static void br_del_file(struct file **to_free, unsigned long long opened,
3586 + aufs_bindex_t br_id)
3588 + unsigned long long ull;
3589 + aufs_bindex_t bindex, bstart, bend, bfound;
3590 + struct file *file;
3591 + struct au_fidir *fidir;
3592 + struct au_hfile *hfile;
3594 + for (ull = 0; ull < opened; ull++) {
3595 + file = to_free[ull];
3596 + if (unlikely(!file))
3599 + /* AuDbg("%pD\n", file); */
3600 + AuDebugOn(!d_is_dir(file->f_path.dentry));
3602 + fidir = au_fi(file)->fi_hdir;
3603 + AuDebugOn(!fidir);
3604 + fi_write_lock(file);
3605 + bstart = au_fbstart(file);
3606 + bend = au_fbend_dir(file);
3607 + for (bindex = bstart; bindex <= bend; bindex++) {
3608 + hfile = fidir->fd_hfile + bindex;
3609 + if (!hfile->hf_file)
3612 + if (hfile->hf_br->br_id == br_id) {
3617 + AuDebugOn(bfound < 0);
3618 + au_set_h_fptr(file, bfound, NULL);
3619 + if (bfound == bstart) {
3620 + for (bstart++; bstart <= bend; bstart++)
3621 + if (au_hf_dir(file, bstart)) {
3622 + au_set_fbstart(file, bstart);
3626 + fi_write_unlock(file);
3630 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3631 + const aufs_bindex_t bindex,
3632 + const aufs_bindex_t bend)
3634 + struct au_branch **brp, **p;
3636 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3638 + brp = sbinfo->si_branch + bindex;
3639 + if (bindex < bend)
3640 + memmove(brp, brp + 1, sizeof(*brp) * (bend - bindex));
3641 + sbinfo->si_branch[0 + bend] = NULL;
3642 + sbinfo->si_bend--;
3644 + p = krealloc(sbinfo->si_branch, sizeof(*p) * bend, AuGFP_SBILIST);
3646 + sbinfo->si_branch = p;
3647 + /* harmless error */
3650 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3651 + const aufs_bindex_t bend)
3653 + struct au_hdentry *hdp, *p;
3655 + AuRwMustWriteLock(&dinfo->di_rwsem);
3657 + hdp = dinfo->di_hdentry;
3658 + if (bindex < bend)
3659 + memmove(hdp + bindex, hdp + bindex + 1,
3660 + sizeof(*hdp) * (bend - bindex));
3661 + hdp[0 + bend].hd_dentry = NULL;
3664 + p = krealloc(hdp, sizeof(*p) * bend, AuGFP_SBILIST);
3666 + dinfo->di_hdentry = p;
3667 + /* harmless error */
3670 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3671 + const aufs_bindex_t bend)
3673 + struct au_hinode *hip, *p;
3675 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3677 + hip = iinfo->ii_hinode + bindex;
3678 + if (bindex < bend)
3679 + memmove(hip, hip + 1, sizeof(*hip) * (bend - bindex));
3680 + iinfo->ii_hinode[0 + bend].hi_inode = NULL;
3681 + au_hn_init(iinfo->ii_hinode + bend);
3684 + p = krealloc(iinfo->ii_hinode, sizeof(*p) * bend, AuGFP_SBILIST);
3686 + iinfo->ii_hinode = p;
3687 + /* harmless error */
3690 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3691 + struct au_branch *br)
3693 + aufs_bindex_t bend;
3694 + struct au_sbinfo *sbinfo;
3695 + struct dentry *root, *h_root;
3696 + struct inode *inode, *h_inode;
3697 + struct au_hinode *hinode;
3699 + SiMustWriteLock(sb);
3701 + root = sb->s_root;
3702 + inode = d_inode(root);
3703 + sbinfo = au_sbi(sb);
3704 + bend = sbinfo->si_bend;
3706 + h_root = au_h_dptr(root, bindex);
3707 + hinode = au_hi(inode, bindex);
3708 + h_inode = au_igrab(hinode->hi_inode);
3711 + au_sbilist_lock();
3712 + au_br_do_del_brp(sbinfo, bindex, bend);
3713 + au_br_do_del_hdp(au_di(root), bindex, bend);
3714 + au_br_do_del_hip(au_ii(inode), bindex, bend);
3715 + au_sbilist_unlock();
3719 + au_br_do_free(br);
3722 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3723 + unsigned long long max, void *arg)
3728 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3731 + unsigned long long opened;
3732 + unsigned int mnt_flags;
3733 + aufs_bindex_t bindex, bend, br_id;
3734 + unsigned char do_wh, verbose;
3735 + struct au_branch *br;
3736 + struct au_wbr *wbr;
3737 + struct dentry *root;
3738 + struct file **to_free;
3743 + root = sb->s_root;
3744 + bindex = au_find_dbindex(root, del->h_path.dentry);
3747 + goto out; /* success */
3749 + pr_err("%s no such branch\n", del->pathname);
3752 + AuDbg("bindex b%d\n", bindex);
3755 + mnt_flags = au_mntflags(sb);
3756 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3757 + bend = au_sbend(sb);
3758 + if (unlikely(!bend)) {
3759 + AuVerbose(verbose, "no more branches left\n");
3762 + br = au_sbr(sb, bindex);
3763 + AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3765 + br_id = br->br_id;
3766 + opened = atomic_read(&br->br_count);
3767 + if (unlikely(opened)) {
3768 + to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3769 + err = PTR_ERR(to_free);
3770 + if (IS_ERR(to_free))
3773 + err = test_file_busy(sb, br_id, to_free, opened);
3774 + if (unlikely(err)) {
3775 + AuVerbose(verbose, "%llu file(s) opened\n", opened);
3781 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3783 + /* instead of WbrWhMustWriteLock(wbr) */
3784 + SiMustWriteLock(sb);
3785 + for (i = 0; i < AuBrWh_Last; i++) {
3786 + dput(wbr->wbr_wh[i]);
3787 + wbr->wbr_wh[i] = NULL;
3791 + err = test_children_busy(root, bindex, verbose);
3792 + if (unlikely(err)) {
3801 + * now we confirmed the branch is deletable.
3802 + * let's free the remaining opened dirs on the branch.
3804 + di_write_unlock(root);
3805 + br_del_file(to_free, opened, br_id);
3806 + di_write_lock_child(root);
3810 + au_br_do_del(sb, bindex, br);
3812 + sysaufs_brs_del(sb, bindex);
3813 + au_br_do_del(sb, bindex, br);
3814 + sysaufs_brs_add(sb, bindex);
3818 + au_cpup_attr_all(d_inode(root), /*force*/1);
3819 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3821 + au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
3822 + if (au_opt_test(mnt_flags, PLINK))
3823 + au_plink_half_refresh(sb, br_id);
3825 + if (au_xino_brid(sb) == br_id)
3826 + au_xino_brid_set(sb, -1);
3827 + goto out; /* success */
3831 + rerr = au_br_init_wh(sb, br, br->br_perm);
3833 + pr_warn("failed re-creating base whiteout, %s. (%d)\n",
3834 + del->pathname, rerr);
3837 + au_farray_free(to_free, opened);
3841 +/* ---------------------------------------------------------------------- */
3843 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
3846 + aufs_bindex_t bstart, bend;
3847 + struct aufs_ibusy ibusy;
3848 + struct inode *inode, *h_inode;
3851 + if (unlikely(!capable(CAP_SYS_ADMIN)))
3854 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
3856 + err = !access_ok(VERIFY_WRITE, &arg->h_ino, sizeof(arg->h_ino));
3857 + if (unlikely(err)) {
3864 + si_read_lock(sb, AuLock_FLUSH);
3865 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbend(sb)))
3869 + ibusy.h_ino = 0; /* invalid */
3870 + inode = ilookup(sb, ibusy.ino);
3872 + || inode->i_ino == AUFS_ROOT_INO
3873 + || is_bad_inode(inode))
3876 + ii_read_lock_child(inode);
3877 + bstart = au_ibstart(inode);
3878 + bend = au_ibend(inode);
3879 + if (bstart <= ibusy.bindex && ibusy.bindex <= bend) {
3880 + h_inode = au_h_iptr(inode, ibusy.bindex);
3881 + if (h_inode && au_test_ibusy(inode, bstart, bend))
3882 + ibusy.h_ino = h_inode->i_ino;
3884 + ii_read_unlock(inode);
3888 + si_read_unlock(sb);
3890 + err = __put_user(ibusy.h_ino, &arg->h_ino);
3891 + if (unlikely(err)) {
3900 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
3902 + return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
3905 +#ifdef CONFIG_COMPAT
3906 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
3908 + return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
3912 +/* ---------------------------------------------------------------------- */
3915 + * change a branch permission
3918 +static void au_warn_ima(void)
3921 + /* since it doesn't support mark_files_ro() */
3922 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
3926 +static int do_need_sigen_inc(int a, int b)
3928 + return au_br_whable(a) && !au_br_whable(b);
3931 +static int need_sigen_inc(int old, int new)
3933 + return do_need_sigen_inc(old, new)
3934 + || do_need_sigen_inc(new, old);
3937 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
3940 + unsigned int mnt_flags;
3941 + unsigned long long ull, max;
3942 + aufs_bindex_t br_id;
3943 + unsigned char verbose, writer;
3944 + struct file *file, *hf, **array;
3945 + struct au_hfile *hfile;
3947 + mnt_flags = au_mntflags(sb);
3948 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3950 + array = au_farray_alloc(sb, &max);
3951 + err = PTR_ERR(array);
3952 + if (IS_ERR(array))
3956 + br_id = au_sbr_id(sb, bindex);
3957 + for (ull = 0; ull < max; ull++) {
3958 + file = array[ull];
3959 + if (unlikely(!file))
3962 + /* AuDbg("%pD\n", file); */
3963 + fi_read_lock(file);
3964 + if (unlikely(au_test_mmapped(file))) {
3966 + AuVerbose(verbose, "mmapped %pD\n", file);
3968 + FiMustNoWaiters(file);
3969 + fi_read_unlock(file);
3973 + hfile = &au_fi(file)->fi_htop;
3974 + hf = hfile->hf_file;
3975 + if (!d_is_reg(file->f_path.dentry)
3976 + || !(file->f_mode & FMODE_WRITE)
3977 + || hfile->hf_br->br_id != br_id
3978 + || !(hf->f_mode & FMODE_WRITE))
3979 + array[ull] = NULL;
3985 + FiMustNoWaiters(file);
3986 + fi_read_unlock(file);
3994 + for (ull = 0; ull < max; ull++) {
3995 + file = array[ull];
3999 + /* todo: already flushed? */
4001 + * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4002 + * approach which resets f_mode and calls mnt_drop_write() and
4003 + * file_release_write() for each file, because the branch
4004 + * attribute in aufs world is totally different from the native
4007 + /* fi_read_lock(file); */
4008 + hfile = &au_fi(file)->fi_htop;
4009 + hf = hfile->hf_file;
4010 + /* fi_read_unlock(file); */
4011 + spin_lock(&hf->f_lock);
4012 + writer = !!(hf->f_mode & FMODE_WRITER);
4013 + hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4014 + spin_unlock(&hf->f_lock);
4016 + put_write_access(file_inode(hf));
4017 + __mnt_drop_write(hf->f_path.mnt);
4022 + au_farray_free(array, max);
4028 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4032 + aufs_bindex_t bindex;
4033 + struct dentry *root;
4034 + struct au_branch *br;
4035 + struct au_br_fhsm *bf;
4037 + root = sb->s_root;
4038 + bindex = au_find_dbindex(root, mod->h_root);
4041 + return 0; /* success */
4043 + pr_err("%s no such branch\n", mod->path);
4046 + AuDbg("bindex b%d\n", bindex);
4048 + err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4049 + if (unlikely(err))
4052 + br = au_sbr(sb, bindex);
4053 + AuDebugOn(mod->h_root != au_br_dentry(br));
4054 + if (br->br_perm == mod->perm)
4055 + return 0; /* success */
4057 + /* pre-allocate for non-fhsm --> fhsm */
4059 + if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4060 + err = au_fhsm_br_alloc(br);
4061 + if (unlikely(err))
4064 + br->br_fhsm = NULL;
4067 + if (au_br_writable(br->br_perm)) {
4068 + /* remove whiteout base */
4069 + err = au_br_init_wh(sb, br, mod->perm);
4070 + if (unlikely(err))
4073 + if (!au_br_writable(mod->perm)) {
4074 + /* rw --> ro, file might be mmapped */
4075 + DiMustNoWaiters(root);
4076 + IiMustNoWaiters(d_inode(root));
4077 + di_write_unlock(root);
4078 + err = au_br_mod_files_ro(sb, bindex);
4079 + /* aufs_write_lock() calls ..._child() */
4080 + di_write_lock_child(root);
4082 + if (unlikely(err)) {
4084 + br->br_wbr = kzalloc(sizeof(*br->br_wbr),
4087 + rerr = au_wbr_init(br, sb, br->br_perm);
4088 + if (unlikely(rerr)) {
4089 + AuIOErr("nested error %d (%d)\n",
4091 + br->br_perm = mod->perm;
4095 + } else if (au_br_writable(mod->perm)) {
4098 + br->br_wbr = kzalloc(sizeof(*br->br_wbr), GFP_NOFS);
4100 + err = au_wbr_init(br, sb, mod->perm);
4101 + if (unlikely(err)) {
4102 + kfree(br->br_wbr);
4103 + br->br_wbr = NULL;
4107 + if (unlikely(err))
4110 + if (au_br_fhsm(br->br_perm)) {
4111 + if (!au_br_fhsm(mod->perm)) {
4112 + /* fhsm --> non-fhsm */
4113 + au_br_fhsm_fin(br->br_fhsm);
4114 + kfree(br->br_fhsm);
4115 + br->br_fhsm = NULL;
4117 + } else if (au_br_fhsm(mod->perm))
4118 + /* non-fhsm --> fhsm */
4121 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4122 + br->br_perm = mod->perm;
4123 + goto out; /* success */
4132 +/* ---------------------------------------------------------------------- */
4134 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4137 + struct kstatfs kstfs;
4139 + err = vfs_statfs(&br->br_path, &kstfs);
4141 + stfs->f_blocks = kstfs.f_blocks;
4142 + stfs->f_bavail = kstfs.f_bavail;
4143 + stfs->f_files = kstfs.f_files;
4144 + stfs->f_ffree = kstfs.f_ffree;
4149 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4150 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
4151 +++ linux/fs/aufs/branch.h 2016-01-13 20:11:11.666426853 +0100
4154 + * Copyright (C) 2005-2015 Junjiro R. Okajima
4156 + * This program, aufs is free software; you can redistribute it and/or modify
4157 + * it under the terms of the GNU General Public License as published by
4158 + * the Free Software Foundation; either version 2 of the License, or
4159 + * (at your option) any later version.
4161 + * This program is distributed in the hope that it will be useful,
4162 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4163 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4164 + * GNU General Public License for more details.
4166 + * You should have received a copy of the GNU General Public License
4167 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4171 + * branch filesystems and xino for them
4174 +#ifndef __AUFS_BRANCH_H__
4175 +#define __AUFS_BRANCH_H__
4179 +#include <linux/mount.h>
4184 +/* ---------------------------------------------------------------------- */
4187 +struct au_xino_file {
4188 + struct file *xi_file;
4189 + struct mutex xi_nondir_mtx;
4191 + /* todo: make xino files an array to support huge inode number */
4193 +#ifdef CONFIG_DEBUG_FS
4194 + struct dentry *xi_dbgaufs;
4198 +/* File-based Hierarchical Storage Management */
4199 +struct au_br_fhsm {
4200 +#ifdef CONFIG_AUFS_FHSM
4201 + struct mutex bf_lock;
4202 + unsigned long bf_jiffy;
4203 + struct aufs_stfs bf_stfs;
4208 +/* members for writable branch only */
4209 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4211 + struct au_rwsem wbr_wh_rwsem;
4212 + struct dentry *wbr_wh[AuBrWh_Last];
4213 + atomic_t wbr_wh_running;
4214 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4215 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4216 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4219 + unsigned long long wbr_bytes;
4222 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4223 +#define AuBrDynOp (AuDyLast * 4)
4225 +#ifdef CONFIG_AUFS_HFSNOTIFY
4226 +/* support for asynchronous destruction */
4227 +struct au_br_hfsnotify {
4228 + struct fsnotify_group *hfsn_group;
4232 +/* sysfs entries */
4233 +struct au_brsysfs {
4235 + struct attribute attr;
4244 +/* protected by superblock rwsem */
4246 + struct au_xino_file br_xino;
4248 + aufs_bindex_t br_id;
4251 + struct path br_path;
4252 + spinlock_t br_dykey_lock;
4253 + struct au_dykey *br_dykey[AuBrDynOp];
4254 + atomic_t br_count;
4256 + struct au_wbr *br_wbr;
4257 + struct au_br_fhsm *br_fhsm;
4259 + /* xino truncation */
4260 + atomic_t br_xino_running;
4262 +#ifdef CONFIG_AUFS_HFSNOTIFY
4263 + struct au_br_hfsnotify *br_hfsn;
4266 +#ifdef CONFIG_SYSFS
4267 + /* entries under sysfs per mount-point */
4268 + struct au_brsysfs br_sysfs[AuBrSysfs_Last];
4272 +/* ---------------------------------------------------------------------- */
4274 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4276 + return br->br_path.mnt;
4279 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4281 + return br->br_path.dentry;
4284 +static inline struct super_block *au_br_sb(struct au_branch *br)
4286 + return au_br_mnt(br)->mnt_sb;
4289 +static inline int au_br_rdonly(struct au_branch *br)
4291 + return ((au_br_sb(br)->s_flags & MS_RDONLY)
4292 + || !au_br_writable(br->br_perm))
4296 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4298 +#ifdef CONFIG_AUFS_HNOTIFY
4299 + return !(brperm & AuBrPerm_RR);
4305 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4307 + int err, exec_flag;
4310 + exec_flag = oflag & __FMODE_EXEC;
4311 + if (unlikely(exec_flag && path_noexec(&br->br_path)))
4317 +/* ---------------------------------------------------------------------- */
4321 +void au_br_free(struct au_sbinfo *sinfo);
4322 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4324 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4326 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4327 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4328 +#ifdef CONFIG_COMPAT
4329 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4332 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4335 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4338 +static const loff_t au_loff_max = LLONG_MAX;
4340 +int au_xib_trunc(struct super_block *sb);
4341 +ssize_t xino_fread(vfs_readf_t func, struct file *file, void *buf, size_t size,
4343 +ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
4344 + size_t size, loff_t *pos);
4345 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src);
4346 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent);
4347 +ino_t au_xino_new_ino(struct super_block *sb);
4348 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4349 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4351 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4353 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4354 + struct file *base_file, int do_test);
4355 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex);
4357 +struct au_opt_xino;
4358 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount);
4359 +void au_xino_clr(struct super_block *sb);
4360 +struct file *au_xino_def(struct super_block *sb);
4361 +int au_xino_path(struct seq_file *seq, struct file *file);
4363 +/* ---------------------------------------------------------------------- */
4365 +/* Superblock to branch */
4367 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4369 + return au_sbr(sb, bindex)->br_id;
4373 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4375 + return au_br_mnt(au_sbr(sb, bindex));
4379 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4381 + return au_br_sb(au_sbr(sb, bindex));
4384 +static inline void au_sbr_put(struct super_block *sb, aufs_bindex_t bindex)
4386 + atomic_dec(&au_sbr(sb, bindex)->br_count);
4389 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4391 + return au_sbr(sb, bindex)->br_perm;
4394 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4396 + return au_br_whable(au_sbr_perm(sb, bindex));
4399 +/* ---------------------------------------------------------------------- */
4402 + * wbr_wh_read_lock, wbr_wh_write_lock
4403 + * wbr_wh_read_unlock, wbr_wh_write_unlock, wbr_wh_downgrade_lock
4405 +AuSimpleRwsemFuncs(wbr_wh, struct au_wbr *wbr, &wbr->wbr_wh_rwsem);
4407 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&wbr->wbr_wh_rwsem)
4408 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&wbr->wbr_wh_rwsem)
4409 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&wbr->wbr_wh_rwsem)
4411 +/* ---------------------------------------------------------------------- */
4413 +#ifdef CONFIG_AUFS_FHSM
4414 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4416 + mutex_init(&brfhsm->bf_lock);
4417 + brfhsm->bf_jiffy = 0;
4418 + brfhsm->bf_readable = 0;
4421 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4423 + mutex_destroy(&brfhsm->bf_lock);
4426 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4427 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4430 +#endif /* __KERNEL__ */
4431 +#endif /* __AUFS_BRANCH_H__ */
4432 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4433 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
4434 +++ linux/fs/aufs/conf.mk 2016-01-13 20:11:11.666426853 +0100
4437 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4441 +AuConfStr += ${1}=${${1}}
4445 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4447 + HNOTIFY HFSNOTIFY \
4458 +$(foreach i, ${AuConfAll}, \
4459 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
4461 +AuConfName = ${obj}/conf.str
4462 +${AuConfName}.tmp: FORCE
4463 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4464 +${AuConfName}: ${AuConfName}.tmp
4465 + @diff -q $< $@ > /dev/null 2>&1 || { \
4466 + echo ' GEN ' $@; \
4470 +clean-files += ${AuConfName} ${AuConfName}.tmp
4471 +${obj}/sysfs.o: ${AuConfName}
4473 +-include ${srctree}/${src}/conf_priv.mk
4474 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4475 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
4476 +++ linux/fs/aufs/cpup.c 2016-01-13 20:11:11.666426853 +0100
4479 + * Copyright (C) 2005-2015 Junjiro R. Okajima
4481 + * This program, aufs is free software; you can redistribute it and/or modify
4482 + * it under the terms of the GNU General Public License as published by
4483 + * the Free Software Foundation; either version 2 of the License, or
4484 + * (at your option) any later version.
4486 + * This program is distributed in the hope that it will be useful,
4487 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4488 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4489 + * GNU General Public License for more details.
4491 + * You should have received a copy of the GNU General Public License
4492 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4496 + * copy-up functions, see wbr_policy.c for copy-down
4499 +#include <linux/fs_stack.h>
4500 +#include <linux/mm.h>
4503 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4505 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4506 + | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4508 + BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4510 + dst->i_flags |= iflags & ~mask;
4511 + if (au_test_fs_notime(dst->i_sb))
4512 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
4515 +void au_cpup_attr_timesizes(struct inode *inode)
4517 + struct inode *h_inode;
4519 + h_inode = au_h_iptr(inode, au_ibstart(inode));
4520 + fsstack_copy_attr_times(inode, h_inode);
4521 + fsstack_copy_inode_size(inode, h_inode);
4524 +void au_cpup_attr_nlink(struct inode *inode, int force)
4526 + struct inode *h_inode;
4527 + struct super_block *sb;
4528 + aufs_bindex_t bindex, bend;
4531 + bindex = au_ibstart(inode);
4532 + h_inode = au_h_iptr(inode, bindex);
4534 + && !S_ISDIR(h_inode->i_mode)
4535 + && au_opt_test(au_mntflags(sb), PLINK)
4536 + && au_plink_test(inode))
4540 + * 0 can happen in revalidating.
4541 + * h_inode->i_mutex may not be held here, but it is harmless since once
4542 + * i_nlink reaches 0, it will never become positive except O_TMPFILE
4544 + * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4545 + * the incorrect link count.
4547 + set_nlink(inode, h_inode->i_nlink);
4550 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4551 + * it may includes whplink directory.
4553 + if (S_ISDIR(h_inode->i_mode)) {
4554 + bend = au_ibend(inode);
4555 + for (bindex++; bindex <= bend; bindex++) {
4556 + h_inode = au_h_iptr(inode, bindex);
4558 + au_add_nlink(inode, h_inode);
4563 +void au_cpup_attr_changeable(struct inode *inode)
4565 + struct inode *h_inode;
4567 + h_inode = au_h_iptr(inode, au_ibstart(inode));
4568 + inode->i_mode = h_inode->i_mode;
4569 + inode->i_uid = h_inode->i_uid;
4570 + inode->i_gid = h_inode->i_gid;
4571 + au_cpup_attr_timesizes(inode);
4572 + au_cpup_attr_flags(inode, h_inode->i_flags);
4575 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4577 + struct au_iinfo *iinfo = au_ii(inode);
4579 + IiMustWriteLock(inode);
4581 + iinfo->ii_higen = h_inode->i_generation;
4582 + iinfo->ii_hsb1 = h_inode->i_sb;
4585 +void au_cpup_attr_all(struct inode *inode, int force)
4587 + struct inode *h_inode;
4589 + h_inode = au_h_iptr(inode, au_ibstart(inode));
4590 + au_cpup_attr_changeable(inode);
4591 + if (inode->i_nlink > 0)
4592 + au_cpup_attr_nlink(inode, force);
4593 + inode->i_rdev = h_inode->i_rdev;
4594 + inode->i_blkbits = h_inode->i_blkbits;
4595 + au_cpup_igen(inode, h_inode);
4598 +/* ---------------------------------------------------------------------- */
4600 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4602 +/* keep the timestamps of the parent dir when cpup */
4603 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4604 + struct path *h_path)
4606 + struct inode *h_inode;
4608 + dt->dt_dentry = dentry;
4609 + dt->dt_h_path = *h_path;
4610 + h_inode = d_inode(h_path->dentry);
4611 + dt->dt_atime = h_inode->i_atime;
4612 + dt->dt_mtime = h_inode->i_mtime;
4616 +void au_dtime_revert(struct au_dtime *dt)
4618 + struct iattr attr;
4621 + attr.ia_atime = dt->dt_atime;
4622 + attr.ia_mtime = dt->dt_mtime;
4623 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4624 + | ATTR_ATIME | ATTR_ATIME_SET;
4626 + /* no delegation since this is a directory */
4627 + err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4628 + if (unlikely(err))
4629 + pr_warn("restoring timestamps failed(%d). ignored\n", err);
4632 +/* ---------------------------------------------------------------------- */
4634 +/* internal use only */
4635 +struct au_cpup_reg_attr {
4638 + unsigned int iflags; /* inode->i_flags */
4641 +static noinline_for_stack
4642 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct dentry *h_src,
4643 + struct au_cpup_reg_attr *h_src_attr)
4645 + int err, sbits, icex;
4646 + unsigned int mnt_flags;
4647 + unsigned char verbose;
4649 + struct path h_path;
4650 + struct inode *h_isrc, *h_idst;
4651 + struct kstat *h_st;
4652 + struct au_branch *br;
4654 + h_path.dentry = au_h_dptr(dst, bindex);
4655 + h_idst = d_inode(h_path.dentry);
4656 + br = au_sbr(dst->d_sb, bindex);
4657 + h_path.mnt = au_br_mnt(br);
4658 + h_isrc = d_inode(h_src);
4659 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4660 + | ATTR_ATIME | ATTR_MTIME
4661 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
4662 + if (h_src_attr && h_src_attr->valid) {
4663 + h_st = &h_src_attr->st;
4664 + ia.ia_uid = h_st->uid;
4665 + ia.ia_gid = h_st->gid;
4666 + ia.ia_atime = h_st->atime;
4667 + ia.ia_mtime = h_st->mtime;
4668 + if (h_idst->i_mode != h_st->mode
4669 + && !S_ISLNK(h_idst->i_mode)) {
4670 + ia.ia_valid |= ATTR_MODE;
4671 + ia.ia_mode = h_st->mode;
4673 + sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4674 + au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4676 + ia.ia_uid = h_isrc->i_uid;
4677 + ia.ia_gid = h_isrc->i_gid;
4678 + ia.ia_atime = h_isrc->i_atime;
4679 + ia.ia_mtime = h_isrc->i_mtime;
4680 + if (h_idst->i_mode != h_isrc->i_mode
4681 + && !S_ISLNK(h_idst->i_mode)) {
4682 + ia.ia_valid |= ATTR_MODE;
4683 + ia.ia_mode = h_isrc->i_mode;
4685 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4686 + au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4688 + /* no delegation since it is just created */
4689 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4691 + /* is this nfs only? */
4692 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4693 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4694 + ia.ia_mode = h_isrc->i_mode;
4695 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4698 + icex = br->br_perm & AuBrAttr_ICEX;
4700 + mnt_flags = au_mntflags(dst->d_sb);
4701 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4702 + err = au_cpup_xattr(h_path.dentry, h_src, icex, verbose);
4708 +/* ---------------------------------------------------------------------- */
4710 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4711 + char *buf, unsigned long blksize)
4714 + size_t sz, rbytes, wbytes;
4715 + unsigned char all_zero;
4717 + struct mutex *h_mtx;
4718 + /* reduce stack usage */
4721 + zp = page_address(ZERO_PAGE(0));
4722 + if (unlikely(!zp))
4723 + return -ENOMEM; /* possible? */
4728 + AuDbg("len %lld\n", len);
4730 + if (len < blksize)
4734 + /* todo: signal_pending? */
4735 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
4736 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
4739 + if (unlikely(err < 0))
4743 + if (len >= rbytes && rbytes == blksize)
4744 + all_zero = !memcmp(buf, zp, rbytes);
4751 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
4753 + /* todo: signal_pending? */
4754 + if (unlikely(err == -EAGAIN || err == -EINTR))
4756 + if (unlikely(err < 0))
4761 + if (unlikely(err < 0))
4767 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
4769 + if (unlikely(res < 0))
4776 + /* the last block may be a hole */
4777 + if (!err && all_zero) {
4778 + AuLabel(last hole);
4781 + if (au_test_nfs(dst->f_path.dentry->d_sb)) {
4782 + /* nfs requires this step to make last hole */
4783 + /* is this only nfs? */
4785 + /* todo: signal_pending? */
4786 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
4787 + } while (err == -EAGAIN || err == -EINTR);
4794 + ia->ia_size = dst->f_pos;
4795 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
4796 + ia->ia_file = dst;
4797 + h_mtx = &file_inode(dst)->i_mutex;
4798 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
4799 + /* no delegation since it is just created */
4800 + err = vfsub_notify_change(&dst->f_path, ia,
4801 + /*delegated*/NULL);
4802 + mutex_unlock(h_mtx);
4809 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
4812 + unsigned long blksize;
4813 + unsigned char do_kfree;
4817 + blksize = dst->f_path.dentry->d_sb->s_blocksize;
4818 + if (!blksize || PAGE_SIZE < blksize)
4819 + blksize = PAGE_SIZE;
4820 + AuDbg("blksize %lu\n", blksize);
4821 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
4823 + buf = kmalloc(blksize, GFP_NOFS);
4825 + buf = (void *)__get_free_page(GFP_NOFS);
4826 + if (unlikely(!buf))
4829 + if (len > (1 << 22))
4830 + AuDbg("copying a large file %lld\n", (long long)len);
4834 + err = au_do_copy_file(dst, src, len, buf, blksize);
4838 + free_page((unsigned long)buf);
4845 + * to support a sparse file which is opened with O_APPEND,
4846 + * we need to close the file.
4848 +static int au_cp_regular(struct au_cp_generic *cpg)
4851 + enum { SRC, DST };
4853 + aufs_bindex_t bindex;
4854 + unsigned int flags;
4855 + struct dentry *dentry;
4857 + struct file *file;
4861 + .bindex = cpg->bsrc,
4862 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
4866 + .bindex = cpg->bdst,
4867 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
4868 + .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
4869 + .label = &&out_src
4872 + struct super_block *sb;
4874 + /* bsrc branch can be ro/rw. */
4875 + sb = cpg->dentry->d_sb;
4877 + for (i = 0; i < 2; i++, f++) {
4878 + f->dentry = au_h_dptr(cpg->dentry, f->bindex);
4879 + f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
4880 + /*file*/NULL, f->force_wr);
4881 + err = PTR_ERR(f->file);
4882 + if (IS_ERR(f->file))
4886 + /* try stopping to update while we copyup */
4887 + IMustLock(d_inode(file[SRC].dentry));
4888 + err = au_copy_file(file[DST].file, file[SRC].file, cpg->len);
4890 + fput(file[DST].file);
4891 + au_sbr_put(sb, file[DST].bindex);
4894 + fput(file[SRC].file);
4895 + au_sbr_put(sb, file[SRC].bindex);
4900 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
4901 + struct au_cpup_reg_attr *h_src_attr)
4905 + struct path h_path;
4906 + struct inode *h_src_inode, *h_dst_inode;
4909 + h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
4910 + l = i_size_read(h_src_inode);
4911 + if (cpg->len == -1 || l < cpg->len)
4914 + /* try stopping to update while we are referencing */
4915 + mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
4916 + au_pin_hdir_unlock(cpg->pin);
4918 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
4919 + h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
4920 + h_src_attr->iflags = h_src_inode->i_flags;
4921 + if (!au_test_nfs(h_src_inode->i_sb))
4922 + err = vfs_getattr(&h_path, &h_src_attr->st);
4924 + mutex_unlock(&h_src_inode->i_mutex);
4925 + err = vfs_getattr(&h_path, &h_src_attr->st);
4926 + mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
4928 + if (unlikely(err)) {
4929 + mutex_unlock(&h_src_inode->i_mutex);
4932 + h_src_attr->valid = 1;
4933 + err = au_cp_regular(cpg);
4934 + mutex_unlock(&h_src_inode->i_mutex);
4935 + rerr = au_pin_hdir_relock(cpg->pin);
4939 + if (!err && (h_src_inode->i_state & I_LINKABLE)) {
4940 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
4941 + h_dst_inode = d_inode(h_path.dentry);
4942 + spin_lock(&h_dst_inode->i_lock);
4943 + h_dst_inode->i_state |= I_LINKABLE;
4944 + spin_unlock(&h_dst_inode->i_lock);
4951 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
4952 + struct inode *h_dir)
4955 + mm_segment_t old_fs;
4960 + struct inode *h_inode = d_inode(h_src);
4961 + const struct inode_operations *h_iop = h_inode->i_op;
4964 + if (unlikely(!h_iop->readlink))
4968 + sym.k = (void *)__get_free_page(GFP_NOFS);
4969 + if (unlikely(!sym.k))
4972 + /* unnecessary to support mmap_sem since symlink is not mmap-able */
4973 + old_fs = get_fs();
4974 + set_fs(KERNEL_DS);
4975 + symlen = h_iop->readlink(h_src, sym.u, PATH_MAX);
4980 + sym.k[symlen] = 0;
4981 + err = vfsub_symlink(h_dir, h_path, sym.k);
4983 + free_page((unsigned long)sym.k);
4989 +static noinline_for_stack
4990 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
4991 + struct au_cpup_reg_attr *h_src_attr)
4995 + unsigned int mnt_flags;
4996 + unsigned char isdir, isreg, force;
4997 + const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
4998 + struct au_dtime dt;
4999 + struct path h_path;
5000 + struct dentry *h_src, *h_dst, *h_parent;
5001 + struct inode *h_inode, *h_dir, *dir, *inode;
5002 + struct super_block *sb;
5004 + /* bsrc branch can be ro/rw. */
5005 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5006 + h_inode = d_inode(h_src);
5007 + AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5009 + /* try stopping to be referenced while we are creating */
5010 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5011 + if (au_ftest_cpup(cpg->flags, RENAME))
5012 + AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5013 + AUFS_WH_PFX_LEN));
5014 + h_parent = h_dst->d_parent; /* dir inode is locked */
5015 + h_dir = d_inode(h_parent);
5017 + AuDebugOn(h_parent != h_dst->d_parent);
5019 + sb = cpg->dentry->d_sb;
5020 + h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5022 + h_path.dentry = h_parent;
5023 + au_dtime_store(&dt, dst_parent, &h_path);
5025 + h_path.dentry = h_dst;
5029 + mode = h_inode->i_mode;
5030 + switch (mode & S_IFMT) {
5033 + err = vfsub_create(h_dir, &h_path, mode | S_IWUSR,
5034 + /*want_excl*/true);
5036 + err = au_do_cpup_regular(cpg, h_src_attr);
5040 + err = vfsub_mkdir(h_dir, &h_path, mode);
5043 + * strange behaviour from the users view,
5044 + * particularry setattr case
5046 + dir = d_inode(dst_parent);
5047 + if (au_ibstart(dir) == cpg->bdst)
5048 + au_cpup_attr_nlink(dir, /*force*/1);
5049 + inode = d_inode(cpg->dentry);
5050 + au_cpup_attr_nlink(inode, /*force*/1);
5054 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5058 + AuDebugOn(!capable(CAP_MKNOD));
5062 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5065 + AuIOErr("Unknown inode type 0%o\n", mode);
5069 + mnt_flags = au_mntflags(sb);
5070 + if (!au_opt_test(mnt_flags, UDBA_NONE)
5072 + && au_opt_test(mnt_flags, XINO)
5073 + && (h_inode->i_nlink == 1
5074 + || (h_inode->i_state & I_LINKABLE))
5075 + /* todo: unnecessary? */
5076 + /* && d_inode(cpg->dentry)->i_nlink == 1 */
5077 + && cpg->bdst < cpg->bsrc
5078 + && !au_ftest_cpup(cpg->flags, KEEPLINO))
5079 + au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5080 + /* ignore this error */
5085 + force = !!cpg->len;
5086 + if (cpg->len == -1)
5087 + force = !!i_size_read(h_inode);
5089 + au_fhsm_wrote(sb, cpg->bdst, force);
5093 + au_dtime_revert(&dt);
5097 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5100 + struct dentry *dentry, *h_dentry, *h_parent, *parent;
5101 + struct inode *h_dir;
5102 + aufs_bindex_t bdst;
5104 + dentry = cpg->dentry;
5106 + h_dentry = au_h_dptr(dentry, bdst);
5107 + if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5109 + au_set_h_dptr(dentry, bdst, NULL);
5110 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
5112 + h_path->dentry = dget(au_h_dptr(dentry, bdst));
5113 + au_set_h_dptr(dentry, bdst, h_dentry);
5116 + parent = dget_parent(dentry);
5117 + h_parent = au_h_dptr(parent, bdst);
5119 + h_path->dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
5120 + if (IS_ERR(h_path->dentry))
5121 + err = PTR_ERR(h_path->dentry);
5123 + if (unlikely(err))
5126 + h_parent = h_dentry->d_parent; /* dir inode is locked */
5127 + h_dir = d_inode(h_parent);
5129 + AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5130 + /* no delegation since it is just created */
5131 + err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL);
5132 + dput(h_path->dentry);
5139 + * copyup the @dentry from @bsrc to @bdst.
5140 + * the caller must set the both of lower dentries.
5141 + * @len is for truncating when it is -1 copyup the entire file.
5142 + * in link/rename cases, @dst_parent may be different from the real one.
5143 + * basic->bsrc can be larger than basic->bdst.
5145 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5148 + aufs_bindex_t old_ibstart;
5149 + unsigned char isdir, plink;
5150 + struct dentry *h_src, *h_dst, *h_parent;
5151 + struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5152 + struct super_block *sb;
5153 + struct au_branch *br;
5154 + /* to reuduce stack size */
5156 + struct au_dtime dt;
5157 + struct path h_path;
5158 + struct au_cpup_reg_attr h_src_attr;
5162 + a = kmalloc(sizeof(*a), GFP_NOFS);
5165 + a->h_src_attr.valid = 0;
5167 + sb = cpg->dentry->d_sb;
5168 + br = au_sbr(sb, cpg->bdst);
5169 + a->h_path.mnt = au_br_mnt(br);
5170 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5171 + h_parent = h_dst->d_parent; /* dir inode is locked */
5172 + h_dir = d_inode(h_parent);
5175 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5176 + inode = d_inode(cpg->dentry);
5179 + dst_parent = dget_parent(cpg->dentry);
5183 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
5184 + dst_inode = au_h_iptr(inode, cpg->bdst);
5186 + if (unlikely(!plink)) {
5188 + AuIOErr("hi%lu(i%lu) exists on b%d "
5189 + "but plink is disabled\n",
5190 + dst_inode->i_ino, inode->i_ino, cpg->bdst);
5194 + if (dst_inode->i_nlink) {
5195 + const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5197 + h_src = au_plink_lkup(inode, cpg->bdst);
5198 + err = PTR_ERR(h_src);
5199 + if (IS_ERR(h_src))
5201 + if (unlikely(d_is_negative(h_src))) {
5203 + AuIOErr("i%lu exists on b%d "
5204 + "but not pseudo-linked\n",
5205 + inode->i_ino, cpg->bdst);
5211 + a->h_path.dentry = h_parent;
5212 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5215 + a->h_path.dentry = h_dst;
5217 + err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5218 + if (!err && au_ftest_cpup(cpg->flags, RENAME))
5219 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5221 + au_dtime_revert(&a->dt);
5222 + if (unlikely(err == -EWOULDBLOCK)) {
5223 + pr_warn("cannot retry for NFSv4 delegation"
5224 + " for an internal link\n");
5230 + /* todo: cpup_wh_file? */
5232 + au_update_ibrange(inode, /*do_put_zero*/1);
5235 + isdir = S_ISDIR(inode->i_mode);
5236 + old_ibstart = au_ibstart(inode);
5237 + err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5238 + if (unlikely(err))
5240 + dst_inode = d_inode(h_dst);
5241 + mutex_lock_nested(&dst_inode->i_mutex, AuLsc_I_CHILD2);
5242 + /* todo: necessary? */
5243 + /* au_pin_hdir_unlock(cpg->pin); */
5245 + err = cpup_iattr(cpg->dentry, cpg->bdst, h_src, &a->h_src_attr);
5246 + if (unlikely(err)) {
5247 + /* todo: necessary? */
5248 + /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5249 + mutex_unlock(&dst_inode->i_mutex);
5253 + if (cpg->bdst < old_ibstart) {
5254 + if (S_ISREG(inode->i_mode)) {
5255 + err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5256 + if (unlikely(err)) {
5257 + /* ignore an error */
5258 + /* au_pin_hdir_relock(cpg->pin); */
5259 + mutex_unlock(&dst_inode->i_mutex);
5263 + au_set_ibstart(inode, cpg->bdst);
5265 + au_set_ibend(inode, cpg->bdst);
5266 + au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5267 + au_hi_flags(inode, isdir));
5269 + /* todo: necessary? */
5270 + /* err = au_pin_hdir_relock(cpg->pin); */
5271 + mutex_unlock(&dst_inode->i_mutex);
5272 + if (unlikely(err))
5275 + src_inode = d_inode(h_src);
5277 + && (src_inode->i_nlink > 1
5278 + || src_inode->i_state & I_LINKABLE)
5280 + au_plink_append(inode, cpg->bdst, h_dst);
5282 + if (au_ftest_cpup(cpg->flags, RENAME)) {
5283 + a->h_path.dentry = h_dst;
5284 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5287 + goto out_parent; /* success */
5291 + a->h_path.dentry = h_parent;
5292 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5293 + a->h_path.dentry = h_dst;
5295 + if (d_is_positive(h_dst)) {
5297 + /* no delegation since it is just created */
5298 + rerr = vfsub_unlink(h_dir, &a->h_path,
5299 + /*delegated*/NULL, /*force*/0);
5301 + rerr = vfsub_rmdir(h_dir, &a->h_path);
5303 + au_dtime_revert(&a->dt);
5305 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5315 +#if 0 /* reserved */
5316 +struct au_cpup_single_args {
5318 + struct au_cp_generic *cpg;
5319 + struct dentry *dst_parent;
5322 +static void au_call_cpup_single(void *args)
5324 + struct au_cpup_single_args *a = args;
5326 + au_pin_hdir_acquire_nest(a->cpg->pin);
5327 + *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5328 + au_pin_hdir_release(a->cpg->pin);
5333 + * prevent SIGXFSZ in copy-up.
5334 + * testing CAP_MKNOD is for generic fs,
5335 + * but CAP_FSETID is for xfs only, currently.
5337 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5340 + struct super_block *sb;
5341 + struct inode *h_dir;
5344 + sb = au_pinned_parent(pin)->d_sb;
5345 + if (!au_wkq_test()
5346 + && (!au_sbi(sb)->si_plink_maint_pid
5347 + || au_plink_maint(sb, AuLock_NOPLM))) {
5348 + switch (mode & S_IFMT) {
5350 + /* no condition about RLIMIT_FSIZE and the file size */
5355 + do_sio = !capable(CAP_MKNOD);
5359 + do_sio = ((mode & (S_ISUID | S_ISGID))
5360 + && !capable(CAP_FSETID));
5361 + /* this workaround may be removed in the future */
5363 + h_dir = au_pinned_h_dir(pin);
5364 + do_sio = h_dir->i_mode & S_ISVTX;
5371 +#if 0 /* reserved */
5372 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5375 + struct dentry *h_dentry;
5377 + h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5378 + if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5379 + err = au_cpup_single(cpg, dst_parent);
5381 + struct au_cpup_single_args args = {
5384 + .dst_parent = dst_parent
5386 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5387 + if (unlikely(wkq_err))
5396 + * copyup the @dentry from the first active lower branch to @bdst,
5397 + * using au_cpup_single().
5399 +static int au_cpup_simple(struct au_cp_generic *cpg)
5402 + unsigned int flags_orig;
5403 + struct dentry *dentry;
5405 + AuDebugOn(cpg->bsrc < 0);
5407 + dentry = cpg->dentry;
5408 + DiMustWriteLock(dentry);
5410 + err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5412 + flags_orig = cpg->flags;
5413 + au_fset_cpup(cpg->flags, RENAME);
5414 + err = au_cpup_single(cpg, NULL);
5415 + cpg->flags = flags_orig;
5417 + return 0; /* success */
5420 + au_set_h_dptr(dentry, cpg->bdst, NULL);
5421 + au_set_dbstart(dentry, cpg->bsrc);
5427 +struct au_cpup_simple_args {
5429 + struct au_cp_generic *cpg;
5432 +static void au_call_cpup_simple(void *args)
5434 + struct au_cpup_simple_args *a = args;
5436 + au_pin_hdir_acquire_nest(a->cpg->pin);
5437 + *a->errp = au_cpup_simple(a->cpg);
5438 + au_pin_hdir_release(a->cpg->pin);
5441 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5444 + struct dentry *dentry, *parent;
5445 + struct file *h_file;
5446 + struct inode *h_dir;
5448 + dentry = cpg->dentry;
5450 + if (au_ftest_cpup(cpg->flags, HOPEN)) {
5451 + AuDebugOn(cpg->bsrc < 0);
5452 + h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5453 + err = PTR_ERR(h_file);
5454 + if (IS_ERR(h_file))
5458 + parent = dget_parent(dentry);
5459 + h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5460 + if (!au_test_h_perm_sio(h_dir, MAY_EXEC | MAY_WRITE)
5461 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5462 + err = au_cpup_simple(cpg);
5464 + struct au_cpup_simple_args args = {
5468 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5469 + if (unlikely(wkq_err))
5475 + au_h_open_post(dentry, cpg->bsrc, h_file);
5481 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5483 + aufs_bindex_t bsrc, bend;
5484 + struct dentry *dentry, *h_dentry;
5486 + if (cpg->bsrc < 0) {
5487 + dentry = cpg->dentry;
5488 + bend = au_dbend(dentry);
5489 + for (bsrc = cpg->bdst + 1; bsrc <= bend; bsrc++) {
5490 + h_dentry = au_h_dptr(dentry, bsrc);
5492 + AuDebugOn(d_is_negative(h_dentry));
5496 + AuDebugOn(bsrc > bend);
5499 + AuDebugOn(cpg->bsrc <= cpg->bdst);
5500 + return au_do_sio_cpup_simple(cpg);
5503 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5505 + AuDebugOn(cpg->bdst <= cpg->bsrc);
5506 + return au_do_sio_cpup_simple(cpg);
5509 +/* ---------------------------------------------------------------------- */
5512 + * copyup the deleted file for writing.
5514 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5515 + struct file *file)
5518 + unsigned int flags_orig;
5519 + aufs_bindex_t bsrc_orig;
5520 + struct dentry *h_d_dst, *h_d_start;
5521 + struct au_dinfo *dinfo;
5522 + struct au_hdentry *hdp;
5524 + dinfo = au_di(cpg->dentry);
5525 + AuRwMustWriteLock(&dinfo->di_rwsem);
5527 + bsrc_orig = cpg->bsrc;
5528 + cpg->bsrc = dinfo->di_bstart;
5529 + hdp = dinfo->di_hdentry;
5530 + h_d_dst = hdp[0 + cpg->bdst].hd_dentry;
5531 + dinfo->di_bstart = cpg->bdst;
5532 + hdp[0 + cpg->bdst].hd_dentry = wh_dentry;
5535 + h_d_start = hdp[0 + cpg->bsrc].hd_dentry;
5536 + hdp[0 + cpg->bsrc].hd_dentry = au_hf_top(file)->f_path.dentry;
5538 + flags_orig = cpg->flags;
5539 + cpg->flags = !AuCpup_DTIME;
5540 + err = au_cpup_single(cpg, /*h_parent*/NULL);
5541 + cpg->flags = flags_orig;
5544 + err = au_reopen_nondir(file);
5545 + hdp[0 + cpg->bsrc].hd_dentry = h_d_start;
5547 + hdp[0 + cpg->bdst].hd_dentry = h_d_dst;
5548 + dinfo->di_bstart = cpg->bsrc;
5549 + cpg->bsrc = bsrc_orig;
5554 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5557 + aufs_bindex_t bdst;
5558 + struct au_dtime dt;
5559 + struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5560 + struct au_branch *br;
5561 + struct path h_path;
5563 + dentry = cpg->dentry;
5565 + br = au_sbr(dentry->d_sb, bdst);
5566 + parent = dget_parent(dentry);
5567 + h_parent = au_h_dptr(parent, bdst);
5568 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5569 + err = PTR_ERR(wh_dentry);
5570 + if (IS_ERR(wh_dentry))
5573 + h_path.dentry = h_parent;
5574 + h_path.mnt = au_br_mnt(br);
5575 + au_dtime_store(&dt, parent, &h_path);
5576 + err = au_do_cpup_wh(cpg, wh_dentry, file);
5577 + if (unlikely(err))
5581 + h_path.dentry = wh_dentry;
5582 + if (!d_is_dir(wh_dentry)) {
5583 + /* no delegation since it is just created */
5584 + err = vfsub_unlink(d_inode(h_parent), &h_path,
5585 + /*delegated*/NULL, /*force*/0);
5587 + err = vfsub_rmdir(d_inode(h_parent), &h_path);
5588 + if (unlikely(err)) {
5589 + AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
5593 + au_dtime_revert(&dt);
5594 + au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
5603 +struct au_cpup_wh_args {
5605 + struct au_cp_generic *cpg;
5606 + struct file *file;
5609 +static void au_call_cpup_wh(void *args)
5611 + struct au_cpup_wh_args *a = args;
5613 + au_pin_hdir_acquire_nest(a->cpg->pin);
5614 + *a->errp = au_cpup_wh(a->cpg, a->file);
5615 + au_pin_hdir_release(a->cpg->pin);
5618 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5621 + aufs_bindex_t bdst;
5622 + struct dentry *dentry, *parent, *h_orph, *h_parent;
5623 + struct inode *dir, *h_dir, *h_tmpdir;
5624 + struct au_wbr *wbr;
5625 + struct au_pin wh_pin, *pin_orig;
5627 + dentry = cpg->dentry;
5629 + parent = dget_parent(dentry);
5630 + dir = d_inode(parent);
5633 + h_dir = au_igrab(au_h_iptr(dir, bdst));
5636 + if (!h_dir->i_nlink) {
5637 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
5638 + h_orph = wbr->wbr_orph;
5640 + h_parent = dget(au_h_dptr(parent, bdst));
5641 + au_set_h_dptr(parent, bdst, dget(h_orph));
5642 + h_tmpdir = d_inode(h_orph);
5643 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
5645 + mutex_lock_nested(&h_tmpdir->i_mutex, AuLsc_I_PARENT3);
5646 + /* todo: au_h_open_pre()? */
5648 + pin_orig = cpg->pin;
5649 + au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
5650 + AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
5651 + cpg->pin = &wh_pin;
5654 + if (!au_test_h_perm_sio(h_tmpdir, MAY_EXEC | MAY_WRITE)
5655 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5656 + err = au_cpup_wh(cpg, file);
5658 + struct au_cpup_wh_args args = {
5663 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
5664 + if (unlikely(wkq_err))
5669 + mutex_unlock(&h_tmpdir->i_mutex);
5670 + /* todo: au_h_open_post()? */
5671 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
5672 + au_set_h_dptr(parent, bdst, h_parent);
5673 + AuDebugOn(!pin_orig);
5674 + cpg->pin = pin_orig;
5682 +/* ---------------------------------------------------------------------- */
5685 + * generic routine for both of copy-up and copy-down.
5687 +/* cf. revalidate function in file.c */
5688 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5689 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
5690 + struct au_pin *pin,
5691 + struct dentry *h_parent, void *arg),
5695 + struct au_pin pin;
5696 + struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
5699 + parent = dget_parent(dentry);
5700 + if (IS_ROOT(parent))
5703 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
5704 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
5706 + /* do not use au_dpage */
5707 + real_parent = parent;
5710 + parent = dget_parent(dentry);
5711 + h_parent = au_h_dptr(parent, bdst);
5713 + goto out; /* success */
5715 + /* find top dir which is necessary to cpup */
5719 + parent = dget_parent(d);
5720 + di_read_lock_parent3(parent, !AuLock_IR);
5721 + h_parent = au_h_dptr(parent, bdst);
5722 + di_read_unlock(parent, !AuLock_IR);
5723 + } while (!h_parent);
5725 + if (d != real_parent)
5726 + di_write_lock_child3(d);
5728 + /* somebody else might create while we were sleeping */
5729 + h_dentry = au_h_dptr(d, bdst);
5730 + if (!h_dentry || d_is_negative(h_dentry)) {
5732 + au_update_dbstart(d);
5734 + au_pin_set_dentry(&pin, d);
5735 + err = au_do_pin(&pin);
5737 + err = cp(d, bdst, &pin, h_parent, arg);
5742 + if (d != real_parent)
5743 + di_write_unlock(d);
5744 + if (unlikely(err))
5753 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
5754 + struct au_pin *pin,
5755 + struct dentry *h_parent __maybe_unused,
5756 + void *arg __maybe_unused)
5758 + struct au_cp_generic cpg = {
5764 + .flags = AuCpup_DTIME
5766 + return au_sio_cpup_simple(&cpg);
5769 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5771 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
5774 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5777 + struct dentry *parent;
5778 + struct inode *dir;
5780 + parent = dget_parent(dentry);
5781 + dir = d_inode(parent);
5783 + if (au_h_iptr(dir, bdst))
5786 + di_read_unlock(parent, AuLock_IR);
5787 + di_write_lock_parent(parent);
5788 + /* someone else might change our inode while we were sleeping */
5789 + if (!au_h_iptr(dir, bdst))
5790 + err = au_cpup_dirs(dentry, bdst);
5791 + di_downgrade_lock(parent, AuLock_IR);
5797 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
5798 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
5799 +++ linux/fs/aufs/cpup.h 2016-01-13 20:11:11.666426853 +0100
5802 + * Copyright (C) 2005-2015 Junjiro R. Okajima
5804 + * This program, aufs is free software; you can redistribute it and/or modify
5805 + * it under the terms of the GNU General Public License as published by
5806 + * the Free Software Foundation; either version 2 of the License, or
5807 + * (at your option) any later version.
5809 + * This program is distributed in the hope that it will be useful,
5810 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5811 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5812 + * GNU General Public License for more details.
5814 + * You should have received a copy of the GNU General Public License
5815 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
5819 + * copy-up/down functions
5822 +#ifndef __AUFS_CPUP_H__
5823 +#define __AUFS_CPUP_H__
5827 +#include <linux/path.h>
5833 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
5834 +void au_cpup_attr_timesizes(struct inode *inode);
5835 +void au_cpup_attr_nlink(struct inode *inode, int force);
5836 +void au_cpup_attr_changeable(struct inode *inode);
5837 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
5838 +void au_cpup_attr_all(struct inode *inode, int force);
5840 +/* ---------------------------------------------------------------------- */
5842 +struct au_cp_generic {
5843 + struct dentry *dentry;
5844 + aufs_bindex_t bdst, bsrc;
5846 + struct au_pin *pin;
5847 + unsigned int flags;
5851 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
5852 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
5854 +#define AuCpup_RENAME (1 << 2) /* rename after cpup */
5855 +#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
5857 +#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
5859 +#define AuCpup_RWDST (1 << 5) /* force write target even if
5860 + the branch is marked as RO */
5862 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
5863 +#define au_fset_cpup(flags, name) \
5864 + do { (flags) |= AuCpup_##name; } while (0)
5865 +#define au_fclr_cpup(flags, name) \
5866 + do { (flags) &= ~AuCpup_##name; } while (0)
5868 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
5869 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
5870 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
5871 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
5873 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5874 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
5875 + struct au_pin *pin,
5876 + struct dentry *h_parent, void *arg),
5878 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5879 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5881 +/* ---------------------------------------------------------------------- */
5883 +/* keep timestamps when copyup */
5885 + struct dentry *dt_dentry;
5886 + struct path dt_h_path;
5887 + struct timespec dt_atime, dt_mtime;
5889 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
5890 + struct path *h_path);
5891 +void au_dtime_revert(struct au_dtime *dt);
5893 +#endif /* __KERNEL__ */
5894 +#endif /* __AUFS_CPUP_H__ */
5895 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
5896 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
5897 +++ linux/fs/aufs/dbgaufs.c 2016-01-13 20:11:11.666426853 +0100
5900 + * Copyright (C) 2005-2015 Junjiro R. Okajima
5902 + * This program, aufs is free software; you can redistribute it and/or modify
5903 + * it under the terms of the GNU General Public License as published by
5904 + * the Free Software Foundation; either version 2 of the License, or
5905 + * (at your option) any later version.
5907 + * This program is distributed in the hope that it will be useful,
5908 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5909 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5910 + * GNU General Public License for more details.
5912 + * You should have received a copy of the GNU General Public License
5913 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
5917 + * debugfs interface
5920 +#include <linux/debugfs.h>
5923 +#ifndef CONFIG_SYSFS
5924 +#error DEBUG_FS depends upon SYSFS
5927 +static struct dentry *dbgaufs;
5928 +static const mode_t dbgaufs_mode = S_IRUSR | S_IRGRP | S_IROTH;
5930 +/* 20 is max digits length of ulong 64 */
5931 +struct dbgaufs_arg {
5937 + * common function for all XINO files
5939 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
5940 + struct file *file)
5942 + kfree(file->private_data);
5946 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt)
5950 + struct dbgaufs_arg *p;
5953 + p = kmalloc(sizeof(*p), GFP_NOFS);
5959 + file->private_data = p;
5963 + err = vfs_getattr(&xf->f_path, &st);
5967 + (p->a, sizeof(p->a), "%ld, %llux%lu %lld\n",
5968 + (long)file_count(xf), st.blocks, st.blksize,
5969 + (long long)st.size);
5971 + p->n = snprintf(p->a, sizeof(p->a), "%llux%lu %lld\n",
5972 + st.blocks, st.blksize,
5973 + (long long)st.size);
5974 + AuDebugOn(p->n >= sizeof(p->a));
5976 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
5985 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
5986 + size_t count, loff_t *ppos)
5988 + struct dbgaufs_arg *p;
5990 + p = file->private_data;
5991 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
5994 +/* ---------------------------------------------------------------------- */
5996 +struct dbgaufs_plink_arg {
6001 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6002 + struct file *file)
6004 + free_page((unsigned long)file->private_data);
6008 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6010 + int err, i, limit;
6011 + unsigned long n, sum;
6012 + struct dbgaufs_plink_arg *p;
6013 + struct au_sbinfo *sbinfo;
6014 + struct super_block *sb;
6015 + struct au_sphlhead *sphl;
6018 + p = (void *)get_zeroed_page(GFP_NOFS);
6023 + sbinfo = inode->i_private;
6024 + sb = sbinfo->si_sb;
6025 + si_noflush_read_lock(sb);
6026 + if (au_opt_test(au_mntflags(sb), PLINK)) {
6027 + limit = PAGE_SIZE - sizeof(p->n);
6029 + /* the number of buckets */
6030 + n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6035 + for (i = 0, sphl = sbinfo->si_plink;
6036 + i < AuPlink_NHASH;
6038 + n = au_sphl_count(sphl);
6041 + n = snprintf(p->a + p->n, limit, "%lu ", n);
6044 + if (unlikely(limit <= 0))
6047 + p->a[p->n - 1] = '\n';
6049 + /* the sum of plinks */
6050 + n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6053 + if (unlikely(limit <= 0))
6056 +#define str "1\n0\n0\n"
6057 + p->n = sizeof(str) - 1;
6058 + strcpy(p->a, str);
6061 + si_read_unlock(sb);
6064 + file->private_data = p;
6065 + goto out; /* success */
6068 + free_page((unsigned long)p);
6073 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6074 + size_t count, loff_t *ppos)
6076 + struct dbgaufs_plink_arg *p;
6078 + p = file->private_data;
6079 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6082 +static const struct file_operations dbgaufs_plink_fop = {
6083 + .owner = THIS_MODULE,
6084 + .open = dbgaufs_plink_open,
6085 + .release = dbgaufs_plink_release,
6086 + .read = dbgaufs_plink_read
6089 +/* ---------------------------------------------------------------------- */
6091 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6094 + struct au_sbinfo *sbinfo;
6095 + struct super_block *sb;
6097 + sbinfo = inode->i_private;
6098 + sb = sbinfo->si_sb;
6099 + si_noflush_read_lock(sb);
6100 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0);
6101 + si_read_unlock(sb);
6105 +static const struct file_operations dbgaufs_xib_fop = {
6106 + .owner = THIS_MODULE,
6107 + .open = dbgaufs_xib_open,
6108 + .release = dbgaufs_xi_release,
6109 + .read = dbgaufs_xi_read
6112 +/* ---------------------------------------------------------------------- */
6114 +#define DbgaufsXi_PREFIX "xi"
6116 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6120 + struct au_sbinfo *sbinfo;
6121 + struct super_block *sb;
6123 + struct qstr *name;
6127 + name = &file->f_path.dentry->d_name;
6128 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6129 + || memcmp(name->name, DbgaufsXi_PREFIX,
6130 + sizeof(DbgaufsXi_PREFIX) - 1)))
6132 + err = kstrtol(name->name + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6133 + if (unlikely(err))
6136 + sbinfo = inode->i_private;
6137 + sb = sbinfo->si_sb;
6138 + si_noflush_read_lock(sb);
6139 + if (l <= au_sbend(sb)) {
6140 + xf = au_sbr(sb, (aufs_bindex_t)l)->br_xino.xi_file;
6141 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1);
6144 + si_read_unlock(sb);
6150 +static const struct file_operations dbgaufs_xino_fop = {
6151 + .owner = THIS_MODULE,
6152 + .open = dbgaufs_xino_open,
6153 + .release = dbgaufs_xi_release,
6154 + .read = dbgaufs_xi_read
6157 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6159 + aufs_bindex_t bend;
6160 + struct au_branch *br;
6161 + struct au_xino_file *xi;
6163 + if (!au_sbi(sb)->si_dbgaufs)
6166 + bend = au_sbend(sb);
6167 + for (; bindex <= bend; bindex++) {
6168 + br = au_sbr(sb, bindex);
6169 + xi = &br->br_xino;
6170 + debugfs_remove(xi->xi_dbgaufs);
6171 + xi->xi_dbgaufs = NULL;
6175 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
6177 + struct au_sbinfo *sbinfo;
6178 + struct dentry *parent;
6179 + struct au_branch *br;
6180 + struct au_xino_file *xi;
6181 + aufs_bindex_t bend;
6182 + char name[sizeof(DbgaufsXi_PREFIX) + 5]; /* "xi" bindex NULL */
6184 + sbinfo = au_sbi(sb);
6185 + parent = sbinfo->si_dbgaufs;
6189 + bend = au_sbend(sb);
6190 + for (; bindex <= bend; bindex++) {
6191 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6192 + br = au_sbr(sb, bindex);
6193 + xi = &br->br_xino;
6194 + AuDebugOn(xi->xi_dbgaufs);
6195 + xi->xi_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6196 + sbinfo, &dbgaufs_xino_fop);
6197 + /* ignore an error */
6198 + if (unlikely(!xi->xi_dbgaufs))
6199 + AuWarn1("failed %s under debugfs\n", name);
6203 +/* ---------------------------------------------------------------------- */
6205 +#ifdef CONFIG_AUFS_EXPORT
6206 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6209 + struct au_sbinfo *sbinfo;
6210 + struct super_block *sb;
6212 + sbinfo = inode->i_private;
6213 + sb = sbinfo->si_sb;
6214 + si_noflush_read_lock(sb);
6215 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0);
6216 + si_read_unlock(sb);
6220 +static const struct file_operations dbgaufs_xigen_fop = {
6221 + .owner = THIS_MODULE,
6222 + .open = dbgaufs_xigen_open,
6223 + .release = dbgaufs_xi_release,
6224 + .read = dbgaufs_xi_read
6227 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6232 + * This function is a dynamic '__init' function actually,
6233 + * so the tiny check for si_rwsem is unnecessary.
6235 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6238 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
6239 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6240 + &dbgaufs_xigen_fop);
6241 + if (sbinfo->si_dbgaufs_xigen)
6247 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6251 +#endif /* CONFIG_AUFS_EXPORT */
6253 +/* ---------------------------------------------------------------------- */
6255 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6258 + * This function is a dynamic '__fin' function actually,
6259 + * so the tiny check for si_rwsem is unnecessary.
6261 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6263 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
6264 + sbinfo->si_dbgaufs = NULL;
6265 + kobject_put(&sbinfo->si_kobj);
6268 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6271 + char name[SysaufsSiNameLen];
6274 + * This function is a dynamic '__init' function actually,
6275 + * so the tiny check for si_rwsem is unnecessary.
6277 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6281 + AuErr1("/debug/aufs is uninitialized\n");
6286 + sysaufs_name(sbinfo, name);
6287 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6288 + if (unlikely(!sbinfo->si_dbgaufs))
6290 + kobject_get(&sbinfo->si_kobj);
6292 + sbinfo->si_dbgaufs_xib = debugfs_create_file
6293 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6294 + &dbgaufs_xib_fop);
6295 + if (unlikely(!sbinfo->si_dbgaufs_xib))
6298 + sbinfo->si_dbgaufs_plink = debugfs_create_file
6299 + ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6300 + &dbgaufs_plink_fop);
6301 + if (unlikely(!sbinfo->si_dbgaufs_plink))
6304 + err = dbgaufs_xigen_init(sbinfo);
6306 + goto out; /* success */
6309 + dbgaufs_si_fin(sbinfo);
6314 +/* ---------------------------------------------------------------------- */
6316 +void dbgaufs_fin(void)
6318 + debugfs_remove(dbgaufs);
6321 +int __init dbgaufs_init(void)
6326 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6331 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6332 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
6333 +++ linux/fs/aufs/dbgaufs.h 2016-01-13 20:11:11.666426853 +0100
6336 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6338 + * This program, aufs is free software; you can redistribute it and/or modify
6339 + * it under the terms of the GNU General Public License as published by
6340 + * the Free Software Foundation; either version 2 of the License, or
6341 + * (at your option) any later version.
6343 + * This program is distributed in the hope that it will be useful,
6344 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6345 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6346 + * GNU General Public License for more details.
6348 + * You should have received a copy of the GNU General Public License
6349 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6353 + * debugfs interface
6356 +#ifndef __DBGAUFS_H__
6357 +#define __DBGAUFS_H__
6361 +struct super_block;
6364 +#ifdef CONFIG_DEBUG_FS
6366 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6367 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
6368 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6369 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6370 +void dbgaufs_fin(void);
6371 +int __init dbgaufs_init(void);
6373 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6374 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
6375 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6376 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6377 +AuStubVoid(dbgaufs_fin, void)
6378 +AuStubInt0(__init dbgaufs_init, void)
6379 +#endif /* CONFIG_DEBUG_FS */
6381 +#endif /* __KERNEL__ */
6382 +#endif /* __DBGAUFS_H__ */
6383 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6384 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
6385 +++ linux/fs/aufs/dcsub.c 2016-01-13 20:11:11.666426853 +0100
6388 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6390 + * This program, aufs is free software; you can redistribute it and/or modify
6391 + * it under the terms of the GNU General Public License as published by
6392 + * the Free Software Foundation; either version 2 of the License, or
6393 + * (at your option) any later version.
6395 + * This program is distributed in the hope that it will be useful,
6396 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6397 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6398 + * GNU General Public License for more details.
6400 + * You should have received a copy of the GNU General Public License
6401 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6405 + * sub-routines for dentry cache
6410 +static void au_dpage_free(struct au_dpage *dpage)
6413 + struct dentry **p;
6415 + p = dpage->dentries;
6416 + for (i = 0; i < dpage->ndentry; i++)
6418 + free_page((unsigned long)dpage->dentries);
6421 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6427 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6428 + if (unlikely(!dpages->dpages))
6431 + p = (void *)__get_free_page(gfp);
6435 + dpages->dpages[0].ndentry = 0;
6436 + dpages->dpages[0].dentries = p;
6437 + dpages->ndpage = 1;
6438 + return 0; /* success */
6441 + kfree(dpages->dpages);
6446 +void au_dpages_free(struct au_dcsub_pages *dpages)
6449 + struct au_dpage *p;
6451 + p = dpages->dpages;
6452 + for (i = 0; i < dpages->ndpage; i++)
6453 + au_dpage_free(p++);
6454 + kfree(dpages->dpages);
6457 +static int au_dpages_append(struct au_dcsub_pages *dpages,
6458 + struct dentry *dentry, gfp_t gfp)
6461 + struct au_dpage *dpage;
6464 + dpage = dpages->dpages + dpages->ndpage - 1;
6465 + sz = PAGE_SIZE / sizeof(dentry);
6466 + if (unlikely(dpage->ndentry >= sz)) {
6467 + AuLabel(new dpage);
6469 + sz = dpages->ndpage * sizeof(*dpages->dpages);
6470 + p = au_kzrealloc(dpages->dpages, sz,
6471 + sz + sizeof(*dpages->dpages), gfp);
6475 + dpages->dpages = p;
6476 + dpage = dpages->dpages + dpages->ndpage;
6477 + p = (void *)__get_free_page(gfp);
6481 + dpage->ndentry = 0;
6482 + dpage->dentries = p;
6486 + AuDebugOn(au_dcount(dentry) <= 0);
6487 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
6488 + return 0; /* success */
6494 +/* todo: BAD approach */
6495 +/* copied from linux/fs/dcache.c */
6503 +extern void d_walk(struct dentry *parent, void *data,
6504 + enum d_walk_ret (*enter)(void *, struct dentry *),
6505 + void (*finish)(void *));
6507 +struct ac_dpages_arg {
6509 + struct au_dcsub_pages *dpages;
6510 + struct super_block *sb;
6511 + au_dpages_test test;
6515 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
6517 + enum d_walk_ret ret;
6518 + struct ac_dpages_arg *arg = _arg;
6520 + ret = D_WALK_CONTINUE;
6521 + if (dentry->d_sb == arg->sb
6522 + && !IS_ROOT(dentry)
6523 + && au_dcount(dentry) > 0
6525 + && (!arg->test || arg->test(dentry, arg->arg))) {
6526 + arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
6527 + if (unlikely(arg->err))
6528 + ret = D_WALK_QUIT;
6534 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6535 + au_dpages_test test, void *arg)
6537 + struct ac_dpages_arg args = {
6545 + d_walk(root, &args, au_call_dpages_append, NULL);
6550 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6551 + int do_include, au_dpages_test test, void *arg)
6556 + write_seqlock(&rename_lock);
6557 + spin_lock(&dentry->d_lock);
6559 + && au_dcount(dentry) > 0
6560 + && (!test || test(dentry, arg)))
6561 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
6562 + spin_unlock(&dentry->d_lock);
6563 + if (unlikely(err))
6567 + * RCU for vfsmount is unnecessary since this is a traverse in a single
6570 + while (!IS_ROOT(dentry)) {
6571 + dentry = dentry->d_parent; /* rename_lock is locked */
6572 + spin_lock(&dentry->d_lock);
6573 + if (au_dcount(dentry) > 0
6574 + && (!test || test(dentry, arg)))
6575 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
6576 + spin_unlock(&dentry->d_lock);
6577 + if (unlikely(err))
6582 + write_sequnlock(&rename_lock);
6586 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
6588 + return au_di(dentry) && dentry->d_sb == arg;
6591 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6592 + struct dentry *dentry, int do_include)
6594 + return au_dcsub_pages_rev(dpages, dentry, do_include,
6595 + au_dcsub_dpages_aufs, dentry->d_sb);
6598 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
6600 + struct path path[2] = {
6609 + return path_is_under(path + 0, path + 1);
6611 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
6612 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
6613 +++ linux/fs/aufs/dcsub.h 2016-01-13 20:11:11.666426853 +0100
6616 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6618 + * This program, aufs is free software; you can redistribute it and/or modify
6619 + * it under the terms of the GNU General Public License as published by
6620 + * the Free Software Foundation; either version 2 of the License, or
6621 + * (at your option) any later version.
6623 + * This program is distributed in the hope that it will be useful,
6624 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6625 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6626 + * GNU General Public License for more details.
6628 + * You should have received a copy of the GNU General Public License
6629 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6633 + * sub-routines for dentry cache
6636 +#ifndef __AUFS_DCSUB_H__
6637 +#define __AUFS_DCSUB_H__
6641 +#include <linux/dcache.h>
6642 +#include <linux/fs.h>
6646 + struct dentry **dentries;
6649 +struct au_dcsub_pages {
6651 + struct au_dpage *dpages;
6654 +/* ---------------------------------------------------------------------- */
6657 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
6658 +void au_dpages_free(struct au_dcsub_pages *dpages);
6659 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
6660 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6661 + au_dpages_test test, void *arg);
6662 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6663 + int do_include, au_dpages_test test, void *arg);
6664 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6665 + struct dentry *dentry, int do_include);
6666 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
6668 +/* ---------------------------------------------------------------------- */
6671 + * todo: in linux-3.13, several similar (but faster) helpers are added to
6672 + * include/linux/dcache.h. Try them (in the future).
6675 +static inline int au_d_hashed_positive(struct dentry *d)
6678 + struct inode *inode = d_inode(d);
6681 + if (unlikely(d_unhashed(d)
6682 + || d_is_negative(d)
6683 + || !inode->i_nlink))
6688 +static inline int au_d_linkable(struct dentry *d)
6691 + struct inode *inode = d_inode(d);
6693 + err = au_d_hashed_positive(d);
6695 + && d_is_positive(d)
6696 + && (inode->i_state & I_LINKABLE))
6701 +static inline int au_d_alive(struct dentry *d)
6704 + struct inode *inode;
6708 + err = au_d_hashed_positive(d);
6710 + inode = d_inode(d);
6711 + if (unlikely(d_unlinked(d)
6712 + || d_is_negative(d)
6713 + || !inode->i_nlink))
6719 +static inline int au_alive_dir(struct dentry *d)
6723 + err = au_d_alive(d);
6724 + if (unlikely(err || IS_DEADDIR(d_inode(d))))
6729 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
6731 + return a->len == b->len
6732 + && !memcmp(a->name, b->name, a->len);
6737 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
6739 + * the type of d_lockref.count became int, but the inlined function d_count()
6740 + * still returns unsigned int.
6741 + * I don't know why. Maybe it is for every d_count() users?
6742 + * Anyway au_dcount() lives on.
6744 +static inline int au_dcount(struct dentry *d)
6746 + return (int)d_count(d);
6749 +#endif /* __KERNEL__ */
6750 +#endif /* __AUFS_DCSUB_H__ */
6751 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
6752 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
6753 +++ linux/fs/aufs/debug.c 2016-01-13 20:11:11.666426853 +0100
6756 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6758 + * This program, aufs is free software; you can redistribute it and/or modify
6759 + * it under the terms of the GNU General Public License as published by
6760 + * the Free Software Foundation; either version 2 of the License, or
6761 + * (at your option) any later version.
6763 + * This program is distributed in the hope that it will be useful,
6764 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6765 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6766 + * GNU General Public License for more details.
6768 + * You should have received a copy of the GNU General Public License
6769 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6773 + * debug print functions
6778 +/* Returns 0, or -errno. arg is in kp->arg. */
6779 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
6783 + err = kstrtoint(val, 0, &n);
6793 +/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
6794 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
6799 + return sprintf(buffer, "%d", atomic_read(a));
6802 +static struct kernel_param_ops param_ops_atomic_t = {
6803 + .set = param_atomic_t_set,
6804 + .get = param_atomic_t_get
6805 + /* void (*free)(void *arg) */
6808 +atomic_t aufs_debug = ATOMIC_INIT(0);
6809 +MODULE_PARM_DESC(debug, "debug print");
6810 +module_param_named(debug, aufs_debug, atomic_t, S_IRUGO | S_IWUSR | S_IWGRP);
6812 +DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
6813 +char *au_plevel = KERN_DEBUG;
6814 +#define dpri(fmt, ...) do { \
6816 + && strcmp(au_plevel, KERN_DEBUG)) \
6817 + || au_debug_test()) \
6818 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
6821 +/* ---------------------------------------------------------------------- */
6823 +void au_dpri_whlist(struct au_nhash *whlist)
6825 + unsigned long ul, n;
6826 + struct hlist_head *head;
6827 + struct au_vdir_wh *pos;
6829 + n = whlist->nh_num;
6830 + head = whlist->nh_head;
6831 + for (ul = 0; ul < n; ul++) {
6832 + hlist_for_each_entry(pos, head, wh_hash)
6833 + dpri("b%d, %.*s, %d\n",
6835 + pos->wh_str.len, pos->wh_str.name,
6841 +void au_dpri_vdir(struct au_vdir *vdir)
6844 + union au_vdir_deblk_p p;
6847 + if (!vdir || IS_ERR(vdir)) {
6848 + dpri("err %ld\n", PTR_ERR(vdir));
6852 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %lu\n",
6853 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
6854 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
6855 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
6856 + p.deblk = vdir->vd_deblk[ul];
6858 + dpri("[%lu]: %p\n", ul, o);
6862 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
6863 + struct dentry *wh)
6868 + if (!inode || IS_ERR(inode)) {
6869 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
6873 + /* the type of i_blocks depends upon CONFIG_LBDAF */
6874 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
6875 + && sizeof(inode->i_blocks) != sizeof(u64));
6877 + n = (void *)wh->d_name.name;
6878 + l = wh->d_name.len;
6881 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
6882 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
6884 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
6885 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
6886 + i_size_read(inode), (unsigned long long)inode->i_blocks,
6887 + hn, (long long)timespec_to_ns(&inode->i_ctime) & 0x0ffff,
6888 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
6889 + inode->i_state, inode->i_flags, inode->i_version,
6890 + inode->i_generation,
6891 + l ? ", wh " : "", l, n);
6895 +void au_dpri_inode(struct inode *inode)
6897 + struct au_iinfo *iinfo;
6898 + aufs_bindex_t bindex;
6901 + err = do_pri_inode(-1, inode, -1, NULL);
6902 + if (err || !au_test_aufs(inode->i_sb))
6905 + iinfo = au_ii(inode);
6908 + dpri("i-1: bstart %d, bend %d, gen %d\n",
6909 + iinfo->ii_bstart, iinfo->ii_bend, au_iigen(inode, NULL));
6910 + if (iinfo->ii_bstart < 0)
6913 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; bindex++) {
6914 + hn = !!au_hn(iinfo->ii_hinode + bindex);
6915 + do_pri_inode(bindex, iinfo->ii_hinode[0 + bindex].hi_inode, hn,
6916 + iinfo->ii_hinode[0 + bindex].hi_whdentry);
6920 +void au_dpri_dalias(struct inode *inode)
6924 + spin_lock(&inode->i_lock);
6925 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
6926 + au_dpri_dentry(d);
6927 + spin_unlock(&inode->i_lock);
6930 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
6932 + struct dentry *wh = NULL;
6934 + struct au_iinfo *iinfo;
6936 + if (!dentry || IS_ERR(dentry)) {
6937 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
6940 + /* do not call dget_parent() here */
6941 + /* note: access d_xxx without d_lock */
6942 + dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
6943 + bindex, dentry, dentry,
6944 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
6945 + au_dcount(dentry), dentry->d_flags,
6946 + d_unhashed(dentry) ? "un" : "");
6949 + && d_is_positive(dentry)
6950 + && au_test_aufs(dentry->d_sb)) {
6951 + iinfo = au_ii(d_inode(dentry));
6953 + hn = !!au_hn(iinfo->ii_hinode + bindex);
6954 + wh = iinfo->ii_hinode[0 + bindex].hi_whdentry;
6957 + do_pri_inode(bindex, d_inode(dentry), hn, wh);
6961 +void au_dpri_dentry(struct dentry *dentry)
6963 + struct au_dinfo *dinfo;
6964 + aufs_bindex_t bindex;
6966 + struct au_hdentry *hdp;
6968 + err = do_pri_dentry(-1, dentry);
6969 + if (err || !au_test_aufs(dentry->d_sb))
6972 + dinfo = au_di(dentry);
6975 + dpri("d-1: bstart %d, bend %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
6976 + dinfo->di_bstart, dinfo->di_bend,
6977 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
6978 + dinfo->di_tmpfile);
6979 + if (dinfo->di_bstart < 0)
6981 + hdp = dinfo->di_hdentry;
6982 + for (bindex = dinfo->di_bstart; bindex <= dinfo->di_bend; bindex++)
6983 + do_pri_dentry(bindex, hdp[0 + bindex].hd_dentry);
6986 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
6990 + if (!file || IS_ERR(file)) {
6991 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
6996 + && !IS_ERR_OR_NULL(file->f_path.dentry)
6997 + && au_test_aufs(file->f_path.dentry->d_sb)
6999 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7000 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7001 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7002 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
7003 + file->f_version, file->f_pos, a);
7004 + if (!IS_ERR_OR_NULL(file->f_path.dentry))
7005 + do_pri_dentry(bindex, file->f_path.dentry);
7009 +void au_dpri_file(struct file *file)
7011 + struct au_finfo *finfo;
7012 + struct au_fidir *fidir;
7013 + struct au_hfile *hfile;
7014 + aufs_bindex_t bindex;
7017 + err = do_pri_file(-1, file);
7019 + || IS_ERR_OR_NULL(file->f_path.dentry)
7020 + || !au_test_aufs(file->f_path.dentry->d_sb))
7023 + finfo = au_fi(file);
7026 + if (finfo->fi_btop < 0)
7028 + fidir = finfo->fi_hdir;
7030 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7032 + for (bindex = finfo->fi_btop;
7033 + bindex >= 0 && bindex <= fidir->fd_bbot;
7035 + hfile = fidir->fd_hfile + bindex;
7036 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7040 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7042 + struct vfsmount *mnt;
7043 + struct super_block *sb;
7045 + if (!br || IS_ERR(br))
7047 + mnt = au_br_mnt(br);
7048 + if (!mnt || IS_ERR(mnt))
7051 + if (!sb || IS_ERR(sb))
7054 + dpri("s%d: {perm 0x%x, id %d, cnt %d, wbr %p}, "
7055 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7057 + bindex, br->br_perm, br->br_id, atomic_read(&br->br_count),
7058 + br->br_wbr, au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7059 + sb->s_flags, sb->s_count,
7060 + atomic_read(&sb->s_active), !!br->br_xino.xi_file);
7064 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7068 +void au_dpri_sb(struct super_block *sb)
7070 + struct au_sbinfo *sbinfo;
7071 + aufs_bindex_t bindex;
7073 + /* to reuduce stack size */
7075 + struct vfsmount mnt;
7076 + struct au_branch fake;
7079 + /* this function can be called from magic sysrq */
7080 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
7081 + if (unlikely(!a)) {
7082 + dpri("no memory\n");
7086 + a->mnt.mnt_sb = sb;
7087 + a->fake.br_path.mnt = &a->mnt;
7088 + atomic_set(&a->fake.br_count, 0);
7089 + smp_mb(); /* atomic_set */
7090 + err = do_pri_br(-1, &a->fake);
7092 + dpri("dev 0x%x\n", sb->s_dev);
7093 + if (err || !au_test_aufs(sb))
7096 + sbinfo = au_sbi(sb);
7099 + dpri("nw %d, gen %u, kobj %d\n",
7100 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7101 + atomic_read(&sbinfo->si_kobj.kref.refcount));
7102 + for (bindex = 0; bindex <= sbinfo->si_bend; bindex++)
7103 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7106 +/* ---------------------------------------------------------------------- */
7108 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7110 + struct inode *h_inode, *inode = d_inode(dentry);
7111 + struct dentry *h_dentry;
7112 + aufs_bindex_t bindex, bend, bi;
7114 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7117 + bend = au_dbend(dentry);
7118 + bi = au_ibend(inode);
7121 + bindex = au_dbstart(dentry);
7122 + bi = au_ibstart(inode);
7126 + for (; bindex <= bend; bindex++) {
7127 + h_dentry = au_h_dptr(dentry, bindex);
7130 + h_inode = au_h_iptr(inode, bindex);
7131 + if (unlikely(h_inode != d_inode(h_dentry))) {
7133 + AuDbg("b%d, %s:%d\n", bindex, func, line);
7134 + AuDbgDentry(dentry);
7135 + AuDbgInode(inode);
7142 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7145 + struct au_dcsub_pages dpages;
7146 + struct au_dpage *dpage;
7147 + struct dentry **dentries;
7149 + err = au_dpages_init(&dpages, GFP_NOFS);
7151 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7153 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7154 + dpage = dpages.dpages + i;
7155 + dentries = dpage->dentries;
7156 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7157 + AuDebugOn(au_digen_test(dentries[j], sigen));
7159 + au_dpages_free(&dpages);
7162 +void au_dbg_verify_kthread(void)
7164 + if (au_wkq_test()) {
7167 + * It may be recursive, but udba=notify between two aufs mounts,
7168 + * where a single ro branch is shared, is not a problem.
7174 +/* ---------------------------------------------------------------------- */
7176 +int __init au_debug_init(void)
7178 + aufs_bindex_t bindex;
7179 + struct au_vdir_destr destr;
7182 + AuDebugOn(bindex >= 0);
7185 + AuDebugOn(destr.len < NAME_MAX);
7187 +#ifdef CONFIG_4KSTACKS
7188 + pr_warn("CONFIG_4KSTACKS is defined.\n");
7193 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7194 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
7195 +++ linux/fs/aufs/debug.h 2016-01-13 20:11:11.666426853 +0100
7198 + * Copyright (C) 2005-2015 Junjiro R. Okajima
7200 + * This program, aufs is free software; you can redistribute it and/or modify
7201 + * it under the terms of the GNU General Public License as published by
7202 + * the Free Software Foundation; either version 2 of the License, or
7203 + * (at your option) any later version.
7205 + * This program is distributed in the hope that it will be useful,
7206 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7207 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7208 + * GNU General Public License for more details.
7210 + * You should have received a copy of the GNU General Public License
7211 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7215 + * debug print functions
7218 +#ifndef __AUFS_DEBUG_H__
7219 +#define __AUFS_DEBUG_H__
7223 +#include <linux/atomic.h>
7224 +#include <linux/module.h>
7225 +#include <linux/kallsyms.h>
7226 +#include <linux/sysrq.h>
7228 +#ifdef CONFIG_AUFS_DEBUG
7229 +#define AuDebugOn(a) BUG_ON(a)
7231 +/* module parameter */
7232 +extern atomic_t aufs_debug;
7233 +static inline void au_debug_on(void)
7235 + atomic_inc(&aufs_debug);
7237 +static inline void au_debug_off(void)
7239 + atomic_dec_if_positive(&aufs_debug);
7242 +static inline int au_debug_test(void)
7244 + return atomic_read(&aufs_debug) > 0;
7247 +#define AuDebugOn(a) do {} while (0)
7248 +AuStubVoid(au_debug_on, void)
7249 +AuStubVoid(au_debug_off, void)
7250 +AuStubInt0(au_debug_test, void)
7251 +#endif /* CONFIG_AUFS_DEBUG */
7253 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7255 +/* ---------------------------------------------------------------------- */
7259 +#define AuDbg(fmt, ...) do { \
7260 + if (au_debug_test()) \
7261 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7263 +#define AuLabel(l) AuDbg(#l "\n")
7264 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7265 +#define AuWarn1(fmt, ...) do { \
7266 + static unsigned char _c; \
7268 + pr_warn(fmt, ##__VA_ARGS__); \
7271 +#define AuErr1(fmt, ...) do { \
7272 + static unsigned char _c; \
7274 + pr_err(fmt, ##__VA_ARGS__); \
7277 +#define AuIOErr1(fmt, ...) do { \
7278 + static unsigned char _c; \
7280 + AuIOErr(fmt, ##__VA_ARGS__); \
7283 +#define AuUnsupportMsg "This operation is not supported." \
7284 + " Please report this application to aufs-users ML."
7285 +#define AuUnsupport(fmt, ...) do { \
7286 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7290 +#define AuTraceErr(e) do { \
7291 + if (unlikely((e) < 0)) \
7292 + AuDbg("err %d\n", (int)(e)); \
7295 +#define AuTraceErrPtr(p) do { \
7297 + AuDbg("err %ld\n", PTR_ERR(p)); \
7300 +/* dirty macros for debug print, use with "%.*s" and caution */
7301 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
7303 +/* ---------------------------------------------------------------------- */
7306 +#ifdef CONFIG_AUFS_DEBUG
7307 +extern struct mutex au_dbg_mtx;
7308 +extern char *au_plevel;
7310 +void au_dpri_whlist(struct au_nhash *whlist);
7312 +void au_dpri_vdir(struct au_vdir *vdir);
7314 +void au_dpri_inode(struct inode *inode);
7315 +void au_dpri_dalias(struct inode *inode);
7316 +void au_dpri_dentry(struct dentry *dentry);
7318 +void au_dpri_file(struct file *filp);
7319 +struct super_block;
7320 +void au_dpri_sb(struct super_block *sb);
7322 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7323 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7324 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7325 +void au_dbg_verify_kthread(void);
7327 +int __init au_debug_init(void);
7329 +#define AuDbgWhlist(w) do { \
7330 + mutex_lock(&au_dbg_mtx); \
7332 + au_dpri_whlist(w); \
7333 + mutex_unlock(&au_dbg_mtx); \
7336 +#define AuDbgVdir(v) do { \
7337 + mutex_lock(&au_dbg_mtx); \
7339 + au_dpri_vdir(v); \
7340 + mutex_unlock(&au_dbg_mtx); \
7343 +#define AuDbgInode(i) do { \
7344 + mutex_lock(&au_dbg_mtx); \
7346 + au_dpri_inode(i); \
7347 + mutex_unlock(&au_dbg_mtx); \
7350 +#define AuDbgDAlias(i) do { \
7351 + mutex_lock(&au_dbg_mtx); \
7353 + au_dpri_dalias(i); \
7354 + mutex_unlock(&au_dbg_mtx); \
7357 +#define AuDbgDentry(d) do { \
7358 + mutex_lock(&au_dbg_mtx); \
7360 + au_dpri_dentry(d); \
7361 + mutex_unlock(&au_dbg_mtx); \
7364 +#define AuDbgFile(f) do { \
7365 + mutex_lock(&au_dbg_mtx); \
7367 + au_dpri_file(f); \
7368 + mutex_unlock(&au_dbg_mtx); \
7371 +#define AuDbgSb(sb) do { \
7372 + mutex_lock(&au_dbg_mtx); \
7373 + AuDbg(#sb "\n"); \
7375 + mutex_unlock(&au_dbg_mtx); \
7378 +#define AuDbgSym(addr) do { \
7379 + char sym[KSYM_SYMBOL_LEN]; \
7380 + sprint_symbol(sym, (unsigned long)addr); \
7381 + AuDbg("%s\n", sym); \
7384 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7385 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7386 +AuStubVoid(au_dbg_verify_kthread, void)
7387 +AuStubInt0(__init au_debug_init, void)
7389 +#define AuDbgWhlist(w) do {} while (0)
7390 +#define AuDbgVdir(v) do {} while (0)
7391 +#define AuDbgInode(i) do {} while (0)
7392 +#define AuDbgDAlias(i) do {} while (0)
7393 +#define AuDbgDentry(d) do {} while (0)
7394 +#define AuDbgFile(f) do {} while (0)
7395 +#define AuDbgSb(sb) do {} while (0)
7396 +#define AuDbgSym(addr) do {} while (0)
7397 +#endif /* CONFIG_AUFS_DEBUG */
7399 +/* ---------------------------------------------------------------------- */
7401 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7402 +int __init au_sysrq_init(void);
7403 +void au_sysrq_fin(void);
7405 +#ifdef CONFIG_HW_CONSOLE
7406 +#define au_dbg_blocked() do { \
7408 + handle_sysrq('w'); \
7411 +AuStubVoid(au_dbg_blocked, void)
7415 +AuStubInt0(__init au_sysrq_init, void)
7416 +AuStubVoid(au_sysrq_fin, void)
7417 +AuStubVoid(au_dbg_blocked, void)
7418 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7420 +#endif /* __KERNEL__ */
7421 +#endif /* __AUFS_DEBUG_H__ */
7422 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7423 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
7424 +++ linux/fs/aufs/dentry.c 2016-01-13 20:11:11.666426853 +0100
7427 + * Copyright (C) 2005-2015 Junjiro R. Okajima
7429 + * This program, aufs is free software; you can redistribute it and/or modify
7430 + * it under the terms of the GNU General Public License as published by
7431 + * the Free Software Foundation; either version 2 of the License, or
7432 + * (at your option) any later version.
7434 + * This program is distributed in the hope that it will be useful,
7435 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7436 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7437 + * GNU General Public License for more details.
7439 + * You should have received a copy of the GNU General Public License
7440 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7444 + * lookup and dentry operations
7447 +#include <linux/namei.h>
7450 +#define AuLkup_ALLOW_NEG 1
7451 +#define AuLkup_IGNORE_PERM (1 << 1)
7452 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
7453 +#define au_fset_lkup(flags, name) \
7454 + do { (flags) |= AuLkup_##name; } while (0)
7455 +#define au_fclr_lkup(flags, name) \
7456 + do { (flags) &= ~AuLkup_##name; } while (0)
7458 +struct au_do_lookup_args {
7459 + unsigned int flags;
7464 + * returns positive/negative dentry, NULL or an error.
7465 + * NULL means whiteout-ed or not-found.
7467 +static struct dentry*
7468 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7469 + aufs_bindex_t bindex, struct qstr *wh_name,
7470 + struct au_do_lookup_args *args)
7472 + struct dentry *h_dentry;
7473 + struct inode *h_inode;
7474 + struct au_branch *br;
7475 + int wh_found, opq;
7476 + unsigned char wh_able;
7477 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
7478 + const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
7482 + br = au_sbr(dentry->d_sb, bindex);
7483 + wh_able = !!au_br_whable(br->br_perm);
7485 + wh_found = au_wh_test(h_parent, wh_name, /*try_sio*/0);
7486 + h_dentry = ERR_PTR(wh_found);
7489 + if (unlikely(wh_found < 0))
7492 + /* We found a whiteout */
7493 + /* au_set_dbend(dentry, bindex); */
7494 + au_set_dbwh(dentry, bindex);
7496 + return NULL; /* success */
7500 + h_dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
7502 + h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
7503 + if (IS_ERR(h_dentry)) {
7504 + if (PTR_ERR(h_dentry) == -ENAMETOOLONG
7510 + h_inode = d_inode(h_dentry);
7511 + if (d_is_negative(h_dentry)) {
7514 + } else if (wh_found
7515 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
7518 + if (au_dbend(dentry) <= bindex)
7519 + au_set_dbend(dentry, bindex);
7520 + if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
7521 + au_set_dbstart(dentry, bindex);
7522 + au_set_h_dptr(dentry, bindex, h_dentry);
7524 + if (!d_is_dir(h_dentry)
7526 + || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
7527 + goto out; /* success */
7529 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
7530 + opq = au_diropq_test(h_dentry);
7531 + mutex_unlock(&h_inode->i_mutex);
7533 + au_set_dbdiropq(dentry, bindex);
7534 + else if (unlikely(opq < 0)) {
7535 + au_set_h_dptr(dentry, bindex, NULL);
7536 + h_dentry = ERR_PTR(opq);
7547 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
7549 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
7550 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
7556 + * returns the number of lower positive dentries,
7557 + * otherwise an error.
7558 + * can be called at unlinking with @type is zero.
7560 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type)
7562 + int npositive, err;
7563 + aufs_bindex_t bindex, btail, bdiropq;
7564 + unsigned char isdir, dirperm1;
7565 + struct qstr whname;
7566 + struct au_do_lookup_args args = {
7570 + const struct qstr *name = &dentry->d_name;
7571 + struct dentry *parent;
7572 + struct super_block *sb;
7574 + sb = dentry->d_sb;
7575 + err = au_test_shwh(sb, name);
7576 + if (unlikely(err))
7579 + err = au_wh_name_alloc(&whname, name);
7580 + if (unlikely(err))
7583 + isdir = !!d_is_dir(dentry);
7585 + au_fset_lkup(args.flags, ALLOW_NEG);
7586 + dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
7589 + parent = dget_parent(dentry);
7590 + btail = au_dbtaildir(parent);
7591 + for (bindex = bstart; bindex <= btail; bindex++) {
7592 + struct dentry *h_parent, *h_dentry;
7593 + struct inode *h_inode, *h_dir;
7595 + h_dentry = au_h_dptr(dentry, bindex);
7597 + if (d_is_positive(h_dentry))
7599 + if (type != S_IFDIR)
7603 + h_parent = au_h_dptr(parent, bindex);
7604 + if (!h_parent || !d_is_dir(h_parent))
7607 + h_dir = d_inode(h_parent);
7608 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
7609 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &whname,
7611 + mutex_unlock(&h_dir->i_mutex);
7612 + err = PTR_ERR(h_dentry);
7613 + if (IS_ERR(h_dentry))
7616 + au_fclr_lkup(args.flags, ALLOW_NEG);
7618 + au_fset_lkup(args.flags, IGNORE_PERM);
7620 + if (au_dbwh(dentry) == bindex)
7624 + if (d_is_negative(h_dentry))
7626 + h_inode = d_inode(h_dentry);
7629 + args.type = h_inode->i_mode & S_IFMT;
7630 + if (args.type != S_IFDIR)
7633 + /* the type of lower may be different */
7634 + bdiropq = au_dbdiropq(dentry);
7635 + if (bdiropq >= 0 && bdiropq <= bindex)
7641 + AuLabel(positive);
7642 + au_update_dbstart(dentry);
7645 + if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
7646 + && au_dbstart(dentry) < 0)) {
7648 + AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
7654 + kfree(whname.name);
7659 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent)
7661 + struct dentry *dentry;
7664 + if (!au_test_h_perm_sio(d_inode(parent), MAY_EXEC))
7665 + dentry = vfsub_lkup_one(name, parent);
7667 + struct vfsub_lkup_one_args args = {
7673 + wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
7674 + if (unlikely(wkq_err))
7675 + dentry = ERR_PTR(wkq_err);
7682 + * lookup @dentry on @bindex which should be negative.
7684 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
7687 + struct dentry *parent, *h_parent, *h_dentry;
7688 + struct au_branch *br;
7690 + parent = dget_parent(dentry);
7691 + h_parent = au_h_dptr(parent, bindex);
7692 + br = au_sbr(dentry->d_sb, bindex);
7694 + h_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
7696 + h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
7697 + err = PTR_ERR(h_dentry);
7698 + if (IS_ERR(h_dentry))
7700 + if (unlikely(d_is_positive(h_dentry))) {
7702 + AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
7708 + if (bindex < au_dbstart(dentry))
7709 + au_set_dbstart(dentry, bindex);
7710 + if (au_dbend(dentry) < bindex)
7711 + au_set_dbend(dentry, bindex);
7712 + au_set_h_dptr(dentry, bindex, h_dentry);
7719 +/* ---------------------------------------------------------------------- */
7721 +/* subset of struct inode */
7723 + unsigned long i_ino;
7724 + /* unsigned int i_nlink; */
7730 + blkcnt_t i_blocks;
7735 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
7737 + ia->i_ino = h_inode->i_ino;
7738 + /* ia->i_nlink = h_inode->i_nlink; */
7739 + ia->i_uid = h_inode->i_uid;
7740 + ia->i_gid = h_inode->i_gid;
7741 + ia->i_version = h_inode->i_version;
7743 + ia->i_size = h_inode->i_size;
7744 + ia->i_blocks = h_inode->i_blocks;
7746 + ia->i_mode = (h_inode->i_mode & S_IFMT);
7749 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
7751 + return ia->i_ino != h_inode->i_ino
7752 + /* || ia->i_nlink != h_inode->i_nlink */
7753 + || !uid_eq(ia->i_uid, h_inode->i_uid)
7754 + || !gid_eq(ia->i_gid, h_inode->i_gid)
7755 + || ia->i_version != h_inode->i_version
7757 + || ia->i_size != h_inode->i_size
7758 + || ia->i_blocks != h_inode->i_blocks
7760 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
7763 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
7764 + struct au_branch *br)
7767 + struct au_iattr ia;
7768 + struct inode *h_inode;
7769 + struct dentry *h_d;
7770 + struct super_block *h_sb;
7773 + memset(&ia, -1, sizeof(ia));
7774 + h_sb = h_dentry->d_sb;
7776 + if (d_is_positive(h_dentry)) {
7777 + h_inode = d_inode(h_dentry);
7778 + au_iattr_save(&ia, h_inode);
7779 + } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
7780 + /* nfs d_revalidate may return 0 for negative dentry */
7781 + /* fuse d_revalidate always return 0 for negative dentry */
7784 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
7785 + h_d = vfsub_lkup_one(&h_dentry->d_name, h_parent);
7786 + err = PTR_ERR(h_d);
7791 + if (unlikely(h_d != h_dentry
7792 + || d_inode(h_d) != h_inode
7793 + || (h_inode && au_iattr_test(&ia, h_inode))))
7794 + err = au_busy_or_stale();
7802 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
7803 + struct dentry *h_parent, struct au_branch *br)
7808 + if (udba == AuOpt_UDBA_REVAL
7809 + && !au_test_fs_remote(h_dentry->d_sb)) {
7811 + err = (d_inode(h_dentry->d_parent) != h_dir);
7812 + } else if (udba != AuOpt_UDBA_NONE)
7813 + err = au_h_verify_dentry(h_dentry, h_parent, br);
7818 +/* ---------------------------------------------------------------------- */
7820 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
7823 + aufs_bindex_t new_bindex, bindex, bend, bwh, bdiropq;
7824 + struct au_hdentry tmp, *p, *q;
7825 + struct au_dinfo *dinfo;
7826 + struct super_block *sb;
7828 + DiMustWriteLock(dentry);
7830 + sb = dentry->d_sb;
7831 + dinfo = au_di(dentry);
7832 + bend = dinfo->di_bend;
7833 + bwh = dinfo->di_bwh;
7834 + bdiropq = dinfo->di_bdiropq;
7835 + p = dinfo->di_hdentry + dinfo->di_bstart;
7836 + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++, p++) {
7837 + if (!p->hd_dentry)
7840 + new_bindex = au_br_index(sb, p->hd_id);
7841 + if (new_bindex == bindex)
7844 + if (dinfo->di_bwh == bindex)
7846 + if (dinfo->di_bdiropq == bindex)
7847 + bdiropq = new_bindex;
7848 + if (new_bindex < 0) {
7850 + p->hd_dentry = NULL;
7854 + /* swap two lower dentries, and loop again */
7855 + q = dinfo->di_hdentry + new_bindex;
7859 + if (tmp.hd_dentry) {
7865 + dinfo->di_bwh = -1;
7866 + if (bwh >= 0 && bwh <= au_sbend(sb) && au_sbr_whable(sb, bwh))
7867 + dinfo->di_bwh = bwh;
7869 + dinfo->di_bdiropq = -1;
7871 + && bdiropq <= au_sbend(sb)
7872 + && au_sbr_whable(sb, bdiropq))
7873 + dinfo->di_bdiropq = bdiropq;
7876 + dinfo->di_bstart = -1;
7877 + dinfo->di_bend = -1;
7878 + bend = au_dbend(parent);
7879 + p = dinfo->di_hdentry;
7880 + for (bindex = 0; bindex <= bend; bindex++, p++)
7881 + if (p->hd_dentry) {
7882 + dinfo->di_bstart = bindex;
7886 + if (dinfo->di_bstart >= 0) {
7887 + p = dinfo->di_hdentry + bend;
7888 + for (bindex = bend; bindex >= 0; bindex--, p--)
7889 + if (p->hd_dentry) {
7890 + dinfo->di_bend = bindex;
7899 +static void au_do_hide(struct dentry *dentry)
7901 + struct inode *inode;
7903 + if (d_really_is_positive(dentry)) {
7904 + inode = d_inode(dentry);
7905 + if (!d_is_dir(dentry)) {
7906 + if (inode->i_nlink && !d_unhashed(dentry))
7907 + drop_nlink(inode);
7909 + clear_nlink(inode);
7910 + /* stop next lookup */
7911 + inode->i_flags |= S_DEAD;
7913 + smp_mb(); /* necessary? */
7918 +static int au_hide_children(struct dentry *parent)
7920 + int err, i, j, ndentry;
7921 + struct au_dcsub_pages dpages;
7922 + struct au_dpage *dpage;
7923 + struct dentry *dentry;
7925 + err = au_dpages_init(&dpages, GFP_NOFS);
7926 + if (unlikely(err))
7928 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
7929 + if (unlikely(err))
7932 + /* in reverse order */
7933 + for (i = dpages.ndpage - 1; i >= 0; i--) {
7934 + dpage = dpages.dpages + i;
7935 + ndentry = dpage->ndentry;
7936 + for (j = ndentry - 1; j >= 0; j--) {
7937 + dentry = dpage->dentries[j];
7938 + if (dentry != parent)
7939 + au_do_hide(dentry);
7944 + au_dpages_free(&dpages);
7949 +static void au_hide(struct dentry *dentry)
7953 + AuDbgDentry(dentry);
7954 + if (d_is_dir(dentry)) {
7955 + /* shrink_dcache_parent(dentry); */
7956 + err = au_hide_children(dentry);
7957 + if (unlikely(err))
7958 + AuIOErr("%pd, failed hiding children, ignored %d\n",
7961 + au_do_hide(dentry);
7965 + * By adding a dirty branch, a cached dentry may be affected in various ways.
7967 + * a dirty branch is added
7968 + * - on the top of layers
7969 + * - in the middle of layers
7970 + * - to the bottom of layers
7972 + * on the added branch there exists
7975 + * - a same named entry
7977 + * * negative --> positive
7978 + * * positive --> positive
7979 + * - type is unchanged
7980 + * - type is changed
7982 + * * negative --> negative
7983 + * * positive --> negative (rejected by au_br_del() for non-dir case)
7986 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
7987 + struct au_dinfo *tmp)
7990 + aufs_bindex_t bindex, bend;
7992 + struct dentry *dentry;
7993 + struct inode *inode;
7995 + } orig_h, tmp_h = {
7998 + struct au_hdentry *hd;
7999 + struct inode *inode, *h_inode;
8000 + struct dentry *h_dentry;
8003 + AuDebugOn(dinfo->di_bstart < 0);
8005 + orig_h.dentry = dinfo->di_hdentry[dinfo->di_bstart].hd_dentry;
8006 + orig_h.inode = NULL;
8007 + if (d_is_positive(orig_h.dentry)) {
8008 + orig_h.inode = d_inode(orig_h.dentry);
8009 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8011 + if (tmp->di_bstart >= 0) {
8012 + tmp_h.dentry = tmp->di_hdentry[tmp->di_bstart].hd_dentry;
8013 + if (d_is_positive(tmp_h.dentry)) {
8014 + tmp_h.inode = d_inode(tmp_h.dentry);
8015 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8020 + if (d_really_is_positive(dentry))
8021 + inode = d_inode(dentry);
8022 + if (!orig_h.inode) {
8023 + AuDbg("nagative originally\n");
8029 + AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
8030 + AuDebugOn(dinfo->di_bdiropq != -1);
8032 + if (!tmp_h.inode) {
8033 + AuDbg("negative --> negative\n");
8034 + /* should have only one negative lower */
8035 + if (tmp->di_bstart >= 0
8036 + && tmp->di_bstart < dinfo->di_bstart) {
8037 + AuDebugOn(tmp->di_bstart != tmp->di_bend);
8038 + AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
8039 + au_set_h_dptr(dentry, dinfo->di_bstart, NULL);
8040 + au_di_cp(dinfo, tmp);
8041 + hd = tmp->di_hdentry + tmp->di_bstart;
8042 + au_set_h_dptr(dentry, tmp->di_bstart,
8043 + dget(hd->hd_dentry));
8045 + au_dbg_verify_dinode(dentry);
8047 + AuDbg("negative --> positive\n");
8049 + * similar to the behaviour of creating with bypassing
8051 + * unhash it in order to force an error in the
8052 + * succeeding create operation.
8053 + * we should not set S_DEAD here.
8056 + /* au_di_swap(tmp, dinfo); */
8057 + au_dbg_verify_dinode(dentry);
8060 + AuDbg("positive originally\n");
8061 + /* inode may be NULL */
8062 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8063 + if (!tmp_h.inode) {
8064 + AuDbg("positive --> negative\n");
8065 + /* or bypassing aufs */
8067 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_bstart)
8068 + dinfo->di_bwh = tmp->di_bwh;
8070 + err = au_refresh_hinode_self(inode);
8071 + au_dbg_verify_dinode(dentry);
8072 + } else if (orig_h.mode == tmp_h.mode) {
8073 + AuDbg("positive --> positive, same type\n");
8074 + if (!S_ISDIR(orig_h.mode)
8075 + && dinfo->di_bstart > tmp->di_bstart) {
8077 + * similar to the behaviour of removing and
8082 + err = au_refresh_hinode_self(inode);
8083 + au_dbg_verify_dinode(dentry);
8085 + /* fill empty slots */
8086 + if (dinfo->di_bstart > tmp->di_bstart)
8087 + dinfo->di_bstart = tmp->di_bstart;
8088 + if (dinfo->di_bend < tmp->di_bend)
8089 + dinfo->di_bend = tmp->di_bend;
8090 + dinfo->di_bwh = tmp->di_bwh;
8091 + dinfo->di_bdiropq = tmp->di_bdiropq;
8092 + hd = tmp->di_hdentry;
8093 + bend = dinfo->di_bend;
8094 + for (bindex = tmp->di_bstart; bindex <= bend;
8096 + if (au_h_dptr(dentry, bindex))
8098 + h_dentry = hd[bindex].hd_dentry;
8101 + AuDebugOn(d_is_negative(h_dentry));
8102 + h_inode = d_inode(h_dentry);
8103 + AuDebugOn(orig_h.mode
8104 + != (h_inode->i_mode
8106 + au_set_h_dptr(dentry, bindex,
8109 + err = au_refresh_hinode(inode, dentry);
8110 + au_dbg_verify_dinode(dentry);
8113 + AuDbg("positive --> positive, different type\n");
8114 + /* similar to the behaviour of removing and creating */
8117 + err = au_refresh_hinode_self(inode);
8118 + au_dbg_verify_dinode(dentry);
8126 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8128 + const struct dentry_operations *dop
8129 + = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8130 + static const unsigned int mask
8131 + = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8133 + BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8135 + if (dentry->d_op == dop)
8138 + AuDbg("%pd\n", dentry);
8139 + spin_lock(&dentry->d_lock);
8140 + if (dop == &aufs_dop)
8141 + dentry->d_flags |= mask;
8143 + dentry->d_flags &= ~mask;
8144 + dentry->d_op = dop;
8145 + spin_unlock(&dentry->d_lock);
8148 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8151 + unsigned int sigen;
8152 + struct au_dinfo *dinfo, *tmp;
8153 + struct super_block *sb;
8154 + struct inode *inode;
8156 + DiMustWriteLock(dentry);
8157 + AuDebugOn(IS_ROOT(dentry));
8158 + AuDebugOn(d_really_is_negative(parent));
8160 + sb = dentry->d_sb;
8161 + sigen = au_sigen(sb);
8162 + err = au_digen_test(parent, sigen);
8163 + if (unlikely(err))
8166 + dinfo = au_di(dentry);
8167 + err = au_di_realloc(dinfo, au_sbend(sb) + 1);
8168 + if (unlikely(err))
8170 + ebrange = au_dbrange_test(dentry);
8172 + ebrange = au_do_refresh_hdentry(dentry, parent);
8174 + if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8175 + AuDebugOn(au_dbstart(dentry) < 0 && au_dbend(dentry) >= 0);
8176 + if (d_really_is_positive(dentry)) {
8177 + inode = d_inode(dentry);
8178 + err = au_refresh_hinode_self(inode);
8180 + au_dbg_verify_dinode(dentry);
8182 + goto out_dgen; /* success */
8186 + /* temporary dinfo */
8187 + AuDbgDentry(dentry);
8189 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8190 + if (unlikely(!tmp))
8192 + au_di_swap(tmp, dinfo);
8193 + /* returns the number of positive dentries */
8195 + * if current working dir is removed, it returns an error.
8196 + * but the dentry is legal.
8198 + err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
8199 + AuDbgDentry(dentry);
8200 + au_di_swap(tmp, dinfo);
8201 + if (err == -ENOENT)
8204 + /* compare/refresh by dinfo */
8205 + AuDbgDentry(dentry);
8206 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8207 + au_dbg_verify_dinode(dentry);
8210 + au_rw_write_unlock(&tmp->di_rwsem);
8212 + if (unlikely(err))
8216 + au_update_digen(dentry);
8218 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8219 + AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8220 + AuDbgDentry(dentry);
8226 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8227 + struct dentry *dentry, aufs_bindex_t bindex)
8232 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8235 + AuDbg("b%d\n", bindex);
8237 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8238 + * due to whiteout and branch permission.
8240 + flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8241 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
8242 + /* it may return tri-state */
8243 + valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8245 + if (unlikely(valid < 0))
8255 +/* todo: remove this */
8256 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8257 + unsigned int flags, int do_udba)
8260 + umode_t mode, h_mode;
8261 + aufs_bindex_t bindex, btail, bstart, ibs, ibe;
8262 + unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8263 + struct inode *h_inode, *h_cached_inode;
8264 + struct dentry *h_dentry;
8265 + struct qstr *name, *h_name;
8272 + unhashed = !!d_unhashed(dentry);
8273 + is_root = !!IS_ROOT(dentry);
8274 + name = &dentry->d_name;
8275 + tmpfile = au_di(dentry)->di_tmpfile;
8278 + * Theoretically, REVAL test should be unnecessary in case of
8280 + * But {fs,i}notify doesn't fire some necessary events,
8281 + * IN_ATTRIB for atime/nlink/pageio
8282 + * Let's do REVAL test too.
8284 + if (do_udba && inode) {
8285 + mode = (inode->i_mode & S_IFMT);
8286 + plus = (inode->i_nlink > 0);
8287 + ibs = au_ibstart(inode);
8288 + ibe = au_ibend(inode);
8291 + bstart = au_dbstart(dentry);
8293 + if (inode && S_ISDIR(inode->i_mode))
8294 + btail = au_dbtaildir(dentry);
8295 + for (bindex = bstart; bindex <= btail; bindex++) {
8296 + h_dentry = au_h_dptr(dentry, bindex);
8300 + AuDbg("b%d, %pd\n", bindex, h_dentry);
8301 + h_nfs = !!au_test_nfs(h_dentry->d_sb);
8302 + spin_lock(&h_dentry->d_lock);
8303 + h_name = &h_dentry->d_name;
8304 + if (unlikely(do_udba
8307 + && (unhashed != !!d_unhashed(h_dentry)
8309 + && !au_qstreq(name, h_name))
8312 + && !(flags & LOOKUP_OPEN)
8313 + && (h_dentry->d_flags
8314 + & DCACHE_NFSFS_RENAMED)))
8318 + h_unhashed = d_unhashed(h_dentry);
8319 + spin_unlock(&h_dentry->d_lock);
8320 + AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8321 + unhashed, h_unhashed, dentry, h_dentry);
8324 + spin_unlock(&h_dentry->d_lock);
8326 + err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8327 + if (unlikely(err))
8328 + /* do not goto err, to keep the errno */
8331 + /* todo: plink too? */
8336 + if (unlikely(!!inode != d_is_positive(h_dentry)))
8340 + if (d_is_positive(h_dentry))
8341 + h_inode = d_inode(h_dentry);
8344 + h_cached_inode = h_inode;
8346 + h_mode = (h_inode->i_mode & S_IFMT);
8347 + h_plus = (h_inode->i_nlink > 0);
8349 + if (inode && ibs <= bindex && bindex <= ibe)
8350 + h_cached_inode = au_h_iptr(inode, bindex);
8353 + if (unlikely(plus != h_plus && !tmpfile))
8356 + if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8358 + && !IS_ROOT(h_dentry)
8359 + && unhashed != d_unhashed(h_dentry)))
8362 + if (unlikely(mode != h_mode
8363 + || h_cached_inode != h_inode))
8376 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8377 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8380 + struct dentry *parent;
8382 + if (!au_digen_test(dentry, sigen))
8385 + parent = dget_parent(dentry);
8386 + di_read_lock_parent(parent, AuLock_IR);
8387 + AuDebugOn(au_digen_test(parent, sigen));
8388 + au_dbg_verify_gen(parent, sigen);
8389 + err = au_refresh_dentry(dentry, parent);
8390 + di_read_unlock(parent, AuLock_IR);
8396 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8399 + struct dentry *d, *parent;
8401 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8402 + return simple_reval_dpath(dentry, sigen);
8404 + /* slow loop, keep it simple and stupid */
8405 + /* cf: au_cpup_dirs() */
8408 + while (au_digen_test(dentry, sigen)) {
8412 + parent = dget_parent(d);
8413 + if (!au_digen_test(parent, sigen))
8419 + di_write_lock_child2(d);
8421 + /* someone might update our dentry while we were sleeping */
8422 + if (au_digen_test(d, sigen)) {
8424 + * todo: consolidate with simple_reval_dpath(),
8425 + * do_refresh() and au_reval_for_attr().
8427 + di_read_lock_parent(parent, AuLock_IR);
8428 + err = au_refresh_dentry(d, parent);
8429 + di_read_unlock(parent, AuLock_IR);
8433 + di_write_unlock(d);
8435 + if (unlikely(err))
8443 + * if valid returns 1, otherwise 0.
8445 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
8448 + unsigned int sigen;
8449 + unsigned char do_udba;
8450 + struct super_block *sb;
8451 + struct inode *inode;
8453 + /* todo: support rcu-walk? */
8454 + if (flags & LOOKUP_RCU)
8458 + if (unlikely(!au_di(dentry)))
8462 + sb = dentry->d_sb;
8465 + * i_mutex of parent dir may be held,
8466 + * but we should not return 'invalid' due to busy.
8468 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
8469 + if (unlikely(err)) {
8475 + if (d_really_is_positive(dentry))
8476 + inode = d_inode(dentry);
8477 + if (unlikely(inode && is_bad_inode(inode))) {
8482 + if (unlikely(au_dbrange_test(dentry))) {
8488 + sigen = au_sigen(sb);
8489 + if (au_digen_test(dentry, sigen)) {
8490 + AuDebugOn(IS_ROOT(dentry));
8491 + err = au_reval_dpath(dentry, sigen);
8492 + if (unlikely(err)) {
8497 + di_downgrade_lock(dentry, AuLock_IR);
8500 + if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
8502 + && !(inode->i_state && I_LINKABLE)
8503 + && (IS_DEADDIR(inode) || !inode->i_nlink)) {
8508 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
8509 + if (do_udba && inode) {
8510 + aufs_bindex_t bstart = au_ibstart(inode);
8511 + struct inode *h_inode;
8513 + if (bstart >= 0) {
8514 + h_inode = au_h_iptr(inode, bstart);
8515 + if (h_inode && au_test_higen(inode, h_inode)) {
8522 + err = h_d_revalidate(dentry, inode, flags, do_udba);
8523 + if (unlikely(!err && do_udba && au_dbstart(dentry) < 0)) {
8525 + AuDbg("both of real entry and whiteout found, %p, err %d\n",
8531 + di_downgrade_lock(dentry, AuLock_IR);
8533 + aufs_read_unlock(dentry, AuLock_IR);
8538 + AuDbg("%pd invalid, %d\n", dentry, valid);
8544 +static void aufs_d_release(struct dentry *dentry)
8546 + if (au_di(dentry)) {
8547 + au_di_fin(dentry);
8548 + au_hn_di_reinit(dentry);
8552 +const struct dentry_operations aufs_dop = {
8553 + .d_revalidate = aufs_d_revalidate,
8554 + .d_weak_revalidate = aufs_d_revalidate,
8555 + .d_release = aufs_d_release
8558 +/* aufs_dop without d_revalidate */
8559 +const struct dentry_operations aufs_dop_noreval = {
8560 + .d_release = aufs_d_release
8562 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
8563 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
8564 +++ linux/fs/aufs/dentry.h 2016-01-13 20:11:11.669760262 +0100
8567 + * Copyright (C) 2005-2015 Junjiro R. Okajima
8569 + * This program, aufs is free software; you can redistribute it and/or modify
8570 + * it under the terms of the GNU General Public License as published by
8571 + * the Free Software Foundation; either version 2 of the License, or
8572 + * (at your option) any later version.
8574 + * This program is distributed in the hope that it will be useful,
8575 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8576 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8577 + * GNU General Public License for more details.
8579 + * You should have received a copy of the GNU General Public License
8580 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
8584 + * lookup and dentry operations
8587 +#ifndef __AUFS_DENTRY_H__
8588 +#define __AUFS_DENTRY_H__
8592 +#include <linux/dcache.h>
8595 +struct au_hdentry {
8596 + struct dentry *hd_dentry;
8597 + aufs_bindex_t hd_id;
8601 + atomic_t di_generation;
8603 + struct au_rwsem di_rwsem;
8604 + aufs_bindex_t di_bstart, di_bend, di_bwh, di_bdiropq;
8605 + unsigned char di_tmpfile; /* to allow the different name */
8606 + struct au_hdentry *di_hdentry;
8607 +} ____cacheline_aligned_in_smp;
8609 +/* ---------------------------------------------------------------------- */
8612 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
8614 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent);
8615 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8616 + struct dentry *h_parent, struct au_branch *br);
8618 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type);
8619 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
8620 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
8621 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
8622 +void au_refresh_dop(struct dentry *dentry, int force_reval);
8625 +void au_di_init_once(void *_di);
8626 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
8627 +void au_di_free(struct au_dinfo *dinfo);
8628 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
8629 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
8630 +int au_di_init(struct dentry *dentry);
8631 +void au_di_fin(struct dentry *dentry);
8632 +int au_di_realloc(struct au_dinfo *dinfo, int nbr);
8634 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
8635 +void di_read_unlock(struct dentry *d, int flags);
8636 +void di_downgrade_lock(struct dentry *d, int flags);
8637 +void di_write_lock(struct dentry *d, unsigned int lsc);
8638 +void di_write_unlock(struct dentry *d);
8639 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
8640 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
8641 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
8643 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
8644 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
8645 +aufs_bindex_t au_dbtail(struct dentry *dentry);
8646 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
8648 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
8649 + struct dentry *h_dentry);
8650 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
8651 +int au_dbrange_test(struct dentry *dentry);
8652 +void au_update_digen(struct dentry *dentry);
8653 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
8654 +void au_update_dbstart(struct dentry *dentry);
8655 +void au_update_dbend(struct dentry *dentry);
8656 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
8658 +/* ---------------------------------------------------------------------- */
8660 +static inline struct au_dinfo *au_di(struct dentry *dentry)
8662 + return dentry->d_fsdata;
8665 +/* ---------------------------------------------------------------------- */
8667 +/* lock subclass for dinfo */
8669 + AuLsc_DI_CHILD, /* child first */
8670 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
8671 + AuLsc_DI_CHILD3, /* copyup dirs */
8675 + AuLsc_DI_TMP /* temp for replacing dinfo */
8679 + * di_read_lock_child, di_write_lock_child,
8680 + * di_read_lock_child2, di_write_lock_child2,
8681 + * di_read_lock_child3, di_write_lock_child3,
8682 + * di_read_lock_parent, di_write_lock_parent,
8683 + * di_read_lock_parent2, di_write_lock_parent2,
8684 + * di_read_lock_parent3, di_write_lock_parent3,
8686 +#define AuReadLockFunc(name, lsc) \
8687 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
8688 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
8690 +#define AuWriteLockFunc(name, lsc) \
8691 +static inline void di_write_lock_##name(struct dentry *d) \
8692 +{ di_write_lock(d, AuLsc_DI_##lsc); }
8694 +#define AuRWLockFuncs(name, lsc) \
8695 + AuReadLockFunc(name, lsc) \
8696 + AuWriteLockFunc(name, lsc)
8698 +AuRWLockFuncs(child, CHILD);
8699 +AuRWLockFuncs(child2, CHILD2);
8700 +AuRWLockFuncs(child3, CHILD3);
8701 +AuRWLockFuncs(parent, PARENT);
8702 +AuRWLockFuncs(parent2, PARENT2);
8703 +AuRWLockFuncs(parent3, PARENT3);
8705 +#undef AuReadLockFunc
8706 +#undef AuWriteLockFunc
8707 +#undef AuRWLockFuncs
8709 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
8710 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
8711 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
8713 +/* ---------------------------------------------------------------------- */
8715 +/* todo: memory barrier? */
8716 +static inline unsigned int au_digen(struct dentry *d)
8718 + return atomic_read(&au_di(d)->di_generation);
8721 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
8723 + hdentry->hd_dentry = NULL;
8726 +static inline void au_hdput(struct au_hdentry *hd)
8729 + dput(hd->hd_dentry);
8732 +static inline aufs_bindex_t au_dbstart(struct dentry *dentry)
8734 + DiMustAnyLock(dentry);
8735 + return au_di(dentry)->di_bstart;
8738 +static inline aufs_bindex_t au_dbend(struct dentry *dentry)
8740 + DiMustAnyLock(dentry);
8741 + return au_di(dentry)->di_bend;
8744 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
8746 + DiMustAnyLock(dentry);
8747 + return au_di(dentry)->di_bwh;
8750 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
8752 + DiMustAnyLock(dentry);
8753 + return au_di(dentry)->di_bdiropq;
8756 +/* todo: hard/soft set? */
8757 +static inline void au_set_dbstart(struct dentry *dentry, aufs_bindex_t bindex)
8759 + DiMustWriteLock(dentry);
8760 + au_di(dentry)->di_bstart = bindex;
8763 +static inline void au_set_dbend(struct dentry *dentry, aufs_bindex_t bindex)
8765 + DiMustWriteLock(dentry);
8766 + au_di(dentry)->di_bend = bindex;
8769 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
8771 + DiMustWriteLock(dentry);
8772 + /* dbwh can be outside of bstart - bend range */
8773 + au_di(dentry)->di_bwh = bindex;
8776 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
8778 + DiMustWriteLock(dentry);
8779 + au_di(dentry)->di_bdiropq = bindex;
8782 +/* ---------------------------------------------------------------------- */
8784 +#ifdef CONFIG_AUFS_HNOTIFY
8785 +static inline void au_digen_dec(struct dentry *d)
8787 + atomic_dec(&au_di(d)->di_generation);
8790 +static inline void au_hn_di_reinit(struct dentry *dentry)
8792 + dentry->d_fsdata = NULL;
8795 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
8796 +#endif /* CONFIG_AUFS_HNOTIFY */
8798 +#endif /* __KERNEL__ */
8799 +#endif /* __AUFS_DENTRY_H__ */
8800 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
8801 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
8802 +++ linux/fs/aufs/dinfo.c 2016-01-13 20:11:11.669760262 +0100
8805 + * Copyright (C) 2005-2015 Junjiro R. Okajima
8807 + * This program, aufs is free software; you can redistribute it and/or modify
8808 + * it under the terms of the GNU General Public License as published by
8809 + * the Free Software Foundation; either version 2 of the License, or
8810 + * (at your option) any later version.
8812 + * This program is distributed in the hope that it will be useful,
8813 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8814 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8815 + * GNU General Public License for more details.
8817 + * You should have received a copy of the GNU General Public License
8818 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
8822 + * dentry private data
8827 +void au_di_init_once(void *_dinfo)
8829 + struct au_dinfo *dinfo = _dinfo;
8830 + static struct lock_class_key aufs_di;
8832 + au_rw_init(&dinfo->di_rwsem);
8833 + au_rw_class(&dinfo->di_rwsem, &aufs_di);
8836 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
8838 + struct au_dinfo *dinfo;
8841 + dinfo = au_cache_alloc_dinfo();
8842 + if (unlikely(!dinfo))
8845 + nbr = au_sbend(sb) + 1;
8848 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
8849 + if (dinfo->di_hdentry) {
8850 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
8851 + dinfo->di_bstart = -1;
8852 + dinfo->di_bend = -1;
8853 + dinfo->di_bwh = -1;
8854 + dinfo->di_bdiropq = -1;
8855 + dinfo->di_tmpfile = 0;
8856 + for (i = 0; i < nbr; i++)
8857 + dinfo->di_hdentry[i].hd_id = -1;
8861 + au_cache_free_dinfo(dinfo);
8868 +void au_di_free(struct au_dinfo *dinfo)
8870 + struct au_hdentry *p;
8871 + aufs_bindex_t bend, bindex;
8873 + /* dentry may not be revalidated */
8874 + bindex = dinfo->di_bstart;
8875 + if (bindex >= 0) {
8876 + bend = dinfo->di_bend;
8877 + p = dinfo->di_hdentry + bindex;
8878 + while (bindex++ <= bend)
8881 + kfree(dinfo->di_hdentry);
8882 + au_cache_free_dinfo(dinfo);
8885 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
8887 + struct au_hdentry *p;
8890 + AuRwMustWriteLock(&a->di_rwsem);
8891 + AuRwMustWriteLock(&b->di_rwsem);
8893 +#define DiSwap(v, name) \
8895 + v = a->di_##name; \
8896 + a->di_##name = b->di_##name; \
8897 + b->di_##name = v; \
8900 + DiSwap(p, hdentry);
8901 + DiSwap(bi, bstart);
8904 + DiSwap(bi, bdiropq);
8910 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
8912 + AuRwMustWriteLock(&dst->di_rwsem);
8913 + AuRwMustWriteLock(&src->di_rwsem);
8915 + dst->di_bstart = src->di_bstart;
8916 + dst->di_bend = src->di_bend;
8917 + dst->di_bwh = src->di_bwh;
8918 + dst->di_bdiropq = src->di_bdiropq;
8922 +int au_di_init(struct dentry *dentry)
8925 + struct super_block *sb;
8926 + struct au_dinfo *dinfo;
8929 + sb = dentry->d_sb;
8930 + dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
8932 + atomic_set(&dinfo->di_generation, au_sigen(sb));
8933 + /* smp_mb(); */ /* atomic_set */
8934 + dentry->d_fsdata = dinfo;
8941 +void au_di_fin(struct dentry *dentry)
8943 + struct au_dinfo *dinfo;
8945 + dinfo = au_di(dentry);
8946 + AuRwDestroy(&dinfo->di_rwsem);
8947 + au_di_free(dinfo);
8950 +int au_di_realloc(struct au_dinfo *dinfo, int nbr)
8953 + struct au_hdentry *hdp;
8955 + AuRwMustWriteLock(&dinfo->di_rwsem);
8958 + sz = sizeof(*hdp) * (dinfo->di_bend + 1);
8960 + sz = sizeof(*hdp);
8961 + hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS);
8963 + dinfo->di_hdentry = hdp;
8970 +/* ---------------------------------------------------------------------- */
8972 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
8975 + case AuLsc_DI_CHILD:
8976 + ii_write_lock_child(inode);
8978 + case AuLsc_DI_CHILD2:
8979 + ii_write_lock_child2(inode);
8981 + case AuLsc_DI_CHILD3:
8982 + ii_write_lock_child3(inode);
8984 + case AuLsc_DI_PARENT:
8985 + ii_write_lock_parent(inode);
8987 + case AuLsc_DI_PARENT2:
8988 + ii_write_lock_parent2(inode);
8990 + case AuLsc_DI_PARENT3:
8991 + ii_write_lock_parent3(inode);
8998 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9001 + case AuLsc_DI_CHILD:
9002 + ii_read_lock_child(inode);
9004 + case AuLsc_DI_CHILD2:
9005 + ii_read_lock_child2(inode);
9007 + case AuLsc_DI_CHILD3:
9008 + ii_read_lock_child3(inode);
9010 + case AuLsc_DI_PARENT:
9011 + ii_read_lock_parent(inode);
9013 + case AuLsc_DI_PARENT2:
9014 + ii_read_lock_parent2(inode);
9016 + case AuLsc_DI_PARENT3:
9017 + ii_read_lock_parent3(inode);
9024 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9026 + struct inode *inode;
9028 + au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9029 + if (d_really_is_positive(d)) {
9030 + inode = d_inode(d);
9031 + if (au_ftest_lock(flags, IW))
9032 + do_ii_write_lock(inode, lsc);
9033 + else if (au_ftest_lock(flags, IR))
9034 + do_ii_read_lock(inode, lsc);
9038 +void di_read_unlock(struct dentry *d, int flags)
9040 + struct inode *inode;
9042 + if (d_really_is_positive(d)) {
9043 + inode = d_inode(d);
9044 + if (au_ftest_lock(flags, IW)) {
9045 + au_dbg_verify_dinode(d);
9046 + ii_write_unlock(inode);
9047 + } else if (au_ftest_lock(flags, IR)) {
9048 + au_dbg_verify_dinode(d);
9049 + ii_read_unlock(inode);
9052 + au_rw_read_unlock(&au_di(d)->di_rwsem);
9055 +void di_downgrade_lock(struct dentry *d, int flags)
9057 + if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9058 + ii_downgrade_lock(d_inode(d));
9059 + au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9062 +void di_write_lock(struct dentry *d, unsigned int lsc)
9064 + au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9065 + if (d_really_is_positive(d))
9066 + do_ii_write_lock(d_inode(d), lsc);
9069 +void di_write_unlock(struct dentry *d)
9071 + au_dbg_verify_dinode(d);
9072 + if (d_really_is_positive(d))
9073 + ii_write_unlock(d_inode(d));
9074 + au_rw_write_unlock(&au_di(d)->di_rwsem);
9077 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9079 + AuDebugOn(d1 == d2
9080 + || d_inode(d1) == d_inode(d2)
9081 + || d1->d_sb != d2->d_sb);
9083 + if (isdir && au_test_subdir(d1, d2)) {
9084 + di_write_lock_child(d1);
9085 + di_write_lock_child2(d2);
9087 + /* there should be no races */
9088 + di_write_lock_child(d2);
9089 + di_write_lock_child2(d1);
9093 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9095 + AuDebugOn(d1 == d2
9096 + || d_inode(d1) == d_inode(d2)
9097 + || d1->d_sb != d2->d_sb);
9099 + if (isdir && au_test_subdir(d1, d2)) {
9100 + di_write_lock_parent(d1);
9101 + di_write_lock_parent2(d2);
9103 + /* there should be no races */
9104 + di_write_lock_parent(d2);
9105 + di_write_lock_parent2(d1);
9109 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9111 + di_write_unlock(d1);
9112 + if (d_inode(d1) == d_inode(d2))
9113 + au_rw_write_unlock(&au_di(d2)->di_rwsem);
9115 + di_write_unlock(d2);
9118 +/* ---------------------------------------------------------------------- */
9120 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9124 + DiMustAnyLock(dentry);
9126 + if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
9128 + AuDebugOn(bindex < 0);
9129 + d = au_di(dentry)->di_hdentry[0 + bindex].hd_dentry;
9130 + AuDebugOn(d && au_dcount(d) <= 0);
9135 + * extended version of au_h_dptr().
9136 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9139 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9141 + struct dentry *h_dentry;
9142 + struct inode *inode, *h_inode;
9144 + AuDebugOn(d_really_is_negative(dentry));
9147 + if (au_dbstart(dentry) <= bindex
9148 + && bindex <= au_dbend(dentry))
9149 + h_dentry = au_h_dptr(dentry, bindex);
9150 + if (h_dentry && !au_d_linkable(h_dentry)) {
9152 + goto out; /* success */
9155 + inode = d_inode(dentry);
9156 + AuDebugOn(bindex < au_ibstart(inode));
9157 + AuDebugOn(au_ibend(inode) < bindex);
9158 + h_inode = au_h_iptr(inode, bindex);
9159 + h_dentry = d_find_alias(h_inode);
9161 + if (!IS_ERR(h_dentry)) {
9162 + if (!au_d_linkable(h_dentry))
9163 + goto out; /* success */
9169 + if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9170 + h_dentry = au_plink_lkup(inode, bindex);
9171 + AuDebugOn(!h_dentry);
9172 + if (!IS_ERR(h_dentry)) {
9173 + if (!au_d_hashed_positive(h_dentry))
9174 + goto out; /* success */
9181 + AuDbgDentry(h_dentry);
9185 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9187 + aufs_bindex_t bend, bwh;
9189 + bend = au_dbend(dentry);
9191 + bwh = au_dbwh(dentry);
9194 + if (0 < bwh && bwh < bend)
9200 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9202 + aufs_bindex_t bend, bopq;
9204 + bend = au_dbtail(dentry);
9206 + bopq = au_dbdiropq(dentry);
9207 + if (0 <= bopq && bopq < bend)
9213 +/* ---------------------------------------------------------------------- */
9215 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9216 + struct dentry *h_dentry)
9218 + struct au_hdentry *hd = au_di(dentry)->di_hdentry + bindex;
9219 + struct au_branch *br;
9221 + DiMustWriteLock(dentry);
9224 + hd->hd_dentry = h_dentry;
9226 + br = au_sbr(dentry->d_sb, bindex);
9227 + hd->hd_id = br->br_id;
9231 +int au_dbrange_test(struct dentry *dentry)
9234 + aufs_bindex_t bstart, bend;
9237 + bstart = au_dbstart(dentry);
9238 + bend = au_dbend(dentry);
9240 + AuDebugOn(bend < 0 && bstart > bend);
9243 + AuDebugOn(bend >= 0);
9249 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9254 + if (unlikely(au_digen(dentry) != sigen
9255 + || au_iigen_test(d_inode(dentry), sigen)))
9261 +void au_update_digen(struct dentry *dentry)
9263 + atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9264 + /* smp_mb(); */ /* atomic_set */
9267 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9269 + struct au_dinfo *dinfo;
9270 + struct dentry *h_d;
9271 + struct au_hdentry *hdp;
9273 + DiMustWriteLock(dentry);
9275 + dinfo = au_di(dentry);
9276 + if (!dinfo || dinfo->di_bstart < 0)
9279 + hdp = dinfo->di_hdentry;
9280 + if (do_put_zero) {
9281 + aufs_bindex_t bindex, bend;
9283 + bend = dinfo->di_bend;
9284 + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++) {
9285 + h_d = hdp[0 + bindex].hd_dentry;
9286 + if (h_d && d_is_negative(h_d))
9287 + au_set_h_dptr(dentry, bindex, NULL);
9291 + dinfo->di_bstart = -1;
9292 + while (++dinfo->di_bstart <= dinfo->di_bend)
9293 + if (hdp[0 + dinfo->di_bstart].hd_dentry)
9295 + if (dinfo->di_bstart > dinfo->di_bend) {
9296 + dinfo->di_bstart = -1;
9297 + dinfo->di_bend = -1;
9302 + while (0 <= --dinfo->di_bend)
9303 + if (hdp[0 + dinfo->di_bend].hd_dentry)
9305 + AuDebugOn(dinfo->di_bstart > dinfo->di_bend || dinfo->di_bend < 0);
9308 +void au_update_dbstart(struct dentry *dentry)
9310 + aufs_bindex_t bindex, bend;
9311 + struct dentry *h_dentry;
9313 + bend = au_dbend(dentry);
9314 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
9315 + h_dentry = au_h_dptr(dentry, bindex);
9318 + if (d_is_positive(h_dentry)) {
9319 + au_set_dbstart(dentry, bindex);
9322 + au_set_h_dptr(dentry, bindex, NULL);
9326 +void au_update_dbend(struct dentry *dentry)
9328 + aufs_bindex_t bindex, bstart;
9329 + struct dentry *h_dentry;
9331 + bstart = au_dbstart(dentry);
9332 + for (bindex = au_dbend(dentry); bindex >= bstart; bindex--) {
9333 + h_dentry = au_h_dptr(dentry, bindex);
9336 + if (d_is_positive(h_dentry)) {
9337 + au_set_dbend(dentry, bindex);
9340 + au_set_h_dptr(dentry, bindex, NULL);
9344 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9346 + aufs_bindex_t bindex, bend;
9348 + bend = au_dbend(dentry);
9349 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++)
9350 + if (au_h_dptr(dentry, bindex) == h_dentry)
9354 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9355 --- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
9356 +++ linux/fs/aufs/dir.c 2016-01-13 20:11:11.669760262 +0100
9359 + * Copyright (C) 2005-2015 Junjiro R. Okajima
9361 + * This program, aufs is free software; you can redistribute it and/or modify
9362 + * it under the terms of the GNU General Public License as published by
9363 + * the Free Software Foundation; either version 2 of the License, or
9364 + * (at your option) any later version.
9366 + * This program is distributed in the hope that it will be useful,
9367 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9368 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9369 + * GNU General Public License for more details.
9371 + * You should have received a copy of the GNU General Public License
9372 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9376 + * directory operations
9379 +#include <linux/fs_stack.h>
9382 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
9384 + unsigned int nlink;
9386 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9388 + nlink = dir->i_nlink;
9389 + nlink += h_dir->i_nlink - 2;
9390 + if (h_dir->i_nlink < 2)
9392 + smp_mb(); /* for i_nlink */
9393 + /* 0 can happen in revaliding */
9394 + set_nlink(dir, nlink);
9397 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
9399 + unsigned int nlink;
9401 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9403 + nlink = dir->i_nlink;
9404 + nlink -= h_dir->i_nlink - 2;
9405 + if (h_dir->i_nlink < 2)
9407 + smp_mb(); /* for i_nlink */
9408 + /* nlink == 0 means the branch-fs is broken */
9409 + set_nlink(dir, nlink);
9412 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
9415 + aufs_bindex_t bindex, bend;
9416 + struct file *h_file;
9417 + struct dentry *h_dentry;
9421 + AuDebugOn(!d_is_dir(file->f_path.dentry));
9423 + bend = au_fbend_dir(file);
9424 + for (bindex = au_fbstart(file);
9425 + bindex <= bend && sz < KMALLOC_MAX_SIZE;
9427 + h_file = au_hf_dir(file, bindex);
9428 + if (h_file && file_inode(h_file))
9429 + sz += vfsub_f_size_read(h_file);
9432 + AuDebugOn(!dentry);
9433 + AuDebugOn(!d_is_dir(dentry));
9435 + bend = au_dbtaildir(dentry);
9436 + for (bindex = au_dbstart(dentry);
9437 + bindex <= bend && sz < KMALLOC_MAX_SIZE;
9439 + h_dentry = au_h_dptr(dentry, bindex);
9440 + if (h_dentry && d_is_positive(h_dentry))
9441 + sz += i_size_read(d_inode(h_dentry));
9444 + if (sz < KMALLOC_MAX_SIZE)
9445 + sz = roundup_pow_of_two(sz);
9446 + if (sz > KMALLOC_MAX_SIZE)
9447 + sz = KMALLOC_MAX_SIZE;
9448 + else if (sz < NAME_MAX) {
9449 + BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
9450 + sz = AUFS_RDBLK_DEF;
9455 +struct au_dir_ts_arg {
9456 + struct dentry *dentry;
9457 + aufs_bindex_t brid;
9460 +static void au_do_dir_ts(void *arg)
9462 + struct au_dir_ts_arg *a = arg;
9463 + struct au_dtime dt;
9464 + struct path h_path;
9465 + struct inode *dir, *h_dir;
9466 + struct super_block *sb;
9467 + struct au_branch *br;
9468 + struct au_hinode *hdir;
9470 + aufs_bindex_t bstart, bindex;
9472 + sb = a->dentry->d_sb;
9473 + if (d_really_is_negative(a->dentry))
9475 + /* no dir->i_mutex lock */
9476 + aufs_read_lock(a->dentry, AuLock_DW); /* noflush */
9478 + dir = d_inode(a->dentry);
9479 + bstart = au_ibstart(dir);
9480 + bindex = au_br_index(sb, a->brid);
9481 + if (bindex < bstart)
9484 + br = au_sbr(sb, bindex);
9485 + h_path.dentry = au_h_dptr(a->dentry, bindex);
9486 + if (!h_path.dentry)
9488 + h_path.mnt = au_br_mnt(br);
9489 + au_dtime_store(&dt, a->dentry, &h_path);
9491 + br = au_sbr(sb, bstart);
9492 + if (!au_br_writable(br->br_perm))
9494 + h_path.dentry = au_h_dptr(a->dentry, bstart);
9495 + h_path.mnt = au_br_mnt(br);
9496 + err = vfsub_mnt_want_write(h_path.mnt);
9499 + hdir = au_hi(dir, bstart);
9500 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
9501 + h_dir = au_h_iptr(dir, bstart);
9502 + if (h_dir->i_nlink
9503 + && timespec_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
9504 + dt.dt_h_path = h_path;
9505 + au_dtime_revert(&dt);
9507 + au_hn_imtx_unlock(hdir);
9508 + vfsub_mnt_drop_write(h_path.mnt);
9509 + au_cpup_attr_timesizes(dir);
9512 + aufs_read_unlock(a->dentry, AuLock_DW);
9515 + au_nwt_done(&au_sbi(sb)->si_nowait);
9519 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
9521 + int perm, wkq_err;
9522 + aufs_bindex_t bstart;
9523 + struct au_dir_ts_arg *arg;
9524 + struct dentry *dentry;
9525 + struct super_block *sb;
9529 + dentry = d_find_any_alias(dir);
9530 + AuDebugOn(!dentry);
9531 + sb = dentry->d_sb;
9532 + bstart = au_ibstart(dir);
9533 + if (bstart == bindex) {
9534 + au_cpup_attr_timesizes(dir);
9538 + perm = au_sbr_perm(sb, bstart);
9539 + if (!au_br_writable(perm))
9542 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
9546 + arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
9547 + arg->brid = au_sbr_id(sb, bindex);
9548 + wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
9549 + if (unlikely(wkq_err)) {
9550 + pr_err("wkq %d\n", wkq_err);
9559 +/* ---------------------------------------------------------------------- */
9561 +static int reopen_dir(struct file *file)
9564 + unsigned int flags;
9565 + aufs_bindex_t bindex, btail, bstart;
9566 + struct dentry *dentry, *h_dentry;
9567 + struct file *h_file;
9569 + /* open all lower dirs */
9570 + dentry = file->f_path.dentry;
9571 + bstart = au_dbstart(dentry);
9572 + for (bindex = au_fbstart(file); bindex < bstart; bindex++)
9573 + au_set_h_fptr(file, bindex, NULL);
9574 + au_set_fbstart(file, bstart);
9576 + btail = au_dbtaildir(dentry);
9577 + for (bindex = au_fbend_dir(file); btail < bindex; bindex--)
9578 + au_set_h_fptr(file, bindex, NULL);
9579 + au_set_fbend_dir(file, btail);
9581 + flags = vfsub_file_flags(file);
9582 + for (bindex = bstart; bindex <= btail; bindex++) {
9583 + h_dentry = au_h_dptr(dentry, bindex);
9586 + h_file = au_hf_dir(file, bindex);
9590 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
9591 + err = PTR_ERR(h_file);
9592 + if (IS_ERR(h_file))
9593 + goto out; /* close all? */
9594 + au_set_h_fptr(file, bindex, h_file);
9596 + au_update_figen(file);
9597 + /* todo: necessary? */
9598 + /* file->f_ra = h_file->f_ra; */
9605 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
9608 + aufs_bindex_t bindex, btail;
9609 + struct dentry *dentry, *h_dentry;
9611 + FiMustWriteLock(file);
9612 + AuDebugOn(h_file);
9615 + dentry = file->f_path.dentry;
9616 + file->f_version = d_inode(dentry)->i_version;
9617 + bindex = au_dbstart(dentry);
9618 + au_set_fbstart(file, bindex);
9619 + btail = au_dbtaildir(dentry);
9620 + au_set_fbend_dir(file, btail);
9621 + for (; !err && bindex <= btail; bindex++) {
9622 + h_dentry = au_h_dptr(dentry, bindex);
9626 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
9627 + if (IS_ERR(h_file)) {
9628 + err = PTR_ERR(h_file);
9631 + au_set_h_fptr(file, bindex, h_file);
9633 + au_update_figen(file);
9634 + /* todo: necessary? */
9635 + /* file->f_ra = h_file->f_ra; */
9637 + return 0; /* success */
9640 + for (bindex = au_fbstart(file); bindex <= btail; bindex++)
9641 + au_set_h_fptr(file, bindex, NULL);
9642 + au_set_fbstart(file, -1);
9643 + au_set_fbend_dir(file, -1);
9648 +static int aufs_open_dir(struct inode *inode __maybe_unused,
9649 + struct file *file)
9652 + struct super_block *sb;
9653 + struct au_fidir *fidir;
9656 + sb = file->f_path.dentry->d_sb;
9657 + si_read_lock(sb, AuLock_FLUSH);
9658 + fidir = au_fidir_alloc(sb);
9660 + struct au_do_open_args args = {
9661 + .open = do_open_dir,
9664 + err = au_do_open(file, &args);
9665 + if (unlikely(err))
9668 + si_read_unlock(sb);
9672 +static int aufs_release_dir(struct inode *inode __maybe_unused,
9673 + struct file *file)
9675 + struct au_vdir *vdir_cache;
9676 + struct au_finfo *finfo;
9677 + struct au_fidir *fidir;
9678 + aufs_bindex_t bindex, bend;
9680 + finfo = au_fi(file);
9681 + fidir = finfo->fi_hdir;
9683 + au_sphl_del(&finfo->fi_hlist,
9684 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
9685 + vdir_cache = fidir->fd_vdir_cache; /* lock-free */
9687 + au_vdir_free(vdir_cache);
9689 + bindex = finfo->fi_btop;
9690 + if (bindex >= 0) {
9692 + * calls fput() instead of filp_close(),
9693 + * since no dnotify or lock for the lower file.
9695 + bend = fidir->fd_bbot;
9696 + for (; bindex <= bend; bindex++)
9697 + au_set_h_fptr(file, bindex, NULL);
9700 + finfo->fi_hdir = NULL;
9702 + au_finfo_fin(file);
9706 +/* ---------------------------------------------------------------------- */
9708 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
9711 + aufs_bindex_t bindex, bend;
9712 + struct file *h_file;
9715 + bend = au_fbend_dir(file);
9716 + for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
9717 + h_file = au_hf_dir(file, bindex);
9719 + err = vfsub_flush(h_file, id);
9724 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
9726 + return au_do_flush(file, id, au_do_flush_dir);
9729 +/* ---------------------------------------------------------------------- */
9731 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
9734 + aufs_bindex_t bend, bindex;
9735 + struct inode *inode;
9736 + struct super_block *sb;
9739 + sb = dentry->d_sb;
9740 + inode = d_inode(dentry);
9742 + bend = au_dbend(dentry);
9743 + for (bindex = au_dbstart(dentry); !err && bindex <= bend; bindex++) {
9744 + struct path h_path;
9746 + if (au_test_ro(sb, bindex, inode))
9748 + h_path.dentry = au_h_dptr(dentry, bindex);
9749 + if (!h_path.dentry)
9752 + h_path.mnt = au_sbr_mnt(sb, bindex);
9753 + err = vfsub_fsync(NULL, &h_path, datasync);
9759 +static int au_do_fsync_dir(struct file *file, int datasync)
9762 + aufs_bindex_t bend, bindex;
9763 + struct file *h_file;
9764 + struct super_block *sb;
9765 + struct inode *inode;
9767 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9768 + if (unlikely(err))
9771 + inode = file_inode(file);
9773 + bend = au_fbend_dir(file);
9774 + for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
9775 + h_file = au_hf_dir(file, bindex);
9776 + if (!h_file || au_test_ro(sb, bindex, inode))
9779 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
9787 + * @file may be NULL
9789 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
9793 + struct dentry *dentry;
9794 + struct inode *inode;
9795 + struct super_block *sb;
9796 + struct mutex *mtx;
9799 + dentry = file->f_path.dentry;
9800 + inode = d_inode(dentry);
9801 + mtx = &inode->i_mutex;
9803 + sb = dentry->d_sb;
9804 + si_noflush_read_lock(sb);
9806 + err = au_do_fsync_dir(file, datasync);
9808 + di_write_lock_child(dentry);
9809 + err = au_do_fsync_dir_no_file(dentry, datasync);
9811 + au_cpup_attr_timesizes(inode);
9812 + di_write_unlock(dentry);
9814 + fi_write_unlock(file);
9816 + si_read_unlock(sb);
9817 + mutex_unlock(mtx);
9821 +/* ---------------------------------------------------------------------- */
9823 +static int aufs_iterate(struct file *file, struct dir_context *ctx)
9826 + struct dentry *dentry;
9827 + struct inode *inode, *h_inode;
9828 + struct super_block *sb;
9830 + AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
9832 + dentry = file->f_path.dentry;
9833 + inode = d_inode(dentry);
9836 + sb = dentry->d_sb;
9837 + si_read_lock(sb, AuLock_FLUSH);
9838 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9839 + if (unlikely(err))
9841 + err = au_alive_dir(dentry);
9843 + err = au_vdir_init(file);
9844 + di_downgrade_lock(dentry, AuLock_IR);
9845 + if (unlikely(err))
9848 + h_inode = au_h_iptr(inode, au_ibstart(inode));
9849 + if (!au_test_nfsd()) {
9850 + err = au_vdir_fill_de(file, ctx);
9851 + fsstack_copy_attr_atime(inode, h_inode);
9854 + * nfsd filldir may call lookup_one_len(), vfs_getattr(),
9855 + * encode_fh() and others.
9857 + atomic_inc(&h_inode->i_count);
9858 + di_read_unlock(dentry, AuLock_IR);
9859 + si_read_unlock(sb);
9860 + err = au_vdir_fill_de(file, ctx);
9861 + fsstack_copy_attr_atime(inode, h_inode);
9862 + fi_write_unlock(file);
9870 + di_read_unlock(dentry, AuLock_IR);
9871 + fi_write_unlock(file);
9873 + si_read_unlock(sb);
9877 +/* ---------------------------------------------------------------------- */
9879 +#define AuTestEmpty_WHONLY 1
9880 +#define AuTestEmpty_CALLED (1 << 1)
9881 +#define AuTestEmpty_SHWH (1 << 2)
9882 +#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
9883 +#define au_fset_testempty(flags, name) \
9884 + do { (flags) |= AuTestEmpty_##name; } while (0)
9885 +#define au_fclr_testempty(flags, name) \
9886 + do { (flags) &= ~AuTestEmpty_##name; } while (0)
9888 +#ifndef CONFIG_AUFS_SHWH
9889 +#undef AuTestEmpty_SHWH
9890 +#define AuTestEmpty_SHWH 0
9893 +struct test_empty_arg {
9894 + struct dir_context ctx;
9895 + struct au_nhash *whlist;
9896 + unsigned int flags;
9898 + aufs_bindex_t bindex;
9901 +static int test_empty_cb(struct dir_context *ctx, const char *__name,
9902 + int namelen, loff_t offset __maybe_unused, u64 ino,
9903 + unsigned int d_type)
9905 + struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
9907 + char *name = (void *)__name;
9910 + au_fset_testempty(arg->flags, CALLED);
9912 + if (name[0] == '.'
9913 + && (namelen == 1 || (name[1] == '.' && namelen == 2)))
9914 + goto out; /* success */
9916 + if (namelen <= AUFS_WH_PFX_LEN
9917 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
9918 + if (au_ftest_testempty(arg->flags, WHONLY)
9919 + && !au_nhash_test_known_wh(arg->whlist, name, namelen))
9920 + arg->err = -ENOTEMPTY;
9924 + name += AUFS_WH_PFX_LEN;
9925 + namelen -= AUFS_WH_PFX_LEN;
9926 + if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
9927 + arg->err = au_nhash_append_wh
9928 + (arg->whlist, name, namelen, ino, d_type, arg->bindex,
9929 + au_ftest_testempty(arg->flags, SHWH));
9933 + AuTraceErr(arg->err);
9937 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9940 + struct file *h_file;
9942 + h_file = au_h_open(dentry, arg->bindex,
9943 + O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
9944 + /*file*/NULL, /*force_wr*/0);
9945 + err = PTR_ERR(h_file);
9946 + if (IS_ERR(h_file))
9950 + if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
9951 + && !file_inode(h_file)->i_nlink)
9956 + au_fclr_testempty(arg->flags, CALLED);
9958 + err = vfsub_iterate_dir(h_file, &arg->ctx);
9961 + } while (!err && au_ftest_testempty(arg->flags, CALLED));
9965 + au_sbr_put(dentry->d_sb, arg->bindex);
9970 +struct do_test_empty_args {
9972 + struct dentry *dentry;
9973 + struct test_empty_arg *arg;
9976 +static void call_do_test_empty(void *args)
9978 + struct do_test_empty_args *a = args;
9979 + *a->errp = do_test_empty(a->dentry, a->arg);
9982 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9985 + struct dentry *h_dentry;
9986 + struct inode *h_inode;
9988 + h_dentry = au_h_dptr(dentry, arg->bindex);
9989 + h_inode = d_inode(h_dentry);
9990 + /* todo: i_mode changes anytime? */
9991 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
9992 + err = au_test_h_perm_sio(h_inode, MAY_EXEC | MAY_READ);
9993 + mutex_unlock(&h_inode->i_mutex);
9995 + err = do_test_empty(dentry, arg);
9997 + struct do_test_empty_args args = {
10002 + unsigned int flags = arg->flags;
10004 + wkq_err = au_wkq_wait(call_do_test_empty, &args);
10005 + if (unlikely(wkq_err))
10007 + arg->flags = flags;
10013 +int au_test_empty_lower(struct dentry *dentry)
10016 + unsigned int rdhash;
10017 + aufs_bindex_t bindex, bstart, btail;
10018 + struct au_nhash whlist;
10019 + struct test_empty_arg arg = {
10021 + .actor = test_empty_cb
10024 + int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10026 + SiMustAnyLock(dentry->d_sb);
10028 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10030 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10031 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10032 + if (unlikely(err))
10036 + arg.whlist = &whlist;
10037 + bstart = au_dbstart(dentry);
10038 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10039 + au_fset_testempty(arg.flags, SHWH);
10040 + test_empty = do_test_empty;
10041 + if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10042 + test_empty = sio_test_empty;
10043 + arg.bindex = bstart;
10044 + err = test_empty(dentry, &arg);
10045 + if (unlikely(err))
10048 + au_fset_testempty(arg.flags, WHONLY);
10049 + btail = au_dbtaildir(dentry);
10050 + for (bindex = bstart + 1; !err && bindex <= btail; bindex++) {
10051 + struct dentry *h_dentry;
10053 + h_dentry = au_h_dptr(dentry, bindex);
10054 + if (h_dentry && d_is_positive(h_dentry)) {
10055 + arg.bindex = bindex;
10056 + err = test_empty(dentry, &arg);
10061 + au_nhash_wh_free(&whlist);
10066 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10069 + struct test_empty_arg arg = {
10071 + .actor = test_empty_cb
10074 + aufs_bindex_t bindex, btail;
10077 + arg.whlist = whlist;
10078 + arg.flags = AuTestEmpty_WHONLY;
10079 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10080 + au_fset_testempty(arg.flags, SHWH);
10081 + btail = au_dbtaildir(dentry);
10082 + for (bindex = au_dbstart(dentry); !err && bindex <= btail; bindex++) {
10083 + struct dentry *h_dentry;
10085 + h_dentry = au_h_dptr(dentry, bindex);
10086 + if (h_dentry && d_is_positive(h_dentry)) {
10087 + arg.bindex = bindex;
10088 + err = sio_test_empty(dentry, &arg);
10095 +/* ---------------------------------------------------------------------- */
10097 +const struct file_operations aufs_dir_fop = {
10098 + .owner = THIS_MODULE,
10099 + .llseek = default_llseek,
10100 + .read = generic_read_dir,
10101 + .iterate = aufs_iterate,
10102 + .unlocked_ioctl = aufs_ioctl_dir,
10103 +#ifdef CONFIG_COMPAT
10104 + .compat_ioctl = aufs_compat_ioctl_dir,
10106 + .open = aufs_open_dir,
10107 + .release = aufs_release_dir,
10108 + .flush = aufs_flush_dir,
10109 + .fsync = aufs_fsync_dir
10111 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10112 --- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
10113 +++ linux/fs/aufs/dir.h 2016-01-13 20:11:11.669760262 +0100
10116 + * Copyright (C) 2005-2015 Junjiro R. Okajima
10118 + * This program, aufs is free software; you can redistribute it and/or modify
10119 + * it under the terms of the GNU General Public License as published by
10120 + * the Free Software Foundation; either version 2 of the License, or
10121 + * (at your option) any later version.
10123 + * This program is distributed in the hope that it will be useful,
10124 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10125 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10126 + * GNU General Public License for more details.
10128 + * You should have received a copy of the GNU General Public License
10129 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10133 + * directory operations
10136 +#ifndef __AUFS_DIR_H__
10137 +#define __AUFS_DIR_H__
10141 +#include <linux/fs.h>
10143 +/* ---------------------------------------------------------------------- */
10145 +/* need to be faster and smaller */
10148 + unsigned int nh_num;
10149 + struct hlist_head *nh_head;
10152 +struct au_vdir_destr {
10153 + unsigned char len;
10154 + unsigned char name[0];
10157 +struct au_vdir_dehstr {
10158 + struct hlist_node hash;
10159 + struct au_vdir_destr *str;
10160 +} ____cacheline_aligned_in_smp;
10162 +struct au_vdir_de {
10164 + unsigned char de_type;
10165 + /* caution: packed */
10166 + struct au_vdir_destr de_str;
10169 +struct au_vdir_wh {
10170 + struct hlist_node wh_hash;
10171 +#ifdef CONFIG_AUFS_SHWH
10173 + aufs_bindex_t wh_bindex;
10174 + unsigned char wh_type;
10176 + aufs_bindex_t wh_bindex;
10178 + /* caution: packed */
10179 + struct au_vdir_destr wh_str;
10182 +union au_vdir_deblk_p {
10183 + unsigned char *deblk;
10184 + struct au_vdir_de *de;
10188 + unsigned char **vd_deblk;
10189 + unsigned long vd_nblk;
10191 + unsigned long ul;
10192 + union au_vdir_deblk_p p;
10195 + unsigned long vd_version;
10196 + unsigned int vd_deblk_sz;
10197 + unsigned long vd_jiffy;
10198 +} ____cacheline_aligned_in_smp;
10200 +/* ---------------------------------------------------------------------- */
10203 +extern const struct file_operations aufs_dir_fop;
10204 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10205 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10206 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10207 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10208 +int au_test_empty_lower(struct dentry *dentry);
10209 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10212 +unsigned int au_rdhash_est(loff_t sz);
10213 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10214 +void au_nhash_wh_free(struct au_nhash *whlist);
10215 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10217 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10218 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10219 + unsigned int d_type, aufs_bindex_t bindex,
10220 + unsigned char shwh);
10221 +void au_vdir_free(struct au_vdir *vdir);
10222 +int au_vdir_init(struct file *file);
10223 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10226 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10228 +#ifdef CONFIG_AUFS_RDU
10230 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10231 +#ifdef CONFIG_COMPAT
10232 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10233 + unsigned long arg);
10236 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10237 + unsigned int cmd, unsigned long arg)
10238 +#ifdef CONFIG_COMPAT
10239 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10240 + unsigned int cmd, unsigned long arg)
10244 +#endif /* __KERNEL__ */
10245 +#endif /* __AUFS_DIR_H__ */
10246 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
10247 --- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
10248 +++ linux/fs/aufs/dynop.c 2016-01-13 20:11:11.669760262 +0100
10251 + * Copyright (C) 2010-2015 Junjiro R. Okajima
10253 + * This program, aufs is free software; you can redistribute it and/or modify
10254 + * it under the terms of the GNU General Public License as published by
10255 + * the Free Software Foundation; either version 2 of the License, or
10256 + * (at your option) any later version.
10258 + * This program is distributed in the hope that it will be useful,
10259 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10260 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10261 + * GNU General Public License for more details.
10263 + * You should have received a copy of the GNU General Public License
10264 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10268 + * dynamically customizable operations for regular files
10273 +#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
10276 + * How large will these lists be?
10277 + * Usually just a few elements, 20-30 at most for each, I guess.
10279 +static struct au_splhead dynop[AuDyLast];
10281 +static struct au_dykey *dy_gfind_get(struct au_splhead *spl, const void *h_op)
10283 + struct au_dykey *key, *tmp;
10284 + struct list_head *head;
10287 + head = &spl->head;
10289 + list_for_each_entry_rcu(tmp, head, dk_list)
10290 + if (tmp->dk_op.dy_hop == h_op) {
10292 + kref_get(&key->dk_kref);
10295 + rcu_read_unlock();
10300 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
10302 + struct au_dykey **k, *found;
10303 + const void *h_op = key->dk_op.dy_hop;
10307 + k = br->br_dykey;
10308 + for (i = 0; i < AuBrDynOp; i++)
10310 + if (k[i]->dk_op.dy_hop == h_op) {
10317 + spin_lock(&br->br_dykey_lock);
10318 + for (; i < AuBrDynOp; i++)
10320 + if (k[i]->dk_op.dy_hop == h_op) {
10328 + spin_unlock(&br->br_dykey_lock);
10329 + BUG_ON(i == AuBrDynOp); /* expand the array */
10335 +/* kref_get() if @key is already added */
10336 +static struct au_dykey *dy_gadd(struct au_splhead *spl, struct au_dykey *key)
10338 + struct au_dykey *tmp, *found;
10339 + struct list_head *head;
10340 + const void *h_op = key->dk_op.dy_hop;
10343 + head = &spl->head;
10344 + spin_lock(&spl->spin);
10345 + list_for_each_entry(tmp, head, dk_list)
10346 + if (tmp->dk_op.dy_hop == h_op) {
10347 + kref_get(&tmp->dk_kref);
10352 + list_add_rcu(&key->dk_list, head);
10353 + spin_unlock(&spl->spin);
10360 +static void dy_free_rcu(struct rcu_head *rcu)
10362 + struct au_dykey *key;
10364 + key = container_of(rcu, struct au_dykey, dk_rcu);
10369 +static void dy_free(struct kref *kref)
10371 + struct au_dykey *key;
10372 + struct au_splhead *spl;
10374 + key = container_of(kref, struct au_dykey, dk_kref);
10375 + spl = dynop + key->dk_op.dy_type;
10376 + au_spl_del_rcu(&key->dk_list, spl);
10377 + call_rcu(&key->dk_rcu, dy_free_rcu);
10380 +void au_dy_put(struct au_dykey *key)
10382 + kref_put(&key->dk_kref, dy_free);
10385 +/* ---------------------------------------------------------------------- */
10387 +#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
10389 +#ifdef CONFIG_AUFS_DEBUG
10390 +#define DyDbgDeclare(cnt) unsigned int cnt = 0
10391 +#define DyDbgInc(cnt) do { cnt++; } while (0)
10393 +#define DyDbgDeclare(cnt) do {} while (0)
10394 +#define DyDbgInc(cnt) do {} while (0)
10397 +#define DySet(func, dst, src, h_op, h_sb) do { \
10399 + if (h_op->func) { \
10401 + dst.func = src.func; \
10403 + AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
10407 +#define DySetForce(func, dst, src) do { \
10408 + AuDebugOn(!src.func); \
10410 + dst.func = src.func; \
10413 +#define DySetAop(func) \
10414 + DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
10415 +#define DySetAopForce(func) \
10416 + DySetForce(func, dyaop->da_op, aufs_aop)
10418 +static void dy_aop(struct au_dykey *key, const void *h_op,
10419 + struct super_block *h_sb __maybe_unused)
10421 + struct au_dyaop *dyaop = (void *)key;
10422 + const struct address_space_operations *h_aop = h_op;
10423 + DyDbgDeclare(cnt);
10425 + AuDbg("%s\n", au_sbtype(h_sb));
10427 + DySetAop(writepage);
10428 + DySetAopForce(readpage); /* force */
10429 + DySetAop(writepages);
10430 + DySetAop(set_page_dirty);
10431 + DySetAop(readpages);
10432 + DySetAop(write_begin);
10433 + DySetAop(write_end);
10435 + DySetAop(invalidatepage);
10436 + DySetAop(releasepage);
10437 + DySetAop(freepage);
10438 + /* this one will be changed according to an aufs mount option */
10439 + DySetAop(direct_IO);
10440 + DySetAop(migratepage);
10441 + DySetAop(launder_page);
10442 + DySetAop(is_partially_uptodate);
10443 + DySetAop(is_dirty_writeback);
10444 + DySetAop(error_remove_page);
10445 + DySetAop(swap_activate);
10446 + DySetAop(swap_deactivate);
10448 + DyDbgSize(cnt, *h_aop);
10451 +/* ---------------------------------------------------------------------- */
10453 +static void dy_bug(struct kref *kref)
10458 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
10460 + struct au_dykey *key, *old;
10461 + struct au_splhead *spl;
10464 + void (*set)(struct au_dykey *key, const void *h_op,
10465 + struct super_block *h_sb __maybe_unused);
10467 + static const struct op a[] = {
10469 + .sz = sizeof(struct au_dyaop),
10473 + const struct op *p;
10475 + spl = dynop + op->dy_type;
10476 + key = dy_gfind_get(spl, op->dy_hop);
10478 + goto out_add; /* success */
10480 + p = a + op->dy_type;
10481 + key = kzalloc(p->sz, GFP_NOFS);
10482 + if (unlikely(!key)) {
10483 + key = ERR_PTR(-ENOMEM);
10487 + key->dk_op.dy_hop = op->dy_hop;
10488 + kref_init(&key->dk_kref);
10489 + p->set(key, op->dy_hop, au_br_sb(br));
10490 + old = dy_gadd(spl, key);
10497 + old = dy_bradd(br, key);
10499 + /* its ref-count should never be zero here */
10500 + kref_put(&key->dk_kref, dy_bug);
10505 +/* ---------------------------------------------------------------------- */
10507 + * Aufs prohibits O_DIRECT by defaut even if the branch supports it.
10508 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
10509 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
10510 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
10511 + * See the aufs manual in detail.
10513 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
10516 + dyaop->da_op.direct_IO = NULL;
10518 + dyaop->da_op.direct_IO = aufs_aop.direct_IO;
10521 +static struct au_dyaop *dy_aget(struct au_branch *br,
10522 + const struct address_space_operations *h_aop,
10525 + struct au_dyaop *dyaop;
10526 + struct au_dynop op;
10528 + op.dy_type = AuDy_AOP;
10529 + op.dy_haop = h_aop;
10530 + dyaop = (void *)dy_get(&op, br);
10531 + if (IS_ERR(dyaop))
10533 + dy_adx(dyaop, do_dx);
10539 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
10540 + struct inode *h_inode)
10543 + struct super_block *sb;
10544 + struct au_branch *br;
10545 + struct au_dyaop *dyaop;
10547 + AuDebugOn(!S_ISREG(h_inode->i_mode));
10548 + IiMustWriteLock(inode);
10550 + sb = inode->i_sb;
10551 + br = au_sbr(sb, bindex);
10552 + do_dx = !!au_opt_test(au_mntflags(sb), DIO);
10553 + dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
10554 + err = PTR_ERR(dyaop);
10555 + if (IS_ERR(dyaop))
10556 + /* unnecessary to call dy_fput() */
10560 + inode->i_mapping->a_ops = &dyaop->da_op;
10567 + * Is it safe to replace a_ops during the inode/file is in operation?
10568 + * Yes, I hope so.
10570 +int au_dy_irefresh(struct inode *inode)
10573 + aufs_bindex_t bstart;
10574 + struct inode *h_inode;
10577 + if (S_ISREG(inode->i_mode)) {
10578 + bstart = au_ibstart(inode);
10579 + h_inode = au_h_iptr(inode, bstart);
10580 + err = au_dy_iaop(inode, bstart, h_inode);
10585 +void au_dy_arefresh(int do_dx)
10587 + struct au_splhead *spl;
10588 + struct list_head *head;
10589 + struct au_dykey *key;
10591 + spl = dynop + AuDy_AOP;
10592 + head = &spl->head;
10593 + spin_lock(&spl->spin);
10594 + list_for_each_entry(key, head, dk_list)
10595 + dy_adx((void *)key, do_dx);
10596 + spin_unlock(&spl->spin);
10599 +/* ---------------------------------------------------------------------- */
10601 +void __init au_dy_init(void)
10605 + /* make sure that 'struct au_dykey *' can be any type */
10606 + BUILD_BUG_ON(offsetof(struct au_dyaop, da_key));
10608 + for (i = 0; i < AuDyLast; i++)
10609 + au_spl_init(dynop + i);
10612 +void au_dy_fin(void)
10616 + for (i = 0; i < AuDyLast; i++)
10617 + WARN_ON(!list_empty(&dynop[i].head));
10619 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
10620 --- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
10621 +++ linux/fs/aufs/dynop.h 2016-01-13 20:11:11.669760262 +0100
10624 + * Copyright (C) 2010-2015 Junjiro R. Okajima
10626 + * This program, aufs is free software; you can redistribute it and/or modify
10627 + * it under the terms of the GNU General Public License as published by
10628 + * the Free Software Foundation; either version 2 of the License, or
10629 + * (at your option) any later version.
10631 + * This program is distributed in the hope that it will be useful,
10632 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10633 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10634 + * GNU General Public License for more details.
10636 + * You should have received a copy of the GNU General Public License
10637 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10641 + * dynamically customizable operations (for regular files only)
10644 +#ifndef __AUFS_DYNOP_H__
10645 +#define __AUFS_DYNOP_H__
10649 +#include <linux/fs.h>
10650 +#include <linux/kref.h>
10652 +enum {AuDy_AOP, AuDyLast};
10657 + const void *dy_hop;
10658 + const struct address_space_operations *dy_haop;
10664 + struct list_head dk_list;
10665 + struct rcu_head dk_rcu;
10667 + struct au_dynop dk_op;
10670 + * during I am in the branch local array, kref is gotten. when the
10671 + * branch is removed, kref is put.
10673 + struct kref dk_kref;
10676 +/* stop unioning since their sizes are very different from each other */
10678 + struct au_dykey da_key;
10679 + struct address_space_operations da_op; /* not const */
10682 +/* ---------------------------------------------------------------------- */
10686 +void au_dy_put(struct au_dykey *key);
10687 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
10688 + struct inode *h_inode);
10689 +int au_dy_irefresh(struct inode *inode);
10690 +void au_dy_arefresh(int do_dio);
10692 +void __init au_dy_init(void);
10693 +void au_dy_fin(void);
10695 +#endif /* __KERNEL__ */
10696 +#endif /* __AUFS_DYNOP_H__ */
10697 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
10698 --- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
10699 +++ linux/fs/aufs/export.c 2016-01-13 20:11:11.669760262 +0100
10702 + * Copyright (C) 2005-2015 Junjiro R. Okajima
10704 + * This program, aufs is free software; you can redistribute it and/or modify
10705 + * it under the terms of the GNU General Public License as published by
10706 + * the Free Software Foundation; either version 2 of the License, or
10707 + * (at your option) any later version.
10709 + * This program is distributed in the hope that it will be useful,
10710 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10711 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10712 + * GNU General Public License for more details.
10714 + * You should have received a copy of the GNU General Public License
10715 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10722 +#include <linux/exportfs.h>
10723 +#include <linux/fs_struct.h>
10724 +#include <linux/namei.h>
10725 +#include <linux/nsproxy.h>
10726 +#include <linux/random.h>
10727 +#include <linux/writeback.h>
10728 +#include "../fs/mount.h"
10732 +#ifdef CONFIG_AUFS_INO_T_64
10740 +static ino_t decode_ino(__u32 *a)
10744 + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
10746 +#ifdef CONFIG_AUFS_INO_T_64
10752 +static void encode_ino(__u32 *a, ino_t ino)
10758 +#ifdef CONFIG_AUFS_INO_T_64
10763 +/* NFS file handle */
10767 +#ifdef CONFIG_AUFS_INO_T_64
10768 + /* support 64bit inode number */
10781 + Fh_ino = Fh_ino1,
10782 + Fh_dir_ino = Fh_dir_ino1
10785 +static int au_test_anon(struct dentry *dentry)
10787 + /* note: read d_flags without d_lock */
10788 + return !!(dentry->d_flags & DCACHE_DISCONNECTED);
10791 +int au_test_nfsd(void)
10794 + struct task_struct *tsk = current;
10795 + char comm[sizeof(tsk->comm)];
10798 + if (tsk->flags & PF_KTHREAD) {
10799 + get_task_comm(comm, tsk);
10800 + ret = !strcmp(comm, "nfsd");
10806 +/* ---------------------------------------------------------------------- */
10807 +/* inode generation external table */
10809 +void au_xigen_inc(struct inode *inode)
10814 + struct super_block *sb;
10815 + struct au_sbinfo *sbinfo;
10817 + sb = inode->i_sb;
10818 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
10820 + sbinfo = au_sbi(sb);
10821 + pos = inode->i_ino;
10822 + pos *= sizeof(igen);
10823 + igen = inode->i_generation + 1;
10824 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xigen, &igen,
10825 + sizeof(igen), &pos);
10826 + if (sz == sizeof(igen))
10827 + return; /* success */
10829 + if (unlikely(sz >= 0))
10830 + AuIOErr("xigen error (%zd)\n", sz);
10833 +int au_xigen_new(struct inode *inode)
10838 + struct super_block *sb;
10839 + struct au_sbinfo *sbinfo;
10840 + struct file *file;
10843 + /* todo: dirty, at mount time */
10844 + if (inode->i_ino == AUFS_ROOT_INO)
10846 + sb = inode->i_sb;
10847 + SiMustAnyLock(sb);
10848 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
10852 + pos = inode->i_ino;
10853 + if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
10854 + AuIOErr1("too large i%lld\n", pos);
10857 + pos *= sizeof(inode->i_generation);
10860 + sbinfo = au_sbi(sb);
10861 + file = sbinfo->si_xigen;
10864 + if (vfsub_f_size_read(file)
10865 + < pos + sizeof(inode->i_generation)) {
10866 + inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
10867 + sz = xino_fwrite(sbinfo->si_xwrite, file, &inode->i_generation,
10868 + sizeof(inode->i_generation), &pos);
10870 + sz = xino_fread(sbinfo->si_xread, file, &inode->i_generation,
10871 + sizeof(inode->i_generation), &pos);
10872 + if (sz == sizeof(inode->i_generation))
10873 + goto out; /* success */
10876 + if (unlikely(sz >= 0)) {
10878 + AuIOErr("xigen error (%zd)\n", sz);
10885 +int au_xigen_set(struct super_block *sb, struct file *base)
10888 + struct au_sbinfo *sbinfo;
10889 + struct file *file;
10891 + SiMustWriteLock(sb);
10893 + sbinfo = au_sbi(sb);
10894 + file = au_xino_create2(base, sbinfo->si_xigen);
10895 + err = PTR_ERR(file);
10896 + if (IS_ERR(file))
10899 + if (sbinfo->si_xigen)
10900 + fput(sbinfo->si_xigen);
10901 + sbinfo->si_xigen = file;
10907 +void au_xigen_clr(struct super_block *sb)
10909 + struct au_sbinfo *sbinfo;
10911 + SiMustWriteLock(sb);
10913 + sbinfo = au_sbi(sb);
10914 + if (sbinfo->si_xigen) {
10915 + fput(sbinfo->si_xigen);
10916 + sbinfo->si_xigen = NULL;
10920 +/* ---------------------------------------------------------------------- */
10922 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
10925 + struct dentry *dentry, *d;
10926 + struct inode *inode;
10927 + unsigned int sigen;
10930 + inode = ilookup(sb, ino);
10934 + dentry = ERR_PTR(-ESTALE);
10935 + sigen = au_sigen(sb);
10936 + if (unlikely(is_bad_inode(inode)
10937 + || IS_DEADDIR(inode)
10938 + || sigen != au_iigen(inode, NULL)))
10942 + if (!dir_ino || S_ISDIR(inode->i_mode))
10943 + dentry = d_find_alias(inode);
10945 + spin_lock(&inode->i_lock);
10946 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
10947 + spin_lock(&d->d_lock);
10948 + if (!au_test_anon(d)
10949 + && d_inode(d->d_parent)->i_ino == dir_ino) {
10950 + dentry = dget_dlock(d);
10951 + spin_unlock(&d->d_lock);
10954 + spin_unlock(&d->d_lock);
10956 + spin_unlock(&inode->i_lock);
10958 + if (unlikely(dentry && au_digen_test(dentry, sigen))) {
10959 + /* need to refresh */
10967 + AuTraceErrPtr(dentry);
10971 +/* ---------------------------------------------------------------------- */
10973 +/* todo: dirty? */
10974 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
10976 +struct au_compare_mnt_args {
10978 + struct super_block *sb;
10981 + struct vfsmount *mnt;
10984 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
10986 + struct au_compare_mnt_args *a = arg;
10988 + if (mnt->mnt_sb != a->sb)
10990 + a->mnt = mntget(mnt);
10994 +static struct vfsmount *au_mnt_get(struct super_block *sb)
10997 + struct path root;
10998 + struct au_compare_mnt_args args = {
11002 + get_fs_root(current->fs, &root);
11004 + err = iterate_mounts(au_compare_mnt, &args, root.mnt);
11005 + rcu_read_unlock();
11008 + AuDebugOn(!args.mnt);
11012 +struct au_nfsd_si_lock {
11013 + unsigned int sigen;
11014 + aufs_bindex_t bindex, br_id;
11015 + unsigned char force_lock;
11018 +static int si_nfsd_read_lock(struct super_block *sb,
11019 + struct au_nfsd_si_lock *nsi_lock)
11022 + aufs_bindex_t bindex;
11024 + si_read_lock(sb, AuLock_FLUSH);
11026 + /* branch id may be wrapped around */
11028 + bindex = au_br_index(sb, nsi_lock->br_id);
11029 + if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
11030 + goto out; /* success */
11034 + if (!nsi_lock->force_lock)
11035 + si_read_unlock(sb);
11038 + nsi_lock->bindex = bindex;
11042 +struct find_name_by_ino {
11043 + struct dir_context ctx;
11044 + int called, found;
11051 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
11052 + loff_t offset, u64 ino, unsigned int d_type)
11054 + struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
11058 + if (a->ino != ino)
11061 + memcpy(a->name, name, namelen);
11062 + a->namelen = namelen;
11067 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
11068 + struct au_nfsd_si_lock *nsi_lock)
11070 + struct dentry *dentry, *parent;
11071 + struct file *file;
11072 + struct inode *dir;
11073 + struct find_name_by_ino arg = {
11075 + .actor = find_name_by_ino
11080 + parent = path->dentry;
11082 + si_read_unlock(parent->d_sb);
11083 + file = vfsub_dentry_open(path, au_dir_roflags);
11084 + dentry = (void *)file;
11085 + if (IS_ERR(file))
11088 + dentry = ERR_PTR(-ENOMEM);
11089 + arg.name = (void *)__get_free_page(GFP_NOFS);
11090 + if (unlikely(!arg.name))
11097 + err = vfsub_iterate_dir(file, &arg.ctx);
11098 + } while (!err && !arg.found && arg.called);
11099 + dentry = ERR_PTR(err);
11100 + if (unlikely(err))
11102 + /* instead of ENOENT */
11103 + dentry = ERR_PTR(-ESTALE);
11107 + /* do not call vfsub_lkup_one() */
11108 + dir = d_inode(parent);
11109 + mutex_lock(&dir->i_mutex);
11110 + dentry = vfsub_lookup_one_len(arg.name, parent, arg.namelen);
11111 + mutex_unlock(&dir->i_mutex);
11112 + AuTraceErrPtr(dentry);
11113 + if (IS_ERR(dentry))
11115 + AuDebugOn(au_test_anon(dentry));
11116 + if (unlikely(d_really_is_negative(dentry))) {
11118 + dentry = ERR_PTR(-ENOENT);
11122 + free_page((unsigned long)arg.name);
11126 + if (unlikely(nsi_lock
11127 + && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
11128 + if (!IS_ERR(dentry)) {
11130 + dentry = ERR_PTR(-ESTALE);
11132 + AuTraceErrPtr(dentry);
11136 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
11138 + struct au_nfsd_si_lock *nsi_lock)
11140 + struct dentry *dentry;
11141 + struct path path;
11143 + if (dir_ino != AUFS_ROOT_INO) {
11144 + path.dentry = decode_by_ino(sb, dir_ino, 0);
11145 + dentry = path.dentry;
11146 + if (!path.dentry || IS_ERR(path.dentry))
11148 + AuDebugOn(au_test_anon(path.dentry));
11150 + path.dentry = dget(sb->s_root);
11152 + path.mnt = au_mnt_get(sb);
11153 + dentry = au_lkup_by_ino(&path, ino, nsi_lock);
11157 + AuTraceErrPtr(dentry);
11161 +/* ---------------------------------------------------------------------- */
11163 +static int h_acceptable(void *expv, struct dentry *dentry)
11168 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
11169 + char *buf, int len, struct super_block *sb)
11173 + struct path path;
11175 + p = d_path(h_rootpath, buf, len);
11180 + path.mnt = h_rootpath->mnt;
11181 + path.dentry = h_parent;
11182 + p = d_path(&path, buf, len);
11188 + path.mnt = au_mnt_get(sb);
11189 + path.dentry = sb->s_root;
11190 + p = d_path(&path, buf, len - strlen(p));
11191 + mntput(path.mnt);
11195 + p[strlen(p)] = '/';
11198 + AuTraceErrPtr(p);
11203 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
11204 + int fh_len, struct au_nfsd_si_lock *nsi_lock)
11206 + struct dentry *dentry, *h_parent, *root;
11207 + struct super_block *h_sb;
11208 + char *pathname, *p;
11209 + struct vfsmount *h_mnt;
11210 + struct au_branch *br;
11212 + struct path path;
11214 + br = au_sbr(sb, nsi_lock->bindex);
11215 + h_mnt = au_br_mnt(br);
11216 + h_sb = h_mnt->mnt_sb;
11217 + /* todo: call lower fh_to_dentry()? fh_to_parent()? */
11218 + h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
11219 + fh_len - Fh_tail, fh[Fh_h_type],
11220 + h_acceptable, /*context*/NULL);
11221 + dentry = h_parent;
11222 + if (unlikely(!h_parent || IS_ERR(h_parent))) {
11223 + AuWarn1("%s decode_fh failed, %ld\n",
11224 + au_sbtype(h_sb), PTR_ERR(h_parent));
11228 + if (unlikely(au_test_anon(h_parent))) {
11229 + AuWarn1("%s decode_fh returned a disconnected dentry\n",
11230 + au_sbtype(h_sb));
11231 + goto out_h_parent;
11234 + dentry = ERR_PTR(-ENOMEM);
11235 + pathname = (void *)__get_free_page(GFP_NOFS);
11236 + if (unlikely(!pathname))
11237 + goto out_h_parent;
11239 + root = sb->s_root;
11240 + path.mnt = h_mnt;
11241 + di_read_lock_parent(root, !AuLock_IR);
11242 + path.dentry = au_h_dptr(root, nsi_lock->bindex);
11243 + di_read_unlock(root, !AuLock_IR);
11244 + p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
11245 + dentry = (void *)p;
11247 + goto out_pathname;
11249 + si_read_unlock(sb);
11250 + err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
11251 + dentry = ERR_PTR(err);
11252 + if (unlikely(err))
11255 + dentry = ERR_PTR(-ENOENT);
11256 + AuDebugOn(au_test_anon(path.dentry));
11257 + if (unlikely(d_really_is_negative(path.dentry)))
11260 + if (ino != d_inode(path.dentry)->i_ino)
11261 + dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
11263 + dentry = dget(path.dentry);
11268 + if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
11269 + if (!IS_ERR(dentry)) {
11271 + dentry = ERR_PTR(-ESTALE);
11274 + free_page((unsigned long)pathname);
11278 + AuTraceErrPtr(dentry);
11282 +/* ---------------------------------------------------------------------- */
11284 +static struct dentry *
11285 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
11288 + struct dentry *dentry;
11289 + __u32 *fh = fid->raw;
11290 + struct au_branch *br;
11291 + ino_t ino, dir_ino;
11292 + struct au_nfsd_si_lock nsi_lock = {
11296 + dentry = ERR_PTR(-ESTALE);
11297 + /* it should never happen, but the file handle is unreliable */
11298 + if (unlikely(fh_len < Fh_tail))
11300 + nsi_lock.sigen = fh[Fh_sigen];
11301 + nsi_lock.br_id = fh[Fh_br_id];
11303 + /* branch id may be wrapped around */
11305 + if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
11307 + nsi_lock.force_lock = 1;
11309 + /* is this inode still cached? */
11310 + ino = decode_ino(fh + Fh_ino);
11311 + /* it should never happen */
11312 + if (unlikely(ino == AUFS_ROOT_INO))
11315 + dir_ino = decode_ino(fh + Fh_dir_ino);
11316 + dentry = decode_by_ino(sb, ino, dir_ino);
11317 + if (IS_ERR(dentry))
11322 + /* is the parent dir cached? */
11323 + br = au_sbr(sb, nsi_lock.bindex);
11324 + atomic_inc(&br->br_count);
11325 + dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
11326 + if (IS_ERR(dentry))
11331 + /* lookup path */
11332 + dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
11333 + if (IS_ERR(dentry))
11335 + if (unlikely(!dentry))
11336 + /* todo?: make it ESTALE */
11340 + if (!au_digen_test(dentry, au_sigen(sb))
11341 + && d_inode(dentry)->i_generation == fh[Fh_igen])
11342 + goto out_unlock; /* success */
11345 + dentry = ERR_PTR(-ESTALE);
11348 + atomic_dec(&br->br_count);
11349 + si_read_unlock(sb);
11351 + AuTraceErrPtr(dentry);
11355 +#if 0 /* reserved for future use */
11356 +/* support subtreecheck option */
11357 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
11358 + int fh_len, int fh_type)
11360 + struct dentry *parent;
11361 + __u32 *fh = fid->raw;
11364 + dir_ino = decode_ino(fh + Fh_dir_ino);
11365 + parent = decode_by_ino(sb, dir_ino, 0);
11366 + if (IS_ERR(parent))
11369 + parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
11370 + dir_ino, fh, fh_len);
11373 + AuTraceErrPtr(parent);
11378 +/* ---------------------------------------------------------------------- */
11380 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
11381 + struct inode *dir)
11384 + aufs_bindex_t bindex;
11385 + struct super_block *sb, *h_sb;
11386 + struct dentry *dentry, *parent, *h_parent;
11387 + struct inode *h_dir;
11388 + struct au_branch *br;
11391 + if (unlikely(*max_len <= Fh_tail)) {
11392 + AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
11396 + err = FILEID_ROOT;
11397 + if (inode->i_ino == AUFS_ROOT_INO) {
11398 + AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
11403 + sb = inode->i_sb;
11404 + err = si_read_lock(sb, AuLock_FLUSH);
11405 + if (unlikely(err))
11408 +#ifdef CONFIG_AUFS_DEBUG
11409 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
11410 + AuWarn1("NFS-exporting requires xino\n");
11414 + ii_read_lock_child(inode);
11415 + bindex = au_ibstart(inode);
11417 + dentry = d_find_any_alias(inode);
11418 + if (unlikely(!dentry))
11420 + AuDebugOn(au_test_anon(dentry));
11421 + parent = dget_parent(dentry);
11423 + if (unlikely(!parent))
11425 + if (d_really_is_positive(parent))
11426 + dir = d_inode(parent);
11429 + ii_read_lock_parent(dir);
11430 + h_dir = au_h_iptr(dir, bindex);
11431 + ii_read_unlock(dir);
11432 + if (unlikely(!h_dir))
11434 + h_parent = d_find_any_alias(h_dir);
11435 + if (unlikely(!h_parent))
11436 + goto out_hparent;
11439 + br = au_sbr(sb, bindex);
11440 + h_sb = au_br_sb(br);
11441 + if (unlikely(!h_sb->s_export_op)) {
11442 + AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
11443 + goto out_hparent;
11446 + fh[Fh_br_id] = br->br_id;
11447 + fh[Fh_sigen] = au_sigen(sb);
11448 + encode_ino(fh + Fh_ino, inode->i_ino);
11449 + encode_ino(fh + Fh_dir_ino, dir->i_ino);
11450 + fh[Fh_igen] = inode->i_generation;
11452 + *max_len -= Fh_tail;
11453 + fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
11455 + /*connectable or subtreecheck*/0);
11456 + err = fh[Fh_h_type];
11457 + *max_len += Fh_tail;
11458 + /* todo: macros? */
11459 + if (err != FILEID_INVALID)
11462 + AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
11469 + ii_read_unlock(inode);
11470 + si_read_unlock(sb);
11472 + if (unlikely(err < 0))
11473 + err = FILEID_INVALID;
11477 +/* ---------------------------------------------------------------------- */
11479 +static int aufs_commit_metadata(struct inode *inode)
11482 + aufs_bindex_t bindex;
11483 + struct super_block *sb;
11484 + struct inode *h_inode;
11485 + int (*f)(struct inode *inode);
11487 + sb = inode->i_sb;
11488 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
11489 + ii_write_lock_child(inode);
11490 + bindex = au_ibstart(inode);
11491 + AuDebugOn(bindex < 0);
11492 + h_inode = au_h_iptr(inode, bindex);
11494 + f = h_inode->i_sb->s_export_op->commit_metadata;
11496 + err = f(h_inode);
11498 + struct writeback_control wbc = {
11499 + .sync_mode = WB_SYNC_ALL,
11500 + .nr_to_write = 0 /* metadata only */
11503 + err = sync_inode(h_inode, &wbc);
11506 + au_cpup_attr_timesizes(inode);
11507 + ii_write_unlock(inode);
11508 + si_read_unlock(sb);
11512 +/* ---------------------------------------------------------------------- */
11514 +static struct export_operations aufs_export_op = {
11515 + .fh_to_dentry = aufs_fh_to_dentry,
11516 + /* .fh_to_parent = aufs_fh_to_parent, */
11517 + .encode_fh = aufs_encode_fh,
11518 + .commit_metadata = aufs_commit_metadata
11521 +void au_export_init(struct super_block *sb)
11523 + struct au_sbinfo *sbinfo;
11526 + sb->s_export_op = &aufs_export_op;
11527 + sbinfo = au_sbi(sb);
11528 + sbinfo->si_xigen = NULL;
11529 + get_random_bytes(&u, sizeof(u));
11530 + BUILD_BUG_ON(sizeof(u) != sizeof(int));
11531 + atomic_set(&sbinfo->si_xigen_next, u);
11533 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
11534 --- /usr/share/empty/fs/aufs/fhsm.c 1970-01-01 01:00:00.000000000 +0100
11535 +++ linux/fs/aufs/fhsm.c 2016-01-13 20:11:11.669760262 +0100
11538 + * Copyright (C) 2011-2015 Junjiro R. Okajima
11540 + * This program, aufs is free software; you can redistribute it and/or modify
11541 + * it under the terms of the GNU General Public License as published by
11542 + * the Free Software Foundation; either version 2 of the License, or
11543 + * (at your option) any later version.
11545 + * This program is distributed in the hope that it will be useful,
11546 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
11547 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11548 + * GNU General Public License for more details.
11550 + * You should have received a copy of the GNU General Public License
11551 + * along with this program; if not, write to the Free Software
11552 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
11556 + * File-based Hierarchy Storage Management
11559 +#include <linux/anon_inodes.h>
11560 +#include <linux/poll.h>
11561 +#include <linux/seq_file.h>
11562 +#include <linux/statfs.h>
11565 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
11567 + struct au_sbinfo *sbinfo;
11568 + struct au_fhsm *fhsm;
11570 + SiMustAnyLock(sb);
11572 + sbinfo = au_sbi(sb);
11573 + fhsm = &sbinfo->si_fhsm;
11574 + AuDebugOn(!fhsm);
11575 + return fhsm->fhsm_bottom;
11578 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
11580 + struct au_sbinfo *sbinfo;
11581 + struct au_fhsm *fhsm;
11583 + SiMustWriteLock(sb);
11585 + sbinfo = au_sbi(sb);
11586 + fhsm = &sbinfo->si_fhsm;
11587 + AuDebugOn(!fhsm);
11588 + fhsm->fhsm_bottom = bindex;
11591 +/* ---------------------------------------------------------------------- */
11593 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
11595 + struct au_br_fhsm *bf;
11597 + bf = br->br_fhsm;
11598 + MtxMustLock(&bf->bf_lock);
11600 + return !bf->bf_readable
11601 + || time_after(jiffies,
11602 + bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
11605 +/* ---------------------------------------------------------------------- */
11607 +static void au_fhsm_notify(struct super_block *sb, int val)
11609 + struct au_sbinfo *sbinfo;
11610 + struct au_fhsm *fhsm;
11612 + SiMustAnyLock(sb);
11614 + sbinfo = au_sbi(sb);
11615 + fhsm = &sbinfo->si_fhsm;
11616 + if (au_fhsm_pid(fhsm)
11617 + && atomic_read(&fhsm->fhsm_readable) != -1) {
11618 + atomic_set(&fhsm->fhsm_readable, val);
11620 + wake_up(&fhsm->fhsm_wqh);
11624 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
11625 + struct aufs_stfs *rstfs, int do_lock, int do_notify)
11628 + struct au_branch *br;
11629 + struct au_br_fhsm *bf;
11631 + br = au_sbr(sb, bindex);
11632 + AuDebugOn(au_br_rdonly(br));
11633 + bf = br->br_fhsm;
11637 + mutex_lock(&bf->bf_lock);
11639 + MtxMustLock(&bf->bf_lock);
11641 + /* sb->s_root for NFS is unreliable */
11642 + err = au_br_stfs(br, &bf->bf_stfs);
11643 + if (unlikely(err)) {
11644 + AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
11648 + bf->bf_jiffy = jiffies;
11649 + bf->bf_readable = 1;
11651 + au_fhsm_notify(sb, /*val*/1);
11653 + *rstfs = bf->bf_stfs;
11657 + mutex_unlock(&bf->bf_lock);
11658 + au_fhsm_notify(sb, /*val*/1);
11663 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
11666 + struct au_sbinfo *sbinfo;
11667 + struct au_fhsm *fhsm;
11668 + struct au_branch *br;
11669 + struct au_br_fhsm *bf;
11671 + AuDbg("b%d, force %d\n", bindex, force);
11672 + SiMustAnyLock(sb);
11674 + sbinfo = au_sbi(sb);
11675 + fhsm = &sbinfo->si_fhsm;
11676 + if (!au_ftest_si(sbinfo, FHSM)
11677 + || fhsm->fhsm_bottom == bindex)
11680 + br = au_sbr(sb, bindex);
11681 + bf = br->br_fhsm;
11683 + mutex_lock(&bf->bf_lock);
11685 + || au_fhsm_pid(fhsm)
11686 + || au_fhsm_test_jiffy(sbinfo, br))
11687 + err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
11689 + mutex_unlock(&bf->bf_lock);
11692 +void au_fhsm_wrote_all(struct super_block *sb, int force)
11694 + aufs_bindex_t bindex, bend;
11695 + struct au_branch *br;
11697 + /* exclude the bottom */
11698 + bend = au_fhsm_bottom(sb);
11699 + for (bindex = 0; bindex < bend; bindex++) {
11700 + br = au_sbr(sb, bindex);
11701 + if (au_br_fhsm(br->br_perm))
11702 + au_fhsm_wrote(sb, bindex, force);
11706 +/* ---------------------------------------------------------------------- */
11708 +static unsigned int au_fhsm_poll(struct file *file,
11709 + struct poll_table_struct *wait)
11711 + unsigned int mask;
11712 + struct au_sbinfo *sbinfo;
11713 + struct au_fhsm *fhsm;
11716 + sbinfo = file->private_data;
11717 + fhsm = &sbinfo->si_fhsm;
11718 + poll_wait(file, &fhsm->fhsm_wqh, wait);
11719 + if (atomic_read(&fhsm->fhsm_readable))
11720 + mask = POLLIN /* | POLLRDNORM */;
11722 + AuTraceErr((int)mask);
11726 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
11727 + struct aufs_stfs *stfs, __s16 brid)
11731 + err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
11733 + err = __put_user(brid, &stbr->brid);
11734 + if (unlikely(err))
11740 +static ssize_t au_fhsm_do_read(struct super_block *sb,
11741 + struct aufs_stbr __user *stbr, size_t count)
11745 + aufs_bindex_t bindex, bend;
11746 + struct au_branch *br;
11747 + struct au_br_fhsm *bf;
11749 + /* except the bottom branch */
11752 + bend = au_fhsm_bottom(sb);
11753 + for (bindex = 0; !err && bindex < bend; bindex++) {
11754 + br = au_sbr(sb, bindex);
11755 + if (!au_br_fhsm(br->br_perm))
11758 + bf = br->br_fhsm;
11759 + mutex_lock(&bf->bf_lock);
11760 + if (bf->bf_readable) {
11762 + if (count >= sizeof(*stbr))
11763 + err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
11766 + bf->bf_readable = 0;
11767 + count -= sizeof(*stbr);
11771 + mutex_unlock(&bf->bf_lock);
11774 + err = sizeof(*stbr) * nstbr;
11779 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
11784 + aufs_bindex_t nfhsm, bindex, bend;
11785 + struct au_sbinfo *sbinfo;
11786 + struct au_fhsm *fhsm;
11787 + struct au_branch *br;
11788 + struct super_block *sb;
11791 + sbinfo = file->private_data;
11792 + fhsm = &sbinfo->si_fhsm;
11794 + spin_lock_irq(&fhsm->fhsm_wqh.lock);
11795 + if (!atomic_read(&fhsm->fhsm_readable)) {
11796 + if (vfsub_file_flags(file) & O_NONBLOCK)
11799 + err = wait_event_interruptible_locked_irq
11801 + atomic_read(&fhsm->fhsm_readable));
11803 + spin_unlock_irq(&fhsm->fhsm_wqh.lock);
11804 + if (unlikely(err))
11807 + /* sb may already be dead */
11808 + au_rw_read_lock(&sbinfo->si_rwsem);
11809 + readable = atomic_read(&fhsm->fhsm_readable);
11810 + if (readable > 0) {
11811 + sb = sbinfo->si_sb;
11813 + /* exclude the bottom branch */
11815 + bend = au_fhsm_bottom(sb);
11816 + for (bindex = 0; bindex < bend; bindex++) {
11817 + br = au_sbr(sb, bindex);
11818 + if (au_br_fhsm(br->br_perm))
11822 + if (nfhsm * sizeof(struct aufs_stbr) <= count) {
11823 + atomic_set(&fhsm->fhsm_readable, 0);
11824 + err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
11828 + au_rw_read_unlock(&sbinfo->si_rwsem);
11836 +static int au_fhsm_release(struct inode *inode, struct file *file)
11838 + struct au_sbinfo *sbinfo;
11839 + struct au_fhsm *fhsm;
11841 + /* sb may already be dead */
11842 + sbinfo = file->private_data;
11843 + fhsm = &sbinfo->si_fhsm;
11844 + spin_lock(&fhsm->fhsm_spin);
11845 + fhsm->fhsm_pid = 0;
11846 + spin_unlock(&fhsm->fhsm_spin);
11847 + kobject_put(&sbinfo->si_kobj);
11852 +static const struct file_operations au_fhsm_fops = {
11853 + .owner = THIS_MODULE,
11854 + .llseek = noop_llseek,
11855 + .read = au_fhsm_read,
11856 + .poll = au_fhsm_poll,
11857 + .release = au_fhsm_release
11860 +int au_fhsm_fd(struct super_block *sb, int oflags)
11863 + struct au_sbinfo *sbinfo;
11864 + struct au_fhsm *fhsm;
11867 + if (unlikely(!capable(CAP_SYS_ADMIN)))
11871 + if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
11875 + sbinfo = au_sbi(sb);
11876 + fhsm = &sbinfo->si_fhsm;
11877 + spin_lock(&fhsm->fhsm_spin);
11878 + if (!fhsm->fhsm_pid)
11879 + fhsm->fhsm_pid = current->pid;
11882 + spin_unlock(&fhsm->fhsm_spin);
11883 + if (unlikely(err))
11886 + oflags |= O_RDONLY;
11887 + /* oflags |= FMODE_NONOTIFY; */
11888 + fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
11890 + if (unlikely(fd < 0))
11893 + /* succeed reglardless 'fhsm' status */
11894 + kobject_get(&sbinfo->si_kobj);
11895 + si_noflush_read_lock(sb);
11896 + if (au_ftest_si(sbinfo, FHSM))
11897 + au_fhsm_wrote_all(sb, /*force*/0);
11898 + si_read_unlock(sb);
11899 + goto out; /* success */
11902 + spin_lock(&fhsm->fhsm_spin);
11903 + fhsm->fhsm_pid = 0;
11904 + spin_unlock(&fhsm->fhsm_spin);
11910 +/* ---------------------------------------------------------------------- */
11912 +int au_fhsm_br_alloc(struct au_branch *br)
11917 + br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
11919 + au_br_fhsm_init(br->br_fhsm);
11926 +/* ---------------------------------------------------------------------- */
11928 +void au_fhsm_fin(struct super_block *sb)
11930 + au_fhsm_notify(sb, /*val*/-1);
11933 +void au_fhsm_init(struct au_sbinfo *sbinfo)
11935 + struct au_fhsm *fhsm;
11937 + fhsm = &sbinfo->si_fhsm;
11938 + spin_lock_init(&fhsm->fhsm_spin);
11939 + init_waitqueue_head(&fhsm->fhsm_wqh);
11940 + atomic_set(&fhsm->fhsm_readable, 0);
11941 + fhsm->fhsm_expire
11942 + = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
11943 + fhsm->fhsm_bottom = -1;
11946 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
11948 + sbinfo->si_fhsm.fhsm_expire
11949 + = msecs_to_jiffies(sec * MSEC_PER_SEC);
11952 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
11956 + if (!au_ftest_si(sbinfo, FHSM))
11959 + u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
11960 + if (u != AUFS_FHSM_CACHE_DEF_SEC)
11961 + seq_printf(seq, ",fhsm_sec=%u", u);
11963 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
11964 --- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
11965 +++ linux/fs/aufs/file.c 2016-01-13 20:11:11.669760262 +0100
11968 + * Copyright (C) 2005-2015 Junjiro R. Okajima
11970 + * This program, aufs is free software; you can redistribute it and/or modify
11971 + * it under the terms of the GNU General Public License as published by
11972 + * the Free Software Foundation; either version 2 of the License, or
11973 + * (at your option) any later version.
11975 + * This program is distributed in the hope that it will be useful,
11976 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
11977 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11978 + * GNU General Public License for more details.
11980 + * You should have received a copy of the GNU General Public License
11981 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
11985 + * handling file/dir, and address_space operation
11988 +#ifdef CONFIG_AUFS_DEBUG
11989 +#include <linux/migrate.h>
11991 +#include <linux/pagemap.h>
11994 +/* drop flags for writing */
11995 +unsigned int au_file_roflags(unsigned int flags)
11997 + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
11998 + flags |= O_RDONLY | O_NOATIME;
12002 +/* common functions to regular file and dir */
12003 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
12004 + struct file *file, int force_wr)
12006 + struct file *h_file;
12007 + struct dentry *h_dentry;
12008 + struct inode *h_inode;
12009 + struct super_block *sb;
12010 + struct au_branch *br;
12011 + struct path h_path;
12014 + /* a race condition can happen between open and unlink/rmdir */
12015 + h_file = ERR_PTR(-ENOENT);
12016 + h_dentry = au_h_dptr(dentry, bindex);
12017 + if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
12019 + h_inode = d_inode(h_dentry);
12020 + spin_lock(&h_dentry->d_lock);
12021 + err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
12022 + /* || !d_inode(dentry)->i_nlink */
12024 + spin_unlock(&h_dentry->d_lock);
12025 + if (unlikely(err))
12028 + sb = dentry->d_sb;
12029 + br = au_sbr(sb, bindex);
12030 + err = au_br_test_oflag(flags, br);
12031 + h_file = ERR_PTR(err);
12032 + if (unlikely(err))
12035 + /* drop flags for writing */
12036 + if (au_test_ro(sb, bindex, d_inode(dentry))) {
12037 + if (force_wr && !(flags & O_WRONLY))
12039 + flags = au_file_roflags(flags);
12041 + h_file = ERR_PTR(-EROFS);
12042 + flags = au_file_roflags(flags);
12043 + if (unlikely(vfsub_native_ro(h_inode)
12044 + || IS_APPEND(h_inode)))
12046 + flags &= ~O_ACCMODE;
12047 + flags |= O_WRONLY;
12050 + flags &= ~O_CREAT;
12051 + atomic_inc(&br->br_count);
12052 + h_path.dentry = h_dentry;
12053 + h_path.mnt = au_br_mnt(br);
12054 + h_file = vfsub_dentry_open(&h_path, flags);
12055 + if (IS_ERR(h_file))
12058 + if (flags & __FMODE_EXEC) {
12059 + err = deny_write_access(h_file);
12060 + if (unlikely(err)) {
12062 + h_file = ERR_PTR(err);
12066 + fsnotify_open(h_file);
12067 + goto out; /* success */
12070 + atomic_dec(&br->br_count);
12075 +static int au_cmoo(struct dentry *dentry)
12078 + unsigned int udba;
12079 + struct path h_path;
12080 + struct au_pin pin;
12081 + struct au_cp_generic cpg = {
12082 + .dentry = dentry,
12087 + .flags = AuCpup_DTIME | AuCpup_HOPEN
12089 + struct inode *delegated;
12090 + struct super_block *sb;
12091 + struct au_sbinfo *sbinfo;
12092 + struct au_fhsm *fhsm;
12094 + struct au_branch *br;
12095 + struct dentry *parent;
12096 + struct au_hinode *hdir;
12098 + DiMustWriteLock(dentry);
12099 + IiMustWriteLock(d_inode(dentry));
12102 + if (IS_ROOT(dentry))
12104 + cpg.bsrc = au_dbstart(dentry);
12108 + sb = dentry->d_sb;
12109 + sbinfo = au_sbi(sb);
12110 + fhsm = &sbinfo->si_fhsm;
12111 + pid = au_fhsm_pid(fhsm);
12113 + && (current->pid == pid
12114 + || current->real_parent->pid == pid))
12117 + br = au_sbr(sb, cpg.bsrc);
12118 + cmoo = au_br_cmoo(br->br_perm);
12121 + if (!d_is_reg(dentry))
12122 + cmoo &= AuBrAttr_COO_ALL;
12126 + parent = dget_parent(dentry);
12127 + di_write_lock_parent(parent);
12128 + err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
12130 + if (unlikely(err < 0)) {
12131 + err = 0; /* there is no upper writable branch */
12134 + AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
12136 + /* do not respect the coo attrib for the target branch */
12137 + err = au_cpup_dirs(dentry, cpg.bdst);
12138 + if (unlikely(err))
12141 + di_downgrade_lock(parent, AuLock_IR);
12142 + udba = au_opt_udba(sb);
12143 + err = au_pin(&pin, dentry, cpg.bdst, udba,
12144 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12145 + if (unlikely(err))
12148 + err = au_sio_cpup_simple(&cpg);
12150 + if (unlikely(err))
12152 + if (!(cmoo & AuBrWAttr_MOO))
12153 + goto out_parent; /* success */
12155 + err = au_pin(&pin, dentry, cpg.bsrc, udba,
12156 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12157 + if (unlikely(err))
12160 + h_path.mnt = au_br_mnt(br);
12161 + h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
12162 + hdir = au_hi(d_inode(parent), cpg.bsrc);
12163 + delegated = NULL;
12164 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
12166 + /* todo: keep h_dentry or not? */
12167 + if (unlikely(err == -EWOULDBLOCK)) {
12168 + pr_warn("cannot retry for NFSv4 delegation"
12169 + " for an internal unlink\n");
12172 + if (unlikely(err)) {
12173 + pr_err("unlink %pd after coo failed (%d), ignored\n",
12177 + goto out_parent; /* success */
12180 + di_downgrade_lock(parent, AuLock_IR);
12182 + di_read_unlock(parent, AuLock_IR);
12189 +int au_do_open(struct file *file, struct au_do_open_args *args)
12191 + int err, no_lock = args->no_lock;
12192 + struct dentry *dentry;
12193 + struct au_finfo *finfo;
12196 + err = au_finfo_init(file, args->fidir);
12199 + err = au_finfo_init(file, args->fidir);
12202 + if (unlikely(err))
12205 + dentry = file->f_path.dentry;
12206 + AuDebugOn(IS_ERR_OR_NULL(dentry));
12208 + di_write_lock_child(dentry);
12209 + err = au_cmoo(dentry);
12210 + di_downgrade_lock(dentry, AuLock_IR);
12212 + err = args->open(file, vfsub_file_flags(file), NULL);
12213 + di_read_unlock(dentry, AuLock_IR);
12215 + err = au_cmoo(dentry);
12217 + err = args->open(file, vfsub_file_flags(file),
12219 + if (!err && au_fbstart(file) != au_dbstart(dentry))
12221 + * cmoo happens after h_file was opened.
12222 + * need to refresh file later.
12224 + atomic_dec(&au_fi(file)->fi_generation);
12227 + finfo = au_fi(file);
12229 + finfo->fi_file = file;
12230 + au_sphl_add(&finfo->fi_hlist,
12231 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
12234 + fi_write_unlock(file);
12237 + fi_write_unlock(file);
12240 + if (unlikely(err)) {
12241 + finfo->fi_hdir = NULL;
12242 + au_finfo_fin(file);
12249 +int au_reopen_nondir(struct file *file)
12252 + aufs_bindex_t bstart;
12253 + struct dentry *dentry;
12254 + struct file *h_file, *h_file_tmp;
12256 + dentry = file->f_path.dentry;
12257 + bstart = au_dbstart(dentry);
12258 + h_file_tmp = NULL;
12259 + if (au_fbstart(file) == bstart) {
12260 + h_file = au_hf_top(file);
12261 + if (file->f_mode == h_file->f_mode)
12262 + return 0; /* success */
12263 + h_file_tmp = h_file;
12264 + get_file(h_file_tmp);
12265 + au_set_h_fptr(file, bstart, NULL);
12267 + AuDebugOn(au_fi(file)->fi_hdir);
12270 + * file exists on both of rw and ro
12271 + * open --> dbstart and fbstart are both 0
12272 + * prepend a branch as rw, "rw" become ro
12274 + * delete the top branch, "rw" becomes rw again
12275 + * --> dbstart is 1, fbstart is still 0
12276 + * write --> fbstart is 0 but dbstart is 1
12278 + /* AuDebugOn(au_fbstart(file) < bstart); */
12280 + h_file = au_h_open(dentry, bstart, vfsub_file_flags(file) & ~O_TRUNC,
12281 + file, /*force_wr*/0);
12282 + err = PTR_ERR(h_file);
12283 + if (IS_ERR(h_file)) {
12284 + if (h_file_tmp) {
12285 + atomic_inc(&au_sbr(dentry->d_sb, bstart)->br_count);
12286 + au_set_h_fptr(file, bstart, h_file_tmp);
12287 + h_file_tmp = NULL;
12289 + goto out; /* todo: close all? */
12293 + au_set_fbstart(file, bstart);
12294 + au_set_h_fptr(file, bstart, h_file);
12295 + au_update_figen(file);
12296 + /* todo: necessary? */
12297 + /* file->f_ra = h_file->f_ra; */
12301 + fput(h_file_tmp);
12305 +/* ---------------------------------------------------------------------- */
12307 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
12308 + struct dentry *hi_wh)
12311 + aufs_bindex_t bstart;
12312 + struct au_dinfo *dinfo;
12313 + struct dentry *h_dentry;
12314 + struct au_hdentry *hdp;
12316 + dinfo = au_di(file->f_path.dentry);
12317 + AuRwMustWriteLock(&dinfo->di_rwsem);
12319 + bstart = dinfo->di_bstart;
12320 + dinfo->di_bstart = btgt;
12321 + hdp = dinfo->di_hdentry;
12322 + h_dentry = hdp[0 + btgt].hd_dentry;
12323 + hdp[0 + btgt].hd_dentry = hi_wh;
12324 + err = au_reopen_nondir(file);
12325 + hdp[0 + btgt].hd_dentry = h_dentry;
12326 + dinfo->di_bstart = bstart;
12331 +static int au_ready_to_write_wh(struct file *file, loff_t len,
12332 + aufs_bindex_t bcpup, struct au_pin *pin)
12335 + struct inode *inode, *h_inode;
12336 + struct dentry *h_dentry, *hi_wh;
12337 + struct au_cp_generic cpg = {
12338 + .dentry = file->f_path.dentry,
12345 + au_update_dbstart(cpg.dentry);
12346 + inode = d_inode(cpg.dentry);
12348 + if (au_dbstart(cpg.dentry) <= bcpup
12349 + && au_dbend(cpg.dentry) >= bcpup) {
12350 + h_dentry = au_h_dptr(cpg.dentry, bcpup);
12351 + if (h_dentry && d_is_positive(h_dentry))
12352 + h_inode = d_inode(h_dentry);
12354 + hi_wh = au_hi_wh(inode, bcpup);
12355 + if (!hi_wh && !h_inode)
12356 + err = au_sio_cpup_wh(&cpg, file);
12358 + /* already copied-up after unlink */
12359 + err = au_reopen_wh(file, bcpup, hi_wh);
12362 + && (inode->i_nlink > 1
12363 + || (inode->i_state & I_LINKABLE))
12364 + && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
12365 + au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
12371 + * prepare the @file for writing.
12373 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
12376 + aufs_bindex_t dbstart;
12377 + struct dentry *parent;
12378 + struct inode *inode;
12379 + struct super_block *sb;
12380 + struct file *h_file;
12381 + struct au_cp_generic cpg = {
12382 + .dentry = file->f_path.dentry,
12387 + .flags = AuCpup_DTIME
12390 + sb = cpg.dentry->d_sb;
12391 + inode = d_inode(cpg.dentry);
12392 + cpg.bsrc = au_fbstart(file);
12393 + err = au_test_ro(sb, cpg.bsrc, inode);
12394 + if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
12395 + err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
12400 + /* need to cpup or reopen */
12401 + parent = dget_parent(cpg.dentry);
12402 + di_write_lock_parent(parent);
12403 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
12405 + if (unlikely(err < 0))
12409 + if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
12410 + err = au_cpup_dirs(cpg.dentry, cpg.bdst);
12411 + if (unlikely(err))
12415 + err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
12416 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12417 + if (unlikely(err))
12420 + dbstart = au_dbstart(cpg.dentry);
12421 + if (dbstart <= cpg.bdst)
12422 + cpg.bsrc = cpg.bdst;
12424 + if (dbstart <= cpg.bdst /* just reopen */
12425 + || !d_unhashed(cpg.dentry) /* copyup and reopen */
12427 + h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
12428 + if (IS_ERR(h_file))
12429 + err = PTR_ERR(h_file);
12431 + di_downgrade_lock(parent, AuLock_IR);
12432 + if (dbstart > cpg.bdst)
12433 + err = au_sio_cpup_simple(&cpg);
12435 + err = au_reopen_nondir(file);
12436 + au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
12438 + } else { /* copyup as wh and reopen */
12440 + * since writable hfsplus branch is not supported,
12441 + * h_open_pre/post() are unnecessary.
12443 + err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
12444 + di_downgrade_lock(parent, AuLock_IR);
12448 + au_pin_set_parent_lflag(pin, /*lflag*/0);
12449 + goto out_dput; /* success */
12455 + di_downgrade_lock(parent, AuLock_IR);
12457 + di_read_unlock(parent, AuLock_IR);
12464 +/* ---------------------------------------------------------------------- */
12466 +int au_do_flush(struct file *file, fl_owner_t id,
12467 + int (*flush)(struct file *file, fl_owner_t id))
12470 + struct super_block *sb;
12471 + struct inode *inode;
12473 + inode = file_inode(file);
12474 + sb = inode->i_sb;
12475 + si_noflush_read_lock(sb);
12476 + fi_read_lock(file);
12477 + ii_read_lock_child(inode);
12479 + err = flush(file, id);
12480 + au_cpup_attr_timesizes(inode);
12482 + ii_read_unlock(inode);
12483 + fi_read_unlock(file);
12484 + si_read_unlock(sb);
12488 +/* ---------------------------------------------------------------------- */
12490 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
12493 + struct au_pin pin;
12494 + struct au_finfo *finfo;
12495 + struct dentry *parent, *hi_wh;
12496 + struct inode *inode;
12497 + struct super_block *sb;
12498 + struct au_cp_generic cpg = {
12499 + .dentry = file->f_path.dentry,
12504 + .flags = AuCpup_DTIME
12507 + FiMustWriteLock(file);
12510 + finfo = au_fi(file);
12511 + sb = cpg.dentry->d_sb;
12512 + inode = d_inode(cpg.dentry);
12513 + cpg.bdst = au_ibstart(inode);
12514 + if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
12517 + parent = dget_parent(cpg.dentry);
12518 + if (au_test_ro(sb, cpg.bdst, inode)) {
12519 + di_read_lock_parent(parent, !AuLock_IR);
12520 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
12522 + di_read_unlock(parent, !AuLock_IR);
12523 + if (unlikely(err < 0))
12528 + di_read_lock_parent(parent, AuLock_IR);
12529 + hi_wh = au_hi_wh(inode, cpg.bdst);
12530 + if (!S_ISDIR(inode->i_mode)
12531 + && au_opt_test(au_mntflags(sb), PLINK)
12532 + && au_plink_test(inode)
12533 + && !d_unhashed(cpg.dentry)
12534 + && cpg.bdst < au_dbstart(cpg.dentry)) {
12535 + err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
12536 + if (unlikely(err))
12539 + /* always superio. */
12540 + err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
12541 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12543 + err = au_sio_cpup_simple(&cpg);
12546 + } else if (hi_wh) {
12547 + /* already copied-up after unlink */
12548 + err = au_reopen_wh(file, cpg.bdst, hi_wh);
12549 + *need_reopen = 0;
12553 + di_read_unlock(parent, AuLock_IR);
12560 +static void au_do_refresh_dir(struct file *file)
12562 + aufs_bindex_t bindex, bend, new_bindex, brid;
12563 + struct au_hfile *p, tmp, *q;
12564 + struct au_finfo *finfo;
12565 + struct super_block *sb;
12566 + struct au_fidir *fidir;
12568 + FiMustWriteLock(file);
12570 + sb = file->f_path.dentry->d_sb;
12571 + finfo = au_fi(file);
12572 + fidir = finfo->fi_hdir;
12573 + AuDebugOn(!fidir);
12574 + p = fidir->fd_hfile + finfo->fi_btop;
12575 + brid = p->hf_br->br_id;
12576 + bend = fidir->fd_bbot;
12577 + for (bindex = finfo->fi_btop; bindex <= bend; bindex++, p++) {
12581 + new_bindex = au_br_index(sb, p->hf_br->br_id);
12582 + if (new_bindex == bindex)
12584 + if (new_bindex < 0) {
12585 + au_set_h_fptr(file, bindex, NULL);
12589 + /* swap two lower inode, and loop again */
12590 + q = fidir->fd_hfile + new_bindex;
12594 + if (tmp.hf_file) {
12600 + p = fidir->fd_hfile;
12601 + if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
12602 + bend = au_sbend(sb);
12603 + for (finfo->fi_btop = 0; finfo->fi_btop <= bend;
12604 + finfo->fi_btop++, p++)
12605 + if (p->hf_file) {
12606 + if (file_inode(p->hf_file))
12608 + au_hfput(p, file);
12611 + bend = au_br_index(sb, brid);
12612 + for (finfo->fi_btop = 0; finfo->fi_btop < bend;
12613 + finfo->fi_btop++, p++)
12615 + au_hfput(p, file);
12616 + bend = au_sbend(sb);
12619 + p = fidir->fd_hfile + bend;
12620 + for (fidir->fd_bbot = bend; fidir->fd_bbot >= finfo->fi_btop;
12621 + fidir->fd_bbot--, p--)
12622 + if (p->hf_file) {
12623 + if (file_inode(p->hf_file))
12625 + au_hfput(p, file);
12627 + AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
12631 + * after branch manipulating, refresh the file.
12633 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
12635 + int err, need_reopen;
12636 + aufs_bindex_t bend, bindex;
12637 + struct dentry *dentry;
12638 + struct au_finfo *finfo;
12639 + struct au_hfile *hfile;
12641 + dentry = file->f_path.dentry;
12642 + finfo = au_fi(file);
12643 + if (!finfo->fi_hdir) {
12644 + hfile = &finfo->fi_htop;
12645 + AuDebugOn(!hfile->hf_file);
12646 + bindex = au_br_index(dentry->d_sb, hfile->hf_br->br_id);
12647 + AuDebugOn(bindex < 0);
12648 + if (bindex != finfo->fi_btop)
12649 + au_set_fbstart(file, bindex);
12651 + err = au_fidir_realloc(finfo, au_sbend(dentry->d_sb) + 1);
12652 + if (unlikely(err))
12654 + au_do_refresh_dir(file);
12659 + if (!au_test_mmapped(file))
12660 + err = au_file_refresh_by_inode(file, &need_reopen);
12661 + if (!err && need_reopen && !d_unlinked(dentry))
12662 + err = reopen(file);
12664 + au_update_figen(file);
12665 + goto out; /* success */
12668 + /* error, close all lower files */
12669 + if (finfo->fi_hdir) {
12670 + bend = au_fbend_dir(file);
12671 + for (bindex = au_fbstart(file); bindex <= bend; bindex++)
12672 + au_set_h_fptr(file, bindex, NULL);
12679 +/* common function to regular file and dir */
12680 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
12684 + unsigned int sigen, figen;
12685 + aufs_bindex_t bstart;
12686 + unsigned char pseudo_link;
12687 + struct dentry *dentry;
12688 + struct inode *inode;
12691 + dentry = file->f_path.dentry;
12692 + inode = d_inode(dentry);
12693 + sigen = au_sigen(dentry->d_sb);
12694 + fi_write_lock(file);
12695 + figen = au_figen(file);
12696 + di_write_lock_child(dentry);
12697 + bstart = au_dbstart(dentry);
12698 + pseudo_link = (bstart != au_ibstart(inode));
12699 + if (sigen == figen && !pseudo_link && au_fbstart(file) == bstart) {
12701 + di_downgrade_lock(dentry, AuLock_IR);
12702 + fi_downgrade_lock(file);
12704 + goto out; /* success */
12707 + AuDbg("sigen %d, figen %d\n", sigen, figen);
12708 + if (au_digen_test(dentry, sigen)) {
12709 + err = au_reval_dpath(dentry, sigen);
12710 + AuDebugOn(!err && au_digen_test(dentry, sigen));
12714 + err = refresh_file(file, reopen);
12717 + di_downgrade_lock(dentry, AuLock_IR);
12718 + fi_downgrade_lock(file);
12721 + di_write_unlock(dentry);
12722 + fi_write_unlock(file);
12729 +/* ---------------------------------------------------------------------- */
12731 +/* cf. aufs_nopage() */
12732 +/* for madvise(2) */
12733 +static int aufs_readpage(struct file *file __maybe_unused, struct page *page)
12735 + unlock_page(page);
12739 +/* it will never be called, but necessary to support O_DIRECT */
12740 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
12742 +{ BUG(); return 0; }
12744 +/* they will never be called. */
12745 +#ifdef CONFIG_AUFS_DEBUG
12746 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
12747 + loff_t pos, unsigned len, unsigned flags,
12748 + struct page **pagep, void **fsdata)
12749 +{ AuUnsupport(); return 0; }
12750 +static int aufs_write_end(struct file *file, struct address_space *mapping,
12751 + loff_t pos, unsigned len, unsigned copied,
12752 + struct page *page, void *fsdata)
12753 +{ AuUnsupport(); return 0; }
12754 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
12755 +{ AuUnsupport(); return 0; }
12757 +static int aufs_set_page_dirty(struct page *page)
12758 +{ AuUnsupport(); return 0; }
12759 +static void aufs_invalidatepage(struct page *page, unsigned int offset,
12760 + unsigned int length)
12761 +{ AuUnsupport(); }
12762 +static int aufs_releasepage(struct page *page, gfp_t gfp)
12763 +{ AuUnsupport(); return 0; }
12764 +#if 0 /* called by memory compaction regardless file */
12765 +static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
12766 + struct page *page, enum migrate_mode mode)
12767 +{ AuUnsupport(); return 0; }
12769 +static int aufs_launder_page(struct page *page)
12770 +{ AuUnsupport(); return 0; }
12771 +static int aufs_is_partially_uptodate(struct page *page,
12772 + unsigned long from,
12773 + unsigned long count)
12774 +{ AuUnsupport(); return 0; }
12775 +static void aufs_is_dirty_writeback(struct page *page, bool *dirty,
12777 +{ AuUnsupport(); }
12778 +static int aufs_error_remove_page(struct address_space *mapping,
12779 + struct page *page)
12780 +{ AuUnsupport(); return 0; }
12781 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
12783 +{ AuUnsupport(); return 0; }
12784 +static void aufs_swap_deactivate(struct file *file)
12785 +{ AuUnsupport(); }
12786 +#endif /* CONFIG_AUFS_DEBUG */
12788 +const struct address_space_operations aufs_aop = {
12789 + .readpage = aufs_readpage,
12790 + .direct_IO = aufs_direct_IO,
12791 +#ifdef CONFIG_AUFS_DEBUG
12792 + .writepage = aufs_writepage,
12793 + /* no writepages, because of writepage */
12794 + .set_page_dirty = aufs_set_page_dirty,
12795 + /* no readpages, because of readpage */
12796 + .write_begin = aufs_write_begin,
12797 + .write_end = aufs_write_end,
12798 + /* no bmap, no block device */
12799 + .invalidatepage = aufs_invalidatepage,
12800 + .releasepage = aufs_releasepage,
12801 + /* is fallback_migrate_page ok? */
12802 + /* .migratepage = aufs_migratepage, */
12803 + .launder_page = aufs_launder_page,
12804 + .is_partially_uptodate = aufs_is_partially_uptodate,
12805 + .is_dirty_writeback = aufs_is_dirty_writeback,
12806 + .error_remove_page = aufs_error_remove_page,
12807 + .swap_activate = aufs_swap_activate,
12808 + .swap_deactivate = aufs_swap_deactivate
12809 +#endif /* CONFIG_AUFS_DEBUG */
12811 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
12812 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
12813 +++ linux/fs/aufs/file.h 2016-01-13 20:11:11.669760262 +0100
12816 + * Copyright (C) 2005-2015 Junjiro R. Okajima
12818 + * This program, aufs is free software; you can redistribute it and/or modify
12819 + * it under the terms of the GNU General Public License as published by
12820 + * the Free Software Foundation; either version 2 of the License, or
12821 + * (at your option) any later version.
12823 + * This program is distributed in the hope that it will be useful,
12824 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12825 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12826 + * GNU General Public License for more details.
12828 + * You should have received a copy of the GNU General Public License
12829 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12833 + * file operations
12836 +#ifndef __AUFS_FILE_H__
12837 +#define __AUFS_FILE_H__
12841 +#include <linux/file.h>
12842 +#include <linux/fs.h>
12843 +#include <linux/poll.h>
12844 +#include "rwsem.h"
12848 + struct file *hf_file;
12849 + struct au_branch *hf_br;
12854 + aufs_bindex_t fd_bbot;
12855 + aufs_bindex_t fd_nent;
12856 + struct au_vdir *fd_vdir_cache;
12857 + struct au_hfile fd_hfile[];
12860 +static inline int au_fidir_sz(int nent)
12862 + AuDebugOn(nent < 0);
12863 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
12867 + atomic_t fi_generation;
12869 + struct au_rwsem fi_rwsem;
12870 + aufs_bindex_t fi_btop;
12872 + /* do not union them */
12873 + struct { /* for non-dir */
12874 + struct au_hfile fi_htop;
12875 + atomic_t fi_mmapped;
12877 + struct au_fidir *fi_hdir; /* for dir only */
12879 + struct hlist_node fi_hlist;
12880 + struct file *fi_file; /* very ugly */
12881 +} ____cacheline_aligned_in_smp;
12883 +/* ---------------------------------------------------------------------- */
12886 +extern const struct address_space_operations aufs_aop;
12887 +unsigned int au_file_roflags(unsigned int flags);
12888 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
12889 + struct file *file, int force_wr);
12890 +struct au_do_open_args {
12892 + int (*open)(struct file *file, int flags,
12893 + struct file *h_file);
12894 + struct au_fidir *fidir;
12895 + struct file *h_file;
12897 +int au_do_open(struct file *file, struct au_do_open_args *args);
12898 +int au_reopen_nondir(struct file *file);
12900 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
12901 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
12903 +int au_do_flush(struct file *file, fl_owner_t id,
12904 + int (*flush)(struct file *file, fl_owner_t id));
12907 +#ifdef CONFIG_AUFS_POLL
12908 +unsigned int aufs_poll(struct file *file, poll_table *wait);
12911 +#ifdef CONFIG_AUFS_BR_HFSPLUS
12913 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
12915 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
12916 + struct file *h_file);
12918 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
12919 + aufs_bindex_t bindex, int force_wr)
12920 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
12921 + struct file *h_file);
12925 +extern const struct file_operations aufs_file_fop;
12926 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
12927 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
12928 +struct file *au_read_pre(struct file *file, int keep_fi);
12931 +void au_hfput(struct au_hfile *hf, struct file *file);
12932 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
12933 + struct file *h_file);
12935 +void au_update_figen(struct file *file);
12936 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
12937 +int au_fidir_realloc(struct au_finfo *finfo, int nbr);
12939 +void au_fi_init_once(void *_fi);
12940 +void au_finfo_fin(struct file *file);
12941 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
12944 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
12945 +#ifdef CONFIG_COMPAT
12946 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
12947 + unsigned long arg);
12948 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
12949 + unsigned long arg);
12952 +/* ---------------------------------------------------------------------- */
12954 +static inline struct au_finfo *au_fi(struct file *file)
12956 + return file->private_data;
12959 +/* ---------------------------------------------------------------------- */
12962 + * fi_read_lock, fi_write_lock,
12963 + * fi_read_unlock, fi_write_unlock, fi_downgrade_lock
12965 +AuSimpleRwsemFuncs(fi, struct file *f, &au_fi(f)->fi_rwsem);
12967 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
12968 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
12969 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
12971 +/* ---------------------------------------------------------------------- */
12973 +/* todo: hard/soft set? */
12974 +static inline aufs_bindex_t au_fbstart(struct file *file)
12976 + FiMustAnyLock(file);
12977 + return au_fi(file)->fi_btop;
12980 +static inline aufs_bindex_t au_fbend_dir(struct file *file)
12982 + FiMustAnyLock(file);
12983 + AuDebugOn(!au_fi(file)->fi_hdir);
12984 + return au_fi(file)->fi_hdir->fd_bbot;
12987 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
12989 + FiMustAnyLock(file);
12990 + AuDebugOn(!au_fi(file)->fi_hdir);
12991 + return au_fi(file)->fi_hdir->fd_vdir_cache;
12994 +static inline void au_set_fbstart(struct file *file, aufs_bindex_t bindex)
12996 + FiMustWriteLock(file);
12997 + au_fi(file)->fi_btop = bindex;
13000 +static inline void au_set_fbend_dir(struct file *file, aufs_bindex_t bindex)
13002 + FiMustWriteLock(file);
13003 + AuDebugOn(!au_fi(file)->fi_hdir);
13004 + au_fi(file)->fi_hdir->fd_bbot = bindex;
13007 +static inline void au_set_fvdir_cache(struct file *file,
13008 + struct au_vdir *vdir_cache)
13010 + FiMustWriteLock(file);
13011 + AuDebugOn(!au_fi(file)->fi_hdir);
13012 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
13015 +static inline struct file *au_hf_top(struct file *file)
13017 + FiMustAnyLock(file);
13018 + AuDebugOn(au_fi(file)->fi_hdir);
13019 + return au_fi(file)->fi_htop.hf_file;
13022 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
13024 + FiMustAnyLock(file);
13025 + AuDebugOn(!au_fi(file)->fi_hdir);
13026 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
13029 +/* todo: memory barrier? */
13030 +static inline unsigned int au_figen(struct file *f)
13032 + return atomic_read(&au_fi(f)->fi_generation);
13035 +static inline void au_set_mmapped(struct file *f)
13037 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
13039 + pr_warn("fi_mmapped wrapped around\n");
13040 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
13044 +static inline void au_unset_mmapped(struct file *f)
13046 + atomic_dec(&au_fi(f)->fi_mmapped);
13049 +static inline int au_test_mmapped(struct file *f)
13051 + return atomic_read(&au_fi(f)->fi_mmapped);
13054 +/* customize vma->vm_file */
13056 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
13057 + struct file *file)
13061 + f = vma->vm_file;
13063 + vma->vm_file = file;
13068 +#define AuDbgVmRegion(file, vma) do {} while (0)
13070 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
13071 + struct file *file)
13073 + au_do_vm_file_reset(vma, file);
13076 +#define AuDbgVmRegion(file, vma) \
13077 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
13079 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
13080 + struct file *file)
13084 + au_do_vm_file_reset(vma, file);
13085 + f = vma->vm_region->vm_file;
13087 + vma->vm_region->vm_file = file;
13090 +#endif /* CONFIG_MMU */
13092 +/* handle vma->vm_prfile */
13093 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
13094 + struct file *file)
13097 + vma->vm_prfile = file;
13098 +#ifndef CONFIG_MMU
13100 + vma->vm_region->vm_prfile = file;
13104 +#endif /* __KERNEL__ */
13105 +#endif /* __AUFS_FILE_H__ */
13106 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
13107 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
13108 +++ linux/fs/aufs/finfo.c 2016-01-13 20:11:11.669760262 +0100
13111 + * Copyright (C) 2005-2015 Junjiro R. Okajima
13113 + * This program, aufs is free software; you can redistribute it and/or modify
13114 + * it under the terms of the GNU General Public License as published by
13115 + * the Free Software Foundation; either version 2 of the License, or
13116 + * (at your option) any later version.
13118 + * This program is distributed in the hope that it will be useful,
13119 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13120 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13121 + * GNU General Public License for more details.
13123 + * You should have received a copy of the GNU General Public License
13124 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
13128 + * file private data
13133 +void au_hfput(struct au_hfile *hf, struct file *file)
13135 + /* todo: direct access f_flags */
13136 + if (vfsub_file_flags(file) & __FMODE_EXEC)
13137 + allow_write_access(hf->hf_file);
13138 + fput(hf->hf_file);
13139 + hf->hf_file = NULL;
13140 + atomic_dec(&hf->hf_br->br_count);
13141 + hf->hf_br = NULL;
13144 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
13146 + struct au_finfo *finfo = au_fi(file);
13147 + struct au_hfile *hf;
13148 + struct au_fidir *fidir;
13150 + fidir = finfo->fi_hdir;
13152 + AuDebugOn(finfo->fi_btop != bindex);
13153 + hf = &finfo->fi_htop;
13155 + hf = fidir->fd_hfile + bindex;
13157 + if (hf && hf->hf_file)
13158 + au_hfput(hf, file);
13160 + FiMustWriteLock(file);
13161 + AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
13162 + hf->hf_file = val;
13163 + hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
13167 +void au_update_figen(struct file *file)
13169 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
13170 + /* smp_mb(); */ /* atomic_set */
13173 +/* ---------------------------------------------------------------------- */
13175 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
13177 + struct au_fidir *fidir;
13180 + nbr = au_sbend(sb) + 1;
13182 + nbr = 2; /* initial allocate for 2 branches */
13183 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
13185 + fidir->fd_bbot = -1;
13186 + fidir->fd_nent = nbr;
13192 +int au_fidir_realloc(struct au_finfo *finfo, int nbr)
13195 + struct au_fidir *fidir, *p;
13197 + AuRwMustWriteLock(&finfo->fi_rwsem);
13198 + fidir = finfo->fi_hdir;
13199 + AuDebugOn(!fidir);
13202 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
13205 + p->fd_nent = nbr;
13206 + finfo->fi_hdir = p;
13213 +/* ---------------------------------------------------------------------- */
13215 +void au_finfo_fin(struct file *file)
13217 + struct au_finfo *finfo;
13219 + au_nfiles_dec(file->f_path.dentry->d_sb);
13221 + finfo = au_fi(file);
13222 + AuDebugOn(finfo->fi_hdir);
13223 + AuRwDestroy(&finfo->fi_rwsem);
13224 + au_cache_free_finfo(finfo);
13227 +void au_fi_init_once(void *_finfo)
13229 + struct au_finfo *finfo = _finfo;
13230 + static struct lock_class_key aufs_fi;
13232 + au_rw_init(&finfo->fi_rwsem);
13233 + au_rw_class(&finfo->fi_rwsem, &aufs_fi);
13236 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
13239 + struct au_finfo *finfo;
13240 + struct dentry *dentry;
13243 + dentry = file->f_path.dentry;
13244 + finfo = au_cache_alloc_finfo();
13245 + if (unlikely(!finfo))
13249 + au_nfiles_inc(dentry->d_sb);
13250 + /* verbose coding for lock class name */
13252 + au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcNonDir_FIINFO);
13254 + au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcDir_FIINFO);
13255 + au_rw_write_lock(&finfo->fi_rwsem);
13256 + finfo->fi_btop = -1;
13257 + finfo->fi_hdir = fidir;
13258 + atomic_set(&finfo->fi_generation, au_digen(dentry));
13259 + /* smp_mb(); */ /* atomic_set */
13261 + file->private_data = finfo;
13266 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
13267 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
13268 +++ linux/fs/aufs/f_op.c 2016-01-13 20:11:11.669760262 +0100
13271 + * Copyright (C) 2005-2015 Junjiro R. Okajima
13273 + * This program, aufs is free software; you can redistribute it and/or modify
13274 + * it under the terms of the GNU General Public License as published by
13275 + * the Free Software Foundation; either version 2 of the License, or
13276 + * (at your option) any later version.
13278 + * This program is distributed in the hope that it will be useful,
13279 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13280 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13281 + * GNU General Public License for more details.
13283 + * You should have received a copy of the GNU General Public License
13284 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
13288 + * file and vm operations
13291 +#include <linux/aio.h>
13292 +#include <linux/fs_stack.h>
13293 +#include <linux/mman.h>
13294 +#include <linux/security.h>
13297 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
13300 + aufs_bindex_t bindex;
13301 + struct dentry *dentry;
13302 + struct au_finfo *finfo;
13303 + struct inode *h_inode;
13305 + FiMustWriteLock(file);
13308 + dentry = file->f_path.dentry;
13309 + AuDebugOn(IS_ERR_OR_NULL(dentry));
13310 + finfo = au_fi(file);
13311 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
13312 + atomic_set(&finfo->fi_mmapped, 0);
13313 + bindex = au_dbstart(dentry);
13315 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
13317 + get_file(h_file);
13318 + if (IS_ERR(h_file))
13319 + err = PTR_ERR(h_file);
13321 + if ((flags & __O_TMPFILE)
13322 + && !(flags & O_EXCL)) {
13323 + h_inode = file_inode(h_file);
13324 + spin_lock(&h_inode->i_lock);
13325 + h_inode->i_state |= I_LINKABLE;
13326 + spin_unlock(&h_inode->i_lock);
13328 + au_set_fbstart(file, bindex);
13329 + au_set_h_fptr(file, bindex, h_file);
13330 + au_update_figen(file);
13331 + /* todo: necessary? */
13332 + /* file->f_ra = h_file->f_ra; */
13338 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
13339 + struct file *file)
13342 + struct super_block *sb;
13343 + struct au_do_open_args args = {
13344 + .open = au_do_open_nondir
13347 + AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
13348 + file, vfsub_file_flags(file), file->f_mode);
13350 + sb = file->f_path.dentry->d_sb;
13351 + si_read_lock(sb, AuLock_FLUSH);
13352 + err = au_do_open(file, &args);
13353 + si_read_unlock(sb);
13357 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
13359 + struct au_finfo *finfo;
13360 + aufs_bindex_t bindex;
13362 + finfo = au_fi(file);
13363 + au_sphl_del(&finfo->fi_hlist,
13364 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
13365 + bindex = finfo->fi_btop;
13367 + au_set_h_fptr(file, bindex, NULL);
13369 + au_finfo_fin(file);
13373 +/* ---------------------------------------------------------------------- */
13375 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
13378 + struct file *h_file;
13381 + h_file = au_hf_top(file);
13383 + err = vfsub_flush(h_file, id);
13387 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
13389 + return au_do_flush(file, id, au_do_flush_nondir);
13392 +/* ---------------------------------------------------------------------- */
13394 + * read and write functions acquire [fdi]_rwsem once, but release before
13395 + * mmap_sem. This is because to stop a race condition between mmap(2).
13396 + * Releasing these aufs-rwsem should be safe, no branch-mamagement (by keeping
13397 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
13398 + * read functions after [fdi]_rwsem are released, but it should be harmless.
13401 +/* Callers should call au_read_post() or fput() in the end */
13402 +struct file *au_read_pre(struct file *file, int keep_fi)
13404 + struct file *h_file;
13407 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
13409 + di_read_unlock(file->f_path.dentry, AuLock_IR);
13410 + h_file = au_hf_top(file);
13411 + get_file(h_file);
13413 + fi_read_unlock(file);
13415 + h_file = ERR_PTR(err);
13420 +static void au_read_post(struct inode *inode, struct file *h_file)
13422 + /* update without lock, I don't think it a problem */
13423 + fsstack_copy_attr_atime(inode, file_inode(h_file));
13427 +struct au_write_pre {
13429 + aufs_bindex_t bstart;
13433 + * return with iinfo is write-locked
13434 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
13437 +static struct file *au_write_pre(struct file *file, int do_ready,
13438 + struct au_write_pre *wpre)
13440 + struct file *h_file;
13441 + struct dentry *dentry;
13443 + struct au_pin pin;
13445 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
13446 + h_file = ERR_PTR(err);
13447 + if (unlikely(err))
13450 + dentry = file->f_path.dentry;
13452 + err = au_ready_to_write(file, -1, &pin);
13453 + if (unlikely(err)) {
13454 + h_file = ERR_PTR(err);
13455 + di_write_unlock(dentry);
13460 + di_downgrade_lock(dentry, /*flags*/0);
13462 + wpre->bstart = au_fbstart(file);
13463 + h_file = au_hf_top(file);
13464 + get_file(h_file);
13466 + wpre->blks = file_inode(h_file)->i_blocks;
13469 + di_read_unlock(dentry, /*flags*/0);
13472 + fi_write_unlock(file);
13477 +static void au_write_post(struct inode *inode, struct file *h_file,
13478 + struct au_write_pre *wpre, ssize_t written)
13480 + struct inode *h_inode;
13482 + au_cpup_attr_timesizes(inode);
13483 + AuDebugOn(au_ibstart(inode) != wpre->bstart);
13484 + h_inode = file_inode(h_file);
13485 + inode->i_mode = h_inode->i_mode;
13486 + ii_write_unlock(inode);
13489 + /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
13491 + au_fhsm_wrote(inode->i_sb, wpre->bstart,
13492 + /*force*/h_inode->i_blocks > wpre->blks);
13495 +static ssize_t aufs_read(struct file *file, char __user *buf, size_t count,
13499 + struct inode *inode;
13500 + struct file *h_file;
13501 + struct super_block *sb;
13503 + inode = file_inode(file);
13504 + sb = inode->i_sb;
13505 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13507 + h_file = au_read_pre(file, /*keep_fi*/0);
13508 + err = PTR_ERR(h_file);
13509 + if (IS_ERR(h_file))
13512 + /* filedata may be obsoleted by concurrent copyup, but no problem */
13513 + err = vfsub_read_u(h_file, buf, count, ppos);
13514 + /* todo: necessary? */
13515 + /* file->f_ra = h_file->f_ra; */
13516 + au_read_post(inode, h_file);
13519 + si_read_unlock(sb);
13524 + * todo: very ugly
13525 + * it locks both of i_mutex and si_rwsem for read in safe.
13526 + * if the plink maintenance mode continues forever (that is the problem),
13527 + * may loop forever.
13529 +static void au_mtx_and_read_lock(struct inode *inode)
13532 + struct super_block *sb = inode->i_sb;
13535 + mutex_lock(&inode->i_mutex);
13536 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
13539 + mutex_unlock(&inode->i_mutex);
13540 + si_read_lock(sb, AuLock_NOPLMW);
13541 + si_read_unlock(sb);
13545 +static ssize_t aufs_write(struct file *file, const char __user *ubuf,
13546 + size_t count, loff_t *ppos)
13549 + struct au_write_pre wpre;
13550 + struct inode *inode;
13551 + struct file *h_file;
13552 + char __user *buf = (char __user *)ubuf;
13554 + inode = file_inode(file);
13555 + au_mtx_and_read_lock(inode);
13557 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13558 + err = PTR_ERR(h_file);
13559 + if (IS_ERR(h_file))
13562 + err = vfsub_write_u(h_file, buf, count, ppos);
13563 + au_write_post(inode, h_file, &wpre, err);
13566 + si_read_unlock(inode->i_sb);
13567 + mutex_unlock(&inode->i_mutex);
13571 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
13572 + struct iov_iter *iov_iter)
13575 + struct file *file;
13576 + ssize_t (*iter)(struct kiocb *, struct iov_iter *);
13578 + err = security_file_permission(h_file, rw);
13579 + if (unlikely(err))
13584 + if (rw == MAY_READ)
13585 + iter = h_file->f_op->read_iter;
13586 + else if (rw == MAY_WRITE)
13587 + iter = h_file->f_op->write_iter;
13589 + file = kio->ki_filp;
13590 + kio->ki_filp = h_file;
13593 + err = iter(kio, iov_iter);
13596 + /* currently there is no such fs */
13598 + kio->ki_filp = file;
13604 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
13607 + struct file *file, *h_file;
13608 + struct inode *inode;
13609 + struct super_block *sb;
13611 + file = kio->ki_filp;
13612 + inode = file_inode(file);
13613 + sb = inode->i_sb;
13614 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13616 + h_file = au_read_pre(file, /*keep_fi*/0);
13617 + err = PTR_ERR(h_file);
13618 + if (IS_ERR(h_file))
13621 + err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
13622 + /* todo: necessary? */
13623 + /* file->f_ra = h_file->f_ra; */
13624 + au_read_post(inode, h_file);
13627 + si_read_unlock(sb);
13631 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
13634 + struct au_write_pre wpre;
13635 + struct inode *inode;
13636 + struct file *file, *h_file;
13638 + file = kio->ki_filp;
13639 + inode = file_inode(file);
13640 + au_mtx_and_read_lock(inode);
13642 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13643 + err = PTR_ERR(h_file);
13644 + if (IS_ERR(h_file))
13647 + err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
13648 + au_write_post(inode, h_file, &wpre, err);
13651 + si_read_unlock(inode->i_sb);
13652 + mutex_unlock(&inode->i_mutex);
13656 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
13657 + struct pipe_inode_info *pipe, size_t len,
13658 + unsigned int flags)
13661 + struct file *h_file;
13662 + struct inode *inode;
13663 + struct super_block *sb;
13665 + inode = file_inode(file);
13666 + sb = inode->i_sb;
13667 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13669 + h_file = au_read_pre(file, /*keep_fi*/1);
13670 + err = PTR_ERR(h_file);
13671 + if (IS_ERR(h_file))
13674 + if (au_test_loopback_kthread()) {
13675 + au_warn_loopback(h_file->f_path.dentry->d_sb);
13676 + if (file->f_mapping != h_file->f_mapping) {
13677 + file->f_mapping = h_file->f_mapping;
13678 + smp_mb(); /* unnecessary? */
13681 + fi_read_unlock(file);
13683 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
13684 + /* todo: necessasry? */
13685 + /* file->f_ra = h_file->f_ra; */
13686 + au_read_post(inode, h_file);
13689 + si_read_unlock(sb);
13694 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
13695 + size_t len, unsigned int flags)
13698 + struct au_write_pre wpre;
13699 + struct inode *inode;
13700 + struct file *h_file;
13702 + inode = file_inode(file);
13703 + au_mtx_and_read_lock(inode);
13705 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13706 + err = PTR_ERR(h_file);
13707 + if (IS_ERR(h_file))
13710 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
13711 + au_write_post(inode, h_file, &wpre, err);
13714 + si_read_unlock(inode->i_sb);
13715 + mutex_unlock(&inode->i_mutex);
13719 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
13723 + struct au_write_pre wpre;
13724 + struct inode *inode;
13725 + struct file *h_file;
13727 + inode = file_inode(file);
13728 + au_mtx_and_read_lock(inode);
13730 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13731 + err = PTR_ERR(h_file);
13732 + if (IS_ERR(h_file))
13736 + err = vfs_fallocate(h_file, mode, offset, len);
13738 + au_write_post(inode, h_file, &wpre, /*written*/1);
13741 + si_read_unlock(inode->i_sb);
13742 + mutex_unlock(&inode->i_mutex);
13746 +/* ---------------------------------------------------------------------- */
13749 + * The locking order around current->mmap_sem.
13750 + * - in most and regular cases
13751 + * file I/O syscall -- aufs_read() or something
13752 + * -- si_rwsem for read -- mmap_sem
13753 + * (Note that [fdi]i_rwsem are released before mmap_sem).
13755 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
13756 + * This AB-BA order is definitly bad, but is not a problem since "si_rwsem for
13757 + * read" allows muliple processes to acquire it and [fdi]i_rwsem are not held in
13758 + * file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
13759 + * It means that when aufs acquires si_rwsem for write, the process should never
13760 + * acquire mmap_sem.
13762 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
13763 + * problem either since any directory is not able to be mmap-ed.
13764 + * The similar scenario is applied to aufs_readlink() too.
13767 +#if 0 /* stop calling security_file_mmap() */
13768 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
13769 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
13771 +static unsigned long au_arch_prot_conv(unsigned long flags)
13773 + /* currently ppc64 only */
13774 +#ifdef CONFIG_PPC64
13775 + /* cf. linux/arch/powerpc/include/asm/mman.h */
13776 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
13777 + return AuConv_VM_PROT(flags, SAO);
13779 + AuDebugOn(arch_calc_vm_prot_bits(-1));
13784 +static unsigned long au_prot_conv(unsigned long flags)
13786 + return AuConv_VM_PROT(flags, READ)
13787 + | AuConv_VM_PROT(flags, WRITE)
13788 + | AuConv_VM_PROT(flags, EXEC)
13789 + | au_arch_prot_conv(flags);
13792 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
13793 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
13795 +static unsigned long au_flag_conv(unsigned long flags)
13797 + return AuConv_VM_MAP(flags, GROWSDOWN)
13798 + | AuConv_VM_MAP(flags, DENYWRITE)
13799 + | AuConv_VM_MAP(flags, LOCKED);
13803 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
13806 + const unsigned char wlock
13807 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
13808 + struct super_block *sb;
13809 + struct file *h_file;
13810 + struct inode *inode;
13812 + AuDbgVmRegion(file, vma);
13814 + inode = file_inode(file);
13815 + sb = inode->i_sb;
13817 + si_read_lock(sb, AuLock_NOPLMW);
13819 + h_file = au_write_pre(file, wlock, /*wpre*/NULL);
13821 + err = PTR_ERR(h_file);
13822 + if (IS_ERR(h_file))
13826 + au_set_mmapped(file);
13827 + au_vm_file_reset(vma, h_file);
13829 + * we cannot call security_mmap_file() here since it may acquire
13830 + * mmap_sem or i_mutex.
13832 + * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
13833 + * au_flag_conv(vma->vm_flags));
13836 + err = h_file->f_op->mmap(h_file, vma);
13838 + au_vm_prfile_set(vma, file);
13839 + fsstack_copy_attr_atime(inode, file_inode(h_file));
13840 + goto out_fput; /* success */
13842 + au_unset_mmapped(file);
13843 + au_vm_file_reset(vma, file);
13847 + ii_write_unlock(inode);
13852 + si_read_unlock(sb);
13858 +/* ---------------------------------------------------------------------- */
13860 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
13864 + struct au_write_pre wpre;
13865 + struct inode *inode;
13866 + struct file *h_file;
13868 + err = 0; /* -EBADF; */ /* posix? */
13869 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
13872 + inode = file_inode(file);
13873 + au_mtx_and_read_lock(inode);
13875 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13876 + err = PTR_ERR(h_file);
13877 + if (IS_ERR(h_file))
13880 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
13881 + au_write_post(inode, h_file, &wpre, /*written*/0);
13884 + si_read_unlock(inode->i_sb);
13885 + mutex_unlock(&inode->i_mutex);
13890 +/* no one supports this operation, currently */
13892 +static int aufs_aio_fsync_nondir(struct kiocb *kio, int datasync)
13895 + struct au_write_pre wpre;
13896 + struct inode *inode;
13897 + struct file *file, *h_file;
13899 + err = 0; /* -EBADF; */ /* posix? */
13900 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
13903 + file = kio->ki_filp;
13904 + inode = file_inode(file);
13905 + au_mtx_and_read_lock(inode);
13907 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13908 + err = PTR_ERR(h_file);
13909 + if (IS_ERR(h_file))
13913 + h_file = au_hf_top(file);
13914 + if (h_file->f_op->aio_fsync) {
13915 + struct mutex *h_mtx;
13917 + h_mtx = &file_inode(h_file)->i_mutex;
13918 + if (!is_sync_kiocb(kio)) {
13919 + get_file(h_file);
13922 + kio->ki_filp = h_file;
13923 + err = h_file->f_op->aio_fsync(kio, datasync);
13924 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
13926 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL);
13928 + mutex_unlock(h_mtx);
13930 + au_write_post(inode, h_file, &wpre, /*written*/0);
13933 + si_read_unlock(inode->sb);
13934 + mutex_unlock(&inode->i_mutex);
13940 +static int aufs_fasync(int fd, struct file *file, int flag)
13943 + struct file *h_file;
13944 + struct super_block *sb;
13946 + sb = file->f_path.dentry->d_sb;
13947 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13949 + h_file = au_read_pre(file, /*keep_fi*/0);
13950 + err = PTR_ERR(h_file);
13951 + if (IS_ERR(h_file))
13954 + if (h_file->f_op->fasync)
13955 + err = h_file->f_op->fasync(fd, h_file, flag);
13956 + fput(h_file); /* instead of au_read_post() */
13959 + si_read_unlock(sb);
13963 +/* ---------------------------------------------------------------------- */
13965 +/* no one supports this operation, currently */
13967 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
13968 + size_t len, loff_t *pos, int more)
13973 +/* ---------------------------------------------------------------------- */
13975 +const struct file_operations aufs_file_fop = {
13976 + .owner = THIS_MODULE,
13978 + .llseek = default_llseek,
13980 + .read = aufs_read,
13981 + .write = aufs_write,
13982 + .read_iter = aufs_read_iter,
13983 + .write_iter = aufs_write_iter,
13985 +#ifdef CONFIG_AUFS_POLL
13986 + .poll = aufs_poll,
13988 + .unlocked_ioctl = aufs_ioctl_nondir,
13989 +#ifdef CONFIG_COMPAT
13990 + .compat_ioctl = aufs_compat_ioctl_nondir,
13992 + .mmap = aufs_mmap,
13993 + .open = aufs_open_nondir,
13994 + .flush = aufs_flush_nondir,
13995 + .release = aufs_release_nondir,
13996 + .fsync = aufs_fsync_nondir,
13997 + /* .aio_fsync = aufs_aio_fsync_nondir, */
13998 + .fasync = aufs_fasync,
13999 + /* .sendpage = aufs_sendpage, */
14000 + .splice_write = aufs_splice_write,
14001 + .splice_read = aufs_splice_read,
14003 + .aio_splice_write = aufs_aio_splice_write,
14004 + .aio_splice_read = aufs_aio_splice_read,
14006 + .fallocate = aufs_fallocate
14008 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
14009 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
14010 +++ linux/fs/aufs/fstype.h 2016-01-13 20:11:11.669760262 +0100
14013 + * Copyright (C) 2005-2015 Junjiro R. Okajima
14015 + * This program, aufs is free software; you can redistribute it and/or modify
14016 + * it under the terms of the GNU General Public License as published by
14017 + * the Free Software Foundation; either version 2 of the License, or
14018 + * (at your option) any later version.
14020 + * This program is distributed in the hope that it will be useful,
14021 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14022 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14023 + * GNU General Public License for more details.
14025 + * You should have received a copy of the GNU General Public License
14026 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14030 + * judging filesystem type
14033 +#ifndef __AUFS_FSTYPE_H__
14034 +#define __AUFS_FSTYPE_H__
14038 +#include <linux/fs.h>
14039 +#include <linux/magic.h>
14040 +#include <linux/nfs_fs.h>
14041 +#include <linux/romfs_fs.h>
14043 +static inline int au_test_aufs(struct super_block *sb)
14045 + return sb->s_magic == AUFS_SUPER_MAGIC;
14048 +static inline const char *au_sbtype(struct super_block *sb)
14050 + return sb->s_type->name;
14053 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
14055 +#if defined(CONFIG_ISO9660_FS) || defined(CONFIG_ISO9660_FS_MODULE)
14056 + return sb->s_magic == ISOFS_SUPER_MAGIC;
14062 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
14064 +#if defined(CONFIG_ROMFS_FS) || defined(CONFIG_ROMFS_FS_MODULE)
14065 + return sb->s_magic == ROMFS_MAGIC;
14071 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
14073 +#if defined(CONFIG_CRAMFS) || defined(CONFIG_CRAMFS_MODULE)
14074 + return sb->s_magic == CRAMFS_MAGIC;
14079 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
14081 +#if defined(CONFIG_NFS_FS) || defined(CONFIG_NFS_FS_MODULE)
14082 + return sb->s_magic == NFS_SUPER_MAGIC;
14088 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
14090 +#if defined(CONFIG_FUSE_FS) || defined(CONFIG_FUSE_FS_MODULE)
14091 + return sb->s_magic == FUSE_SUPER_MAGIC;
14097 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
14099 +#if defined(CONFIG_XFS_FS) || defined(CONFIG_XFS_FS_MODULE)
14100 + return sb->s_magic == XFS_SB_MAGIC;
14106 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
14108 +#ifdef CONFIG_TMPFS
14109 + return sb->s_magic == TMPFS_MAGIC;
14115 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
14117 +#if defined(CONFIG_ECRYPT_FS) || defined(CONFIG_ECRYPT_FS_MODULE)
14118 + return !strcmp(au_sbtype(sb), "ecryptfs");
14124 +static inline int au_test_ramfs(struct super_block *sb)
14126 + return sb->s_magic == RAMFS_MAGIC;
14129 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
14131 +#if defined(CONFIG_UBIFS_FS) || defined(CONFIG_UBIFS_FS_MODULE)
14132 + return sb->s_magic == UBIFS_SUPER_MAGIC;
14138 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
14140 +#ifdef CONFIG_PROC_FS
14141 + return sb->s_magic == PROC_SUPER_MAGIC;
14147 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
14149 +#ifdef CONFIG_SYSFS
14150 + return sb->s_magic == SYSFS_MAGIC;
14156 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
14158 +#if defined(CONFIG_CONFIGFS_FS) || defined(CONFIG_CONFIGFS_FS_MODULE)
14159 + return sb->s_magic == CONFIGFS_MAGIC;
14165 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
14167 +#if defined(CONFIG_MINIX_FS) || defined(CONFIG_MINIX_FS_MODULE)
14168 + return sb->s_magic == MINIX3_SUPER_MAGIC
14169 + || sb->s_magic == MINIX2_SUPER_MAGIC
14170 + || sb->s_magic == MINIX2_SUPER_MAGIC2
14171 + || sb->s_magic == MINIX_SUPER_MAGIC
14172 + || sb->s_magic == MINIX_SUPER_MAGIC2;
14178 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
14180 +#if defined(CONFIG_FAT_FS) || defined(CONFIG_FAT_FS_MODULE)
14181 + return sb->s_magic == MSDOS_SUPER_MAGIC;
14187 +static inline int au_test_msdos(struct super_block *sb)
14189 + return au_test_fat(sb);
14192 +static inline int au_test_vfat(struct super_block *sb)
14194 + return au_test_fat(sb);
14197 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
14199 +#ifdef CONFIG_SECURITYFS
14200 + return sb->s_magic == SECURITYFS_MAGIC;
14206 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
14208 +#if defined(CONFIG_SQUASHFS) || defined(CONFIG_SQUASHFS_MODULE)
14209 + return sb->s_magic == SQUASHFS_MAGIC;
14215 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
14217 +#if defined(CONFIG_BTRFS_FS) || defined(CONFIG_BTRFS_FS_MODULE)
14218 + return sb->s_magic == BTRFS_SUPER_MAGIC;
14224 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
14226 +#if defined(CONFIG_XENFS) || defined(CONFIG_XENFS_MODULE)
14227 + return sb->s_magic == XENFS_SUPER_MAGIC;
14233 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
14235 +#ifdef CONFIG_DEBUG_FS
14236 + return sb->s_magic == DEBUGFS_MAGIC;
14242 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
14244 +#if defined(CONFIG_NILFS) || defined(CONFIG_NILFS_MODULE)
14245 + return sb->s_magic == NILFS_SUPER_MAGIC;
14251 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
14253 +#if defined(CONFIG_HFSPLUS_FS) || defined(CONFIG_HFSPLUS_FS_MODULE)
14254 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
14260 +/* ---------------------------------------------------------------------- */
14262 + * they can't be an aufs branch.
14264 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
14267 +#ifndef CONFIG_AUFS_BR_RAMFS
14268 + au_test_ramfs(sb) ||
14270 + au_test_procfs(sb)
14271 + || au_test_sysfs(sb)
14272 + || au_test_configfs(sb)
14273 + || au_test_debugfs(sb)
14274 + || au_test_securityfs(sb)
14275 + || au_test_xenfs(sb)
14276 + || au_test_ecryptfs(sb)
14277 + /* || !strcmp(au_sbtype(sb), "unionfs") */
14278 + || au_test_aufs(sb); /* will be supported in next version */
14281 +static inline int au_test_fs_remote(struct super_block *sb)
14283 + return !au_test_tmpfs(sb)
14284 +#ifdef CONFIG_AUFS_BR_RAMFS
14285 + && !au_test_ramfs(sb)
14287 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
14290 +/* ---------------------------------------------------------------------- */
14293 + * Note: these functions (below) are created after reading ->getattr() in all
14294 + * filesystems under linux/fs. it means we have to do so in every update...
14298 + * some filesystems require getattr to refresh the inode attributes before
14300 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
14301 + * and leave the work for d_revalidate()
14303 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
14305 + return au_test_nfs(sb)
14306 + || au_test_fuse(sb)
14307 + /* || au_test_btrfs(sb) */ /* untested */
14312 + * filesystems which don't maintain i_size or i_blocks.
14314 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
14316 + return au_test_xfs(sb)
14317 + || au_test_btrfs(sb)
14318 + || au_test_ubifs(sb)
14319 + || au_test_hfsplus(sb) /* maintained, but incorrect */
14320 + /* || au_test_minix(sb) */ /* untested */
14325 + * filesystems which don't store the correct value in some of their inode
14328 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
14330 + return au_test_fs_bad_iattr_size(sb)
14331 + || au_test_fat(sb)
14332 + || au_test_msdos(sb)
14333 + || au_test_vfat(sb);
14336 +/* they don't check i_nlink in link(2) */
14337 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
14339 + return au_test_tmpfs(sb)
14340 +#ifdef CONFIG_AUFS_BR_RAMFS
14341 + || au_test_ramfs(sb)
14343 + || au_test_ubifs(sb)
14344 + || au_test_hfsplus(sb);
14348 + * filesystems which sets S_NOATIME and S_NOCMTIME.
14350 +static inline int au_test_fs_notime(struct super_block *sb)
14352 + return au_test_nfs(sb)
14353 + || au_test_fuse(sb)
14354 + || au_test_ubifs(sb)
14358 +/* temporary support for i#1 in cramfs */
14359 +static inline int au_test_fs_unique_ino(struct inode *inode)
14361 + if (au_test_cramfs(inode->i_sb))
14362 + return inode->i_ino != 1;
14366 +/* ---------------------------------------------------------------------- */
14369 + * the filesystem where the xino files placed must support i/o after unlink and
14370 + * maintain i_size and i_blocks.
14372 +static inline int au_test_fs_bad_xino(struct super_block *sb)
14374 + return au_test_fs_remote(sb)
14375 + || au_test_fs_bad_iattr_size(sb)
14376 + /* don't want unnecessary work for xino */
14377 + || au_test_aufs(sb)
14378 + || au_test_ecryptfs(sb)
14379 + || au_test_nilfs(sb);
14382 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
14384 + return au_test_tmpfs(sb)
14385 + || au_test_ramfs(sb);
14389 + * test if the @sb is real-readonly.
14391 +static inline int au_test_fs_rr(struct super_block *sb)
14393 + return au_test_squashfs(sb)
14394 + || au_test_iso9660(sb)
14395 + || au_test_cramfs(sb)
14396 + || au_test_romfs(sb);
14400 + * test if the @inode is nfs with 'noacl' option
14401 + * NFS always sets MS_POSIXACL regardless its mount option 'noacl.'
14403 +static inline int au_test_nfs_noacl(struct inode *inode)
14405 + return au_test_nfs(inode->i_sb)
14406 + /* && IS_POSIXACL(inode) */
14407 + && !nfs_server_capable(inode, NFS_CAP_ACLS);
14410 +#endif /* __KERNEL__ */
14411 +#endif /* __AUFS_FSTYPE_H__ */
14412 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
14413 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
14414 +++ linux/fs/aufs/hfsnotify.c 2016-01-13 20:11:11.669760262 +0100
14417 + * Copyright (C) 2005-2015 Junjiro R. Okajima
14419 + * This program, aufs is free software; you can redistribute it and/or modify
14420 + * it under the terms of the GNU General Public License as published by
14421 + * the Free Software Foundation; either version 2 of the License, or
14422 + * (at your option) any later version.
14424 + * This program is distributed in the hope that it will be useful,
14425 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14426 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14427 + * GNU General Public License for more details.
14429 + * You should have received a copy of the GNU General Public License
14430 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14434 + * fsnotify for the lower directories
14439 +/* FS_IN_IGNORED is unnecessary */
14440 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
14441 + | FS_CREATE | FS_EVENT_ON_CHILD);
14442 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
14443 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
14445 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
14447 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
14450 + au_cache_free_hnotify(hn);
14451 + smp_mb__before_atomic();
14452 + if (atomic64_dec_and_test(&au_hfsn_ifree))
14453 + wake_up(&au_hfsn_wq);
14456 +static int au_hfsn_alloc(struct au_hinode *hinode)
14459 + struct au_hnotify *hn;
14460 + struct super_block *sb;
14461 + struct au_branch *br;
14462 + struct fsnotify_mark *mark;
14463 + aufs_bindex_t bindex;
14465 + hn = hinode->hi_notify;
14466 + sb = hn->hn_aufs_inode->i_sb;
14467 + bindex = au_br_index(sb, hinode->hi_id);
14468 + br = au_sbr(sb, bindex);
14469 + AuDebugOn(!br->br_hfsn);
14471 + mark = &hn->hn_mark;
14472 + fsnotify_init_mark(mark, au_hfsn_free_mark);
14473 + mark->mask = AuHfsnMask;
14475 + * by udba rename or rmdir, aufs assign a new inode to the known
14476 + * h_inode, so specify 1 to allow dups.
14479 + err = fsnotify_add_mark(mark, br->br_hfsn->hfsn_group, hinode->hi_inode,
14480 + /*mnt*/NULL, /*allow_dups*/1);
14481 + /* even if err */
14482 + fsnotify_put_mark(mark);
14488 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
14490 + struct fsnotify_mark *mark;
14491 + unsigned long long ull;
14492 + struct fsnotify_group *group;
14494 + ull = atomic64_inc_return(&au_hfsn_ifree);
14497 + mark = &hn->hn_mark;
14498 + spin_lock(&mark->lock);
14499 + group = mark->group;
14500 + fsnotify_get_group(group);
14501 + spin_unlock(&mark->lock);
14503 + fsnotify_destroy_mark(mark, group);
14504 + fsnotify_put_group(group);
14507 + /* free hn by myself */
14511 +/* ---------------------------------------------------------------------- */
14513 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
14515 + struct fsnotify_mark *mark;
14517 + mark = &hinode->hi_notify->hn_mark;
14518 + spin_lock(&mark->lock);
14520 + AuDebugOn(mark->mask & AuHfsnMask);
14521 + mark->mask |= AuHfsnMask;
14523 + AuDebugOn(!(mark->mask & AuHfsnMask));
14524 + mark->mask &= ~AuHfsnMask;
14526 + spin_unlock(&mark->lock);
14527 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
14530 +/* ---------------------------------------------------------------------- */
14532 +/* #define AuDbgHnotify */
14533 +#ifdef AuDbgHnotify
14534 +static char *au_hfsn_name(u32 mask)
14536 +#ifdef CONFIG_AUFS_DEBUG
14537 +#define test_ret(flag) \
14539 + if (mask & flag) \
14542 + test_ret(FS_ACCESS);
14543 + test_ret(FS_MODIFY);
14544 + test_ret(FS_ATTRIB);
14545 + test_ret(FS_CLOSE_WRITE);
14546 + test_ret(FS_CLOSE_NOWRITE);
14547 + test_ret(FS_OPEN);
14548 + test_ret(FS_MOVED_FROM);
14549 + test_ret(FS_MOVED_TO);
14550 + test_ret(FS_CREATE);
14551 + test_ret(FS_DELETE);
14552 + test_ret(FS_DELETE_SELF);
14553 + test_ret(FS_MOVE_SELF);
14554 + test_ret(FS_UNMOUNT);
14555 + test_ret(FS_Q_OVERFLOW);
14556 + test_ret(FS_IN_IGNORED);
14557 + test_ret(FS_ISDIR);
14558 + test_ret(FS_IN_ONESHOT);
14559 + test_ret(FS_EVENT_ON_CHILD);
14568 +/* ---------------------------------------------------------------------- */
14570 +static void au_hfsn_free_group(struct fsnotify_group *group)
14572 + struct au_br_hfsnotify *hfsn = group->private;
14578 +static int au_hfsn_handle_event(struct fsnotify_group *group,
14579 + struct inode *inode,
14580 + struct fsnotify_mark *inode_mark,
14581 + struct fsnotify_mark *vfsmount_mark,
14582 + u32 mask, void *data, int data_type,
14583 + const unsigned char *file_name, u32 cookie)
14586 + struct au_hnotify *hnotify;
14587 + struct inode *h_dir, *h_inode;
14588 + struct qstr h_child_qstr = QSTR_INIT(file_name, strlen(file_name));
14590 + AuDebugOn(data_type != FSNOTIFY_EVENT_INODE);
14593 + /* if FS_UNMOUNT happens, there must be another bug */
14594 + AuDebugOn(mask & FS_UNMOUNT);
14595 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
14600 +#ifdef AuDbgHnotify
14602 + if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
14603 + || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
14604 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
14605 + h_dir->i_ino, mask, au_hfsn_name(mask),
14606 + AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
14607 + /* WARN_ON(1); */
14612 + AuDebugOn(!inode_mark);
14613 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
14614 + err = au_hnotify(h_dir, hnotify, mask, &h_child_qstr, h_inode);
14620 +static struct fsnotify_ops au_hfsn_ops = {
14621 + .handle_event = au_hfsn_handle_event,
14622 + .free_group_priv = au_hfsn_free_group
14625 +/* ---------------------------------------------------------------------- */
14627 +static void au_hfsn_fin_br(struct au_branch *br)
14629 + struct au_br_hfsnotify *hfsn;
14631 + hfsn = br->br_hfsn;
14634 + fsnotify_put_group(hfsn->hfsn_group);
14639 +static int au_hfsn_init_br(struct au_branch *br, int perm)
14642 + struct fsnotify_group *group;
14643 + struct au_br_hfsnotify *hfsn;
14646 + br->br_hfsn = NULL;
14647 + if (!au_br_hnotifyable(perm))
14651 + hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
14652 + if (unlikely(!hfsn))
14656 + group = fsnotify_alloc_group(&au_hfsn_ops);
14657 + if (IS_ERR(group)) {
14658 + err = PTR_ERR(group);
14659 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
14663 + group->private = hfsn;
14664 + hfsn->hfsn_group = group;
14665 + br->br_hfsn = hfsn;
14666 + goto out; /* success */
14674 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
14679 + if (!br->br_hfsn)
14680 + err = au_hfsn_init_br(br, perm);
14685 +/* ---------------------------------------------------------------------- */
14687 +static void au_hfsn_fin(void)
14689 + AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
14690 + wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
14693 +const struct au_hnotify_op au_hnotify_op = {
14694 + .ctl = au_hfsn_ctl,
14695 + .alloc = au_hfsn_alloc,
14696 + .free = au_hfsn_free,
14698 + .fin = au_hfsn_fin,
14700 + .reset_br = au_hfsn_reset_br,
14701 + .fin_br = au_hfsn_fin_br,
14702 + .init_br = au_hfsn_init_br
14704 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
14705 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
14706 +++ linux/fs/aufs/hfsplus.c 2016-01-13 20:11:11.669760262 +0100
14709 + * Copyright (C) 2010-2015 Junjiro R. Okajima
14711 + * This program, aufs is free software; you can redistribute it and/or modify
14712 + * it under the terms of the GNU General Public License as published by
14713 + * the Free Software Foundation; either version 2 of the License, or
14714 + * (at your option) any later version.
14716 + * This program is distributed in the hope that it will be useful,
14717 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14718 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14719 + * GNU General Public License for more details.
14721 + * You should have received a copy of the GNU General Public License
14722 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14726 + * special support for filesystems which aqucires an inode mutex
14727 + * at final closing a file, eg, hfsplus.
14729 + * This trick is very simple and stupid, just to open the file before really
14730 + * neceeary open to tell hfsplus that this is not the final closing.
14731 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
14732 + * and au_h_open_post() after releasing it.
14737 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
14740 + struct file *h_file;
14741 + struct dentry *h_dentry;
14743 + h_dentry = au_h_dptr(dentry, bindex);
14744 + AuDebugOn(!h_dentry);
14745 + AuDebugOn(d_is_negative(h_dentry));
14748 + if (au_test_hfsplus(h_dentry->d_sb)
14749 + && d_is_reg(h_dentry))
14750 + h_file = au_h_open(dentry, bindex,
14751 + O_RDONLY | O_NOATIME | O_LARGEFILE,
14752 + /*file*/NULL, force_wr);
14756 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
14757 + struct file *h_file)
14761 + au_sbr_put(dentry->d_sb, bindex);
14764 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
14765 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
14766 +++ linux/fs/aufs/hnotify.c 2016-01-13 20:11:11.669760262 +0100
14769 + * Copyright (C) 2005-2015 Junjiro R. Okajima
14771 + * This program, aufs is free software; you can redistribute it and/or modify
14772 + * it under the terms of the GNU General Public License as published by
14773 + * the Free Software Foundation; either version 2 of the License, or
14774 + * (at your option) any later version.
14776 + * This program is distributed in the hope that it will be useful,
14777 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14778 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14779 + * GNU General Public License for more details.
14781 + * You should have received a copy of the GNU General Public License
14782 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14786 + * abstraction to notify the direct changes on lower directories
14791 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
14794 + struct au_hnotify *hn;
14797 + hn = au_cache_alloc_hnotify();
14799 + hn->hn_aufs_inode = inode;
14800 + hinode->hi_notify = hn;
14801 + err = au_hnotify_op.alloc(hinode);
14803 + if (unlikely(err)) {
14804 + hinode->hi_notify = NULL;
14805 + au_cache_free_hnotify(hn);
14807 + * The upper dir was removed by udba, but the same named
14808 + * dir left. In this case, aufs assignes a new inode
14809 + * number and set the monitor again.
14810 + * For the lower dir, the old monitnor is still left.
14812 + if (err == -EEXIST)
14821 +void au_hn_free(struct au_hinode *hinode)
14823 + struct au_hnotify *hn;
14825 + hn = hinode->hi_notify;
14827 + hinode->hi_notify = NULL;
14828 + if (au_hnotify_op.free(hinode, hn))
14829 + au_cache_free_hnotify(hn);
14833 +/* ---------------------------------------------------------------------- */
14835 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
14837 + if (hinode->hi_notify)
14838 + au_hnotify_op.ctl(hinode, do_set);
14841 +void au_hn_reset(struct inode *inode, unsigned int flags)
14843 + aufs_bindex_t bindex, bend;
14844 + struct inode *hi;
14845 + struct dentry *iwhdentry;
14847 + bend = au_ibend(inode);
14848 + for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
14849 + hi = au_h_iptr(inode, bindex);
14853 + /* mutex_lock_nested(&hi->i_mutex, AuLsc_I_CHILD); */
14854 + iwhdentry = au_hi_wh(inode, bindex);
14858 + au_set_h_iptr(inode, bindex, NULL, 0);
14859 + au_set_h_iptr(inode, bindex, au_igrab(hi),
14860 + flags & ~AuHi_XINO);
14863 + /* mutex_unlock(&hi->i_mutex); */
14867 +/* ---------------------------------------------------------------------- */
14869 +static int hn_xino(struct inode *inode, struct inode *h_inode)
14872 + aufs_bindex_t bindex, bend, bfound, bstart;
14873 + struct inode *h_i;
14876 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
14877 + pr_warn("branch root dir was changed\n");
14882 + bend = au_ibend(inode);
14883 + bstart = au_ibstart(inode);
14884 +#if 0 /* reserved for future use */
14885 + if (bindex == bend) {
14886 + /* keep this ino in rename case */
14890 + for (bindex = bstart; bindex <= bend; bindex++)
14891 + if (au_h_iptr(inode, bindex) == h_inode) {
14898 + for (bindex = bstart; bindex <= bend; bindex++) {
14899 + h_i = au_h_iptr(inode, bindex);
14903 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
14904 + /* ignore this error */
14905 + /* bad action? */
14908 + /* children inode number will be broken */
14915 +static int hn_gen_tree(struct dentry *dentry)
14917 + int err, i, j, ndentry;
14918 + struct au_dcsub_pages dpages;
14919 + struct au_dpage *dpage;
14920 + struct dentry **dentries;
14922 + err = au_dpages_init(&dpages, GFP_NOFS);
14923 + if (unlikely(err))
14925 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
14926 + if (unlikely(err))
14929 + for (i = 0; i < dpages.ndpage; i++) {
14930 + dpage = dpages.dpages + i;
14931 + dentries = dpage->dentries;
14932 + ndentry = dpage->ndentry;
14933 + for (j = 0; j < ndentry; j++) {
14934 + struct dentry *d;
14941 + if (d_really_is_positive(d))
14942 + /* todo: reset children xino?
14943 + cached children only? */
14944 + au_iigen_dec(d_inode(d));
14949 + au_dpages_free(&dpages);
14952 + /* discard children */
14953 + dentry_unhash(dentry);
14961 + * return 0 if processed.
14963 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
14964 + const unsigned int isdir)
14967 + struct dentry *d;
14968 + struct qstr *dname;
14971 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
14972 + pr_warn("branch root dir was changed\n");
14978 + AuDebugOn(!name);
14979 + au_iigen_dec(inode);
14980 + spin_lock(&inode->i_lock);
14981 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
14982 + spin_lock(&d->d_lock);
14983 + dname = &d->d_name;
14984 + if (dname->len != nlen
14985 + && memcmp(dname->name, name, nlen)) {
14986 + spin_unlock(&d->d_lock);
14991 + spin_unlock(&d->d_lock);
14994 + spin_unlock(&inode->i_lock);
14996 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
14997 + d = d_find_any_alias(inode);
14999 + au_iigen_dec(inode);
15003 + spin_lock(&d->d_lock);
15004 + dname = &d->d_name;
15005 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
15006 + spin_unlock(&d->d_lock);
15007 + err = hn_gen_tree(d);
15008 + spin_lock(&d->d_lock);
15010 + spin_unlock(&d->d_lock);
15019 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
15023 + if (IS_ROOT(dentry)) {
15024 + pr_warn("branch root dir was changed\n");
15030 + au_digen_dec(dentry);
15031 + if (d_really_is_positive(dentry))
15032 + au_iigen_dec(d_inode(dentry));
15034 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
15035 + if (d_really_is_positive(dentry))
15036 + err = hn_gen_tree(dentry);
15043 +/* ---------------------------------------------------------------------- */
15045 +/* hnotify job flags */
15046 +#define AuHnJob_XINO0 1
15047 +#define AuHnJob_GEN (1 << 1)
15048 +#define AuHnJob_DIRENT (1 << 2)
15049 +#define AuHnJob_ISDIR (1 << 3)
15050 +#define AuHnJob_TRYXINO0 (1 << 4)
15051 +#define AuHnJob_MNTPNT (1 << 5)
15052 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
15053 +#define au_fset_hnjob(flags, name) \
15054 + do { (flags) |= AuHnJob_##name; } while (0)
15055 +#define au_fclr_hnjob(flags, name) \
15056 + do { (flags) &= ~AuHnJob_##name; } while (0)
15064 +struct au_hnotify_args {
15065 + struct inode *h_dir, *dir, *h_child_inode;
15067 + unsigned int flags[AuHnLast];
15068 + unsigned int h_child_nlen;
15069 + char h_child_name[];
15072 +struct hn_job_args {
15073 + unsigned int flags;
15074 + struct inode *inode, *h_inode, *dir, *h_dir;
15075 + struct dentry *dentry;
15080 +static int hn_job(struct hn_job_args *a)
15082 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
15086 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
15087 + hn_xino(a->inode, a->h_inode); /* ignore this error */
15089 + if (au_ftest_hnjob(a->flags, TRYXINO0)
15092 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
15093 + if (!a->h_inode->i_nlink
15094 + && !(a->h_inode->i_state & I_LINKABLE))
15095 + hn_xino(a->inode, a->h_inode); /* ignore this error */
15096 + mutex_unlock(&a->h_inode->i_mutex);
15099 + /* make the generation obsolete */
15100 + if (au_ftest_hnjob(a->flags, GEN)) {
15103 + e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
15105 + if (e && a->dentry)
15106 + hn_gen_by_name(a->dentry, isdir);
15107 + /* ignore this error */
15110 + /* make dir entries obsolete */
15111 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
15112 + struct au_vdir *vdir;
15114 + vdir = au_ivdir(a->inode);
15116 + vdir->vd_jiffy = 0;
15117 + /* IMustLock(a->inode); */
15118 + /* a->inode->i_version++; */
15121 + /* can do nothing but warn */
15122 + if (au_ftest_hnjob(a->flags, MNTPNT)
15124 + && d_mountpoint(a->dentry))
15125 + pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
15130 +/* ---------------------------------------------------------------------- */
15132 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
15133 + struct inode *dir)
15135 + struct dentry *dentry, *d, *parent;
15136 + struct qstr *dname;
15138 + parent = d_find_any_alias(dir);
15143 + spin_lock(&parent->d_lock);
15144 + list_for_each_entry(d, &parent->d_subdirs, d_child) {
15145 + /* AuDbg("%pd\n", d); */
15146 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
15147 + dname = &d->d_name;
15148 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
15149 + goto cont_unlock;
15153 + goto cont_unlock;
15154 + if (au_dcount(d) > 0) {
15155 + dentry = dget_dlock(d);
15156 + spin_unlock(&d->d_lock);
15161 + spin_unlock(&d->d_lock);
15163 + spin_unlock(&parent->d_lock);
15167 + di_write_lock_child(dentry);
15172 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
15173 + aufs_bindex_t bindex, ino_t h_ino)
15175 + struct inode *inode;
15180 + err = au_xino_read(sb, bindex, h_ino, &ino);
15182 + inode = ilookup(sb, ino);
15186 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
15187 + pr_warn("wrong root branch\n");
15193 + ii_write_lock_child(inode);
15199 +static void au_hn_bh(void *_args)
15201 + struct au_hnotify_args *a = _args;
15202 + struct super_block *sb;
15203 + aufs_bindex_t bindex, bend, bfound;
15204 + unsigned char xino, try_iput;
15206 + struct inode *inode;
15208 + struct hn_job_args args;
15209 + struct dentry *dentry;
15210 + struct au_sbinfo *sbinfo;
15212 + AuDebugOn(!_args);
15213 + AuDebugOn(!a->h_dir);
15214 + AuDebugOn(!a->dir);
15215 + AuDebugOn(!a->mask);
15216 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
15217 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
15218 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
15223 + * do not lock a->dir->i_mutex here
15224 + * because of d_revalidate() may cause a deadlock.
15226 + sb = a->dir->i_sb;
15228 + sbinfo = au_sbi(sb);
15229 + AuDebugOn(!sbinfo);
15230 + si_write_lock(sb, AuLock_NOPLMW);
15232 + ii_read_lock_parent(a->dir);
15234 + bend = au_ibend(a->dir);
15235 + for (bindex = au_ibstart(a->dir); bindex <= bend; bindex++)
15236 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
15240 + ii_read_unlock(a->dir);
15241 + if (unlikely(bfound < 0))
15244 + xino = !!au_opt_test(au_mntflags(sb), XINO);
15246 + if (a->h_child_inode)
15247 + h_ino = a->h_child_inode->i_ino;
15249 + if (a->h_child_nlen
15250 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
15251 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
15252 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
15255 + if (dentry && d_really_is_positive(dentry))
15256 + inode = d_inode(dentry);
15257 + if (xino && !inode && h_ino
15258 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
15259 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
15260 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
15261 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
15265 + args.flags = a->flags[AuHn_CHILD];
15266 + args.dentry = dentry;
15267 + args.inode = inode;
15268 + args.h_inode = a->h_child_inode;
15269 + args.dir = a->dir;
15270 + args.h_dir = a->h_dir;
15271 + args.h_name = a->h_child_name;
15272 + args.h_nlen = a->h_child_nlen;
15273 + err = hn_job(&args);
15275 + if (au_di(dentry))
15276 + di_write_unlock(dentry);
15279 + if (inode && try_iput) {
15280 + ii_write_unlock(inode);
15284 + ii_write_lock_parent(a->dir);
15285 + args.flags = a->flags[AuHn_PARENT];
15286 + args.dentry = NULL;
15287 + args.inode = a->dir;
15288 + args.h_inode = a->h_dir;
15290 + args.h_dir = NULL;
15291 + args.h_name = NULL;
15293 + err = hn_job(&args);
15294 + ii_write_unlock(a->dir);
15297 + iput(a->h_child_inode);
15300 + si_write_unlock(sb);
15301 + au_nwt_done(&sbinfo->si_nowait);
15305 +/* ---------------------------------------------------------------------- */
15307 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
15308 + struct qstr *h_child_qstr, struct inode *h_child_inode)
15311 + unsigned int flags[AuHnLast], f;
15312 + unsigned char isdir, isroot, wh;
15313 + struct inode *dir;
15314 + struct au_hnotify_args *args;
15315 + char *p, *h_child_name;
15318 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
15319 + dir = igrab(hnotify->hn_aufs_inode);
15323 + isroot = (dir->i_ino == AUFS_ROOT_INO);
15325 + h_child_name = (void *)h_child_qstr->name;
15326 + len = h_child_qstr->len;
15327 + if (h_child_name) {
15328 + if (len > AUFS_WH_PFX_LEN
15329 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
15330 + h_child_name += AUFS_WH_PFX_LEN;
15331 + len -= AUFS_WH_PFX_LEN;
15337 + if (h_child_inode)
15338 + isdir = !!S_ISDIR(h_child_inode->i_mode);
15339 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
15340 + flags[AuHn_CHILD] = 0;
15342 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
15343 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
15344 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
15345 + switch (mask & FS_EVENTS_POSS_ON_CHILD) {
15346 + case FS_MOVED_FROM:
15347 + case FS_MOVED_TO:
15348 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
15349 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
15352 + AuDebugOn(!h_child_name);
15357 + * aufs never be able to get this child inode.
15358 + * revalidation should be in d_revalidate()
15359 + * by checking i_nlink, i_generation or d_unhashed().
15361 + AuDebugOn(!h_child_name);
15362 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
15363 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
15371 + h_child_inode = NULL;
15374 + /* iput() and kfree() will be called in au_hnotify() */
15375 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
15376 + if (unlikely(!args)) {
15377 + AuErr1("no memory\n");
15381 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
15382 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
15383 + args->mask = mask;
15385 + args->h_dir = igrab(h_dir);
15386 + if (h_child_inode)
15387 + h_child_inode = igrab(h_child_inode); /* can be NULL */
15388 + args->h_child_inode = h_child_inode;
15389 + args->h_child_nlen = len;
15391 + p = (void *)args;
15392 + p += sizeof(*args);
15393 + memcpy(p, h_child_name, len);
15397 + /* NFS fires the event for silly-renamed one from kworker */
15399 + if (!dir->i_nlink
15400 + || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
15402 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
15403 + if (unlikely(err)) {
15404 + pr_err("wkq %d\n", err);
15405 + iput(args->h_child_inode);
15406 + iput(args->h_dir);
15415 +/* ---------------------------------------------------------------------- */
15417 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
15421 + AuDebugOn(!(udba & AuOptMask_UDBA));
15424 + if (au_hnotify_op.reset_br)
15425 + err = au_hnotify_op.reset_br(udba, br, perm);
15430 +int au_hnotify_init_br(struct au_branch *br, int perm)
15435 + if (au_hnotify_op.init_br)
15436 + err = au_hnotify_op.init_br(br, perm);
15441 +void au_hnotify_fin_br(struct au_branch *br)
15443 + if (au_hnotify_op.fin_br)
15444 + au_hnotify_op.fin_br(br);
15447 +static void au_hn_destroy_cache(void)
15449 + kmem_cache_destroy(au_cachep[AuCache_HNOTIFY]);
15450 + au_cachep[AuCache_HNOTIFY] = NULL;
15453 +int __init au_hnotify_init(void)
15458 + au_cachep[AuCache_HNOTIFY] = AuCache(au_hnotify);
15459 + if (au_cachep[AuCache_HNOTIFY]) {
15461 + if (au_hnotify_op.init)
15462 + err = au_hnotify_op.init();
15463 + if (unlikely(err))
15464 + au_hn_destroy_cache();
15470 +void au_hnotify_fin(void)
15472 + if (au_hnotify_op.fin)
15473 + au_hnotify_op.fin();
15474 + /* cf. au_cache_fin() */
15475 + if (au_cachep[AuCache_HNOTIFY])
15476 + au_hn_destroy_cache();
15478 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
15479 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
15480 +++ linux/fs/aufs/iinfo.c 2016-01-13 20:11:11.669760262 +0100
15483 + * Copyright (C) 2005-2015 Junjiro R. Okajima
15485 + * This program, aufs is free software; you can redistribute it and/or modify
15486 + * it under the terms of the GNU General Public License as published by
15487 + * the Free Software Foundation; either version 2 of the License, or
15488 + * (at your option) any later version.
15490 + * This program is distributed in the hope that it will be useful,
15491 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15492 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15493 + * GNU General Public License for more details.
15495 + * You should have received a copy of the GNU General Public License
15496 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15500 + * inode private data
15505 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
15507 + struct inode *h_inode;
15509 + IiMustAnyLock(inode);
15511 + h_inode = au_ii(inode)->ii_hinode[0 + bindex].hi_inode;
15512 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
15516 +/* todo: hard/soft set? */
15517 +void au_hiput(struct au_hinode *hinode)
15519 + au_hn_free(hinode);
15520 + dput(hinode->hi_whdentry);
15521 + iput(hinode->hi_inode);
15524 +unsigned int au_hi_flags(struct inode *inode, int isdir)
15526 + unsigned int flags;
15527 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
15530 + if (au_opt_test(mnt_flags, XINO))
15531 + au_fset_hi(flags, XINO);
15532 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
15533 + au_fset_hi(flags, HNOTIFY);
15537 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
15538 + struct inode *h_inode, unsigned int flags)
15540 + struct au_hinode *hinode;
15541 + struct inode *hi;
15542 + struct au_iinfo *iinfo = au_ii(inode);
15544 + IiMustWriteLock(inode);
15546 + hinode = iinfo->ii_hinode + bindex;
15547 + hi = hinode->hi_inode;
15548 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
15551 + au_hiput(hinode);
15552 + hinode->hi_inode = h_inode;
15555 + struct super_block *sb = inode->i_sb;
15556 + struct au_branch *br;
15558 + AuDebugOn(inode->i_mode
15559 + && (h_inode->i_mode & S_IFMT)
15560 + != (inode->i_mode & S_IFMT));
15561 + if (bindex == iinfo->ii_bstart)
15562 + au_cpup_igen(inode, h_inode);
15563 + br = au_sbr(sb, bindex);
15564 + hinode->hi_id = br->br_id;
15565 + if (au_ftest_hi(flags, XINO)) {
15566 + err = au_xino_write(sb, bindex, h_inode->i_ino,
15568 + if (unlikely(err))
15569 + AuIOErr1("failed au_xino_write() %d\n", err);
15572 + if (au_ftest_hi(flags, HNOTIFY)
15573 + && au_br_hnotifyable(br->br_perm)) {
15574 + err = au_hn_alloc(hinode, inode);
15575 + if (unlikely(err))
15576 + AuIOErr1("au_hn_alloc() %d\n", err);
15581 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
15582 + struct dentry *h_wh)
15584 + struct au_hinode *hinode;
15586 + IiMustWriteLock(inode);
15588 + hinode = au_ii(inode)->ii_hinode + bindex;
15589 + AuDebugOn(hinode->hi_whdentry);
15590 + hinode->hi_whdentry = h_wh;
15593 +void au_update_iigen(struct inode *inode, int half)
15595 + struct au_iinfo *iinfo;
15596 + struct au_iigen *iigen;
15597 + unsigned int sigen;
15599 + sigen = au_sigen(inode->i_sb);
15600 + iinfo = au_ii(inode);
15601 + iigen = &iinfo->ii_generation;
15602 + spin_lock(&iigen->ig_spin);
15603 + iigen->ig_generation = sigen;
15605 + au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
15607 + au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
15608 + spin_unlock(&iigen->ig_spin);
15611 +/* it may be called at remount time, too */
15612 +void au_update_ibrange(struct inode *inode, int do_put_zero)
15614 + struct au_iinfo *iinfo;
15615 + aufs_bindex_t bindex, bend;
15617 + iinfo = au_ii(inode);
15621 + IiMustWriteLock(inode);
15623 + if (do_put_zero && iinfo->ii_bstart >= 0) {
15624 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
15626 + struct inode *h_i;
15628 + h_i = iinfo->ii_hinode[0 + bindex].hi_inode;
15631 + && !(h_i->i_state & I_LINKABLE))
15632 + au_set_h_iptr(inode, bindex, NULL, 0);
15636 + iinfo->ii_bstart = -1;
15637 + iinfo->ii_bend = -1;
15638 + bend = au_sbend(inode->i_sb);
15639 + for (bindex = 0; bindex <= bend; bindex++)
15640 + if (iinfo->ii_hinode[0 + bindex].hi_inode) {
15641 + iinfo->ii_bstart = bindex;
15644 + if (iinfo->ii_bstart >= 0)
15645 + for (bindex = bend; bindex >= iinfo->ii_bstart; bindex--)
15646 + if (iinfo->ii_hinode[0 + bindex].hi_inode) {
15647 + iinfo->ii_bend = bindex;
15650 + AuDebugOn(iinfo->ii_bstart > iinfo->ii_bend);
15653 +/* ---------------------------------------------------------------------- */
15655 +void au_icntnr_init_once(void *_c)
15657 + struct au_icntnr *c = _c;
15658 + struct au_iinfo *iinfo = &c->iinfo;
15659 + static struct lock_class_key aufs_ii;
15661 + spin_lock_init(&iinfo->ii_generation.ig_spin);
15662 + au_rw_init(&iinfo->ii_rwsem);
15663 + au_rw_class(&iinfo->ii_rwsem, &aufs_ii);
15664 + inode_init_once(&c->vfs_inode);
15667 +int au_iinfo_init(struct inode *inode)
15669 + struct au_iinfo *iinfo;
15670 + struct super_block *sb;
15673 + sb = inode->i_sb;
15674 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
15675 + nbr = au_sbend(sb) + 1;
15676 + if (unlikely(nbr <= 0))
15678 + iinfo->ii_hinode = kcalloc(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
15679 + if (iinfo->ii_hinode) {
15680 + au_ninodes_inc(sb);
15681 + for (i = 0; i < nbr; i++)
15682 + iinfo->ii_hinode[i].hi_id = -1;
15684 + iinfo->ii_generation.ig_generation = au_sigen(sb);
15685 + iinfo->ii_bstart = -1;
15686 + iinfo->ii_bend = -1;
15687 + iinfo->ii_vdir = NULL;
15693 +int au_ii_realloc(struct au_iinfo *iinfo, int nbr)
15696 + struct au_hinode *hip;
15698 + AuRwMustWriteLock(&iinfo->ii_rwsem);
15701 + sz = sizeof(*hip) * (iinfo->ii_bend + 1);
15703 + sz = sizeof(*hip);
15704 + hip = au_kzrealloc(iinfo->ii_hinode, sz, sizeof(*hip) * nbr, GFP_NOFS);
15706 + iinfo->ii_hinode = hip;
15713 +void au_iinfo_fin(struct inode *inode)
15715 + struct au_iinfo *iinfo;
15716 + struct au_hinode *hi;
15717 + struct super_block *sb;
15718 + aufs_bindex_t bindex, bend;
15719 + const unsigned char unlinked = !inode->i_nlink;
15721 + iinfo = au_ii(inode);
15722 + /* bad_inode case */
15726 + sb = inode->i_sb;
15727 + au_ninodes_dec(sb);
15728 + if (si_pid_test(sb))
15729 + au_xino_delete_inode(inode, unlinked);
15732 + * it is safe to hide the dependency between sbinfo and
15736 + si_noflush_read_lock(sb);
15737 + au_xino_delete_inode(inode, unlinked);
15738 + si_read_unlock(sb);
15742 + if (iinfo->ii_vdir)
15743 + au_vdir_free(iinfo->ii_vdir);
15745 + bindex = iinfo->ii_bstart;
15746 + if (bindex >= 0) {
15747 + hi = iinfo->ii_hinode + bindex;
15748 + bend = iinfo->ii_bend;
15749 + while (bindex++ <= bend) {
15750 + if (hi->hi_inode)
15755 + kfree(iinfo->ii_hinode);
15756 + iinfo->ii_hinode = NULL;
15757 + AuRwDestroy(&iinfo->ii_rwsem);
15759 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
15760 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
15761 +++ linux/fs/aufs/inode.c 2016-01-13 20:11:11.669760262 +0100
15764 + * Copyright (C) 2005-2015 Junjiro R. Okajima
15766 + * This program, aufs is free software; you can redistribute it and/or modify
15767 + * it under the terms of the GNU General Public License as published by
15768 + * the Free Software Foundation; either version 2 of the License, or
15769 + * (at your option) any later version.
15771 + * This program is distributed in the hope that it will be useful,
15772 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15773 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15774 + * GNU General Public License for more details.
15776 + * You should have received a copy of the GNU General Public License
15777 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15781 + * inode functions
15786 +struct inode *au_igrab(struct inode *inode)
15789 + AuDebugOn(!atomic_read(&inode->i_count));
15795 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
15797 + au_cpup_attr_all(inode, /*force*/0);
15798 + au_update_iigen(inode, /*half*/1);
15800 + inode->i_version++;
15803 +static int au_ii_refresh(struct inode *inode, int *update)
15807 + aufs_bindex_t bindex, new_bindex;
15808 + struct super_block *sb;
15809 + struct au_iinfo *iinfo;
15810 + struct au_hinode *p, *q, tmp;
15812 + IiMustWriteLock(inode);
15815 + sb = inode->i_sb;
15816 + type = inode->i_mode & S_IFMT;
15817 + iinfo = au_ii(inode);
15818 + err = au_ii_realloc(iinfo, au_sbend(sb) + 1);
15819 + if (unlikely(err))
15822 + AuDebugOn(iinfo->ii_bstart < 0);
15823 + p = iinfo->ii_hinode + iinfo->ii_bstart;
15824 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
15826 + if (!p->hi_inode)
15829 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
15830 + new_bindex = au_br_index(sb, p->hi_id);
15831 + if (new_bindex == bindex)
15834 + if (new_bindex < 0) {
15837 + p->hi_inode = NULL;
15841 + if (new_bindex < iinfo->ii_bstart)
15842 + iinfo->ii_bstart = new_bindex;
15843 + if (iinfo->ii_bend < new_bindex)
15844 + iinfo->ii_bend = new_bindex;
15845 + /* swap two lower inode, and loop again */
15846 + q = iinfo->ii_hinode + new_bindex;
15850 + if (tmp.hi_inode) {
15855 + au_update_ibrange(inode, /*do_put_zero*/0);
15856 + e = au_dy_irefresh(inode);
15857 + if (unlikely(e && !err))
15865 +void au_refresh_iop(struct inode *inode, int force_getattr)
15868 + struct au_sbinfo *sbi = au_sbi(inode->i_sb);
15869 + const struct inode_operations *iop
15870 + = force_getattr ? aufs_iop : sbi->si_iop_array;
15872 + if (inode->i_op == iop)
15875 + switch (inode->i_mode & S_IFMT) {
15877 + type = AuIop_DIR;
15880 + type = AuIop_SYMLINK;
15883 + type = AuIop_OTHER;
15887 + inode->i_op = iop + type;
15888 + /* unnecessary smp_wmb() */
15891 +int au_refresh_hinode_self(struct inode *inode)
15895 + err = au_ii_refresh(inode, &update);
15897 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
15903 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
15905 + int err, e, update;
15906 + unsigned int flags;
15908 + aufs_bindex_t bindex, bend;
15909 + unsigned char isdir;
15910 + struct au_hinode *p;
15911 + struct au_iinfo *iinfo;
15913 + err = au_ii_refresh(inode, &update);
15914 + if (unlikely(err))
15918 + iinfo = au_ii(inode);
15919 + p = iinfo->ii_hinode + iinfo->ii_bstart;
15920 + mode = (inode->i_mode & S_IFMT);
15921 + isdir = S_ISDIR(mode);
15922 + flags = au_hi_flags(inode, isdir);
15923 + bend = au_dbend(dentry);
15924 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
15925 + struct inode *h_i, *h_inode;
15926 + struct dentry *h_d;
15928 + h_d = au_h_dptr(dentry, bindex);
15929 + if (!h_d || d_is_negative(h_d))
15932 + h_inode = d_inode(h_d);
15933 + AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
15934 + if (iinfo->ii_bstart <= bindex && bindex <= iinfo->ii_bend) {
15935 + h_i = au_h_iptr(inode, bindex);
15937 + if (h_i == h_inode)
15943 + if (bindex < iinfo->ii_bstart)
15944 + iinfo->ii_bstart = bindex;
15945 + if (iinfo->ii_bend < bindex)
15946 + iinfo->ii_bend = bindex;
15947 + au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
15950 + au_update_ibrange(inode, /*do_put_zero*/0);
15951 + e = au_dy_irefresh(inode);
15952 + if (unlikely(e && !err))
15955 + au_refresh_hinode_attr(inode, update && isdir);
15962 +static int set_inode(struct inode *inode, struct dentry *dentry)
15965 + unsigned int flags;
15967 + aufs_bindex_t bindex, bstart, btail;
15968 + unsigned char isdir;
15969 + struct dentry *h_dentry;
15970 + struct inode *h_inode;
15971 + struct au_iinfo *iinfo;
15972 + struct inode_operations *iop;
15974 + IiMustWriteLock(inode);
15978 + iop = au_sbi(inode->i_sb)->si_iop_array;
15979 + bstart = au_dbstart(dentry);
15980 + h_dentry = au_h_dptr(dentry, bstart);
15981 + h_inode = d_inode(h_dentry);
15982 + mode = h_inode->i_mode;
15983 + switch (mode & S_IFMT) {
15985 + btail = au_dbtail(dentry);
15986 + inode->i_op = iop + AuIop_OTHER;
15987 + inode->i_fop = &aufs_file_fop;
15988 + err = au_dy_iaop(inode, bstart, h_inode);
15989 + if (unlikely(err))
15994 + btail = au_dbtaildir(dentry);
15995 + inode->i_op = iop + AuIop_DIR;
15996 + inode->i_fop = &aufs_dir_fop;
15999 + btail = au_dbtail(dentry);
16000 + inode->i_op = iop + AuIop_SYMLINK;
16006 + btail = au_dbtail(dentry);
16007 + inode->i_op = iop + AuIop_OTHER;
16008 + init_special_inode(inode, mode, h_inode->i_rdev);
16011 + AuIOErr("Unknown file type 0%o\n", mode);
16016 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
16017 + flags = au_hi_flags(inode, isdir);
16018 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
16019 + && au_ftest_hi(flags, HNOTIFY)
16020 + && dentry->d_name.len > AUFS_WH_PFX_LEN
16021 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
16022 + au_fclr_hi(flags, HNOTIFY);
16023 + iinfo = au_ii(inode);
16024 + iinfo->ii_bstart = bstart;
16025 + iinfo->ii_bend = btail;
16026 + for (bindex = bstart; bindex <= btail; bindex++) {
16027 + h_dentry = au_h_dptr(dentry, bindex);
16029 + au_set_h_iptr(inode, bindex,
16030 + au_igrab(d_inode(h_dentry)), flags);
16032 + au_cpup_attr_all(inode, /*force*/1);
16034 + * to force calling aufs_get_acl() every time,
16035 + * do not call cache_no_acl() for aufs inode.
16043 + * successful returns with iinfo write_locked
16045 + * zero: success, matched
16046 + * plus: no error, but unmatched
16048 +static int reval_inode(struct inode *inode, struct dentry *dentry)
16051 + unsigned int gen;
16052 + struct au_iigen iigen;
16053 + aufs_bindex_t bindex, bend;
16054 + struct inode *h_inode, *h_dinode;
16055 + struct dentry *h_dentry;
16058 + * before this function, if aufs got any iinfo lock, it must be only
16059 + * one, the parent dir.
16060 + * it can happen by UDBA and the obsoleted inode number.
16063 + if (unlikely(inode->i_ino == parent_ino(dentry)))
16067 + ii_write_lock_new_child(inode);
16068 + h_dentry = au_h_dptr(dentry, au_dbstart(dentry));
16069 + h_dinode = d_inode(h_dentry);
16070 + bend = au_ibend(inode);
16071 + for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
16072 + h_inode = au_h_iptr(inode, bindex);
16073 + if (!h_inode || h_inode != h_dinode)
16077 + gen = au_iigen(inode, &iigen);
16078 + if (gen == au_digen(dentry)
16079 + && !au_ig_ftest(iigen.ig_flags, HALF_REFRESHED))
16082 + /* fully refresh inode using dentry */
16083 + err = au_refresh_hinode(inode, dentry);
16085 + au_update_iigen(inode, /*half*/0);
16089 + if (unlikely(err))
16090 + ii_write_unlock(inode);
16095 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
16096 + unsigned int d_type, ino_t *ino)
16099 + struct mutex *mtx;
16101 + /* prevent hardlinked inode number from race condition */
16103 + if (d_type != DT_DIR) {
16104 + mtx = &au_sbr(sb, bindex)->br_xino.xi_nondir_mtx;
16107 + err = au_xino_read(sb, bindex, h_ino, ino);
16108 + if (unlikely(err))
16113 + *ino = au_xino_new_ino(sb);
16114 + if (unlikely(!*ino))
16116 + err = au_xino_write(sb, bindex, h_ino, *ino);
16117 + if (unlikely(err))
16123 + mutex_unlock(mtx);
16127 +/* successful returns with iinfo write_locked */
16128 +/* todo: return with unlocked? */
16129 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
16131 + struct inode *inode, *h_inode;
16132 + struct dentry *h_dentry;
16133 + struct super_block *sb;
16134 + struct mutex *mtx;
16135 + ino_t h_ino, ino;
16137 + aufs_bindex_t bstart;
16139 + sb = dentry->d_sb;
16140 + bstart = au_dbstart(dentry);
16141 + h_dentry = au_h_dptr(dentry, bstart);
16142 + h_inode = d_inode(h_dentry);
16143 + h_ino = h_inode->i_ino;
16146 + * stop 'race'-ing between hardlinks under different
16150 + if (!d_is_dir(h_dentry))
16151 + mtx = &au_sbr(sb, bstart)->br_xino.xi_nondir_mtx;
16156 + err = au_xino_read(sb, bstart, h_ino, &ino);
16157 + inode = ERR_PTR(err);
16158 + if (unlikely(err))
16162 + ino = au_xino_new_ino(sb);
16163 + if (unlikely(!ino)) {
16164 + inode = ERR_PTR(-EIO);
16169 + AuDbg("i%lu\n", (unsigned long)ino);
16170 + inode = au_iget_locked(sb, ino);
16171 + err = PTR_ERR(inode);
16172 + if (IS_ERR(inode))
16175 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
16176 + if (inode->i_state & I_NEW) {
16177 + /* verbose coding for lock class name */
16178 + if (unlikely(d_is_symlink(h_dentry)))
16179 + au_rw_class(&au_ii(inode)->ii_rwsem,
16180 + au_lc_key + AuLcSymlink_IIINFO);
16181 + else if (unlikely(d_is_dir(h_dentry)))
16182 + au_rw_class(&au_ii(inode)->ii_rwsem,
16183 + au_lc_key + AuLcDir_IIINFO);
16184 + else /* likely */
16185 + au_rw_class(&au_ii(inode)->ii_rwsem,
16186 + au_lc_key + AuLcNonDir_IIINFO);
16188 + ii_write_lock_new_child(inode);
16189 + err = set_inode(inode, dentry);
16191 + unlock_new_inode(inode);
16192 + goto out; /* success */
16196 + * iget_failed() calls iput(), but we need to call
16197 + * ii_write_unlock() after iget_failed(). so dirty hack for
16200 + atomic_inc(&inode->i_count);
16201 + iget_failed(inode);
16202 + ii_write_unlock(inode);
16203 + au_xino_write(sb, bstart, h_ino, /*ino*/0);
16204 + /* ignore this error */
16206 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
16208 + * horrible race condition between lookup, readdir and copyup
16209 + * (or something).
16212 + mutex_unlock(mtx);
16213 + err = reval_inode(inode, dentry);
16214 + if (unlikely(err < 0)) {
16221 + goto out; /* success */
16226 + if (unlikely(au_test_fs_unique_ino(h_inode)))
16227 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
16228 + " b%d, %s, %pd, hi%lu, i%lu.\n",
16229 + bstart, au_sbtype(h_dentry->d_sb), dentry,
16230 + (unsigned long)h_ino, (unsigned long)ino);
16232 + err = au_xino_write(sb, bstart, h_ino, /*ino*/0);
16236 + mutex_unlock(mtx);
16242 + inode = ERR_PTR(err);
16245 + mutex_unlock(mtx);
16249 +/* ---------------------------------------------------------------------- */
16251 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
16252 + struct inode *inode)
16255 + struct inode *hi;
16257 + err = au_br_rdonly(au_sbr(sb, bindex));
16259 + /* pseudo-link after flushed may happen out of bounds */
16262 + && au_ibstart(inode) <= bindex
16263 + && bindex <= au_ibend(inode)) {
16265 + * permission check is unnecessary since vfsub routine
16266 + * will be called later
16268 + hi = au_h_iptr(inode, bindex);
16270 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
16276 +int au_test_h_perm(struct inode *h_inode, int mask)
16278 + if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
16280 + return inode_permission(h_inode, mask);
16283 +int au_test_h_perm_sio(struct inode *h_inode, int mask)
16285 + if (au_test_nfs(h_inode->i_sb)
16286 + && (mask & MAY_WRITE)
16287 + && S_ISDIR(h_inode->i_mode))
16288 + mask |= MAY_READ; /* force permission check */
16289 + return au_test_h_perm(h_inode, mask);
16291 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
16292 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
16293 +++ linux/fs/aufs/inode.h 2016-01-13 20:11:11.669760262 +0100
16296 + * Copyright (C) 2005-2015 Junjiro R. Okajima
16298 + * This program, aufs is free software; you can redistribute it and/or modify
16299 + * it under the terms of the GNU General Public License as published by
16300 + * the Free Software Foundation; either version 2 of the License, or
16301 + * (at your option) any later version.
16303 + * This program is distributed in the hope that it will be useful,
16304 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16305 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16306 + * GNU General Public License for more details.
16308 + * You should have received a copy of the GNU General Public License
16309 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16313 + * inode operations
16316 +#ifndef __AUFS_INODE_H__
16317 +#define __AUFS_INODE_H__
16321 +#include <linux/fsnotify.h>
16322 +#include "rwsem.h"
16326 +struct au_hnotify {
16327 +#ifdef CONFIG_AUFS_HNOTIFY
16328 +#ifdef CONFIG_AUFS_HFSNOTIFY
16329 + /* never use fsnotify_add_vfsmount_mark() */
16330 + struct fsnotify_mark hn_mark;
16332 + struct inode *hn_aufs_inode; /* no get/put */
16334 +} ____cacheline_aligned_in_smp;
16336 +struct au_hinode {
16337 + struct inode *hi_inode;
16338 + aufs_bindex_t hi_id;
16339 +#ifdef CONFIG_AUFS_HNOTIFY
16340 + struct au_hnotify *hi_notify;
16343 + /* reference to the copied-up whiteout with get/put */
16344 + struct dentry *hi_whdentry;
16348 +#define AuIG_HALF_REFRESHED 1
16349 +#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
16350 +#define au_ig_fset(flags, name) \
16351 + do { (flags) |= AuIG_##name; } while (0)
16352 +#define au_ig_fclr(flags, name) \
16353 + do { (flags) &= ~AuIG_##name; } while (0)
16356 + spinlock_t ig_spin;
16357 + __u32 ig_generation, ig_flags;
16362 + struct au_iigen ii_generation;
16363 + struct super_block *ii_hsb1; /* no get/put */
16365 + struct au_rwsem ii_rwsem;
16366 + aufs_bindex_t ii_bstart, ii_bend;
16368 + struct au_hinode *ii_hinode;
16369 + struct au_vdir *ii_vdir;
16372 +struct au_icntnr {
16373 + struct au_iinfo iinfo;
16374 + struct inode vfs_inode;
16375 +} ____cacheline_aligned_in_smp;
16377 +/* au_pin flags */
16378 +#define AuPin_DI_LOCKED 1
16379 +#define AuPin_MNT_WRITE (1 << 1)
16380 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
16381 +#define au_fset_pin(flags, name) \
16382 + do { (flags) |= AuPin_##name; } while (0)
16383 +#define au_fclr_pin(flags, name) \
16384 + do { (flags) &= ~AuPin_##name; } while (0)
16388 + struct dentry *dentry;
16389 + unsigned int udba;
16390 + unsigned char lsc_di, lsc_hi, flags;
16391 + aufs_bindex_t bindex;
16394 + struct dentry *parent;
16395 + struct au_hinode *hdir;
16396 + struct vfsmount *h_mnt;
16398 + /* temporary unlock/relock for copyup */
16399 + struct dentry *h_dentry, *h_parent;
16400 + struct au_branch *br;
16401 + struct task_struct *task;
16404 +void au_pin_hdir_unlock(struct au_pin *p);
16405 +int au_pin_hdir_lock(struct au_pin *p);
16406 +int au_pin_hdir_relock(struct au_pin *p);
16407 +void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task);
16408 +void au_pin_hdir_acquire_nest(struct au_pin *p);
16409 +void au_pin_hdir_release(struct au_pin *p);
16411 +/* ---------------------------------------------------------------------- */
16413 +static inline struct au_iinfo *au_ii(struct inode *inode)
16415 + struct au_iinfo *iinfo;
16417 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
16418 + if (iinfo->ii_hinode)
16420 + return NULL; /* debugging bad_inode case */
16423 +/* ---------------------------------------------------------------------- */
16426 +struct inode *au_igrab(struct inode *inode);
16427 +void au_refresh_iop(struct inode *inode, int force_getattr);
16428 +int au_refresh_hinode_self(struct inode *inode);
16429 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
16430 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
16431 + unsigned int d_type, ino_t *ino);
16432 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
16433 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
16434 + struct inode *inode);
16435 +int au_test_h_perm(struct inode *h_inode, int mask);
16436 +int au_test_h_perm_sio(struct inode *h_inode, int mask);
16438 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
16439 + ino_t h_ino, unsigned int d_type, ino_t *ino)
16441 +#ifdef CONFIG_AUFS_SHWH
16442 + return au_ino(sb, bindex, h_ino, d_type, ino);
16455 +extern struct inode_operations aufs_iop[AuIop_Last],
16456 + aufs_iop_nogetattr[AuIop_Last];
16458 +/* au_wr_dir flags */
16459 +#define AuWrDir_ADD_ENTRY 1
16460 +#define AuWrDir_ISDIR (1 << 1)
16461 +#define AuWrDir_TMPFILE (1 << 2)
16462 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
16463 +#define au_fset_wrdir(flags, name) \
16464 + do { (flags) |= AuWrDir_##name; } while (0)
16465 +#define au_fclr_wrdir(flags, name) \
16466 + do { (flags) &= ~AuWrDir_##name; } while (0)
16468 +struct au_wr_dir_args {
16469 + aufs_bindex_t force_btgt;
16470 + unsigned char flags;
16472 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
16473 + struct au_wr_dir_args *args);
16475 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
16476 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
16477 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
16478 + unsigned int udba, unsigned char flags);
16479 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
16480 + unsigned int udba, unsigned char flags) __must_check;
16481 +int au_do_pin(struct au_pin *pin) __must_check;
16482 +void au_unpin(struct au_pin *pin);
16483 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
16485 +#define AuIcpup_DID_CPUP 1
16486 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
16487 +#define au_fset_icpup(flags, name) \
16488 + do { (flags) |= AuIcpup_##name; } while (0)
16489 +#define au_fclr_icpup(flags, name) \
16490 + do { (flags) &= ~AuIcpup_##name; } while (0)
16492 +struct au_icpup_args {
16493 + unsigned char flags;
16494 + unsigned char pin_flags;
16495 + aufs_bindex_t btgt;
16496 + unsigned int udba;
16497 + struct au_pin pin;
16498 + struct path h_path;
16499 + struct inode *h_inode;
16502 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
16503 + struct au_icpup_args *a);
16505 +int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path);
16508 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
16509 + struct dentry *h_parent, int isdir);
16510 +int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
16512 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname);
16513 +int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
16515 +struct vfsub_aopen_args;
16516 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
16517 + struct vfsub_aopen_args *args);
16518 +int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode);
16519 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
16520 + struct dentry *dentry);
16521 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
16524 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
16525 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
16526 + struct dentry *h_parent, int isdir);
16527 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
16528 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
16531 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
16532 +int aufs_rename(struct inode *src_dir, struct dentry *src_dentry,
16533 + struct inode *dir, struct dentry *dentry);
16536 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
16537 +void au_hiput(struct au_hinode *hinode);
16538 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
16539 + struct dentry *h_wh);
16540 +unsigned int au_hi_flags(struct inode *inode, int isdir);
16542 +/* hinode flags */
16543 +#define AuHi_XINO 1
16544 +#define AuHi_HNOTIFY (1 << 1)
16545 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
16546 +#define au_fset_hi(flags, name) \
16547 + do { (flags) |= AuHi_##name; } while (0)
16548 +#define au_fclr_hi(flags, name) \
16549 + do { (flags) &= ~AuHi_##name; } while (0)
16551 +#ifndef CONFIG_AUFS_HNOTIFY
16552 +#undef AuHi_HNOTIFY
16553 +#define AuHi_HNOTIFY 0
16556 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
16557 + struct inode *h_inode, unsigned int flags);
16559 +void au_update_iigen(struct inode *inode, int half);
16560 +void au_update_ibrange(struct inode *inode, int do_put_zero);
16562 +void au_icntnr_init_once(void *_c);
16563 +int au_iinfo_init(struct inode *inode);
16564 +void au_iinfo_fin(struct inode *inode);
16565 +int au_ii_realloc(struct au_iinfo *iinfo, int nbr);
16567 +#ifdef CONFIG_PROC_FS
16569 +int au_plink_maint(struct super_block *sb, int flags);
16571 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
16572 +int au_plink_maint_enter(struct super_block *sb);
16573 +#ifdef CONFIG_AUFS_DEBUG
16574 +void au_plink_list(struct super_block *sb);
16576 +AuStubVoid(au_plink_list, struct super_block *sb)
16578 +int au_plink_test(struct inode *inode);
16579 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
16580 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
16581 + struct dentry *h_dentry);
16582 +void au_plink_put(struct super_block *sb, int verbose);
16583 +void au_plink_clean(struct super_block *sb, int verbose);
16584 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
16586 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
16587 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
16588 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
16589 +AuStubVoid(au_plink_list, struct super_block *sb);
16590 +AuStubInt0(au_plink_test, struct inode *inode);
16591 +AuStub(struct dentry *, au_plink_lkup, return NULL,
16592 + struct inode *inode, aufs_bindex_t bindex);
16593 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
16594 + struct dentry *h_dentry);
16595 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
16596 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
16597 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
16598 +#endif /* CONFIG_PROC_FS */
16600 +#ifdef CONFIG_AUFS_XATTR
16602 +int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
16603 + unsigned int verbose);
16604 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
16605 +ssize_t aufs_getxattr(struct dentry *dentry, const char *name, void *value,
16607 +int aufs_setxattr(struct dentry *dentry, const char *name, const void *value,
16608 + size_t size, int flags);
16609 +int aufs_removexattr(struct dentry *dentry, const char *name);
16611 +/* void au_xattr_init(struct super_block *sb); */
16613 +AuStubInt0(au_cpup_xattr, struct dentry *h_dst, struct dentry *h_src,
16614 + int ignore_flags, unsigned int verbose);
16615 +/* AuStubVoid(au_xattr_init, struct super_block *sb); */
16618 +#ifdef CONFIG_FS_POSIX_ACL
16619 +struct posix_acl *aufs_get_acl(struct inode *inode, int type);
16620 +int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
16623 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
16630 +struct au_srxattr {
16634 + const char *name;
16635 + const void *value;
16640 + const char *name;
16643 + struct posix_acl *acl;
16648 +ssize_t au_srxattr(struct dentry *dentry, struct au_srxattr *arg);
16651 +/* ---------------------------------------------------------------------- */
16653 +/* lock subclass for iinfo */
16655 + AuLsc_II_CHILD, /* child first */
16656 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
16657 + AuLsc_II_CHILD3, /* copyup dirs */
16658 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
16659 + AuLsc_II_PARENT2,
16660 + AuLsc_II_PARENT3, /* copyup dirs */
16661 + AuLsc_II_NEW_CHILD
16665 + * ii_read_lock_child, ii_write_lock_child,
16666 + * ii_read_lock_child2, ii_write_lock_child2,
16667 + * ii_read_lock_child3, ii_write_lock_child3,
16668 + * ii_read_lock_parent, ii_write_lock_parent,
16669 + * ii_read_lock_parent2, ii_write_lock_parent2,
16670 + * ii_read_lock_parent3, ii_write_lock_parent3,
16671 + * ii_read_lock_new_child, ii_write_lock_new_child,
16673 +#define AuReadLockFunc(name, lsc) \
16674 +static inline void ii_read_lock_##name(struct inode *i) \
16676 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
16679 +#define AuWriteLockFunc(name, lsc) \
16680 +static inline void ii_write_lock_##name(struct inode *i) \
16682 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
16685 +#define AuRWLockFuncs(name, lsc) \
16686 + AuReadLockFunc(name, lsc) \
16687 + AuWriteLockFunc(name, lsc)
16689 +AuRWLockFuncs(child, CHILD);
16690 +AuRWLockFuncs(child2, CHILD2);
16691 +AuRWLockFuncs(child3, CHILD3);
16692 +AuRWLockFuncs(parent, PARENT);
16693 +AuRWLockFuncs(parent2, PARENT2);
16694 +AuRWLockFuncs(parent3, PARENT3);
16695 +AuRWLockFuncs(new_child, NEW_CHILD);
16697 +#undef AuReadLockFunc
16698 +#undef AuWriteLockFunc
16699 +#undef AuRWLockFuncs
16702 + * ii_read_unlock, ii_write_unlock, ii_downgrade_lock
16704 +AuSimpleUnlockRwsemFuncs(ii, struct inode *i, &au_ii(i)->ii_rwsem);
16706 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
16707 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
16708 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
16710 +/* ---------------------------------------------------------------------- */
16712 +static inline void au_icntnr_init(struct au_icntnr *c)
16714 +#ifdef CONFIG_AUFS_DEBUG
16715 + c->vfs_inode.i_mode = 0;
16719 +static inline unsigned int au_iigen(struct inode *inode, struct au_iigen *iigen_arg)
16721 + unsigned int gen;
16722 + struct au_iinfo *iinfo;
16723 + struct au_iigen *iigen;
16725 + iinfo = au_ii(inode);
16726 + iigen = &iinfo->ii_generation;
16727 + spin_lock(&iigen->ig_spin);
16729 + *iigen_arg = *iigen;
16730 + gen = iigen->ig_generation;
16731 + spin_unlock(&iigen->ig_spin);
16736 +/* tiny test for inode number */
16737 +/* tmpfs generation is too rough */
16738 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
16740 + struct au_iinfo *iinfo;
16742 + iinfo = au_ii(inode);
16743 + AuRwMustAnyLock(&iinfo->ii_rwsem);
16744 + return !(iinfo->ii_hsb1 == h_inode->i_sb
16745 + && iinfo->ii_higen == h_inode->i_generation);
16748 +static inline void au_iigen_dec(struct inode *inode)
16750 + struct au_iinfo *iinfo;
16751 + struct au_iigen *iigen;
16753 + iinfo = au_ii(inode);
16754 + iigen = &iinfo->ii_generation;
16755 + spin_lock(&iigen->ig_spin);
16756 + iigen->ig_generation--;
16757 + spin_unlock(&iigen->ig_spin);
16760 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
16765 + if (unlikely(inode && au_iigen(inode, NULL) != sigen))
16771 +/* ---------------------------------------------------------------------- */
16773 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
16774 + aufs_bindex_t bindex)
16776 + IiMustAnyLock(inode);
16777 + return au_ii(inode)->ii_hinode[0 + bindex].hi_id;
16780 +static inline aufs_bindex_t au_ibstart(struct inode *inode)
16782 + IiMustAnyLock(inode);
16783 + return au_ii(inode)->ii_bstart;
16786 +static inline aufs_bindex_t au_ibend(struct inode *inode)
16788 + IiMustAnyLock(inode);
16789 + return au_ii(inode)->ii_bend;
16792 +static inline struct au_vdir *au_ivdir(struct inode *inode)
16794 + IiMustAnyLock(inode);
16795 + return au_ii(inode)->ii_vdir;
16798 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
16800 + IiMustAnyLock(inode);
16801 + return au_ii(inode)->ii_hinode[0 + bindex].hi_whdentry;
16804 +static inline void au_set_ibstart(struct inode *inode, aufs_bindex_t bindex)
16806 + IiMustWriteLock(inode);
16807 + au_ii(inode)->ii_bstart = bindex;
16810 +static inline void au_set_ibend(struct inode *inode, aufs_bindex_t bindex)
16812 + IiMustWriteLock(inode);
16813 + au_ii(inode)->ii_bend = bindex;
16816 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
16818 + IiMustWriteLock(inode);
16819 + au_ii(inode)->ii_vdir = vdir;
16822 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
16824 + IiMustAnyLock(inode);
16825 + return au_ii(inode)->ii_hinode + bindex;
16828 +/* ---------------------------------------------------------------------- */
16830 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
16833 + return pin->parent;
16837 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
16839 + if (pin && pin->hdir)
16840 + return pin->hdir->hi_inode;
16844 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
16847 + return pin->hdir;
16851 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
16854 + pin->dentry = dentry;
16857 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
16858 + unsigned char lflag)
16862 + au_fset_pin(pin->flags, DI_LOCKED);
16864 + au_fclr_pin(pin->flags, DI_LOCKED);
16868 +#if 0 /* reserved */
16869 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
16872 + dput(pin->parent);
16873 + pin->parent = dget(parent);
16878 +/* ---------------------------------------------------------------------- */
16881 +#ifdef CONFIG_AUFS_HNOTIFY
16882 +struct au_hnotify_op {
16883 + void (*ctl)(struct au_hinode *hinode, int do_set);
16884 + int (*alloc)(struct au_hinode *hinode);
16887 + * if it returns true, the the caller should free hinode->hi_notify,
16888 + * otherwise ->free() frees it.
16890 + int (*free)(struct au_hinode *hinode,
16891 + struct au_hnotify *hn) __must_check;
16893 + void (*fin)(void);
16894 + int (*init)(void);
16896 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
16897 + void (*fin_br)(struct au_branch *br);
16898 + int (*init_br)(struct au_branch *br, int perm);
16902 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
16903 +void au_hn_free(struct au_hinode *hinode);
16904 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
16905 +void au_hn_reset(struct inode *inode, unsigned int flags);
16906 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
16907 + struct qstr *h_child_qstr, struct inode *h_child_inode);
16908 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
16909 +int au_hnotify_init_br(struct au_branch *br, int perm);
16910 +void au_hnotify_fin_br(struct au_branch *br);
16911 +int __init au_hnotify_init(void);
16912 +void au_hnotify_fin(void);
16915 +extern const struct au_hnotify_op au_hnotify_op;
16918 +void au_hn_init(struct au_hinode *hinode)
16920 + hinode->hi_notify = NULL;
16923 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
16925 + return hinode->hi_notify;
16929 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
16930 + struct au_hinode *hinode __maybe_unused,
16931 + struct inode *inode __maybe_unused)
16932 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
16933 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
16934 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
16935 + int do_set __maybe_unused)
16936 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
16937 + unsigned int flags __maybe_unused)
16938 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
16939 + struct au_branch *br __maybe_unused,
16940 + int perm __maybe_unused)
16941 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
16942 + int perm __maybe_unused)
16943 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
16944 +AuStubInt0(__init au_hnotify_init, void)
16945 +AuStubVoid(au_hnotify_fin, void)
16946 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
16947 +#endif /* CONFIG_AUFS_HNOTIFY */
16949 +static inline void au_hn_suspend(struct au_hinode *hdir)
16951 + au_hn_ctl(hdir, /*do_set*/0);
16954 +static inline void au_hn_resume(struct au_hinode *hdir)
16956 + au_hn_ctl(hdir, /*do_set*/1);
16959 +static inline void au_hn_imtx_lock(struct au_hinode *hdir)
16961 + mutex_lock(&hdir->hi_inode->i_mutex);
16962 + au_hn_suspend(hdir);
16965 +static inline void au_hn_imtx_lock_nested(struct au_hinode *hdir,
16966 + unsigned int sc __maybe_unused)
16968 + mutex_lock_nested(&hdir->hi_inode->i_mutex, sc);
16969 + au_hn_suspend(hdir);
16972 +static inline void au_hn_imtx_unlock(struct au_hinode *hdir)
16974 + au_hn_resume(hdir);
16975 + mutex_unlock(&hdir->hi_inode->i_mutex);
16978 +#endif /* __KERNEL__ */
16979 +#endif /* __AUFS_INODE_H__ */
16980 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
16981 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
16982 +++ linux/fs/aufs/ioctl.c 2016-01-13 20:11:11.669760262 +0100
16985 + * Copyright (C) 2005-2015 Junjiro R. Okajima
16987 + * This program, aufs is free software; you can redistribute it and/or modify
16988 + * it under the terms of the GNU General Public License as published by
16989 + * the Free Software Foundation; either version 2 of the License, or
16990 + * (at your option) any later version.
16992 + * This program is distributed in the hope that it will be useful,
16993 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16994 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16995 + * GNU General Public License for more details.
16997 + * You should have received a copy of the GNU General Public License
16998 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17003 + * plink-management and readdir in userspace.
17004 + * assist the pathconf(3) wrapper library.
17006 + * File-based Hierarchical Storage Management.
17009 +#include <linux/compat.h>
17010 +#include <linux/file.h>
17013 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
17016 + aufs_bindex_t wbi, bindex, bend;
17017 + struct file *h_file;
17018 + struct super_block *sb;
17019 + struct dentry *root;
17020 + struct au_branch *br;
17021 + struct aufs_wbr_fd wbrfd = {
17022 + .oflags = au_dir_roflags,
17025 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
17026 + | O_NOATIME | O_CLOEXEC;
17028 + AuDebugOn(wbrfd.oflags & ~valid);
17031 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
17032 + if (unlikely(err)) {
17038 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
17039 + wbrfd.oflags |= au_dir_roflags;
17040 + AuDbg("0%o\n", wbrfd.oflags);
17041 + if (unlikely(wbrfd.oflags & ~valid))
17045 + fd = get_unused_fd_flags(0);
17047 + if (unlikely(fd < 0))
17050 + h_file = ERR_PTR(-EINVAL);
17053 + sb = path->dentry->d_sb;
17054 + root = sb->s_root;
17055 + aufs_read_lock(root, AuLock_IR);
17056 + bend = au_sbend(sb);
17057 + if (wbrfd.brid >= 0) {
17058 + wbi = au_br_index(sb, wbrfd.brid);
17059 + if (unlikely(wbi < 0 || wbi > bend))
17063 + h_file = ERR_PTR(-ENOENT);
17064 + br = au_sbr(sb, wbi);
17065 + if (!au_br_writable(br->br_perm)) {
17069 + bindex = wbi + 1;
17071 + for (; bindex <= bend; bindex++) {
17072 + br = au_sbr(sb, bindex);
17073 + if (au_br_writable(br->br_perm)) {
17075 + br = au_sbr(sb, wbi);
17080 + AuDbg("wbi %d\n", wbi);
17082 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
17086 + aufs_read_unlock(root, AuLock_IR);
17087 + err = PTR_ERR(h_file);
17088 + if (IS_ERR(h_file))
17091 + atomic_dec(&br->br_count); /* cf. au_h_open() */
17092 + fd_install(fd, h_file);
17094 + goto out; /* success */
17097 + put_unused_fd(fd);
17103 +/* ---------------------------------------------------------------------- */
17105 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
17108 + struct dentry *dentry;
17111 + case AUFS_CTL_RDU:
17112 + case AUFS_CTL_RDU_INO:
17113 + err = au_rdu_ioctl(file, cmd, arg);
17116 + case AUFS_CTL_WBR_FD:
17117 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
17120 + case AUFS_CTL_IBUSY:
17121 + err = au_ibusy_ioctl(file, arg);
17124 + case AUFS_CTL_BRINFO:
17125 + err = au_brinfo_ioctl(file, arg);
17128 + case AUFS_CTL_FHSM_FD:
17129 + dentry = file->f_path.dentry;
17130 + if (IS_ROOT(dentry))
17131 + err = au_fhsm_fd(dentry->d_sb, arg);
17137 + /* do not call the lower */
17138 + AuDbg("0x%x\n", cmd);
17146 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
17151 + case AUFS_CTL_MVDOWN:
17152 + err = au_mvdown(file->f_path.dentry, (void __user *)arg);
17155 + case AUFS_CTL_WBR_FD:
17156 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
17160 + /* do not call the lower */
17161 + AuDbg("0x%x\n", cmd);
17169 +#ifdef CONFIG_COMPAT
17170 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
17171 + unsigned long arg)
17176 + case AUFS_CTL_RDU:
17177 + case AUFS_CTL_RDU_INO:
17178 + err = au_rdu_compat_ioctl(file, cmd, arg);
17181 + case AUFS_CTL_IBUSY:
17182 + err = au_ibusy_compat_ioctl(file, arg);
17185 + case AUFS_CTL_BRINFO:
17186 + err = au_brinfo_compat_ioctl(file, arg);
17190 + err = aufs_ioctl_dir(file, cmd, arg);
17197 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
17198 + unsigned long arg)
17200 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
17203 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
17204 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
17205 +++ linux/fs/aufs/i_op_add.c 2016-01-13 20:11:11.669760262 +0100
17208 + * Copyright (C) 2005-2015 Junjiro R. Okajima
17210 + * This program, aufs is free software; you can redistribute it and/or modify
17211 + * it under the terms of the GNU General Public License as published by
17212 + * the Free Software Foundation; either version 2 of the License, or
17213 + * (at your option) any later version.
17215 + * This program is distributed in the hope that it will be useful,
17216 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17217 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17218 + * GNU General Public License for more details.
17220 + * You should have received a copy of the GNU General Public License
17221 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17225 + * inode operations (add entry)
17231 + * final procedure of adding a new entry, except link(2).
17232 + * remove whiteout, instantiate, copyup the parent dir's times and size
17233 + * and update version.
17234 + * if it failed, re-create the removed whiteout.
17236 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
17237 + struct dentry *wh_dentry, struct dentry *dentry)
17240 + aufs_bindex_t bwh;
17241 + struct path h_path;
17242 + struct super_block *sb;
17243 + struct inode *inode, *h_dir;
17244 + struct dentry *wh;
17249 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
17250 + IMustLock(h_dir);
17251 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
17252 + bwh = au_dbwh(dentry);
17253 + h_path.dentry = wh_dentry;
17254 + h_path.mnt = au_sbr_mnt(sb, bindex);
17255 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
17257 + if (unlikely(err))
17261 + inode = au_new_inode(dentry, /*must_new*/1);
17262 + if (!IS_ERR(inode)) {
17263 + d_instantiate(dentry, inode);
17264 + dir = d_inode(dentry->d_parent); /* dir inode is locked */
17266 + au_dir_ts(dir, bindex);
17267 + dir->i_version++;
17268 + au_fhsm_wrote(sb, bindex, /*force*/0);
17269 + return 0; /* success */
17272 + err = PTR_ERR(inode);
17277 + /* dir inode is locked */
17278 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
17279 + rerr = PTR_ERR(wh);
17280 + if (IS_ERR(wh)) {
17281 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
17282 + dentry, err, rerr);
17291 +static int au_d_may_add(struct dentry *dentry)
17296 + if (unlikely(d_unhashed(dentry)))
17298 + if (unlikely(d_really_is_positive(dentry)))
17304 + * simple tests for the adding inode operations.
17305 + * following the checks in vfs, plus the parent-child relationship.
17307 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
17308 + struct dentry *h_parent, int isdir)
17312 + struct dentry *h_dentry;
17313 + struct inode *h_inode;
17315 + err = -ENAMETOOLONG;
17316 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
17319 + h_dentry = au_h_dptr(dentry, bindex);
17320 + if (d_really_is_negative(dentry)) {
17322 + if (unlikely(d_is_positive(h_dentry)))
17325 + /* rename(2) case */
17327 + if (unlikely(d_is_negative(h_dentry)))
17329 + h_inode = d_inode(h_dentry);
17330 + if (unlikely(!h_inode->i_nlink))
17333 + h_mode = h_inode->i_mode;
17336 + if (unlikely(S_ISDIR(h_mode)))
17338 + } else if (unlikely(!S_ISDIR(h_mode))) {
17345 + /* expected parent dir is locked */
17346 + if (unlikely(h_parent != h_dentry->d_parent))
17355 + * initial procedure of adding a new entry.
17356 + * prepare writable branch and the parent dir, lock it,
17357 + * and lookup whiteout for the new entry.
17359 +static struct dentry*
17360 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
17361 + struct dentry *src_dentry, struct au_pin *pin,
17362 + struct au_wr_dir_args *wr_dir_args)
17364 + struct dentry *wh_dentry, *h_parent;
17365 + struct super_block *sb;
17366 + struct au_branch *br;
17368 + unsigned int udba;
17369 + aufs_bindex_t bcpup;
17371 + AuDbg("%pd\n", dentry);
17373 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
17375 + wh_dentry = ERR_PTR(err);
17376 + if (unlikely(err < 0))
17379 + sb = dentry->d_sb;
17380 + udba = au_opt_udba(sb);
17381 + err = au_pin(pin, dentry, bcpup, udba,
17382 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17383 + wh_dentry = ERR_PTR(err);
17384 + if (unlikely(err))
17387 + h_parent = au_pinned_h_parent(pin);
17388 + if (udba != AuOpt_UDBA_NONE
17389 + && au_dbstart(dentry) == bcpup)
17390 + err = au_may_add(dentry, bcpup, h_parent,
17391 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
17392 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
17393 + err = -ENAMETOOLONG;
17394 + wh_dentry = ERR_PTR(err);
17395 + if (unlikely(err))
17398 + br = au_sbr(sb, bcpup);
17400 + struct path tmp = {
17401 + .dentry = h_parent,
17402 + .mnt = au_br_mnt(br)
17404 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
17407 + wh_dentry = NULL;
17408 + if (bcpup != au_dbwh(dentry))
17409 + goto out; /* success */
17412 + * ENAMETOOLONG here means that if we allowed create such name, then it
17413 + * would not be able to removed in the future. So we don't allow such
17414 + * name here and we don't handle ENAMETOOLONG differently here.
17416 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
17419 + if (IS_ERR(wh_dentry))
17422 + return wh_dentry;
17425 +/* ---------------------------------------------------------------------- */
17427 +enum { Mknod, Symlink, Creat };
17428 +struct simple_arg {
17435 + struct vfsub_aopen_args *aopen;
17438 + const char *symname;
17447 +static int add_simple(struct inode *dir, struct dentry *dentry,
17448 + struct simple_arg *arg)
17451 + aufs_bindex_t bstart;
17452 + unsigned char created;
17453 + const unsigned char try_aopen
17454 + = (arg->type == Creat && arg->u.c.try_aopen);
17455 + struct dentry *wh_dentry, *parent;
17456 + struct inode *h_dir;
17457 + struct super_block *sb;
17458 + struct au_branch *br;
17459 + /* to reuduce stack size */
17461 + struct au_dtime dt;
17462 + struct au_pin pin;
17463 + struct path h_path;
17464 + struct au_wr_dir_args wr_dir_args;
17467 + AuDbg("%pd\n", dentry);
17471 + a = kmalloc(sizeof(*a), GFP_NOFS);
17472 + if (unlikely(!a))
17474 + a->wr_dir_args.force_btgt = -1;
17475 + a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
17477 + parent = dentry->d_parent; /* dir inode is locked */
17478 + if (!try_aopen) {
17479 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
17480 + if (unlikely(err))
17483 + err = au_d_may_add(dentry);
17484 + if (unlikely(err))
17487 + di_write_lock_parent(parent);
17488 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
17489 + &a->pin, &a->wr_dir_args);
17490 + err = PTR_ERR(wh_dentry);
17491 + if (IS_ERR(wh_dentry))
17494 + bstart = au_dbstart(dentry);
17495 + sb = dentry->d_sb;
17496 + br = au_sbr(sb, bstart);
17497 + a->h_path.dentry = au_h_dptr(dentry, bstart);
17498 + a->h_path.mnt = au_br_mnt(br);
17499 + h_dir = au_pinned_h_dir(&a->pin);
17500 + switch (arg->type) {
17503 + if (!try_aopen || !h_dir->i_op->atomic_open)
17504 + err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
17505 + arg->u.c.want_excl);
17507 + err = vfsub_atomic_open(h_dir, a->h_path.dentry,
17508 + arg->u.c.aopen, br);
17511 + err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
17514 + err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
17522 + err = epilog(dir, bstart, wh_dentry, dentry);
17525 + if (unlikely(created && err && d_is_positive(a->h_path.dentry))) {
17526 + /* no delegation since it is just created */
17527 + rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
17530 + AuIOErr("%pd revert failure(%d, %d)\n",
17531 + dentry, err, rerr);
17534 + au_dtime_revert(&a->dt);
17537 + if (!err && try_aopen && !h_dir->i_op->atomic_open)
17538 + *arg->u.c.aopen->opened |= FILE_CREATED;
17540 + au_unpin(&a->pin);
17545 + di_write_unlock(parent);
17547 + if (unlikely(err)) {
17548 + au_update_dbstart(dentry);
17552 + aufs_read_unlock(dentry, AuLock_DW);
17559 +int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
17562 + struct simple_arg arg = {
17569 + return add_simple(dir, dentry, &arg);
17572 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
17574 + struct simple_arg arg = {
17576 + .u.s.symname = symname
17578 + return add_simple(dir, dentry, &arg);
17581 +int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
17584 + struct simple_arg arg = {
17588 + .want_excl = want_excl
17591 + return add_simple(dir, dentry, &arg);
17594 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
17595 + struct vfsub_aopen_args *aopen_args)
17597 + struct simple_arg arg = {
17600 + .mode = aopen_args->create_mode,
17601 + .want_excl = aopen_args->open_flag & O_EXCL,
17602 + .try_aopen = true,
17603 + .aopen = aopen_args
17606 + return add_simple(dir, dentry, &arg);
17609 +int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
17612 + aufs_bindex_t bindex;
17613 + struct super_block *sb;
17614 + struct dentry *parent, *h_parent, *h_dentry;
17615 + struct inode *h_dir, *inode;
17616 + struct vfsmount *h_mnt;
17617 + struct au_wr_dir_args wr_dir_args = {
17618 + .force_btgt = -1,
17619 + .flags = AuWrDir_TMPFILE
17622 + /* copy-up may happen */
17623 + mutex_lock(&dir->i_mutex);
17626 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
17627 + if (unlikely(err))
17630 + err = au_di_init(dentry);
17631 + if (unlikely(err))
17635 + parent = d_find_any_alias(dir);
17636 + AuDebugOn(!parent);
17637 + di_write_lock_parent(parent);
17638 + if (unlikely(d_inode(parent) != dir))
17641 + err = au_digen_test(parent, au_sigen(sb));
17642 + if (unlikely(err))
17645 + bindex = au_dbstart(parent);
17646 + au_set_dbstart(dentry, bindex);
17647 + au_set_dbend(dentry, bindex);
17648 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
17650 + if (unlikely(err < 0))
17653 + err = -EOPNOTSUPP;
17654 + h_dir = au_h_iptr(dir, bindex);
17655 + if (unlikely(!h_dir->i_op->tmpfile))
17658 + h_mnt = au_sbr_mnt(sb, bindex);
17659 + err = vfsub_mnt_want_write(h_mnt);
17660 + if (unlikely(err))
17663 + h_parent = au_h_dptr(parent, bindex);
17664 + err = inode_permission(d_inode(h_parent), MAY_WRITE | MAY_EXEC);
17665 + if (unlikely(err))
17669 + h_dentry = d_alloc(h_parent, &dentry->d_name);
17670 + if (unlikely(!h_dentry))
17673 + err = h_dir->i_op->tmpfile(h_dir, h_dentry, mode);
17674 + if (unlikely(err))
17677 + au_set_dbstart(dentry, bindex);
17678 + au_set_dbend(dentry, bindex);
17679 + au_set_h_dptr(dentry, bindex, dget(h_dentry));
17680 + inode = au_new_inode(dentry, /*must_new*/1);
17681 + if (IS_ERR(inode)) {
17682 + err = PTR_ERR(inode);
17683 + au_set_h_dptr(dentry, bindex, NULL);
17684 + au_set_dbstart(dentry, -1);
17685 + au_set_dbend(dentry, -1);
17687 + if (!inode->i_nlink)
17688 + set_nlink(inode, 1);
17689 + d_tmpfile(dentry, inode);
17690 + au_di(dentry)->di_tmpfile = 1;
17692 + /* update without i_mutex */
17693 + if (au_ibstart(dir) == au_dbstart(dentry))
17694 + au_cpup_attr_timesizes(dir);
17700 + vfsub_mnt_drop_write(h_mnt);
17702 + di_write_unlock(parent);
17704 + di_write_unlock(dentry);
17707 + /* verbose coding for lock class name */
17708 + au_rw_class(&au_di(dentry)->di_rwsem,
17709 + au_lc_key + AuLcNonDir_DIINFO);
17714 + au_di_fin(dentry);
17715 + dentry->d_fsdata = NULL;
17718 + si_read_unlock(sb);
17720 + mutex_unlock(&dir->i_mutex);
17724 +/* ---------------------------------------------------------------------- */
17726 +struct au_link_args {
17727 + aufs_bindex_t bdst, bsrc;
17728 + struct au_pin pin;
17729 + struct path h_path;
17730 + struct dentry *src_parent, *parent;
17733 +static int au_cpup_before_link(struct dentry *src_dentry,
17734 + struct au_link_args *a)
17737 + struct dentry *h_src_dentry;
17738 + struct au_cp_generic cpg = {
17739 + .dentry = src_dentry,
17744 + .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
17747 + di_read_lock_parent(a->src_parent, AuLock_IR);
17748 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
17749 + if (unlikely(err))
17752 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
17753 + err = au_pin(&a->pin, src_dentry, a->bdst,
17754 + au_opt_udba(src_dentry->d_sb),
17755 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17756 + if (unlikely(err))
17759 + err = au_sio_cpup_simple(&cpg);
17760 + au_unpin(&a->pin);
17763 + di_read_unlock(a->src_parent, AuLock_IR);
17767 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
17768 + struct au_link_args *a)
17771 + unsigned char plink;
17772 + aufs_bindex_t bend;
17773 + struct dentry *h_src_dentry;
17774 + struct inode *h_inode, *inode, *delegated;
17775 + struct super_block *sb;
17776 + struct file *h_file;
17780 + sb = src_dentry->d_sb;
17781 + inode = d_inode(src_dentry);
17782 + if (au_ibstart(inode) <= a->bdst)
17783 + h_inode = au_h_iptr(inode, a->bdst);
17784 + if (!h_inode || !h_inode->i_nlink) {
17785 + /* copyup src_dentry as the name of dentry. */
17786 + bend = au_dbend(dentry);
17787 + if (bend < a->bsrc)
17788 + au_set_dbend(dentry, a->bsrc);
17789 + au_set_h_dptr(dentry, a->bsrc,
17790 + dget(au_h_dptr(src_dentry, a->bsrc)));
17791 + dget(a->h_path.dentry);
17792 + au_set_h_dptr(dentry, a->bdst, NULL);
17793 + AuDbg("temporary d_inode...\n");
17794 + spin_lock(&dentry->d_lock);
17795 + dentry->d_inode = d_inode(src_dentry); /* tmp */
17796 + spin_unlock(&dentry->d_lock);
17797 + h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
17798 + if (IS_ERR(h_file))
17799 + err = PTR_ERR(h_file);
17801 + struct au_cp_generic cpg = {
17802 + .dentry = dentry,
17807 + .flags = AuCpup_KEEPLINO
17809 + err = au_sio_cpup_simple(&cpg);
17810 + au_h_open_post(dentry, a->bsrc, h_file);
17812 + dput(a->h_path.dentry);
17813 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
17815 + au_set_h_dptr(dentry, a->bdst,
17816 + a->h_path.dentry);
17818 + spin_lock(&dentry->d_lock);
17819 + dentry->d_inode = NULL; /* restore */
17820 + spin_unlock(&dentry->d_lock);
17821 + AuDbg("temporary d_inode...done\n");
17822 + au_set_h_dptr(dentry, a->bsrc, NULL);
17823 + au_set_dbend(dentry, bend);
17825 + /* the inode of src_dentry already exists on a.bdst branch */
17826 + h_src_dentry = d_find_alias(h_inode);
17827 + if (!h_src_dentry && au_plink_test(inode)) {
17829 + h_src_dentry = au_plink_lkup(inode, a->bdst);
17830 + err = PTR_ERR(h_src_dentry);
17831 + if (IS_ERR(h_src_dentry))
17834 + if (unlikely(d_is_negative(h_src_dentry))) {
17835 + dput(h_src_dentry);
17836 + h_src_dentry = NULL;
17840 + if (h_src_dentry) {
17841 + delegated = NULL;
17842 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
17843 + &a->h_path, &delegated);
17844 + if (unlikely(err == -EWOULDBLOCK)) {
17845 + pr_warn("cannot retry for NFSv4 delegation"
17846 + " for an internal link\n");
17849 + dput(h_src_dentry);
17851 + AuIOErr("no dentry found for hi%lu on b%d\n",
17852 + h_inode->i_ino, a->bdst);
17857 + if (!err && !plink)
17858 + au_plink_append(inode, a->bdst, a->h_path.dentry);
17865 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
17866 + struct dentry *dentry)
17869 + struct au_dtime dt;
17870 + struct au_link_args *a;
17871 + struct dentry *wh_dentry, *h_src_dentry;
17872 + struct inode *inode, *delegated;
17873 + struct super_block *sb;
17874 + struct au_wr_dir_args wr_dir_args = {
17875 + /* .force_btgt = -1, */
17876 + .flags = AuWrDir_ADD_ENTRY
17880 + inode = d_inode(src_dentry);
17881 + IMustLock(inode);
17884 + a = kzalloc(sizeof(*a), GFP_NOFS);
17885 + if (unlikely(!a))
17888 + a->parent = dentry->d_parent; /* dir inode is locked */
17889 + err = aufs_read_and_write_lock2(dentry, src_dentry,
17890 + AuLock_NOPLM | AuLock_GEN);
17891 + if (unlikely(err))
17893 + err = au_d_linkable(src_dentry);
17894 + if (unlikely(err))
17896 + err = au_d_may_add(dentry);
17897 + if (unlikely(err))
17900 + a->src_parent = dget_parent(src_dentry);
17901 + wr_dir_args.force_btgt = au_ibstart(inode);
17903 + di_write_lock_parent(a->parent);
17904 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
17905 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
17907 + err = PTR_ERR(wh_dentry);
17908 + if (IS_ERR(wh_dentry))
17912 + sb = dentry->d_sb;
17913 + a->bdst = au_dbstart(dentry);
17914 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
17915 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
17916 + a->bsrc = au_ibstart(inode);
17917 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
17918 + if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
17919 + h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
17920 + if (!h_src_dentry) {
17921 + a->bsrc = au_dbstart(src_dentry);
17922 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
17923 + AuDebugOn(!h_src_dentry);
17924 + } else if (IS_ERR(h_src_dentry)) {
17925 + err = PTR_ERR(h_src_dentry);
17929 + if (au_opt_test(au_mntflags(sb), PLINK)) {
17930 + if (a->bdst < a->bsrc
17931 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
17932 + err = au_cpup_or_link(src_dentry, dentry, a);
17934 + delegated = NULL;
17935 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
17936 + &a->h_path, &delegated);
17937 + if (unlikely(err == -EWOULDBLOCK)) {
17938 + pr_warn("cannot retry for NFSv4 delegation"
17939 + " for an internal link\n");
17943 + dput(h_src_dentry);
17946 + * copyup src_dentry to the branch we process,
17947 + * and then link(2) to it.
17949 + dput(h_src_dentry);
17950 + if (a->bdst < a->bsrc
17951 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
17952 + au_unpin(&a->pin);
17953 + di_write_unlock(a->parent);
17954 + err = au_cpup_before_link(src_dentry, a);
17955 + di_write_lock_parent(a->parent);
17957 + err = au_pin(&a->pin, dentry, a->bdst,
17959 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17960 + if (unlikely(err))
17964 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
17966 + if (h_src_dentry && d_is_positive(h_src_dentry)) {
17967 + delegated = NULL;
17968 + err = vfsub_link(h_src_dentry,
17969 + au_pinned_h_dir(&a->pin),
17970 + &a->h_path, &delegated);
17971 + if (unlikely(err == -EWOULDBLOCK)) {
17972 + pr_warn("cannot retry"
17973 + " for NFSv4 delegation"
17974 + " for an internal link\n");
17980 + if (unlikely(err))
17984 + a->h_path.dentry = wh_dentry;
17985 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
17987 + if (unlikely(err))
17991 + au_dir_ts(dir, a->bdst);
17992 + dir->i_version++;
17993 + inc_nlink(inode);
17994 + inode->i_ctime = dir->i_ctime;
17995 + d_instantiate(dentry, au_igrab(inode));
17996 + if (d_unhashed(a->h_path.dentry))
17997 + /* some filesystem calls d_drop() */
17999 + /* some filesystems consume an inode even hardlink */
18000 + au_fhsm_wrote(sb, a->bdst, /*force*/0);
18001 + goto out_unpin; /* success */
18004 + /* no delegation since it is just created */
18005 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
18006 + /*delegated*/NULL, /*force*/0);
18007 + if (unlikely(rerr)) {
18008 + AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
18011 + au_dtime_revert(&dt);
18013 + au_unpin(&a->pin);
18017 + di_write_unlock(a->parent);
18018 + dput(a->src_parent);
18020 + if (unlikely(err)) {
18021 + au_update_dbstart(dentry);
18024 + aufs_read_and_write_unlock2(dentry, src_dentry);
18032 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
18035 + aufs_bindex_t bindex;
18036 + unsigned char diropq;
18037 + struct path h_path;
18038 + struct dentry *wh_dentry, *parent, *opq_dentry;
18039 + struct mutex *h_mtx;
18040 + struct super_block *sb;
18042 + struct au_pin pin;
18043 + struct au_dtime dt;
18044 + } *a; /* reduce the stack usage */
18045 + struct au_wr_dir_args wr_dir_args = {
18046 + .force_btgt = -1,
18047 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
18053 + a = kmalloc(sizeof(*a), GFP_NOFS);
18054 + if (unlikely(!a))
18057 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
18058 + if (unlikely(err))
18060 + err = au_d_may_add(dentry);
18061 + if (unlikely(err))
18064 + parent = dentry->d_parent; /* dir inode is locked */
18065 + di_write_lock_parent(parent);
18066 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
18067 + &a->pin, &wr_dir_args);
18068 + err = PTR_ERR(wh_dentry);
18069 + if (IS_ERR(wh_dentry))
18072 + sb = dentry->d_sb;
18073 + bindex = au_dbstart(dentry);
18074 + h_path.dentry = au_h_dptr(dentry, bindex);
18075 + h_path.mnt = au_sbr_mnt(sb, bindex);
18076 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
18077 + if (unlikely(err))
18080 + /* make the dir opaque */
18082 + h_mtx = &d_inode(h_path.dentry)->i_mutex;
18084 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
18085 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18086 + opq_dentry = au_diropq_create(dentry, bindex);
18087 + mutex_unlock(h_mtx);
18088 + err = PTR_ERR(opq_dentry);
18089 + if (IS_ERR(opq_dentry))
18091 + dput(opq_dentry);
18095 + err = epilog(dir, bindex, wh_dentry, dentry);
18098 + goto out_unpin; /* success */
18103 + AuLabel(revert opq);
18104 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18105 + rerr = au_diropq_remove(dentry, bindex);
18106 + mutex_unlock(h_mtx);
18108 + AuIOErr("%pd reverting diropq failed(%d, %d)\n",
18109 + dentry, err, rerr);
18115 + AuLabel(revert dir);
18116 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
18118 + AuIOErr("%pd reverting dir failed(%d, %d)\n",
18119 + dentry, err, rerr);
18122 + au_dtime_revert(&a->dt);
18124 + au_unpin(&a->pin);
18127 + di_write_unlock(parent);
18129 + if (unlikely(err)) {
18130 + au_update_dbstart(dentry);
18133 + aufs_read_unlock(dentry, AuLock_DW);
18139 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
18140 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
18141 +++ linux/fs/aufs/i_op.c 2016-01-13 20:11:11.669760262 +0100
18144 + * Copyright (C) 2005-2015 Junjiro R. Okajima
18146 + * This program, aufs is free software; you can redistribute it and/or modify
18147 + * it under the terms of the GNU General Public License as published by
18148 + * the Free Software Foundation; either version 2 of the License, or
18149 + * (at your option) any later version.
18151 + * This program is distributed in the hope that it will be useful,
18152 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18153 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18154 + * GNU General Public License for more details.
18156 + * You should have received a copy of the GNU General Public License
18157 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18161 + * inode operations (except add/del/rename)
18164 +#include <linux/device_cgroup.h>
18165 +#include <linux/fs_stack.h>
18166 +#include <linux/namei.h>
18167 +#include <linux/security.h>
18170 +static int h_permission(struct inode *h_inode, int mask,
18171 + struct path *h_path, int brperm)
18174 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
18177 + if ((write_mask && IS_IMMUTABLE(h_inode))
18178 + || ((mask & MAY_EXEC)
18179 + && S_ISREG(h_inode->i_mode)
18180 + && (path_noexec(h_path)
18181 + || !(h_inode->i_mode & S_IXUGO))))
18185 + * - skip the lower fs test in the case of write to ro branch.
18186 + * - nfs dir permission write check is optimized, but a policy for
18187 + * link/rename requires a real check.
18188 + * - nfs always sets MS_POSIXACL regardless its mount option 'noacl.'
18189 + * in this case, generic_permission() returns -EOPNOTSUPP.
18191 + if ((write_mask && !au_br_writable(brperm))
18192 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
18193 + && write_mask && !(mask & MAY_READ))
18194 + || !h_inode->i_op->permission) {
18195 + /* AuLabel(generic_permission); */
18196 + /* AuDbg("get_acl %pf\n", h_inode->i_op->get_acl); */
18197 + err = generic_permission(h_inode, mask);
18198 + if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
18199 + err = h_inode->i_op->permission(h_inode, mask);
18202 + /* AuLabel(h_inode->permission); */
18203 + err = h_inode->i_op->permission(h_inode, mask);
18208 + err = devcgroup_inode_permission(h_inode, mask);
18210 + err = security_inode_permission(h_inode, mask);
18214 + /* todo: do we need to call ima_path_check()? */
18215 + struct path h_path = {
18219 + err = ima_path_check(&h_path,
18220 + mask & (MAY_READ | MAY_WRITE | MAY_EXEC),
18221 + IMA_COUNT_LEAVE);
18229 +static int aufs_permission(struct inode *inode, int mask)
18232 + aufs_bindex_t bindex, bend;
18233 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
18234 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
18235 + struct inode *h_inode;
18236 + struct super_block *sb;
18237 + struct au_branch *br;
18239 + /* todo: support rcu-walk? */
18240 + if (mask & MAY_NOT_BLOCK)
18243 + sb = inode->i_sb;
18244 + si_read_lock(sb, AuLock_FLUSH);
18245 + ii_read_lock_child(inode);
18247 + err = au_iigen_test(inode, au_sigen(sb));
18248 + if (unlikely(err))
18254 + || au_opt_test(au_mntflags(sb), DIRPERM1)) {
18255 + err = au_busy_or_stale();
18256 + h_inode = au_h_iptr(inode, au_ibstart(inode));
18257 + if (unlikely(!h_inode
18258 + || (h_inode->i_mode & S_IFMT)
18259 + != (inode->i_mode & S_IFMT)))
18263 + bindex = au_ibstart(inode);
18264 + br = au_sbr(sb, bindex);
18265 + err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
18268 + && !special_file(h_inode->i_mode)) {
18269 + /* test whether the upper writable branch exists */
18271 + for (; bindex >= 0; bindex--)
18272 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
18280 + /* non-write to dir */
18282 + bend = au_ibend(inode);
18283 + for (bindex = au_ibstart(inode); !err && bindex <= bend; bindex++) {
18284 + h_inode = au_h_iptr(inode, bindex);
18286 + err = au_busy_or_stale();
18287 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
18290 + br = au_sbr(sb, bindex);
18291 + err = h_permission(h_inode, mask, &br->br_path,
18297 + ii_read_unlock(inode);
18298 + si_read_unlock(sb);
18302 +/* ---------------------------------------------------------------------- */
18304 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
18305 + unsigned int flags)
18307 + struct dentry *ret, *parent;
18308 + struct inode *inode;
18309 + struct super_block *sb;
18310 + int err, npositive;
18314 + /* todo: support rcu-walk? */
18315 + ret = ERR_PTR(-ECHILD);
18316 + if (flags & LOOKUP_RCU)
18319 + ret = ERR_PTR(-ENAMETOOLONG);
18320 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
18324 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
18325 + ret = ERR_PTR(err);
18326 + if (unlikely(err))
18329 + err = au_di_init(dentry);
18330 + ret = ERR_PTR(err);
18331 + if (unlikely(err))
18335 + npositive = 0; /* suppress a warning */
18336 + parent = dentry->d_parent; /* dir inode is locked */
18337 + di_read_lock_parent(parent, AuLock_IR);
18338 + err = au_alive_dir(parent);
18340 + err = au_digen_test(parent, au_sigen(sb));
18342 + npositive = au_lkup_dentry(dentry, au_dbstart(parent),
18346 + di_read_unlock(parent, AuLock_IR);
18347 + ret = ERR_PTR(err);
18348 + if (unlikely(err < 0))
18352 + inode = au_new_inode(dentry, /*must_new*/0);
18353 + if (IS_ERR(inode)) {
18354 + ret = (void *)inode;
18361 + atomic_inc(&inode->i_count);
18362 + ret = d_splice_alias(inode, dentry);
18364 + if (unlikely(d_need_lookup(dentry))) {
18365 + spin_lock(&dentry->d_lock);
18366 + dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
18367 + spin_unlock(&dentry->d_lock);
18371 + if (!IS_ERR(ret)) {
18373 + if (ret && ret != dentry)
18374 + ii_write_unlock(inode);
18376 + ii_write_unlock(inode);
18383 + di_write_unlock(dentry);
18385 + /* verbose coding for lock class name */
18386 + if (unlikely(S_ISLNK(inode->i_mode)))
18387 + au_rw_class(&au_di(dentry)->di_rwsem,
18388 + au_lc_key + AuLcSymlink_DIINFO);
18389 + else if (unlikely(S_ISDIR(inode->i_mode)))
18390 + au_rw_class(&au_di(dentry)->di_rwsem,
18391 + au_lc_key + AuLcDir_DIINFO);
18392 + else /* likely */
18393 + au_rw_class(&au_di(dentry)->di_rwsem,
18394 + au_lc_key + AuLcNonDir_DIINFO);
18397 + si_read_unlock(sb);
18402 +/* ---------------------------------------------------------------------- */
18404 +struct aopen_node {
18405 + struct hlist_node hlist;
18406 + struct file *file, *h_file;
18409 +static int au_do_aopen(struct inode *inode, struct file *file)
18411 + struct au_sphlhead *aopen;
18412 + struct aopen_node *node;
18413 + struct au_do_open_args args = {
18415 + .open = au_do_open_nondir
18418 + aopen = &au_sbi(inode->i_sb)->si_aopen;
18419 + spin_lock(&aopen->spin);
18420 + hlist_for_each_entry(node, &aopen->head, hlist)
18421 + if (node->file == file) {
18422 + args.h_file = node->h_file;
18425 + spin_unlock(&aopen->spin);
18426 + /* AuDebugOn(!args.h_file); */
18428 + return au_do_open(file, &args);
18431 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
18432 + struct file *file, unsigned int open_flag,
18433 + umode_t create_mode, int *opened)
18435 + int err, h_opened = *opened;
18436 + struct dentry *parent;
18437 + struct dentry *d;
18438 + struct au_sphlhead *aopen;
18439 + struct vfsub_aopen_args args = {
18440 + .open_flag = open_flag,
18441 + .create_mode = create_mode,
18442 + .opened = &h_opened
18444 + struct aopen_node aopen_node = {
18449 + AuDbg("open_flag 0x%x\n", open_flag);
18450 + AuDbgDentry(dentry);
18453 + if (!au_di(dentry)) {
18454 + d = aufs_lookup(dir, dentry, /*flags*/0);
18456 + err = PTR_ERR(d);
18460 + * obsoleted dentry found.
18461 + * another error will be returned later.
18467 + AuDbgDentry(dentry);
18470 + if (d_is_positive(dentry)
18471 + || d_unhashed(dentry)
18472 + || d_unlinked(dentry)
18473 + || !(open_flag & O_CREAT))
18474 + goto out_no_open;
18476 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
18477 + if (unlikely(err))
18480 + parent = dentry->d_parent; /* dir is locked */
18481 + di_write_lock_parent(parent);
18482 + err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
18483 + if (unlikely(err))
18486 + AuDbgDentry(dentry);
18487 + if (d_is_positive(dentry))
18490 + args.file = get_empty_filp();
18491 + err = PTR_ERR(args.file);
18492 + if (IS_ERR(args.file))
18495 + args.file->f_flags = file->f_flags;
18496 + err = au_aopen_or_create(dir, dentry, &args);
18498 + AuDbgFile(args.file);
18499 + if (unlikely(err < 0)) {
18500 + if (h_opened & FILE_OPENED)
18503 + put_filp(args.file);
18507 + /* some filesystems don't set FILE_CREATED while succeeded? */
18508 + *opened |= FILE_CREATED;
18509 + if (h_opened & FILE_OPENED)
18510 + aopen_node.h_file = args.file;
18512 + put_filp(args.file);
18513 + args.file = NULL;
18515 + aopen = &au_sbi(dir->i_sb)->si_aopen;
18516 + au_sphl_add(&aopen_node.hlist, aopen);
18517 + err = finish_open(file, dentry, au_do_aopen, opened);
18518 + au_sphl_del(&aopen_node.hlist, aopen);
18521 + if (aopen_node.h_file)
18522 + fput(aopen_node.h_file);
18525 + di_write_unlock(parent);
18526 + aufs_read_unlock(dentry, AuLock_DW);
18527 + AuDbgDentry(dentry);
18528 + if (unlikely(err))
18531 + if (!err && !(*opened & FILE_CREATED)) {
18532 + AuLabel(out_no_open);
18534 + err = finish_no_open(file, dentry);
18537 + AuDbg("%pd%s%s\n", dentry,
18538 + (*opened & FILE_CREATED) ? " created" : "",
18539 + (*opened & FILE_OPENED) ? " opened" : "");
18545 +/* ---------------------------------------------------------------------- */
18547 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
18548 + const unsigned char add_entry, aufs_bindex_t bcpup,
18549 + aufs_bindex_t bstart)
18552 + struct dentry *h_parent;
18553 + struct inode *h_dir;
18556 + IMustLock(d_inode(parent));
18558 + di_write_lock_parent(parent);
18561 + if (!au_h_dptr(parent, bcpup)) {
18562 + if (bstart > bcpup)
18563 + err = au_cpup_dirs(dentry, bcpup);
18564 + else if (bstart < bcpup)
18565 + err = au_cpdown_dirs(dentry, bcpup);
18569 + if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
18570 + h_parent = au_h_dptr(parent, bcpup);
18571 + h_dir = d_inode(h_parent);
18572 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
18573 + err = au_lkup_neg(dentry, bcpup, /*wh*/0);
18574 + /* todo: no unlock here */
18575 + mutex_unlock(&h_dir->i_mutex);
18577 + AuDbg("bcpup %d\n", bcpup);
18579 + if (d_really_is_negative(dentry))
18580 + au_set_h_dptr(dentry, bstart, NULL);
18581 + au_update_dbrange(dentry, /*do_put_zero*/0);
18586 + di_write_unlock(parent);
18588 + err = bcpup; /* success */
18595 + * decide the branch and the parent dir where we will create a new entry.
18596 + * returns new bindex or an error.
18597 + * copyup the parent dir if needed.
18599 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
18600 + struct au_wr_dir_args *args)
18603 + unsigned int flags;
18604 + aufs_bindex_t bcpup, bstart, src_bstart;
18605 + const unsigned char add_entry
18606 + = au_ftest_wrdir(args->flags, ADD_ENTRY)
18607 + | au_ftest_wrdir(args->flags, TMPFILE);
18608 + struct super_block *sb;
18609 + struct dentry *parent;
18610 + struct au_sbinfo *sbinfo;
18612 + sb = dentry->d_sb;
18613 + sbinfo = au_sbi(sb);
18614 + parent = dget_parent(dentry);
18615 + bstart = au_dbstart(dentry);
18617 + if (args->force_btgt < 0) {
18618 + if (src_dentry) {
18619 + src_bstart = au_dbstart(src_dentry);
18620 + if (src_bstart < bstart)
18621 + bcpup = src_bstart;
18622 + } else if (add_entry) {
18624 + if (au_ftest_wrdir(args->flags, ISDIR))
18625 + au_fset_wbr(flags, DIR);
18626 + err = AuWbrCreate(sbinfo, dentry, flags);
18630 + if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
18632 + err = AuWbrCopyup(sbinfo, dentry);
18634 + if (!IS_ROOT(dentry)) {
18635 + di_read_lock_parent(parent, !AuLock_IR);
18636 + err = AuWbrCopyup(sbinfo, dentry);
18637 + di_read_unlock(parent, !AuLock_IR);
18639 + err = AuWbrCopyup(sbinfo, dentry);
18642 + if (unlikely(err < 0))
18646 + bcpup = args->force_btgt;
18647 + AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
18650 + AuDbg("bstart %d, bcpup %d\n", bstart, bcpup);
18652 + if (bcpup == bstart)
18653 + goto out; /* success */
18655 + /* copyup the new parent into the branch we process */
18656 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, bstart);
18658 + if (d_really_is_negative(dentry)) {
18659 + au_set_h_dptr(dentry, bstart, NULL);
18660 + au_set_dbstart(dentry, bcpup);
18661 + au_set_dbend(dentry, bcpup);
18663 + AuDebugOn(add_entry
18664 + && !au_ftest_wrdir(args->flags, TMPFILE)
18665 + && !au_h_dptr(dentry, bcpup));
18673 +/* ---------------------------------------------------------------------- */
18675 +void au_pin_hdir_unlock(struct au_pin *p)
18678 + au_hn_imtx_unlock(p->hdir);
18681 +int au_pin_hdir_lock(struct au_pin *p)
18689 + /* even if an error happens later, keep this lock */
18690 + au_hn_imtx_lock_nested(p->hdir, p->lsc_hi);
18693 + if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
18698 + err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
18699 + p->h_parent, p->br);
18705 +int au_pin_hdir_relock(struct au_pin *p)
18708 + struct inode *h_i;
18709 + struct dentry *h_d[] = {
18714 + err = au_pin_hdir_lock(p);
18715 + if (unlikely(err))
18718 + for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
18721 + if (d_is_positive(h_d[i])) {
18722 + h_i = d_inode(h_d[i]);
18723 + err = !h_i->i_nlink;
18731 +void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
18733 +#if defined(CONFIG_DEBUG_MUTEXES) || defined(CONFIG_SMP)
18734 + p->hdir->hi_inode->i_mutex.owner = task;
18738 +void au_pin_hdir_acquire_nest(struct au_pin *p)
18741 + mutex_acquire_nest(&p->hdir->hi_inode->i_mutex.dep_map,
18742 + p->lsc_hi, 0, NULL, _RET_IP_);
18743 + au_pin_hdir_set_owner(p, current);
18747 +void au_pin_hdir_release(struct au_pin *p)
18750 + au_pin_hdir_set_owner(p, p->task);
18751 + mutex_release(&p->hdir->hi_inode->i_mutex.dep_map, 1, _RET_IP_);
18755 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
18757 + if (pin && pin->parent)
18758 + return au_h_dptr(pin->parent, pin->bindex);
18762 +void au_unpin(struct au_pin *p)
18765 + au_pin_hdir_unlock(p);
18766 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
18767 + vfsub_mnt_drop_write(p->h_mnt);
18771 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18772 + di_read_unlock(p->parent, AuLock_IR);
18773 + iput(p->hdir->hi_inode);
18775 + p->parent = NULL;
18778 + /* do not clear p->task */
18781 +int au_do_pin(struct au_pin *p)
18784 + struct super_block *sb;
18785 + struct inode *h_dir;
18788 + sb = p->dentry->d_sb;
18789 + p->br = au_sbr(sb, p->bindex);
18790 + if (IS_ROOT(p->dentry)) {
18791 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
18792 + p->h_mnt = au_br_mnt(p->br);
18793 + err = vfsub_mnt_want_write(p->h_mnt);
18794 + if (unlikely(err)) {
18795 + au_fclr_pin(p->flags, MNT_WRITE);
18802 + p->h_dentry = NULL;
18803 + if (p->bindex <= au_dbend(p->dentry))
18804 + p->h_dentry = au_h_dptr(p->dentry, p->bindex);
18806 + p->parent = dget_parent(p->dentry);
18807 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18808 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
18811 + p->h_parent = au_h_dptr(p->parent, p->bindex);
18812 + p->hdir = au_hi(d_inode(p->parent), p->bindex);
18814 + h_dir = p->hdir->hi_inode;
18818 + * if DI_LOCKED is not set, then p->parent may be different
18819 + * and h_parent can be NULL.
18821 + if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
18823 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18824 + di_read_unlock(p->parent, AuLock_IR);
18826 + p->parent = NULL;
18830 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
18831 + p->h_mnt = au_br_mnt(p->br);
18832 + err = vfsub_mnt_want_write(p->h_mnt);
18833 + if (unlikely(err)) {
18834 + au_fclr_pin(p->flags, MNT_WRITE);
18835 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18836 + di_read_unlock(p->parent, AuLock_IR);
18838 + p->parent = NULL;
18844 + err = au_pin_hdir_lock(p);
18846 + goto out; /* success */
18851 + pr_err("err %d\n", err);
18852 + err = au_busy_or_stale();
18857 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
18858 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
18859 + unsigned int udba, unsigned char flags)
18861 + p->dentry = dentry;
18863 + p->lsc_di = lsc_di;
18864 + p->lsc_hi = lsc_hi;
18865 + p->flags = flags;
18866 + p->bindex = bindex;
18868 + p->parent = NULL;
18872 + p->h_dentry = NULL;
18873 + p->h_parent = NULL;
18875 + p->task = current;
18878 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
18879 + unsigned int udba, unsigned char flags)
18881 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
18883 + return au_do_pin(pin);
18886 +/* ---------------------------------------------------------------------- */
18889 + * ->setattr() and ->getattr() are called in various cases.
18890 + * chmod, stat: dentry is revalidated.
18891 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
18893 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
18895 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
18896 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
18899 + struct dentry *parent;
18902 + if (au_digen_test(dentry, sigen)) {
18903 + parent = dget_parent(dentry);
18904 + di_read_lock_parent(parent, AuLock_IR);
18905 + err = au_refresh_dentry(dentry, parent);
18906 + di_read_unlock(parent, AuLock_IR);
18914 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
18915 + struct au_icpup_args *a)
18919 + aufs_bindex_t bstart, ibstart;
18920 + struct dentry *hi_wh, *parent;
18921 + struct inode *inode;
18922 + struct au_wr_dir_args wr_dir_args = {
18923 + .force_btgt = -1,
18927 + if (d_is_dir(dentry))
18928 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
18929 + /* plink or hi_wh() case */
18930 + bstart = au_dbstart(dentry);
18931 + inode = d_inode(dentry);
18932 + ibstart = au_ibstart(inode);
18933 + if (bstart != ibstart && !au_test_ro(inode->i_sb, ibstart, inode))
18934 + wr_dir_args.force_btgt = ibstart;
18935 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
18936 + if (unlikely(err < 0))
18939 + if (err != bstart)
18940 + au_fset_icpup(a->flags, DID_CPUP);
18943 + a->pin_flags = AuPin_MNT_WRITE;
18945 + if (!IS_ROOT(dentry)) {
18946 + au_fset_pin(a->pin_flags, DI_LOCKED);
18947 + parent = dget_parent(dentry);
18948 + di_write_lock_parent(parent);
18951 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
18952 + if (unlikely(err))
18955 + a->h_path.dentry = au_h_dptr(dentry, bstart);
18957 + a->h_inode = d_inode(a->h_path.dentry);
18958 + if (ia && (ia->ia_valid & ATTR_SIZE)) {
18959 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
18960 + if (ia->ia_size < i_size_read(a->h_inode))
18961 + sz = ia->ia_size;
18962 + mutex_unlock(&a->h_inode->i_mutex);
18966 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
18967 + hi_wh = au_hi_wh(inode, a->btgt);
18969 + struct au_cp_generic cpg = {
18970 + .dentry = dentry,
18976 + err = au_sio_cpup_wh(&cpg, /*file*/NULL);
18977 + if (unlikely(err))
18979 + hi_wh = au_hi_wh(inode, a->btgt);
18980 + /* todo: revalidate hi_wh? */
18985 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
18986 + di_downgrade_lock(parent, AuLock_IR);
18990 + if (!au_ftest_icpup(a->flags, DID_CPUP))
18991 + goto out; /* success */
18993 + if (!d_unhashed(dentry)) {
18994 + struct au_cp_generic cpg = {
18995 + .dentry = dentry,
19000 + .flags = AuCpup_DTIME | AuCpup_HOPEN
19002 + err = au_sio_cpup_simple(&cpg);
19004 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
19005 + } else if (!hi_wh)
19006 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
19008 + a->h_path.dentry = hi_wh; /* do not dget here */
19011 + a->h_inode = d_inode(a->h_path.dentry);
19013 + goto out; /* success */
19014 + au_unpin(&a->pin);
19017 + di_write_unlock(parent);
19022 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
19026 +static int aufs_setattr(struct dentry *dentry, struct iattr *ia)
19029 + struct inode *inode, *delegated;
19030 + struct super_block *sb;
19031 + struct file *file;
19032 + struct au_icpup_args *a;
19034 + inode = d_inode(dentry);
19035 + IMustLock(inode);
19038 + a = kzalloc(sizeof(*a), GFP_NOFS);
19039 + if (unlikely(!a))
19042 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
19043 + ia->ia_valid &= ~ATTR_MODE;
19046 + sb = dentry->d_sb;
19047 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19048 + if (unlikely(err))
19051 + if (ia->ia_valid & ATTR_FILE) {
19052 + /* currently ftruncate(2) only */
19053 + AuDebugOn(!d_is_reg(dentry));
19054 + file = ia->ia_file;
19055 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
19056 + if (unlikely(err))
19058 + ia->ia_file = au_hf_top(file);
19059 + a->udba = AuOpt_UDBA_NONE;
19061 + /* fchmod() doesn't pass ia_file */
19062 + a->udba = au_opt_udba(sb);
19063 + di_write_lock_child(dentry);
19064 + /* no d_unlinked(), to set UDBA_NONE for root */
19065 + if (d_unhashed(dentry))
19066 + a->udba = AuOpt_UDBA_NONE;
19067 + if (a->udba != AuOpt_UDBA_NONE) {
19068 + AuDebugOn(IS_ROOT(dentry));
19069 + err = au_reval_for_attr(dentry, au_sigen(sb));
19070 + if (unlikely(err))
19075 + err = au_pin_and_icpup(dentry, ia, a);
19076 + if (unlikely(err < 0))
19078 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
19079 + ia->ia_file = NULL;
19080 + ia->ia_valid &= ~ATTR_FILE;
19083 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
19084 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
19085 + == (ATTR_MODE | ATTR_CTIME)) {
19086 + err = security_path_chmod(&a->h_path, ia->ia_mode);
19087 + if (unlikely(err))
19089 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
19090 + && (ia->ia_valid & ATTR_CTIME)) {
19091 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
19092 + if (unlikely(err))
19096 + if (ia->ia_valid & ATTR_SIZE) {
19099 + if (ia->ia_size < i_size_read(inode))
19101 + truncate_setsize(inode, ia->ia_size);
19104 + if (ia->ia_valid & ATTR_FILE)
19106 + mutex_unlock(&a->h_inode->i_mutex);
19107 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
19108 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
19110 + delegated = NULL;
19112 + err = vfsub_notify_change(&a->h_path, ia, &delegated);
19114 + err = break_deleg_wait(&delegated);
19122 + au_cpup_attr_changeable(inode);
19125 + mutex_unlock(&a->h_inode->i_mutex);
19126 + au_unpin(&a->pin);
19127 + if (unlikely(err))
19128 + au_update_dbstart(dentry);
19130 + di_write_unlock(dentry);
19132 + fi_write_unlock(file);
19133 + ia->ia_file = file;
19134 + ia->ia_valid |= ATTR_FILE;
19137 + si_read_unlock(sb);
19145 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
19146 +static int au_h_path_to_set_attr(struct dentry *dentry,
19147 + struct au_icpup_args *a, struct path *h_path)
19150 + struct super_block *sb;
19152 + sb = dentry->d_sb;
19153 + a->udba = au_opt_udba(sb);
19154 + /* no d_unlinked(), to set UDBA_NONE for root */
19155 + if (d_unhashed(dentry))
19156 + a->udba = AuOpt_UDBA_NONE;
19157 + if (a->udba != AuOpt_UDBA_NONE) {
19158 + AuDebugOn(IS_ROOT(dentry));
19159 + err = au_reval_for_attr(dentry, au_sigen(sb));
19160 + if (unlikely(err))
19163 + err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
19164 + if (unlikely(err < 0))
19167 + h_path->dentry = a->h_path.dentry;
19168 + h_path->mnt = au_sbr_mnt(sb, a->btgt);
19174 +ssize_t au_srxattr(struct dentry *dentry, struct au_srxattr *arg)
19177 + struct path h_path;
19178 + struct super_block *sb;
19179 + struct au_icpup_args *a;
19180 + struct inode *inode, *h_inode;
19182 + inode = d_inode(dentry);
19183 + IMustLock(inode);
19186 + a = kzalloc(sizeof(*a), GFP_NOFS);
19187 + if (unlikely(!a))
19190 + sb = dentry->d_sb;
19191 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19192 + if (unlikely(err))
19195 + h_path.dentry = NULL; /* silence gcc */
19196 + di_write_lock_child(dentry);
19197 + err = au_h_path_to_set_attr(dentry, a, &h_path);
19198 + if (unlikely(err))
19201 + mutex_unlock(&a->h_inode->i_mutex);
19202 + switch (arg->type) {
19203 + case AU_XATTR_SET:
19204 + err = vfsub_setxattr(h_path.dentry,
19205 + arg->u.set.name, arg->u.set.value,
19206 + arg->u.set.size, arg->u.set.flags);
19208 + case AU_XATTR_REMOVE:
19209 + err = vfsub_removexattr(h_path.dentry, arg->u.remove.name);
19212 + err = -EOPNOTSUPP;
19213 + h_inode = d_inode(h_path.dentry);
19214 + if (h_inode->i_op->set_acl)
19215 + err = h_inode->i_op->set_acl(h_inode,
19216 + arg->u.acl_set.acl,
19217 + arg->u.acl_set.type);
19221 + au_cpup_attr_timesizes(inode);
19223 + au_unpin(&a->pin);
19224 + if (unlikely(err))
19225 + au_update_dbstart(dentry);
19228 + di_write_unlock(dentry);
19229 + si_read_unlock(sb);
19238 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
19239 + unsigned int nlink)
19243 + inode->i_mode = st->mode;
19244 + /* don't i_[ug]id_write() here */
19245 + inode->i_uid = st->uid;
19246 + inode->i_gid = st->gid;
19247 + inode->i_atime = st->atime;
19248 + inode->i_mtime = st->mtime;
19249 + inode->i_ctime = st->ctime;
19251 + au_cpup_attr_nlink(inode, /*force*/0);
19252 + if (S_ISDIR(inode->i_mode)) {
19253 + n = inode->i_nlink;
19256 + smp_mb(); /* for i_nlink */
19257 + /* 0 can happen */
19258 + set_nlink(inode, n);
19261 + spin_lock(&inode->i_lock);
19262 + inode->i_blocks = st->blocks;
19263 + i_size_write(inode, st->size);
19264 + spin_unlock(&inode->i_lock);
19268 + * common routine for aufs_getattr() and aufs_getxattr().
19269 + * returns zero or negative (an error).
19270 + * @dentry will be read-locked in success.
19272 +int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path)
19275 + unsigned int mnt_flags, sigen;
19276 + unsigned char udba_none;
19277 + aufs_bindex_t bindex;
19278 + struct super_block *sb, *h_sb;
19279 + struct inode *inode;
19281 + h_path->mnt = NULL;
19282 + h_path->dentry = NULL;
19285 + sb = dentry->d_sb;
19286 + mnt_flags = au_mntflags(sb);
19287 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
19289 + /* support fstat(2) */
19290 + if (!d_unlinked(dentry) && !udba_none) {
19291 + sigen = au_sigen(sb);
19292 + err = au_digen_test(dentry, sigen);
19294 + di_read_lock_child(dentry, AuLock_IR);
19295 + err = au_dbrange_test(dentry);
19296 + if (unlikely(err)) {
19297 + di_read_unlock(dentry, AuLock_IR);
19301 + AuDebugOn(IS_ROOT(dentry));
19302 + di_write_lock_child(dentry);
19303 + err = au_dbrange_test(dentry);
19305 + err = au_reval_for_attr(dentry, sigen);
19307 + di_downgrade_lock(dentry, AuLock_IR);
19309 + di_write_unlock(dentry);
19314 + di_read_lock_child(dentry, AuLock_IR);
19316 + inode = d_inode(dentry);
19317 + bindex = au_ibstart(inode);
19318 + h_path->mnt = au_sbr_mnt(sb, bindex);
19319 + h_sb = h_path->mnt->mnt_sb;
19321 + && !au_test_fs_bad_iattr(h_sb)
19323 + goto out; /* success */
19325 + if (au_dbstart(dentry) == bindex)
19326 + h_path->dentry = au_h_dptr(dentry, bindex);
19327 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
19328 + h_path->dentry = au_plink_lkup(inode, bindex);
19329 + if (IS_ERR(h_path->dentry))
19330 + /* pretending success */
19331 + h_path->dentry = NULL;
19333 + dput(h_path->dentry);
19340 +static int aufs_getattr(struct vfsmount *mnt __maybe_unused,
19341 + struct dentry *dentry, struct kstat *st)
19344 + unsigned char positive;
19345 + struct path h_path;
19346 + struct inode *inode;
19347 + struct super_block *sb;
19349 + inode = d_inode(dentry);
19350 + sb = dentry->d_sb;
19351 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19352 + if (unlikely(err))
19354 + err = au_h_path_getattr(dentry, /*force*/0, &h_path);
19355 + if (unlikely(err))
19357 + if (unlikely(!h_path.dentry))
19358 + /* illegally overlapped or something */
19359 + goto out_fill; /* pretending success */
19361 + positive = d_is_positive(h_path.dentry);
19363 + err = vfs_getattr(&h_path, st);
19366 + au_refresh_iattr(inode, st,
19367 + d_inode(h_path.dentry)->i_nlink);
19368 + goto out_fill; /* success */
19374 + generic_fillattr(inode, st);
19376 + di_read_unlock(dentry, AuLock_IR);
19378 + si_read_unlock(sb);
19384 +/* ---------------------------------------------------------------------- */
19388 + * - the number of symlinks is not so many.
19391 + * - sbinfo (instead of iinfo) contains an hlist of struct au_symlink.
19392 + * If iinfo contained the hlist, then it would be rather large waste of memory
19394 + * - struct au_symlink contains the necessary info for h_inode follow_link() and
19398 +struct au_symlink {
19400 + struct hlist_node hlist;
19401 + struct rcu_head rcu;
19404 + struct inode *h_inode;
19408 +static void au_symlink_add(struct super_block *sb, struct au_symlink *slink,
19409 + struct inode *h_inode, void *cookie)
19411 + struct au_sbinfo *sbinfo;
19414 + slink->h_inode = h_inode;
19415 + slink->h_cookie = cookie;
19416 + sbinfo = au_sbi(sb);
19417 + au_sphl_add(&slink->hlist, &sbinfo->si_symlink);
19420 +static void au_symlink_del(struct super_block *sb, struct au_symlink *slink)
19422 + struct au_sbinfo *sbinfo;
19424 + /* do not iput() within rcu */
19425 + iput(slink->h_inode);
19426 + slink->h_inode = NULL;
19427 + sbinfo = au_sbi(sb);
19428 + au_sphl_del_rcu(&slink->hlist, &sbinfo->si_symlink);
19429 + kfree_rcu(slink, rcu);
19432 +static const char *aufs_follow_link(struct dentry *dentry, void **cookie)
19435 + struct inode *inode, *h_inode;
19436 + struct dentry *h_dentry;
19437 + struct au_symlink *slink;
19439 + aufs_bindex_t bindex;
19441 + ret = NULL; /* suppress a warning */
19442 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
19443 + if (unlikely(err))
19446 + err = au_d_hashed_positive(dentry);
19447 + if (unlikely(err))
19451 + inode = d_inode(dentry);
19452 + bindex = au_ibstart(inode);
19453 + h_inode = au_h_iptr(inode, bindex);
19454 + if (unlikely(!h_inode->i_op->follow_link))
19458 + slink = kmalloc(sizeof(*slink), GFP_NOFS);
19459 + if (unlikely(!slink))
19464 + if (au_dbstart(dentry) <= bindex) {
19465 + h_dentry = au_h_dptr(dentry, bindex);
19470 + h_dentry = d_find_any_alias(h_inode);
19471 + if (IS_ERR(h_dentry)) {
19472 + err = PTR_ERR(h_dentry);
19476 + if (unlikely(!h_dentry))
19480 + AuDbg("%pf\n", h_inode->i_op->follow_link);
19481 + AuDbgDentry(h_dentry);
19482 + ret = h_inode->i_op->follow_link(h_dentry, cookie);
19485 + if (!IS_ERR_OR_NULL(ret)) {
19486 + au_symlink_add(inode->i_sb, slink, h_inode, *cookie);
19488 + AuDbg("slink %p\n", slink);
19489 + goto out_unlock; /* success */
19493 + slink->h_inode = NULL;
19494 + kfree_rcu(slink, rcu);
19496 + aufs_read_unlock(dentry, AuLock_IR);
19498 + if (unlikely(err))
19499 + ret = ERR_PTR(err);
19500 + AuTraceErrPtr(ret);
19504 +static void aufs_put_link(struct inode *inode, void *cookie)
19506 + struct au_symlink *slink;
19507 + struct inode *h_inode;
19510 + AuDbg("slink %p\n", slink);
19511 + h_inode = slink->h_inode;
19512 + AuDbg("%pf\n", h_inode->i_op->put_link);
19513 + AuDbgInode(h_inode);
19514 + if (h_inode->i_op->put_link)
19515 + h_inode->i_op->put_link(h_inode, slink->h_cookie);
19516 + au_symlink_del(inode->i_sb, slink);
19519 +/* ---------------------------------------------------------------------- */
19521 +static int aufs_update_time(struct inode *inode, struct timespec *ts, int flags)
19524 + struct super_block *sb;
19525 + struct inode *h_inode;
19527 + sb = inode->i_sb;
19528 + /* mmap_sem might be acquired already, cf. aufs_mmap() */
19530 + si_read_lock(sb, AuLock_FLUSH);
19531 + ii_write_lock_child(inode);
19533 + h_inode = au_h_iptr(inode, au_ibstart(inode));
19534 + err = vfsub_update_time(h_inode, ts, flags);
19537 + au_cpup_attr_timesizes(inode);
19538 + ii_write_unlock(inode);
19539 + si_read_unlock(sb);
19542 + if (!err && (flags & S_VERSION))
19543 + inode_inc_iversion(inode);
19548 +/* ---------------------------------------------------------------------- */
19550 +/* no getattr version will be set by module.c:aufs_init() */
19551 +struct inode_operations aufs_iop_nogetattr[AuIop_Last],
19553 + [AuIop_SYMLINK] = {
19554 + .permission = aufs_permission,
19555 +#ifdef CONFIG_FS_POSIX_ACL
19556 + .get_acl = aufs_get_acl,
19557 + .set_acl = aufs_set_acl, /* unsupport for symlink? */
19560 + .setattr = aufs_setattr,
19561 + .getattr = aufs_getattr,
19563 +#ifdef CONFIG_AUFS_XATTR
19564 + .setxattr = aufs_setxattr,
19565 + .getxattr = aufs_getxattr,
19566 + .listxattr = aufs_listxattr,
19567 + .removexattr = aufs_removexattr,
19570 + .readlink = generic_readlink,
19571 + .follow_link = aufs_follow_link,
19572 + .put_link = aufs_put_link,
19574 + /* .update_time = aufs_update_time */
19577 + .create = aufs_create,
19578 + .lookup = aufs_lookup,
19579 + .link = aufs_link,
19580 + .unlink = aufs_unlink,
19581 + .symlink = aufs_symlink,
19582 + .mkdir = aufs_mkdir,
19583 + .rmdir = aufs_rmdir,
19584 + .mknod = aufs_mknod,
19585 + .rename = aufs_rename,
19587 + .permission = aufs_permission,
19588 +#ifdef CONFIG_FS_POSIX_ACL
19589 + .get_acl = aufs_get_acl,
19590 + .set_acl = aufs_set_acl,
19593 + .setattr = aufs_setattr,
19594 + .getattr = aufs_getattr,
19596 +#ifdef CONFIG_AUFS_XATTR
19597 + .setxattr = aufs_setxattr,
19598 + .getxattr = aufs_getxattr,
19599 + .listxattr = aufs_listxattr,
19600 + .removexattr = aufs_removexattr,
19603 + .update_time = aufs_update_time,
19604 + .atomic_open = aufs_atomic_open,
19605 + .tmpfile = aufs_tmpfile
19607 + [AuIop_OTHER] = {
19608 + .permission = aufs_permission,
19609 +#ifdef CONFIG_FS_POSIX_ACL
19610 + .get_acl = aufs_get_acl,
19611 + .set_acl = aufs_set_acl,
19614 + .setattr = aufs_setattr,
19615 + .getattr = aufs_getattr,
19617 +#ifdef CONFIG_AUFS_XATTR
19618 + .setxattr = aufs_setxattr,
19619 + .getxattr = aufs_getxattr,
19620 + .listxattr = aufs_listxattr,
19621 + .removexattr = aufs_removexattr,
19624 + .update_time = aufs_update_time
19627 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
19628 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
19629 +++ linux/fs/aufs/i_op_del.c 2016-01-13 20:11:11.669760262 +0100
19632 + * Copyright (C) 2005-2015 Junjiro R. Okajima
19634 + * This program, aufs is free software; you can redistribute it and/or modify
19635 + * it under the terms of the GNU General Public License as published by
19636 + * the Free Software Foundation; either version 2 of the License, or
19637 + * (at your option) any later version.
19639 + * This program is distributed in the hope that it will be useful,
19640 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19641 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19642 + * GNU General Public License for more details.
19644 + * You should have received a copy of the GNU General Public License
19645 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19649 + * inode operations (del entry)
19655 + * decide if a new whiteout for @dentry is necessary or not.
19656 + * when it is necessary, prepare the parent dir for the upper branch whose
19657 + * branch index is @bcpup for creation. the actual creation of the whiteout will
19658 + * be done by caller.
19660 + * 0: wh is unnecessary
19661 + * plus: wh is necessary
19664 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
19666 + int need_wh, err;
19667 + aufs_bindex_t bstart;
19668 + struct super_block *sb;
19670 + sb = dentry->d_sb;
19671 + bstart = au_dbstart(dentry);
19672 + if (*bcpup < 0) {
19674 + if (au_test_ro(sb, bstart, d_inode(dentry))) {
19675 + err = AuWbrCopyup(au_sbi(sb), dentry);
19677 + if (unlikely(err < 0))
19681 + AuDebugOn(bstart < *bcpup
19682 + || au_test_ro(sb, *bcpup, d_inode(dentry)));
19683 + AuDbg("bcpup %d, bstart %d\n", *bcpup, bstart);
19685 + if (*bcpup != bstart) {
19686 + err = au_cpup_dirs(dentry, *bcpup);
19687 + if (unlikely(err))
19691 + struct au_dinfo *dinfo, *tmp;
19693 + need_wh = -ENOMEM;
19694 + dinfo = au_di(dentry);
19695 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
19697 + au_di_cp(tmp, dinfo);
19698 + au_di_swap(tmp, dinfo);
19699 + /* returns the number of positive dentries */
19700 + need_wh = au_lkup_dentry(dentry, bstart + 1, /*type*/0);
19701 + au_di_swap(tmp, dinfo);
19702 + au_rw_write_unlock(&tmp->di_rwsem);
19706 + AuDbg("need_wh %d\n", need_wh);
19714 + * simple tests for the del-entry operations.
19715 + * following the checks in vfs, plus the parent-child relationship.
19717 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
19718 + struct dentry *h_parent, int isdir)
19722 + struct dentry *h_dentry, *h_latest;
19723 + struct inode *h_inode;
19725 + h_dentry = au_h_dptr(dentry, bindex);
19726 + if (d_really_is_positive(dentry)) {
19728 + if (unlikely(d_is_negative(h_dentry)))
19730 + h_inode = d_inode(h_dentry);
19731 + if (unlikely(!h_inode->i_nlink))
19734 + h_mode = h_inode->i_mode;
19737 + if (unlikely(S_ISDIR(h_mode)))
19739 + } else if (unlikely(!S_ISDIR(h_mode))) {
19744 + /* rename(2) case */
19746 + if (unlikely(d_is_positive(h_dentry)))
19751 + /* expected parent dir is locked */
19752 + if (unlikely(h_parent != h_dentry->d_parent))
19757 + * rmdir a dir may break the consistency on some filesystem.
19758 + * let's try heavy test.
19761 + if (unlikely(!au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1)
19762 + && au_test_h_perm(d_inode(h_parent),
19763 + MAY_EXEC | MAY_WRITE)))
19766 + h_latest = au_sio_lkup_one(&dentry->d_name, h_parent);
19768 + if (IS_ERR(h_latest))
19770 + if (h_latest == h_dentry)
19779 + * decide the branch where we operate for @dentry. the branch index will be set
19780 + * @rbcpup. after diciding it, 'pin' it and store the timestamps of the parent
19781 + * dir for reverting.
19782 + * when a new whiteout is necessary, create it.
19784 +static struct dentry*
19785 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
19786 + struct au_dtime *dt, struct au_pin *pin)
19788 + struct dentry *wh_dentry;
19789 + struct super_block *sb;
19790 + struct path h_path;
19791 + int err, need_wh;
19792 + unsigned int udba;
19793 + aufs_bindex_t bcpup;
19795 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
19796 + wh_dentry = ERR_PTR(need_wh);
19797 + if (unlikely(need_wh < 0))
19800 + sb = dentry->d_sb;
19801 + udba = au_opt_udba(sb);
19803 + err = au_pin(pin, dentry, bcpup, udba,
19804 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
19805 + wh_dentry = ERR_PTR(err);
19806 + if (unlikely(err))
19809 + h_path.dentry = au_pinned_h_parent(pin);
19810 + if (udba != AuOpt_UDBA_NONE
19811 + && au_dbstart(dentry) == bcpup) {
19812 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
19813 + wh_dentry = ERR_PTR(err);
19814 + if (unlikely(err))
19818 + h_path.mnt = au_sbr_mnt(sb, bcpup);
19819 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
19820 + wh_dentry = NULL;
19822 + goto out; /* success, no need to create whiteout */
19824 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
19825 + if (IS_ERR(wh_dentry))
19828 + /* returns with the parent is locked and wh_dentry is dget-ed */
19829 + goto out; /* success */
19834 + return wh_dentry;
19838 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
19839 + * in order to be revertible and save time for removing many child whiteouts
19841 + * returns 1 when there are too many child whiteout and caller should remove
19842 + * them asynchronously. returns 0 when the number of children is enough small to
19843 + * remove now or the branch fs is a remote fs.
19844 + * otherwise return an error.
19846 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
19847 + struct au_nhash *whlist, struct inode *dir)
19849 + int rmdir_later, err, dirwh;
19850 + struct dentry *h_dentry;
19851 + struct super_block *sb;
19852 + struct inode *inode;
19854 + sb = dentry->d_sb;
19855 + SiMustAnyLock(sb);
19856 + h_dentry = au_h_dptr(dentry, bindex);
19857 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
19858 + if (unlikely(err))
19861 + /* stop monitoring */
19862 + inode = d_inode(dentry);
19863 + au_hn_free(au_hi(inode, bindex));
19865 + if (!au_test_fs_remote(h_dentry->d_sb)) {
19866 + dirwh = au_sbi(sb)->si_dirwh;
19867 + rmdir_later = (dirwh <= 1);
19868 + if (!rmdir_later)
19869 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
19872 + return rmdir_later;
19875 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
19876 + if (unlikely(err)) {
19877 + AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
19878 + h_dentry, bindex, err);
19888 + * final procedure for deleting a entry.
19889 + * maintain dentry and iattr.
19891 +static void epilog(struct inode *dir, struct dentry *dentry,
19892 + aufs_bindex_t bindex)
19894 + struct inode *inode;
19896 + inode = d_inode(dentry);
19898 + inode->i_ctime = dir->i_ctime;
19900 + au_dir_ts(dir, bindex);
19901 + dir->i_version++;
19905 + * when an error happened, remove the created whiteout and revert everything.
19907 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
19908 + aufs_bindex_t bwh, struct dentry *wh_dentry,
19909 + struct dentry *dentry, struct au_dtime *dt)
19912 + struct path h_path = {
19913 + .dentry = wh_dentry,
19914 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
19917 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
19919 + au_set_dbwh(dentry, bwh);
19920 + au_dtime_revert(dt);
19924 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
19928 +/* ---------------------------------------------------------------------- */
19930 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
19933 + aufs_bindex_t bwh, bindex, bstart;
19934 + struct inode *inode, *h_dir, *delegated;
19935 + struct dentry *parent, *wh_dentry;
19936 + /* to reuduce stack size */
19938 + struct au_dtime dt;
19939 + struct au_pin pin;
19940 + struct path h_path;
19946 + a = kmalloc(sizeof(*a), GFP_NOFS);
19947 + if (unlikely(!a))
19950 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
19951 + if (unlikely(err))
19953 + err = au_d_hashed_positive(dentry);
19954 + if (unlikely(err))
19956 + inode = d_inode(dentry);
19957 + IMustLock(inode);
19959 + if (unlikely(d_is_dir(dentry)))
19960 + goto out_unlock; /* possible? */
19962 + bstart = au_dbstart(dentry);
19963 + bwh = au_dbwh(dentry);
19965 + parent = dentry->d_parent; /* dir inode is locked */
19966 + di_write_lock_parent(parent);
19967 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
19969 + err = PTR_ERR(wh_dentry);
19970 + if (IS_ERR(wh_dentry))
19973 + a->h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
19974 + a->h_path.dentry = au_h_dptr(dentry, bstart);
19975 + dget(a->h_path.dentry);
19976 + if (bindex == bstart) {
19977 + h_dir = au_pinned_h_dir(&a->pin);
19978 + delegated = NULL;
19979 + err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
19980 + if (unlikely(err == -EWOULDBLOCK)) {
19981 + pr_warn("cannot retry for NFSv4 delegation"
19982 + " for an internal unlink\n");
19986 + /* dir inode is locked */
19987 + h_dir = d_inode(wh_dentry->d_parent);
19988 + IMustLock(h_dir);
19993 + vfsub_drop_nlink(inode);
19994 + epilog(dir, dentry, bindex);
19996 + /* update target timestamps */
19997 + if (bindex == bstart) {
19998 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
20000 + inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
20002 + /* todo: this timestamp may be reverted later */
20003 + inode->i_ctime = h_dir->i_ctime;
20004 + goto out_unpin; /* success */
20011 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
20018 + au_unpin(&a->pin);
20020 + dput(a->h_path.dentry);
20022 + di_write_unlock(parent);
20024 + aufs_read_unlock(dentry, AuLock_DW);
20031 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
20033 + int err, rmdir_later;
20034 + aufs_bindex_t bwh, bindex, bstart;
20035 + struct inode *inode;
20036 + struct dentry *parent, *wh_dentry, *h_dentry;
20037 + struct au_whtmp_rmdir *args;
20038 + /* to reuduce stack size */
20040 + struct au_dtime dt;
20041 + struct au_pin pin;
20047 + a = kmalloc(sizeof(*a), GFP_NOFS);
20048 + if (unlikely(!a))
20051 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
20052 + if (unlikely(err))
20054 + err = au_alive_dir(dentry);
20055 + if (unlikely(err))
20057 + inode = d_inode(dentry);
20058 + IMustLock(inode);
20060 + if (unlikely(!d_is_dir(dentry)))
20061 + goto out_unlock; /* possible? */
20064 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
20065 + if (unlikely(!args))
20068 + parent = dentry->d_parent; /* dir inode is locked */
20069 + di_write_lock_parent(parent);
20070 + err = au_test_empty(dentry, &args->whlist);
20071 + if (unlikely(err))
20074 + bstart = au_dbstart(dentry);
20075 + bwh = au_dbwh(dentry);
20077 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
20079 + err = PTR_ERR(wh_dentry);
20080 + if (IS_ERR(wh_dentry))
20083 + h_dentry = au_h_dptr(dentry, bstart);
20086 + if (bindex == bstart) {
20087 + err = renwh_and_rmdir(dentry, bstart, &args->whlist, dir);
20089 + rmdir_later = err;
20093 + /* stop monitoring */
20094 + au_hn_free(au_hi(inode, bstart));
20096 + /* dir inode is locked */
20097 + IMustLock(d_inode(wh_dentry->d_parent));
20102 + vfsub_dead_dir(inode);
20103 + au_set_dbdiropq(dentry, -1);
20104 + epilog(dir, dentry, bindex);
20106 + if (rmdir_later) {
20107 + au_whtmp_kick_rmdir(dir, bstart, h_dentry, args);
20111 + goto out_unpin; /* success */
20119 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
20126 + au_unpin(&a->pin);
20130 + di_write_unlock(parent);
20132 + au_whtmp_rmdir_free(args);
20134 + aufs_read_unlock(dentry, AuLock_DW);
20141 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
20142 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
20143 +++ linux/fs/aufs/i_op_ren.c 2016-01-13 20:11:11.669760262 +0100
20146 + * Copyright (C) 2005-2015 Junjiro R. Okajima
20148 + * This program, aufs is free software; you can redistribute it and/or modify
20149 + * it under the terms of the GNU General Public License as published by
20150 + * the Free Software Foundation; either version 2 of the License, or
20151 + * (at your option) any later version.
20153 + * This program is distributed in the hope that it will be useful,
20154 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20155 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20156 + * GNU General Public License for more details.
20158 + * You should have received a copy of the GNU General Public License
20159 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20163 + * inode operation (rename entry)
20164 + * todo: this is crazy monster
20169 +enum { AuSRC, AuDST, AuSrcDst };
20170 +enum { AuPARENT, AuCHILD, AuParentChild };
20172 +#define AuRen_ISDIR 1
20173 +#define AuRen_ISSAMEDIR (1 << 1)
20174 +#define AuRen_WHSRC (1 << 2)
20175 +#define AuRen_WHDST (1 << 3)
20176 +#define AuRen_MNT_WRITE (1 << 4)
20177 +#define AuRen_DT_DSTDIR (1 << 5)
20178 +#define AuRen_DIROPQ (1 << 6)
20179 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
20180 +#define au_fset_ren(flags, name) \
20181 + do { (flags) |= AuRen_##name; } while (0)
20182 +#define au_fclr_ren(flags, name) \
20183 + do { (flags) &= ~AuRen_##name; } while (0)
20185 +struct au_ren_args {
20187 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
20189 + struct inode *dir, *inode;
20190 + struct au_hinode *hdir;
20191 + struct au_dtime dt[AuParentChild];
20192 + aufs_bindex_t bstart;
20195 +#define src_dentry sd[AuSRC].dentry
20196 +#define src_dir sd[AuSRC].dir
20197 +#define src_inode sd[AuSRC].inode
20198 +#define src_h_dentry sd[AuSRC].h_dentry
20199 +#define src_parent sd[AuSRC].parent
20200 +#define src_h_parent sd[AuSRC].h_parent
20201 +#define src_wh_dentry sd[AuSRC].wh_dentry
20202 +#define src_hdir sd[AuSRC].hdir
20203 +#define src_h_dir sd[AuSRC].hdir->hi_inode
20204 +#define src_dt sd[AuSRC].dt
20205 +#define src_bstart sd[AuSRC].bstart
20207 +#define dst_dentry sd[AuDST].dentry
20208 +#define dst_dir sd[AuDST].dir
20209 +#define dst_inode sd[AuDST].inode
20210 +#define dst_h_dentry sd[AuDST].h_dentry
20211 +#define dst_parent sd[AuDST].parent
20212 +#define dst_h_parent sd[AuDST].h_parent
20213 +#define dst_wh_dentry sd[AuDST].wh_dentry
20214 +#define dst_hdir sd[AuDST].hdir
20215 +#define dst_h_dir sd[AuDST].hdir->hi_inode
20216 +#define dst_dt sd[AuDST].dt
20217 +#define dst_bstart sd[AuDST].bstart
20219 + struct dentry *h_trap;
20220 + struct au_branch *br;
20221 + struct au_hinode *src_hinode;
20222 + struct path h_path;
20223 + struct au_nhash whlist;
20224 + aufs_bindex_t btgt, src_bwh, src_bdiropq;
20226 + unsigned int flags;
20228 + struct au_whtmp_rmdir *thargs;
20229 + struct dentry *h_dst;
20232 +/* ---------------------------------------------------------------------- */
20235 + * functions for reverting.
20236 + * when an error happened in a single rename systemcall, we should revert
20237 + * everything as if nothing happened.
20238 + * we don't need to revert the copied-up/down the parent dir since they are
20242 +#define RevertFailure(fmt, ...) do { \
20243 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
20244 + ##__VA_ARGS__, err, rerr); \
20248 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
20252 + au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
20253 + rerr = au_diropq_remove(a->src_dentry, a->btgt);
20254 + au_hn_imtx_unlock(a->src_hinode);
20255 + au_set_dbdiropq(a->src_dentry, a->src_bdiropq);
20257 + RevertFailure("remove diropq %pd", a->src_dentry);
20260 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
20263 + struct inode *delegated;
20265 + a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name,
20266 + a->src_h_parent);
20267 + rerr = PTR_ERR(a->h_path.dentry);
20268 + if (IS_ERR(a->h_path.dentry)) {
20269 + RevertFailure("lkup one %pd", a->src_dentry);
20273 + delegated = NULL;
20274 + rerr = vfsub_rename(a->dst_h_dir,
20275 + au_h_dptr(a->src_dentry, a->btgt),
20276 + a->src_h_dir, &a->h_path, &delegated);
20277 + if (unlikely(rerr == -EWOULDBLOCK)) {
20278 + pr_warn("cannot retry for NFSv4 delegation"
20279 + " for an internal rename\n");
20282 + d_drop(a->h_path.dentry);
20283 + dput(a->h_path.dentry);
20284 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
20286 + RevertFailure("rename %pd", a->src_dentry);
20289 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
20292 + struct inode *delegated;
20294 + a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name,
20295 + a->dst_h_parent);
20296 + rerr = PTR_ERR(a->h_path.dentry);
20297 + if (IS_ERR(a->h_path.dentry)) {
20298 + RevertFailure("lkup one %pd", a->dst_dentry);
20301 + if (d_is_positive(a->h_path.dentry)) {
20302 + d_drop(a->h_path.dentry);
20303 + dput(a->h_path.dentry);
20307 + delegated = NULL;
20308 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
20310 + if (unlikely(rerr == -EWOULDBLOCK)) {
20311 + pr_warn("cannot retry for NFSv4 delegation"
20312 + " for an internal rename\n");
20315 + d_drop(a->h_path.dentry);
20316 + dput(a->h_path.dentry);
20318 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
20320 + RevertFailure("rename %pd", a->h_dst);
20323 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
20327 + a->h_path.dentry = a->src_wh_dentry;
20328 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
20329 + au_set_dbwh(a->src_dentry, a->src_bwh);
20331 + RevertFailure("unlink %pd", a->src_wh_dentry);
20333 +#undef RevertFailure
20335 +/* ---------------------------------------------------------------------- */
20338 + * when we have to copyup the renaming entry, do it with the rename-target name
20339 + * in order to minimize the cost (the later actual rename is unnecessary).
20340 + * otherwise rename it on the target branch.
20342 +static int au_ren_or_cpup(struct au_ren_args *a)
20345 + struct dentry *d;
20346 + struct inode *delegated;
20348 + d = a->src_dentry;
20349 + if (au_dbstart(d) == a->btgt) {
20350 + a->h_path.dentry = a->dst_h_dentry;
20351 + if (au_ftest_ren(a->flags, DIROPQ)
20352 + && au_dbdiropq(d) == a->btgt)
20353 + au_fclr_ren(a->flags, DIROPQ);
20354 + AuDebugOn(au_dbstart(d) != a->btgt);
20355 + delegated = NULL;
20356 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
20357 + a->dst_h_dir, &a->h_path, &delegated);
20358 + if (unlikely(err == -EWOULDBLOCK)) {
20359 + pr_warn("cannot retry for NFSv4 delegation"
20360 + " for an internal rename\n");
20366 + if (!err && a->h_dst)
20367 + /* it will be set to dinfo later */
20373 +/* cf. aufs_rmdir() */
20374 +static int au_ren_del_whtmp(struct au_ren_args *a)
20377 + struct inode *dir;
20379 + dir = a->dst_dir;
20380 + SiMustAnyLock(dir->i_sb);
20381 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
20382 + au_sbi(dir->i_sb)->si_dirwh)
20383 + || au_test_fs_remote(a->h_dst->d_sb)) {
20384 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
20385 + if (unlikely(err))
20386 + pr_warn("failed removing whtmp dir %pd (%d), "
20387 + "ignored.\n", a->h_dst, err);
20389 + au_nhash_wh_free(&a->thargs->whlist);
20390 + a->thargs->whlist = a->whlist;
20391 + a->whlist.nh_num = 0;
20392 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
20394 + a->thargs = NULL;
20400 +/* make it 'opaque' dir. */
20401 +static int au_ren_diropq(struct au_ren_args *a)
20404 + struct dentry *diropq;
20407 + a->src_bdiropq = au_dbdiropq(a->src_dentry);
20408 + a->src_hinode = au_hi(a->src_inode, a->btgt);
20409 + au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
20410 + diropq = au_diropq_create(a->src_dentry, a->btgt);
20411 + au_hn_imtx_unlock(a->src_hinode);
20412 + if (IS_ERR(diropq))
20413 + err = PTR_ERR(diropq);
20420 +static int do_rename(struct au_ren_args *a)
20423 + struct dentry *d, *h_d;
20425 + /* prepare workqueue args for asynchronous rmdir */
20426 + h_d = a->dst_h_dentry;
20427 + if (au_ftest_ren(a->flags, ISDIR) && d_is_positive(h_d)) {
20429 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb, GFP_NOFS);
20430 + if (unlikely(!a->thargs))
20432 + a->h_dst = dget(h_d);
20435 + /* create whiteout for src_dentry */
20436 + if (au_ftest_ren(a->flags, WHSRC)) {
20437 + a->src_bwh = au_dbwh(a->src_dentry);
20438 + AuDebugOn(a->src_bwh >= 0);
20440 + = au_wh_create(a->src_dentry, a->btgt, a->src_h_parent);
20441 + err = PTR_ERR(a->src_wh_dentry);
20442 + if (IS_ERR(a->src_wh_dentry))
20446 + /* lookup whiteout for dentry */
20447 + if (au_ftest_ren(a->flags, WHDST)) {
20448 + h_d = au_wh_lkup(a->dst_h_parent, &a->dst_dentry->d_name,
20450 + err = PTR_ERR(h_d);
20453 + if (d_is_negative(h_d))
20456 + a->dst_wh_dentry = h_d;
20459 + /* rename dentry to tmpwh */
20461 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
20462 + if (unlikely(err))
20465 + d = a->dst_dentry;
20466 + au_set_h_dptr(d, a->btgt, NULL);
20467 + err = au_lkup_neg(d, a->btgt, /*wh*/0);
20468 + if (unlikely(err))
20470 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
20473 + BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_bstart != a->btgt);
20475 + /* rename by vfs_rename or cpup */
20476 + d = a->dst_dentry;
20477 + if (au_ftest_ren(a->flags, ISDIR)
20478 + && (a->dst_wh_dentry
20479 + || au_dbdiropq(d) == a->btgt
20480 + /* hide the lower to keep xino */
20481 + || a->btgt < au_dbend(d)
20482 + || au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ)))
20483 + au_fset_ren(a->flags, DIROPQ);
20484 + err = au_ren_or_cpup(a);
20485 + if (unlikely(err))
20486 + /* leave the copied-up one */
20489 + /* make dir opaque */
20490 + if (au_ftest_ren(a->flags, DIROPQ)) {
20491 + err = au_ren_diropq(a);
20492 + if (unlikely(err))
20496 + /* update target timestamps */
20497 + AuDebugOn(au_dbstart(a->src_dentry) != a->btgt);
20498 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
20499 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
20500 + a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
20502 + /* remove whiteout for dentry */
20503 + if (a->dst_wh_dentry) {
20504 + a->h_path.dentry = a->dst_wh_dentry;
20505 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
20507 + if (unlikely(err))
20511 + /* remove whtmp */
20513 + au_ren_del_whtmp(a); /* ignore this error */
20515 + au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
20517 + goto out_success;
20520 + if (au_ftest_ren(a->flags, DIROPQ))
20521 + au_ren_rev_diropq(err, a);
20523 + au_ren_rev_rename(err, a);
20527 + au_ren_rev_whtmp(err, a);
20529 + dput(a->dst_wh_dentry);
20530 + a->dst_wh_dentry = NULL;
20532 + if (a->src_wh_dentry)
20533 + au_ren_rev_whsrc(err, a);
20535 + dput(a->src_wh_dentry);
20536 + dput(a->dst_wh_dentry);
20540 + au_whtmp_rmdir_free(a->thargs);
20541 + a->thargs = NULL;
20547 +/* ---------------------------------------------------------------------- */
20550 + * test if @dentry dir can be rename destination or not.
20551 + * success means, it is a logically empty dir.
20553 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
20555 + return au_test_empty(dentry, whlist);
20559 + * test if @dentry dir can be rename source or not.
20560 + * if it can, return 0 and @children is filled.
20562 + * - it is a logically empty dir.
20563 + * - or, it exists on writable branch and has no children including whiteouts
20564 + * on the lower branch.
20566 +static int may_rename_srcdir(struct dentry *dentry, aufs_bindex_t btgt)
20569 + unsigned int rdhash;
20570 + aufs_bindex_t bstart;
20572 + bstart = au_dbstart(dentry);
20573 + if (bstart != btgt) {
20574 + struct au_nhash whlist;
20576 + SiMustAnyLock(dentry->d_sb);
20577 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
20579 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
20581 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
20582 + if (unlikely(err))
20584 + err = au_test_empty(dentry, &whlist);
20585 + au_nhash_wh_free(&whlist);
20589 + if (bstart == au_dbtaildir(dentry))
20590 + return 0; /* success */
20592 + err = au_test_empty_lower(dentry);
20595 + if (err == -ENOTEMPTY) {
20596 + AuWarn1("renaming dir who has child(ren) on multiple branches,"
20597 + " is not supported\n");
20603 +/* side effect: sets whlist and h_dentry */
20604 +static int au_ren_may_dir(struct au_ren_args *a)
20607 + unsigned int rdhash;
20608 + struct dentry *d;
20610 + d = a->dst_dentry;
20611 + SiMustAnyLock(d->d_sb);
20614 + if (au_ftest_ren(a->flags, ISDIR) && a->dst_inode) {
20615 + rdhash = au_sbi(d->d_sb)->si_rdhash;
20617 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
20618 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
20619 + if (unlikely(err))
20622 + au_set_dbstart(d, a->dst_bstart);
20623 + err = may_rename_dstdir(d, &a->whlist);
20624 + au_set_dbstart(d, a->btgt);
20626 + a->dst_h_dentry = au_h_dptr(d, au_dbstart(d));
20627 + if (unlikely(err))
20630 + d = a->src_dentry;
20631 + a->src_h_dentry = au_h_dptr(d, au_dbstart(d));
20632 + if (au_ftest_ren(a->flags, ISDIR)) {
20633 + err = may_rename_srcdir(d, a->btgt);
20634 + if (unlikely(err)) {
20635 + au_nhash_wh_free(&a->whlist);
20636 + a->whlist.nh_num = 0;
20643 +/* ---------------------------------------------------------------------- */
20646 + * simple tests for rename.
20647 + * following the checks in vfs, plus the parent-child relationship.
20649 +static int au_may_ren(struct au_ren_args *a)
20652 + struct inode *h_inode;
20654 + if (a->src_bstart == a->btgt) {
20655 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
20656 + au_ftest_ren(a->flags, ISDIR));
20657 + if (unlikely(err))
20660 + if (unlikely(a->src_h_dentry == a->h_trap))
20665 + if (a->dst_bstart != a->btgt)
20668 + err = -ENOTEMPTY;
20669 + if (unlikely(a->dst_h_dentry == a->h_trap))
20673 + isdir = !!au_ftest_ren(a->flags, ISDIR);
20674 + if (d_really_is_negative(a->dst_dentry)) {
20675 + if (d_is_negative(a->dst_h_dentry))
20676 + err = au_may_add(a->dst_dentry, a->btgt,
20677 + a->dst_h_parent, isdir);
20679 + if (unlikely(d_is_negative(a->dst_h_dentry)))
20681 + h_inode = d_inode(a->dst_h_dentry);
20682 + if (h_inode->i_nlink)
20683 + err = au_may_del(a->dst_dentry, a->btgt,
20684 + a->dst_h_parent, isdir);
20688 + if (unlikely(err == -ENOENT || err == -EEXIST))
20694 +/* ---------------------------------------------------------------------- */
20699 + * - src_dir and dir by lock_rename()
20700 + * - inode if exitsts
20703 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
20705 + * + di_write_lock2_child()
20706 + * + di_write_lock_child()
20707 + * + ii_write_lock_child()
20708 + * + di_write_lock_child2()
20709 + * + ii_write_lock_child2()
20710 + * + src_parent and parent
20711 + * + di_write_lock_parent()
20712 + * + ii_write_lock_parent()
20713 + * + di_write_lock_parent2()
20714 + * + ii_write_lock_parent2()
20715 + * + lower src_dir and dir by vfsub_lock_rename()
20716 + * + verify the every relationships between child and parent. if any
20717 + * of them failed, unlock all and return -EBUSY.
20719 +static void au_ren_unlock(struct au_ren_args *a)
20721 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
20722 + a->dst_h_parent, a->dst_hdir);
20723 + if (au_ftest_ren(a->flags, MNT_WRITE))
20724 + vfsub_mnt_drop_write(au_br_mnt(a->br));
20727 +static int au_ren_lock(struct au_ren_args *a)
20730 + unsigned int udba;
20733 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
20734 + a->src_hdir = au_hi(a->src_dir, a->btgt);
20735 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
20736 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
20738 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
20739 + if (unlikely(err))
20741 + au_fset_ren(a->flags, MNT_WRITE);
20742 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
20743 + a->dst_h_parent, a->dst_hdir);
20744 + udba = au_opt_udba(a->src_dentry->d_sb);
20745 + if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
20746 + || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
20747 + err = au_busy_or_stale();
20748 + if (!err && au_dbstart(a->src_dentry) == a->btgt)
20749 + err = au_h_verify(a->src_h_dentry, udba,
20750 + d_inode(a->src_h_parent), a->src_h_parent,
20752 + if (!err && au_dbstart(a->dst_dentry) == a->btgt)
20753 + err = au_h_verify(a->dst_h_dentry, udba,
20754 + d_inode(a->dst_h_parent), a->dst_h_parent,
20757 + goto out; /* success */
20759 + err = au_busy_or_stale();
20760 + au_ren_unlock(a);
20766 +/* ---------------------------------------------------------------------- */
20768 +static void au_ren_refresh_dir(struct au_ren_args *a)
20770 + struct inode *dir;
20772 + dir = a->dst_dir;
20773 + dir->i_version++;
20774 + if (au_ftest_ren(a->flags, ISDIR)) {
20775 + /* is this updating defined in POSIX? */
20776 + au_cpup_attr_timesizes(a->src_inode);
20777 + au_cpup_attr_nlink(dir, /*force*/1);
20780 + au_dir_ts(dir, a->btgt);
20782 + if (au_ftest_ren(a->flags, ISSAMEDIR))
20785 + dir = a->src_dir;
20786 + dir->i_version++;
20787 + if (au_ftest_ren(a->flags, ISDIR))
20788 + au_cpup_attr_nlink(dir, /*force*/1);
20789 + au_dir_ts(dir, a->btgt);
20792 +static void au_ren_refresh(struct au_ren_args *a)
20794 + aufs_bindex_t bend, bindex;
20795 + struct dentry *d, *h_d;
20796 + struct inode *i, *h_i;
20797 + struct super_block *sb;
20799 + d = a->dst_dentry;
20802 + /* already dget-ed by au_ren_or_cpup() */
20803 + au_set_h_dptr(d, a->btgt, a->h_dst);
20805 + i = a->dst_inode;
20807 + if (!au_ftest_ren(a->flags, ISDIR))
20808 + vfsub_drop_nlink(i);
20810 + vfsub_dead_dir(i);
20811 + au_cpup_attr_timesizes(i);
20813 + au_update_dbrange(d, /*do_put_zero*/1);
20816 + for (bindex = au_dbstart(d); bindex < bend; bindex++)
20817 + au_set_h_dptr(d, bindex, NULL);
20818 + bend = au_dbend(d);
20819 + for (bindex = a->btgt + 1; bindex <= bend; bindex++)
20820 + au_set_h_dptr(d, bindex, NULL);
20821 + au_update_dbrange(d, /*do_put_zero*/0);
20824 + d = a->src_dentry;
20825 + au_set_dbwh(d, -1);
20826 + bend = au_dbend(d);
20827 + for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
20828 + h_d = au_h_dptr(d, bindex);
20830 + au_set_h_dptr(d, bindex, NULL);
20832 + au_set_dbend(d, a->btgt);
20835 + i = a->src_inode;
20836 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
20837 + return; /* success */
20839 + bend = au_ibend(i);
20840 + for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
20841 + h_i = au_h_iptr(i, bindex);
20843 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
20844 + /* ignore this error */
20845 + au_set_h_iptr(i, bindex, NULL, 0);
20848 + au_set_ibend(i, a->btgt);
20851 +/* ---------------------------------------------------------------------- */
20853 +/* mainly for link(2) and rename(2) */
20854 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
20856 + aufs_bindex_t bdiropq, bwh;
20857 + struct dentry *parent;
20858 + struct au_branch *br;
20860 + parent = dentry->d_parent;
20861 + IMustLock(d_inode(parent)); /* dir is locked */
20863 + bdiropq = au_dbdiropq(parent);
20864 + bwh = au_dbwh(dentry);
20865 + br = au_sbr(dentry->d_sb, btgt);
20866 + if (au_br_rdonly(br)
20867 + || (0 <= bdiropq && bdiropq < btgt)
20868 + || (0 <= bwh && bwh < btgt))
20871 + AuDbg("btgt %d\n", btgt);
20875 +/* sets src_bstart, dst_bstart and btgt */
20876 +static int au_ren_wbr(struct au_ren_args *a)
20879 + struct au_wr_dir_args wr_dir_args = {
20880 + /* .force_btgt = -1, */
20881 + .flags = AuWrDir_ADD_ENTRY
20884 + a->src_bstart = au_dbstart(a->src_dentry);
20885 + a->dst_bstart = au_dbstart(a->dst_dentry);
20886 + if (au_ftest_ren(a->flags, ISDIR))
20887 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
20888 + wr_dir_args.force_btgt = a->src_bstart;
20889 + if (a->dst_inode && a->dst_bstart < a->src_bstart)
20890 + wr_dir_args.force_btgt = a->dst_bstart;
20891 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
20892 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
20898 +static void au_ren_dt(struct au_ren_args *a)
20900 + a->h_path.dentry = a->src_h_parent;
20901 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
20902 + if (!au_ftest_ren(a->flags, ISSAMEDIR)) {
20903 + a->h_path.dentry = a->dst_h_parent;
20904 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
20907 + au_fclr_ren(a->flags, DT_DSTDIR);
20908 + if (!au_ftest_ren(a->flags, ISDIR))
20911 + a->h_path.dentry = a->src_h_dentry;
20912 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
20913 + if (d_is_positive(a->dst_h_dentry)) {
20914 + au_fset_ren(a->flags, DT_DSTDIR);
20915 + a->h_path.dentry = a->dst_h_dentry;
20916 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
20920 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
20922 + struct dentry *h_d;
20923 + struct mutex *h_mtx;
20925 + au_dtime_revert(a->src_dt + AuPARENT);
20926 + if (!au_ftest_ren(a->flags, ISSAMEDIR))
20927 + au_dtime_revert(a->dst_dt + AuPARENT);
20929 + if (au_ftest_ren(a->flags, ISDIR) && err != -EIO) {
20930 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
20931 + h_mtx = &d_inode(h_d)->i_mutex;
20932 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
20933 + au_dtime_revert(a->src_dt + AuCHILD);
20934 + mutex_unlock(h_mtx);
20936 + if (au_ftest_ren(a->flags, DT_DSTDIR)) {
20937 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
20938 + h_mtx = &d_inode(h_d)->i_mutex;
20939 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
20940 + au_dtime_revert(a->dst_dt + AuCHILD);
20941 + mutex_unlock(h_mtx);
20946 +/* ---------------------------------------------------------------------- */
20948 +int aufs_rename(struct inode *_src_dir, struct dentry *_src_dentry,
20949 + struct inode *_dst_dir, struct dentry *_dst_dentry)
20952 + /* reduce stack space */
20953 + struct au_ren_args *a;
20955 + AuDbg("%pd, %pd\n", _src_dentry, _dst_dentry);
20956 + IMustLock(_src_dir);
20957 + IMustLock(_dst_dir);
20960 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
20961 + a = kzalloc(sizeof(*a), GFP_NOFS);
20962 + if (unlikely(!a))
20965 + a->src_dir = _src_dir;
20966 + a->src_dentry = _src_dentry;
20967 + a->src_inode = NULL;
20968 + if (d_really_is_positive(a->src_dentry))
20969 + a->src_inode = d_inode(a->src_dentry);
20970 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
20971 + a->dst_dir = _dst_dir;
20972 + a->dst_dentry = _dst_dentry;
20973 + a->dst_inode = NULL;
20974 + if (d_really_is_positive(a->dst_dentry))
20975 + a->dst_inode = d_inode(a->dst_dentry);
20976 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
20977 + if (a->dst_inode) {
20978 + IMustLock(a->dst_inode);
20979 + au_igrab(a->dst_inode);
20983 + flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
20984 + if (d_is_dir(a->src_dentry)) {
20985 + au_fset_ren(a->flags, ISDIR);
20986 + if (unlikely(d_really_is_positive(a->dst_dentry)
20987 + && !d_is_dir(a->dst_dentry)))
20989 + flags |= AuLock_DIRS;
20991 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry, flags);
20992 + if (unlikely(err))
20995 + err = au_d_hashed_positive(a->src_dentry);
20996 + if (unlikely(err))
20999 + if (a->dst_inode) {
21001 + * If it is a dir, VFS unhash dst_dentry before this
21002 + * function. It means we cannot rely upon d_unhashed().
21004 + if (unlikely(!a->dst_inode->i_nlink))
21006 + if (!S_ISDIR(a->dst_inode->i_mode)) {
21007 + err = au_d_hashed_positive(a->dst_dentry);
21008 + if (unlikely(err))
21010 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
21012 + } else if (unlikely(d_unhashed(a->dst_dentry)))
21016 + * is it possible?
21017 + * yes, it happened (in linux-3.3-rcN) but I don't know why.
21018 + * there may exist a problem somewhere else.
21021 + if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
21024 + au_fset_ren(a->flags, ISSAMEDIR); /* temporary */
21025 + di_write_lock_parent(a->dst_parent);
21027 + /* which branch we process */
21028 + err = au_ren_wbr(a);
21029 + if (unlikely(err < 0))
21031 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
21032 + a->h_path.mnt = au_br_mnt(a->br);
21034 + /* are they available to be renamed */
21035 + err = au_ren_may_dir(a);
21036 + if (unlikely(err))
21037 + goto out_children;
21039 + /* prepare the writable parent dir on the same branch */
21040 + if (a->dst_bstart == a->btgt) {
21041 + au_fset_ren(a->flags, WHDST);
21043 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
21044 + if (unlikely(err))
21045 + goto out_children;
21048 + if (a->src_dir != a->dst_dir) {
21050 + * this temporary unlock is safe,
21051 + * because both dir->i_mutex are locked.
21053 + di_write_unlock(a->dst_parent);
21054 + di_write_lock_parent(a->src_parent);
21055 + err = au_wr_dir_need_wh(a->src_dentry,
21056 + au_ftest_ren(a->flags, ISDIR),
21058 + di_write_unlock(a->src_parent);
21059 + di_write_lock2_parent(a->src_parent, a->dst_parent, /*isdir*/1);
21060 + au_fclr_ren(a->flags, ISSAMEDIR);
21062 + err = au_wr_dir_need_wh(a->src_dentry,
21063 + au_ftest_ren(a->flags, ISDIR),
21065 + if (unlikely(err < 0))
21066 + goto out_children;
21068 + au_fset_ren(a->flags, WHSRC);
21071 + if (a->src_bstart != a->btgt) {
21072 + struct au_pin pin;
21074 + err = au_pin(&pin, a->src_dentry, a->btgt,
21075 + au_opt_udba(a->src_dentry->d_sb),
21076 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21078 + struct au_cp_generic cpg = {
21079 + .dentry = a->src_dentry,
21081 + .bsrc = a->src_bstart,
21084 + .flags = AuCpup_DTIME | AuCpup_HOPEN
21086 + AuDebugOn(au_dbstart(a->src_dentry) != a->src_bstart);
21087 + err = au_sio_cpup_simple(&cpg);
21090 + if (unlikely(err))
21091 + goto out_children;
21092 + a->src_bstart = a->btgt;
21093 + a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
21094 + au_fset_ren(a->flags, WHSRC);
21097 + /* lock them all */
21098 + err = au_ren_lock(a);
21099 + if (unlikely(err))
21100 + /* leave the copied-up one */
21101 + goto out_children;
21103 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
21104 + err = au_may_ren(a);
21105 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
21106 + err = -ENAMETOOLONG;
21107 + if (unlikely(err))
21110 + /* store timestamps to be revertible */
21114 + err = do_rename(a);
21115 + if (unlikely(err))
21118 + /* update dir attributes */
21119 + au_ren_refresh_dir(a);
21121 + /* dput/iput all lower dentries */
21122 + au_ren_refresh(a);
21124 + goto out_hdir; /* success */
21127 + au_ren_rev_dt(err, a);
21129 + au_ren_unlock(a);
21131 + au_nhash_wh_free(&a->whlist);
21132 + if (err && a->dst_inode && a->dst_bstart != a->btgt) {
21133 + AuDbg("bstart %d, btgt %d\n", a->dst_bstart, a->btgt);
21134 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
21135 + au_set_dbstart(a->dst_dentry, a->dst_bstart);
21139 + d_move(a->src_dentry, a->dst_dentry);
21141 + au_update_dbstart(a->dst_dentry);
21142 + if (!a->dst_inode)
21143 + d_drop(a->dst_dentry);
21145 + if (au_ftest_ren(a->flags, ISSAMEDIR))
21146 + di_write_unlock(a->dst_parent);
21148 + di_write_unlock2(a->src_parent, a->dst_parent);
21150 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
21152 + iput(a->dst_inode);
21154 + au_whtmp_rmdir_free(a->thargs);
21160 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
21161 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
21162 +++ linux/fs/aufs/Kconfig 2016-01-13 20:11:11.666426853 +0100
21165 + tristate "Aufs (Advanced multi layered unification filesystem) support"
21167 + Aufs is a stackable unification filesystem such as Unionfs,
21168 + which unifies several directories and provides a merged single
21170 + In the early days, aufs was entirely re-designed and
21171 + re-implemented Unionfs Version 1.x series. Introducing many
21172 + original ideas, approaches and improvements, it becomes totally
21173 + different from Unionfs while keeping the basic features.
21177 + prompt "Maximum number of branches"
21178 + default AUFS_BRANCH_MAX_127
21180 + Specifies the maximum number of branches (or member directories)
21181 + in a single aufs. The larger value consumes more system
21182 + resources and has a minor impact to performance.
21183 +config AUFS_BRANCH_MAX_127
21186 + Specifies the maximum number of branches (or member directories)
21187 + in a single aufs. The larger value consumes more system
21188 + resources and has a minor impact to performance.
21189 +config AUFS_BRANCH_MAX_511
21192 + Specifies the maximum number of branches (or member directories)
21193 + in a single aufs. The larger value consumes more system
21194 + resources and has a minor impact to performance.
21195 +config AUFS_BRANCH_MAX_1023
21198 + Specifies the maximum number of branches (or member directories)
21199 + in a single aufs. The larger value consumes more system
21200 + resources and has a minor impact to performance.
21201 +config AUFS_BRANCH_MAX_32767
21204 + Specifies the maximum number of branches (or member directories)
21205 + in a single aufs. The larger value consumes more system
21206 + resources and has a minor impact to performance.
21209 +config AUFS_SBILIST
21211 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
21214 + Automatic configuration for internal use.
21215 + When aufs supports Magic SysRq or /proc, enabled automatically.
21217 +config AUFS_HNOTIFY
21218 + bool "Detect direct branch access (bypassing aufs)"
21220 + If you want to modify files on branches directly, eg. bypassing aufs,
21221 + and want aufs to detect the changes of them fully, then enable this
21222 + option and use 'udba=notify' mount option.
21223 + Currently there is only one available configuration, "fsnotify".
21224 + It will have a negative impact to the performance.
21225 + See detail in aufs.5.
21228 + prompt "method" if AUFS_HNOTIFY
21229 + default AUFS_HFSNOTIFY
21230 +config AUFS_HFSNOTIFY
21235 +config AUFS_EXPORT
21236 + bool "NFS-exportable aufs"
21237 + depends on EXPORTFS
21239 + If you want to export your mounted aufs via NFS, then enable this
21240 + option. There are several requirements for this configuration.
21241 + See detail in aufs.5.
21243 +config AUFS_INO_T_64
21245 + depends on AUFS_EXPORT
21246 + depends on 64BIT && !(ALPHA || S390)
21249 + Automatic configuration for internal use.
21250 + /* typedef unsigned long/int __kernel_ino_t */
21251 + /* alpha and s390x are int */
21254 + bool "support for XATTR/EA (including Security Labels)"
21256 + If your branch fs supports XATTR/EA and you want to make them
21257 + available in aufs too, then enable this opsion and specify the
21258 + branch attributes for EA.
21259 + See detail in aufs.5.
21262 + bool "File-based Hierarchical Storage Management"
21264 + Hierarchical Storage Management (or HSM) is a well-known feature
21265 + in the storage world. Aufs provides this feature as file-based.
21266 + with multiple branches.
21267 + These multiple branches are prioritized, ie. the topmost one
21268 + should be the fastest drive and be used heavily.
21271 + bool "Readdir in userspace"
21273 + Aufs has two methods to provide a merged view for a directory,
21274 + by a user-space library and by kernel-space natively. The latter
21275 + is always enabled but sometimes large and slow.
21276 + If you enable this option, install the library in aufs2-util
21277 + package, and set some environment variables for your readdir(3),
21278 + then the work will be handled in user-space which generally
21279 + shows better performance in most cases.
21280 + See detail in aufs.5.
21283 + bool "Show whiteouts"
21285 + If you want to make the whiteouts in aufs visible, then enable
21286 + this option and specify 'shwh' mount option. Although it may
21287 + sounds like philosophy or something, but in technically it
21288 + simply shows the name of whiteout with keeping its behaviour.
21290 +config AUFS_BR_RAMFS
21291 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
21293 + If you want to use ramfs as an aufs branch fs, then enable this
21294 + option. Generally tmpfs is recommended.
21295 + Aufs prohibited them to be a branch fs by default, because
21296 + initramfs becomes unusable after switch_root or something
21297 + generally. If you sets initramfs as an aufs branch and boot your
21298 + system by switch_root, you will meet a problem easily since the
21299 + files in initramfs may be inaccessible.
21300 + Unless you are going to use ramfs as an aufs branch fs without
21301 + switch_root or something, leave it N.
21303 +config AUFS_BR_FUSE
21304 + bool "Fuse fs as an aufs branch"
21305 + depends on FUSE_FS
21308 + If you want to use fuse-based userspace filesystem as an aufs
21309 + branch fs, then enable this option.
21310 + It implements the internal poll(2) operation which is
21311 + implemented by fuse only (curretnly).
21316 + Automatic configuration for internal use.
21318 +config AUFS_BR_HFSPLUS
21319 + bool "Hfsplus as an aufs branch"
21320 + depends on HFSPLUS_FS
21323 + If you want to use hfsplus fs as an aufs branch fs, then enable
21324 + this option. This option introduces a small overhead at
21325 + copying-up a file on hfsplus.
21327 +config AUFS_BDEV_LOOP
21329 + depends on BLK_DEV_LOOP
21332 + Automatic configuration for internal use.
21333 + Convert =[ym] into =y.
21336 + bool "Debug aufs"
21338 + Enable this to compile aufs internal debug code.
21339 + It will have a negative impact to the performance.
21341 +config AUFS_MAGIC_SYSRQ
21343 + depends on AUFS_DEBUG && MAGIC_SYSRQ
21346 + Automatic configuration for internal use.
21347 + When aufs supports Magic SysRq, enabled automatically.
21349 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
21350 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
21351 +++ linux/fs/aufs/loop.c 2016-01-13 20:11:11.669760262 +0100
21354 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21356 + * This program, aufs is free software; you can redistribute it and/or modify
21357 + * it under the terms of the GNU General Public License as published by
21358 + * the Free Software Foundation; either version 2 of the License, or
21359 + * (at your option) any later version.
21361 + * This program is distributed in the hope that it will be useful,
21362 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21363 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21364 + * GNU General Public License for more details.
21366 + * You should have received a copy of the GNU General Public License
21367 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21371 + * support for loopback block device as a branch
21376 +/* added into drivers/block/loop.c */
21377 +static struct file *(*backing_file_func)(struct super_block *sb);
21380 + * test if two lower dentries have overlapping branches.
21382 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
21384 + struct super_block *h_sb;
21385 + struct file *backing_file;
21387 + if (unlikely(!backing_file_func)) {
21388 + /* don't load "loop" module here */
21389 + backing_file_func = symbol_get(loop_backing_file);
21390 + if (unlikely(!backing_file_func))
21391 + /* "loop" module is not loaded */
21395 + h_sb = h_adding->d_sb;
21396 + backing_file = backing_file_func(h_sb);
21397 + if (!backing_file)
21400 + h_adding = backing_file->f_path.dentry;
21402 + * h_adding can be local NFS.
21403 + * in this case aufs cannot detect the loop.
21405 + if (unlikely(h_adding->d_sb == sb))
21407 + return !!au_test_subdir(h_adding, sb->s_root);
21410 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
21411 +int au_test_loopback_kthread(void)
21414 + struct task_struct *tsk = current;
21415 + char c, comm[sizeof(tsk->comm)];
21418 + if (tsk->flags & PF_KTHREAD) {
21419 + get_task_comm(comm, tsk);
21421 + ret = ('0' <= c && c <= '9'
21422 + && !strncmp(comm, "loop", 4));
21428 +/* ---------------------------------------------------------------------- */
21430 +#define au_warn_loopback_step 16
21431 +static int au_warn_loopback_nelem = au_warn_loopback_step;
21432 +static unsigned long *au_warn_loopback_array;
21434 +void au_warn_loopback(struct super_block *h_sb)
21436 + int i, new_nelem;
21437 + unsigned long *a, magic;
21438 + static DEFINE_SPINLOCK(spin);
21440 + magic = h_sb->s_magic;
21441 + spin_lock(&spin);
21442 + a = au_warn_loopback_array;
21443 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
21444 + if (a[i] == magic) {
21445 + spin_unlock(&spin);
21449 + /* h_sb is new to us, print it */
21450 + if (i < au_warn_loopback_nelem) {
21455 + /* expand the array */
21456 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
21457 + a = au_kzrealloc(au_warn_loopback_array,
21458 + au_warn_loopback_nelem * sizeof(unsigned long),
21459 + new_nelem * sizeof(unsigned long), GFP_ATOMIC);
21461 + au_warn_loopback_nelem = new_nelem;
21462 + au_warn_loopback_array = a;
21467 + spin_unlock(&spin);
21468 + AuWarn1("realloc failed, ignored\n");
21472 + spin_unlock(&spin);
21473 + pr_warn("you may want to try another patch for loopback file "
21474 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
21477 +int au_loopback_init(void)
21480 + struct super_block *sb __maybe_unused;
21482 + BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(unsigned long));
21485 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
21486 + sizeof(unsigned long), GFP_NOFS);
21487 + if (unlikely(!au_warn_loopback_array))
21493 +void au_loopback_fin(void)
21495 + if (backing_file_func)
21496 + symbol_put(loop_backing_file);
21497 + kfree(au_warn_loopback_array);
21499 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
21500 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
21501 +++ linux/fs/aufs/loop.h 2016-01-13 20:11:11.669760262 +0100
21504 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21506 + * This program, aufs is free software; you can redistribute it and/or modify
21507 + * it under the terms of the GNU General Public License as published by
21508 + * the Free Software Foundation; either version 2 of the License, or
21509 + * (at your option) any later version.
21511 + * This program is distributed in the hope that it will be useful,
21512 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21513 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21514 + * GNU General Public License for more details.
21516 + * You should have received a copy of the GNU General Public License
21517 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21521 + * support for loopback mount as a branch
21524 +#ifndef __AUFS_LOOP_H__
21525 +#define __AUFS_LOOP_H__
21530 +struct super_block;
21532 +#ifdef CONFIG_AUFS_BDEV_LOOP
21533 +/* drivers/block/loop.c */
21534 +struct file *loop_backing_file(struct super_block *sb);
21537 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
21538 +int au_test_loopback_kthread(void);
21539 +void au_warn_loopback(struct super_block *h_sb);
21541 +int au_loopback_init(void);
21542 +void au_loopback_fin(void);
21544 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
21545 + struct dentry *h_adding)
21546 +AuStubInt0(au_test_loopback_kthread, void)
21547 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
21549 +AuStubInt0(au_loopback_init, void)
21550 +AuStubVoid(au_loopback_fin, void)
21551 +#endif /* BLK_DEV_LOOP */
21553 +#endif /* __KERNEL__ */
21554 +#endif /* __AUFS_LOOP_H__ */
21555 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
21556 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
21557 +++ linux/fs/aufs/magic.mk 2016-01-13 20:11:11.669760262 +0100
21560 +# defined in ${srctree}/fs/fuse/inode.c
21562 +ifdef CONFIG_FUSE_FS
21563 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
21566 +# defined in ${srctree}/fs/xfs/xfs_sb.h
21568 +ifdef CONFIG_XFS_FS
21569 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
21572 +# defined in ${srctree}/fs/configfs/mount.c
21574 +ifdef CONFIG_CONFIGFS_FS
21575 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
21578 +# defined in ${srctree}/fs/ubifs/ubifs.h
21580 +ifdef CONFIG_UBIFS_FS
21581 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
21584 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
21586 +ifdef CONFIG_HFSPLUS_FS
21587 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
21589 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
21590 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
21591 +++ linux/fs/aufs/Makefile 2016-01-13 20:11:11.666426853 +0100
21594 +include ${src}/magic.mk
21595 +ifeq (${CONFIG_AUFS_FS},m)
21596 +include ${src}/conf.mk
21598 +-include ${src}/priv_def.mk
21600 +# cf. include/linux/kernel.h
21602 +ccflags-y += -DDEBUG
21603 +# sparse requires the full pathname
21605 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
21607 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
21610 +obj-$(CONFIG_AUFS_FS) += aufs.o
21611 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o \
21612 + wkq.o vfsub.o dcsub.o \
21613 + cpup.o whout.o wbr_policy.o \
21614 + dinfo.o dentry.o \
21616 + finfo.o file.o f_op.o \
21618 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
21622 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
21623 +aufs-$(CONFIG_SYSFS) += sysfs.o
21624 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
21625 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
21626 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
21627 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
21628 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
21629 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
21630 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
21631 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
21632 +aufs-$(CONFIG_AUFS_POLL) += poll.o
21633 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
21634 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
21635 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
21636 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
21637 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
21638 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
21639 +++ linux/fs/aufs/module.c 2016-01-13 20:11:11.669760262 +0100
21642 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21644 + * This program, aufs is free software; you can redistribute it and/or modify
21645 + * it under the terms of the GNU General Public License as published by
21646 + * the Free Software Foundation; either version 2 of the License, or
21647 + * (at your option) any later version.
21649 + * This program is distributed in the hope that it will be useful,
21650 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21651 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21652 + * GNU General Public License for more details.
21654 + * You should have received a copy of the GNU General Public License
21655 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21659 + * module global variables and operations
21662 +#include <linux/module.h>
21663 +#include <linux/seq_file.h>
21666 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp)
21668 + if (new_sz <= nused)
21671 + p = krealloc(p, new_sz, gfp);
21673 + memset(p + nused, 0, new_sz - nused);
21677 +/* ---------------------------------------------------------------------- */
21682 +struct kmem_cache *au_cachep[AuCache_Last];
21683 +static int __init au_cache_init(void)
21685 + au_cachep[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
21686 + if (au_cachep[AuCache_DINFO])
21687 + /* SLAB_DESTROY_BY_RCU */
21688 + au_cachep[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
21689 + au_icntnr_init_once);
21690 + if (au_cachep[AuCache_ICNTNR])
21691 + au_cachep[AuCache_FINFO] = AuCacheCtor(au_finfo,
21692 + au_fi_init_once);
21693 + if (au_cachep[AuCache_FINFO])
21694 + au_cachep[AuCache_VDIR] = AuCache(au_vdir);
21695 + if (au_cachep[AuCache_VDIR])
21696 + au_cachep[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
21697 + if (au_cachep[AuCache_DEHSTR])
21703 +static void au_cache_fin(void)
21708 + * Make sure all delayed rcu free inodes are flushed before we
21713 + /* excluding AuCache_HNOTIFY */
21714 + BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
21715 + for (i = 0; i < AuCache_HNOTIFY; i++) {
21716 + kmem_cache_destroy(au_cachep[i]);
21717 + au_cachep[i] = NULL;
21721 +/* ---------------------------------------------------------------------- */
21723 +int au_dir_roflags;
21725 +#ifdef CONFIG_AUFS_SBILIST
21727 + * iterate_supers_type() doesn't protect us from
21728 + * remounting (branch management)
21730 +struct au_splhead au_sbilist;
21733 +struct lock_class_key au_lc_key[AuLcKey_Last];
21736 + * functions for module interface.
21738 +MODULE_LICENSE("GPL");
21739 +/* MODULE_LICENSE("GPL v2"); */
21740 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
21741 +MODULE_DESCRIPTION(AUFS_NAME
21742 + " -- Advanced multi layered unification filesystem");
21743 +MODULE_VERSION(AUFS_VERSION);
21744 +MODULE_ALIAS_FS(AUFS_NAME);
21746 +/* this module parameter has no meaning when SYSFS is disabled */
21747 +int sysaufs_brs = 1;
21748 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
21749 +module_param_named(brs, sysaufs_brs, int, S_IRUGO);
21751 +/* this module parameter has no meaning when USER_NS is disabled */
21752 +static bool au_userns;
21753 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
21754 +module_param_named(allow_userns, au_userns, bool, S_IRUGO);
21756 +/* ---------------------------------------------------------------------- */
21758 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
21760 +int au_seq_path(struct seq_file *seq, struct path *path)
21764 + err = seq_path(seq, path, au_esc_chars);
21767 + else if (err < 0)
21773 +/* ---------------------------------------------------------------------- */
21775 +static int __init aufs_init(void)
21780 + p = au_esc_chars;
21781 + for (i = 1; i <= ' '; i++)
21787 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
21789 + memcpy(aufs_iop_nogetattr, aufs_iop, sizeof(aufs_iop));
21790 + for (i = 0; i < AuIop_Last; i++)
21791 + aufs_iop_nogetattr[i].getattr = NULL;
21793 + au_sbilist_init();
21794 + sysaufs_brs_init();
21797 + err = sysaufs_init();
21798 + if (unlikely(err))
21800 + err = au_procfs_init();
21801 + if (unlikely(err))
21802 + goto out_sysaufs;
21803 + err = au_wkq_init();
21804 + if (unlikely(err))
21806 + err = au_loopback_init();
21807 + if (unlikely(err))
21809 + err = au_hnotify_init();
21810 + if (unlikely(err))
21811 + goto out_loopback;
21812 + err = au_sysrq_init();
21813 + if (unlikely(err))
21815 + err = au_cache_init();
21816 + if (unlikely(err))
21819 + aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
21820 + err = register_filesystem(&aufs_fs_type);
21821 + if (unlikely(err))
21824 + /* since we define pr_fmt, call printk directly */
21825 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
21826 + goto out; /* success */
21833 + au_hnotify_fin();
21835 + au_loopback_fin();
21847 +static void __exit aufs_exit(void)
21849 + unregister_filesystem(&aufs_fs_type);
21852 + au_hnotify_fin();
21853 + au_loopback_fin();
21860 +module_init(aufs_init);
21861 +module_exit(aufs_exit);
21862 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
21863 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
21864 +++ linux/fs/aufs/module.h 2016-01-13 20:11:11.669760262 +0100
21867 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21869 + * This program, aufs is free software; you can redistribute it and/or modify
21870 + * it under the terms of the GNU General Public License as published by
21871 + * the Free Software Foundation; either version 2 of the License, or
21872 + * (at your option) any later version.
21874 + * This program is distributed in the hope that it will be useful,
21875 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21876 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21877 + * GNU General Public License for more details.
21879 + * You should have received a copy of the GNU General Public License
21880 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21884 + * module initialization and module-global
21887 +#ifndef __AUFS_MODULE_H__
21888 +#define __AUFS_MODULE_H__
21892 +#include <linux/slab.h>
21897 +/* module parameters */
21898 +extern int sysaufs_brs;
21900 +/* ---------------------------------------------------------------------- */
21902 +extern int au_dir_roflags;
21905 + AuLcNonDir_FIINFO,
21906 + AuLcNonDir_DIINFO,
21907 + AuLcNonDir_IIINFO,
21913 + AuLcSymlink_DIINFO,
21914 + AuLcSymlink_IIINFO,
21918 +extern struct lock_class_key au_lc_key[AuLcKey_Last];
21920 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp);
21921 +int au_seq_path(struct seq_file *seq, struct path *path);
21923 +#ifdef CONFIG_PROC_FS
21925 +int __init au_procfs_init(void);
21926 +void au_procfs_fin(void);
21928 +AuStubInt0(au_procfs_init, void);
21929 +AuStubVoid(au_procfs_fin, void);
21932 +/* ---------------------------------------------------------------------- */
21941 + AuCache_HNOTIFY, /* must be last */
21945 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
21946 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
21947 +#define AuCacheCtor(type, ctor) \
21948 + kmem_cache_create(#type, sizeof(struct type), \
21949 + __alignof__(struct type), AuCacheFlags, ctor)
21951 +extern struct kmem_cache *au_cachep[];
21953 +#define AuCacheFuncs(name, index) \
21954 +static inline struct au_##name *au_cache_alloc_##name(void) \
21955 +{ return kmem_cache_alloc(au_cachep[AuCache_##index], GFP_NOFS); } \
21956 +static inline void au_cache_free_##name(struct au_##name *p) \
21957 +{ kmem_cache_free(au_cachep[AuCache_##index], p); }
21959 +AuCacheFuncs(dinfo, DINFO);
21960 +AuCacheFuncs(icntnr, ICNTNR);
21961 +AuCacheFuncs(finfo, FINFO);
21962 +AuCacheFuncs(vdir, VDIR);
21963 +AuCacheFuncs(vdir_dehstr, DEHSTR);
21964 +#ifdef CONFIG_AUFS_HNOTIFY
21965 +AuCacheFuncs(hnotify, HNOTIFY);
21968 +#endif /* __KERNEL__ */
21969 +#endif /* __AUFS_MODULE_H__ */
21970 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
21971 --- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
21972 +++ linux/fs/aufs/mvdown.c 2016-01-13 20:11:11.669760262 +0100
21975 + * Copyright (C) 2011-2015 Junjiro R. Okajima
21977 + * This program, aufs is free software; you can redistribute it and/or modify
21978 + * it under the terms of the GNU General Public License as published by
21979 + * the Free Software Foundation; either version 2 of the License, or
21980 + * (at your option) any later version.
21982 + * This program is distributed in the hope that it will be useful,
21983 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21984 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21985 + * GNU General Public License for more details.
21987 + * You should have received a copy of the GNU General Public License
21988 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21992 + * move-down, opposite of copy-up
21997 +struct au_mvd_args {
21999 + struct super_block *h_sb;
22000 + struct dentry *h_parent;
22001 + struct au_hinode *hdir;
22002 + struct inode *h_dir, *h_inode;
22003 + struct au_pin pin;
22004 + } info[AUFS_MVDOWN_NARRAY];
22006 + struct aufs_mvdown mvdown;
22007 + struct dentry *dentry, *parent;
22008 + struct inode *inode, *dir;
22009 + struct super_block *sb;
22010 + aufs_bindex_t bopq, bwh, bfound;
22011 + unsigned char rename_lock;
22014 +#define mvd_errno mvdown.au_errno
22015 +#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
22016 +#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
22017 +#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
22018 +#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
22020 +#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
22021 +#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
22022 +#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
22023 +#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
22024 +#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
22025 +#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
22027 +#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
22028 +#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
22029 +#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
22030 +#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
22031 +#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
22032 +#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
22034 +#define AU_MVD_PR(flag, ...) do { \
22036 + pr_err(__VA_ARGS__); \
22039 +static int find_lower_writable(struct au_mvd_args *a)
22041 + struct super_block *sb;
22042 + aufs_bindex_t bindex, bend;
22043 + struct au_branch *br;
22046 + bindex = a->mvd_bsrc;
22047 + bend = au_sbend(sb);
22048 + if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
22049 + for (bindex++; bindex <= bend; bindex++) {
22050 + br = au_sbr(sb, bindex);
22051 + if (au_br_fhsm(br->br_perm)
22052 + && (!(au_br_sb(br)->s_flags & MS_RDONLY)))
22055 + else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
22056 + for (bindex++; bindex <= bend; bindex++) {
22057 + br = au_sbr(sb, bindex);
22058 + if (!au_br_rdonly(br))
22062 + for (bindex++; bindex <= bend; bindex++) {
22063 + br = au_sbr(sb, bindex);
22064 + if (!(au_br_sb(br)->s_flags & MS_RDONLY)) {
22065 + if (au_br_rdonly(br))
22067 + |= AUFS_MVDOWN_ROLOWER_R;
22075 +/* make the parent dir on bdst */
22076 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
22081 + a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
22082 + a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
22083 + a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
22084 + a->mvd_h_dst_parent = NULL;
22085 + if (au_dbend(a->parent) >= a->mvd_bdst)
22086 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
22087 + if (!a->mvd_h_dst_parent) {
22088 + err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
22089 + if (unlikely(err)) {
22090 + AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
22093 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
22101 +/* lock them all */
22102 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
22105 + struct dentry *h_trap;
22107 + a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
22108 + a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
22109 + err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
22110 + au_opt_udba(a->sb),
22111 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22113 + if (unlikely(err)) {
22114 + AU_MVD_PR(dmsg, "pin_dst failed\n");
22118 + if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
22119 + a->rename_lock = 0;
22120 + au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
22121 + AuLsc_DI_PARENT, AuLsc_I_PARENT3,
22122 + au_opt_udba(a->sb),
22123 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22124 + err = au_do_pin(&a->mvd_pin_src);
22126 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
22127 + if (unlikely(err)) {
22128 + AU_MVD_PR(dmsg, "pin_src failed\n");
22131 + goto out; /* success */
22134 + a->rename_lock = 1;
22135 + au_pin_hdir_unlock(&a->mvd_pin_dst);
22136 + err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
22137 + au_opt_udba(a->sb),
22138 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22140 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
22141 + if (unlikely(err)) {
22142 + AU_MVD_PR(dmsg, "pin_src failed\n");
22143 + au_pin_hdir_lock(&a->mvd_pin_dst);
22146 + au_pin_hdir_unlock(&a->mvd_pin_src);
22147 + h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22148 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
22150 + err = (h_trap != a->mvd_h_src_parent);
22152 + err = (h_trap != a->mvd_h_dst_parent);
22154 + BUG_ON(err); /* it should never happen */
22155 + if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
22158 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22159 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
22160 + au_pin_hdir_lock(&a->mvd_pin_src);
22161 + au_unpin(&a->mvd_pin_src);
22162 + au_pin_hdir_lock(&a->mvd_pin_dst);
22165 + goto out; /* success */
22168 + au_unpin(&a->mvd_pin_dst);
22174 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
22176 + if (!a->rename_lock)
22177 + au_unpin(&a->mvd_pin_src);
22179 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22180 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
22181 + au_pin_hdir_lock(&a->mvd_pin_src);
22182 + au_unpin(&a->mvd_pin_src);
22183 + au_pin_hdir_lock(&a->mvd_pin_dst);
22185 + au_unpin(&a->mvd_pin_dst);
22188 +/* copy-down the file */
22189 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
22192 + struct au_cp_generic cpg = {
22193 + .dentry = a->dentry,
22194 + .bdst = a->mvd_bdst,
22195 + .bsrc = a->mvd_bsrc,
22197 + .pin = &a->mvd_pin_dst,
22198 + .flags = AuCpup_DTIME | AuCpup_HOPEN
22201 + AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
22202 + if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
22203 + au_fset_cpup(cpg.flags, OVERWRITE);
22204 + if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
22205 + au_fset_cpup(cpg.flags, RWDST);
22206 + err = au_sio_cpdown_simple(&cpg);
22207 + if (unlikely(err))
22208 + AU_MVD_PR(dmsg, "cpdown failed\n");
22215 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
22218 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
22221 + struct path h_path;
22222 + struct au_branch *br;
22223 + struct inode *delegated;
22225 + br = au_sbr(a->sb, a->mvd_bdst);
22226 + h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
22227 + err = PTR_ERR(h_path.dentry);
22228 + if (IS_ERR(h_path.dentry)) {
22229 + AU_MVD_PR(dmsg, "wh_lkup failed\n");
22234 + if (d_is_positive(h_path.dentry)) {
22235 + h_path.mnt = au_br_mnt(br);
22236 + delegated = NULL;
22237 + err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
22238 + &delegated, /*force*/0);
22239 + if (unlikely(err == -EWOULDBLOCK)) {
22240 + pr_warn("cannot retry for NFSv4 delegation"
22241 + " for an internal unlink\n");
22244 + if (unlikely(err))
22245 + AU_MVD_PR(dmsg, "wh_unlink failed\n");
22247 + dput(h_path.dentry);
22255 + * unlink the topmost h_dentry
22257 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
22260 + struct path h_path;
22261 + struct inode *delegated;
22263 + h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
22264 + h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
22265 + delegated = NULL;
22266 + err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
22267 + if (unlikely(err == -EWOULDBLOCK)) {
22268 + pr_warn("cannot retry for NFSv4 delegation"
22269 + " for an internal unlink\n");
22272 + if (unlikely(err))
22273 + AU_MVD_PR(dmsg, "unlink failed\n");
22279 +/* Since mvdown succeeded, we ignore an error of this function */
22280 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
22283 + struct au_branch *br;
22285 + a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
22286 + br = au_sbr(a->sb, a->mvd_bsrc);
22287 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
22289 + br = au_sbr(a->sb, a->mvd_bdst);
22290 + a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
22291 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
22294 + a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
22296 + AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
22300 + * copy-down the file and unlink the bsrc file.
22301 + * - unlink the bdst whout if exist
22302 + * - copy-down the file (with whtmp name and rename)
22303 + * - unlink the bsrc file
22305 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
22309 + err = au_do_mkdir(dmsg, a);
22311 + err = au_do_lock(dmsg, a);
22312 + if (unlikely(err))
22316 + * do not revert the activities we made on bdst since they should be
22317 + * harmless in aufs.
22320 + err = au_do_cpdown(dmsg, a);
22322 + err = au_do_unlink_wh(dmsg, a);
22323 + if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
22324 + err = au_do_unlink(dmsg, a);
22325 + if (unlikely(err))
22328 + AuDbg("%pd2, 0x%x, %d --> %d\n",
22329 + a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
22330 + if (find_lower_writable(a) < 0)
22331 + a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
22333 + if (a->mvdown.flags & AUFS_MVDOWN_STFS)
22334 + au_do_stfs(dmsg, a);
22336 + /* maintain internal array */
22337 + if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
22338 + au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
22339 + au_set_dbstart(a->dentry, a->mvd_bdst);
22340 + au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
22341 + au_set_ibstart(a->inode, a->mvd_bdst);
22343 + /* hide the lower */
22344 + au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
22345 + au_set_dbend(a->dentry, a->mvd_bsrc);
22346 + au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
22347 + au_set_ibend(a->inode, a->mvd_bsrc);
22349 + if (au_dbend(a->dentry) < a->mvd_bdst)
22350 + au_set_dbend(a->dentry, a->mvd_bdst);
22351 + if (au_ibend(a->inode) < a->mvd_bdst)
22352 + au_set_ibend(a->inode, a->mvd_bdst);
22355 + au_do_unlock(dmsg, a);
22361 +/* ---------------------------------------------------------------------- */
22363 +/* make sure the file is idle */
22364 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
22366 + int err, plinked;
22369 + plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
22370 + if (au_dbstart(a->dentry) == a->mvd_bsrc
22371 + && au_dcount(a->dentry) == 1
22372 + && atomic_read(&a->inode->i_count) == 1
22373 + /* && a->mvd_h_src_inode->i_nlink == 1 */
22374 + && (!plinked || !au_plink_test(a->inode))
22375 + && a->inode->i_nlink == 1)
22380 + "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
22381 + a->mvd_bsrc, au_dbstart(a->dentry), au_dcount(a->dentry),
22382 + atomic_read(&a->inode->i_count), a->inode->i_nlink,
22383 + a->mvd_h_src_inode->i_nlink,
22384 + plinked, plinked ? au_plink_test(a->inode) : 0);
22391 +/* make sure the parent dir is fine */
22392 +static int au_mvd_args_parent(const unsigned char dmsg,
22393 + struct au_mvd_args *a)
22396 + aufs_bindex_t bindex;
22399 + if (unlikely(au_alive_dir(a->parent))) {
22401 + AU_MVD_PR(dmsg, "parent dir is dead\n");
22405 + a->bopq = au_dbdiropq(a->parent);
22406 + bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
22407 + AuDbg("b%d\n", bindex);
22408 + if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
22409 + || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
22411 + a->mvd_errno = EAU_MVDOWN_OPAQUE;
22412 + AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
22413 + a->bopq, a->mvd_bdst);
22421 +static int au_mvd_args_intermediate(const unsigned char dmsg,
22422 + struct au_mvd_args *a)
22425 + struct au_dinfo *dinfo, *tmp;
22427 + /* lookup the next lower positive entry */
22429 + tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
22430 + if (unlikely(!tmp))
22435 + dinfo = au_di(a->dentry);
22436 + au_di_cp(tmp, dinfo);
22437 + au_di_swap(tmp, dinfo);
22439 + /* returns the number of positive dentries */
22440 + err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1, /*type*/0);
22442 + a->bwh = au_dbwh(a->dentry);
22443 + else if (err > 0)
22444 + a->bfound = au_dbstart(a->dentry);
22446 + au_di_swap(tmp, dinfo);
22447 + au_rw_write_unlock(&tmp->di_rwsem);
22449 + if (unlikely(err < 0))
22450 + AU_MVD_PR(dmsg, "failed look-up lower\n");
22453 + * here, we have these cases.
22455 + * no positive dentry under bsrc. there are more sub-cases.
22457 + * there no whiteout, we can safely move-down.
22460 + * bsrc < bwh && bwh < bdst
22461 + * there is a whiteout on RO branch. cannot proceed.
22463 + * there is a whiteout on the RW target branch. it should
22466 + * there is a whiteout somewhere unrelated branch.
22467 + * -1 < bfound && bfound <= bsrc
22470 + * found, but it is on RO branch between bsrc and bdst. cannot
22473 + * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
22476 + * found, after we create the file on bdst, it will be hidden.
22479 + AuDebugOn(a->bfound == -1
22481 + && a->bwh <= a->mvd_bsrc);
22482 + AuDebugOn(-1 < a->bfound
22483 + && a->bfound <= a->mvd_bsrc);
22486 + if (a->bfound == -1
22487 + && a->mvd_bsrc < a->bwh
22489 + && a->bwh < a->mvd_bdst) {
22490 + a->mvd_errno = EAU_MVDOWN_WHITEOUT;
22491 + AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
22492 + a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
22494 + } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
22495 + a->mvd_errno = EAU_MVDOWN_UPPER;
22496 + AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
22497 + a->mvd_bdst, a->bfound);
22501 + err = 0; /* success */
22508 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
22513 + if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
22514 + && a->bfound == a->mvd_bdst)
22520 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
22523 + struct au_branch *br;
22526 + if (unlikely(S_ISDIR(a->inode->i_mode)))
22530 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
22531 + a->mvd_bsrc = au_ibstart(a->inode);
22533 + a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
22534 + if (unlikely(a->mvd_bsrc < 0
22535 + || (a->mvd_bsrc < au_dbstart(a->dentry)
22536 + || au_dbend(a->dentry) < a->mvd_bsrc
22537 + || !au_h_dptr(a->dentry, a->mvd_bsrc))
22538 + || (a->mvd_bsrc < au_ibstart(a->inode)
22539 + || au_ibend(a->inode) < a->mvd_bsrc
22540 + || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
22541 + a->mvd_errno = EAU_MVDOWN_NOUPPER;
22542 + AU_MVD_PR(dmsg, "no upper\n");
22546 + if (unlikely(a->mvd_bsrc == au_sbend(a->sb))) {
22547 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
22548 + AU_MVD_PR(dmsg, "on the bottom\n");
22551 + a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
22552 + br = au_sbr(a->sb, a->mvd_bsrc);
22553 + err = au_br_rdonly(br);
22554 + if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
22555 + if (unlikely(err))
22557 + } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
22558 + || IS_APPEND(a->mvd_h_src_inode))) {
22560 + a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
22566 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
22567 + a->mvd_bdst = find_lower_writable(a);
22568 + if (unlikely(a->mvd_bdst < 0)) {
22569 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
22570 + AU_MVD_PR(dmsg, "no writable lower branch\n");
22574 + a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
22575 + if (unlikely(a->mvd_bdst < 0
22576 + || au_sbend(a->sb) < a->mvd_bdst)) {
22577 + a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
22578 + AU_MVD_PR(dmsg, "no lower brid\n");
22583 + err = au_mvd_args_busy(dmsg, a);
22585 + err = au_mvd_args_parent(dmsg, a);
22587 + err = au_mvd_args_intermediate(dmsg, a);
22589 + err = au_mvd_args_exist(dmsg, a);
22591 + AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
22598 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
22601 + unsigned char dmsg;
22602 + struct au_mvd_args *args;
22603 + struct inode *inode;
22605 + inode = d_inode(dentry);
22607 + if (unlikely(!capable(CAP_SYS_ADMIN)))
22611 + args = kmalloc(sizeof(*args), GFP_NOFS);
22612 + if (unlikely(!args))
22615 + err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
22617 + err = !access_ok(VERIFY_WRITE, uarg, sizeof(*uarg));
22618 + if (unlikely(err)) {
22623 + AuDbg("flags 0x%x\n", args->mvdown.flags);
22624 + args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
22625 + args->mvdown.au_errno = 0;
22626 + args->dentry = dentry;
22627 + args->inode = inode;
22628 + args->sb = dentry->d_sb;
22631 + dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
22632 + args->parent = dget_parent(dentry);
22633 + args->dir = d_inode(args->parent);
22634 + mutex_lock_nested(&args->dir->i_mutex, I_MUTEX_PARENT);
22635 + dput(args->parent);
22636 + if (unlikely(args->parent != dentry->d_parent)) {
22637 + AU_MVD_PR(dmsg, "parent dir is moved\n");
22641 + mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
22642 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_NOPLMW);
22643 + if (unlikely(err))
22646 + di_write_lock_parent(args->parent);
22647 + err = au_mvd_args(dmsg, args);
22648 + if (unlikely(err))
22651 + err = au_do_mvdown(dmsg, args);
22652 + if (unlikely(err))
22655 + au_cpup_attr_timesizes(args->dir);
22656 + au_cpup_attr_timesizes(inode);
22657 + if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
22658 + au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
22659 + /* au_digen_dec(dentry); */
22662 + di_write_unlock(args->parent);
22663 + aufs_read_unlock(dentry, AuLock_DW);
22665 + mutex_unlock(&inode->i_mutex);
22667 + mutex_unlock(&args->dir->i_mutex);
22669 + e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
22677 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
22678 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
22679 +++ linux/fs/aufs/opts.c 2016-01-13 20:11:11.673093671 +0100
22682 + * Copyright (C) 2005-2015 Junjiro R. Okajima
22684 + * This program, aufs is free software; you can redistribute it and/or modify
22685 + * it under the terms of the GNU General Public License as published by
22686 + * the Free Software Foundation; either version 2 of the License, or
22687 + * (at your option) any later version.
22689 + * This program is distributed in the hope that it will be useful,
22690 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22691 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22692 + * GNU General Public License for more details.
22694 + * You should have received a copy of the GNU General Public License
22695 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
22699 + * mount options/flags
22702 +#include <linux/namei.h>
22703 +#include <linux/types.h> /* a distribution requires */
22704 +#include <linux/parser.h>
22707 +/* ---------------------------------------------------------------------- */
22711 + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
22712 + Opt_idel, Opt_imod,
22713 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
22714 + Opt_rdblk_def, Opt_rdhash_def,
22715 + Opt_xino, Opt_noxino,
22716 + Opt_trunc_xino, Opt_trunc_xino_v, Opt_notrunc_xino,
22717 + Opt_trunc_xino_path, Opt_itrunc_xino,
22718 + Opt_trunc_xib, Opt_notrunc_xib,
22719 + Opt_shwh, Opt_noshwh,
22720 + Opt_plink, Opt_noplink, Opt_list_plink,
22722 + Opt_dio, Opt_nodio,
22723 + Opt_diropq_a, Opt_diropq_w,
22724 + Opt_warn_perm, Opt_nowarn_perm,
22725 + Opt_wbr_copyup, Opt_wbr_create,
22727 + Opt_verbose, Opt_noverbose,
22728 + Opt_sum, Opt_nosum, Opt_wsum,
22729 + Opt_dirperm1, Opt_nodirperm1,
22730 + Opt_acl, Opt_noacl,
22731 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
22734 +static match_table_t options = {
22735 + {Opt_br, "br=%s"},
22736 + {Opt_br, "br:%s"},
22738 + {Opt_add, "add=%d:%s"},
22739 + {Opt_add, "add:%d:%s"},
22740 + {Opt_add, "ins=%d:%s"},
22741 + {Opt_add, "ins:%d:%s"},
22742 + {Opt_append, "append=%s"},
22743 + {Opt_append, "append:%s"},
22744 + {Opt_prepend, "prepend=%s"},
22745 + {Opt_prepend, "prepend:%s"},
22747 + {Opt_del, "del=%s"},
22748 + {Opt_del, "del:%s"},
22749 + /* {Opt_idel, "idel:%d"}, */
22750 + {Opt_mod, "mod=%s"},
22751 + {Opt_mod, "mod:%s"},
22752 + /* {Opt_imod, "imod:%d:%s"}, */
22754 + {Opt_dirwh, "dirwh=%d"},
22756 + {Opt_xino, "xino=%s"},
22757 + {Opt_noxino, "noxino"},
22758 + {Opt_trunc_xino, "trunc_xino"},
22759 + {Opt_trunc_xino_v, "trunc_xino_v=%d:%d"},
22760 + {Opt_notrunc_xino, "notrunc_xino"},
22761 + {Opt_trunc_xino_path, "trunc_xino=%s"},
22762 + {Opt_itrunc_xino, "itrunc_xino=%d"},
22763 + /* {Opt_zxino, "zxino=%s"}, */
22764 + {Opt_trunc_xib, "trunc_xib"},
22765 + {Opt_notrunc_xib, "notrunc_xib"},
22767 +#ifdef CONFIG_PROC_FS
22768 + {Opt_plink, "plink"},
22770 + {Opt_ignore_silent, "plink"},
22773 + {Opt_noplink, "noplink"},
22775 +#ifdef CONFIG_AUFS_DEBUG
22776 + {Opt_list_plink, "list_plink"},
22779 + {Opt_udba, "udba=%s"},
22781 + {Opt_dio, "dio"},
22782 + {Opt_nodio, "nodio"},
22784 +#ifdef CONFIG_AUFS_FHSM
22785 + {Opt_fhsm_sec, "fhsm_sec=%d"},
22787 + {Opt_ignore_silent, "fhsm_sec=%d"},
22790 + {Opt_diropq_a, "diropq=always"},
22791 + {Opt_diropq_a, "diropq=a"},
22792 + {Opt_diropq_w, "diropq=whiteouted"},
22793 + {Opt_diropq_w, "diropq=w"},
22795 + {Opt_warn_perm, "warn_perm"},
22796 + {Opt_nowarn_perm, "nowarn_perm"},
22798 + /* keep them temporary */
22799 + {Opt_ignore_silent, "nodlgt"},
22800 + {Opt_ignore_silent, "clean_plink"},
22802 +#ifdef CONFIG_AUFS_SHWH
22803 + {Opt_shwh, "shwh"},
22805 + {Opt_noshwh, "noshwh"},
22807 + {Opt_dirperm1, "dirperm1"},
22808 + {Opt_nodirperm1, "nodirperm1"},
22810 + {Opt_verbose, "verbose"},
22811 + {Opt_verbose, "v"},
22812 + {Opt_noverbose, "noverbose"},
22813 + {Opt_noverbose, "quiet"},
22814 + {Opt_noverbose, "q"},
22815 + {Opt_noverbose, "silent"},
22817 + {Opt_sum, "sum"},
22818 + {Opt_nosum, "nosum"},
22819 + {Opt_wsum, "wsum"},
22821 + {Opt_rdcache, "rdcache=%d"},
22822 + {Opt_rdblk, "rdblk=%d"},
22823 + {Opt_rdblk_def, "rdblk=def"},
22824 + {Opt_rdhash, "rdhash=%d"},
22825 + {Opt_rdhash_def, "rdhash=def"},
22827 + {Opt_wbr_create, "create=%s"},
22828 + {Opt_wbr_create, "create_policy=%s"},
22829 + {Opt_wbr_copyup, "cpup=%s"},
22830 + {Opt_wbr_copyup, "copyup=%s"},
22831 + {Opt_wbr_copyup, "copyup_policy=%s"},
22833 + /* generic VFS flag */
22834 +#ifdef CONFIG_FS_POSIX_ACL
22835 + {Opt_acl, "acl"},
22836 + {Opt_noacl, "noacl"},
22838 + {Opt_ignore_silent, "acl"},
22839 + {Opt_ignore_silent, "noacl"},
22842 + /* internal use for the scripts */
22843 + {Opt_ignore_silent, "si=%s"},
22845 + {Opt_br, "dirs=%s"},
22846 + {Opt_ignore, "debug=%d"},
22847 + {Opt_ignore, "delete=whiteout"},
22848 + {Opt_ignore, "delete=all"},
22849 + {Opt_ignore, "imap=%s"},
22851 + /* temporary workaround, due to old mount(8)? */
22852 + {Opt_ignore_silent, "relatime"},
22857 +/* ---------------------------------------------------------------------- */
22859 +static const char *au_parser_pattern(int val, match_table_t tbl)
22861 + struct match_token *p;
22864 + while (p->pattern) {
22865 + if (p->token == val)
22866 + return p->pattern;
22873 +static const char *au_optstr(int *val, match_table_t tbl)
22875 + struct match_token *p;
22882 + while (p->pattern) {
22884 + && (v & p->token) == p->token) {
22885 + *val &= ~p->token;
22886 + return p->pattern;
22895 +/* ---------------------------------------------------------------------- */
22897 +static match_table_t brperm = {
22898 + {AuBrPerm_RO, AUFS_BRPERM_RO},
22899 + {AuBrPerm_RR, AUFS_BRPERM_RR},
22900 + {AuBrPerm_RW, AUFS_BRPERM_RW},
22904 +static match_table_t brattr = {
22906 + {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
22907 + {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
22908 + /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
22909 + {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
22910 +#ifdef CONFIG_AUFS_FHSM
22911 + {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
22913 +#ifdef CONFIG_AUFS_XATTR
22914 + {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
22915 + {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
22916 + {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
22917 + {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
22918 + {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
22919 + {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
22922 + /* ro/rr branch */
22923 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
22926 + {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
22927 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
22932 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
22939 + p = strchr(str, '+');
22942 + v = match_token(str, table, args);
22944 + if (v & AuBrAttr_CMOO_Mask)
22945 + attr &= ~AuBrAttr_CMOO_Mask;
22950 + pr_warn("ignored branch attribute %s\n", str);
22960 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
22968 + p = au_optstr(&perm, brattr);
22971 + memcpy(q, p, sz + 1);
22977 + p = au_optstr(&perm, brattr);
22981 + memcpy(q, p, sz + 1);
22987 + return q - str->a;
22990 +static int noinline_for_stack br_perm_val(char *perm)
22992 + int val, bad, sz;
22994 + substring_t args[MAX_OPT_ARGS];
22995 + au_br_perm_str_t attr;
22997 + p = strchr(perm, '+');
23000 + val = match_token(perm, brperm, args);
23004 + pr_warn("ignored branch permission %s\n", perm);
23005 + val = AuBrPerm_RO;
23011 + val |= br_attr_val(p + 1, brattr, args);
23014 + switch (val & AuBrPerm_Mask) {
23015 + case AuBrPerm_RO:
23016 + case AuBrPerm_RR:
23017 + bad = val & AuBrWAttr_Mask;
23018 + val &= ~AuBrWAttr_Mask;
23020 + case AuBrPerm_RW:
23021 + bad = val & AuBrRAttr_Mask;
23022 + val &= ~AuBrRAttr_Mask;
23027 + * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
23028 + * does not treat it as an error, just warning.
23029 + * this is a tiny guard for the user operation.
23031 + if (val & AuBrAttr_UNPIN) {
23032 + bad |= AuBrAttr_UNPIN;
23033 + val &= ~AuBrAttr_UNPIN;
23036 + if (unlikely(bad)) {
23037 + sz = au_do_optstr_br_attr(&attr, bad);
23039 + pr_warn("ignored branch attribute %s\n", attr.a);
23046 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
23048 + au_br_perm_str_t attr;
23054 + p = au_optstr(&perm, brperm);
23055 + AuDebugOn(!p || !*p);
23057 + memcpy(q, p, sz + 1);
23060 + sz = au_do_optstr_br_attr(&attr, perm);
23063 + memcpy(q, attr.a, sz + 1);
23066 + AuDebugOn(strlen(str->a) >= sizeof(str->a));
23069 +/* ---------------------------------------------------------------------- */
23071 +static match_table_t udbalevel = {
23072 + {AuOpt_UDBA_REVAL, "reval"},
23073 + {AuOpt_UDBA_NONE, "none"},
23074 +#ifdef CONFIG_AUFS_HNOTIFY
23075 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
23076 +#ifdef CONFIG_AUFS_HFSNOTIFY
23077 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
23083 +static int noinline_for_stack udba_val(char *str)
23085 + substring_t args[MAX_OPT_ARGS];
23087 + return match_token(str, udbalevel, args);
23090 +const char *au_optstr_udba(int udba)
23092 + return au_parser_pattern(udba, udbalevel);
23095 +/* ---------------------------------------------------------------------- */
23097 +static match_table_t au_wbr_create_policy = {
23098 + {AuWbrCreate_TDP, "tdp"},
23099 + {AuWbrCreate_TDP, "top-down-parent"},
23100 + {AuWbrCreate_RR, "rr"},
23101 + {AuWbrCreate_RR, "round-robin"},
23102 + {AuWbrCreate_MFS, "mfs"},
23103 + {AuWbrCreate_MFS, "most-free-space"},
23104 + {AuWbrCreate_MFSV, "mfs:%d"},
23105 + {AuWbrCreate_MFSV, "most-free-space:%d"},
23107 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
23108 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
23109 + {AuWbrCreate_PMFS, "pmfs"},
23110 + {AuWbrCreate_PMFSV, "pmfs:%d"},
23111 + {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
23112 + {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
23118 + * cf. linux/lib/parser.c and cmdline.c
23119 + * gave up calling memparse() since it uses simple_strtoull() instead of
23122 +static int noinline_for_stack
23123 +au_match_ull(substring_t *s, unsigned long long *result)
23126 + unsigned int len;
23130 + len = s->to - s->from;
23131 + if (len + 1 <= sizeof(a)) {
23132 + memcpy(a, s->from, len);
23134 + err = kstrtoull(a, 0, result);
23139 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
23140 + struct au_opt_wbr_create *create)
23143 + unsigned long long ull;
23146 + if (!au_match_ull(arg, &ull))
23147 + create->mfsrr_watermark = ull;
23149 + pr_err("bad integer in %s\n", str);
23156 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
23157 + struct au_opt_wbr_create *create)
23162 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
23163 + create->mfs_second = n;
23165 + pr_err("bad integer in %s\n", str);
23172 +static int noinline_for_stack
23173 +au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
23176 + substring_t args[MAX_OPT_ARGS];
23178 + err = match_token(str, au_wbr_create_policy, args);
23179 + create->wbr_create = err;
23181 + case AuWbrCreate_MFSRRV:
23182 + case AuWbrCreate_PMFSRRV:
23183 + e = au_wbr_mfs_wmark(&args[0], str, create);
23185 + e = au_wbr_mfs_sec(&args[1], str, create);
23189 + case AuWbrCreate_MFSRR:
23190 + case AuWbrCreate_PMFSRR:
23191 + e = au_wbr_mfs_wmark(&args[0], str, create);
23192 + if (unlikely(e)) {
23197 + case AuWbrCreate_MFS:
23198 + case AuWbrCreate_PMFS:
23199 + create->mfs_second = AUFS_MFS_DEF_SEC;
23201 + case AuWbrCreate_MFSV:
23202 + case AuWbrCreate_PMFSV:
23203 + e = au_wbr_mfs_sec(&args[0], str, create);
23212 +const char *au_optstr_wbr_create(int wbr_create)
23214 + return au_parser_pattern(wbr_create, au_wbr_create_policy);
23217 +static match_table_t au_wbr_copyup_policy = {
23218 + {AuWbrCopyup_TDP, "tdp"},
23219 + {AuWbrCopyup_TDP, "top-down-parent"},
23220 + {AuWbrCopyup_BUP, "bup"},
23221 + {AuWbrCopyup_BUP, "bottom-up-parent"},
23222 + {AuWbrCopyup_BU, "bu"},
23223 + {AuWbrCopyup_BU, "bottom-up"},
23227 +static int noinline_for_stack au_wbr_copyup_val(char *str)
23229 + substring_t args[MAX_OPT_ARGS];
23231 + return match_token(str, au_wbr_copyup_policy, args);
23234 +const char *au_optstr_wbr_copyup(int wbr_copyup)
23236 + return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
23239 +/* ---------------------------------------------------------------------- */
23241 +static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
23243 +static void dump_opts(struct au_opts *opts)
23245 +#ifdef CONFIG_AUFS_DEBUG
23246 + /* reduce stack space */
23248 + struct au_opt_add *add;
23249 + struct au_opt_del *del;
23250 + struct au_opt_mod *mod;
23251 + struct au_opt_xino *xino;
23252 + struct au_opt_xino_itrunc *xino_itrunc;
23253 + struct au_opt_wbr_create *create;
23255 + struct au_opt *opt;
23258 + while (opt->type != Opt_tail) {
23259 + switch (opt->type) {
23261 + u.add = &opt->add;
23262 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
23263 + u.add->bindex, u.add->pathname, u.add->perm,
23264 + u.add->path.dentry);
23268 + u.del = &opt->del;
23269 + AuDbg("del {%s, %p}\n",
23270 + u.del->pathname, u.del->h_path.dentry);
23274 + u.mod = &opt->mod;
23275 + AuDbg("mod {%s, 0x%x, %p}\n",
23276 + u.mod->path, u.mod->perm, u.mod->h_root);
23279 + u.add = &opt->add;
23280 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
23281 + u.add->bindex, u.add->pathname, u.add->perm,
23282 + u.add->path.dentry);
23284 + case Opt_prepend:
23285 + u.add = &opt->add;
23286 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
23287 + u.add->bindex, u.add->pathname, u.add->perm,
23288 + u.add->path.dentry);
23291 + AuDbg("dirwh %d\n", opt->dirwh);
23293 + case Opt_rdcache:
23294 + AuDbg("rdcache %d\n", opt->rdcache);
23297 + AuDbg("rdblk %u\n", opt->rdblk);
23299 + case Opt_rdblk_def:
23300 + AuDbg("rdblk_def\n");
23303 + AuDbg("rdhash %u\n", opt->rdhash);
23305 + case Opt_rdhash_def:
23306 + AuDbg("rdhash_def\n");
23309 + u.xino = &opt->xino;
23310 + AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
23312 + case Opt_trunc_xino:
23313 + AuLabel(trunc_xino);
23315 + case Opt_notrunc_xino:
23316 + AuLabel(notrunc_xino);
23318 + case Opt_trunc_xino_path:
23319 + case Opt_itrunc_xino:
23320 + u.xino_itrunc = &opt->xino_itrunc;
23321 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
23326 + case Opt_trunc_xib:
23327 + AuLabel(trunc_xib);
23329 + case Opt_notrunc_xib:
23330 + AuLabel(notrunc_xib);
23338 + case Opt_dirperm1:
23339 + AuLabel(dirperm1);
23341 + case Opt_nodirperm1:
23342 + AuLabel(nodirperm1);
23347 + case Opt_noplink:
23348 + AuLabel(noplink);
23350 + case Opt_list_plink:
23351 + AuLabel(list_plink);
23354 + AuDbg("udba %d, %s\n",
23355 + opt->udba, au_optstr_udba(opt->udba));
23363 + case Opt_diropq_a:
23364 + AuLabel(diropq_a);
23366 + case Opt_diropq_w:
23367 + AuLabel(diropq_w);
23369 + case Opt_warn_perm:
23370 + AuLabel(warn_perm);
23372 + case Opt_nowarn_perm:
23373 + AuLabel(nowarn_perm);
23375 + case Opt_verbose:
23376 + AuLabel(verbose);
23378 + case Opt_noverbose:
23379 + AuLabel(noverbose);
23390 + case Opt_wbr_create:
23391 + u.create = &opt->wbr_create;
23392 + AuDbg("create %d, %s\n", u.create->wbr_create,
23393 + au_optstr_wbr_create(u.create->wbr_create));
23394 + switch (u.create->wbr_create) {
23395 + case AuWbrCreate_MFSV:
23396 + case AuWbrCreate_PMFSV:
23397 + AuDbg("%d sec\n", u.create->mfs_second);
23399 + case AuWbrCreate_MFSRR:
23400 + AuDbg("%llu watermark\n",
23401 + u.create->mfsrr_watermark);
23403 + case AuWbrCreate_MFSRRV:
23404 + case AuWbrCreate_PMFSRRV:
23405 + AuDbg("%llu watermark, %d sec\n",
23406 + u.create->mfsrr_watermark,
23407 + u.create->mfs_second);
23411 + case Opt_wbr_copyup:
23412 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
23413 + au_optstr_wbr_copyup(opt->wbr_copyup));
23415 + case Opt_fhsm_sec:
23416 + AuDbg("fhsm_sec %u\n", opt->fhsm_second);
23432 +void au_opts_free(struct au_opts *opts)
23434 + struct au_opt *opt;
23437 + while (opt->type != Opt_tail) {
23438 + switch (opt->type) {
23441 + case Opt_prepend:
23442 + path_put(&opt->add.path);
23446 + path_put(&opt->del.h_path);
23450 + dput(opt->mod.h_root);
23453 + fput(opt->xino.file);
23460 +static int opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
23461 + aufs_bindex_t bindex)
23464 + struct au_opt_add *add = &opt->add;
23467 + add->bindex = bindex;
23468 + add->perm = AuBrPerm_RO;
23469 + add->pathname = opt_str;
23470 + p = strchr(opt_str, '=');
23474 + add->perm = br_perm_val(p);
23477 + err = vfsub_kern_path(add->pathname, lkup_dirflags, &add->path);
23480 + add->perm = AuBrPerm_RO;
23481 + if (au_test_fs_rr(add->path.dentry->d_sb))
23482 + add->perm = AuBrPerm_RR;
23483 + else if (!bindex && !(sb_flags & MS_RDONLY))
23484 + add->perm = AuBrPerm_RW;
23486 + opt->type = Opt_add;
23489 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
23496 +static int au_opts_parse_del(struct au_opt_del *del, substring_t args[])
23500 + del->pathname = args[0].from;
23501 + AuDbg("del path %s\n", del->pathname);
23503 + err = vfsub_kern_path(del->pathname, lkup_dirflags, &del->h_path);
23504 + if (unlikely(err))
23505 + pr_err("lookup failed %s (%d)\n", del->pathname, err);
23510 +#if 0 /* reserved for future use */
23511 +static int au_opts_parse_idel(struct super_block *sb, aufs_bindex_t bindex,
23512 + struct au_opt_del *del, substring_t args[])
23515 + struct dentry *root;
23518 + root = sb->s_root;
23519 + aufs_read_lock(root, AuLock_FLUSH);
23520 + if (bindex < 0 || au_sbend(sb) < bindex) {
23521 + pr_err("out of bounds, %d\n", bindex);
23526 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
23527 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
23530 + aufs_read_unlock(root, !AuLock_IR);
23535 +static int noinline_for_stack
23536 +au_opts_parse_mod(struct au_opt_mod *mod, substring_t args[])
23539 + struct path path;
23543 + mod->path = args[0].from;
23544 + p = strchr(mod->path, '=');
23545 + if (unlikely(!p)) {
23546 + pr_err("no permssion %s\n", args[0].from);
23551 + err = vfsub_kern_path(mod->path, lkup_dirflags, &path);
23552 + if (unlikely(err)) {
23553 + pr_err("lookup failed %s (%d)\n", mod->path, err);
23557 + mod->perm = br_perm_val(p);
23558 + AuDbg("mod path %s, perm 0x%x, %s\n", mod->path, mod->perm, p);
23559 + mod->h_root = dget(path.dentry);
23566 +#if 0 /* reserved for future use */
23567 +static int au_opts_parse_imod(struct super_block *sb, aufs_bindex_t bindex,
23568 + struct au_opt_mod *mod, substring_t args[])
23571 + struct dentry *root;
23574 + root = sb->s_root;
23575 + aufs_read_lock(root, AuLock_FLUSH);
23576 + if (bindex < 0 || au_sbend(sb) < bindex) {
23577 + pr_err("out of bounds, %d\n", bindex);
23582 + mod->perm = br_perm_val(args[1].from);
23583 + AuDbg("mod path %s, perm 0x%x, %s\n",
23584 + mod->path, mod->perm, args[1].from);
23585 + mod->h_root = dget(au_h_dptr(root, bindex));
23588 + aufs_read_unlock(root, !AuLock_IR);
23593 +static int au_opts_parse_xino(struct super_block *sb, struct au_opt_xino *xino,
23594 + substring_t args[])
23597 + struct file *file;
23599 + file = au_xino_create(sb, args[0].from, /*silent*/0);
23600 + err = PTR_ERR(file);
23601 + if (IS_ERR(file))
23605 + if (unlikely(file->f_path.dentry->d_sb == sb)) {
23607 + pr_err("%s must be outside\n", args[0].from);
23612 + xino->file = file;
23613 + xino->path = args[0].from;
23619 +static int noinline_for_stack
23620 +au_opts_parse_xino_itrunc_path(struct super_block *sb,
23621 + struct au_opt_xino_itrunc *xino_itrunc,
23622 + substring_t args[])
23625 + aufs_bindex_t bend, bindex;
23626 + struct path path;
23627 + struct dentry *root;
23629 + err = vfsub_kern_path(args[0].from, lkup_dirflags, &path);
23630 + if (unlikely(err)) {
23631 + pr_err("lookup failed %s (%d)\n", args[0].from, err);
23635 + xino_itrunc->bindex = -1;
23636 + root = sb->s_root;
23637 + aufs_read_lock(root, AuLock_FLUSH);
23638 + bend = au_sbend(sb);
23639 + for (bindex = 0; bindex <= bend; bindex++) {
23640 + if (au_h_dptr(root, bindex) == path.dentry) {
23641 + xino_itrunc->bindex = bindex;
23645 + aufs_read_unlock(root, !AuLock_IR);
23648 + if (unlikely(xino_itrunc->bindex < 0)) {
23649 + pr_err("no such branch %s\n", args[0].from);
23657 +/* called without aufs lock */
23658 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts)
23660 + int err, n, token;
23661 + aufs_bindex_t bindex;
23662 + unsigned char skipped;
23663 + struct dentry *root;
23664 + struct au_opt *opt, *opt_tail;
23666 + /* reduce the stack space */
23668 + struct au_opt_xino_itrunc *xino_itrunc;
23669 + struct au_opt_wbr_create *create;
23672 + substring_t args[MAX_OPT_ARGS];
23676 + a = kmalloc(sizeof(*a), GFP_NOFS);
23677 + if (unlikely(!a))
23680 + root = sb->s_root;
23684 + opt_tail = opt + opts->max_opt - 1;
23685 + opt->type = Opt_tail;
23686 + while (!err && (opt_str = strsep(&str, ",")) && *opt_str) {
23689 + token = match_token(opt_str, options, a->args);
23693 + while (!err && (opt_str = strsep(&a->args[0].from, ":"))
23695 + err = opt_add(opt, opt_str, opts->sb_flags,
23697 + if (unlikely(!err && ++opt > opt_tail)) {
23701 + opt->type = Opt_tail;
23706 + if (unlikely(match_int(&a->args[0], &n))) {
23707 + pr_err("bad integer in %s\n", opt_str);
23711 + err = opt_add(opt, a->args[1].from, opts->sb_flags,
23714 + opt->type = token;
23717 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
23718 + /*dummy bindex*/1);
23720 + opt->type = token;
23722 + case Opt_prepend:
23723 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
23726 + opt->type = token;
23729 + err = au_opts_parse_del(&opt->del, a->args);
23731 + opt->type = token;
23733 +#if 0 /* reserved for future use */
23735 + del->pathname = "(indexed)";
23736 + if (unlikely(match_int(&args[0], &n))) {
23737 + pr_err("bad integer in %s\n", opt_str);
23740 + err = au_opts_parse_idel(sb, n, &opt->del, a->args);
23742 + opt->type = token;
23746 + err = au_opts_parse_mod(&opt->mod, a->args);
23748 + opt->type = token;
23750 +#ifdef IMOD /* reserved for future use */
23752 + u.mod->path = "(indexed)";
23753 + if (unlikely(match_int(&a->args[0], &n))) {
23754 + pr_err("bad integer in %s\n", opt_str);
23757 + err = au_opts_parse_imod(sb, n, &opt->mod, a->args);
23759 + opt->type = token;
23763 + err = au_opts_parse_xino(sb, &opt->xino, a->args);
23765 + opt->type = token;
23768 + case Opt_trunc_xino_path:
23769 + err = au_opts_parse_xino_itrunc_path
23770 + (sb, &opt->xino_itrunc, a->args);
23772 + opt->type = token;
23775 + case Opt_itrunc_xino:
23776 + u.xino_itrunc = &opt->xino_itrunc;
23777 + if (unlikely(match_int(&a->args[0], &n))) {
23778 + pr_err("bad integer in %s\n", opt_str);
23781 + u.xino_itrunc->bindex = n;
23782 + aufs_read_lock(root, AuLock_FLUSH);
23783 + if (n < 0 || au_sbend(sb) < n) {
23784 + pr_err("out of bounds, %d\n", n);
23785 + aufs_read_unlock(root, !AuLock_IR);
23788 + aufs_read_unlock(root, !AuLock_IR);
23790 + opt->type = token;
23794 + if (unlikely(match_int(&a->args[0], &opt->dirwh)))
23797 + opt->type = token;
23800 + case Opt_rdcache:
23801 + if (unlikely(match_int(&a->args[0], &n))) {
23802 + pr_err("bad integer in %s\n", opt_str);
23805 + if (unlikely(n > AUFS_RDCACHE_MAX)) {
23806 + pr_err("rdcache must be smaller than %d\n",
23807 + AUFS_RDCACHE_MAX);
23810 + opt->rdcache = n;
23812 + opt->type = token;
23815 + if (unlikely(match_int(&a->args[0], &n)
23817 + || n > KMALLOC_MAX_SIZE)) {
23818 + pr_err("bad integer in %s\n", opt_str);
23821 + if (unlikely(n && n < NAME_MAX)) {
23822 + pr_err("rdblk must be larger than %d\n",
23828 + opt->type = token;
23831 + if (unlikely(match_int(&a->args[0], &n)
23833 + || n * sizeof(struct hlist_head)
23834 + > KMALLOC_MAX_SIZE)) {
23835 + pr_err("bad integer in %s\n", opt_str);
23840 + opt->type = token;
23843 + case Opt_trunc_xino:
23844 + case Opt_notrunc_xino:
23846 + case Opt_trunc_xib:
23847 + case Opt_notrunc_xib:
23850 + case Opt_dirperm1:
23851 + case Opt_nodirperm1:
23853 + case Opt_noplink:
23854 + case Opt_list_plink:
23857 + case Opt_diropq_a:
23858 + case Opt_diropq_w:
23859 + case Opt_warn_perm:
23860 + case Opt_nowarn_perm:
23861 + case Opt_verbose:
23862 + case Opt_noverbose:
23866 + case Opt_rdblk_def:
23867 + case Opt_rdhash_def:
23871 + opt->type = token;
23875 + opt->udba = udba_val(a->args[0].from);
23876 + if (opt->udba >= 0) {
23878 + opt->type = token;
23880 + pr_err("wrong value, %s\n", opt_str);
23883 + case Opt_wbr_create:
23884 + u.create = &opt->wbr_create;
23885 + u.create->wbr_create
23886 + = au_wbr_create_val(a->args[0].from, u.create);
23887 + if (u.create->wbr_create >= 0) {
23889 + opt->type = token;
23891 + pr_err("wrong value, %s\n", opt_str);
23893 + case Opt_wbr_copyup:
23894 + opt->wbr_copyup = au_wbr_copyup_val(a->args[0].from);
23895 + if (opt->wbr_copyup >= 0) {
23897 + opt->type = token;
23899 + pr_err("wrong value, %s\n", opt_str);
23902 + case Opt_fhsm_sec:
23903 + if (unlikely(match_int(&a->args[0], &n)
23905 + pr_err("bad integer in %s\n", opt_str);
23908 + if (sysaufs_brs) {
23909 + opt->fhsm_second = n;
23910 + opt->type = token;
23912 + pr_warn("ignored %s\n", opt_str);
23917 + pr_warn("ignored %s\n", opt_str);
23919 + case Opt_ignore_silent:
23924 + pr_err("unknown option %s\n", opt_str);
23928 + if (!err && !skipped) {
23929 + if (unlikely(++opt > opt_tail)) {
23932 + opt->type = Opt_tail;
23935 + opt->type = Opt_tail;
23941 + if (unlikely(err))
23942 + au_opts_free(opts);
23948 +static int au_opt_wbr_create(struct super_block *sb,
23949 + struct au_opt_wbr_create *create)
23952 + struct au_sbinfo *sbinfo;
23954 + SiMustWriteLock(sb);
23956 + err = 1; /* handled */
23957 + sbinfo = au_sbi(sb);
23958 + if (sbinfo->si_wbr_create_ops->fin) {
23959 + err = sbinfo->si_wbr_create_ops->fin(sb);
23964 + sbinfo->si_wbr_create = create->wbr_create;
23965 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
23966 + switch (create->wbr_create) {
23967 + case AuWbrCreate_MFSRRV:
23968 + case AuWbrCreate_MFSRR:
23969 + case AuWbrCreate_PMFSRR:
23970 + case AuWbrCreate_PMFSRRV:
23971 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
23973 + case AuWbrCreate_MFS:
23974 + case AuWbrCreate_MFSV:
23975 + case AuWbrCreate_PMFS:
23976 + case AuWbrCreate_PMFSV:
23977 + sbinfo->si_wbr_mfs.mfs_expire
23978 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
23982 + if (sbinfo->si_wbr_create_ops->init)
23983 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
23990 + * plus: processed without an error
23991 + * zero: unprocessed
23993 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
23994 + struct au_opts *opts)
23997 + struct au_sbinfo *sbinfo;
23999 + SiMustWriteLock(sb);
24001 + err = 1; /* handled */
24002 + sbinfo = au_sbi(sb);
24003 + switch (opt->type) {
24005 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
24006 + sbinfo->si_mntflags |= opt->udba;
24007 + opts->given_udba |= opt->udba;
24011 + au_opt_set(sbinfo->si_mntflags, PLINK);
24013 + case Opt_noplink:
24014 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
24015 + au_plink_put(sb, /*verbose*/1);
24016 + au_opt_clr(sbinfo->si_mntflags, PLINK);
24018 + case Opt_list_plink:
24019 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
24020 + au_plink_list(sb);
24024 + au_opt_set(sbinfo->si_mntflags, DIO);
24025 + au_fset_opts(opts->flags, REFRESH_DYAOP);
24028 + au_opt_clr(sbinfo->si_mntflags, DIO);
24029 + au_fset_opts(opts->flags, REFRESH_DYAOP);
24032 + case Opt_fhsm_sec:
24033 + au_fhsm_set(sbinfo, opt->fhsm_second);
24036 + case Opt_diropq_a:
24037 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
24039 + case Opt_diropq_w:
24040 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
24043 + case Opt_warn_perm:
24044 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
24046 + case Opt_nowarn_perm:
24047 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
24050 + case Opt_verbose:
24051 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
24053 + case Opt_noverbose:
24054 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
24058 + au_opt_set(sbinfo->si_mntflags, SUM);
24061 + au_opt_clr(sbinfo->si_mntflags, SUM);
24062 + au_opt_set(sbinfo->si_mntflags, SUM_W);
24064 + au_opt_clr(sbinfo->si_mntflags, SUM);
24065 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
24068 + case Opt_wbr_create:
24069 + err = au_opt_wbr_create(sb, &opt->wbr_create);
24071 + case Opt_wbr_copyup:
24072 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
24073 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
24077 + sbinfo->si_dirwh = opt->dirwh;
24080 + case Opt_rdcache:
24081 + sbinfo->si_rdcache
24082 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
24085 + sbinfo->si_rdblk = opt->rdblk;
24087 + case Opt_rdblk_def:
24088 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
24091 + sbinfo->si_rdhash = opt->rdhash;
24093 + case Opt_rdhash_def:
24094 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
24098 + au_opt_set(sbinfo->si_mntflags, SHWH);
24101 + au_opt_clr(sbinfo->si_mntflags, SHWH);
24104 + case Opt_dirperm1:
24105 + au_opt_set(sbinfo->si_mntflags, DIRPERM1);
24107 + case Opt_nodirperm1:
24108 + au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
24111 + case Opt_trunc_xino:
24112 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
24114 + case Opt_notrunc_xino:
24115 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
24118 + case Opt_trunc_xino_path:
24119 + case Opt_itrunc_xino:
24120 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex);
24125 + case Opt_trunc_xib:
24126 + au_fset_opts(opts->flags, TRUNC_XIB);
24128 + case Opt_notrunc_xib:
24129 + au_fclr_opts(opts->flags, TRUNC_XIB);
24133 + sb->s_flags |= MS_POSIXACL;
24136 + sb->s_flags &= ~MS_POSIXACL;
24148 + * returns tri-state.
24149 + * plus: processed without an error
24150 + * zero: unprocessed
24153 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
24154 + struct au_opts *opts)
24156 + int err, do_refresh;
24159 + switch (opt->type) {
24161 + opt->add.bindex = au_sbend(sb) + 1;
24162 + if (opt->add.bindex < 0)
24163 + opt->add.bindex = 0;
24165 + case Opt_prepend:
24166 + opt->add.bindex = 0;
24167 + add: /* indented label */
24169 + err = au_br_add(sb, &opt->add,
24170 + au_ftest_opts(opts->flags, REMOUNT));
24173 + au_fset_opts(opts->flags, REFRESH);
24179 + err = au_br_del(sb, &opt->del,
24180 + au_ftest_opts(opts->flags, REMOUNT));
24183 + au_fset_opts(opts->flags, TRUNC_XIB);
24184 + au_fset_opts(opts->flags, REFRESH);
24190 + err = au_br_mod(sb, &opt->mod,
24191 + au_ftest_opts(opts->flags, REMOUNT),
24196 + au_fset_opts(opts->flags, REFRESH);
24204 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
24205 + struct au_opt_xino **opt_xino,
24206 + struct au_opts *opts)
24209 + aufs_bindex_t bend, bindex;
24210 + struct dentry *root, *parent, *h_root;
24213 + switch (opt->type) {
24215 + err = au_xino_set(sb, &opt->xino,
24216 + !!au_ftest_opts(opts->flags, REMOUNT));
24217 + if (unlikely(err))
24220 + *opt_xino = &opt->xino;
24221 + au_xino_brid_set(sb, -1);
24223 + /* safe d_parent access */
24224 + parent = opt->xino.file->f_path.dentry->d_parent;
24225 + root = sb->s_root;
24226 + bend = au_sbend(sb);
24227 + for (bindex = 0; bindex <= bend; bindex++) {
24228 + h_root = au_h_dptr(root, bindex);
24229 + if (h_root == parent) {
24230 + au_xino_brid_set(sb, au_sbr_id(sb, bindex));
24238 + au_xino_brid_set(sb, -1);
24239 + *opt_xino = (void *)-1;
24246 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
24247 + unsigned int pending)
24250 + aufs_bindex_t bindex, bend;
24251 + unsigned char do_plink, skip, do_free, can_no_dreval;
24252 + struct au_branch *br;
24253 + struct au_wbr *wbr;
24254 + struct dentry *root, *dentry;
24255 + struct inode *dir, *h_dir;
24256 + struct au_sbinfo *sbinfo;
24257 + struct au_hinode *hdir;
24259 + SiMustAnyLock(sb);
24261 + sbinfo = au_sbi(sb);
24262 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
24264 + if (!(sb_flags & MS_RDONLY)) {
24265 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
24266 + pr_warn("first branch should be rw\n");
24267 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
24268 + pr_warn("shwh should be used with ro\n");
24271 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
24272 + && !au_opt_test(sbinfo->si_mntflags, XINO))
24273 + pr_warn("udba=*notify requires xino\n");
24275 + if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
24276 + pr_warn("dirperm1 breaks the protection"
24277 + " by the permission bits on the lower branch\n");
24281 + root = sb->s_root;
24282 + dir = d_inode(root);
24283 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
24284 + can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
24286 + bend = au_sbend(sb);
24287 + for (bindex = 0; !err && bindex <= bend; bindex++) {
24289 + h_dir = au_h_iptr(dir, bindex);
24290 + br = au_sbr(sb, bindex);
24292 + if ((br->br_perm & AuBrAttr_ICEX)
24293 + && !h_dir->i_op->listxattr)
24294 + br->br_perm &= ~AuBrAttr_ICEX;
24296 + if ((br->br_perm & AuBrAttr_ICEX_SEC)
24297 + && (au_br_sb(br)->s_flags & MS_NOSEC))
24298 + br->br_perm &= ~AuBrAttr_ICEX_SEC;
24302 + wbr = br->br_wbr;
24304 + wbr_wh_read_lock(wbr);
24306 + if (!au_br_writable(br->br_perm)) {
24309 + || (!wbr->wbr_whbase
24310 + && !wbr->wbr_plink
24311 + && !wbr->wbr_orph));
24312 + } else if (!au_br_wh_linkable(br->br_perm)) {
24313 + /* skip = (!br->br_whbase && !br->br_orph); */
24314 + skip = (!wbr || !wbr->wbr_whbase);
24315 + if (skip && wbr) {
24317 + skip = !!wbr->wbr_plink;
24319 + skip = !wbr->wbr_plink;
24322 + /* skip = (br->br_whbase && br->br_ohph); */
24323 + skip = (wbr && wbr->wbr_whbase);
24326 + skip = !!wbr->wbr_plink;
24328 + skip = !wbr->wbr_plink;
24332 + wbr_wh_read_unlock(wbr);
24334 + if (can_no_dreval) {
24335 + dentry = br->br_path.dentry;
24336 + spin_lock(&dentry->d_lock);
24337 + if (dentry->d_flags &
24338 + (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
24339 + can_no_dreval = 0;
24340 + spin_unlock(&dentry->d_lock);
24343 + if (au_br_fhsm(br->br_perm)) {
24345 + AuDebugOn(!br->br_fhsm);
24351 + hdir = au_hi(dir, bindex);
24352 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
24354 + wbr_wh_write_lock(wbr);
24355 + err = au_wh_init(br, sb);
24357 + wbr_wh_write_unlock(wbr);
24358 + au_hn_imtx_unlock(hdir);
24360 + if (!err && do_free) {
24362 + br->br_wbr = NULL;
24366 + if (can_no_dreval)
24367 + au_fset_si(sbinfo, NO_DREVAL);
24369 + au_fclr_si(sbinfo, NO_DREVAL);
24372 + au_fset_si(sbinfo, FHSM);
24373 + for (bindex = bend; bindex >= 0; bindex--) {
24374 + br = au_sbr(sb, bindex);
24375 + if (au_br_fhsm(br->br_perm)) {
24376 + au_fhsm_set_bottom(sb, bindex);
24381 + au_fclr_si(sbinfo, FHSM);
24382 + au_fhsm_set_bottom(sb, -1);
24388 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
24391 + unsigned int tmp;
24392 + aufs_bindex_t bindex, bend;
24393 + struct au_opt *opt;
24394 + struct au_opt_xino *opt_xino, xino;
24395 + struct au_sbinfo *sbinfo;
24396 + struct au_branch *br;
24397 + struct inode *dir;
24399 + SiMustWriteLock(sb);
24404 + while (err >= 0 && opt->type != Opt_tail)
24405 + err = au_opt_simple(sb, opt++, opts);
24408 + else if (unlikely(err < 0))
24411 + /* disable xino and udba temporary */
24412 + sbinfo = au_sbi(sb);
24413 + tmp = sbinfo->si_mntflags;
24414 + au_opt_clr(sbinfo->si_mntflags, XINO);
24415 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
24418 + while (err >= 0 && opt->type != Opt_tail)
24419 + err = au_opt_br(sb, opt++, opts);
24422 + else if (unlikely(err < 0))
24425 + bend = au_sbend(sb);
24426 + if (unlikely(bend < 0)) {
24428 + pr_err("no branches\n");
24432 + if (au_opt_test(tmp, XINO))
24433 + au_opt_set(sbinfo->si_mntflags, XINO);
24435 + while (!err && opt->type != Opt_tail)
24436 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
24437 + if (unlikely(err))
24440 + err = au_opts_verify(sb, sb->s_flags, tmp);
24441 + if (unlikely(err))
24444 + /* restore xino */
24445 + if (au_opt_test(tmp, XINO) && !opt_xino) {
24446 + xino.file = au_xino_def(sb);
24447 + err = PTR_ERR(xino.file);
24448 + if (IS_ERR(xino.file))
24451 + err = au_xino_set(sb, &xino, /*remount*/0);
24453 + if (unlikely(err))
24457 + /* restore udba */
24458 + tmp &= AuOptMask_UDBA;
24459 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
24460 + sbinfo->si_mntflags |= tmp;
24461 + bend = au_sbend(sb);
24462 + for (bindex = 0; bindex <= bend; bindex++) {
24463 + br = au_sbr(sb, bindex);
24464 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
24465 + if (unlikely(err))
24466 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
24468 + /* go on even if err */
24470 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
24471 + dir = d_inode(sb->s_root);
24472 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
24479 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
24482 + unsigned char no_dreval;
24483 + struct inode *dir;
24484 + struct au_opt_xino *opt_xino;
24485 + struct au_opt *opt;
24486 + struct au_sbinfo *sbinfo;
24488 + SiMustWriteLock(sb);
24491 + dir = d_inode(sb->s_root);
24492 + sbinfo = au_sbi(sb);
24495 + while (err >= 0 && opt->type != Opt_tail) {
24496 + err = au_opt_simple(sb, opt, opts);
24498 + err = au_opt_br(sb, opt, opts);
24500 + err = au_opt_xino(sb, opt, &opt_xino, opts);
24506 + /* go on even err */
24508 + no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
24509 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
24510 + if (unlikely(rerr && !err))
24513 + if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
24514 + au_fset_opts(opts->flags, REFRESH_IDOP);
24516 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
24517 + rerr = au_xib_trunc(sb);
24518 + if (unlikely(rerr && !err))
24522 + /* will be handled by the caller */
24523 + if (!au_ftest_opts(opts->flags, REFRESH)
24524 + && (opts->given_udba
24525 + || au_opt_test(sbinfo->si_mntflags, XINO)
24526 + || au_ftest_opts(opts->flags, REFRESH_IDOP)
24528 + au_fset_opts(opts->flags, REFRESH);
24530 + AuDbg("status 0x%x\n", opts->flags);
24534 +/* ---------------------------------------------------------------------- */
24536 +unsigned int au_opt_udba(struct super_block *sb)
24538 + return au_mntflags(sb) & AuOptMask_UDBA;
24540 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
24541 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
24542 +++ linux/fs/aufs/opts.h 2016-01-13 20:11:11.673093671 +0100
24545 + * Copyright (C) 2005-2015 Junjiro R. Okajima
24547 + * This program, aufs is free software; you can redistribute it and/or modify
24548 + * it under the terms of the GNU General Public License as published by
24549 + * the Free Software Foundation; either version 2 of the License, or
24550 + * (at your option) any later version.
24552 + * This program is distributed in the hope that it will be useful,
24553 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24554 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24555 + * GNU General Public License for more details.
24557 + * You should have received a copy of the GNU General Public License
24558 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24562 + * mount options/flags
24565 +#ifndef __AUFS_OPTS_H__
24566 +#define __AUFS_OPTS_H__
24570 +#include <linux/path.h>
24573 +struct super_block;
24575 +/* ---------------------------------------------------------------------- */
24578 +#define AuOpt_XINO 1 /* external inode number bitmap
24579 + and translation table */
24580 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
24581 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
24582 +#define AuOpt_UDBA_REVAL (1 << 3)
24583 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
24584 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
24585 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
24586 +#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
24588 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
24589 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
24590 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
24591 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
24592 +#define AuOpt_VERBOSE (1 << 13) /* busy inode when del-branch */
24593 +#define AuOpt_DIO (1 << 14) /* direct io */
24595 +#ifndef CONFIG_AUFS_HNOTIFY
24596 +#undef AuOpt_UDBA_HNOTIFY
24597 +#define AuOpt_UDBA_HNOTIFY 0
24599 +#ifndef CONFIG_AUFS_SHWH
24601 +#define AuOpt_SHWH 0
24604 +#define AuOpt_Def (AuOpt_XINO \
24605 + | AuOpt_UDBA_REVAL \
24607 + /* | AuOpt_DIRPERM1 */ \
24608 + | AuOpt_WARN_PERM)
24609 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
24610 + | AuOpt_UDBA_REVAL \
24611 + | AuOpt_UDBA_HNOTIFY)
24613 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
24614 +#define au_opt_set(flags, name) do { \
24615 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
24616 + ((flags) |= AuOpt_##name); \
24618 +#define au_opt_set_udba(flags, name) do { \
24619 + (flags) &= ~AuOptMask_UDBA; \
24620 + ((flags) |= AuOpt_##name); \
24622 +#define au_opt_clr(flags, name) do { \
24623 + ((flags) &= ~AuOpt_##name); \
24626 +static inline unsigned int au_opts_plink(unsigned int mntflags)
24628 +#ifdef CONFIG_PROC_FS
24631 + return mntflags & ~AuOpt_PLINK;
24635 +/* ---------------------------------------------------------------------- */
24637 +/* policies to select one among multiple writable branches */
24639 + AuWbrCreate_TDP, /* top down parent */
24640 + AuWbrCreate_RR, /* round robin */
24641 + AuWbrCreate_MFS, /* most free space */
24642 + AuWbrCreate_MFSV, /* mfs with seconds */
24643 + AuWbrCreate_MFSRR, /* mfs then rr */
24644 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
24645 + AuWbrCreate_PMFS, /* parent and mfs */
24646 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
24647 + AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
24648 + AuWbrCreate_PMFSRRV, /* plus seconds */
24650 + AuWbrCreate_Def = AuWbrCreate_TDP
24654 + AuWbrCopyup_TDP, /* top down parent */
24655 + AuWbrCopyup_BUP, /* bottom up parent */
24656 + AuWbrCopyup_BU, /* bottom up */
24658 + AuWbrCopyup_Def = AuWbrCopyup_TDP
24661 +/* ---------------------------------------------------------------------- */
24663 +struct au_opt_add {
24664 + aufs_bindex_t bindex;
24667 + struct path path;
24670 +struct au_opt_del {
24672 + struct path h_path;
24675 +struct au_opt_mod {
24678 + struct dentry *h_root;
24681 +struct au_opt_xino {
24683 + struct file *file;
24686 +struct au_opt_xino_itrunc {
24687 + aufs_bindex_t bindex;
24690 +struct au_opt_wbr_create {
24693 + unsigned long long mfsrr_watermark;
24699 + struct au_opt_xino xino;
24700 + struct au_opt_xino_itrunc xino_itrunc;
24701 + struct au_opt_add add;
24702 + struct au_opt_del del;
24703 + struct au_opt_mod mod;
24706 + unsigned int rdblk;
24707 + unsigned int rdhash;
24709 + struct au_opt_wbr_create wbr_create;
24711 + unsigned int fhsm_second;
24716 +#define AuOpts_REMOUNT 1
24717 +#define AuOpts_REFRESH (1 << 1)
24718 +#define AuOpts_TRUNC_XIB (1 << 2)
24719 +#define AuOpts_REFRESH_DYAOP (1 << 3)
24720 +#define AuOpts_REFRESH_IDOP (1 << 4)
24721 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
24722 +#define au_fset_opts(flags, name) \
24723 + do { (flags) |= AuOpts_##name; } while (0)
24724 +#define au_fclr_opts(flags, name) \
24725 + do { (flags) &= ~AuOpts_##name; } while (0)
24728 + struct au_opt *opt;
24731 + unsigned int given_udba;
24732 + unsigned int flags;
24733 + unsigned long sb_flags;
24736 +/* ---------------------------------------------------------------------- */
24739 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
24740 +const char *au_optstr_udba(int udba);
24741 +const char *au_optstr_wbr_copyup(int wbr_copyup);
24742 +const char *au_optstr_wbr_create(int wbr_create);
24744 +void au_opts_free(struct au_opts *opts);
24745 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts);
24746 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
24747 + unsigned int pending);
24748 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
24749 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
24751 +unsigned int au_opt_udba(struct super_block *sb);
24753 +#endif /* __KERNEL__ */
24754 +#endif /* __AUFS_OPTS_H__ */
24755 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
24756 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
24757 +++ linux/fs/aufs/plink.c 2016-01-13 20:11:11.673093671 +0100
24760 + * Copyright (C) 2005-2015 Junjiro R. Okajima
24762 + * This program, aufs is free software; you can redistribute it and/or modify
24763 + * it under the terms of the GNU General Public License as published by
24764 + * the Free Software Foundation; either version 2 of the License, or
24765 + * (at your option) any later version.
24767 + * This program is distributed in the hope that it will be useful,
24768 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24769 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24770 + * GNU General Public License for more details.
24772 + * You should have received a copy of the GNU General Public License
24773 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24783 + * the pseudo-link maintenance mode.
24784 + * during a user process maintains the pseudo-links,
24785 + * prohibit adding a new plink and branch manipulation.
24789 + * For entry functions which will handle plink, and i_mutex is already held
24791 + * They cannot wait and should return an error at once.
24792 + * Callers has to check the error.
24794 + * For entry functions which will handle plink, but i_mutex is not held
24796 + * They can wait the plink maintenance mode to finish.
24798 + * They behave like F_SETLK and F_SETLKW.
24799 + * If the caller never handle plink, then both flags are unnecessary.
24802 +int au_plink_maint(struct super_block *sb, int flags)
24806 + struct au_sbinfo *sbi;
24808 + SiMustAnyLock(sb);
24811 + if (!au_opt_test(au_mntflags(sb), PLINK))
24814 + sbi = au_sbi(sb);
24815 + pid = sbi->si_plink_maint_pid;
24816 + if (!pid || pid == current->pid)
24819 + /* todo: it highly depends upon /sbin/mount.aufs */
24821 + ppid = task_pid_vnr(rcu_dereference(current->real_parent));
24822 + rcu_read_unlock();
24826 + if (au_ftest_lock(flags, NOPLMW)) {
24827 + /* if there is no i_mutex lock in VFS, we don't need to wait */
24828 + /* AuDebugOn(!lockdep_depth(current)); */
24829 + while (sbi->si_plink_maint_pid) {
24830 + si_read_unlock(sb);
24831 + /* gave up wake_up_bit() */
24832 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
24834 + if (au_ftest_lock(flags, FLUSH))
24835 + au_nwt_flush(&sbi->si_nowait);
24836 + si_noflush_read_lock(sb);
24838 + } else if (au_ftest_lock(flags, NOPLM)) {
24839 + AuDbg("ppid %d, pid %d\n", ppid, pid);
24847 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
24849 + spin_lock(&sbinfo->si_plink_maint_lock);
24850 + sbinfo->si_plink_maint_pid = 0;
24851 + spin_unlock(&sbinfo->si_plink_maint_lock);
24852 + wake_up_all(&sbinfo->si_plink_wq);
24855 +int au_plink_maint_enter(struct super_block *sb)
24858 + struct au_sbinfo *sbinfo;
24861 + sbinfo = au_sbi(sb);
24862 + /* make sure i am the only one in this fs */
24863 + si_write_lock(sb, AuLock_FLUSH);
24864 + if (au_opt_test(au_mntflags(sb), PLINK)) {
24865 + spin_lock(&sbinfo->si_plink_maint_lock);
24866 + if (!sbinfo->si_plink_maint_pid)
24867 + sbinfo->si_plink_maint_pid = current->pid;
24870 + spin_unlock(&sbinfo->si_plink_maint_lock);
24872 + si_write_unlock(sb);
24877 +/* ---------------------------------------------------------------------- */
24879 +#ifdef CONFIG_AUFS_DEBUG
24880 +void au_plink_list(struct super_block *sb)
24883 + struct au_sbinfo *sbinfo;
24884 + struct hlist_head *plink_hlist;
24885 + struct pseudo_link *plink;
24887 + SiMustAnyLock(sb);
24889 + sbinfo = au_sbi(sb);
24890 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
24891 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
24893 + for (i = 0; i < AuPlink_NHASH; i++) {
24894 + plink_hlist = &sbinfo->si_plink[i].head;
24896 + hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
24897 + AuDbg("%lu\n", plink->inode->i_ino);
24898 + rcu_read_unlock();
24903 +/* is the inode pseudo-linked? */
24904 +int au_plink_test(struct inode *inode)
24907 + struct au_sbinfo *sbinfo;
24908 + struct hlist_head *plink_hlist;
24909 + struct pseudo_link *plink;
24911 + sbinfo = au_sbi(inode->i_sb);
24912 + AuRwMustAnyLock(&sbinfo->si_rwsem);
24913 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
24914 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
24917 + i = au_plink_hash(inode->i_ino);
24918 + plink_hlist = &sbinfo->si_plink[i].head;
24920 + hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
24921 + if (plink->inode == inode) {
24925 + rcu_read_unlock();
24929 +/* ---------------------------------------------------------------------- */
24932 + * generate a name for plink.
24933 + * the file will be stored under AUFS_WH_PLINKDIR.
24935 +/* 20 is max digits length of ulong 64 */
24936 +#define PLINK_NAME_LEN ((20 + 1) * 2)
24938 +static int plink_name(char *name, int len, struct inode *inode,
24939 + aufs_bindex_t bindex)
24942 + struct inode *h_inode;
24944 + h_inode = au_h_iptr(inode, bindex);
24945 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
24949 +struct au_do_plink_lkup_args {
24950 + struct dentry **errp;
24951 + struct qstr *tgtname;
24952 + struct dentry *h_parent;
24953 + struct au_branch *br;
24956 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
24957 + struct dentry *h_parent,
24958 + struct au_branch *br)
24960 + struct dentry *h_dentry;
24961 + struct mutex *h_mtx;
24963 + h_mtx = &d_inode(h_parent)->i_mutex;
24964 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
24965 + h_dentry = vfsub_lkup_one(tgtname, h_parent);
24966 + mutex_unlock(h_mtx);
24970 +static void au_call_do_plink_lkup(void *args)
24972 + struct au_do_plink_lkup_args *a = args;
24973 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br);
24976 +/* lookup the plink-ed @inode under the branch at @bindex */
24977 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
24979 + struct dentry *h_dentry, *h_parent;
24980 + struct au_branch *br;
24982 + char a[PLINK_NAME_LEN];
24983 + struct qstr tgtname = QSTR_INIT(a, 0);
24985 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
24987 + br = au_sbr(inode->i_sb, bindex);
24988 + h_parent = br->br_wbr->wbr_plink;
24989 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
24991 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
24992 + struct au_do_plink_lkup_args args = {
24993 + .errp = &h_dentry,
24994 + .tgtname = &tgtname,
24995 + .h_parent = h_parent,
24999 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
25000 + if (unlikely(wkq_err))
25001 + h_dentry = ERR_PTR(wkq_err);
25003 + h_dentry = au_do_plink_lkup(&tgtname, h_parent, br);
25008 +/* create a pseudo-link */
25009 +static int do_whplink(struct qstr *tgt, struct dentry *h_parent,
25010 + struct dentry *h_dentry, struct au_branch *br)
25013 + struct path h_path = {
25014 + .mnt = au_br_mnt(br)
25016 + struct inode *h_dir, *delegated;
25018 + h_dir = d_inode(h_parent);
25019 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_CHILD2);
25021 + h_path.dentry = vfsub_lkup_one(tgt, h_parent);
25022 + err = PTR_ERR(h_path.dentry);
25023 + if (IS_ERR(h_path.dentry))
25027 + /* wh.plink dir is not monitored */
25028 + /* todo: is it really safe? */
25029 + if (d_is_positive(h_path.dentry)
25030 + && d_inode(h_path.dentry) != d_inode(h_dentry)) {
25031 + delegated = NULL;
25032 + err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
25033 + if (unlikely(err == -EWOULDBLOCK)) {
25034 + pr_warn("cannot retry for NFSv4 delegation"
25035 + " for an internal unlink\n");
25038 + dput(h_path.dentry);
25039 + h_path.dentry = NULL;
25043 + if (!err && d_is_negative(h_path.dentry)) {
25044 + delegated = NULL;
25045 + err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
25046 + if (unlikely(err == -EWOULDBLOCK)) {
25047 + pr_warn("cannot retry for NFSv4 delegation"
25048 + " for an internal link\n");
25052 + dput(h_path.dentry);
25055 + mutex_unlock(&h_dir->i_mutex);
25059 +struct do_whplink_args {
25061 + struct qstr *tgt;
25062 + struct dentry *h_parent;
25063 + struct dentry *h_dentry;
25064 + struct au_branch *br;
25067 +static void call_do_whplink(void *args)
25069 + struct do_whplink_args *a = args;
25070 + *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br);
25073 +static int whplink(struct dentry *h_dentry, struct inode *inode,
25074 + aufs_bindex_t bindex, struct au_branch *br)
25076 + int err, wkq_err;
25077 + struct au_wbr *wbr;
25078 + struct dentry *h_parent;
25079 + char a[PLINK_NAME_LEN];
25080 + struct qstr tgtname = QSTR_INIT(a, 0);
25082 + wbr = au_sbr(inode->i_sb, bindex)->br_wbr;
25083 + h_parent = wbr->wbr_plink;
25084 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
25086 + /* always superio. */
25087 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
25088 + struct do_whplink_args args = {
25091 + .h_parent = h_parent,
25092 + .h_dentry = h_dentry,
25095 + wkq_err = au_wkq_wait(call_do_whplink, &args);
25096 + if (unlikely(wkq_err))
25099 + err = do_whplink(&tgtname, h_parent, h_dentry, br);
25104 +/* free a single plink */
25105 +static void do_put_plink(struct pseudo_link *plink, int do_del)
25108 + hlist_del(&plink->hlist);
25109 + iput(plink->inode);
25113 +static void do_put_plink_rcu(struct rcu_head *rcu)
25115 + struct pseudo_link *plink;
25117 + plink = container_of(rcu, struct pseudo_link, rcu);
25118 + iput(plink->inode);
25123 + * create a new pseudo-link for @h_dentry on @bindex.
25124 + * the linked inode is held in aufs @inode.
25126 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
25127 + struct dentry *h_dentry)
25129 + struct super_block *sb;
25130 + struct au_sbinfo *sbinfo;
25131 + struct hlist_head *plink_hlist;
25132 + struct pseudo_link *plink, *tmp;
25133 + struct au_sphlhead *sphl;
25134 + int found, err, cnt, i;
25136 + sb = inode->i_sb;
25137 + sbinfo = au_sbi(sb);
25138 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
25139 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
25141 + found = au_plink_test(inode);
25145 + i = au_plink_hash(inode->i_ino);
25146 + sphl = sbinfo->si_plink + i;
25147 + plink_hlist = &sphl->head;
25148 + tmp = kmalloc(sizeof(*plink), GFP_NOFS);
25150 + tmp->inode = au_igrab(inode);
25156 + spin_lock(&sphl->spin);
25157 + hlist_for_each_entry(plink, plink_hlist, hlist) {
25158 + if (plink->inode == inode) {
25164 + hlist_add_head_rcu(&tmp->hlist, plink_hlist);
25165 + spin_unlock(&sphl->spin);
25167 + cnt = au_sphl_count(sphl);
25168 +#define msg "unexpectedly unblanced or too many pseudo-links"
25169 + if (cnt > AUFS_PLINK_WARN)
25170 + AuWarn1(msg ", %d\n", cnt);
25172 + err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex));
25174 + do_put_plink(tmp, 0);
25179 + if (unlikely(err)) {
25180 + pr_warn("err %d, damaged pseudo link.\n", err);
25182 + au_sphl_del_rcu(&tmp->hlist, sphl);
25183 + call_rcu(&tmp->rcu, do_put_plink_rcu);
25188 +/* free all plinks */
25189 +void au_plink_put(struct super_block *sb, int verbose)
25192 + struct au_sbinfo *sbinfo;
25193 + struct hlist_head *plink_hlist;
25194 + struct hlist_node *tmp;
25195 + struct pseudo_link *plink;
25197 + SiMustWriteLock(sb);
25199 + sbinfo = au_sbi(sb);
25200 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
25201 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
25203 + /* no spin_lock since sbinfo is write-locked */
25205 + for (i = 0; i < AuPlink_NHASH; i++) {
25206 + plink_hlist = &sbinfo->si_plink[i].head;
25207 + if (!warned && verbose && !hlist_empty(plink_hlist)) {
25208 + pr_warn("pseudo-link is not flushed");
25211 + hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist)
25212 + do_put_plink(plink, 0);
25213 + INIT_HLIST_HEAD(plink_hlist);
25217 +void au_plink_clean(struct super_block *sb, int verbose)
25219 + struct dentry *root;
25221 + root = sb->s_root;
25222 + aufs_write_lock(root);
25223 + if (au_opt_test(au_mntflags(sb), PLINK))
25224 + au_plink_put(sb, verbose);
25225 + aufs_write_unlock(root);
25228 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
25231 + aufs_bindex_t bstart, bend, bindex;
25234 + bstart = au_ibstart(inode);
25235 + bend = au_ibend(inode);
25236 + if (bstart >= 0) {
25237 + for (bindex = bstart; bindex <= bend; bindex++) {
25238 + if (!au_h_iptr(inode, bindex)
25239 + || au_ii_br_id(inode, bindex) != br_id)
25241 + au_set_h_iptr(inode, bindex, NULL, 0);
25246 + for (bindex = bstart; bindex <= bend; bindex++)
25247 + if (au_h_iptr(inode, bindex)) {
25257 +/* free the plinks on a branch specified by @br_id */
25258 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
25260 + struct au_sbinfo *sbinfo;
25261 + struct hlist_head *plink_hlist;
25262 + struct hlist_node *tmp;
25263 + struct pseudo_link *plink;
25264 + struct inode *inode;
25267 + SiMustWriteLock(sb);
25269 + sbinfo = au_sbi(sb);
25270 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
25271 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
25273 + /* no spin_lock since sbinfo is write-locked */
25274 + for (i = 0; i < AuPlink_NHASH; i++) {
25275 + plink_hlist = &sbinfo->si_plink[i].head;
25276 + hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist) {
25277 + inode = au_igrab(plink->inode);
25278 + ii_write_lock_child(inode);
25279 + do_put = au_plink_do_half_refresh(inode, br_id);
25281 + do_put_plink(plink, 1);
25282 + ii_write_unlock(inode);
25287 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
25288 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
25289 +++ linux/fs/aufs/poll.c 2016-01-13 20:11:11.673093671 +0100
25292 + * Copyright (C) 2005-2015 Junjiro R. Okajima
25294 + * This program, aufs is free software; you can redistribute it and/or modify
25295 + * it under the terms of the GNU General Public License as published by
25296 + * the Free Software Foundation; either version 2 of the License, or
25297 + * (at your option) any later version.
25299 + * This program is distributed in the hope that it will be useful,
25300 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25301 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25302 + * GNU General Public License for more details.
25304 + * You should have received a copy of the GNU General Public License
25305 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25310 + * There is only one filesystem which implements ->poll operation, currently.
25315 +unsigned int aufs_poll(struct file *file, poll_table *wait)
25317 + unsigned int mask;
25319 + struct file *h_file;
25320 + struct super_block *sb;
25322 + /* We should pretend an error happened. */
25323 + mask = POLLERR /* | POLLIN | POLLOUT */;
25324 + sb = file->f_path.dentry->d_sb;
25325 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
25327 + h_file = au_read_pre(file, /*keep_fi*/0);
25328 + err = PTR_ERR(h_file);
25329 + if (IS_ERR(h_file))
25332 + /* it is not an error if h_file has no operation */
25333 + mask = DEFAULT_POLLMASK;
25334 + if (h_file->f_op->poll)
25335 + mask = h_file->f_op->poll(h_file, wait);
25336 + fput(h_file); /* instead of au_read_post() */
25339 + si_read_unlock(sb);
25340 + AuTraceErr((int)mask);
25343 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
25344 --- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
25345 +++ linux/fs/aufs/posix_acl.c 2016-01-13 20:11:11.673093671 +0100
25348 + * Copyright (C) 2014-2015 Junjiro R. Okajima
25350 + * This program, aufs is free software; you can redistribute it and/or modify
25351 + * it under the terms of the GNU General Public License as published by
25352 + * the Free Software Foundation; either version 2 of the License, or
25353 + * (at your option) any later version.
25355 + * This program is distributed in the hope that it will be useful,
25356 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25357 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25358 + * GNU General Public License for more details.
25360 + * You should have received a copy of the GNU General Public License
25361 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25365 + * posix acl operations
25368 +#include <linux/fs.h>
25369 +#include <linux/posix_acl.h>
25372 +struct posix_acl *aufs_get_acl(struct inode *inode, int type)
25374 + struct posix_acl *acl;
25376 + aufs_bindex_t bindex;
25377 + struct inode *h_inode;
25378 + struct super_block *sb;
25381 + sb = inode->i_sb;
25382 + si_read_lock(sb, AuLock_FLUSH);
25383 + ii_read_lock_child(inode);
25384 + if (!(sb->s_flags & MS_POSIXACL))
25387 + bindex = au_ibstart(inode);
25388 + h_inode = au_h_iptr(inode, bindex);
25389 + if (unlikely(!h_inode
25390 + || ((h_inode->i_mode & S_IFMT)
25391 + != (inode->i_mode & S_IFMT)))) {
25392 + err = au_busy_or_stale();
25393 + acl = ERR_PTR(err);
25397 + /* always topmost only */
25398 + acl = get_acl(h_inode, type);
25401 + ii_read_unlock(inode);
25402 + si_read_unlock(sb);
25404 + AuTraceErrPtr(acl);
25408 +int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
25412 + struct dentry *dentry;
25413 + struct au_srxattr arg = {
25414 + .type = AU_ACL_SET,
25421 + mutex_lock(&inode->i_mutex);
25422 + if (inode->i_ino == AUFS_ROOT_INO)
25423 + dentry = dget(inode->i_sb->s_root);
25425 + dentry = d_find_alias(inode);
25427 + dentry = d_find_any_alias(inode);
25429 + pr_warn("cannot handle this inode, "
25430 + "please report to aufs-users ML\n");
25436 + ssz = au_srxattr(dentry, &arg);
25443 + mutex_unlock(&inode->i_mutex);
25446 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
25447 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
25448 +++ linux/fs/aufs/procfs.c 2016-01-13 20:11:11.673093671 +0100
25451 + * Copyright (C) 2010-2015 Junjiro R. Okajima
25453 + * This program, aufs is free software; you can redistribute it and/or modify
25454 + * it under the terms of the GNU General Public License as published by
25455 + * the Free Software Foundation; either version 2 of the License, or
25456 + * (at your option) any later version.
25458 + * This program is distributed in the hope that it will be useful,
25459 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25460 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25461 + * GNU General Public License for more details.
25463 + * You should have received a copy of the GNU General Public License
25464 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25468 + * procfs interfaces
25471 +#include <linux/proc_fs.h>
25474 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
25476 + struct au_sbinfo *sbinfo;
25478 + sbinfo = file->private_data;
25480 + au_plink_maint_leave(sbinfo);
25481 + kobject_put(&sbinfo->si_kobj);
25487 +static void au_procfs_plm_write_clean(struct file *file)
25489 + struct au_sbinfo *sbinfo;
25491 + sbinfo = file->private_data;
25493 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
25496 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
25499 + struct super_block *sb;
25500 + struct au_sbinfo *sbinfo;
25503 + if (unlikely(file->private_data))
25507 + /* don't use au_sbilist_lock() here */
25508 + spin_lock(&au_sbilist.spin);
25509 + list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
25510 + if (id == sysaufs_si_id(sbinfo)) {
25511 + kobject_get(&sbinfo->si_kobj);
25512 + sb = sbinfo->si_sb;
25515 + spin_unlock(&au_sbilist.spin);
25518 + if (unlikely(!sb))
25521 + err = au_plink_maint_enter(sb);
25523 + /* keep kobject_get() */
25524 + file->private_data = sbinfo;
25526 + kobject_put(&sbinfo->si_kobj);
25532 + * Accept a valid "si=xxxx" only.
25533 + * Once it is accepted successfully, accept "clean" too.
25535 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
25536 + size_t count, loff_t *ppos)
25539 + unsigned long id;
25540 + /* last newline is allowed */
25541 + char buf[3 + sizeof(unsigned long) * 2 + 1];
25544 + if (unlikely(!capable(CAP_SYS_ADMIN)))
25548 + if (unlikely(count > sizeof(buf)))
25551 + err = copy_from_user(buf, ubuf, count);
25552 + if (unlikely(err)) {
25559 + if (!strcmp("clean", buf)) {
25560 + au_procfs_plm_write_clean(file);
25561 + goto out_success;
25562 + } else if (unlikely(strncmp("si=", buf, 3)))
25565 + err = kstrtoul(buf + 3, 16, &id);
25566 + if (unlikely(err))
25569 + err = au_procfs_plm_write_si(file, id);
25570 + if (unlikely(err))
25574 + err = count; /* success */
25579 +static const struct file_operations au_procfs_plm_fop = {
25580 + .write = au_procfs_plm_write,
25581 + .release = au_procfs_plm_release,
25582 + .owner = THIS_MODULE
25585 +/* ---------------------------------------------------------------------- */
25587 +static struct proc_dir_entry *au_procfs_dir;
25589 +void au_procfs_fin(void)
25591 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
25592 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
25595 +int __init au_procfs_init(void)
25598 + struct proc_dir_entry *entry;
25601 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
25602 + if (unlikely(!au_procfs_dir))
25605 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | S_IWUSR,
25606 + au_procfs_dir, &au_procfs_plm_fop);
25607 + if (unlikely(!entry))
25611 + goto out; /* success */
25615 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
25619 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
25620 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
25621 +++ linux/fs/aufs/rdu.c 2016-01-13 20:11:11.673093671 +0100
25624 + * Copyright (C) 2005-2015 Junjiro R. Okajima
25626 + * This program, aufs is free software; you can redistribute it and/or modify
25627 + * it under the terms of the GNU General Public License as published by
25628 + * the Free Software Foundation; either version 2 of the License, or
25629 + * (at your option) any later version.
25631 + * This program is distributed in the hope that it will be useful,
25632 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25633 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25634 + * GNU General Public License for more details.
25636 + * You should have received a copy of the GNU General Public License
25637 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25641 + * readdir in userspace.
25644 +#include <linux/compat.h>
25645 +#include <linux/fs_stack.h>
25646 +#include <linux/security.h>
25649 +/* bits for struct aufs_rdu.flags */
25650 +#define AuRdu_CALLED 1
25651 +#define AuRdu_CONT (1 << 1)
25652 +#define AuRdu_FULL (1 << 2)
25653 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
25654 +#define au_fset_rdu(flags, name) \
25655 + do { (flags) |= AuRdu_##name; } while (0)
25656 +#define au_fclr_rdu(flags, name) \
25657 + do { (flags) &= ~AuRdu_##name; } while (0)
25659 +struct au_rdu_arg {
25660 + struct dir_context ctx;
25661 + struct aufs_rdu *rdu;
25662 + union au_rdu_ent_ul ent;
25663 + unsigned long end;
25665 + struct super_block *sb;
25669 +static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
25670 + loff_t offset, u64 h_ino, unsigned int d_type)
25673 + struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
25674 + struct aufs_rdu *rdu = arg->rdu;
25675 + struct au_rdu_ent ent;
25679 + au_fset_rdu(rdu->cookie.flags, CALLED);
25680 + len = au_rdu_len(nlen);
25681 + if (arg->ent.ul + len < arg->end) {
25683 + ent.bindex = rdu->cookie.bindex;
25684 + ent.type = d_type;
25686 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
25687 + ent.type = DT_UNKNOWN;
25689 + /* unnecessary to support mmap_sem since this is a dir */
25691 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
25693 + if (copy_to_user(arg->ent.e->name, name, nlen))
25695 + /* the terminating NULL */
25696 + if (__put_user(0, arg->ent.e->name + nlen))
25699 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
25700 + arg->ent.ul += len;
25704 + au_fset_rdu(rdu->cookie.flags, FULL);
25706 + rdu->tail = arg->ent;
25710 + /* AuTraceErr(err); */
25714 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
25718 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
25720 + /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
25721 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
25723 + if (unlikely(offset != cookie->h_pos))
25729 + au_fclr_rdu(cookie->flags, CALLED);
25731 + err = vfsub_iterate_dir(h_file, &arg->ctx);
25735 + && au_ftest_rdu(cookie->flags, CALLED)
25736 + && !au_ftest_rdu(cookie->flags, FULL));
25737 + cookie->h_pos = h_file->f_pos;
25744 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
25747 + aufs_bindex_t bend;
25748 + struct au_rdu_arg arg = {
25750 + .actor = au_rdu_fill
25753 + struct dentry *dentry;
25754 + struct inode *inode;
25755 + struct file *h_file;
25756 + struct au_rdu_cookie *cookie = &rdu->cookie;
25758 + err = !access_ok(VERIFY_WRITE, rdu->ent.e, rdu->sz);
25759 + if (unlikely(err)) {
25765 + rdu->tail = rdu->ent;
25768 + arg.ent = rdu->ent;
25769 + arg.end = arg.ent.ul;
25770 + arg.end += rdu->sz;
25773 + if (unlikely(!file->f_op->iterate))
25776 + err = security_file_permission(file, MAY_READ);
25778 + if (unlikely(err))
25781 + dentry = file->f_path.dentry;
25782 + inode = d_inode(dentry);
25784 + mutex_lock(&inode->i_mutex);
25786 + err = mutex_lock_killable(&inode->i_mutex);
25788 + if (unlikely(err))
25792 + arg.sb = inode->i_sb;
25793 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
25794 + if (unlikely(err))
25796 + err = au_alive_dir(dentry);
25797 + if (unlikely(err))
25799 + /* todo: reval? */
25800 + fi_read_lock(file);
25803 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
25804 + && cookie->generation != au_figen(file)))
25809 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
25811 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
25813 + bend = au_fbstart(file);
25814 + if (cookie->bindex < bend)
25815 + cookie->bindex = bend;
25816 + bend = au_fbend_dir(file);
25817 + /* AuDbg("b%d, b%d\n", cookie->bindex, bend); */
25818 + for (; !err && cookie->bindex <= bend;
25819 + cookie->bindex++, cookie->h_pos = 0) {
25820 + h_file = au_hf_dir(file, cookie->bindex);
25824 + au_fclr_rdu(cookie->flags, FULL);
25825 + err = au_rdu_do(h_file, &arg);
25827 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
25830 + AuDbg("rent %llu\n", rdu->rent);
25832 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
25833 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
25834 + au_fset_rdu(cookie->flags, CONT);
25835 + cookie->generation = au_figen(file);
25838 + ii_read_lock_child(inode);
25839 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibstart(inode)));
25840 + ii_read_unlock(inode);
25843 + fi_read_unlock(file);
25845 + si_read_unlock(arg.sb);
25847 + mutex_unlock(&inode->i_mutex);
25853 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
25857 + unsigned long long nent;
25858 + union au_rdu_ent_ul *u;
25859 + struct au_rdu_ent ent;
25860 + struct super_block *sb;
25863 + nent = rdu->nent;
25865 + sb = file->f_path.dentry->d_sb;
25866 + si_read_lock(sb, AuLock_FLUSH);
25867 + while (nent-- > 0) {
25868 + /* unnecessary to support mmap_sem since this is a dir */
25869 + err = copy_from_user(&ent, u->e, sizeof(ent));
25871 + err = !access_ok(VERIFY_WRITE, &u->e->ino, sizeof(ino));
25872 + if (unlikely(err)) {
25878 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
25880 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
25882 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
25884 + if (unlikely(err)) {
25889 + err = __put_user(ino, &u->e->ino);
25890 + if (unlikely(err)) {
25895 + u->ul += au_rdu_len(ent.nlen);
25897 + si_read_unlock(sb);
25902 +/* ---------------------------------------------------------------------- */
25904 +static int au_rdu_verify(struct aufs_rdu *rdu)
25906 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
25907 + "%llu, b%d, 0x%x, g%u}\n",
25908 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
25910 + rdu->rent, rdu->shwh, rdu->full,
25911 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
25912 + rdu->cookie.generation);
25914 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
25918 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
25922 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
25925 + struct aufs_rdu rdu;
25926 + void __user *p = (void __user *)arg;
25928 + err = copy_from_user(&rdu, p, sizeof(rdu));
25929 + if (unlikely(err)) {
25934 + err = au_rdu_verify(&rdu);
25935 + if (unlikely(err))
25939 + case AUFS_CTL_RDU:
25940 + err = au_rdu(file, &rdu);
25941 + if (unlikely(err))
25944 + e = copy_to_user(p, &rdu, sizeof(rdu));
25945 + if (unlikely(e)) {
25950 + case AUFS_CTL_RDU_INO:
25951 + err = au_rdu_ino(file, &rdu);
25955 + /* err = -ENOTTY; */
25964 +#ifdef CONFIG_COMPAT
25965 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
25968 + struct aufs_rdu rdu;
25969 + void __user *p = compat_ptr(arg);
25971 + /* todo: get_user()? */
25972 + err = copy_from_user(&rdu, p, sizeof(rdu));
25973 + if (unlikely(err)) {
25978 + rdu.ent.e = compat_ptr(rdu.ent.ul);
25979 + err = au_rdu_verify(&rdu);
25980 + if (unlikely(err))
25984 + case AUFS_CTL_RDU:
25985 + err = au_rdu(file, &rdu);
25986 + if (unlikely(err))
25989 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
25990 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
25991 + e = copy_to_user(p, &rdu, sizeof(rdu));
25992 + if (unlikely(e)) {
25997 + case AUFS_CTL_RDU_INO:
25998 + err = au_rdu_ino(file, &rdu);
26002 + /* err = -ENOTTY; */
26011 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
26012 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
26013 +++ linux/fs/aufs/rwsem.h 2016-01-13 20:11:11.673093671 +0100
26016 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26018 + * This program, aufs is free software; you can redistribute it and/or modify
26019 + * it under the terms of the GNU General Public License as published by
26020 + * the Free Software Foundation; either version 2 of the License, or
26021 + * (at your option) any later version.
26023 + * This program is distributed in the hope that it will be useful,
26024 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26025 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26026 + * GNU General Public License for more details.
26028 + * You should have received a copy of the GNU General Public License
26029 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26033 + * simple read-write semaphore wrappers
26036 +#ifndef __AUFS_RWSEM_H__
26037 +#define __AUFS_RWSEM_H__
26041 +#include "debug.h"
26044 + struct rw_semaphore rwsem;
26045 +#ifdef CONFIG_AUFS_DEBUG
26046 + /* just for debugging, not almighty counter */
26047 + atomic_t rcnt, wcnt;
26051 +#ifdef CONFIG_AUFS_DEBUG
26052 +#define AuDbgCntInit(rw) do { \
26053 + atomic_set(&(rw)->rcnt, 0); \
26054 + atomic_set(&(rw)->wcnt, 0); \
26055 + smp_mb(); /* atomic set */ \
26058 +#define AuDbgRcntInc(rw) atomic_inc(&(rw)->rcnt)
26059 +#define AuDbgRcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->rcnt) < 0)
26060 +#define AuDbgWcntInc(rw) atomic_inc(&(rw)->wcnt)
26061 +#define AuDbgWcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->wcnt) < 0)
26063 +#define AuDbgCntInit(rw) do {} while (0)
26064 +#define AuDbgRcntInc(rw) do {} while (0)
26065 +#define AuDbgRcntDec(rw) do {} while (0)
26066 +#define AuDbgWcntInc(rw) do {} while (0)
26067 +#define AuDbgWcntDec(rw) do {} while (0)
26068 +#endif /* CONFIG_AUFS_DEBUG */
26070 +/* to debug easier, do not make them inlined functions */
26071 +#define AuRwMustNoWaiters(rw) AuDebugOn(!list_empty(&(rw)->rwsem.wait_list))
26072 +/* rwsem_is_locked() is unusable */
26073 +#define AuRwMustReadLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0)
26074 +#define AuRwMustWriteLock(rw) AuDebugOn(atomic_read(&(rw)->wcnt) <= 0)
26075 +#define AuRwMustAnyLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0 \
26076 + && atomic_read(&(rw)->wcnt) <= 0)
26077 +#define AuRwDestroy(rw) AuDebugOn(atomic_read(&(rw)->rcnt) \
26078 + || atomic_read(&(rw)->wcnt))
26080 +#define au_rw_class(rw, key) lockdep_set_class(&(rw)->rwsem, key)
26082 +static inline void au_rw_init(struct au_rwsem *rw)
26084 + AuDbgCntInit(rw);
26085 + init_rwsem(&rw->rwsem);
26088 +static inline void au_rw_init_wlock(struct au_rwsem *rw)
26091 + down_write(&rw->rwsem);
26092 + AuDbgWcntInc(rw);
26095 +static inline void au_rw_init_wlock_nested(struct au_rwsem *rw,
26096 + unsigned int lsc)
26099 + down_write_nested(&rw->rwsem, lsc);
26100 + AuDbgWcntInc(rw);
26103 +static inline void au_rw_read_lock(struct au_rwsem *rw)
26105 + down_read(&rw->rwsem);
26106 + AuDbgRcntInc(rw);
26109 +static inline void au_rw_read_lock_nested(struct au_rwsem *rw, unsigned int lsc)
26111 + down_read_nested(&rw->rwsem, lsc);
26112 + AuDbgRcntInc(rw);
26115 +static inline void au_rw_read_unlock(struct au_rwsem *rw)
26117 + AuRwMustReadLock(rw);
26118 + AuDbgRcntDec(rw);
26119 + up_read(&rw->rwsem);
26122 +static inline void au_rw_dgrade_lock(struct au_rwsem *rw)
26124 + AuRwMustWriteLock(rw);
26125 + AuDbgRcntInc(rw);
26126 + AuDbgWcntDec(rw);
26127 + downgrade_write(&rw->rwsem);
26130 +static inline void au_rw_write_lock(struct au_rwsem *rw)
26132 + down_write(&rw->rwsem);
26133 + AuDbgWcntInc(rw);
26136 +static inline void au_rw_write_lock_nested(struct au_rwsem *rw,
26137 + unsigned int lsc)
26139 + down_write_nested(&rw->rwsem, lsc);
26140 + AuDbgWcntInc(rw);
26143 +static inline void au_rw_write_unlock(struct au_rwsem *rw)
26145 + AuRwMustWriteLock(rw);
26146 + AuDbgWcntDec(rw);
26147 + up_write(&rw->rwsem);
26150 +/* why is not _nested version defined */
26151 +static inline int au_rw_read_trylock(struct au_rwsem *rw)
26155 + ret = down_read_trylock(&rw->rwsem);
26157 + AuDbgRcntInc(rw);
26161 +static inline int au_rw_write_trylock(struct au_rwsem *rw)
26165 + ret = down_write_trylock(&rw->rwsem);
26167 + AuDbgWcntInc(rw);
26171 +#undef AuDbgCntInit
26172 +#undef AuDbgRcntInc
26173 +#undef AuDbgRcntDec
26174 +#undef AuDbgWcntInc
26175 +#undef AuDbgWcntDec
26177 +#define AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
26178 +static inline void prefix##_read_lock(param) \
26179 +{ au_rw_read_lock(rwsem); } \
26180 +static inline void prefix##_write_lock(param) \
26181 +{ au_rw_write_lock(rwsem); } \
26182 +static inline int prefix##_read_trylock(param) \
26183 +{ return au_rw_read_trylock(rwsem); } \
26184 +static inline int prefix##_write_trylock(param) \
26185 +{ return au_rw_write_trylock(rwsem); }
26186 +/* why is not _nested version defined */
26187 +/* static inline void prefix##_read_trylock_nested(param, lsc)
26188 +{ au_rw_read_trylock_nested(rwsem, lsc)); }
26189 +static inline void prefix##_write_trylock_nestd(param, lsc)
26190 +{ au_rw_write_trylock_nested(rwsem, lsc); } */
26192 +#define AuSimpleUnlockRwsemFuncs(prefix, param, rwsem) \
26193 +static inline void prefix##_read_unlock(param) \
26194 +{ au_rw_read_unlock(rwsem); } \
26195 +static inline void prefix##_write_unlock(param) \
26196 +{ au_rw_write_unlock(rwsem); } \
26197 +static inline void prefix##_downgrade_lock(param) \
26198 +{ au_rw_dgrade_lock(rwsem); }
26200 +#define AuSimpleRwsemFuncs(prefix, param, rwsem) \
26201 + AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
26202 + AuSimpleUnlockRwsemFuncs(prefix, param, rwsem)
26204 +#endif /* __KERNEL__ */
26205 +#endif /* __AUFS_RWSEM_H__ */
26206 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
26207 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
26208 +++ linux/fs/aufs/sbinfo.c 2016-01-13 20:11:11.673093671 +0100
26211 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26213 + * This program, aufs is free software; you can redistribute it and/or modify
26214 + * it under the terms of the GNU General Public License as published by
26215 + * the Free Software Foundation; either version 2 of the License, or
26216 + * (at your option) any later version.
26218 + * This program is distributed in the hope that it will be useful,
26219 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26220 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26221 + * GNU General Public License for more details.
26223 + * You should have received a copy of the GNU General Public License
26224 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26228 + * superblock private data
26234 + * they are necessary regardless sysfs is disabled.
26236 +void au_si_free(struct kobject *kobj)
26239 + struct au_sbinfo *sbinfo;
26240 + char *locked __maybe_unused; /* debug only */
26242 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
26243 + for (i = 0; i < AuPlink_NHASH; i++)
26244 + AuDebugOn(!hlist_empty(&sbinfo->si_plink[i].head));
26245 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
26247 + AuDebugOn(!hlist_empty(&sbinfo->si_symlink.head));
26249 + au_rw_write_lock(&sbinfo->si_rwsem);
26250 + au_br_free(sbinfo);
26251 + au_rw_write_unlock(&sbinfo->si_rwsem);
26253 + AuDebugOn(radix_tree_gang_lookup
26254 + (&sbinfo->au_si_pid.tree, (void **)&locked,
26255 + /*first_index*/PID_MAX_DEFAULT - 1,
26256 + /*max_items*/sizeof(locked)/sizeof(*locked)));
26258 + kfree(sbinfo->si_branch);
26259 + kfree(sbinfo->au_si_pid.bitmap);
26260 + mutex_destroy(&sbinfo->si_xib_mtx);
26261 + AuRwDestroy(&sbinfo->si_rwsem);
26266 +int au_si_alloc(struct super_block *sb)
26269 + struct au_sbinfo *sbinfo;
26270 + static struct lock_class_key aufs_si;
26273 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
26274 + if (unlikely(!sbinfo))
26277 + BUILD_BUG_ON(sizeof(unsigned long) !=
26278 + sizeof(*sbinfo->au_si_pid.bitmap));
26279 + sbinfo->au_si_pid.bitmap = kcalloc(BITS_TO_LONGS(PID_MAX_DEFAULT),
26280 + sizeof(*sbinfo->au_si_pid.bitmap),
26282 + if (unlikely(!sbinfo->au_si_pid.bitmap))
26285 + /* will be reallocated separately */
26286 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
26287 + if (unlikely(!sbinfo->si_branch))
26290 + err = sysaufs_si_init(sbinfo);
26291 + if (unlikely(err))
26294 + au_nwt_init(&sbinfo->si_nowait);
26295 + au_rw_init_wlock(&sbinfo->si_rwsem);
26296 + au_rw_class(&sbinfo->si_rwsem, &aufs_si);
26297 + spin_lock_init(&sbinfo->au_si_pid.tree_lock);
26298 + INIT_RADIX_TREE(&sbinfo->au_si_pid.tree, GFP_ATOMIC | __GFP_NOFAIL);
26300 + atomic_long_set(&sbinfo->si_ninodes, 0);
26301 + atomic_long_set(&sbinfo->si_nfiles, 0);
26303 + sbinfo->si_bend = -1;
26304 + sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
26306 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
26307 + sbinfo->si_wbr_create = AuWbrCreate_Def;
26308 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
26309 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
26311 + au_fhsm_init(sbinfo);
26313 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
26315 + au_sphl_init(&sbinfo->si_symlink);
26317 + sbinfo->si_xino_jiffy = jiffies;
26318 + sbinfo->si_xino_expire
26319 + = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
26320 + mutex_init(&sbinfo->si_xib_mtx);
26321 + sbinfo->si_xino_brid = -1;
26322 + /* leave si_xib_last_pindex and si_xib_next_bit */
26324 + au_sphl_init(&sbinfo->si_aopen);
26326 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
26327 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
26328 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
26329 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
26331 + for (i = 0; i < AuPlink_NHASH; i++)
26332 + au_sphl_init(sbinfo->si_plink + i);
26333 + init_waitqueue_head(&sbinfo->si_plink_wq);
26334 + spin_lock_init(&sbinfo->si_plink_maint_lock);
26336 + au_sphl_init(&sbinfo->si_files);
26338 + /* with getattr by default */
26339 + sbinfo->si_iop_array = aufs_iop;
26341 + /* leave other members for sysaufs and si_mnt. */
26342 + sbinfo->si_sb = sb;
26343 + sb->s_fs_info = sbinfo;
26345 + return 0; /* success */
26348 + kfree(sbinfo->si_branch);
26350 + kfree(sbinfo->au_si_pid.bitmap);
26357 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr)
26360 + struct au_branch **brp;
26362 + AuRwMustWriteLock(&sbinfo->si_rwsem);
26365 + sz = sizeof(*brp) * (sbinfo->si_bend + 1);
26366 + if (unlikely(!sz))
26367 + sz = sizeof(*brp);
26368 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS);
26370 + sbinfo->si_branch = brp;
26377 +/* ---------------------------------------------------------------------- */
26379 +unsigned int au_sigen_inc(struct super_block *sb)
26381 + unsigned int gen;
26382 + struct inode *inode;
26384 + SiMustWriteLock(sb);
26386 + gen = ++au_sbi(sb)->si_generation;
26387 + au_update_digen(sb->s_root);
26388 + inode = d_inode(sb->s_root);
26389 + au_update_iigen(inode, /*half*/0);
26390 + inode->i_version++;
26394 +aufs_bindex_t au_new_br_id(struct super_block *sb)
26396 + aufs_bindex_t br_id;
26398 + struct au_sbinfo *sbinfo;
26400 + SiMustWriteLock(sb);
26402 + sbinfo = au_sbi(sb);
26403 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
26404 + br_id = ++sbinfo->si_last_br_id;
26405 + AuDebugOn(br_id < 0);
26406 + if (br_id && au_br_index(sb, br_id) < 0)
26413 +/* ---------------------------------------------------------------------- */
26415 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
26416 +int si_read_lock(struct super_block *sb, int flags)
26421 + if (au_ftest_lock(flags, FLUSH))
26422 + au_nwt_flush(&au_sbi(sb)->si_nowait);
26424 + si_noflush_read_lock(sb);
26425 + err = au_plink_maint(sb, flags);
26426 + if (unlikely(err))
26427 + si_read_unlock(sb);
26432 +int si_write_lock(struct super_block *sb, int flags)
26436 + if (au_ftest_lock(flags, FLUSH))
26437 + au_nwt_flush(&au_sbi(sb)->si_nowait);
26439 + si_noflush_write_lock(sb);
26440 + err = au_plink_maint(sb, flags);
26441 + if (unlikely(err))
26442 + si_write_unlock(sb);
26447 +/* dentry and super_block lock. call at entry point */
26448 +int aufs_read_lock(struct dentry *dentry, int flags)
26451 + struct super_block *sb;
26453 + sb = dentry->d_sb;
26454 + err = si_read_lock(sb, flags);
26455 + if (unlikely(err))
26458 + if (au_ftest_lock(flags, DW))
26459 + di_write_lock_child(dentry);
26461 + di_read_lock_child(dentry, flags);
26463 + if (au_ftest_lock(flags, GEN)) {
26464 + err = au_digen_test(dentry, au_sigen(sb));
26465 + if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
26466 + AuDebugOn(!err && au_dbrange_test(dentry));
26468 + err = au_dbrange_test(dentry);
26469 + if (unlikely(err))
26470 + aufs_read_unlock(dentry, flags);
26477 +void aufs_read_unlock(struct dentry *dentry, int flags)
26479 + if (au_ftest_lock(flags, DW))
26480 + di_write_unlock(dentry);
26482 + di_read_unlock(dentry, flags);
26483 + si_read_unlock(dentry->d_sb);
26486 +void aufs_write_lock(struct dentry *dentry)
26488 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
26489 + di_write_lock_child(dentry);
26492 +void aufs_write_unlock(struct dentry *dentry)
26494 + di_write_unlock(dentry);
26495 + si_write_unlock(dentry->d_sb);
26498 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
26501 + unsigned int sigen;
26502 + struct super_block *sb;
26505 + err = si_read_lock(sb, flags);
26506 + if (unlikely(err))
26509 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIRS));
26511 + if (au_ftest_lock(flags, GEN)) {
26512 + sigen = au_sigen(sb);
26513 + err = au_digen_test(d1, sigen);
26514 + AuDebugOn(!err && au_dbrange_test(d1));
26516 + err = au_digen_test(d2, sigen);
26517 + AuDebugOn(!err && au_dbrange_test(d2));
26519 + if (unlikely(err))
26520 + aufs_read_and_write_unlock2(d1, d2);
26527 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
26529 + di_write_unlock2(d1, d2);
26530 + si_read_unlock(d1->d_sb);
26533 +/* ---------------------------------------------------------------------- */
26535 +int si_pid_test_slow(struct super_block *sb)
26540 + p = radix_tree_lookup(&au_sbi(sb)->au_si_pid.tree, current->pid);
26541 + rcu_read_unlock();
26543 + return (long)!!p;
26546 +void si_pid_set_slow(struct super_block *sb)
26549 + struct au_sbinfo *sbinfo;
26551 + AuDebugOn(si_pid_test_slow(sb));
26553 + sbinfo = au_sbi(sb);
26554 + err = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
26556 + spin_lock(&sbinfo->au_si_pid.tree_lock);
26557 + err = radix_tree_insert(&sbinfo->au_si_pid.tree, current->pid,
26558 + /*any valid ptr*/sb);
26559 + spin_unlock(&sbinfo->au_si_pid.tree_lock);
26561 + radix_tree_preload_end();
26564 +void si_pid_clr_slow(struct super_block *sb)
26567 + struct au_sbinfo *sbinfo;
26569 + AuDebugOn(!si_pid_test_slow(sb));
26571 + sbinfo = au_sbi(sb);
26572 + spin_lock(&sbinfo->au_si_pid.tree_lock);
26573 + p = radix_tree_delete(&sbinfo->au_si_pid.tree, current->pid);
26574 + spin_unlock(&sbinfo->au_si_pid.tree_lock);
26576 diff -urN /usr/share/empty/fs/aufs/spl.h linux/fs/aufs/spl.h
26577 --- /usr/share/empty/fs/aufs/spl.h 1970-01-01 01:00:00.000000000 +0100
26578 +++ linux/fs/aufs/spl.h 2016-01-13 20:11:11.673093671 +0100
26581 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26583 + * This program, aufs is free software; you can redistribute it and/or modify
26584 + * it under the terms of the GNU General Public License as published by
26585 + * the Free Software Foundation; either version 2 of the License, or
26586 + * (at your option) any later version.
26588 + * This program is distributed in the hope that it will be useful,
26589 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26590 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26591 + * GNU General Public License for more details.
26593 + * You should have received a copy of the GNU General Public License
26594 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26598 + * simple list protected by a spinlock
26601 +#ifndef __AUFS_SPL_H__
26602 +#define __AUFS_SPL_H__
26606 +struct au_splhead {
26608 + struct list_head head;
26611 +static inline void au_spl_init(struct au_splhead *spl)
26613 + spin_lock_init(&spl->spin);
26614 + INIT_LIST_HEAD(&spl->head);
26617 +static inline void au_spl_add(struct list_head *list, struct au_splhead *spl)
26619 + spin_lock(&spl->spin);
26620 + list_add(list, &spl->head);
26621 + spin_unlock(&spl->spin);
26624 +static inline void au_spl_del(struct list_head *list, struct au_splhead *spl)
26626 + spin_lock(&spl->spin);
26628 + spin_unlock(&spl->spin);
26631 +static inline void au_spl_del_rcu(struct list_head *list,
26632 + struct au_splhead *spl)
26634 + spin_lock(&spl->spin);
26635 + list_del_rcu(list);
26636 + spin_unlock(&spl->spin);
26639 +/* ---------------------------------------------------------------------- */
26641 +struct au_sphlhead {
26643 + struct hlist_head head;
26646 +static inline void au_sphl_init(struct au_sphlhead *sphl)
26648 + spin_lock_init(&sphl->spin);
26649 + INIT_HLIST_HEAD(&sphl->head);
26652 +static inline void au_sphl_add(struct hlist_node *hlist,
26653 + struct au_sphlhead *sphl)
26655 + spin_lock(&sphl->spin);
26656 + hlist_add_head(hlist, &sphl->head);
26657 + spin_unlock(&sphl->spin);
26660 +static inline void au_sphl_del(struct hlist_node *hlist,
26661 + struct au_sphlhead *sphl)
26663 + spin_lock(&sphl->spin);
26664 + hlist_del(hlist);
26665 + spin_unlock(&sphl->spin);
26668 +static inline void au_sphl_del_rcu(struct hlist_node *hlist,
26669 + struct au_sphlhead *sphl)
26671 + spin_lock(&sphl->spin);
26672 + hlist_del_rcu(hlist);
26673 + spin_unlock(&sphl->spin);
26676 +static inline unsigned long au_sphl_count(struct au_sphlhead *sphl)
26678 + unsigned long cnt;
26679 + struct hlist_node *pos;
26682 + spin_lock(&sphl->spin);
26683 + hlist_for_each(pos, &sphl->head)
26685 + spin_unlock(&sphl->spin);
26689 +#endif /* __KERNEL__ */
26690 +#endif /* __AUFS_SPL_H__ */
26691 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
26692 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
26693 +++ linux/fs/aufs/super.c 2016-01-13 20:11:11.673093671 +0100
26696 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26698 + * This program, aufs is free software; you can redistribute it and/or modify
26699 + * it under the terms of the GNU General Public License as published by
26700 + * the Free Software Foundation; either version 2 of the License, or
26701 + * (at your option) any later version.
26703 + * This program is distributed in the hope that it will be useful,
26704 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26705 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26706 + * GNU General Public License for more details.
26708 + * You should have received a copy of the GNU General Public License
26709 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26713 + * mount and super_block operations
26716 +#include <linux/mm.h>
26717 +#include <linux/seq_file.h>
26718 +#include <linux/statfs.h>
26719 +#include <linux/vmalloc.h>
26723 + * super_operations
26725 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
26727 + struct au_icntnr *c;
26729 + c = au_cache_alloc_icntnr();
26731 + au_icntnr_init(c);
26732 + c->vfs_inode.i_version = 1; /* sigen(sb); */
26733 + c->iinfo.ii_hinode = NULL;
26734 + return &c->vfs_inode;
26739 +static void aufs_destroy_inode_cb(struct rcu_head *head)
26741 + struct inode *inode = container_of(head, struct inode, i_rcu);
26743 + INIT_HLIST_HEAD(&inode->i_dentry);
26744 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
26747 +static void aufs_destroy_inode(struct inode *inode)
26749 + au_iinfo_fin(inode);
26750 + call_rcu(&inode->i_rcu, aufs_destroy_inode_cb);
26753 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
26755 + struct inode *inode;
26758 + inode = iget_locked(sb, ino);
26759 + if (unlikely(!inode)) {
26760 + inode = ERR_PTR(-ENOMEM);
26763 + if (!(inode->i_state & I_NEW))
26766 + err = au_xigen_new(inode);
26768 + err = au_iinfo_init(inode);
26770 + inode->i_version++;
26772 + iget_failed(inode);
26773 + inode = ERR_PTR(err);
26777 + /* never return NULL */
26778 + AuDebugOn(!inode);
26779 + AuTraceErrPtr(inode);
26783 +/* lock free root dinfo */
26784 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
26787 + aufs_bindex_t bindex, bend;
26788 + struct path path;
26789 + struct au_hdentry *hdp;
26790 + struct au_branch *br;
26791 + au_br_perm_str_t perm;
26794 + bend = au_sbend(sb);
26795 + hdp = au_di(sb->s_root)->di_hdentry;
26796 + for (bindex = 0; !err && bindex <= bend; bindex++) {
26797 + br = au_sbr(sb, bindex);
26798 + path.mnt = au_br_mnt(br);
26799 + path.dentry = hdp[bindex].hd_dentry;
26800 + err = au_seq_path(seq, &path);
26802 + au_optstr_br_perm(&perm, br->br_perm);
26803 + seq_printf(seq, "=%s", perm.a);
26804 + if (bindex != bend)
26805 + seq_putc(seq, ':');
26808 + if (unlikely(err || seq_has_overflowed(seq)))
26814 +static void au_show_wbr_create(struct seq_file *m, int v,
26815 + struct au_sbinfo *sbinfo)
26819 + AuRwMustAnyLock(&sbinfo->si_rwsem);
26821 + seq_puts(m, ",create=");
26822 + pat = au_optstr_wbr_create(v);
26824 + case AuWbrCreate_TDP:
26825 + case AuWbrCreate_RR:
26826 + case AuWbrCreate_MFS:
26827 + case AuWbrCreate_PMFS:
26828 + seq_puts(m, pat);
26830 + case AuWbrCreate_MFSV:
26831 + seq_printf(m, /*pat*/"mfs:%lu",
26832 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26835 + case AuWbrCreate_PMFSV:
26836 + seq_printf(m, /*pat*/"pmfs:%lu",
26837 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26840 + case AuWbrCreate_MFSRR:
26841 + seq_printf(m, /*pat*/"mfsrr:%llu",
26842 + sbinfo->si_wbr_mfs.mfsrr_watermark);
26844 + case AuWbrCreate_MFSRRV:
26845 + seq_printf(m, /*pat*/"mfsrr:%llu:%lu",
26846 + sbinfo->si_wbr_mfs.mfsrr_watermark,
26847 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26850 + case AuWbrCreate_PMFSRR:
26851 + seq_printf(m, /*pat*/"pmfsrr:%llu",
26852 + sbinfo->si_wbr_mfs.mfsrr_watermark);
26854 + case AuWbrCreate_PMFSRRV:
26855 + seq_printf(m, /*pat*/"pmfsrr:%llu:%lu",
26856 + sbinfo->si_wbr_mfs.mfsrr_watermark,
26857 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26863 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
26865 +#ifdef CONFIG_SYSFS
26869 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
26870 + aufs_bindex_t bindex, brid;
26871 + struct qstr *name;
26873 + struct dentry *d, *h_root;
26874 + struct au_hdentry *hdp;
26876 + AuRwMustAnyLock(&sbinfo->si_rwsem);
26879 + f = au_sbi(sb)->si_xib;
26883 + /* stop printing the default xino path on the first writable branch */
26885 + brid = au_xino_brid(sb);
26887 + bindex = au_br_index(sb, brid);
26888 + hdp = au_di(sb->s_root)->di_hdentry;
26889 + h_root = hdp[0 + bindex].hd_dentry;
26891 + d = f->f_path.dentry;
26892 + name = &d->d_name;
26893 + /* safe ->d_parent because the file is unlinked */
26894 + if (d->d_parent == h_root
26895 + && name->len == len
26896 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
26899 + seq_puts(seq, ",xino=");
26900 + err = au_xino_path(seq, f);
26907 +/* seq_file will re-call me in case of too long string */
26908 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
26911 + unsigned int mnt_flags, v;
26912 + struct super_block *sb;
26913 + struct au_sbinfo *sbinfo;
26915 +#define AuBool(name, str) do { \
26916 + v = au_opt_test(mnt_flags, name); \
26917 + if (v != au_opt_test(AuOpt_Def, name)) \
26918 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
26921 +#define AuStr(name, str) do { \
26922 + v = mnt_flags & AuOptMask_##name; \
26923 + if (v != (AuOpt_Def & AuOptMask_##name)) \
26924 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
26927 +#define AuUInt(name, str, val) do { \
26928 + if (val != AUFS_##name##_DEF) \
26929 + seq_printf(m, "," #str "=%u", val); \
26932 + sb = dentry->d_sb;
26933 + if (sb->s_flags & MS_POSIXACL)
26934 + seq_puts(m, ",acl");
26936 + /* lock free root dinfo */
26937 + si_noflush_read_lock(sb);
26938 + sbinfo = au_sbi(sb);
26939 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
26941 + mnt_flags = au_mntflags(sb);
26942 + if (au_opt_test(mnt_flags, XINO)) {
26943 + err = au_show_xino(m, sb);
26944 + if (unlikely(err))
26947 + seq_puts(m, ",noxino");
26949 + AuBool(TRUNC_XINO, trunc_xino);
26950 + AuStr(UDBA, udba);
26951 + AuBool(SHWH, shwh);
26952 + AuBool(PLINK, plink);
26953 + AuBool(DIO, dio);
26954 + AuBool(DIRPERM1, dirperm1);
26956 + v = sbinfo->si_wbr_create;
26957 + if (v != AuWbrCreate_Def)
26958 + au_show_wbr_create(m, v, sbinfo);
26960 + v = sbinfo->si_wbr_copyup;
26961 + if (v != AuWbrCopyup_Def)
26962 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
26964 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
26965 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
26966 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
26968 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
26970 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
26971 + AuUInt(RDCACHE, rdcache, v);
26973 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
26974 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
26976 + au_fhsm_show(m, sbinfo);
26978 + AuBool(SUM, sum);
26979 + /* AuBool(SUM_W, wsum); */
26980 + AuBool(WARN_PERM, warn_perm);
26981 + AuBool(VERBOSE, verbose);
26984 + /* be sure to print "br:" last */
26985 + if (!sysaufs_brs) {
26986 + seq_puts(m, ",br:");
26987 + au_show_brs(m, sb);
26989 + si_read_unlock(sb);
26997 +/* ---------------------------------------------------------------------- */
26999 +/* sum mode which returns the summation for statfs(2) */
27001 +static u64 au_add_till_max(u64 a, u64 b)
27009 + return ULLONG_MAX;
27012 +static u64 au_mul_till_max(u64 a, long mul)
27020 + return ULLONG_MAX;
27023 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
27026 + long bsize, factor;
27027 + u64 blocks, bfree, bavail, files, ffree;
27028 + aufs_bindex_t bend, bindex, i;
27029 + unsigned char shared;
27030 + struct path h_path;
27031 + struct super_block *h_sb;
27034 + bsize = LONG_MAX;
27040 + bend = au_sbend(sb);
27041 + for (bindex = 0; bindex <= bend; bindex++) {
27042 + h_path.mnt = au_sbr_mnt(sb, bindex);
27043 + h_sb = h_path.mnt->mnt_sb;
27045 + for (i = 0; !shared && i < bindex; i++)
27046 + shared = (au_sbr_sb(sb, i) == h_sb);
27050 + /* sb->s_root for NFS is unreliable */
27051 + h_path.dentry = h_path.mnt->mnt_root;
27052 + err = vfs_statfs(&h_path, buf);
27053 + if (unlikely(err))
27056 + if (bsize > buf->f_bsize) {
27058 + * we will reduce bsize, so we have to expand blocks
27059 + * etc. to match them again
27061 + factor = (bsize / buf->f_bsize);
27062 + blocks = au_mul_till_max(blocks, factor);
27063 + bfree = au_mul_till_max(bfree, factor);
27064 + bavail = au_mul_till_max(bavail, factor);
27065 + bsize = buf->f_bsize;
27068 + factor = (buf->f_bsize / bsize);
27069 + blocks = au_add_till_max(blocks,
27070 + au_mul_till_max(buf->f_blocks, factor));
27071 + bfree = au_add_till_max(bfree,
27072 + au_mul_till_max(buf->f_bfree, factor));
27073 + bavail = au_add_till_max(bavail,
27074 + au_mul_till_max(buf->f_bavail, factor));
27075 + files = au_add_till_max(files, buf->f_files);
27076 + ffree = au_add_till_max(ffree, buf->f_ffree);
27079 + buf->f_bsize = bsize;
27080 + buf->f_blocks = blocks;
27081 + buf->f_bfree = bfree;
27082 + buf->f_bavail = bavail;
27083 + buf->f_files = files;
27084 + buf->f_ffree = ffree;
27085 + buf->f_frsize = 0;
27091 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
27094 + struct path h_path;
27095 + struct super_block *sb;
27097 + /* lock free root dinfo */
27098 + sb = dentry->d_sb;
27099 + si_noflush_read_lock(sb);
27100 + if (!au_opt_test(au_mntflags(sb), SUM)) {
27101 + /* sb->s_root for NFS is unreliable */
27102 + h_path.mnt = au_sbr_mnt(sb, 0);
27103 + h_path.dentry = h_path.mnt->mnt_root;
27104 + err = vfs_statfs(&h_path, buf);
27106 + err = au_statfs_sum(sb, buf);
27107 + si_read_unlock(sb);
27110 + buf->f_type = AUFS_SUPER_MAGIC;
27111 + buf->f_namelen = AUFS_MAX_NAMELEN;
27112 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
27114 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
27119 +/* ---------------------------------------------------------------------- */
27121 +static int aufs_sync_fs(struct super_block *sb, int wait)
27124 + aufs_bindex_t bend, bindex;
27125 + struct au_branch *br;
27126 + struct super_block *h_sb;
27129 + si_noflush_read_lock(sb);
27130 + bend = au_sbend(sb);
27131 + for (bindex = 0; bindex <= bend; bindex++) {
27132 + br = au_sbr(sb, bindex);
27133 + if (!au_br_writable(br->br_perm))
27136 + h_sb = au_sbr_sb(sb, bindex);
27137 + if (h_sb->s_op->sync_fs) {
27138 + e = h_sb->s_op->sync_fs(h_sb, wait);
27139 + if (unlikely(e && !err))
27141 + /* go on even if an error happens */
27144 + si_read_unlock(sb);
27149 +/* ---------------------------------------------------------------------- */
27151 +/* final actions when unmounting a file system */
27152 +static void aufs_put_super(struct super_block *sb)
27154 + struct au_sbinfo *sbinfo;
27156 + sbinfo = au_sbi(sb);
27160 + dbgaufs_si_fin(sbinfo);
27161 + kobject_put(&sbinfo->si_kobj);
27164 +/* ---------------------------------------------------------------------- */
27166 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
27167 + struct super_block *sb, void *arg)
27170 + unsigned long long n, sz;
27177 + if (*hint > ULLONG_MAX / sizeof(array)) {
27178 + array = ERR_PTR(-EMFILE);
27179 + pr_err("hint %llu\n", *hint);
27183 + sz = sizeof(array) * *hint;
27184 + array = kzalloc(sz, GFP_NOFS);
27185 + if (unlikely(!array))
27186 + array = vzalloc(sz);
27187 + if (unlikely(!array)) {
27188 + array = ERR_PTR(-ENOMEM);
27192 + n = cb(sb, array, *hint, arg);
27193 + AuDebugOn(n > *hint);
27200 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
27201 + unsigned long long max __maybe_unused,
27204 + unsigned long long n;
27205 + struct inode **p, *inode;
27206 + struct list_head *head;
27211 + spin_lock(&sb->s_inode_list_lock);
27212 + list_for_each_entry(inode, head, i_sb_list) {
27213 + if (!is_bad_inode(inode)
27214 + && au_ii(inode)->ii_bstart >= 0) {
27215 + spin_lock(&inode->i_lock);
27216 + if (atomic_read(&inode->i_count)) {
27220 + AuDebugOn(n > max);
27222 + spin_unlock(&inode->i_lock);
27225 + spin_unlock(&sb->s_inode_list_lock);
27230 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
27232 + *max = atomic_long_read(&au_sbi(sb)->si_ninodes);
27233 + return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
27236 +void au_iarray_free(struct inode **a, unsigned long long max)
27238 + unsigned long long ull;
27240 + for (ull = 0; ull < max; ull++)
27245 +/* ---------------------------------------------------------------------- */
27248 + * refresh dentry and inode at remount time.
27250 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
27251 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
27252 + struct dentry *parent)
27256 + di_write_lock_child(dentry);
27257 + di_read_lock_parent(parent, AuLock_IR);
27258 + err = au_refresh_dentry(dentry, parent);
27259 + if (!err && dir_flags)
27260 + au_hn_reset(d_inode(dentry), dir_flags);
27261 + di_read_unlock(parent, AuLock_IR);
27262 + di_write_unlock(dentry);
27267 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
27268 + struct au_sbinfo *sbinfo,
27269 + const unsigned int dir_flags, unsigned int do_idop)
27272 + struct dentry *parent;
27275 + parent = dget_parent(dentry);
27276 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
27277 + if (d_really_is_positive(dentry)) {
27278 + if (!d_is_dir(dentry))
27279 + err = au_do_refresh(dentry, /*dir_flags*/0,
27282 + err = au_do_refresh(dentry, dir_flags, parent);
27283 + if (unlikely(err))
27284 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
27287 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
27288 + AuDbgDentry(dentry);
27294 + au_refresh_dop(dentry, /*force_reval*/0);
27296 + au_refresh_dop(dentry, /*force_reval*/1);
27302 +static int au_refresh_d(struct super_block *sb, unsigned int do_idop)
27304 + int err, i, j, ndentry, e;
27305 + unsigned int sigen;
27306 + struct au_dcsub_pages dpages;
27307 + struct au_dpage *dpage;
27308 + struct dentry **dentries, *d;
27309 + struct au_sbinfo *sbinfo;
27310 + struct dentry *root = sb->s_root;
27311 + const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
27314 + au_refresh_dop(root, /*force_reval*/0);
27316 + err = au_dpages_init(&dpages, GFP_NOFS);
27317 + if (unlikely(err))
27319 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
27320 + if (unlikely(err))
27323 + sigen = au_sigen(sb);
27324 + sbinfo = au_sbi(sb);
27325 + for (i = 0; i < dpages.ndpage; i++) {
27326 + dpage = dpages.dpages + i;
27327 + dentries = dpage->dentries;
27328 + ndentry = dpage->ndentry;
27329 + for (j = 0; j < ndentry; j++) {
27331 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
27333 + if (unlikely(e && !err))
27335 + /* go on even err */
27340 + au_dpages_free(&dpages);
27345 +static int au_refresh_i(struct super_block *sb, unsigned int do_idop)
27348 + unsigned int sigen;
27349 + unsigned long long max, ull;
27350 + struct inode *inode, **array;
27352 + array = au_iarray_alloc(sb, &max);
27353 + err = PTR_ERR(array);
27354 + if (IS_ERR(array))
27358 + sigen = au_sigen(sb);
27359 + for (ull = 0; ull < max; ull++) {
27360 + inode = array[ull];
27361 + if (unlikely(!inode))
27365 + ii_write_lock_child(inode);
27366 + if (au_iigen(inode, NULL) != sigen) {
27367 + e = au_refresh_hinode_self(inode);
27368 + if (unlikely(e)) {
27369 + au_refresh_iop(inode, /*force_getattr*/1);
27370 + pr_err("error %d, i%lu\n", e, inode->i_ino);
27373 + /* go on even if err */
27376 + if (!e && do_idop)
27377 + au_refresh_iop(inode, /*force_getattr*/0);
27378 + ii_write_unlock(inode);
27381 + au_iarray_free(array, max);
27387 +static void au_remount_refresh(struct super_block *sb, unsigned int do_idop)
27390 + unsigned int udba;
27391 + aufs_bindex_t bindex, bend;
27392 + struct dentry *root;
27393 + struct inode *inode;
27394 + struct au_branch *br;
27395 + struct au_sbinfo *sbi;
27397 + au_sigen_inc(sb);
27398 + sbi = au_sbi(sb);
27399 + au_fclr_si(sbi, FAILED_REFRESH_DIR);
27401 + root = sb->s_root;
27402 + DiMustNoWaiters(root);
27403 + inode = d_inode(root);
27404 + IiMustNoWaiters(inode);
27406 + udba = au_opt_udba(sb);
27407 + bend = au_sbend(sb);
27408 + for (bindex = 0; bindex <= bend; bindex++) {
27409 + br = au_sbr(sb, bindex);
27410 + err = au_hnotify_reset_br(udba, br, br->br_perm);
27411 + if (unlikely(err))
27412 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
27414 + /* go on even if err */
27416 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
27419 + if (au_ftest_si(sbi, NO_DREVAL)) {
27420 + AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
27421 + sb->s_d_op = &aufs_dop_noreval;
27422 + AuDebugOn(sbi->si_iop_array == aufs_iop_nogetattr);
27423 + sbi->si_iop_array = aufs_iop_nogetattr;
27425 + AuDebugOn(sb->s_d_op == &aufs_dop);
27426 + sb->s_d_op = &aufs_dop;
27427 + AuDebugOn(sbi->si_iop_array == aufs_iop);
27428 + sbi->si_iop_array = aufs_iop;
27430 + pr_info("reset to %pf and %pf\n",
27431 + sb->s_d_op, sbi->si_iop_array);
27434 + di_write_unlock(root);
27435 + err = au_refresh_d(sb, do_idop);
27436 + e = au_refresh_i(sb, do_idop);
27437 + if (unlikely(e && !err))
27439 + /* aufs_write_lock() calls ..._child() */
27440 + di_write_lock_child(root);
27442 + au_cpup_attr_all(inode, /*force*/1);
27444 + if (unlikely(err))
27445 + AuIOErr("refresh failed, ignored, %d\n", err);
27448 +/* stop extra interpretation of errno in mount(8), and strange error messages */
27449 +static int cvt_err(int err)
27463 +static int aufs_remount_fs(struct super_block *sb, int *flags, char *data)
27466 + unsigned int mntflags;
27467 + struct au_opts opts = {
27470 + struct dentry *root;
27471 + struct inode *inode;
27472 + struct au_sbinfo *sbinfo;
27475 + root = sb->s_root;
27476 + if (!data || !*data) {
27477 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
27479 + di_write_lock_child(root);
27480 + err = au_opts_verify(sb, *flags, /*pending*/0);
27481 + aufs_write_unlock(root);
27487 + opts.opt = (void *)__get_free_page(GFP_NOFS);
27488 + if (unlikely(!opts.opt))
27490 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
27491 + opts.flags = AuOpts_REMOUNT;
27492 + opts.sb_flags = *flags;
27494 + /* parse it before aufs lock */
27495 + err = au_opts_parse(sb, data, &opts);
27496 + if (unlikely(err))
27499 + sbinfo = au_sbi(sb);
27500 + inode = d_inode(root);
27501 + mutex_lock(&inode->i_mutex);
27502 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
27503 + if (unlikely(err))
27505 + di_write_lock_child(root);
27507 + /* au_opts_remount() may return an error */
27508 + err = au_opts_remount(sb, &opts);
27509 + au_opts_free(&opts);
27511 + if (au_ftest_opts(opts.flags, REFRESH))
27512 + au_remount_refresh(sb, au_ftest_opts(opts.flags, REFRESH_IDOP));
27514 + if (au_ftest_opts(opts.flags, REFRESH_DYAOP)) {
27515 + mntflags = au_mntflags(sb);
27516 + do_dx = !!au_opt_test(mntflags, DIO);
27517 + au_dy_arefresh(do_dx);
27520 + au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
27521 + aufs_write_unlock(root);
27524 + mutex_unlock(&inode->i_mutex);
27526 + free_page((unsigned long)opts.opt);
27528 + err = cvt_err(err);
27533 +static const struct super_operations aufs_sop = {
27534 + .alloc_inode = aufs_alloc_inode,
27535 + .destroy_inode = aufs_destroy_inode,
27536 + /* always deleting, no clearing */
27537 + .drop_inode = generic_delete_inode,
27538 + .show_options = aufs_show_options,
27539 + .statfs = aufs_statfs,
27540 + .put_super = aufs_put_super,
27541 + .sync_fs = aufs_sync_fs,
27542 + .remount_fs = aufs_remount_fs
27545 +/* ---------------------------------------------------------------------- */
27547 +static int alloc_root(struct super_block *sb)
27550 + struct inode *inode;
27551 + struct dentry *root;
27554 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
27555 + err = PTR_ERR(inode);
27556 + if (IS_ERR(inode))
27559 + inode->i_op = aufs_iop + AuIop_DIR; /* with getattr by default */
27560 + inode->i_fop = &aufs_dir_fop;
27561 + inode->i_mode = S_IFDIR;
27562 + set_nlink(inode, 2);
27563 + unlock_new_inode(inode);
27565 + root = d_make_root(inode);
27566 + if (unlikely(!root))
27568 + err = PTR_ERR(root);
27569 + if (IS_ERR(root))
27572 + err = au_di_init(root);
27574 + sb->s_root = root;
27575 + return 0; /* success */
27583 +static int aufs_fill_super(struct super_block *sb, void *raw_data,
27584 + int silent __maybe_unused)
27587 + struct au_opts opts = {
27590 + struct au_sbinfo *sbinfo;
27591 + struct dentry *root;
27592 + struct inode *inode;
27593 + char *arg = raw_data;
27595 + if (unlikely(!arg || !*arg)) {
27597 + pr_err("no arg\n");
27602 + opts.opt = (void *)__get_free_page(GFP_NOFS);
27603 + if (unlikely(!opts.opt))
27605 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
27606 + opts.sb_flags = sb->s_flags;
27608 + err = au_si_alloc(sb);
27609 + if (unlikely(err))
27611 + sbinfo = au_sbi(sb);
27613 + /* all timestamps always follow the ones on the branch */
27614 + sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
27615 + sb->s_op = &aufs_sop;
27616 + sb->s_d_op = &aufs_dop;
27617 + sb->s_magic = AUFS_SUPER_MAGIC;
27618 + sb->s_maxbytes = 0;
27619 + sb->s_stack_depth = 1;
27620 + au_export_init(sb);
27621 + /* au_xattr_init(sb); */
27623 + err = alloc_root(sb);
27624 + if (unlikely(err)) {
27625 + si_write_unlock(sb);
27628 + root = sb->s_root;
27629 + inode = d_inode(root);
27632 + * actually we can parse options regardless aufs lock here.
27633 + * but at remount time, parsing must be done before aufs lock.
27634 + * so we follow the same rule.
27636 + ii_write_lock_parent(inode);
27637 + aufs_write_unlock(root);
27638 + err = au_opts_parse(sb, arg, &opts);
27639 + if (unlikely(err))
27642 + /* lock vfs_inode first, then aufs. */
27643 + mutex_lock(&inode->i_mutex);
27644 + aufs_write_lock(root);
27645 + err = au_opts_mount(sb, &opts);
27646 + au_opts_free(&opts);
27647 + if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
27648 + sb->s_d_op = &aufs_dop_noreval;
27649 + pr_info("%pf\n", sb->s_d_op);
27650 + au_refresh_dop(root, /*force_reval*/0);
27651 + sbinfo->si_iop_array = aufs_iop_nogetattr;
27652 + au_refresh_iop(inode, /*force_getattr*/0);
27654 + aufs_write_unlock(root);
27655 + mutex_unlock(&inode->i_mutex);
27657 + goto out_opts; /* success */
27661 + sb->s_root = NULL;
27663 + dbgaufs_si_fin(sbinfo);
27664 + kobject_put(&sbinfo->si_kobj);
27665 + sb->s_fs_info = NULL;
27667 + free_page((unsigned long)opts.opt);
27670 + err = cvt_err(err);
27675 +/* ---------------------------------------------------------------------- */
27677 +static struct dentry *aufs_mount(struct file_system_type *fs_type, int flags,
27678 + const char *dev_name __maybe_unused,
27681 + struct dentry *root;
27682 + struct super_block *sb;
27684 + /* all timestamps always follow the ones on the branch */
27685 + /* mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; */
27686 + root = mount_nodev(fs_type, flags, raw_data, aufs_fill_super);
27687 + if (IS_ERR(root))
27691 + si_write_lock(sb, !AuLock_FLUSH);
27692 + sysaufs_brs_add(sb, 0);
27693 + si_write_unlock(sb);
27694 + au_sbilist_add(sb);
27700 +static void aufs_kill_sb(struct super_block *sb)
27702 + struct au_sbinfo *sbinfo;
27704 + sbinfo = au_sbi(sb);
27706 + au_sbilist_del(sb);
27707 + aufs_write_lock(sb->s_root);
27709 + if (sbinfo->si_wbr_create_ops->fin)
27710 + sbinfo->si_wbr_create_ops->fin(sb);
27711 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
27712 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
27713 + au_remount_refresh(sb, /*do_idop*/0);
27715 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27716 + au_plink_put(sb, /*verbose*/1);
27718 + sbinfo->si_sb = NULL;
27719 + aufs_write_unlock(sb->s_root);
27720 + au_nwt_flush(&sbinfo->si_nowait);
27722 + kill_anon_super(sb);
27725 +struct file_system_type aufs_fs_type = {
27726 + .name = AUFS_FSTYPE,
27727 + /* a race between rename and others */
27728 + .fs_flags = FS_RENAME_DOES_D_MOVE,
27729 + .mount = aufs_mount,
27730 + .kill_sb = aufs_kill_sb,
27731 + /* no need to __module_get() and module_put(). */
27732 + .owner = THIS_MODULE,
27734 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
27735 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
27736 +++ linux/fs/aufs/super.h 2016-01-13 20:11:11.673093671 +0100
27739 + * Copyright (C) 2005-2015 Junjiro R. Okajima
27741 + * This program, aufs is free software; you can redistribute it and/or modify
27742 + * it under the terms of the GNU General Public License as published by
27743 + * the Free Software Foundation; either version 2 of the License, or
27744 + * (at your option) any later version.
27746 + * This program is distributed in the hope that it will be useful,
27747 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27748 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27749 + * GNU General Public License for more details.
27751 + * You should have received a copy of the GNU General Public License
27752 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27756 + * super_block operations
27759 +#ifndef __AUFS_SUPER_H__
27760 +#define __AUFS_SUPER_H__
27764 +#include <linux/fs.h>
27765 +#include <linux/kobject.h>
27766 +#include "rwsem.h"
27770 +/* policies to select one among multiple writable branches */
27771 +struct au_wbr_copyup_operations {
27772 + int (*copyup)(struct dentry *dentry);
27775 +#define AuWbr_DIR 1 /* target is a dir */
27776 +#define AuWbr_PARENT (1 << 1) /* always require a parent */
27778 +#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
27779 +#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
27780 +#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
27782 +struct au_wbr_create_operations {
27783 + int (*create)(struct dentry *dentry, unsigned int flags);
27784 + int (*init)(struct super_block *sb);
27785 + int (*fin)(struct super_block *sb);
27788 +struct au_wbr_mfs {
27789 + struct mutex mfs_lock; /* protect this structure */
27790 + unsigned long mfs_jiffy;
27791 + unsigned long mfs_expire;
27792 + aufs_bindex_t mfs_bindex;
27794 + unsigned long long mfsrr_bytes;
27795 + unsigned long long mfsrr_watermark;
27798 +struct pseudo_link {
27800 + struct hlist_node hlist;
27801 + struct rcu_head rcu;
27803 + struct inode *inode;
27806 +#define AuPlink_NHASH 100
27807 +static inline int au_plink_hash(ino_t ino)
27809 + return ino % AuPlink_NHASH;
27812 +/* File-based Hierarchical Storage Management */
27814 +#ifdef CONFIG_AUFS_FHSM
27815 + /* allow only one process who can receive the notification */
27816 + spinlock_t fhsm_spin;
27818 + wait_queue_head_t fhsm_wqh;
27819 + atomic_t fhsm_readable;
27821 + /* these are protected by si_rwsem */
27822 + unsigned long fhsm_expire;
27823 + aufs_bindex_t fhsm_bottom;
27828 +struct au_sbinfo {
27829 + /* nowait tasks in the system-wide workqueue */
27830 + struct au_nowait_tasks si_nowait;
27833 + * tried sb->s_umount, but failed due to the dependecy between i_mutex.
27834 + * rwsem for au_sbinfo is necessary.
27836 + struct au_rwsem si_rwsem;
27838 + /* prevent recursive locking in deleting inode */
27840 + unsigned long *bitmap;
27841 + spinlock_t tree_lock;
27842 + struct radix_tree_root tree;
27846 + * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
27849 + atomic_long_t si_ninodes, si_nfiles;
27851 + /* branch management */
27852 + unsigned int si_generation;
27854 + /* see AuSi_ flags */
27855 + unsigned char au_si_status;
27857 + aufs_bindex_t si_bend;
27859 + /* dirty trick to keep br_id plus */
27860 + unsigned int si_last_br_id :
27861 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
27862 + struct au_branch **si_branch;
27864 + /* policy to select a writable branch */
27865 + unsigned char si_wbr_copyup;
27866 + unsigned char si_wbr_create;
27867 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
27868 + struct au_wbr_create_operations *si_wbr_create_ops;
27870 + /* round robin */
27871 + atomic_t si_wbr_rr_next;
27873 + /* most free space */
27874 + struct au_wbr_mfs si_wbr_mfs;
27876 + /* File-based Hierarchical Storage Management */
27877 + struct au_fhsm si_fhsm;
27879 + /* mount flags */
27880 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
27881 + unsigned int si_mntflags;
27883 + /* symlink to follow_link() and put_link() */
27884 + struct au_sphlhead si_symlink;
27886 + /* external inode number (bitmap and translation table) */
27887 + vfs_readf_t si_xread;
27888 + vfs_writef_t si_xwrite;
27889 + struct file *si_xib;
27890 + struct mutex si_xib_mtx; /* protect xib members */
27891 + unsigned long *si_xib_buf;
27892 + unsigned long si_xib_last_pindex;
27893 + int si_xib_next_bit;
27894 + aufs_bindex_t si_xino_brid;
27895 + unsigned long si_xino_jiffy;
27896 + unsigned long si_xino_expire;
27897 + /* reserved for future use */
27898 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
27900 +#ifdef CONFIG_AUFS_EXPORT
27901 + /* i_generation */
27902 + struct file *si_xigen;
27903 + atomic_t si_xigen_next;
27906 + /* dirty trick to suppoer atomic_open */
27907 + struct au_sphlhead si_aopen;
27909 + /* vdir parameters */
27910 + unsigned long si_rdcache; /* max cache time in jiffies */
27911 + unsigned int si_rdblk; /* deblk size */
27912 + unsigned int si_rdhash; /* hash size */
27915 + * If the number of whiteouts are larger than si_dirwh, leave all of
27916 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
27917 + * future fsck.aufs or kernel thread will remove them later.
27918 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
27920 + unsigned int si_dirwh;
27922 + /* pseudo_link list */
27923 + struct au_sphlhead si_plink[AuPlink_NHASH];
27924 + wait_queue_head_t si_plink_wq;
27925 + spinlock_t si_plink_maint_lock;
27926 + pid_t si_plink_maint_pid;
27929 + struct au_sphlhead si_files;
27931 + /* with/without getattr, brother of sb->s_d_op */
27932 + struct inode_operations *si_iop_array;
27935 + * sysfs and lifetime management.
27936 + * this is not a small structure and it may be a waste of memory in case
27937 + * of sysfs is disabled, particulary when many aufs-es are mounted.
27938 + * but using sysfs is majority.
27940 + struct kobject si_kobj;
27941 +#ifdef CONFIG_DEBUG_FS
27942 + struct dentry *si_dbgaufs;
27943 + struct dentry *si_dbgaufs_plink;
27944 + struct dentry *si_dbgaufs_xib;
27945 +#ifdef CONFIG_AUFS_EXPORT
27946 + struct dentry *si_dbgaufs_xigen;
27950 +#ifdef CONFIG_AUFS_SBILIST
27951 + struct list_head si_list;
27954 + /* dirty, necessary for unmounting, sysfs and sysrq */
27955 + struct super_block *si_sb;
27958 +/* sbinfo status flags */
27960 + * set true when refresh_dirs() failed at remount time.
27961 + * then try refreshing dirs at access time again.
27962 + * if it is false, refreshing dirs at access time is unnecesary
27964 +#define AuSi_FAILED_REFRESH_DIR 1
27965 +#define AuSi_FHSM (1 << 1) /* fhsm is active now */
27966 +#define AuSi_NO_DREVAL (1 << 2) /* disable all d_revalidate */
27968 +#ifndef CONFIG_AUFS_FHSM
27970 +#define AuSi_FHSM 0
27973 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
27974 + unsigned int flag)
27976 + AuRwMustAnyLock(&sbi->si_rwsem);
27977 + return sbi->au_si_status & flag;
27979 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
27980 +#define au_fset_si(sbinfo, name) do { \
27981 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
27982 + (sbinfo)->au_si_status |= AuSi_##name; \
27984 +#define au_fclr_si(sbinfo, name) do { \
27985 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
27986 + (sbinfo)->au_si_status &= ~AuSi_##name; \
27989 +/* ---------------------------------------------------------------------- */
27991 +/* policy to select one among writable branches */
27992 +#define AuWbrCopyup(sbinfo, ...) \
27993 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
27994 +#define AuWbrCreate(sbinfo, ...) \
27995 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
27997 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
27998 +#define AuLock_DW 1 /* write-lock dentry */
27999 +#define AuLock_IR (1 << 1) /* read-lock inode */
28000 +#define AuLock_IW (1 << 2) /* write-lock inode */
28001 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
28002 +#define AuLock_DIRS (1 << 4) /* target is a pair of dirs */
28003 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
28004 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
28005 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
28006 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
28007 +#define au_fset_lock(flags, name) \
28008 + do { (flags) |= AuLock_##name; } while (0)
28009 +#define au_fclr_lock(flags, name) \
28010 + do { (flags) &= ~AuLock_##name; } while (0)
28012 +/* ---------------------------------------------------------------------- */
28015 +extern struct file_system_type aufs_fs_type;
28016 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
28017 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
28018 + unsigned long long max, void *arg);
28019 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
28020 + struct super_block *sb, void *arg);
28021 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
28022 +void au_iarray_free(struct inode **a, unsigned long long max);
28025 +void au_si_free(struct kobject *kobj);
28026 +int au_si_alloc(struct super_block *sb);
28027 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr);
28029 +unsigned int au_sigen_inc(struct super_block *sb);
28030 +aufs_bindex_t au_new_br_id(struct super_block *sb);
28032 +int si_read_lock(struct super_block *sb, int flags);
28033 +int si_write_lock(struct super_block *sb, int flags);
28034 +int aufs_read_lock(struct dentry *dentry, int flags);
28035 +void aufs_read_unlock(struct dentry *dentry, int flags);
28036 +void aufs_write_lock(struct dentry *dentry);
28037 +void aufs_write_unlock(struct dentry *dentry);
28038 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
28039 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
28041 +int si_pid_test_slow(struct super_block *sb);
28042 +void si_pid_set_slow(struct super_block *sb);
28043 +void si_pid_clr_slow(struct super_block *sb);
28045 +/* wbr_policy.c */
28046 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
28047 +extern struct au_wbr_create_operations au_wbr_create_ops[];
28048 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
28049 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
28050 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t bstart);
28053 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
28055 +#ifdef CONFIG_AUFS_FHSM
28058 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
28062 + spin_lock(&fhsm->fhsm_spin);
28063 + pid = fhsm->fhsm_pid;
28064 + spin_unlock(&fhsm->fhsm_spin);
28069 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
28070 +void au_fhsm_wrote_all(struct super_block *sb, int force);
28071 +int au_fhsm_fd(struct super_block *sb, int oflags);
28072 +int au_fhsm_br_alloc(struct au_branch *br);
28073 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
28074 +void au_fhsm_fin(struct super_block *sb);
28075 +void au_fhsm_init(struct au_sbinfo *sbinfo);
28076 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
28077 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
28079 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
28081 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
28082 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
28083 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
28084 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
28085 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
28086 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
28087 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
28088 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
28089 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
28092 +/* ---------------------------------------------------------------------- */
28094 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
28096 + return sb->s_fs_info;
28099 +/* ---------------------------------------------------------------------- */
28101 +#ifdef CONFIG_AUFS_EXPORT
28102 +int au_test_nfsd(void);
28103 +void au_export_init(struct super_block *sb);
28104 +void au_xigen_inc(struct inode *inode);
28105 +int au_xigen_new(struct inode *inode);
28106 +int au_xigen_set(struct super_block *sb, struct file *base);
28107 +void au_xigen_clr(struct super_block *sb);
28109 +static inline int au_busy_or_stale(void)
28111 + if (!au_test_nfsd())
28116 +AuStubInt0(au_test_nfsd, void)
28117 +AuStubVoid(au_export_init, struct super_block *sb)
28118 +AuStubVoid(au_xigen_inc, struct inode *inode)
28119 +AuStubInt0(au_xigen_new, struct inode *inode)
28120 +AuStubInt0(au_xigen_set, struct super_block *sb, struct file *base)
28121 +AuStubVoid(au_xigen_clr, struct super_block *sb)
28122 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
28123 +#endif /* CONFIG_AUFS_EXPORT */
28125 +/* ---------------------------------------------------------------------- */
28127 +#ifdef CONFIG_AUFS_SBILIST
28129 +extern struct au_splhead au_sbilist;
28131 +static inline void au_sbilist_init(void)
28133 + au_spl_init(&au_sbilist);
28136 +static inline void au_sbilist_add(struct super_block *sb)
28138 + au_spl_add(&au_sbi(sb)->si_list, &au_sbilist);
28141 +static inline void au_sbilist_del(struct super_block *sb)
28143 + au_spl_del(&au_sbi(sb)->si_list, &au_sbilist);
28146 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
28147 +static inline void au_sbilist_lock(void)
28149 + spin_lock(&au_sbilist.spin);
28152 +static inline void au_sbilist_unlock(void)
28154 + spin_unlock(&au_sbilist.spin);
28156 +#define AuGFP_SBILIST GFP_ATOMIC
28158 +AuStubVoid(au_sbilist_lock, void)
28159 +AuStubVoid(au_sbilist_unlock, void)
28160 +#define AuGFP_SBILIST GFP_NOFS
28161 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
28163 +AuStubVoid(au_sbilist_init, void)
28164 +AuStubVoid(au_sbilist_add, struct super_block *sb)
28165 +AuStubVoid(au_sbilist_del, struct super_block *sb)
28166 +AuStubVoid(au_sbilist_lock, void)
28167 +AuStubVoid(au_sbilist_unlock, void)
28168 +#define AuGFP_SBILIST GFP_NOFS
28171 +/* ---------------------------------------------------------------------- */
28173 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
28176 + * This function is a dynamic '__init' function actually,
28177 + * so the tiny check for si_rwsem is unnecessary.
28179 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
28180 +#ifdef CONFIG_DEBUG_FS
28181 + sbinfo->si_dbgaufs = NULL;
28182 + sbinfo->si_dbgaufs_plink = NULL;
28183 + sbinfo->si_dbgaufs_xib = NULL;
28184 +#ifdef CONFIG_AUFS_EXPORT
28185 + sbinfo->si_dbgaufs_xigen = NULL;
28190 +/* ---------------------------------------------------------------------- */
28192 +static inline pid_t si_pid_bit(void)
28194 + /* the origin of pid is 1, but the bitmap's is 0 */
28195 + return current->pid - 1;
28198 +static inline int si_pid_test(struct super_block *sb)
28202 + bit = si_pid_bit();
28203 + if (bit < PID_MAX_DEFAULT)
28204 + return test_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28205 + return si_pid_test_slow(sb);
28208 +static inline void si_pid_set(struct super_block *sb)
28212 + bit = si_pid_bit();
28213 + if (bit < PID_MAX_DEFAULT) {
28214 + AuDebugOn(test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
28215 + set_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28218 + si_pid_set_slow(sb);
28221 +static inline void si_pid_clr(struct super_block *sb)
28225 + bit = si_pid_bit();
28226 + if (bit < PID_MAX_DEFAULT) {
28227 + AuDebugOn(!test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
28228 + clear_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28231 + si_pid_clr_slow(sb);
28234 +/* ---------------------------------------------------------------------- */
28236 +/* lock superblock. mainly for entry point functions */
28238 + * __si_read_lock, __si_write_lock,
28239 + * __si_read_unlock, __si_write_unlock, __si_downgrade_lock
28241 +AuSimpleRwsemFuncs(__si, struct super_block *sb, &au_sbi(sb)->si_rwsem);
28243 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
28244 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
28245 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
28247 +static inline void si_noflush_read_lock(struct super_block *sb)
28249 + __si_read_lock(sb);
28253 +static inline int si_noflush_read_trylock(struct super_block *sb)
28257 + locked = __si_read_trylock(sb);
28263 +static inline void si_noflush_write_lock(struct super_block *sb)
28265 + __si_write_lock(sb);
28269 +static inline int si_noflush_write_trylock(struct super_block *sb)
28273 + locked = __si_write_trylock(sb);
28279 +#if 0 /* reserved */
28280 +static inline int si_read_trylock(struct super_block *sb, int flags)
28282 + if (au_ftest_lock(flags, FLUSH))
28283 + au_nwt_flush(&au_sbi(sb)->si_nowait);
28284 + return si_noflush_read_trylock(sb);
28288 +static inline void si_read_unlock(struct super_block *sb)
28291 + __si_read_unlock(sb);
28294 +#if 0 /* reserved */
28295 +static inline int si_write_trylock(struct super_block *sb, int flags)
28297 + if (au_ftest_lock(flags, FLUSH))
28298 + au_nwt_flush(&au_sbi(sb)->si_nowait);
28299 + return si_noflush_write_trylock(sb);
28303 +static inline void si_write_unlock(struct super_block *sb)
28306 + __si_write_unlock(sb);
28309 +#if 0 /* reserved */
28310 +static inline void si_downgrade_lock(struct super_block *sb)
28312 + __si_downgrade_lock(sb);
28316 +/* ---------------------------------------------------------------------- */
28318 +static inline aufs_bindex_t au_sbend(struct super_block *sb)
28320 + SiMustAnyLock(sb);
28321 + return au_sbi(sb)->si_bend;
28324 +static inline unsigned int au_mntflags(struct super_block *sb)
28326 + SiMustAnyLock(sb);
28327 + return au_sbi(sb)->si_mntflags;
28330 +static inline unsigned int au_sigen(struct super_block *sb)
28332 + SiMustAnyLock(sb);
28333 + return au_sbi(sb)->si_generation;
28336 +static inline void au_ninodes_inc(struct super_block *sb)
28338 + atomic_long_inc(&au_sbi(sb)->si_ninodes);
28341 +static inline void au_ninodes_dec(struct super_block *sb)
28343 + AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_ninodes));
28344 + atomic_long_dec(&au_sbi(sb)->si_ninodes);
28347 +static inline void au_nfiles_inc(struct super_block *sb)
28349 + atomic_long_inc(&au_sbi(sb)->si_nfiles);
28352 +static inline void au_nfiles_dec(struct super_block *sb)
28354 + AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_nfiles));
28355 + atomic_long_dec(&au_sbi(sb)->si_nfiles);
28358 +static inline struct au_branch *au_sbr(struct super_block *sb,
28359 + aufs_bindex_t bindex)
28361 + SiMustAnyLock(sb);
28362 + return au_sbi(sb)->si_branch[0 + bindex];
28365 +static inline void au_xino_brid_set(struct super_block *sb, aufs_bindex_t brid)
28367 + SiMustWriteLock(sb);
28368 + au_sbi(sb)->si_xino_brid = brid;
28371 +static inline aufs_bindex_t au_xino_brid(struct super_block *sb)
28373 + SiMustAnyLock(sb);
28374 + return au_sbi(sb)->si_xino_brid;
28377 +#endif /* __KERNEL__ */
28378 +#endif /* __AUFS_SUPER_H__ */
28379 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
28380 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
28381 +++ linux/fs/aufs/sysaufs.c 2016-01-13 20:11:11.673093671 +0100
28384 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28386 + * This program, aufs is free software; you can redistribute it and/or modify
28387 + * it under the terms of the GNU General Public License as published by
28388 + * the Free Software Foundation; either version 2 of the License, or
28389 + * (at your option) any later version.
28391 + * This program is distributed in the hope that it will be useful,
28392 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28393 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28394 + * GNU General Public License for more details.
28396 + * You should have received a copy of the GNU General Public License
28397 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28401 + * sysfs interface and lifetime management
28402 + * they are necessary regardless sysfs is disabled.
28405 +#include <linux/random.h>
28408 +unsigned long sysaufs_si_mask;
28409 +struct kset *sysaufs_kset;
28411 +#define AuSiAttr(_name) { \
28412 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
28413 + .show = sysaufs_si_##_name, \
28416 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
28417 +struct attribute *sysaufs_si_attrs[] = {
28418 + &sysaufs_si_attr_xi_path.attr,
28422 +static const struct sysfs_ops au_sbi_ops = {
28423 + .show = sysaufs_si_show
28426 +static struct kobj_type au_sbi_ktype = {
28427 + .release = au_si_free,
28428 + .sysfs_ops = &au_sbi_ops,
28429 + .default_attrs = sysaufs_si_attrs
28432 +/* ---------------------------------------------------------------------- */
28434 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
28438 + sbinfo->si_kobj.kset = sysaufs_kset;
28439 + /* cf. sysaufs_name() */
28440 + err = kobject_init_and_add
28441 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
28442 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
28444 + dbgaufs_si_null(sbinfo);
28446 + err = dbgaufs_si_init(sbinfo);
28447 + if (unlikely(err))
28448 + kobject_put(&sbinfo->si_kobj);
28453 +void sysaufs_fin(void)
28456 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
28457 + kset_unregister(sysaufs_kset);
28460 +int __init sysaufs_init(void)
28465 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
28466 + } while (!sysaufs_si_mask);
28469 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
28470 + if (unlikely(!sysaufs_kset))
28472 + err = PTR_ERR(sysaufs_kset);
28473 + if (IS_ERR(sysaufs_kset))
28475 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
28476 + if (unlikely(err)) {
28477 + kset_unregister(sysaufs_kset);
28481 + err = dbgaufs_init();
28482 + if (unlikely(err))
28487 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
28488 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
28489 +++ linux/fs/aufs/sysaufs.h 2016-01-13 20:11:11.673093671 +0100
28492 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28494 + * This program, aufs is free software; you can redistribute it and/or modify
28495 + * it under the terms of the GNU General Public License as published by
28496 + * the Free Software Foundation; either version 2 of the License, or
28497 + * (at your option) any later version.
28499 + * This program is distributed in the hope that it will be useful,
28500 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28501 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28502 + * GNU General Public License for more details.
28504 + * You should have received a copy of the GNU General Public License
28505 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28509 + * sysfs interface and mount lifetime management
28512 +#ifndef __SYSAUFS_H__
28513 +#define __SYSAUFS_H__
28517 +#include <linux/sysfs.h>
28518 +#include "module.h"
28520 +struct super_block;
28523 +struct sysaufs_si_attr {
28524 + struct attribute attr;
28525 + int (*show)(struct seq_file *seq, struct super_block *sb);
28528 +/* ---------------------------------------------------------------------- */
28531 +extern unsigned long sysaufs_si_mask;
28532 +extern struct kset *sysaufs_kset;
28533 +extern struct attribute *sysaufs_si_attrs[];
28534 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
28535 +int __init sysaufs_init(void);
28536 +void sysaufs_fin(void);
28538 +/* ---------------------------------------------------------------------- */
28540 +/* some people doesn't like to show a pointer in kernel */
28541 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
28543 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
28546 +#define SysaufsSiNamePrefix "si_"
28547 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
28548 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
28550 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
28551 + sysaufs_si_id(sbinfo));
28555 +#ifdef CONFIG_SYSFS
28557 +extern struct attribute_group *sysaufs_attr_group;
28559 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
28560 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
28562 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
28563 +#ifdef CONFIG_COMPAT
28564 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
28567 +void sysaufs_br_init(struct au_branch *br);
28568 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
28569 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
28571 +#define sysaufs_brs_init() do {} while (0)
28574 +#define sysaufs_attr_group NULL
28576 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
28577 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
28578 + struct attribute *attr, char *buf)
28579 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
28580 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
28581 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
28583 +static inline void sysaufs_brs_init(void)
28588 +#endif /* CONFIG_SYSFS */
28590 +#endif /* __KERNEL__ */
28591 +#endif /* __SYSAUFS_H__ */
28592 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
28593 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
28594 +++ linux/fs/aufs/sysfs.c 2016-01-13 20:11:11.673093671 +0100
28597 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28599 + * This program, aufs is free software; you can redistribute it and/or modify
28600 + * it under the terms of the GNU General Public License as published by
28601 + * the Free Software Foundation; either version 2 of the License, or
28602 + * (at your option) any later version.
28604 + * This program is distributed in the hope that it will be useful,
28605 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28606 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28607 + * GNU General Public License for more details.
28609 + * You should have received a copy of the GNU General Public License
28610 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28614 + * sysfs interface
28617 +#include <linux/compat.h>
28618 +#include <linux/seq_file.h>
28621 +#ifdef CONFIG_AUFS_FS_MODULE
28622 +/* this entry violates the "one line per file" policy of sysfs */
28623 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
28627 + static char *conf =
28628 +/* this file is generated at compiling */
28629 +#include "conf.str"
28632 + err = snprintf(buf, PAGE_SIZE, conf);
28633 + if (unlikely(err >= PAGE_SIZE))
28638 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
28641 +static struct attribute *au_attr[] = {
28642 +#ifdef CONFIG_AUFS_FS_MODULE
28643 + &au_config_attr.attr,
28645 + NULL, /* need to NULL terminate the list of attributes */
28648 +static struct attribute_group sysaufs_attr_group_body = {
28652 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
28654 +/* ---------------------------------------------------------------------- */
28656 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
28660 + SiMustAnyLock(sb);
28663 + if (au_opt_test(au_mntflags(sb), XINO)) {
28664 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
28665 + seq_putc(seq, '\n');
28671 + * the lifetime of branch is independent from the entry under sysfs.
28672 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
28675 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
28676 + aufs_bindex_t bindex, int idx)
28679 + struct path path;
28680 + struct dentry *root;
28681 + struct au_branch *br;
28682 + au_br_perm_str_t perm;
28684 + AuDbg("b%d\n", bindex);
28687 + root = sb->s_root;
28688 + di_read_lock_parent(root, !AuLock_IR);
28689 + br = au_sbr(sb, bindex);
28692 + case AuBrSysfs_BR:
28693 + path.mnt = au_br_mnt(br);
28694 + path.dentry = au_h_dptr(root, bindex);
28695 + err = au_seq_path(seq, &path);
28697 + au_optstr_br_perm(&perm, br->br_perm);
28698 + seq_printf(seq, "=%s\n", perm.a);
28701 + case AuBrSysfs_BRID:
28702 + seq_printf(seq, "%d\n", br->br_id);
28705 + di_read_unlock(root, !AuLock_IR);
28706 + if (unlikely(err || seq_has_overflowed(seq)))
28712 +/* ---------------------------------------------------------------------- */
28714 +static struct seq_file *au_seq(char *p, ssize_t len)
28716 + struct seq_file *seq;
28718 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
28720 + /* mutex_init(&seq.lock); */
28723 + return seq; /* success */
28726 + seq = ERR_PTR(-ENOMEM);
28730 +#define SysaufsBr_PREFIX "br"
28731 +#define SysaufsBrid_PREFIX "brid"
28733 +/* todo: file size may exceed PAGE_SIZE */
28734 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
28740 + aufs_bindex_t bend;
28741 + struct au_sbinfo *sbinfo;
28742 + struct super_block *sb;
28743 + struct seq_file *seq;
28745 + struct attribute **cattr;
28747 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
28748 + sb = sbinfo->si_sb;
28751 + * prevent a race condition between sysfs and aufs.
28752 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
28753 + * prohibits maintaining the sysfs entries.
28754 + * hew we acquire read lock after sysfs_get_active_two().
28755 + * on the other hand, the remount process may maintain the sysfs/aufs
28756 + * entries after acquiring write lock.
28757 + * it can cause a deadlock.
28758 + * simply we gave up processing read here.
28761 + if (unlikely(!si_noflush_read_trylock(sb)))
28764 + seq = au_seq(buf, PAGE_SIZE);
28765 + err = PTR_ERR(seq);
28769 + name = (void *)attr->name;
28770 + cattr = sysaufs_si_attrs;
28772 + if (!strcmp(name, (*cattr)->name)) {
28773 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
28780 + if (!strncmp(name, SysaufsBrid_PREFIX,
28781 + sizeof(SysaufsBrid_PREFIX) - 1)) {
28782 + idx = AuBrSysfs_BRID;
28783 + name += sizeof(SysaufsBrid_PREFIX) - 1;
28784 + } else if (!strncmp(name, SysaufsBr_PREFIX,
28785 + sizeof(SysaufsBr_PREFIX) - 1)) {
28786 + idx = AuBrSysfs_BR;
28787 + name += sizeof(SysaufsBr_PREFIX) - 1;
28791 + err = kstrtol(name, 10, &l);
28793 + bend = au_sbend(sb);
28795 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
28802 + err = seq->count;
28803 + /* sysfs limit */
28804 + if (unlikely(err == PAGE_SIZE))
28809 + si_read_unlock(sb);
28814 +/* ---------------------------------------------------------------------- */
28816 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
28820 + aufs_bindex_t bindex, bend;
28823 + struct seq_file *seq;
28824 + struct au_branch *br;
28826 + si_read_lock(sb, AuLock_FLUSH);
28827 + bend = au_sbend(sb);
28833 + buf = (void *)__get_free_page(GFP_NOFS);
28834 + if (unlikely(!buf))
28837 + seq = au_seq(buf, PAGE_SIZE);
28838 + err = PTR_ERR(seq);
28842 + sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
28843 + for (bindex = 0; bindex <= bend; bindex++, arg++) {
28844 + err = !access_ok(VERIFY_WRITE, arg, sizeof(*arg));
28845 + if (unlikely(err))
28848 + br = au_sbr(sb, bindex);
28849 + brid = br->br_id;
28850 + BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
28851 + err = __put_user(brid, &arg->id);
28852 + if (unlikely(err))
28855 + BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
28856 + err = __put_user(br->br_perm, &arg->perm);
28857 + if (unlikely(err))
28860 + err = au_seq_path(seq, &br->br_path);
28861 + if (unlikely(err))
28863 + seq_putc(seq, '\0');
28864 + if (!seq_has_overflowed(seq)) {
28865 + err = copy_to_user(arg->path, seq->buf, seq->count);
28867 + if (unlikely(err))
28874 + if (unlikely(err))
28880 + free_page((unsigned long)buf);
28882 + si_read_unlock(sb);
28886 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
28888 + return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
28891 +#ifdef CONFIG_COMPAT
28892 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
28894 + return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
28898 +/* ---------------------------------------------------------------------- */
28900 +void sysaufs_br_init(struct au_branch *br)
28903 + struct au_brsysfs *br_sysfs;
28904 + struct attribute *attr;
28906 + br_sysfs = br->br_sysfs;
28907 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28908 + attr = &br_sysfs->attr;
28909 + sysfs_attr_init(attr);
28910 + attr->name = br_sysfs->name;
28911 + attr->mode = S_IRUGO;
28916 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
28918 + struct au_branch *br;
28919 + struct kobject *kobj;
28920 + struct au_brsysfs *br_sysfs;
28922 + aufs_bindex_t bend;
28924 + dbgaufs_brs_del(sb, bindex);
28926 + if (!sysaufs_brs)
28929 + kobj = &au_sbi(sb)->si_kobj;
28930 + bend = au_sbend(sb);
28931 + for (; bindex <= bend; bindex++) {
28932 + br = au_sbr(sb, bindex);
28933 + br_sysfs = br->br_sysfs;
28934 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28935 + sysfs_remove_file(kobj, &br_sysfs->attr);
28941 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
28944 + aufs_bindex_t bend;
28945 + struct kobject *kobj;
28946 + struct au_branch *br;
28947 + struct au_brsysfs *br_sysfs;
28949 + dbgaufs_brs_add(sb, bindex);
28951 + if (!sysaufs_brs)
28954 + kobj = &au_sbi(sb)->si_kobj;
28955 + bend = au_sbend(sb);
28956 + for (; bindex <= bend; bindex++) {
28957 + br = au_sbr(sb, bindex);
28958 + br_sysfs = br->br_sysfs;
28959 + snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
28960 + SysaufsBr_PREFIX "%d", bindex);
28961 + snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
28962 + SysaufsBrid_PREFIX "%d", bindex);
28963 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28964 + err = sysfs_create_file(kobj, &br_sysfs->attr);
28965 + if (unlikely(err))
28966 + pr_warn("failed %s under sysfs(%d)\n",
28967 + br_sysfs->name, err);
28972 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
28973 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
28974 +++ linux/fs/aufs/sysrq.c 2016-01-13 20:11:11.673093671 +0100
28977 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28979 + * This program, aufs is free software; you can redistribute it and/or modify
28980 + * it under the terms of the GNU General Public License as published by
28981 + * the Free Software Foundation; either version 2 of the License, or
28982 + * (at your option) any later version.
28984 + * This program is distributed in the hope that it will be useful,
28985 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28986 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28987 + * GNU General Public License for more details.
28989 + * You should have received a copy of the GNU General Public License
28990 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28994 + * magic sysrq hanlder
28997 +/* #include <linux/sysrq.h> */
28998 +#include <linux/writeback.h>
29001 +/* ---------------------------------------------------------------------- */
29003 +static void sysrq_sb(struct super_block *sb)
29006 + struct au_sbinfo *sbinfo;
29007 + struct file *file;
29008 + struct au_sphlhead *files;
29009 + struct au_finfo *finfo;
29011 + plevel = au_plevel;
29012 + au_plevel = KERN_WARNING;
29014 + /* since we define pr_fmt, call printk directly */
29015 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
29017 + sbinfo = au_sbi(sb);
29018 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
29019 + pr("superblock\n");
29023 + pr("root dentry\n");
29024 + au_dpri_dentry(sb->s_root);
29025 + pr("root inode\n");
29026 + au_dpri_inode(d_inode(sb->s_root));
29031 + int err, i, j, ndentry;
29032 + struct au_dcsub_pages dpages;
29033 + struct au_dpage *dpage;
29035 + err = au_dpages_init(&dpages, GFP_ATOMIC);
29036 + if (unlikely(err))
29038 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
29040 + for (i = 0; i < dpages.ndpage; i++) {
29041 + dpage = dpages.dpages + i;
29042 + ndentry = dpage->ndentry;
29043 + for (j = 0; j < ndentry; j++)
29044 + au_dpri_dentry(dpage->dentries[j]);
29046 + au_dpages_free(&dpages);
29054 + pr("isolated inode\n");
29055 + spin_lock(&sb->s_inode_list_lock);
29056 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
29057 + spin_lock(&i->i_lock);
29058 + if (1 || hlist_empty(&i->i_dentry))
29059 + au_dpri_inode(i);
29060 + spin_unlock(&i->i_lock);
29062 + spin_unlock(&sb->s_inode_list_lock);
29066 + files = &au_sbi(sb)->si_files;
29067 + spin_lock(&files->spin);
29068 + hlist_for_each_entry(finfo, &files->head, fi_hlist) {
29071 + file = finfo->fi_file;
29072 + mode = file_inode(file)->i_mode;
29073 + if (!special_file(mode))
29074 + au_dpri_file(file);
29076 + spin_unlock(&files->spin);
29080 + au_plevel = plevel;
29083 +/* ---------------------------------------------------------------------- */
29085 +/* module parameter */
29086 +static char *aufs_sysrq_key = "a";
29087 +module_param_named(sysrq, aufs_sysrq_key, charp, S_IRUGO);
29088 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
29090 +static void au_sysrq(int key __maybe_unused)
29092 + struct au_sbinfo *sbinfo;
29095 + au_sbilist_lock();
29096 + list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
29097 + sysrq_sb(sbinfo->si_sb);
29098 + au_sbilist_unlock();
29102 +static struct sysrq_key_op au_sysrq_op = {
29103 + .handler = au_sysrq,
29104 + .help_msg = "Aufs",
29105 + .action_msg = "Aufs",
29106 + .enable_mask = SYSRQ_ENABLE_DUMP
29109 +/* ---------------------------------------------------------------------- */
29111 +int __init au_sysrq_init(void)
29117 + key = *aufs_sysrq_key;
29118 + if ('a' <= key && key <= 'z')
29119 + err = register_sysrq_key(key, &au_sysrq_op);
29120 + if (unlikely(err))
29121 + pr_err("err %d, sysrq=%c\n", err, key);
29125 +void au_sysrq_fin(void)
29129 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
29130 + if (unlikely(err))
29131 + pr_err("err %d (ignored)\n", err);
29133 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
29134 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
29135 +++ linux/fs/aufs/vdir.c 2016-01-13 20:11:11.673093671 +0100
29138 + * Copyright (C) 2005-2015 Junjiro R. Okajima
29140 + * This program, aufs is free software; you can redistribute it and/or modify
29141 + * it under the terms of the GNU General Public License as published by
29142 + * the Free Software Foundation; either version 2 of the License, or
29143 + * (at your option) any later version.
29145 + * This program is distributed in the hope that it will be useful,
29146 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29147 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29148 + * GNU General Public License for more details.
29150 + * You should have received a copy of the GNU General Public License
29151 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29155 + * virtual or vertical directory
29160 +static unsigned int calc_size(int nlen)
29162 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
29165 +static int set_deblk_end(union au_vdir_deblk_p *p,
29166 + union au_vdir_deblk_p *deblk_end)
29168 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
29169 + p->de->de_str.len = 0;
29173 + return -1; /* error */
29176 +/* returns true or false */
29177 +static int is_deblk_end(union au_vdir_deblk_p *p,
29178 + union au_vdir_deblk_p *deblk_end)
29180 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
29181 + return !p->de->de_str.len;
29185 +static unsigned char *last_deblk(struct au_vdir *vdir)
29187 + return vdir->vd_deblk[vdir->vd_nblk - 1];
29190 +/* ---------------------------------------------------------------------- */
29192 +/* estimate the appropriate size for name hash table */
29193 +unsigned int au_rdhash_est(loff_t sz)
29201 + if (sz < AUFS_RDHASH_DEF)
29202 + n = AUFS_RDHASH_DEF;
29203 + /* pr_info("n %u\n", n); */
29208 + * the allocated memory has to be freed by
29209 + * au_nhash_wh_free() or au_nhash_de_free().
29211 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
29213 + struct hlist_head *head;
29217 + sz = sizeof(*nhash->nh_head) * num_hash;
29218 + head = kmalloc(sz, gfp);
29220 + nhash->nh_num = num_hash;
29221 + nhash->nh_head = head;
29222 + for (u = 0; u < num_hash; u++)
29223 + INIT_HLIST_HEAD(head++);
29224 + return 0; /* success */
29230 +static void nhash_count(struct hlist_head *head)
29234 + struct hlist_node *pos;
29237 + hlist_for_each(pos, head)
29239 + pr_info("%lu\n", n);
29243 +static void au_nhash_wh_do_free(struct hlist_head *head)
29245 + struct au_vdir_wh *pos;
29246 + struct hlist_node *node;
29248 + hlist_for_each_entry_safe(pos, node, head, wh_hash)
29252 +static void au_nhash_de_do_free(struct hlist_head *head)
29254 + struct au_vdir_dehstr *pos;
29255 + struct hlist_node *node;
29257 + hlist_for_each_entry_safe(pos, node, head, hash)
29258 + au_cache_free_vdir_dehstr(pos);
29261 +static void au_nhash_do_free(struct au_nhash *nhash,
29262 + void (*free)(struct hlist_head *head))
29265 + struct hlist_head *head;
29267 + n = nhash->nh_num;
29271 + head = nhash->nh_head;
29272 + while (n-- > 0) {
29273 + nhash_count(head);
29276 + kfree(nhash->nh_head);
29279 +void au_nhash_wh_free(struct au_nhash *whlist)
29281 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
29284 +static void au_nhash_de_free(struct au_nhash *delist)
29286 + au_nhash_do_free(delist, au_nhash_de_do_free);
29289 +/* ---------------------------------------------------------------------- */
29291 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
29295 + unsigned int u, n;
29296 + struct hlist_head *head;
29297 + struct au_vdir_wh *pos;
29300 + n = whlist->nh_num;
29301 + head = whlist->nh_head;
29302 + for (u = 0; u < n; u++, head++)
29303 + hlist_for_each_entry(pos, head, wh_hash)
29304 + if (pos->wh_bindex == btgt && ++num > limit)
29309 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
29310 + unsigned char *name,
29311 + unsigned int len)
29314 + /* const unsigned int magic_bit = 12; */
29316 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
29321 + /* v = hash_long(v, magic_bit); */
29322 + v %= nhash->nh_num;
29323 + return nhash->nh_head + v;
29326 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
29329 + return str->len == nlen && !memcmp(str->name, name, nlen);
29332 +/* returns found or not */
29333 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
29335 + struct hlist_head *head;
29336 + struct au_vdir_wh *pos;
29337 + struct au_vdir_destr *str;
29339 + head = au_name_hash(whlist, name, nlen);
29340 + hlist_for_each_entry(pos, head, wh_hash) {
29341 + str = &pos->wh_str;
29342 + AuDbg("%.*s\n", str->len, str->name);
29343 + if (au_nhash_test_name(str, name, nlen))
29349 +/* returns found(true) or not */
29350 +static int test_known(struct au_nhash *delist, char *name, int nlen)
29352 + struct hlist_head *head;
29353 + struct au_vdir_dehstr *pos;
29354 + struct au_vdir_destr *str;
29356 + head = au_name_hash(delist, name, nlen);
29357 + hlist_for_each_entry(pos, head, hash) {
29359 + AuDbg("%.*s\n", str->len, str->name);
29360 + if (au_nhash_test_name(str, name, nlen))
29366 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
29367 + unsigned char d_type)
29369 +#ifdef CONFIG_AUFS_SHWH
29370 + wh->wh_ino = ino;
29371 + wh->wh_type = d_type;
29375 +/* ---------------------------------------------------------------------- */
29377 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
29378 + unsigned int d_type, aufs_bindex_t bindex,
29379 + unsigned char shwh)
29382 + struct au_vdir_destr *str;
29383 + struct au_vdir_wh *wh;
29385 + AuDbg("%.*s\n", nlen, name);
29386 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
29389 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
29390 + if (unlikely(!wh))
29394 + wh->wh_bindex = bindex;
29396 + au_shwh_init_wh(wh, ino, d_type);
29397 + str = &wh->wh_str;
29399 + memcpy(str->name, name, nlen);
29400 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
29407 +static int append_deblk(struct au_vdir *vdir)
29410 + unsigned long ul;
29411 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
29412 + union au_vdir_deblk_p p, deblk_end;
29413 + unsigned char **o;
29416 + o = krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
29418 + if (unlikely(!o))
29421 + vdir->vd_deblk = o;
29422 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
29424 + ul = vdir->vd_nblk++;
29425 + vdir->vd_deblk[ul] = p.deblk;
29426 + vdir->vd_last.ul = ul;
29427 + vdir->vd_last.p.deblk = p.deblk;
29428 + deblk_end.deblk = p.deblk + deblk_sz;
29429 + err = set_deblk_end(&p, &deblk_end);
29436 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
29437 + unsigned int d_type, struct au_nhash *delist)
29441 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
29442 + union au_vdir_deblk_p p, *room, deblk_end;
29443 + struct au_vdir_dehstr *dehstr;
29445 + p.deblk = last_deblk(vdir);
29446 + deblk_end.deblk = p.deblk + deblk_sz;
29447 + room = &vdir->vd_last.p;
29448 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
29449 + || !is_deblk_end(room, &deblk_end));
29451 + sz = calc_size(nlen);
29452 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
29453 + err = append_deblk(vdir);
29454 + if (unlikely(err))
29457 + p.deblk = last_deblk(vdir);
29458 + deblk_end.deblk = p.deblk + deblk_sz;
29460 + AuDebugOn(room->deblk != p.deblk);
29464 + dehstr = au_cache_alloc_vdir_dehstr();
29465 + if (unlikely(!dehstr))
29468 + dehstr->str = &room->de->de_str;
29469 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
29470 + room->de->de_ino = ino;
29471 + room->de->de_type = d_type;
29472 + room->de->de_str.len = nlen;
29473 + memcpy(room->de->de_str.name, name, nlen);
29476 + room->deblk += sz;
29477 + if (unlikely(set_deblk_end(room, &deblk_end)))
29478 + err = append_deblk(vdir);
29485 +/* ---------------------------------------------------------------------- */
29487 +void au_vdir_free(struct au_vdir *vdir)
29489 + unsigned char **deblk;
29491 + deblk = vdir->vd_deblk;
29492 + while (vdir->vd_nblk--)
29494 + kfree(vdir->vd_deblk);
29495 + au_cache_free_vdir(vdir);
29498 +static struct au_vdir *alloc_vdir(struct file *file)
29500 + struct au_vdir *vdir;
29501 + struct super_block *sb;
29504 + sb = file->f_path.dentry->d_sb;
29505 + SiMustAnyLock(sb);
29508 + vdir = au_cache_alloc_vdir();
29509 + if (unlikely(!vdir))
29512 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
29513 + if (unlikely(!vdir->vd_deblk))
29516 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
29517 + if (!vdir->vd_deblk_sz) {
29518 + /* estimate the appropriate size for deblk */
29519 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
29520 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
29522 + vdir->vd_nblk = 0;
29523 + vdir->vd_version = 0;
29524 + vdir->vd_jiffy = 0;
29525 + err = append_deblk(vdir);
29527 + return vdir; /* success */
29529 + kfree(vdir->vd_deblk);
29532 + au_cache_free_vdir(vdir);
29534 + vdir = ERR_PTR(err);
29538 +static int reinit_vdir(struct au_vdir *vdir)
29541 + union au_vdir_deblk_p p, deblk_end;
29543 + while (vdir->vd_nblk > 1) {
29544 + kfree(vdir->vd_deblk[vdir->vd_nblk - 1]);
29545 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
29548 + p.deblk = vdir->vd_deblk[0];
29549 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
29550 + err = set_deblk_end(&p, &deblk_end);
29551 + /* keep vd_dblk_sz */
29552 + vdir->vd_last.ul = 0;
29553 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
29554 + vdir->vd_version = 0;
29555 + vdir->vd_jiffy = 0;
29560 +/* ---------------------------------------------------------------------- */
29562 +#define AuFillVdir_CALLED 1
29563 +#define AuFillVdir_WHABLE (1 << 1)
29564 +#define AuFillVdir_SHWH (1 << 2)
29565 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
29566 +#define au_fset_fillvdir(flags, name) \
29567 + do { (flags) |= AuFillVdir_##name; } while (0)
29568 +#define au_fclr_fillvdir(flags, name) \
29569 + do { (flags) &= ~AuFillVdir_##name; } while (0)
29571 +#ifndef CONFIG_AUFS_SHWH
29572 +#undef AuFillVdir_SHWH
29573 +#define AuFillVdir_SHWH 0
29576 +struct fillvdir_arg {
29577 + struct dir_context ctx;
29578 + struct file *file;
29579 + struct au_vdir *vdir;
29580 + struct au_nhash delist;
29581 + struct au_nhash whlist;
29582 + aufs_bindex_t bindex;
29583 + unsigned int flags;
29587 +static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
29588 + loff_t offset __maybe_unused, u64 h_ino,
29589 + unsigned int d_type)
29591 + struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
29592 + char *name = (void *)__name;
29593 + struct super_block *sb;
29595 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
29598 + sb = arg->file->f_path.dentry->d_sb;
29599 + au_fset_fillvdir(arg->flags, CALLED);
29601 + if (nlen <= AUFS_WH_PFX_LEN
29602 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
29603 + if (test_known(&arg->delist, name, nlen)
29604 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
29605 + goto out; /* already exists or whiteouted */
29607 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
29609 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
29610 + d_type = DT_UNKNOWN;
29611 + arg->err = append_de(arg->vdir, name, nlen, ino,
29612 + d_type, &arg->delist);
29614 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
29615 + name += AUFS_WH_PFX_LEN;
29616 + nlen -= AUFS_WH_PFX_LEN;
29617 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
29618 + goto out; /* already whiteouted */
29621 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
29624 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
29625 + d_type = DT_UNKNOWN;
29626 + arg->err = au_nhash_append_wh
29627 + (&arg->whlist, name, nlen, ino, d_type,
29628 + arg->bindex, shwh);
29634 + arg->vdir->vd_jiffy = jiffies;
29636 + AuTraceErr(arg->err);
29640 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
29641 + struct au_nhash *whlist, struct au_nhash *delist)
29643 +#ifdef CONFIG_AUFS_SHWH
29645 + unsigned int nh, u;
29646 + struct hlist_head *head;
29647 + struct au_vdir_wh *pos;
29648 + struct hlist_node *n;
29650 + struct au_vdir_destr *destr;
29652 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
29655 + o = p = (void *)__get_free_page(GFP_NOFS);
29656 + if (unlikely(!p))
29660 + nh = whlist->nh_num;
29661 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
29662 + p += AUFS_WH_PFX_LEN;
29663 + for (u = 0; u < nh; u++) {
29664 + head = whlist->nh_head + u;
29665 + hlist_for_each_entry_safe(pos, n, head, wh_hash) {
29666 + destr = &pos->wh_str;
29667 + memcpy(p, destr->name, destr->len);
29668 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
29669 + pos->wh_ino, pos->wh_type, delist);
29670 + if (unlikely(err))
29675 + free_page((unsigned long)o);
29685 +static int au_do_read_vdir(struct fillvdir_arg *arg)
29688 + unsigned int rdhash;
29690 + aufs_bindex_t bend, bindex, bstart;
29691 + unsigned char shwh;
29692 + struct file *hf, *file;
29693 + struct super_block *sb;
29695 + file = arg->file;
29696 + sb = file->f_path.dentry->d_sb;
29697 + SiMustAnyLock(sb);
29699 + rdhash = au_sbi(sb)->si_rdhash;
29701 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
29702 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
29703 + if (unlikely(err))
29705 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
29706 + if (unlikely(err))
29712 + if (au_opt_test(au_mntflags(sb), SHWH)) {
29714 + au_fset_fillvdir(arg->flags, SHWH);
29716 + bstart = au_fbstart(file);
29717 + bend = au_fbend_dir(file);
29718 + for (bindex = bstart; !err && bindex <= bend; bindex++) {
29719 + hf = au_hf_dir(file, bindex);
29723 + offset = vfsub_llseek(hf, 0, SEEK_SET);
29725 + if (unlikely(offset))
29728 + arg->bindex = bindex;
29729 + au_fclr_fillvdir(arg->flags, WHABLE);
29731 + || (bindex != bend
29732 + && au_br_whable(au_sbr_perm(sb, bindex))))
29733 + au_fset_fillvdir(arg->flags, WHABLE);
29736 + au_fclr_fillvdir(arg->flags, CALLED);
29738 + err = vfsub_iterate_dir(hf, &arg->ctx);
29741 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
29744 + * dir_relax() may be good for concurrency, but aufs should not
29745 + * use it since it will cause a lockdep problem.
29749 + if (!err && shwh)
29750 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
29752 + au_nhash_wh_free(&arg->whlist);
29755 + au_nhash_de_free(&arg->delist);
29760 +static int read_vdir(struct file *file, int may_read)
29763 + unsigned long expire;
29764 + unsigned char do_read;
29765 + struct fillvdir_arg arg = {
29767 + .actor = fillvdir
29770 + struct inode *inode;
29771 + struct au_vdir *vdir, *allocated;
29774 + inode = file_inode(file);
29775 + IMustLock(inode);
29776 + SiMustAnyLock(inode->i_sb);
29778 + allocated = NULL;
29780 + expire = au_sbi(inode->i_sb)->si_rdcache;
29781 + vdir = au_ivdir(inode);
29784 + vdir = alloc_vdir(file);
29785 + err = PTR_ERR(vdir);
29786 + if (IS_ERR(vdir))
29789 + allocated = vdir;
29790 + } else if (may_read
29791 + && (inode->i_version != vdir->vd_version
29792 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
29794 + err = reinit_vdir(vdir);
29795 + if (unlikely(err))
29800 + return 0; /* success */
29804 + err = au_do_read_vdir(&arg);
29806 + /* file->f_pos = 0; */ /* todo: ctx->pos? */
29807 + vdir->vd_version = inode->i_version;
29808 + vdir->vd_last.ul = 0;
29809 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
29811 + au_set_ivdir(inode, allocated);
29812 + } else if (allocated)
29813 + au_vdir_free(allocated);
29819 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
29822 + unsigned long ul, n;
29823 + const unsigned int deblk_sz = src->vd_deblk_sz;
29825 + AuDebugOn(tgt->vd_nblk != 1);
29828 + if (tgt->vd_nblk < src->vd_nblk) {
29829 + unsigned char **p;
29831 + p = krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
29833 + if (unlikely(!p))
29835 + tgt->vd_deblk = p;
29838 + if (tgt->vd_deblk_sz != deblk_sz) {
29839 + unsigned char *p;
29841 + tgt->vd_deblk_sz = deblk_sz;
29842 + p = krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS);
29843 + if (unlikely(!p))
29845 + tgt->vd_deblk[0] = p;
29847 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
29848 + tgt->vd_version = src->vd_version;
29849 + tgt->vd_jiffy = src->vd_jiffy;
29851 + n = src->vd_nblk;
29852 + for (ul = 1; ul < n; ul++) {
29853 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
29855 + if (unlikely(!tgt->vd_deblk[ul]))
29859 + tgt->vd_nblk = n;
29860 + tgt->vd_last.ul = tgt->vd_last.ul;
29861 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
29862 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
29863 + - src->vd_deblk[src->vd_last.ul];
29865 + return 0; /* success */
29868 + rerr = reinit_vdir(tgt);
29873 +int au_vdir_init(struct file *file)
29876 + struct inode *inode;
29877 + struct au_vdir *vdir_cache, *allocated;
29879 + /* test file->f_pos here instead of ctx->pos */
29880 + err = read_vdir(file, !file->f_pos);
29881 + if (unlikely(err))
29884 + allocated = NULL;
29885 + vdir_cache = au_fvdir_cache(file);
29886 + if (!vdir_cache) {
29887 + vdir_cache = alloc_vdir(file);
29888 + err = PTR_ERR(vdir_cache);
29889 + if (IS_ERR(vdir_cache))
29891 + allocated = vdir_cache;
29892 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
29893 + /* test file->f_pos here instead of ctx->pos */
29894 + err = reinit_vdir(vdir_cache);
29895 + if (unlikely(err))
29898 + return 0; /* success */
29900 + inode = file_inode(file);
29901 + err = copy_vdir(vdir_cache, au_ivdir(inode));
29903 + file->f_version = inode->i_version;
29905 + au_set_fvdir_cache(file, allocated);
29906 + } else if (allocated)
29907 + au_vdir_free(allocated);
29913 +static loff_t calc_offset(struct au_vdir *vdir)
29916 + union au_vdir_deblk_p p;
29918 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
29919 + offset = vdir->vd_last.p.deblk - p.deblk;
29920 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
29924 +/* returns true or false */
29925 +static int seek_vdir(struct file *file, struct dir_context *ctx)
29928 + unsigned int deblk_sz;
29929 + unsigned long ul, n;
29931 + union au_vdir_deblk_p p, deblk_end;
29932 + struct au_vdir *vdir_cache;
29935 + vdir_cache = au_fvdir_cache(file);
29936 + offset = calc_offset(vdir_cache);
29937 + AuDbg("offset %lld\n", offset);
29938 + if (ctx->pos == offset)
29941 + vdir_cache->vd_last.ul = 0;
29942 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
29947 + deblk_sz = vdir_cache->vd_deblk_sz;
29948 + ul = div64_u64(ctx->pos, deblk_sz);
29949 + AuDbg("ul %lu\n", ul);
29950 + if (ul >= vdir_cache->vd_nblk)
29953 + n = vdir_cache->vd_nblk;
29954 + for (; ul < n; ul++) {
29955 + p.deblk = vdir_cache->vd_deblk[ul];
29956 + deblk_end.deblk = p.deblk + deblk_sz;
29958 + offset *= deblk_sz;
29959 + while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
29962 + l = calc_size(p.de->de_str.len);
29966 + if (!is_deblk_end(&p, &deblk_end)) {
29968 + vdir_cache->vd_last.ul = ul;
29969 + vdir_cache->vd_last.p = p;
29976 + AuTraceErr(!valid);
29980 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
29982 + unsigned int l, deblk_sz;
29983 + union au_vdir_deblk_p deblk_end;
29984 + struct au_vdir *vdir_cache;
29985 + struct au_vdir_de *de;
29987 + vdir_cache = au_fvdir_cache(file);
29988 + if (!seek_vdir(file, ctx))
29991 + deblk_sz = vdir_cache->vd_deblk_sz;
29993 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
29994 + deblk_end.deblk += deblk_sz;
29995 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
29996 + de = vdir_cache->vd_last.p.de;
29997 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
29998 + de->de_str.len, de->de_str.name, ctx->pos,
29999 + (unsigned long)de->de_ino, de->de_type);
30000 + if (unlikely(!dir_emit(ctx, de->de_str.name,
30001 + de->de_str.len, de->de_ino,
30003 + /* todo: ignore the error caused by udba? */
30004 + /* return err; */
30008 + l = calc_size(de->de_str.len);
30009 + vdir_cache->vd_last.p.deblk += l;
30012 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
30013 + vdir_cache->vd_last.ul++;
30014 + vdir_cache->vd_last.p.deblk
30015 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
30016 + ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
30025 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
30026 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
30027 +++ linux/fs/aufs/vfsub.c 2016-01-13 20:11:11.673093671 +0100
30030 + * Copyright (C) 2005-2015 Junjiro R. Okajima
30032 + * This program, aufs is free software; you can redistribute it and/or modify
30033 + * it under the terms of the GNU General Public License as published by
30034 + * the Free Software Foundation; either version 2 of the License, or
30035 + * (at your option) any later version.
30037 + * This program is distributed in the hope that it will be useful,
30038 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30039 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30040 + * GNU General Public License for more details.
30042 + * You should have received a copy of the GNU General Public License
30043 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30047 + * sub-routines for VFS
30050 +#include <linux/namei.h>
30051 +#include <linux/security.h>
30052 +#include <linux/splice.h>
30055 +int vfsub_update_h_iattr(struct path *h_path, int *did)
30059 + struct super_block *h_sb;
30061 + /* for remote fs, leave work for its getattr or d_revalidate */
30062 + /* for bad i_attr fs, handle them in aufs_getattr() */
30063 + /* still some fs may acquire i_mutex. we need to skip them */
30067 + h_sb = h_path->dentry->d_sb;
30068 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
30070 + err = vfs_getattr(h_path, &st);
30075 +/* ---------------------------------------------------------------------- */
30077 +struct file *vfsub_dentry_open(struct path *path, int flags)
30079 + struct file *file;
30081 + file = dentry_open(path, flags /* | __FMODE_NONOTIFY */,
30083 + if (!IS_ERR_OR_NULL(file)
30084 + && (file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
30085 + i_readcount_inc(d_inode(path->dentry));
30090 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
30092 + struct file *file;
30095 + file = filp_open(path,
30096 + oflags /* | __FMODE_NONOTIFY */,
30099 + if (IS_ERR(file))
30101 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30108 + * Ideally this function should call VFS:do_last() in order to keep all its
30109 + * checkings. But it is very hard for aufs to regenerate several VFS internal
30110 + * structure such as nameidata. This is a second (or third) best approach.
30111 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
30113 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
30114 + struct vfsub_aopen_args *args, struct au_branch *br)
30117 + struct file *file = args->file;
30118 + /* copied from linux/fs/namei.c:atomic_open() */
30119 + struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
30122 + AuDebugOn(!dir->i_op->atomic_open);
30124 + err = au_br_test_oflag(args->open_flag, br);
30125 + if (unlikely(err))
30128 + args->file->f_path.dentry = DENTRY_NOT_SET;
30129 + args->file->f_path.mnt = au_br_mnt(br);
30130 + err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
30131 + args->create_mode, args->opened);
30133 + /* some filesystems don't set FILE_CREATED while succeeded? */
30134 + if (*args->opened & FILE_CREATED)
30135 + fsnotify_create(dir, dentry);
30141 + /* todo: call VFS:may_open() here */
30142 + err = open_check_o_direct(file);
30143 + /* todo: ima_file_check() too? */
30144 + if (!err && (args->open_flag & __FMODE_EXEC))
30145 + err = deny_write_access(file);
30146 + if (unlikely(err))
30147 + /* note that the file is created and still opened */
30151 + atomic_inc(&br->br_count);
30152 + fsnotify_open(file);
30158 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
30162 + err = kern_path(name, flags, path);
30163 + if (!err && d_is_positive(path->dentry))
30164 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
30168 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
30171 + struct path path = {
30175 + /* VFS checks it too, but by WARN_ON_ONCE() */
30176 + IMustLock(d_inode(parent));
30178 + path.dentry = lookup_one_len(name, parent, len);
30179 + if (IS_ERR(path.dentry))
30181 + if (d_is_positive(path.dentry))
30182 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
30185 + AuTraceErrPtr(path.dentry);
30186 + return path.dentry;
30189 +void vfsub_call_lkup_one(void *args)
30191 + struct vfsub_lkup_one_args *a = args;
30192 + *a->errp = vfsub_lkup_one(a->name, a->parent);
30195 +/* ---------------------------------------------------------------------- */
30197 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
30198 + struct dentry *d2, struct au_hinode *hdir2)
30200 + struct dentry *d;
30203 + d = lock_rename(d1, d2);
30205 + au_hn_suspend(hdir1);
30206 + if (hdir1 != hdir2)
30207 + au_hn_suspend(hdir2);
30212 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
30213 + struct dentry *d2, struct au_hinode *hdir2)
30215 + au_hn_resume(hdir1);
30216 + if (hdir1 != hdir2)
30217 + au_hn_resume(hdir2);
30219 + unlock_rename(d1, d2);
30223 +/* ---------------------------------------------------------------------- */
30225 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
30228 + struct dentry *d;
30232 + d = path->dentry;
30233 + path->dentry = d->d_parent;
30234 + err = security_path_mknod(path, d, mode, 0);
30235 + path->dentry = d;
30236 + if (unlikely(err))
30240 + err = vfs_create(dir, path->dentry, mode, want_excl);
30243 + struct path tmp = *path;
30246 + vfsub_update_h_iattr(&tmp, &did);
30248 + tmp.dentry = path->dentry->d_parent;
30249 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30258 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
30261 + struct dentry *d;
30265 + d = path->dentry;
30266 + path->dentry = d->d_parent;
30267 + err = security_path_symlink(path, d, symname);
30268 + path->dentry = d;
30269 + if (unlikely(err))
30273 + err = vfs_symlink(dir, path->dentry, symname);
30276 + struct path tmp = *path;
30279 + vfsub_update_h_iattr(&tmp, &did);
30281 + tmp.dentry = path->dentry->d_parent;
30282 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30291 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
30294 + struct dentry *d;
30298 + d = path->dentry;
30299 + path->dentry = d->d_parent;
30300 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
30301 + path->dentry = d;
30302 + if (unlikely(err))
30306 + err = vfs_mknod(dir, path->dentry, mode, dev);
30309 + struct path tmp = *path;
30312 + vfsub_update_h_iattr(&tmp, &did);
30314 + tmp.dentry = path->dentry->d_parent;
30315 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30324 +static int au_test_nlink(struct inode *inode)
30326 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
30328 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
30329 + || inode->i_nlink < link_max)
30334 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
30335 + struct inode **delegated_inode)
30338 + struct dentry *d;
30342 + err = au_test_nlink(d_inode(src_dentry));
30343 + if (unlikely(err))
30346 + /* we don't call may_linkat() */
30347 + d = path->dentry;
30348 + path->dentry = d->d_parent;
30349 + err = security_path_link(src_dentry, path, d);
30350 + path->dentry = d;
30351 + if (unlikely(err))
30355 + err = vfs_link(src_dentry, dir, path->dentry, delegated_inode);
30358 + struct path tmp = *path;
30361 + /* fuse has different memory inode for the same inumber */
30362 + vfsub_update_h_iattr(&tmp, &did);
30364 + tmp.dentry = path->dentry->d_parent;
30365 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30366 + tmp.dentry = src_dentry;
30367 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30376 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
30377 + struct inode *dir, struct path *path,
30378 + struct inode **delegated_inode)
30381 + struct path tmp = {
30384 + struct dentry *d;
30387 + IMustLock(src_dir);
30389 + d = path->dentry;
30390 + path->dentry = d->d_parent;
30391 + tmp.dentry = src_dentry->d_parent;
30392 + err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
30393 + path->dentry = d;
30394 + if (unlikely(err))
30398 + err = vfs_rename(src_dir, src_dentry, dir, path->dentry,
30399 + delegated_inode, /*flags*/0);
30404 + tmp.dentry = d->d_parent;
30405 + vfsub_update_h_iattr(&tmp, &did);
30407 + tmp.dentry = src_dentry;
30408 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30409 + tmp.dentry = src_dentry->d_parent;
30410 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30419 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
30422 + struct dentry *d;
30426 + d = path->dentry;
30427 + path->dentry = d->d_parent;
30428 + err = security_path_mkdir(path, d, mode);
30429 + path->dentry = d;
30430 + if (unlikely(err))
30434 + err = vfs_mkdir(dir, path->dentry, mode);
30437 + struct path tmp = *path;
30440 + vfsub_update_h_iattr(&tmp, &did);
30442 + tmp.dentry = path->dentry->d_parent;
30443 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30452 +int vfsub_rmdir(struct inode *dir, struct path *path)
30455 + struct dentry *d;
30459 + d = path->dentry;
30460 + path->dentry = d->d_parent;
30461 + err = security_path_rmdir(path, d);
30462 + path->dentry = d;
30463 + if (unlikely(err))
30467 + err = vfs_rmdir(dir, path->dentry);
30470 + struct path tmp = {
30471 + .dentry = path->dentry->d_parent,
30475 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
30482 +/* ---------------------------------------------------------------------- */
30484 +/* todo: support mmap_sem? */
30485 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
30491 + err = vfs_read(file, ubuf, count, ppos);
30494 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30498 +/* todo: kernel_read()? */
30499 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
30503 + mm_segment_t oldfs;
30510 + oldfs = get_fs();
30511 + set_fs(KERNEL_DS);
30512 + err = vfsub_read_u(file, buf.u, count, ppos);
30517 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
30523 + err = vfs_write(file, ubuf, count, ppos);
30526 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30530 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
30533 + mm_segment_t oldfs;
30536 + const char __user *u;
30540 + oldfs = get_fs();
30541 + set_fs(KERNEL_DS);
30542 + err = vfsub_write_u(file, buf.u, count, ppos);
30547 +int vfsub_flush(struct file *file, fl_owner_t id)
30552 + if (file->f_op->flush) {
30553 + if (!au_test_nfs(file->f_path.dentry->d_sb))
30554 + err = file->f_op->flush(file, id);
30557 + err = file->f_op->flush(file, id);
30561 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
30567 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
30571 + AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
30574 + err = iterate_dir(file, ctx);
30577 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30581 +long vfsub_splice_to(struct file *in, loff_t *ppos,
30582 + struct pipe_inode_info *pipe, size_t len,
30583 + unsigned int flags)
30588 + err = do_splice_to(in, ppos, pipe, len, flags);
30590 + file_accessed(in);
30592 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
30596 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
30597 + loff_t *ppos, size_t len, unsigned int flags)
30602 + err = do_splice_from(pipe, out, ppos, len, flags);
30605 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
30609 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
30613 + /* file can be NULL */
30615 + err = vfs_fsync(file, datasync);
30619 + AuDebugOn(!file);
30620 + path = &file->f_path;
30622 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
30627 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
30628 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
30629 + struct file *h_file)
30632 + struct inode *h_inode;
30633 + struct super_block *h_sb;
30636 + err = vfsub_truncate(h_path, length);
30640 + h_inode = d_inode(h_path->dentry);
30641 + h_sb = h_inode->i_sb;
30643 + sb_start_write(h_sb);
30645 + err = locks_verify_truncate(h_inode, h_file, length);
30647 + err = security_path_truncate(h_path);
30650 + err = do_truncate(h_path->dentry, length, attr, h_file);
30654 + sb_end_write(h_sb);
30661 +/* ---------------------------------------------------------------------- */
30663 +struct au_vfsub_mkdir_args {
30665 + struct inode *dir;
30666 + struct path *path;
30670 +static void au_call_vfsub_mkdir(void *args)
30672 + struct au_vfsub_mkdir_args *a = args;
30673 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
30676 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
30678 + int err, do_sio, wkq_err;
30680 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
30683 + err = vfsub_mkdir(dir, path, mode);
30686 + struct au_vfsub_mkdir_args args = {
30692 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
30693 + if (unlikely(wkq_err))
30700 +struct au_vfsub_rmdir_args {
30702 + struct inode *dir;
30703 + struct path *path;
30706 +static void au_call_vfsub_rmdir(void *args)
30708 + struct au_vfsub_rmdir_args *a = args;
30709 + *a->errp = vfsub_rmdir(a->dir, a->path);
30712 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
30714 + int err, do_sio, wkq_err;
30716 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
30719 + err = vfsub_rmdir(dir, path);
30722 + struct au_vfsub_rmdir_args args = {
30727 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
30728 + if (unlikely(wkq_err))
30735 +/* ---------------------------------------------------------------------- */
30737 +struct notify_change_args {
30739 + struct path *path;
30740 + struct iattr *ia;
30741 + struct inode **delegated_inode;
30744 +static void call_notify_change(void *args)
30746 + struct notify_change_args *a = args;
30747 + struct inode *h_inode;
30749 + h_inode = d_inode(a->path->dentry);
30750 + IMustLock(h_inode);
30752 + *a->errp = -EPERM;
30753 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
30755 + *a->errp = notify_change(a->path->dentry, a->ia,
30756 + a->delegated_inode);
30759 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
30761 + AuTraceErr(*a->errp);
30764 +int vfsub_notify_change(struct path *path, struct iattr *ia,
30765 + struct inode **delegated_inode)
30768 + struct notify_change_args args = {
30772 + .delegated_inode = delegated_inode
30775 + call_notify_change(&args);
30780 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
30781 + struct inode **delegated_inode)
30783 + int err, wkq_err;
30784 + struct notify_change_args args = {
30788 + .delegated_inode = delegated_inode
30791 + wkq_err = au_wkq_wait(call_notify_change, &args);
30792 + if (unlikely(wkq_err))
30798 +/* ---------------------------------------------------------------------- */
30800 +struct unlink_args {
30802 + struct inode *dir;
30803 + struct path *path;
30804 + struct inode **delegated_inode;
30807 +static void call_unlink(void *args)
30809 + struct unlink_args *a = args;
30810 + struct dentry *d = a->path->dentry;
30811 + struct inode *h_inode;
30812 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
30813 + && au_dcount(d) == 1);
30815 + IMustLock(a->dir);
30817 + a->path->dentry = d->d_parent;
30818 + *a->errp = security_path_unlink(a->path, d);
30819 + a->path->dentry = d;
30820 + if (unlikely(*a->errp))
30823 + if (!stop_sillyrename)
30826 + if (d_is_positive(d)) {
30827 + h_inode = d_inode(d);
30832 + *a->errp = vfs_unlink(a->dir, d, a->delegated_inode);
30835 + struct path tmp = {
30836 + .dentry = d->d_parent,
30837 + .mnt = a->path->mnt
30839 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
30842 + if (!stop_sillyrename)
30847 + AuTraceErr(*a->errp);
30851 + * @dir: must be locked.
30852 + * @dentry: target dentry.
30854 +int vfsub_unlink(struct inode *dir, struct path *path,
30855 + struct inode **delegated_inode, int force)
30858 + struct unlink_args args = {
30862 + .delegated_inode = delegated_inode
30866 + call_unlink(&args);
30870 + wkq_err = au_wkq_wait(call_unlink, &args);
30871 + if (unlikely(wkq_err))
30877 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
30878 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
30879 +++ linux/fs/aufs/vfsub.h 2016-01-13 20:11:11.673093671 +0100
30882 + * Copyright (C) 2005-2015 Junjiro R. Okajima
30884 + * This program, aufs is free software; you can redistribute it and/or modify
30885 + * it under the terms of the GNU General Public License as published by
30886 + * the Free Software Foundation; either version 2 of the License, or
30887 + * (at your option) any later version.
30889 + * This program is distributed in the hope that it will be useful,
30890 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30891 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30892 + * GNU General Public License for more details.
30894 + * You should have received a copy of the GNU General Public License
30895 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30899 + * sub-routines for VFS
30902 +#ifndef __AUFS_VFSUB_H__
30903 +#define __AUFS_VFSUB_H__
30907 +#include <linux/fs.h>
30908 +#include <linux/mount.h>
30909 +#include <linux/xattr.h>
30910 +#include "debug.h"
30912 +/* copied from linux/fs/internal.h */
30913 +/* todo: BAD approach!! */
30914 +extern void __mnt_drop_write(struct vfsmount *);
30915 +extern int open_check_o_direct(struct file *f);
30917 +/* ---------------------------------------------------------------------- */
30919 +/* lock subclass for lower inode */
30920 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
30921 +/* reduce? gave up. */
30923 + AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
30924 + AuLsc_I_PARENT, /* lower inode, parent first */
30925 + AuLsc_I_PARENT2, /* copyup dirs */
30926 + AuLsc_I_PARENT3, /* copyup wh */
30932 +/* to debug easier, do not make them inlined functions */
30933 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
30934 +#define IMustLock(i) MtxMustLock(&(i)->i_mutex)
30936 +/* ---------------------------------------------------------------------- */
30938 +static inline void vfsub_drop_nlink(struct inode *inode)
30940 + AuDebugOn(!inode->i_nlink);
30941 + drop_nlink(inode);
30944 +static inline void vfsub_dead_dir(struct inode *inode)
30946 + AuDebugOn(!S_ISDIR(inode->i_mode));
30947 + inode->i_flags |= S_DEAD;
30948 + clear_nlink(inode);
30951 +static inline int vfsub_native_ro(struct inode *inode)
30953 + return (inode->i_sb->s_flags & MS_RDONLY)
30954 + || IS_RDONLY(inode)
30955 + /* || IS_APPEND(inode) */
30956 + || IS_IMMUTABLE(inode);
30959 +/* ---------------------------------------------------------------------- */
30961 +int vfsub_update_h_iattr(struct path *h_path, int *did);
30962 +struct file *vfsub_dentry_open(struct path *path, int flags);
30963 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
30964 +struct vfsub_aopen_args {
30965 + struct file *file;
30966 + unsigned int open_flag;
30967 + umode_t create_mode;
30971 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
30972 + struct vfsub_aopen_args *args, struct au_branch *br);
30973 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
30975 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
30978 +struct vfsub_lkup_one_args {
30979 + struct dentry **errp;
30980 + struct qstr *name;
30981 + struct dentry *parent;
30984 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
30985 + struct dentry *parent)
30987 + return vfsub_lookup_one_len(name->name, parent, name->len);
30990 +void vfsub_call_lkup_one(void *args);
30992 +/* ---------------------------------------------------------------------- */
30994 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
30999 + err = mnt_want_write(mnt);
31004 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
31007 + mnt_drop_write(mnt);
31011 +#if 0 /* reserved */
31012 +static inline void vfsub_mnt_drop_write_file(struct file *file)
31015 + mnt_drop_write_file(file);
31020 +/* ---------------------------------------------------------------------- */
31023 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
31024 + struct dentry *d2, struct au_hinode *hdir2);
31025 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
31026 + struct dentry *d2, struct au_hinode *hdir2);
31028 +int vfsub_create(struct inode *dir, struct path *path, int mode,
31030 +int vfsub_symlink(struct inode *dir, struct path *path,
31031 + const char *symname);
31032 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
31033 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
31034 + struct path *path, struct inode **delegated_inode);
31035 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
31036 + struct inode *hdir, struct path *path,
31037 + struct inode **delegated_inode);
31038 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
31039 +int vfsub_rmdir(struct inode *dir, struct path *path);
31041 +/* ---------------------------------------------------------------------- */
31043 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
31045 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
31047 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
31049 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
31051 +int vfsub_flush(struct file *file, fl_owner_t id);
31052 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
31054 +static inline loff_t vfsub_f_size_read(struct file *file)
31056 + return i_size_read(file_inode(file));
31059 +static inline unsigned int vfsub_file_flags(struct file *file)
31061 + unsigned int flags;
31063 + spin_lock(&file->f_lock);
31064 + flags = file->f_flags;
31065 + spin_unlock(&file->f_lock);
31070 +#if 0 /* reserved */
31071 +static inline void vfsub_file_accessed(struct file *h_file)
31073 + file_accessed(h_file);
31074 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
31078 +#if 0 /* reserved */
31079 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
31080 + struct dentry *h_dentry)
31082 + struct path h_path = {
31083 + .dentry = h_dentry,
31086 + touch_atime(&h_path);
31087 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
31091 +static inline int vfsub_update_time(struct inode *h_inode, struct timespec *ts,
31094 + return generic_update_time(h_inode, ts, flags);
31095 + /* no vfsub_update_h_iattr() since we don't have struct path */
31098 +long vfsub_splice_to(struct file *in, loff_t *ppos,
31099 + struct pipe_inode_info *pipe, size_t len,
31100 + unsigned int flags);
31101 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
31102 + loff_t *ppos, size_t len, unsigned int flags);
31104 +static inline long vfsub_truncate(struct path *path, loff_t length)
31109 + err = vfs_truncate(path, length);
31114 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
31115 + struct file *h_file);
31116 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
31118 +/* ---------------------------------------------------------------------- */
31120 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
31125 + err = vfs_llseek(file, offset, origin);
31130 +/* ---------------------------------------------------------------------- */
31132 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
31133 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
31134 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
31135 + struct inode **delegated_inode);
31136 +int vfsub_notify_change(struct path *path, struct iattr *ia,
31137 + struct inode **delegated_inode);
31138 +int vfsub_unlink(struct inode *dir, struct path *path,
31139 + struct inode **delegated_inode, int force);
31141 +/* ---------------------------------------------------------------------- */
31143 +static inline int vfsub_setxattr(struct dentry *dentry, const char *name,
31144 + const void *value, size_t size, int flags)
31149 + err = vfs_setxattr(dentry, name, value, size, flags);
31155 +static inline int vfsub_removexattr(struct dentry *dentry, const char *name)
31160 + err = vfs_removexattr(dentry, name);
31166 +#endif /* __KERNEL__ */
31167 +#endif /* __AUFS_VFSUB_H__ */
31168 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
31169 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
31170 +++ linux/fs/aufs/wbr_policy.c 2016-01-13 20:11:11.673093671 +0100
31173 + * Copyright (C) 2005-2015 Junjiro R. Okajima
31175 + * This program, aufs is free software; you can redistribute it and/or modify
31176 + * it under the terms of the GNU General Public License as published by
31177 + * the Free Software Foundation; either version 2 of the License, or
31178 + * (at your option) any later version.
31180 + * This program is distributed in the hope that it will be useful,
31181 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31182 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31183 + * GNU General Public License for more details.
31185 + * You should have received a copy of the GNU General Public License
31186 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31190 + * policies for selecting one among multiple writable branches
31193 +#include <linux/statfs.h>
31196 +/* subset of cpup_attr() */
31197 +static noinline_for_stack
31198 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
31202 + struct inode *h_isrc;
31204 + h_isrc = d_inode(h_src);
31205 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
31206 + ia.ia_mode = h_isrc->i_mode;
31207 + ia.ia_uid = h_isrc->i_uid;
31208 + ia.ia_gid = h_isrc->i_gid;
31209 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
31210 + au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
31211 + /* no delegation since it is just created */
31212 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
31214 + /* is this nfs only? */
31215 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
31216 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
31217 + ia.ia_mode = h_isrc->i_mode;
31218 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
31224 +#define AuCpdown_PARENT_OPQ 1
31225 +#define AuCpdown_WHED (1 << 1)
31226 +#define AuCpdown_MADE_DIR (1 << 2)
31227 +#define AuCpdown_DIROPQ (1 << 3)
31228 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
31229 +#define au_fset_cpdown(flags, name) \
31230 + do { (flags) |= AuCpdown_##name; } while (0)
31231 +#define au_fclr_cpdown(flags, name) \
31232 + do { (flags) &= ~AuCpdown_##name; } while (0)
31234 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
31235 + unsigned int *flags)
31238 + struct dentry *opq_dentry;
31240 + opq_dentry = au_diropq_create(dentry, bdst);
31241 + err = PTR_ERR(opq_dentry);
31242 + if (IS_ERR(opq_dentry))
31244 + dput(opq_dentry);
31245 + au_fset_cpdown(*flags, DIROPQ);
31251 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
31252 + struct inode *dir, aufs_bindex_t bdst)
31255 + struct path h_path;
31256 + struct au_branch *br;
31258 + br = au_sbr(dentry->d_sb, bdst);
31259 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
31260 + err = PTR_ERR(h_path.dentry);
31261 + if (IS_ERR(h_path.dentry))
31265 + if (d_is_positive(h_path.dentry)) {
31266 + h_path.mnt = au_br_mnt(br);
31267 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
31270 + dput(h_path.dentry);
31276 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
31277 + struct au_pin *pin,
31278 + struct dentry *h_parent, void *arg)
31281 + aufs_bindex_t bopq, bstart;
31282 + struct path h_path;
31283 + struct dentry *parent;
31284 + struct inode *h_dir, *h_inode, *inode, *dir;
31285 + unsigned int *flags = arg;
31287 + bstart = au_dbstart(dentry);
31288 + /* dentry is di-locked */
31289 + parent = dget_parent(dentry);
31290 + dir = d_inode(parent);
31291 + h_dir = d_inode(h_parent);
31292 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
31293 + IMustLock(h_dir);
31295 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
31296 + if (unlikely(err < 0))
31298 + h_path.dentry = au_h_dptr(dentry, bdst);
31299 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
31300 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path,
31301 + S_IRWXU | S_IRUGO | S_IXUGO);
31302 + if (unlikely(err))
31304 + au_fset_cpdown(*flags, MADE_DIR);
31306 + bopq = au_dbdiropq(dentry);
31307 + au_fclr_cpdown(*flags, WHED);
31308 + au_fclr_cpdown(*flags, DIROPQ);
31309 + if (au_dbwh(dentry) == bdst)
31310 + au_fset_cpdown(*flags, WHED);
31311 + if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
31312 + au_fset_cpdown(*flags, PARENT_OPQ);
31313 + h_inode = d_inode(h_path.dentry);
31314 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
31315 + if (au_ftest_cpdown(*flags, WHED)) {
31316 + err = au_cpdown_dir_opq(dentry, bdst, flags);
31317 + if (unlikely(err)) {
31318 + mutex_unlock(&h_inode->i_mutex);
31323 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, bstart));
31324 + mutex_unlock(&h_inode->i_mutex);
31325 + if (unlikely(err))
31328 + if (au_ftest_cpdown(*flags, WHED)) {
31329 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
31330 + if (unlikely(err))
31334 + inode = d_inode(dentry);
31335 + if (au_ibend(inode) < bdst)
31336 + au_set_ibend(inode, bdst);
31337 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
31338 + au_hi_flags(inode, /*isdir*/1));
31339 + au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
31340 + goto out; /* success */
31344 + if (au_ftest_cpdown(*flags, DIROPQ)) {
31345 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
31346 + rerr = au_diropq_remove(dentry, bdst);
31347 + mutex_unlock(&h_inode->i_mutex);
31348 + if (unlikely(rerr)) {
31349 + AuIOErr("failed removing diropq for %pd b%d (%d)\n",
31350 + dentry, bdst, rerr);
31356 + if (au_ftest_cpdown(*flags, MADE_DIR)) {
31357 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
31358 + if (unlikely(rerr)) {
31359 + AuIOErr("failed removing %pd b%d (%d)\n",
31360 + dentry, bdst, rerr);
31365 + au_set_h_dptr(dentry, bdst, NULL);
31366 + if (au_dbend(dentry) == bdst)
31367 + au_update_dbend(dentry);
31373 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
31376 + unsigned int flags;
31379 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
31384 +/* ---------------------------------------------------------------------- */
31386 +/* policies for create */
31388 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
31390 + int err, i, j, ndentry;
31391 + aufs_bindex_t bopq;
31392 + struct au_dcsub_pages dpages;
31393 + struct au_dpage *dpage;
31394 + struct dentry **dentries, *parent, *d;
31396 + err = au_dpages_init(&dpages, GFP_NOFS);
31397 + if (unlikely(err))
31399 + parent = dget_parent(dentry);
31400 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
31401 + if (unlikely(err))
31405 + for (i = 0; i < dpages.ndpage; i++) {
31406 + dpage = dpages.dpages + i;
31407 + dentries = dpage->dentries;
31408 + ndentry = dpage->ndentry;
31409 + for (j = 0; j < ndentry; j++) {
31411 + di_read_lock_parent2(d, !AuLock_IR);
31412 + bopq = au_dbdiropq(d);
31413 + di_read_unlock(d, !AuLock_IR);
31414 + if (bopq >= 0 && bopq < err)
31421 + au_dpages_free(&dpages);
31426 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
31428 + for (; bindex >= 0; bindex--)
31429 + if (!au_br_rdonly(au_sbr(sb, bindex)))
31434 +/* top down parent */
31435 +static int au_wbr_create_tdp(struct dentry *dentry,
31436 + unsigned int flags __maybe_unused)
31439 + aufs_bindex_t bstart, bindex;
31440 + struct super_block *sb;
31441 + struct dentry *parent, *h_parent;
31443 + sb = dentry->d_sb;
31444 + bstart = au_dbstart(dentry);
31446 + if (!au_br_rdonly(au_sbr(sb, bstart)))
31450 + parent = dget_parent(dentry);
31451 + for (bindex = au_dbstart(parent); bindex < bstart; bindex++) {
31452 + h_parent = au_h_dptr(parent, bindex);
31453 + if (!h_parent || d_is_negative(h_parent))
31456 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
31463 + /* bottom up here */
31464 + if (unlikely(err < 0)) {
31465 + err = au_wbr_bu(sb, bstart - 1);
31467 + err = au_wbr_nonopq(dentry, err);
31471 + AuDbg("b%d\n", err);
31475 +/* ---------------------------------------------------------------------- */
31477 +/* an exception for the policy other than tdp */
31478 +static int au_wbr_create_exp(struct dentry *dentry)
31481 + aufs_bindex_t bwh, bdiropq;
31482 + struct dentry *parent;
31485 + bwh = au_dbwh(dentry);
31486 + parent = dget_parent(dentry);
31487 + bdiropq = au_dbdiropq(parent);
31489 + if (bdiropq >= 0)
31490 + err = min(bdiropq, bwh);
31493 + AuDbg("%d\n", err);
31494 + } else if (bdiropq >= 0) {
31496 + AuDbg("%d\n", err);
31501 + err = au_wbr_nonopq(dentry, err);
31503 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
31506 + AuDbg("%d\n", err);
31510 +/* ---------------------------------------------------------------------- */
31513 +static int au_wbr_create_init_rr(struct super_block *sb)
31517 + err = au_wbr_bu(sb, au_sbend(sb));
31518 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
31521 + AuDbg("b%d\n", err);
31525 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
31529 + aufs_bindex_t bindex, bend;
31530 + struct super_block *sb;
31533 + err = au_wbr_create_exp(dentry);
31537 + sb = dentry->d_sb;
31538 + next = &au_sbi(sb)->si_wbr_rr_next;
31539 + bend = au_sbend(sb);
31541 + for (bindex = 0; bindex <= bend; bindex++) {
31542 + if (!au_ftest_wbr(flags, DIR)) {
31543 + err = atomic_dec_return(next) + 1;
31544 + /* modulo for 0 is meaningless */
31545 + if (unlikely(!err))
31546 + err = atomic_dec_return(next) + 1;
31548 + err = atomic_read(next);
31549 + AuDbg("%d\n", err);
31552 + AuDbg("%d\n", err);
31553 + if (!au_br_rdonly(au_sbr(sb, err)))
31559 + err = au_wbr_nonopq(dentry, err);
31562 + AuDbg("%d\n", err);
31566 +/* ---------------------------------------------------------------------- */
31568 +/* most free space */
31569 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
31571 + struct super_block *sb;
31572 + struct au_branch *br;
31573 + struct au_wbr_mfs *mfs;
31574 + struct dentry *h_parent;
31575 + aufs_bindex_t bindex, bend;
31577 + unsigned long long b, bavail;
31578 + struct path h_path;
31579 + /* reduce the stack usage */
31580 + struct kstatfs *st;
31582 + st = kmalloc(sizeof(*st), GFP_NOFS);
31583 + if (unlikely(!st)) {
31584 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
31589 + sb = dentry->d_sb;
31590 + mfs = &au_sbi(sb)->si_wbr_mfs;
31591 + MtxMustLock(&mfs->mfs_lock);
31592 + mfs->mfs_bindex = -EROFS;
31593 + mfs->mfsrr_bytes = 0;
31596 + bend = au_sbend(sb);
31598 + bindex = au_dbstart(parent);
31599 + bend = au_dbtaildir(parent);
31602 + for (; bindex <= bend; bindex++) {
31604 + h_parent = au_h_dptr(parent, bindex);
31605 + if (!h_parent || d_is_negative(h_parent))
31608 + br = au_sbr(sb, bindex);
31609 + if (au_br_rdonly(br))
31612 + /* sb->s_root for NFS is unreliable */
31613 + h_path.mnt = au_br_mnt(br);
31614 + h_path.dentry = h_path.mnt->mnt_root;
31615 + err = vfs_statfs(&h_path, st);
31616 + if (unlikely(err)) {
31617 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
31621 + /* when the available size is equal, select the lower one */
31622 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
31623 + || sizeof(b) < sizeof(st->f_bsize));
31624 + b = st->f_bavail * st->f_bsize;
31625 + br->br_wbr->wbr_bytes = b;
31626 + if (b >= bavail) {
31628 + mfs->mfs_bindex = bindex;
31629 + mfs->mfs_jiffy = jiffies;
31633 + mfs->mfsrr_bytes = bavail;
31634 + AuDbg("b%d\n", mfs->mfs_bindex);
31638 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
31641 + struct dentry *parent;
31642 + struct super_block *sb;
31643 + struct au_wbr_mfs *mfs;
31645 + err = au_wbr_create_exp(dentry);
31649 + sb = dentry->d_sb;
31651 + if (au_ftest_wbr(flags, PARENT))
31652 + parent = dget_parent(dentry);
31653 + mfs = &au_sbi(sb)->si_wbr_mfs;
31654 + mutex_lock(&mfs->mfs_lock);
31655 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
31656 + || mfs->mfs_bindex < 0
31657 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
31658 + au_mfs(dentry, parent);
31659 + mutex_unlock(&mfs->mfs_lock);
31660 + err = mfs->mfs_bindex;
31664 + err = au_wbr_nonopq(dentry, err);
31667 + AuDbg("b%d\n", err);
31671 +static int au_wbr_create_init_mfs(struct super_block *sb)
31673 + struct au_wbr_mfs *mfs;
31675 + mfs = &au_sbi(sb)->si_wbr_mfs;
31676 + mutex_init(&mfs->mfs_lock);
31677 + mfs->mfs_jiffy = 0;
31678 + mfs->mfs_bindex = -EROFS;
31683 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
31685 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
31689 +/* ---------------------------------------------------------------------- */
31691 +/* most free space and then round robin */
31692 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
31695 + struct au_wbr_mfs *mfs;
31697 + err = au_wbr_create_mfs(dentry, flags);
31699 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
31700 + mutex_lock(&mfs->mfs_lock);
31701 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
31702 + err = au_wbr_create_rr(dentry, flags);
31703 + mutex_unlock(&mfs->mfs_lock);
31706 + AuDbg("b%d\n", err);
31710 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
31714 + au_wbr_create_init_mfs(sb); /* ignore */
31715 + err = au_wbr_create_init_rr(sb);
31720 +/* ---------------------------------------------------------------------- */
31722 +/* top down parent and most free space */
31723 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
31726 + unsigned long long b;
31727 + aufs_bindex_t bindex, bstart, bend;
31728 + struct super_block *sb;
31729 + struct dentry *parent, *h_parent;
31730 + struct au_branch *br;
31732 + err = au_wbr_create_tdp(dentry, flags);
31733 + if (unlikely(err < 0))
31735 + parent = dget_parent(dentry);
31736 + bstart = au_dbstart(parent);
31737 + bend = au_dbtaildir(parent);
31738 + if (bstart == bend)
31739 + goto out_parent; /* success */
31741 + e2 = au_wbr_create_mfs(dentry, flags);
31743 + goto out_parent; /* success */
31745 + /* when the available size is equal, select upper one */
31746 + sb = dentry->d_sb;
31747 + br = au_sbr(sb, err);
31748 + b = br->br_wbr->wbr_bytes;
31749 + AuDbg("b%d, %llu\n", err, b);
31751 + for (bindex = bstart; bindex <= bend; bindex++) {
31752 + h_parent = au_h_dptr(parent, bindex);
31753 + if (!h_parent || d_is_negative(h_parent))
31756 + br = au_sbr(sb, bindex);
31757 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
31758 + b = br->br_wbr->wbr_bytes;
31760 + AuDbg("b%d, %llu\n", err, b);
31765 + err = au_wbr_nonopq(dentry, err);
31770 + AuDbg("b%d\n", err);
31774 +/* ---------------------------------------------------------------------- */
31777 + * - top down parent
31778 + * - most free space with parent
31779 + * - most free space round-robin regardless parent
31781 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
31784 + unsigned long long watermark;
31785 + struct super_block *sb;
31786 + struct au_branch *br;
31787 + struct au_wbr_mfs *mfs;
31789 + err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
31790 + if (unlikely(err < 0))
31793 + sb = dentry->d_sb;
31794 + br = au_sbr(sb, err);
31795 + mfs = &au_sbi(sb)->si_wbr_mfs;
31796 + mutex_lock(&mfs->mfs_lock);
31797 + watermark = mfs->mfsrr_watermark;
31798 + mutex_unlock(&mfs->mfs_lock);
31799 + if (br->br_wbr->wbr_bytes < watermark)
31800 + /* regardless the parent dir */
31801 + err = au_wbr_create_mfsrr(dentry, flags);
31804 + AuDbg("b%d\n", err);
31808 +/* ---------------------------------------------------------------------- */
31810 +/* policies for copyup */
31812 +/* top down parent */
31813 +static int au_wbr_copyup_tdp(struct dentry *dentry)
31815 + return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
31818 +/* bottom up parent */
31819 +static int au_wbr_copyup_bup(struct dentry *dentry)
31822 + aufs_bindex_t bindex, bstart;
31823 + struct dentry *parent, *h_parent;
31824 + struct super_block *sb;
31827 + sb = dentry->d_sb;
31828 + parent = dget_parent(dentry);
31829 + bstart = au_dbstart(parent);
31830 + for (bindex = au_dbstart(dentry); bindex >= bstart; bindex--) {
31831 + h_parent = au_h_dptr(parent, bindex);
31832 + if (!h_parent || d_is_negative(h_parent))
31835 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
31842 + /* bottom up here */
31843 + if (unlikely(err < 0))
31844 + err = au_wbr_bu(sb, bstart - 1);
31846 + AuDbg("b%d\n", err);
31851 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t bstart)
31855 + err = au_wbr_bu(dentry->d_sb, bstart);
31856 + AuDbg("b%d\n", err);
31857 + if (err > bstart)
31858 + err = au_wbr_nonopq(dentry, err);
31860 + AuDbg("b%d\n", err);
31864 +static int au_wbr_copyup_bu(struct dentry *dentry)
31867 + aufs_bindex_t bstart;
31869 + bstart = au_dbstart(dentry);
31870 + err = au_wbr_do_copyup_bu(dentry, bstart);
31874 +/* ---------------------------------------------------------------------- */
31876 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
31877 + [AuWbrCopyup_TDP] = {
31878 + .copyup = au_wbr_copyup_tdp
31880 + [AuWbrCopyup_BUP] = {
31881 + .copyup = au_wbr_copyup_bup
31883 + [AuWbrCopyup_BU] = {
31884 + .copyup = au_wbr_copyup_bu
31888 +struct au_wbr_create_operations au_wbr_create_ops[] = {
31889 + [AuWbrCreate_TDP] = {
31890 + .create = au_wbr_create_tdp
31892 + [AuWbrCreate_RR] = {
31893 + .create = au_wbr_create_rr,
31894 + .init = au_wbr_create_init_rr
31896 + [AuWbrCreate_MFS] = {
31897 + .create = au_wbr_create_mfs,
31898 + .init = au_wbr_create_init_mfs,
31899 + .fin = au_wbr_create_fin_mfs
31901 + [AuWbrCreate_MFSV] = {
31902 + .create = au_wbr_create_mfs,
31903 + .init = au_wbr_create_init_mfs,
31904 + .fin = au_wbr_create_fin_mfs
31906 + [AuWbrCreate_MFSRR] = {
31907 + .create = au_wbr_create_mfsrr,
31908 + .init = au_wbr_create_init_mfsrr,
31909 + .fin = au_wbr_create_fin_mfs
31911 + [AuWbrCreate_MFSRRV] = {
31912 + .create = au_wbr_create_mfsrr,
31913 + .init = au_wbr_create_init_mfsrr,
31914 + .fin = au_wbr_create_fin_mfs
31916 + [AuWbrCreate_PMFS] = {
31917 + .create = au_wbr_create_pmfs,
31918 + .init = au_wbr_create_init_mfs,
31919 + .fin = au_wbr_create_fin_mfs
31921 + [AuWbrCreate_PMFSV] = {
31922 + .create = au_wbr_create_pmfs,
31923 + .init = au_wbr_create_init_mfs,
31924 + .fin = au_wbr_create_fin_mfs
31926 + [AuWbrCreate_PMFSRR] = {
31927 + .create = au_wbr_create_pmfsrr,
31928 + .init = au_wbr_create_init_mfsrr,
31929 + .fin = au_wbr_create_fin_mfs
31931 + [AuWbrCreate_PMFSRRV] = {
31932 + .create = au_wbr_create_pmfsrr,
31933 + .init = au_wbr_create_init_mfsrr,
31934 + .fin = au_wbr_create_fin_mfs
31937 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
31938 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
31939 +++ linux/fs/aufs/whout.c 2016-01-13 20:11:11.673093671 +0100
31942 + * Copyright (C) 2005-2015 Junjiro R. Okajima
31944 + * This program, aufs is free software; you can redistribute it and/or modify
31945 + * it under the terms of the GNU General Public License as published by
31946 + * the Free Software Foundation; either version 2 of the License, or
31947 + * (at your option) any later version.
31949 + * This program is distributed in the hope that it will be useful,
31950 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31951 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31952 + * GNU General Public License for more details.
31954 + * You should have received a copy of the GNU General Public License
31955 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31959 + * whiteout for logical deletion and opaque directory
31964 +#define WH_MASK S_IRUGO
31967 + * If a directory contains this file, then it is opaque. We start with the
31968 + * .wh. flag so that it is blocked by lookup.
31970 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
31971 + sizeof(AUFS_WH_DIROPQ) - 1);
31974 + * generate whiteout name, which is NOT terminated by NULL.
31975 + * @name: original d_name.name
31976 + * @len: original d_name.len
31977 + * @wh: whiteout qstr
31978 + * returns zero when succeeds, otherwise error.
31979 + * succeeded value as wh->name should be freed by kfree().
31981 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
31985 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
31986 + return -ENAMETOOLONG;
31988 + wh->len = name->len + AUFS_WH_PFX_LEN;
31989 + p = kmalloc(wh->len, GFP_NOFS);
31992 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
31993 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
32000 +/* ---------------------------------------------------------------------- */
32003 + * test if the @wh_name exists under @h_parent.
32004 + * @try_sio specifies the necessary of super-io.
32006 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio)
32009 + struct dentry *wh_dentry;
32012 + wh_dentry = vfsub_lkup_one(wh_name, h_parent);
32014 + wh_dentry = au_sio_lkup_one(wh_name, h_parent);
32015 + err = PTR_ERR(wh_dentry);
32016 + if (IS_ERR(wh_dentry)) {
32017 + if (err == -ENAMETOOLONG)
32023 + if (d_is_negative(wh_dentry))
32024 + goto out_wh; /* success */
32027 + if (d_is_reg(wh_dentry))
32028 + goto out_wh; /* success */
32031 + AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
32032 + wh_dentry, d_inode(wh_dentry)->i_mode);
32041 + * test if the @h_dentry sets opaque or not.
32043 +int au_diropq_test(struct dentry *h_dentry)
32046 + struct inode *h_dir;
32048 + h_dir = d_inode(h_dentry);
32049 + err = au_wh_test(h_dentry, &diropq_name,
32050 + au_test_h_perm_sio(h_dir, MAY_EXEC));
32055 + * returns a negative dentry whose name is unique and temporary.
32057 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
32058 + struct qstr *prefix)
32060 + struct dentry *dentry;
32062 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
32064 + /* strict atomic_t is unnecessary here */
32065 + static unsigned short cnt;
32068 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
32071 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
32072 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
32073 + dentry = ERR_PTR(-ENAMETOOLONG);
32074 + if (unlikely(qs.len > NAME_MAX))
32076 + dentry = ERR_PTR(-ENOMEM);
32077 + name = kmalloc(qs.len + 1, GFP_NOFS);
32078 + if (unlikely(!name))
32082 + /* doubly whiteout-ed */
32083 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
32084 + p = name + AUFS_WH_PFX_LEN * 2;
32085 + memcpy(p, prefix->name, prefix->len);
32086 + p += prefix->len;
32088 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
32091 + for (i = 0; i < 3; i++) {
32092 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
32093 + dentry = au_sio_lkup_one(&qs, h_parent);
32094 + if (IS_ERR(dentry) || d_is_negative(dentry))
32098 + /* pr_warn("could not get random name\n"); */
32099 + dentry = ERR_PTR(-EEXIST);
32100 + AuDbg("%.*s\n", AuLNPair(&qs));
32104 + if (name != defname)
32107 + AuTraceErrPtr(dentry);
32112 + * rename the @h_dentry on @br to the whiteouted temporary name.
32114 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
32117 + struct path h_path = {
32118 + .mnt = au_br_mnt(br)
32120 + struct inode *h_dir, *delegated;
32121 + struct dentry *h_parent;
32123 + h_parent = h_dentry->d_parent; /* dir inode is locked */
32124 + h_dir = d_inode(h_parent);
32125 + IMustLock(h_dir);
32127 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
32128 + err = PTR_ERR(h_path.dentry);
32129 + if (IS_ERR(h_path.dentry))
32132 + /* under the same dir, no need to lock_rename() */
32133 + delegated = NULL;
32134 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated);
32136 + if (unlikely(err == -EWOULDBLOCK)) {
32137 + pr_warn("cannot retry for NFSv4 delegation"
32138 + " for an internal rename\n");
32141 + dput(h_path.dentry);
32148 +/* ---------------------------------------------------------------------- */
32150 + * functions for removing a whiteout
32153 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
32156 + struct inode *delegated;
32159 + * forces superio when the dir has a sticky bit.
32160 + * this may be a violation of unix fs semantics.
32162 + force = (h_dir->i_mode & S_ISVTX)
32163 + && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
32164 + delegated = NULL;
32165 + err = vfsub_unlink(h_dir, h_path, &delegated, force);
32166 + if (unlikely(err == -EWOULDBLOCK)) {
32167 + pr_warn("cannot retry for NFSv4 delegation"
32168 + " for an internal unlink\n");
32174 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
32175 + struct dentry *dentry)
32179 + err = do_unlink_wh(h_dir, h_path);
32180 + if (!err && dentry)
32181 + au_set_dbwh(dentry, -1);
32186 +static int unlink_wh_name(struct dentry *h_parent, struct qstr *wh,
32187 + struct au_branch *br)
32190 + struct path h_path = {
32191 + .mnt = au_br_mnt(br)
32195 + h_path.dentry = vfsub_lkup_one(wh, h_parent);
32196 + if (IS_ERR(h_path.dentry))
32197 + err = PTR_ERR(h_path.dentry);
32199 + if (d_is_reg(h_path.dentry))
32200 + err = do_unlink_wh(d_inode(h_parent), &h_path);
32201 + dput(h_path.dentry);
32207 +/* ---------------------------------------------------------------------- */
32209 + * initialize/clean whiteout for a branch
32212 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
32216 + struct inode *delegated;
32218 + if (d_is_negative(whpath->dentry))
32222 + err = vfsub_rmdir(h_dir, whpath);
32224 + delegated = NULL;
32225 + err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
32226 + if (unlikely(err == -EWOULDBLOCK)) {
32227 + pr_warn("cannot retry for NFSv4 delegation"
32228 + " for an internal unlink\n");
32232 + if (unlikely(err))
32233 + pr_warn("failed removing %pd (%d), ignored.\n",
32234 + whpath->dentry, err);
32237 +static int test_linkable(struct dentry *h_root)
32239 + struct inode *h_dir = d_inode(h_root);
32241 + if (h_dir->i_op->link)
32244 + pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
32245 + h_root, au_sbtype(h_root->d_sb));
32249 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
32250 +static int au_whdir(struct inode *h_dir, struct path *path)
32255 + if (d_is_negative(path->dentry)) {
32256 + int mode = S_IRWXU;
32258 + if (au_test_nfs(path->dentry->d_sb))
32260 + err = vfsub_mkdir(h_dir, path, mode);
32261 + } else if (d_is_dir(path->dentry))
32264 + pr_err("unknown %pd exists\n", path->dentry);
32269 +struct au_wh_base {
32270 + const struct qstr *name;
32271 + struct dentry *dentry;
32274 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
32275 + struct path *h_path)
32277 + h_path->dentry = base[AuBrWh_BASE].dentry;
32278 + au_wh_clean(h_dir, h_path, /*isdir*/0);
32279 + h_path->dentry = base[AuBrWh_PLINK].dentry;
32280 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32281 + h_path->dentry = base[AuBrWh_ORPH].dentry;
32282 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32286 + * returns tri-state,
32287 + * minus: error, caller should print the message
32289 + * plus: error, caller should NOT print the message
32291 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
32292 + int do_plink, struct au_wh_base base[],
32293 + struct path *h_path)
32296 + struct inode *h_dir;
32298 + h_dir = d_inode(h_root);
32299 + h_path->dentry = base[AuBrWh_BASE].dentry;
32300 + au_wh_clean(h_dir, h_path, /*isdir*/0);
32301 + h_path->dentry = base[AuBrWh_PLINK].dentry;
32303 + err = test_linkable(h_root);
32304 + if (unlikely(err)) {
32309 + err = au_whdir(h_dir, h_path);
32310 + if (unlikely(err))
32312 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
32314 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32315 + h_path->dentry = base[AuBrWh_ORPH].dentry;
32316 + err = au_whdir(h_dir, h_path);
32317 + if (unlikely(err))
32319 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
32326 + * for the moment, aufs supports the branch filesystem which does not support
32327 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
32328 + * copyup failed. finally, such filesystem will not be used as the writable
32331 + * returns tri-state, see above.
32333 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
32334 + int do_plink, struct au_wh_base base[],
32335 + struct path *h_path)
32338 + struct inode *h_dir;
32340 + WbrWhMustWriteLock(wbr);
32342 + err = test_linkable(h_root);
32343 + if (unlikely(err)) {
32349 + * todo: should this create be done in /sbin/mount.aufs helper?
32352 + h_dir = d_inode(h_root);
32353 + if (d_is_negative(base[AuBrWh_BASE].dentry)) {
32354 + h_path->dentry = base[AuBrWh_BASE].dentry;
32355 + err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
32356 + } else if (d_is_reg(base[AuBrWh_BASE].dentry))
32359 + pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
32360 + if (unlikely(err))
32363 + h_path->dentry = base[AuBrWh_PLINK].dentry;
32365 + err = au_whdir(h_dir, h_path);
32366 + if (unlikely(err))
32368 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
32370 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32371 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
32373 + h_path->dentry = base[AuBrWh_ORPH].dentry;
32374 + err = au_whdir(h_dir, h_path);
32375 + if (unlikely(err))
32377 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
32384 + * initialize the whiteout base file/dir for @br.
32386 +int au_wh_init(struct au_branch *br, struct super_block *sb)
32389 + const unsigned char do_plink
32390 + = !!au_opt_test(au_mntflags(sb), PLINK);
32391 + struct inode *h_dir;
32392 + struct path path = br->br_path;
32393 + struct dentry *h_root = path.dentry;
32394 + struct au_wbr *wbr = br->br_wbr;
32395 + static const struct qstr base_name[] = {
32396 + [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
32397 + sizeof(AUFS_BASE_NAME) - 1),
32398 + [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
32399 + sizeof(AUFS_PLINKDIR_NAME) - 1),
32400 + [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
32401 + sizeof(AUFS_ORPHDIR_NAME) - 1)
32403 + struct au_wh_base base[] = {
32404 + [AuBrWh_BASE] = {
32405 + .name = base_name + AuBrWh_BASE,
32408 + [AuBrWh_PLINK] = {
32409 + .name = base_name + AuBrWh_PLINK,
32412 + [AuBrWh_ORPH] = {
32413 + .name = base_name + AuBrWh_ORPH,
32419 + WbrWhMustWriteLock(wbr);
32421 + for (i = 0; i < AuBrWh_Last; i++) {
32422 + /* doubly whiteouted */
32423 + struct dentry *d;
32425 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
32426 + err = PTR_ERR(d);
32430 + base[i].dentry = d;
32432 + && wbr->wbr_wh[i]
32433 + && wbr->wbr_wh[i] != base[i].dentry);
32437 + for (i = 0; i < AuBrWh_Last; i++) {
32438 + dput(wbr->wbr_wh[i]);
32439 + wbr->wbr_wh[i] = NULL;
32443 + if (!au_br_writable(br->br_perm)) {
32444 + h_dir = d_inode(h_root);
32445 + au_wh_init_ro(h_dir, base, &path);
32446 + } else if (!au_br_wh_linkable(br->br_perm)) {
32447 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
32453 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
32459 + goto out; /* success */
32462 + pr_err("an error(%d) on the writable branch %pd(%s)\n",
32463 + err, h_root, au_sbtype(h_root->d_sb));
32465 + for (i = 0; i < AuBrWh_Last; i++)
32466 + dput(base[i].dentry);
32470 +/* ---------------------------------------------------------------------- */
32472 + * whiteouts are all hard-linked usually.
32473 + * when its link count reaches a ceiling, we create a new whiteout base
32474 + * asynchronously.
32477 +struct reinit_br_wh {
32478 + struct super_block *sb;
32479 + struct au_branch *br;
32482 +static void reinit_br_wh(void *arg)
32485 + aufs_bindex_t bindex;
32486 + struct path h_path;
32487 + struct reinit_br_wh *a = arg;
32488 + struct au_wbr *wbr;
32489 + struct inode *dir, *delegated;
32490 + struct dentry *h_root;
32491 + struct au_hinode *hdir;
32494 + wbr = a->br->br_wbr;
32495 + /* big aufs lock */
32496 + si_noflush_write_lock(a->sb);
32497 + if (!au_br_writable(a->br->br_perm))
32499 + bindex = au_br_index(a->sb, a->br->br_id);
32500 + if (unlikely(bindex < 0))
32503 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
32504 + dir = d_inode(a->sb->s_root);
32505 + hdir = au_hi(dir, bindex);
32506 + h_root = au_h_dptr(a->sb->s_root, bindex);
32507 + AuDebugOn(h_root != au_br_dentry(a->br));
32509 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
32510 + wbr_wh_write_lock(wbr);
32511 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
32514 + h_path.dentry = wbr->wbr_whbase;
32515 + h_path.mnt = au_br_mnt(a->br);
32516 + delegated = NULL;
32517 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
32519 + if (unlikely(err == -EWOULDBLOCK)) {
32520 + pr_warn("cannot retry for NFSv4 delegation"
32521 + " for an internal unlink\n");
32525 + pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
32528 + dput(wbr->wbr_whbase);
32529 + wbr->wbr_whbase = NULL;
32531 + err = au_wh_init(a->br, a->sb);
32532 + wbr_wh_write_unlock(wbr);
32533 + au_hn_imtx_unlock(hdir);
32534 + di_read_unlock(a->sb->s_root, AuLock_IR);
32536 + au_fhsm_wrote(a->sb, bindex, /*force*/0);
32540 + atomic_dec(&wbr->wbr_wh_running);
32541 + atomic_dec(&a->br->br_count);
32542 + si_write_unlock(a->sb);
32543 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
32545 + if (unlikely(err))
32546 + AuIOErr("err %d\n", err);
32549 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
32551 + int do_dec, wkq_err;
32552 + struct reinit_br_wh *arg;
32555 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
32558 + /* ignore ENOMEM */
32559 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
32562 + * dec(wh_running), kfree(arg) and dec(br_count)
32563 + * in reinit function
32567 + atomic_inc(&br->br_count);
32568 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
32569 + if (unlikely(wkq_err)) {
32570 + atomic_dec(&br->br_wbr->wbr_wh_running);
32571 + atomic_dec(&br->br_count);
32579 + atomic_dec(&br->br_wbr->wbr_wh_running);
32582 +/* ---------------------------------------------------------------------- */
32585 + * create the whiteout @wh.
32587 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
32588 + struct dentry *wh)
32591 + struct path h_path = {
32594 + struct au_branch *br;
32595 + struct au_wbr *wbr;
32596 + struct dentry *h_parent;
32597 + struct inode *h_dir, *delegated;
32599 + h_parent = wh->d_parent; /* dir inode is locked */
32600 + h_dir = d_inode(h_parent);
32601 + IMustLock(h_dir);
32603 + br = au_sbr(sb, bindex);
32604 + h_path.mnt = au_br_mnt(br);
32605 + wbr = br->br_wbr;
32606 + wbr_wh_read_lock(wbr);
32607 + if (wbr->wbr_whbase) {
32608 + delegated = NULL;
32609 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
32610 + if (unlikely(err == -EWOULDBLOCK)) {
32611 + pr_warn("cannot retry for NFSv4 delegation"
32612 + " for an internal link\n");
32615 + if (!err || err != -EMLINK)
32618 + /* link count full. re-initialize br_whbase. */
32619 + kick_reinit_br_wh(sb, br);
32622 + /* return this error in this context */
32623 + err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
32625 + au_fhsm_wrote(sb, bindex, /*force*/0);
32628 + wbr_wh_read_unlock(wbr);
32632 +/* ---------------------------------------------------------------------- */
32635 + * create or remove the diropq.
32637 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
32638 + unsigned int flags)
32640 + struct dentry *opq_dentry, *h_dentry;
32641 + struct super_block *sb;
32642 + struct au_branch *br;
32645 + sb = dentry->d_sb;
32646 + br = au_sbr(sb, bindex);
32647 + h_dentry = au_h_dptr(dentry, bindex);
32648 + opq_dentry = vfsub_lkup_one(&diropq_name, h_dentry);
32649 + if (IS_ERR(opq_dentry))
32652 + if (au_ftest_diropq(flags, CREATE)) {
32653 + err = link_or_create_wh(sb, bindex, opq_dentry);
32655 + au_set_dbdiropq(dentry, bindex);
32656 + goto out; /* success */
32659 + struct path tmp = {
32660 + .dentry = opq_dentry,
32661 + .mnt = au_br_mnt(br)
32663 + err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &tmp);
32665 + au_set_dbdiropq(dentry, -1);
32667 + dput(opq_dentry);
32668 + opq_dentry = ERR_PTR(err);
32671 + return opq_dentry;
32674 +struct do_diropq_args {
32675 + struct dentry **errp;
32676 + struct dentry *dentry;
32677 + aufs_bindex_t bindex;
32678 + unsigned int flags;
32681 +static void call_do_diropq(void *args)
32683 + struct do_diropq_args *a = args;
32684 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
32687 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
32688 + unsigned int flags)
32690 + struct dentry *diropq, *h_dentry;
32692 + h_dentry = au_h_dptr(dentry, bindex);
32693 + if (!au_test_h_perm_sio(d_inode(h_dentry), MAY_EXEC | MAY_WRITE))
32694 + diropq = do_diropq(dentry, bindex, flags);
32697 + struct do_diropq_args args = {
32699 + .dentry = dentry,
32700 + .bindex = bindex,
32704 + wkq_err = au_wkq_wait(call_do_diropq, &args);
32705 + if (unlikely(wkq_err))
32706 + diropq = ERR_PTR(wkq_err);
32712 +/* ---------------------------------------------------------------------- */
32715 + * lookup whiteout dentry.
32716 + * @h_parent: lower parent dentry which must exist and be locked
32717 + * @base_name: name of dentry which will be whiteouted
32718 + * returns dentry for whiteout.
32720 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
32721 + struct au_branch *br)
32724 + struct qstr wh_name;
32725 + struct dentry *wh_dentry;
32727 + err = au_wh_name_alloc(&wh_name, base_name);
32728 + wh_dentry = ERR_PTR(err);
32730 + wh_dentry = vfsub_lkup_one(&wh_name, h_parent);
32731 + kfree(wh_name.name);
32733 + return wh_dentry;
32737 + * link/create a whiteout for @dentry on @bindex.
32739 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
32740 + struct dentry *h_parent)
32742 + struct dentry *wh_dentry;
32743 + struct super_block *sb;
32746 + sb = dentry->d_sb;
32747 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
32748 + if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
32749 + err = link_or_create_wh(sb, bindex, wh_dentry);
32751 + au_set_dbwh(dentry, bindex);
32752 + au_fhsm_wrote(sb, bindex, /*force*/0);
32755 + wh_dentry = ERR_PTR(err);
32759 + return wh_dentry;
32762 +/* ---------------------------------------------------------------------- */
32764 +/* Delete all whiteouts in this directory on branch bindex. */
32765 +static int del_wh_children(struct dentry *h_dentry, struct au_nhash *whlist,
32766 + aufs_bindex_t bindex, struct au_branch *br)
32769 + unsigned long ul, n;
32770 + struct qstr wh_name;
32772 + struct hlist_head *head;
32773 + struct au_vdir_wh *pos;
32774 + struct au_vdir_destr *str;
32777 + p = (void *)__get_free_page(GFP_NOFS);
32778 + wh_name.name = p;
32779 + if (unlikely(!wh_name.name))
32783 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32784 + p += AUFS_WH_PFX_LEN;
32785 + n = whlist->nh_num;
32786 + head = whlist->nh_head;
32787 + for (ul = 0; !err && ul < n; ul++, head++) {
32788 + hlist_for_each_entry(pos, head, wh_hash) {
32789 + if (pos->wh_bindex != bindex)
32792 + str = &pos->wh_str;
32793 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
32794 + memcpy(p, str->name, str->len);
32795 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
32796 + err = unlink_wh_name(h_dentry, &wh_name, br);
32801 + AuIOErr("whiteout name too long %.*s\n",
32802 + str->len, str->name);
32807 + free_page((unsigned long)wh_name.name);
32813 +struct del_wh_children_args {
32815 + struct dentry *h_dentry;
32816 + struct au_nhash *whlist;
32817 + aufs_bindex_t bindex;
32818 + struct au_branch *br;
32821 +static void call_del_wh_children(void *args)
32823 + struct del_wh_children_args *a = args;
32824 + *a->errp = del_wh_children(a->h_dentry, a->whlist, a->bindex, a->br);
32827 +/* ---------------------------------------------------------------------- */
32829 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
32831 + struct au_whtmp_rmdir *whtmp;
32833 + unsigned int rdhash;
32835 + SiMustAnyLock(sb);
32837 + whtmp = kzalloc(sizeof(*whtmp), gfp);
32838 + if (unlikely(!whtmp)) {
32839 + whtmp = ERR_PTR(-ENOMEM);
32843 + /* no estimation for dir size */
32844 + rdhash = au_sbi(sb)->si_rdhash;
32846 + rdhash = AUFS_RDHASH_DEF;
32847 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
32848 + if (unlikely(err)) {
32850 + whtmp = ERR_PTR(err);
32857 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
32860 + atomic_dec(&whtmp->br->br_count);
32861 + dput(whtmp->wh_dentry);
32862 + iput(whtmp->dir);
32863 + au_nhash_wh_free(&whtmp->whlist);
32868 + * rmdir the whiteouted temporary named dir @h_dentry.
32869 + * @whlist: whiteouted children.
32871 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
32872 + struct dentry *wh_dentry, struct au_nhash *whlist)
32875 + unsigned int h_nlink;
32876 + struct path h_tmp;
32877 + struct inode *wh_inode, *h_dir;
32878 + struct au_branch *br;
32880 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
32881 + IMustLock(h_dir);
32883 + br = au_sbr(dir->i_sb, bindex);
32884 + wh_inode = d_inode(wh_dentry);
32885 + mutex_lock_nested(&wh_inode->i_mutex, AuLsc_I_CHILD);
32888 + * someone else might change some whiteouts while we were sleeping.
32889 + * it means this whlist may have an obsoleted entry.
32891 + if (!au_test_h_perm_sio(wh_inode, MAY_EXEC | MAY_WRITE))
32892 + err = del_wh_children(wh_dentry, whlist, bindex, br);
32895 + struct del_wh_children_args args = {
32897 + .h_dentry = wh_dentry,
32898 + .whlist = whlist,
32899 + .bindex = bindex,
32903 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
32904 + if (unlikely(wkq_err))
32907 + mutex_unlock(&wh_inode->i_mutex);
32910 + h_tmp.dentry = wh_dentry;
32911 + h_tmp.mnt = au_br_mnt(br);
32912 + h_nlink = h_dir->i_nlink;
32913 + err = vfsub_rmdir(h_dir, &h_tmp);
32914 + /* some fs doesn't change the parent nlink in some cases */
32915 + h_nlink -= h_dir->i_nlink;
32919 + if (au_ibstart(dir) == bindex) {
32920 + /* todo: dir->i_mutex is necessary */
32921 + au_cpup_attr_timesizes(dir);
32923 + vfsub_drop_nlink(dir);
32925 + return 0; /* success */
32928 + pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
32932 +static void call_rmdir_whtmp(void *args)
32935 + aufs_bindex_t bindex;
32936 + struct au_whtmp_rmdir *a = args;
32937 + struct super_block *sb;
32938 + struct dentry *h_parent;
32939 + struct inode *h_dir;
32940 + struct au_hinode *hdir;
32942 + /* rmdir by nfsd may cause deadlock with this i_mutex */
32943 + /* mutex_lock(&a->dir->i_mutex); */
32945 + sb = a->dir->i_sb;
32946 + si_read_lock(sb, !AuLock_FLUSH);
32947 + if (!au_br_writable(a->br->br_perm))
32949 + bindex = au_br_index(sb, a->br->br_id);
32950 + if (unlikely(bindex < 0))
32954 + ii_write_lock_parent(a->dir);
32955 + h_parent = dget_parent(a->wh_dentry);
32956 + h_dir = d_inode(h_parent);
32957 + hdir = au_hi(a->dir, bindex);
32958 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
32959 + if (unlikely(err))
32961 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
32962 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
32965 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
32966 + au_hn_imtx_unlock(hdir);
32967 + vfsub_mnt_drop_write(au_br_mnt(a->br));
32971 + ii_write_unlock(a->dir);
32973 + /* mutex_unlock(&a->dir->i_mutex); */
32974 + au_whtmp_rmdir_free(a);
32975 + si_read_unlock(sb);
32976 + au_nwt_done(&au_sbi(sb)->si_nowait);
32977 + if (unlikely(err))
32978 + AuIOErr("err %d\n", err);
32981 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
32982 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
32985 + struct super_block *sb;
32989 + /* all post-process will be done in do_rmdir_whtmp(). */
32991 + args->dir = au_igrab(dir);
32992 + args->br = au_sbr(sb, bindex);
32993 + atomic_inc(&args->br->br_count);
32994 + args->wh_dentry = dget(wh_dentry);
32995 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
32996 + if (unlikely(wkq_err)) {
32997 + pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
32998 + au_whtmp_rmdir_free(args);
33001 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
33002 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
33003 +++ linux/fs/aufs/whout.h 2016-01-13 20:11:11.673093671 +0100
33006 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33008 + * This program, aufs is free software; you can redistribute it and/or modify
33009 + * it under the terms of the GNU General Public License as published by
33010 + * the Free Software Foundation; either version 2 of the License, or
33011 + * (at your option) any later version.
33013 + * This program is distributed in the hope that it will be useful,
33014 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33015 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33016 + * GNU General Public License for more details.
33018 + * You should have received a copy of the GNU General Public License
33019 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33023 + * whiteout for logical deletion and opaque directory
33026 +#ifndef __AUFS_WHOUT_H__
33027 +#define __AUFS_WHOUT_H__
33034 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
33035 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio);
33036 +int au_diropq_test(struct dentry *h_dentry);
33038 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
33039 + struct qstr *prefix);
33040 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
33041 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
33042 + struct dentry *dentry);
33043 +int au_wh_init(struct au_branch *br, struct super_block *sb);
33045 +/* diropq flags */
33046 +#define AuDiropq_CREATE 1
33047 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
33048 +#define au_fset_diropq(flags, name) \
33049 + do { (flags) |= AuDiropq_##name; } while (0)
33050 +#define au_fclr_diropq(flags, name) \
33051 + do { (flags) &= ~AuDiropq_##name; } while (0)
33053 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
33054 + unsigned int flags);
33055 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
33056 + struct au_branch *br);
33057 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
33058 + struct dentry *h_parent);
33060 +/* real rmdir for the whiteout-ed dir */
33061 +struct au_whtmp_rmdir {
33062 + struct inode *dir;
33063 + struct au_branch *br;
33064 + struct dentry *wh_dentry;
33065 + struct au_nhash whlist;
33068 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
33069 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
33070 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
33071 + struct dentry *wh_dentry, struct au_nhash *whlist);
33072 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
33073 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
33075 +/* ---------------------------------------------------------------------- */
33077 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
33078 + aufs_bindex_t bindex)
33080 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
33083 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
33085 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
33088 +#endif /* __KERNEL__ */
33089 +#endif /* __AUFS_WHOUT_H__ */
33090 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
33091 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
33092 +++ linux/fs/aufs/wkq.c 2016-01-13 20:11:11.673093671 +0100
33095 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33097 + * This program, aufs is free software; you can redistribute it and/or modify
33098 + * it under the terms of the GNU General Public License as published by
33099 + * the Free Software Foundation; either version 2 of the License, or
33100 + * (at your option) any later version.
33102 + * This program is distributed in the hope that it will be useful,
33103 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33104 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33105 + * GNU General Public License for more details.
33107 + * You should have received a copy of the GNU General Public License
33108 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33112 + * workqueue for asynchronous/super-io operations
33113 + * todo: try new dredential scheme
33116 +#include <linux/module.h>
33119 +/* internal workqueue named AUFS_WKQ_NAME */
33121 +static struct workqueue_struct *au_wkq;
33123 +struct au_wkinfo {
33124 + struct work_struct wk;
33125 + struct kobject *kobj;
33127 + unsigned int flags; /* see wkq.h */
33129 + au_wkq_func_t func;
33132 + struct completion *comp;
33135 +/* ---------------------------------------------------------------------- */
33137 +static void wkq_func(struct work_struct *wk)
33139 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
33141 + AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
33142 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
33144 + wkinfo->func(wkinfo->args);
33145 + if (au_ftest_wkq(wkinfo->flags, WAIT))
33146 + complete(wkinfo->comp);
33148 + kobject_put(wkinfo->kobj);
33149 + module_put(THIS_MODULE); /* todo: ?? */
33155 + * Since struct completion is large, try allocating it dynamically.
33157 +#if 1 /* defined(CONFIG_4KSTACKS) || defined(AuTest4KSTACKS) */
33158 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
33160 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
33162 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
33164 + init_completion(*comp);
33165 + wkinfo->comp = *comp;
33171 +static void au_wkq_comp_free(struct completion *comp)
33179 +#define AuWkqCompDeclare(name) \
33180 + DECLARE_COMPLETION_ONSTACK(_ ## name); \
33181 + struct completion *comp = &_ ## name
33183 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
33185 + wkinfo->comp = *comp;
33189 +static void au_wkq_comp_free(struct completion *comp __maybe_unused)
33193 +#endif /* 4KSTACKS */
33195 +static void au_wkq_run(struct au_wkinfo *wkinfo)
33197 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
33198 + if (au_wkq_test()) {
33199 + AuWarn1("wkq from wkq, unless silly-rename on NFS,"
33200 + " due to a dead dir by UDBA?\n");
33201 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
33204 + au_dbg_verify_kthread();
33206 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
33207 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
33208 + queue_work(au_wkq, &wkinfo->wk);
33210 + INIT_WORK(&wkinfo->wk, wkq_func);
33211 + schedule_work(&wkinfo->wk);
33216 + * Be careful. It is easy to make deadlock happen.
33217 + * processA: lock, wkq and wait
33218 + * processB: wkq and wait, lock in wkq
33221 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
33224 + AuWkqCompDeclare(comp);
33225 + struct au_wkinfo wkinfo = {
33231 + err = au_wkq_comp_alloc(&wkinfo, &comp);
33233 + au_wkq_run(&wkinfo);
33234 + /* no timeout, no interrupt */
33235 + wait_for_completion(wkinfo.comp);
33236 + au_wkq_comp_free(comp);
33237 + destroy_work_on_stack(&wkinfo.wk);
33245 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
33246 + * problem in a concurrent umounting.
33248 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
33249 + unsigned int flags)
33252 + struct au_wkinfo *wkinfo;
33254 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
33257 + * wkq_func() must free this wkinfo.
33258 + * it highly depends upon the implementation of workqueue.
33261 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
33263 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
33264 + wkinfo->flags = flags & ~AuWkq_WAIT;
33265 + wkinfo->func = func;
33266 + wkinfo->args = args;
33267 + wkinfo->comp = NULL;
33268 + kobject_get(wkinfo->kobj);
33269 + __module_get(THIS_MODULE); /* todo: ?? */
33271 + au_wkq_run(wkinfo);
33274 + au_nwt_done(&au_sbi(sb)->si_nowait);
33280 +/* ---------------------------------------------------------------------- */
33282 +void au_nwt_init(struct au_nowait_tasks *nwt)
33284 + atomic_set(&nwt->nw_len, 0);
33285 + /* smp_mb(); */ /* atomic_set */
33286 + init_waitqueue_head(&nwt->nw_wq);
33289 +void au_wkq_fin(void)
33291 + destroy_workqueue(au_wkq);
33294 +int __init au_wkq_init(void)
33299 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
33300 + if (IS_ERR(au_wkq))
33301 + err = PTR_ERR(au_wkq);
33302 + else if (!au_wkq)
33307 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
33308 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
33309 +++ linux/fs/aufs/wkq.h 2016-01-13 20:11:11.673093671 +0100
33312 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33314 + * This program, aufs is free software; you can redistribute it and/or modify
33315 + * it under the terms of the GNU General Public License as published by
33316 + * the Free Software Foundation; either version 2 of the License, or
33317 + * (at your option) any later version.
33319 + * This program is distributed in the hope that it will be useful,
33320 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33321 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33322 + * GNU General Public License for more details.
33324 + * You should have received a copy of the GNU General Public License
33325 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33329 + * workqueue for asynchronous/super-io operations
33330 + * todo: try new credentials management scheme
33333 +#ifndef __AUFS_WKQ_H__
33334 +#define __AUFS_WKQ_H__
33338 +struct super_block;
33340 +/* ---------------------------------------------------------------------- */
33343 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
33345 +struct au_nowait_tasks {
33347 + wait_queue_head_t nw_wq;
33350 +/* ---------------------------------------------------------------------- */
33352 +typedef void (*au_wkq_func_t)(void *args);
33355 +#define AuWkq_WAIT 1
33356 +#define AuWkq_NEST (1 << 1)
33357 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
33358 +#define au_fset_wkq(flags, name) \
33359 + do { (flags) |= AuWkq_##name; } while (0)
33360 +#define au_fclr_wkq(flags, name) \
33361 + do { (flags) &= ~AuWkq_##name; } while (0)
33363 +#ifndef CONFIG_AUFS_HNOTIFY
33365 +#define AuWkq_NEST 0
33369 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
33370 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
33371 + unsigned int flags);
33372 +void au_nwt_init(struct au_nowait_tasks *nwt);
33373 +int __init au_wkq_init(void);
33374 +void au_wkq_fin(void);
33376 +/* ---------------------------------------------------------------------- */
33378 +static inline int au_wkq_test(void)
33380 + return current->flags & PF_WQ_WORKER;
33383 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
33385 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
33388 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
33390 + if (atomic_dec_and_test(&nwt->nw_len))
33391 + wake_up_all(&nwt->nw_wq);
33394 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
33396 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
33400 +#endif /* __KERNEL__ */
33401 +#endif /* __AUFS_WKQ_H__ */
33402 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
33403 --- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
33404 +++ linux/fs/aufs/xattr.c 2016-01-13 20:11:11.673093671 +0100
33407 + * Copyright (C) 2014-2015 Junjiro R. Okajima
33409 + * This program, aufs is free software; you can redistribute it and/or modify
33410 + * it under the terms of the GNU General Public License as published by
33411 + * the Free Software Foundation; either version 2 of the License, or
33412 + * (at your option) any later version.
33414 + * This program is distributed in the hope that it will be useful,
33415 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33416 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33417 + * GNU General Public License for more details.
33419 + * You should have received a copy of the GNU General Public License
33420 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33424 + * handling xattr functions
33427 +#include <linux/xattr.h>
33430 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
33432 + if (!ignore_flags)
33440 + if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
33445 +#define cmp(brattr, prefix) do { \
33446 + if (!strncmp(name, XATTR_##prefix##_PREFIX, \
33447 + XATTR_##prefix##_PREFIX_LEN)) { \
33448 + if (ignore_flags & AuBrAttr_ICEX_##brattr) \
33454 + cmp(SEC, SECURITY);
33455 + cmp(SYS, SYSTEM);
33456 + cmp(TR, TRUSTED);
33460 + if (ignore_flags & AuBrAttr_ICEX_OTH)
33467 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
33469 +static int au_do_cpup_xattr(struct dentry *h_dst, struct dentry *h_src,
33470 + char *name, char **buf, unsigned int ignore_flags,
33471 + unsigned int verbose)
33475 + struct inode *h_idst;
33477 + ssz = vfs_getxattr_alloc(h_src, name, buf, 0, GFP_NOFS);
33479 + if (unlikely(err <= 0)) {
33480 + if (err == -ENODATA
33481 + || (err == -EOPNOTSUPP
33482 + && ((ignore_flags & au_xattr_out_of_list)
33483 + || (au_test_nfs_noacl(d_inode(h_src))
33484 + && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
33486 + XATTR_NAME_POSIX_ACL_DEFAULT))))
33489 + if (err && (verbose || au_debug_test()))
33490 + pr_err("%s, err %d\n", name, err);
33494 + /* unlock it temporary */
33495 + h_idst = d_inode(h_dst);
33496 + mutex_unlock(&h_idst->i_mutex);
33497 + err = vfsub_setxattr(h_dst, name, *buf, ssz, /*flags*/0);
33498 + mutex_lock_nested(&h_idst->i_mutex, AuLsc_I_CHILD2);
33499 + if (unlikely(err)) {
33500 + if (verbose || au_debug_test())
33501 + pr_err("%s, err %d\n", name, err);
33502 + err = au_xattr_ignore(err, name, ignore_flags);
33509 +int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
33510 + unsigned int verbose)
33512 + int err, unlocked, acl_access, acl_default;
33514 + struct inode *h_isrc, *h_idst;
33515 + char *value, *p, *o, *e;
33517 + /* try stopping to update the source inode while we are referencing */
33518 + /* there should not be the parent-child relationship between them */
33519 + h_isrc = d_inode(h_src);
33520 + h_idst = d_inode(h_dst);
33521 + mutex_unlock(&h_idst->i_mutex);
33522 + mutex_lock_nested(&h_isrc->i_mutex, AuLsc_I_CHILD);
33523 + mutex_lock_nested(&h_idst->i_mutex, AuLsc_I_CHILD2);
33526 + /* some filesystems don't list POSIX ACL, for example tmpfs */
33527 + ssz = vfs_listxattr(h_src, NULL, 0);
33529 + if (unlikely(err < 0)) {
33531 + if (err == -ENODATA
33532 + || err == -EOPNOTSUPP)
33533 + err = 0; /* ignore */
33542 + p = kmalloc(ssz, GFP_NOFS);
33544 + if (unlikely(!p))
33546 + err = vfs_listxattr(h_src, p, ssz);
33548 + mutex_unlock(&h_isrc->i_mutex);
33550 + AuDbg("err %d, ssz %zd\n", err, ssz);
33551 + if (unlikely(err < 0))
33559 + while (!err && p < e) {
33560 + acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
33561 + sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
33562 + acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
33563 + sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
33565 + err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
33567 + p += strlen(p) + 1;
33570 + ignore_flags |= au_xattr_out_of_list;
33571 + if (!err && !acl_access) {
33572 + err = au_do_cpup_xattr(h_dst, h_src,
33573 + XATTR_NAME_POSIX_ACL_ACCESS, &value,
33574 + ignore_flags, verbose);
33577 + if (!err && !acl_default) {
33578 + err = au_do_cpup_xattr(h_dst, h_src,
33579 + XATTR_NAME_POSIX_ACL_DEFAULT, &value,
33580 + ignore_flags, verbose);
33590 + mutex_unlock(&h_isrc->i_mutex);
33595 +/* ---------------------------------------------------------------------- */
33602 +struct au_lgxattr {
33610 + const char *name;
33617 +static ssize_t au_lgxattr(struct dentry *dentry, struct au_lgxattr *arg)
33620 + struct path h_path;
33621 + struct super_block *sb;
33623 + sb = dentry->d_sb;
33624 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
33625 + if (unlikely(err))
33627 + err = au_h_path_getattr(dentry, /*force*/1, &h_path);
33628 + if (unlikely(err))
33630 + if (unlikely(!h_path.dentry))
33631 + /* illegally overlapped or something */
33632 + goto out_di; /* pretending success */
33634 + /* always topmost entry only */
33635 + switch (arg->type) {
33636 + case AU_XATTR_LIST:
33637 + err = vfs_listxattr(h_path.dentry,
33638 + arg->u.list.list, arg->u.list.size);
33640 + case AU_XATTR_GET:
33641 + err = vfs_getxattr(h_path.dentry,
33642 + arg->u.get.name, arg->u.get.value,
33643 + arg->u.get.size);
33648 + di_read_unlock(dentry, AuLock_IR);
33650 + si_read_unlock(sb);
33656 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
33658 + struct au_lgxattr arg = {
33659 + .type = AU_XATTR_LIST,
33666 + return au_lgxattr(dentry, &arg);
33669 +ssize_t aufs_getxattr(struct dentry *dentry, const char *name, void *value,
33672 + struct au_lgxattr arg = {
33673 + .type = AU_XATTR_GET,
33681 + return au_lgxattr(dentry, &arg);
33684 +int aufs_setxattr(struct dentry *dentry, const char *name, const void *value,
33685 + size_t size, int flags)
33687 + struct au_srxattr arg = {
33688 + .type = AU_XATTR_SET,
33697 + return au_srxattr(dentry, &arg);
33700 +int aufs_removexattr(struct dentry *dentry, const char *name)
33702 + struct au_srxattr arg = {
33703 + .type = AU_XATTR_REMOVE,
33709 + return au_srxattr(dentry, &arg);
33712 +/* ---------------------------------------------------------------------- */
33715 +static size_t au_xattr_list(struct dentry *dentry, char *list, size_t list_size,
33716 + const char *name, size_t name_len, int type)
33718 + return aufs_listxattr(dentry, list, list_size);
33721 +static int au_xattr_get(struct dentry *dentry, const char *name, void *buffer,
33722 + size_t size, int type)
33724 + return aufs_getxattr(dentry, name, buffer, size);
33727 +static int au_xattr_set(struct dentry *dentry, const char *name,
33728 + const void *value, size_t size, int flags, int type)
33730 + return aufs_setxattr(dentry, name, value, size, flags);
33733 +static const struct xattr_handler au_xattr_handler = {
33734 + /* no prefix, no flags */
33735 + .list = au_xattr_list,
33736 + .get = au_xattr_get,
33737 + .set = au_xattr_set
33738 + /* why no remove? */
33741 +static const struct xattr_handler *au_xattr_handlers[] = {
33742 + &au_xattr_handler
33745 +void au_xattr_init(struct super_block *sb)
33747 + /* sb->s_xattr = au_xattr_handlers; */
33750 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
33751 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
33752 +++ linux/fs/aufs/xino.c 2016-01-13 20:11:11.673093671 +0100
33755 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33757 + * This program, aufs is free software; you can redistribute it and/or modify
33758 + * it under the terms of the GNU General Public License as published by
33759 + * the Free Software Foundation; either version 2 of the License, or
33760 + * (at your option) any later version.
33762 + * This program is distributed in the hope that it will be useful,
33763 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33764 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33765 + * GNU General Public License for more details.
33767 + * You should have received a copy of the GNU General Public License
33768 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33772 + * external inode number translation table and bitmap
33775 +#include <linux/seq_file.h>
33776 +#include <linux/statfs.h>
33779 +/* todo: unnecessary to support mmap_sem since kernel-space? */
33780 +ssize_t xino_fread(vfs_readf_t func, struct file *file, void *kbuf, size_t size,
33784 + mm_segment_t oldfs;
33791 + oldfs = get_fs();
33792 + set_fs(KERNEL_DS);
33794 + /* todo: signal_pending? */
33795 + err = func(file, buf.u, size, pos);
33796 + } while (err == -EAGAIN || err == -EINTR);
33799 +#if 0 /* reserved for future use */
33801 + fsnotify_access(file->f_path.dentry);
33807 +/* ---------------------------------------------------------------------- */
33809 +static ssize_t xino_fwrite_wkq(vfs_writef_t func, struct file *file, void *buf,
33810 + size_t size, loff_t *pos);
33812 +static ssize_t do_xino_fwrite(vfs_writef_t func, struct file *file, void *kbuf,
33813 + size_t size, loff_t *pos)
33816 + mm_segment_t oldfs;
33819 + const char __user *u;
33822 + const int prevent_endless = 10;
33826 + oldfs = get_fs();
33827 + set_fs(KERNEL_DS);
33829 + err = func(file, buf.u, size, pos);
33830 + if (err == -EINTR
33831 + && !au_wkq_test()
33832 + && fatal_signal_pending(current)) {
33834 + err = xino_fwrite_wkq(func, file, kbuf, size, pos);
33835 + BUG_ON(err == -EINTR);
33836 + oldfs = get_fs();
33837 + set_fs(KERNEL_DS);
33839 + } while (i++ < prevent_endless
33840 + && (err == -EAGAIN || err == -EINTR));
33843 +#if 0 /* reserved for future use */
33845 + fsnotify_modify(file->f_path.dentry);
33851 +struct do_xino_fwrite_args {
33853 + vfs_writef_t func;
33854 + struct file *file;
33860 +static void call_do_xino_fwrite(void *args)
33862 + struct do_xino_fwrite_args *a = args;
33863 + *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos);
33866 +static ssize_t xino_fwrite_wkq(vfs_writef_t func, struct file *file, void *buf,
33867 + size_t size, loff_t *pos)
33871 + struct do_xino_fwrite_args args = {
33881 + * it breaks RLIMIT_FSIZE and normal user's limit,
33882 + * users should care about quota and real 'filesystem full.'
33884 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
33885 + if (unlikely(wkq_err))
33891 +ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
33892 + size_t size, loff_t *pos)
33896 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
33898 + err = do_xino_fwrite(func, file, buf, size, pos);
33901 + err = xino_fwrite_wkq(func, file, buf, size, pos);
33906 +/* ---------------------------------------------------------------------- */
33909 + * create a new xinofile at the same place/path as @base_file.
33911 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src)
33913 + struct file *file;
33914 + struct dentry *base, *parent;
33915 + struct inode *dir, *delegated;
33916 + struct qstr *name;
33917 + struct path path;
33920 + base = base_file->f_path.dentry;
33921 + parent = base->d_parent; /* dir inode is locked */
33922 + dir = d_inode(parent);
33925 + file = ERR_PTR(-EINVAL);
33926 + name = &base->d_name;
33927 + path.dentry = vfsub_lookup_one_len(name->name, parent, name->len);
33928 + if (IS_ERR(path.dentry)) {
33929 + file = (void *)path.dentry;
33930 + pr_err("%pd lookup err %ld\n",
33931 + base, PTR_ERR(path.dentry));
33935 + /* no need to mnt_want_write() since we call dentry_open() later */
33936 + err = vfs_create(dir, path.dentry, S_IRUGO | S_IWUGO, NULL);
33937 + if (unlikely(err)) {
33938 + file = ERR_PTR(err);
33939 + pr_err("%pd create err %d\n", base, err);
33943 + path.mnt = base_file->f_path.mnt;
33944 + file = vfsub_dentry_open(&path,
33945 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
33946 + /* | __FMODE_NONOTIFY */);
33947 + if (IS_ERR(file)) {
33948 + pr_err("%pd open err %ld\n", base, PTR_ERR(file));
33952 + delegated = NULL;
33953 + err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
33954 + if (unlikely(err == -EWOULDBLOCK)) {
33955 + pr_warn("cannot retry for NFSv4 delegation"
33956 + " for an internal unlink\n");
33959 + if (unlikely(err)) {
33960 + pr_err("%pd unlink err %d\n", base, err);
33965 + /* no one can touch copy_src xino */
33966 + err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
33967 + if (unlikely(err)) {
33968 + pr_err("%pd copy err %d\n", base, err);
33972 + goto out_dput; /* success */
33976 + file = ERR_PTR(err);
33978 + dput(path.dentry);
33983 +struct au_xino_lock_dir {
33984 + struct au_hinode *hdir;
33985 + struct dentry *parent;
33986 + struct mutex *mtx;
33989 +static void au_xino_lock_dir(struct super_block *sb, struct file *xino,
33990 + struct au_xino_lock_dir *ldir)
33992 + aufs_bindex_t brid, bindex;
33994 + ldir->hdir = NULL;
33996 + brid = au_xino_brid(sb);
33998 + bindex = au_br_index(sb, brid);
33999 + if (bindex >= 0) {
34000 + ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
34001 + au_hn_imtx_lock_nested(ldir->hdir, AuLsc_I_PARENT);
34003 + ldir->parent = dget_parent(xino->f_path.dentry);
34004 + ldir->mtx = &d_inode(ldir->parent)->i_mutex;
34005 + mutex_lock_nested(ldir->mtx, AuLsc_I_PARENT);
34009 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
34012 + au_hn_imtx_unlock(ldir->hdir);
34014 + mutex_unlock(ldir->mtx);
34015 + dput(ldir->parent);
34019 +/* ---------------------------------------------------------------------- */
34021 +/* trucate xino files asynchronously */
34023 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex)
34026 + unsigned long jiffy;
34028 + aufs_bindex_t bi, bend;
34029 + struct kstatfs *st;
34030 + struct au_branch *br;
34031 + struct file *new_xino, *file;
34032 + struct super_block *h_sb;
34033 + struct au_xino_lock_dir ldir;
34036 + st = kmalloc(sizeof(*st), GFP_NOFS);
34037 + if (unlikely(!st))
34041 + bend = au_sbend(sb);
34042 + if (unlikely(bindex < 0 || bend < bindex))
34044 + br = au_sbr(sb, bindex);
34045 + file = br->br_xino.xi_file;
34049 + err = vfs_statfs(&file->f_path, st);
34050 + if (unlikely(err))
34051 + AuErr1("statfs err %d, ignored\n", err);
34053 + blocks = file_inode(file)->i_blocks;
34054 + pr_info("begin truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
34055 + bindex, (u64)blocks, st->f_bfree, st->f_blocks);
34057 + au_xino_lock_dir(sb, file, &ldir);
34058 + /* mnt_want_write() is unnecessary here */
34059 + new_xino = au_xino_create2(file, file);
34060 + au_xino_unlock_dir(&ldir);
34061 + err = PTR_ERR(new_xino);
34062 + if (IS_ERR(new_xino)) {
34063 + pr_err("err %d, ignored\n", err);
34068 + br->br_xino.xi_file = new_xino;
34070 + h_sb = au_br_sb(br);
34071 + for (bi = 0; bi <= bend; bi++) {
34072 + if (unlikely(bi == bindex))
34074 + br = au_sbr(sb, bi);
34075 + if (au_br_sb(br) != h_sb)
34078 + fput(br->br_xino.xi_file);
34079 + br->br_xino.xi_file = new_xino;
34080 + get_file(new_xino);
34083 + err = vfs_statfs(&new_xino->f_path, st);
34085 + pr_info("end truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
34086 + bindex, (u64)file_inode(new_xino)->i_blocks,
34087 + st->f_bfree, st->f_blocks);
34088 + if (file_inode(new_xino)->i_blocks < blocks)
34089 + au_sbi(sb)->si_xino_jiffy = jiffy;
34091 + AuErr1("statfs err %d, ignored\n", err);
34099 +struct xino_do_trunc_args {
34100 + struct super_block *sb;
34101 + struct au_branch *br;
34104 +static void xino_do_trunc(void *_args)
34106 + struct xino_do_trunc_args *args = _args;
34107 + struct super_block *sb;
34108 + struct au_branch *br;
34109 + struct inode *dir;
34111 + aufs_bindex_t bindex;
34115 + dir = d_inode(sb->s_root);
34118 + si_noflush_write_lock(sb);
34119 + ii_read_lock_parent(dir);
34120 + bindex = au_br_index(sb, br->br_id);
34121 + err = au_xino_trunc(sb, bindex);
34122 + ii_read_unlock(dir);
34123 + if (unlikely(err))
34124 + pr_warn("err b%d, (%d)\n", bindex, err);
34125 + atomic_dec(&br->br_xino_running);
34126 + atomic_dec(&br->br_count);
34127 + si_write_unlock(sb);
34128 + au_nwt_done(&au_sbi(sb)->si_nowait);
34132 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
34135 + struct kstatfs st;
34136 + struct au_sbinfo *sbinfo;
34138 + /* todo: si_xino_expire and the ratio should be customizable */
34139 + sbinfo = au_sbi(sb);
34140 + if (time_before(jiffies,
34141 + sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
34144 + /* truncation border */
34145 + err = vfs_statfs(&br->br_xino.xi_file->f_path, &st);
34146 + if (unlikely(err)) {
34147 + AuErr1("statfs err %d, ignored\n", err);
34150 + if (div64_u64(st.f_bfree * 100, st.f_blocks) >= AUFS_XINO_DEF_TRUNC)
34156 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
34158 + struct xino_do_trunc_args *args;
34161 + if (!xino_trunc_test(sb, br))
34164 + if (atomic_inc_return(&br->br_xino_running) > 1)
34167 + /* lock and kfree() will be called in trunc_xino() */
34168 + args = kmalloc(sizeof(*args), GFP_NOFS);
34169 + if (unlikely(!args)) {
34170 + AuErr1("no memory\n");
34174 + atomic_inc(&br->br_count);
34177 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
34179 + return; /* success */
34181 + pr_err("wkq %d\n", wkq_err);
34182 + atomic_dec(&br->br_count);
34187 + atomic_dec(&br->br_xino_running);
34190 +/* ---------------------------------------------------------------------- */
34192 +static int au_xino_do_write(vfs_writef_t write, struct file *file,
34193 + ino_t h_ino, ino_t ino)
34199 + if (unlikely(au_loff_max / sizeof(ino) - 1 < pos)) {
34200 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
34203 + pos *= sizeof(ino);
34204 + sz = xino_fwrite(write, file, &ino, sizeof(ino), &pos);
34205 + if (sz == sizeof(ino))
34206 + return 0; /* success */
34208 + AuIOErr("write failed (%zd)\n", sz);
34213 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
34214 + * at the position of @h_ino.
34215 + * even if @ino is zero, it is written to the xinofile and means no entry.
34216 + * if the size of the xino file on a specific filesystem exceeds the watermark,
34217 + * try truncating it.
34219 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
34223 + unsigned int mnt_flags;
34224 + struct au_branch *br;
34226 + BUILD_BUG_ON(sizeof(long long) != sizeof(au_loff_max)
34227 + || ((loff_t)-1) > 0);
34228 + SiMustAnyLock(sb);
34230 + mnt_flags = au_mntflags(sb);
34231 + if (!au_opt_test(mnt_flags, XINO))
34234 + br = au_sbr(sb, bindex);
34235 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
34238 + if (au_opt_test(mnt_flags, TRUNC_XINO)
34239 + && au_test_fs_trunc_xino(au_br_sb(br)))
34240 + xino_try_trunc(sb, br);
34241 + return 0; /* success */
34244 + AuIOErr("write failed (%d)\n", err);
34248 +/* ---------------------------------------------------------------------- */
34250 +/* aufs inode number bitmap */
34252 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
34253 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
34257 + AuDebugOn(bit < 0 || page_bits <= bit);
34258 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
34262 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
34264 + AuDebugOn(ino < AUFS_FIRST_INO);
34265 + ino -= AUFS_FIRST_INO;
34266 + *pindex = ino / page_bits;
34267 + *bit = ino % page_bits;
34270 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
34275 + struct au_sbinfo *sbinfo;
34276 + struct file *xib;
34277 + unsigned long *p;
34279 + sbinfo = au_sbi(sb);
34280 + MtxMustLock(&sbinfo->si_xib_mtx);
34281 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
34282 + || !au_opt_test(sbinfo->si_mntflags, XINO));
34284 + if (pindex == sbinfo->si_xib_last_pindex)
34287 + xib = sbinfo->si_xib;
34288 + p = sbinfo->si_xib_buf;
34289 + pos = sbinfo->si_xib_last_pindex;
34290 + pos *= PAGE_SIZE;
34291 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
34292 + if (unlikely(sz != PAGE_SIZE))
34296 + pos *= PAGE_SIZE;
34297 + if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
34298 + sz = xino_fread(sbinfo->si_xread, xib, p, PAGE_SIZE, &pos);
34300 + memset(p, 0, PAGE_SIZE);
34301 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
34303 + if (sz == PAGE_SIZE) {
34304 + sbinfo->si_xib_last_pindex = pindex;
34305 + return 0; /* success */
34309 + AuIOErr1("write failed (%zd)\n", sz);
34316 +/* ---------------------------------------------------------------------- */
34318 +static void au_xib_clear_bit(struct inode *inode)
34321 + unsigned long pindex;
34322 + struct super_block *sb;
34323 + struct au_sbinfo *sbinfo;
34325 + AuDebugOn(inode->i_nlink);
34327 + sb = inode->i_sb;
34328 + xib_calc_bit(inode->i_ino, &pindex, &bit);
34329 + AuDebugOn(page_bits <= bit);
34330 + sbinfo = au_sbi(sb);
34331 + mutex_lock(&sbinfo->si_xib_mtx);
34332 + err = xib_pindex(sb, pindex);
34334 + clear_bit(bit, sbinfo->si_xib_buf);
34335 + sbinfo->si_xib_next_bit = bit;
34337 + mutex_unlock(&sbinfo->si_xib_mtx);
34340 +/* for s_op->delete_inode() */
34341 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
34344 + unsigned int mnt_flags;
34345 + aufs_bindex_t bindex, bend, bi;
34346 + unsigned char try_trunc;
34347 + struct au_iinfo *iinfo;
34348 + struct super_block *sb;
34349 + struct au_hinode *hi;
34350 + struct inode *h_inode;
34351 + struct au_branch *br;
34352 + vfs_writef_t xwrite;
34354 + sb = inode->i_sb;
34355 + mnt_flags = au_mntflags(sb);
34356 + if (!au_opt_test(mnt_flags, XINO)
34357 + || inode->i_ino == AUFS_ROOT_INO)
34361 + au_xigen_inc(inode);
34362 + au_xib_clear_bit(inode);
34365 + iinfo = au_ii(inode);
34369 + bindex = iinfo->ii_bstart;
34373 + xwrite = au_sbi(sb)->si_xwrite;
34374 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
34375 + hi = iinfo->ii_hinode + bindex;
34376 + bend = iinfo->ii_bend;
34377 + for (; bindex <= bend; bindex++, hi++) {
34378 + h_inode = hi->hi_inode;
34380 + || (!unlinked && h_inode->i_nlink))
34383 + /* inode may not be revalidated */
34384 + bi = au_br_index(sb, hi->hi_id);
34388 + br = au_sbr(sb, bi);
34389 + err = au_xino_do_write(xwrite, br->br_xino.xi_file,
34390 + h_inode->i_ino, /*ino*/0);
34391 + if (!err && try_trunc
34392 + && au_test_fs_trunc_xino(au_br_sb(br)))
34393 + xino_try_trunc(sb, br);
34397 +/* get an unused inode number from bitmap */
34398 +ino_t au_xino_new_ino(struct super_block *sb)
34401 + unsigned long *p, pindex, ul, pend;
34402 + struct au_sbinfo *sbinfo;
34403 + struct file *file;
34404 + int free_bit, err;
34406 + if (!au_opt_test(au_mntflags(sb), XINO))
34407 + return iunique(sb, AUFS_FIRST_INO);
34409 + sbinfo = au_sbi(sb);
34410 + mutex_lock(&sbinfo->si_xib_mtx);
34411 + p = sbinfo->si_xib_buf;
34412 + free_bit = sbinfo->si_xib_next_bit;
34413 + if (free_bit < page_bits && !test_bit(free_bit, p))
34414 + goto out; /* success */
34415 + free_bit = find_first_zero_bit(p, page_bits);
34416 + if (free_bit < page_bits)
34417 + goto out; /* success */
34419 + pindex = sbinfo->si_xib_last_pindex;
34420 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
34421 + err = xib_pindex(sb, ul);
34422 + if (unlikely(err))
34424 + free_bit = find_first_zero_bit(p, page_bits);
34425 + if (free_bit < page_bits)
34426 + goto out; /* success */
34429 + file = sbinfo->si_xib;
34430 + pend = vfsub_f_size_read(file) / PAGE_SIZE;
34431 + for (ul = pindex + 1; ul <= pend; ul++) {
34432 + err = xib_pindex(sb, ul);
34433 + if (unlikely(err))
34435 + free_bit = find_first_zero_bit(p, page_bits);
34436 + if (free_bit < page_bits)
34437 + goto out; /* success */
34442 + set_bit(free_bit, p);
34443 + sbinfo->si_xib_next_bit = free_bit + 1;
34444 + pindex = sbinfo->si_xib_last_pindex;
34445 + mutex_unlock(&sbinfo->si_xib_mtx);
34446 + ino = xib_calc_ino(pindex, free_bit);
34447 + AuDbg("i%lu\n", (unsigned long)ino);
34450 + mutex_unlock(&sbinfo->si_xib_mtx);
34456 + * read @ino from xinofile for the specified branch{@sb, @bindex}
34457 + * at the position of @h_ino.
34458 + * if @ino does not exist and @do_new is true, get new one.
34460 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
34466 + struct file *file;
34467 + struct au_sbinfo *sbinfo;
34470 + if (!au_opt_test(au_mntflags(sb), XINO))
34471 + return 0; /* no xino */
34474 + sbinfo = au_sbi(sb);
34476 + if (unlikely(au_loff_max / sizeof(*ino) - 1 < pos)) {
34477 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
34480 + pos *= sizeof(*ino);
34482 + file = au_sbr(sb, bindex)->br_xino.xi_file;
34483 + if (vfsub_f_size_read(file) < pos + sizeof(*ino))
34484 + return 0; /* no ino */
34486 + sz = xino_fread(sbinfo->si_xread, file, ino, sizeof(*ino), &pos);
34487 + if (sz == sizeof(*ino))
34488 + return 0; /* success */
34491 + if (unlikely(sz >= 0)) {
34493 + AuIOErr("xino read error (%zd)\n", sz);
34499 +/* ---------------------------------------------------------------------- */
34501 +/* create and set a new xino file */
34503 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent)
34505 + struct file *file;
34506 + struct dentry *h_parent, *d;
34507 + struct inode *h_dir, *inode;
34511 + * at mount-time, and the xino file is the default path,
34512 + * hnotify is disabled so we have no notify events to ignore.
34513 + * when a user specified the xino, we cannot get au_hdir to be ignored.
34515 + file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
34516 + /* | __FMODE_NONOTIFY */,
34517 + S_IRUGO | S_IWUGO);
34518 + if (IS_ERR(file)) {
34520 + pr_err("open %s(%ld)\n", fname, PTR_ERR(file));
34524 + /* keep file count */
34526 + inode = file_inode(file);
34527 + h_parent = dget_parent(file->f_path.dentry);
34528 + h_dir = d_inode(h_parent);
34529 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
34530 + /* mnt_want_write() is unnecessary here */
34531 + /* no delegation since it is just created */
34532 + if (inode->i_nlink)
34533 + err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
34535 + mutex_unlock(&h_dir->i_mutex);
34537 + if (unlikely(err)) {
34539 + pr_err("unlink %s(%d)\n", fname, err);
34544 + d = file->f_path.dentry;
34545 + if (unlikely(sb == d->d_sb)) {
34547 + pr_err("%s must be outside\n", fname);
34550 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
34552 + pr_err("xino doesn't support %s(%s)\n",
34553 + fname, au_sbtype(d->d_sb));
34556 + return file; /* success */
34560 + file = ERR_PTR(err);
34565 + * find another branch who is on the same filesystem of the specified
34566 + * branch{@btgt}. search until @bend.
34568 +static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
34569 + aufs_bindex_t bend)
34571 + aufs_bindex_t bindex;
34572 + struct super_block *tgt_sb = au_sbr_sb(sb, btgt);
34574 + for (bindex = 0; bindex < btgt; bindex++)
34575 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
34577 + for (bindex++; bindex <= bend; bindex++)
34578 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
34583 +/* ---------------------------------------------------------------------- */
34586 + * initialize the xinofile for the specified branch @br
34587 + * at the place/path where @base_file indicates.
34588 + * test whether another branch is on the same filesystem or not,
34589 + * if @do_test is true.
34591 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
34592 + struct file *base_file, int do_test)
34596 + aufs_bindex_t bend, bindex;
34597 + struct au_branch *shared_br, *b;
34598 + struct file *file;
34599 + struct super_block *tgt_sb;
34601 + shared_br = NULL;
34602 + bend = au_sbend(sb);
34604 + tgt_sb = au_br_sb(br);
34605 + for (bindex = 0; bindex <= bend; bindex++) {
34606 + b = au_sbr(sb, bindex);
34607 + if (tgt_sb == au_br_sb(b)) {
34614 + if (!shared_br || !shared_br->br_xino.xi_file) {
34615 + struct au_xino_lock_dir ldir;
34617 + au_xino_lock_dir(sb, base_file, &ldir);
34618 + /* mnt_want_write() is unnecessary here */
34619 + file = au_xino_create2(base_file, NULL);
34620 + au_xino_unlock_dir(&ldir);
34621 + err = PTR_ERR(file);
34622 + if (IS_ERR(file))
34624 + br->br_xino.xi_file = file;
34626 + br->br_xino.xi_file = shared_br->br_xino.xi_file;
34627 + get_file(br->br_xino.xi_file);
34630 + ino = AUFS_ROOT_INO;
34631 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
34633 + if (unlikely(err)) {
34634 + fput(br->br_xino.xi_file);
34635 + br->br_xino.xi_file = NULL;
34642 +/* ---------------------------------------------------------------------- */
34644 +/* trucate a xino bitmap file */
34647 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
34651 + unsigned long pindex;
34652 + loff_t pos, pend;
34653 + struct au_sbinfo *sbinfo;
34654 + vfs_readf_t func;
34656 + unsigned long *p;
34659 + sbinfo = au_sbi(sb);
34660 + MtxMustLock(&sbinfo->si_xib_mtx);
34661 + p = sbinfo->si_xib_buf;
34662 + func = sbinfo->si_xread;
34663 + pend = vfsub_f_size_read(file);
34665 + while (pos < pend) {
34666 + sz = xino_fread(func, file, page, PAGE_SIZE, &pos);
34668 + if (unlikely(sz <= 0))
34672 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
34673 + if (unlikely(*ino < AUFS_FIRST_INO))
34676 + xib_calc_bit(*ino, &pindex, &bit);
34677 + AuDebugOn(page_bits <= bit);
34678 + err = xib_pindex(sb, pindex);
34690 +static int xib_restore(struct super_block *sb)
34693 + aufs_bindex_t bindex, bend;
34697 + page = (void *)__get_free_page(GFP_NOFS);
34698 + if (unlikely(!page))
34702 + bend = au_sbend(sb);
34703 + for (bindex = 0; !err && bindex <= bend; bindex++)
34704 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0)
34705 + err = do_xib_restore
34706 + (sb, au_sbr(sb, bindex)->br_xino.xi_file, page);
34708 + AuDbg("b%d\n", bindex);
34709 + free_page((unsigned long)page);
34715 +int au_xib_trunc(struct super_block *sb)
34720 + struct au_xino_lock_dir ldir;
34721 + struct au_sbinfo *sbinfo;
34722 + unsigned long *p;
34723 + struct file *file;
34725 + SiMustWriteLock(sb);
34728 + sbinfo = au_sbi(sb);
34729 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
34732 + file = sbinfo->si_xib;
34733 + if (vfsub_f_size_read(file) <= PAGE_SIZE)
34736 + au_xino_lock_dir(sb, file, &ldir);
34737 + /* mnt_want_write() is unnecessary here */
34738 + file = au_xino_create2(sbinfo->si_xib, NULL);
34739 + au_xino_unlock_dir(&ldir);
34740 + err = PTR_ERR(file);
34741 + if (IS_ERR(file))
34743 + fput(sbinfo->si_xib);
34744 + sbinfo->si_xib = file;
34746 + p = sbinfo->si_xib_buf;
34747 + memset(p, 0, PAGE_SIZE);
34749 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xib, p, PAGE_SIZE, &pos);
34750 + if (unlikely(sz != PAGE_SIZE)) {
34752 + AuIOErr("err %d\n", err);
34758 + mutex_lock(&sbinfo->si_xib_mtx);
34759 + /* mnt_want_write() is unnecessary here */
34760 + err = xib_restore(sb);
34761 + mutex_unlock(&sbinfo->si_xib_mtx);
34767 +/* ---------------------------------------------------------------------- */
34770 + * xino mount option handlers
34774 +static void xino_clear_xib(struct super_block *sb)
34776 + struct au_sbinfo *sbinfo;
34778 + SiMustWriteLock(sb);
34780 + sbinfo = au_sbi(sb);
34781 + sbinfo->si_xread = NULL;
34782 + sbinfo->si_xwrite = NULL;
34783 + if (sbinfo->si_xib)
34784 + fput(sbinfo->si_xib);
34785 + sbinfo->si_xib = NULL;
34786 + free_page((unsigned long)sbinfo->si_xib_buf);
34787 + sbinfo->si_xib_buf = NULL;
34790 +static int au_xino_set_xib(struct super_block *sb, struct file *base)
34794 + struct au_sbinfo *sbinfo;
34795 + struct file *file;
34797 + SiMustWriteLock(sb);
34799 + sbinfo = au_sbi(sb);
34800 + file = au_xino_create2(base, sbinfo->si_xib);
34801 + err = PTR_ERR(file);
34802 + if (IS_ERR(file))
34804 + if (sbinfo->si_xib)
34805 + fput(sbinfo->si_xib);
34806 + sbinfo->si_xib = file;
34807 + sbinfo->si_xread = vfs_readf(file);
34808 + sbinfo->si_xwrite = vfs_writef(file);
34811 + if (!sbinfo->si_xib_buf)
34812 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
34813 + if (unlikely(!sbinfo->si_xib_buf))
34816 + sbinfo->si_xib_last_pindex = 0;
34817 + sbinfo->si_xib_next_bit = 0;
34818 + if (vfsub_f_size_read(file) < PAGE_SIZE) {
34820 + err = xino_fwrite(sbinfo->si_xwrite, file, sbinfo->si_xib_buf,
34821 + PAGE_SIZE, &pos);
34822 + if (unlikely(err != PAGE_SIZE))
34826 + goto out; /* success */
34829 + free_page((unsigned long)sbinfo->si_xib_buf);
34830 + sbinfo->si_xib_buf = NULL;
34834 + fput(sbinfo->si_xib);
34835 + sbinfo->si_xib = NULL;
34836 + sbinfo->si_xread = NULL;
34837 + sbinfo->si_xwrite = NULL;
34842 +/* xino for each branch */
34843 +static void xino_clear_br(struct super_block *sb)
34845 + aufs_bindex_t bindex, bend;
34846 + struct au_branch *br;
34848 + bend = au_sbend(sb);
34849 + for (bindex = 0; bindex <= bend; bindex++) {
34850 + br = au_sbr(sb, bindex);
34851 + if (!br || !br->br_xino.xi_file)
34854 + fput(br->br_xino.xi_file);
34855 + br->br_xino.xi_file = NULL;
34859 +static int au_xino_set_br(struct super_block *sb, struct file *base)
34863 + aufs_bindex_t bindex, bend, bshared;
34865 + struct file *old, *new;
34867 + struct au_branch *br;
34868 + struct inode *inode;
34869 + vfs_writef_t writef;
34871 + SiMustWriteLock(sb);
34874 + bend = au_sbend(sb);
34875 + fpair = kcalloc(bend + 1, sizeof(*fpair), GFP_NOFS);
34876 + if (unlikely(!fpair))
34879 + inode = d_inode(sb->s_root);
34880 + ino = AUFS_ROOT_INO;
34881 + writef = au_sbi(sb)->si_xwrite;
34882 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
34883 + br = au_sbr(sb, bindex);
34884 + bshared = is_sb_shared(sb, bindex, bindex - 1);
34885 + if (bshared >= 0) {
34886 + /* shared xino */
34887 + *p = fpair[bshared];
34888 + get_file(p->new);
34893 + p->old = br->br_xino.xi_file;
34894 + p->new = au_xino_create2(base, br->br_xino.xi_file);
34895 + err = PTR_ERR(p->new);
34896 + if (IS_ERR(p->new)) {
34902 + err = au_xino_do_write(writef, p->new,
34903 + au_h_iptr(inode, bindex)->i_ino, ino);
34904 + if (unlikely(err))
34908 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
34909 + br = au_sbr(sb, bindex);
34910 + if (br->br_xino.xi_file)
34911 + fput(br->br_xino.xi_file);
34912 + get_file(p->new);
34913 + br->br_xino.xi_file = p->new;
34917 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++)
34927 +void au_xino_clr(struct super_block *sb)
34929 + struct au_sbinfo *sbinfo;
34931 + au_xigen_clr(sb);
34932 + xino_clear_xib(sb);
34933 + xino_clear_br(sb);
34934 + sbinfo = au_sbi(sb);
34935 + /* lvalue, do not call au_mntflags() */
34936 + au_opt_clr(sbinfo->si_mntflags, XINO);
34939 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount)
34942 + struct dentry *parent, *cur_parent;
34943 + struct qstr *dname, *cur_name;
34944 + struct file *cur_xino;
34945 + struct inode *dir;
34946 + struct au_sbinfo *sbinfo;
34948 + SiMustWriteLock(sb);
34951 + sbinfo = au_sbi(sb);
34952 + parent = dget_parent(xino->file->f_path.dentry);
34955 + dname = &xino->file->f_path.dentry->d_name;
34956 + cur_xino = sbinfo->si_xib;
34958 + cur_parent = dget_parent(cur_xino->f_path.dentry);
34959 + cur_name = &cur_xino->f_path.dentry->d_name;
34960 + skip = (cur_parent == parent
34961 + && au_qstreq(dname, cur_name));
34962 + dput(cur_parent);
34968 + au_opt_set(sbinfo->si_mntflags, XINO);
34969 + dir = d_inode(parent);
34970 + mutex_lock_nested(&dir->i_mutex, AuLsc_I_PARENT);
34971 + /* mnt_want_write() is unnecessary here */
34972 + err = au_xino_set_xib(sb, xino->file);
34974 + err = au_xigen_set(sb, xino->file);
34976 + err = au_xino_set_br(sb, xino->file);
34977 + mutex_unlock(&dir->i_mutex);
34979 + goto out; /* success */
34982 + AuIOErr("failed creating xino(%d).\n", err);
34983 + au_xigen_clr(sb);
34984 + xino_clear_xib(sb);
34991 +/* ---------------------------------------------------------------------- */
34994 + * create a xinofile at the default place/path.
34996 +struct file *au_xino_def(struct super_block *sb)
34998 + struct file *file;
35000 + struct au_branch *br;
35001 + struct super_block *h_sb;
35002 + struct path path;
35003 + aufs_bindex_t bend, bindex, bwr;
35006 + bend = au_sbend(sb);
35008 + for (bindex = 0; bindex <= bend; bindex++) {
35009 + br = au_sbr(sb, bindex);
35010 + if (au_br_writable(br->br_perm)
35011 + && !au_test_fs_bad_xino(au_br_sb(br))) {
35018 + file = ERR_PTR(-ENOMEM);
35019 + page = (void *)__get_free_page(GFP_NOFS);
35020 + if (unlikely(!page))
35022 + path.mnt = au_br_mnt(br);
35023 + path.dentry = au_h_dptr(sb->s_root, bwr);
35024 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
35025 + file = (void *)p;
35026 + if (!IS_ERR(p)) {
35027 + strcat(p, "/" AUFS_XINO_FNAME);
35028 + AuDbg("%s\n", p);
35029 + file = au_xino_create(sb, p, /*silent*/0);
35030 + if (!IS_ERR(file))
35031 + au_xino_brid_set(sb, br->br_id);
35033 + free_page((unsigned long)page);
35035 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0);
35036 + if (IS_ERR(file))
35038 + h_sb = file->f_path.dentry->d_sb;
35039 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
35040 + pr_err("xino doesn't support %s(%s)\n",
35041 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
35043 + file = ERR_PTR(-EINVAL);
35045 + if (!IS_ERR(file))
35046 + au_xino_brid_set(sb, -1);
35053 +/* ---------------------------------------------------------------------- */
35055 +int au_xino_path(struct seq_file *seq, struct file *file)
35059 + err = au_seq_path(seq, &file->f_path);
35060 + if (unlikely(err))
35063 +#define Deleted "\\040(deleted)"
35064 + seq->count -= sizeof(Deleted) - 1;
35065 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
35066 + sizeof(Deleted) - 1));
35072 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
35073 --- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
35074 +++ linux/include/uapi/linux/aufs_type.h 2016-01-13 20:11:11.673093671 +0100
35077 + * Copyright (C) 2005-2015 Junjiro R. Okajima
35079 + * This program, aufs is free software; you can redistribute it and/or modify
35080 + * it under the terms of the GNU General Public License as published by
35081 + * the Free Software Foundation; either version 2 of the License, or
35082 + * (at your option) any later version.
35084 + * This program is distributed in the hope that it will be useful,
35085 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35086 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35087 + * GNU General Public License for more details.
35089 + * You should have received a copy of the GNU General Public License
35090 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35093 +#ifndef __AUFS_TYPE_H__
35094 +#define __AUFS_TYPE_H__
35096 +#define AUFS_NAME "aufs"
35100 + * define it before including all other headers.
35101 + * sched.h may use pr_* macros before defining "current", so define the
35102 + * no-current version first, and re-define later.
35104 +#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
35105 +#include <linux/sched.h>
35107 +#define pr_fmt(fmt) \
35108 + AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
35109 + (int)sizeof(current->comm), current->comm, current->pid
35111 +#include <stdint.h>
35112 +#include <sys/types.h>
35113 +#endif /* __KERNEL__ */
35115 +#include <linux/limits.h>
35117 +#define AUFS_VERSION "4.x-rcN-20160111"
35119 +/* todo? move this to linux-2.6.19/include/magic.h */
35120 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
35122 +/* ---------------------------------------------------------------------- */
35124 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
35125 +typedef int8_t aufs_bindex_t;
35126 +#define AUFS_BRANCH_MAX 127
35128 +typedef int16_t aufs_bindex_t;
35129 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
35130 +#define AUFS_BRANCH_MAX 511
35131 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
35132 +#define AUFS_BRANCH_MAX 1023
35133 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
35134 +#define AUFS_BRANCH_MAX 32767
35139 +#ifndef AUFS_BRANCH_MAX
35140 +#error unknown CONFIG_AUFS_BRANCH_MAX value
35142 +#endif /* __KERNEL__ */
35144 +/* ---------------------------------------------------------------------- */
35146 +#define AUFS_FSTYPE AUFS_NAME
35148 +#define AUFS_ROOT_INO 2
35149 +#define AUFS_FIRST_INO 11
35151 +#define AUFS_WH_PFX ".wh."
35152 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
35153 +#define AUFS_WH_TMP_LEN 4
35154 +/* a limit for rmdir/rename a dir and copyup */
35155 +#define AUFS_MAX_NAMELEN (NAME_MAX \
35156 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
35158 + - AUFS_WH_TMP_LEN) /* hex */
35159 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
35160 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
35161 +#define AUFS_XINO_DEF_SEC 30 /* seconds */
35162 +#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
35163 +#define AUFS_DIRWH_DEF 3
35164 +#define AUFS_RDCACHE_DEF 10 /* seconds */
35165 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
35166 +#define AUFS_RDBLK_DEF 512 /* bytes */
35167 +#define AUFS_RDHASH_DEF 32
35168 +#define AUFS_WKQ_NAME AUFS_NAME "d"
35169 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
35170 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
35171 +#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
35172 +#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
35174 +/* pseudo-link maintenace under /proc */
35175 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
35176 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
35177 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
35179 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
35180 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
35182 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
35183 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
35184 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
35186 +/* doubly whiteouted */
35187 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
35188 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
35189 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
35191 +/* branch permissions and attributes */
35192 +#define AUFS_BRPERM_RW "rw"
35193 +#define AUFS_BRPERM_RO "ro"
35194 +#define AUFS_BRPERM_RR "rr"
35195 +#define AUFS_BRATTR_COO_REG "coo_reg"
35196 +#define AUFS_BRATTR_COO_ALL "coo_all"
35197 +#define AUFS_BRATTR_FHSM "fhsm"
35198 +#define AUFS_BRATTR_UNPIN "unpin"
35199 +#define AUFS_BRATTR_ICEX "icex"
35200 +#define AUFS_BRATTR_ICEX_SEC "icexsec"
35201 +#define AUFS_BRATTR_ICEX_SYS "icexsys"
35202 +#define AUFS_BRATTR_ICEX_TR "icextr"
35203 +#define AUFS_BRATTR_ICEX_USR "icexusr"
35204 +#define AUFS_BRATTR_ICEX_OTH "icexoth"
35205 +#define AUFS_BRRATTR_WH "wh"
35206 +#define AUFS_BRWATTR_NLWH "nolwh"
35207 +#define AUFS_BRWATTR_MOO "moo"
35209 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
35210 +#define AuBrPerm_RO (1 << 1) /* readonly */
35211 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
35212 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
35214 +#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
35215 +#define AuBrAttr_COO_ALL (1 << 4)
35216 +#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
35218 +#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
35219 +#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
35220 + branch. meaningless since
35221 + linux-3.18-rc1 */
35223 +/* ignore error in copying XATTR */
35224 +#define AuBrAttr_ICEX_SEC (1 << 7)
35225 +#define AuBrAttr_ICEX_SYS (1 << 8)
35226 +#define AuBrAttr_ICEX_TR (1 << 9)
35227 +#define AuBrAttr_ICEX_USR (1 << 10)
35228 +#define AuBrAttr_ICEX_OTH (1 << 11)
35229 +#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
35230 + | AuBrAttr_ICEX_SYS \
35231 + | AuBrAttr_ICEX_TR \
35232 + | AuBrAttr_ICEX_USR \
35233 + | AuBrAttr_ICEX_OTH)
35235 +#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
35236 +#define AuBrRAttr_Mask AuBrRAttr_WH
35238 +#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
35239 +#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
35240 +#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
35242 +#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
35244 +/* #warning test userspace */
35246 +#ifndef CONFIG_AUFS_FHSM
35247 +#undef AuBrAttr_FHSM
35248 +#define AuBrAttr_FHSM 0
35250 +#ifndef CONFIG_AUFS_XATTR
35251 +#undef AuBrAttr_ICEX
35252 +#define AuBrAttr_ICEX 0
35253 +#undef AuBrAttr_ICEX_SEC
35254 +#define AuBrAttr_ICEX_SEC 0
35255 +#undef AuBrAttr_ICEX_SYS
35256 +#define AuBrAttr_ICEX_SYS 0
35257 +#undef AuBrAttr_ICEX_TR
35258 +#define AuBrAttr_ICEX_TR 0
35259 +#undef AuBrAttr_ICEX_USR
35260 +#define AuBrAttr_ICEX_USR 0
35261 +#undef AuBrAttr_ICEX_OTH
35262 +#define AuBrAttr_ICEX_OTH 0
35266 +/* the longest combination */
35267 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
35268 +#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
35269 + "+" AUFS_BRATTR_COO_REG \
35270 + "+" AUFS_BRATTR_FHSM \
35271 + "+" AUFS_BRATTR_UNPIN \
35272 + "+" AUFS_BRATTR_ICEX_SEC \
35273 + "+" AUFS_BRATTR_ICEX_SYS \
35274 + "+" AUFS_BRATTR_ICEX_USR \
35275 + "+" AUFS_BRATTR_ICEX_OTH \
35276 + "+" AUFS_BRWATTR_NLWH)
35279 + char a[AuBrPermStrSz];
35280 +} au_br_perm_str_t;
35282 +static inline int au_br_writable(int brperm)
35284 + return brperm & AuBrPerm_RW;
35287 +static inline int au_br_whable(int brperm)
35289 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
35292 +static inline int au_br_wh_linkable(int brperm)
35294 + return !(brperm & AuBrWAttr_NoLinkWH);
35297 +static inline int au_br_cmoo(int brperm)
35299 + return brperm & AuBrAttr_CMOO_Mask;
35302 +static inline int au_br_fhsm(int brperm)
35304 + return brperm & AuBrAttr_FHSM;
35307 +/* ---------------------------------------------------------------------- */
35311 + /* readdir in userspace */
35315 + AuCtl_WBR_FD, /* pathconf wrapper */
35316 + AuCtl_IBUSY, /* busy inode */
35317 + AuCtl_MVDOWN, /* move-down */
35318 + AuCtl_BR, /* info about branches */
35319 + AuCtl_FHSM_FD /* connection for fhsm */
35322 +/* borrowed from linux/include/linux/kernel.h */
35324 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
35325 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
35328 +/* borrowed from linux/include/linux/compiler-gcc3.h */
35330 +#define __aligned(x) __attribute__((aligned(x)))
35335 +#define __packed __attribute__((packed))
35339 +struct au_rdu_cookie {
35344 + uint32_t generation;
35347 +struct au_rdu_ent {
35356 +static inline int au_rdu_len(int nlen)
35358 + /* include the terminating NULL */
35359 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
35360 + sizeof(uint64_t));
35363 +union au_rdu_ent_ul {
35364 + struct au_rdu_ent __user *e;
35376 + uint64_t sz; /* AuCtl_RDU */
35377 + uint64_t nent; /* AuCtl_RDU_INO */
35379 + union au_rdu_ent_ul ent;
35380 + uint16_t verify[AufsCtlRduV_End];
35382 + /* input/output */
35386 + union au_rdu_ent_ul tail;
35387 + /* number of entries which were added in a single call */
35392 + struct au_rdu_cookie cookie;
35395 +/* ---------------------------------------------------------------------- */
35397 +struct aufs_wbr_fd {
35402 +/* ---------------------------------------------------------------------- */
35404 +struct aufs_ibusy {
35405 + uint64_t ino, h_ino;
35409 +/* ---------------------------------------------------------------------- */
35411 +/* error code for move-down */
35412 +/* the actual message strings are implemented in aufs-util.git */
35414 + EAU_MVDOWN_OPAQUE = 1,
35415 + EAU_MVDOWN_WHITEOUT,
35416 + EAU_MVDOWN_UPPER,
35417 + EAU_MVDOWN_BOTTOM,
35418 + EAU_MVDOWN_NOUPPER,
35419 + EAU_MVDOWN_NOLOWERBR,
35423 +/* flags for move-down */
35424 +#define AUFS_MVDOWN_DMSG 1
35425 +#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
35426 +#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
35427 +#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
35428 +#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
35429 +#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
35430 +#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
35431 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
35432 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
35433 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
35434 +#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
35435 +#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
35436 +#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
35438 +/* index for move-down */
35440 + AUFS_MVDOWN_UPPER,
35441 + AUFS_MVDOWN_LOWER,
35442 + AUFS_MVDOWN_NARRAY
35446 + * additional info of move-down
35447 + * number of free blocks and inodes.
35448 + * subset of struct kstatfs, but smaller and always 64bit.
35450 +struct aufs_stfs {
35451 + uint64_t f_blocks;
35452 + uint64_t f_bavail;
35453 + uint64_t f_files;
35454 + uint64_t f_ffree;
35457 +struct aufs_stbr {
35458 + int16_t brid; /* optional input */
35459 + int16_t bindex; /* output */
35460 + struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
35463 +struct aufs_mvdown {
35464 + uint32_t flags; /* input/output */
35465 + struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
35466 + int8_t au_errno; /* output */
35469 +/* ---------------------------------------------------------------------- */
35471 +union aufs_brinfo {
35472 + /* PATH_MAX may differ between kernel-space and user-space */
35473 + char _spacer[4096];
35481 +/* ---------------------------------------------------------------------- */
35483 +#define AuCtlType 'A'
35484 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
35485 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
35486 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
35487 + struct aufs_wbr_fd)
35488 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
35489 +#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
35490 + struct aufs_mvdown)
35491 +#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
35492 +#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
35494 +#endif /* __AUFS_TYPE_H__ */
35495 aufs4.x-rcN loopback patch
35497 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
35498 index abfdd2b..a2e3c43 100644
35499 --- a/drivers/block/loop.c
35500 +++ b/drivers/block/loop.c
35501 @@ -556,7 +556,7 @@ static int do_req_filebacked(struct loop_device *lo, struct request *rq)
35504 struct switch_request {
35505 - struct file *file;
35506 + struct file *file, *virt_file;
35507 struct completion wait;
35510 @@ -582,6 +582,7 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
35511 mapping = file->f_mapping;
35512 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
35513 lo->lo_backing_file = file;
35514 + lo->lo_backing_virt_file = p->virt_file;
35515 lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ?
35516 mapping->host->i_bdev->bd_block_size : PAGE_SIZE;
35517 lo->old_gfp_mask = mapping_gfp_mask(mapping);
35518 @@ -594,11 +595,13 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
35519 * First it needs to flush existing IO, it does this by sending a magic
35520 * BIO down the pipe. The completion of this BIO does the actual switch.
35522 -static int loop_switch(struct loop_device *lo, struct file *file)
35523 +static int loop_switch(struct loop_device *lo, struct file *file,
35524 + struct file *virt_file)
35526 struct switch_request w;
35529 + w.virt_file = virt_file;
35531 /* freeze queue and wait for completion of scheduled requests */
35532 blk_mq_freeze_queue(lo->lo_queue);
35533 @@ -617,7 +620,16 @@ static int loop_switch(struct loop_device *lo, struct file *file)
35535 static int loop_flush(struct loop_device *lo)
35537 - return loop_switch(lo, NULL);
35538 + return loop_switch(lo, NULL, NULL);
35541 +static struct file *loop_real_file(struct file *file)
35543 + struct file *f = NULL;
35545 + if (file->f_path.dentry->d_sb->s_op->real_loop)
35546 + f = file->f_path.dentry->d_sb->s_op->real_loop(file);
35550 static void loop_reread_partitions(struct loop_device *lo,
35551 @@ -654,6 +666,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
35554 struct file *file, *old_file;
35555 + struct file *f, *virt_file = NULL, *old_virt_file;
35556 struct inode *inode;
35559 @@ -670,9 +683,16 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
35563 + f = loop_real_file(file);
35565 + virt_file = file;
35570 inode = file->f_mapping->host;
35571 old_file = lo->lo_backing_file;
35572 + old_virt_file = lo->lo_backing_virt_file;
35576 @@ -684,17 +704,21 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
35579 /* and ... switch */
35580 - error = loop_switch(lo, file);
35581 + error = loop_switch(lo, file, virt_file);
35586 + if (old_virt_file)
35587 + fput(old_virt_file);
35588 if (lo->lo_flags & LO_FLAGS_PARTSCAN)
35589 loop_reread_partitions(lo, bdev);
35599 @@ -881,7 +905,7 @@ static int loop_prepare_queue(struct loop_device *lo)
35600 static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35601 struct block_device *bdev, unsigned int arg)
35603 - struct file *file, *f;
35604 + struct file *file, *f, *virt_file = NULL;
35605 struct inode *inode;
35606 struct address_space *mapping;
35607 unsigned lo_blocksize;
35608 @@ -896,6 +920,12 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35612 + f = loop_real_file(file);
35614 + virt_file = file;
35620 if (lo->lo_state != Lo_unbound)
35621 @@ -948,6 +978,7 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35622 lo->lo_device = bdev;
35623 lo->lo_flags = lo_flags;
35624 lo->lo_backing_file = file;
35625 + lo->lo_backing_virt_file = virt_file;
35626 lo->transfer = NULL;
35628 lo->lo_sizelimit = 0;
35629 @@ -980,6 +1011,8 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35636 /* This is safe: open() is still holding a reference. */
35637 module_put(THIS_MODULE);
35638 @@ -1026,6 +1059,7 @@ loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
35639 static int loop_clr_fd(struct loop_device *lo)
35641 struct file *filp = lo->lo_backing_file;
35642 + struct file *virt_filp = lo->lo_backing_virt_file;
35643 gfp_t gfp = lo->old_gfp_mask;
35644 struct block_device *bdev = lo->lo_device;
35646 @@ -1057,6 +1091,7 @@ static int loop_clr_fd(struct loop_device *lo)
35647 spin_lock_irq(&lo->lo_lock);
35648 lo->lo_state = Lo_rundown;
35649 lo->lo_backing_file = NULL;
35650 + lo->lo_backing_virt_file = NULL;
35651 spin_unlock_irq(&lo->lo_lock);
35653 loop_release_xfer(lo);
35654 @@ -1101,6 +1136,8 @@ static int loop_clr_fd(struct loop_device *lo)
35655 * bd_mutex which is usually taken before lo_ctl_mutex.
35663 diff --git a/drivers/block/loop.h b/drivers/block/loop.h
35664 index fb2237c..c3888c5 100644
35665 --- a/drivers/block/loop.h
35666 +++ b/drivers/block/loop.h
35667 @@ -46,7 +46,7 @@ struct loop_device {
35668 int (*ioctl)(struct loop_device *, int cmd,
35669 unsigned long arg);
35671 - struct file * lo_backing_file;
35672 + struct file * lo_backing_file, *lo_backing_virt_file;
35673 struct block_device *lo_device;
35674 unsigned lo_blocksize;
35676 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
35677 index 91c2ce7..d4ee5a7 100644
35678 --- a/fs/aufs/f_op.c
35679 +++ b/fs/aufs/f_op.c
35680 @@ -389,7 +389,7 @@ static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
35681 if (IS_ERR(h_file))
35684 - if (au_test_loopback_kthread()) {
35685 + if (0 && au_test_loopback_kthread()) {
35686 au_warn_loopback(h_file->f_path.dentry->d_sb);
35687 if (file->f_mapping != h_file->f_mapping) {
35688 file->f_mapping = h_file->f_mapping;
35689 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
35690 index f324758..4555e7b 100644
35691 --- a/fs/aufs/loop.c
35692 +++ b/fs/aufs/loop.c
35693 @@ -131,3 +131,19 @@ void au_loopback_fin(void)
35694 symbol_put(loop_backing_file);
35695 kfree(au_warn_loopback_array);
35698 +/* ---------------------------------------------------------------------- */
35700 +/* support the loopback block device insude aufs */
35702 +struct file *aufs_real_loop(struct file *file)
35706 + BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
35707 + fi_read_lock(file);
35708 + f = au_hf_top(file);
35709 + fi_read_unlock(file);
35713 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
35714 index 6d9864d..3322557 100644
35715 --- a/fs/aufs/loop.h
35716 +++ b/fs/aufs/loop.h
35717 @@ -25,7 +25,11 @@ void au_warn_loopback(struct super_block *h_sb);
35719 int au_loopback_init(void);
35720 void au_loopback_fin(void);
35722 +struct file *aufs_real_loop(struct file *file);
35724 +AuStub(struct file *, loop_backing_file, return NULL)
35726 AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
35727 struct dentry *h_adding)
35728 AuStubInt0(au_test_loopback_kthread, void)
35729 @@ -33,6 +37,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
35731 AuStubInt0(au_loopback_init, void)
35732 AuStubVoid(au_loopback_fin, void)
35734 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
35735 #endif /* BLK_DEV_LOOP */
35737 #endif /* __KERNEL__ */
35738 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
35739 index 98cfd64..a705a0e 100644
35740 --- a/fs/aufs/super.c
35741 +++ b/fs/aufs/super.c
35742 @@ -832,7 +832,10 @@ static const struct super_operations aufs_sop = {
35743 .statfs = aufs_statfs,
35744 .put_super = aufs_put_super,
35745 .sync_fs = aufs_sync_fs,
35746 - .remount_fs = aufs_remount_fs
35747 + .remount_fs = aufs_remount_fs,
35748 +#ifdef CONFIG_AUFS_BDEV_LOOP
35749 + .real_loop = aufs_real_loop
35753 /* ---------------------------------------------------------------------- */
35754 diff --git a/include/linux/fs.h b/include/linux/fs.h
35755 index 8d48506..5246785 100644
35756 --- a/include/linux/fs.h
35757 +++ b/include/linux/fs.h
35758 @@ -1719,6 +1719,10 @@ struct super_operations {
35759 struct shrink_control *);
35760 long (*free_cached_objects)(struct super_block *,
35761 struct shrink_control *);
35762 +#if defined(CONFIG_BLK_DEV_LOOP) || defined(CONFIG_BLK_DEV_LOOP_MODULE)
35764 + struct file *(*real_loop)(struct file *);