3 diff --git a/fs/Kconfig b/fs/Kconfig
4 index da3f32f..b9879fe 100644
7 @@ -215,6 +215,7 @@ 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 f79cf40..7562a4d 100644
19 @@ -125,3 +125,4 @@ 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 f7b2db4..47098aed 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
38 diff --git a/MAINTAINERS b/MAINTAINERS
39 index 747c653..53ecc33 100644
42 @@ -1985,6 +1985,19 @@ 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 674f800a..291ec9e 100644
64 --- a/drivers/block/loop.c
65 +++ b/drivers/block/loop.c
66 @@ -560,6 +560,24 @@ 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 5fc1e50..5f8385a 100644
141 @@ -1102,8 +1102,8 @@ 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 @@ -1119,9 +1119,9 @@ 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 72d8a84..fabd9d7a 100644
179 --- a/include/linux/fs.h
180 +++ b/include/linux/fs.h
181 @@ -1687,6 +1687,12 @@ 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);
211 diff --git a/fs/buffer.c b/fs/buffer.c
212 index 82283ab..477e5f3 100644
215 @@ -2473,7 +2473,7 @@ int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
216 * Update file times before taking page lock. We may end up failing the
217 * fault so this update may be superfluous but who really cares...
219 - file_update_time(vma->vm_file);
220 + vma_file_update_time(vma);
222 ret = __block_page_mkwrite(vma, vmf, get_block);
223 sb_end_pagefault(sb);
224 diff --git a/fs/proc/base.c b/fs/proc/base.c
225 index b25eee4..c83d588 100644
228 @@ -1914,7 +1914,7 @@ static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
229 down_read(&mm->mmap_sem);
230 vma = find_exact_vma(mm, vm_start, vm_end);
231 if (vma && vma->vm_file) {
232 - *path = vma->vm_file->f_path;
233 + *path = vma_pr_or_file(vma)->f_path;
237 diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
238 index f8595e8..cb8eda0 100644
239 --- a/fs/proc/nommu.c
240 +++ b/fs/proc/nommu.c
241 @@ -45,7 +45,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
242 file = region->vm_file;
245 - struct inode *inode = file_inode(region->vm_file);
246 + struct inode *inode;
248 + file = vmr_pr_or_file(region);
249 + inode = file_inode(file);
250 dev = inode->i_sb->s_dev;
253 diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
254 index e2d46ad..5e7e631 100644
255 --- a/fs/proc/task_mmu.c
256 +++ b/fs/proc/task_mmu.c
257 @@ -280,7 +280,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid)
258 const char *name = NULL;
261 - struct inode *inode = file_inode(vma->vm_file);
262 + struct inode *inode;
264 + file = vma_pr_or_file(vma);
265 + inode = file_inode(file);
266 dev = inode->i_sb->s_dev;
268 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
269 @@ -1465,7 +1468,7 @@ static int show_numa_map(struct seq_file *m, void *v, int is_pid)
270 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
271 struct vm_area_struct *vma = v;
272 struct numa_maps *md = &numa_priv->md;
273 - struct file *file = vma->vm_file;
274 + struct file *file = vma_pr_or_file(vma);
275 struct mm_struct *mm = vma->vm_mm;
276 struct mm_walk walk = {
277 .hugetlb_entry = gather_hugetlb_stats,
278 diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
279 index e0d64c9..7aa92db 100644
280 --- a/fs/proc/task_nommu.c
281 +++ b/fs/proc/task_nommu.c
282 @@ -160,7 +160,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma,
286 - struct inode *inode = file_inode(vma->vm_file);
287 + struct inode *inode;
289 + file = vma_pr_or_file(vma);
290 + inode = file_inode(file);
291 dev = inode->i_sb->s_dev;
293 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
294 diff --git a/include/linux/mm.h b/include/linux/mm.h
295 index 80001de..9248b97 100644
296 --- a/include/linux/mm.h
297 +++ b/include/linux/mm.h
298 @@ -1211,6 +1211,28 @@ static inline int fixup_user_fault(struct task_struct *tsk,
302 +extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
303 +extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
305 +extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
306 +extern void vma_do_fput(struct vm_area_struct *, const char[], int);
308 +#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
310 +#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
312 +#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
313 +#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
316 +extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
317 +extern void vmr_do_fput(struct vm_region *, const char[], int);
319 +#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
321 +#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
322 +#endif /* !CONFIG_MMU */
324 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
325 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
326 void *buf, int len, int write);
327 diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
328 index 3d6baa7..750ca95 100644
329 --- a/include/linux/mm_types.h
330 +++ b/include/linux/mm_types.h
331 @@ -250,6 +250,7 @@ struct vm_region {
332 unsigned long vm_top; /* region allocated to here */
333 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
334 struct file *vm_file; /* the backing file or NULL */
335 + struct file *vm_prfile; /* the virtual backing file or NULL */
337 int vm_usage; /* region usage count (access under nommu_region_sem) */
338 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
339 @@ -324,6 +325,7 @@ struct vm_area_struct {
340 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
341 units, *not* PAGE_CACHE_SIZE */
342 struct file * vm_file; /* File we map to (can be NULL). */
343 + struct file *vm_prfile; /* shadow of vm_file */
344 void * vm_private_data; /* was vm_pte (shared mem) */
347 diff --git a/kernel/fork.c b/kernel/fork.c
348 index 2845623..71004bd 100644
351 @@ -462,7 +462,7 @@ static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
352 struct inode *inode = file_inode(file);
353 struct address_space *mapping = file->f_mapping;
357 if (tmp->vm_flags & VM_DENYWRITE)
358 atomic_dec(&inode->i_writecount);
359 i_mmap_lock_write(mapping);
360 diff --git a/mm/Makefile b/mm/Makefile
361 index 2ed4319..e3a53f5 100644
364 @@ -21,7 +21,7 @@ obj-y := filemap.o mempool.o oom_kill.o \
365 mm_init.o mmu_context.o percpu.o slab_common.o \
366 compaction.o vmacache.o \
367 interval_tree.o list_lru.o workingset.o \
369 + prfile.o debug.o $(mmu-y)
373 diff --git a/mm/filemap.c b/mm/filemap.c
374 index 327910c..7bbc372 100644
377 @@ -2089,7 +2089,7 @@ int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
378 int ret = VM_FAULT_LOCKED;
380 sb_start_pagefault(inode->i_sb);
381 - file_update_time(vma->vm_file);
382 + vma_file_update_time(vma);
384 if (page->mapping != inode->i_mapping) {
386 diff --git a/mm/memory.c b/mm/memory.c
387 index deb679c..df2ce3e 100644
390 @@ -2035,7 +2035,7 @@ static inline int wp_page_reuse(struct mm_struct *mm,
394 - file_update_time(vma->vm_file);
395 + vma_file_update_time(vma);
398 return VM_FAULT_WRITE;
399 diff --git a/mm/mmap.c b/mm/mmap.c
400 index 79bcc9f..da28c8a 100644
403 @@ -275,7 +275,7 @@ static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
404 if (vma->vm_ops && vma->vm_ops->close)
405 vma->vm_ops->close(vma);
407 - fput(vma->vm_file);
409 mpol_put(vma_policy(vma));
410 kmem_cache_free(vm_area_cachep, vma);
412 @@ -887,7 +887,7 @@ again: remove_next = 1 + (end > next->vm_end);
415 uprobe_munmap(next, next->vm_start, next->vm_end);
420 anon_vma_merge(vma, next);
421 @@ -1683,8 +1683,8 @@ out:
429 /* Undo any partial mapping done by a device driver. */
430 unmap_region(mm, vma, prev, vma->vm_start, vma->vm_end);
431 @@ -2485,7 +2485,7 @@ static int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
435 - get_file(new->vm_file);
438 if (new->vm_ops && new->vm_ops->open)
439 new->vm_ops->open(new);
440 @@ -2504,7 +2504,7 @@ static int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
441 if (new->vm_ops && new->vm_ops->close)
442 new->vm_ops->close(new);
444 - fput(new->vm_file);
446 unlink_anon_vmas(new);
448 mpol_put(vma_policy(new));
449 @@ -2646,7 +2646,6 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
450 struct vm_area_struct *vma;
451 unsigned long populate = 0;
452 unsigned long ret = -EINVAL;
455 pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. "
456 "See Documentation/vm/remap_file_pages.txt.\n",
457 @@ -2690,10 +2689,10 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
458 munlock_vma_pages_range(vma, start, start + size);
461 - file = get_file(vma->vm_file);
463 ret = do_mmap_pgoff(vma->vm_file, start, size,
464 prot, flags, pgoff, &populate);
468 up_write(&mm->mmap_sem);
470 @@ -2963,7 +2962,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
471 if (anon_vma_clone(new_vma, vma))
472 goto out_free_mempol;
473 if (new_vma->vm_file)
474 - get_file(new_vma->vm_file);
475 + vma_get_file(new_vma);
476 if (new_vma->vm_ops && new_vma->vm_ops->open)
477 new_vma->vm_ops->open(new_vma);
478 vma_link(mm, new_vma, prev, rb_link, rb_parent);
479 diff --git a/mm/nommu.c b/mm/nommu.c
480 index ab14a20..fffc566 100644
483 @@ -671,7 +671,7 @@ static void __put_nommu_region(struct vm_region *region)
484 up_write(&nommu_region_sem);
487 - fput(region->vm_file);
490 /* IO memory and memory shared directly out of the pagecache
491 * from ramfs/tmpfs mustn't be released here */
492 @@ -829,7 +829,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
493 if (vma->vm_ops && vma->vm_ops->close)
494 vma->vm_ops->close(vma);
496 - fput(vma->vm_file);
498 put_nommu_region(vma->vm_region);
499 kmem_cache_free(vm_area_cachep, vma);
501 @@ -1355,7 +1355,7 @@ unsigned long do_mmap(struct file *file,
502 goto error_just_free;
505 - fput(region->vm_file);
507 kmem_cache_free(vm_region_jar, region);
510 @@ -1430,10 +1430,10 @@ error_just_free:
511 up_write(&nommu_region_sem);
514 - fput(region->vm_file);
516 kmem_cache_free(vm_region_jar, region);
518 - fput(vma->vm_file);
520 kmem_cache_free(vm_area_cachep, vma);
523 diff --git a/mm/prfile.c b/mm/prfile.c
525 index 0000000..b323b8a
530 + * Mainly for aufs which mmap(2) diffrent file and wants to print different path
531 + * in /proc/PID/maps.
532 + * Call these functions via macros defined in linux/mm.h.
534 + * See Documentation/filesystems/aufs/design/06mmap.txt
536 + * Copyright (c) 2014 Junjro R. Okajima
537 + * Copyright (c) 2014 Ian Campbell
540 +#include <linux/mm.h>
541 +#include <linux/file.h>
542 +#include <linux/fs.h>
544 +/* #define PRFILE_TRACE */
545 +static inline void prfile_trace(struct file *f, struct file *pr,
546 + const char func[], int line, const char func2[])
550 + pr_info("%s:%d: %s, %s\n", func, line, func2,
551 + f ? (char *)f->f_path.dentry->d_name.name : "(null)");
555 +void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
558 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
560 + prfile_trace(f, pr, func, line, __func__);
561 + file_update_time(f);
563 + file_update_time(pr);
566 +struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
569 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
571 + prfile_trace(f, pr, func, line, __func__);
572 + return (f && pr) ? pr : f;
575 +void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
577 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
579 + prfile_trace(f, pr, func, line, __func__);
585 +void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
587 + struct file *f = vma->vm_file, *pr = vma->vm_prfile;
589 + prfile_trace(f, pr, func, line, __func__);
596 +struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
599 + struct file *f = region->vm_file, *pr = region->vm_prfile;
601 + prfile_trace(f, pr, func, line, __func__);
602 + return (f && pr) ? pr : f;
605 +void vmr_do_fput(struct vm_region *region, const char func[], int line)
607 + struct file *f = region->vm_file, *pr = region->vm_prfile;
609 + prfile_trace(f, pr, func, line, __func__);
614 +#endif /* !CONFIG_MMU */
615 aufs4.3 standalone patch
617 diff --git a/fs/dcache.c b/fs/dcache.c
618 index 8aa7f26..f997345 100644
621 @@ -1272,6 +1272,7 @@ rename_retry:
625 +EXPORT_SYMBOL(d_walk);
628 * Search for at least 1 mount point in the dentry's subdirs.
629 diff --git a/fs/exec.c b/fs/exec.c
630 index b06623a..b9206c5 100644
633 @@ -103,6 +103,7 @@ bool path_noexec(const struct path *path)
634 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
635 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
637 +EXPORT_SYMBOL(path_noexec);
641 diff --git a/fs/file_table.c b/fs/file_table.c
642 index ad17e05..df66450 100644
643 --- a/fs/file_table.c
644 +++ b/fs/file_table.c
645 @@ -147,6 +147,7 @@ over:
647 return ERR_PTR(-ENFILE);
649 +EXPORT_SYMBOL(get_empty_filp);
652 * alloc_file - allocate and initialize a 'struct file'
653 @@ -308,6 +309,7 @@ void put_filp(struct file *file)
657 +EXPORT_SYMBOL(put_filp);
659 void __init files_init(void)
661 diff --git a/fs/namespace.c b/fs/namespace.c
662 index 0570729..ec560d8 100644
665 @@ -463,6 +463,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
666 mnt_dec_writers(real_mount(mnt));
669 +EXPORT_SYMBOL_GPL(__mnt_drop_write);
672 * mnt_drop_write - give up write access to a mount
673 @@ -1803,6 +1804,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
677 +EXPORT_SYMBOL(iterate_mounts);
679 static void cleanup_group_ids(struct mount *mnt, struct mount *end)
681 diff --git a/fs/notify/group.c b/fs/notify/group.c
682 index d16b62c..06ca6bc 100644
683 --- a/fs/notify/group.c
684 +++ b/fs/notify/group.c
686 #include <linux/srcu.h>
687 #include <linux/rculist.h>
688 #include <linux/wait.h>
689 +#include <linux/module.h>
691 #include <linux/fsnotify_backend.h>
692 #include "fsnotify.h"
693 @@ -72,6 +73,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
695 atomic_inc(&group->refcnt);
697 +EXPORT_SYMBOL(fsnotify_get_group);
700 * Drop a reference to a group. Free it if it's through.
701 @@ -81,6 +83,7 @@ void fsnotify_put_group(struct fsnotify_group *group)
702 if (atomic_dec_and_test(&group->refcnt))
703 fsnotify_final_destroy_group(group);
705 +EXPORT_SYMBOL(fsnotify_put_group);
708 * Create a new fsnotify_group and hold a reference for the group returned.
709 @@ -109,6 +112,7 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
713 +EXPORT_SYMBOL(fsnotify_alloc_group);
715 int fsnotify_fasync(int fd, struct file *file, int on)
717 diff --git a/fs/notify/mark.c b/fs/notify/mark.c
718 index fc0df44..325b5c6 100644
719 --- a/fs/notify/mark.c
720 +++ b/fs/notify/mark.c
721 @@ -109,6 +109,7 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
722 mark->free_mark(mark);
725 +EXPORT_SYMBOL(fsnotify_put_mark);
727 /* Calculate mask of events for a list of marks */
728 u32 fsnotify_recalc_mask(struct hlist_head *head)
729 @@ -208,6 +209,7 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark,
730 mutex_unlock(&group->mark_mutex);
731 fsnotify_free_mark(mark);
733 +EXPORT_SYMBOL(fsnotify_destroy_mark);
735 void fsnotify_destroy_marks(struct hlist_head *head, spinlock_t *lock)
737 @@ -392,6 +394,7 @@ err:
741 +EXPORT_SYMBOL(fsnotify_add_mark);
743 int fsnotify_add_mark(struct fsnotify_mark *mark, struct fsnotify_group *group,
744 struct inode *inode, struct vfsmount *mnt, int allow_dups)
745 @@ -492,6 +495,7 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
746 atomic_set(&mark->refcnt, 1);
747 mark->free_mark = free_mark;
749 +EXPORT_SYMBOL(fsnotify_init_mark);
751 static int fsnotify_mark_destroy(void *ignored)
753 diff --git a/fs/open.c b/fs/open.c
754 index b6f1e96..4ab0d4e 100644
757 @@ -64,6 +64,7 @@ int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
758 mutex_unlock(&dentry->d_inode->i_mutex);
761 +EXPORT_SYMBOL(do_truncate);
763 long vfs_truncate(struct path *path, loff_t length)
765 @@ -678,6 +679,7 @@ int open_check_o_direct(struct file *f)
769 +EXPORT_SYMBOL(open_check_o_direct);
771 static int do_dentry_open(struct file *f,
773 diff --git a/fs/read_write.c b/fs/read_write.c
774 index fd0414e..8ace6ec 100644
775 --- a/fs/read_write.c
776 +++ b/fs/read_write.c
777 @@ -504,6 +504,7 @@ vfs_readf_t vfs_readf(struct file *file)
778 return new_sync_read;
779 return ERR_PTR(-ENOSYS);
781 +EXPORT_SYMBOL(vfs_readf);
783 vfs_writef_t vfs_writef(struct file *file)
785 @@ -515,6 +516,7 @@ vfs_writef_t vfs_writef(struct file *file)
786 return new_sync_write;
787 return ERR_PTR(-ENOSYS);
789 +EXPORT_SYMBOL(vfs_writef);
791 ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
793 diff --git a/fs/splice.c b/fs/splice.c
794 index 5f8385a..f76067e 100644
797 @@ -1115,6 +1115,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
799 return splice_write(pipe, out, ppos, len, flags);
801 +EXPORT_SYMBOL(do_splice_from);
804 * Attempt to initiate a splice from a file to a pipe.
805 @@ -1141,6 +1142,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
807 return splice_read(in, ppos, pipe, len, flags);
809 +EXPORT_SYMBOL(do_splice_to);
812 * splice_direct_to_actor - splices data directly between two non-pipes
813 diff --git a/fs/xattr.c b/fs/xattr.c
814 index 072fee1..a7677af 100644
817 @@ -207,6 +207,7 @@ vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
818 *xattr_value = value;
821 +EXPORT_SYMBOL(vfs_getxattr_alloc);
823 /* Compare an extended attribute value with the given value */
824 int vfs_xattr_cmp(struct dentry *dentry, const char *xattr_name,
825 diff --git a/security/commoncap.c b/security/commoncap.c
826 index 1832cf7..987ff5f 100644
827 --- a/security/commoncap.c
828 +++ b/security/commoncap.c
829 @@ -1053,12 +1053,14 @@ int cap_mmap_addr(unsigned long addr)
833 +EXPORT_SYMBOL(cap_mmap_addr);
835 int cap_mmap_file(struct file *file, unsigned long reqprot,
836 unsigned long prot, unsigned long flags)
840 +EXPORT_SYMBOL(cap_mmap_file);
842 #ifdef CONFIG_SECURITY
844 diff --git a/security/device_cgroup.c b/security/device_cgroup.c
845 index 03c1652..b00aa76 100644
846 --- a/security/device_cgroup.c
847 +++ b/security/device_cgroup.c
849 #include <linux/device_cgroup.h>
850 #include <linux/cgroup.h>
851 #include <linux/ctype.h>
852 +#include <linux/export.h>
853 #include <linux/list.h>
854 #include <linux/uaccess.h>
855 #include <linux/seq_file.h>
856 @@ -849,6 +850,7 @@ int __devcgroup_inode_permission(struct inode *inode, int mask)
857 return __devcgroup_check_permission(type, imajor(inode), iminor(inode),
860 +EXPORT_SYMBOL(__devcgroup_inode_permission);
862 int devcgroup_inode_mknod(int mode, dev_t dev)
864 diff --git a/security/security.c b/security/security.c
865 index 46f405c..54488b0 100644
866 --- a/security/security.c
867 +++ b/security/security.c
868 @@ -433,6 +433,7 @@ int security_path_rmdir(struct path *dir, struct dentry *dentry)
870 return call_int_hook(path_rmdir, 0, dir, dentry);
872 +EXPORT_SYMBOL(security_path_rmdir);
874 int security_path_unlink(struct path *dir, struct dentry *dentry)
876 @@ -449,6 +450,7 @@ int security_path_symlink(struct path *dir, struct dentry *dentry,
878 return call_int_hook(path_symlink, 0, dir, dentry, old_name);
880 +EXPORT_SYMBOL(security_path_symlink);
882 int security_path_link(struct dentry *old_dentry, struct path *new_dir,
883 struct dentry *new_dentry)
884 @@ -457,6 +459,7 @@ int security_path_link(struct dentry *old_dentry, struct path *new_dir,
886 return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
888 +EXPORT_SYMBOL(security_path_link);
890 int security_path_rename(struct path *old_dir, struct dentry *old_dentry,
891 struct path *new_dir, struct dentry *new_dentry,
892 @@ -484,6 +487,7 @@ int security_path_truncate(struct path *path)
894 return call_int_hook(path_truncate, 0, path);
896 +EXPORT_SYMBOL(security_path_truncate);
898 int security_path_chmod(struct path *path, umode_t mode)
900 @@ -491,6 +495,7 @@ int security_path_chmod(struct path *path, umode_t mode)
902 return call_int_hook(path_chmod, 0, path, mode);
904 +EXPORT_SYMBOL(security_path_chmod);
906 int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
908 @@ -498,6 +503,7 @@ int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
910 return call_int_hook(path_chown, 0, path, uid, gid);
912 +EXPORT_SYMBOL(security_path_chown);
914 int security_path_chroot(struct path *path)
916 @@ -583,6 +589,7 @@ int security_inode_readlink(struct dentry *dentry)
918 return call_int_hook(inode_readlink, 0, dentry);
920 +EXPORT_SYMBOL(security_inode_readlink);
922 int security_inode_follow_link(struct dentry *dentry, struct inode *inode,
924 @@ -598,6 +605,7 @@ int security_inode_permission(struct inode *inode, int mask)
926 return call_int_hook(inode_permission, 0, inode, mask);
928 +EXPORT_SYMBOL(security_inode_permission);
930 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
932 @@ -736,6 +744,7 @@ int security_file_permission(struct file *file, int mask)
934 return fsnotify_perm(file, mask);
936 +EXPORT_SYMBOL(security_file_permission);
938 int security_file_alloc(struct file *file)
940 @@ -795,6 +804,7 @@ int security_mmap_file(struct file *file, unsigned long prot,
942 return ima_file_mmap(file, prot);
944 +EXPORT_SYMBOL(security_mmap_file);
946 int security_mmap_addr(unsigned long addr)
948 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
949 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
950 +++ linux/Documentation/ABI/testing/debugfs-aufs 2015-09-24 10:47:58.244719488 +0200
952 +What: /debug/aufs/si_<id>/
954 +Contact: J. R. Okajima <hooanon05g@gmail.com>
956 + Under /debug/aufs, a directory named si_<id> is created
957 + per aufs mount, where <id> is a unique id generated
960 +What: /debug/aufs/si_<id>/plink
962 +Contact: J. R. Okajima <hooanon05g@gmail.com>
964 + It has three lines and shows the information about the
965 + pseudo-link. The first line is a single number
966 + representing a number of buckets. The second line is a
967 + number of pseudo-links per buckets (separated by a
968 + blank). The last line is a single number representing a
969 + total number of psedo-links.
970 + When the aufs mount option 'noplink' is specified, it
971 + will show "1\n0\n0\n".
973 +What: /debug/aufs/si_<id>/xib
975 +Contact: J. R. Okajima <hooanon05g@gmail.com>
977 + It shows the consumed blocks by xib (External Inode Number
978 + Bitmap), its block size and file size.
979 + When the aufs mount option 'noxino' is specified, it
980 + will be empty. About XINO files, see the aufs manual.
982 +What: /debug/aufs/si_<id>/xino0, xino1 ... xinoN
984 +Contact: J. R. Okajima <hooanon05g@gmail.com>
986 + It shows the consumed blocks by xino (External Inode Number
987 + Translation Table), its link count, block size and file
989 + When the aufs mount option 'noxino' is specified, it
990 + will be empty. About XINO files, see the aufs manual.
992 +What: /debug/aufs/si_<id>/xigen
994 +Contact: J. R. Okajima <hooanon05g@gmail.com>
996 + It shows the consumed blocks by xigen (External Inode
997 + Generation Table), its block size and file size.
998 + If CONFIG_AUFS_EXPORT is disabled, this entry will not
1000 + When the aufs mount option 'noxino' is specified, it
1001 + will be empty. About XINO files, see the aufs manual.
1002 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
1003 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
1004 +++ linux/Documentation/ABI/testing/sysfs-aufs 2015-09-24 10:47:58.244719488 +0200
1006 +What: /sys/fs/aufs/si_<id>/
1008 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1010 + Under /sys/fs/aufs, a directory named si_<id> is created
1011 + per aufs mount, where <id> is a unique id generated
1014 +What: /sys/fs/aufs/si_<id>/br0, br1 ... brN
1016 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1018 + It shows the abolute path of a member directory (which
1019 + is called branch) in aufs, and its permission.
1021 +What: /sys/fs/aufs/si_<id>/brid0, brid1 ... bridN
1023 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1025 + It shows the id of a member directory (which is called
1028 +What: /sys/fs/aufs/si_<id>/xi_path
1030 +Contact: J. R. Okajima <hooanon05g@gmail.com>
1032 + It shows the abolute path of XINO (External Inode Number
1033 + Bitmap, Translation Table and Generation Table) file
1034 + even if it is the default path.
1035 + When the aufs mount option 'noxino' is specified, it
1036 + will be empty. About XINO files, see the aufs manual.
1037 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1038 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
1039 +++ linux/Documentation/filesystems/aufs/design/01intro.txt 2015-09-24 10:47:58.244719488 +0200
1042 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1044 +# This program is free software; you can redistribute it and/or modify
1045 +# it under the terms of the GNU General Public License as published by
1046 +# the Free Software Foundation; either version 2 of the License, or
1047 +# (at your option) any later version.
1049 +# This program is distributed in the hope that it will be useful,
1050 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1051 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1052 +# GNU General Public License for more details.
1054 +# You should have received a copy of the GNU General Public License
1055 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1058 +----------------------------------------
1060 +aufs [ei ju: ef es] | [a u f s]
1061 +1. abbrev. for "advanced multi-layered unification filesystem".
1062 +2. abbrev. for "another unionfs".
1063 +3. abbrev. for "auf das" in German which means "on the" in English.
1064 + Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1065 + But "Filesystem aufs Filesystem" is hard to understand.
1067 +AUFS is a filesystem with features:
1068 +- multi layered stackable unification filesystem, the member directory
1069 + is called as a branch.
1070 +- branch permission and attribute, 'readonly', 'real-readonly',
1071 + 'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
1073 +- internal "file copy-on-write".
1074 +- logical deletion, whiteout.
1075 +- dynamic branch manipulation, adding, deleting and changing permission.
1076 +- allow bypassing aufs, user's direct branch access.
1077 +- external inode number translation table and bitmap which maintains the
1078 + persistent aufs inode number.
1079 +- seekable directory, including NFS readdir.
1080 +- file mapping, mmap and sharing pages.
1081 +- pseudo-link, hardlink over branches.
1082 +- loopback mounted filesystem as a branch.
1083 +- several policies to select one among multiple writable branches.
1084 +- revert a single systemcall when an error occurs in aufs.
1088 +Multi Layered Stackable Unification Filesystem
1089 +----------------------------------------------------------------------
1090 +Most people already knows what it is.
1091 +It is a filesystem which unifies several directories and provides a
1092 +merged single directory. When users access a file, the access will be
1093 +passed/re-directed/converted (sorry, I am not sure which English word is
1094 +correct) to the real file on the member filesystem. The member
1095 +filesystem is called 'lower filesystem' or 'branch' and has a mode
1096 +'readonly' and 'readwrite.' And the deletion for a file on the lower
1097 +readonly branch is handled by creating 'whiteout' on the upper writable
1100 +On LKML, there have been discussions about UnionMount (Jan Blunck,
1101 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1102 +different approaches to implement the merged-view.
1103 +The former tries putting it into VFS, and the latter implements as a
1104 +separate filesystem.
1105 +(If I misunderstand about these implementations, please let me know and
1106 +I shall correct it. Because it is a long time ago when I read their
1107 +source files last time).
1109 +UnionMount's approach will be able to small, but may be hard to share
1110 +branches between several UnionMount since the whiteout in it is
1111 +implemented in the inode on branch filesystem and always
1112 +shared. According to Bharata's post, readdir does not seems to be
1114 +There are several missing features known in this implementations such as
1115 +- for users, the inode number may change silently. eg. copy-up.
1116 +- link(2) may break by copy-up.
1117 +- read(2) may get an obsoleted filedata (fstat(2) too).
1118 +- fcntl(F_SETLK) may be broken by copy-up.
1119 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1122 +In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1123 +merged into mainline. This is another implementation of UnionMount as a
1124 +separated filesystem. All the limitations and known problems which
1125 +UnionMount are equally inherited to "overlay" filesystem.
1127 +Unionfs has a longer history. When I started implementing a stackable
1128 +filesystem (Aug 2005), it already existed. It has virtual super_block,
1129 +inode, dentry and file objects and they have an array pointing lower
1130 +same kind objects. After contributing many patches for Unionfs, I
1131 +re-started my project AUFS (Jun 2006).
1133 +In AUFS, the structure of filesystem resembles to Unionfs, but I
1134 +implemented my own ideas, approaches and enhancements and it became
1135 +totally different one.
1137 +Comparing DM snapshot and fs based implementation
1138 +- the number of bytes to be copied between devices is much smaller.
1139 +- the type of filesystem must be one and only.
1140 +- the fs must be writable, no readonly fs, even for the lower original
1141 + device. so the compression fs will not be usable. but if we use
1142 + loopback mount, we may address this issue.
1144 + mount /cdrom/squashfs.img /sq
1145 + losetup /sq/ext2.img
1146 + losetup /somewhere/cow
1147 + dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1148 +- it will be difficult (or needs more operations) to extract the
1149 + difference between the original device and COW.
1150 +- DM snapshot-merge may help a lot when users try merging. in the
1151 + fs-layer union, users will use rsync(1).
1153 +You may want to read my old paper "Filesystems in LiveCD"
1154 +(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
1157 +Several characters/aspects/persona of aufs
1158 +----------------------------------------------------------------------
1160 +Aufs has several characters, aspects or persona.
1161 +1. a filesystem, callee of VFS helper
1162 +2. sub-VFS, caller of VFS helper for branches
1163 +3. a virtual filesystem which maintains persistent inode number
1164 +4. reader/writer of files on branches such like an application
1166 +1. Callee of VFS Helper
1167 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1168 +unlink(2) from an application reaches sys_unlink() kernel function and
1169 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1170 +calls filesystem specific unlink operation. Actually aufs implements the
1171 +unlink operation but it behaves like a redirector.
1173 +2. Caller of VFS Helper for Branches
1174 +aufs_unlink() passes the unlink request to the branch filesystem as if
1175 +it were called from VFS. So the called unlink operation of the branch
1176 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
1177 +every necessary pre/post operation for the branch filesystem.
1178 +- acquire the lock for the parent dir on a branch
1179 +- lookup in a branch
1180 +- revalidate dentry on a branch
1181 +- mnt_want_write() for a branch
1182 +- vfs_unlink() for a branch
1183 +- mnt_drop_write() for a branch
1184 +- release the lock on a branch
1186 +3. Persistent Inode Number
1187 +One of the most important issue for a filesystem is to maintain inode
1188 +numbers. This is particularly important to support exporting a
1189 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1190 +backend block device for its own. But some storage is necessary to
1191 +keep and maintain the inode numbers. It may be a large space and may not
1192 +suit to keep in memory. Aufs rents some space from its first writable
1193 +branch filesystem (by default) and creates file(s) on it. These files
1194 +are created by aufs internally and removed soon (currently) keeping
1196 +Note: Because these files are removed, they are totally gone after
1197 + unmounting aufs. It means the inode numbers are not persistent
1198 + across unmount or reboot. I have a plan to make them really
1199 + persistent which will be important for aufs on NFS server.
1201 +4. Read/Write Files Internally (copy-on-write)
1202 +Because a branch can be readonly, when you write a file on it, aufs will
1203 +"copy-up" it to the upper writable branch internally. And then write the
1204 +originally requested thing to the file. Generally kernel doesn't
1205 +open/read/write file actively. In aufs, even a single write may cause a
1206 +internal "file copy". This behaviour is very similar to cp(1) command.
1208 +Some people may think it is better to pass such work to user space
1209 +helper, instead of doing in kernel space. Actually I am still thinking
1210 +about it. But currently I have implemented it in kernel space.
1211 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1212 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
1213 +++ linux/Documentation/filesystems/aufs/design/02struct.txt 2015-09-24 10:47:58.244719488 +0200
1216 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1218 +# This program is free software; you can redistribute it and/or modify
1219 +# it under the terms of the GNU General Public License as published by
1220 +# the Free Software Foundation; either version 2 of the License, or
1221 +# (at your option) any later version.
1223 +# This program is distributed in the hope that it will be useful,
1224 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1225 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1226 +# GNU General Public License for more details.
1228 +# You should have received a copy of the GNU General Public License
1229 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1231 +Basic Aufs Internal Structure
1233 +Superblock/Inode/Dentry/File Objects
1234 +----------------------------------------------------------------------
1235 +As like an ordinary filesystem, aufs has its own
1236 +superblock/inode/dentry/file objects. All these objects have a
1237 +dynamically allocated array and store the same kind of pointers to the
1238 +lower filesystem, branch.
1239 +For example, when you build a union with one readwrite branch and one
1240 +readonly, mounted /au, /rw and /ro respectively.
1242 +- /ro/fileA exists but /rw/fileA
1244 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1245 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
1246 +- [0] = NULL (because /rw/fileA doesn't exist)
1249 +This style of an array is essentially same to the aufs
1250 +superblock/inode/dentry/file objects.
1252 +Because aufs supports manipulating branches, ie. add/delete/change
1253 +branches dynamically, these objects has its own generation. When
1254 +branches are changed, the generation in aufs superblock is
1255 +incremented. And a generation in other object are compared when it is
1256 +accessed. When a generation in other objects are obsoleted, aufs
1257 +refreshes the internal array.
1261 +----------------------------------------------------------------------
1262 +Additionally aufs superblock has some data for policies to select one
1263 +among multiple writable branches, XIB files, pseudo-links and kobject.
1264 +See below in detail.
1265 +About the policies which supports copy-down a directory, see
1266 +wbr_policy.txt too.
1269 +Branch and XINO(External Inode Number Translation Table)
1270 +----------------------------------------------------------------------
1271 +Every branch has its own xino (external inode number translation table)
1272 +file. The xino file is created and unlinked by aufs internally. When two
1273 +members of a union exist on the same filesystem, they share the single
1275 +The struct of a xino file is simple, just a sequence of aufs inode
1276 +numbers which is indexed by the lower inode number.
1277 +In the above sample, assume the inode number of /ro/fileA is i111 and
1278 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1279 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1281 +When the inode numbers are not contiguous, the xino file will be sparse
1282 +which has a hole in it and doesn't consume as much disk space as it
1283 +might appear. If your branch filesystem consumes disk space for such
1284 +holes, then you should specify 'xino=' option at mounting aufs.
1286 +Aufs has a mount option to free the disk blocks for such holes in XINO
1287 +files on tmpfs or ramdisk. But it is not so effective actually. If you
1288 +meet a problem of disk shortage due to XINO files, then you should try
1289 +"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1290 +The patch localizes the assignment inumbers per tmpfs-mount and avoid
1291 +the holes in XINO files.
1293 +Also a writable branch has three kinds of "whiteout bases". All these
1294 +are existed when the branch is joined to aufs, and their names are
1295 +whiteout-ed doubly, so that users will never see their names in aufs
1297 +1. a regular file which will be hardlinked to all whiteouts.
1298 +2. a directory to store a pseudo-link.
1299 +3. a directory to store an "orphan"-ed file temporary.
1302 + When you remove a file on a readonly branch, aufs handles it as a
1303 + logical deletion and creates a whiteout on the upper writable branch
1304 + as a hardlink of this file in order not to consume inode on the
1307 + See below, Pseudo-link.
1309 + When "fileC" exists on the lower readonly branch only and it is
1310 + opened and removed with its parent dir, and then user writes
1311 + something into it, then aufs copies-up fileC to this
1312 + directory. Because there is no other dir to store fileC. After
1313 + creating a file under this dir, the file is unlinked.
1315 +Because aufs supports manipulating branches, ie. add/delete/change
1316 +dynamically, a branch has its own id. When the branch order changes,
1317 +aufs finds the new index by searching the branch id.
1321 +----------------------------------------------------------------------
1322 +Assume "fileA" exists on the lower readonly branch only and it is
1323 +hardlinked to "fileB" on the branch. When you write something to fileA,
1324 +aufs copies-up it to the upper writable branch. Additionally aufs
1325 +creates a hardlink under the Pseudo-link Directory of the writable
1326 +branch. The inode of a pseudo-link is kept in aufs super_block as a
1327 +simple list. If fileB is read after unlinking fileA, aufs returns
1328 +filedata from the pseudo-link instead of the lower readonly
1329 +branch. Because the pseudo-link is based upon the inode, to keep the
1330 +inode number by xino (see above) is essentially necessary.
1332 +All the hardlinks under the Pseudo-link Directory of the writable branch
1333 +should be restored in a proper location later. Aufs provides a utility
1334 +to do this. The userspace helpers executed at remounting and unmounting
1336 +During this utility is running, it puts aufs into the pseudo-link
1337 +maintenance mode. In this mode, only the process which began the
1338 +maintenance mode (and its child processes) is allowed to operate in
1339 +aufs. Some other processes which are not related to the pseudo-link will
1340 +be allowed to run too, but the rest have to return an error or wait
1341 +until the maintenance mode ends. If a process already acquires an inode
1342 +mutex (in VFS), it has to return an error.
1345 +XIB(external inode number bitmap)
1346 +----------------------------------------------------------------------
1347 +Addition to the xino file per a branch, aufs has an external inode number
1348 +bitmap in a superblock object. It is also an internal file such like a
1350 +It is a simple bitmap to mark whether the aufs inode number is in-use or
1352 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
1354 +As well as XINO files, aufs has a feature to truncate/refresh XIB to
1355 +reduce the number of consumed disk blocks for these files.
1358 +Virtual or Vertical Dir, and Readdir in Userspace
1359 +----------------------------------------------------------------------
1360 +In order to support multiple layers (branches), aufs readdir operation
1361 +constructs a virtual dir block on memory. For readdir, aufs calls
1362 +vfs_readdir() internally for each dir on branches, merges their entries
1363 +with eliminating the whiteout-ed ones, and sets it to file (dir)
1364 +object. So the file object has its entry list until it is closed. The
1365 +entry list will be updated when the file position is zero and becomes
1366 +obsoleted. This decision is made in aufs automatically.
1368 +The dynamically allocated memory block for the name of entries has a
1369 +unit of 512 bytes (by default) and stores the names contiguously (no
1370 +padding). Another block for each entry is handled by kmem_cache too.
1371 +During building dir blocks, aufs creates hash list and judging whether
1372 +the entry is whiteouted by its upper branch or already listed.
1373 +The merged result is cached in the corresponding inode object and
1374 +maintained by a customizable life-time option.
1376 +Some people may call it can be a security hole or invite DoS attack
1377 +since the opened and once readdir-ed dir (file object) holds its entry
1378 +list and becomes a pressure for system memory. But I'd say it is similar
1379 +to files under /proc or /sys. The virtual files in them also holds a
1380 +memory page (generally) while they are opened. When an idea to reduce
1381 +memory for them is introduced, it will be applied to aufs too.
1382 +For those who really hate this situation, I've developed readdir(3)
1383 +library which operates this merging in userspace. You just need to set
1384 +LD_PRELOAD environment variable, and aufs will not consume no memory in
1385 +kernel space for readdir(3).
1389 +----------------------------------------------------------------------
1390 +Aufs sometimes requires privilege access to a branch. For instance,
1391 +in copy-up/down operation. When a user process is going to make changes
1392 +to a file which exists in the lower readonly branch only, and the mode
1393 +of one of ancestor directories may not be writable by a user
1394 +process. Here aufs copy-up the file with its ancestors and they may
1395 +require privilege to set its owner/group/mode/etc.
1396 +This is a typical case of a application character of aufs (see
1399 +Aufs uses workqueue synchronously for this case. It creates its own
1400 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
1401 +passes the request to call mkdir or write (for example), and wait for
1402 +its completion. This approach solves a problem of a signal handler
1404 +If aufs didn't adopt the workqueue and changed the privilege of the
1405 +process, then the process may receive the unexpected SIGXFSZ or other
1408 +Also aufs uses the system global workqueue ("events" kernel thread) too
1409 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1410 +whiteout base and etc. This is unrelated to a privilege.
1411 +Most of aufs operation tries acquiring a rw_semaphore for aufs
1412 +superblock at the beginning, at the same time waits for the completion
1413 +of all queued asynchronous tasks.
1417 +----------------------------------------------------------------------
1418 +The whiteout in aufs is very similar to Unionfs's. That is represented
1419 +by its filename. UnionMount takes an approach of a file mode, but I am
1420 +afraid several utilities (find(1) or something) will have to support it.
1422 +Basically the whiteout represents "logical deletion" which stops aufs to
1423 +lookup further, but also it represents "dir is opaque" which also stop
1426 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1427 +In order to make several functions in a single systemcall to be
1428 +revertible, aufs adopts an approach to rename a directory to a temporary
1429 +unique whiteouted name.
1430 +For example, in rename(2) dir where the target dir already existed, aufs
1431 +renames the target dir to a temporary unique whiteouted name before the
1432 +actual rename on a branch, and then handles other actions (make it opaque,
1433 +update the attributes, etc). If an error happens in these actions, aufs
1434 +simply renames the whiteouted name back and returns an error. If all are
1435 +succeeded, aufs registers a function to remove the whiteouted unique
1436 +temporary name completely and asynchronously to the system global
1441 +----------------------------------------------------------------------
1442 +It is a well-known feature or concept.
1443 +When user modifies a file on a readonly branch, aufs operate "copy-up"
1444 +internally and makes change to the new file on the upper writable branch.
1445 +When the trigger systemcall does not update the timestamps of the parent
1446 +dir, aufs reverts it after copy-up.
1449 +Move-down (aufs3.9 and later)
1450 +----------------------------------------------------------------------
1451 +"Copy-up" is one of the essential feature in aufs. It copies a file from
1452 +the lower readonly branch to the upper writable branch when a user
1453 +changes something about the file.
1454 +"Move-down" is an opposite action of copy-up. Basically this action is
1455 +ran manually instead of automatically and internally.
1456 +For desgin and implementation, aufs has to consider these issues.
1457 +- whiteout for the file may exist on the lower branch.
1458 +- ancestor directories may not exist on the lower branch.
1459 +- diropq for the ancestor directories may exist on the upper branch.
1460 +- free space on the lower branch will reduce.
1461 +- another access to the file may happen during moving-down, including
1462 + UDBA (see "Revalidate Dentry and UDBA").
1463 +- the file should not be hard-linked nor pseudo-linked. they should be
1464 + handled by auplink utility later.
1466 +Sometimes users want to move-down a file from the upper writable branch
1467 +to the lower readonly or writable branch. For instance,
1468 +- the free space of the upper writable branch is going to run out.
1469 +- create a new intermediate branch between the upper and lower branch.
1472 +For this purpose, use "aumvdown" command in aufs-util.git.
1473 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1474 --- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt 1970-01-01 01:00:00.000000000 +0100
1475 +++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt 2015-09-24 10:47:58.244719488 +0200
1478 +# Copyright (C) 2015 Junjiro R. Okajima
1480 +# This program is free software; you can redistribute it and/or modify
1481 +# it under the terms of the GNU General Public License as published by
1482 +# the Free Software Foundation; either version 2 of the License, or
1483 +# (at your option) any later version.
1485 +# This program is distributed in the hope that it will be useful,
1486 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1487 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1488 +# GNU General Public License for more details.
1490 +# You should have received a copy of the GNU General Public License
1491 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1493 +Support for a branch who has its ->atomic_open()
1494 +----------------------------------------------------------------------
1495 +The filesystems who implement its ->atomic_open() are not majority. For
1496 +example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1497 +particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1498 +->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1499 +sure whether all filesystems who have ->atomic_open() behave like this,
1500 +but NFSv4 surely returns the error.
1502 +In order to support ->atomic_open() for aufs, there are a few
1505 +A. Introduce aufs_atomic_open()
1506 + - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1508 +B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1509 + an aufs user Pip Cet's approach
1510 + - calls aufs_create(), VFS finish_open() and notify_change().
1511 + - pass fake-mode to finish_open(), and then correct the mode by
1513 +C. Extend aufs_open() to call branch fs's ->atomic_open()
1514 + - no aufs_atomic_open().
1515 + - aufs_lookup() registers the TID to an aufs internal object.
1516 + - aufs_create() does nothing when the matching TID is registered, but
1517 + registers the mode.
1518 + - aufs_open() calls branch fs's ->atomic_open() when the matching
1519 + TID is registered.
1520 +D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1522 + - no aufs_atomic_open().
1523 + - aufs_create() registers the TID to an internal object. this info
1524 + represents "this process created this file just now."
1525 + - when aufs gets EACCES from branch fs's ->open(), then confirm the
1526 + registered TID and re-try open() with superuser's credential.
1528 +Pros and cons for each approach.
1531 + - straightforward but highly depends upon VFS internal.
1532 + - the atomic behavaiour is kept.
1533 + - some of parameters such as nameidata are hard to reproduce for
1537 + - easy to implement.
1538 + - the atomic behavaiour is lost.
1540 + - the atomic behavaiour is kept.
1541 + - dirty and tricky.
1542 + - VFS checks whether the file is created correctly after calling
1543 + ->create(), which means this approach doesn't work.
1545 + - easy to implement.
1546 + - the atomic behavaiour is lost.
1547 + - to open a file with superuser's credential and give it to a user
1548 + process is a bad idea, since the file object keeps the credential
1549 + in it. It may affect LSM or something. This approach doesn't work
1552 +The approach A is ideal, but it hard to implement. So here is a
1553 +variation of A, which is to be implemented.
1555 +A-1. Introduce aufs_atomic_open()
1556 + - calls branch fs ->atomic_open() if exists. otherwise calls
1557 + vfs_create() and finish_open().
1558 + - the demerit is that the several checks after branch fs
1559 + ->atomic_open() are lost. in the ordinary case, the checks are
1560 + done by VFS:do_last(), lookup_open() and atomic_open(). some can
1561 + be implemented in aufs, but not all I am afraid.
1562 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1563 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
1564 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2015-09-24 10:47:58.244719488 +0200
1567 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1569 +# This program is free software; you can redistribute it and/or modify
1570 +# it under the terms of the GNU General Public License as published by
1571 +# the Free Software Foundation; either version 2 of the License, or
1572 +# (at your option) any later version.
1574 +# This program is distributed in the hope that it will be useful,
1575 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1576 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1577 +# GNU General Public License for more details.
1579 +# You should have received a copy of the GNU General Public License
1580 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1583 +----------------------------------------------------------------------
1584 +Since aufs has a character of sub-VFS (see Introduction), it operates
1585 +lookup for branches as VFS does. It may be a heavy work. But almost all
1586 +lookup operation in aufs is the simplest case, ie. lookup only an entry
1587 +directly connected to its parent. Digging down the directory hierarchy
1588 +is unnecessary. VFS has a function lookup_one_len() for that use, and
1591 +When a branch is a remote filesystem, aufs basically relies upon its
1592 +->d_revalidate(), also aufs forces the hardest revalidate tests for
1594 +For d_revalidate, aufs implements three levels of revalidate tests. See
1595 +"Revalidate Dentry and UDBA" in detail.
1598 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1599 +----------------------------------------------------------------------
1600 +Let's try case study.
1601 +- aufs has two branches, upper readwrite and lower readonly.
1603 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1604 +- user invoked "chmod a+rx /au/dirA"
1605 +- the internal copy-up is activated and "/rw/dirA" is created and its
1606 + permission bits are set to world readable.
1607 +- then "/au/dirA" becomes world readable?
1609 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1610 +or it may be a natively readonly filesystem. If aufs respects the lower
1611 +branch, it should not respond readdir request from other users. But user
1612 +allowed it by chmod. Should really aufs rejects showing the entries
1615 +To be honest, I don't have a good solution for this case. So aufs
1616 +implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1618 +When dirperm1 is specified, aufs checks only the highest one for the
1619 +directory permission, and shows the entries. Otherwise, as usual, checks
1620 +every dir existing on all branches and rejects the request.
1622 +As a side effect, dirperm1 option improves the performance of aufs
1623 +because the number of permission check is reduced when the number of
1627 +Revalidate Dentry and UDBA (User's Direct Branch Access)
1628 +----------------------------------------------------------------------
1629 +Generally VFS helpers re-validate a dentry as a part of lookup.
1630 +0. digging down the directory hierarchy.
1631 +1. lock the parent dir by its i_mutex.
1632 +2. lookup the final (child) entry.
1634 +4. call the actual operation (create, unlink, etc.)
1635 +5. unlock the parent dir
1637 +If the filesystem implements its ->d_revalidate() (step 3), then it is
1638 +called. Actually aufs implements it and checks the dentry on a branch is
1640 +But it is not enough. Because aufs has to release the lock for the
1641 +parent dir on a branch at the end of ->lookup() (step 2) and
1642 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1644 +If the file on a branch is changed directly, eg. bypassing aufs, after
1645 +aufs released the lock, then the subsequent operation may cause
1646 +something unpleasant result.
1648 +This situation is a result of VFS architecture, ->lookup() and
1649 +->d_revalidate() is separated. But I never say it is wrong. It is a good
1650 +design from VFS's point of view. It is just not suitable for sub-VFS
1653 +Aufs supports such case by three level of revalidation which is
1654 +selectable by user.
1655 +1. Simple Revalidate
1656 + Addition to the native flow in VFS's, confirm the child-parent
1657 + relationship on the branch just after locking the parent dir on the
1658 + branch in the "actual operation" (step 4). When this validation
1659 + fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1660 + checks the validation of the dentry on branches.
1661 +2. Monitor Changes Internally by Inotify/Fsnotify
1662 + Addition to above, in the "actual operation" (step 4) aufs re-lookup
1663 + the dentry on the branch, and returns EBUSY if it finds different
1665 + Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1666 + during it is in cache. When the event is notified, aufs registers a
1667 + function to kernel 'events' thread by schedule_work(). And the
1668 + function sets some special status to the cached aufs dentry and inode
1669 + private data. If they are not cached, then aufs has nothing to
1670 + do. When the same file is accessed through aufs (step 0-3) later,
1671 + aufs will detect the status and refresh all necessary data.
1672 + In this mode, aufs has to ignore the event which is fired by aufs
1674 +3. No Extra Validation
1675 + This is the simplest test and doesn't add any additional revalidation
1676 + test, and skip the revalidation in step 4. It is useful and improves
1677 + aufs performance when system surely hide the aufs branches from user,
1678 + by over-mounting something (or another method).
1679 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1680 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
1681 +++ linux/Documentation/filesystems/aufs/design/04branch.txt 2015-09-24 10:47:58.244719488 +0200
1684 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1686 +# This program is free software; you can redistribute it and/or modify
1687 +# it under the terms of the GNU General Public License as published by
1688 +# the Free Software Foundation; either version 2 of the License, or
1689 +# (at your option) any later version.
1691 +# This program is distributed in the hope that it will be useful,
1692 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1693 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1694 +# GNU General Public License for more details.
1696 +# You should have received a copy of the GNU General Public License
1697 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1699 +Branch Manipulation
1701 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1702 +and changing its permission/attribute, there are a lot of works to do.
1706 +----------------------------------------------------------------------
1707 +o Confirm the adding dir exists outside of aufs, including loopback
1708 + mount, and its various attributes.
1709 +o Initialize the xino file and whiteout bases if necessary.
1712 +o Check the owner/group/mode of the directory
1713 + When the owner/group/mode of the adding directory differs from the
1714 + existing branch, aufs issues a warning because it may impose a
1716 + For example, when a upper writable branch has a world writable empty
1717 + top directory, a malicious user can create any files on the writable
1718 + branch directly, like copy-up and modify manually. If something like
1719 + /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1720 + writable branch, and the writable branch is world-writable, then a
1721 + malicious guy may create /etc/passwd on the writable branch directly
1722 + and the infected file will be valid in aufs.
1723 + I am afraid it can be a security issue, but aufs can do nothing except
1724 + producing a warning.
1728 +----------------------------------------------------------------------
1729 +o Confirm the deleting branch is not busy
1730 + To be general, there is one merit to adopt "remount" interface to
1731 + manipulate branches. It is to discard caches. At deleting a branch,
1732 + aufs checks the still cached (and connected) dentries and inodes. If
1733 + there are any, then they are all in-use. An inode without its
1734 + corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1736 + For the cached one, aufs checks whether the same named entry exists on
1738 + If the cached one is a directory, because aufs provides a merged view
1739 + to users, as long as one dir is left on any branch aufs can show the
1740 + dir to users. In this case, the branch can be removed from aufs.
1741 + Otherwise aufs rejects deleting the branch.
1743 + If any file on the deleting branch is opened by aufs, then aufs
1747 +Modify the Permission of a Branch
1748 +----------------------------------------------------------------------
1749 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1752 +o rw --> ro: Confirm the modifying branch is not busy
1753 + Aufs rejects the request if any of these conditions are true.
1754 + - a file on the branch is mmap-ed.
1755 + - a regular file on the branch is opened for write and there is no
1756 + same named entry on the upper branch.
1757 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1758 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
1759 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2015-09-24 10:47:58.244719488 +0200
1762 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1764 +# This program is free software; you can redistribute it and/or modify
1765 +# it under the terms of the GNU General Public License as published by
1766 +# the Free Software Foundation; either version 2 of the License, or
1767 +# (at your option) any later version.
1769 +# This program is distributed in the hope that it will be useful,
1770 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1771 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1772 +# GNU General Public License for more details.
1774 +# You should have received a copy of the GNU General Public License
1775 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1777 +Policies to Select One among Multiple Writable Branches
1778 +----------------------------------------------------------------------
1779 +When the number of writable branch is more than one, aufs has to decide
1780 +the target branch for file creation or copy-up. By default, the highest
1781 +writable branch which has the parent (or ancestor) dir of the target
1782 +file is chosen (top-down-parent policy).
1783 +By user's request, aufs implements some other policies to select the
1784 +writable branch, for file creation several policies, round-robin,
1785 +most-free-space, and other policies. For copy-up, top-down-parent,
1786 +bottom-up-parent, bottom-up and others.
1788 +As expected, the round-robin policy selects the branch in circular. When
1789 +you have two writable branches and creates 10 new files, 5 files will be
1790 +created for each branch. mkdir(2) systemcall is an exception. When you
1791 +create 10 new directories, all will be created on the same branch.
1792 +And the most-free-space policy selects the one which has most free
1793 +space among the writable branches. The amount of free space will be
1794 +checked by aufs internally, and users can specify its time interval.
1796 +The policies for copy-up is more simple,
1797 +top-down-parent is equivalent to the same named on in create policy,
1798 +bottom-up-parent selects the writable branch where the parent dir
1799 +exists and the nearest upper one from the copyup-source,
1800 +bottom-up selects the nearest upper writable branch from the
1801 +copyup-source, regardless the existence of the parent dir.
1803 +There are some rules or exceptions to apply these policies.
1804 +- If there is a readonly branch above the policy-selected branch and
1805 + the parent dir is marked as opaque (a variation of whiteout), or the
1806 + target (creating) file is whiteout-ed on the upper readonly branch,
1807 + then the result of the policy is ignored and the target file will be
1808 + created on the nearest upper writable branch than the readonly branch.
1809 +- If there is a writable branch above the policy-selected branch and
1810 + the parent dir is marked as opaque or the target file is whiteouted
1811 + on the branch, then the result of the policy is ignored and the target
1812 + file will be created on the highest one among the upper writable
1813 + branches who has diropq or whiteout. In case of whiteout, aufs removes
1815 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1816 + They try selecting the branch where the source exists as possible
1817 + since copyup a large file will take long time. If it can't be,
1818 + ie. the branch where the source exists is readonly, then they will
1819 + follow the copyup policy.
1820 +- There is an exception for rename(2) when the target exists.
1821 + If the rename target exists, aufs compares the index of the branches
1822 + where the source and the target exists and selects the higher
1823 + one. If the selected branch is readonly, then aufs follows the
1825 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1826 --- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt 1970-01-01 01:00:00.000000000 +0100
1827 +++ linux/Documentation/filesystems/aufs/design/06fhsm.txt 2015-09-24 10:47:58.244719488 +0200
1830 +# Copyright (C) 2011-2015 Junjiro R. Okajima
1832 +# This program is free software; you can redistribute it and/or modify
1833 +# it under the terms of the GNU General Public License as published by
1834 +# the Free Software Foundation; either version 2 of the License, or
1835 +# (at your option) any later version.
1837 +# This program is distributed in the hope that it will be useful,
1838 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1839 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1840 +# GNU General Public License for more details.
1842 +# You should have received a copy of the GNU General Public License
1843 +# along with this program; if not, write to the Free Software
1844 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1847 +File-based Hierarchical Storage Management (FHSM)
1848 +----------------------------------------------------------------------
1849 +Hierarchical Storage Management (or HSM) is a well-known feature in the
1850 +storage world. Aufs provides this feature as file-based with multiple
1851 +writable branches, based upon the principle of "Colder, the Lower".
1852 +Here the word "colder" means that the less used files, and "lower" means
1853 +that the position in the order of the stacked branches vertically.
1854 +These multiple writable branches are prioritized, ie. the topmost one
1855 +should be the fastest drive and be used heavily.
1857 +o Characters in aufs FHSM story
1858 +- aufs itself and a new branch attribute.
1859 +- a new ioctl interface to move-down and to establish a connection with
1860 + the daemon ("move-down" is a converse of "copy-up").
1861 +- userspace tool and daemon.
1863 +The userspace daemon establishes a connection with aufs and waits for
1864 +the notification. The notified information is very similar to struct
1865 +statfs containing the number of consumed blocks and inodes.
1866 +When the consumed blocks/inodes of a branch exceeds the user-specified
1867 +upper watermark, the daemon activates its move-down process until the
1868 +consumed blocks/inodes reaches the user-specified lower watermark.
1870 +The actual move-down is done by aufs based upon the request from
1871 +user-space since we need to maintain the inode number and the internal
1872 +pointer arrays in aufs.
1874 +Currently aufs FHSM handles the regular files only. Additionally they
1875 +must not be hard-linked nor pseudo-linked.
1878 +o Cowork of aufs and the user-space daemon
1879 + During the userspace daemon established the connection, aufs sends a
1880 + small notification to it whenever aufs writes something into the
1881 + writable branch. But it may cost high since aufs issues statfs(2)
1882 + internally. So user can specify a new option to cache the
1883 + info. Actually the notification is controlled by these factors.
1884 + + the specified cache time.
1885 + + classified as "force" by aufs internally.
1886 + Until the specified time expires, aufs doesn't send the info
1887 + except the forced cases. When aufs decide forcing, the info is always
1888 + notified to userspace.
1889 + For example, the number of free inodes is generally large enough and
1890 + the shortage of it happens rarely. So aufs doesn't force the
1891 + notification when creating a new file, directory and others. This is
1892 + the typical case which aufs doesn't force.
1893 + When aufs writes the actual filedata and the files consumes any of new
1894 + blocks, the aufs forces notifying.
1897 +o Interfaces in aufs
1898 +- New branch attribute.
1900 + Specifies that the branch is managed by FHSM feature. In other word,
1901 + participant in the FHSM.
1902 + When nofhsm is set to the branch, it will not be the source/target
1903 + branch of the move-down operation. This attribute is set
1904 + independently from coo and moo attributes, and if you want full
1905 + FHSM, you should specify them as well.
1906 +- New mount option.
1908 + Specifies a second to suppress many less important info to be
1911 + + AUFS_CTL_FHSM_FD
1912 + create a new file descriptor which userspace can read the notification
1913 + (a subset of struct statfs) from aufs.
1914 +- Module parameter 'brs'
1915 + It has to be set to 1. Otherwise the new mount option 'fhsm' will not
1917 +- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
1918 + When there are two or more branches with fhsm attributes,
1919 + /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
1920 + terminates it. As a result of remounting and branch-manipulation, the
1921 + number of branches with fhsm attribute can be one. In this case,
1922 + /sbin/mount.aufs will terminate the user-space daemon.
1925 +Finally the operation is done as these steps in kernel-space.
1927 + + no one else is using the file.
1928 + + the file is not hard-linked.
1929 + + the file is not pseudo-linked.
1930 + + the file is a regular file.
1931 + + the parent dir is not opaqued.
1932 +- find the target writable branch.
1933 +- make sure the file is not whiteout-ed by the upper (than the target)
1935 +- make the parent dir on the target branch.
1936 +- mutex lock the inode on the branch.
1937 +- unlink the whiteout on the target branch (if exists).
1938 +- lookup and create the whiteout-ed temporary name on the target branch.
1939 +- copy the file as the whiteout-ed temporary name on the target branch.
1940 +- rename the whiteout-ed temporary name to the original name.
1941 +- unlink the file on the source branch.
1942 +- maintain the internal pointer array and the external inode number
1944 +- maintain the timestamps and other attributes of the parent dir and the
1947 +And of course, in every step, an error may happen. So the operation
1948 +should restore the original file state after an error happens.
1949 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
1950 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
1951 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2015-09-24 10:47:58.244719488 +0200
1954 +# Copyright (C) 2005-2015 Junjiro R. Okajima
1956 +# This program is free software; you can redistribute it and/or modify
1957 +# it under the terms of the GNU General Public License as published by
1958 +# the Free Software Foundation; either version 2 of the License, or
1959 +# (at your option) any later version.
1961 +# This program is distributed in the hope that it will be useful,
1962 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1963 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1964 +# GNU General Public License for more details.
1966 +# You should have received a copy of the GNU General Public License
1967 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
1969 +mmap(2) -- File Memory Mapping
1970 +----------------------------------------------------------------------
1971 +In aufs, the file-mapped pages are handled by a branch fs directly, no
1972 +interaction with aufs. It means aufs_mmap() calls the branch fs's
1974 +This approach is simple and good, but there is one problem.
1975 +Under /proc, several entries show the mmapped files by its path (with
1976 +device and inode number), and the printed path will be the path on the
1977 +branch fs's instead of virtual aufs's.
1978 +This is not a problem in most cases, but some utilities lsof(1) (and its
1979 +user) may expect the path on aufs.
1981 +To address this issue, aufs adds a new member called vm_prfile in struct
1982 +vm_area_struct (and struct vm_region). The original vm_file points to
1983 +the file on the branch fs in order to handle everything correctly as
1984 +usual. The new vm_prfile points to a virtual file in aufs, and the
1985 +show-functions in procfs refers to vm_prfile if it is set.
1986 +Also we need to maintain several other places where touching vm_file
1988 +- fork()/clone() copies vma and the reference count of vm_file is
1990 +- merging vma maintains the ref count too.
1992 +This is not a good approach. It just fakes the printed path. But it
1993 +leaves all behaviour around f_mapping unchanged. This is surely an
1995 +Actually aufs had adopted another complicated approach which calls
1996 +generic_file_mmap() and handles struct vm_operations_struct. In this
1997 +approach, aufs met a hard problem and I could not solve it without
1998 +switching the approach.
2000 +There may be one more another approach which is
2001 +- bind-mount the branch-root onto the aufs-root internally
2002 +- grab the new vfsmount (ie. struct mount)
2003 +- lazy-umount the branch-root internally
2004 +- in open(2) the aufs-file, open the branch-file with the hidden
2005 + vfsmount (instead of the original branch's vfsmount)
2006 +- ideally this "bind-mount and lazy-umount" should be done atomically,
2007 + but it may be possible from userspace by the mount helper.
2009 +Adding the internal hidden vfsmount and using it in opening a file, the
2010 +file path under /proc will be printed correctly. This approach looks
2011 +smarter, but is not possible I am afraid.
2012 +- aufs-root may be bind-mount later. when it happens, another hidden
2013 + vfsmount will be required.
2014 +- it is hard to get the chance to bind-mount and lazy-umount
2015 + + in kernel-space, FS can have vfsmount in open(2) via
2016 + file->f_path, and aufs can know its vfsmount. But several locks are
2017 + already acquired, and if aufs tries to bind-mount and lazy-umount
2018 + here, then it may cause a deadlock.
2019 + + in user-space, bind-mount doesn't invoke the mount helper.
2020 +- since /proc shows dev and ino, aufs has to give vma these info. it
2021 + means a new member vm_prinode will be necessary. this is essentially
2022 + equivalent to vm_prfile described above.
2024 +I have to give up this "looks-smater" approach.
2025 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2026 --- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt 1970-01-01 01:00:00.000000000 +0100
2027 +++ linux/Documentation/filesystems/aufs/design/06xattr.txt 2015-09-24 10:47:58.244719488 +0200
2030 +# Copyright (C) 2014-2015 Junjiro R. Okajima
2032 +# This program is free software; you can redistribute it and/or modify
2033 +# it under the terms of the GNU General Public License as published by
2034 +# the Free Software Foundation; either version 2 of the License, or
2035 +# (at your option) any later version.
2037 +# This program is distributed in the hope that it will be useful,
2038 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2039 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2040 +# GNU General Public License for more details.
2042 +# You should have received a copy of the GNU General Public License
2043 +# along with this program; if not, write to the Free Software
2044 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2047 +Listing XATTR/EA and getting the value
2048 +----------------------------------------------------------------------
2049 +For the inode standard attributes (owner, group, timestamps, etc.), aufs
2050 +shows the values from the topmost existing file. This behaviour is good
2051 +for the non-dir entries since the bahaviour exactly matches the shown
2052 +information. But for the directories, aufs considers all the same named
2053 +entries on the lower branches. Which means, if one of the lower entry
2054 +rejects readdir call, then aufs returns an error even if the topmost
2055 +entry allows it. This behaviour is necessary to respect the branch fs's
2056 +security, but can make users confused since the user-visible standard
2057 +attributes don't match the behaviour.
2058 +To address this issue, aufs has a mount option called dirperm1 which
2059 +checks the permission for the topmost entry only, and ignores the lower
2060 +entry's permission.
2062 +A similar issue can happen around XATTR.
2063 +getxattr(2) and listxattr(2) families behave as if dirperm1 option is
2064 +always set. Otherwise these very unpleasant situation would happen.
2065 +- listxattr(2) may return the duplicated entries.
2066 +- users may not be able to remove or reset the XATTR forever,
2069 +XATTR/EA support in the internal (copy,move)-(up,down)
2070 +----------------------------------------------------------------------
2071 +Generally the extended attributes of inode are categorized as these.
2072 +- "security" for LSM and capability.
2073 +- "system" for posix ACL, 'acl' mount option is required for the branch
2075 +- "trusted" for userspace, CAP_SYS_ADMIN is required.
2076 +- "user" for userspace, 'user_xattr' mount option is required for the
2077 + branch fs generally.
2079 +Moreover there are some other categories. Aufs handles these rather
2080 +unpopular categories as the ordinary ones, ie. there is no special
2081 +condition nor exception.
2083 +In copy-up, the support for XATTR on the dst branch may differ from the
2084 +src branch. In this case, the copy-up operation will get an error and
2085 +the original user operation which triggered the copy-up will fail. It
2086 +can happen that even all copy-up will fail.
2087 +When both of src and dst branches support XATTR and if an error occurs
2088 +during copying XATTR, then the copy-up should fail obviously. That is a
2089 +good reason and aufs should return an error to userspace. But when only
2090 +the src branch support that XATTR, aufs should not return an error.
2091 +For example, the src branch supports ACL but the dst branch doesn't
2092 +because the dst branch may natively un-support it or temporary
2093 +un-support it due to "noacl" mount option. Of course, the dst branch fs
2094 +may NOT return an error even if the XATTR is not supported. It is
2095 +totally up to the branch fs.
2097 +Anyway when the aufs internal copy-up gets an error from the dst branch
2098 +fs, then aufs tries removing the just copied entry and returns the error
2099 +to the userspace. The worst case of this situation will be all copy-up
2102 +For the copy-up operation, there two basic approaches.
2103 +- copy the specified XATTR only (by category above), and return the
2104 + error unconditionally if it happens.
2105 +- copy all XATTR, and ignore the error on the specified category only.
2107 +In order to support XATTR and to implement the correct behaviour, aufs
2108 +chooses the latter approach and introduces some new branch attributes,
2109 +"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
2110 +They correspond to the XATTR namespaces (see above). Additionally, to be
2111 +convenient, "icex" is also provided which means all "icex*" attributes
2112 +are set (here the word "icex" stands for "ignore copy-error on XATTR").
2114 +The meaning of these attributes is to ignore the error from setting
2115 +XATTR on that branch.
2116 +Note that aufs tries copying all XATTR unconditionally, and ignores the
2117 +error from the dst branch according to the specified attributes.
2119 +Some XATTR may have its default value. The default value may come from
2120 +the parent dir or the environment. If the default value is set at the
2121 +file creating-time, it will be overwritten by copy-up.
2122 +Some contradiction may happen I am afraid.
2123 +Do we need another attribute to stop copying XATTR? I am unsure. For
2124 +now, aufs implements the branch attributes to ignore the error.
2125 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2126 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
2127 +++ linux/Documentation/filesystems/aufs/design/07export.txt 2015-09-24 10:47:58.248052907 +0200
2130 +# Copyright (C) 2005-2015 Junjiro R. Okajima
2132 +# This program is free software; you can redistribute it and/or modify
2133 +# it under the terms of the GNU General Public License as published by
2134 +# the Free Software Foundation; either version 2 of the License, or
2135 +# (at your option) any later version.
2137 +# This program is distributed in the hope that it will be useful,
2138 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2139 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2140 +# GNU General Public License for more details.
2142 +# You should have received a copy of the GNU General Public License
2143 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2145 +Export Aufs via NFS
2146 +----------------------------------------------------------------------
2147 +Here is an approach.
2148 +- like xino/xib, add a new file 'xigen' which stores aufs inode
2150 +- iget_locked(): initialize aufs inode generation for a new inode, and
2151 + store it in xigen file.
2152 +- destroy_inode(): increment aufs inode generation and store it in xigen
2153 + file. it is necessary even if it is not unlinked, because any data of
2154 + inode may be changed by UDBA.
2155 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2156 + build file handle by
2157 + + branch id (4 bytes)
2158 + + superblock generation (4 bytes)
2159 + + inode number (4 or 8 bytes)
2160 + + parent dir inode number (4 or 8 bytes)
2161 + + inode generation (4 bytes))
2162 + + return value of exportfs_encode_fh() for the parent on a branch (4
2164 + + file handle for a branch (by exportfs_encode_fh())
2166 + + find the index of a branch from its id in handle, and check it is
2167 + still exist in aufs.
2168 + + 1st level: get the inode number from handle and search it in cache.
2169 + + 2nd level: if not found in cache, get the parent inode number from
2170 + the handle and search it in cache. and then open the found parent
2171 + dir, find the matching inode number by vfs_readdir() and get its
2172 + name, and call lookup_one_len() for the target dentry.
2173 + + 3rd level: if the parent dir is not cached, call
2174 + exportfs_decode_fh() for a branch and get the parent on a branch,
2175 + build a pathname of it, convert it a pathname in aufs, call
2176 + path_lookup(). now aufs gets a parent dir dentry, then handle it as
2178 + + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2179 + for every branch, but not itself. to get this, (currently) aufs
2180 + searches in current->nsproxy->mnt_ns list. it may not be a good
2181 + idea, but I didn't get other approach.
2182 + + test the generation of the gotten inode.
2183 +- every inode operation: they may get EBUSY due to UDBA. in this case,
2184 + convert it into ESTALE for NFSD.
2185 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
2186 + lookup_one_len(), vfs_getattr(), encode_fh() and others.
2187 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2188 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
2189 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2015-09-24 10:47:58.248052907 +0200
2192 +# Copyright (C) 2005-2015 Junjiro R. Okajima
2194 +# This program is free software; you can redistribute it and/or modify
2195 +# it under the terms of the GNU General Public License as published by
2196 +# the Free Software Foundation; either version 2 of the License, or
2197 +# (at your option) any later version.
2199 +# This program is distributed in the hope that it will be useful,
2200 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2201 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2202 +# GNU General Public License for more details.
2204 +# You should have received a copy of the GNU General Public License
2205 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2207 +Show Whiteout Mode (shwh)
2208 +----------------------------------------------------------------------
2209 +Generally aufs hides the name of whiteouts. But in some cases, to show
2210 +them is very useful for users. For instance, creating a new middle layer
2211 +(branch) by merging existing layers.
2213 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
2214 +When you have three branches,
2215 +- Bottom: 'system', squashfs (underlying base system), read-only
2216 +- Middle: 'mods', squashfs, read-only
2217 +- Top: 'overlay', ram (tmpfs), read-write
2219 +The top layer is loaded at boot time and saved at shutdown, to preserve
2220 +the changes made to the system during the session.
2221 +When larger changes have been made, or smaller changes have accumulated,
2222 +the size of the saved top layer data grows. At this point, it would be
2223 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
2224 +and rewrite the 'mods' squashfs, clearing the top layer and thus
2225 +restoring save and load speed.
2227 +This merging is simplified by the use of another aufs mount, of just the
2228 +two overlay branches using the 'shwh' option.
2229 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2230 + aufs /livesys/merge_union
2232 +A merged view of these two branches is then available at
2233 +/livesys/merge_union, and the new feature is that the whiteouts are
2235 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2236 +writing to all branches. Also the default mode for all branches is 'ro'.
2237 +It is now possible to save the combined contents of the two overlay
2238 +branches to a new squashfs, e.g.:
2239 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
2241 +This new squashfs archive can be stored on the boot device and the
2242 +initramfs will use it to replace the old one at the next boot.
2243 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2244 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
2245 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2015-09-24 10:47:58.248052907 +0200
2248 +# Copyright (C) 2010-2015 Junjiro R. Okajima
2250 +# This program is free software; you can redistribute it and/or modify
2251 +# it under the terms of the GNU General Public License as published by
2252 +# the Free Software Foundation; either version 2 of the License, or
2253 +# (at your option) any later version.
2255 +# This program is distributed in the hope that it will be useful,
2256 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
2257 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2258 +# GNU General Public License for more details.
2260 +# You should have received a copy of the GNU General Public License
2261 +# along with this program. If not, see <http://www.gnu.org/licenses/>.
2263 +Dynamically customizable FS operations
2264 +----------------------------------------------------------------------
2265 +Generally FS operations (struct inode_operations, struct
2266 +address_space_operations, struct file_operations, etc.) are defined as
2267 +"static const", but it never means that FS have only one set of
2268 +operation. Some FS have multiple sets of them. For instance, ext2 has
2269 +three sets, one for XIP, for NOBH, and for normal.
2270 +Since aufs overrides and redirects these operations, sometimes aufs has
2271 +to change its behaviour according to the branch FS type. More importantly
2272 +VFS acts differently if a function (member in the struct) is set or
2273 +not. It means aufs should have several sets of operations and select one
2274 +among them according to the branch FS definition.
2276 +In order to solve this problem and not to affect the behaviour of VFS,
2277 +aufs defines these operations dynamically. For instance, aufs defines
2278 +dummy direct_IO function for struct address_space_operations, but it may
2279 +not be set to the address_space_operations actually. When the branch FS
2280 +doesn't have it, aufs doesn't set it to its address_space_operations
2281 +while the function definition itself is still alive. So the behaviour
2282 +itself will not change, and it will return an error when direct_IO is
2285 +The lifetime of these dynamically generated operation object is
2286 +maintained by aufs branch object. When the branch is removed from aufs,
2287 +the reference counter of the object is decremented. When it reaches
2288 +zero, the dynamically generated operation object will be freed.
2290 +This approach is designed to support AIO (io_submit), Direct I/O and
2292 +Currently this approach is applied to address_space_operations for
2293 +regular files only.
2294 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2295 --- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
2296 +++ linux/Documentation/filesystems/aufs/README 2015-09-24 10:47:58.244719488 +0200
2299 +Aufs4 -- advanced multi layered unification filesystem version 4.x
2305 +----------------------------------------
2306 +In the early days, aufs was entirely re-designed and re-implemented
2307 +Unionfs Version 1.x series. Adding many original ideas, approaches,
2308 +improvements and implementations, it becomes totally different from
2309 +Unionfs while keeping the basic features.
2310 +Recently, Unionfs Version 2.x series begin taking some of the same
2311 +approaches to aufs1's.
2312 +Unionfs is being developed by Professor Erez Zadok at Stony Brook
2313 +University and his team.
2315 +Aufs4 supports linux-4.0 and later, and for linux-3.x series try aufs3.
2316 +If you want older kernel version support, try aufs2-2.6.git or
2317 +aufs2-standalone.git repository, aufs1 from CVS on SourceForge.
2319 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
2320 + According to Christoph Hellwig, linux rejects all union-type
2321 + filesystems but UnionMount.
2322 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2324 +PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2325 + UnionMount, and he pointed out an issue around a directory mutex
2326 + lock and aufs addressed it. But it is still unsure whether aufs will
2327 + be merged (or any other union solution).
2328 +<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
2332 +----------------------------------------
2333 +- unite several directories into a single virtual filesystem. The member
2334 + directory is called as a branch.
2335 +- you can specify the permission flags to the branch, which are 'readonly',
2336 + 'readwrite' and 'whiteout-able.'
2337 +- by upper writable branch, internal copyup and whiteout, files/dirs on
2338 + readonly branch are modifiable logically.
2339 +- dynamic branch manipulation, add, del.
2342 +Also there are many enhancements in aufs, such as:
2343 +- test only the highest one for the directory permission (dirperm1)
2344 +- copyup on open (coo=)
2345 +- 'move' policy for copy-up between two writable branches, after
2346 + checking free space.
2348 +- readdir(3) in userspace.
2349 +- keep inode number by external inode number table
2350 +- keep the timestamps of file/dir in internal copyup operation
2351 +- seekable directory, supporting NFS readdir.
2352 +- whiteout is hardlinked in order to reduce the consumption of inodes
2354 +- do not copyup, nor create a whiteout when it is unnecessary
2355 +- revert a single systemcall when an error occurs in aufs
2356 +- remount interface instead of ioctl
2357 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2358 +- loopback mounted filesystem as a branch
2359 +- kernel thread for removing the dir who has a plenty of whiteouts
2360 +- support copyup sparse file (a file which has a 'hole' in it)
2361 +- default permission flags for branches
2362 +- selectable permission flags for ro branch, whether whiteout can
2365 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
2366 +- support multiple writable branches, some policies to select one
2367 + among multiple writable branches.
2368 +- a new semantics for link(2) and rename(2) to support multiple
2369 + writable branches.
2370 +- no glibc changes are required.
2371 +- pseudo hardlink (hardlink over branches)
2372 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2373 + including NFS or remote filesystem branch.
2374 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2377 +Currently these features are dropped temporary from aufs4.
2378 +See design/08plan.txt in detail.
2379 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2381 +- statistics of aufs thread (/sys/fs/aufs/stat)
2383 +Features or just an idea in the future (see also design/*.txt),
2384 +- reorder the branch index without del/re-add.
2385 +- permanent xino files for NFSD
2386 +- an option for refreshing the opened files after add/del branches
2387 +- light version, without branch manipulation. (unnecessary?)
2388 +- copyup in userspace
2389 +- inotify in userspace
2394 +----------------------------------------
2395 +There are three GIT trees for aufs4, aufs4-linux.git,
2396 +aufs4-standalone.git, and aufs-util.git. Note that there is no "4" in
2398 +While the aufs-util is always necessary, you need either of aufs4-linux
2399 +or aufs4-standalone.
2401 +The aufs4-linux tree includes the whole linux mainline GIT tree,
2402 +git://git.kernel.org/.../torvalds/linux.git.
2403 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
2404 +build aufs4 as an external kernel module.
2405 +Several extra patches are not included in this tree. Only
2406 +aufs4-standalone tree contains them. They are describe in the later
2407 +section "Configuration and Compilation."
2409 +On the other hand, the aufs4-standalone tree has only aufs source files
2410 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
2411 +But you need to apply all aufs patches manually.
2413 +You will find GIT branches whose name is in form of "aufs4.x" where "x"
2414 +represents the linux kernel version, "linux-4.x". For instance,
2415 +"aufs4.0" is for linux-4.0. For latest "linux-4.x-rcN", use
2416 +"aufs4.x-rcN" branch.
2419 +$ git clone --reference /your/linux/git/tree \
2420 + git://github.com/sfjro/aufs4-linux.git aufs4-linux.git
2421 +- if you don't have linux GIT tree, then remove "--reference ..."
2422 +$ cd aufs4-linux.git
2423 +$ git checkout origin/aufs4.0
2425 +Or You may want to directly git-pull aufs into your linux GIT tree, and
2426 +leave the patch-work to GIT.
2427 +$ cd /your/linux/git/tree
2428 +$ git remote add aufs4 git://github.com/sfjro/aufs4-linux.git
2430 +$ git checkout -b my4.0 v4.0
2431 +$ (add your local change...)
2432 +$ git pull aufs4 aufs4.0
2433 +- now you have v4.0 + your_changes + aufs4.0 in you my4.0 branch.
2434 +- you may need to solve some conflicts between your_changes and
2435 + aufs4.0. in this case, git-rerere is recommended so that you can
2436 + solve the similar conflicts automatically when you upgrade to 4.1 or
2437 + later in the future.
2439 +o aufs4-standalone tree
2440 +$ git clone git://github.com/sfjro/aufs4-standalone.git aufs4-standalone.git
2441 +$ cd aufs4-standalone.git
2442 +$ git checkout origin/aufs4.0
2445 +$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2446 +- note that the public aufs-util.git is on SourceForge instead of
2449 +$ git checkout origin/aufs4.0
2451 +Note: The 4.x-rcN branch is to be used with `rc' kernel versions ONLY.
2452 +The minor version number, 'x' in '4.x', of aufs may not always
2453 +follow the minor version number of the kernel.
2454 +Because changes in the kernel that cause the use of a new
2455 +minor version number do not always require changes to aufs-util.
2457 +Since aufs-util has its own minor version number, you may not be
2458 +able to find a GIT branch in aufs-util for your kernel's
2459 +exact minor version number.
2460 +In this case, you should git-checkout the branch for the
2461 +nearest lower number.
2463 +For (an unreleased) example:
2464 +If you are using "linux-4.10" and the "aufs4.10" branch
2465 +does not exist in aufs-util repository, then "aufs4.9", "aufs4.8"
2466 +or something numerically smaller is the branch for your kernel.
2468 +Also you can view all branches by
2472 +3. Configuration and Compilation
2473 +----------------------------------------
2474 +Make sure you have git-checkout'ed the correct branch.
2476 +For aufs4-linux tree,
2477 +- enable CONFIG_AUFS_FS.
2478 +- set other aufs configurations if necessary.
2480 +For aufs4-standalone tree,
2481 +There are several ways to build.
2484 +- apply ./aufs4-kbuild.patch to your kernel source files.
2485 +- apply ./aufs4-base.patch too.
2486 +- apply ./aufs4-mmap.patch too.
2487 +- apply ./aufs4-standalone.patch too, if you have a plan to set
2488 + CONFIG_AUFS_FS=m. otherwise you don't need ./aufs4-standalone.patch.
2489 +- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2490 + kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
2491 +- enable CONFIG_AUFS_FS, you can select either
2493 +- and build your kernel as usual.
2494 +- install the built kernel.
2495 + Note: Since linux-3.9, every filesystem module requires an alias
2496 + "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2497 + modules.aliases file if you set CONFIG_AUFS_FS=m.
2498 +- install the header files too by "make headers_install" to the
2499 + directory where you specify. By default, it is $PWD/usr.
2500 + "make help" shows a brief note for headers_install.
2501 +- and reboot your system.
2504 +- module only (CONFIG_AUFS_FS=m).
2505 +- apply ./aufs4-base.patch to your kernel source files.
2506 +- apply ./aufs4-mmap.patch too.
2507 +- apply ./aufs4-standalone.patch too.
2508 +- build your kernel, don't forget "make headers_install", and reboot.
2509 +- edit ./config.mk and set other aufs configurations if necessary.
2510 + Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
2511 + every aufs configurations.
2512 +- build the module by simple "make".
2513 + Note: Since linux-3.9, every filesystem module requires an alias
2514 + "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2515 + modules.aliases file.
2516 +- you can specify ${KDIR} make variable which points to your kernel
2518 +- install the files
2519 + + run "make install" to install the aufs module, or copy the built
2520 + $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2521 + + run "make install_headers" (instead of headers_install) to install
2522 + the modified aufs header file (you can specify DESTDIR which is
2523 + available in aufs standalone version's Makefile only), or copy
2524 + $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2525 + you like manually. By default, the target directory is $PWD/usr.
2526 +- no need to apply aufs4-kbuild.patch, nor copying source files to your
2527 + kernel source tree.
2529 +Note: The header file aufs_type.h is necessary to build aufs-util
2530 + as well as "make headers_install" in the kernel source tree.
2531 + headers_install is subject to be forgotten, but it is essentially
2532 + necessary, not only for building aufs-util.
2533 + You may not meet problems without headers_install in some older
2537 +- read README in aufs-util, build and install it
2538 +- note that your distribution may contain an obsoleted version of
2539 + aufs_type.h in /usr/include/linux or something. When you build aufs
2540 + utilities, make sure that your compiler refers the correct aufs header
2541 + file which is built by "make headers_install."
2542 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2543 + then run "make install_ulib" too. And refer to the aufs manual in
2546 +There several other patches in aufs4-standalone.git. They are all
2547 +optional. When you meet some problems, they will help you.
2548 +- aufs4-loopback.patch
2549 + Supports a nested loopback mount in a branch-fs. This patch is
2550 + unnecessary until aufs produces a message like "you may want to try
2551 + another patch for loopback file".
2553 + Modifies a system global kernel internal function get_next_ino() in
2554 + order to stop assigning 0 for an inode-number. Not directly related to
2555 + aufs, but recommended generally.
2557 + Keeps the tmpfs inode number as the lowest value. Effective to reduce
2558 + the size of aufs XINO files for tmpfs branch. Also it prevents the
2559 + duplication of inode number, which is important for backup tools and
2560 + other utilities. When you find aufs XINO files for tmpfs branch
2561 + growing too much, try this patch.
2565 +----------------------------------------
2566 +At first, make sure aufs-util are installed, and please read the aufs
2567 +manual, aufs.5 in aufs-util.git tree.
2571 +$ mkdir /tmp/rw /tmp/aufs
2572 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2574 +Here is another example. The result is equivalent.
2575 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2577 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2578 +# mount -o remount,append:${HOME} /tmp/aufs
2580 +Then, you can see whole tree of your home dir through /tmp/aufs. If
2581 +you modify a file under /tmp/aufs, the one on your home directory is
2582 +not affected, instead the same named file will be newly created under
2583 +/tmp/rw. And all of your modification to a file will be applied to
2584 +the one under /tmp/rw. This is called the file based Copy on Write
2586 +Aufs mount options are described in aufs.5.
2587 +If you run chroot or something and make your aufs as a root directory,
2588 +then you need to customize the shutdown script. See the aufs manual in
2591 +Additionally, there are some sample usages of aufs which are a
2592 +diskless system with network booting, and LiveCD over NFS.
2593 +See sample dir in CVS tree on SourceForge.
2597 +----------------------------------------
2598 +When you have any problems or strange behaviour in aufs, please let me
2600 +- /proc/mounts (instead of the output of mount(8))
2601 +- /sys/module/aufs/*
2602 +- /sys/fs/aufs/* (if you have them)
2603 +- /debug/aufs/* (if you have them)
2604 +- linux kernel version
2605 + if your kernel is not plain, for example modified by distributor,
2606 + the url where i can download its source is necessary too.
2607 +- aufs version which was printed at loading the module or booting the
2608 + system, instead of the date you downloaded.
2609 +- configuration (define/undefine CONFIG_AUFS_xxx)
2610 +- kernel configuration or /proc/config.gz (if you have it)
2611 +- behaviour which you think to be incorrect
2612 +- actual operation, reproducible one is better
2613 +- mailto: aufs-users at lists.sourceforge.net
2615 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2616 +and Feature Requests) on SourceForge. Please join and write to
2620 +6. Acknowledgements
2621 +----------------------------------------
2622 +Thanks to everyone who have tried and are using aufs, whoever
2623 +have reported a bug or any feedback.
2625 +Especially donators:
2626 +Tomas Matejicek(slax.org) made a donation (much more than once).
2627 + Since Apr 2010, Tomas M (the author of Slax and Linux Live
2628 + scripts) is making "doubling" donations.
2629 + Unfortunately I cannot list all of the donators, but I really
2631 + It ends Aug 2010, but the ordinary donation URL is still available.
2632 + <http://sourceforge.net/donate/index.php?group_id=167503>
2633 +Dai Itasaka made a donation (2007/8).
2634 +Chuck Smith made a donation (2008/4, 10 and 12).
2635 +Henk Schoneveld made a donation (2008/9).
2636 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2637 +Francois Dupoux made a donation (2008/11).
2638 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2639 + aufs2 GIT tree (2009/2).
2640 +William Grant made a donation (2009/3).
2641 +Patrick Lane made a donation (2009/4).
2642 +The Mail Archive (mail-archive.com) made donations (2009/5).
2643 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2644 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2645 +Pavel Pronskiy made a donation (2011/2).
2646 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2647 + Networks (Ed Wildgoose) made a donation for hardware (2011/3).
2648 +Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2650 +Sam Liddicott made a donation (2011/9).
2651 +Era Scarecrow made a donation (2013/4).
2652 +Bor Ratajc made a donation (2013/4).
2653 +Alessandro Gorreta made a donation (2013/4).
2654 +POIRETTE Marc made a donation (2013/4).
2655 +Alessandro Gorreta made a donation (2013/4).
2656 +lauri kasvandik made a donation (2013/5).
2657 +"pemasu from Finland" made a donation (2013/7).
2658 +The Parted Magic Project made a donation (2013/9 and 11).
2659 +Pavel Barta made a donation (2013/10).
2660 +Nikolay Pertsev made a donation (2014/5).
2661 +James B made a donation (2014/7 and 2015/7).
2662 +Stefano Di Biase made a donation (2014/8).
2663 +Daniel Epellei made a donation (2015/1).
2665 +Thank you very much.
2666 +Donations are always, including future donations, very important and
2667 +helpful for me to keep on developing aufs.
2671 +----------------------------------------
2672 +If you are an experienced user, no explanation is needed. Aufs is
2673 +just a linux filesystem.
2678 +# Local variables: ;
2681 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2682 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
2683 +++ linux/fs/aufs/aufs.h 2015-09-24 10:47:58.248052907 +0200
2686 + * Copyright (C) 2005-2015 Junjiro R. Okajima
2688 + * This program, aufs is free software; you can redistribute it and/or modify
2689 + * it under the terms of the GNU General Public License as published by
2690 + * the Free Software Foundation; either version 2 of the License, or
2691 + * (at your option) any later version.
2693 + * This program is distributed in the hope that it will be useful,
2694 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2695 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2696 + * GNU General Public License for more details.
2698 + * You should have received a copy of the GNU General Public License
2699 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2703 + * all header files
2711 +#define AuStub(type, name, body, ...) \
2712 + static inline type name(__VA_ARGS__) { body; }
2714 +#define AuStubVoid(name, ...) \
2715 + AuStub(void, name, , __VA_ARGS__)
2716 +#define AuStubInt0(name, ...) \
2717 + AuStub(int, name, return 0, __VA_ARGS__)
2721 +#include "branch.h"
2724 +#include "dbgaufs.h"
2725 +#include "dentry.h"
2729 +#include "fstype.h"
2732 +#include "module.h"
2737 +#include "sysaufs.h"
2742 +#endif /* __KERNEL__ */
2743 +#endif /* __AUFS_H__ */
2744 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2745 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
2746 +++ linux/fs/aufs/branch.c 2015-11-11 17:21:46.915530388 +0100
2749 + * Copyright (C) 2005-2015 Junjiro R. Okajima
2751 + * This program, aufs is free software; you can redistribute it and/or modify
2752 + * it under the terms of the GNU General Public License as published by
2753 + * the Free Software Foundation; either version 2 of the License, or
2754 + * (at your option) any later version.
2756 + * This program is distributed in the hope that it will be useful,
2757 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2758 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2759 + * GNU General Public License for more details.
2761 + * You should have received a copy of the GNU General Public License
2762 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
2766 + * branch management
2769 +#include <linux/compat.h>
2770 +#include <linux/statfs.h>
2774 + * free a single branch
2776 +static void au_br_do_free(struct au_branch *br)
2779 + struct au_wbr *wbr;
2780 + struct au_dykey **key;
2782 + au_hnotify_fin_br(br);
2784 + if (br->br_xino.xi_file)
2785 + fput(br->br_xino.xi_file);
2786 + mutex_destroy(&br->br_xino.xi_nondir_mtx);
2788 + AuDebugOn(atomic_read(&br->br_count));
2792 + for (i = 0; i < AuBrWh_Last; i++)
2793 + dput(wbr->wbr_wh[i]);
2794 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
2795 + AuRwDestroy(&wbr->wbr_wh_rwsem);
2798 + if (br->br_fhsm) {
2799 + au_br_fhsm_fin(br->br_fhsm);
2800 + kfree(br->br_fhsm);
2803 + key = br->br_dykey;
2804 + for (i = 0; i < AuBrDynOp; i++, key++)
2810 + /* recursive lock, s_umount of branch's */
2812 + path_put(&br->br_path);
2819 + * frees all branches
2821 +void au_br_free(struct au_sbinfo *sbinfo)
2823 + aufs_bindex_t bmax;
2824 + struct au_branch **br;
2826 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2828 + bmax = sbinfo->si_bend + 1;
2829 + br = sbinfo->si_branch;
2831 + au_br_do_free(*br++);
2835 + * find the index of a branch which is specified by @br_id.
2837 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2839 + aufs_bindex_t bindex, bend;
2841 + bend = au_sbend(sb);
2842 + for (bindex = 0; bindex <= bend; bindex++)
2843 + if (au_sbr_id(sb, bindex) == br_id)
2848 +/* ---------------------------------------------------------------------- */
2854 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
2855 + struct dentry *h_root)
2857 + if (unlikely(h_adding == h_root
2858 + || au_test_loopback_overlap(sb, h_adding)))
2860 + if (h_adding->d_sb != h_root->d_sb)
2862 + return au_test_subdir(h_adding, h_root)
2863 + || au_test_subdir(h_root, h_adding);
2867 + * returns a newly allocated branch. @new_nbranch is a number of branches
2868 + * after adding a branch.
2870 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
2873 + struct au_branch *add_branch;
2874 + struct dentry *root;
2875 + struct inode *inode;
2879 + root = sb->s_root;
2880 + add_branch = kmalloc(sizeof(*add_branch), GFP_NOFS);
2881 + if (unlikely(!add_branch))
2884 + err = au_hnotify_init_br(add_branch, perm);
2885 + if (unlikely(err))
2888 + add_branch->br_wbr = NULL;
2889 + if (au_br_writable(perm)) {
2890 + /* may be freed separately at changing the branch permission */
2891 + add_branch->br_wbr = kmalloc(sizeof(*add_branch->br_wbr),
2893 + if (unlikely(!add_branch->br_wbr))
2897 + add_branch->br_fhsm = NULL;
2898 + if (au_br_fhsm(perm)) {
2899 + err = au_fhsm_br_alloc(add_branch);
2900 + if (unlikely(err))
2904 + err = au_sbr_realloc(au_sbi(sb), new_nbranch);
2906 + err = au_di_realloc(au_di(root), new_nbranch);
2908 + inode = d_inode(root);
2909 + err = au_ii_realloc(au_ii(inode), new_nbranch);
2912 + return add_branch; /* success */
2915 + kfree(add_branch->br_wbr);
2917 + au_hnotify_fin_br(add_branch);
2919 + kfree(add_branch);
2921 + return ERR_PTR(err);
2925 + * test if the branch permission is legal or not.
2927 +static int test_br(struct inode *inode, int brperm, char *path)
2931 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
2936 + pr_err("write permission for readonly mount or inode, %s\n", path);
2944 + * 0: success, the caller will add it
2945 + * plus: success, it is already unified, the caller should ignore it
2948 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
2951 + aufs_bindex_t bend, bindex;
2952 + struct dentry *root, *h_dentry;
2953 + struct inode *inode, *h_inode;
2955 + root = sb->s_root;
2956 + bend = au_sbend(sb);
2957 + if (unlikely(bend >= 0
2958 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
2962 + pr_err("%s duplicated\n", add->pathname);
2967 + err = -ENOSPC; /* -E2BIG; */
2968 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
2969 + || AUFS_BRANCH_MAX - 1 <= bend)) {
2970 + pr_err("number of branches exceeded %s\n", add->pathname);
2975 + if (unlikely(add->bindex < 0 || bend + 1 < add->bindex)) {
2976 + pr_err("bad index %d\n", add->bindex);
2980 + inode = d_inode(add->path.dentry);
2982 + if (unlikely(!inode->i_nlink)) {
2983 + pr_err("no existence %s\n", add->pathname);
2988 + if (unlikely(inode->i_sb == sb)) {
2989 + pr_err("%s must be outside\n", add->pathname);
2993 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
2994 + pr_err("unsupported filesystem, %s (%s)\n",
2995 + add->pathname, au_sbtype(inode->i_sb));
2999 + if (unlikely(inode->i_sb->s_stack_depth)) {
3000 + pr_err("already stacked, %s (%s)\n",
3001 + add->pathname, au_sbtype(inode->i_sb));
3005 + err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
3006 + if (unlikely(err))
3010 + return 0; /* success */
3013 + for (bindex = 0; bindex <= bend; bindex++)
3014 + if (unlikely(test_overlap(sb, add->path.dentry,
3015 + au_h_dptr(root, bindex)))) {
3016 + pr_err("%s is overlapped\n", add->pathname);
3021 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
3022 + h_dentry = au_h_dptr(root, 0);
3023 + h_inode = d_inode(h_dentry);
3024 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
3025 + || !uid_eq(h_inode->i_uid, inode->i_uid)
3026 + || !gid_eq(h_inode->i_gid, inode->i_gid))
3027 + pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3029 + i_uid_read(inode), i_gid_read(inode),
3030 + (inode->i_mode & S_IALLUGO),
3031 + i_uid_read(h_inode), i_gid_read(h_inode),
3032 + (h_inode->i_mode & S_IALLUGO));
3040 + * initialize or clean the whiteouts for an adding branch
3042 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
3045 + int err, old_perm;
3046 + aufs_bindex_t bindex;
3047 + struct mutex *h_mtx;
3048 + struct au_wbr *wbr;
3049 + struct au_hinode *hdir;
3050 + struct dentry *h_dentry;
3052 + err = vfsub_mnt_want_write(au_br_mnt(br));
3053 + if (unlikely(err))
3057 + old_perm = br->br_perm;
3058 + br->br_perm = new_perm;
3061 + bindex = au_br_index(sb, br->br_id);
3062 + if (0 <= bindex) {
3063 + hdir = au_hi(d_inode(sb->s_root), bindex);
3064 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
3066 + h_dentry = au_br_dentry(br);
3067 + h_mtx = &d_inode(h_dentry)->i_mutex;
3068 + mutex_lock_nested(h_mtx, AuLsc_I_PARENT);
3071 + err = au_wh_init(br, sb);
3073 + wbr_wh_write_lock(wbr);
3074 + err = au_wh_init(br, sb);
3075 + wbr_wh_write_unlock(wbr);
3078 + au_hn_imtx_unlock(hdir);
3080 + mutex_unlock(h_mtx);
3081 + vfsub_mnt_drop_write(au_br_mnt(br));
3082 + br->br_perm = old_perm;
3084 + if (!err && wbr && !au_br_writable(new_perm)) {
3086 + br->br_wbr = NULL;
3093 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
3097 + struct kstatfs kst;
3098 + struct au_wbr *wbr;
3101 + au_rw_init(&wbr->wbr_wh_rwsem);
3102 + memset(wbr->wbr_wh, 0, sizeof(wbr->wbr_wh));
3103 + atomic_set(&wbr->wbr_wh_running, 0);
3104 + wbr->wbr_bytes = 0;
3107 + * a limit for rmdir/rename a dir
3108 + * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
3110 + err = vfs_statfs(&br->br_path, &kst);
3111 + if (unlikely(err))
3114 + if (kst.f_namelen >= NAME_MAX)
3115 + err = au_br_init_wh(sb, br, perm);
3117 + pr_err("%pd(%s), unsupported namelen %ld\n",
3119 + au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
3125 +/* initialize a new branch */
3126 +static int au_br_init(struct au_branch *br, struct super_block *sb,
3127 + struct au_opt_add *add)
3130 + struct inode *h_inode;
3133 + memset(&br->br_xino, 0, sizeof(br->br_xino));
3134 + mutex_init(&br->br_xino.xi_nondir_mtx);
3135 + br->br_perm = add->perm;
3136 + br->br_path = add->path; /* set first, path_get() later */
3137 + spin_lock_init(&br->br_dykey_lock);
3138 + memset(br->br_dykey, 0, sizeof(br->br_dykey));
3139 + atomic_set(&br->br_count, 0);
3140 + atomic_set(&br->br_xino_running, 0);
3141 + br->br_id = au_new_br_id(sb);
3142 + AuDebugOn(br->br_id < 0);
3144 + if (au_br_writable(add->perm)) {
3145 + err = au_wbr_init(br, sb, add->perm);
3146 + if (unlikely(err))
3150 + if (au_opt_test(au_mntflags(sb), XINO)) {
3151 + h_inode = d_inode(add->path.dentry);
3152 + err = au_xino_br(sb, br, h_inode->i_ino,
3153 + au_sbr(sb, 0)->br_xino.xi_file, /*do_test*/1);
3154 + if (unlikely(err)) {
3155 + AuDebugOn(br->br_xino.xi_file);
3160 + sysaufs_br_init(br);
3161 + path_get(&br->br_path);
3162 + goto out; /* success */
3165 + memset(&br->br_path, 0, sizeof(br->br_path));
3170 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3171 + struct au_branch *br, aufs_bindex_t bend,
3172 + aufs_bindex_t amount)
3174 + struct au_branch **brp;
3176 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3178 + brp = sbinfo->si_branch + bindex;
3179 + memmove(brp + 1, brp, sizeof(*brp) * amount);
3181 + sbinfo->si_bend++;
3182 + if (unlikely(bend < 0))
3183 + sbinfo->si_bend = 0;
3186 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3187 + aufs_bindex_t bend, aufs_bindex_t amount)
3189 + struct au_hdentry *hdp;
3191 + AuRwMustWriteLock(&dinfo->di_rwsem);
3193 + hdp = dinfo->di_hdentry + bindex;
3194 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3195 + au_h_dentry_init(hdp);
3197 + if (unlikely(bend < 0))
3198 + dinfo->di_bstart = 0;
3201 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3202 + aufs_bindex_t bend, aufs_bindex_t amount)
3204 + struct au_hinode *hip;
3206 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3208 + hip = iinfo->ii_hinode + bindex;
3209 + memmove(hip + 1, hip, sizeof(*hip) * amount);
3210 + hip->hi_inode = NULL;
3213 + if (unlikely(bend < 0))
3214 + iinfo->ii_bstart = 0;
3217 +static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3218 + aufs_bindex_t bindex)
3220 + struct dentry *root, *h_dentry;
3221 + struct inode *root_inode, *h_inode;
3222 + aufs_bindex_t bend, amount;
3224 + root = sb->s_root;
3225 + root_inode = d_inode(root);
3226 + bend = au_sbend(sb);
3227 + amount = bend + 1 - bindex;
3228 + h_dentry = au_br_dentry(br);
3229 + au_sbilist_lock();
3230 + au_br_do_add_brp(au_sbi(sb), bindex, br, bend, amount);
3231 + au_br_do_add_hdp(au_di(root), bindex, bend, amount);
3232 + au_br_do_add_hip(au_ii(root_inode), bindex, bend, amount);
3233 + au_set_h_dptr(root, bindex, dget(h_dentry));
3234 + h_inode = d_inode(h_dentry);
3235 + au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
3236 + au_sbilist_unlock();
3239 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3242 + aufs_bindex_t bend, add_bindex;
3243 + struct dentry *root, *h_dentry;
3244 + struct inode *root_inode;
3245 + struct au_branch *add_branch;
3247 + root = sb->s_root;
3248 + root_inode = d_inode(root);
3249 + IMustLock(root_inode);
3250 + err = test_add(sb, add, remount);
3251 + if (unlikely(err < 0))
3255 + goto out; /* success */
3258 + bend = au_sbend(sb);
3259 + add_branch = au_br_alloc(sb, bend + 2, add->perm);
3260 + err = PTR_ERR(add_branch);
3261 + if (IS_ERR(add_branch))
3264 + err = au_br_init(add_branch, sb, add);
3265 + if (unlikely(err)) {
3266 + au_br_do_free(add_branch);
3270 + add_bindex = add->bindex;
3272 + au_br_do_add(sb, add_branch, add_bindex);
3274 + sysaufs_brs_del(sb, add_bindex);
3275 + au_br_do_add(sb, add_branch, add_bindex);
3276 + sysaufs_brs_add(sb, add_bindex);
3279 + h_dentry = add->path.dentry;
3280 + if (!add_bindex) {
3281 + au_cpup_attr_all(root_inode, /*force*/1);
3282 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3284 + au_add_nlink(root_inode, d_inode(h_dentry));
3287 + * this test/set prevents aufs from handling unnecesary notify events
3288 + * of xino files, in case of re-adding a writable branch which was
3289 + * once detached from aufs.
3291 + if (au_xino_brid(sb) < 0
3292 + && au_br_writable(add_branch->br_perm)
3293 + && !au_test_fs_bad_xino(h_dentry->d_sb)
3294 + && add_branch->br_xino.xi_file
3295 + && add_branch->br_xino.xi_file->f_path.dentry->d_parent == h_dentry)
3296 + au_xino_brid_set(sb, add_branch->br_id);
3302 +/* ---------------------------------------------------------------------- */
3304 +static unsigned long long au_farray_cb(struct super_block *sb, void *a,
3305 + unsigned long long max __maybe_unused,
3308 + unsigned long long n;
3309 + struct file **p, *f;
3310 + struct au_sphlhead *files;
3311 + struct au_finfo *finfo;
3315 + files = &au_sbi(sb)->si_files;
3316 + spin_lock(&files->spin);
3317 + hlist_for_each_entry(finfo, &files->head, fi_hlist) {
3318 + f = finfo->fi_file;
3320 + && !special_file(file_inode(f)->i_mode)) {
3324 + AuDebugOn(n > max);
3327 + spin_unlock(&files->spin);
3332 +static struct file **au_farray_alloc(struct super_block *sb,
3333 + unsigned long long *max)
3335 + *max = atomic_long_read(&au_sbi(sb)->si_nfiles);
3336 + return au_array_alloc(max, au_farray_cb, sb, /*arg*/NULL);
3339 +static void au_farray_free(struct file **a, unsigned long long max)
3341 + unsigned long long ull;
3343 + for (ull = 0; ull < max; ull++)
3349 +/* ---------------------------------------------------------------------- */
3355 +/* to show the line number, do not make it inlined function */
3356 +#define AuVerbose(do_info, fmt, ...) do { \
3358 + pr_info(fmt, ##__VA_ARGS__); \
3361 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t bstart,
3362 + aufs_bindex_t bend)
3364 + return (inode && !S_ISDIR(inode->i_mode)) || bstart == bend;
3367 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t bstart,
3368 + aufs_bindex_t bend)
3370 + return au_test_ibusy(d_inode(dentry), bstart, bend);
3374 + * test if the branch is deletable or not.
3376 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
3377 + unsigned int sigen, const unsigned int verbose)
3379 + int err, i, j, ndentry;
3380 + aufs_bindex_t bstart, bend;
3381 + struct au_dcsub_pages dpages;
3382 + struct au_dpage *dpage;
3385 + err = au_dpages_init(&dpages, GFP_NOFS);
3386 + if (unlikely(err))
3388 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
3389 + if (unlikely(err))
3392 + for (i = 0; !err && i < dpages.ndpage; i++) {
3393 + dpage = dpages.dpages + i;
3394 + ndentry = dpage->ndentry;
3395 + for (j = 0; !err && j < ndentry; j++) {
3396 + d = dpage->dentries[j];
3397 + AuDebugOn(au_dcount(d) <= 0);
3398 + if (!au_digen_test(d, sigen)) {
3399 + di_read_lock_child(d, AuLock_IR);
3400 + if (unlikely(au_dbrange_test(d))) {
3401 + di_read_unlock(d, AuLock_IR);
3405 + di_write_lock_child(d);
3406 + if (unlikely(au_dbrange_test(d))) {
3407 + di_write_unlock(d);
3410 + err = au_reval_dpath(d, sigen);
3412 + di_downgrade_lock(d, AuLock_IR);
3414 + di_write_unlock(d);
3419 + /* AuDbgDentry(d); */
3420 + bstart = au_dbstart(d);
3421 + bend = au_dbend(d);
3422 + if (bstart <= bindex
3424 + && au_h_dptr(d, bindex)
3425 + && au_test_dbusy(d, bstart, bend)) {
3427 + AuVerbose(verbose, "busy %pd\n", d);
3430 + di_read_unlock(d, AuLock_IR);
3435 + au_dpages_free(&dpages);
3440 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
3441 + unsigned int sigen, const unsigned int verbose)
3444 + unsigned long long max, ull;
3445 + struct inode *i, **array;
3446 + aufs_bindex_t bstart, bend;
3448 + array = au_iarray_alloc(sb, &max);
3449 + err = PTR_ERR(array);
3450 + if (IS_ERR(array))
3454 + AuDbg("b%d\n", bindex);
3455 + for (ull = 0; !err && ull < max; ull++) {
3459 + if (i->i_ino == AUFS_ROOT_INO)
3462 + /* AuDbgInode(i); */
3463 + if (au_iigen(i, NULL) == sigen)
3464 + ii_read_lock_child(i);
3466 + ii_write_lock_child(i);
3467 + err = au_refresh_hinode_self(i);
3470 + ii_downgrade_lock(i);
3472 + ii_write_unlock(i);
3477 + bstart = au_ibstart(i);
3478 + bend = au_ibend(i);
3479 + if (bstart <= bindex
3481 + && au_h_iptr(i, bindex)
3482 + && au_test_ibusy(i, bstart, bend)) {
3484 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
3487 + ii_read_unlock(i);
3489 + au_iarray_free(array, max);
3495 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3496 + const unsigned int verbose)
3499 + unsigned int sigen;
3501 + sigen = au_sigen(root->d_sb);
3502 + DiMustNoWaiters(root);
3503 + IiMustNoWaiters(d_inode(root));
3504 + di_write_unlock(root);
3505 + err = test_dentry_busy(root, bindex, sigen, verbose);
3507 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
3508 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3513 +static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3514 + struct file **to_free, int *idx)
3517 + unsigned char matched, root;
3518 + aufs_bindex_t bindex, bend;
3519 + struct au_fidir *fidir;
3520 + struct au_hfile *hfile;
3523 + root = IS_ROOT(file->f_path.dentry);
3526 + to_free[*idx] = file;
3532 + fidir = au_fi(file)->fi_hdir;
3533 + AuDebugOn(!fidir);
3534 + bend = au_fbend_dir(file);
3535 + for (bindex = au_fbstart(file); bindex <= bend; bindex++) {
3536 + hfile = fidir->fd_hfile + bindex;
3537 + if (!hfile->hf_file)
3540 + if (hfile->hf_br->br_id == br_id) {
3552 +static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3553 + struct file **to_free, int opened)
3556 + unsigned long long ull, max;
3557 + aufs_bindex_t bstart;
3558 + struct file *file, **array;
3559 + struct dentry *root;
3560 + struct au_hfile *hfile;
3562 + array = au_farray_alloc(sb, &max);
3563 + err = PTR_ERR(array);
3564 + if (IS_ERR(array))
3569 + root = sb->s_root;
3570 + di_write_unlock(root);
3571 + for (ull = 0; ull < max; ull++) {
3572 + file = array[ull];
3573 + if (unlikely(!file))
3576 + /* AuDbg("%pD\n", file); */
3577 + fi_read_lock(file);
3578 + bstart = au_fbstart(file);
3579 + if (!d_is_dir(file->f_path.dentry)) {
3580 + hfile = &au_fi(file)->fi_htop;
3581 + if (hfile->hf_br->br_id == br_id)
3584 + err = test_dir_busy(file, br_id, to_free, &idx);
3585 + fi_read_unlock(file);
3586 + if (unlikely(err))
3589 + di_write_lock_child(root);
3590 + au_farray_free(array, max);
3591 + AuDebugOn(idx > opened);
3597 +static void br_del_file(struct file **to_free, unsigned long long opened,
3598 + aufs_bindex_t br_id)
3600 + unsigned long long ull;
3601 + aufs_bindex_t bindex, bstart, bend, bfound;
3602 + struct file *file;
3603 + struct au_fidir *fidir;
3604 + struct au_hfile *hfile;
3606 + for (ull = 0; ull < opened; ull++) {
3607 + file = to_free[ull];
3608 + if (unlikely(!file))
3611 + /* AuDbg("%pD\n", file); */
3612 + AuDebugOn(!d_is_dir(file->f_path.dentry));
3614 + fidir = au_fi(file)->fi_hdir;
3615 + AuDebugOn(!fidir);
3616 + fi_write_lock(file);
3617 + bstart = au_fbstart(file);
3618 + bend = au_fbend_dir(file);
3619 + for (bindex = bstart; bindex <= bend; bindex++) {
3620 + hfile = fidir->fd_hfile + bindex;
3621 + if (!hfile->hf_file)
3624 + if (hfile->hf_br->br_id == br_id) {
3629 + AuDebugOn(bfound < 0);
3630 + au_set_h_fptr(file, bfound, NULL);
3631 + if (bfound == bstart) {
3632 + for (bstart++; bstart <= bend; bstart++)
3633 + if (au_hf_dir(file, bstart)) {
3634 + au_set_fbstart(file, bstart);
3638 + fi_write_unlock(file);
3642 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3643 + const aufs_bindex_t bindex,
3644 + const aufs_bindex_t bend)
3646 + struct au_branch **brp, **p;
3648 + AuRwMustWriteLock(&sbinfo->si_rwsem);
3650 + brp = sbinfo->si_branch + bindex;
3651 + if (bindex < bend)
3652 + memmove(brp, brp + 1, sizeof(*brp) * (bend - bindex));
3653 + sbinfo->si_branch[0 + bend] = NULL;
3654 + sbinfo->si_bend--;
3656 + p = krealloc(sbinfo->si_branch, sizeof(*p) * bend, AuGFP_SBILIST);
3658 + sbinfo->si_branch = p;
3659 + /* harmless error */
3662 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3663 + const aufs_bindex_t bend)
3665 + struct au_hdentry *hdp, *p;
3667 + AuRwMustWriteLock(&dinfo->di_rwsem);
3669 + hdp = dinfo->di_hdentry;
3670 + if (bindex < bend)
3671 + memmove(hdp + bindex, hdp + bindex + 1,
3672 + sizeof(*hdp) * (bend - bindex));
3673 + hdp[0 + bend].hd_dentry = NULL;
3676 + p = krealloc(hdp, sizeof(*p) * bend, AuGFP_SBILIST);
3678 + dinfo->di_hdentry = p;
3679 + /* harmless error */
3682 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3683 + const aufs_bindex_t bend)
3685 + struct au_hinode *hip, *p;
3687 + AuRwMustWriteLock(&iinfo->ii_rwsem);
3689 + hip = iinfo->ii_hinode + bindex;
3690 + if (bindex < bend)
3691 + memmove(hip, hip + 1, sizeof(*hip) * (bend - bindex));
3692 + iinfo->ii_hinode[0 + bend].hi_inode = NULL;
3693 + au_hn_init(iinfo->ii_hinode + bend);
3696 + p = krealloc(iinfo->ii_hinode, sizeof(*p) * bend, AuGFP_SBILIST);
3698 + iinfo->ii_hinode = p;
3699 + /* harmless error */
3702 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3703 + struct au_branch *br)
3705 + aufs_bindex_t bend;
3706 + struct au_sbinfo *sbinfo;
3707 + struct dentry *root, *h_root;
3708 + struct inode *inode, *h_inode;
3709 + struct au_hinode *hinode;
3711 + SiMustWriteLock(sb);
3713 + root = sb->s_root;
3714 + inode = d_inode(root);
3715 + sbinfo = au_sbi(sb);
3716 + bend = sbinfo->si_bend;
3718 + h_root = au_h_dptr(root, bindex);
3719 + hinode = au_hi(inode, bindex);
3720 + h_inode = au_igrab(hinode->hi_inode);
3723 + au_sbilist_lock();
3724 + au_br_do_del_brp(sbinfo, bindex, bend);
3725 + au_br_do_del_hdp(au_di(root), bindex, bend);
3726 + au_br_do_del_hip(au_ii(inode), bindex, bend);
3727 + au_sbilist_unlock();
3731 + au_br_do_free(br);
3734 +static unsigned long long empty_cb(struct super_block *sb, void *array,
3735 + unsigned long long max, void *arg)
3740 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3743 + unsigned long long opened;
3744 + unsigned int mnt_flags;
3745 + aufs_bindex_t bindex, bend, br_id;
3746 + unsigned char do_wh, verbose;
3747 + struct au_branch *br;
3748 + struct au_wbr *wbr;
3749 + struct dentry *root;
3750 + struct file **to_free;
3755 + root = sb->s_root;
3756 + bindex = au_find_dbindex(root, del->h_path.dentry);
3759 + goto out; /* success */
3761 + pr_err("%s no such branch\n", del->pathname);
3764 + AuDbg("bindex b%d\n", bindex);
3767 + mnt_flags = au_mntflags(sb);
3768 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3769 + bend = au_sbend(sb);
3770 + if (unlikely(!bend)) {
3771 + AuVerbose(verbose, "no more branches left\n");
3774 + br = au_sbr(sb, bindex);
3775 + AuDebugOn(!path_equal(&br->br_path, &del->h_path));
3777 + br_id = br->br_id;
3778 + opened = atomic_read(&br->br_count);
3779 + if (unlikely(opened)) {
3780 + to_free = au_array_alloc(&opened, empty_cb, sb, NULL);
3781 + err = PTR_ERR(to_free);
3782 + if (IS_ERR(to_free))
3785 + err = test_file_busy(sb, br_id, to_free, opened);
3786 + if (unlikely(err)) {
3787 + AuVerbose(verbose, "%llu file(s) opened\n", opened);
3793 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3795 + /* instead of WbrWhMustWriteLock(wbr) */
3796 + SiMustWriteLock(sb);
3797 + for (i = 0; i < AuBrWh_Last; i++) {
3798 + dput(wbr->wbr_wh[i]);
3799 + wbr->wbr_wh[i] = NULL;
3803 + err = test_children_busy(root, bindex, verbose);
3804 + if (unlikely(err)) {
3813 + * now we confirmed the branch is deletable.
3814 + * let's free the remaining opened dirs on the branch.
3816 + di_write_unlock(root);
3817 + br_del_file(to_free, opened, br_id);
3818 + di_write_lock_child(root);
3822 + au_br_do_del(sb, bindex, br);
3824 + sysaufs_brs_del(sb, bindex);
3825 + au_br_do_del(sb, bindex, br);
3826 + sysaufs_brs_add(sb, bindex);
3830 + au_cpup_attr_all(d_inode(root), /*force*/1);
3831 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3833 + au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
3834 + if (au_opt_test(mnt_flags, PLINK))
3835 + au_plink_half_refresh(sb, br_id);
3837 + if (au_xino_brid(sb) == br_id)
3838 + au_xino_brid_set(sb, -1);
3839 + goto out; /* success */
3843 + rerr = au_br_init_wh(sb, br, br->br_perm);
3845 + pr_warn("failed re-creating base whiteout, %s. (%d)\n",
3846 + del->pathname, rerr);
3849 + au_farray_free(to_free, opened);
3853 +/* ---------------------------------------------------------------------- */
3855 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
3858 + aufs_bindex_t bstart, bend;
3859 + struct aufs_ibusy ibusy;
3860 + struct inode *inode, *h_inode;
3863 + if (unlikely(!capable(CAP_SYS_ADMIN)))
3866 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
3868 + err = !access_ok(VERIFY_WRITE, &arg->h_ino, sizeof(arg->h_ino));
3869 + if (unlikely(err)) {
3876 + si_read_lock(sb, AuLock_FLUSH);
3877 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbend(sb)))
3881 + ibusy.h_ino = 0; /* invalid */
3882 + inode = ilookup(sb, ibusy.ino);
3884 + || inode->i_ino == AUFS_ROOT_INO
3885 + || is_bad_inode(inode))
3888 + ii_read_lock_child(inode);
3889 + bstart = au_ibstart(inode);
3890 + bend = au_ibend(inode);
3891 + if (bstart <= ibusy.bindex && ibusy.bindex <= bend) {
3892 + h_inode = au_h_iptr(inode, ibusy.bindex);
3893 + if (h_inode && au_test_ibusy(inode, bstart, bend))
3894 + ibusy.h_ino = h_inode->i_ino;
3896 + ii_read_unlock(inode);
3900 + si_read_unlock(sb);
3902 + err = __put_user(ibusy.h_ino, &arg->h_ino);
3903 + if (unlikely(err)) {
3912 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
3914 + return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
3917 +#ifdef CONFIG_COMPAT
3918 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
3920 + return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
3924 +/* ---------------------------------------------------------------------- */
3927 + * change a branch permission
3930 +static void au_warn_ima(void)
3933 + /* since it doesn't support mark_files_ro() */
3934 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
3938 +static int do_need_sigen_inc(int a, int b)
3940 + return au_br_whable(a) && !au_br_whable(b);
3943 +static int need_sigen_inc(int old, int new)
3945 + return do_need_sigen_inc(old, new)
3946 + || do_need_sigen_inc(new, old);
3949 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
3952 + unsigned int mnt_flags;
3953 + unsigned long long ull, max;
3954 + aufs_bindex_t br_id;
3955 + unsigned char verbose, writer;
3956 + struct file *file, *hf, **array;
3957 + struct au_hfile *hfile;
3959 + mnt_flags = au_mntflags(sb);
3960 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
3962 + array = au_farray_alloc(sb, &max);
3963 + err = PTR_ERR(array);
3964 + if (IS_ERR(array))
3968 + br_id = au_sbr_id(sb, bindex);
3969 + for (ull = 0; ull < max; ull++) {
3970 + file = array[ull];
3971 + if (unlikely(!file))
3974 + /* AuDbg("%pD\n", file); */
3975 + fi_read_lock(file);
3976 + if (unlikely(au_test_mmapped(file))) {
3978 + AuVerbose(verbose, "mmapped %pD\n", file);
3980 + FiMustNoWaiters(file);
3981 + fi_read_unlock(file);
3985 + hfile = &au_fi(file)->fi_htop;
3986 + hf = hfile->hf_file;
3987 + if (!d_is_reg(file->f_path.dentry)
3988 + || !(file->f_mode & FMODE_WRITE)
3989 + || hfile->hf_br->br_id != br_id
3990 + || !(hf->f_mode & FMODE_WRITE))
3991 + array[ull] = NULL;
3997 + FiMustNoWaiters(file);
3998 + fi_read_unlock(file);
4006 + for (ull = 0; ull < max; ull++) {
4007 + file = array[ull];
4011 + /* todo: already flushed? */
4013 + * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4014 + * approach which resets f_mode and calls mnt_drop_write() and
4015 + * file_release_write() for each file, because the branch
4016 + * attribute in aufs world is totally different from the native
4019 + /* fi_read_lock(file); */
4020 + hfile = &au_fi(file)->fi_htop;
4021 + hf = hfile->hf_file;
4022 + /* fi_read_unlock(file); */
4023 + spin_lock(&hf->f_lock);
4024 + writer = !!(hf->f_mode & FMODE_WRITER);
4025 + hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
4026 + spin_unlock(&hf->f_lock);
4028 + put_write_access(file_inode(hf));
4029 + __mnt_drop_write(hf->f_path.mnt);
4034 + au_farray_free(array, max);
4040 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4044 + aufs_bindex_t bindex;
4045 + struct dentry *root;
4046 + struct au_branch *br;
4047 + struct au_br_fhsm *bf;
4049 + root = sb->s_root;
4050 + bindex = au_find_dbindex(root, mod->h_root);
4053 + return 0; /* success */
4055 + pr_err("%s no such branch\n", mod->path);
4058 + AuDbg("bindex b%d\n", bindex);
4060 + err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
4061 + if (unlikely(err))
4064 + br = au_sbr(sb, bindex);
4065 + AuDebugOn(mod->h_root != au_br_dentry(br));
4066 + if (br->br_perm == mod->perm)
4067 + return 0; /* success */
4069 + /* pre-allocate for non-fhsm --> fhsm */
4071 + if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4072 + err = au_fhsm_br_alloc(br);
4073 + if (unlikely(err))
4076 + br->br_fhsm = NULL;
4079 + if (au_br_writable(br->br_perm)) {
4080 + /* remove whiteout base */
4081 + err = au_br_init_wh(sb, br, mod->perm);
4082 + if (unlikely(err))
4085 + if (!au_br_writable(mod->perm)) {
4086 + /* rw --> ro, file might be mmapped */
4087 + DiMustNoWaiters(root);
4088 + IiMustNoWaiters(d_inode(root));
4089 + di_write_unlock(root);
4090 + err = au_br_mod_files_ro(sb, bindex);
4091 + /* aufs_write_lock() calls ..._child() */
4092 + di_write_lock_child(root);
4094 + if (unlikely(err)) {
4096 + br->br_wbr = kmalloc(sizeof(*br->br_wbr),
4099 + rerr = au_wbr_init(br, sb, br->br_perm);
4100 + if (unlikely(rerr)) {
4101 + AuIOErr("nested error %d (%d)\n",
4103 + br->br_perm = mod->perm;
4107 + } else if (au_br_writable(mod->perm)) {
4110 + br->br_wbr = kmalloc(sizeof(*br->br_wbr), GFP_NOFS);
4112 + err = au_wbr_init(br, sb, mod->perm);
4113 + if (unlikely(err)) {
4114 + kfree(br->br_wbr);
4115 + br->br_wbr = NULL;
4119 + if (unlikely(err))
4122 + if (au_br_fhsm(br->br_perm)) {
4123 + if (!au_br_fhsm(mod->perm)) {
4124 + /* fhsm --> non-fhsm */
4125 + au_br_fhsm_fin(br->br_fhsm);
4126 + kfree(br->br_fhsm);
4127 + br->br_fhsm = NULL;
4129 + } else if (au_br_fhsm(mod->perm))
4130 + /* non-fhsm --> fhsm */
4133 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4134 + br->br_perm = mod->perm;
4135 + goto out; /* success */
4144 +/* ---------------------------------------------------------------------- */
4146 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4149 + struct kstatfs kstfs;
4151 + err = vfs_statfs(&br->br_path, &kstfs);
4153 + stfs->f_blocks = kstfs.f_blocks;
4154 + stfs->f_bavail = kstfs.f_bavail;
4155 + stfs->f_files = kstfs.f_files;
4156 + stfs->f_ffree = kstfs.f_ffree;
4161 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4162 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
4163 +++ linux/fs/aufs/branch.h 2015-11-11 17:21:46.915530388 +0100
4166 + * Copyright (C) 2005-2015 Junjiro R. Okajima
4168 + * This program, aufs is free software; you can redistribute it and/or modify
4169 + * it under the terms of the GNU General Public License as published by
4170 + * the Free Software Foundation; either version 2 of the License, or
4171 + * (at your option) any later version.
4173 + * This program is distributed in the hope that it will be useful,
4174 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4175 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4176 + * GNU General Public License for more details.
4178 + * You should have received a copy of the GNU General Public License
4179 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4183 + * branch filesystems and xino for them
4186 +#ifndef __AUFS_BRANCH_H__
4187 +#define __AUFS_BRANCH_H__
4191 +#include <linux/mount.h>
4196 +/* ---------------------------------------------------------------------- */
4199 +struct au_xino_file {
4200 + struct file *xi_file;
4201 + struct mutex xi_nondir_mtx;
4203 + /* todo: make xino files an array to support huge inode number */
4205 +#ifdef CONFIG_DEBUG_FS
4206 + struct dentry *xi_dbgaufs;
4210 +/* File-based Hierarchical Storage Management */
4211 +struct au_br_fhsm {
4212 +#ifdef CONFIG_AUFS_FHSM
4213 + struct mutex bf_lock;
4214 + unsigned long bf_jiffy;
4215 + struct aufs_stfs bf_stfs;
4220 +/* members for writable branch only */
4221 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4223 + struct au_rwsem wbr_wh_rwsem;
4224 + struct dentry *wbr_wh[AuBrWh_Last];
4225 + atomic_t wbr_wh_running;
4226 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4227 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4228 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4231 + unsigned long long wbr_bytes;
4234 +/* ext2 has 3 types of operations at least, ext3 has 4 */
4235 +#define AuBrDynOp (AuDyLast * 4)
4237 +#ifdef CONFIG_AUFS_HFSNOTIFY
4238 +/* support for asynchronous destruction */
4239 +struct au_br_hfsnotify {
4240 + struct fsnotify_group *hfsn_group;
4244 +/* sysfs entries */
4245 +struct au_brsysfs {
4247 + struct attribute attr;
4256 +/* protected by superblock rwsem */
4258 + struct au_xino_file br_xino;
4260 + aufs_bindex_t br_id;
4263 + struct path br_path;
4264 + spinlock_t br_dykey_lock;
4265 + struct au_dykey *br_dykey[AuBrDynOp];
4266 + atomic_t br_count;
4268 + struct au_wbr *br_wbr;
4269 + struct au_br_fhsm *br_fhsm;
4271 + /* xino truncation */
4272 + atomic_t br_xino_running;
4274 +#ifdef CONFIG_AUFS_HFSNOTIFY
4275 + struct au_br_hfsnotify *br_hfsn;
4278 +#ifdef CONFIG_SYSFS
4279 + /* entries under sysfs per mount-point */
4280 + struct au_brsysfs br_sysfs[AuBrSysfs_Last];
4284 +/* ---------------------------------------------------------------------- */
4286 +static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4288 + return br->br_path.mnt;
4291 +static inline struct dentry *au_br_dentry(struct au_branch *br)
4293 + return br->br_path.dentry;
4296 +static inline struct super_block *au_br_sb(struct au_branch *br)
4298 + return au_br_mnt(br)->mnt_sb;
4301 +static inline int au_br_rdonly(struct au_branch *br)
4303 + return ((au_br_sb(br)->s_flags & MS_RDONLY)
4304 + || !au_br_writable(br->br_perm))
4308 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
4310 +#ifdef CONFIG_AUFS_HNOTIFY
4311 + return !(brperm & AuBrPerm_RR);
4317 +static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4319 + int err, exec_flag;
4322 + exec_flag = oflag & __FMODE_EXEC;
4323 + if (unlikely(exec_flag && path_noexec(&br->br_path)))
4329 +/* ---------------------------------------------------------------------- */
4333 +void au_br_free(struct au_sbinfo *sinfo);
4334 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4336 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4338 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
4339 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
4340 +#ifdef CONFIG_COMPAT
4341 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4344 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
4347 +int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
4350 +static const loff_t au_loff_max = LLONG_MAX;
4352 +int au_xib_trunc(struct super_block *sb);
4353 +ssize_t xino_fread(vfs_readf_t func, struct file *file, void *buf, size_t size,
4355 +ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
4356 + size_t size, loff_t *pos);
4357 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src);
4358 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent);
4359 +ino_t au_xino_new_ino(struct super_block *sb);
4360 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
4361 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4363 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4365 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4366 + struct file *base_file, int do_test);
4367 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex);
4369 +struct au_opt_xino;
4370 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount);
4371 +void au_xino_clr(struct super_block *sb);
4372 +struct file *au_xino_def(struct super_block *sb);
4373 +int au_xino_path(struct seq_file *seq, struct file *file);
4375 +/* ---------------------------------------------------------------------- */
4377 +/* Superblock to branch */
4379 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4381 + return au_sbr(sb, bindex)->br_id;
4385 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4387 + return au_br_mnt(au_sbr(sb, bindex));
4391 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4393 + return au_br_sb(au_sbr(sb, bindex));
4396 +static inline void au_sbr_put(struct super_block *sb, aufs_bindex_t bindex)
4398 + atomic_dec(&au_sbr(sb, bindex)->br_count);
4401 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4403 + return au_sbr(sb, bindex)->br_perm;
4406 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4408 + return au_br_whable(au_sbr_perm(sb, bindex));
4411 +/* ---------------------------------------------------------------------- */
4414 + * wbr_wh_read_lock, wbr_wh_write_lock
4415 + * wbr_wh_read_unlock, wbr_wh_write_unlock, wbr_wh_downgrade_lock
4417 +AuSimpleRwsemFuncs(wbr_wh, struct au_wbr *wbr, &wbr->wbr_wh_rwsem);
4419 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&wbr->wbr_wh_rwsem)
4420 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&wbr->wbr_wh_rwsem)
4421 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&wbr->wbr_wh_rwsem)
4423 +/* ---------------------------------------------------------------------- */
4425 +#ifdef CONFIG_AUFS_FHSM
4426 +static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4428 + mutex_init(&brfhsm->bf_lock);
4429 + brfhsm->bf_jiffy = 0;
4430 + brfhsm->bf_readable = 0;
4433 +static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4435 + mutex_destroy(&brfhsm->bf_lock);
4438 +AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4439 +AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4442 +#endif /* __KERNEL__ */
4443 +#endif /* __AUFS_BRANCH_H__ */
4444 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4445 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
4446 +++ linux/fs/aufs/conf.mk 2015-09-24 10:47:58.248052907 +0200
4449 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4453 +AuConfStr += ${1}=${${1}}
4457 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
4459 + HNOTIFY HFSNOTIFY \
4470 +$(foreach i, ${AuConfAll}, \
4471 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
4473 +AuConfName = ${obj}/conf.str
4474 +${AuConfName}.tmp: FORCE
4475 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4476 +${AuConfName}: ${AuConfName}.tmp
4477 + @diff -q $< $@ > /dev/null 2>&1 || { \
4478 + echo ' GEN ' $@; \
4482 +clean-files += ${AuConfName} ${AuConfName}.tmp
4483 +${obj}/sysfs.o: ${AuConfName}
4485 +-include ${srctree}/${src}/conf_priv.mk
4486 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4487 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
4488 +++ linux/fs/aufs/cpup.c 2015-11-11 17:21:46.915530388 +0100
4491 + * Copyright (C) 2005-2015 Junjiro R. Okajima
4493 + * This program, aufs is free software; you can redistribute it and/or modify
4494 + * it under the terms of the GNU General Public License as published by
4495 + * the Free Software Foundation; either version 2 of the License, or
4496 + * (at your option) any later version.
4498 + * This program is distributed in the hope that it will be useful,
4499 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4500 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4501 + * GNU General Public License for more details.
4503 + * You should have received a copy of the GNU General Public License
4504 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
4508 + * copy-up functions, see wbr_policy.c for copy-down
4511 +#include <linux/fs_stack.h>
4512 +#include <linux/mm.h>
4515 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
4517 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
4518 + | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
4520 + BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4522 + dst->i_flags |= iflags & ~mask;
4523 + if (au_test_fs_notime(dst->i_sb))
4524 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
4527 +void au_cpup_attr_timesizes(struct inode *inode)
4529 + struct inode *h_inode;
4531 + h_inode = au_h_iptr(inode, au_ibstart(inode));
4532 + fsstack_copy_attr_times(inode, h_inode);
4533 + fsstack_copy_inode_size(inode, h_inode);
4536 +void au_cpup_attr_nlink(struct inode *inode, int force)
4538 + struct inode *h_inode;
4539 + struct super_block *sb;
4540 + aufs_bindex_t bindex, bend;
4543 + bindex = au_ibstart(inode);
4544 + h_inode = au_h_iptr(inode, bindex);
4546 + && !S_ISDIR(h_inode->i_mode)
4547 + && au_opt_test(au_mntflags(sb), PLINK)
4548 + && au_plink_test(inode))
4552 + * 0 can happen in revalidating.
4553 + * h_inode->i_mutex may not be held here, but it is harmless since once
4554 + * i_nlink reaches 0, it will never become positive except O_TMPFILE
4556 + * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4557 + * the incorrect link count.
4559 + set_nlink(inode, h_inode->i_nlink);
4562 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4563 + * it may includes whplink directory.
4565 + if (S_ISDIR(h_inode->i_mode)) {
4566 + bend = au_ibend(inode);
4567 + for (bindex++; bindex <= bend; bindex++) {
4568 + h_inode = au_h_iptr(inode, bindex);
4570 + au_add_nlink(inode, h_inode);
4575 +void au_cpup_attr_changeable(struct inode *inode)
4577 + struct inode *h_inode;
4579 + h_inode = au_h_iptr(inode, au_ibstart(inode));
4580 + inode->i_mode = h_inode->i_mode;
4581 + inode->i_uid = h_inode->i_uid;
4582 + inode->i_gid = h_inode->i_gid;
4583 + au_cpup_attr_timesizes(inode);
4584 + au_cpup_attr_flags(inode, h_inode->i_flags);
4587 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4589 + struct au_iinfo *iinfo = au_ii(inode);
4591 + IiMustWriteLock(inode);
4593 + iinfo->ii_higen = h_inode->i_generation;
4594 + iinfo->ii_hsb1 = h_inode->i_sb;
4597 +void au_cpup_attr_all(struct inode *inode, int force)
4599 + struct inode *h_inode;
4601 + h_inode = au_h_iptr(inode, au_ibstart(inode));
4602 + au_cpup_attr_changeable(inode);
4603 + if (inode->i_nlink > 0)
4604 + au_cpup_attr_nlink(inode, force);
4605 + inode->i_rdev = h_inode->i_rdev;
4606 + inode->i_blkbits = h_inode->i_blkbits;
4607 + au_cpup_igen(inode, h_inode);
4610 +/* ---------------------------------------------------------------------- */
4612 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4614 +/* keep the timestamps of the parent dir when cpup */
4615 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4616 + struct path *h_path)
4618 + struct inode *h_inode;
4620 + dt->dt_dentry = dentry;
4621 + dt->dt_h_path = *h_path;
4622 + h_inode = d_inode(h_path->dentry);
4623 + dt->dt_atime = h_inode->i_atime;
4624 + dt->dt_mtime = h_inode->i_mtime;
4628 +void au_dtime_revert(struct au_dtime *dt)
4630 + struct iattr attr;
4633 + attr.ia_atime = dt->dt_atime;
4634 + attr.ia_mtime = dt->dt_mtime;
4635 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4636 + | ATTR_ATIME | ATTR_ATIME_SET;
4638 + /* no delegation since this is a directory */
4639 + err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
4640 + if (unlikely(err))
4641 + pr_warn("restoring timestamps failed(%d). ignored\n", err);
4644 +/* ---------------------------------------------------------------------- */
4646 +/* internal use only */
4647 +struct au_cpup_reg_attr {
4650 + unsigned int iflags; /* inode->i_flags */
4653 +static noinline_for_stack
4654 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct dentry *h_src,
4655 + struct au_cpup_reg_attr *h_src_attr)
4657 + int err, sbits, icex;
4658 + unsigned int mnt_flags;
4659 + unsigned char verbose;
4661 + struct path h_path;
4662 + struct inode *h_isrc, *h_idst;
4663 + struct kstat *h_st;
4664 + struct au_branch *br;
4666 + h_path.dentry = au_h_dptr(dst, bindex);
4667 + h_idst = d_inode(h_path.dentry);
4668 + br = au_sbr(dst->d_sb, bindex);
4669 + h_path.mnt = au_br_mnt(br);
4670 + h_isrc = d_inode(h_src);
4671 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
4672 + | ATTR_ATIME | ATTR_MTIME
4673 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
4674 + if (h_src_attr && h_src_attr->valid) {
4675 + h_st = &h_src_attr->st;
4676 + ia.ia_uid = h_st->uid;
4677 + ia.ia_gid = h_st->gid;
4678 + ia.ia_atime = h_st->atime;
4679 + ia.ia_mtime = h_st->mtime;
4680 + if (h_idst->i_mode != h_st->mode
4681 + && !S_ISLNK(h_idst->i_mode)) {
4682 + ia.ia_valid |= ATTR_MODE;
4683 + ia.ia_mode = h_st->mode;
4685 + sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4686 + au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4688 + ia.ia_uid = h_isrc->i_uid;
4689 + ia.ia_gid = h_isrc->i_gid;
4690 + ia.ia_atime = h_isrc->i_atime;
4691 + ia.ia_mtime = h_isrc->i_mtime;
4692 + if (h_idst->i_mode != h_isrc->i_mode
4693 + && !S_ISLNK(h_idst->i_mode)) {
4694 + ia.ia_valid |= ATTR_MODE;
4695 + ia.ia_mode = h_isrc->i_mode;
4697 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4698 + au_cpup_attr_flags(h_idst, h_isrc->i_flags);
4700 + /* no delegation since it is just created */
4701 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4703 + /* is this nfs only? */
4704 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4705 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4706 + ia.ia_mode = h_isrc->i_mode;
4707 + err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
4710 + icex = br->br_perm & AuBrAttr_ICEX;
4712 + mnt_flags = au_mntflags(dst->d_sb);
4713 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
4714 + err = au_cpup_xattr(h_path.dentry, h_src, icex, verbose);
4720 +/* ---------------------------------------------------------------------- */
4722 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4723 + char *buf, unsigned long blksize)
4726 + size_t sz, rbytes, wbytes;
4727 + unsigned char all_zero;
4729 + struct mutex *h_mtx;
4730 + /* reduce stack usage */
4733 + zp = page_address(ZERO_PAGE(0));
4734 + if (unlikely(!zp))
4735 + return -ENOMEM; /* possible? */
4740 + AuDbg("len %lld\n", len);
4742 + if (len < blksize)
4746 + /* todo: signal_pending? */
4747 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
4748 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
4751 + if (unlikely(err < 0))
4755 + if (len >= rbytes && rbytes == blksize)
4756 + all_zero = !memcmp(buf, zp, rbytes);
4763 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
4765 + /* todo: signal_pending? */
4766 + if (unlikely(err == -EAGAIN || err == -EINTR))
4768 + if (unlikely(err < 0))
4773 + if (unlikely(err < 0))
4779 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
4781 + if (unlikely(res < 0))
4788 + /* the last block may be a hole */
4789 + if (!err && all_zero) {
4790 + AuLabel(last hole);
4793 + if (au_test_nfs(dst->f_path.dentry->d_sb)) {
4794 + /* nfs requires this step to make last hole */
4795 + /* is this only nfs? */
4797 + /* todo: signal_pending? */
4798 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
4799 + } while (err == -EAGAIN || err == -EINTR);
4806 + ia->ia_size = dst->f_pos;
4807 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
4808 + ia->ia_file = dst;
4809 + h_mtx = &file_inode(dst)->i_mutex;
4810 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
4811 + /* no delegation since it is just created */
4812 + err = vfsub_notify_change(&dst->f_path, ia,
4813 + /*delegated*/NULL);
4814 + mutex_unlock(h_mtx);
4821 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
4824 + unsigned long blksize;
4825 + unsigned char do_kfree;
4829 + blksize = dst->f_path.dentry->d_sb->s_blocksize;
4830 + if (!blksize || PAGE_SIZE < blksize)
4831 + blksize = PAGE_SIZE;
4832 + AuDbg("blksize %lu\n", blksize);
4833 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
4835 + buf = kmalloc(blksize, GFP_NOFS);
4837 + buf = (void *)__get_free_page(GFP_NOFS);
4838 + if (unlikely(!buf))
4841 + if (len > (1 << 22))
4842 + AuDbg("copying a large file %lld\n", (long long)len);
4846 + err = au_do_copy_file(dst, src, len, buf, blksize);
4850 + free_page((unsigned long)buf);
4857 + * to support a sparse file which is opened with O_APPEND,
4858 + * we need to close the file.
4860 +static int au_cp_regular(struct au_cp_generic *cpg)
4863 + enum { SRC, DST };
4865 + aufs_bindex_t bindex;
4866 + unsigned int flags;
4867 + struct dentry *dentry;
4869 + struct file *file;
4873 + .bindex = cpg->bsrc,
4874 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
4878 + .bindex = cpg->bdst,
4879 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
4880 + .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
4881 + .label = &&out_src
4884 + struct super_block *sb;
4886 + /* bsrc branch can be ro/rw. */
4887 + sb = cpg->dentry->d_sb;
4889 + for (i = 0; i < 2; i++, f++) {
4890 + f->dentry = au_h_dptr(cpg->dentry, f->bindex);
4891 + f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
4892 + /*file*/NULL, f->force_wr);
4893 + err = PTR_ERR(f->file);
4894 + if (IS_ERR(f->file))
4898 + /* try stopping to update while we copyup */
4899 + IMustLock(d_inode(file[SRC].dentry));
4900 + err = au_copy_file(file[DST].file, file[SRC].file, cpg->len);
4902 + fput(file[DST].file);
4903 + au_sbr_put(sb, file[DST].bindex);
4906 + fput(file[SRC].file);
4907 + au_sbr_put(sb, file[SRC].bindex);
4912 +static int au_do_cpup_regular(struct au_cp_generic *cpg,
4913 + struct au_cpup_reg_attr *h_src_attr)
4917 + struct path h_path;
4918 + struct inode *h_src_inode, *h_dst_inode;
4921 + h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
4922 + l = i_size_read(h_src_inode);
4923 + if (cpg->len == -1 || l < cpg->len)
4926 + /* try stopping to update while we are referencing */
4927 + mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
4928 + au_pin_hdir_unlock(cpg->pin);
4930 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
4931 + h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
4932 + h_src_attr->iflags = h_src_inode->i_flags;
4933 + if (!au_test_nfs(h_src_inode->i_sb))
4934 + err = vfs_getattr(&h_path, &h_src_attr->st);
4936 + mutex_unlock(&h_src_inode->i_mutex);
4937 + err = vfs_getattr(&h_path, &h_src_attr->st);
4938 + mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
4940 + if (unlikely(err)) {
4941 + mutex_unlock(&h_src_inode->i_mutex);
4944 + h_src_attr->valid = 1;
4945 + err = au_cp_regular(cpg);
4946 + mutex_unlock(&h_src_inode->i_mutex);
4947 + rerr = au_pin_hdir_relock(cpg->pin);
4951 + if (!err && (h_src_inode->i_state & I_LINKABLE)) {
4952 + h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
4953 + h_dst_inode = d_inode(h_path.dentry);
4954 + spin_lock(&h_dst_inode->i_lock);
4955 + h_dst_inode->i_state |= I_LINKABLE;
4956 + spin_unlock(&h_dst_inode->i_lock);
4963 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
4964 + struct inode *h_dir)
4967 + mm_segment_t old_fs;
4972 + struct inode *h_inode = d_inode(h_src);
4973 + const struct inode_operations *h_iop = h_inode->i_op;
4976 + if (unlikely(!h_iop->readlink))
4980 + sym.k = (void *)__get_free_page(GFP_NOFS);
4981 + if (unlikely(!sym.k))
4984 + /* unnecessary to support mmap_sem since symlink is not mmap-able */
4985 + old_fs = get_fs();
4986 + set_fs(KERNEL_DS);
4987 + symlen = h_iop->readlink(h_src, sym.u, PATH_MAX);
4992 + sym.k[symlen] = 0;
4993 + err = vfsub_symlink(h_dir, h_path, sym.k);
4995 + free_page((unsigned long)sym.k);
5001 +static noinline_for_stack
5002 +int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
5003 + struct au_cpup_reg_attr *h_src_attr)
5007 + unsigned int mnt_flags;
5008 + unsigned char isdir, isreg, force;
5009 + const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
5010 + struct au_dtime dt;
5011 + struct path h_path;
5012 + struct dentry *h_src, *h_dst, *h_parent;
5013 + struct inode *h_inode, *h_dir, *dir, *inode;
5014 + struct super_block *sb;
5016 + /* bsrc branch can be ro/rw. */
5017 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5018 + h_inode = d_inode(h_src);
5019 + AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
5021 + /* try stopping to be referenced while we are creating */
5022 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5023 + if (au_ftest_cpup(cpg->flags, RENAME))
5024 + AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5025 + AUFS_WH_PFX_LEN));
5026 + h_parent = h_dst->d_parent; /* dir inode is locked */
5027 + h_dir = d_inode(h_parent);
5029 + AuDebugOn(h_parent != h_dst->d_parent);
5031 + sb = cpg->dentry->d_sb;
5032 + h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
5034 + h_path.dentry = h_parent;
5035 + au_dtime_store(&dt, dst_parent, &h_path);
5037 + h_path.dentry = h_dst;
5041 + mode = h_inode->i_mode;
5042 + switch (mode & S_IFMT) {
5045 + err = vfsub_create(h_dir, &h_path, mode | S_IWUSR,
5046 + /*want_excl*/true);
5048 + err = au_do_cpup_regular(cpg, h_src_attr);
5052 + err = vfsub_mkdir(h_dir, &h_path, mode);
5055 + * strange behaviour from the users view,
5056 + * particularry setattr case
5058 + dir = d_inode(dst_parent);
5059 + if (au_ibstart(dir) == cpg->bdst)
5060 + au_cpup_attr_nlink(dir, /*force*/1);
5061 + inode = d_inode(cpg->dentry);
5062 + au_cpup_attr_nlink(inode, /*force*/1);
5066 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5070 + AuDebugOn(!capable(CAP_MKNOD));
5074 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5077 + AuIOErr("Unknown inode type 0%o\n", mode);
5081 + mnt_flags = au_mntflags(sb);
5082 + if (!au_opt_test(mnt_flags, UDBA_NONE)
5084 + && au_opt_test(mnt_flags, XINO)
5085 + && (h_inode->i_nlink == 1
5086 + || (h_inode->i_state & I_LINKABLE))
5087 + /* todo: unnecessary? */
5088 + /* && d_inode(cpg->dentry)->i_nlink == 1 */
5089 + && cpg->bdst < cpg->bsrc
5090 + && !au_ftest_cpup(cpg->flags, KEEPLINO))
5091 + au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
5092 + /* ignore this error */
5097 + force = !!cpg->len;
5098 + if (cpg->len == -1)
5099 + force = !!i_size_read(h_inode);
5101 + au_fhsm_wrote(sb, cpg->bdst, force);
5105 + au_dtime_revert(&dt);
5109 +static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
5112 + struct dentry *dentry, *h_dentry, *h_parent, *parent;
5113 + struct inode *h_dir;
5114 + aufs_bindex_t bdst;
5116 + dentry = cpg->dentry;
5118 + h_dentry = au_h_dptr(dentry, bdst);
5119 + if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5121 + au_set_h_dptr(dentry, bdst, NULL);
5122 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
5124 + h_path->dentry = dget(au_h_dptr(dentry, bdst));
5125 + au_set_h_dptr(dentry, bdst, h_dentry);
5128 + parent = dget_parent(dentry);
5129 + h_parent = au_h_dptr(parent, bdst);
5131 + h_path->dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
5132 + if (IS_ERR(h_path->dentry))
5133 + err = PTR_ERR(h_path->dentry);
5135 + if (unlikely(err))
5138 + h_parent = h_dentry->d_parent; /* dir inode is locked */
5139 + h_dir = d_inode(h_parent);
5141 + AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5142 + /* no delegation since it is just created */
5143 + err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL);
5144 + dput(h_path->dentry);
5151 + * copyup the @dentry from @bsrc to @bdst.
5152 + * the caller must set the both of lower dentries.
5153 + * @len is for truncating when it is -1 copyup the entire file.
5154 + * in link/rename cases, @dst_parent may be different from the real one.
5155 + * basic->bsrc can be larger than basic->bdst.
5157 +static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5160 + aufs_bindex_t old_ibstart;
5161 + unsigned char isdir, plink;
5162 + struct dentry *h_src, *h_dst, *h_parent;
5163 + struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
5164 + struct super_block *sb;
5165 + struct au_branch *br;
5166 + /* to reuduce stack size */
5168 + struct au_dtime dt;
5169 + struct path h_path;
5170 + struct au_cpup_reg_attr h_src_attr;
5174 + a = kmalloc(sizeof(*a), GFP_NOFS);
5177 + a->h_src_attr.valid = 0;
5179 + sb = cpg->dentry->d_sb;
5180 + br = au_sbr(sb, cpg->bdst);
5181 + a->h_path.mnt = au_br_mnt(br);
5182 + h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5183 + h_parent = h_dst->d_parent; /* dir inode is locked */
5184 + h_dir = d_inode(h_parent);
5187 + h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5188 + inode = d_inode(cpg->dentry);
5191 + dst_parent = dget_parent(cpg->dentry);
5195 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
5196 + dst_inode = au_h_iptr(inode, cpg->bdst);
5198 + if (unlikely(!plink)) {
5200 + AuIOErr("hi%lu(i%lu) exists on b%d "
5201 + "but plink is disabled\n",
5202 + dst_inode->i_ino, inode->i_ino, cpg->bdst);
5206 + if (dst_inode->i_nlink) {
5207 + const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
5209 + h_src = au_plink_lkup(inode, cpg->bdst);
5210 + err = PTR_ERR(h_src);
5211 + if (IS_ERR(h_src))
5213 + if (unlikely(d_is_negative(h_src))) {
5215 + AuIOErr("i%lu exists on b%d "
5216 + "but not pseudo-linked\n",
5217 + inode->i_ino, cpg->bdst);
5223 + a->h_path.dentry = h_parent;
5224 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5227 + a->h_path.dentry = h_dst;
5229 + err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
5230 + if (!err && au_ftest_cpup(cpg->flags, RENAME))
5231 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5233 + au_dtime_revert(&a->dt);
5234 + if (unlikely(err == -EWOULDBLOCK)) {
5235 + pr_warn("cannot retry for NFSv4 delegation"
5236 + " for an internal link\n");
5242 + /* todo: cpup_wh_file? */
5244 + au_update_ibrange(inode, /*do_put_zero*/1);
5247 + isdir = S_ISDIR(inode->i_mode);
5248 + old_ibstart = au_ibstart(inode);
5249 + err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
5250 + if (unlikely(err))
5252 + dst_inode = d_inode(h_dst);
5253 + mutex_lock_nested(&dst_inode->i_mutex, AuLsc_I_CHILD2);
5254 + /* todo: necessary? */
5255 + /* au_pin_hdir_unlock(cpg->pin); */
5257 + err = cpup_iattr(cpg->dentry, cpg->bdst, h_src, &a->h_src_attr);
5258 + if (unlikely(err)) {
5259 + /* todo: necessary? */
5260 + /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
5261 + mutex_unlock(&dst_inode->i_mutex);
5265 + if (cpg->bdst < old_ibstart) {
5266 + if (S_ISREG(inode->i_mode)) {
5267 + err = au_dy_iaop(inode, cpg->bdst, dst_inode);
5268 + if (unlikely(err)) {
5269 + /* ignore an error */
5270 + /* au_pin_hdir_relock(cpg->pin); */
5271 + mutex_unlock(&dst_inode->i_mutex);
5275 + au_set_ibstart(inode, cpg->bdst);
5277 + au_set_ibend(inode, cpg->bdst);
5278 + au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
5279 + au_hi_flags(inode, isdir));
5281 + /* todo: necessary? */
5282 + /* err = au_pin_hdir_relock(cpg->pin); */
5283 + mutex_unlock(&dst_inode->i_mutex);
5284 + if (unlikely(err))
5287 + src_inode = d_inode(h_src);
5289 + && (src_inode->i_nlink > 1
5290 + || src_inode->i_state & I_LINKABLE)
5292 + au_plink_append(inode, cpg->bdst, h_dst);
5294 + if (au_ftest_cpup(cpg->flags, RENAME)) {
5295 + a->h_path.dentry = h_dst;
5296 + err = au_do_ren_after_cpup(cpg, &a->h_path);
5299 + goto out_parent; /* success */
5303 + a->h_path.dentry = h_parent;
5304 + au_dtime_store(&a->dt, dst_parent, &a->h_path);
5305 + a->h_path.dentry = h_dst;
5307 + if (d_is_positive(h_dst)) {
5309 + /* no delegation since it is just created */
5310 + rerr = vfsub_unlink(h_dir, &a->h_path,
5311 + /*delegated*/NULL, /*force*/0);
5313 + rerr = vfsub_rmdir(h_dir, &a->h_path);
5315 + au_dtime_revert(&a->dt);
5317 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5327 +#if 0 /* reserved */
5328 +struct au_cpup_single_args {
5330 + struct au_cp_generic *cpg;
5331 + struct dentry *dst_parent;
5334 +static void au_call_cpup_single(void *args)
5336 + struct au_cpup_single_args *a = args;
5338 + au_pin_hdir_acquire_nest(a->cpg->pin);
5339 + *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5340 + au_pin_hdir_release(a->cpg->pin);
5345 + * prevent SIGXFSZ in copy-up.
5346 + * testing CAP_MKNOD is for generic fs,
5347 + * but CAP_FSETID is for xfs only, currently.
5349 +static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
5352 + struct super_block *sb;
5353 + struct inode *h_dir;
5356 + sb = au_pinned_parent(pin)->d_sb;
5357 + if (!au_wkq_test()
5358 + && (!au_sbi(sb)->si_plink_maint_pid
5359 + || au_plink_maint(sb, AuLock_NOPLM))) {
5360 + switch (mode & S_IFMT) {
5362 + /* no condition about RLIMIT_FSIZE and the file size */
5367 + do_sio = !capable(CAP_MKNOD);
5371 + do_sio = ((mode & (S_ISUID | S_ISGID))
5372 + && !capable(CAP_FSETID));
5373 + /* this workaround may be removed in the future */
5375 + h_dir = au_pinned_h_dir(pin);
5376 + do_sio = h_dir->i_mode & S_ISVTX;
5383 +#if 0 /* reserved */
5384 +int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
5387 + struct dentry *h_dentry;
5389 + h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5390 + if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
5391 + err = au_cpup_single(cpg, dst_parent);
5393 + struct au_cpup_single_args args = {
5396 + .dst_parent = dst_parent
5398 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5399 + if (unlikely(wkq_err))
5408 + * copyup the @dentry from the first active lower branch to @bdst,
5409 + * using au_cpup_single().
5411 +static int au_cpup_simple(struct au_cp_generic *cpg)
5414 + unsigned int flags_orig;
5415 + struct dentry *dentry;
5417 + AuDebugOn(cpg->bsrc < 0);
5419 + dentry = cpg->dentry;
5420 + DiMustWriteLock(dentry);
5422 + err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
5424 + flags_orig = cpg->flags;
5425 + au_fset_cpup(cpg->flags, RENAME);
5426 + err = au_cpup_single(cpg, NULL);
5427 + cpg->flags = flags_orig;
5429 + return 0; /* success */
5432 + au_set_h_dptr(dentry, cpg->bdst, NULL);
5433 + au_set_dbstart(dentry, cpg->bsrc);
5439 +struct au_cpup_simple_args {
5441 + struct au_cp_generic *cpg;
5444 +static void au_call_cpup_simple(void *args)
5446 + struct au_cpup_simple_args *a = args;
5448 + au_pin_hdir_acquire_nest(a->cpg->pin);
5449 + *a->errp = au_cpup_simple(a->cpg);
5450 + au_pin_hdir_release(a->cpg->pin);
5453 +static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
5456 + struct dentry *dentry, *parent;
5457 + struct file *h_file;
5458 + struct inode *h_dir;
5460 + dentry = cpg->dentry;
5462 + if (au_ftest_cpup(cpg->flags, HOPEN)) {
5463 + AuDebugOn(cpg->bsrc < 0);
5464 + h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
5465 + err = PTR_ERR(h_file);
5466 + if (IS_ERR(h_file))
5470 + parent = dget_parent(dentry);
5471 + h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
5472 + if (!au_test_h_perm_sio(h_dir, MAY_EXEC | MAY_WRITE)
5473 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5474 + err = au_cpup_simple(cpg);
5476 + struct au_cpup_simple_args args = {
5480 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5481 + if (unlikely(wkq_err))
5487 + au_h_open_post(dentry, cpg->bsrc, h_file);
5493 +int au_sio_cpup_simple(struct au_cp_generic *cpg)
5495 + aufs_bindex_t bsrc, bend;
5496 + struct dentry *dentry, *h_dentry;
5498 + if (cpg->bsrc < 0) {
5499 + dentry = cpg->dentry;
5500 + bend = au_dbend(dentry);
5501 + for (bsrc = cpg->bdst + 1; bsrc <= bend; bsrc++) {
5502 + h_dentry = au_h_dptr(dentry, bsrc);
5504 + AuDebugOn(d_is_negative(h_dentry));
5508 + AuDebugOn(bsrc > bend);
5511 + AuDebugOn(cpg->bsrc <= cpg->bdst);
5512 + return au_do_sio_cpup_simple(cpg);
5515 +int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5517 + AuDebugOn(cpg->bdst <= cpg->bsrc);
5518 + return au_do_sio_cpup_simple(cpg);
5521 +/* ---------------------------------------------------------------------- */
5524 + * copyup the deleted file for writing.
5526 +static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5527 + struct file *file)
5530 + unsigned int flags_orig;
5531 + aufs_bindex_t bsrc_orig;
5532 + struct dentry *h_d_dst, *h_d_start;
5533 + struct au_dinfo *dinfo;
5534 + struct au_hdentry *hdp;
5536 + dinfo = au_di(cpg->dentry);
5537 + AuRwMustWriteLock(&dinfo->di_rwsem);
5539 + bsrc_orig = cpg->bsrc;
5540 + cpg->bsrc = dinfo->di_bstart;
5541 + hdp = dinfo->di_hdentry;
5542 + h_d_dst = hdp[0 + cpg->bdst].hd_dentry;
5543 + dinfo->di_bstart = cpg->bdst;
5544 + hdp[0 + cpg->bdst].hd_dentry = wh_dentry;
5547 + h_d_start = hdp[0 + cpg->bsrc].hd_dentry;
5548 + hdp[0 + cpg->bsrc].hd_dentry = au_hf_top(file)->f_path.dentry;
5550 + flags_orig = cpg->flags;
5551 + cpg->flags = !AuCpup_DTIME;
5552 + err = au_cpup_single(cpg, /*h_parent*/NULL);
5553 + cpg->flags = flags_orig;
5556 + err = au_reopen_nondir(file);
5557 + hdp[0 + cpg->bsrc].hd_dentry = h_d_start;
5559 + hdp[0 + cpg->bdst].hd_dentry = h_d_dst;
5560 + dinfo->di_bstart = cpg->bsrc;
5561 + cpg->bsrc = bsrc_orig;
5566 +static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5569 + aufs_bindex_t bdst;
5570 + struct au_dtime dt;
5571 + struct dentry *dentry, *parent, *h_parent, *wh_dentry;
5572 + struct au_branch *br;
5573 + struct path h_path;
5575 + dentry = cpg->dentry;
5577 + br = au_sbr(dentry->d_sb, bdst);
5578 + parent = dget_parent(dentry);
5579 + h_parent = au_h_dptr(parent, bdst);
5580 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5581 + err = PTR_ERR(wh_dentry);
5582 + if (IS_ERR(wh_dentry))
5585 + h_path.dentry = h_parent;
5586 + h_path.mnt = au_br_mnt(br);
5587 + au_dtime_store(&dt, parent, &h_path);
5588 + err = au_do_cpup_wh(cpg, wh_dentry, file);
5589 + if (unlikely(err))
5593 + h_path.dentry = wh_dentry;
5594 + if (!d_is_dir(wh_dentry)) {
5595 + /* no delegation since it is just created */
5596 + err = vfsub_unlink(d_inode(h_parent), &h_path,
5597 + /*delegated*/NULL, /*force*/0);
5599 + err = vfsub_rmdir(d_inode(h_parent), &h_path);
5600 + if (unlikely(err)) {
5601 + AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
5605 + au_dtime_revert(&dt);
5606 + au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
5615 +struct au_cpup_wh_args {
5617 + struct au_cp_generic *cpg;
5618 + struct file *file;
5621 +static void au_call_cpup_wh(void *args)
5623 + struct au_cpup_wh_args *a = args;
5625 + au_pin_hdir_acquire_nest(a->cpg->pin);
5626 + *a->errp = au_cpup_wh(a->cpg, a->file);
5627 + au_pin_hdir_release(a->cpg->pin);
5630 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
5633 + aufs_bindex_t bdst;
5634 + struct dentry *dentry, *parent, *h_orph, *h_parent;
5635 + struct inode *dir, *h_dir, *h_tmpdir;
5636 + struct au_wbr *wbr;
5637 + struct au_pin wh_pin, *pin_orig;
5639 + dentry = cpg->dentry;
5641 + parent = dget_parent(dentry);
5642 + dir = d_inode(parent);
5645 + h_dir = au_igrab(au_h_iptr(dir, bdst));
5648 + if (!h_dir->i_nlink) {
5649 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
5650 + h_orph = wbr->wbr_orph;
5652 + h_parent = dget(au_h_dptr(parent, bdst));
5653 + au_set_h_dptr(parent, bdst, dget(h_orph));
5654 + h_tmpdir = d_inode(h_orph);
5655 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
5657 + mutex_lock_nested(&h_tmpdir->i_mutex, AuLsc_I_PARENT3);
5658 + /* todo: au_h_open_pre()? */
5660 + pin_orig = cpg->pin;
5661 + au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
5662 + AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
5663 + cpg->pin = &wh_pin;
5666 + if (!au_test_h_perm_sio(h_tmpdir, MAY_EXEC | MAY_WRITE)
5667 + && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
5668 + err = au_cpup_wh(cpg, file);
5670 + struct au_cpup_wh_args args = {
5675 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
5676 + if (unlikely(wkq_err))
5681 + mutex_unlock(&h_tmpdir->i_mutex);
5682 + /* todo: au_h_open_post()? */
5683 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
5684 + au_set_h_dptr(parent, bdst, h_parent);
5685 + AuDebugOn(!pin_orig);
5686 + cpg->pin = pin_orig;
5694 +/* ---------------------------------------------------------------------- */
5697 + * generic routine for both of copy-up and copy-down.
5699 +/* cf. revalidate function in file.c */
5700 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5701 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
5702 + struct au_pin *pin,
5703 + struct dentry *h_parent, void *arg),
5707 + struct au_pin pin;
5708 + struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
5711 + parent = dget_parent(dentry);
5712 + if (IS_ROOT(parent))
5715 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
5716 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
5718 + /* do not use au_dpage */
5719 + real_parent = parent;
5722 + parent = dget_parent(dentry);
5723 + h_parent = au_h_dptr(parent, bdst);
5725 + goto out; /* success */
5727 + /* find top dir which is necessary to cpup */
5731 + parent = dget_parent(d);
5732 + di_read_lock_parent3(parent, !AuLock_IR);
5733 + h_parent = au_h_dptr(parent, bdst);
5734 + di_read_unlock(parent, !AuLock_IR);
5735 + } while (!h_parent);
5737 + if (d != real_parent)
5738 + di_write_lock_child3(d);
5740 + /* somebody else might create while we were sleeping */
5741 + h_dentry = au_h_dptr(d, bdst);
5742 + if (!h_dentry || d_is_negative(h_dentry)) {
5744 + au_update_dbstart(d);
5746 + au_pin_set_dentry(&pin, d);
5747 + err = au_do_pin(&pin);
5749 + err = cp(d, bdst, &pin, h_parent, arg);
5754 + if (d != real_parent)
5755 + di_write_unlock(d);
5756 + if (unlikely(err))
5765 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
5766 + struct au_pin *pin,
5767 + struct dentry *h_parent __maybe_unused,
5768 + void *arg __maybe_unused)
5770 + struct au_cp_generic cpg = {
5776 + .flags = AuCpup_DTIME
5778 + return au_sio_cpup_simple(&cpg);
5781 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5783 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
5786 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5789 + struct dentry *parent;
5790 + struct inode *dir;
5792 + parent = dget_parent(dentry);
5793 + dir = d_inode(parent);
5795 + if (au_h_iptr(dir, bdst))
5798 + di_read_unlock(parent, AuLock_IR);
5799 + di_write_lock_parent(parent);
5800 + /* someone else might change our inode while we were sleeping */
5801 + if (!au_h_iptr(dir, bdst))
5802 + err = au_cpup_dirs(dentry, bdst);
5803 + di_downgrade_lock(parent, AuLock_IR);
5809 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
5810 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
5811 +++ linux/fs/aufs/cpup.h 2015-09-24 10:47:58.248052907 +0200
5814 + * Copyright (C) 2005-2015 Junjiro R. Okajima
5816 + * This program, aufs is free software; you can redistribute it and/or modify
5817 + * it under the terms of the GNU General Public License as published by
5818 + * the Free Software Foundation; either version 2 of the License, or
5819 + * (at your option) any later version.
5821 + * This program is distributed in the hope that it will be useful,
5822 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5823 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5824 + * GNU General Public License for more details.
5826 + * You should have received a copy of the GNU General Public License
5827 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
5831 + * copy-up/down functions
5834 +#ifndef __AUFS_CPUP_H__
5835 +#define __AUFS_CPUP_H__
5839 +#include <linux/path.h>
5845 +void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
5846 +void au_cpup_attr_timesizes(struct inode *inode);
5847 +void au_cpup_attr_nlink(struct inode *inode, int force);
5848 +void au_cpup_attr_changeable(struct inode *inode);
5849 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
5850 +void au_cpup_attr_all(struct inode *inode, int force);
5852 +/* ---------------------------------------------------------------------- */
5854 +struct au_cp_generic {
5855 + struct dentry *dentry;
5856 + aufs_bindex_t bdst, bsrc;
5858 + struct au_pin *pin;
5859 + unsigned int flags;
5863 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
5864 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
5866 +#define AuCpup_RENAME (1 << 2) /* rename after cpup */
5867 +#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
5869 +#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
5871 +#define AuCpup_RWDST (1 << 5) /* force write target even if
5872 + the branch is marked as RO */
5874 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
5875 +#define au_fset_cpup(flags, name) \
5876 + do { (flags) |= AuCpup_##name; } while (0)
5877 +#define au_fclr_cpup(flags, name) \
5878 + do { (flags) &= ~AuCpup_##name; } while (0)
5880 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
5881 +int au_sio_cpup_simple(struct au_cp_generic *cpg);
5882 +int au_sio_cpdown_simple(struct au_cp_generic *cpg);
5883 +int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
5885 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5886 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
5887 + struct au_pin *pin,
5888 + struct dentry *h_parent, void *arg),
5890 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5891 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5893 +/* ---------------------------------------------------------------------- */
5895 +/* keep timestamps when copyup */
5897 + struct dentry *dt_dentry;
5898 + struct path dt_h_path;
5899 + struct timespec dt_atime, dt_mtime;
5901 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
5902 + struct path *h_path);
5903 +void au_dtime_revert(struct au_dtime *dt);
5905 +#endif /* __KERNEL__ */
5906 +#endif /* __AUFS_CPUP_H__ */
5907 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
5908 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
5909 +++ linux/fs/aufs/dbgaufs.c 2015-09-24 10:47:58.248052907 +0200
5912 + * Copyright (C) 2005-2015 Junjiro R. Okajima
5914 + * This program, aufs is free software; you can redistribute it and/or modify
5915 + * it under the terms of the GNU General Public License as published by
5916 + * the Free Software Foundation; either version 2 of the License, or
5917 + * (at your option) any later version.
5919 + * This program is distributed in the hope that it will be useful,
5920 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5921 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5922 + * GNU General Public License for more details.
5924 + * You should have received a copy of the GNU General Public License
5925 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
5929 + * debugfs interface
5932 +#include <linux/debugfs.h>
5935 +#ifndef CONFIG_SYSFS
5936 +#error DEBUG_FS depends upon SYSFS
5939 +static struct dentry *dbgaufs;
5940 +static const mode_t dbgaufs_mode = S_IRUSR | S_IRGRP | S_IROTH;
5942 +/* 20 is max digits length of ulong 64 */
5943 +struct dbgaufs_arg {
5949 + * common function for all XINO files
5951 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
5952 + struct file *file)
5954 + kfree(file->private_data);
5958 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt)
5962 + struct dbgaufs_arg *p;
5965 + p = kmalloc(sizeof(*p), GFP_NOFS);
5971 + file->private_data = p;
5975 + err = vfs_getattr(&xf->f_path, &st);
5979 + (p->a, sizeof(p->a), "%ld, %llux%lu %lld\n",
5980 + (long)file_count(xf), st.blocks, st.blksize,
5981 + (long long)st.size);
5983 + p->n = snprintf(p->a, sizeof(p->a), "%llux%lu %lld\n",
5984 + st.blocks, st.blksize,
5985 + (long long)st.size);
5986 + AuDebugOn(p->n >= sizeof(p->a));
5988 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
5997 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
5998 + size_t count, loff_t *ppos)
6000 + struct dbgaufs_arg *p;
6002 + p = file->private_data;
6003 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6006 +/* ---------------------------------------------------------------------- */
6008 +struct dbgaufs_plink_arg {
6013 +static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6014 + struct file *file)
6016 + free_page((unsigned long)file->private_data);
6020 +static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6022 + int err, i, limit;
6023 + unsigned long n, sum;
6024 + struct dbgaufs_plink_arg *p;
6025 + struct au_sbinfo *sbinfo;
6026 + struct super_block *sb;
6027 + struct au_sphlhead *sphl;
6030 + p = (void *)get_zeroed_page(GFP_NOFS);
6035 + sbinfo = inode->i_private;
6036 + sb = sbinfo->si_sb;
6037 + si_noflush_read_lock(sb);
6038 + if (au_opt_test(au_mntflags(sb), PLINK)) {
6039 + limit = PAGE_SIZE - sizeof(p->n);
6041 + /* the number of buckets */
6042 + n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6047 + for (i = 0, sphl = sbinfo->si_plink;
6048 + i < AuPlink_NHASH;
6050 + n = au_sphl_count(sphl);
6053 + n = snprintf(p->a + p->n, limit, "%lu ", n);
6056 + if (unlikely(limit <= 0))
6059 + p->a[p->n - 1] = '\n';
6061 + /* the sum of plinks */
6062 + n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6065 + if (unlikely(limit <= 0))
6068 +#define str "1\n0\n0\n"
6069 + p->n = sizeof(str) - 1;
6070 + strcpy(p->a, str);
6073 + si_read_unlock(sb);
6076 + file->private_data = p;
6077 + goto out; /* success */
6080 + free_page((unsigned long)p);
6085 +static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6086 + size_t count, loff_t *ppos)
6088 + struct dbgaufs_plink_arg *p;
6090 + p = file->private_data;
6091 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6094 +static const struct file_operations dbgaufs_plink_fop = {
6095 + .owner = THIS_MODULE,
6096 + .open = dbgaufs_plink_open,
6097 + .release = dbgaufs_plink_release,
6098 + .read = dbgaufs_plink_read
6101 +/* ---------------------------------------------------------------------- */
6103 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6106 + struct au_sbinfo *sbinfo;
6107 + struct super_block *sb;
6109 + sbinfo = inode->i_private;
6110 + sb = sbinfo->si_sb;
6111 + si_noflush_read_lock(sb);
6112 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0);
6113 + si_read_unlock(sb);
6117 +static const struct file_operations dbgaufs_xib_fop = {
6118 + .owner = THIS_MODULE,
6119 + .open = dbgaufs_xib_open,
6120 + .release = dbgaufs_xi_release,
6121 + .read = dbgaufs_xi_read
6124 +/* ---------------------------------------------------------------------- */
6126 +#define DbgaufsXi_PREFIX "xi"
6128 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6132 + struct au_sbinfo *sbinfo;
6133 + struct super_block *sb;
6135 + struct qstr *name;
6139 + name = &file->f_path.dentry->d_name;
6140 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6141 + || memcmp(name->name, DbgaufsXi_PREFIX,
6142 + sizeof(DbgaufsXi_PREFIX) - 1)))
6144 + err = kstrtol(name->name + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
6145 + if (unlikely(err))
6148 + sbinfo = inode->i_private;
6149 + sb = sbinfo->si_sb;
6150 + si_noflush_read_lock(sb);
6151 + if (l <= au_sbend(sb)) {
6152 + xf = au_sbr(sb, (aufs_bindex_t)l)->br_xino.xi_file;
6153 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1);
6156 + si_read_unlock(sb);
6162 +static const struct file_operations dbgaufs_xino_fop = {
6163 + .owner = THIS_MODULE,
6164 + .open = dbgaufs_xino_open,
6165 + .release = dbgaufs_xi_release,
6166 + .read = dbgaufs_xi_read
6169 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6171 + aufs_bindex_t bend;
6172 + struct au_branch *br;
6173 + struct au_xino_file *xi;
6175 + if (!au_sbi(sb)->si_dbgaufs)
6178 + bend = au_sbend(sb);
6179 + for (; bindex <= bend; bindex++) {
6180 + br = au_sbr(sb, bindex);
6181 + xi = &br->br_xino;
6182 + debugfs_remove(xi->xi_dbgaufs);
6183 + xi->xi_dbgaufs = NULL;
6187 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
6189 + struct au_sbinfo *sbinfo;
6190 + struct dentry *parent;
6191 + struct au_branch *br;
6192 + struct au_xino_file *xi;
6193 + aufs_bindex_t bend;
6194 + char name[sizeof(DbgaufsXi_PREFIX) + 5]; /* "xi" bindex NULL */
6196 + sbinfo = au_sbi(sb);
6197 + parent = sbinfo->si_dbgaufs;
6201 + bend = au_sbend(sb);
6202 + for (; bindex <= bend; bindex++) {
6203 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6204 + br = au_sbr(sb, bindex);
6205 + xi = &br->br_xino;
6206 + AuDebugOn(xi->xi_dbgaufs);
6207 + xi->xi_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6208 + sbinfo, &dbgaufs_xino_fop);
6209 + /* ignore an error */
6210 + if (unlikely(!xi->xi_dbgaufs))
6211 + AuWarn1("failed %s under debugfs\n", name);
6215 +/* ---------------------------------------------------------------------- */
6217 +#ifdef CONFIG_AUFS_EXPORT
6218 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6221 + struct au_sbinfo *sbinfo;
6222 + struct super_block *sb;
6224 + sbinfo = inode->i_private;
6225 + sb = sbinfo->si_sb;
6226 + si_noflush_read_lock(sb);
6227 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0);
6228 + si_read_unlock(sb);
6232 +static const struct file_operations dbgaufs_xigen_fop = {
6233 + .owner = THIS_MODULE,
6234 + .open = dbgaufs_xigen_open,
6235 + .release = dbgaufs_xi_release,
6236 + .read = dbgaufs_xi_read
6239 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6244 + * This function is a dynamic '__init' function actually,
6245 + * so the tiny check for si_rwsem is unnecessary.
6247 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6250 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
6251 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6252 + &dbgaufs_xigen_fop);
6253 + if (sbinfo->si_dbgaufs_xigen)
6259 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6263 +#endif /* CONFIG_AUFS_EXPORT */
6265 +/* ---------------------------------------------------------------------- */
6267 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6270 + * This function is a dynamic '__fin' function actually,
6271 + * so the tiny check for si_rwsem is unnecessary.
6273 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6275 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
6276 + sbinfo->si_dbgaufs = NULL;
6277 + kobject_put(&sbinfo->si_kobj);
6280 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6283 + char name[SysaufsSiNameLen];
6286 + * This function is a dynamic '__init' function actually,
6287 + * so the tiny check for si_rwsem is unnecessary.
6289 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6293 + AuErr1("/debug/aufs is uninitialized\n");
6298 + sysaufs_name(sbinfo, name);
6299 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6300 + if (unlikely(!sbinfo->si_dbgaufs))
6302 + kobject_get(&sbinfo->si_kobj);
6304 + sbinfo->si_dbgaufs_xib = debugfs_create_file
6305 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6306 + &dbgaufs_xib_fop);
6307 + if (unlikely(!sbinfo->si_dbgaufs_xib))
6310 + sbinfo->si_dbgaufs_plink = debugfs_create_file
6311 + ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6312 + &dbgaufs_plink_fop);
6313 + if (unlikely(!sbinfo->si_dbgaufs_plink))
6316 + err = dbgaufs_xigen_init(sbinfo);
6318 + goto out; /* success */
6321 + dbgaufs_si_fin(sbinfo);
6326 +/* ---------------------------------------------------------------------- */
6328 +void dbgaufs_fin(void)
6330 + debugfs_remove(dbgaufs);
6333 +int __init dbgaufs_init(void)
6338 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6343 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6344 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
6345 +++ linux/fs/aufs/dbgaufs.h 2015-09-24 10:47:58.248052907 +0200
6348 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6350 + * This program, aufs is free software; you can redistribute it and/or modify
6351 + * it under the terms of the GNU General Public License as published by
6352 + * the Free Software Foundation; either version 2 of the License, or
6353 + * (at your option) any later version.
6355 + * This program is distributed in the hope that it will be useful,
6356 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6357 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6358 + * GNU General Public License for more details.
6360 + * You should have received a copy of the GNU General Public License
6361 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6365 + * debugfs interface
6368 +#ifndef __DBGAUFS_H__
6369 +#define __DBGAUFS_H__
6373 +struct super_block;
6376 +#ifdef CONFIG_DEBUG_FS
6378 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6379 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
6380 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6381 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6382 +void dbgaufs_fin(void);
6383 +int __init dbgaufs_init(void);
6385 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6386 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
6387 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6388 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6389 +AuStubVoid(dbgaufs_fin, void)
6390 +AuStubInt0(__init dbgaufs_init, void)
6391 +#endif /* CONFIG_DEBUG_FS */
6393 +#endif /* __KERNEL__ */
6394 +#endif /* __DBGAUFS_H__ */
6395 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6396 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
6397 +++ linux/fs/aufs/dcsub.c 2015-09-24 10:47:58.248052907 +0200
6400 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6402 + * This program, aufs is free software; you can redistribute it and/or modify
6403 + * it under the terms of the GNU General Public License as published by
6404 + * the Free Software Foundation; either version 2 of the License, or
6405 + * (at your option) any later version.
6407 + * This program is distributed in the hope that it will be useful,
6408 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6409 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6410 + * GNU General Public License for more details.
6412 + * You should have received a copy of the GNU General Public License
6413 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6417 + * sub-routines for dentry cache
6422 +static void au_dpage_free(struct au_dpage *dpage)
6425 + struct dentry **p;
6427 + p = dpage->dentries;
6428 + for (i = 0; i < dpage->ndentry; i++)
6430 + free_page((unsigned long)dpage->dentries);
6433 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6439 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6440 + if (unlikely(!dpages->dpages))
6443 + p = (void *)__get_free_page(gfp);
6447 + dpages->dpages[0].ndentry = 0;
6448 + dpages->dpages[0].dentries = p;
6449 + dpages->ndpage = 1;
6450 + return 0; /* success */
6453 + kfree(dpages->dpages);
6458 +void au_dpages_free(struct au_dcsub_pages *dpages)
6461 + struct au_dpage *p;
6463 + p = dpages->dpages;
6464 + for (i = 0; i < dpages->ndpage; i++)
6465 + au_dpage_free(p++);
6466 + kfree(dpages->dpages);
6469 +static int au_dpages_append(struct au_dcsub_pages *dpages,
6470 + struct dentry *dentry, gfp_t gfp)
6473 + struct au_dpage *dpage;
6476 + dpage = dpages->dpages + dpages->ndpage - 1;
6477 + sz = PAGE_SIZE / sizeof(dentry);
6478 + if (unlikely(dpage->ndentry >= sz)) {
6479 + AuLabel(new dpage);
6481 + sz = dpages->ndpage * sizeof(*dpages->dpages);
6482 + p = au_kzrealloc(dpages->dpages, sz,
6483 + sz + sizeof(*dpages->dpages), gfp);
6487 + dpages->dpages = p;
6488 + dpage = dpages->dpages + dpages->ndpage;
6489 + p = (void *)__get_free_page(gfp);
6493 + dpage->ndentry = 0;
6494 + dpage->dentries = p;
6498 + AuDebugOn(au_dcount(dentry) <= 0);
6499 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
6500 + return 0; /* success */
6506 +/* todo: BAD approach */
6507 +/* copied from linux/fs/dcache.c */
6515 +extern void d_walk(struct dentry *parent, void *data,
6516 + enum d_walk_ret (*enter)(void *, struct dentry *),
6517 + void (*finish)(void *));
6519 +struct ac_dpages_arg {
6521 + struct au_dcsub_pages *dpages;
6522 + struct super_block *sb;
6523 + au_dpages_test test;
6527 +static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
6529 + enum d_walk_ret ret;
6530 + struct ac_dpages_arg *arg = _arg;
6532 + ret = D_WALK_CONTINUE;
6533 + if (dentry->d_sb == arg->sb
6534 + && !IS_ROOT(dentry)
6535 + && au_dcount(dentry) > 0
6537 + && (!arg->test || arg->test(dentry, arg->arg))) {
6538 + arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
6539 + if (unlikely(arg->err))
6540 + ret = D_WALK_QUIT;
6546 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6547 + au_dpages_test test, void *arg)
6549 + struct ac_dpages_arg args = {
6557 + d_walk(root, &args, au_call_dpages_append, NULL);
6562 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6563 + int do_include, au_dpages_test test, void *arg)
6568 + write_seqlock(&rename_lock);
6569 + spin_lock(&dentry->d_lock);
6571 + && au_dcount(dentry) > 0
6572 + && (!test || test(dentry, arg)))
6573 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
6574 + spin_unlock(&dentry->d_lock);
6575 + if (unlikely(err))
6579 + * RCU for vfsmount is unnecessary since this is a traverse in a single
6582 + while (!IS_ROOT(dentry)) {
6583 + dentry = dentry->d_parent; /* rename_lock is locked */
6584 + spin_lock(&dentry->d_lock);
6585 + if (au_dcount(dentry) > 0
6586 + && (!test || test(dentry, arg)))
6587 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
6588 + spin_unlock(&dentry->d_lock);
6589 + if (unlikely(err))
6594 + write_sequnlock(&rename_lock);
6598 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
6600 + return au_di(dentry) && dentry->d_sb == arg;
6603 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6604 + struct dentry *dentry, int do_include)
6606 + return au_dcsub_pages_rev(dpages, dentry, do_include,
6607 + au_dcsub_dpages_aufs, dentry->d_sb);
6610 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
6612 + struct path path[2] = {
6621 + return path_is_under(path + 0, path + 1);
6623 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
6624 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
6625 +++ linux/fs/aufs/dcsub.h 2015-09-24 10:47:58.251386326 +0200
6628 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6630 + * This program, aufs is free software; you can redistribute it and/or modify
6631 + * it under the terms of the GNU General Public License as published by
6632 + * the Free Software Foundation; either version 2 of the License, or
6633 + * (at your option) any later version.
6635 + * This program is distributed in the hope that it will be useful,
6636 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6637 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6638 + * GNU General Public License for more details.
6640 + * You should have received a copy of the GNU General Public License
6641 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6645 + * sub-routines for dentry cache
6648 +#ifndef __AUFS_DCSUB_H__
6649 +#define __AUFS_DCSUB_H__
6653 +#include <linux/dcache.h>
6654 +#include <linux/fs.h>
6658 + struct dentry **dentries;
6661 +struct au_dcsub_pages {
6663 + struct au_dpage *dpages;
6666 +/* ---------------------------------------------------------------------- */
6669 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
6670 +void au_dpages_free(struct au_dcsub_pages *dpages);
6671 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
6672 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6673 + au_dpages_test test, void *arg);
6674 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6675 + int do_include, au_dpages_test test, void *arg);
6676 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6677 + struct dentry *dentry, int do_include);
6678 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
6680 +/* ---------------------------------------------------------------------- */
6683 + * todo: in linux-3.13, several similar (but faster) helpers are added to
6684 + * include/linux/dcache.h. Try them (in the future).
6687 +static inline int au_d_hashed_positive(struct dentry *d)
6690 + struct inode *inode = d_inode(d);
6693 + if (unlikely(d_unhashed(d)
6694 + || d_is_negative(d)
6695 + || !inode->i_nlink))
6700 +static inline int au_d_linkable(struct dentry *d)
6703 + struct inode *inode = d_inode(d);
6705 + err = au_d_hashed_positive(d);
6707 + && d_is_positive(d)
6708 + && (inode->i_state & I_LINKABLE))
6713 +static inline int au_d_alive(struct dentry *d)
6716 + struct inode *inode;
6720 + err = au_d_hashed_positive(d);
6722 + inode = d_inode(d);
6723 + if (unlikely(d_unlinked(d)
6724 + || d_is_negative(d)
6725 + || !inode->i_nlink))
6731 +static inline int au_alive_dir(struct dentry *d)
6735 + err = au_d_alive(d);
6736 + if (unlikely(err || IS_DEADDIR(d_inode(d))))
6741 +static inline int au_qstreq(struct qstr *a, struct qstr *b)
6743 + return a->len == b->len
6744 + && !memcmp(a->name, b->name, a->len);
6749 + * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
6751 + * the type of d_lockref.count became int, but the inlined function d_count()
6752 + * still returns unsigned int.
6753 + * I don't know why. Maybe it is for every d_count() users?
6754 + * Anyway au_dcount() lives on.
6756 +static inline int au_dcount(struct dentry *d)
6758 + return (int)d_count(d);
6761 +#endif /* __KERNEL__ */
6762 +#endif /* __AUFS_DCSUB_H__ */
6763 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
6764 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
6765 +++ linux/fs/aufs/debug.c 2015-09-24 10:47:58.251386326 +0200
6768 + * Copyright (C) 2005-2015 Junjiro R. Okajima
6770 + * This program, aufs is free software; you can redistribute it and/or modify
6771 + * it under the terms of the GNU General Public License as published by
6772 + * the Free Software Foundation; either version 2 of the License, or
6773 + * (at your option) any later version.
6775 + * This program is distributed in the hope that it will be useful,
6776 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
6777 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6778 + * GNU General Public License for more details.
6780 + * You should have received a copy of the GNU General Public License
6781 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
6785 + * debug print functions
6790 +/* Returns 0, or -errno. arg is in kp->arg. */
6791 +static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
6795 + err = kstrtoint(val, 0, &n);
6805 +/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
6806 +static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
6811 + return sprintf(buffer, "%d", atomic_read(a));
6814 +static struct kernel_param_ops param_ops_atomic_t = {
6815 + .set = param_atomic_t_set,
6816 + .get = param_atomic_t_get
6817 + /* void (*free)(void *arg) */
6820 +atomic_t aufs_debug = ATOMIC_INIT(0);
6821 +MODULE_PARM_DESC(debug, "debug print");
6822 +module_param_named(debug, aufs_debug, atomic_t, S_IRUGO | S_IWUSR | S_IWGRP);
6824 +DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
6825 +char *au_plevel = KERN_DEBUG;
6826 +#define dpri(fmt, ...) do { \
6828 + && strcmp(au_plevel, KERN_DEBUG)) \
6829 + || au_debug_test()) \
6830 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
6833 +/* ---------------------------------------------------------------------- */
6835 +void au_dpri_whlist(struct au_nhash *whlist)
6837 + unsigned long ul, n;
6838 + struct hlist_head *head;
6839 + struct au_vdir_wh *pos;
6841 + n = whlist->nh_num;
6842 + head = whlist->nh_head;
6843 + for (ul = 0; ul < n; ul++) {
6844 + hlist_for_each_entry(pos, head, wh_hash)
6845 + dpri("b%d, %.*s, %d\n",
6847 + pos->wh_str.len, pos->wh_str.name,
6853 +void au_dpri_vdir(struct au_vdir *vdir)
6856 + union au_vdir_deblk_p p;
6859 + if (!vdir || IS_ERR(vdir)) {
6860 + dpri("err %ld\n", PTR_ERR(vdir));
6864 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %lu\n",
6865 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
6866 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
6867 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
6868 + p.deblk = vdir->vd_deblk[ul];
6870 + dpri("[%lu]: %p\n", ul, o);
6874 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
6875 + struct dentry *wh)
6880 + if (!inode || IS_ERR(inode)) {
6881 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
6885 + /* the type of i_blocks depends upon CONFIG_LBDAF */
6886 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
6887 + && sizeof(inode->i_blocks) != sizeof(u64));
6889 + n = (void *)wh->d_name.name;
6890 + l = wh->d_name.len;
6893 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
6894 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
6896 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
6897 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
6898 + i_size_read(inode), (unsigned long long)inode->i_blocks,
6899 + hn, (long long)timespec_to_ns(&inode->i_ctime) & 0x0ffff,
6900 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
6901 + inode->i_state, inode->i_flags, inode->i_version,
6902 + inode->i_generation,
6903 + l ? ", wh " : "", l, n);
6907 +void au_dpri_inode(struct inode *inode)
6909 + struct au_iinfo *iinfo;
6910 + aufs_bindex_t bindex;
6913 + err = do_pri_inode(-1, inode, -1, NULL);
6914 + if (err || !au_test_aufs(inode->i_sb))
6917 + iinfo = au_ii(inode);
6920 + dpri("i-1: bstart %d, bend %d, gen %d\n",
6921 + iinfo->ii_bstart, iinfo->ii_bend, au_iigen(inode, NULL));
6922 + if (iinfo->ii_bstart < 0)
6925 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; bindex++) {
6926 + hn = !!au_hn(iinfo->ii_hinode + bindex);
6927 + do_pri_inode(bindex, iinfo->ii_hinode[0 + bindex].hi_inode, hn,
6928 + iinfo->ii_hinode[0 + bindex].hi_whdentry);
6932 +void au_dpri_dalias(struct inode *inode)
6936 + spin_lock(&inode->i_lock);
6937 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
6938 + au_dpri_dentry(d);
6939 + spin_unlock(&inode->i_lock);
6942 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
6944 + struct dentry *wh = NULL;
6946 + struct au_iinfo *iinfo;
6948 + if (!dentry || IS_ERR(dentry)) {
6949 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
6952 + /* do not call dget_parent() here */
6953 + /* note: access d_xxx without d_lock */
6954 + dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
6955 + bindex, dentry, dentry,
6956 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
6957 + au_dcount(dentry), dentry->d_flags,
6958 + d_unhashed(dentry) ? "un" : "");
6961 + && d_is_positive(dentry)
6962 + && au_test_aufs(dentry->d_sb)) {
6963 + iinfo = au_ii(d_inode(dentry));
6965 + hn = !!au_hn(iinfo->ii_hinode + bindex);
6966 + wh = iinfo->ii_hinode[0 + bindex].hi_whdentry;
6969 + do_pri_inode(bindex, d_inode(dentry), hn, wh);
6973 +void au_dpri_dentry(struct dentry *dentry)
6975 + struct au_dinfo *dinfo;
6976 + aufs_bindex_t bindex;
6978 + struct au_hdentry *hdp;
6980 + err = do_pri_dentry(-1, dentry);
6981 + if (err || !au_test_aufs(dentry->d_sb))
6984 + dinfo = au_di(dentry);
6987 + dpri("d-1: bstart %d, bend %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
6988 + dinfo->di_bstart, dinfo->di_bend,
6989 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
6990 + dinfo->di_tmpfile);
6991 + if (dinfo->di_bstart < 0)
6993 + hdp = dinfo->di_hdentry;
6994 + for (bindex = dinfo->di_bstart; bindex <= dinfo->di_bend; bindex++)
6995 + do_pri_dentry(bindex, hdp[0 + bindex].hd_dentry);
6998 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7002 + if (!file || IS_ERR(file)) {
7003 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7008 + && !IS_ERR_OR_NULL(file->f_path.dentry)
7009 + && au_test_aufs(file->f_path.dentry->d_sb)
7011 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
7012 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
7013 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
7014 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
7015 + file->f_version, file->f_pos, a);
7016 + if (!IS_ERR_OR_NULL(file->f_path.dentry))
7017 + do_pri_dentry(bindex, file->f_path.dentry);
7021 +void au_dpri_file(struct file *file)
7023 + struct au_finfo *finfo;
7024 + struct au_fidir *fidir;
7025 + struct au_hfile *hfile;
7026 + aufs_bindex_t bindex;
7029 + err = do_pri_file(-1, file);
7031 + || IS_ERR_OR_NULL(file->f_path.dentry)
7032 + || !au_test_aufs(file->f_path.dentry->d_sb))
7035 + finfo = au_fi(file);
7038 + if (finfo->fi_btop < 0)
7040 + fidir = finfo->fi_hdir;
7042 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7044 + for (bindex = finfo->fi_btop;
7045 + bindex >= 0 && bindex <= fidir->fd_bbot;
7047 + hfile = fidir->fd_hfile + bindex;
7048 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7052 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7054 + struct vfsmount *mnt;
7055 + struct super_block *sb;
7057 + if (!br || IS_ERR(br))
7059 + mnt = au_br_mnt(br);
7060 + if (!mnt || IS_ERR(mnt))
7063 + if (!sb || IS_ERR(sb))
7066 + dpri("s%d: {perm 0x%x, id %d, cnt %d, wbr %p}, "
7067 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
7069 + bindex, br->br_perm, br->br_id, atomic_read(&br->br_count),
7070 + br->br_wbr, au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
7071 + sb->s_flags, sb->s_count,
7072 + atomic_read(&sb->s_active), !!br->br_xino.xi_file);
7076 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7080 +void au_dpri_sb(struct super_block *sb)
7082 + struct au_sbinfo *sbinfo;
7083 + aufs_bindex_t bindex;
7085 + /* to reuduce stack size */
7087 + struct vfsmount mnt;
7088 + struct au_branch fake;
7091 + /* this function can be called from magic sysrq */
7092 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
7093 + if (unlikely(!a)) {
7094 + dpri("no memory\n");
7098 + a->mnt.mnt_sb = sb;
7099 + a->fake.br_perm = 0;
7100 + a->fake.br_path.mnt = &a->mnt;
7101 + a->fake.br_xino.xi_file = NULL;
7102 + atomic_set(&a->fake.br_count, 0);
7103 + smp_mb(); /* atomic_set */
7104 + err = do_pri_br(-1, &a->fake);
7106 + dpri("dev 0x%x\n", sb->s_dev);
7107 + if (err || !au_test_aufs(sb))
7110 + sbinfo = au_sbi(sb);
7113 + dpri("nw %d, gen %u, kobj %d\n",
7114 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7115 + atomic_read(&sbinfo->si_kobj.kref.refcount));
7116 + for (bindex = 0; bindex <= sbinfo->si_bend; bindex++)
7117 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7120 +/* ---------------------------------------------------------------------- */
7122 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7124 + struct inode *h_inode, *inode = d_inode(dentry);
7125 + struct dentry *h_dentry;
7126 + aufs_bindex_t bindex, bend, bi;
7128 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7131 + bend = au_dbend(dentry);
7132 + bi = au_ibend(inode);
7135 + bindex = au_dbstart(dentry);
7136 + bi = au_ibstart(inode);
7140 + for (; bindex <= bend; bindex++) {
7141 + h_dentry = au_h_dptr(dentry, bindex);
7144 + h_inode = au_h_iptr(inode, bindex);
7145 + if (unlikely(h_inode != d_inode(h_dentry))) {
7147 + AuDbg("b%d, %s:%d\n", bindex, func, line);
7148 + AuDbgDentry(dentry);
7149 + AuDbgInode(inode);
7156 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7159 + struct au_dcsub_pages dpages;
7160 + struct au_dpage *dpage;
7161 + struct dentry **dentries;
7163 + err = au_dpages_init(&dpages, GFP_NOFS);
7165 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
7167 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7168 + dpage = dpages.dpages + i;
7169 + dentries = dpage->dentries;
7170 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
7171 + AuDebugOn(au_digen_test(dentries[j], sigen));
7173 + au_dpages_free(&dpages);
7176 +void au_dbg_verify_kthread(void)
7178 + if (au_wkq_test()) {
7181 + * It may be recursive, but udba=notify between two aufs mounts,
7182 + * where a single ro branch is shared, is not a problem.
7188 +/* ---------------------------------------------------------------------- */
7190 +int __init au_debug_init(void)
7192 + aufs_bindex_t bindex;
7193 + struct au_vdir_destr destr;
7196 + AuDebugOn(bindex >= 0);
7199 + AuDebugOn(destr.len < NAME_MAX);
7201 +#ifdef CONFIG_4KSTACKS
7202 + pr_warn("CONFIG_4KSTACKS is defined.\n");
7207 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7208 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
7209 +++ linux/fs/aufs/debug.h 2015-09-24 10:47:58.251386326 +0200
7212 + * Copyright (C) 2005-2015 Junjiro R. Okajima
7214 + * This program, aufs is free software; you can redistribute it and/or modify
7215 + * it under the terms of the GNU General Public License as published by
7216 + * the Free Software Foundation; either version 2 of the License, or
7217 + * (at your option) any later version.
7219 + * This program is distributed in the hope that it will be useful,
7220 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7221 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7222 + * GNU General Public License for more details.
7224 + * You should have received a copy of the GNU General Public License
7225 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7229 + * debug print functions
7232 +#ifndef __AUFS_DEBUG_H__
7233 +#define __AUFS_DEBUG_H__
7237 +#include <linux/atomic.h>
7238 +#include <linux/module.h>
7239 +#include <linux/kallsyms.h>
7240 +#include <linux/sysrq.h>
7242 +#ifdef CONFIG_AUFS_DEBUG
7243 +#define AuDebugOn(a) BUG_ON(a)
7245 +/* module parameter */
7246 +extern atomic_t aufs_debug;
7247 +static inline void au_debug_on(void)
7249 + atomic_inc(&aufs_debug);
7251 +static inline void au_debug_off(void)
7253 + atomic_dec_if_positive(&aufs_debug);
7256 +static inline int au_debug_test(void)
7258 + return atomic_read(&aufs_debug) > 0;
7261 +#define AuDebugOn(a) do {} while (0)
7262 +AuStubVoid(au_debug_on, void)
7263 +AuStubVoid(au_debug_off, void)
7264 +AuStubInt0(au_debug_test, void)
7265 +#endif /* CONFIG_AUFS_DEBUG */
7267 +#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7269 +/* ---------------------------------------------------------------------- */
7273 +#define AuDbg(fmt, ...) do { \
7274 + if (au_debug_test()) \
7275 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
7277 +#define AuLabel(l) AuDbg(#l "\n")
7278 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7279 +#define AuWarn1(fmt, ...) do { \
7280 + static unsigned char _c; \
7282 + pr_warn(fmt, ##__VA_ARGS__); \
7285 +#define AuErr1(fmt, ...) do { \
7286 + static unsigned char _c; \
7288 + pr_err(fmt, ##__VA_ARGS__); \
7291 +#define AuIOErr1(fmt, ...) do { \
7292 + static unsigned char _c; \
7294 + AuIOErr(fmt, ##__VA_ARGS__); \
7297 +#define AuUnsupportMsg "This operation is not supported." \
7298 + " Please report this application to aufs-users ML."
7299 +#define AuUnsupport(fmt, ...) do { \
7300 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
7304 +#define AuTraceErr(e) do { \
7305 + if (unlikely((e) < 0)) \
7306 + AuDbg("err %d\n", (int)(e)); \
7309 +#define AuTraceErrPtr(p) do { \
7311 + AuDbg("err %ld\n", PTR_ERR(p)); \
7314 +/* dirty macros for debug print, use with "%.*s" and caution */
7315 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
7317 +/* ---------------------------------------------------------------------- */
7320 +#ifdef CONFIG_AUFS_DEBUG
7321 +extern struct mutex au_dbg_mtx;
7322 +extern char *au_plevel;
7324 +void au_dpri_whlist(struct au_nhash *whlist);
7326 +void au_dpri_vdir(struct au_vdir *vdir);
7328 +void au_dpri_inode(struct inode *inode);
7329 +void au_dpri_dalias(struct inode *inode);
7330 +void au_dpri_dentry(struct dentry *dentry);
7332 +void au_dpri_file(struct file *filp);
7333 +struct super_block;
7334 +void au_dpri_sb(struct super_block *sb);
7336 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7337 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
7338 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
7339 +void au_dbg_verify_kthread(void);
7341 +int __init au_debug_init(void);
7343 +#define AuDbgWhlist(w) do { \
7344 + mutex_lock(&au_dbg_mtx); \
7346 + au_dpri_whlist(w); \
7347 + mutex_unlock(&au_dbg_mtx); \
7350 +#define AuDbgVdir(v) do { \
7351 + mutex_lock(&au_dbg_mtx); \
7353 + au_dpri_vdir(v); \
7354 + mutex_unlock(&au_dbg_mtx); \
7357 +#define AuDbgInode(i) do { \
7358 + mutex_lock(&au_dbg_mtx); \
7360 + au_dpri_inode(i); \
7361 + mutex_unlock(&au_dbg_mtx); \
7364 +#define AuDbgDAlias(i) do { \
7365 + mutex_lock(&au_dbg_mtx); \
7367 + au_dpri_dalias(i); \
7368 + mutex_unlock(&au_dbg_mtx); \
7371 +#define AuDbgDentry(d) do { \
7372 + mutex_lock(&au_dbg_mtx); \
7374 + au_dpri_dentry(d); \
7375 + mutex_unlock(&au_dbg_mtx); \
7378 +#define AuDbgFile(f) do { \
7379 + mutex_lock(&au_dbg_mtx); \
7381 + au_dpri_file(f); \
7382 + mutex_unlock(&au_dbg_mtx); \
7385 +#define AuDbgSb(sb) do { \
7386 + mutex_lock(&au_dbg_mtx); \
7387 + AuDbg(#sb "\n"); \
7389 + mutex_unlock(&au_dbg_mtx); \
7392 +#define AuDbgSym(addr) do { \
7393 + char sym[KSYM_SYMBOL_LEN]; \
7394 + sprint_symbol(sym, (unsigned long)addr); \
7395 + AuDbg("%s\n", sym); \
7398 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
7399 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7400 +AuStubVoid(au_dbg_verify_kthread, void)
7401 +AuStubInt0(__init au_debug_init, void)
7403 +#define AuDbgWhlist(w) do {} while (0)
7404 +#define AuDbgVdir(v) do {} while (0)
7405 +#define AuDbgInode(i) do {} while (0)
7406 +#define AuDbgDAlias(i) do {} while (0)
7407 +#define AuDbgDentry(d) do {} while (0)
7408 +#define AuDbgFile(f) do {} while (0)
7409 +#define AuDbgSb(sb) do {} while (0)
7410 +#define AuDbgSym(addr) do {} while (0)
7411 +#endif /* CONFIG_AUFS_DEBUG */
7413 +/* ---------------------------------------------------------------------- */
7415 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7416 +int __init au_sysrq_init(void);
7417 +void au_sysrq_fin(void);
7419 +#ifdef CONFIG_HW_CONSOLE
7420 +#define au_dbg_blocked() do { \
7422 + handle_sysrq('w'); \
7425 +AuStubVoid(au_dbg_blocked, void)
7429 +AuStubInt0(__init au_sysrq_init, void)
7430 +AuStubVoid(au_sysrq_fin, void)
7431 +AuStubVoid(au_dbg_blocked, void)
7432 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7434 +#endif /* __KERNEL__ */
7435 +#endif /* __AUFS_DEBUG_H__ */
7436 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7437 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
7438 +++ linux/fs/aufs/dentry.c 2015-11-11 17:21:46.918863802 +0100
7441 + * Copyright (C) 2005-2015 Junjiro R. Okajima
7443 + * This program, aufs is free software; you can redistribute it and/or modify
7444 + * it under the terms of the GNU General Public License as published by
7445 + * the Free Software Foundation; either version 2 of the License, or
7446 + * (at your option) any later version.
7448 + * This program is distributed in the hope that it will be useful,
7449 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7450 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7451 + * GNU General Public License for more details.
7453 + * You should have received a copy of the GNU General Public License
7454 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
7458 + * lookup and dentry operations
7461 +#include <linux/namei.h>
7464 +#define AuLkup_ALLOW_NEG 1
7465 +#define AuLkup_IGNORE_PERM (1 << 1)
7466 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
7467 +#define au_fset_lkup(flags, name) \
7468 + do { (flags) |= AuLkup_##name; } while (0)
7469 +#define au_fclr_lkup(flags, name) \
7470 + do { (flags) &= ~AuLkup_##name; } while (0)
7472 +struct au_do_lookup_args {
7473 + unsigned int flags;
7478 + * returns positive/negative dentry, NULL or an error.
7479 + * NULL means whiteout-ed or not-found.
7481 +static struct dentry*
7482 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7483 + aufs_bindex_t bindex, struct qstr *wh_name,
7484 + struct au_do_lookup_args *args)
7486 + struct dentry *h_dentry;
7487 + struct inode *h_inode;
7488 + struct au_branch *br;
7489 + int wh_found, opq;
7490 + unsigned char wh_able;
7491 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
7492 + const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
7496 + br = au_sbr(dentry->d_sb, bindex);
7497 + wh_able = !!au_br_whable(br->br_perm);
7499 + wh_found = au_wh_test(h_parent, wh_name, /*try_sio*/0);
7500 + h_dentry = ERR_PTR(wh_found);
7503 + if (unlikely(wh_found < 0))
7506 + /* We found a whiteout */
7507 + /* au_set_dbend(dentry, bindex); */
7508 + au_set_dbwh(dentry, bindex);
7510 + return NULL; /* success */
7514 + h_dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
7516 + h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
7517 + if (IS_ERR(h_dentry)) {
7518 + if (PTR_ERR(h_dentry) == -ENAMETOOLONG
7524 + h_inode = d_inode(h_dentry);
7525 + if (d_is_negative(h_dentry)) {
7528 + } else if (wh_found
7529 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
7532 + if (au_dbend(dentry) <= bindex)
7533 + au_set_dbend(dentry, bindex);
7534 + if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
7535 + au_set_dbstart(dentry, bindex);
7536 + au_set_h_dptr(dentry, bindex, h_dentry);
7538 + if (!d_is_dir(h_dentry)
7540 + || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
7541 + goto out; /* success */
7543 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
7544 + opq = au_diropq_test(h_dentry);
7545 + mutex_unlock(&h_inode->i_mutex);
7547 + au_set_dbdiropq(dentry, bindex);
7548 + else if (unlikely(opq < 0)) {
7549 + au_set_h_dptr(dentry, bindex, NULL);
7550 + h_dentry = ERR_PTR(opq);
7561 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
7563 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
7564 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
7570 + * returns the number of lower positive dentries,
7571 + * otherwise an error.
7572 + * can be called at unlinking with @type is zero.
7574 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type)
7576 + int npositive, err;
7577 + aufs_bindex_t bindex, btail, bdiropq;
7578 + unsigned char isdir, dirperm1;
7579 + struct qstr whname;
7580 + struct au_do_lookup_args args = {
7584 + const struct qstr *name = &dentry->d_name;
7585 + struct dentry *parent;
7586 + struct super_block *sb;
7588 + sb = dentry->d_sb;
7589 + err = au_test_shwh(sb, name);
7590 + if (unlikely(err))
7593 + err = au_wh_name_alloc(&whname, name);
7594 + if (unlikely(err))
7597 + isdir = !!d_is_dir(dentry);
7599 + au_fset_lkup(args.flags, ALLOW_NEG);
7600 + dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
7603 + parent = dget_parent(dentry);
7604 + btail = au_dbtaildir(parent);
7605 + for (bindex = bstart; bindex <= btail; bindex++) {
7606 + struct dentry *h_parent, *h_dentry;
7607 + struct inode *h_inode, *h_dir;
7609 + h_dentry = au_h_dptr(dentry, bindex);
7611 + if (d_is_positive(h_dentry))
7613 + if (type != S_IFDIR)
7617 + h_parent = au_h_dptr(parent, bindex);
7618 + if (!h_parent || !d_is_dir(h_parent))
7621 + h_dir = d_inode(h_parent);
7622 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
7623 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &whname,
7625 + mutex_unlock(&h_dir->i_mutex);
7626 + err = PTR_ERR(h_dentry);
7627 + if (IS_ERR(h_dentry))
7630 + au_fclr_lkup(args.flags, ALLOW_NEG);
7632 + au_fset_lkup(args.flags, IGNORE_PERM);
7634 + if (au_dbwh(dentry) == bindex)
7638 + if (d_is_negative(h_dentry))
7640 + h_inode = d_inode(h_dentry);
7643 + args.type = h_inode->i_mode & S_IFMT;
7644 + if (args.type != S_IFDIR)
7647 + /* the type of lower may be different */
7648 + bdiropq = au_dbdiropq(dentry);
7649 + if (bdiropq >= 0 && bdiropq <= bindex)
7655 + AuLabel(positive);
7656 + au_update_dbstart(dentry);
7659 + if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
7660 + && au_dbstart(dentry) < 0)) {
7662 + AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
7668 + kfree(whname.name);
7673 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent)
7675 + struct dentry *dentry;
7678 + if (!au_test_h_perm_sio(d_inode(parent), MAY_EXEC))
7679 + dentry = vfsub_lkup_one(name, parent);
7681 + struct vfsub_lkup_one_args args = {
7687 + wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
7688 + if (unlikely(wkq_err))
7689 + dentry = ERR_PTR(wkq_err);
7696 + * lookup @dentry on @bindex which should be negative.
7698 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
7701 + struct dentry *parent, *h_parent, *h_dentry;
7702 + struct au_branch *br;
7704 + parent = dget_parent(dentry);
7705 + h_parent = au_h_dptr(parent, bindex);
7706 + br = au_sbr(dentry->d_sb, bindex);
7708 + h_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
7710 + h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
7711 + err = PTR_ERR(h_dentry);
7712 + if (IS_ERR(h_dentry))
7714 + if (unlikely(d_is_positive(h_dentry))) {
7716 + AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
7722 + if (bindex < au_dbstart(dentry))
7723 + au_set_dbstart(dentry, bindex);
7724 + if (au_dbend(dentry) < bindex)
7725 + au_set_dbend(dentry, bindex);
7726 + au_set_h_dptr(dentry, bindex, h_dentry);
7733 +/* ---------------------------------------------------------------------- */
7735 +/* subset of struct inode */
7737 + unsigned long i_ino;
7738 + /* unsigned int i_nlink; */
7744 + blkcnt_t i_blocks;
7749 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
7751 + ia->i_ino = h_inode->i_ino;
7752 + /* ia->i_nlink = h_inode->i_nlink; */
7753 + ia->i_uid = h_inode->i_uid;
7754 + 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_iattr_test(struct au_iattr *ia, struct inode *h_inode)
7765 + return ia->i_ino != h_inode->i_ino
7766 + /* || ia->i_nlink != h_inode->i_nlink */
7767 + || !uid_eq(ia->i_uid, h_inode->i_uid)
7768 + || !gid_eq(ia->i_gid, h_inode->i_gid)
7769 + || ia->i_version != h_inode->i_version
7771 + || ia->i_size != h_inode->i_size
7772 + || ia->i_blocks != h_inode->i_blocks
7774 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
7777 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
7778 + struct au_branch *br)
7781 + struct au_iattr ia;
7782 + struct inode *h_inode;
7783 + struct dentry *h_d;
7784 + struct super_block *h_sb;
7787 + memset(&ia, -1, sizeof(ia));
7788 + h_sb = h_dentry->d_sb;
7790 + if (d_is_positive(h_dentry)) {
7791 + h_inode = d_inode(h_dentry);
7792 + au_iattr_save(&ia, h_inode);
7793 + } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
7794 + /* nfs d_revalidate may return 0 for negative dentry */
7795 + /* fuse d_revalidate always return 0 for negative dentry */
7798 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
7799 + h_d = vfsub_lkup_one(&h_dentry->d_name, h_parent);
7800 + err = PTR_ERR(h_d);
7805 + if (unlikely(h_d != h_dentry
7806 + || d_inode(h_d) != h_inode
7807 + || (h_inode && au_iattr_test(&ia, h_inode))))
7808 + err = au_busy_or_stale();
7816 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
7817 + struct dentry *h_parent, struct au_branch *br)
7822 + if (udba == AuOpt_UDBA_REVAL
7823 + && !au_test_fs_remote(h_dentry->d_sb)) {
7825 + err = (d_inode(h_dentry->d_parent) != h_dir);
7826 + } else if (udba != AuOpt_UDBA_NONE)
7827 + err = au_h_verify_dentry(h_dentry, h_parent, br);
7832 +/* ---------------------------------------------------------------------- */
7834 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
7837 + aufs_bindex_t new_bindex, bindex, bend, bwh, bdiropq;
7838 + struct au_hdentry tmp, *p, *q;
7839 + struct au_dinfo *dinfo;
7840 + struct super_block *sb;
7842 + DiMustWriteLock(dentry);
7844 + sb = dentry->d_sb;
7845 + dinfo = au_di(dentry);
7846 + bend = dinfo->di_bend;
7847 + bwh = dinfo->di_bwh;
7848 + bdiropq = dinfo->di_bdiropq;
7849 + p = dinfo->di_hdentry + dinfo->di_bstart;
7850 + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++, p++) {
7851 + if (!p->hd_dentry)
7854 + new_bindex = au_br_index(sb, p->hd_id);
7855 + if (new_bindex == bindex)
7858 + if (dinfo->di_bwh == bindex)
7860 + if (dinfo->di_bdiropq == bindex)
7861 + bdiropq = new_bindex;
7862 + if (new_bindex < 0) {
7864 + p->hd_dentry = NULL;
7868 + /* swap two lower dentries, and loop again */
7869 + q = dinfo->di_hdentry + new_bindex;
7873 + if (tmp.hd_dentry) {
7879 + dinfo->di_bwh = -1;
7880 + if (bwh >= 0 && bwh <= au_sbend(sb) && au_sbr_whable(sb, bwh))
7881 + dinfo->di_bwh = bwh;
7883 + dinfo->di_bdiropq = -1;
7885 + && bdiropq <= au_sbend(sb)
7886 + && au_sbr_whable(sb, bdiropq))
7887 + dinfo->di_bdiropq = bdiropq;
7890 + dinfo->di_bstart = -1;
7891 + dinfo->di_bend = -1;
7892 + bend = au_dbend(parent);
7893 + p = dinfo->di_hdentry;
7894 + for (bindex = 0; bindex <= bend; bindex++, p++)
7895 + if (p->hd_dentry) {
7896 + dinfo->di_bstart = bindex;
7900 + if (dinfo->di_bstart >= 0) {
7901 + p = dinfo->di_hdentry + bend;
7902 + for (bindex = bend; bindex >= 0; bindex--, p--)
7903 + if (p->hd_dentry) {
7904 + dinfo->di_bend = bindex;
7913 +static void au_do_hide(struct dentry *dentry)
7915 + struct inode *inode;
7917 + if (d_really_is_positive(dentry)) {
7918 + inode = d_inode(dentry);
7919 + if (!d_is_dir(dentry)) {
7920 + if (inode->i_nlink && !d_unhashed(dentry))
7921 + drop_nlink(inode);
7923 + clear_nlink(inode);
7924 + /* stop next lookup */
7925 + inode->i_flags |= S_DEAD;
7927 + smp_mb(); /* necessary? */
7932 +static int au_hide_children(struct dentry *parent)
7934 + int err, i, j, ndentry;
7935 + struct au_dcsub_pages dpages;
7936 + struct au_dpage *dpage;
7937 + struct dentry *dentry;
7939 + err = au_dpages_init(&dpages, GFP_NOFS);
7940 + if (unlikely(err))
7942 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
7943 + if (unlikely(err))
7946 + /* in reverse order */
7947 + for (i = dpages.ndpage - 1; i >= 0; i--) {
7948 + dpage = dpages.dpages + i;
7949 + ndentry = dpage->ndentry;
7950 + for (j = ndentry - 1; j >= 0; j--) {
7951 + dentry = dpage->dentries[j];
7952 + if (dentry != parent)
7953 + au_do_hide(dentry);
7958 + au_dpages_free(&dpages);
7963 +static void au_hide(struct dentry *dentry)
7967 + AuDbgDentry(dentry);
7968 + if (d_is_dir(dentry)) {
7969 + /* shrink_dcache_parent(dentry); */
7970 + err = au_hide_children(dentry);
7971 + if (unlikely(err))
7972 + AuIOErr("%pd, failed hiding children, ignored %d\n",
7975 + au_do_hide(dentry);
7979 + * By adding a dirty branch, a cached dentry may be affected in various ways.
7981 + * a dirty branch is added
7982 + * - on the top of layers
7983 + * - in the middle of layers
7984 + * - to the bottom of layers
7986 + * on the added branch there exists
7989 + * - a same named entry
7991 + * * negative --> positive
7992 + * * positive --> positive
7993 + * - type is unchanged
7994 + * - type is changed
7996 + * * negative --> negative
7997 + * * positive --> negative (rejected by au_br_del() for non-dir case)
8000 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8001 + struct au_dinfo *tmp)
8004 + aufs_bindex_t bindex, bend;
8006 + struct dentry *dentry;
8007 + struct inode *inode;
8010 + struct au_hdentry *hd;
8011 + struct inode *inode, *h_inode;
8012 + struct dentry *h_dentry;
8015 + AuDebugOn(dinfo->di_bstart < 0);
8017 + orig_h.dentry = dinfo->di_hdentry[dinfo->di_bstart].hd_dentry;
8018 + orig_h.inode = NULL;
8019 + if (d_is_positive(orig_h.dentry)) {
8020 + orig_h.inode = d_inode(orig_h.dentry);
8021 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
8023 + memset(&tmp_h, 0, sizeof(tmp_h));
8024 + if (tmp->di_bstart >= 0) {
8025 + tmp_h.dentry = tmp->di_hdentry[tmp->di_bstart].hd_dentry;
8026 + tmp_h.inode = NULL;
8027 + if (d_is_positive(tmp_h.dentry)) {
8028 + tmp_h.inode = d_inode(tmp_h.dentry);
8029 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
8034 + if (d_really_is_positive(dentry))
8035 + inode = d_inode(dentry);
8036 + if (!orig_h.inode) {
8037 + AuDbg("nagative originally\n");
8043 + AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
8044 + AuDebugOn(dinfo->di_bdiropq != -1);
8046 + if (!tmp_h.inode) {
8047 + AuDbg("negative --> negative\n");
8048 + /* should have only one negative lower */
8049 + if (tmp->di_bstart >= 0
8050 + && tmp->di_bstart < dinfo->di_bstart) {
8051 + AuDebugOn(tmp->di_bstart != tmp->di_bend);
8052 + AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
8053 + au_set_h_dptr(dentry, dinfo->di_bstart, NULL);
8054 + au_di_cp(dinfo, tmp);
8055 + hd = tmp->di_hdentry + tmp->di_bstart;
8056 + au_set_h_dptr(dentry, tmp->di_bstart,
8057 + dget(hd->hd_dentry));
8059 + au_dbg_verify_dinode(dentry);
8061 + AuDbg("negative --> positive\n");
8063 + * similar to the behaviour of creating with bypassing
8065 + * unhash it in order to force an error in the
8066 + * succeeding create operation.
8067 + * we should not set S_DEAD here.
8070 + /* au_di_swap(tmp, dinfo); */
8071 + au_dbg_verify_dinode(dentry);
8074 + AuDbg("positive originally\n");
8075 + /* inode may be NULL */
8076 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8077 + if (!tmp_h.inode) {
8078 + AuDbg("positive --> negative\n");
8079 + /* or bypassing aufs */
8081 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_bstart)
8082 + dinfo->di_bwh = tmp->di_bwh;
8084 + err = au_refresh_hinode_self(inode);
8085 + au_dbg_verify_dinode(dentry);
8086 + } else if (orig_h.mode == tmp_h.mode) {
8087 + AuDbg("positive --> positive, same type\n");
8088 + if (!S_ISDIR(orig_h.mode)
8089 + && dinfo->di_bstart > tmp->di_bstart) {
8091 + * similar to the behaviour of removing and
8096 + err = au_refresh_hinode_self(inode);
8097 + au_dbg_verify_dinode(dentry);
8099 + /* fill empty slots */
8100 + if (dinfo->di_bstart > tmp->di_bstart)
8101 + dinfo->di_bstart = tmp->di_bstart;
8102 + if (dinfo->di_bend < tmp->di_bend)
8103 + dinfo->di_bend = tmp->di_bend;
8104 + dinfo->di_bwh = tmp->di_bwh;
8105 + dinfo->di_bdiropq = tmp->di_bdiropq;
8106 + hd = tmp->di_hdentry;
8107 + bend = dinfo->di_bend;
8108 + for (bindex = tmp->di_bstart; bindex <= bend;
8110 + if (au_h_dptr(dentry, bindex))
8112 + h_dentry = hd[bindex].hd_dentry;
8115 + AuDebugOn(d_is_negative(h_dentry));
8116 + h_inode = d_inode(h_dentry);
8117 + AuDebugOn(orig_h.mode
8118 + != (h_inode->i_mode
8120 + au_set_h_dptr(dentry, bindex,
8123 + err = au_refresh_hinode(inode, dentry);
8124 + au_dbg_verify_dinode(dentry);
8127 + AuDbg("positive --> positive, different type\n");
8128 + /* similar to the behaviour of removing and creating */
8131 + err = au_refresh_hinode_self(inode);
8132 + au_dbg_verify_dinode(dentry);
8140 +void au_refresh_dop(struct dentry *dentry, int force_reval)
8142 + const struct dentry_operations *dop
8143 + = force_reval ? &aufs_dop : dentry->d_sb->s_d_op;
8144 + static const unsigned int mask
8145 + = DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE;
8147 + BUILD_BUG_ON(sizeof(mask) != sizeof(dentry->d_flags));
8149 + if (dentry->d_op == dop)
8152 + AuDbg("%pd\n", dentry);
8153 + spin_lock(&dentry->d_lock);
8154 + if (dop == &aufs_dop)
8155 + dentry->d_flags |= mask;
8157 + dentry->d_flags &= ~mask;
8158 + dentry->d_op = dop;
8159 + spin_unlock(&dentry->d_lock);
8162 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8165 + unsigned int sigen;
8166 + struct au_dinfo *dinfo, *tmp;
8167 + struct super_block *sb;
8168 + struct inode *inode;
8170 + DiMustWriteLock(dentry);
8171 + AuDebugOn(IS_ROOT(dentry));
8172 + AuDebugOn(d_really_is_negative(parent));
8174 + sb = dentry->d_sb;
8175 + sigen = au_sigen(sb);
8176 + err = au_digen_test(parent, sigen);
8177 + if (unlikely(err))
8180 + dinfo = au_di(dentry);
8181 + err = au_di_realloc(dinfo, au_sbend(sb) + 1);
8182 + if (unlikely(err))
8184 + ebrange = au_dbrange_test(dentry);
8186 + ebrange = au_do_refresh_hdentry(dentry, parent);
8188 + if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
8189 + AuDebugOn(au_dbstart(dentry) < 0 && au_dbend(dentry) >= 0);
8190 + if (d_really_is_positive(dentry)) {
8191 + inode = d_inode(dentry);
8192 + err = au_refresh_hinode_self(inode);
8194 + au_dbg_verify_dinode(dentry);
8196 + goto out_dgen; /* success */
8200 + /* temporary dinfo */
8201 + AuDbgDentry(dentry);
8203 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8204 + if (unlikely(!tmp))
8206 + au_di_swap(tmp, dinfo);
8207 + /* returns the number of positive dentries */
8209 + * if current working dir is removed, it returns an error.
8210 + * but the dentry is legal.
8212 + err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
8213 + AuDbgDentry(dentry);
8214 + au_di_swap(tmp, dinfo);
8215 + if (err == -ENOENT)
8218 + /* compare/refresh by dinfo */
8219 + AuDbgDentry(dentry);
8220 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8221 + au_dbg_verify_dinode(dentry);
8224 + au_rw_write_unlock(&tmp->di_rwsem);
8226 + if (unlikely(err))
8230 + au_update_digen(dentry);
8232 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
8233 + AuIOErr("failed refreshing %pd, %d\n", dentry, err);
8234 + AuDbgDentry(dentry);
8240 +static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8241 + struct dentry *dentry, aufs_bindex_t bindex)
8246 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8249 + AuDbg("b%d\n", bindex);
8251 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8252 + * due to whiteout and branch permission.
8254 + flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8255 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
8256 + /* it may return tri-state */
8257 + valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
8259 + if (unlikely(valid < 0))
8269 +/* todo: remove this */
8270 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
8271 + unsigned int flags, int do_udba)
8274 + umode_t mode, h_mode;
8275 + aufs_bindex_t bindex, btail, bstart, ibs, ibe;
8276 + unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
8277 + struct inode *h_inode, *h_cached_inode;
8278 + struct dentry *h_dentry;
8279 + struct qstr *name, *h_name;
8286 + unhashed = !!d_unhashed(dentry);
8287 + is_root = !!IS_ROOT(dentry);
8288 + name = &dentry->d_name;
8289 + tmpfile = au_di(dentry)->di_tmpfile;
8292 + * Theoretically, REVAL test should be unnecessary in case of
8294 + * But {fs,i}notify doesn't fire some necessary events,
8295 + * IN_ATTRIB for atime/nlink/pageio
8296 + * Let's do REVAL test too.
8298 + if (do_udba && inode) {
8299 + mode = (inode->i_mode & S_IFMT);
8300 + plus = (inode->i_nlink > 0);
8301 + ibs = au_ibstart(inode);
8302 + ibe = au_ibend(inode);
8305 + bstart = au_dbstart(dentry);
8307 + if (inode && S_ISDIR(inode->i_mode))
8308 + btail = au_dbtaildir(dentry);
8309 + for (bindex = bstart; bindex <= btail; bindex++) {
8310 + h_dentry = au_h_dptr(dentry, bindex);
8314 + AuDbg("b%d, %pd\n", bindex, h_dentry);
8315 + h_nfs = !!au_test_nfs(h_dentry->d_sb);
8316 + spin_lock(&h_dentry->d_lock);
8317 + h_name = &h_dentry->d_name;
8318 + if (unlikely(do_udba
8321 + && (unhashed != !!d_unhashed(h_dentry)
8323 + && !au_qstreq(name, h_name))
8326 + && !(flags & LOOKUP_OPEN)
8327 + && (h_dentry->d_flags
8328 + & DCACHE_NFSFS_RENAMED)))
8332 + h_unhashed = d_unhashed(h_dentry);
8333 + spin_unlock(&h_dentry->d_lock);
8334 + AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8335 + unhashed, h_unhashed, dentry, h_dentry);
8338 + spin_unlock(&h_dentry->d_lock);
8340 + err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
8341 + if (unlikely(err))
8342 + /* do not goto err, to keep the errno */
8345 + /* todo: plink too? */
8350 + if (unlikely(!!inode != d_is_positive(h_dentry)))
8354 + if (d_is_positive(h_dentry))
8355 + h_inode = d_inode(h_dentry);
8358 + h_cached_inode = h_inode;
8360 + h_mode = (h_inode->i_mode & S_IFMT);
8361 + h_plus = (h_inode->i_nlink > 0);
8363 + if (inode && ibs <= bindex && bindex <= ibe)
8364 + h_cached_inode = au_h_iptr(inode, bindex);
8367 + if (unlikely(plus != h_plus && !tmpfile))
8370 + if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8372 + && !IS_ROOT(h_dentry)
8373 + && unhashed != d_unhashed(h_dentry)))
8376 + if (unlikely(mode != h_mode
8377 + || h_cached_inode != h_inode))
8390 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
8391 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8394 + struct dentry *parent;
8396 + if (!au_digen_test(dentry, sigen))
8399 + parent = dget_parent(dentry);
8400 + di_read_lock_parent(parent, AuLock_IR);
8401 + AuDebugOn(au_digen_test(parent, sigen));
8402 + au_dbg_verify_gen(parent, sigen);
8403 + err = au_refresh_dentry(dentry, parent);
8404 + di_read_unlock(parent, AuLock_IR);
8410 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8413 + struct dentry *d, *parent;
8415 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
8416 + return simple_reval_dpath(dentry, sigen);
8418 + /* slow loop, keep it simple and stupid */
8419 + /* cf: au_cpup_dirs() */
8422 + while (au_digen_test(dentry, sigen)) {
8426 + parent = dget_parent(d);
8427 + if (!au_digen_test(parent, sigen))
8433 + di_write_lock_child2(d);
8435 + /* someone might update our dentry while we were sleeping */
8436 + if (au_digen_test(d, sigen)) {
8438 + * todo: consolidate with simple_reval_dpath(),
8439 + * do_refresh() and au_reval_for_attr().
8441 + di_read_lock_parent(parent, AuLock_IR);
8442 + err = au_refresh_dentry(d, parent);
8443 + di_read_unlock(parent, AuLock_IR);
8447 + di_write_unlock(d);
8449 + if (unlikely(err))
8457 + * if valid returns 1, otherwise 0.
8459 +static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
8462 + unsigned int sigen;
8463 + unsigned char do_udba;
8464 + struct super_block *sb;
8465 + struct inode *inode;
8467 + /* todo: support rcu-walk? */
8468 + if (flags & LOOKUP_RCU)
8472 + if (unlikely(!au_di(dentry)))
8476 + sb = dentry->d_sb;
8479 + * i_mutex of parent dir may be held,
8480 + * but we should not return 'invalid' due to busy.
8482 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
8483 + if (unlikely(err)) {
8489 + if (d_really_is_positive(dentry))
8490 + inode = d_inode(dentry);
8491 + if (unlikely(inode && is_bad_inode(inode))) {
8496 + if (unlikely(au_dbrange_test(dentry))) {
8502 + sigen = au_sigen(sb);
8503 + if (au_digen_test(dentry, sigen)) {
8504 + AuDebugOn(IS_ROOT(dentry));
8505 + err = au_reval_dpath(dentry, sigen);
8506 + if (unlikely(err)) {
8511 + di_downgrade_lock(dentry, AuLock_IR);
8514 + if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
8516 + && !(inode->i_state && I_LINKABLE)
8517 + && (IS_DEADDIR(inode) || !inode->i_nlink)) {
8522 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
8523 + if (do_udba && inode) {
8524 + aufs_bindex_t bstart = au_ibstart(inode);
8525 + struct inode *h_inode;
8527 + if (bstart >= 0) {
8528 + h_inode = au_h_iptr(inode, bstart);
8529 + if (h_inode && au_test_higen(inode, h_inode)) {
8536 + err = h_d_revalidate(dentry, inode, flags, do_udba);
8537 + if (unlikely(!err && do_udba && au_dbstart(dentry) < 0)) {
8539 + AuDbg("both of real entry and whiteout found, %p, err %d\n",
8545 + di_downgrade_lock(dentry, AuLock_IR);
8547 + aufs_read_unlock(dentry, AuLock_IR);
8552 + AuDbg("%pd invalid, %d\n", dentry, valid);
8558 +static void aufs_d_release(struct dentry *dentry)
8560 + if (au_di(dentry)) {
8561 + au_di_fin(dentry);
8562 + au_hn_di_reinit(dentry);
8566 +const struct dentry_operations aufs_dop = {
8567 + .d_revalidate = aufs_d_revalidate,
8568 + .d_weak_revalidate = aufs_d_revalidate,
8569 + .d_release = aufs_d_release
8572 +/* aufs_dop without d_revalidate */
8573 +const struct dentry_operations aufs_dop_noreval = {
8574 + .d_release = aufs_d_release
8576 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
8577 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
8578 +++ linux/fs/aufs/dentry.h 2015-11-11 17:21:46.918863802 +0100
8581 + * Copyright (C) 2005-2015 Junjiro R. Okajima
8583 + * This program, aufs is free software; you can redistribute it and/or modify
8584 + * it under the terms of the GNU General Public License as published by
8585 + * the Free Software Foundation; either version 2 of the License, or
8586 + * (at your option) any later version.
8588 + * This program is distributed in the hope that it will be useful,
8589 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8590 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8591 + * GNU General Public License for more details.
8593 + * You should have received a copy of the GNU General Public License
8594 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
8598 + * lookup and dentry operations
8601 +#ifndef __AUFS_DENTRY_H__
8602 +#define __AUFS_DENTRY_H__
8606 +#include <linux/dcache.h>
8609 +struct au_hdentry {
8610 + struct dentry *hd_dentry;
8611 + aufs_bindex_t hd_id;
8615 + atomic_t di_generation;
8617 + struct au_rwsem di_rwsem;
8618 + aufs_bindex_t di_bstart, di_bend, di_bwh, di_bdiropq;
8619 + unsigned char di_tmpfile; /* to allow the different name */
8620 + struct au_hdentry *di_hdentry;
8621 +} ____cacheline_aligned_in_smp;
8623 +/* ---------------------------------------------------------------------- */
8626 +extern const struct dentry_operations aufs_dop, aufs_dop_noreval;
8628 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent);
8629 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8630 + struct dentry *h_parent, struct au_branch *br);
8632 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type);
8633 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
8634 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
8635 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
8636 +void au_refresh_dop(struct dentry *dentry, int force_reval);
8639 +void au_di_init_once(void *_di);
8640 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
8641 +void au_di_free(struct au_dinfo *dinfo);
8642 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
8643 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
8644 +int au_di_init(struct dentry *dentry);
8645 +void au_di_fin(struct dentry *dentry);
8646 +int au_di_realloc(struct au_dinfo *dinfo, int nbr);
8648 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
8649 +void di_read_unlock(struct dentry *d, int flags);
8650 +void di_downgrade_lock(struct dentry *d, int flags);
8651 +void di_write_lock(struct dentry *d, unsigned int lsc);
8652 +void di_write_unlock(struct dentry *d);
8653 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
8654 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
8655 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
8657 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
8658 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
8659 +aufs_bindex_t au_dbtail(struct dentry *dentry);
8660 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
8662 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
8663 + struct dentry *h_dentry);
8664 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
8665 +int au_dbrange_test(struct dentry *dentry);
8666 +void au_update_digen(struct dentry *dentry);
8667 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
8668 +void au_update_dbstart(struct dentry *dentry);
8669 +void au_update_dbend(struct dentry *dentry);
8670 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
8672 +/* ---------------------------------------------------------------------- */
8674 +static inline struct au_dinfo *au_di(struct dentry *dentry)
8676 + return dentry->d_fsdata;
8679 +/* ---------------------------------------------------------------------- */
8681 +/* lock subclass for dinfo */
8683 + AuLsc_DI_CHILD, /* child first */
8684 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
8685 + AuLsc_DI_CHILD3, /* copyup dirs */
8689 + AuLsc_DI_TMP /* temp for replacing dinfo */
8693 + * di_read_lock_child, di_write_lock_child,
8694 + * di_read_lock_child2, di_write_lock_child2,
8695 + * di_read_lock_child3, di_write_lock_child3,
8696 + * di_read_lock_parent, di_write_lock_parent,
8697 + * di_read_lock_parent2, di_write_lock_parent2,
8698 + * di_read_lock_parent3, di_write_lock_parent3,
8700 +#define AuReadLockFunc(name, lsc) \
8701 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
8702 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
8704 +#define AuWriteLockFunc(name, lsc) \
8705 +static inline void di_write_lock_##name(struct dentry *d) \
8706 +{ di_write_lock(d, AuLsc_DI_##lsc); }
8708 +#define AuRWLockFuncs(name, lsc) \
8709 + AuReadLockFunc(name, lsc) \
8710 + AuWriteLockFunc(name, lsc)
8712 +AuRWLockFuncs(child, CHILD);
8713 +AuRWLockFuncs(child2, CHILD2);
8714 +AuRWLockFuncs(child3, CHILD3);
8715 +AuRWLockFuncs(parent, PARENT);
8716 +AuRWLockFuncs(parent2, PARENT2);
8717 +AuRWLockFuncs(parent3, PARENT3);
8719 +#undef AuReadLockFunc
8720 +#undef AuWriteLockFunc
8721 +#undef AuRWLockFuncs
8723 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
8724 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
8725 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
8727 +/* ---------------------------------------------------------------------- */
8729 +/* todo: memory barrier? */
8730 +static inline unsigned int au_digen(struct dentry *d)
8732 + return atomic_read(&au_di(d)->di_generation);
8735 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
8737 + hdentry->hd_dentry = NULL;
8740 +static inline void au_hdput(struct au_hdentry *hd)
8743 + dput(hd->hd_dentry);
8746 +static inline aufs_bindex_t au_dbstart(struct dentry *dentry)
8748 + DiMustAnyLock(dentry);
8749 + return au_di(dentry)->di_bstart;
8752 +static inline aufs_bindex_t au_dbend(struct dentry *dentry)
8754 + DiMustAnyLock(dentry);
8755 + return au_di(dentry)->di_bend;
8758 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
8760 + DiMustAnyLock(dentry);
8761 + return au_di(dentry)->di_bwh;
8764 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
8766 + DiMustAnyLock(dentry);
8767 + return au_di(dentry)->di_bdiropq;
8770 +/* todo: hard/soft set? */
8771 +static inline void au_set_dbstart(struct dentry *dentry, aufs_bindex_t bindex)
8773 + DiMustWriteLock(dentry);
8774 + au_di(dentry)->di_bstart = bindex;
8777 +static inline void au_set_dbend(struct dentry *dentry, aufs_bindex_t bindex)
8779 + DiMustWriteLock(dentry);
8780 + au_di(dentry)->di_bend = bindex;
8783 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
8785 + DiMustWriteLock(dentry);
8786 + /* dbwh can be outside of bstart - bend range */
8787 + au_di(dentry)->di_bwh = bindex;
8790 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
8792 + DiMustWriteLock(dentry);
8793 + au_di(dentry)->di_bdiropq = bindex;
8796 +/* ---------------------------------------------------------------------- */
8798 +#ifdef CONFIG_AUFS_HNOTIFY
8799 +static inline void au_digen_dec(struct dentry *d)
8801 + atomic_dec(&au_di(d)->di_generation);
8804 +static inline void au_hn_di_reinit(struct dentry *dentry)
8806 + dentry->d_fsdata = NULL;
8809 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
8810 +#endif /* CONFIG_AUFS_HNOTIFY */
8812 +#endif /* __KERNEL__ */
8813 +#endif /* __AUFS_DENTRY_H__ */
8814 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
8815 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
8816 +++ linux/fs/aufs/dinfo.c 2015-09-24 10:47:58.251386326 +0200
8819 + * Copyright (C) 2005-2015 Junjiro R. Okajima
8821 + * This program, aufs is free software; you can redistribute it and/or modify
8822 + * it under the terms of the GNU General Public License as published by
8823 + * the Free Software Foundation; either version 2 of the License, or
8824 + * (at your option) any later version.
8826 + * This program is distributed in the hope that it will be useful,
8827 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8828 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8829 + * GNU General Public License for more details.
8831 + * You should have received a copy of the GNU General Public License
8832 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
8836 + * dentry private data
8841 +void au_di_init_once(void *_dinfo)
8843 + struct au_dinfo *dinfo = _dinfo;
8844 + static struct lock_class_key aufs_di;
8846 + au_rw_init(&dinfo->di_rwsem);
8847 + au_rw_class(&dinfo->di_rwsem, &aufs_di);
8850 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
8852 + struct au_dinfo *dinfo;
8855 + dinfo = au_cache_alloc_dinfo();
8856 + if (unlikely(!dinfo))
8859 + nbr = au_sbend(sb) + 1;
8862 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
8863 + if (dinfo->di_hdentry) {
8864 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
8865 + dinfo->di_bstart = -1;
8866 + dinfo->di_bend = -1;
8867 + dinfo->di_bwh = -1;
8868 + dinfo->di_bdiropq = -1;
8869 + dinfo->di_tmpfile = 0;
8870 + for (i = 0; i < nbr; i++)
8871 + dinfo->di_hdentry[i].hd_id = -1;
8875 + au_cache_free_dinfo(dinfo);
8882 +void au_di_free(struct au_dinfo *dinfo)
8884 + struct au_hdentry *p;
8885 + aufs_bindex_t bend, bindex;
8887 + /* dentry may not be revalidated */
8888 + bindex = dinfo->di_bstart;
8889 + if (bindex >= 0) {
8890 + bend = dinfo->di_bend;
8891 + p = dinfo->di_hdentry + bindex;
8892 + while (bindex++ <= bend)
8895 + kfree(dinfo->di_hdentry);
8896 + au_cache_free_dinfo(dinfo);
8899 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
8901 + struct au_hdentry *p;
8904 + AuRwMustWriteLock(&a->di_rwsem);
8905 + AuRwMustWriteLock(&b->di_rwsem);
8907 +#define DiSwap(v, name) \
8909 + v = a->di_##name; \
8910 + a->di_##name = b->di_##name; \
8911 + b->di_##name = v; \
8914 + DiSwap(p, hdentry);
8915 + DiSwap(bi, bstart);
8918 + DiSwap(bi, bdiropq);
8924 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
8926 + AuRwMustWriteLock(&dst->di_rwsem);
8927 + AuRwMustWriteLock(&src->di_rwsem);
8929 + dst->di_bstart = src->di_bstart;
8930 + dst->di_bend = src->di_bend;
8931 + dst->di_bwh = src->di_bwh;
8932 + dst->di_bdiropq = src->di_bdiropq;
8936 +int au_di_init(struct dentry *dentry)
8939 + struct super_block *sb;
8940 + struct au_dinfo *dinfo;
8943 + sb = dentry->d_sb;
8944 + dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
8946 + atomic_set(&dinfo->di_generation, au_sigen(sb));
8947 + /* smp_mb(); */ /* atomic_set */
8948 + dentry->d_fsdata = dinfo;
8955 +void au_di_fin(struct dentry *dentry)
8957 + struct au_dinfo *dinfo;
8959 + dinfo = au_di(dentry);
8960 + AuRwDestroy(&dinfo->di_rwsem);
8961 + au_di_free(dinfo);
8964 +int au_di_realloc(struct au_dinfo *dinfo, int nbr)
8967 + struct au_hdentry *hdp;
8969 + AuRwMustWriteLock(&dinfo->di_rwsem);
8972 + sz = sizeof(*hdp) * (dinfo->di_bend + 1);
8974 + sz = sizeof(*hdp);
8975 + hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS);
8977 + dinfo->di_hdentry = hdp;
8984 +/* ---------------------------------------------------------------------- */
8986 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
8989 + case AuLsc_DI_CHILD:
8990 + ii_write_lock_child(inode);
8992 + case AuLsc_DI_CHILD2:
8993 + ii_write_lock_child2(inode);
8995 + case AuLsc_DI_CHILD3:
8996 + ii_write_lock_child3(inode);
8998 + case AuLsc_DI_PARENT:
8999 + ii_write_lock_parent(inode);
9001 + case AuLsc_DI_PARENT2:
9002 + ii_write_lock_parent2(inode);
9004 + case AuLsc_DI_PARENT3:
9005 + ii_write_lock_parent3(inode);
9012 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
9015 + case AuLsc_DI_CHILD:
9016 + ii_read_lock_child(inode);
9018 + case AuLsc_DI_CHILD2:
9019 + ii_read_lock_child2(inode);
9021 + case AuLsc_DI_CHILD3:
9022 + ii_read_lock_child3(inode);
9024 + case AuLsc_DI_PARENT:
9025 + ii_read_lock_parent(inode);
9027 + case AuLsc_DI_PARENT2:
9028 + ii_read_lock_parent2(inode);
9030 + case AuLsc_DI_PARENT3:
9031 + ii_read_lock_parent3(inode);
9038 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9040 + struct inode *inode;
9042 + au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
9043 + if (d_really_is_positive(d)) {
9044 + inode = d_inode(d);
9045 + if (au_ftest_lock(flags, IW))
9046 + do_ii_write_lock(inode, lsc);
9047 + else if (au_ftest_lock(flags, IR))
9048 + do_ii_read_lock(inode, lsc);
9052 +void di_read_unlock(struct dentry *d, int flags)
9054 + struct inode *inode;
9056 + if (d_really_is_positive(d)) {
9057 + inode = d_inode(d);
9058 + if (au_ftest_lock(flags, IW)) {
9059 + au_dbg_verify_dinode(d);
9060 + ii_write_unlock(inode);
9061 + } else if (au_ftest_lock(flags, IR)) {
9062 + au_dbg_verify_dinode(d);
9063 + ii_read_unlock(inode);
9066 + au_rw_read_unlock(&au_di(d)->di_rwsem);
9069 +void di_downgrade_lock(struct dentry *d, int flags)
9071 + if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9072 + ii_downgrade_lock(d_inode(d));
9073 + au_rw_dgrade_lock(&au_di(d)->di_rwsem);
9076 +void di_write_lock(struct dentry *d, unsigned int lsc)
9078 + au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
9079 + if (d_really_is_positive(d))
9080 + do_ii_write_lock(d_inode(d), lsc);
9083 +void di_write_unlock(struct dentry *d)
9085 + au_dbg_verify_dinode(d);
9086 + if (d_really_is_positive(d))
9087 + ii_write_unlock(d_inode(d));
9088 + au_rw_write_unlock(&au_di(d)->di_rwsem);
9091 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9093 + AuDebugOn(d1 == d2
9094 + || d_inode(d1) == d_inode(d2)
9095 + || d1->d_sb != d2->d_sb);
9097 + if (isdir && au_test_subdir(d1, d2)) {
9098 + di_write_lock_child(d1);
9099 + di_write_lock_child2(d2);
9101 + /* there should be no races */
9102 + di_write_lock_child(d2);
9103 + di_write_lock_child2(d1);
9107 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9109 + AuDebugOn(d1 == d2
9110 + || d_inode(d1) == d_inode(d2)
9111 + || d1->d_sb != d2->d_sb);
9113 + if (isdir && au_test_subdir(d1, d2)) {
9114 + di_write_lock_parent(d1);
9115 + di_write_lock_parent2(d2);
9117 + /* there should be no races */
9118 + di_write_lock_parent(d2);
9119 + di_write_lock_parent2(d1);
9123 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9125 + di_write_unlock(d1);
9126 + if (d_inode(d1) == d_inode(d2))
9127 + au_rw_write_unlock(&au_di(d2)->di_rwsem);
9129 + di_write_unlock(d2);
9132 +/* ---------------------------------------------------------------------- */
9134 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9138 + DiMustAnyLock(dentry);
9140 + if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
9142 + AuDebugOn(bindex < 0);
9143 + d = au_di(dentry)->di_hdentry[0 + bindex].hd_dentry;
9144 + AuDebugOn(d && au_dcount(d) <= 0);
9149 + * extended version of au_h_dptr().
9150 + * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9153 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9155 + struct dentry *h_dentry;
9156 + struct inode *inode, *h_inode;
9158 + AuDebugOn(d_really_is_negative(dentry));
9161 + if (au_dbstart(dentry) <= bindex
9162 + && bindex <= au_dbend(dentry))
9163 + h_dentry = au_h_dptr(dentry, bindex);
9164 + if (h_dentry && !au_d_linkable(h_dentry)) {
9166 + goto out; /* success */
9169 + inode = d_inode(dentry);
9170 + AuDebugOn(bindex < au_ibstart(inode));
9171 + AuDebugOn(au_ibend(inode) < bindex);
9172 + h_inode = au_h_iptr(inode, bindex);
9173 + h_dentry = d_find_alias(h_inode);
9175 + if (!IS_ERR(h_dentry)) {
9176 + if (!au_d_linkable(h_dentry))
9177 + goto out; /* success */
9183 + if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9184 + h_dentry = au_plink_lkup(inode, bindex);
9185 + AuDebugOn(!h_dentry);
9186 + if (!IS_ERR(h_dentry)) {
9187 + if (!au_d_hashed_positive(h_dentry))
9188 + goto out; /* success */
9195 + AuDbgDentry(h_dentry);
9199 +aufs_bindex_t au_dbtail(struct dentry *dentry)
9201 + aufs_bindex_t bend, bwh;
9203 + bend = au_dbend(dentry);
9205 + bwh = au_dbwh(dentry);
9208 + if (0 < bwh && bwh < bend)
9214 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9216 + aufs_bindex_t bend, bopq;
9218 + bend = au_dbtail(dentry);
9220 + bopq = au_dbdiropq(dentry);
9221 + if (0 <= bopq && bopq < bend)
9227 +/* ---------------------------------------------------------------------- */
9229 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9230 + struct dentry *h_dentry)
9232 + struct au_hdentry *hd = au_di(dentry)->di_hdentry + bindex;
9233 + struct au_branch *br;
9235 + DiMustWriteLock(dentry);
9238 + hd->hd_dentry = h_dentry;
9240 + br = au_sbr(dentry->d_sb, bindex);
9241 + hd->hd_id = br->br_id;
9245 +int au_dbrange_test(struct dentry *dentry)
9248 + aufs_bindex_t bstart, bend;
9251 + bstart = au_dbstart(dentry);
9252 + bend = au_dbend(dentry);
9254 + AuDebugOn(bend < 0 && bstart > bend);
9257 + AuDebugOn(bend >= 0);
9263 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
9268 + if (unlikely(au_digen(dentry) != sigen
9269 + || au_iigen_test(d_inode(dentry), sigen)))
9275 +void au_update_digen(struct dentry *dentry)
9277 + atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9278 + /* smp_mb(); */ /* atomic_set */
9281 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9283 + struct au_dinfo *dinfo;
9284 + struct dentry *h_d;
9285 + struct au_hdentry *hdp;
9287 + DiMustWriteLock(dentry);
9289 + dinfo = au_di(dentry);
9290 + if (!dinfo || dinfo->di_bstart < 0)
9293 + hdp = dinfo->di_hdentry;
9294 + if (do_put_zero) {
9295 + aufs_bindex_t bindex, bend;
9297 + bend = dinfo->di_bend;
9298 + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++) {
9299 + h_d = hdp[0 + bindex].hd_dentry;
9300 + if (h_d && d_is_negative(h_d))
9301 + au_set_h_dptr(dentry, bindex, NULL);
9305 + dinfo->di_bstart = -1;
9306 + while (++dinfo->di_bstart <= dinfo->di_bend)
9307 + if (hdp[0 + dinfo->di_bstart].hd_dentry)
9309 + if (dinfo->di_bstart > dinfo->di_bend) {
9310 + dinfo->di_bstart = -1;
9311 + dinfo->di_bend = -1;
9316 + while (0 <= --dinfo->di_bend)
9317 + if (hdp[0 + dinfo->di_bend].hd_dentry)
9319 + AuDebugOn(dinfo->di_bstart > dinfo->di_bend || dinfo->di_bend < 0);
9322 +void au_update_dbstart(struct dentry *dentry)
9324 + aufs_bindex_t bindex, bend;
9325 + struct dentry *h_dentry;
9327 + bend = au_dbend(dentry);
9328 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
9329 + h_dentry = au_h_dptr(dentry, bindex);
9332 + if (d_is_positive(h_dentry)) {
9333 + au_set_dbstart(dentry, bindex);
9336 + au_set_h_dptr(dentry, bindex, NULL);
9340 +void au_update_dbend(struct dentry *dentry)
9342 + aufs_bindex_t bindex, bstart;
9343 + struct dentry *h_dentry;
9345 + bstart = au_dbstart(dentry);
9346 + for (bindex = au_dbend(dentry); bindex >= bstart; bindex--) {
9347 + h_dentry = au_h_dptr(dentry, bindex);
9350 + if (d_is_positive(h_dentry)) {
9351 + au_set_dbend(dentry, bindex);
9354 + au_set_h_dptr(dentry, bindex, NULL);
9358 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9360 + aufs_bindex_t bindex, bend;
9362 + bend = au_dbend(dentry);
9363 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++)
9364 + if (au_h_dptr(dentry, bindex) == h_dentry)
9368 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9369 --- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
9370 +++ linux/fs/aufs/dir.c 2015-09-24 10:47:58.251386326 +0200
9373 + * Copyright (C) 2005-2015 Junjiro R. Okajima
9375 + * This program, aufs is free software; you can redistribute it and/or modify
9376 + * it under the terms of the GNU General Public License as published by
9377 + * the Free Software Foundation; either version 2 of the License, or
9378 + * (at your option) any later version.
9380 + * This program is distributed in the hope that it will be useful,
9381 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9382 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9383 + * GNU General Public License for more details.
9385 + * You should have received a copy of the GNU General Public License
9386 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
9390 + * directory operations
9393 +#include <linux/fs_stack.h>
9396 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
9398 + unsigned int nlink;
9400 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9402 + nlink = dir->i_nlink;
9403 + nlink += h_dir->i_nlink - 2;
9404 + if (h_dir->i_nlink < 2)
9406 + smp_mb(); /* for i_nlink */
9407 + /* 0 can happen in revaliding */
9408 + set_nlink(dir, nlink);
9411 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
9413 + unsigned int nlink;
9415 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9417 + nlink = dir->i_nlink;
9418 + nlink -= h_dir->i_nlink - 2;
9419 + if (h_dir->i_nlink < 2)
9421 + smp_mb(); /* for i_nlink */
9422 + /* nlink == 0 means the branch-fs is broken */
9423 + set_nlink(dir, nlink);
9426 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
9429 + aufs_bindex_t bindex, bend;
9430 + struct file *h_file;
9431 + struct dentry *h_dentry;
9435 + AuDebugOn(!d_is_dir(file->f_path.dentry));
9437 + bend = au_fbend_dir(file);
9438 + for (bindex = au_fbstart(file);
9439 + bindex <= bend && sz < KMALLOC_MAX_SIZE;
9441 + h_file = au_hf_dir(file, bindex);
9442 + if (h_file && file_inode(h_file))
9443 + sz += vfsub_f_size_read(h_file);
9446 + AuDebugOn(!dentry);
9447 + AuDebugOn(!d_is_dir(dentry));
9449 + bend = au_dbtaildir(dentry);
9450 + for (bindex = au_dbstart(dentry);
9451 + bindex <= bend && sz < KMALLOC_MAX_SIZE;
9453 + h_dentry = au_h_dptr(dentry, bindex);
9454 + if (h_dentry && d_is_positive(h_dentry))
9455 + sz += i_size_read(d_inode(h_dentry));
9458 + if (sz < KMALLOC_MAX_SIZE)
9459 + sz = roundup_pow_of_two(sz);
9460 + if (sz > KMALLOC_MAX_SIZE)
9461 + sz = KMALLOC_MAX_SIZE;
9462 + else if (sz < NAME_MAX) {
9463 + BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
9464 + sz = AUFS_RDBLK_DEF;
9469 +struct au_dir_ts_arg {
9470 + struct dentry *dentry;
9471 + aufs_bindex_t brid;
9474 +static void au_do_dir_ts(void *arg)
9476 + struct au_dir_ts_arg *a = arg;
9477 + struct au_dtime dt;
9478 + struct path h_path;
9479 + struct inode *dir, *h_dir;
9480 + struct super_block *sb;
9481 + struct au_branch *br;
9482 + struct au_hinode *hdir;
9484 + aufs_bindex_t bstart, bindex;
9486 + sb = a->dentry->d_sb;
9487 + if (d_really_is_negative(a->dentry))
9489 + aufs_read_lock(a->dentry, AuLock_DW | AuLock_DIR); /* noflush */
9491 + /* no dir->i_mutex lock */
9492 + dir = d_inode(a->dentry);
9493 + bstart = au_ibstart(dir);
9494 + bindex = au_br_index(sb, a->brid);
9495 + if (bindex < bstart)
9498 + br = au_sbr(sb, bindex);
9499 + h_path.dentry = au_h_dptr(a->dentry, bindex);
9500 + if (!h_path.dentry)
9502 + h_path.mnt = au_br_mnt(br);
9503 + au_dtime_store(&dt, a->dentry, &h_path);
9505 + br = au_sbr(sb, bstart);
9506 + if (!au_br_writable(br->br_perm))
9508 + h_path.dentry = au_h_dptr(a->dentry, bstart);
9509 + h_path.mnt = au_br_mnt(br);
9510 + err = vfsub_mnt_want_write(h_path.mnt);
9513 + hdir = au_hi(dir, bstart);
9514 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
9515 + h_dir = au_h_iptr(dir, bstart);
9516 + if (h_dir->i_nlink
9517 + && timespec_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
9518 + dt.dt_h_path = h_path;
9519 + au_dtime_revert(&dt);
9521 + au_hn_imtx_unlock(hdir);
9522 + vfsub_mnt_drop_write(h_path.mnt);
9523 + au_cpup_attr_timesizes(dir);
9526 + aufs_read_unlock(a->dentry, AuLock_DW);
9529 + au_nwt_done(&au_sbi(sb)->si_nowait);
9533 +void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
9535 + int perm, wkq_err;
9536 + aufs_bindex_t bstart;
9537 + struct au_dir_ts_arg *arg;
9538 + struct dentry *dentry;
9539 + struct super_block *sb;
9543 + dentry = d_find_any_alias(dir);
9544 + AuDebugOn(!dentry);
9545 + sb = dentry->d_sb;
9546 + bstart = au_ibstart(dir);
9547 + if (bstart == bindex) {
9548 + au_cpup_attr_timesizes(dir);
9552 + perm = au_sbr_perm(sb, bstart);
9553 + if (!au_br_writable(perm))
9556 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
9560 + arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
9561 + arg->brid = au_sbr_id(sb, bindex);
9562 + wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
9563 + if (unlikely(wkq_err)) {
9564 + pr_err("wkq %d\n", wkq_err);
9573 +/* ---------------------------------------------------------------------- */
9575 +static int reopen_dir(struct file *file)
9578 + unsigned int flags;
9579 + aufs_bindex_t bindex, btail, bstart;
9580 + struct dentry *dentry, *h_dentry;
9581 + struct file *h_file;
9583 + /* open all lower dirs */
9584 + dentry = file->f_path.dentry;
9585 + bstart = au_dbstart(dentry);
9586 + for (bindex = au_fbstart(file); bindex < bstart; bindex++)
9587 + au_set_h_fptr(file, bindex, NULL);
9588 + au_set_fbstart(file, bstart);
9590 + btail = au_dbtaildir(dentry);
9591 + for (bindex = au_fbend_dir(file); btail < bindex; bindex--)
9592 + au_set_h_fptr(file, bindex, NULL);
9593 + au_set_fbend_dir(file, btail);
9595 + flags = vfsub_file_flags(file);
9596 + for (bindex = bstart; bindex <= btail; bindex++) {
9597 + h_dentry = au_h_dptr(dentry, bindex);
9600 + h_file = au_hf_dir(file, bindex);
9604 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
9605 + err = PTR_ERR(h_file);
9606 + if (IS_ERR(h_file))
9607 + goto out; /* close all? */
9608 + au_set_h_fptr(file, bindex, h_file);
9610 + au_update_figen(file);
9611 + /* todo: necessary? */
9612 + /* file->f_ra = h_file->f_ra; */
9619 +static int do_open_dir(struct file *file, int flags, struct file *h_file)
9622 + aufs_bindex_t bindex, btail;
9623 + struct dentry *dentry, *h_dentry;
9625 + FiMustWriteLock(file);
9626 + AuDebugOn(h_file);
9629 + dentry = file->f_path.dentry;
9630 + file->f_version = d_inode(dentry)->i_version;
9631 + bindex = au_dbstart(dentry);
9632 + au_set_fbstart(file, bindex);
9633 + btail = au_dbtaildir(dentry);
9634 + au_set_fbend_dir(file, btail);
9635 + for (; !err && bindex <= btail; bindex++) {
9636 + h_dentry = au_h_dptr(dentry, bindex);
9640 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
9641 + if (IS_ERR(h_file)) {
9642 + err = PTR_ERR(h_file);
9645 + au_set_h_fptr(file, bindex, h_file);
9647 + au_update_figen(file);
9648 + /* todo: necessary? */
9649 + /* file->f_ra = h_file->f_ra; */
9651 + return 0; /* success */
9654 + for (bindex = au_fbstart(file); bindex <= btail; bindex++)
9655 + au_set_h_fptr(file, bindex, NULL);
9656 + au_set_fbstart(file, -1);
9657 + au_set_fbend_dir(file, -1);
9662 +static int aufs_open_dir(struct inode *inode __maybe_unused,
9663 + struct file *file)
9666 + struct super_block *sb;
9667 + struct au_fidir *fidir;
9670 + sb = file->f_path.dentry->d_sb;
9671 + si_read_lock(sb, AuLock_FLUSH);
9672 + fidir = au_fidir_alloc(sb);
9674 + struct au_do_open_args args = {
9675 + .open = do_open_dir,
9678 + err = au_do_open(file, &args);
9679 + if (unlikely(err))
9682 + si_read_unlock(sb);
9686 +static int aufs_release_dir(struct inode *inode __maybe_unused,
9687 + struct file *file)
9689 + struct au_vdir *vdir_cache;
9690 + struct au_finfo *finfo;
9691 + struct au_fidir *fidir;
9692 + aufs_bindex_t bindex, bend;
9694 + finfo = au_fi(file);
9695 + fidir = finfo->fi_hdir;
9697 + au_sphl_del(&finfo->fi_hlist,
9698 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
9699 + vdir_cache = fidir->fd_vdir_cache; /* lock-free */
9701 + au_vdir_free(vdir_cache);
9703 + bindex = finfo->fi_btop;
9704 + if (bindex >= 0) {
9706 + * calls fput() instead of filp_close(),
9707 + * since no dnotify or lock for the lower file.
9709 + bend = fidir->fd_bbot;
9710 + for (; bindex <= bend; bindex++)
9711 + au_set_h_fptr(file, bindex, NULL);
9714 + finfo->fi_hdir = NULL;
9716 + au_finfo_fin(file);
9720 +/* ---------------------------------------------------------------------- */
9722 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
9725 + aufs_bindex_t bindex, bend;
9726 + struct file *h_file;
9729 + bend = au_fbend_dir(file);
9730 + for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
9731 + h_file = au_hf_dir(file, bindex);
9733 + err = vfsub_flush(h_file, id);
9738 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
9740 + return au_do_flush(file, id, au_do_flush_dir);
9743 +/* ---------------------------------------------------------------------- */
9745 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
9748 + aufs_bindex_t bend, bindex;
9749 + struct inode *inode;
9750 + struct super_block *sb;
9753 + sb = dentry->d_sb;
9754 + inode = d_inode(dentry);
9756 + bend = au_dbend(dentry);
9757 + for (bindex = au_dbstart(dentry); !err && bindex <= bend; bindex++) {
9758 + struct path h_path;
9760 + if (au_test_ro(sb, bindex, inode))
9762 + h_path.dentry = au_h_dptr(dentry, bindex);
9763 + if (!h_path.dentry)
9766 + h_path.mnt = au_sbr_mnt(sb, bindex);
9767 + err = vfsub_fsync(NULL, &h_path, datasync);
9773 +static int au_do_fsync_dir(struct file *file, int datasync)
9776 + aufs_bindex_t bend, bindex;
9777 + struct file *h_file;
9778 + struct super_block *sb;
9779 + struct inode *inode;
9781 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9782 + if (unlikely(err))
9785 + inode = file_inode(file);
9787 + bend = au_fbend_dir(file);
9788 + for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
9789 + h_file = au_hf_dir(file, bindex);
9790 + if (!h_file || au_test_ro(sb, bindex, inode))
9793 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
9801 + * @file may be NULL
9803 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
9807 + struct dentry *dentry;
9808 + struct inode *inode;
9809 + struct super_block *sb;
9810 + struct mutex *mtx;
9813 + dentry = file->f_path.dentry;
9814 + inode = d_inode(dentry);
9815 + mtx = &inode->i_mutex;
9817 + sb = dentry->d_sb;
9818 + si_noflush_read_lock(sb);
9820 + err = au_do_fsync_dir(file, datasync);
9822 + di_write_lock_child(dentry);
9823 + err = au_do_fsync_dir_no_file(dentry, datasync);
9825 + au_cpup_attr_timesizes(inode);
9826 + di_write_unlock(dentry);
9828 + fi_write_unlock(file);
9830 + si_read_unlock(sb);
9831 + mutex_unlock(mtx);
9835 +/* ---------------------------------------------------------------------- */
9837 +static int aufs_iterate(struct file *file, struct dir_context *ctx)
9840 + struct dentry *dentry;
9841 + struct inode *inode, *h_inode;
9842 + struct super_block *sb;
9844 + AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
9846 + dentry = file->f_path.dentry;
9847 + inode = d_inode(dentry);
9850 + sb = dentry->d_sb;
9851 + si_read_lock(sb, AuLock_FLUSH);
9852 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9853 + if (unlikely(err))
9855 + err = au_alive_dir(dentry);
9857 + err = au_vdir_init(file);
9858 + di_downgrade_lock(dentry, AuLock_IR);
9859 + if (unlikely(err))
9862 + h_inode = au_h_iptr(inode, au_ibstart(inode));
9863 + if (!au_test_nfsd()) {
9864 + err = au_vdir_fill_de(file, ctx);
9865 + fsstack_copy_attr_atime(inode, h_inode);
9868 + * nfsd filldir may call lookup_one_len(), vfs_getattr(),
9869 + * encode_fh() and others.
9871 + atomic_inc(&h_inode->i_count);
9872 + di_read_unlock(dentry, AuLock_IR);
9873 + si_read_unlock(sb);
9874 + err = au_vdir_fill_de(file, ctx);
9875 + fsstack_copy_attr_atime(inode, h_inode);
9876 + fi_write_unlock(file);
9884 + di_read_unlock(dentry, AuLock_IR);
9885 + fi_write_unlock(file);
9887 + si_read_unlock(sb);
9891 +/* ---------------------------------------------------------------------- */
9893 +#define AuTestEmpty_WHONLY 1
9894 +#define AuTestEmpty_CALLED (1 << 1)
9895 +#define AuTestEmpty_SHWH (1 << 2)
9896 +#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
9897 +#define au_fset_testempty(flags, name) \
9898 + do { (flags) |= AuTestEmpty_##name; } while (0)
9899 +#define au_fclr_testempty(flags, name) \
9900 + do { (flags) &= ~AuTestEmpty_##name; } while (0)
9902 +#ifndef CONFIG_AUFS_SHWH
9903 +#undef AuTestEmpty_SHWH
9904 +#define AuTestEmpty_SHWH 0
9907 +struct test_empty_arg {
9908 + struct dir_context ctx;
9909 + struct au_nhash *whlist;
9910 + unsigned int flags;
9912 + aufs_bindex_t bindex;
9915 +static int test_empty_cb(struct dir_context *ctx, const char *__name,
9916 + int namelen, loff_t offset __maybe_unused, u64 ino,
9917 + unsigned int d_type)
9919 + struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
9921 + char *name = (void *)__name;
9924 + au_fset_testempty(arg->flags, CALLED);
9926 + if (name[0] == '.'
9927 + && (namelen == 1 || (name[1] == '.' && namelen == 2)))
9928 + goto out; /* success */
9930 + if (namelen <= AUFS_WH_PFX_LEN
9931 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
9932 + if (au_ftest_testempty(arg->flags, WHONLY)
9933 + && !au_nhash_test_known_wh(arg->whlist, name, namelen))
9934 + arg->err = -ENOTEMPTY;
9938 + name += AUFS_WH_PFX_LEN;
9939 + namelen -= AUFS_WH_PFX_LEN;
9940 + if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
9941 + arg->err = au_nhash_append_wh
9942 + (arg->whlist, name, namelen, ino, d_type, arg->bindex,
9943 + au_ftest_testempty(arg->flags, SHWH));
9947 + AuTraceErr(arg->err);
9951 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9954 + struct file *h_file;
9956 + h_file = au_h_open(dentry, arg->bindex,
9957 + O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
9958 + /*file*/NULL, /*force_wr*/0);
9959 + err = PTR_ERR(h_file);
9960 + if (IS_ERR(h_file))
9964 + if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
9965 + && !file_inode(h_file)->i_nlink)
9970 + au_fclr_testempty(arg->flags, CALLED);
9972 + err = vfsub_iterate_dir(h_file, &arg->ctx);
9975 + } while (!err && au_ftest_testempty(arg->flags, CALLED));
9979 + au_sbr_put(dentry->d_sb, arg->bindex);
9984 +struct do_test_empty_args {
9986 + struct dentry *dentry;
9987 + struct test_empty_arg *arg;
9990 +static void call_do_test_empty(void *args)
9992 + struct do_test_empty_args *a = args;
9993 + *a->errp = do_test_empty(a->dentry, a->arg);
9996 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9999 + struct dentry *h_dentry;
10000 + struct inode *h_inode;
10002 + h_dentry = au_h_dptr(dentry, arg->bindex);
10003 + h_inode = d_inode(h_dentry);
10004 + /* todo: i_mode changes anytime? */
10005 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
10006 + err = au_test_h_perm_sio(h_inode, MAY_EXEC | MAY_READ);
10007 + mutex_unlock(&h_inode->i_mutex);
10009 + err = do_test_empty(dentry, arg);
10011 + struct do_test_empty_args args = {
10013 + .dentry = dentry,
10016 + unsigned int flags = arg->flags;
10018 + wkq_err = au_wkq_wait(call_do_test_empty, &args);
10019 + if (unlikely(wkq_err))
10021 + arg->flags = flags;
10027 +int au_test_empty_lower(struct dentry *dentry)
10030 + unsigned int rdhash;
10031 + aufs_bindex_t bindex, bstart, btail;
10032 + struct au_nhash whlist;
10033 + struct test_empty_arg arg = {
10035 + .actor = test_empty_cb
10038 + int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
10040 + SiMustAnyLock(dentry->d_sb);
10042 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10044 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10045 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
10046 + if (unlikely(err))
10050 + arg.whlist = &whlist;
10051 + bstart = au_dbstart(dentry);
10052 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10053 + au_fset_testempty(arg.flags, SHWH);
10054 + test_empty = do_test_empty;
10055 + if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10056 + test_empty = sio_test_empty;
10057 + arg.bindex = bstart;
10058 + err = test_empty(dentry, &arg);
10059 + if (unlikely(err))
10062 + au_fset_testempty(arg.flags, WHONLY);
10063 + btail = au_dbtaildir(dentry);
10064 + for (bindex = bstart + 1; !err && bindex <= btail; bindex++) {
10065 + struct dentry *h_dentry;
10067 + h_dentry = au_h_dptr(dentry, bindex);
10068 + if (h_dentry && d_is_positive(h_dentry)) {
10069 + arg.bindex = bindex;
10070 + err = test_empty(dentry, &arg);
10075 + au_nhash_wh_free(&whlist);
10080 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10083 + struct test_empty_arg arg = {
10085 + .actor = test_empty_cb
10088 + aufs_bindex_t bindex, btail;
10091 + arg.whlist = whlist;
10092 + arg.flags = AuTestEmpty_WHONLY;
10093 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10094 + au_fset_testempty(arg.flags, SHWH);
10095 + btail = au_dbtaildir(dentry);
10096 + for (bindex = au_dbstart(dentry); !err && bindex <= btail; bindex++) {
10097 + struct dentry *h_dentry;
10099 + h_dentry = au_h_dptr(dentry, bindex);
10100 + if (h_dentry && d_is_positive(h_dentry)) {
10101 + arg.bindex = bindex;
10102 + err = sio_test_empty(dentry, &arg);
10109 +/* ---------------------------------------------------------------------- */
10111 +const struct file_operations aufs_dir_fop = {
10112 + .owner = THIS_MODULE,
10113 + .llseek = default_llseek,
10114 + .read = generic_read_dir,
10115 + .iterate = aufs_iterate,
10116 + .unlocked_ioctl = aufs_ioctl_dir,
10117 +#ifdef CONFIG_COMPAT
10118 + .compat_ioctl = aufs_compat_ioctl_dir,
10120 + .open = aufs_open_dir,
10121 + .release = aufs_release_dir,
10122 + .flush = aufs_flush_dir,
10123 + .fsync = aufs_fsync_dir
10125 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10126 --- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
10127 +++ linux/fs/aufs/dir.h 2015-09-24 10:47:58.251386326 +0200
10130 + * Copyright (C) 2005-2015 Junjiro R. Okajima
10132 + * This program, aufs is free software; you can redistribute it and/or modify
10133 + * it under the terms of the GNU General Public License as published by
10134 + * the Free Software Foundation; either version 2 of the License, or
10135 + * (at your option) any later version.
10137 + * This program is distributed in the hope that it will be useful,
10138 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10139 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10140 + * GNU General Public License for more details.
10142 + * You should have received a copy of the GNU General Public License
10143 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10147 + * directory operations
10150 +#ifndef __AUFS_DIR_H__
10151 +#define __AUFS_DIR_H__
10155 +#include <linux/fs.h>
10157 +/* ---------------------------------------------------------------------- */
10159 +/* need to be faster and smaller */
10162 + unsigned int nh_num;
10163 + struct hlist_head *nh_head;
10166 +struct au_vdir_destr {
10167 + unsigned char len;
10168 + unsigned char name[0];
10171 +struct au_vdir_dehstr {
10172 + struct hlist_node hash;
10173 + struct au_vdir_destr *str;
10174 +} ____cacheline_aligned_in_smp;
10176 +struct au_vdir_de {
10178 + unsigned char de_type;
10179 + /* caution: packed */
10180 + struct au_vdir_destr de_str;
10183 +struct au_vdir_wh {
10184 + struct hlist_node wh_hash;
10185 +#ifdef CONFIG_AUFS_SHWH
10187 + aufs_bindex_t wh_bindex;
10188 + unsigned char wh_type;
10190 + aufs_bindex_t wh_bindex;
10192 + /* caution: packed */
10193 + struct au_vdir_destr wh_str;
10196 +union au_vdir_deblk_p {
10197 + unsigned char *deblk;
10198 + struct au_vdir_de *de;
10202 + unsigned char **vd_deblk;
10203 + unsigned long vd_nblk;
10205 + unsigned long ul;
10206 + union au_vdir_deblk_p p;
10209 + unsigned long vd_version;
10210 + unsigned int vd_deblk_sz;
10211 + unsigned long vd_jiffy;
10212 +} ____cacheline_aligned_in_smp;
10214 +/* ---------------------------------------------------------------------- */
10217 +extern const struct file_operations aufs_dir_fop;
10218 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
10219 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
10220 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
10221 +void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
10222 +int au_test_empty_lower(struct dentry *dentry);
10223 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10226 +unsigned int au_rdhash_est(loff_t sz);
10227 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10228 +void au_nhash_wh_free(struct au_nhash *whlist);
10229 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10231 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10232 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10233 + unsigned int d_type, aufs_bindex_t bindex,
10234 + unsigned char shwh);
10235 +void au_vdir_free(struct au_vdir *vdir);
10236 +int au_vdir_init(struct file *file);
10237 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
10240 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10242 +#ifdef CONFIG_AUFS_RDU
10244 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
10245 +#ifdef CONFIG_COMPAT
10246 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10247 + unsigned long arg);
10250 +AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10251 + unsigned int cmd, unsigned long arg)
10252 +#ifdef CONFIG_COMPAT
10253 +AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10254 + unsigned int cmd, unsigned long arg)
10258 +#endif /* __KERNEL__ */
10259 +#endif /* __AUFS_DIR_H__ */
10260 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
10261 --- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
10262 +++ linux/fs/aufs/dynop.c 2015-09-24 10:47:58.251386326 +0200
10265 + * Copyright (C) 2010-2015 Junjiro R. Okajima
10267 + * This program, aufs is free software; you can redistribute it and/or modify
10268 + * it under the terms of the GNU General Public License as published by
10269 + * the Free Software Foundation; either version 2 of the License, or
10270 + * (at your option) any later version.
10272 + * This program is distributed in the hope that it will be useful,
10273 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10274 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10275 + * GNU General Public License for more details.
10277 + * You should have received a copy of the GNU General Public License
10278 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10282 + * dynamically customizable operations for regular files
10287 +#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
10290 + * How large will these lists be?
10291 + * Usually just a few elements, 20-30 at most for each, I guess.
10293 +static struct au_splhead dynop[AuDyLast];
10295 +static struct au_dykey *dy_gfind_get(struct au_splhead *spl, const void *h_op)
10297 + struct au_dykey *key, *tmp;
10298 + struct list_head *head;
10301 + head = &spl->head;
10303 + list_for_each_entry_rcu(tmp, head, dk_list)
10304 + if (tmp->dk_op.dy_hop == h_op) {
10306 + kref_get(&key->dk_kref);
10309 + rcu_read_unlock();
10314 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
10316 + struct au_dykey **k, *found;
10317 + const void *h_op = key->dk_op.dy_hop;
10321 + k = br->br_dykey;
10322 + for (i = 0; i < AuBrDynOp; i++)
10324 + if (k[i]->dk_op.dy_hop == h_op) {
10331 + spin_lock(&br->br_dykey_lock);
10332 + for (; i < AuBrDynOp; i++)
10334 + if (k[i]->dk_op.dy_hop == h_op) {
10342 + spin_unlock(&br->br_dykey_lock);
10343 + BUG_ON(i == AuBrDynOp); /* expand the array */
10349 +/* kref_get() if @key is already added */
10350 +static struct au_dykey *dy_gadd(struct au_splhead *spl, struct au_dykey *key)
10352 + struct au_dykey *tmp, *found;
10353 + struct list_head *head;
10354 + const void *h_op = key->dk_op.dy_hop;
10357 + head = &spl->head;
10358 + spin_lock(&spl->spin);
10359 + list_for_each_entry(tmp, head, dk_list)
10360 + if (tmp->dk_op.dy_hop == h_op) {
10361 + kref_get(&tmp->dk_kref);
10366 + list_add_rcu(&key->dk_list, head);
10367 + spin_unlock(&spl->spin);
10374 +static void dy_free_rcu(struct rcu_head *rcu)
10376 + struct au_dykey *key;
10378 + key = container_of(rcu, struct au_dykey, dk_rcu);
10383 +static void dy_free(struct kref *kref)
10385 + struct au_dykey *key;
10386 + struct au_splhead *spl;
10388 + key = container_of(kref, struct au_dykey, dk_kref);
10389 + spl = dynop + key->dk_op.dy_type;
10390 + au_spl_del_rcu(&key->dk_list, spl);
10391 + call_rcu(&key->dk_rcu, dy_free_rcu);
10394 +void au_dy_put(struct au_dykey *key)
10396 + kref_put(&key->dk_kref, dy_free);
10399 +/* ---------------------------------------------------------------------- */
10401 +#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
10403 +#ifdef CONFIG_AUFS_DEBUG
10404 +#define DyDbgDeclare(cnt) unsigned int cnt = 0
10405 +#define DyDbgInc(cnt) do { cnt++; } while (0)
10407 +#define DyDbgDeclare(cnt) do {} while (0)
10408 +#define DyDbgInc(cnt) do {} while (0)
10411 +#define DySet(func, dst, src, h_op, h_sb) do { \
10413 + if (h_op->func) { \
10415 + dst.func = src.func; \
10417 + AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
10421 +#define DySetForce(func, dst, src) do { \
10422 + AuDebugOn(!src.func); \
10424 + dst.func = src.func; \
10427 +#define DySetAop(func) \
10428 + DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
10429 +#define DySetAopForce(func) \
10430 + DySetForce(func, dyaop->da_op, aufs_aop)
10432 +static void dy_aop(struct au_dykey *key, const void *h_op,
10433 + struct super_block *h_sb __maybe_unused)
10435 + struct au_dyaop *dyaop = (void *)key;
10436 + const struct address_space_operations *h_aop = h_op;
10437 + DyDbgDeclare(cnt);
10439 + AuDbg("%s\n", au_sbtype(h_sb));
10441 + DySetAop(writepage);
10442 + DySetAopForce(readpage); /* force */
10443 + DySetAop(writepages);
10444 + DySetAop(set_page_dirty);
10445 + DySetAop(readpages);
10446 + DySetAop(write_begin);
10447 + DySetAop(write_end);
10449 + DySetAop(invalidatepage);
10450 + DySetAop(releasepage);
10451 + DySetAop(freepage);
10452 + /* this one will be changed according to an aufs mount option */
10453 + DySetAop(direct_IO);
10454 + DySetAop(migratepage);
10455 + DySetAop(launder_page);
10456 + DySetAop(is_partially_uptodate);
10457 + DySetAop(is_dirty_writeback);
10458 + DySetAop(error_remove_page);
10459 + DySetAop(swap_activate);
10460 + DySetAop(swap_deactivate);
10462 + DyDbgSize(cnt, *h_aop);
10465 +/* ---------------------------------------------------------------------- */
10467 +static void dy_bug(struct kref *kref)
10472 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
10474 + struct au_dykey *key, *old;
10475 + struct au_splhead *spl;
10478 + void (*set)(struct au_dykey *key, const void *h_op,
10479 + struct super_block *h_sb __maybe_unused);
10481 + static const struct op a[] = {
10483 + .sz = sizeof(struct au_dyaop),
10487 + const struct op *p;
10489 + spl = dynop + op->dy_type;
10490 + key = dy_gfind_get(spl, op->dy_hop);
10492 + goto out_add; /* success */
10494 + p = a + op->dy_type;
10495 + key = kzalloc(p->sz, GFP_NOFS);
10496 + if (unlikely(!key)) {
10497 + key = ERR_PTR(-ENOMEM);
10501 + key->dk_op.dy_hop = op->dy_hop;
10502 + kref_init(&key->dk_kref);
10503 + p->set(key, op->dy_hop, au_br_sb(br));
10504 + old = dy_gadd(spl, key);
10511 + old = dy_bradd(br, key);
10513 + /* its ref-count should never be zero here */
10514 + kref_put(&key->dk_kref, dy_bug);
10519 +/* ---------------------------------------------------------------------- */
10521 + * Aufs prohibits O_DIRECT by defaut even if the branch supports it.
10522 + * This behaviour is necessary to return an error from open(O_DIRECT) instead
10523 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
10524 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
10525 + * See the aufs manual in detail.
10527 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
10530 + dyaop->da_op.direct_IO = NULL;
10532 + dyaop->da_op.direct_IO = aufs_aop.direct_IO;
10535 +static struct au_dyaop *dy_aget(struct au_branch *br,
10536 + const struct address_space_operations *h_aop,
10539 + struct au_dyaop *dyaop;
10540 + struct au_dynop op;
10542 + op.dy_type = AuDy_AOP;
10543 + op.dy_haop = h_aop;
10544 + dyaop = (void *)dy_get(&op, br);
10545 + if (IS_ERR(dyaop))
10547 + dy_adx(dyaop, do_dx);
10553 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
10554 + struct inode *h_inode)
10557 + struct super_block *sb;
10558 + struct au_branch *br;
10559 + struct au_dyaop *dyaop;
10561 + AuDebugOn(!S_ISREG(h_inode->i_mode));
10562 + IiMustWriteLock(inode);
10564 + sb = inode->i_sb;
10565 + br = au_sbr(sb, bindex);
10566 + do_dx = !!au_opt_test(au_mntflags(sb), DIO);
10567 + dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
10568 + err = PTR_ERR(dyaop);
10569 + if (IS_ERR(dyaop))
10570 + /* unnecessary to call dy_fput() */
10574 + inode->i_mapping->a_ops = &dyaop->da_op;
10581 + * Is it safe to replace a_ops during the inode/file is in operation?
10582 + * Yes, I hope so.
10584 +int au_dy_irefresh(struct inode *inode)
10587 + aufs_bindex_t bstart;
10588 + struct inode *h_inode;
10591 + if (S_ISREG(inode->i_mode)) {
10592 + bstart = au_ibstart(inode);
10593 + h_inode = au_h_iptr(inode, bstart);
10594 + err = au_dy_iaop(inode, bstart, h_inode);
10599 +void au_dy_arefresh(int do_dx)
10601 + struct au_splhead *spl;
10602 + struct list_head *head;
10603 + struct au_dykey *key;
10605 + spl = dynop + AuDy_AOP;
10606 + head = &spl->head;
10607 + spin_lock(&spl->spin);
10608 + list_for_each_entry(key, head, dk_list)
10609 + dy_adx((void *)key, do_dx);
10610 + spin_unlock(&spl->spin);
10613 +/* ---------------------------------------------------------------------- */
10615 +void __init au_dy_init(void)
10619 + /* make sure that 'struct au_dykey *' can be any type */
10620 + BUILD_BUG_ON(offsetof(struct au_dyaop, da_key));
10622 + for (i = 0; i < AuDyLast; i++)
10623 + au_spl_init(dynop + i);
10626 +void au_dy_fin(void)
10630 + for (i = 0; i < AuDyLast; i++)
10631 + WARN_ON(!list_empty(&dynop[i].head));
10633 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
10634 --- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
10635 +++ linux/fs/aufs/dynop.h 2015-09-24 10:47:58.251386326 +0200
10638 + * Copyright (C) 2010-2015 Junjiro R. Okajima
10640 + * This program, aufs is free software; you can redistribute it and/or modify
10641 + * it under the terms of the GNU General Public License as published by
10642 + * the Free Software Foundation; either version 2 of the License, or
10643 + * (at your option) any later version.
10645 + * This program is distributed in the hope that it will be useful,
10646 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10647 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10648 + * GNU General Public License for more details.
10650 + * You should have received a copy of the GNU General Public License
10651 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10655 + * dynamically customizable operations (for regular files only)
10658 +#ifndef __AUFS_DYNOP_H__
10659 +#define __AUFS_DYNOP_H__
10663 +#include <linux/fs.h>
10664 +#include <linux/kref.h>
10666 +enum {AuDy_AOP, AuDyLast};
10671 + const void *dy_hop;
10672 + const struct address_space_operations *dy_haop;
10678 + struct list_head dk_list;
10679 + struct rcu_head dk_rcu;
10681 + struct au_dynop dk_op;
10684 + * during I am in the branch local array, kref is gotten. when the
10685 + * branch is removed, kref is put.
10687 + struct kref dk_kref;
10690 +/* stop unioning since their sizes are very different from each other */
10692 + struct au_dykey da_key;
10693 + struct address_space_operations da_op; /* not const */
10696 +/* ---------------------------------------------------------------------- */
10700 +void au_dy_put(struct au_dykey *key);
10701 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
10702 + struct inode *h_inode);
10703 +int au_dy_irefresh(struct inode *inode);
10704 +void au_dy_arefresh(int do_dio);
10706 +void __init au_dy_init(void);
10707 +void au_dy_fin(void);
10709 +#endif /* __KERNEL__ */
10710 +#endif /* __AUFS_DYNOP_H__ */
10711 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
10712 --- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
10713 +++ linux/fs/aufs/export.c 2015-09-24 10:47:58.251386326 +0200
10716 + * Copyright (C) 2005-2015 Junjiro R. Okajima
10718 + * This program, aufs is free software; you can redistribute it and/or modify
10719 + * it under the terms of the GNU General Public License as published by
10720 + * the Free Software Foundation; either version 2 of the License, or
10721 + * (at your option) any later version.
10723 + * This program is distributed in the hope that it will be useful,
10724 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10725 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10726 + * GNU General Public License for more details.
10728 + * You should have received a copy of the GNU General Public License
10729 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
10736 +#include <linux/exportfs.h>
10737 +#include <linux/fs_struct.h>
10738 +#include <linux/namei.h>
10739 +#include <linux/nsproxy.h>
10740 +#include <linux/random.h>
10741 +#include <linux/writeback.h>
10742 +#include "../fs/mount.h"
10746 +#ifdef CONFIG_AUFS_INO_T_64
10754 +static ino_t decode_ino(__u32 *a)
10758 + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
10760 +#ifdef CONFIG_AUFS_INO_T_64
10766 +static void encode_ino(__u32 *a, ino_t ino)
10772 +#ifdef CONFIG_AUFS_INO_T_64
10777 +/* NFS file handle */
10781 +#ifdef CONFIG_AUFS_INO_T_64
10782 + /* support 64bit inode number */
10795 + Fh_ino = Fh_ino1,
10796 + Fh_dir_ino = Fh_dir_ino1
10799 +static int au_test_anon(struct dentry *dentry)
10801 + /* note: read d_flags without d_lock */
10802 + return !!(dentry->d_flags & DCACHE_DISCONNECTED);
10805 +int au_test_nfsd(void)
10808 + struct task_struct *tsk = current;
10809 + char comm[sizeof(tsk->comm)];
10812 + if (tsk->flags & PF_KTHREAD) {
10813 + get_task_comm(comm, tsk);
10814 + ret = !strcmp(comm, "nfsd");
10820 +/* ---------------------------------------------------------------------- */
10821 +/* inode generation external table */
10823 +void au_xigen_inc(struct inode *inode)
10828 + struct super_block *sb;
10829 + struct au_sbinfo *sbinfo;
10831 + sb = inode->i_sb;
10832 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
10834 + sbinfo = au_sbi(sb);
10835 + pos = inode->i_ino;
10836 + pos *= sizeof(igen);
10837 + igen = inode->i_generation + 1;
10838 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xigen, &igen,
10839 + sizeof(igen), &pos);
10840 + if (sz == sizeof(igen))
10841 + return; /* success */
10843 + if (unlikely(sz >= 0))
10844 + AuIOErr("xigen error (%zd)\n", sz);
10847 +int au_xigen_new(struct inode *inode)
10852 + struct super_block *sb;
10853 + struct au_sbinfo *sbinfo;
10854 + struct file *file;
10857 + /* todo: dirty, at mount time */
10858 + if (inode->i_ino == AUFS_ROOT_INO)
10860 + sb = inode->i_sb;
10861 + SiMustAnyLock(sb);
10862 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
10866 + pos = inode->i_ino;
10867 + if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
10868 + AuIOErr1("too large i%lld\n", pos);
10871 + pos *= sizeof(inode->i_generation);
10874 + sbinfo = au_sbi(sb);
10875 + file = sbinfo->si_xigen;
10878 + if (vfsub_f_size_read(file)
10879 + < pos + sizeof(inode->i_generation)) {
10880 + inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
10881 + sz = xino_fwrite(sbinfo->si_xwrite, file, &inode->i_generation,
10882 + sizeof(inode->i_generation), &pos);
10884 + sz = xino_fread(sbinfo->si_xread, file, &inode->i_generation,
10885 + sizeof(inode->i_generation), &pos);
10886 + if (sz == sizeof(inode->i_generation))
10887 + goto out; /* success */
10890 + if (unlikely(sz >= 0)) {
10892 + AuIOErr("xigen error (%zd)\n", sz);
10899 +int au_xigen_set(struct super_block *sb, struct file *base)
10902 + struct au_sbinfo *sbinfo;
10903 + struct file *file;
10905 + SiMustWriteLock(sb);
10907 + sbinfo = au_sbi(sb);
10908 + file = au_xino_create2(base, sbinfo->si_xigen);
10909 + err = PTR_ERR(file);
10910 + if (IS_ERR(file))
10913 + if (sbinfo->si_xigen)
10914 + fput(sbinfo->si_xigen);
10915 + sbinfo->si_xigen = file;
10921 +void au_xigen_clr(struct super_block *sb)
10923 + struct au_sbinfo *sbinfo;
10925 + SiMustWriteLock(sb);
10927 + sbinfo = au_sbi(sb);
10928 + if (sbinfo->si_xigen) {
10929 + fput(sbinfo->si_xigen);
10930 + sbinfo->si_xigen = NULL;
10934 +/* ---------------------------------------------------------------------- */
10936 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
10939 + struct dentry *dentry, *d;
10940 + struct inode *inode;
10941 + unsigned int sigen;
10944 + inode = ilookup(sb, ino);
10948 + dentry = ERR_PTR(-ESTALE);
10949 + sigen = au_sigen(sb);
10950 + if (unlikely(is_bad_inode(inode)
10951 + || IS_DEADDIR(inode)
10952 + || sigen != au_iigen(inode, NULL)))
10956 + if (!dir_ino || S_ISDIR(inode->i_mode))
10957 + dentry = d_find_alias(inode);
10959 + spin_lock(&inode->i_lock);
10960 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
10961 + spin_lock(&d->d_lock);
10962 + if (!au_test_anon(d)
10963 + && d_inode(d->d_parent)->i_ino == dir_ino) {
10964 + dentry = dget_dlock(d);
10965 + spin_unlock(&d->d_lock);
10968 + spin_unlock(&d->d_lock);
10970 + spin_unlock(&inode->i_lock);
10972 + if (unlikely(dentry && au_digen_test(dentry, sigen))) {
10973 + /* need to refresh */
10981 + AuTraceErrPtr(dentry);
10985 +/* ---------------------------------------------------------------------- */
10987 +/* todo: dirty? */
10988 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
10990 +struct au_compare_mnt_args {
10992 + struct super_block *sb;
10995 + struct vfsmount *mnt;
10998 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
11000 + struct au_compare_mnt_args *a = arg;
11002 + if (mnt->mnt_sb != a->sb)
11004 + a->mnt = mntget(mnt);
11008 +static struct vfsmount *au_mnt_get(struct super_block *sb)
11011 + struct path root;
11012 + struct au_compare_mnt_args args = {
11016 + get_fs_root(current->fs, &root);
11018 + err = iterate_mounts(au_compare_mnt, &args, root.mnt);
11019 + rcu_read_unlock();
11022 + AuDebugOn(!args.mnt);
11026 +struct au_nfsd_si_lock {
11027 + unsigned int sigen;
11028 + aufs_bindex_t bindex, br_id;
11029 + unsigned char force_lock;
11032 +static int si_nfsd_read_lock(struct super_block *sb,
11033 + struct au_nfsd_si_lock *nsi_lock)
11036 + aufs_bindex_t bindex;
11038 + si_read_lock(sb, AuLock_FLUSH);
11040 + /* branch id may be wrapped around */
11042 + bindex = au_br_index(sb, nsi_lock->br_id);
11043 + if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
11044 + goto out; /* success */
11048 + if (!nsi_lock->force_lock)
11049 + si_read_unlock(sb);
11052 + nsi_lock->bindex = bindex;
11056 +struct find_name_by_ino {
11057 + struct dir_context ctx;
11058 + int called, found;
11065 +find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
11066 + loff_t offset, u64 ino, unsigned int d_type)
11068 + struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
11072 + if (a->ino != ino)
11075 + memcpy(a->name, name, namelen);
11076 + a->namelen = namelen;
11081 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
11082 + struct au_nfsd_si_lock *nsi_lock)
11084 + struct dentry *dentry, *parent;
11085 + struct file *file;
11086 + struct inode *dir;
11087 + struct find_name_by_ino arg = {
11089 + .actor = find_name_by_ino
11094 + parent = path->dentry;
11096 + si_read_unlock(parent->d_sb);
11097 + file = vfsub_dentry_open(path, au_dir_roflags);
11098 + dentry = (void *)file;
11099 + if (IS_ERR(file))
11102 + dentry = ERR_PTR(-ENOMEM);
11103 + arg.name = (void *)__get_free_page(GFP_NOFS);
11104 + if (unlikely(!arg.name))
11111 + err = vfsub_iterate_dir(file, &arg.ctx);
11112 + } while (!err && !arg.found && arg.called);
11113 + dentry = ERR_PTR(err);
11114 + if (unlikely(err))
11116 + /* instead of ENOENT */
11117 + dentry = ERR_PTR(-ESTALE);
11121 + /* do not call vfsub_lkup_one() */
11122 + dir = d_inode(parent);
11123 + mutex_lock(&dir->i_mutex);
11124 + dentry = vfsub_lookup_one_len(arg.name, parent, arg.namelen);
11125 + mutex_unlock(&dir->i_mutex);
11126 + AuTraceErrPtr(dentry);
11127 + if (IS_ERR(dentry))
11129 + AuDebugOn(au_test_anon(dentry));
11130 + if (unlikely(d_really_is_negative(dentry))) {
11132 + dentry = ERR_PTR(-ENOENT);
11136 + free_page((unsigned long)arg.name);
11140 + if (unlikely(nsi_lock
11141 + && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
11142 + if (!IS_ERR(dentry)) {
11144 + dentry = ERR_PTR(-ESTALE);
11146 + AuTraceErrPtr(dentry);
11150 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
11152 + struct au_nfsd_si_lock *nsi_lock)
11154 + struct dentry *dentry;
11155 + struct path path;
11157 + if (dir_ino != AUFS_ROOT_INO) {
11158 + path.dentry = decode_by_ino(sb, dir_ino, 0);
11159 + dentry = path.dentry;
11160 + if (!path.dentry || IS_ERR(path.dentry))
11162 + AuDebugOn(au_test_anon(path.dentry));
11164 + path.dentry = dget(sb->s_root);
11166 + path.mnt = au_mnt_get(sb);
11167 + dentry = au_lkup_by_ino(&path, ino, nsi_lock);
11171 + AuTraceErrPtr(dentry);
11175 +/* ---------------------------------------------------------------------- */
11177 +static int h_acceptable(void *expv, struct dentry *dentry)
11182 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
11183 + char *buf, int len, struct super_block *sb)
11187 + struct path path;
11189 + p = d_path(h_rootpath, buf, len);
11194 + path.mnt = h_rootpath->mnt;
11195 + path.dentry = h_parent;
11196 + p = d_path(&path, buf, len);
11202 + path.mnt = au_mnt_get(sb);
11203 + path.dentry = sb->s_root;
11204 + p = d_path(&path, buf, len - strlen(p));
11205 + mntput(path.mnt);
11209 + p[strlen(p)] = '/';
11212 + AuTraceErrPtr(p);
11217 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
11218 + int fh_len, struct au_nfsd_si_lock *nsi_lock)
11220 + struct dentry *dentry, *h_parent, *root;
11221 + struct super_block *h_sb;
11222 + char *pathname, *p;
11223 + struct vfsmount *h_mnt;
11224 + struct au_branch *br;
11226 + struct path path;
11228 + br = au_sbr(sb, nsi_lock->bindex);
11229 + h_mnt = au_br_mnt(br);
11230 + h_sb = h_mnt->mnt_sb;
11231 + /* todo: call lower fh_to_dentry()? fh_to_parent()? */
11232 + h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
11233 + fh_len - Fh_tail, fh[Fh_h_type],
11234 + h_acceptable, /*context*/NULL);
11235 + dentry = h_parent;
11236 + if (unlikely(!h_parent || IS_ERR(h_parent))) {
11237 + AuWarn1("%s decode_fh failed, %ld\n",
11238 + au_sbtype(h_sb), PTR_ERR(h_parent));
11242 + if (unlikely(au_test_anon(h_parent))) {
11243 + AuWarn1("%s decode_fh returned a disconnected dentry\n",
11244 + au_sbtype(h_sb));
11245 + goto out_h_parent;
11248 + dentry = ERR_PTR(-ENOMEM);
11249 + pathname = (void *)__get_free_page(GFP_NOFS);
11250 + if (unlikely(!pathname))
11251 + goto out_h_parent;
11253 + root = sb->s_root;
11254 + path.mnt = h_mnt;
11255 + di_read_lock_parent(root, !AuLock_IR);
11256 + path.dentry = au_h_dptr(root, nsi_lock->bindex);
11257 + di_read_unlock(root, !AuLock_IR);
11258 + p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
11259 + dentry = (void *)p;
11261 + goto out_pathname;
11263 + si_read_unlock(sb);
11264 + err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
11265 + dentry = ERR_PTR(err);
11266 + if (unlikely(err))
11269 + dentry = ERR_PTR(-ENOENT);
11270 + AuDebugOn(au_test_anon(path.dentry));
11271 + if (unlikely(d_really_is_negative(path.dentry)))
11274 + if (ino != d_inode(path.dentry)->i_ino)
11275 + dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
11277 + dentry = dget(path.dentry);
11282 + if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
11283 + if (!IS_ERR(dentry)) {
11285 + dentry = ERR_PTR(-ESTALE);
11288 + free_page((unsigned long)pathname);
11292 + AuTraceErrPtr(dentry);
11296 +/* ---------------------------------------------------------------------- */
11298 +static struct dentry *
11299 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
11302 + struct dentry *dentry;
11303 + __u32 *fh = fid->raw;
11304 + struct au_branch *br;
11305 + ino_t ino, dir_ino;
11306 + struct au_nfsd_si_lock nsi_lock = {
11310 + dentry = ERR_PTR(-ESTALE);
11311 + /* it should never happen, but the file handle is unreliable */
11312 + if (unlikely(fh_len < Fh_tail))
11314 + nsi_lock.sigen = fh[Fh_sigen];
11315 + nsi_lock.br_id = fh[Fh_br_id];
11317 + /* branch id may be wrapped around */
11319 + if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
11321 + nsi_lock.force_lock = 1;
11323 + /* is this inode still cached? */
11324 + ino = decode_ino(fh + Fh_ino);
11325 + /* it should never happen */
11326 + if (unlikely(ino == AUFS_ROOT_INO))
11329 + dir_ino = decode_ino(fh + Fh_dir_ino);
11330 + dentry = decode_by_ino(sb, ino, dir_ino);
11331 + if (IS_ERR(dentry))
11336 + /* is the parent dir cached? */
11337 + br = au_sbr(sb, nsi_lock.bindex);
11338 + atomic_inc(&br->br_count);
11339 + dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
11340 + if (IS_ERR(dentry))
11345 + /* lookup path */
11346 + dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
11347 + if (IS_ERR(dentry))
11349 + if (unlikely(!dentry))
11350 + /* todo?: make it ESTALE */
11354 + if (!au_digen_test(dentry, au_sigen(sb))
11355 + && d_inode(dentry)->i_generation == fh[Fh_igen])
11356 + goto out_unlock; /* success */
11359 + dentry = ERR_PTR(-ESTALE);
11362 + atomic_dec(&br->br_count);
11363 + si_read_unlock(sb);
11365 + AuTraceErrPtr(dentry);
11369 +#if 0 /* reserved for future use */
11370 +/* support subtreecheck option */
11371 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
11372 + int fh_len, int fh_type)
11374 + struct dentry *parent;
11375 + __u32 *fh = fid->raw;
11378 + dir_ino = decode_ino(fh + Fh_dir_ino);
11379 + parent = decode_by_ino(sb, dir_ino, 0);
11380 + if (IS_ERR(parent))
11383 + parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
11384 + dir_ino, fh, fh_len);
11387 + AuTraceErrPtr(parent);
11392 +/* ---------------------------------------------------------------------- */
11394 +static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
11395 + struct inode *dir)
11398 + aufs_bindex_t bindex;
11399 + struct super_block *sb, *h_sb;
11400 + struct dentry *dentry, *parent, *h_parent;
11401 + struct inode *h_dir;
11402 + struct au_branch *br;
11405 + if (unlikely(*max_len <= Fh_tail)) {
11406 + AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
11410 + err = FILEID_ROOT;
11411 + if (inode->i_ino == AUFS_ROOT_INO) {
11412 + AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
11417 + sb = inode->i_sb;
11418 + err = si_read_lock(sb, AuLock_FLUSH);
11419 + if (unlikely(err))
11422 +#ifdef CONFIG_AUFS_DEBUG
11423 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
11424 + AuWarn1("NFS-exporting requires xino\n");
11428 + ii_read_lock_child(inode);
11429 + bindex = au_ibstart(inode);
11431 + dentry = d_find_any_alias(inode);
11432 + if (unlikely(!dentry))
11434 + AuDebugOn(au_test_anon(dentry));
11435 + parent = dget_parent(dentry);
11437 + if (unlikely(!parent))
11439 + if (d_really_is_positive(parent))
11440 + dir = d_inode(parent);
11443 + ii_read_lock_parent(dir);
11444 + h_dir = au_h_iptr(dir, bindex);
11445 + ii_read_unlock(dir);
11446 + if (unlikely(!h_dir))
11448 + h_parent = d_find_any_alias(h_dir);
11449 + if (unlikely(!h_parent))
11450 + goto out_hparent;
11453 + br = au_sbr(sb, bindex);
11454 + h_sb = au_br_sb(br);
11455 + if (unlikely(!h_sb->s_export_op)) {
11456 + AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
11457 + goto out_hparent;
11460 + fh[Fh_br_id] = br->br_id;
11461 + fh[Fh_sigen] = au_sigen(sb);
11462 + encode_ino(fh + Fh_ino, inode->i_ino);
11463 + encode_ino(fh + Fh_dir_ino, dir->i_ino);
11464 + fh[Fh_igen] = inode->i_generation;
11466 + *max_len -= Fh_tail;
11467 + fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
11469 + /*connectable or subtreecheck*/0);
11470 + err = fh[Fh_h_type];
11471 + *max_len += Fh_tail;
11472 + /* todo: macros? */
11473 + if (err != FILEID_INVALID)
11476 + AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
11483 + ii_read_unlock(inode);
11484 + si_read_unlock(sb);
11486 + if (unlikely(err < 0))
11487 + err = FILEID_INVALID;
11491 +/* ---------------------------------------------------------------------- */
11493 +static int aufs_commit_metadata(struct inode *inode)
11496 + aufs_bindex_t bindex;
11497 + struct super_block *sb;
11498 + struct inode *h_inode;
11499 + int (*f)(struct inode *inode);
11501 + sb = inode->i_sb;
11502 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
11503 + ii_write_lock_child(inode);
11504 + bindex = au_ibstart(inode);
11505 + AuDebugOn(bindex < 0);
11506 + h_inode = au_h_iptr(inode, bindex);
11508 + f = h_inode->i_sb->s_export_op->commit_metadata;
11510 + err = f(h_inode);
11512 + struct writeback_control wbc = {
11513 + .sync_mode = WB_SYNC_ALL,
11514 + .nr_to_write = 0 /* metadata only */
11517 + err = sync_inode(h_inode, &wbc);
11520 + au_cpup_attr_timesizes(inode);
11521 + ii_write_unlock(inode);
11522 + si_read_unlock(sb);
11526 +/* ---------------------------------------------------------------------- */
11528 +static struct export_operations aufs_export_op = {
11529 + .fh_to_dentry = aufs_fh_to_dentry,
11530 + /* .fh_to_parent = aufs_fh_to_parent, */
11531 + .encode_fh = aufs_encode_fh,
11532 + .commit_metadata = aufs_commit_metadata
11535 +void au_export_init(struct super_block *sb)
11537 + struct au_sbinfo *sbinfo;
11540 + sb->s_export_op = &aufs_export_op;
11541 + sbinfo = au_sbi(sb);
11542 + sbinfo->si_xigen = NULL;
11543 + get_random_bytes(&u, sizeof(u));
11544 + BUILD_BUG_ON(sizeof(u) != sizeof(int));
11545 + atomic_set(&sbinfo->si_xigen_next, u);
11547 diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
11548 --- /usr/share/empty/fs/aufs/fhsm.c 1970-01-01 01:00:00.000000000 +0100
11549 +++ linux/fs/aufs/fhsm.c 2015-09-24 10:47:58.251386326 +0200
11552 + * Copyright (C) 2011-2015 Junjiro R. Okajima
11554 + * This program, aufs is free software; you can redistribute it and/or modify
11555 + * it under the terms of the GNU General Public License as published by
11556 + * the Free Software Foundation; either version 2 of the License, or
11557 + * (at your option) any later version.
11559 + * This program is distributed in the hope that it will be useful,
11560 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
11561 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11562 + * GNU General Public License for more details.
11564 + * You should have received a copy of the GNU General Public License
11565 + * along with this program; if not, write to the Free Software
11566 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
11570 + * File-based Hierarchy Storage Management
11573 +#include <linux/anon_inodes.h>
11574 +#include <linux/poll.h>
11575 +#include <linux/seq_file.h>
11576 +#include <linux/statfs.h>
11579 +static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
11581 + struct au_sbinfo *sbinfo;
11582 + struct au_fhsm *fhsm;
11584 + SiMustAnyLock(sb);
11586 + sbinfo = au_sbi(sb);
11587 + fhsm = &sbinfo->si_fhsm;
11588 + AuDebugOn(!fhsm);
11589 + return fhsm->fhsm_bottom;
11592 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
11594 + struct au_sbinfo *sbinfo;
11595 + struct au_fhsm *fhsm;
11597 + SiMustWriteLock(sb);
11599 + sbinfo = au_sbi(sb);
11600 + fhsm = &sbinfo->si_fhsm;
11601 + AuDebugOn(!fhsm);
11602 + fhsm->fhsm_bottom = bindex;
11605 +/* ---------------------------------------------------------------------- */
11607 +static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
11609 + struct au_br_fhsm *bf;
11611 + bf = br->br_fhsm;
11612 + MtxMustLock(&bf->bf_lock);
11614 + return !bf->bf_readable
11615 + || time_after(jiffies,
11616 + bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
11619 +/* ---------------------------------------------------------------------- */
11621 +static void au_fhsm_notify(struct super_block *sb, int val)
11623 + struct au_sbinfo *sbinfo;
11624 + struct au_fhsm *fhsm;
11626 + SiMustAnyLock(sb);
11628 + sbinfo = au_sbi(sb);
11629 + fhsm = &sbinfo->si_fhsm;
11630 + if (au_fhsm_pid(fhsm)
11631 + && atomic_read(&fhsm->fhsm_readable) != -1) {
11632 + atomic_set(&fhsm->fhsm_readable, val);
11634 + wake_up(&fhsm->fhsm_wqh);
11638 +static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
11639 + struct aufs_stfs *rstfs, int do_lock, int do_notify)
11642 + struct au_branch *br;
11643 + struct au_br_fhsm *bf;
11645 + br = au_sbr(sb, bindex);
11646 + AuDebugOn(au_br_rdonly(br));
11647 + bf = br->br_fhsm;
11651 + mutex_lock(&bf->bf_lock);
11653 + MtxMustLock(&bf->bf_lock);
11655 + /* sb->s_root for NFS is unreliable */
11656 + err = au_br_stfs(br, &bf->bf_stfs);
11657 + if (unlikely(err)) {
11658 + AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
11662 + bf->bf_jiffy = jiffies;
11663 + bf->bf_readable = 1;
11665 + au_fhsm_notify(sb, /*val*/1);
11667 + *rstfs = bf->bf_stfs;
11671 + mutex_unlock(&bf->bf_lock);
11672 + au_fhsm_notify(sb, /*val*/1);
11677 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
11680 + struct au_sbinfo *sbinfo;
11681 + struct au_fhsm *fhsm;
11682 + struct au_branch *br;
11683 + struct au_br_fhsm *bf;
11685 + AuDbg("b%d, force %d\n", bindex, force);
11686 + SiMustAnyLock(sb);
11688 + sbinfo = au_sbi(sb);
11689 + fhsm = &sbinfo->si_fhsm;
11690 + if (!au_ftest_si(sbinfo, FHSM)
11691 + || fhsm->fhsm_bottom == bindex)
11694 + br = au_sbr(sb, bindex);
11695 + bf = br->br_fhsm;
11697 + mutex_lock(&bf->bf_lock);
11699 + || au_fhsm_pid(fhsm)
11700 + || au_fhsm_test_jiffy(sbinfo, br))
11701 + err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
11703 + mutex_unlock(&bf->bf_lock);
11706 +void au_fhsm_wrote_all(struct super_block *sb, int force)
11708 + aufs_bindex_t bindex, bend;
11709 + struct au_branch *br;
11711 + /* exclude the bottom */
11712 + bend = au_fhsm_bottom(sb);
11713 + for (bindex = 0; bindex < bend; bindex++) {
11714 + br = au_sbr(sb, bindex);
11715 + if (au_br_fhsm(br->br_perm))
11716 + au_fhsm_wrote(sb, bindex, force);
11720 +/* ---------------------------------------------------------------------- */
11722 +static unsigned int au_fhsm_poll(struct file *file,
11723 + struct poll_table_struct *wait)
11725 + unsigned int mask;
11726 + struct au_sbinfo *sbinfo;
11727 + struct au_fhsm *fhsm;
11730 + sbinfo = file->private_data;
11731 + fhsm = &sbinfo->si_fhsm;
11732 + poll_wait(file, &fhsm->fhsm_wqh, wait);
11733 + if (atomic_read(&fhsm->fhsm_readable))
11734 + mask = POLLIN /* | POLLRDNORM */;
11736 + AuTraceErr((int)mask);
11740 +static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
11741 + struct aufs_stfs *stfs, __s16 brid)
11745 + err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
11747 + err = __put_user(brid, &stbr->brid);
11748 + if (unlikely(err))
11754 +static ssize_t au_fhsm_do_read(struct super_block *sb,
11755 + struct aufs_stbr __user *stbr, size_t count)
11759 + aufs_bindex_t bindex, bend;
11760 + struct au_branch *br;
11761 + struct au_br_fhsm *bf;
11763 + /* except the bottom branch */
11766 + bend = au_fhsm_bottom(sb);
11767 + for (bindex = 0; !err && bindex < bend; bindex++) {
11768 + br = au_sbr(sb, bindex);
11769 + if (!au_br_fhsm(br->br_perm))
11772 + bf = br->br_fhsm;
11773 + mutex_lock(&bf->bf_lock);
11774 + if (bf->bf_readable) {
11776 + if (count >= sizeof(*stbr))
11777 + err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
11780 + bf->bf_readable = 0;
11781 + count -= sizeof(*stbr);
11785 + mutex_unlock(&bf->bf_lock);
11788 + err = sizeof(*stbr) * nstbr;
11793 +static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
11798 + aufs_bindex_t nfhsm, bindex, bend;
11799 + struct au_sbinfo *sbinfo;
11800 + struct au_fhsm *fhsm;
11801 + struct au_branch *br;
11802 + struct super_block *sb;
11805 + sbinfo = file->private_data;
11806 + fhsm = &sbinfo->si_fhsm;
11808 + spin_lock_irq(&fhsm->fhsm_wqh.lock);
11809 + if (!atomic_read(&fhsm->fhsm_readable)) {
11810 + if (vfsub_file_flags(file) & O_NONBLOCK)
11813 + err = wait_event_interruptible_locked_irq
11815 + atomic_read(&fhsm->fhsm_readable));
11817 + spin_unlock_irq(&fhsm->fhsm_wqh.lock);
11818 + if (unlikely(err))
11821 + /* sb may already be dead */
11822 + au_rw_read_lock(&sbinfo->si_rwsem);
11823 + readable = atomic_read(&fhsm->fhsm_readable);
11824 + if (readable > 0) {
11825 + sb = sbinfo->si_sb;
11827 + /* exclude the bottom branch */
11829 + bend = au_fhsm_bottom(sb);
11830 + for (bindex = 0; bindex < bend; bindex++) {
11831 + br = au_sbr(sb, bindex);
11832 + if (au_br_fhsm(br->br_perm))
11836 + if (nfhsm * sizeof(struct aufs_stbr) <= count) {
11837 + atomic_set(&fhsm->fhsm_readable, 0);
11838 + err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
11842 + au_rw_read_unlock(&sbinfo->si_rwsem);
11850 +static int au_fhsm_release(struct inode *inode, struct file *file)
11852 + struct au_sbinfo *sbinfo;
11853 + struct au_fhsm *fhsm;
11855 + /* sb may already be dead */
11856 + sbinfo = file->private_data;
11857 + fhsm = &sbinfo->si_fhsm;
11858 + spin_lock(&fhsm->fhsm_spin);
11859 + fhsm->fhsm_pid = 0;
11860 + spin_unlock(&fhsm->fhsm_spin);
11861 + kobject_put(&sbinfo->si_kobj);
11866 +static const struct file_operations au_fhsm_fops = {
11867 + .owner = THIS_MODULE,
11868 + .llseek = noop_llseek,
11869 + .read = au_fhsm_read,
11870 + .poll = au_fhsm_poll,
11871 + .release = au_fhsm_release
11874 +int au_fhsm_fd(struct super_block *sb, int oflags)
11877 + struct au_sbinfo *sbinfo;
11878 + struct au_fhsm *fhsm;
11881 + if (unlikely(!capable(CAP_SYS_ADMIN)))
11885 + if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
11889 + sbinfo = au_sbi(sb);
11890 + fhsm = &sbinfo->si_fhsm;
11891 + spin_lock(&fhsm->fhsm_spin);
11892 + if (!fhsm->fhsm_pid)
11893 + fhsm->fhsm_pid = current->pid;
11896 + spin_unlock(&fhsm->fhsm_spin);
11897 + if (unlikely(err))
11900 + oflags |= O_RDONLY;
11901 + /* oflags |= FMODE_NONOTIFY; */
11902 + fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
11904 + if (unlikely(fd < 0))
11907 + /* succeed reglardless 'fhsm' status */
11908 + kobject_get(&sbinfo->si_kobj);
11909 + si_noflush_read_lock(sb);
11910 + if (au_ftest_si(sbinfo, FHSM))
11911 + au_fhsm_wrote_all(sb, /*force*/0);
11912 + si_read_unlock(sb);
11913 + goto out; /* success */
11916 + spin_lock(&fhsm->fhsm_spin);
11917 + fhsm->fhsm_pid = 0;
11918 + spin_unlock(&fhsm->fhsm_spin);
11924 +/* ---------------------------------------------------------------------- */
11926 +int au_fhsm_br_alloc(struct au_branch *br)
11931 + br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
11933 + au_br_fhsm_init(br->br_fhsm);
11940 +/* ---------------------------------------------------------------------- */
11942 +void au_fhsm_fin(struct super_block *sb)
11944 + au_fhsm_notify(sb, /*val*/-1);
11947 +void au_fhsm_init(struct au_sbinfo *sbinfo)
11949 + struct au_fhsm *fhsm;
11951 + fhsm = &sbinfo->si_fhsm;
11952 + spin_lock_init(&fhsm->fhsm_spin);
11953 + init_waitqueue_head(&fhsm->fhsm_wqh);
11954 + atomic_set(&fhsm->fhsm_readable, 0);
11955 + fhsm->fhsm_expire
11956 + = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
11957 + fhsm->fhsm_bottom = -1;
11960 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
11962 + sbinfo->si_fhsm.fhsm_expire
11963 + = msecs_to_jiffies(sec * MSEC_PER_SEC);
11966 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
11970 + if (!au_ftest_si(sbinfo, FHSM))
11973 + u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
11974 + if (u != AUFS_FHSM_CACHE_DEF_SEC)
11975 + seq_printf(seq, ",fhsm_sec=%u", u);
11977 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
11978 --- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
11979 +++ linux/fs/aufs/file.c 2015-11-11 17:21:46.918863802 +0100
11982 + * Copyright (C) 2005-2015 Junjiro R. Okajima
11984 + * This program, aufs is free software; you can redistribute it and/or modify
11985 + * it under the terms of the GNU General Public License as published by
11986 + * the Free Software Foundation; either version 2 of the License, or
11987 + * (at your option) any later version.
11989 + * This program is distributed in the hope that it will be useful,
11990 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
11991 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11992 + * GNU General Public License for more details.
11994 + * You should have received a copy of the GNU General Public License
11995 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
11999 + * handling file/dir, and address_space operation
12002 +#ifdef CONFIG_AUFS_DEBUG
12003 +#include <linux/migrate.h>
12005 +#include <linux/pagemap.h>
12008 +/* drop flags for writing */
12009 +unsigned int au_file_roflags(unsigned int flags)
12011 + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
12012 + flags |= O_RDONLY | O_NOATIME;
12016 +/* common functions to regular file and dir */
12017 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
12018 + struct file *file, int force_wr)
12020 + struct file *h_file;
12021 + struct dentry *h_dentry;
12022 + struct inode *h_inode;
12023 + struct super_block *sb;
12024 + struct au_branch *br;
12025 + struct path h_path;
12028 + /* a race condition can happen between open and unlink/rmdir */
12029 + h_file = ERR_PTR(-ENOENT);
12030 + h_dentry = au_h_dptr(dentry, bindex);
12031 + if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
12033 + h_inode = d_inode(h_dentry);
12034 + spin_lock(&h_dentry->d_lock);
12035 + err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
12036 + /* || !d_inode(dentry)->i_nlink */
12038 + spin_unlock(&h_dentry->d_lock);
12039 + if (unlikely(err))
12042 + sb = dentry->d_sb;
12043 + br = au_sbr(sb, bindex);
12044 + err = au_br_test_oflag(flags, br);
12045 + h_file = ERR_PTR(err);
12046 + if (unlikely(err))
12049 + /* drop flags for writing */
12050 + if (au_test_ro(sb, bindex, d_inode(dentry))) {
12051 + if (force_wr && !(flags & O_WRONLY))
12053 + flags = au_file_roflags(flags);
12055 + h_file = ERR_PTR(-EROFS);
12056 + flags = au_file_roflags(flags);
12057 + if (unlikely(vfsub_native_ro(h_inode)
12058 + || IS_APPEND(h_inode)))
12060 + flags &= ~O_ACCMODE;
12061 + flags |= O_WRONLY;
12064 + flags &= ~O_CREAT;
12065 + atomic_inc(&br->br_count);
12066 + h_path.dentry = h_dentry;
12067 + h_path.mnt = au_br_mnt(br);
12068 + h_file = vfsub_dentry_open(&h_path, flags);
12069 + if (IS_ERR(h_file))
12072 + if (flags & __FMODE_EXEC) {
12073 + err = deny_write_access(h_file);
12074 + if (unlikely(err)) {
12076 + h_file = ERR_PTR(err);
12080 + fsnotify_open(h_file);
12081 + goto out; /* success */
12084 + atomic_dec(&br->br_count);
12089 +static int au_cmoo(struct dentry *dentry)
12092 + unsigned int udba;
12093 + struct path h_path;
12094 + struct au_pin pin;
12095 + struct au_cp_generic cpg = {
12096 + .dentry = dentry,
12101 + .flags = AuCpup_DTIME | AuCpup_HOPEN
12103 + struct inode *delegated;
12104 + struct super_block *sb;
12105 + struct au_sbinfo *sbinfo;
12106 + struct au_fhsm *fhsm;
12108 + struct au_branch *br;
12109 + struct dentry *parent;
12110 + struct au_hinode *hdir;
12112 + DiMustWriteLock(dentry);
12113 + IiMustWriteLock(d_inode(dentry));
12116 + if (IS_ROOT(dentry))
12118 + cpg.bsrc = au_dbstart(dentry);
12122 + sb = dentry->d_sb;
12123 + sbinfo = au_sbi(sb);
12124 + fhsm = &sbinfo->si_fhsm;
12125 + pid = au_fhsm_pid(fhsm);
12127 + && (current->pid == pid
12128 + || current->real_parent->pid == pid))
12131 + br = au_sbr(sb, cpg.bsrc);
12132 + cmoo = au_br_cmoo(br->br_perm);
12135 + if (!d_is_reg(dentry))
12136 + cmoo &= AuBrAttr_COO_ALL;
12140 + parent = dget_parent(dentry);
12141 + di_write_lock_parent(parent);
12142 + err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
12144 + if (unlikely(err < 0)) {
12145 + err = 0; /* there is no upper writable branch */
12148 + AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
12150 + /* do not respect the coo attrib for the target branch */
12151 + err = au_cpup_dirs(dentry, cpg.bdst);
12152 + if (unlikely(err))
12155 + di_downgrade_lock(parent, AuLock_IR);
12156 + udba = au_opt_udba(sb);
12157 + err = au_pin(&pin, dentry, cpg.bdst, udba,
12158 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12159 + if (unlikely(err))
12162 + err = au_sio_cpup_simple(&cpg);
12164 + if (unlikely(err))
12166 + if (!(cmoo & AuBrWAttr_MOO))
12167 + goto out_parent; /* success */
12169 + err = au_pin(&pin, dentry, cpg.bsrc, udba,
12170 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12171 + if (unlikely(err))
12174 + h_path.mnt = au_br_mnt(br);
12175 + h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
12176 + hdir = au_hi(d_inode(parent), cpg.bsrc);
12177 + delegated = NULL;
12178 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
12180 + /* todo: keep h_dentry or not? */
12181 + if (unlikely(err == -EWOULDBLOCK)) {
12182 + pr_warn("cannot retry for NFSv4 delegation"
12183 + " for an internal unlink\n");
12186 + if (unlikely(err)) {
12187 + pr_err("unlink %pd after coo failed (%d), ignored\n",
12191 + goto out_parent; /* success */
12194 + di_downgrade_lock(parent, AuLock_IR);
12196 + di_read_unlock(parent, AuLock_IR);
12203 +int au_do_open(struct file *file, struct au_do_open_args *args)
12205 + int err, no_lock = args->no_lock;
12206 + struct dentry *dentry;
12207 + struct au_finfo *finfo;
12210 + err = au_finfo_init(file, args->fidir);
12213 + err = au_finfo_init(file, args->fidir);
12216 + if (unlikely(err))
12219 + dentry = file->f_path.dentry;
12220 + AuDebugOn(IS_ERR_OR_NULL(dentry));
12222 + di_write_lock_child(dentry);
12223 + err = au_cmoo(dentry);
12224 + di_downgrade_lock(dentry, AuLock_IR);
12226 + err = args->open(file, vfsub_file_flags(file), NULL);
12227 + di_read_unlock(dentry, AuLock_IR);
12229 + err = au_cmoo(dentry);
12231 + err = args->open(file, vfsub_file_flags(file),
12233 + if (!err && au_fbstart(file) != au_dbstart(dentry))
12235 + * cmoo happens after h_file was opened.
12236 + * need to refresh file later.
12238 + atomic_dec(&au_fi(file)->fi_generation);
12241 + finfo = au_fi(file);
12243 + finfo->fi_file = file;
12244 + au_sphl_add(&finfo->fi_hlist,
12245 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
12248 + fi_write_unlock(file);
12251 + fi_write_unlock(file);
12254 + if (unlikely(err)) {
12255 + finfo->fi_hdir = NULL;
12256 + au_finfo_fin(file);
12263 +int au_reopen_nondir(struct file *file)
12266 + aufs_bindex_t bstart;
12267 + struct dentry *dentry;
12268 + struct file *h_file, *h_file_tmp;
12270 + dentry = file->f_path.dentry;
12271 + bstart = au_dbstart(dentry);
12272 + h_file_tmp = NULL;
12273 + if (au_fbstart(file) == bstart) {
12274 + h_file = au_hf_top(file);
12275 + if (file->f_mode == h_file->f_mode)
12276 + return 0; /* success */
12277 + h_file_tmp = h_file;
12278 + get_file(h_file_tmp);
12279 + au_set_h_fptr(file, bstart, NULL);
12281 + AuDebugOn(au_fi(file)->fi_hdir);
12284 + * file exists on both of rw and ro
12285 + * open --> dbstart and fbstart are both 0
12286 + * prepend a branch as rw, "rw" become ro
12288 + * delete the top branch, "rw" becomes rw again
12289 + * --> dbstart is 1, fbstart is still 0
12290 + * write --> fbstart is 0 but dbstart is 1
12292 + /* AuDebugOn(au_fbstart(file) < bstart); */
12294 + h_file = au_h_open(dentry, bstart, vfsub_file_flags(file) & ~O_TRUNC,
12295 + file, /*force_wr*/0);
12296 + err = PTR_ERR(h_file);
12297 + if (IS_ERR(h_file)) {
12298 + if (h_file_tmp) {
12299 + atomic_inc(&au_sbr(dentry->d_sb, bstart)->br_count);
12300 + au_set_h_fptr(file, bstart, h_file_tmp);
12301 + h_file_tmp = NULL;
12303 + goto out; /* todo: close all? */
12307 + au_set_fbstart(file, bstart);
12308 + au_set_h_fptr(file, bstart, h_file);
12309 + au_update_figen(file);
12310 + /* todo: necessary? */
12311 + /* file->f_ra = h_file->f_ra; */
12315 + fput(h_file_tmp);
12319 +/* ---------------------------------------------------------------------- */
12321 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
12322 + struct dentry *hi_wh)
12325 + aufs_bindex_t bstart;
12326 + struct au_dinfo *dinfo;
12327 + struct dentry *h_dentry;
12328 + struct au_hdentry *hdp;
12330 + dinfo = au_di(file->f_path.dentry);
12331 + AuRwMustWriteLock(&dinfo->di_rwsem);
12333 + bstart = dinfo->di_bstart;
12334 + dinfo->di_bstart = btgt;
12335 + hdp = dinfo->di_hdentry;
12336 + h_dentry = hdp[0 + btgt].hd_dentry;
12337 + hdp[0 + btgt].hd_dentry = hi_wh;
12338 + err = au_reopen_nondir(file);
12339 + hdp[0 + btgt].hd_dentry = h_dentry;
12340 + dinfo->di_bstart = bstart;
12345 +static int au_ready_to_write_wh(struct file *file, loff_t len,
12346 + aufs_bindex_t bcpup, struct au_pin *pin)
12349 + struct inode *inode, *h_inode;
12350 + struct dentry *h_dentry, *hi_wh;
12351 + struct au_cp_generic cpg = {
12352 + .dentry = file->f_path.dentry,
12359 + au_update_dbstart(cpg.dentry);
12360 + inode = d_inode(cpg.dentry);
12362 + if (au_dbstart(cpg.dentry) <= bcpup
12363 + && au_dbend(cpg.dentry) >= bcpup) {
12364 + h_dentry = au_h_dptr(cpg.dentry, bcpup);
12365 + if (h_dentry && d_is_positive(h_dentry))
12366 + h_inode = d_inode(h_dentry);
12368 + hi_wh = au_hi_wh(inode, bcpup);
12369 + if (!hi_wh && !h_inode)
12370 + err = au_sio_cpup_wh(&cpg, file);
12372 + /* already copied-up after unlink */
12373 + err = au_reopen_wh(file, bcpup, hi_wh);
12376 + && (inode->i_nlink > 1
12377 + || (inode->i_state & I_LINKABLE))
12378 + && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
12379 + au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
12385 + * prepare the @file for writing.
12387 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
12390 + aufs_bindex_t dbstart;
12391 + struct dentry *parent;
12392 + struct inode *inode;
12393 + struct super_block *sb;
12394 + struct file *h_file;
12395 + struct au_cp_generic cpg = {
12396 + .dentry = file->f_path.dentry,
12401 + .flags = AuCpup_DTIME
12404 + sb = cpg.dentry->d_sb;
12405 + inode = d_inode(cpg.dentry);
12406 + cpg.bsrc = au_fbstart(file);
12407 + err = au_test_ro(sb, cpg.bsrc, inode);
12408 + if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
12409 + err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
12414 + /* need to cpup or reopen */
12415 + parent = dget_parent(cpg.dentry);
12416 + di_write_lock_parent(parent);
12417 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
12419 + if (unlikely(err < 0))
12423 + if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
12424 + err = au_cpup_dirs(cpg.dentry, cpg.bdst);
12425 + if (unlikely(err))
12429 + err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
12430 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12431 + if (unlikely(err))
12434 + dbstart = au_dbstart(cpg.dentry);
12435 + if (dbstart <= cpg.bdst)
12436 + cpg.bsrc = cpg.bdst;
12438 + if (dbstart <= cpg.bdst /* just reopen */
12439 + || !d_unhashed(cpg.dentry) /* copyup and reopen */
12441 + h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
12442 + if (IS_ERR(h_file))
12443 + err = PTR_ERR(h_file);
12445 + di_downgrade_lock(parent, AuLock_IR);
12446 + if (dbstart > cpg.bdst)
12447 + err = au_sio_cpup_simple(&cpg);
12449 + err = au_reopen_nondir(file);
12450 + au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
12452 + } else { /* copyup as wh and reopen */
12454 + * since writable hfsplus branch is not supported,
12455 + * h_open_pre/post() are unnecessary.
12457 + err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
12458 + di_downgrade_lock(parent, AuLock_IR);
12462 + au_pin_set_parent_lflag(pin, /*lflag*/0);
12463 + goto out_dput; /* success */
12469 + di_downgrade_lock(parent, AuLock_IR);
12471 + di_read_unlock(parent, AuLock_IR);
12478 +/* ---------------------------------------------------------------------- */
12480 +int au_do_flush(struct file *file, fl_owner_t id,
12481 + int (*flush)(struct file *file, fl_owner_t id))
12484 + struct super_block *sb;
12485 + struct inode *inode;
12487 + inode = file_inode(file);
12488 + sb = inode->i_sb;
12489 + si_noflush_read_lock(sb);
12490 + fi_read_lock(file);
12491 + ii_read_lock_child(inode);
12493 + err = flush(file, id);
12494 + au_cpup_attr_timesizes(inode);
12496 + ii_read_unlock(inode);
12497 + fi_read_unlock(file);
12498 + si_read_unlock(sb);
12502 +/* ---------------------------------------------------------------------- */
12504 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
12507 + struct au_pin pin;
12508 + struct au_finfo *finfo;
12509 + struct dentry *parent, *hi_wh;
12510 + struct inode *inode;
12511 + struct super_block *sb;
12512 + struct au_cp_generic cpg = {
12513 + .dentry = file->f_path.dentry,
12518 + .flags = AuCpup_DTIME
12521 + FiMustWriteLock(file);
12524 + finfo = au_fi(file);
12525 + sb = cpg.dentry->d_sb;
12526 + inode = d_inode(cpg.dentry);
12527 + cpg.bdst = au_ibstart(inode);
12528 + if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
12531 + parent = dget_parent(cpg.dentry);
12532 + if (au_test_ro(sb, cpg.bdst, inode)) {
12533 + di_read_lock_parent(parent, !AuLock_IR);
12534 + err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
12536 + di_read_unlock(parent, !AuLock_IR);
12537 + if (unlikely(err < 0))
12542 + di_read_lock_parent(parent, AuLock_IR);
12543 + hi_wh = au_hi_wh(inode, cpg.bdst);
12544 + if (!S_ISDIR(inode->i_mode)
12545 + && au_opt_test(au_mntflags(sb), PLINK)
12546 + && au_plink_test(inode)
12547 + && !d_unhashed(cpg.dentry)
12548 + && cpg.bdst < au_dbstart(cpg.dentry)) {
12549 + err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
12550 + if (unlikely(err))
12553 + /* always superio. */
12554 + err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
12555 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12557 + err = au_sio_cpup_simple(&cpg);
12560 + } else if (hi_wh) {
12561 + /* already copied-up after unlink */
12562 + err = au_reopen_wh(file, cpg.bdst, hi_wh);
12563 + *need_reopen = 0;
12567 + di_read_unlock(parent, AuLock_IR);
12574 +static void au_do_refresh_dir(struct file *file)
12576 + aufs_bindex_t bindex, bend, new_bindex, brid;
12577 + struct au_hfile *p, tmp, *q;
12578 + struct au_finfo *finfo;
12579 + struct super_block *sb;
12580 + struct au_fidir *fidir;
12582 + FiMustWriteLock(file);
12584 + sb = file->f_path.dentry->d_sb;
12585 + finfo = au_fi(file);
12586 + fidir = finfo->fi_hdir;
12587 + AuDebugOn(!fidir);
12588 + p = fidir->fd_hfile + finfo->fi_btop;
12589 + brid = p->hf_br->br_id;
12590 + bend = fidir->fd_bbot;
12591 + for (bindex = finfo->fi_btop; bindex <= bend; bindex++, p++) {
12595 + new_bindex = au_br_index(sb, p->hf_br->br_id);
12596 + if (new_bindex == bindex)
12598 + if (new_bindex < 0) {
12599 + au_set_h_fptr(file, bindex, NULL);
12603 + /* swap two lower inode, and loop again */
12604 + q = fidir->fd_hfile + new_bindex;
12608 + if (tmp.hf_file) {
12614 + p = fidir->fd_hfile;
12615 + if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
12616 + bend = au_sbend(sb);
12617 + for (finfo->fi_btop = 0; finfo->fi_btop <= bend;
12618 + finfo->fi_btop++, p++)
12619 + if (p->hf_file) {
12620 + if (file_inode(p->hf_file))
12622 + au_hfput(p, file);
12625 + bend = au_br_index(sb, brid);
12626 + for (finfo->fi_btop = 0; finfo->fi_btop < bend;
12627 + finfo->fi_btop++, p++)
12629 + au_hfput(p, file);
12630 + bend = au_sbend(sb);
12633 + p = fidir->fd_hfile + bend;
12634 + for (fidir->fd_bbot = bend; fidir->fd_bbot >= finfo->fi_btop;
12635 + fidir->fd_bbot--, p--)
12636 + if (p->hf_file) {
12637 + if (file_inode(p->hf_file))
12639 + au_hfput(p, file);
12641 + AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
12645 + * after branch manipulating, refresh the file.
12647 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
12649 + int err, need_reopen;
12650 + aufs_bindex_t bend, bindex;
12651 + struct dentry *dentry;
12652 + struct au_finfo *finfo;
12653 + struct au_hfile *hfile;
12655 + dentry = file->f_path.dentry;
12656 + finfo = au_fi(file);
12657 + if (!finfo->fi_hdir) {
12658 + hfile = &finfo->fi_htop;
12659 + AuDebugOn(!hfile->hf_file);
12660 + bindex = au_br_index(dentry->d_sb, hfile->hf_br->br_id);
12661 + AuDebugOn(bindex < 0);
12662 + if (bindex != finfo->fi_btop)
12663 + au_set_fbstart(file, bindex);
12665 + err = au_fidir_realloc(finfo, au_sbend(dentry->d_sb) + 1);
12666 + if (unlikely(err))
12668 + au_do_refresh_dir(file);
12673 + if (!au_test_mmapped(file))
12674 + err = au_file_refresh_by_inode(file, &need_reopen);
12675 + if (!err && need_reopen && !d_unlinked(dentry))
12676 + err = reopen(file);
12678 + au_update_figen(file);
12679 + goto out; /* success */
12682 + /* error, close all lower files */
12683 + if (finfo->fi_hdir) {
12684 + bend = au_fbend_dir(file);
12685 + for (bindex = au_fbstart(file); bindex <= bend; bindex++)
12686 + au_set_h_fptr(file, bindex, NULL);
12693 +/* common function to regular file and dir */
12694 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
12698 + unsigned int sigen, figen;
12699 + aufs_bindex_t bstart;
12700 + unsigned char pseudo_link;
12701 + struct dentry *dentry;
12702 + struct inode *inode;
12705 + dentry = file->f_path.dentry;
12706 + inode = d_inode(dentry);
12707 + sigen = au_sigen(dentry->d_sb);
12708 + fi_write_lock(file);
12709 + figen = au_figen(file);
12710 + di_write_lock_child(dentry);
12711 + bstart = au_dbstart(dentry);
12712 + pseudo_link = (bstart != au_ibstart(inode));
12713 + if (sigen == figen && !pseudo_link && au_fbstart(file) == bstart) {
12715 + di_downgrade_lock(dentry, AuLock_IR);
12716 + fi_downgrade_lock(file);
12718 + goto out; /* success */
12721 + AuDbg("sigen %d, figen %d\n", sigen, figen);
12722 + if (au_digen_test(dentry, sigen)) {
12723 + err = au_reval_dpath(dentry, sigen);
12724 + AuDebugOn(!err && au_digen_test(dentry, sigen));
12728 + err = refresh_file(file, reopen);
12731 + di_downgrade_lock(dentry, AuLock_IR);
12732 + fi_downgrade_lock(file);
12735 + di_write_unlock(dentry);
12736 + fi_write_unlock(file);
12743 +/* ---------------------------------------------------------------------- */
12745 +/* cf. aufs_nopage() */
12746 +/* for madvise(2) */
12747 +static int aufs_readpage(struct file *file __maybe_unused, struct page *page)
12749 + unlock_page(page);
12753 +/* it will never be called, but necessary to support O_DIRECT */
12754 +static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
12756 +{ BUG(); return 0; }
12758 +/* they will never be called. */
12759 +#ifdef CONFIG_AUFS_DEBUG
12760 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
12761 + loff_t pos, unsigned len, unsigned flags,
12762 + struct page **pagep, void **fsdata)
12763 +{ AuUnsupport(); return 0; }
12764 +static int aufs_write_end(struct file *file, struct address_space *mapping,
12765 + loff_t pos, unsigned len, unsigned copied,
12766 + struct page *page, void *fsdata)
12767 +{ AuUnsupport(); return 0; }
12768 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
12769 +{ AuUnsupport(); return 0; }
12771 +static int aufs_set_page_dirty(struct page *page)
12772 +{ AuUnsupport(); return 0; }
12773 +static void aufs_invalidatepage(struct page *page, unsigned int offset,
12774 + unsigned int length)
12775 +{ AuUnsupport(); }
12776 +static int aufs_releasepage(struct page *page, gfp_t gfp)
12777 +{ AuUnsupport(); return 0; }
12778 +#if 0 /* called by memory compaction regardless file */
12779 +static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
12780 + struct page *page, enum migrate_mode mode)
12781 +{ AuUnsupport(); return 0; }
12783 +static int aufs_launder_page(struct page *page)
12784 +{ AuUnsupport(); return 0; }
12785 +static int aufs_is_partially_uptodate(struct page *page,
12786 + unsigned long from,
12787 + unsigned long count)
12788 +{ AuUnsupport(); return 0; }
12789 +static void aufs_is_dirty_writeback(struct page *page, bool *dirty,
12791 +{ AuUnsupport(); }
12792 +static int aufs_error_remove_page(struct address_space *mapping,
12793 + struct page *page)
12794 +{ AuUnsupport(); return 0; }
12795 +static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
12797 +{ AuUnsupport(); return 0; }
12798 +static void aufs_swap_deactivate(struct file *file)
12799 +{ AuUnsupport(); }
12800 +#endif /* CONFIG_AUFS_DEBUG */
12802 +const struct address_space_operations aufs_aop = {
12803 + .readpage = aufs_readpage,
12804 + .direct_IO = aufs_direct_IO,
12805 +#ifdef CONFIG_AUFS_DEBUG
12806 + .writepage = aufs_writepage,
12807 + /* no writepages, because of writepage */
12808 + .set_page_dirty = aufs_set_page_dirty,
12809 + /* no readpages, because of readpage */
12810 + .write_begin = aufs_write_begin,
12811 + .write_end = aufs_write_end,
12812 + /* no bmap, no block device */
12813 + .invalidatepage = aufs_invalidatepage,
12814 + .releasepage = aufs_releasepage,
12815 + /* is fallback_migrate_page ok? */
12816 + /* .migratepage = aufs_migratepage, */
12817 + .launder_page = aufs_launder_page,
12818 + .is_partially_uptodate = aufs_is_partially_uptodate,
12819 + .is_dirty_writeback = aufs_is_dirty_writeback,
12820 + .error_remove_page = aufs_error_remove_page,
12821 + .swap_activate = aufs_swap_activate,
12822 + .swap_deactivate = aufs_swap_deactivate
12823 +#endif /* CONFIG_AUFS_DEBUG */
12825 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
12826 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
12827 +++ linux/fs/aufs/file.h 2015-09-24 10:47:58.251386326 +0200
12830 + * Copyright (C) 2005-2015 Junjiro R. Okajima
12832 + * This program, aufs is free software; you can redistribute it and/or modify
12833 + * it under the terms of the GNU General Public License as published by
12834 + * the Free Software Foundation; either version 2 of the License, or
12835 + * (at your option) any later version.
12837 + * This program is distributed in the hope that it will be useful,
12838 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12839 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12840 + * GNU General Public License for more details.
12842 + * You should have received a copy of the GNU General Public License
12843 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
12847 + * file operations
12850 +#ifndef __AUFS_FILE_H__
12851 +#define __AUFS_FILE_H__
12855 +#include <linux/file.h>
12856 +#include <linux/fs.h>
12857 +#include <linux/poll.h>
12858 +#include "rwsem.h"
12862 + struct file *hf_file;
12863 + struct au_branch *hf_br;
12868 + aufs_bindex_t fd_bbot;
12869 + aufs_bindex_t fd_nent;
12870 + struct au_vdir *fd_vdir_cache;
12871 + struct au_hfile fd_hfile[];
12874 +static inline int au_fidir_sz(int nent)
12876 + AuDebugOn(nent < 0);
12877 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
12881 + atomic_t fi_generation;
12883 + struct au_rwsem fi_rwsem;
12884 + aufs_bindex_t fi_btop;
12886 + /* do not union them */
12887 + struct { /* for non-dir */
12888 + struct au_hfile fi_htop;
12889 + atomic_t fi_mmapped;
12891 + struct au_fidir *fi_hdir; /* for dir only */
12893 + struct hlist_node fi_hlist;
12894 + struct file *fi_file; /* very ugly */
12895 +} ____cacheline_aligned_in_smp;
12897 +/* ---------------------------------------------------------------------- */
12900 +extern const struct address_space_operations aufs_aop;
12901 +unsigned int au_file_roflags(unsigned int flags);
12902 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
12903 + struct file *file, int force_wr);
12904 +struct au_do_open_args {
12906 + int (*open)(struct file *file, int flags,
12907 + struct file *h_file);
12908 + struct au_fidir *fidir;
12909 + struct file *h_file;
12911 +int au_do_open(struct file *file, struct au_do_open_args *args);
12912 +int au_reopen_nondir(struct file *file);
12914 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
12915 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
12917 +int au_do_flush(struct file *file, fl_owner_t id,
12918 + int (*flush)(struct file *file, fl_owner_t id));
12921 +#ifdef CONFIG_AUFS_POLL
12922 +unsigned int aufs_poll(struct file *file, poll_table *wait);
12925 +#ifdef CONFIG_AUFS_BR_HFSPLUS
12927 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
12929 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
12930 + struct file *h_file);
12932 +AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
12933 + aufs_bindex_t bindex, int force_wr)
12934 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
12935 + struct file *h_file);
12939 +extern const struct file_operations aufs_file_fop;
12940 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
12941 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
12942 +struct file *au_read_pre(struct file *file, int keep_fi);
12945 +void au_hfput(struct au_hfile *hf, struct file *file);
12946 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
12947 + struct file *h_file);
12949 +void au_update_figen(struct file *file);
12950 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
12951 +int au_fidir_realloc(struct au_finfo *finfo, int nbr);
12953 +void au_fi_init_once(void *_fi);
12954 +void au_finfo_fin(struct file *file);
12955 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
12958 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
12959 +#ifdef CONFIG_COMPAT
12960 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
12961 + unsigned long arg);
12962 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
12963 + unsigned long arg);
12966 +/* ---------------------------------------------------------------------- */
12968 +static inline struct au_finfo *au_fi(struct file *file)
12970 + return file->private_data;
12973 +/* ---------------------------------------------------------------------- */
12976 + * fi_read_lock, fi_write_lock,
12977 + * fi_read_unlock, fi_write_unlock, fi_downgrade_lock
12979 +AuSimpleRwsemFuncs(fi, struct file *f, &au_fi(f)->fi_rwsem);
12981 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
12982 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
12983 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
12985 +/* ---------------------------------------------------------------------- */
12987 +/* todo: hard/soft set? */
12988 +static inline aufs_bindex_t au_fbstart(struct file *file)
12990 + FiMustAnyLock(file);
12991 + return au_fi(file)->fi_btop;
12994 +static inline aufs_bindex_t au_fbend_dir(struct file *file)
12996 + FiMustAnyLock(file);
12997 + AuDebugOn(!au_fi(file)->fi_hdir);
12998 + return au_fi(file)->fi_hdir->fd_bbot;
13001 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
13003 + FiMustAnyLock(file);
13004 + AuDebugOn(!au_fi(file)->fi_hdir);
13005 + return au_fi(file)->fi_hdir->fd_vdir_cache;
13008 +static inline void au_set_fbstart(struct file *file, aufs_bindex_t bindex)
13010 + FiMustWriteLock(file);
13011 + au_fi(file)->fi_btop = bindex;
13014 +static inline void au_set_fbend_dir(struct file *file, aufs_bindex_t bindex)
13016 + FiMustWriteLock(file);
13017 + AuDebugOn(!au_fi(file)->fi_hdir);
13018 + au_fi(file)->fi_hdir->fd_bbot = bindex;
13021 +static inline void au_set_fvdir_cache(struct file *file,
13022 + struct au_vdir *vdir_cache)
13024 + FiMustWriteLock(file);
13025 + AuDebugOn(!au_fi(file)->fi_hdir);
13026 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
13029 +static inline struct file *au_hf_top(struct file *file)
13031 + FiMustAnyLock(file);
13032 + AuDebugOn(au_fi(file)->fi_hdir);
13033 + return au_fi(file)->fi_htop.hf_file;
13036 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
13038 + FiMustAnyLock(file);
13039 + AuDebugOn(!au_fi(file)->fi_hdir);
13040 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
13043 +/* todo: memory barrier? */
13044 +static inline unsigned int au_figen(struct file *f)
13046 + return atomic_read(&au_fi(f)->fi_generation);
13049 +static inline void au_set_mmapped(struct file *f)
13051 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
13053 + pr_warn("fi_mmapped wrapped around\n");
13054 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
13058 +static inline void au_unset_mmapped(struct file *f)
13060 + atomic_dec(&au_fi(f)->fi_mmapped);
13063 +static inline int au_test_mmapped(struct file *f)
13065 + return atomic_read(&au_fi(f)->fi_mmapped);
13068 +/* customize vma->vm_file */
13070 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
13071 + struct file *file)
13075 + f = vma->vm_file;
13077 + vma->vm_file = file;
13082 +#define AuDbgVmRegion(file, vma) do {} while (0)
13084 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
13085 + struct file *file)
13087 + au_do_vm_file_reset(vma, file);
13090 +#define AuDbgVmRegion(file, vma) \
13091 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
13093 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
13094 + struct file *file)
13098 + au_do_vm_file_reset(vma, file);
13099 + f = vma->vm_region->vm_file;
13101 + vma->vm_region->vm_file = file;
13104 +#endif /* CONFIG_MMU */
13106 +/* handle vma->vm_prfile */
13107 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
13108 + struct file *file)
13111 + vma->vm_prfile = file;
13112 +#ifndef CONFIG_MMU
13114 + vma->vm_region->vm_prfile = file;
13118 +#endif /* __KERNEL__ */
13119 +#endif /* __AUFS_FILE_H__ */
13120 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
13121 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
13122 +++ linux/fs/aufs/finfo.c 2015-09-24 10:47:58.251386326 +0200
13125 + * Copyright (C) 2005-2015 Junjiro R. Okajima
13127 + * This program, aufs is free software; you can redistribute it and/or modify
13128 + * it under the terms of the GNU General Public License as published by
13129 + * the Free Software Foundation; either version 2 of the License, or
13130 + * (at your option) any later version.
13132 + * This program is distributed in the hope that it will be useful,
13133 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13134 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13135 + * GNU General Public License for more details.
13137 + * You should have received a copy of the GNU General Public License
13138 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
13142 + * file private data
13147 +void au_hfput(struct au_hfile *hf, struct file *file)
13149 + /* todo: direct access f_flags */
13150 + if (vfsub_file_flags(file) & __FMODE_EXEC)
13151 + allow_write_access(hf->hf_file);
13152 + fput(hf->hf_file);
13153 + hf->hf_file = NULL;
13154 + atomic_dec(&hf->hf_br->br_count);
13155 + hf->hf_br = NULL;
13158 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
13160 + struct au_finfo *finfo = au_fi(file);
13161 + struct au_hfile *hf;
13162 + struct au_fidir *fidir;
13164 + fidir = finfo->fi_hdir;
13166 + AuDebugOn(finfo->fi_btop != bindex);
13167 + hf = &finfo->fi_htop;
13169 + hf = fidir->fd_hfile + bindex;
13171 + if (hf && hf->hf_file)
13172 + au_hfput(hf, file);
13174 + FiMustWriteLock(file);
13175 + AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
13176 + hf->hf_file = val;
13177 + hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
13181 +void au_update_figen(struct file *file)
13183 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
13184 + /* smp_mb(); */ /* atomic_set */
13187 +/* ---------------------------------------------------------------------- */
13189 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
13191 + struct au_fidir *fidir;
13194 + nbr = au_sbend(sb) + 1;
13196 + nbr = 2; /* initial allocate for 2 branches */
13197 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
13199 + fidir->fd_bbot = -1;
13200 + fidir->fd_nent = nbr;
13201 + fidir->fd_vdir_cache = NULL;
13207 +int au_fidir_realloc(struct au_finfo *finfo, int nbr)
13210 + struct au_fidir *fidir, *p;
13212 + AuRwMustWriteLock(&finfo->fi_rwsem);
13213 + fidir = finfo->fi_hdir;
13214 + AuDebugOn(!fidir);
13217 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
13220 + p->fd_nent = nbr;
13221 + finfo->fi_hdir = p;
13228 +/* ---------------------------------------------------------------------- */
13230 +void au_finfo_fin(struct file *file)
13232 + struct au_finfo *finfo;
13234 + au_nfiles_dec(file->f_path.dentry->d_sb);
13236 + finfo = au_fi(file);
13237 + AuDebugOn(finfo->fi_hdir);
13238 + AuRwDestroy(&finfo->fi_rwsem);
13239 + au_cache_free_finfo(finfo);
13242 +void au_fi_init_once(void *_finfo)
13244 + struct au_finfo *finfo = _finfo;
13245 + static struct lock_class_key aufs_fi;
13247 + au_rw_init(&finfo->fi_rwsem);
13248 + au_rw_class(&finfo->fi_rwsem, &aufs_fi);
13251 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
13254 + struct au_finfo *finfo;
13255 + struct dentry *dentry;
13258 + dentry = file->f_path.dentry;
13259 + finfo = au_cache_alloc_finfo();
13260 + if (unlikely(!finfo))
13264 + au_nfiles_inc(dentry->d_sb);
13265 + /* verbose coding for lock class name */
13267 + au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcNonDir_FIINFO);
13269 + au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcDir_FIINFO);
13270 + au_rw_write_lock(&finfo->fi_rwsem);
13271 + finfo->fi_btop = -1;
13272 + finfo->fi_hdir = fidir;
13273 + atomic_set(&finfo->fi_generation, au_digen(dentry));
13274 + /* smp_mb(); */ /* atomic_set */
13276 + file->private_data = finfo;
13281 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
13282 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
13283 +++ linux/fs/aufs/f_op.c 2015-09-24 10:47:58.251386326 +0200
13286 + * Copyright (C) 2005-2015 Junjiro R. Okajima
13288 + * This program, aufs is free software; you can redistribute it and/or modify
13289 + * it under the terms of the GNU General Public License as published by
13290 + * the Free Software Foundation; either version 2 of the License, or
13291 + * (at your option) any later version.
13293 + * This program is distributed in the hope that it will be useful,
13294 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13295 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13296 + * GNU General Public License for more details.
13298 + * You should have received a copy of the GNU General Public License
13299 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
13303 + * file and vm operations
13306 +#include <linux/aio.h>
13307 +#include <linux/fs_stack.h>
13308 +#include <linux/mman.h>
13309 +#include <linux/security.h>
13312 +int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
13315 + aufs_bindex_t bindex;
13316 + struct dentry *dentry;
13317 + struct au_finfo *finfo;
13318 + struct inode *h_inode;
13320 + FiMustWriteLock(file);
13323 + dentry = file->f_path.dentry;
13324 + AuDebugOn(IS_ERR_OR_NULL(dentry));
13325 + finfo = au_fi(file);
13326 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
13327 + atomic_set(&finfo->fi_mmapped, 0);
13328 + bindex = au_dbstart(dentry);
13330 + h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
13332 + get_file(h_file);
13333 + if (IS_ERR(h_file))
13334 + err = PTR_ERR(h_file);
13336 + if ((flags & __O_TMPFILE)
13337 + && !(flags & O_EXCL)) {
13338 + h_inode = file_inode(h_file);
13339 + spin_lock(&h_inode->i_lock);
13340 + h_inode->i_state |= I_LINKABLE;
13341 + spin_unlock(&h_inode->i_lock);
13343 + au_set_fbstart(file, bindex);
13344 + au_set_h_fptr(file, bindex, h_file);
13345 + au_update_figen(file);
13346 + /* todo: necessary? */
13347 + /* file->f_ra = h_file->f_ra; */
13353 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
13354 + struct file *file)
13357 + struct super_block *sb;
13358 + struct au_do_open_args args = {
13359 + .open = au_do_open_nondir
13362 + AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
13363 + file, vfsub_file_flags(file), file->f_mode);
13365 + sb = file->f_path.dentry->d_sb;
13366 + si_read_lock(sb, AuLock_FLUSH);
13367 + err = au_do_open(file, &args);
13368 + si_read_unlock(sb);
13372 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
13374 + struct au_finfo *finfo;
13375 + aufs_bindex_t bindex;
13377 + finfo = au_fi(file);
13378 + au_sphl_del(&finfo->fi_hlist,
13379 + &au_sbi(file->f_path.dentry->d_sb)->si_files);
13380 + bindex = finfo->fi_btop;
13382 + au_set_h_fptr(file, bindex, NULL);
13384 + au_finfo_fin(file);
13388 +/* ---------------------------------------------------------------------- */
13390 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
13393 + struct file *h_file;
13396 + h_file = au_hf_top(file);
13398 + err = vfsub_flush(h_file, id);
13402 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
13404 + return au_do_flush(file, id, au_do_flush_nondir);
13407 +/* ---------------------------------------------------------------------- */
13409 + * read and write functions acquire [fdi]_rwsem once, but release before
13410 + * mmap_sem. This is because to stop a race condition between mmap(2).
13411 + * Releasing these aufs-rwsem should be safe, no branch-mamagement (by keeping
13412 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
13413 + * read functions after [fdi]_rwsem are released, but it should be harmless.
13416 +/* Callers should call au_read_post() or fput() in the end */
13417 +struct file *au_read_pre(struct file *file, int keep_fi)
13419 + struct file *h_file;
13422 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
13424 + di_read_unlock(file->f_path.dentry, AuLock_IR);
13425 + h_file = au_hf_top(file);
13426 + get_file(h_file);
13428 + fi_read_unlock(file);
13430 + h_file = ERR_PTR(err);
13435 +static void au_read_post(struct inode *inode, struct file *h_file)
13437 + /* update without lock, I don't think it a problem */
13438 + fsstack_copy_attr_atime(inode, file_inode(h_file));
13442 +struct au_write_pre {
13444 + aufs_bindex_t bstart;
13448 + * return with iinfo is write-locked
13449 + * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
13452 +static struct file *au_write_pre(struct file *file, int do_ready,
13453 + struct au_write_pre *wpre)
13455 + struct file *h_file;
13456 + struct dentry *dentry;
13458 + struct au_pin pin;
13460 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
13461 + h_file = ERR_PTR(err);
13462 + if (unlikely(err))
13465 + dentry = file->f_path.dentry;
13467 + err = au_ready_to_write(file, -1, &pin);
13468 + if (unlikely(err)) {
13469 + h_file = ERR_PTR(err);
13470 + di_write_unlock(dentry);
13475 + di_downgrade_lock(dentry, /*flags*/0);
13477 + wpre->bstart = au_fbstart(file);
13478 + h_file = au_hf_top(file);
13479 + get_file(h_file);
13481 + wpre->blks = file_inode(h_file)->i_blocks;
13484 + di_read_unlock(dentry, /*flags*/0);
13487 + fi_write_unlock(file);
13492 +static void au_write_post(struct inode *inode, struct file *h_file,
13493 + struct au_write_pre *wpre, ssize_t written)
13495 + struct inode *h_inode;
13497 + au_cpup_attr_timesizes(inode);
13498 + AuDebugOn(au_ibstart(inode) != wpre->bstart);
13499 + h_inode = file_inode(h_file);
13500 + inode->i_mode = h_inode->i_mode;
13501 + ii_write_unlock(inode);
13504 + /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
13506 + au_fhsm_wrote(inode->i_sb, wpre->bstart,
13507 + /*force*/h_inode->i_blocks > wpre->blks);
13510 +static ssize_t aufs_read(struct file *file, char __user *buf, size_t count,
13514 + struct inode *inode;
13515 + struct file *h_file;
13516 + struct super_block *sb;
13518 + inode = file_inode(file);
13519 + sb = inode->i_sb;
13520 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13522 + h_file = au_read_pre(file, /*keep_fi*/0);
13523 + err = PTR_ERR(h_file);
13524 + if (IS_ERR(h_file))
13527 + /* filedata may be obsoleted by concurrent copyup, but no problem */
13528 + err = vfsub_read_u(h_file, buf, count, ppos);
13529 + /* todo: necessary? */
13530 + /* file->f_ra = h_file->f_ra; */
13531 + au_read_post(inode, h_file);
13534 + si_read_unlock(sb);
13539 + * todo: very ugly
13540 + * it locks both of i_mutex and si_rwsem for read in safe.
13541 + * if the plink maintenance mode continues forever (that is the problem),
13542 + * may loop forever.
13544 +static void au_mtx_and_read_lock(struct inode *inode)
13547 + struct super_block *sb = inode->i_sb;
13550 + mutex_lock(&inode->i_mutex);
13551 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
13554 + mutex_unlock(&inode->i_mutex);
13555 + si_read_lock(sb, AuLock_NOPLMW);
13556 + si_read_unlock(sb);
13560 +static ssize_t aufs_write(struct file *file, const char __user *ubuf,
13561 + size_t count, loff_t *ppos)
13564 + struct au_write_pre wpre;
13565 + struct inode *inode;
13566 + struct file *h_file;
13567 + char __user *buf = (char __user *)ubuf;
13569 + inode = file_inode(file);
13570 + au_mtx_and_read_lock(inode);
13572 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13573 + err = PTR_ERR(h_file);
13574 + if (IS_ERR(h_file))
13577 + err = vfsub_write_u(h_file, buf, count, ppos);
13578 + au_write_post(inode, h_file, &wpre, err);
13581 + si_read_unlock(inode->i_sb);
13582 + mutex_unlock(&inode->i_mutex);
13586 +static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
13587 + struct iov_iter *iov_iter)
13590 + struct file *file;
13591 + ssize_t (*iter)(struct kiocb *, struct iov_iter *);
13593 + err = security_file_permission(h_file, rw);
13594 + if (unlikely(err))
13599 + if (rw == MAY_READ)
13600 + iter = h_file->f_op->read_iter;
13601 + else if (rw == MAY_WRITE)
13602 + iter = h_file->f_op->write_iter;
13604 + file = kio->ki_filp;
13605 + kio->ki_filp = h_file;
13608 + err = iter(kio, iov_iter);
13611 + /* currently there is no such fs */
13613 + kio->ki_filp = file;
13619 +static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
13622 + struct file *file, *h_file;
13623 + struct inode *inode;
13624 + struct super_block *sb;
13626 + file = kio->ki_filp;
13627 + inode = file_inode(file);
13628 + sb = inode->i_sb;
13629 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13631 + h_file = au_read_pre(file, /*keep_fi*/0);
13632 + err = PTR_ERR(h_file);
13633 + if (IS_ERR(h_file))
13636 + err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
13637 + /* todo: necessary? */
13638 + /* file->f_ra = h_file->f_ra; */
13639 + au_read_post(inode, h_file);
13642 + si_read_unlock(sb);
13646 +static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
13649 + struct au_write_pre wpre;
13650 + struct inode *inode;
13651 + struct file *file, *h_file;
13653 + file = kio->ki_filp;
13654 + inode = file_inode(file);
13655 + au_mtx_and_read_lock(inode);
13657 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13658 + err = PTR_ERR(h_file);
13659 + if (IS_ERR(h_file))
13662 + err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
13663 + au_write_post(inode, h_file, &wpre, err);
13666 + si_read_unlock(inode->i_sb);
13667 + mutex_unlock(&inode->i_mutex);
13671 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
13672 + struct pipe_inode_info *pipe, size_t len,
13673 + unsigned int flags)
13676 + struct file *h_file;
13677 + struct inode *inode;
13678 + struct super_block *sb;
13680 + inode = file_inode(file);
13681 + sb = inode->i_sb;
13682 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13684 + h_file = au_read_pre(file, /*keep_fi*/1);
13685 + err = PTR_ERR(h_file);
13686 + if (IS_ERR(h_file))
13689 + if (au_test_loopback_kthread()) {
13690 + au_warn_loopback(h_file->f_path.dentry->d_sb);
13691 + if (file->f_mapping != h_file->f_mapping) {
13692 + file->f_mapping = h_file->f_mapping;
13693 + smp_mb(); /* unnecessary? */
13696 + fi_read_unlock(file);
13698 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
13699 + /* todo: necessasry? */
13700 + /* file->f_ra = h_file->f_ra; */
13701 + au_read_post(inode, h_file);
13704 + si_read_unlock(sb);
13709 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
13710 + size_t len, unsigned int flags)
13713 + struct au_write_pre wpre;
13714 + struct inode *inode;
13715 + struct file *h_file;
13717 + inode = file_inode(file);
13718 + au_mtx_and_read_lock(inode);
13720 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13721 + err = PTR_ERR(h_file);
13722 + if (IS_ERR(h_file))
13725 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
13726 + au_write_post(inode, h_file, &wpre, err);
13729 + si_read_unlock(inode->i_sb);
13730 + mutex_unlock(&inode->i_mutex);
13734 +static long aufs_fallocate(struct file *file, int mode, loff_t offset,
13738 + struct au_write_pre wpre;
13739 + struct inode *inode;
13740 + struct file *h_file;
13742 + inode = file_inode(file);
13743 + au_mtx_and_read_lock(inode);
13745 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13746 + err = PTR_ERR(h_file);
13747 + if (IS_ERR(h_file))
13751 + err = vfs_fallocate(h_file, mode, offset, len);
13753 + au_write_post(inode, h_file, &wpre, /*written*/1);
13756 + si_read_unlock(inode->i_sb);
13757 + mutex_unlock(&inode->i_mutex);
13761 +/* ---------------------------------------------------------------------- */
13764 + * The locking order around current->mmap_sem.
13765 + * - in most and regular cases
13766 + * file I/O syscall -- aufs_read() or something
13767 + * -- si_rwsem for read -- mmap_sem
13768 + * (Note that [fdi]i_rwsem are released before mmap_sem).
13770 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
13771 + * This AB-BA order is definitly bad, but is not a problem since "si_rwsem for
13772 + * read" allows muliple processes to acquire it and [fdi]i_rwsem are not held in
13773 + * file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
13774 + * It means that when aufs acquires si_rwsem for write, the process should never
13775 + * acquire mmap_sem.
13777 + * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
13778 + * problem either since any directory is not able to be mmap-ed.
13779 + * The similar scenario is applied to aufs_readlink() too.
13782 +#if 0 /* stop calling security_file_mmap() */
13783 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
13784 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
13786 +static unsigned long au_arch_prot_conv(unsigned long flags)
13788 + /* currently ppc64 only */
13789 +#ifdef CONFIG_PPC64
13790 + /* cf. linux/arch/powerpc/include/asm/mman.h */
13791 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
13792 + return AuConv_VM_PROT(flags, SAO);
13794 + AuDebugOn(arch_calc_vm_prot_bits(-1));
13799 +static unsigned long au_prot_conv(unsigned long flags)
13801 + return AuConv_VM_PROT(flags, READ)
13802 + | AuConv_VM_PROT(flags, WRITE)
13803 + | AuConv_VM_PROT(flags, EXEC)
13804 + | au_arch_prot_conv(flags);
13807 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
13808 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
13810 +static unsigned long au_flag_conv(unsigned long flags)
13812 + return AuConv_VM_MAP(flags, GROWSDOWN)
13813 + | AuConv_VM_MAP(flags, DENYWRITE)
13814 + | AuConv_VM_MAP(flags, LOCKED);
13818 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
13821 + const unsigned char wlock
13822 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
13823 + struct super_block *sb;
13824 + struct file *h_file;
13825 + struct inode *inode;
13827 + AuDbgVmRegion(file, vma);
13829 + inode = file_inode(file);
13830 + sb = inode->i_sb;
13832 + si_read_lock(sb, AuLock_NOPLMW);
13834 + h_file = au_write_pre(file, wlock, /*wpre*/NULL);
13836 + err = PTR_ERR(h_file);
13837 + if (IS_ERR(h_file))
13841 + au_set_mmapped(file);
13842 + au_vm_file_reset(vma, h_file);
13844 + * we cannot call security_mmap_file() here since it may acquire
13845 + * mmap_sem or i_mutex.
13847 + * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
13848 + * au_flag_conv(vma->vm_flags));
13851 + err = h_file->f_op->mmap(h_file, vma);
13853 + au_vm_prfile_set(vma, file);
13854 + fsstack_copy_attr_atime(inode, file_inode(h_file));
13855 + goto out_fput; /* success */
13857 + au_unset_mmapped(file);
13858 + au_vm_file_reset(vma, file);
13862 + ii_write_unlock(inode);
13867 + si_read_unlock(sb);
13873 +/* ---------------------------------------------------------------------- */
13875 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
13879 + struct au_write_pre wpre;
13880 + struct inode *inode;
13881 + struct file *h_file;
13883 + err = 0; /* -EBADF; */ /* posix? */
13884 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
13887 + inode = file_inode(file);
13888 + au_mtx_and_read_lock(inode);
13890 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13891 + err = PTR_ERR(h_file);
13892 + if (IS_ERR(h_file))
13895 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
13896 + au_write_post(inode, h_file, &wpre, /*written*/0);
13899 + si_read_unlock(inode->i_sb);
13900 + mutex_unlock(&inode->i_mutex);
13905 +/* no one supports this operation, currently */
13907 +static int aufs_aio_fsync_nondir(struct kiocb *kio, int datasync)
13910 + struct au_write_pre wpre;
13911 + struct inode *inode;
13912 + struct file *file, *h_file;
13914 + err = 0; /* -EBADF; */ /* posix? */
13915 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
13918 + file = kio->ki_filp;
13919 + inode = file_inode(file);
13920 + au_mtx_and_read_lock(inode);
13922 + h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13923 + err = PTR_ERR(h_file);
13924 + if (IS_ERR(h_file))
13928 + h_file = au_hf_top(file);
13929 + if (h_file->f_op->aio_fsync) {
13930 + struct mutex *h_mtx;
13932 + h_mtx = &file_inode(h_file)->i_mutex;
13933 + if (!is_sync_kiocb(kio)) {
13934 + get_file(h_file);
13937 + kio->ki_filp = h_file;
13938 + err = h_file->f_op->aio_fsync(kio, datasync);
13939 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
13941 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL);
13943 + mutex_unlock(h_mtx);
13945 + au_write_post(inode, h_file, &wpre, /*written*/0);
13948 + si_read_unlock(inode->sb);
13949 + mutex_unlock(&inode->i_mutex);
13955 +static int aufs_fasync(int fd, struct file *file, int flag)
13958 + struct file *h_file;
13959 + struct super_block *sb;
13961 + sb = file->f_path.dentry->d_sb;
13962 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13964 + h_file = au_read_pre(file, /*keep_fi*/0);
13965 + err = PTR_ERR(h_file);
13966 + if (IS_ERR(h_file))
13969 + if (h_file->f_op->fasync)
13970 + err = h_file->f_op->fasync(fd, h_file, flag);
13971 + fput(h_file); /* instead of au_read_post() */
13974 + si_read_unlock(sb);
13978 +/* ---------------------------------------------------------------------- */
13980 +/* no one supports this operation, currently */
13982 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
13983 + size_t len, loff_t *pos, int more)
13988 +/* ---------------------------------------------------------------------- */
13990 +const struct file_operations aufs_file_fop = {
13991 + .owner = THIS_MODULE,
13993 + .llseek = default_llseek,
13995 + .read = aufs_read,
13996 + .write = aufs_write,
13997 + .read_iter = aufs_read_iter,
13998 + .write_iter = aufs_write_iter,
14000 +#ifdef CONFIG_AUFS_POLL
14001 + .poll = aufs_poll,
14003 + .unlocked_ioctl = aufs_ioctl_nondir,
14004 +#ifdef CONFIG_COMPAT
14005 + .compat_ioctl = aufs_compat_ioctl_nondir,
14007 + .mmap = aufs_mmap,
14008 + .open = aufs_open_nondir,
14009 + .flush = aufs_flush_nondir,
14010 + .release = aufs_release_nondir,
14011 + .fsync = aufs_fsync_nondir,
14012 + /* .aio_fsync = aufs_aio_fsync_nondir, */
14013 + .fasync = aufs_fasync,
14014 + /* .sendpage = aufs_sendpage, */
14015 + .splice_write = aufs_splice_write,
14016 + .splice_read = aufs_splice_read,
14018 + .aio_splice_write = aufs_aio_splice_write,
14019 + .aio_splice_read = aufs_aio_splice_read,
14021 + .fallocate = aufs_fallocate
14023 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
14024 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
14025 +++ linux/fs/aufs/fstype.h 2015-09-24 10:47:58.254719746 +0200
14028 + * Copyright (C) 2005-2015 Junjiro R. Okajima
14030 + * This program, aufs is free software; you can redistribute it and/or modify
14031 + * it under the terms of the GNU General Public License as published by
14032 + * the Free Software Foundation; either version 2 of the License, or
14033 + * (at your option) any later version.
14035 + * This program is distributed in the hope that it will be useful,
14036 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14037 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14038 + * GNU General Public License for more details.
14040 + * You should have received a copy of the GNU General Public License
14041 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14045 + * judging filesystem type
14048 +#ifndef __AUFS_FSTYPE_H__
14049 +#define __AUFS_FSTYPE_H__
14053 +#include <linux/fs.h>
14054 +#include <linux/magic.h>
14055 +#include <linux/romfs_fs.h>
14056 +#include <linux/nfs_fs.h>
14058 +static inline int au_test_aufs(struct super_block *sb)
14060 + return sb->s_magic == AUFS_SUPER_MAGIC;
14063 +static inline const char *au_sbtype(struct super_block *sb)
14065 + return sb->s_type->name;
14068 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
14070 +#if defined(CONFIG_ISO9660_FS) || defined(CONFIG_ISO9660_FS_MODULE)
14071 + return sb->s_magic == ISOFS_SUPER_MAGIC;
14077 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
14079 +#if defined(CONFIG_ROMFS_FS) || defined(CONFIG_ROMFS_FS_MODULE)
14080 + return sb->s_magic == ROMFS_MAGIC;
14086 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
14088 +#if defined(CONFIG_CRAMFS) || defined(CONFIG_CRAMFS_MODULE)
14089 + return sb->s_magic == CRAMFS_MAGIC;
14094 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
14096 +#if defined(CONFIG_NFS_FS) || defined(CONFIG_NFS_FS_MODULE)
14097 + return sb->s_magic == NFS_SUPER_MAGIC;
14103 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
14105 +#if defined(CONFIG_FUSE_FS) || defined(CONFIG_FUSE_FS_MODULE)
14106 + return sb->s_magic == FUSE_SUPER_MAGIC;
14112 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
14114 +#if defined(CONFIG_XFS_FS) || defined(CONFIG_XFS_FS_MODULE)
14115 + return sb->s_magic == XFS_SB_MAGIC;
14121 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
14123 +#ifdef CONFIG_TMPFS
14124 + return sb->s_magic == TMPFS_MAGIC;
14130 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
14132 +#if defined(CONFIG_ECRYPT_FS) || defined(CONFIG_ECRYPT_FS_MODULE)
14133 + return !strcmp(au_sbtype(sb), "ecryptfs");
14139 +static inline int au_test_ramfs(struct super_block *sb)
14141 + return sb->s_magic == RAMFS_MAGIC;
14144 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
14146 +#if defined(CONFIG_UBIFS_FS) || defined(CONFIG_UBIFS_FS_MODULE)
14147 + return sb->s_magic == UBIFS_SUPER_MAGIC;
14153 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
14155 +#ifdef CONFIG_PROC_FS
14156 + return sb->s_magic == PROC_SUPER_MAGIC;
14162 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
14164 +#ifdef CONFIG_SYSFS
14165 + return sb->s_magic == SYSFS_MAGIC;
14171 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
14173 +#if defined(CONFIG_CONFIGFS_FS) || defined(CONFIG_CONFIGFS_FS_MODULE)
14174 + return sb->s_magic == CONFIGFS_MAGIC;
14180 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
14182 +#if defined(CONFIG_MINIX_FS) || defined(CONFIG_MINIX_FS_MODULE)
14183 + return sb->s_magic == MINIX3_SUPER_MAGIC
14184 + || sb->s_magic == MINIX2_SUPER_MAGIC
14185 + || sb->s_magic == MINIX2_SUPER_MAGIC2
14186 + || sb->s_magic == MINIX_SUPER_MAGIC
14187 + || sb->s_magic == MINIX_SUPER_MAGIC2;
14193 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
14195 +#if defined(CONFIG_FAT_FS) || defined(CONFIG_FAT_FS_MODULE)
14196 + return sb->s_magic == MSDOS_SUPER_MAGIC;
14202 +static inline int au_test_msdos(struct super_block *sb)
14204 + return au_test_fat(sb);
14207 +static inline int au_test_vfat(struct super_block *sb)
14209 + return au_test_fat(sb);
14212 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
14214 +#ifdef CONFIG_SECURITYFS
14215 + return sb->s_magic == SECURITYFS_MAGIC;
14221 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
14223 +#if defined(CONFIG_SQUASHFS) || defined(CONFIG_SQUASHFS_MODULE)
14224 + return sb->s_magic == SQUASHFS_MAGIC;
14230 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
14232 +#if defined(CONFIG_BTRFS_FS) || defined(CONFIG_BTRFS_FS_MODULE)
14233 + return sb->s_magic == BTRFS_SUPER_MAGIC;
14239 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
14241 +#if defined(CONFIG_XENFS) || defined(CONFIG_XENFS_MODULE)
14242 + return sb->s_magic == XENFS_SUPER_MAGIC;
14248 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
14250 +#ifdef CONFIG_DEBUG_FS
14251 + return sb->s_magic == DEBUGFS_MAGIC;
14257 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
14259 +#if defined(CONFIG_NILFS) || defined(CONFIG_NILFS_MODULE)
14260 + return sb->s_magic == NILFS_SUPER_MAGIC;
14266 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
14268 +#if defined(CONFIG_HFSPLUS_FS) || defined(CONFIG_HFSPLUS_FS_MODULE)
14269 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
14275 +/* ---------------------------------------------------------------------- */
14277 + * they can't be an aufs branch.
14279 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
14282 +#ifndef CONFIG_AUFS_BR_RAMFS
14283 + au_test_ramfs(sb) ||
14285 + au_test_procfs(sb)
14286 + || au_test_sysfs(sb)
14287 + || au_test_configfs(sb)
14288 + || au_test_debugfs(sb)
14289 + || au_test_securityfs(sb)
14290 + || au_test_xenfs(sb)
14291 + || au_test_ecryptfs(sb)
14292 + /* || !strcmp(au_sbtype(sb), "unionfs") */
14293 + || au_test_aufs(sb); /* will be supported in next version */
14296 +static inline int au_test_fs_remote(struct super_block *sb)
14298 + return !au_test_tmpfs(sb)
14299 +#ifdef CONFIG_AUFS_BR_RAMFS
14300 + && !au_test_ramfs(sb)
14302 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
14305 +/* ---------------------------------------------------------------------- */
14308 + * Note: these functions (below) are created after reading ->getattr() in all
14309 + * filesystems under linux/fs. it means we have to do so in every update...
14313 + * some filesystems require getattr to refresh the inode attributes before
14315 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
14316 + * and leave the work for d_revalidate()
14318 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
14320 + return au_test_nfs(sb)
14321 + || au_test_fuse(sb)
14322 + /* || au_test_btrfs(sb) */ /* untested */
14327 + * filesystems which don't maintain i_size or i_blocks.
14329 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
14331 + return au_test_xfs(sb)
14332 + || au_test_btrfs(sb)
14333 + || au_test_ubifs(sb)
14334 + || au_test_hfsplus(sb) /* maintained, but incorrect */
14335 + /* || au_test_minix(sb) */ /* untested */
14340 + * filesystems which don't store the correct value in some of their inode
14343 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
14345 + return au_test_fs_bad_iattr_size(sb)
14346 + || au_test_fat(sb)
14347 + || au_test_msdos(sb)
14348 + || au_test_vfat(sb);
14351 +/* they don't check i_nlink in link(2) */
14352 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
14354 + return au_test_tmpfs(sb)
14355 +#ifdef CONFIG_AUFS_BR_RAMFS
14356 + || au_test_ramfs(sb)
14358 + || au_test_ubifs(sb)
14359 + || au_test_hfsplus(sb);
14363 + * filesystems which sets S_NOATIME and S_NOCMTIME.
14365 +static inline int au_test_fs_notime(struct super_block *sb)
14367 + return au_test_nfs(sb)
14368 + || au_test_fuse(sb)
14369 + || au_test_ubifs(sb)
14373 +/* temporary support for i#1 in cramfs */
14374 +static inline int au_test_fs_unique_ino(struct inode *inode)
14376 + if (au_test_cramfs(inode->i_sb))
14377 + return inode->i_ino != 1;
14381 +/* ---------------------------------------------------------------------- */
14384 + * the filesystem where the xino files placed must support i/o after unlink and
14385 + * maintain i_size and i_blocks.
14387 +static inline int au_test_fs_bad_xino(struct super_block *sb)
14389 + return au_test_fs_remote(sb)
14390 + || au_test_fs_bad_iattr_size(sb)
14391 + /* don't want unnecessary work for xino */
14392 + || au_test_aufs(sb)
14393 + || au_test_ecryptfs(sb)
14394 + || au_test_nilfs(sb);
14397 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
14399 + return au_test_tmpfs(sb)
14400 + || au_test_ramfs(sb);
14404 + * test if the @sb is real-readonly.
14406 +static inline int au_test_fs_rr(struct super_block *sb)
14408 + return au_test_squashfs(sb)
14409 + || au_test_iso9660(sb)
14410 + || au_test_cramfs(sb)
14411 + || au_test_romfs(sb);
14415 + * test if the @inode is nfs with 'noacl' option
14416 + * NFS always sets MS_POSIXACL regardless its mount option 'noacl.'
14418 +static inline int au_test_nfs_noacl(struct inode *inode)
14420 + return au_test_nfs(inode->i_sb)
14421 + /* && IS_POSIXACL(inode) */
14422 + && !nfs_server_capable(inode, NFS_CAP_ACLS);
14425 +#endif /* __KERNEL__ */
14426 +#endif /* __AUFS_FSTYPE_H__ */
14427 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
14428 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
14429 +++ linux/fs/aufs/hfsnotify.c 2015-09-24 10:47:58.254719746 +0200
14432 + * Copyright (C) 2005-2015 Junjiro R. Okajima
14434 + * This program, aufs is free software; you can redistribute it and/or modify
14435 + * it under the terms of the GNU General Public License as published by
14436 + * the Free Software Foundation; either version 2 of the License, or
14437 + * (at your option) any later version.
14439 + * This program is distributed in the hope that it will be useful,
14440 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14441 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14442 + * GNU General Public License for more details.
14444 + * You should have received a copy of the GNU General Public License
14445 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14449 + * fsnotify for the lower directories
14454 +/* FS_IN_IGNORED is unnecessary */
14455 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
14456 + | FS_CREATE | FS_EVENT_ON_CHILD);
14457 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
14458 +static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
14460 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
14462 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
14465 + au_cache_free_hnotify(hn);
14466 + smp_mb__before_atomic();
14467 + if (atomic64_dec_and_test(&au_hfsn_ifree))
14468 + wake_up(&au_hfsn_wq);
14471 +static int au_hfsn_alloc(struct au_hinode *hinode)
14474 + struct au_hnotify *hn;
14475 + struct super_block *sb;
14476 + struct au_branch *br;
14477 + struct fsnotify_mark *mark;
14478 + aufs_bindex_t bindex;
14480 + hn = hinode->hi_notify;
14481 + sb = hn->hn_aufs_inode->i_sb;
14482 + bindex = au_br_index(sb, hinode->hi_id);
14483 + br = au_sbr(sb, bindex);
14484 + AuDebugOn(!br->br_hfsn);
14486 + mark = &hn->hn_mark;
14487 + fsnotify_init_mark(mark, au_hfsn_free_mark);
14488 + mark->mask = AuHfsnMask;
14490 + * by udba rename or rmdir, aufs assign a new inode to the known
14491 + * h_inode, so specify 1 to allow dups.
14494 + err = fsnotify_add_mark(mark, br->br_hfsn->hfsn_group, hinode->hi_inode,
14495 + /*mnt*/NULL, /*allow_dups*/1);
14496 + /* even if err */
14497 + fsnotify_put_mark(mark);
14503 +static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
14505 + struct fsnotify_mark *mark;
14506 + unsigned long long ull;
14507 + struct fsnotify_group *group;
14509 + ull = atomic64_inc_return(&au_hfsn_ifree);
14512 + mark = &hn->hn_mark;
14513 + spin_lock(&mark->lock);
14514 + group = mark->group;
14515 + fsnotify_get_group(group);
14516 + spin_unlock(&mark->lock);
14518 + fsnotify_destroy_mark(mark, group);
14519 + fsnotify_put_group(group);
14522 + /* free hn by myself */
14526 +/* ---------------------------------------------------------------------- */
14528 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
14530 + struct fsnotify_mark *mark;
14532 + mark = &hinode->hi_notify->hn_mark;
14533 + spin_lock(&mark->lock);
14535 + AuDebugOn(mark->mask & AuHfsnMask);
14536 + mark->mask |= AuHfsnMask;
14538 + AuDebugOn(!(mark->mask & AuHfsnMask));
14539 + mark->mask &= ~AuHfsnMask;
14541 + spin_unlock(&mark->lock);
14542 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
14545 +/* ---------------------------------------------------------------------- */
14547 +/* #define AuDbgHnotify */
14548 +#ifdef AuDbgHnotify
14549 +static char *au_hfsn_name(u32 mask)
14551 +#ifdef CONFIG_AUFS_DEBUG
14552 +#define test_ret(flag) \
14554 + if (mask & flag) \
14557 + test_ret(FS_ACCESS);
14558 + test_ret(FS_MODIFY);
14559 + test_ret(FS_ATTRIB);
14560 + test_ret(FS_CLOSE_WRITE);
14561 + test_ret(FS_CLOSE_NOWRITE);
14562 + test_ret(FS_OPEN);
14563 + test_ret(FS_MOVED_FROM);
14564 + test_ret(FS_MOVED_TO);
14565 + test_ret(FS_CREATE);
14566 + test_ret(FS_DELETE);
14567 + test_ret(FS_DELETE_SELF);
14568 + test_ret(FS_MOVE_SELF);
14569 + test_ret(FS_UNMOUNT);
14570 + test_ret(FS_Q_OVERFLOW);
14571 + test_ret(FS_IN_IGNORED);
14572 + test_ret(FS_ISDIR);
14573 + test_ret(FS_IN_ONESHOT);
14574 + test_ret(FS_EVENT_ON_CHILD);
14583 +/* ---------------------------------------------------------------------- */
14585 +static void au_hfsn_free_group(struct fsnotify_group *group)
14587 + struct au_br_hfsnotify *hfsn = group->private;
14593 +static int au_hfsn_handle_event(struct fsnotify_group *group,
14594 + struct inode *inode,
14595 + struct fsnotify_mark *inode_mark,
14596 + struct fsnotify_mark *vfsmount_mark,
14597 + u32 mask, void *data, int data_type,
14598 + const unsigned char *file_name, u32 cookie)
14601 + struct au_hnotify *hnotify;
14602 + struct inode *h_dir, *h_inode;
14603 + struct qstr h_child_qstr = QSTR_INIT(file_name, strlen(file_name));
14605 + AuDebugOn(data_type != FSNOTIFY_EVENT_INODE);
14608 + /* if FS_UNMOUNT happens, there must be another bug */
14609 + AuDebugOn(mask & FS_UNMOUNT);
14610 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
14615 +#ifdef AuDbgHnotify
14617 + if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
14618 + || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
14619 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
14620 + h_dir->i_ino, mask, au_hfsn_name(mask),
14621 + AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
14622 + /* WARN_ON(1); */
14627 + AuDebugOn(!inode_mark);
14628 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
14629 + err = au_hnotify(h_dir, hnotify, mask, &h_child_qstr, h_inode);
14635 +static struct fsnotify_ops au_hfsn_ops = {
14636 + .handle_event = au_hfsn_handle_event,
14637 + .free_group_priv = au_hfsn_free_group
14640 +/* ---------------------------------------------------------------------- */
14642 +static void au_hfsn_fin_br(struct au_branch *br)
14644 + struct au_br_hfsnotify *hfsn;
14646 + hfsn = br->br_hfsn;
14649 + fsnotify_put_group(hfsn->hfsn_group);
14654 +static int au_hfsn_init_br(struct au_branch *br, int perm)
14657 + struct fsnotify_group *group;
14658 + struct au_br_hfsnotify *hfsn;
14661 + br->br_hfsn = NULL;
14662 + if (!au_br_hnotifyable(perm))
14666 + hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
14667 + if (unlikely(!hfsn))
14671 + group = fsnotify_alloc_group(&au_hfsn_ops);
14672 + if (IS_ERR(group)) {
14673 + err = PTR_ERR(group);
14674 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
14678 + group->private = hfsn;
14679 + hfsn->hfsn_group = group;
14680 + br->br_hfsn = hfsn;
14681 + goto out; /* success */
14689 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
14694 + if (!br->br_hfsn)
14695 + err = au_hfsn_init_br(br, perm);
14700 +/* ---------------------------------------------------------------------- */
14702 +static void au_hfsn_fin(void)
14704 + AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
14705 + wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
14708 +const struct au_hnotify_op au_hnotify_op = {
14709 + .ctl = au_hfsn_ctl,
14710 + .alloc = au_hfsn_alloc,
14711 + .free = au_hfsn_free,
14713 + .fin = au_hfsn_fin,
14715 + .reset_br = au_hfsn_reset_br,
14716 + .fin_br = au_hfsn_fin_br,
14717 + .init_br = au_hfsn_init_br
14719 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
14720 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
14721 +++ linux/fs/aufs/hfsplus.c 2015-09-24 10:47:58.254719746 +0200
14724 + * Copyright (C) 2010-2015 Junjiro R. Okajima
14726 + * This program, aufs is free software; you can redistribute it and/or modify
14727 + * it under the terms of the GNU General Public License as published by
14728 + * the Free Software Foundation; either version 2 of the License, or
14729 + * (at your option) any later version.
14731 + * This program is distributed in the hope that it will be useful,
14732 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14733 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14734 + * GNU General Public License for more details.
14736 + * You should have received a copy of the GNU General Public License
14737 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14741 + * special support for filesystems which aqucires an inode mutex
14742 + * at final closing a file, eg, hfsplus.
14744 + * This trick is very simple and stupid, just to open the file before really
14745 + * neceeary open to tell hfsplus that this is not the final closing.
14746 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
14747 + * and au_h_open_post() after releasing it.
14752 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
14755 + struct file *h_file;
14756 + struct dentry *h_dentry;
14758 + h_dentry = au_h_dptr(dentry, bindex);
14759 + AuDebugOn(!h_dentry);
14760 + AuDebugOn(d_is_negative(h_dentry));
14763 + if (au_test_hfsplus(h_dentry->d_sb)
14764 + && d_is_reg(h_dentry))
14765 + h_file = au_h_open(dentry, bindex,
14766 + O_RDONLY | O_NOATIME | O_LARGEFILE,
14767 + /*file*/NULL, force_wr);
14771 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
14772 + struct file *h_file)
14776 + au_sbr_put(dentry->d_sb, bindex);
14779 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
14780 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
14781 +++ linux/fs/aufs/hnotify.c 2015-09-24 10:47:58.254719746 +0200
14784 + * Copyright (C) 2005-2015 Junjiro R. Okajima
14786 + * This program, aufs is free software; you can redistribute it and/or modify
14787 + * it under the terms of the GNU General Public License as published by
14788 + * the Free Software Foundation; either version 2 of the License, or
14789 + * (at your option) any later version.
14791 + * This program is distributed in the hope that it will be useful,
14792 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14793 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14794 + * GNU General Public License for more details.
14796 + * You should have received a copy of the GNU General Public License
14797 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
14801 + * abstraction to notify the direct changes on lower directories
14806 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
14809 + struct au_hnotify *hn;
14812 + hn = au_cache_alloc_hnotify();
14814 + hn->hn_aufs_inode = inode;
14815 + hinode->hi_notify = hn;
14816 + err = au_hnotify_op.alloc(hinode);
14818 + if (unlikely(err)) {
14819 + hinode->hi_notify = NULL;
14820 + au_cache_free_hnotify(hn);
14822 + * The upper dir was removed by udba, but the same named
14823 + * dir left. In this case, aufs assignes a new inode
14824 + * number and set the monitor again.
14825 + * For the lower dir, the old monitnor is still left.
14827 + if (err == -EEXIST)
14836 +void au_hn_free(struct au_hinode *hinode)
14838 + struct au_hnotify *hn;
14840 + hn = hinode->hi_notify;
14842 + hinode->hi_notify = NULL;
14843 + if (au_hnotify_op.free(hinode, hn))
14844 + au_cache_free_hnotify(hn);
14848 +/* ---------------------------------------------------------------------- */
14850 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
14852 + if (hinode->hi_notify)
14853 + au_hnotify_op.ctl(hinode, do_set);
14856 +void au_hn_reset(struct inode *inode, unsigned int flags)
14858 + aufs_bindex_t bindex, bend;
14859 + struct inode *hi;
14860 + struct dentry *iwhdentry;
14862 + bend = au_ibend(inode);
14863 + for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
14864 + hi = au_h_iptr(inode, bindex);
14868 + /* mutex_lock_nested(&hi->i_mutex, AuLsc_I_CHILD); */
14869 + iwhdentry = au_hi_wh(inode, bindex);
14873 + au_set_h_iptr(inode, bindex, NULL, 0);
14874 + au_set_h_iptr(inode, bindex, au_igrab(hi),
14875 + flags & ~AuHi_XINO);
14878 + /* mutex_unlock(&hi->i_mutex); */
14882 +/* ---------------------------------------------------------------------- */
14884 +static int hn_xino(struct inode *inode, struct inode *h_inode)
14887 + aufs_bindex_t bindex, bend, bfound, bstart;
14888 + struct inode *h_i;
14891 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
14892 + pr_warn("branch root dir was changed\n");
14897 + bend = au_ibend(inode);
14898 + bstart = au_ibstart(inode);
14899 +#if 0 /* reserved for future use */
14900 + if (bindex == bend) {
14901 + /* keep this ino in rename case */
14905 + for (bindex = bstart; bindex <= bend; bindex++)
14906 + if (au_h_iptr(inode, bindex) == h_inode) {
14913 + for (bindex = bstart; bindex <= bend; bindex++) {
14914 + h_i = au_h_iptr(inode, bindex);
14918 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
14919 + /* ignore this error */
14920 + /* bad action? */
14923 + /* children inode number will be broken */
14930 +static int hn_gen_tree(struct dentry *dentry)
14932 + int err, i, j, ndentry;
14933 + struct au_dcsub_pages dpages;
14934 + struct au_dpage *dpage;
14935 + struct dentry **dentries;
14937 + err = au_dpages_init(&dpages, GFP_NOFS);
14938 + if (unlikely(err))
14940 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
14941 + if (unlikely(err))
14944 + for (i = 0; i < dpages.ndpage; i++) {
14945 + dpage = dpages.dpages + i;
14946 + dentries = dpage->dentries;
14947 + ndentry = dpage->ndentry;
14948 + for (j = 0; j < ndentry; j++) {
14949 + struct dentry *d;
14956 + if (d_really_is_positive(d))
14957 + /* todo: reset children xino?
14958 + cached children only? */
14959 + au_iigen_dec(d_inode(d));
14964 + au_dpages_free(&dpages);
14967 + /* discard children */
14968 + dentry_unhash(dentry);
14976 + * return 0 if processed.
14978 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
14979 + const unsigned int isdir)
14982 + struct dentry *d;
14983 + struct qstr *dname;
14986 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
14987 + pr_warn("branch root dir was changed\n");
14993 + AuDebugOn(!name);
14994 + au_iigen_dec(inode);
14995 + spin_lock(&inode->i_lock);
14996 + hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
14997 + spin_lock(&d->d_lock);
14998 + dname = &d->d_name;
14999 + if (dname->len != nlen
15000 + && memcmp(dname->name, name, nlen)) {
15001 + spin_unlock(&d->d_lock);
15006 + spin_unlock(&d->d_lock);
15009 + spin_unlock(&inode->i_lock);
15011 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
15012 + d = d_find_any_alias(inode);
15014 + au_iigen_dec(inode);
15018 + spin_lock(&d->d_lock);
15019 + dname = &d->d_name;
15020 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
15021 + spin_unlock(&d->d_lock);
15022 + err = hn_gen_tree(d);
15023 + spin_lock(&d->d_lock);
15025 + spin_unlock(&d->d_lock);
15034 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
15038 + if (IS_ROOT(dentry)) {
15039 + pr_warn("branch root dir was changed\n");
15045 + au_digen_dec(dentry);
15046 + if (d_really_is_positive(dentry))
15047 + au_iigen_dec(d_inode(dentry));
15049 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
15050 + if (d_really_is_positive(dentry))
15051 + err = hn_gen_tree(dentry);
15058 +/* ---------------------------------------------------------------------- */
15060 +/* hnotify job flags */
15061 +#define AuHnJob_XINO0 1
15062 +#define AuHnJob_GEN (1 << 1)
15063 +#define AuHnJob_DIRENT (1 << 2)
15064 +#define AuHnJob_ISDIR (1 << 3)
15065 +#define AuHnJob_TRYXINO0 (1 << 4)
15066 +#define AuHnJob_MNTPNT (1 << 5)
15067 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
15068 +#define au_fset_hnjob(flags, name) \
15069 + do { (flags) |= AuHnJob_##name; } while (0)
15070 +#define au_fclr_hnjob(flags, name) \
15071 + do { (flags) &= ~AuHnJob_##name; } while (0)
15079 +struct au_hnotify_args {
15080 + struct inode *h_dir, *dir, *h_child_inode;
15082 + unsigned int flags[AuHnLast];
15083 + unsigned int h_child_nlen;
15084 + char h_child_name[];
15087 +struct hn_job_args {
15088 + unsigned int flags;
15089 + struct inode *inode, *h_inode, *dir, *h_dir;
15090 + struct dentry *dentry;
15095 +static int hn_job(struct hn_job_args *a)
15097 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
15101 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
15102 + hn_xino(a->inode, a->h_inode); /* ignore this error */
15104 + if (au_ftest_hnjob(a->flags, TRYXINO0)
15107 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
15108 + if (!a->h_inode->i_nlink
15109 + && !(a->h_inode->i_state & I_LINKABLE))
15110 + hn_xino(a->inode, a->h_inode); /* ignore this error */
15111 + mutex_unlock(&a->h_inode->i_mutex);
15114 + /* make the generation obsolete */
15115 + if (au_ftest_hnjob(a->flags, GEN)) {
15118 + e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
15120 + if (e && a->dentry)
15121 + hn_gen_by_name(a->dentry, isdir);
15122 + /* ignore this error */
15125 + /* make dir entries obsolete */
15126 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
15127 + struct au_vdir *vdir;
15129 + vdir = au_ivdir(a->inode);
15131 + vdir->vd_jiffy = 0;
15132 + /* IMustLock(a->inode); */
15133 + /* a->inode->i_version++; */
15136 + /* can do nothing but warn */
15137 + if (au_ftest_hnjob(a->flags, MNTPNT)
15139 + && d_mountpoint(a->dentry))
15140 + pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
15145 +/* ---------------------------------------------------------------------- */
15147 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
15148 + struct inode *dir)
15150 + struct dentry *dentry, *d, *parent;
15151 + struct qstr *dname;
15153 + parent = d_find_any_alias(dir);
15158 + spin_lock(&parent->d_lock);
15159 + list_for_each_entry(d, &parent->d_subdirs, d_child) {
15160 + /* AuDbg("%pd\n", d); */
15161 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
15162 + dname = &d->d_name;
15163 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
15164 + goto cont_unlock;
15168 + goto cont_unlock;
15169 + if (au_dcount(d) > 0) {
15170 + dentry = dget_dlock(d);
15171 + spin_unlock(&d->d_lock);
15176 + spin_unlock(&d->d_lock);
15178 + spin_unlock(&parent->d_lock);
15182 + di_write_lock_child(dentry);
15187 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
15188 + aufs_bindex_t bindex, ino_t h_ino)
15190 + struct inode *inode;
15195 + err = au_xino_read(sb, bindex, h_ino, &ino);
15197 + inode = ilookup(sb, ino);
15201 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
15202 + pr_warn("wrong root branch\n");
15208 + ii_write_lock_child(inode);
15214 +static void au_hn_bh(void *_args)
15216 + struct au_hnotify_args *a = _args;
15217 + struct super_block *sb;
15218 + aufs_bindex_t bindex, bend, bfound;
15219 + unsigned char xino, try_iput;
15221 + struct inode *inode;
15223 + struct hn_job_args args;
15224 + struct dentry *dentry;
15225 + struct au_sbinfo *sbinfo;
15227 + AuDebugOn(!_args);
15228 + AuDebugOn(!a->h_dir);
15229 + AuDebugOn(!a->dir);
15230 + AuDebugOn(!a->mask);
15231 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
15232 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
15233 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
15238 + * do not lock a->dir->i_mutex here
15239 + * because of d_revalidate() may cause a deadlock.
15241 + sb = a->dir->i_sb;
15243 + sbinfo = au_sbi(sb);
15244 + AuDebugOn(!sbinfo);
15245 + si_write_lock(sb, AuLock_NOPLMW);
15247 + ii_read_lock_parent(a->dir);
15249 + bend = au_ibend(a->dir);
15250 + for (bindex = au_ibstart(a->dir); bindex <= bend; bindex++)
15251 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
15255 + ii_read_unlock(a->dir);
15256 + if (unlikely(bfound < 0))
15259 + xino = !!au_opt_test(au_mntflags(sb), XINO);
15261 + if (a->h_child_inode)
15262 + h_ino = a->h_child_inode->i_ino;
15264 + if (a->h_child_nlen
15265 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
15266 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
15267 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
15270 + if (dentry && d_really_is_positive(dentry))
15271 + inode = d_inode(dentry);
15272 + if (xino && !inode && h_ino
15273 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
15274 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
15275 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
15276 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
15280 + args.flags = a->flags[AuHn_CHILD];
15281 + args.dentry = dentry;
15282 + args.inode = inode;
15283 + args.h_inode = a->h_child_inode;
15284 + args.dir = a->dir;
15285 + args.h_dir = a->h_dir;
15286 + args.h_name = a->h_child_name;
15287 + args.h_nlen = a->h_child_nlen;
15288 + err = hn_job(&args);
15290 + if (au_di(dentry))
15291 + di_write_unlock(dentry);
15294 + if (inode && try_iput) {
15295 + ii_write_unlock(inode);
15299 + ii_write_lock_parent(a->dir);
15300 + args.flags = a->flags[AuHn_PARENT];
15301 + args.dentry = NULL;
15302 + args.inode = a->dir;
15303 + args.h_inode = a->h_dir;
15305 + args.h_dir = NULL;
15306 + args.h_name = NULL;
15308 + err = hn_job(&args);
15309 + ii_write_unlock(a->dir);
15312 + iput(a->h_child_inode);
15315 + si_write_unlock(sb);
15316 + au_nwt_done(&sbinfo->si_nowait);
15320 +/* ---------------------------------------------------------------------- */
15322 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
15323 + struct qstr *h_child_qstr, struct inode *h_child_inode)
15326 + unsigned int flags[AuHnLast], f;
15327 + unsigned char isdir, isroot, wh;
15328 + struct inode *dir;
15329 + struct au_hnotify_args *args;
15330 + char *p, *h_child_name;
15333 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
15334 + dir = igrab(hnotify->hn_aufs_inode);
15338 + isroot = (dir->i_ino == AUFS_ROOT_INO);
15340 + h_child_name = (void *)h_child_qstr->name;
15341 + len = h_child_qstr->len;
15342 + if (h_child_name) {
15343 + if (len > AUFS_WH_PFX_LEN
15344 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
15345 + h_child_name += AUFS_WH_PFX_LEN;
15346 + len -= AUFS_WH_PFX_LEN;
15352 + if (h_child_inode)
15353 + isdir = !!S_ISDIR(h_child_inode->i_mode);
15354 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
15355 + flags[AuHn_CHILD] = 0;
15357 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
15358 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
15359 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
15360 + switch (mask & FS_EVENTS_POSS_ON_CHILD) {
15361 + case FS_MOVED_FROM:
15362 + case FS_MOVED_TO:
15363 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
15364 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
15367 + AuDebugOn(!h_child_name);
15372 + * aufs never be able to get this child inode.
15373 + * revalidation should be in d_revalidate()
15374 + * by checking i_nlink, i_generation or d_unhashed().
15376 + AuDebugOn(!h_child_name);
15377 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
15378 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
15386 + h_child_inode = NULL;
15389 + /* iput() and kfree() will be called in au_hnotify() */
15390 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
15391 + if (unlikely(!args)) {
15392 + AuErr1("no memory\n");
15396 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
15397 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
15398 + args->mask = mask;
15400 + args->h_dir = igrab(h_dir);
15401 + if (h_child_inode)
15402 + h_child_inode = igrab(h_child_inode); /* can be NULL */
15403 + args->h_child_inode = h_child_inode;
15404 + args->h_child_nlen = len;
15406 + p = (void *)args;
15407 + p += sizeof(*args);
15408 + memcpy(p, h_child_name, len);
15412 + /* NFS fires the event for silly-renamed one from kworker */
15414 + if (!dir->i_nlink
15415 + || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
15417 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
15418 + if (unlikely(err)) {
15419 + pr_err("wkq %d\n", err);
15420 + iput(args->h_child_inode);
15421 + iput(args->h_dir);
15430 +/* ---------------------------------------------------------------------- */
15432 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
15436 + AuDebugOn(!(udba & AuOptMask_UDBA));
15439 + if (au_hnotify_op.reset_br)
15440 + err = au_hnotify_op.reset_br(udba, br, perm);
15445 +int au_hnotify_init_br(struct au_branch *br, int perm)
15450 + if (au_hnotify_op.init_br)
15451 + err = au_hnotify_op.init_br(br, perm);
15456 +void au_hnotify_fin_br(struct au_branch *br)
15458 + if (au_hnotify_op.fin_br)
15459 + au_hnotify_op.fin_br(br);
15462 +static void au_hn_destroy_cache(void)
15464 + kmem_cache_destroy(au_cachep[AuCache_HNOTIFY]);
15465 + au_cachep[AuCache_HNOTIFY] = NULL;
15468 +int __init au_hnotify_init(void)
15473 + au_cachep[AuCache_HNOTIFY] = AuCache(au_hnotify);
15474 + if (au_cachep[AuCache_HNOTIFY]) {
15476 + if (au_hnotify_op.init)
15477 + err = au_hnotify_op.init();
15478 + if (unlikely(err))
15479 + au_hn_destroy_cache();
15485 +void au_hnotify_fin(void)
15487 + if (au_hnotify_op.fin)
15488 + au_hnotify_op.fin();
15489 + /* cf. au_cache_fin() */
15490 + if (au_cachep[AuCache_HNOTIFY])
15491 + au_hn_destroy_cache();
15493 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
15494 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
15495 +++ linux/fs/aufs/iinfo.c 2015-09-24 10:47:58.254719746 +0200
15498 + * Copyright (C) 2005-2015 Junjiro R. Okajima
15500 + * This program, aufs is free software; you can redistribute it and/or modify
15501 + * it under the terms of the GNU General Public License as published by
15502 + * the Free Software Foundation; either version 2 of the License, or
15503 + * (at your option) any later version.
15505 + * This program is distributed in the hope that it will be useful,
15506 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15507 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15508 + * GNU General Public License for more details.
15510 + * You should have received a copy of the GNU General Public License
15511 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15515 + * inode private data
15520 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
15522 + struct inode *h_inode;
15524 + IiMustAnyLock(inode);
15526 + h_inode = au_ii(inode)->ii_hinode[0 + bindex].hi_inode;
15527 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
15531 +/* todo: hard/soft set? */
15532 +void au_hiput(struct au_hinode *hinode)
15534 + au_hn_free(hinode);
15535 + dput(hinode->hi_whdentry);
15536 + iput(hinode->hi_inode);
15539 +unsigned int au_hi_flags(struct inode *inode, int isdir)
15541 + unsigned int flags;
15542 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
15545 + if (au_opt_test(mnt_flags, XINO))
15546 + au_fset_hi(flags, XINO);
15547 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
15548 + au_fset_hi(flags, HNOTIFY);
15552 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
15553 + struct inode *h_inode, unsigned int flags)
15555 + struct au_hinode *hinode;
15556 + struct inode *hi;
15557 + struct au_iinfo *iinfo = au_ii(inode);
15559 + IiMustWriteLock(inode);
15561 + hinode = iinfo->ii_hinode + bindex;
15562 + hi = hinode->hi_inode;
15563 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
15566 + au_hiput(hinode);
15567 + hinode->hi_inode = h_inode;
15570 + struct super_block *sb = inode->i_sb;
15571 + struct au_branch *br;
15573 + AuDebugOn(inode->i_mode
15574 + && (h_inode->i_mode & S_IFMT)
15575 + != (inode->i_mode & S_IFMT));
15576 + if (bindex == iinfo->ii_bstart)
15577 + au_cpup_igen(inode, h_inode);
15578 + br = au_sbr(sb, bindex);
15579 + hinode->hi_id = br->br_id;
15580 + if (au_ftest_hi(flags, XINO)) {
15581 + err = au_xino_write(sb, bindex, h_inode->i_ino,
15583 + if (unlikely(err))
15584 + AuIOErr1("failed au_xino_write() %d\n", err);
15587 + if (au_ftest_hi(flags, HNOTIFY)
15588 + && au_br_hnotifyable(br->br_perm)) {
15589 + err = au_hn_alloc(hinode, inode);
15590 + if (unlikely(err))
15591 + AuIOErr1("au_hn_alloc() %d\n", err);
15596 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
15597 + struct dentry *h_wh)
15599 + struct au_hinode *hinode;
15601 + IiMustWriteLock(inode);
15603 + hinode = au_ii(inode)->ii_hinode + bindex;
15604 + AuDebugOn(hinode->hi_whdentry);
15605 + hinode->hi_whdentry = h_wh;
15608 +void au_update_iigen(struct inode *inode, int half)
15610 + struct au_iinfo *iinfo;
15611 + struct au_iigen *iigen;
15612 + unsigned int sigen;
15614 + sigen = au_sigen(inode->i_sb);
15615 + iinfo = au_ii(inode);
15616 + iigen = &iinfo->ii_generation;
15617 + spin_lock(&iinfo->ii_genspin);
15618 + iigen->ig_generation = sigen;
15620 + au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
15622 + au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
15623 + spin_unlock(&iinfo->ii_genspin);
15626 +/* it may be called at remount time, too */
15627 +void au_update_ibrange(struct inode *inode, int do_put_zero)
15629 + struct au_iinfo *iinfo;
15630 + aufs_bindex_t bindex, bend;
15632 + iinfo = au_ii(inode);
15636 + IiMustWriteLock(inode);
15638 + if (do_put_zero && iinfo->ii_bstart >= 0) {
15639 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
15641 + struct inode *h_i;
15643 + h_i = iinfo->ii_hinode[0 + bindex].hi_inode;
15646 + && !(h_i->i_state & I_LINKABLE))
15647 + au_set_h_iptr(inode, bindex, NULL, 0);
15651 + iinfo->ii_bstart = -1;
15652 + iinfo->ii_bend = -1;
15653 + bend = au_sbend(inode->i_sb);
15654 + for (bindex = 0; bindex <= bend; bindex++)
15655 + if (iinfo->ii_hinode[0 + bindex].hi_inode) {
15656 + iinfo->ii_bstart = bindex;
15659 + if (iinfo->ii_bstart >= 0)
15660 + for (bindex = bend; bindex >= iinfo->ii_bstart; bindex--)
15661 + if (iinfo->ii_hinode[0 + bindex].hi_inode) {
15662 + iinfo->ii_bend = bindex;
15665 + AuDebugOn(iinfo->ii_bstart > iinfo->ii_bend);
15668 +/* ---------------------------------------------------------------------- */
15670 +void au_icntnr_init_once(void *_c)
15672 + struct au_icntnr *c = _c;
15673 + struct au_iinfo *iinfo = &c->iinfo;
15674 + static struct lock_class_key aufs_ii;
15676 + spin_lock_init(&iinfo->ii_genspin);
15677 + au_rw_init(&iinfo->ii_rwsem);
15678 + au_rw_class(&iinfo->ii_rwsem, &aufs_ii);
15679 + inode_init_once(&c->vfs_inode);
15682 +int au_iinfo_init(struct inode *inode)
15684 + struct au_iinfo *iinfo;
15685 + struct super_block *sb;
15688 + sb = inode->i_sb;
15689 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
15690 + nbr = au_sbend(sb) + 1;
15691 + if (unlikely(nbr <= 0))
15693 + iinfo->ii_hinode = kcalloc(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
15694 + if (iinfo->ii_hinode) {
15695 + au_ninodes_inc(sb);
15696 + for (i = 0; i < nbr; i++)
15697 + iinfo->ii_hinode[i].hi_id = -1;
15699 + iinfo->ii_generation.ig_generation = au_sigen(sb);
15700 + iinfo->ii_bstart = -1;
15701 + iinfo->ii_bend = -1;
15702 + iinfo->ii_vdir = NULL;
15708 +int au_ii_realloc(struct au_iinfo *iinfo, int nbr)
15711 + struct au_hinode *hip;
15713 + AuRwMustWriteLock(&iinfo->ii_rwsem);
15716 + sz = sizeof(*hip) * (iinfo->ii_bend + 1);
15718 + sz = sizeof(*hip);
15719 + hip = au_kzrealloc(iinfo->ii_hinode, sz, sizeof(*hip) * nbr, GFP_NOFS);
15721 + iinfo->ii_hinode = hip;
15728 +void au_iinfo_fin(struct inode *inode)
15730 + struct au_iinfo *iinfo;
15731 + struct au_hinode *hi;
15732 + struct super_block *sb;
15733 + aufs_bindex_t bindex, bend;
15734 + const unsigned char unlinked = !inode->i_nlink;
15736 + iinfo = au_ii(inode);
15737 + /* bad_inode case */
15741 + sb = inode->i_sb;
15742 + au_ninodes_dec(sb);
15743 + if (si_pid_test(sb))
15744 + au_xino_delete_inode(inode, unlinked);
15747 + * it is safe to hide the dependency between sbinfo and
15751 + si_noflush_read_lock(sb);
15752 + au_xino_delete_inode(inode, unlinked);
15753 + si_read_unlock(sb);
15757 + if (iinfo->ii_vdir)
15758 + au_vdir_free(iinfo->ii_vdir);
15760 + bindex = iinfo->ii_bstart;
15761 + if (bindex >= 0) {
15762 + hi = iinfo->ii_hinode + bindex;
15763 + bend = iinfo->ii_bend;
15764 + while (bindex++ <= bend) {
15765 + if (hi->hi_inode)
15770 + kfree(iinfo->ii_hinode);
15771 + iinfo->ii_hinode = NULL;
15772 + AuRwDestroy(&iinfo->ii_rwsem);
15774 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
15775 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
15776 +++ linux/fs/aufs/inode.c 2015-09-24 10:47:58.254719746 +0200
15779 + * Copyright (C) 2005-2015 Junjiro R. Okajima
15781 + * This program, aufs is free software; you can redistribute it and/or modify
15782 + * it under the terms of the GNU General Public License as published by
15783 + * the Free Software Foundation; either version 2 of the License, or
15784 + * (at your option) any later version.
15786 + * This program is distributed in the hope that it will be useful,
15787 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15788 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15789 + * GNU General Public License for more details.
15791 + * You should have received a copy of the GNU General Public License
15792 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
15796 + * inode functions
15801 +struct inode *au_igrab(struct inode *inode)
15804 + AuDebugOn(!atomic_read(&inode->i_count));
15810 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
15812 + au_cpup_attr_all(inode, /*force*/0);
15813 + au_update_iigen(inode, /*half*/1);
15815 + inode->i_version++;
15818 +static int au_ii_refresh(struct inode *inode, int *update)
15822 + aufs_bindex_t bindex, new_bindex;
15823 + struct super_block *sb;
15824 + struct au_iinfo *iinfo;
15825 + struct au_hinode *p, *q, tmp;
15827 + IiMustWriteLock(inode);
15830 + sb = inode->i_sb;
15831 + type = inode->i_mode & S_IFMT;
15832 + iinfo = au_ii(inode);
15833 + err = au_ii_realloc(iinfo, au_sbend(sb) + 1);
15834 + if (unlikely(err))
15837 + AuDebugOn(iinfo->ii_bstart < 0);
15838 + p = iinfo->ii_hinode + iinfo->ii_bstart;
15839 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
15841 + if (!p->hi_inode)
15844 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
15845 + new_bindex = au_br_index(sb, p->hi_id);
15846 + if (new_bindex == bindex)
15849 + if (new_bindex < 0) {
15852 + p->hi_inode = NULL;
15856 + if (new_bindex < iinfo->ii_bstart)
15857 + iinfo->ii_bstart = new_bindex;
15858 + if (iinfo->ii_bend < new_bindex)
15859 + iinfo->ii_bend = new_bindex;
15860 + /* swap two lower inode, and loop again */
15861 + q = iinfo->ii_hinode + new_bindex;
15865 + if (tmp.hi_inode) {
15870 + au_update_ibrange(inode, /*do_put_zero*/0);
15871 + e = au_dy_irefresh(inode);
15872 + if (unlikely(e && !err))
15880 +int au_refresh_hinode_self(struct inode *inode)
15884 + err = au_ii_refresh(inode, &update);
15886 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
15892 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
15894 + int err, e, update;
15895 + unsigned int flags;
15897 + aufs_bindex_t bindex, bend;
15898 + unsigned char isdir;
15899 + struct au_hinode *p;
15900 + struct au_iinfo *iinfo;
15902 + err = au_ii_refresh(inode, &update);
15903 + if (unlikely(err))
15907 + iinfo = au_ii(inode);
15908 + p = iinfo->ii_hinode + iinfo->ii_bstart;
15909 + mode = (inode->i_mode & S_IFMT);
15910 + isdir = S_ISDIR(mode);
15911 + flags = au_hi_flags(inode, isdir);
15912 + bend = au_dbend(dentry);
15913 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
15914 + struct inode *h_i, *h_inode;
15915 + struct dentry *h_d;
15917 + h_d = au_h_dptr(dentry, bindex);
15918 + if (!h_d || d_is_negative(h_d))
15921 + h_inode = d_inode(h_d);
15922 + AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
15923 + if (iinfo->ii_bstart <= bindex && bindex <= iinfo->ii_bend) {
15924 + h_i = au_h_iptr(inode, bindex);
15926 + if (h_i == h_inode)
15932 + if (bindex < iinfo->ii_bstart)
15933 + iinfo->ii_bstart = bindex;
15934 + if (iinfo->ii_bend < bindex)
15935 + iinfo->ii_bend = bindex;
15936 + au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
15939 + au_update_ibrange(inode, /*do_put_zero*/0);
15940 + e = au_dy_irefresh(inode);
15941 + if (unlikely(e && !err))
15944 + au_refresh_hinode_attr(inode, update && isdir);
15951 +static int set_inode(struct inode *inode, struct dentry *dentry)
15954 + unsigned int flags;
15956 + aufs_bindex_t bindex, bstart, btail;
15957 + unsigned char isdir;
15958 + struct dentry *h_dentry;
15959 + struct inode *h_inode;
15960 + struct au_iinfo *iinfo;
15962 + IiMustWriteLock(inode);
15966 + bstart = au_dbstart(dentry);
15967 + h_dentry = au_h_dptr(dentry, bstart);
15968 + h_inode = d_inode(h_dentry);
15969 + mode = h_inode->i_mode;
15970 + switch (mode & S_IFMT) {
15972 + btail = au_dbtail(dentry);
15973 + inode->i_op = &aufs_iop;
15974 + inode->i_fop = &aufs_file_fop;
15975 + err = au_dy_iaop(inode, bstart, h_inode);
15976 + if (unlikely(err))
15981 + btail = au_dbtaildir(dentry);
15982 + inode->i_op = &aufs_dir_iop;
15983 + inode->i_fop = &aufs_dir_fop;
15986 + btail = au_dbtail(dentry);
15987 + inode->i_op = &aufs_symlink_iop;
15993 + btail = au_dbtail(dentry);
15994 + inode->i_op = &aufs_iop;
15995 + init_special_inode(inode, mode, h_inode->i_rdev);
15998 + AuIOErr("Unknown file type 0%o\n", mode);
16003 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
16004 + flags = au_hi_flags(inode, isdir);
16005 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
16006 + && au_ftest_hi(flags, HNOTIFY)
16007 + && dentry->d_name.len > AUFS_WH_PFX_LEN
16008 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
16009 + au_fclr_hi(flags, HNOTIFY);
16010 + iinfo = au_ii(inode);
16011 + iinfo->ii_bstart = bstart;
16012 + iinfo->ii_bend = btail;
16013 + for (bindex = bstart; bindex <= btail; bindex++) {
16014 + h_dentry = au_h_dptr(dentry, bindex);
16016 + au_set_h_iptr(inode, bindex,
16017 + au_igrab(d_inode(h_dentry)), flags);
16019 + au_cpup_attr_all(inode, /*force*/1);
16021 + * to force calling aufs_get_acl() every time,
16022 + * do not call cache_no_acl() for aufs inode.
16030 + * successful returns with iinfo write_locked
16032 + * zero: success, matched
16033 + * plus: no error, but unmatched
16035 +static int reval_inode(struct inode *inode, struct dentry *dentry)
16038 + unsigned int gen;
16039 + struct au_iigen iigen;
16040 + aufs_bindex_t bindex, bend;
16041 + struct inode *h_inode, *h_dinode;
16042 + struct dentry *h_dentry;
16045 + * before this function, if aufs got any iinfo lock, it must be only
16046 + * one, the parent dir.
16047 + * it can happen by UDBA and the obsoleted inode number.
16050 + if (unlikely(inode->i_ino == parent_ino(dentry)))
16054 + ii_write_lock_new_child(inode);
16055 + h_dentry = au_h_dptr(dentry, au_dbstart(dentry));
16056 + h_dinode = d_inode(h_dentry);
16057 + bend = au_ibend(inode);
16058 + for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
16059 + h_inode = au_h_iptr(inode, bindex);
16060 + if (!h_inode || h_inode != h_dinode)
16064 + gen = au_iigen(inode, &iigen);
16065 + if (gen == au_digen(dentry)
16066 + && !au_ig_ftest(iigen.ig_flags, HALF_REFRESHED))
16069 + /* fully refresh inode using dentry */
16070 + err = au_refresh_hinode(inode, dentry);
16072 + au_update_iigen(inode, /*half*/0);
16076 + if (unlikely(err))
16077 + ii_write_unlock(inode);
16082 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
16083 + unsigned int d_type, ino_t *ino)
16086 + struct mutex *mtx;
16088 + /* prevent hardlinked inode number from race condition */
16090 + if (d_type != DT_DIR) {
16091 + mtx = &au_sbr(sb, bindex)->br_xino.xi_nondir_mtx;
16094 + err = au_xino_read(sb, bindex, h_ino, ino);
16095 + if (unlikely(err))
16100 + *ino = au_xino_new_ino(sb);
16101 + if (unlikely(!*ino))
16103 + err = au_xino_write(sb, bindex, h_ino, *ino);
16104 + if (unlikely(err))
16110 + mutex_unlock(mtx);
16114 +/* successful returns with iinfo write_locked */
16115 +/* todo: return with unlocked? */
16116 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
16118 + struct inode *inode, *h_inode;
16119 + struct dentry *h_dentry;
16120 + struct super_block *sb;
16121 + struct mutex *mtx;
16122 + ino_t h_ino, ino;
16124 + aufs_bindex_t bstart;
16126 + sb = dentry->d_sb;
16127 + bstart = au_dbstart(dentry);
16128 + h_dentry = au_h_dptr(dentry, bstart);
16129 + h_inode = d_inode(h_dentry);
16130 + h_ino = h_inode->i_ino;
16133 + * stop 'race'-ing between hardlinks under different
16137 + if (!d_is_dir(h_dentry))
16138 + mtx = &au_sbr(sb, bstart)->br_xino.xi_nondir_mtx;
16143 + err = au_xino_read(sb, bstart, h_ino, &ino);
16144 + inode = ERR_PTR(err);
16145 + if (unlikely(err))
16149 + ino = au_xino_new_ino(sb);
16150 + if (unlikely(!ino)) {
16151 + inode = ERR_PTR(-EIO);
16156 + AuDbg("i%lu\n", (unsigned long)ino);
16157 + inode = au_iget_locked(sb, ino);
16158 + err = PTR_ERR(inode);
16159 + if (IS_ERR(inode))
16162 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
16163 + if (inode->i_state & I_NEW) {
16164 + /* verbose coding for lock class name */
16165 + if (unlikely(d_is_symlink(h_dentry)))
16166 + au_rw_class(&au_ii(inode)->ii_rwsem,
16167 + au_lc_key + AuLcSymlink_IIINFO);
16168 + else if (unlikely(d_is_dir(h_dentry)))
16169 + au_rw_class(&au_ii(inode)->ii_rwsem,
16170 + au_lc_key + AuLcDir_IIINFO);
16171 + else /* likely */
16172 + au_rw_class(&au_ii(inode)->ii_rwsem,
16173 + au_lc_key + AuLcNonDir_IIINFO);
16175 + ii_write_lock_new_child(inode);
16176 + err = set_inode(inode, dentry);
16178 + unlock_new_inode(inode);
16179 + goto out; /* success */
16183 + * iget_failed() calls iput(), but we need to call
16184 + * ii_write_unlock() after iget_failed(). so dirty hack for
16187 + atomic_inc(&inode->i_count);
16188 + iget_failed(inode);
16189 + ii_write_unlock(inode);
16190 + au_xino_write(sb, bstart, h_ino, /*ino*/0);
16191 + /* ignore this error */
16193 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
16195 + * horrible race condition between lookup, readdir and copyup
16196 + * (or something).
16199 + mutex_unlock(mtx);
16200 + err = reval_inode(inode, dentry);
16201 + if (unlikely(err < 0)) {
16208 + goto out; /* success */
16213 + if (unlikely(au_test_fs_unique_ino(h_inode)))
16214 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
16215 + " b%d, %s, %pd, hi%lu, i%lu.\n",
16216 + bstart, au_sbtype(h_dentry->d_sb), dentry,
16217 + (unsigned long)h_ino, (unsigned long)ino);
16219 + err = au_xino_write(sb, bstart, h_ino, /*ino*/0);
16223 + mutex_unlock(mtx);
16229 + inode = ERR_PTR(err);
16232 + mutex_unlock(mtx);
16236 +/* ---------------------------------------------------------------------- */
16238 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
16239 + struct inode *inode)
16242 + struct inode *hi;
16244 + err = au_br_rdonly(au_sbr(sb, bindex));
16246 + /* pseudo-link after flushed may happen out of bounds */
16249 + && au_ibstart(inode) <= bindex
16250 + && bindex <= au_ibend(inode)) {
16252 + * permission check is unnecessary since vfsub routine
16253 + * will be called later
16255 + hi = au_h_iptr(inode, bindex);
16257 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
16263 +int au_test_h_perm(struct inode *h_inode, int mask)
16265 + if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
16267 + return inode_permission(h_inode, mask);
16270 +int au_test_h_perm_sio(struct inode *h_inode, int mask)
16272 + if (au_test_nfs(h_inode->i_sb)
16273 + && (mask & MAY_WRITE)
16274 + && S_ISDIR(h_inode->i_mode))
16275 + mask |= MAY_READ; /* force permission check */
16276 + return au_test_h_perm(h_inode, mask);
16278 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
16279 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
16280 +++ linux/fs/aufs/inode.h 2015-09-24 10:47:58.254719746 +0200
16283 + * Copyright (C) 2005-2015 Junjiro R. Okajima
16285 + * This program, aufs is free software; you can redistribute it and/or modify
16286 + * it under the terms of the GNU General Public License as published by
16287 + * the Free Software Foundation; either version 2 of the License, or
16288 + * (at your option) any later version.
16290 + * This program is distributed in the hope that it will be useful,
16291 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16292 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16293 + * GNU General Public License for more details.
16295 + * You should have received a copy of the GNU General Public License
16296 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16300 + * inode operations
16303 +#ifndef __AUFS_INODE_H__
16304 +#define __AUFS_INODE_H__
16308 +#include <linux/fsnotify.h>
16309 +#include "rwsem.h"
16313 +struct au_hnotify {
16314 +#ifdef CONFIG_AUFS_HNOTIFY
16315 +#ifdef CONFIG_AUFS_HFSNOTIFY
16316 + /* never use fsnotify_add_vfsmount_mark() */
16317 + struct fsnotify_mark hn_mark;
16319 + struct inode *hn_aufs_inode; /* no get/put */
16321 +} ____cacheline_aligned_in_smp;
16323 +struct au_hinode {
16324 + struct inode *hi_inode;
16325 + aufs_bindex_t hi_id;
16326 +#ifdef CONFIG_AUFS_HNOTIFY
16327 + struct au_hnotify *hi_notify;
16330 + /* reference to the copied-up whiteout with get/put */
16331 + struct dentry *hi_whdentry;
16335 +#define AuIG_HALF_REFRESHED 1
16336 +#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
16337 +#define au_ig_fset(flags, name) \
16338 + do { (flags) |= AuIG_##name; } while (0)
16339 +#define au_ig_fclr(flags, name) \
16340 + do { (flags) &= ~AuIG_##name; } while (0)
16343 + __u32 ig_generation, ig_flags;
16348 + spinlock_t ii_genspin;
16349 + struct au_iigen ii_generation;
16350 + struct super_block *ii_hsb1; /* no get/put */
16352 + struct au_rwsem ii_rwsem;
16353 + aufs_bindex_t ii_bstart, ii_bend;
16355 + struct au_hinode *ii_hinode;
16356 + struct au_vdir *ii_vdir;
16359 +struct au_icntnr {
16360 + struct au_iinfo iinfo;
16361 + struct inode vfs_inode;
16362 +} ____cacheline_aligned_in_smp;
16364 +/* au_pin flags */
16365 +#define AuPin_DI_LOCKED 1
16366 +#define AuPin_MNT_WRITE (1 << 1)
16367 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
16368 +#define au_fset_pin(flags, name) \
16369 + do { (flags) |= AuPin_##name; } while (0)
16370 +#define au_fclr_pin(flags, name) \
16371 + do { (flags) &= ~AuPin_##name; } while (0)
16375 + struct dentry *dentry;
16376 + unsigned int udba;
16377 + unsigned char lsc_di, lsc_hi, flags;
16378 + aufs_bindex_t bindex;
16381 + struct dentry *parent;
16382 + struct au_hinode *hdir;
16383 + struct vfsmount *h_mnt;
16385 + /* temporary unlock/relock for copyup */
16386 + struct dentry *h_dentry, *h_parent;
16387 + struct au_branch *br;
16388 + struct task_struct *task;
16391 +void au_pin_hdir_unlock(struct au_pin *p);
16392 +int au_pin_hdir_lock(struct au_pin *p);
16393 +int au_pin_hdir_relock(struct au_pin *p);
16394 +void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task);
16395 +void au_pin_hdir_acquire_nest(struct au_pin *p);
16396 +void au_pin_hdir_release(struct au_pin *p);
16398 +/* ---------------------------------------------------------------------- */
16400 +static inline struct au_iinfo *au_ii(struct inode *inode)
16402 + struct au_iinfo *iinfo;
16404 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
16405 + if (iinfo->ii_hinode)
16407 + return NULL; /* debugging bad_inode case */
16410 +/* ---------------------------------------------------------------------- */
16413 +struct inode *au_igrab(struct inode *inode);
16414 +int au_refresh_hinode_self(struct inode *inode);
16415 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
16416 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
16417 + unsigned int d_type, ino_t *ino);
16418 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
16419 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
16420 + struct inode *inode);
16421 +int au_test_h_perm(struct inode *h_inode, int mask);
16422 +int au_test_h_perm_sio(struct inode *h_inode, int mask);
16424 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
16425 + ino_t h_ino, unsigned int d_type, ino_t *ino)
16427 +#ifdef CONFIG_AUFS_SHWH
16428 + return au_ino(sb, bindex, h_ino, d_type, ino);
16435 +extern struct inode_operations aufs_iop, aufs_symlink_iop, aufs_dir_iop;
16437 +/* au_wr_dir flags */
16438 +#define AuWrDir_ADD_ENTRY 1
16439 +#define AuWrDir_ISDIR (1 << 1)
16440 +#define AuWrDir_TMPFILE (1 << 2)
16441 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
16442 +#define au_fset_wrdir(flags, name) \
16443 + do { (flags) |= AuWrDir_##name; } while (0)
16444 +#define au_fclr_wrdir(flags, name) \
16445 + do { (flags) &= ~AuWrDir_##name; } while (0)
16447 +struct au_wr_dir_args {
16448 + aufs_bindex_t force_btgt;
16449 + unsigned char flags;
16451 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
16452 + struct au_wr_dir_args *args);
16454 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
16455 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
16456 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
16457 + unsigned int udba, unsigned char flags);
16458 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
16459 + unsigned int udba, unsigned char flags) __must_check;
16460 +int au_do_pin(struct au_pin *pin) __must_check;
16461 +void au_unpin(struct au_pin *pin);
16462 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
16464 +#define AuIcpup_DID_CPUP 1
16465 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
16466 +#define au_fset_icpup(flags, name) \
16467 + do { (flags) |= AuIcpup_##name; } while (0)
16468 +#define au_fclr_icpup(flags, name) \
16469 + do { (flags) &= ~AuIcpup_##name; } while (0)
16471 +struct au_icpup_args {
16472 + unsigned char flags;
16473 + unsigned char pin_flags;
16474 + aufs_bindex_t btgt;
16475 + unsigned int udba;
16476 + struct au_pin pin;
16477 + struct path h_path;
16478 + struct inode *h_inode;
16481 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
16482 + struct au_icpup_args *a);
16484 +int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path);
16487 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
16488 + struct dentry *h_parent, int isdir);
16489 +int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
16491 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname);
16492 +int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
16494 +struct vfsub_aopen_args;
16495 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
16496 + struct vfsub_aopen_args *args);
16497 +int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode);
16498 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
16499 + struct dentry *dentry);
16500 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
16503 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
16504 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
16505 + struct dentry *h_parent, int isdir);
16506 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
16507 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
16510 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
16511 +int aufs_rename(struct inode *src_dir, struct dentry *src_dentry,
16512 + struct inode *dir, struct dentry *dentry);
16515 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
16516 +void au_hiput(struct au_hinode *hinode);
16517 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
16518 + struct dentry *h_wh);
16519 +unsigned int au_hi_flags(struct inode *inode, int isdir);
16521 +/* hinode flags */
16522 +#define AuHi_XINO 1
16523 +#define AuHi_HNOTIFY (1 << 1)
16524 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
16525 +#define au_fset_hi(flags, name) \
16526 + do { (flags) |= AuHi_##name; } while (0)
16527 +#define au_fclr_hi(flags, name) \
16528 + do { (flags) &= ~AuHi_##name; } while (0)
16530 +#ifndef CONFIG_AUFS_HNOTIFY
16531 +#undef AuHi_HNOTIFY
16532 +#define AuHi_HNOTIFY 0
16535 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
16536 + struct inode *h_inode, unsigned int flags);
16538 +void au_update_iigen(struct inode *inode, int half);
16539 +void au_update_ibrange(struct inode *inode, int do_put_zero);
16541 +void au_icntnr_init_once(void *_c);
16542 +int au_iinfo_init(struct inode *inode);
16543 +void au_iinfo_fin(struct inode *inode);
16544 +int au_ii_realloc(struct au_iinfo *iinfo, int nbr);
16546 +#ifdef CONFIG_PROC_FS
16548 +int au_plink_maint(struct super_block *sb, int flags);
16550 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
16551 +int au_plink_maint_enter(struct super_block *sb);
16552 +#ifdef CONFIG_AUFS_DEBUG
16553 +void au_plink_list(struct super_block *sb);
16555 +AuStubVoid(au_plink_list, struct super_block *sb)
16557 +int au_plink_test(struct inode *inode);
16558 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
16559 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
16560 + struct dentry *h_dentry);
16561 +void au_plink_put(struct super_block *sb, int verbose);
16562 +void au_plink_clean(struct super_block *sb, int verbose);
16563 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
16565 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
16566 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
16567 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
16568 +AuStubVoid(au_plink_list, struct super_block *sb);
16569 +AuStubInt0(au_plink_test, struct inode *inode);
16570 +AuStub(struct dentry *, au_plink_lkup, return NULL,
16571 + struct inode *inode, aufs_bindex_t bindex);
16572 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
16573 + struct dentry *h_dentry);
16574 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
16575 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
16576 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
16577 +#endif /* CONFIG_PROC_FS */
16579 +#ifdef CONFIG_AUFS_XATTR
16581 +int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
16582 + unsigned int verbose);
16583 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
16584 +ssize_t aufs_getxattr(struct dentry *dentry, const char *name, void *value,
16586 +int aufs_setxattr(struct dentry *dentry, const char *name, const void *value,
16587 + size_t size, int flags);
16588 +int aufs_removexattr(struct dentry *dentry, const char *name);
16590 +/* void au_xattr_init(struct super_block *sb); */
16592 +AuStubInt0(au_cpup_xattr, struct dentry *h_dst, struct dentry *h_src,
16593 + int ignore_flags, unsigned int verbose);
16594 +/* AuStubVoid(au_xattr_init, struct super_block *sb); */
16597 +#ifdef CONFIG_FS_POSIX_ACL
16598 +struct posix_acl *aufs_get_acl(struct inode *inode, int type);
16599 +int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
16602 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
16609 +struct au_srxattr {
16613 + const char *name;
16614 + const void *value;
16619 + const char *name;
16622 + struct posix_acl *acl;
16627 +ssize_t au_srxattr(struct dentry *dentry, struct au_srxattr *arg);
16630 +/* ---------------------------------------------------------------------- */
16632 +/* lock subclass for iinfo */
16634 + AuLsc_II_CHILD, /* child first */
16635 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
16636 + AuLsc_II_CHILD3, /* copyup dirs */
16637 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
16638 + AuLsc_II_PARENT2,
16639 + AuLsc_II_PARENT3, /* copyup dirs */
16640 + AuLsc_II_NEW_CHILD
16644 + * ii_read_lock_child, ii_write_lock_child,
16645 + * ii_read_lock_child2, ii_write_lock_child2,
16646 + * ii_read_lock_child3, ii_write_lock_child3,
16647 + * ii_read_lock_parent, ii_write_lock_parent,
16648 + * ii_read_lock_parent2, ii_write_lock_parent2,
16649 + * ii_read_lock_parent3, ii_write_lock_parent3,
16650 + * ii_read_lock_new_child, ii_write_lock_new_child,
16652 +#define AuReadLockFunc(name, lsc) \
16653 +static inline void ii_read_lock_##name(struct inode *i) \
16655 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
16658 +#define AuWriteLockFunc(name, lsc) \
16659 +static inline void ii_write_lock_##name(struct inode *i) \
16661 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
16664 +#define AuRWLockFuncs(name, lsc) \
16665 + AuReadLockFunc(name, lsc) \
16666 + AuWriteLockFunc(name, lsc)
16668 +AuRWLockFuncs(child, CHILD);
16669 +AuRWLockFuncs(child2, CHILD2);
16670 +AuRWLockFuncs(child3, CHILD3);
16671 +AuRWLockFuncs(parent, PARENT);
16672 +AuRWLockFuncs(parent2, PARENT2);
16673 +AuRWLockFuncs(parent3, PARENT3);
16674 +AuRWLockFuncs(new_child, NEW_CHILD);
16676 +#undef AuReadLockFunc
16677 +#undef AuWriteLockFunc
16678 +#undef AuRWLockFuncs
16681 + * ii_read_unlock, ii_write_unlock, ii_downgrade_lock
16683 +AuSimpleUnlockRwsemFuncs(ii, struct inode *i, &au_ii(i)->ii_rwsem);
16685 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
16686 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
16687 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
16689 +/* ---------------------------------------------------------------------- */
16691 +static inline void au_icntnr_init(struct au_icntnr *c)
16693 +#ifdef CONFIG_AUFS_DEBUG
16694 + c->vfs_inode.i_mode = 0;
16698 +static inline unsigned int au_iigen(struct inode *inode, struct au_iigen *iigen)
16700 + unsigned int gen;
16701 + struct au_iinfo *iinfo;
16703 + iinfo = au_ii(inode);
16704 + spin_lock(&iinfo->ii_genspin);
16706 + *iigen = iinfo->ii_generation;
16707 + gen = iinfo->ii_generation.ig_generation;
16708 + spin_unlock(&iinfo->ii_genspin);
16713 +/* tiny test for inode number */
16714 +/* tmpfs generation is too rough */
16715 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
16717 + struct au_iinfo *iinfo;
16719 + iinfo = au_ii(inode);
16720 + AuRwMustAnyLock(&iinfo->ii_rwsem);
16721 + return !(iinfo->ii_hsb1 == h_inode->i_sb
16722 + && iinfo->ii_higen == h_inode->i_generation);
16725 +static inline void au_iigen_dec(struct inode *inode)
16727 + struct au_iinfo *iinfo;
16729 + iinfo = au_ii(inode);
16730 + spin_lock(&iinfo->ii_genspin);
16731 + iinfo->ii_generation.ig_generation--;
16732 + spin_unlock(&iinfo->ii_genspin);
16735 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
16740 + if (unlikely(inode && au_iigen(inode, NULL) != sigen))
16746 +/* ---------------------------------------------------------------------- */
16748 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
16749 + aufs_bindex_t bindex)
16751 + IiMustAnyLock(inode);
16752 + return au_ii(inode)->ii_hinode[0 + bindex].hi_id;
16755 +static inline aufs_bindex_t au_ibstart(struct inode *inode)
16757 + IiMustAnyLock(inode);
16758 + return au_ii(inode)->ii_bstart;
16761 +static inline aufs_bindex_t au_ibend(struct inode *inode)
16763 + IiMustAnyLock(inode);
16764 + return au_ii(inode)->ii_bend;
16767 +static inline struct au_vdir *au_ivdir(struct inode *inode)
16769 + IiMustAnyLock(inode);
16770 + return au_ii(inode)->ii_vdir;
16773 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
16775 + IiMustAnyLock(inode);
16776 + return au_ii(inode)->ii_hinode[0 + bindex].hi_whdentry;
16779 +static inline void au_set_ibstart(struct inode *inode, aufs_bindex_t bindex)
16781 + IiMustWriteLock(inode);
16782 + au_ii(inode)->ii_bstart = bindex;
16785 +static inline void au_set_ibend(struct inode *inode, aufs_bindex_t bindex)
16787 + IiMustWriteLock(inode);
16788 + au_ii(inode)->ii_bend = bindex;
16791 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
16793 + IiMustWriteLock(inode);
16794 + au_ii(inode)->ii_vdir = vdir;
16797 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
16799 + IiMustAnyLock(inode);
16800 + return au_ii(inode)->ii_hinode + bindex;
16803 +/* ---------------------------------------------------------------------- */
16805 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
16808 + return pin->parent;
16812 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
16814 + if (pin && pin->hdir)
16815 + return pin->hdir->hi_inode;
16819 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
16822 + return pin->hdir;
16826 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
16829 + pin->dentry = dentry;
16832 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
16833 + unsigned char lflag)
16837 + au_fset_pin(pin->flags, DI_LOCKED);
16839 + au_fclr_pin(pin->flags, DI_LOCKED);
16843 +#if 0 /* reserved */
16844 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
16847 + dput(pin->parent);
16848 + pin->parent = dget(parent);
16853 +/* ---------------------------------------------------------------------- */
16856 +#ifdef CONFIG_AUFS_HNOTIFY
16857 +struct au_hnotify_op {
16858 + void (*ctl)(struct au_hinode *hinode, int do_set);
16859 + int (*alloc)(struct au_hinode *hinode);
16862 + * if it returns true, the the caller should free hinode->hi_notify,
16863 + * otherwise ->free() frees it.
16865 + int (*free)(struct au_hinode *hinode,
16866 + struct au_hnotify *hn) __must_check;
16868 + void (*fin)(void);
16869 + int (*init)(void);
16871 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
16872 + void (*fin_br)(struct au_branch *br);
16873 + int (*init_br)(struct au_branch *br, int perm);
16877 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
16878 +void au_hn_free(struct au_hinode *hinode);
16879 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
16880 +void au_hn_reset(struct inode *inode, unsigned int flags);
16881 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
16882 + struct qstr *h_child_qstr, struct inode *h_child_inode);
16883 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
16884 +int au_hnotify_init_br(struct au_branch *br, int perm);
16885 +void au_hnotify_fin_br(struct au_branch *br);
16886 +int __init au_hnotify_init(void);
16887 +void au_hnotify_fin(void);
16890 +extern const struct au_hnotify_op au_hnotify_op;
16893 +void au_hn_init(struct au_hinode *hinode)
16895 + hinode->hi_notify = NULL;
16898 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
16900 + return hinode->hi_notify;
16904 +AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
16905 + struct au_hinode *hinode __maybe_unused,
16906 + struct inode *inode __maybe_unused)
16907 +AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
16908 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
16909 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
16910 + int do_set __maybe_unused)
16911 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
16912 + unsigned int flags __maybe_unused)
16913 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
16914 + struct au_branch *br __maybe_unused,
16915 + int perm __maybe_unused)
16916 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
16917 + int perm __maybe_unused)
16918 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
16919 +AuStubInt0(__init au_hnotify_init, void)
16920 +AuStubVoid(au_hnotify_fin, void)
16921 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
16922 +#endif /* CONFIG_AUFS_HNOTIFY */
16924 +static inline void au_hn_suspend(struct au_hinode *hdir)
16926 + au_hn_ctl(hdir, /*do_set*/0);
16929 +static inline void au_hn_resume(struct au_hinode *hdir)
16931 + au_hn_ctl(hdir, /*do_set*/1);
16934 +static inline void au_hn_imtx_lock(struct au_hinode *hdir)
16936 + mutex_lock(&hdir->hi_inode->i_mutex);
16937 + au_hn_suspend(hdir);
16940 +static inline void au_hn_imtx_lock_nested(struct au_hinode *hdir,
16941 + unsigned int sc __maybe_unused)
16943 + mutex_lock_nested(&hdir->hi_inode->i_mutex, sc);
16944 + au_hn_suspend(hdir);
16947 +static inline void au_hn_imtx_unlock(struct au_hinode *hdir)
16949 + au_hn_resume(hdir);
16950 + mutex_unlock(&hdir->hi_inode->i_mutex);
16953 +#endif /* __KERNEL__ */
16954 +#endif /* __AUFS_INODE_H__ */
16955 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
16956 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
16957 +++ linux/fs/aufs/ioctl.c 2015-09-24 10:47:58.254719746 +0200
16960 + * Copyright (C) 2005-2015 Junjiro R. Okajima
16962 + * This program, aufs is free software; you can redistribute it and/or modify
16963 + * it under the terms of the GNU General Public License as published by
16964 + * the Free Software Foundation; either version 2 of the License, or
16965 + * (at your option) any later version.
16967 + * This program is distributed in the hope that it will be useful,
16968 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16969 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16970 + * GNU General Public License for more details.
16972 + * You should have received a copy of the GNU General Public License
16973 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
16978 + * plink-management and readdir in userspace.
16979 + * assist the pathconf(3) wrapper library.
16981 + * File-based Hierarchical Storage Management.
16984 +#include <linux/compat.h>
16985 +#include <linux/file.h>
16988 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
16991 + aufs_bindex_t wbi, bindex, bend;
16992 + struct file *h_file;
16993 + struct super_block *sb;
16994 + struct dentry *root;
16995 + struct au_branch *br;
16996 + struct aufs_wbr_fd wbrfd = {
16997 + .oflags = au_dir_roflags,
17000 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
17001 + | O_NOATIME | O_CLOEXEC;
17003 + AuDebugOn(wbrfd.oflags & ~valid);
17006 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
17007 + if (unlikely(err)) {
17013 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
17014 + wbrfd.oflags |= au_dir_roflags;
17015 + AuDbg("0%o\n", wbrfd.oflags);
17016 + if (unlikely(wbrfd.oflags & ~valid))
17020 + fd = get_unused_fd_flags(0);
17022 + if (unlikely(fd < 0))
17025 + h_file = ERR_PTR(-EINVAL);
17028 + sb = path->dentry->d_sb;
17029 + root = sb->s_root;
17030 + aufs_read_lock(root, AuLock_IR);
17031 + bend = au_sbend(sb);
17032 + if (wbrfd.brid >= 0) {
17033 + wbi = au_br_index(sb, wbrfd.brid);
17034 + if (unlikely(wbi < 0 || wbi > bend))
17038 + h_file = ERR_PTR(-ENOENT);
17039 + br = au_sbr(sb, wbi);
17040 + if (!au_br_writable(br->br_perm)) {
17044 + bindex = wbi + 1;
17046 + for (; bindex <= bend; bindex++) {
17047 + br = au_sbr(sb, bindex);
17048 + if (au_br_writable(br->br_perm)) {
17050 + br = au_sbr(sb, wbi);
17055 + AuDbg("wbi %d\n", wbi);
17057 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
17061 + aufs_read_unlock(root, AuLock_IR);
17062 + err = PTR_ERR(h_file);
17063 + if (IS_ERR(h_file))
17066 + atomic_dec(&br->br_count); /* cf. au_h_open() */
17067 + fd_install(fd, h_file);
17069 + goto out; /* success */
17072 + put_unused_fd(fd);
17078 +/* ---------------------------------------------------------------------- */
17080 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
17083 + struct dentry *dentry;
17086 + case AUFS_CTL_RDU:
17087 + case AUFS_CTL_RDU_INO:
17088 + err = au_rdu_ioctl(file, cmd, arg);
17091 + case AUFS_CTL_WBR_FD:
17092 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
17095 + case AUFS_CTL_IBUSY:
17096 + err = au_ibusy_ioctl(file, arg);
17099 + case AUFS_CTL_BRINFO:
17100 + err = au_brinfo_ioctl(file, arg);
17103 + case AUFS_CTL_FHSM_FD:
17104 + dentry = file->f_path.dentry;
17105 + if (IS_ROOT(dentry))
17106 + err = au_fhsm_fd(dentry->d_sb, arg);
17112 + /* do not call the lower */
17113 + AuDbg("0x%x\n", cmd);
17121 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
17126 + case AUFS_CTL_MVDOWN:
17127 + err = au_mvdown(file->f_path.dentry, (void __user *)arg);
17130 + case AUFS_CTL_WBR_FD:
17131 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
17135 + /* do not call the lower */
17136 + AuDbg("0x%x\n", cmd);
17144 +#ifdef CONFIG_COMPAT
17145 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
17146 + unsigned long arg)
17151 + case AUFS_CTL_RDU:
17152 + case AUFS_CTL_RDU_INO:
17153 + err = au_rdu_compat_ioctl(file, cmd, arg);
17156 + case AUFS_CTL_IBUSY:
17157 + err = au_ibusy_compat_ioctl(file, arg);
17160 + case AUFS_CTL_BRINFO:
17161 + err = au_brinfo_compat_ioctl(file, arg);
17165 + err = aufs_ioctl_dir(file, cmd, arg);
17172 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
17173 + unsigned long arg)
17175 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
17178 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
17179 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
17180 +++ linux/fs/aufs/i_op_add.c 2015-09-24 10:47:58.254719746 +0200
17183 + * Copyright (C) 2005-2015 Junjiro R. Okajima
17185 + * This program, aufs is free software; you can redistribute it and/or modify
17186 + * it under the terms of the GNU General Public License as published by
17187 + * the Free Software Foundation; either version 2 of the License, or
17188 + * (at your option) any later version.
17190 + * This program is distributed in the hope that it will be useful,
17191 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17192 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17193 + * GNU General Public License for more details.
17195 + * You should have received a copy of the GNU General Public License
17196 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
17200 + * inode operations (add entry)
17206 + * final procedure of adding a new entry, except link(2).
17207 + * remove whiteout, instantiate, copyup the parent dir's times and size
17208 + * and update version.
17209 + * if it failed, re-create the removed whiteout.
17211 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
17212 + struct dentry *wh_dentry, struct dentry *dentry)
17215 + aufs_bindex_t bwh;
17216 + struct path h_path;
17217 + struct super_block *sb;
17218 + struct inode *inode, *h_dir;
17219 + struct dentry *wh;
17224 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
17225 + IMustLock(h_dir);
17226 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
17227 + bwh = au_dbwh(dentry);
17228 + h_path.dentry = wh_dentry;
17229 + h_path.mnt = au_sbr_mnt(sb, bindex);
17230 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
17232 + if (unlikely(err))
17236 + inode = au_new_inode(dentry, /*must_new*/1);
17237 + if (!IS_ERR(inode)) {
17238 + d_instantiate(dentry, inode);
17239 + dir = d_inode(dentry->d_parent); /* dir inode is locked */
17241 + au_dir_ts(dir, bindex);
17242 + dir->i_version++;
17243 + au_fhsm_wrote(sb, bindex, /*force*/0);
17244 + return 0; /* success */
17247 + err = PTR_ERR(inode);
17252 + /* dir inode is locked */
17253 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
17254 + rerr = PTR_ERR(wh);
17255 + if (IS_ERR(wh)) {
17256 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
17257 + dentry, err, rerr);
17266 +static int au_d_may_add(struct dentry *dentry)
17271 + if (unlikely(d_unhashed(dentry)))
17273 + if (unlikely(d_really_is_positive(dentry)))
17279 + * simple tests for the adding inode operations.
17280 + * following the checks in vfs, plus the parent-child relationship.
17282 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
17283 + struct dentry *h_parent, int isdir)
17287 + struct dentry *h_dentry;
17288 + struct inode *h_inode;
17290 + err = -ENAMETOOLONG;
17291 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
17294 + h_dentry = au_h_dptr(dentry, bindex);
17295 + if (d_really_is_negative(dentry)) {
17297 + if (unlikely(d_is_positive(h_dentry)))
17300 + /* rename(2) case */
17302 + if (unlikely(d_is_negative(h_dentry)))
17304 + h_inode = d_inode(h_dentry);
17305 + if (unlikely(!h_inode->i_nlink))
17308 + h_mode = h_inode->i_mode;
17311 + if (unlikely(S_ISDIR(h_mode)))
17313 + } else if (unlikely(!S_ISDIR(h_mode))) {
17320 + /* expected parent dir is locked */
17321 + if (unlikely(h_parent != h_dentry->d_parent))
17330 + * initial procedure of adding a new entry.
17331 + * prepare writable branch and the parent dir, lock it,
17332 + * and lookup whiteout for the new entry.
17334 +static struct dentry*
17335 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
17336 + struct dentry *src_dentry, struct au_pin *pin,
17337 + struct au_wr_dir_args *wr_dir_args)
17339 + struct dentry *wh_dentry, *h_parent;
17340 + struct super_block *sb;
17341 + struct au_branch *br;
17343 + unsigned int udba;
17344 + aufs_bindex_t bcpup;
17346 + AuDbg("%pd\n", dentry);
17348 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
17350 + wh_dentry = ERR_PTR(err);
17351 + if (unlikely(err < 0))
17354 + sb = dentry->d_sb;
17355 + udba = au_opt_udba(sb);
17356 + err = au_pin(pin, dentry, bcpup, udba,
17357 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17358 + wh_dentry = ERR_PTR(err);
17359 + if (unlikely(err))
17362 + h_parent = au_pinned_h_parent(pin);
17363 + if (udba != AuOpt_UDBA_NONE
17364 + && au_dbstart(dentry) == bcpup)
17365 + err = au_may_add(dentry, bcpup, h_parent,
17366 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
17367 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
17368 + err = -ENAMETOOLONG;
17369 + wh_dentry = ERR_PTR(err);
17370 + if (unlikely(err))
17373 + br = au_sbr(sb, bcpup);
17375 + struct path tmp = {
17376 + .dentry = h_parent,
17377 + .mnt = au_br_mnt(br)
17379 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
17382 + wh_dentry = NULL;
17383 + if (bcpup != au_dbwh(dentry))
17384 + goto out; /* success */
17387 + * ENAMETOOLONG here means that if we allowed create such name, then it
17388 + * would not be able to removed in the future. So we don't allow such
17389 + * name here and we don't handle ENAMETOOLONG differently here.
17391 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
17394 + if (IS_ERR(wh_dentry))
17397 + return wh_dentry;
17400 +/* ---------------------------------------------------------------------- */
17402 +enum { Mknod, Symlink, Creat };
17403 +struct simple_arg {
17410 + struct vfsub_aopen_args *aopen;
17413 + const char *symname;
17422 +static int add_simple(struct inode *dir, struct dentry *dentry,
17423 + struct simple_arg *arg)
17426 + aufs_bindex_t bstart;
17427 + unsigned char created;
17428 + const unsigned char try_aopen
17429 + = (arg->type == Creat && arg->u.c.try_aopen);
17430 + struct dentry *wh_dentry, *parent;
17431 + struct inode *h_dir;
17432 + struct super_block *sb;
17433 + struct au_branch *br;
17434 + /* to reuduce stack size */
17436 + struct au_dtime dt;
17437 + struct au_pin pin;
17438 + struct path h_path;
17439 + struct au_wr_dir_args wr_dir_args;
17442 + AuDbg("%pd\n", dentry);
17446 + a = kmalloc(sizeof(*a), GFP_NOFS);
17447 + if (unlikely(!a))
17449 + a->wr_dir_args.force_btgt = -1;
17450 + a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
17452 + parent = dentry->d_parent; /* dir inode is locked */
17453 + if (!try_aopen) {
17454 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
17455 + if (unlikely(err))
17458 + err = au_d_may_add(dentry);
17459 + if (unlikely(err))
17462 + di_write_lock_parent(parent);
17463 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
17464 + &a->pin, &a->wr_dir_args);
17465 + err = PTR_ERR(wh_dentry);
17466 + if (IS_ERR(wh_dentry))
17469 + bstart = au_dbstart(dentry);
17470 + sb = dentry->d_sb;
17471 + br = au_sbr(sb, bstart);
17472 + a->h_path.dentry = au_h_dptr(dentry, bstart);
17473 + a->h_path.mnt = au_br_mnt(br);
17474 + h_dir = au_pinned_h_dir(&a->pin);
17475 + switch (arg->type) {
17478 + if (!try_aopen || !h_dir->i_op->atomic_open)
17479 + err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
17480 + arg->u.c.want_excl);
17482 + err = vfsub_atomic_open(h_dir, a->h_path.dentry,
17483 + arg->u.c.aopen, br);
17486 + err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
17489 + err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
17497 + err = epilog(dir, bstart, wh_dentry, dentry);
17500 + if (unlikely(created && err && d_is_positive(a->h_path.dentry))) {
17501 + /* no delegation since it is just created */
17502 + rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
17505 + AuIOErr("%pd revert failure(%d, %d)\n",
17506 + dentry, err, rerr);
17509 + au_dtime_revert(&a->dt);
17512 + if (!err && try_aopen && !h_dir->i_op->atomic_open)
17513 + *arg->u.c.aopen->opened |= FILE_CREATED;
17515 + au_unpin(&a->pin);
17520 + di_write_unlock(parent);
17522 + if (unlikely(err)) {
17523 + au_update_dbstart(dentry);
17527 + aufs_read_unlock(dentry, AuLock_DW);
17534 +int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
17537 + struct simple_arg arg = {
17544 + return add_simple(dir, dentry, &arg);
17547 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
17549 + struct simple_arg arg = {
17551 + .u.s.symname = symname
17553 + return add_simple(dir, dentry, &arg);
17556 +int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
17559 + struct simple_arg arg = {
17563 + .want_excl = want_excl
17566 + return add_simple(dir, dentry, &arg);
17569 +int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
17570 + struct vfsub_aopen_args *aopen_args)
17572 + struct simple_arg arg = {
17575 + .mode = aopen_args->create_mode,
17576 + .want_excl = aopen_args->open_flag & O_EXCL,
17577 + .try_aopen = true,
17578 + .aopen = aopen_args
17581 + return add_simple(dir, dentry, &arg);
17584 +int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
17587 + aufs_bindex_t bindex;
17588 + struct super_block *sb;
17589 + struct dentry *parent, *h_parent, *h_dentry;
17590 + struct inode *h_dir, *inode;
17591 + struct vfsmount *h_mnt;
17592 + struct au_wr_dir_args wr_dir_args = {
17593 + .force_btgt = -1,
17594 + .flags = AuWrDir_TMPFILE
17597 + /* copy-up may happen */
17598 + mutex_lock(&dir->i_mutex);
17601 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
17602 + if (unlikely(err))
17605 + err = au_di_init(dentry);
17606 + if (unlikely(err))
17610 + parent = d_find_any_alias(dir);
17611 + AuDebugOn(!parent);
17612 + di_write_lock_parent(parent);
17613 + if (unlikely(d_inode(parent) != dir))
17616 + err = au_digen_test(parent, au_sigen(sb));
17617 + if (unlikely(err))
17620 + bindex = au_dbstart(parent);
17621 + au_set_dbstart(dentry, bindex);
17622 + au_set_dbend(dentry, bindex);
17623 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
17625 + if (unlikely(err < 0))
17628 + err = -EOPNOTSUPP;
17629 + h_dir = au_h_iptr(dir, bindex);
17630 + if (unlikely(!h_dir->i_op->tmpfile))
17633 + h_mnt = au_sbr_mnt(sb, bindex);
17634 + err = vfsub_mnt_want_write(h_mnt);
17635 + if (unlikely(err))
17638 + h_parent = au_h_dptr(parent, bindex);
17639 + err = inode_permission(d_inode(h_parent), MAY_WRITE | MAY_EXEC);
17640 + if (unlikely(err))
17644 + h_dentry = d_alloc(h_parent, &dentry->d_name);
17645 + if (unlikely(!h_dentry))
17648 + err = h_dir->i_op->tmpfile(h_dir, h_dentry, mode);
17649 + if (unlikely(err))
17652 + au_set_dbstart(dentry, bindex);
17653 + au_set_dbend(dentry, bindex);
17654 + au_set_h_dptr(dentry, bindex, dget(h_dentry));
17655 + inode = au_new_inode(dentry, /*must_new*/1);
17656 + if (IS_ERR(inode)) {
17657 + err = PTR_ERR(inode);
17658 + au_set_h_dptr(dentry, bindex, NULL);
17659 + au_set_dbstart(dentry, -1);
17660 + au_set_dbend(dentry, -1);
17662 + if (!inode->i_nlink)
17663 + set_nlink(inode, 1);
17664 + d_tmpfile(dentry, inode);
17665 + au_di(dentry)->di_tmpfile = 1;
17667 + /* update without i_mutex */
17668 + if (au_ibstart(dir) == au_dbstart(dentry))
17669 + au_cpup_attr_timesizes(dir);
17675 + vfsub_mnt_drop_write(h_mnt);
17677 + di_write_unlock(parent);
17679 + di_write_unlock(dentry);
17682 + /* verbose coding for lock class name */
17683 + au_rw_class(&au_di(dentry)->di_rwsem,
17684 + au_lc_key + AuLcNonDir_DIINFO);
17689 + au_di_fin(dentry);
17690 + dentry->d_fsdata = NULL;
17693 + si_read_unlock(sb);
17695 + mutex_unlock(&dir->i_mutex);
17699 +/* ---------------------------------------------------------------------- */
17701 +struct au_link_args {
17702 + aufs_bindex_t bdst, bsrc;
17703 + struct au_pin pin;
17704 + struct path h_path;
17705 + struct dentry *src_parent, *parent;
17708 +static int au_cpup_before_link(struct dentry *src_dentry,
17709 + struct au_link_args *a)
17712 + struct dentry *h_src_dentry;
17713 + struct au_cp_generic cpg = {
17714 + .dentry = src_dentry,
17719 + .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
17722 + di_read_lock_parent(a->src_parent, AuLock_IR);
17723 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
17724 + if (unlikely(err))
17727 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
17728 + err = au_pin(&a->pin, src_dentry, a->bdst,
17729 + au_opt_udba(src_dentry->d_sb),
17730 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17731 + if (unlikely(err))
17734 + err = au_sio_cpup_simple(&cpg);
17735 + au_unpin(&a->pin);
17738 + di_read_unlock(a->src_parent, AuLock_IR);
17742 +static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
17743 + struct au_link_args *a)
17746 + unsigned char plink;
17747 + aufs_bindex_t bend;
17748 + struct dentry *h_src_dentry;
17749 + struct inode *h_inode, *inode, *delegated;
17750 + struct super_block *sb;
17751 + struct file *h_file;
17755 + sb = src_dentry->d_sb;
17756 + inode = d_inode(src_dentry);
17757 + if (au_ibstart(inode) <= a->bdst)
17758 + h_inode = au_h_iptr(inode, a->bdst);
17759 + if (!h_inode || !h_inode->i_nlink) {
17760 + /* copyup src_dentry as the name of dentry. */
17761 + bend = au_dbend(dentry);
17762 + if (bend < a->bsrc)
17763 + au_set_dbend(dentry, a->bsrc);
17764 + au_set_h_dptr(dentry, a->bsrc,
17765 + dget(au_h_dptr(src_dentry, a->bsrc)));
17766 + dget(a->h_path.dentry);
17767 + au_set_h_dptr(dentry, a->bdst, NULL);
17768 + AuDbg("temporary d_inode...\n");
17769 + spin_lock(&dentry->d_lock);
17770 + dentry->d_inode = d_inode(src_dentry); /* tmp */
17771 + spin_unlock(&dentry->d_lock);
17772 + h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
17773 + if (IS_ERR(h_file))
17774 + err = PTR_ERR(h_file);
17776 + struct au_cp_generic cpg = {
17777 + .dentry = dentry,
17782 + .flags = AuCpup_KEEPLINO
17784 + err = au_sio_cpup_simple(&cpg);
17785 + au_h_open_post(dentry, a->bsrc, h_file);
17787 + dput(a->h_path.dentry);
17788 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
17790 + au_set_h_dptr(dentry, a->bdst,
17791 + a->h_path.dentry);
17793 + spin_lock(&dentry->d_lock);
17794 + dentry->d_inode = NULL; /* restore */
17795 + spin_unlock(&dentry->d_lock);
17796 + AuDbg("temporary d_inode...done\n");
17797 + au_set_h_dptr(dentry, a->bsrc, NULL);
17798 + au_set_dbend(dentry, bend);
17800 + /* the inode of src_dentry already exists on a.bdst branch */
17801 + h_src_dentry = d_find_alias(h_inode);
17802 + if (!h_src_dentry && au_plink_test(inode)) {
17804 + h_src_dentry = au_plink_lkup(inode, a->bdst);
17805 + err = PTR_ERR(h_src_dentry);
17806 + if (IS_ERR(h_src_dentry))
17809 + if (unlikely(d_is_negative(h_src_dentry))) {
17810 + dput(h_src_dentry);
17811 + h_src_dentry = NULL;
17815 + if (h_src_dentry) {
17816 + delegated = NULL;
17817 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
17818 + &a->h_path, &delegated);
17819 + if (unlikely(err == -EWOULDBLOCK)) {
17820 + pr_warn("cannot retry for NFSv4 delegation"
17821 + " for an internal link\n");
17824 + dput(h_src_dentry);
17826 + AuIOErr("no dentry found for hi%lu on b%d\n",
17827 + h_inode->i_ino, a->bdst);
17832 + if (!err && !plink)
17833 + au_plink_append(inode, a->bdst, a->h_path.dentry);
17840 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
17841 + struct dentry *dentry)
17844 + struct au_dtime dt;
17845 + struct au_link_args *a;
17846 + struct dentry *wh_dentry, *h_src_dentry;
17847 + struct inode *inode, *delegated;
17848 + struct super_block *sb;
17849 + struct au_wr_dir_args wr_dir_args = {
17850 + /* .force_btgt = -1, */
17851 + .flags = AuWrDir_ADD_ENTRY
17855 + inode = d_inode(src_dentry);
17856 + IMustLock(inode);
17859 + a = kzalloc(sizeof(*a), GFP_NOFS);
17860 + if (unlikely(!a))
17863 + a->parent = dentry->d_parent; /* dir inode is locked */
17864 + err = aufs_read_and_write_lock2(dentry, src_dentry,
17865 + AuLock_NOPLM | AuLock_GEN);
17866 + if (unlikely(err))
17868 + err = au_d_linkable(src_dentry);
17869 + if (unlikely(err))
17871 + err = au_d_may_add(dentry);
17872 + if (unlikely(err))
17875 + a->src_parent = dget_parent(src_dentry);
17876 + wr_dir_args.force_btgt = au_ibstart(inode);
17878 + di_write_lock_parent(a->parent);
17879 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
17880 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
17882 + err = PTR_ERR(wh_dentry);
17883 + if (IS_ERR(wh_dentry))
17887 + sb = dentry->d_sb;
17888 + a->bdst = au_dbstart(dentry);
17889 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
17890 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
17891 + a->bsrc = au_ibstart(inode);
17892 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
17893 + if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
17894 + h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
17895 + if (!h_src_dentry) {
17896 + a->bsrc = au_dbstart(src_dentry);
17897 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
17898 + AuDebugOn(!h_src_dentry);
17899 + } else if (IS_ERR(h_src_dentry)) {
17900 + err = PTR_ERR(h_src_dentry);
17904 + if (au_opt_test(au_mntflags(sb), PLINK)) {
17905 + if (a->bdst < a->bsrc
17906 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
17907 + err = au_cpup_or_link(src_dentry, dentry, a);
17909 + delegated = NULL;
17910 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
17911 + &a->h_path, &delegated);
17912 + if (unlikely(err == -EWOULDBLOCK)) {
17913 + pr_warn("cannot retry for NFSv4 delegation"
17914 + " for an internal link\n");
17918 + dput(h_src_dentry);
17921 + * copyup src_dentry to the branch we process,
17922 + * and then link(2) to it.
17924 + dput(h_src_dentry);
17925 + if (a->bdst < a->bsrc
17926 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
17927 + au_unpin(&a->pin);
17928 + di_write_unlock(a->parent);
17929 + err = au_cpup_before_link(src_dentry, a);
17930 + di_write_lock_parent(a->parent);
17932 + err = au_pin(&a->pin, dentry, a->bdst,
17934 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17935 + if (unlikely(err))
17939 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
17941 + if (h_src_dentry && d_is_positive(h_src_dentry)) {
17942 + delegated = NULL;
17943 + err = vfsub_link(h_src_dentry,
17944 + au_pinned_h_dir(&a->pin),
17945 + &a->h_path, &delegated);
17946 + if (unlikely(err == -EWOULDBLOCK)) {
17947 + pr_warn("cannot retry"
17948 + " for NFSv4 delegation"
17949 + " for an internal link\n");
17955 + if (unlikely(err))
17959 + a->h_path.dentry = wh_dentry;
17960 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
17962 + if (unlikely(err))
17966 + au_dir_ts(dir, a->bdst);
17967 + dir->i_version++;
17968 + inc_nlink(inode);
17969 + inode->i_ctime = dir->i_ctime;
17970 + d_instantiate(dentry, au_igrab(inode));
17971 + if (d_unhashed(a->h_path.dentry))
17972 + /* some filesystem calls d_drop() */
17974 + /* some filesystems consume an inode even hardlink */
17975 + au_fhsm_wrote(sb, a->bdst, /*force*/0);
17976 + goto out_unpin; /* success */
17979 + /* no delegation since it is just created */
17980 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
17981 + /*delegated*/NULL, /*force*/0);
17982 + if (unlikely(rerr)) {
17983 + AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
17986 + au_dtime_revert(&dt);
17988 + au_unpin(&a->pin);
17992 + di_write_unlock(a->parent);
17993 + dput(a->src_parent);
17995 + if (unlikely(err)) {
17996 + au_update_dbstart(dentry);
17999 + aufs_read_and_write_unlock2(dentry, src_dentry);
18007 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
18010 + aufs_bindex_t bindex;
18011 + unsigned char diropq;
18012 + struct path h_path;
18013 + struct dentry *wh_dentry, *parent, *opq_dentry;
18014 + struct mutex *h_mtx;
18015 + struct super_block *sb;
18017 + struct au_pin pin;
18018 + struct au_dtime dt;
18019 + } *a; /* reduce the stack usage */
18020 + struct au_wr_dir_args wr_dir_args = {
18021 + .force_btgt = -1,
18022 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
18028 + a = kmalloc(sizeof(*a), GFP_NOFS);
18029 + if (unlikely(!a))
18032 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
18033 + if (unlikely(err))
18035 + err = au_d_may_add(dentry);
18036 + if (unlikely(err))
18039 + parent = dentry->d_parent; /* dir inode is locked */
18040 + di_write_lock_parent(parent);
18041 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
18042 + &a->pin, &wr_dir_args);
18043 + err = PTR_ERR(wh_dentry);
18044 + if (IS_ERR(wh_dentry))
18047 + sb = dentry->d_sb;
18048 + bindex = au_dbstart(dentry);
18049 + h_path.dentry = au_h_dptr(dentry, bindex);
18050 + h_path.mnt = au_sbr_mnt(sb, bindex);
18051 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
18052 + if (unlikely(err))
18055 + /* make the dir opaque */
18057 + h_mtx = &d_inode(h_path.dentry)->i_mutex;
18059 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
18060 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18061 + opq_dentry = au_diropq_create(dentry, bindex);
18062 + mutex_unlock(h_mtx);
18063 + err = PTR_ERR(opq_dentry);
18064 + if (IS_ERR(opq_dentry))
18066 + dput(opq_dentry);
18070 + err = epilog(dir, bindex, wh_dentry, dentry);
18073 + goto out_unpin; /* success */
18078 + AuLabel(revert opq);
18079 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18080 + rerr = au_diropq_remove(dentry, bindex);
18081 + mutex_unlock(h_mtx);
18083 + AuIOErr("%pd reverting diropq failed(%d, %d)\n",
18084 + dentry, err, rerr);
18090 + AuLabel(revert dir);
18091 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
18093 + AuIOErr("%pd reverting dir failed(%d, %d)\n",
18094 + dentry, err, rerr);
18097 + au_dtime_revert(&a->dt);
18099 + au_unpin(&a->pin);
18102 + di_write_unlock(parent);
18104 + if (unlikely(err)) {
18105 + au_update_dbstart(dentry);
18108 + aufs_read_unlock(dentry, AuLock_DW);
18114 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
18115 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
18116 +++ linux/fs/aufs/i_op.c 2015-11-11 17:21:46.918863802 +0100
18119 + * Copyright (C) 2005-2015 Junjiro R. Okajima
18121 + * This program, aufs is free software; you can redistribute it and/or modify
18122 + * it under the terms of the GNU General Public License as published by
18123 + * the Free Software Foundation; either version 2 of the License, or
18124 + * (at your option) any later version.
18126 + * This program is distributed in the hope that it will be useful,
18127 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18128 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18129 + * GNU General Public License for more details.
18131 + * You should have received a copy of the GNU General Public License
18132 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
18136 + * inode operations (except add/del/rename)
18139 +#include <linux/device_cgroup.h>
18140 +#include <linux/fs_stack.h>
18141 +#include <linux/namei.h>
18142 +#include <linux/security.h>
18145 +static int h_permission(struct inode *h_inode, int mask,
18146 + struct path *h_path, int brperm)
18149 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
18152 + if ((write_mask && IS_IMMUTABLE(h_inode))
18153 + || ((mask & MAY_EXEC)
18154 + && S_ISREG(h_inode->i_mode)
18155 + && (path_noexec(h_path)
18156 + || !(h_inode->i_mode & S_IXUGO))))
18160 + * - skip the lower fs test in the case of write to ro branch.
18161 + * - nfs dir permission write check is optimized, but a policy for
18162 + * link/rename requires a real check.
18163 + * - nfs always sets MS_POSIXACL regardless its mount option 'noacl.'
18164 + * in this case, generic_permission() returns -EOPNOTSUPP.
18166 + if ((write_mask && !au_br_writable(brperm))
18167 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
18168 + && write_mask && !(mask & MAY_READ))
18169 + || !h_inode->i_op->permission) {
18170 + /* AuLabel(generic_permission); */
18171 + /* AuDbg("get_acl %pf\n", h_inode->i_op->get_acl); */
18172 + err = generic_permission(h_inode, mask);
18173 + if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
18174 + err = h_inode->i_op->permission(h_inode, mask);
18177 + /* AuLabel(h_inode->permission); */
18178 + err = h_inode->i_op->permission(h_inode, mask);
18183 + err = devcgroup_inode_permission(h_inode, mask);
18185 + err = security_inode_permission(h_inode, mask);
18189 + /* todo: do we need to call ima_path_check()? */
18190 + struct path h_path = {
18194 + err = ima_path_check(&h_path,
18195 + mask & (MAY_READ | MAY_WRITE | MAY_EXEC),
18196 + IMA_COUNT_LEAVE);
18204 +static int aufs_permission(struct inode *inode, int mask)
18207 + aufs_bindex_t bindex, bend;
18208 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
18209 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
18210 + struct inode *h_inode;
18211 + struct super_block *sb;
18212 + struct au_branch *br;
18214 + /* todo: support rcu-walk? */
18215 + if (mask & MAY_NOT_BLOCK)
18218 + sb = inode->i_sb;
18219 + si_read_lock(sb, AuLock_FLUSH);
18220 + ii_read_lock_child(inode);
18222 + err = au_iigen_test(inode, au_sigen(sb));
18223 + if (unlikely(err))
18229 + || au_opt_test(au_mntflags(sb), DIRPERM1)) {
18230 + err = au_busy_or_stale();
18231 + h_inode = au_h_iptr(inode, au_ibstart(inode));
18232 + if (unlikely(!h_inode
18233 + || (h_inode->i_mode & S_IFMT)
18234 + != (inode->i_mode & S_IFMT)))
18238 + bindex = au_ibstart(inode);
18239 + br = au_sbr(sb, bindex);
18240 + err = h_permission(h_inode, mask, &br->br_path, br->br_perm);
18243 + && !special_file(h_inode->i_mode)) {
18244 + /* test whether the upper writable branch exists */
18246 + for (; bindex >= 0; bindex--)
18247 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
18255 + /* non-write to dir */
18257 + bend = au_ibend(inode);
18258 + for (bindex = au_ibstart(inode); !err && bindex <= bend; bindex++) {
18259 + h_inode = au_h_iptr(inode, bindex);
18261 + err = au_busy_or_stale();
18262 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
18265 + br = au_sbr(sb, bindex);
18266 + err = h_permission(h_inode, mask, &br->br_path,
18272 + ii_read_unlock(inode);
18273 + si_read_unlock(sb);
18277 +/* ---------------------------------------------------------------------- */
18279 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
18280 + unsigned int flags)
18282 + struct dentry *ret, *parent;
18283 + struct inode *inode;
18284 + struct super_block *sb;
18285 + int err, npositive;
18289 + /* todo: support rcu-walk? */
18290 + ret = ERR_PTR(-ECHILD);
18291 + if (flags & LOOKUP_RCU)
18294 + ret = ERR_PTR(-ENAMETOOLONG);
18295 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
18299 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
18300 + ret = ERR_PTR(err);
18301 + if (unlikely(err))
18304 + err = au_di_init(dentry);
18305 + ret = ERR_PTR(err);
18306 + if (unlikely(err))
18310 + npositive = 0; /* suppress a warning */
18311 + parent = dentry->d_parent; /* dir inode is locked */
18312 + di_read_lock_parent(parent, AuLock_IR);
18313 + err = au_alive_dir(parent);
18315 + err = au_digen_test(parent, au_sigen(sb));
18317 + npositive = au_lkup_dentry(dentry, au_dbstart(parent),
18321 + di_read_unlock(parent, AuLock_IR);
18322 + ret = ERR_PTR(err);
18323 + if (unlikely(err < 0))
18327 + inode = au_new_inode(dentry, /*must_new*/0);
18328 + if (IS_ERR(inode)) {
18329 + ret = (void *)inode;
18336 + atomic_inc(&inode->i_count);
18337 + ret = d_splice_alias(inode, dentry);
18339 + if (unlikely(d_need_lookup(dentry))) {
18340 + spin_lock(&dentry->d_lock);
18341 + dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
18342 + spin_unlock(&dentry->d_lock);
18346 + if (!IS_ERR(ret)) {
18348 + if (ret && ret != dentry)
18349 + ii_write_unlock(inode);
18351 + ii_write_unlock(inode);
18358 + di_write_unlock(dentry);
18360 + /* verbose coding for lock class name */
18361 + if (unlikely(S_ISLNK(inode->i_mode)))
18362 + au_rw_class(&au_di(dentry)->di_rwsem,
18363 + au_lc_key + AuLcSymlink_DIINFO);
18364 + else if (unlikely(S_ISDIR(inode->i_mode)))
18365 + au_rw_class(&au_di(dentry)->di_rwsem,
18366 + au_lc_key + AuLcDir_DIINFO);
18367 + else /* likely */
18368 + au_rw_class(&au_di(dentry)->di_rwsem,
18369 + au_lc_key + AuLcNonDir_DIINFO);
18372 + si_read_unlock(sb);
18377 +/* ---------------------------------------------------------------------- */
18379 +struct aopen_node {
18380 + struct hlist_node hlist;
18381 + struct file *file, *h_file;
18384 +static int au_do_aopen(struct inode *inode, struct file *file)
18386 + struct au_sphlhead *aopen;
18387 + struct aopen_node *node;
18388 + struct au_do_open_args args = {
18390 + .open = au_do_open_nondir
18393 + aopen = &au_sbi(inode->i_sb)->si_aopen;
18394 + spin_lock(&aopen->spin);
18395 + hlist_for_each_entry(node, &aopen->head, hlist)
18396 + if (node->file == file) {
18397 + args.h_file = node->h_file;
18400 + spin_unlock(&aopen->spin);
18401 + /* AuDebugOn(!args.h_file); */
18403 + return au_do_open(file, &args);
18406 +static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
18407 + struct file *file, unsigned int open_flag,
18408 + umode_t create_mode, int *opened)
18410 + int err, h_opened = *opened;
18411 + struct dentry *parent;
18412 + struct dentry *d;
18413 + struct au_sphlhead *aopen;
18414 + struct vfsub_aopen_args args = {
18415 + .open_flag = open_flag,
18416 + .create_mode = create_mode,
18417 + .opened = &h_opened
18419 + struct aopen_node aopen_node = {
18424 + AuDbg("open_flag 0x%x\n", open_flag);
18425 + AuDbgDentry(dentry);
18428 + if (!au_di(dentry)) {
18429 + d = aufs_lookup(dir, dentry, /*flags*/0);
18431 + err = PTR_ERR(d);
18435 + * obsoleted dentry found.
18436 + * another error will be returned later.
18442 + AuDbgDentry(dentry);
18445 + if (d_is_positive(dentry)
18446 + || d_unhashed(dentry)
18447 + || d_unlinked(dentry)
18448 + || !(open_flag & O_CREAT))
18449 + goto out_no_open;
18451 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
18452 + if (unlikely(err))
18455 + parent = dentry->d_parent; /* dir is locked */
18456 + di_write_lock_parent(parent);
18457 + err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
18458 + if (unlikely(err))
18461 + AuDbgDentry(dentry);
18462 + if (d_is_positive(dentry))
18465 + args.file = get_empty_filp();
18466 + err = PTR_ERR(args.file);
18467 + if (IS_ERR(args.file))
18470 + args.file->f_flags = file->f_flags;
18471 + err = au_aopen_or_create(dir, dentry, &args);
18473 + AuDbgFile(args.file);
18474 + if (unlikely(err < 0)) {
18475 + if (h_opened & FILE_OPENED)
18478 + put_filp(args.file);
18482 + /* some filesystems don't set FILE_CREATED while succeeded? */
18483 + *opened |= FILE_CREATED;
18484 + if (h_opened & FILE_OPENED)
18485 + aopen_node.h_file = args.file;
18487 + put_filp(args.file);
18488 + args.file = NULL;
18490 + aopen = &au_sbi(dir->i_sb)->si_aopen;
18491 + au_sphl_add(&aopen_node.hlist, aopen);
18492 + err = finish_open(file, dentry, au_do_aopen, opened);
18493 + au_sphl_del(&aopen_node.hlist, aopen);
18496 + if (aopen_node.h_file)
18497 + fput(aopen_node.h_file);
18500 + di_write_unlock(parent);
18501 + aufs_read_unlock(dentry, AuLock_DW);
18502 + AuDbgDentry(dentry);
18503 + if (unlikely(err))
18506 + if (!err && !(*opened & FILE_CREATED)) {
18507 + AuLabel(out_no_open);
18509 + err = finish_no_open(file, dentry);
18512 + AuDbg("%pd%s%s\n", dentry,
18513 + (*opened & FILE_CREATED) ? " created" : "",
18514 + (*opened & FILE_OPENED) ? " opened" : "");
18520 +/* ---------------------------------------------------------------------- */
18522 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
18523 + const unsigned char add_entry, aufs_bindex_t bcpup,
18524 + aufs_bindex_t bstart)
18527 + struct dentry *h_parent;
18528 + struct inode *h_dir;
18531 + IMustLock(d_inode(parent));
18533 + di_write_lock_parent(parent);
18536 + if (!au_h_dptr(parent, bcpup)) {
18537 + if (bstart > bcpup)
18538 + err = au_cpup_dirs(dentry, bcpup);
18539 + else if (bstart < bcpup)
18540 + err = au_cpdown_dirs(dentry, bcpup);
18544 + if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
18545 + h_parent = au_h_dptr(parent, bcpup);
18546 + h_dir = d_inode(h_parent);
18547 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
18548 + err = au_lkup_neg(dentry, bcpup, /*wh*/0);
18549 + /* todo: no unlock here */
18550 + mutex_unlock(&h_dir->i_mutex);
18552 + AuDbg("bcpup %d\n", bcpup);
18554 + if (d_really_is_negative(dentry))
18555 + au_set_h_dptr(dentry, bstart, NULL);
18556 + au_update_dbrange(dentry, /*do_put_zero*/0);
18561 + di_write_unlock(parent);
18563 + err = bcpup; /* success */
18570 + * decide the branch and the parent dir where we will create a new entry.
18571 + * returns new bindex or an error.
18572 + * copyup the parent dir if needed.
18574 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
18575 + struct au_wr_dir_args *args)
18578 + unsigned int flags;
18579 + aufs_bindex_t bcpup, bstart, src_bstart;
18580 + const unsigned char add_entry
18581 + = au_ftest_wrdir(args->flags, ADD_ENTRY)
18582 + | au_ftest_wrdir(args->flags, TMPFILE);
18583 + struct super_block *sb;
18584 + struct dentry *parent;
18585 + struct au_sbinfo *sbinfo;
18587 + sb = dentry->d_sb;
18588 + sbinfo = au_sbi(sb);
18589 + parent = dget_parent(dentry);
18590 + bstart = au_dbstart(dentry);
18592 + if (args->force_btgt < 0) {
18593 + if (src_dentry) {
18594 + src_bstart = au_dbstart(src_dentry);
18595 + if (src_bstart < bstart)
18596 + bcpup = src_bstart;
18597 + } else if (add_entry) {
18599 + if (au_ftest_wrdir(args->flags, ISDIR))
18600 + au_fset_wbr(flags, DIR);
18601 + err = AuWbrCreate(sbinfo, dentry, flags);
18605 + if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
18607 + err = AuWbrCopyup(sbinfo, dentry);
18609 + if (!IS_ROOT(dentry)) {
18610 + di_read_lock_parent(parent, !AuLock_IR);
18611 + err = AuWbrCopyup(sbinfo, dentry);
18612 + di_read_unlock(parent, !AuLock_IR);
18614 + err = AuWbrCopyup(sbinfo, dentry);
18617 + if (unlikely(err < 0))
18621 + bcpup = args->force_btgt;
18622 + AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
18625 + AuDbg("bstart %d, bcpup %d\n", bstart, bcpup);
18627 + if (bcpup == bstart)
18628 + goto out; /* success */
18630 + /* copyup the new parent into the branch we process */
18631 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, bstart);
18633 + if (d_really_is_negative(dentry)) {
18634 + au_set_h_dptr(dentry, bstart, NULL);
18635 + au_set_dbstart(dentry, bcpup);
18636 + au_set_dbend(dentry, bcpup);
18638 + AuDebugOn(add_entry
18639 + && !au_ftest_wrdir(args->flags, TMPFILE)
18640 + && !au_h_dptr(dentry, bcpup));
18648 +/* ---------------------------------------------------------------------- */
18650 +void au_pin_hdir_unlock(struct au_pin *p)
18653 + au_hn_imtx_unlock(p->hdir);
18656 +int au_pin_hdir_lock(struct au_pin *p)
18664 + /* even if an error happens later, keep this lock */
18665 + au_hn_imtx_lock_nested(p->hdir, p->lsc_hi);
18668 + if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
18673 + err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
18674 + p->h_parent, p->br);
18680 +int au_pin_hdir_relock(struct au_pin *p)
18683 + struct inode *h_i;
18684 + struct dentry *h_d[] = {
18689 + err = au_pin_hdir_lock(p);
18690 + if (unlikely(err))
18693 + for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
18696 + if (d_is_positive(h_d[i])) {
18697 + h_i = d_inode(h_d[i]);
18698 + err = !h_i->i_nlink;
18706 +void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
18708 +#if defined(CONFIG_DEBUG_MUTEXES) || defined(CONFIG_SMP)
18709 + p->hdir->hi_inode->i_mutex.owner = task;
18713 +void au_pin_hdir_acquire_nest(struct au_pin *p)
18716 + mutex_acquire_nest(&p->hdir->hi_inode->i_mutex.dep_map,
18717 + p->lsc_hi, 0, NULL, _RET_IP_);
18718 + au_pin_hdir_set_owner(p, current);
18722 +void au_pin_hdir_release(struct au_pin *p)
18725 + au_pin_hdir_set_owner(p, p->task);
18726 + mutex_release(&p->hdir->hi_inode->i_mutex.dep_map, 1, _RET_IP_);
18730 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
18732 + if (pin && pin->parent)
18733 + return au_h_dptr(pin->parent, pin->bindex);
18737 +void au_unpin(struct au_pin *p)
18740 + au_pin_hdir_unlock(p);
18741 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
18742 + vfsub_mnt_drop_write(p->h_mnt);
18746 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18747 + di_read_unlock(p->parent, AuLock_IR);
18748 + iput(p->hdir->hi_inode);
18750 + p->parent = NULL;
18753 + /* do not clear p->task */
18756 +int au_do_pin(struct au_pin *p)
18759 + struct super_block *sb;
18760 + struct inode *h_dir;
18763 + sb = p->dentry->d_sb;
18764 + p->br = au_sbr(sb, p->bindex);
18765 + if (IS_ROOT(p->dentry)) {
18766 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
18767 + p->h_mnt = au_br_mnt(p->br);
18768 + err = vfsub_mnt_want_write(p->h_mnt);
18769 + if (unlikely(err)) {
18770 + au_fclr_pin(p->flags, MNT_WRITE);
18777 + p->h_dentry = NULL;
18778 + if (p->bindex <= au_dbend(p->dentry))
18779 + p->h_dentry = au_h_dptr(p->dentry, p->bindex);
18781 + p->parent = dget_parent(p->dentry);
18782 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18783 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
18786 + p->h_parent = au_h_dptr(p->parent, p->bindex);
18787 + p->hdir = au_hi(d_inode(p->parent), p->bindex);
18789 + h_dir = p->hdir->hi_inode;
18793 + * if DI_LOCKED is not set, then p->parent may be different
18794 + * and h_parent can be NULL.
18796 + if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
18798 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18799 + di_read_unlock(p->parent, AuLock_IR);
18801 + p->parent = NULL;
18805 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
18806 + p->h_mnt = au_br_mnt(p->br);
18807 + err = vfsub_mnt_want_write(p->h_mnt);
18808 + if (unlikely(err)) {
18809 + au_fclr_pin(p->flags, MNT_WRITE);
18810 + if (!au_ftest_pin(p->flags, DI_LOCKED))
18811 + di_read_unlock(p->parent, AuLock_IR);
18813 + p->parent = NULL;
18819 + err = au_pin_hdir_lock(p);
18821 + goto out; /* success */
18826 + pr_err("err %d\n", err);
18827 + err = au_busy_or_stale();
18832 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
18833 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
18834 + unsigned int udba, unsigned char flags)
18836 + p->dentry = dentry;
18838 + p->lsc_di = lsc_di;
18839 + p->lsc_hi = lsc_hi;
18840 + p->flags = flags;
18841 + p->bindex = bindex;
18843 + p->parent = NULL;
18847 + p->h_dentry = NULL;
18848 + p->h_parent = NULL;
18850 + p->task = current;
18853 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
18854 + unsigned int udba, unsigned char flags)
18856 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
18858 + return au_do_pin(pin);
18861 +/* ---------------------------------------------------------------------- */
18864 + * ->setattr() and ->getattr() are called in various cases.
18865 + * chmod, stat: dentry is revalidated.
18866 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
18868 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
18870 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
18871 +int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
18874 + struct dentry *parent;
18877 + if (au_digen_test(dentry, sigen)) {
18878 + parent = dget_parent(dentry);
18879 + di_read_lock_parent(parent, AuLock_IR);
18880 + err = au_refresh_dentry(dentry, parent);
18881 + di_read_unlock(parent, AuLock_IR);
18889 +int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
18890 + struct au_icpup_args *a)
18894 + aufs_bindex_t bstart, ibstart;
18895 + struct dentry *hi_wh, *parent;
18896 + struct inode *inode;
18897 + struct au_wr_dir_args wr_dir_args = {
18898 + .force_btgt = -1,
18902 + if (d_is_dir(dentry))
18903 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
18904 + /* plink or hi_wh() case */
18905 + bstart = au_dbstart(dentry);
18906 + inode = d_inode(dentry);
18907 + ibstart = au_ibstart(inode);
18908 + if (bstart != ibstart && !au_test_ro(inode->i_sb, ibstart, inode))
18909 + wr_dir_args.force_btgt = ibstart;
18910 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
18911 + if (unlikely(err < 0))
18914 + if (err != bstart)
18915 + au_fset_icpup(a->flags, DID_CPUP);
18918 + a->pin_flags = AuPin_MNT_WRITE;
18920 + if (!IS_ROOT(dentry)) {
18921 + au_fset_pin(a->pin_flags, DI_LOCKED);
18922 + parent = dget_parent(dentry);
18923 + di_write_lock_parent(parent);
18926 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
18927 + if (unlikely(err))
18930 + a->h_path.dentry = au_h_dptr(dentry, bstart);
18932 + a->h_inode = d_inode(a->h_path.dentry);
18933 + if (ia && (ia->ia_valid & ATTR_SIZE)) {
18934 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
18935 + if (ia->ia_size < i_size_read(a->h_inode))
18936 + sz = ia->ia_size;
18937 + mutex_unlock(&a->h_inode->i_mutex);
18941 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
18942 + hi_wh = au_hi_wh(inode, a->btgt);
18944 + struct au_cp_generic cpg = {
18945 + .dentry = dentry,
18951 + err = au_sio_cpup_wh(&cpg, /*file*/NULL);
18952 + if (unlikely(err))
18954 + hi_wh = au_hi_wh(inode, a->btgt);
18955 + /* todo: revalidate hi_wh? */
18960 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
18961 + di_downgrade_lock(parent, AuLock_IR);
18965 + if (!au_ftest_icpup(a->flags, DID_CPUP))
18966 + goto out; /* success */
18968 + if (!d_unhashed(dentry)) {
18969 + struct au_cp_generic cpg = {
18970 + .dentry = dentry,
18975 + .flags = AuCpup_DTIME | AuCpup_HOPEN
18977 + err = au_sio_cpup_simple(&cpg);
18979 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
18980 + } else if (!hi_wh)
18981 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
18983 + a->h_path.dentry = hi_wh; /* do not dget here */
18986 + a->h_inode = d_inode(a->h_path.dentry);
18988 + goto out; /* success */
18989 + au_unpin(&a->pin);
18992 + di_write_unlock(parent);
18997 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
19001 +static int aufs_setattr(struct dentry *dentry, struct iattr *ia)
19004 + struct inode *inode, *delegated;
19005 + struct super_block *sb;
19006 + struct file *file;
19007 + struct au_icpup_args *a;
19009 + inode = d_inode(dentry);
19010 + IMustLock(inode);
19013 + a = kzalloc(sizeof(*a), GFP_NOFS);
19014 + if (unlikely(!a))
19017 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
19018 + ia->ia_valid &= ~ATTR_MODE;
19021 + sb = dentry->d_sb;
19022 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19023 + if (unlikely(err))
19026 + if (ia->ia_valid & ATTR_FILE) {
19027 + /* currently ftruncate(2) only */
19028 + AuDebugOn(!d_is_reg(dentry));
19029 + file = ia->ia_file;
19030 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
19031 + if (unlikely(err))
19033 + ia->ia_file = au_hf_top(file);
19034 + a->udba = AuOpt_UDBA_NONE;
19036 + /* fchmod() doesn't pass ia_file */
19037 + a->udba = au_opt_udba(sb);
19038 + di_write_lock_child(dentry);
19039 + /* no d_unlinked(), to set UDBA_NONE for root */
19040 + if (d_unhashed(dentry))
19041 + a->udba = AuOpt_UDBA_NONE;
19042 + if (a->udba != AuOpt_UDBA_NONE) {
19043 + AuDebugOn(IS_ROOT(dentry));
19044 + err = au_reval_for_attr(dentry, au_sigen(sb));
19045 + if (unlikely(err))
19050 + err = au_pin_and_icpup(dentry, ia, a);
19051 + if (unlikely(err < 0))
19053 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
19054 + ia->ia_file = NULL;
19055 + ia->ia_valid &= ~ATTR_FILE;
19058 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
19059 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
19060 + == (ATTR_MODE | ATTR_CTIME)) {
19061 + err = security_path_chmod(&a->h_path, ia->ia_mode);
19062 + if (unlikely(err))
19064 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
19065 + && (ia->ia_valid & ATTR_CTIME)) {
19066 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
19067 + if (unlikely(err))
19071 + if (ia->ia_valid & ATTR_SIZE) {
19074 + if (ia->ia_size < i_size_read(inode))
19076 + truncate_setsize(inode, ia->ia_size);
19079 + if (ia->ia_valid & ATTR_FILE)
19081 + mutex_unlock(&a->h_inode->i_mutex);
19082 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
19083 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
19085 + delegated = NULL;
19087 + err = vfsub_notify_change(&a->h_path, ia, &delegated);
19089 + err = break_deleg_wait(&delegated);
19097 + au_cpup_attr_changeable(inode);
19100 + mutex_unlock(&a->h_inode->i_mutex);
19101 + au_unpin(&a->pin);
19102 + if (unlikely(err))
19103 + au_update_dbstart(dentry);
19105 + di_write_unlock(dentry);
19107 + fi_write_unlock(file);
19108 + ia->ia_file = file;
19109 + ia->ia_valid |= ATTR_FILE;
19112 + si_read_unlock(sb);
19120 +#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
19121 +static int au_h_path_to_set_attr(struct dentry *dentry,
19122 + struct au_icpup_args *a, struct path *h_path)
19125 + struct super_block *sb;
19127 + sb = dentry->d_sb;
19128 + a->udba = au_opt_udba(sb);
19129 + /* no d_unlinked(), to set UDBA_NONE for root */
19130 + if (d_unhashed(dentry))
19131 + a->udba = AuOpt_UDBA_NONE;
19132 + if (a->udba != AuOpt_UDBA_NONE) {
19133 + AuDebugOn(IS_ROOT(dentry));
19134 + err = au_reval_for_attr(dentry, au_sigen(sb));
19135 + if (unlikely(err))
19138 + err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
19139 + if (unlikely(err < 0))
19142 + h_path->dentry = a->h_path.dentry;
19143 + h_path->mnt = au_sbr_mnt(sb, a->btgt);
19149 +ssize_t au_srxattr(struct dentry *dentry, struct au_srxattr *arg)
19152 + struct path h_path;
19153 + struct super_block *sb;
19154 + struct au_icpup_args *a;
19155 + struct inode *inode, *h_inode;
19157 + inode = d_inode(dentry);
19158 + IMustLock(inode);
19161 + a = kzalloc(sizeof(*a), GFP_NOFS);
19162 + if (unlikely(!a))
19165 + sb = dentry->d_sb;
19166 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19167 + if (unlikely(err))
19170 + h_path.dentry = NULL; /* silence gcc */
19171 + di_write_lock_child(dentry);
19172 + err = au_h_path_to_set_attr(dentry, a, &h_path);
19173 + if (unlikely(err))
19176 + mutex_unlock(&a->h_inode->i_mutex);
19177 + switch (arg->type) {
19178 + case AU_XATTR_SET:
19179 + err = vfsub_setxattr(h_path.dentry,
19180 + arg->u.set.name, arg->u.set.value,
19181 + arg->u.set.size, arg->u.set.flags);
19183 + case AU_XATTR_REMOVE:
19184 + err = vfsub_removexattr(h_path.dentry, arg->u.remove.name);
19187 + err = -EOPNOTSUPP;
19188 + h_inode = d_inode(h_path.dentry);
19189 + if (h_inode->i_op->set_acl)
19190 + err = h_inode->i_op->set_acl(h_inode,
19191 + arg->u.acl_set.acl,
19192 + arg->u.acl_set.type);
19196 + au_cpup_attr_timesizes(inode);
19198 + au_unpin(&a->pin);
19199 + if (unlikely(err))
19200 + au_update_dbstart(dentry);
19203 + di_write_unlock(dentry);
19204 + si_read_unlock(sb);
19213 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
19214 + unsigned int nlink)
19218 + inode->i_mode = st->mode;
19219 + /* don't i_[ug]id_write() here */
19220 + inode->i_uid = st->uid;
19221 + inode->i_gid = st->gid;
19222 + inode->i_atime = st->atime;
19223 + inode->i_mtime = st->mtime;
19224 + inode->i_ctime = st->ctime;
19226 + au_cpup_attr_nlink(inode, /*force*/0);
19227 + if (S_ISDIR(inode->i_mode)) {
19228 + n = inode->i_nlink;
19231 + smp_mb(); /* for i_nlink */
19232 + /* 0 can happen */
19233 + set_nlink(inode, n);
19236 + spin_lock(&inode->i_lock);
19237 + inode->i_blocks = st->blocks;
19238 + i_size_write(inode, st->size);
19239 + spin_unlock(&inode->i_lock);
19243 + * common routine for aufs_getattr() and aufs_getxattr().
19244 + * returns zero or negative (an error).
19245 + * @dentry will be read-locked in success.
19247 +int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path)
19250 + unsigned int mnt_flags, sigen;
19251 + unsigned char udba_none;
19252 + aufs_bindex_t bindex;
19253 + struct super_block *sb, *h_sb;
19254 + struct inode *inode;
19256 + h_path->mnt = NULL;
19257 + h_path->dentry = NULL;
19260 + sb = dentry->d_sb;
19261 + mnt_flags = au_mntflags(sb);
19262 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
19264 + /* support fstat(2) */
19265 + if (!d_unlinked(dentry) && !udba_none) {
19266 + sigen = au_sigen(sb);
19267 + err = au_digen_test(dentry, sigen);
19269 + di_read_lock_child(dentry, AuLock_IR);
19270 + err = au_dbrange_test(dentry);
19271 + if (unlikely(err)) {
19272 + di_read_unlock(dentry, AuLock_IR);
19276 + AuDebugOn(IS_ROOT(dentry));
19277 + di_write_lock_child(dentry);
19278 + err = au_dbrange_test(dentry);
19280 + err = au_reval_for_attr(dentry, sigen);
19282 + di_downgrade_lock(dentry, AuLock_IR);
19284 + di_write_unlock(dentry);
19289 + di_read_lock_child(dentry, AuLock_IR);
19291 + inode = d_inode(dentry);
19292 + bindex = au_ibstart(inode);
19293 + h_path->mnt = au_sbr_mnt(sb, bindex);
19294 + h_sb = h_path->mnt->mnt_sb;
19296 + && !au_test_fs_bad_iattr(h_sb)
19298 + goto out; /* success */
19300 + if (au_dbstart(dentry) == bindex)
19301 + h_path->dentry = au_h_dptr(dentry, bindex);
19302 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
19303 + h_path->dentry = au_plink_lkup(inode, bindex);
19304 + if (IS_ERR(h_path->dentry))
19305 + /* pretending success */
19306 + h_path->dentry = NULL;
19308 + dput(h_path->dentry);
19315 +static int aufs_getattr(struct vfsmount *mnt __maybe_unused,
19316 + struct dentry *dentry, struct kstat *st)
19319 + unsigned char positive;
19320 + struct path h_path;
19321 + struct inode *inode;
19322 + struct super_block *sb;
19324 + inode = d_inode(dentry);
19325 + sb = dentry->d_sb;
19326 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19327 + if (unlikely(err))
19329 + err = au_h_path_getattr(dentry, /*force*/0, &h_path);
19330 + if (unlikely(err))
19332 + if (unlikely(!h_path.dentry))
19333 + /* illegally overlapped or something */
19334 + goto out_fill; /* pretending success */
19336 + positive = d_is_positive(h_path.dentry);
19338 + err = vfs_getattr(&h_path, st);
19341 + au_refresh_iattr(inode, st,
19342 + d_inode(h_path.dentry)->i_nlink);
19343 + goto out_fill; /* success */
19349 + generic_fillattr(inode, st);
19351 + di_read_unlock(dentry, AuLock_IR);
19353 + si_read_unlock(sb);
19359 +/* ---------------------------------------------------------------------- */
19363 + * - the number of symlinks is not so many.
19366 + * - sbinfo (instead of iinfo) contains an hlist of struct au_symlink.
19367 + * If iinfo contained the hlist, then it would be rather large waste of memory
19369 + * - struct au_symlink contains the necessary info for h_inode follow_link() and
19373 +struct au_symlink {
19375 + struct hlist_node hlist;
19376 + struct rcu_head rcu;
19379 + struct inode *h_inode;
19383 +static void au_symlink_add(struct super_block *sb, struct au_symlink *slink,
19384 + struct inode *h_inode, void *cookie)
19386 + struct au_sbinfo *sbinfo;
19389 + slink->h_inode = h_inode;
19390 + slink->h_cookie = cookie;
19391 + sbinfo = au_sbi(sb);
19392 + au_sphl_add(&slink->hlist, &sbinfo->si_symlink);
19395 +static void au_symlink_del(struct super_block *sb, struct au_symlink *slink)
19397 + struct au_sbinfo *sbinfo;
19399 + /* do not iput() within rcu */
19400 + iput(slink->h_inode);
19401 + slink->h_inode = NULL;
19402 + sbinfo = au_sbi(sb);
19403 + au_sphl_del_rcu(&slink->hlist, &sbinfo->si_symlink);
19404 + kfree_rcu(slink, rcu);
19407 +static const char *aufs_follow_link(struct dentry *dentry, void **cookie)
19410 + struct inode *inode, *h_inode;
19411 + struct dentry *h_dentry;
19412 + struct au_symlink *slink;
19414 + aufs_bindex_t bindex;
19416 + ret = NULL; /* suppress a warning */
19417 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
19418 + if (unlikely(err))
19421 + err = au_d_hashed_positive(dentry);
19422 + if (unlikely(err))
19426 + inode = d_inode(dentry);
19427 + bindex = au_ibstart(inode);
19428 + h_inode = au_h_iptr(inode, bindex);
19429 + if (unlikely(!h_inode->i_op->follow_link))
19433 + slink = kmalloc(sizeof(*slink), GFP_NOFS);
19434 + if (unlikely(!slink))
19439 + if (au_dbstart(dentry) <= bindex) {
19440 + h_dentry = au_h_dptr(dentry, bindex);
19445 + h_dentry = d_find_any_alias(h_inode);
19446 + if (IS_ERR(h_dentry)) {
19447 + err = PTR_ERR(h_dentry);
19451 + if (unlikely(!h_dentry))
19455 + AuDbg("%pf\n", h_inode->i_op->follow_link);
19456 + AuDbgDentry(h_dentry);
19457 + ret = h_inode->i_op->follow_link(h_dentry, cookie);
19460 + if (!IS_ERR_OR_NULL(ret)) {
19461 + au_symlink_add(inode->i_sb, slink, h_inode, *cookie);
19463 + AuDbg("slink %p\n", slink);
19464 + goto out_unlock; /* success */
19468 + slink->h_inode = NULL;
19469 + kfree_rcu(slink, rcu);
19471 + aufs_read_unlock(dentry, AuLock_IR);
19473 + if (unlikely(err))
19474 + ret = ERR_PTR(err);
19475 + AuTraceErrPtr(ret);
19479 +static void aufs_put_link(struct inode *inode, void *cookie)
19481 + struct au_symlink *slink;
19482 + struct inode *h_inode;
19485 + AuDbg("slink %p\n", slink);
19486 + h_inode = slink->h_inode;
19487 + AuDbg("%pf\n", h_inode->i_op->put_link);
19488 + AuDbgInode(h_inode);
19489 + if (h_inode->i_op->put_link)
19490 + h_inode->i_op->put_link(h_inode, slink->h_cookie);
19491 + au_symlink_del(inode->i_sb, slink);
19494 +/* ---------------------------------------------------------------------- */
19496 +static int aufs_update_time(struct inode *inode, struct timespec *ts, int flags)
19499 + struct super_block *sb;
19500 + struct inode *h_inode;
19502 + sb = inode->i_sb;
19503 + /* mmap_sem might be acquired already, cf. aufs_mmap() */
19505 + si_read_lock(sb, AuLock_FLUSH);
19506 + ii_write_lock_child(inode);
19508 + h_inode = au_h_iptr(inode, au_ibstart(inode));
19509 + err = vfsub_update_time(h_inode, ts, flags);
19512 + au_cpup_attr_timesizes(inode);
19513 + ii_write_unlock(inode);
19514 + si_read_unlock(sb);
19517 + if (!err && (flags & S_VERSION))
19518 + inode_inc_iversion(inode);
19523 +/* ---------------------------------------------------------------------- */
19525 +struct inode_operations aufs_symlink_iop = {
19526 + .permission = aufs_permission,
19527 +#ifdef CONFIG_FS_POSIX_ACL
19528 + .get_acl = aufs_get_acl,
19529 + .set_acl = aufs_set_acl, /* unsupport for symlink? */
19532 + .setattr = aufs_setattr,
19533 + .getattr = aufs_getattr,
19535 +#ifdef CONFIG_AUFS_XATTR
19536 + .setxattr = aufs_setxattr,
19537 + .getxattr = aufs_getxattr,
19538 + .listxattr = aufs_listxattr,
19539 + .removexattr = aufs_removexattr,
19542 + .readlink = generic_readlink,
19543 + .follow_link = aufs_follow_link,
19544 + .put_link = aufs_put_link,
19546 + /* .update_time = aufs_update_time */
19549 +struct inode_operations aufs_dir_iop = {
19550 + .create = aufs_create,
19551 + .lookup = aufs_lookup,
19552 + .link = aufs_link,
19553 + .unlink = aufs_unlink,
19554 + .symlink = aufs_symlink,
19555 + .mkdir = aufs_mkdir,
19556 + .rmdir = aufs_rmdir,
19557 + .mknod = aufs_mknod,
19558 + .rename = aufs_rename,
19560 + .permission = aufs_permission,
19561 +#ifdef CONFIG_FS_POSIX_ACL
19562 + .get_acl = aufs_get_acl,
19563 + .set_acl = aufs_set_acl,
19566 + .setattr = aufs_setattr,
19567 + .getattr = aufs_getattr,
19569 +#ifdef CONFIG_AUFS_XATTR
19570 + .setxattr = aufs_setxattr,
19571 + .getxattr = aufs_getxattr,
19572 + .listxattr = aufs_listxattr,
19573 + .removexattr = aufs_removexattr,
19576 + .update_time = aufs_update_time,
19577 + .atomic_open = aufs_atomic_open,
19578 + .tmpfile = aufs_tmpfile
19581 +struct inode_operations aufs_iop = {
19582 + .permission = aufs_permission,
19583 +#ifdef CONFIG_FS_POSIX_ACL
19584 + .get_acl = aufs_get_acl,
19585 + .set_acl = aufs_set_acl,
19588 + .setattr = aufs_setattr,
19589 + .getattr = aufs_getattr,
19591 +#ifdef CONFIG_AUFS_XATTR
19592 + .setxattr = aufs_setxattr,
19593 + .getxattr = aufs_getxattr,
19594 + .listxattr = aufs_listxattr,
19595 + .removexattr = aufs_removexattr,
19598 + .update_time = aufs_update_time
19600 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
19601 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
19602 +++ linux/fs/aufs/i_op_del.c 2015-09-24 10:47:58.254719746 +0200
19605 + * Copyright (C) 2005-2015 Junjiro R. Okajima
19607 + * This program, aufs is free software; you can redistribute it and/or modify
19608 + * it under the terms of the GNU General Public License as published by
19609 + * the Free Software Foundation; either version 2 of the License, or
19610 + * (at your option) any later version.
19612 + * This program is distributed in the hope that it will be useful,
19613 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19614 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19615 + * GNU General Public License for more details.
19617 + * You should have received a copy of the GNU General Public License
19618 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
19622 + * inode operations (del entry)
19628 + * decide if a new whiteout for @dentry is necessary or not.
19629 + * when it is necessary, prepare the parent dir for the upper branch whose
19630 + * branch index is @bcpup for creation. the actual creation of the whiteout will
19631 + * be done by caller.
19633 + * 0: wh is unnecessary
19634 + * plus: wh is necessary
19637 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
19639 + int need_wh, err;
19640 + aufs_bindex_t bstart;
19641 + struct super_block *sb;
19643 + sb = dentry->d_sb;
19644 + bstart = au_dbstart(dentry);
19645 + if (*bcpup < 0) {
19647 + if (au_test_ro(sb, bstart, d_inode(dentry))) {
19648 + err = AuWbrCopyup(au_sbi(sb), dentry);
19650 + if (unlikely(err < 0))
19654 + AuDebugOn(bstart < *bcpup
19655 + || au_test_ro(sb, *bcpup, d_inode(dentry)));
19656 + AuDbg("bcpup %d, bstart %d\n", *bcpup, bstart);
19658 + if (*bcpup != bstart) {
19659 + err = au_cpup_dirs(dentry, *bcpup);
19660 + if (unlikely(err))
19664 + struct au_dinfo *dinfo, *tmp;
19666 + need_wh = -ENOMEM;
19667 + dinfo = au_di(dentry);
19668 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
19670 + au_di_cp(tmp, dinfo);
19671 + au_di_swap(tmp, dinfo);
19672 + /* returns the number of positive dentries */
19673 + need_wh = au_lkup_dentry(dentry, bstart + 1, /*type*/0);
19674 + au_di_swap(tmp, dinfo);
19675 + au_rw_write_unlock(&tmp->di_rwsem);
19679 + AuDbg("need_wh %d\n", need_wh);
19687 + * simple tests for the del-entry operations.
19688 + * following the checks in vfs, plus the parent-child relationship.
19690 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
19691 + struct dentry *h_parent, int isdir)
19695 + struct dentry *h_dentry, *h_latest;
19696 + struct inode *h_inode;
19698 + h_dentry = au_h_dptr(dentry, bindex);
19699 + if (d_really_is_positive(dentry)) {
19701 + if (unlikely(d_is_negative(h_dentry)))
19703 + h_inode = d_inode(h_dentry);
19704 + if (unlikely(!h_inode->i_nlink))
19707 + h_mode = h_inode->i_mode;
19710 + if (unlikely(S_ISDIR(h_mode)))
19712 + } else if (unlikely(!S_ISDIR(h_mode))) {
19717 + /* rename(2) case */
19719 + if (unlikely(d_is_positive(h_dentry)))
19724 + /* expected parent dir is locked */
19725 + if (unlikely(h_parent != h_dentry->d_parent))
19730 + * rmdir a dir may break the consistency on some filesystem.
19731 + * let's try heavy test.
19734 + if (unlikely(!au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1)
19735 + && au_test_h_perm(d_inode(h_parent),
19736 + MAY_EXEC | MAY_WRITE)))
19739 + h_latest = au_sio_lkup_one(&dentry->d_name, h_parent);
19741 + if (IS_ERR(h_latest))
19743 + if (h_latest == h_dentry)
19752 + * decide the branch where we operate for @dentry. the branch index will be set
19753 + * @rbcpup. after diciding it, 'pin' it and store the timestamps of the parent
19754 + * dir for reverting.
19755 + * when a new whiteout is necessary, create it.
19757 +static struct dentry*
19758 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
19759 + struct au_dtime *dt, struct au_pin *pin)
19761 + struct dentry *wh_dentry;
19762 + struct super_block *sb;
19763 + struct path h_path;
19764 + int err, need_wh;
19765 + unsigned int udba;
19766 + aufs_bindex_t bcpup;
19768 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
19769 + wh_dentry = ERR_PTR(need_wh);
19770 + if (unlikely(need_wh < 0))
19773 + sb = dentry->d_sb;
19774 + udba = au_opt_udba(sb);
19776 + err = au_pin(pin, dentry, bcpup, udba,
19777 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
19778 + wh_dentry = ERR_PTR(err);
19779 + if (unlikely(err))
19782 + h_path.dentry = au_pinned_h_parent(pin);
19783 + if (udba != AuOpt_UDBA_NONE
19784 + && au_dbstart(dentry) == bcpup) {
19785 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
19786 + wh_dentry = ERR_PTR(err);
19787 + if (unlikely(err))
19791 + h_path.mnt = au_sbr_mnt(sb, bcpup);
19792 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
19793 + wh_dentry = NULL;
19795 + goto out; /* success, no need to create whiteout */
19797 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
19798 + if (IS_ERR(wh_dentry))
19801 + /* returns with the parent is locked and wh_dentry is dget-ed */
19802 + goto out; /* success */
19807 + return wh_dentry;
19811 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
19812 + * in order to be revertible and save time for removing many child whiteouts
19814 + * returns 1 when there are too many child whiteout and caller should remove
19815 + * them asynchronously. returns 0 when the number of children is enough small to
19816 + * remove now or the branch fs is a remote fs.
19817 + * otherwise return an error.
19819 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
19820 + struct au_nhash *whlist, struct inode *dir)
19822 + int rmdir_later, err, dirwh;
19823 + struct dentry *h_dentry;
19824 + struct super_block *sb;
19825 + struct inode *inode;
19827 + sb = dentry->d_sb;
19828 + SiMustAnyLock(sb);
19829 + h_dentry = au_h_dptr(dentry, bindex);
19830 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
19831 + if (unlikely(err))
19834 + /* stop monitoring */
19835 + inode = d_inode(dentry);
19836 + au_hn_free(au_hi(inode, bindex));
19838 + if (!au_test_fs_remote(h_dentry->d_sb)) {
19839 + dirwh = au_sbi(sb)->si_dirwh;
19840 + rmdir_later = (dirwh <= 1);
19841 + if (!rmdir_later)
19842 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
19845 + return rmdir_later;
19848 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
19849 + if (unlikely(err)) {
19850 + AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
19851 + h_dentry, bindex, err);
19861 + * final procedure for deleting a entry.
19862 + * maintain dentry and iattr.
19864 +static void epilog(struct inode *dir, struct dentry *dentry,
19865 + aufs_bindex_t bindex)
19867 + struct inode *inode;
19869 + inode = d_inode(dentry);
19871 + inode->i_ctime = dir->i_ctime;
19873 + au_dir_ts(dir, bindex);
19874 + dir->i_version++;
19878 + * when an error happened, remove the created whiteout and revert everything.
19880 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
19881 + aufs_bindex_t bwh, struct dentry *wh_dentry,
19882 + struct dentry *dentry, struct au_dtime *dt)
19885 + struct path h_path = {
19886 + .dentry = wh_dentry,
19887 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
19890 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
19892 + au_set_dbwh(dentry, bwh);
19893 + au_dtime_revert(dt);
19897 + AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
19901 +/* ---------------------------------------------------------------------- */
19903 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
19906 + aufs_bindex_t bwh, bindex, bstart;
19907 + struct inode *inode, *h_dir, *delegated;
19908 + struct dentry *parent, *wh_dentry;
19909 + /* to reuduce stack size */
19911 + struct au_dtime dt;
19912 + struct au_pin pin;
19913 + struct path h_path;
19919 + a = kmalloc(sizeof(*a), GFP_NOFS);
19920 + if (unlikely(!a))
19923 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
19924 + if (unlikely(err))
19926 + err = au_d_hashed_positive(dentry);
19927 + if (unlikely(err))
19929 + inode = d_inode(dentry);
19930 + IMustLock(inode);
19932 + if (unlikely(d_is_dir(dentry)))
19933 + goto out_unlock; /* possible? */
19935 + bstart = au_dbstart(dentry);
19936 + bwh = au_dbwh(dentry);
19938 + parent = dentry->d_parent; /* dir inode is locked */
19939 + di_write_lock_parent(parent);
19940 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
19942 + err = PTR_ERR(wh_dentry);
19943 + if (IS_ERR(wh_dentry))
19946 + a->h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
19947 + a->h_path.dentry = au_h_dptr(dentry, bstart);
19948 + dget(a->h_path.dentry);
19949 + if (bindex == bstart) {
19950 + h_dir = au_pinned_h_dir(&a->pin);
19951 + delegated = NULL;
19952 + err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
19953 + if (unlikely(err == -EWOULDBLOCK)) {
19954 + pr_warn("cannot retry for NFSv4 delegation"
19955 + " for an internal unlink\n");
19959 + /* dir inode is locked */
19960 + h_dir = d_inode(wh_dentry->d_parent);
19961 + IMustLock(h_dir);
19966 + vfsub_drop_nlink(inode);
19967 + epilog(dir, dentry, bindex);
19969 + /* update target timestamps */
19970 + if (bindex == bstart) {
19971 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
19973 + inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
19975 + /* todo: this timestamp may be reverted later */
19976 + inode->i_ctime = h_dir->i_ctime;
19977 + goto out_unpin; /* success */
19984 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
19991 + au_unpin(&a->pin);
19993 + dput(a->h_path.dentry);
19995 + di_write_unlock(parent);
19997 + aufs_read_unlock(dentry, AuLock_DW);
20004 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
20006 + int err, rmdir_later;
20007 + aufs_bindex_t bwh, bindex, bstart;
20008 + struct inode *inode;
20009 + struct dentry *parent, *wh_dentry, *h_dentry;
20010 + struct au_whtmp_rmdir *args;
20011 + /* to reuduce stack size */
20013 + struct au_dtime dt;
20014 + struct au_pin pin;
20020 + a = kmalloc(sizeof(*a), GFP_NOFS);
20021 + if (unlikely(!a))
20024 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
20025 + if (unlikely(err))
20027 + err = au_alive_dir(dentry);
20028 + if (unlikely(err))
20030 + inode = d_inode(dentry);
20031 + IMustLock(inode);
20033 + if (unlikely(!d_is_dir(dentry)))
20034 + goto out_unlock; /* possible? */
20037 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
20038 + if (unlikely(!args))
20041 + parent = dentry->d_parent; /* dir inode is locked */
20042 + di_write_lock_parent(parent);
20043 + err = au_test_empty(dentry, &args->whlist);
20044 + if (unlikely(err))
20047 + bstart = au_dbstart(dentry);
20048 + bwh = au_dbwh(dentry);
20050 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
20052 + err = PTR_ERR(wh_dentry);
20053 + if (IS_ERR(wh_dentry))
20056 + h_dentry = au_h_dptr(dentry, bstart);
20059 + if (bindex == bstart) {
20060 + err = renwh_and_rmdir(dentry, bstart, &args->whlist, dir);
20062 + rmdir_later = err;
20066 + /* stop monitoring */
20067 + au_hn_free(au_hi(inode, bstart));
20069 + /* dir inode is locked */
20070 + IMustLock(d_inode(wh_dentry->d_parent));
20075 + vfsub_dead_dir(inode);
20076 + au_set_dbdiropq(dentry, -1);
20077 + epilog(dir, dentry, bindex);
20079 + if (rmdir_later) {
20080 + au_whtmp_kick_rmdir(dir, bstart, h_dentry, args);
20084 + goto out_unpin; /* success */
20092 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
20099 + au_unpin(&a->pin);
20103 + di_write_unlock(parent);
20105 + au_whtmp_rmdir_free(args);
20107 + aufs_read_unlock(dentry, AuLock_DW);
20114 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
20115 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
20116 +++ linux/fs/aufs/i_op_ren.c 2015-11-11 17:21:46.918863802 +0100
20119 + * Copyright (C) 2005-2015 Junjiro R. Okajima
20121 + * This program, aufs is free software; you can redistribute it and/or modify
20122 + * it under the terms of the GNU General Public License as published by
20123 + * the Free Software Foundation; either version 2 of the License, or
20124 + * (at your option) any later version.
20126 + * This program is distributed in the hope that it will be useful,
20127 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20128 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20129 + * GNU General Public License for more details.
20131 + * You should have received a copy of the GNU General Public License
20132 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
20136 + * inode operation (rename entry)
20137 + * todo: this is crazy monster
20142 +enum { AuSRC, AuDST, AuSrcDst };
20143 +enum { AuPARENT, AuCHILD, AuParentChild };
20145 +#define AuRen_ISDIR 1
20146 +#define AuRen_ISSAMEDIR (1 << 1)
20147 +#define AuRen_WHSRC (1 << 2)
20148 +#define AuRen_WHDST (1 << 3)
20149 +#define AuRen_MNT_WRITE (1 << 4)
20150 +#define AuRen_DT_DSTDIR (1 << 5)
20151 +#define AuRen_DIROPQ (1 << 6)
20152 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
20153 +#define au_fset_ren(flags, name) \
20154 + do { (flags) |= AuRen_##name; } while (0)
20155 +#define au_fclr_ren(flags, name) \
20156 + do { (flags) &= ~AuRen_##name; } while (0)
20158 +struct au_ren_args {
20160 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
20162 + struct inode *dir, *inode;
20163 + struct au_hinode *hdir;
20164 + struct au_dtime dt[AuParentChild];
20165 + aufs_bindex_t bstart;
20168 +#define src_dentry sd[AuSRC].dentry
20169 +#define src_dir sd[AuSRC].dir
20170 +#define src_inode sd[AuSRC].inode
20171 +#define src_h_dentry sd[AuSRC].h_dentry
20172 +#define src_parent sd[AuSRC].parent
20173 +#define src_h_parent sd[AuSRC].h_parent
20174 +#define src_wh_dentry sd[AuSRC].wh_dentry
20175 +#define src_hdir sd[AuSRC].hdir
20176 +#define src_h_dir sd[AuSRC].hdir->hi_inode
20177 +#define src_dt sd[AuSRC].dt
20178 +#define src_bstart sd[AuSRC].bstart
20180 +#define dst_dentry sd[AuDST].dentry
20181 +#define dst_dir sd[AuDST].dir
20182 +#define dst_inode sd[AuDST].inode
20183 +#define dst_h_dentry sd[AuDST].h_dentry
20184 +#define dst_parent sd[AuDST].parent
20185 +#define dst_h_parent sd[AuDST].h_parent
20186 +#define dst_wh_dentry sd[AuDST].wh_dentry
20187 +#define dst_hdir sd[AuDST].hdir
20188 +#define dst_h_dir sd[AuDST].hdir->hi_inode
20189 +#define dst_dt sd[AuDST].dt
20190 +#define dst_bstart sd[AuDST].bstart
20192 + struct dentry *h_trap;
20193 + struct au_branch *br;
20194 + struct au_hinode *src_hinode;
20195 + struct path h_path;
20196 + struct au_nhash whlist;
20197 + aufs_bindex_t btgt, src_bwh, src_bdiropq;
20199 + unsigned int flags;
20201 + struct au_whtmp_rmdir *thargs;
20202 + struct dentry *h_dst;
20205 +/* ---------------------------------------------------------------------- */
20208 + * functions for reverting.
20209 + * when an error happened in a single rename systemcall, we should revert
20210 + * everything as if nothing happened.
20211 + * we don't need to revert the copied-up/down the parent dir since they are
20215 +#define RevertFailure(fmt, ...) do { \
20216 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
20217 + ##__VA_ARGS__, err, rerr); \
20221 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
20225 + au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
20226 + rerr = au_diropq_remove(a->src_dentry, a->btgt);
20227 + au_hn_imtx_unlock(a->src_hinode);
20228 + au_set_dbdiropq(a->src_dentry, a->src_bdiropq);
20230 + RevertFailure("remove diropq %pd", a->src_dentry);
20233 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
20236 + struct inode *delegated;
20238 + a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name,
20239 + a->src_h_parent);
20240 + rerr = PTR_ERR(a->h_path.dentry);
20241 + if (IS_ERR(a->h_path.dentry)) {
20242 + RevertFailure("lkup one %pd", a->src_dentry);
20246 + delegated = NULL;
20247 + rerr = vfsub_rename(a->dst_h_dir,
20248 + au_h_dptr(a->src_dentry, a->btgt),
20249 + a->src_h_dir, &a->h_path, &delegated);
20250 + if (unlikely(rerr == -EWOULDBLOCK)) {
20251 + pr_warn("cannot retry for NFSv4 delegation"
20252 + " for an internal rename\n");
20255 + d_drop(a->h_path.dentry);
20256 + dput(a->h_path.dentry);
20257 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
20259 + RevertFailure("rename %pd", a->src_dentry);
20262 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
20265 + struct inode *delegated;
20267 + a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name,
20268 + a->dst_h_parent);
20269 + rerr = PTR_ERR(a->h_path.dentry);
20270 + if (IS_ERR(a->h_path.dentry)) {
20271 + RevertFailure("lkup one %pd", a->dst_dentry);
20274 + if (d_is_positive(a->h_path.dentry)) {
20275 + d_drop(a->h_path.dentry);
20276 + dput(a->h_path.dentry);
20280 + delegated = NULL;
20281 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
20283 + if (unlikely(rerr == -EWOULDBLOCK)) {
20284 + pr_warn("cannot retry for NFSv4 delegation"
20285 + " for an internal rename\n");
20288 + d_drop(a->h_path.dentry);
20289 + dput(a->h_path.dentry);
20291 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
20293 + RevertFailure("rename %pd", a->h_dst);
20296 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
20300 + a->h_path.dentry = a->src_wh_dentry;
20301 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
20302 + au_set_dbwh(a->src_dentry, a->src_bwh);
20304 + RevertFailure("unlink %pd", a->src_wh_dentry);
20306 +#undef RevertFailure
20308 +/* ---------------------------------------------------------------------- */
20311 + * when we have to copyup the renaming entry, do it with the rename-target name
20312 + * in order to minimize the cost (the later actual rename is unnecessary).
20313 + * otherwise rename it on the target branch.
20315 +static int au_ren_or_cpup(struct au_ren_args *a)
20318 + struct dentry *d;
20319 + struct inode *delegated;
20321 + d = a->src_dentry;
20322 + if (au_dbstart(d) == a->btgt) {
20323 + a->h_path.dentry = a->dst_h_dentry;
20324 + if (au_ftest_ren(a->flags, DIROPQ)
20325 + && au_dbdiropq(d) == a->btgt)
20326 + au_fclr_ren(a->flags, DIROPQ);
20327 + AuDebugOn(au_dbstart(d) != a->btgt);
20328 + delegated = NULL;
20329 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
20330 + a->dst_h_dir, &a->h_path, &delegated);
20331 + if (unlikely(err == -EWOULDBLOCK)) {
20332 + pr_warn("cannot retry for NFSv4 delegation"
20333 + " for an internal rename\n");
20339 + if (!err && a->h_dst)
20340 + /* it will be set to dinfo later */
20346 +/* cf. aufs_rmdir() */
20347 +static int au_ren_del_whtmp(struct au_ren_args *a)
20350 + struct inode *dir;
20352 + dir = a->dst_dir;
20353 + SiMustAnyLock(dir->i_sb);
20354 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
20355 + au_sbi(dir->i_sb)->si_dirwh)
20356 + || au_test_fs_remote(a->h_dst->d_sb)) {
20357 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
20358 + if (unlikely(err))
20359 + pr_warn("failed removing whtmp dir %pd (%d), "
20360 + "ignored.\n", a->h_dst, err);
20362 + au_nhash_wh_free(&a->thargs->whlist);
20363 + a->thargs->whlist = a->whlist;
20364 + a->whlist.nh_num = 0;
20365 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
20367 + a->thargs = NULL;
20373 +/* make it 'opaque' dir. */
20374 +static int au_ren_diropq(struct au_ren_args *a)
20377 + struct dentry *diropq;
20380 + a->src_bdiropq = au_dbdiropq(a->src_dentry);
20381 + a->src_hinode = au_hi(a->src_inode, a->btgt);
20382 + au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
20383 + diropq = au_diropq_create(a->src_dentry, a->btgt);
20384 + au_hn_imtx_unlock(a->src_hinode);
20385 + if (IS_ERR(diropq))
20386 + err = PTR_ERR(diropq);
20393 +static int do_rename(struct au_ren_args *a)
20396 + struct dentry *d, *h_d;
20398 + /* prepare workqueue args for asynchronous rmdir */
20399 + h_d = a->dst_h_dentry;
20400 + if (au_ftest_ren(a->flags, ISDIR) && d_is_positive(h_d)) {
20402 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb, GFP_NOFS);
20403 + if (unlikely(!a->thargs))
20405 + a->h_dst = dget(h_d);
20408 + /* create whiteout for src_dentry */
20409 + if (au_ftest_ren(a->flags, WHSRC)) {
20410 + a->src_bwh = au_dbwh(a->src_dentry);
20411 + AuDebugOn(a->src_bwh >= 0);
20413 + = au_wh_create(a->src_dentry, a->btgt, a->src_h_parent);
20414 + err = PTR_ERR(a->src_wh_dentry);
20415 + if (IS_ERR(a->src_wh_dentry))
20419 + /* lookup whiteout for dentry */
20420 + if (au_ftest_ren(a->flags, WHDST)) {
20421 + h_d = au_wh_lkup(a->dst_h_parent, &a->dst_dentry->d_name,
20423 + err = PTR_ERR(h_d);
20426 + if (d_is_negative(h_d))
20429 + a->dst_wh_dentry = h_d;
20432 + /* rename dentry to tmpwh */
20434 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
20435 + if (unlikely(err))
20438 + d = a->dst_dentry;
20439 + au_set_h_dptr(d, a->btgt, NULL);
20440 + err = au_lkup_neg(d, a->btgt, /*wh*/0);
20441 + if (unlikely(err))
20443 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
20446 + BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_bstart != a->btgt);
20448 + /* rename by vfs_rename or cpup */
20449 + d = a->dst_dentry;
20450 + if (au_ftest_ren(a->flags, ISDIR)
20451 + && (a->dst_wh_dentry
20452 + || au_dbdiropq(d) == a->btgt
20453 + /* hide the lower to keep xino */
20454 + || a->btgt < au_dbend(d)
20455 + || au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ)))
20456 + au_fset_ren(a->flags, DIROPQ);
20457 + err = au_ren_or_cpup(a);
20458 + if (unlikely(err))
20459 + /* leave the copied-up one */
20462 + /* make dir opaque */
20463 + if (au_ftest_ren(a->flags, DIROPQ)) {
20464 + err = au_ren_diropq(a);
20465 + if (unlikely(err))
20469 + /* update target timestamps */
20470 + AuDebugOn(au_dbstart(a->src_dentry) != a->btgt);
20471 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
20472 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
20473 + a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
20475 + /* remove whiteout for dentry */
20476 + if (a->dst_wh_dentry) {
20477 + a->h_path.dentry = a->dst_wh_dentry;
20478 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
20480 + if (unlikely(err))
20484 + /* remove whtmp */
20486 + au_ren_del_whtmp(a); /* ignore this error */
20488 + au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
20490 + goto out_success;
20493 + if (au_ftest_ren(a->flags, DIROPQ))
20494 + au_ren_rev_diropq(err, a);
20496 + au_ren_rev_rename(err, a);
20500 + au_ren_rev_whtmp(err, a);
20502 + dput(a->dst_wh_dentry);
20503 + a->dst_wh_dentry = NULL;
20505 + if (a->src_wh_dentry)
20506 + au_ren_rev_whsrc(err, a);
20508 + dput(a->src_wh_dentry);
20509 + dput(a->dst_wh_dentry);
20513 + au_whtmp_rmdir_free(a->thargs);
20514 + a->thargs = NULL;
20520 +/* ---------------------------------------------------------------------- */
20523 + * test if @dentry dir can be rename destination or not.
20524 + * success means, it is a logically empty dir.
20526 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
20528 + return au_test_empty(dentry, whlist);
20532 + * test if @dentry dir can be rename source or not.
20533 + * if it can, return 0 and @children is filled.
20535 + * - it is a logically empty dir.
20536 + * - or, it exists on writable branch and has no children including whiteouts
20537 + * on the lower branch.
20539 +static int may_rename_srcdir(struct dentry *dentry, aufs_bindex_t btgt)
20542 + unsigned int rdhash;
20543 + aufs_bindex_t bstart;
20545 + bstart = au_dbstart(dentry);
20546 + if (bstart != btgt) {
20547 + struct au_nhash whlist;
20549 + SiMustAnyLock(dentry->d_sb);
20550 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
20552 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
20554 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
20555 + if (unlikely(err))
20557 + err = au_test_empty(dentry, &whlist);
20558 + au_nhash_wh_free(&whlist);
20562 + if (bstart == au_dbtaildir(dentry))
20563 + return 0; /* success */
20565 + err = au_test_empty_lower(dentry);
20568 + if (err == -ENOTEMPTY) {
20569 + AuWarn1("renaming dir who has child(ren) on multiple branches,"
20570 + " is not supported\n");
20576 +/* side effect: sets whlist and h_dentry */
20577 +static int au_ren_may_dir(struct au_ren_args *a)
20580 + unsigned int rdhash;
20581 + struct dentry *d;
20583 + d = a->dst_dentry;
20584 + SiMustAnyLock(d->d_sb);
20587 + if (au_ftest_ren(a->flags, ISDIR) && a->dst_inode) {
20588 + rdhash = au_sbi(d->d_sb)->si_rdhash;
20590 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
20591 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
20592 + if (unlikely(err))
20595 + au_set_dbstart(d, a->dst_bstart);
20596 + err = may_rename_dstdir(d, &a->whlist);
20597 + au_set_dbstart(d, a->btgt);
20599 + a->dst_h_dentry = au_h_dptr(d, au_dbstart(d));
20600 + if (unlikely(err))
20603 + d = a->src_dentry;
20604 + a->src_h_dentry = au_h_dptr(d, au_dbstart(d));
20605 + if (au_ftest_ren(a->flags, ISDIR)) {
20606 + err = may_rename_srcdir(d, a->btgt);
20607 + if (unlikely(err)) {
20608 + au_nhash_wh_free(&a->whlist);
20609 + a->whlist.nh_num = 0;
20616 +/* ---------------------------------------------------------------------- */
20619 + * simple tests for rename.
20620 + * following the checks in vfs, plus the parent-child relationship.
20622 +static int au_may_ren(struct au_ren_args *a)
20625 + struct inode *h_inode;
20627 + if (a->src_bstart == a->btgt) {
20628 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
20629 + au_ftest_ren(a->flags, ISDIR));
20630 + if (unlikely(err))
20633 + if (unlikely(a->src_h_dentry == a->h_trap))
20638 + if (a->dst_bstart != a->btgt)
20641 + err = -ENOTEMPTY;
20642 + if (unlikely(a->dst_h_dentry == a->h_trap))
20646 + isdir = !!au_ftest_ren(a->flags, ISDIR);
20647 + if (d_really_is_negative(a->dst_dentry)) {
20648 + if (d_is_negative(a->dst_h_dentry))
20649 + err = au_may_add(a->dst_dentry, a->btgt,
20650 + a->dst_h_parent, isdir);
20652 + if (unlikely(d_is_negative(a->dst_h_dentry)))
20654 + h_inode = d_inode(a->dst_h_dentry);
20655 + if (h_inode->i_nlink)
20656 + err = au_may_del(a->dst_dentry, a->btgt,
20657 + a->dst_h_parent, isdir);
20661 + if (unlikely(err == -ENOENT || err == -EEXIST))
20667 +/* ---------------------------------------------------------------------- */
20672 + * - src_dir and dir by lock_rename()
20673 + * - inode if exitsts
20676 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
20678 + * + di_write_lock2_child()
20679 + * + di_write_lock_child()
20680 + * + ii_write_lock_child()
20681 + * + di_write_lock_child2()
20682 + * + ii_write_lock_child2()
20683 + * + src_parent and parent
20684 + * + di_write_lock_parent()
20685 + * + ii_write_lock_parent()
20686 + * + di_write_lock_parent2()
20687 + * + ii_write_lock_parent2()
20688 + * + lower src_dir and dir by vfsub_lock_rename()
20689 + * + verify the every relationships between child and parent. if any
20690 + * of them failed, unlock all and return -EBUSY.
20692 +static void au_ren_unlock(struct au_ren_args *a)
20694 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
20695 + a->dst_h_parent, a->dst_hdir);
20696 + if (au_ftest_ren(a->flags, MNT_WRITE))
20697 + vfsub_mnt_drop_write(au_br_mnt(a->br));
20700 +static int au_ren_lock(struct au_ren_args *a)
20703 + unsigned int udba;
20706 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
20707 + a->src_hdir = au_hi(a->src_dir, a->btgt);
20708 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
20709 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
20711 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
20712 + if (unlikely(err))
20714 + au_fset_ren(a->flags, MNT_WRITE);
20715 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
20716 + a->dst_h_parent, a->dst_hdir);
20717 + udba = au_opt_udba(a->src_dentry->d_sb);
20718 + if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
20719 + || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
20720 + err = au_busy_or_stale();
20721 + if (!err && au_dbstart(a->src_dentry) == a->btgt)
20722 + err = au_h_verify(a->src_h_dentry, udba,
20723 + d_inode(a->src_h_parent), a->src_h_parent,
20725 + if (!err && au_dbstart(a->dst_dentry) == a->btgt)
20726 + err = au_h_verify(a->dst_h_dentry, udba,
20727 + d_inode(a->dst_h_parent), a->dst_h_parent,
20730 + goto out; /* success */
20732 + err = au_busy_or_stale();
20733 + au_ren_unlock(a);
20739 +/* ---------------------------------------------------------------------- */
20741 +static void au_ren_refresh_dir(struct au_ren_args *a)
20743 + struct inode *dir;
20745 + dir = a->dst_dir;
20746 + dir->i_version++;
20747 + if (au_ftest_ren(a->flags, ISDIR)) {
20748 + /* is this updating defined in POSIX? */
20749 + au_cpup_attr_timesizes(a->src_inode);
20750 + au_cpup_attr_nlink(dir, /*force*/1);
20753 + au_dir_ts(dir, a->btgt);
20755 + if (au_ftest_ren(a->flags, ISSAMEDIR))
20758 + dir = a->src_dir;
20759 + dir->i_version++;
20760 + if (au_ftest_ren(a->flags, ISDIR))
20761 + au_cpup_attr_nlink(dir, /*force*/1);
20762 + au_dir_ts(dir, a->btgt);
20765 +static void au_ren_refresh(struct au_ren_args *a)
20767 + aufs_bindex_t bend, bindex;
20768 + struct dentry *d, *h_d;
20769 + struct inode *i, *h_i;
20770 + struct super_block *sb;
20772 + d = a->dst_dentry;
20775 + /* already dget-ed by au_ren_or_cpup() */
20776 + au_set_h_dptr(d, a->btgt, a->h_dst);
20778 + i = a->dst_inode;
20780 + if (!au_ftest_ren(a->flags, ISDIR))
20781 + vfsub_drop_nlink(i);
20783 + vfsub_dead_dir(i);
20784 + au_cpup_attr_timesizes(i);
20786 + au_update_dbrange(d, /*do_put_zero*/1);
20789 + for (bindex = au_dbstart(d); bindex < bend; bindex++)
20790 + au_set_h_dptr(d, bindex, NULL);
20791 + bend = au_dbend(d);
20792 + for (bindex = a->btgt + 1; bindex <= bend; bindex++)
20793 + au_set_h_dptr(d, bindex, NULL);
20794 + au_update_dbrange(d, /*do_put_zero*/0);
20797 + d = a->src_dentry;
20798 + au_set_dbwh(d, -1);
20799 + bend = au_dbend(d);
20800 + for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
20801 + h_d = au_h_dptr(d, bindex);
20803 + au_set_h_dptr(d, bindex, NULL);
20805 + au_set_dbend(d, a->btgt);
20808 + i = a->src_inode;
20809 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
20810 + return; /* success */
20812 + bend = au_ibend(i);
20813 + for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
20814 + h_i = au_h_iptr(i, bindex);
20816 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
20817 + /* ignore this error */
20818 + au_set_h_iptr(i, bindex, NULL, 0);
20821 + au_set_ibend(i, a->btgt);
20824 +/* ---------------------------------------------------------------------- */
20826 +/* mainly for link(2) and rename(2) */
20827 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
20829 + aufs_bindex_t bdiropq, bwh;
20830 + struct dentry *parent;
20831 + struct au_branch *br;
20833 + parent = dentry->d_parent;
20834 + IMustLock(d_inode(parent)); /* dir is locked */
20836 + bdiropq = au_dbdiropq(parent);
20837 + bwh = au_dbwh(dentry);
20838 + br = au_sbr(dentry->d_sb, btgt);
20839 + if (au_br_rdonly(br)
20840 + || (0 <= bdiropq && bdiropq < btgt)
20841 + || (0 <= bwh && bwh < btgt))
20844 + AuDbg("btgt %d\n", btgt);
20848 +/* sets src_bstart, dst_bstart and btgt */
20849 +static int au_ren_wbr(struct au_ren_args *a)
20852 + struct au_wr_dir_args wr_dir_args = {
20853 + /* .force_btgt = -1, */
20854 + .flags = AuWrDir_ADD_ENTRY
20857 + a->src_bstart = au_dbstart(a->src_dentry);
20858 + a->dst_bstart = au_dbstart(a->dst_dentry);
20859 + if (au_ftest_ren(a->flags, ISDIR))
20860 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
20861 + wr_dir_args.force_btgt = a->src_bstart;
20862 + if (a->dst_inode && a->dst_bstart < a->src_bstart)
20863 + wr_dir_args.force_btgt = a->dst_bstart;
20864 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
20865 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
20871 +static void au_ren_dt(struct au_ren_args *a)
20873 + a->h_path.dentry = a->src_h_parent;
20874 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
20875 + if (!au_ftest_ren(a->flags, ISSAMEDIR)) {
20876 + a->h_path.dentry = a->dst_h_parent;
20877 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
20880 + au_fclr_ren(a->flags, DT_DSTDIR);
20881 + if (!au_ftest_ren(a->flags, ISDIR))
20884 + a->h_path.dentry = a->src_h_dentry;
20885 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
20886 + if (d_is_positive(a->dst_h_dentry)) {
20887 + au_fset_ren(a->flags, DT_DSTDIR);
20888 + a->h_path.dentry = a->dst_h_dentry;
20889 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
20893 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
20895 + struct dentry *h_d;
20896 + struct mutex *h_mtx;
20898 + au_dtime_revert(a->src_dt + AuPARENT);
20899 + if (!au_ftest_ren(a->flags, ISSAMEDIR))
20900 + au_dtime_revert(a->dst_dt + AuPARENT);
20902 + if (au_ftest_ren(a->flags, ISDIR) && err != -EIO) {
20903 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
20904 + h_mtx = &d_inode(h_d)->i_mutex;
20905 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
20906 + au_dtime_revert(a->src_dt + AuCHILD);
20907 + mutex_unlock(h_mtx);
20909 + if (au_ftest_ren(a->flags, DT_DSTDIR)) {
20910 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
20911 + h_mtx = &d_inode(h_d)->i_mutex;
20912 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
20913 + au_dtime_revert(a->dst_dt + AuCHILD);
20914 + mutex_unlock(h_mtx);
20919 +/* ---------------------------------------------------------------------- */
20921 +int aufs_rename(struct inode *_src_dir, struct dentry *_src_dentry,
20922 + struct inode *_dst_dir, struct dentry *_dst_dentry)
20925 + /* reduce stack space */
20926 + struct au_ren_args *a;
20928 + AuDbg("%pd, %pd\n", _src_dentry, _dst_dentry);
20929 + IMustLock(_src_dir);
20930 + IMustLock(_dst_dir);
20933 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
20934 + a = kzalloc(sizeof(*a), GFP_NOFS);
20935 + if (unlikely(!a))
20938 + a->src_dir = _src_dir;
20939 + a->src_dentry = _src_dentry;
20940 + a->src_inode = NULL;
20941 + if (d_really_is_positive(a->src_dentry))
20942 + a->src_inode = d_inode(a->src_dentry);
20943 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
20944 + a->dst_dir = _dst_dir;
20945 + a->dst_dentry = _dst_dentry;
20946 + a->dst_inode = NULL;
20947 + if (d_really_is_positive(a->dst_dentry))
20948 + a->dst_inode = d_inode(a->dst_dentry);
20949 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
20950 + if (a->dst_inode) {
20951 + IMustLock(a->dst_inode);
20952 + au_igrab(a->dst_inode);
20956 + flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
20957 + if (d_is_dir(a->src_dentry)) {
20958 + au_fset_ren(a->flags, ISDIR);
20959 + if (unlikely(d_really_is_positive(a->dst_dentry)
20960 + && !d_is_dir(a->dst_dentry)))
20962 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
20963 + AuLock_DIR | flags);
20965 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
20967 + if (unlikely(err))
20970 + err = au_d_hashed_positive(a->src_dentry);
20971 + if (unlikely(err))
20974 + if (a->dst_inode) {
20976 + * If it is a dir, VFS unhash dst_dentry before this
20977 + * function. It means we cannot rely upon d_unhashed().
20979 + if (unlikely(!a->dst_inode->i_nlink))
20981 + if (!S_ISDIR(a->dst_inode->i_mode)) {
20982 + err = au_d_hashed_positive(a->dst_dentry);
20983 + if (unlikely(err))
20985 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
20987 + } else if (unlikely(d_unhashed(a->dst_dentry)))
20991 + * is it possible?
20992 + * yes, it happened (in linux-3.3-rcN) but I don't know why.
20993 + * there may exist a problem somewhere else.
20996 + if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
20999 + au_fset_ren(a->flags, ISSAMEDIR); /* temporary */
21000 + di_write_lock_parent(a->dst_parent);
21002 + /* which branch we process */
21003 + err = au_ren_wbr(a);
21004 + if (unlikely(err < 0))
21006 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
21007 + a->h_path.mnt = au_br_mnt(a->br);
21009 + /* are they available to be renamed */
21010 + err = au_ren_may_dir(a);
21011 + if (unlikely(err))
21012 + goto out_children;
21014 + /* prepare the writable parent dir on the same branch */
21015 + if (a->dst_bstart == a->btgt) {
21016 + au_fset_ren(a->flags, WHDST);
21018 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
21019 + if (unlikely(err))
21020 + goto out_children;
21023 + if (a->src_dir != a->dst_dir) {
21025 + * this temporary unlock is safe,
21026 + * because both dir->i_mutex are locked.
21028 + di_write_unlock(a->dst_parent);
21029 + di_write_lock_parent(a->src_parent);
21030 + err = au_wr_dir_need_wh(a->src_dentry,
21031 + au_ftest_ren(a->flags, ISDIR),
21033 + di_write_unlock(a->src_parent);
21034 + di_write_lock2_parent(a->src_parent, a->dst_parent, /*isdir*/1);
21035 + au_fclr_ren(a->flags, ISSAMEDIR);
21037 + err = au_wr_dir_need_wh(a->src_dentry,
21038 + au_ftest_ren(a->flags, ISDIR),
21040 + if (unlikely(err < 0))
21041 + goto out_children;
21043 + au_fset_ren(a->flags, WHSRC);
21046 + if (a->src_bstart != a->btgt) {
21047 + struct au_pin pin;
21049 + err = au_pin(&pin, a->src_dentry, a->btgt,
21050 + au_opt_udba(a->src_dentry->d_sb),
21051 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
21053 + struct au_cp_generic cpg = {
21054 + .dentry = a->src_dentry,
21056 + .bsrc = a->src_bstart,
21059 + .flags = AuCpup_DTIME | AuCpup_HOPEN
21061 + AuDebugOn(au_dbstart(a->src_dentry) != a->src_bstart);
21062 + err = au_sio_cpup_simple(&cpg);
21065 + if (unlikely(err))
21066 + goto out_children;
21067 + a->src_bstart = a->btgt;
21068 + a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
21069 + au_fset_ren(a->flags, WHSRC);
21072 + /* lock them all */
21073 + err = au_ren_lock(a);
21074 + if (unlikely(err))
21075 + /* leave the copied-up one */
21076 + goto out_children;
21078 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
21079 + err = au_may_ren(a);
21080 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
21081 + err = -ENAMETOOLONG;
21082 + if (unlikely(err))
21085 + /* store timestamps to be revertible */
21089 + err = do_rename(a);
21090 + if (unlikely(err))
21093 + /* update dir attributes */
21094 + au_ren_refresh_dir(a);
21096 + /* dput/iput all lower dentries */
21097 + au_ren_refresh(a);
21099 + goto out_hdir; /* success */
21102 + au_ren_rev_dt(err, a);
21104 + au_ren_unlock(a);
21106 + au_nhash_wh_free(&a->whlist);
21107 + if (err && a->dst_inode && a->dst_bstart != a->btgt) {
21108 + AuDbg("bstart %d, btgt %d\n", a->dst_bstart, a->btgt);
21109 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
21110 + au_set_dbstart(a->dst_dentry, a->dst_bstart);
21114 + d_move(a->src_dentry, a->dst_dentry);
21116 + au_update_dbstart(a->dst_dentry);
21117 + if (!a->dst_inode)
21118 + d_drop(a->dst_dentry);
21120 + if (au_ftest_ren(a->flags, ISSAMEDIR))
21121 + di_write_unlock(a->dst_parent);
21123 + di_write_unlock2(a->src_parent, a->dst_parent);
21125 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
21127 + iput(a->dst_inode);
21129 + au_whtmp_rmdir_free(a->thargs);
21135 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
21136 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
21137 +++ linux/fs/aufs/Kconfig 2015-09-24 10:47:58.248052907 +0200
21140 + tristate "Aufs (Advanced multi layered unification filesystem) support"
21142 + Aufs is a stackable unification filesystem such as Unionfs,
21143 + which unifies several directories and provides a merged single
21145 + In the early days, aufs was entirely re-designed and
21146 + re-implemented Unionfs Version 1.x series. Introducing many
21147 + original ideas, approaches and improvements, it becomes totally
21148 + different from Unionfs while keeping the basic features.
21152 + prompt "Maximum number of branches"
21153 + default AUFS_BRANCH_MAX_127
21155 + Specifies the maximum number of branches (or member directories)
21156 + in a single aufs. The larger value consumes more system
21157 + resources and has a minor impact to performance.
21158 +config AUFS_BRANCH_MAX_127
21161 + Specifies the maximum number of branches (or member directories)
21162 + in a single aufs. The larger value consumes more system
21163 + resources and has a minor impact to performance.
21164 +config AUFS_BRANCH_MAX_511
21167 + Specifies the maximum number of branches (or member directories)
21168 + in a single aufs. The larger value consumes more system
21169 + resources and has a minor impact to performance.
21170 +config AUFS_BRANCH_MAX_1023
21173 + Specifies the maximum number of branches (or member directories)
21174 + in a single aufs. The larger value consumes more system
21175 + resources and has a minor impact to performance.
21176 +config AUFS_BRANCH_MAX_32767
21179 + Specifies the maximum number of branches (or member directories)
21180 + in a single aufs. The larger value consumes more system
21181 + resources and has a minor impact to performance.
21184 +config AUFS_SBILIST
21186 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
21189 + Automatic configuration for internal use.
21190 + When aufs supports Magic SysRq or /proc, enabled automatically.
21192 +config AUFS_HNOTIFY
21193 + bool "Detect direct branch access (bypassing aufs)"
21195 + If you want to modify files on branches directly, eg. bypassing aufs,
21196 + and want aufs to detect the changes of them fully, then enable this
21197 + option and use 'udba=notify' mount option.
21198 + Currently there is only one available configuration, "fsnotify".
21199 + It will have a negative impact to the performance.
21200 + See detail in aufs.5.
21203 + prompt "method" if AUFS_HNOTIFY
21204 + default AUFS_HFSNOTIFY
21205 +config AUFS_HFSNOTIFY
21210 +config AUFS_EXPORT
21211 + bool "NFS-exportable aufs"
21212 + depends on EXPORTFS
21214 + If you want to export your mounted aufs via NFS, then enable this
21215 + option. There are several requirements for this configuration.
21216 + See detail in aufs.5.
21218 +config AUFS_INO_T_64
21220 + depends on AUFS_EXPORT
21221 + depends on 64BIT && !(ALPHA || S390)
21224 + Automatic configuration for internal use.
21225 + /* typedef unsigned long/int __kernel_ino_t */
21226 + /* alpha and s390x are int */
21229 + bool "support for XATTR/EA (including Security Labels)"
21231 + If your branch fs supports XATTR/EA and you want to make them
21232 + available in aufs too, then enable this opsion and specify the
21233 + branch attributes for EA.
21234 + See detail in aufs.5.
21237 + bool "File-based Hierarchical Storage Management"
21239 + Hierarchical Storage Management (or HSM) is a well-known feature
21240 + in the storage world. Aufs provides this feature as file-based.
21241 + with multiple branches.
21242 + These multiple branches are prioritized, ie. the topmost one
21243 + should be the fastest drive and be used heavily.
21246 + bool "Readdir in userspace"
21248 + Aufs has two methods to provide a merged view for a directory,
21249 + by a user-space library and by kernel-space natively. The latter
21250 + is always enabled but sometimes large and slow.
21251 + If you enable this option, install the library in aufs2-util
21252 + package, and set some environment variables for your readdir(3),
21253 + then the work will be handled in user-space which generally
21254 + shows better performance in most cases.
21255 + See detail in aufs.5.
21258 + bool "Show whiteouts"
21260 + If you want to make the whiteouts in aufs visible, then enable
21261 + this option and specify 'shwh' mount option. Although it may
21262 + sounds like philosophy or something, but in technically it
21263 + simply shows the name of whiteout with keeping its behaviour.
21265 +config AUFS_BR_RAMFS
21266 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
21268 + If you want to use ramfs as an aufs branch fs, then enable this
21269 + option. Generally tmpfs is recommended.
21270 + Aufs prohibited them to be a branch fs by default, because
21271 + initramfs becomes unusable after switch_root or something
21272 + generally. If you sets initramfs as an aufs branch and boot your
21273 + system by switch_root, you will meet a problem easily since the
21274 + files in initramfs may be inaccessible.
21275 + Unless you are going to use ramfs as an aufs branch fs without
21276 + switch_root or something, leave it N.
21278 +config AUFS_BR_FUSE
21279 + bool "Fuse fs as an aufs branch"
21280 + depends on FUSE_FS
21283 + If you want to use fuse-based userspace filesystem as an aufs
21284 + branch fs, then enable this option.
21285 + It implements the internal poll(2) operation which is
21286 + implemented by fuse only (curretnly).
21291 + Automatic configuration for internal use.
21293 +config AUFS_BR_HFSPLUS
21294 + bool "Hfsplus as an aufs branch"
21295 + depends on HFSPLUS_FS
21298 + If you want to use hfsplus fs as an aufs branch fs, then enable
21299 + this option. This option introduces a small overhead at
21300 + copying-up a file on hfsplus.
21302 +config AUFS_BDEV_LOOP
21304 + depends on BLK_DEV_LOOP
21307 + Automatic configuration for internal use.
21308 + Convert =[ym] into =y.
21311 + bool "Debug aufs"
21313 + Enable this to compile aufs internal debug code.
21314 + It will have a negative impact to the performance.
21316 +config AUFS_MAGIC_SYSRQ
21318 + depends on AUFS_DEBUG && MAGIC_SYSRQ
21321 + Automatic configuration for internal use.
21322 + When aufs supports Magic SysRq, enabled automatically.
21324 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
21325 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
21326 +++ linux/fs/aufs/loop.c 2015-11-11 17:21:46.918863802 +0100
21329 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21331 + * This program, aufs is free software; you can redistribute it and/or modify
21332 + * it under the terms of the GNU General Public License as published by
21333 + * the Free Software Foundation; either version 2 of the License, or
21334 + * (at your option) any later version.
21336 + * This program is distributed in the hope that it will be useful,
21337 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21338 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21339 + * GNU General Public License for more details.
21341 + * You should have received a copy of the GNU General Public License
21342 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21346 + * support for loopback block device as a branch
21351 +/* added into drivers/block/loop.c */
21352 +static struct file *(*backing_file_func)(struct super_block *sb);
21355 + * test if two lower dentries have overlapping branches.
21357 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
21359 + struct super_block *h_sb;
21360 + struct file *backing_file;
21362 + if (unlikely(!backing_file_func)) {
21363 + /* don't load "loop" module here */
21364 + backing_file_func = symbol_get(loop_backing_file);
21365 + if (unlikely(!backing_file_func))
21366 + /* "loop" module is not loaded */
21370 + h_sb = h_adding->d_sb;
21371 + backing_file = backing_file_func(h_sb);
21372 + if (!backing_file)
21375 + h_adding = backing_file->f_path.dentry;
21377 + * h_adding can be local NFS.
21378 + * in this case aufs cannot detect the loop.
21380 + if (unlikely(h_adding->d_sb == sb))
21382 + return !!au_test_subdir(h_adding, sb->s_root);
21385 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
21386 +int au_test_loopback_kthread(void)
21389 + struct task_struct *tsk = current;
21390 + char c, comm[sizeof(tsk->comm)];
21393 + if (tsk->flags & PF_KTHREAD) {
21394 + get_task_comm(comm, tsk);
21396 + ret = ('0' <= c && c <= '9'
21397 + && !strncmp(comm, "loop", 4));
21403 +/* ---------------------------------------------------------------------- */
21405 +#define au_warn_loopback_step 16
21406 +static int au_warn_loopback_nelem = au_warn_loopback_step;
21407 +static unsigned long *au_warn_loopback_array;
21409 +void au_warn_loopback(struct super_block *h_sb)
21411 + int i, new_nelem;
21412 + unsigned long *a, magic;
21413 + static DEFINE_SPINLOCK(spin);
21415 + magic = h_sb->s_magic;
21416 + spin_lock(&spin);
21417 + a = au_warn_loopback_array;
21418 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
21419 + if (a[i] == magic) {
21420 + spin_unlock(&spin);
21424 + /* h_sb is new to us, print it */
21425 + if (i < au_warn_loopback_nelem) {
21430 + /* expand the array */
21431 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
21432 + a = au_kzrealloc(au_warn_loopback_array,
21433 + au_warn_loopback_nelem * sizeof(unsigned long),
21434 + new_nelem * sizeof(unsigned long), GFP_ATOMIC);
21436 + au_warn_loopback_nelem = new_nelem;
21437 + au_warn_loopback_array = a;
21442 + spin_unlock(&spin);
21443 + AuWarn1("realloc failed, ignored\n");
21447 + spin_unlock(&spin);
21448 + pr_warn("you may want to try another patch for loopback file "
21449 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
21452 +int au_loopback_init(void)
21455 + struct super_block *sb __maybe_unused;
21457 + BUILD_BUG_ON(sizeof(sb->s_magic) != sizeof(unsigned long));
21460 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
21461 + sizeof(unsigned long), GFP_NOFS);
21462 + if (unlikely(!au_warn_loopback_array))
21468 +void au_loopback_fin(void)
21470 + if (backing_file_func)
21471 + symbol_put(loop_backing_file);
21472 + kfree(au_warn_loopback_array);
21474 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
21475 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
21476 +++ linux/fs/aufs/loop.h 2015-09-24 10:47:58.254719746 +0200
21479 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21481 + * This program, aufs is free software; you can redistribute it and/or modify
21482 + * it under the terms of the GNU General Public License as published by
21483 + * the Free Software Foundation; either version 2 of the License, or
21484 + * (at your option) any later version.
21486 + * This program is distributed in the hope that it will be useful,
21487 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21488 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21489 + * GNU General Public License for more details.
21491 + * You should have received a copy of the GNU General Public License
21492 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21496 + * support for loopback mount as a branch
21499 +#ifndef __AUFS_LOOP_H__
21500 +#define __AUFS_LOOP_H__
21505 +struct super_block;
21507 +#ifdef CONFIG_AUFS_BDEV_LOOP
21508 +/* drivers/block/loop.c */
21509 +struct file *loop_backing_file(struct super_block *sb);
21512 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
21513 +int au_test_loopback_kthread(void);
21514 +void au_warn_loopback(struct super_block *h_sb);
21516 +int au_loopback_init(void);
21517 +void au_loopback_fin(void);
21519 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
21520 + struct dentry *h_adding)
21521 +AuStubInt0(au_test_loopback_kthread, void)
21522 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
21524 +AuStubInt0(au_loopback_init, void)
21525 +AuStubVoid(au_loopback_fin, void)
21526 +#endif /* BLK_DEV_LOOP */
21528 +#endif /* __KERNEL__ */
21529 +#endif /* __AUFS_LOOP_H__ */
21530 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
21531 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
21532 +++ linux/fs/aufs/magic.mk 2015-09-24 10:47:58.254719746 +0200
21535 +# defined in ${srctree}/fs/fuse/inode.c
21537 +ifdef CONFIG_FUSE_FS
21538 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
21541 +# defined in ${srctree}/fs/xfs/xfs_sb.h
21543 +ifdef CONFIG_XFS_FS
21544 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
21547 +# defined in ${srctree}/fs/configfs/mount.c
21549 +ifdef CONFIG_CONFIGFS_FS
21550 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
21553 +# defined in ${srctree}/fs/ubifs/ubifs.h
21555 +ifdef CONFIG_UBIFS_FS
21556 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
21559 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
21561 +ifdef CONFIG_HFSPLUS_FS
21562 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
21564 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
21565 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
21566 +++ linux/fs/aufs/Makefile 2015-09-24 10:47:58.248052907 +0200
21569 +include ${src}/magic.mk
21570 +ifeq (${CONFIG_AUFS_FS},m)
21571 +include ${src}/conf.mk
21573 +-include ${src}/priv_def.mk
21575 +# cf. include/linux/kernel.h
21577 +ccflags-y += -DDEBUG
21578 +# sparse requires the full pathname
21580 +ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
21582 +ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
21585 +obj-$(CONFIG_AUFS_FS) += aufs.o
21586 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o \
21587 + wkq.o vfsub.o dcsub.o \
21588 + cpup.o whout.o wbr_policy.o \
21589 + dinfo.o dentry.o \
21591 + finfo.o file.o f_op.o \
21593 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
21597 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
21598 +aufs-$(CONFIG_SYSFS) += sysfs.o
21599 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
21600 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
21601 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
21602 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
21603 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
21604 +aufs-$(CONFIG_AUFS_XATTR) += xattr.o
21605 +aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
21606 +aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
21607 +aufs-$(CONFIG_AUFS_POLL) += poll.o
21608 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
21609 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
21610 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
21611 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
21612 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
21613 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
21614 +++ linux/fs/aufs/module.c 2015-11-11 17:21:46.918863802 +0100
21617 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21619 + * This program, aufs is free software; you can redistribute it and/or modify
21620 + * it under the terms of the GNU General Public License as published by
21621 + * the Free Software Foundation; either version 2 of the License, or
21622 + * (at your option) any later version.
21624 + * This program is distributed in the hope that it will be useful,
21625 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21626 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21627 + * GNU General Public License for more details.
21629 + * You should have received a copy of the GNU General Public License
21630 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21634 + * module global variables and operations
21637 +#include <linux/module.h>
21638 +#include <linux/seq_file.h>
21641 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp)
21643 + if (new_sz <= nused)
21646 + p = krealloc(p, new_sz, gfp);
21648 + memset(p + nused, 0, new_sz - nused);
21652 +/* ---------------------------------------------------------------------- */
21657 +struct kmem_cache *au_cachep[AuCache_Last];
21658 +static int __init au_cache_init(void)
21660 + au_cachep[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
21661 + if (au_cachep[AuCache_DINFO])
21662 + /* SLAB_DESTROY_BY_RCU */
21663 + au_cachep[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
21664 + au_icntnr_init_once);
21665 + if (au_cachep[AuCache_ICNTNR])
21666 + au_cachep[AuCache_FINFO] = AuCacheCtor(au_finfo,
21667 + au_fi_init_once);
21668 + if (au_cachep[AuCache_FINFO])
21669 + au_cachep[AuCache_VDIR] = AuCache(au_vdir);
21670 + if (au_cachep[AuCache_VDIR])
21671 + au_cachep[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
21672 + if (au_cachep[AuCache_DEHSTR])
21678 +static void au_cache_fin(void)
21683 + * Make sure all delayed rcu free inodes are flushed before we
21688 + /* excluding AuCache_HNOTIFY */
21689 + BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
21690 + for (i = 0; i < AuCache_HNOTIFY; i++) {
21691 + kmem_cache_destroy(au_cachep[i]);
21692 + au_cachep[i] = NULL;
21696 +/* ---------------------------------------------------------------------- */
21698 +int au_dir_roflags;
21700 +#ifdef CONFIG_AUFS_SBILIST
21702 + * iterate_supers_type() doesn't protect us from
21703 + * remounting (branch management)
21705 +struct au_splhead au_sbilist;
21708 +struct lock_class_key au_lc_key[AuLcKey_Last];
21711 + * functions for module interface.
21713 +MODULE_LICENSE("GPL");
21714 +/* MODULE_LICENSE("GPL v2"); */
21715 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
21716 +MODULE_DESCRIPTION(AUFS_NAME
21717 + " -- Advanced multi layered unification filesystem");
21718 +MODULE_VERSION(AUFS_VERSION);
21719 +MODULE_ALIAS_FS(AUFS_NAME);
21721 +/* this module parameter has no meaning when SYSFS is disabled */
21722 +int sysaufs_brs = 1;
21723 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
21724 +module_param_named(brs, sysaufs_brs, int, S_IRUGO);
21726 +/* this module parameter has no meaning when USER_NS is disabled */
21727 +static bool au_userns;
21728 +MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
21729 +module_param_named(allow_userns, au_userns, bool, S_IRUGO);
21731 +/* ---------------------------------------------------------------------- */
21733 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
21735 +int au_seq_path(struct seq_file *seq, struct path *path)
21739 + err = seq_path(seq, path, au_esc_chars);
21742 + else if (err < 0)
21748 +/* ---------------------------------------------------------------------- */
21750 +static int __init aufs_init(void)
21755 + p = au_esc_chars;
21756 + for (i = 1; i <= ' '; i++)
21762 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
21764 + au_sbilist_init();
21765 + sysaufs_brs_init();
21768 + err = sysaufs_init();
21769 + if (unlikely(err))
21771 + err = au_procfs_init();
21772 + if (unlikely(err))
21773 + goto out_sysaufs;
21774 + err = au_wkq_init();
21775 + if (unlikely(err))
21777 + err = au_loopback_init();
21778 + if (unlikely(err))
21780 + err = au_hnotify_init();
21781 + if (unlikely(err))
21782 + goto out_loopback;
21783 + err = au_sysrq_init();
21784 + if (unlikely(err))
21786 + err = au_cache_init();
21787 + if (unlikely(err))
21790 + aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
21791 + err = register_filesystem(&aufs_fs_type);
21792 + if (unlikely(err))
21795 + /* since we define pr_fmt, call printk directly */
21796 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
21797 + goto out; /* success */
21804 + au_hnotify_fin();
21806 + au_loopback_fin();
21818 +static void __exit aufs_exit(void)
21820 + unregister_filesystem(&aufs_fs_type);
21823 + au_hnotify_fin();
21824 + au_loopback_fin();
21831 +module_init(aufs_init);
21832 +module_exit(aufs_exit);
21833 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
21834 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
21835 +++ linux/fs/aufs/module.h 2015-09-24 10:47:58.254719746 +0200
21838 + * Copyright (C) 2005-2015 Junjiro R. Okajima
21840 + * This program, aufs is free software; you can redistribute it and/or modify
21841 + * it under the terms of the GNU General Public License as published by
21842 + * the Free Software Foundation; either version 2 of the License, or
21843 + * (at your option) any later version.
21845 + * This program is distributed in the hope that it will be useful,
21846 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21847 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21848 + * GNU General Public License for more details.
21850 + * You should have received a copy of the GNU General Public License
21851 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21855 + * module initialization and module-global
21858 +#ifndef __AUFS_MODULE_H__
21859 +#define __AUFS_MODULE_H__
21863 +#include <linux/slab.h>
21868 +/* module parameters */
21869 +extern int sysaufs_brs;
21871 +/* ---------------------------------------------------------------------- */
21873 +extern int au_dir_roflags;
21876 + AuLcNonDir_FIINFO,
21877 + AuLcNonDir_DIINFO,
21878 + AuLcNonDir_IIINFO,
21884 + AuLcSymlink_DIINFO,
21885 + AuLcSymlink_IIINFO,
21889 +extern struct lock_class_key au_lc_key[AuLcKey_Last];
21891 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp);
21892 +int au_seq_path(struct seq_file *seq, struct path *path);
21894 +#ifdef CONFIG_PROC_FS
21896 +int __init au_procfs_init(void);
21897 +void au_procfs_fin(void);
21899 +AuStubInt0(au_procfs_init, void);
21900 +AuStubVoid(au_procfs_fin, void);
21903 +/* ---------------------------------------------------------------------- */
21912 + AuCache_HNOTIFY, /* must be last */
21916 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
21917 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
21918 +#define AuCacheCtor(type, ctor) \
21919 + kmem_cache_create(#type, sizeof(struct type), \
21920 + __alignof__(struct type), AuCacheFlags, ctor)
21922 +extern struct kmem_cache *au_cachep[];
21924 +#define AuCacheFuncs(name, index) \
21925 +static inline struct au_##name *au_cache_alloc_##name(void) \
21926 +{ return kmem_cache_alloc(au_cachep[AuCache_##index], GFP_NOFS); } \
21927 +static inline void au_cache_free_##name(struct au_##name *p) \
21928 +{ kmem_cache_free(au_cachep[AuCache_##index], p); }
21930 +AuCacheFuncs(dinfo, DINFO);
21931 +AuCacheFuncs(icntnr, ICNTNR);
21932 +AuCacheFuncs(finfo, FINFO);
21933 +AuCacheFuncs(vdir, VDIR);
21934 +AuCacheFuncs(vdir_dehstr, DEHSTR);
21935 +#ifdef CONFIG_AUFS_HNOTIFY
21936 +AuCacheFuncs(hnotify, HNOTIFY);
21939 +#endif /* __KERNEL__ */
21940 +#endif /* __AUFS_MODULE_H__ */
21941 diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
21942 --- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
21943 +++ linux/fs/aufs/mvdown.c 2015-11-11 17:21:46.918863802 +0100
21946 + * Copyright (C) 2011-2015 Junjiro R. Okajima
21948 + * This program, aufs is free software; you can redistribute it and/or modify
21949 + * it under the terms of the GNU General Public License as published by
21950 + * the Free Software Foundation; either version 2 of the License, or
21951 + * (at your option) any later version.
21953 + * This program is distributed in the hope that it will be useful,
21954 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21955 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21956 + * GNU General Public License for more details.
21958 + * You should have received a copy of the GNU General Public License
21959 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
21963 + * move-down, opposite of copy-up
21968 +struct au_mvd_args {
21970 + struct super_block *h_sb;
21971 + struct dentry *h_parent;
21972 + struct au_hinode *hdir;
21973 + struct inode *h_dir, *h_inode;
21974 + struct au_pin pin;
21975 + } info[AUFS_MVDOWN_NARRAY];
21977 + struct aufs_mvdown mvdown;
21978 + struct dentry *dentry, *parent;
21979 + struct inode *inode, *dir;
21980 + struct super_block *sb;
21981 + aufs_bindex_t bopq, bwh, bfound;
21982 + unsigned char rename_lock;
21985 +#define mvd_errno mvdown.au_errno
21986 +#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
21987 +#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
21988 +#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
21989 +#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
21991 +#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
21992 +#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
21993 +#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
21994 +#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
21995 +#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
21996 +#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
21998 +#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
21999 +#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
22000 +#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
22001 +#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
22002 +#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
22003 +#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
22005 +#define AU_MVD_PR(flag, ...) do { \
22007 + pr_err(__VA_ARGS__); \
22010 +static int find_lower_writable(struct au_mvd_args *a)
22012 + struct super_block *sb;
22013 + aufs_bindex_t bindex, bend;
22014 + struct au_branch *br;
22017 + bindex = a->mvd_bsrc;
22018 + bend = au_sbend(sb);
22019 + if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
22020 + for (bindex++; bindex <= bend; bindex++) {
22021 + br = au_sbr(sb, bindex);
22022 + if (au_br_fhsm(br->br_perm)
22023 + && (!(au_br_sb(br)->s_flags & MS_RDONLY)))
22026 + else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
22027 + for (bindex++; bindex <= bend; bindex++) {
22028 + br = au_sbr(sb, bindex);
22029 + if (!au_br_rdonly(br))
22033 + for (bindex++; bindex <= bend; bindex++) {
22034 + br = au_sbr(sb, bindex);
22035 + if (!(au_br_sb(br)->s_flags & MS_RDONLY)) {
22036 + if (au_br_rdonly(br))
22038 + |= AUFS_MVDOWN_ROLOWER_R;
22046 +/* make the parent dir on bdst */
22047 +static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
22052 + a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
22053 + a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
22054 + a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
22055 + a->mvd_h_dst_parent = NULL;
22056 + if (au_dbend(a->parent) >= a->mvd_bdst)
22057 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
22058 + if (!a->mvd_h_dst_parent) {
22059 + err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
22060 + if (unlikely(err)) {
22061 + AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
22064 + a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
22072 +/* lock them all */
22073 +static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
22076 + struct dentry *h_trap;
22078 + a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
22079 + a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
22080 + err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
22081 + au_opt_udba(a->sb),
22082 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22084 + if (unlikely(err)) {
22085 + AU_MVD_PR(dmsg, "pin_dst failed\n");
22089 + if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
22090 + a->rename_lock = 0;
22091 + au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
22092 + AuLsc_DI_PARENT, AuLsc_I_PARENT3,
22093 + au_opt_udba(a->sb),
22094 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22095 + err = au_do_pin(&a->mvd_pin_src);
22097 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
22098 + if (unlikely(err)) {
22099 + AU_MVD_PR(dmsg, "pin_src failed\n");
22102 + goto out; /* success */
22105 + a->rename_lock = 1;
22106 + au_pin_hdir_unlock(&a->mvd_pin_dst);
22107 + err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
22108 + au_opt_udba(a->sb),
22109 + AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22111 + a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
22112 + if (unlikely(err)) {
22113 + AU_MVD_PR(dmsg, "pin_src failed\n");
22114 + au_pin_hdir_lock(&a->mvd_pin_dst);
22117 + au_pin_hdir_unlock(&a->mvd_pin_src);
22118 + h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22119 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
22121 + err = (h_trap != a->mvd_h_src_parent);
22123 + err = (h_trap != a->mvd_h_dst_parent);
22125 + BUG_ON(err); /* it should never happen */
22126 + if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
22129 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22130 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
22131 + au_pin_hdir_lock(&a->mvd_pin_src);
22132 + au_unpin(&a->mvd_pin_src);
22133 + au_pin_hdir_lock(&a->mvd_pin_dst);
22136 + goto out; /* success */
22139 + au_unpin(&a->mvd_pin_dst);
22145 +static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
22147 + if (!a->rename_lock)
22148 + au_unpin(&a->mvd_pin_src);
22150 + vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22151 + a->mvd_h_dst_parent, a->mvd_hdir_dst);
22152 + au_pin_hdir_lock(&a->mvd_pin_src);
22153 + au_unpin(&a->mvd_pin_src);
22154 + au_pin_hdir_lock(&a->mvd_pin_dst);
22156 + au_unpin(&a->mvd_pin_dst);
22159 +/* copy-down the file */
22160 +static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
22163 + struct au_cp_generic cpg = {
22164 + .dentry = a->dentry,
22165 + .bdst = a->mvd_bdst,
22166 + .bsrc = a->mvd_bsrc,
22168 + .pin = &a->mvd_pin_dst,
22169 + .flags = AuCpup_DTIME | AuCpup_HOPEN
22172 + AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
22173 + if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
22174 + au_fset_cpup(cpg.flags, OVERWRITE);
22175 + if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
22176 + au_fset_cpup(cpg.flags, RWDST);
22177 + err = au_sio_cpdown_simple(&cpg);
22178 + if (unlikely(err))
22179 + AU_MVD_PR(dmsg, "cpdown failed\n");
22186 + * unlink the whiteout on bdst if exist which may be created by UDBA while we
22189 +static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
22192 + struct path h_path;
22193 + struct au_branch *br;
22194 + struct inode *delegated;
22196 + br = au_sbr(a->sb, a->mvd_bdst);
22197 + h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
22198 + err = PTR_ERR(h_path.dentry);
22199 + if (IS_ERR(h_path.dentry)) {
22200 + AU_MVD_PR(dmsg, "wh_lkup failed\n");
22205 + if (d_is_positive(h_path.dentry)) {
22206 + h_path.mnt = au_br_mnt(br);
22207 + delegated = NULL;
22208 + err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
22209 + &delegated, /*force*/0);
22210 + if (unlikely(err == -EWOULDBLOCK)) {
22211 + pr_warn("cannot retry for NFSv4 delegation"
22212 + " for an internal unlink\n");
22215 + if (unlikely(err))
22216 + AU_MVD_PR(dmsg, "wh_unlink failed\n");
22218 + dput(h_path.dentry);
22226 + * unlink the topmost h_dentry
22228 +static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
22231 + struct path h_path;
22232 + struct inode *delegated;
22234 + h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
22235 + h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
22236 + delegated = NULL;
22237 + err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
22238 + if (unlikely(err == -EWOULDBLOCK)) {
22239 + pr_warn("cannot retry for NFSv4 delegation"
22240 + " for an internal unlink\n");
22243 + if (unlikely(err))
22244 + AU_MVD_PR(dmsg, "unlink failed\n");
22250 +/* Since mvdown succeeded, we ignore an error of this function */
22251 +static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
22254 + struct au_branch *br;
22256 + a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
22257 + br = au_sbr(a->sb, a->mvd_bsrc);
22258 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
22260 + br = au_sbr(a->sb, a->mvd_bdst);
22261 + a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
22262 + err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
22265 + a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
22267 + AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
22271 + * copy-down the file and unlink the bsrc file.
22272 + * - unlink the bdst whout if exist
22273 + * - copy-down the file (with whtmp name and rename)
22274 + * - unlink the bsrc file
22276 +static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
22280 + err = au_do_mkdir(dmsg, a);
22282 + err = au_do_lock(dmsg, a);
22283 + if (unlikely(err))
22287 + * do not revert the activities we made on bdst since they should be
22288 + * harmless in aufs.
22291 + err = au_do_cpdown(dmsg, a);
22293 + err = au_do_unlink_wh(dmsg, a);
22294 + if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
22295 + err = au_do_unlink(dmsg, a);
22296 + if (unlikely(err))
22299 + AuDbg("%pd2, 0x%x, %d --> %d\n",
22300 + a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
22301 + if (find_lower_writable(a) < 0)
22302 + a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
22304 + if (a->mvdown.flags & AUFS_MVDOWN_STFS)
22305 + au_do_stfs(dmsg, a);
22307 + /* maintain internal array */
22308 + if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
22309 + au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
22310 + au_set_dbstart(a->dentry, a->mvd_bdst);
22311 + au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
22312 + au_set_ibstart(a->inode, a->mvd_bdst);
22314 + /* hide the lower */
22315 + au_set_h_dptr(a->dentry, a->mvd_bdst, NULL);
22316 + au_set_dbend(a->dentry, a->mvd_bsrc);
22317 + au_set_h_iptr(a->inode, a->mvd_bdst, NULL, /*flags*/0);
22318 + au_set_ibend(a->inode, a->mvd_bsrc);
22320 + if (au_dbend(a->dentry) < a->mvd_bdst)
22321 + au_set_dbend(a->dentry, a->mvd_bdst);
22322 + if (au_ibend(a->inode) < a->mvd_bdst)
22323 + au_set_ibend(a->inode, a->mvd_bdst);
22326 + au_do_unlock(dmsg, a);
22332 +/* ---------------------------------------------------------------------- */
22334 +/* make sure the file is idle */
22335 +static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
22337 + int err, plinked;
22340 + plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
22341 + if (au_dbstart(a->dentry) == a->mvd_bsrc
22342 + && au_dcount(a->dentry) == 1
22343 + && atomic_read(&a->inode->i_count) == 1
22344 + /* && a->mvd_h_src_inode->i_nlink == 1 */
22345 + && (!plinked || !au_plink_test(a->inode))
22346 + && a->inode->i_nlink == 1)
22351 + "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
22352 + a->mvd_bsrc, au_dbstart(a->dentry), au_dcount(a->dentry),
22353 + atomic_read(&a->inode->i_count), a->inode->i_nlink,
22354 + a->mvd_h_src_inode->i_nlink,
22355 + plinked, plinked ? au_plink_test(a->inode) : 0);
22362 +/* make sure the parent dir is fine */
22363 +static int au_mvd_args_parent(const unsigned char dmsg,
22364 + struct au_mvd_args *a)
22367 + aufs_bindex_t bindex;
22370 + if (unlikely(au_alive_dir(a->parent))) {
22372 + AU_MVD_PR(dmsg, "parent dir is dead\n");
22376 + a->bopq = au_dbdiropq(a->parent);
22377 + bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
22378 + AuDbg("b%d\n", bindex);
22379 + if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
22380 + || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
22382 + a->mvd_errno = EAU_MVDOWN_OPAQUE;
22383 + AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
22384 + a->bopq, a->mvd_bdst);
22392 +static int au_mvd_args_intermediate(const unsigned char dmsg,
22393 + struct au_mvd_args *a)
22396 + struct au_dinfo *dinfo, *tmp;
22398 + /* lookup the next lower positive entry */
22400 + tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
22401 + if (unlikely(!tmp))
22406 + dinfo = au_di(a->dentry);
22407 + au_di_cp(tmp, dinfo);
22408 + au_di_swap(tmp, dinfo);
22410 + /* returns the number of positive dentries */
22411 + err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1, /*type*/0);
22413 + a->bwh = au_dbwh(a->dentry);
22414 + else if (err > 0)
22415 + a->bfound = au_dbstart(a->dentry);
22417 + au_di_swap(tmp, dinfo);
22418 + au_rw_write_unlock(&tmp->di_rwsem);
22420 + if (unlikely(err < 0))
22421 + AU_MVD_PR(dmsg, "failed look-up lower\n");
22424 + * here, we have these cases.
22426 + * no positive dentry under bsrc. there are more sub-cases.
22428 + * there no whiteout, we can safely move-down.
22431 + * bsrc < bwh && bwh < bdst
22432 + * there is a whiteout on RO branch. cannot proceed.
22434 + * there is a whiteout on the RW target branch. it should
22437 + * there is a whiteout somewhere unrelated branch.
22438 + * -1 < bfound && bfound <= bsrc
22441 + * found, but it is on RO branch between bsrc and bdst. cannot
22444 + * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
22447 + * found, after we create the file on bdst, it will be hidden.
22450 + AuDebugOn(a->bfound == -1
22452 + && a->bwh <= a->mvd_bsrc);
22453 + AuDebugOn(-1 < a->bfound
22454 + && a->bfound <= a->mvd_bsrc);
22457 + if (a->bfound == -1
22458 + && a->mvd_bsrc < a->bwh
22460 + && a->bwh < a->mvd_bdst) {
22461 + a->mvd_errno = EAU_MVDOWN_WHITEOUT;
22462 + AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
22463 + a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
22465 + } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
22466 + a->mvd_errno = EAU_MVDOWN_UPPER;
22467 + AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
22468 + a->mvd_bdst, a->bfound);
22472 + err = 0; /* success */
22479 +static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
22484 + if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
22485 + && a->bfound == a->mvd_bdst)
22491 +static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
22494 + struct au_branch *br;
22497 + if (unlikely(S_ISDIR(a->inode->i_mode)))
22501 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
22502 + a->mvd_bsrc = au_ibstart(a->inode);
22504 + a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
22505 + if (unlikely(a->mvd_bsrc < 0
22506 + || (a->mvd_bsrc < au_dbstart(a->dentry)
22507 + || au_dbend(a->dentry) < a->mvd_bsrc
22508 + || !au_h_dptr(a->dentry, a->mvd_bsrc))
22509 + || (a->mvd_bsrc < au_ibstart(a->inode)
22510 + || au_ibend(a->inode) < a->mvd_bsrc
22511 + || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
22512 + a->mvd_errno = EAU_MVDOWN_NOUPPER;
22513 + AU_MVD_PR(dmsg, "no upper\n");
22517 + if (unlikely(a->mvd_bsrc == au_sbend(a->sb))) {
22518 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
22519 + AU_MVD_PR(dmsg, "on the bottom\n");
22522 + a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
22523 + br = au_sbr(a->sb, a->mvd_bsrc);
22524 + err = au_br_rdonly(br);
22525 + if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
22526 + if (unlikely(err))
22528 + } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
22529 + || IS_APPEND(a->mvd_h_src_inode))) {
22531 + a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
22537 + if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
22538 + a->mvd_bdst = find_lower_writable(a);
22539 + if (unlikely(a->mvd_bdst < 0)) {
22540 + a->mvd_errno = EAU_MVDOWN_BOTTOM;
22541 + AU_MVD_PR(dmsg, "no writable lower branch\n");
22545 + a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
22546 + if (unlikely(a->mvd_bdst < 0
22547 + || au_sbend(a->sb) < a->mvd_bdst)) {
22548 + a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
22549 + AU_MVD_PR(dmsg, "no lower brid\n");
22554 + err = au_mvd_args_busy(dmsg, a);
22556 + err = au_mvd_args_parent(dmsg, a);
22558 + err = au_mvd_args_intermediate(dmsg, a);
22560 + err = au_mvd_args_exist(dmsg, a);
22562 + AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
22569 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
22572 + unsigned char dmsg;
22573 + struct au_mvd_args *args;
22574 + struct inode *inode;
22576 + inode = d_inode(dentry);
22578 + if (unlikely(!capable(CAP_SYS_ADMIN)))
22582 + args = kmalloc(sizeof(*args), GFP_NOFS);
22583 + if (unlikely(!args))
22586 + err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
22588 + err = !access_ok(VERIFY_WRITE, uarg, sizeof(*uarg));
22589 + if (unlikely(err)) {
22594 + AuDbg("flags 0x%x\n", args->mvdown.flags);
22595 + args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
22596 + args->mvdown.au_errno = 0;
22597 + args->dentry = dentry;
22598 + args->inode = inode;
22599 + args->sb = dentry->d_sb;
22602 + dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
22603 + args->parent = dget_parent(dentry);
22604 + args->dir = d_inode(args->parent);
22605 + mutex_lock_nested(&args->dir->i_mutex, I_MUTEX_PARENT);
22606 + dput(args->parent);
22607 + if (unlikely(args->parent != dentry->d_parent)) {
22608 + AU_MVD_PR(dmsg, "parent dir is moved\n");
22612 + mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
22613 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH);
22614 + if (unlikely(err))
22617 + di_write_lock_parent(args->parent);
22618 + err = au_mvd_args(dmsg, args);
22619 + if (unlikely(err))
22622 + err = au_do_mvdown(dmsg, args);
22623 + if (unlikely(err))
22626 + au_cpup_attr_timesizes(args->dir);
22627 + au_cpup_attr_timesizes(inode);
22628 + if (!(args->mvdown.flags & AUFS_MVDOWN_KUPPER))
22629 + au_cpup_igen(inode, au_h_iptr(inode, args->mvd_bdst));
22630 + /* au_digen_dec(dentry); */
22633 + di_write_unlock(args->parent);
22634 + aufs_read_unlock(dentry, AuLock_DW);
22636 + mutex_unlock(&inode->i_mutex);
22638 + mutex_unlock(&args->dir->i_mutex);
22640 + e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
22648 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
22649 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
22650 +++ linux/fs/aufs/opts.c 2015-11-11 17:21:46.918863802 +0100
22653 + * Copyright (C) 2005-2015 Junjiro R. Okajima
22655 + * This program, aufs is free software; you can redistribute it and/or modify
22656 + * it under the terms of the GNU General Public License as published by
22657 + * the Free Software Foundation; either version 2 of the License, or
22658 + * (at your option) any later version.
22660 + * This program is distributed in the hope that it will be useful,
22661 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22662 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22663 + * GNU General Public License for more details.
22665 + * You should have received a copy of the GNU General Public License
22666 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
22670 + * mount options/flags
22673 +#include <linux/namei.h>
22674 +#include <linux/types.h> /* a distribution requires */
22675 +#include <linux/parser.h>
22678 +/* ---------------------------------------------------------------------- */
22682 + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
22683 + Opt_idel, Opt_imod,
22684 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
22685 + Opt_rdblk_def, Opt_rdhash_def,
22686 + Opt_xino, Opt_noxino,
22687 + Opt_trunc_xino, Opt_trunc_xino_v, Opt_notrunc_xino,
22688 + Opt_trunc_xino_path, Opt_itrunc_xino,
22689 + Opt_trunc_xib, Opt_notrunc_xib,
22690 + Opt_shwh, Opt_noshwh,
22691 + Opt_plink, Opt_noplink, Opt_list_plink,
22693 + Opt_dio, Opt_nodio,
22694 + Opt_diropq_a, Opt_diropq_w,
22695 + Opt_warn_perm, Opt_nowarn_perm,
22696 + Opt_wbr_copyup, Opt_wbr_create,
22698 + Opt_verbose, Opt_noverbose,
22699 + Opt_sum, Opt_nosum, Opt_wsum,
22700 + Opt_dirperm1, Opt_nodirperm1,
22701 + Opt_acl, Opt_noacl,
22702 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
22705 +static match_table_t options = {
22706 + {Opt_br, "br=%s"},
22707 + {Opt_br, "br:%s"},
22709 + {Opt_add, "add=%d:%s"},
22710 + {Opt_add, "add:%d:%s"},
22711 + {Opt_add, "ins=%d:%s"},
22712 + {Opt_add, "ins:%d:%s"},
22713 + {Opt_append, "append=%s"},
22714 + {Opt_append, "append:%s"},
22715 + {Opt_prepend, "prepend=%s"},
22716 + {Opt_prepend, "prepend:%s"},
22718 + {Opt_del, "del=%s"},
22719 + {Opt_del, "del:%s"},
22720 + /* {Opt_idel, "idel:%d"}, */
22721 + {Opt_mod, "mod=%s"},
22722 + {Opt_mod, "mod:%s"},
22723 + /* {Opt_imod, "imod:%d:%s"}, */
22725 + {Opt_dirwh, "dirwh=%d"},
22727 + {Opt_xino, "xino=%s"},
22728 + {Opt_noxino, "noxino"},
22729 + {Opt_trunc_xino, "trunc_xino"},
22730 + {Opt_trunc_xino_v, "trunc_xino_v=%d:%d"},
22731 + {Opt_notrunc_xino, "notrunc_xino"},
22732 + {Opt_trunc_xino_path, "trunc_xino=%s"},
22733 + {Opt_itrunc_xino, "itrunc_xino=%d"},
22734 + /* {Opt_zxino, "zxino=%s"}, */
22735 + {Opt_trunc_xib, "trunc_xib"},
22736 + {Opt_notrunc_xib, "notrunc_xib"},
22738 +#ifdef CONFIG_PROC_FS
22739 + {Opt_plink, "plink"},
22741 + {Opt_ignore_silent, "plink"},
22744 + {Opt_noplink, "noplink"},
22746 +#ifdef CONFIG_AUFS_DEBUG
22747 + {Opt_list_plink, "list_plink"},
22750 + {Opt_udba, "udba=%s"},
22752 + {Opt_dio, "dio"},
22753 + {Opt_nodio, "nodio"},
22755 +#ifdef CONFIG_AUFS_FHSM
22756 + {Opt_fhsm_sec, "fhsm_sec=%d"},
22758 + {Opt_ignore_silent, "fhsm_sec=%d"},
22761 + {Opt_diropq_a, "diropq=always"},
22762 + {Opt_diropq_a, "diropq=a"},
22763 + {Opt_diropq_w, "diropq=whiteouted"},
22764 + {Opt_diropq_w, "diropq=w"},
22766 + {Opt_warn_perm, "warn_perm"},
22767 + {Opt_nowarn_perm, "nowarn_perm"},
22769 + /* keep them temporary */
22770 + {Opt_ignore_silent, "nodlgt"},
22771 + {Opt_ignore_silent, "clean_plink"},
22773 +#ifdef CONFIG_AUFS_SHWH
22774 + {Opt_shwh, "shwh"},
22776 + {Opt_noshwh, "noshwh"},
22778 + {Opt_dirperm1, "dirperm1"},
22779 + {Opt_nodirperm1, "nodirperm1"},
22781 + {Opt_verbose, "verbose"},
22782 + {Opt_verbose, "v"},
22783 + {Opt_noverbose, "noverbose"},
22784 + {Opt_noverbose, "quiet"},
22785 + {Opt_noverbose, "q"},
22786 + {Opt_noverbose, "silent"},
22788 + {Opt_sum, "sum"},
22789 + {Opt_nosum, "nosum"},
22790 + {Opt_wsum, "wsum"},
22792 + {Opt_rdcache, "rdcache=%d"},
22793 + {Opt_rdblk, "rdblk=%d"},
22794 + {Opt_rdblk_def, "rdblk=def"},
22795 + {Opt_rdhash, "rdhash=%d"},
22796 + {Opt_rdhash_def, "rdhash=def"},
22798 + {Opt_wbr_create, "create=%s"},
22799 + {Opt_wbr_create, "create_policy=%s"},
22800 + {Opt_wbr_copyup, "cpup=%s"},
22801 + {Opt_wbr_copyup, "copyup=%s"},
22802 + {Opt_wbr_copyup, "copyup_policy=%s"},
22804 + /* generic VFS flag */
22805 +#ifdef CONFIG_FS_POSIX_ACL
22806 + {Opt_acl, "acl"},
22807 + {Opt_noacl, "noacl"},
22809 + {Opt_ignore_silent, "acl"},
22810 + {Opt_ignore_silent, "noacl"},
22813 + /* internal use for the scripts */
22814 + {Opt_ignore_silent, "si=%s"},
22816 + {Opt_br, "dirs=%s"},
22817 + {Opt_ignore, "debug=%d"},
22818 + {Opt_ignore, "delete=whiteout"},
22819 + {Opt_ignore, "delete=all"},
22820 + {Opt_ignore, "imap=%s"},
22822 + /* temporary workaround, due to old mount(8)? */
22823 + {Opt_ignore_silent, "relatime"},
22828 +/* ---------------------------------------------------------------------- */
22830 +static const char *au_parser_pattern(int val, match_table_t tbl)
22832 + struct match_token *p;
22835 + while (p->pattern) {
22836 + if (p->token == val)
22837 + return p->pattern;
22844 +static const char *au_optstr(int *val, match_table_t tbl)
22846 + struct match_token *p;
22853 + while (p->pattern) {
22855 + && (v & p->token) == p->token) {
22856 + *val &= ~p->token;
22857 + return p->pattern;
22866 +/* ---------------------------------------------------------------------- */
22868 +static match_table_t brperm = {
22869 + {AuBrPerm_RO, AUFS_BRPERM_RO},
22870 + {AuBrPerm_RR, AUFS_BRPERM_RR},
22871 + {AuBrPerm_RW, AUFS_BRPERM_RW},
22875 +static match_table_t brattr = {
22877 + {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
22878 + {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
22879 + /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
22880 + {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
22881 +#ifdef CONFIG_AUFS_FHSM
22882 + {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
22884 +#ifdef CONFIG_AUFS_XATTR
22885 + {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
22886 + {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
22887 + {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
22888 + {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
22889 + {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
22890 + {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
22893 + /* ro/rr branch */
22894 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
22897 + {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
22898 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
22903 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
22910 + p = strchr(str, '+');
22913 + v = match_token(str, table, args);
22915 + if (v & AuBrAttr_CMOO_Mask)
22916 + attr &= ~AuBrAttr_CMOO_Mask;
22921 + pr_warn("ignored branch attribute %s\n", str);
22931 +static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
22939 + p = au_optstr(&perm, brattr);
22942 + memcpy(q, p, sz + 1);
22948 + p = au_optstr(&perm, brattr);
22952 + memcpy(q, p, sz + 1);
22958 + return q - str->a;
22961 +static int noinline_for_stack br_perm_val(char *perm)
22963 + int val, bad, sz;
22965 + substring_t args[MAX_OPT_ARGS];
22966 + au_br_perm_str_t attr;
22968 + p = strchr(perm, '+');
22971 + val = match_token(perm, brperm, args);
22975 + pr_warn("ignored branch permission %s\n", perm);
22976 + val = AuBrPerm_RO;
22982 + val |= br_attr_val(p + 1, brattr, args);
22985 + switch (val & AuBrPerm_Mask) {
22986 + case AuBrPerm_RO:
22987 + case AuBrPerm_RR:
22988 + bad = val & AuBrWAttr_Mask;
22989 + val &= ~AuBrWAttr_Mask;
22991 + case AuBrPerm_RW:
22992 + bad = val & AuBrRAttr_Mask;
22993 + val &= ~AuBrRAttr_Mask;
22998 + * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
22999 + * does not treat it as an error, just warning.
23000 + * this is a tiny guard for the user operation.
23002 + if (val & AuBrAttr_UNPIN) {
23003 + bad |= AuBrAttr_UNPIN;
23004 + val &= ~AuBrAttr_UNPIN;
23007 + if (unlikely(bad)) {
23008 + sz = au_do_optstr_br_attr(&attr, bad);
23010 + pr_warn("ignored branch attribute %s\n", attr.a);
23017 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
23019 + au_br_perm_str_t attr;
23025 + p = au_optstr(&perm, brperm);
23026 + AuDebugOn(!p || !*p);
23028 + memcpy(q, p, sz + 1);
23031 + sz = au_do_optstr_br_attr(&attr, perm);
23034 + memcpy(q, attr.a, sz + 1);
23037 + AuDebugOn(strlen(str->a) >= sizeof(str->a));
23040 +/* ---------------------------------------------------------------------- */
23042 +static match_table_t udbalevel = {
23043 + {AuOpt_UDBA_REVAL, "reval"},
23044 + {AuOpt_UDBA_NONE, "none"},
23045 +#ifdef CONFIG_AUFS_HNOTIFY
23046 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
23047 +#ifdef CONFIG_AUFS_HFSNOTIFY
23048 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
23054 +static int noinline_for_stack udba_val(char *str)
23056 + substring_t args[MAX_OPT_ARGS];
23058 + return match_token(str, udbalevel, args);
23061 +const char *au_optstr_udba(int udba)
23063 + return au_parser_pattern(udba, udbalevel);
23066 +/* ---------------------------------------------------------------------- */
23068 +static match_table_t au_wbr_create_policy = {
23069 + {AuWbrCreate_TDP, "tdp"},
23070 + {AuWbrCreate_TDP, "top-down-parent"},
23071 + {AuWbrCreate_RR, "rr"},
23072 + {AuWbrCreate_RR, "round-robin"},
23073 + {AuWbrCreate_MFS, "mfs"},
23074 + {AuWbrCreate_MFS, "most-free-space"},
23075 + {AuWbrCreate_MFSV, "mfs:%d"},
23076 + {AuWbrCreate_MFSV, "most-free-space:%d"},
23078 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
23079 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
23080 + {AuWbrCreate_PMFS, "pmfs"},
23081 + {AuWbrCreate_PMFSV, "pmfs:%d"},
23082 + {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
23083 + {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
23089 + * cf. linux/lib/parser.c and cmdline.c
23090 + * gave up calling memparse() since it uses simple_strtoull() instead of
23093 +static int noinline_for_stack
23094 +au_match_ull(substring_t *s, unsigned long long *result)
23097 + unsigned int len;
23101 + len = s->to - s->from;
23102 + if (len + 1 <= sizeof(a)) {
23103 + memcpy(a, s->from, len);
23105 + err = kstrtoull(a, 0, result);
23110 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
23111 + struct au_opt_wbr_create *create)
23114 + unsigned long long ull;
23117 + if (!au_match_ull(arg, &ull))
23118 + create->mfsrr_watermark = ull;
23120 + pr_err("bad integer in %s\n", str);
23127 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
23128 + struct au_opt_wbr_create *create)
23133 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
23134 + create->mfs_second = n;
23136 + pr_err("bad integer in %s\n", str);
23143 +static int noinline_for_stack
23144 +au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
23147 + substring_t args[MAX_OPT_ARGS];
23149 + err = match_token(str, au_wbr_create_policy, args);
23150 + create->wbr_create = err;
23152 + case AuWbrCreate_MFSRRV:
23153 + case AuWbrCreate_PMFSRRV:
23154 + e = au_wbr_mfs_wmark(&args[0], str, create);
23156 + e = au_wbr_mfs_sec(&args[1], str, create);
23160 + case AuWbrCreate_MFSRR:
23161 + case AuWbrCreate_PMFSRR:
23162 + e = au_wbr_mfs_wmark(&args[0], str, create);
23163 + if (unlikely(e)) {
23168 + case AuWbrCreate_MFS:
23169 + case AuWbrCreate_PMFS:
23170 + create->mfs_second = AUFS_MFS_DEF_SEC;
23172 + case AuWbrCreate_MFSV:
23173 + case AuWbrCreate_PMFSV:
23174 + e = au_wbr_mfs_sec(&args[0], str, create);
23183 +const char *au_optstr_wbr_create(int wbr_create)
23185 + return au_parser_pattern(wbr_create, au_wbr_create_policy);
23188 +static match_table_t au_wbr_copyup_policy = {
23189 + {AuWbrCopyup_TDP, "tdp"},
23190 + {AuWbrCopyup_TDP, "top-down-parent"},
23191 + {AuWbrCopyup_BUP, "bup"},
23192 + {AuWbrCopyup_BUP, "bottom-up-parent"},
23193 + {AuWbrCopyup_BU, "bu"},
23194 + {AuWbrCopyup_BU, "bottom-up"},
23198 +static int noinline_for_stack au_wbr_copyup_val(char *str)
23200 + substring_t args[MAX_OPT_ARGS];
23202 + return match_token(str, au_wbr_copyup_policy, args);
23205 +const char *au_optstr_wbr_copyup(int wbr_copyup)
23207 + return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
23210 +/* ---------------------------------------------------------------------- */
23212 +static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
23214 +static void dump_opts(struct au_opts *opts)
23216 +#ifdef CONFIG_AUFS_DEBUG
23217 + /* reduce stack space */
23219 + struct au_opt_add *add;
23220 + struct au_opt_del *del;
23221 + struct au_opt_mod *mod;
23222 + struct au_opt_xino *xino;
23223 + struct au_opt_xino_itrunc *xino_itrunc;
23224 + struct au_opt_wbr_create *create;
23226 + struct au_opt *opt;
23229 + while (opt->type != Opt_tail) {
23230 + switch (opt->type) {
23232 + u.add = &opt->add;
23233 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
23234 + u.add->bindex, u.add->pathname, u.add->perm,
23235 + u.add->path.dentry);
23239 + u.del = &opt->del;
23240 + AuDbg("del {%s, %p}\n",
23241 + u.del->pathname, u.del->h_path.dentry);
23245 + u.mod = &opt->mod;
23246 + AuDbg("mod {%s, 0x%x, %p}\n",
23247 + u.mod->path, u.mod->perm, u.mod->h_root);
23250 + u.add = &opt->add;
23251 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
23252 + u.add->bindex, u.add->pathname, u.add->perm,
23253 + u.add->path.dentry);
23255 + case Opt_prepend:
23256 + u.add = &opt->add;
23257 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
23258 + u.add->bindex, u.add->pathname, u.add->perm,
23259 + u.add->path.dentry);
23262 + AuDbg("dirwh %d\n", opt->dirwh);
23264 + case Opt_rdcache:
23265 + AuDbg("rdcache %d\n", opt->rdcache);
23268 + AuDbg("rdblk %u\n", opt->rdblk);
23270 + case Opt_rdblk_def:
23271 + AuDbg("rdblk_def\n");
23274 + AuDbg("rdhash %u\n", opt->rdhash);
23276 + case Opt_rdhash_def:
23277 + AuDbg("rdhash_def\n");
23280 + u.xino = &opt->xino;
23281 + AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
23283 + case Opt_trunc_xino:
23284 + AuLabel(trunc_xino);
23286 + case Opt_notrunc_xino:
23287 + AuLabel(notrunc_xino);
23289 + case Opt_trunc_xino_path:
23290 + case Opt_itrunc_xino:
23291 + u.xino_itrunc = &opt->xino_itrunc;
23292 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
23297 + case Opt_trunc_xib:
23298 + AuLabel(trunc_xib);
23300 + case Opt_notrunc_xib:
23301 + AuLabel(notrunc_xib);
23309 + case Opt_dirperm1:
23310 + AuLabel(dirperm1);
23312 + case Opt_nodirperm1:
23313 + AuLabel(nodirperm1);
23318 + case Opt_noplink:
23319 + AuLabel(noplink);
23321 + case Opt_list_plink:
23322 + AuLabel(list_plink);
23325 + AuDbg("udba %d, %s\n",
23326 + opt->udba, au_optstr_udba(opt->udba));
23334 + case Opt_diropq_a:
23335 + AuLabel(diropq_a);
23337 + case Opt_diropq_w:
23338 + AuLabel(diropq_w);
23340 + case Opt_warn_perm:
23341 + AuLabel(warn_perm);
23343 + case Opt_nowarn_perm:
23344 + AuLabel(nowarn_perm);
23346 + case Opt_verbose:
23347 + AuLabel(verbose);
23349 + case Opt_noverbose:
23350 + AuLabel(noverbose);
23361 + case Opt_wbr_create:
23362 + u.create = &opt->wbr_create;
23363 + AuDbg("create %d, %s\n", u.create->wbr_create,
23364 + au_optstr_wbr_create(u.create->wbr_create));
23365 + switch (u.create->wbr_create) {
23366 + case AuWbrCreate_MFSV:
23367 + case AuWbrCreate_PMFSV:
23368 + AuDbg("%d sec\n", u.create->mfs_second);
23370 + case AuWbrCreate_MFSRR:
23371 + AuDbg("%llu watermark\n",
23372 + u.create->mfsrr_watermark);
23374 + case AuWbrCreate_MFSRRV:
23375 + case AuWbrCreate_PMFSRRV:
23376 + AuDbg("%llu watermark, %d sec\n",
23377 + u.create->mfsrr_watermark,
23378 + u.create->mfs_second);
23382 + case Opt_wbr_copyup:
23383 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
23384 + au_optstr_wbr_copyup(opt->wbr_copyup));
23386 + case Opt_fhsm_sec:
23387 + AuDbg("fhsm_sec %u\n", opt->fhsm_second);
23403 +void au_opts_free(struct au_opts *opts)
23405 + struct au_opt *opt;
23408 + while (opt->type != Opt_tail) {
23409 + switch (opt->type) {
23412 + case Opt_prepend:
23413 + path_put(&opt->add.path);
23417 + path_put(&opt->del.h_path);
23421 + dput(opt->mod.h_root);
23424 + fput(opt->xino.file);
23431 +static int opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
23432 + aufs_bindex_t bindex)
23435 + struct au_opt_add *add = &opt->add;
23438 + add->bindex = bindex;
23439 + add->perm = AuBrPerm_RO;
23440 + add->pathname = opt_str;
23441 + p = strchr(opt_str, '=');
23445 + add->perm = br_perm_val(p);
23448 + err = vfsub_kern_path(add->pathname, lkup_dirflags, &add->path);
23451 + add->perm = AuBrPerm_RO;
23452 + if (au_test_fs_rr(add->path.dentry->d_sb))
23453 + add->perm = AuBrPerm_RR;
23454 + else if (!bindex && !(sb_flags & MS_RDONLY))
23455 + add->perm = AuBrPerm_RW;
23457 + opt->type = Opt_add;
23460 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
23467 +static int au_opts_parse_del(struct au_opt_del *del, substring_t args[])
23471 + del->pathname = args[0].from;
23472 + AuDbg("del path %s\n", del->pathname);
23474 + err = vfsub_kern_path(del->pathname, lkup_dirflags, &del->h_path);
23475 + if (unlikely(err))
23476 + pr_err("lookup failed %s (%d)\n", del->pathname, err);
23481 +#if 0 /* reserved for future use */
23482 +static int au_opts_parse_idel(struct super_block *sb, aufs_bindex_t bindex,
23483 + struct au_opt_del *del, substring_t args[])
23486 + struct dentry *root;
23489 + root = sb->s_root;
23490 + aufs_read_lock(root, AuLock_FLUSH);
23491 + if (bindex < 0 || au_sbend(sb) < bindex) {
23492 + pr_err("out of bounds, %d\n", bindex);
23497 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
23498 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
23501 + aufs_read_unlock(root, !AuLock_IR);
23506 +static int noinline_for_stack
23507 +au_opts_parse_mod(struct au_opt_mod *mod, substring_t args[])
23510 + struct path path;
23514 + mod->path = args[0].from;
23515 + p = strchr(mod->path, '=');
23516 + if (unlikely(!p)) {
23517 + pr_err("no permssion %s\n", args[0].from);
23522 + err = vfsub_kern_path(mod->path, lkup_dirflags, &path);
23523 + if (unlikely(err)) {
23524 + pr_err("lookup failed %s (%d)\n", mod->path, err);
23528 + mod->perm = br_perm_val(p);
23529 + AuDbg("mod path %s, perm 0x%x, %s\n", mod->path, mod->perm, p);
23530 + mod->h_root = dget(path.dentry);
23537 +#if 0 /* reserved for future use */
23538 +static int au_opts_parse_imod(struct super_block *sb, aufs_bindex_t bindex,
23539 + struct au_opt_mod *mod, substring_t args[])
23542 + struct dentry *root;
23545 + root = sb->s_root;
23546 + aufs_read_lock(root, AuLock_FLUSH);
23547 + if (bindex < 0 || au_sbend(sb) < bindex) {
23548 + pr_err("out of bounds, %d\n", bindex);
23553 + mod->perm = br_perm_val(args[1].from);
23554 + AuDbg("mod path %s, perm 0x%x, %s\n",
23555 + mod->path, mod->perm, args[1].from);
23556 + mod->h_root = dget(au_h_dptr(root, bindex));
23559 + aufs_read_unlock(root, !AuLock_IR);
23564 +static int au_opts_parse_xino(struct super_block *sb, struct au_opt_xino *xino,
23565 + substring_t args[])
23568 + struct file *file;
23570 + file = au_xino_create(sb, args[0].from, /*silent*/0);
23571 + err = PTR_ERR(file);
23572 + if (IS_ERR(file))
23576 + if (unlikely(file->f_path.dentry->d_sb == sb)) {
23578 + pr_err("%s must be outside\n", args[0].from);
23583 + xino->file = file;
23584 + xino->path = args[0].from;
23590 +static int noinline_for_stack
23591 +au_opts_parse_xino_itrunc_path(struct super_block *sb,
23592 + struct au_opt_xino_itrunc *xino_itrunc,
23593 + substring_t args[])
23596 + aufs_bindex_t bend, bindex;
23597 + struct path path;
23598 + struct dentry *root;
23600 + err = vfsub_kern_path(args[0].from, lkup_dirflags, &path);
23601 + if (unlikely(err)) {
23602 + pr_err("lookup failed %s (%d)\n", args[0].from, err);
23606 + xino_itrunc->bindex = -1;
23607 + root = sb->s_root;
23608 + aufs_read_lock(root, AuLock_FLUSH);
23609 + bend = au_sbend(sb);
23610 + for (bindex = 0; bindex <= bend; bindex++) {
23611 + if (au_h_dptr(root, bindex) == path.dentry) {
23612 + xino_itrunc->bindex = bindex;
23616 + aufs_read_unlock(root, !AuLock_IR);
23619 + if (unlikely(xino_itrunc->bindex < 0)) {
23620 + pr_err("no such branch %s\n", args[0].from);
23628 +/* called without aufs lock */
23629 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts)
23631 + int err, n, token;
23632 + aufs_bindex_t bindex;
23633 + unsigned char skipped;
23634 + struct dentry *root;
23635 + struct au_opt *opt, *opt_tail;
23637 + /* reduce the stack space */
23639 + struct au_opt_xino_itrunc *xino_itrunc;
23640 + struct au_opt_wbr_create *create;
23643 + substring_t args[MAX_OPT_ARGS];
23647 + a = kmalloc(sizeof(*a), GFP_NOFS);
23648 + if (unlikely(!a))
23651 + root = sb->s_root;
23655 + opt_tail = opt + opts->max_opt - 1;
23656 + opt->type = Opt_tail;
23657 + while (!err && (opt_str = strsep(&str, ",")) && *opt_str) {
23660 + token = match_token(opt_str, options, a->args);
23664 + while (!err && (opt_str = strsep(&a->args[0].from, ":"))
23666 + err = opt_add(opt, opt_str, opts->sb_flags,
23668 + if (unlikely(!err && ++opt > opt_tail)) {
23672 + opt->type = Opt_tail;
23677 + if (unlikely(match_int(&a->args[0], &n))) {
23678 + pr_err("bad integer in %s\n", opt_str);
23682 + err = opt_add(opt, a->args[1].from, opts->sb_flags,
23685 + opt->type = token;
23688 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
23689 + /*dummy bindex*/1);
23691 + opt->type = token;
23693 + case Opt_prepend:
23694 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
23697 + opt->type = token;
23700 + err = au_opts_parse_del(&opt->del, a->args);
23702 + opt->type = token;
23704 +#if 0 /* reserved for future use */
23706 + del->pathname = "(indexed)";
23707 + if (unlikely(match_int(&args[0], &n))) {
23708 + pr_err("bad integer in %s\n", opt_str);
23711 + err = au_opts_parse_idel(sb, n, &opt->del, a->args);
23713 + opt->type = token;
23717 + err = au_opts_parse_mod(&opt->mod, a->args);
23719 + opt->type = token;
23721 +#ifdef IMOD /* reserved for future use */
23723 + u.mod->path = "(indexed)";
23724 + if (unlikely(match_int(&a->args[0], &n))) {
23725 + pr_err("bad integer in %s\n", opt_str);
23728 + err = au_opts_parse_imod(sb, n, &opt->mod, a->args);
23730 + opt->type = token;
23734 + err = au_opts_parse_xino(sb, &opt->xino, a->args);
23736 + opt->type = token;
23739 + case Opt_trunc_xino_path:
23740 + err = au_opts_parse_xino_itrunc_path
23741 + (sb, &opt->xino_itrunc, a->args);
23743 + opt->type = token;
23746 + case Opt_itrunc_xino:
23747 + u.xino_itrunc = &opt->xino_itrunc;
23748 + if (unlikely(match_int(&a->args[0], &n))) {
23749 + pr_err("bad integer in %s\n", opt_str);
23752 + u.xino_itrunc->bindex = n;
23753 + aufs_read_lock(root, AuLock_FLUSH);
23754 + if (n < 0 || au_sbend(sb) < n) {
23755 + pr_err("out of bounds, %d\n", n);
23756 + aufs_read_unlock(root, !AuLock_IR);
23759 + aufs_read_unlock(root, !AuLock_IR);
23761 + opt->type = token;
23765 + if (unlikely(match_int(&a->args[0], &opt->dirwh)))
23768 + opt->type = token;
23771 + case Opt_rdcache:
23772 + if (unlikely(match_int(&a->args[0], &n))) {
23773 + pr_err("bad integer in %s\n", opt_str);
23776 + if (unlikely(n > AUFS_RDCACHE_MAX)) {
23777 + pr_err("rdcache must be smaller than %d\n",
23778 + AUFS_RDCACHE_MAX);
23781 + opt->rdcache = n;
23783 + opt->type = token;
23786 + if (unlikely(match_int(&a->args[0], &n)
23788 + || n > KMALLOC_MAX_SIZE)) {
23789 + pr_err("bad integer in %s\n", opt_str);
23792 + if (unlikely(n && n < NAME_MAX)) {
23793 + pr_err("rdblk must be larger than %d\n",
23799 + opt->type = token;
23802 + if (unlikely(match_int(&a->args[0], &n)
23804 + || n * sizeof(struct hlist_head)
23805 + > KMALLOC_MAX_SIZE)) {
23806 + pr_err("bad integer in %s\n", opt_str);
23811 + opt->type = token;
23814 + case Opt_trunc_xino:
23815 + case Opt_notrunc_xino:
23817 + case Opt_trunc_xib:
23818 + case Opt_notrunc_xib:
23821 + case Opt_dirperm1:
23822 + case Opt_nodirperm1:
23824 + case Opt_noplink:
23825 + case Opt_list_plink:
23828 + case Opt_diropq_a:
23829 + case Opt_diropq_w:
23830 + case Opt_warn_perm:
23831 + case Opt_nowarn_perm:
23832 + case Opt_verbose:
23833 + case Opt_noverbose:
23837 + case Opt_rdblk_def:
23838 + case Opt_rdhash_def:
23842 + opt->type = token;
23846 + opt->udba = udba_val(a->args[0].from);
23847 + if (opt->udba >= 0) {
23849 + opt->type = token;
23851 + pr_err("wrong value, %s\n", opt_str);
23854 + case Opt_wbr_create:
23855 + u.create = &opt->wbr_create;
23856 + u.create->wbr_create
23857 + = au_wbr_create_val(a->args[0].from, u.create);
23858 + if (u.create->wbr_create >= 0) {
23860 + opt->type = token;
23862 + pr_err("wrong value, %s\n", opt_str);
23864 + case Opt_wbr_copyup:
23865 + opt->wbr_copyup = au_wbr_copyup_val(a->args[0].from);
23866 + if (opt->wbr_copyup >= 0) {
23868 + opt->type = token;
23870 + pr_err("wrong value, %s\n", opt_str);
23873 + case Opt_fhsm_sec:
23874 + if (unlikely(match_int(&a->args[0], &n)
23876 + pr_err("bad integer in %s\n", opt_str);
23879 + if (sysaufs_brs) {
23880 + opt->fhsm_second = n;
23881 + opt->type = token;
23883 + pr_warn("ignored %s\n", opt_str);
23888 + pr_warn("ignored %s\n", opt_str);
23890 + case Opt_ignore_silent:
23895 + pr_err("unknown option %s\n", opt_str);
23899 + if (!err && !skipped) {
23900 + if (unlikely(++opt > opt_tail)) {
23903 + opt->type = Opt_tail;
23906 + opt->type = Opt_tail;
23912 + if (unlikely(err))
23913 + au_opts_free(opts);
23919 +static int au_opt_wbr_create(struct super_block *sb,
23920 + struct au_opt_wbr_create *create)
23923 + struct au_sbinfo *sbinfo;
23925 + SiMustWriteLock(sb);
23927 + err = 1; /* handled */
23928 + sbinfo = au_sbi(sb);
23929 + if (sbinfo->si_wbr_create_ops->fin) {
23930 + err = sbinfo->si_wbr_create_ops->fin(sb);
23935 + sbinfo->si_wbr_create = create->wbr_create;
23936 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
23937 + switch (create->wbr_create) {
23938 + case AuWbrCreate_MFSRRV:
23939 + case AuWbrCreate_MFSRR:
23940 + case AuWbrCreate_PMFSRR:
23941 + case AuWbrCreate_PMFSRRV:
23942 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
23944 + case AuWbrCreate_MFS:
23945 + case AuWbrCreate_MFSV:
23946 + case AuWbrCreate_PMFS:
23947 + case AuWbrCreate_PMFSV:
23948 + sbinfo->si_wbr_mfs.mfs_expire
23949 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
23953 + if (sbinfo->si_wbr_create_ops->init)
23954 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
23961 + * plus: processed without an error
23962 + * zero: unprocessed
23964 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
23965 + struct au_opts *opts)
23968 + struct au_sbinfo *sbinfo;
23970 + SiMustWriteLock(sb);
23972 + err = 1; /* handled */
23973 + sbinfo = au_sbi(sb);
23974 + switch (opt->type) {
23976 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
23977 + sbinfo->si_mntflags |= opt->udba;
23978 + opts->given_udba |= opt->udba;
23982 + au_opt_set(sbinfo->si_mntflags, PLINK);
23984 + case Opt_noplink:
23985 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
23986 + au_plink_put(sb, /*verbose*/1);
23987 + au_opt_clr(sbinfo->si_mntflags, PLINK);
23989 + case Opt_list_plink:
23990 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
23991 + au_plink_list(sb);
23995 + au_opt_set(sbinfo->si_mntflags, DIO);
23996 + au_fset_opts(opts->flags, REFRESH_DYAOP);
23999 + au_opt_clr(sbinfo->si_mntflags, DIO);
24000 + au_fset_opts(opts->flags, REFRESH_DYAOP);
24003 + case Opt_fhsm_sec:
24004 + au_fhsm_set(sbinfo, opt->fhsm_second);
24007 + case Opt_diropq_a:
24008 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
24010 + case Opt_diropq_w:
24011 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
24014 + case Opt_warn_perm:
24015 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
24017 + case Opt_nowarn_perm:
24018 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
24021 + case Opt_verbose:
24022 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
24024 + case Opt_noverbose:
24025 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
24029 + au_opt_set(sbinfo->si_mntflags, SUM);
24032 + au_opt_clr(sbinfo->si_mntflags, SUM);
24033 + au_opt_set(sbinfo->si_mntflags, SUM_W);
24035 + au_opt_clr(sbinfo->si_mntflags, SUM);
24036 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
24039 + case Opt_wbr_create:
24040 + err = au_opt_wbr_create(sb, &opt->wbr_create);
24042 + case Opt_wbr_copyup:
24043 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
24044 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
24048 + sbinfo->si_dirwh = opt->dirwh;
24051 + case Opt_rdcache:
24052 + sbinfo->si_rdcache
24053 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
24056 + sbinfo->si_rdblk = opt->rdblk;
24058 + case Opt_rdblk_def:
24059 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
24062 + sbinfo->si_rdhash = opt->rdhash;
24064 + case Opt_rdhash_def:
24065 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
24069 + au_opt_set(sbinfo->si_mntflags, SHWH);
24072 + au_opt_clr(sbinfo->si_mntflags, SHWH);
24075 + case Opt_dirperm1:
24076 + au_opt_set(sbinfo->si_mntflags, DIRPERM1);
24078 + case Opt_nodirperm1:
24079 + au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
24082 + case Opt_trunc_xino:
24083 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
24085 + case Opt_notrunc_xino:
24086 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
24089 + case Opt_trunc_xino_path:
24090 + case Opt_itrunc_xino:
24091 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex);
24096 + case Opt_trunc_xib:
24097 + au_fset_opts(opts->flags, TRUNC_XIB);
24099 + case Opt_notrunc_xib:
24100 + au_fclr_opts(opts->flags, TRUNC_XIB);
24104 + sb->s_flags |= MS_POSIXACL;
24107 + sb->s_flags &= ~MS_POSIXACL;
24119 + * returns tri-state.
24120 + * plus: processed without an error
24121 + * zero: unprocessed
24124 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
24125 + struct au_opts *opts)
24127 + int err, do_refresh;
24130 + switch (opt->type) {
24132 + opt->add.bindex = au_sbend(sb) + 1;
24133 + if (opt->add.bindex < 0)
24134 + opt->add.bindex = 0;
24136 + case Opt_prepend:
24137 + opt->add.bindex = 0;
24138 + add: /* indented label */
24140 + err = au_br_add(sb, &opt->add,
24141 + au_ftest_opts(opts->flags, REMOUNT));
24144 + au_fset_opts(opts->flags, REFRESH);
24150 + err = au_br_del(sb, &opt->del,
24151 + au_ftest_opts(opts->flags, REMOUNT));
24154 + au_fset_opts(opts->flags, TRUNC_XIB);
24155 + au_fset_opts(opts->flags, REFRESH);
24161 + err = au_br_mod(sb, &opt->mod,
24162 + au_ftest_opts(opts->flags, REMOUNT),
24167 + au_fset_opts(opts->flags, REFRESH);
24175 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
24176 + struct au_opt_xino **opt_xino,
24177 + struct au_opts *opts)
24180 + aufs_bindex_t bend, bindex;
24181 + struct dentry *root, *parent, *h_root;
24184 + switch (opt->type) {
24186 + err = au_xino_set(sb, &opt->xino,
24187 + !!au_ftest_opts(opts->flags, REMOUNT));
24188 + if (unlikely(err))
24191 + *opt_xino = &opt->xino;
24192 + au_xino_brid_set(sb, -1);
24194 + /* safe d_parent access */
24195 + parent = opt->xino.file->f_path.dentry->d_parent;
24196 + root = sb->s_root;
24197 + bend = au_sbend(sb);
24198 + for (bindex = 0; bindex <= bend; bindex++) {
24199 + h_root = au_h_dptr(root, bindex);
24200 + if (h_root == parent) {
24201 + au_xino_brid_set(sb, au_sbr_id(sb, bindex));
24209 + au_xino_brid_set(sb, -1);
24210 + *opt_xino = (void *)-1;
24217 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
24218 + unsigned int pending)
24221 + aufs_bindex_t bindex, bend;
24222 + unsigned char do_plink, skip, do_free, can_no_dreval;
24223 + struct au_branch *br;
24224 + struct au_wbr *wbr;
24225 + struct dentry *root, *dentry;
24226 + struct inode *dir, *h_dir;
24227 + struct au_sbinfo *sbinfo;
24228 + struct au_hinode *hdir;
24230 + SiMustAnyLock(sb);
24232 + sbinfo = au_sbi(sb);
24233 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
24235 + if (!(sb_flags & MS_RDONLY)) {
24236 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
24237 + pr_warn("first branch should be rw\n");
24238 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
24239 + pr_warn("shwh should be used with ro\n");
24242 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
24243 + && !au_opt_test(sbinfo->si_mntflags, XINO))
24244 + pr_warn("udba=*notify requires xino\n");
24246 + if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
24247 + pr_warn("dirperm1 breaks the protection"
24248 + " by the permission bits on the lower branch\n");
24252 + root = sb->s_root;
24253 + dir = d_inode(root);
24254 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
24255 + can_no_dreval = !!au_opt_test((sbinfo->si_mntflags | pending),
24257 + bend = au_sbend(sb);
24258 + for (bindex = 0; !err && bindex <= bend; bindex++) {
24260 + h_dir = au_h_iptr(dir, bindex);
24261 + br = au_sbr(sb, bindex);
24263 + if ((br->br_perm & AuBrAttr_ICEX)
24264 + && !h_dir->i_op->listxattr)
24265 + br->br_perm &= ~AuBrAttr_ICEX;
24267 + if ((br->br_perm & AuBrAttr_ICEX_SEC)
24268 + && (au_br_sb(br)->s_flags & MS_NOSEC))
24269 + br->br_perm &= ~AuBrAttr_ICEX_SEC;
24273 + wbr = br->br_wbr;
24275 + wbr_wh_read_lock(wbr);
24277 + if (!au_br_writable(br->br_perm)) {
24280 + || (!wbr->wbr_whbase
24281 + && !wbr->wbr_plink
24282 + && !wbr->wbr_orph));
24283 + } else if (!au_br_wh_linkable(br->br_perm)) {
24284 + /* skip = (!br->br_whbase && !br->br_orph); */
24285 + skip = (!wbr || !wbr->wbr_whbase);
24286 + if (skip && wbr) {
24288 + skip = !!wbr->wbr_plink;
24290 + skip = !wbr->wbr_plink;
24293 + /* skip = (br->br_whbase && br->br_ohph); */
24294 + skip = (wbr && wbr->wbr_whbase);
24297 + skip = !!wbr->wbr_plink;
24299 + skip = !wbr->wbr_plink;
24303 + wbr_wh_read_unlock(wbr);
24305 + if (can_no_dreval) {
24306 + dentry = br->br_path.dentry;
24307 + spin_lock(&dentry->d_lock);
24308 + if (dentry->d_flags &
24309 + (DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE))
24310 + can_no_dreval = 0;
24311 + spin_unlock(&dentry->d_lock);
24314 + if (au_br_fhsm(br->br_perm)) {
24316 + AuDebugOn(!br->br_fhsm);
24322 + hdir = au_hi(dir, bindex);
24323 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
24325 + wbr_wh_write_lock(wbr);
24326 + err = au_wh_init(br, sb);
24328 + wbr_wh_write_unlock(wbr);
24329 + au_hn_imtx_unlock(hdir);
24331 + if (!err && do_free) {
24333 + br->br_wbr = NULL;
24337 + if (can_no_dreval)
24338 + au_fset_si(sbinfo, NO_DREVAL);
24340 + au_fclr_si(sbinfo, NO_DREVAL);
24343 + au_fset_si(sbinfo, FHSM);
24344 + for (bindex = bend; bindex >= 0; bindex--) {
24345 + br = au_sbr(sb, bindex);
24346 + if (au_br_fhsm(br->br_perm)) {
24347 + au_fhsm_set_bottom(sb, bindex);
24352 + au_fclr_si(sbinfo, FHSM);
24353 + au_fhsm_set_bottom(sb, -1);
24359 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
24362 + unsigned int tmp;
24363 + aufs_bindex_t bindex, bend;
24364 + struct au_opt *opt;
24365 + struct au_opt_xino *opt_xino, xino;
24366 + struct au_sbinfo *sbinfo;
24367 + struct au_branch *br;
24368 + struct inode *dir;
24370 + SiMustWriteLock(sb);
24375 + while (err >= 0 && opt->type != Opt_tail)
24376 + err = au_opt_simple(sb, opt++, opts);
24379 + else if (unlikely(err < 0))
24382 + /* disable xino and udba temporary */
24383 + sbinfo = au_sbi(sb);
24384 + tmp = sbinfo->si_mntflags;
24385 + au_opt_clr(sbinfo->si_mntflags, XINO);
24386 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
24389 + while (err >= 0 && opt->type != Opt_tail)
24390 + err = au_opt_br(sb, opt++, opts);
24393 + else if (unlikely(err < 0))
24396 + bend = au_sbend(sb);
24397 + if (unlikely(bend < 0)) {
24399 + pr_err("no branches\n");
24403 + if (au_opt_test(tmp, XINO))
24404 + au_opt_set(sbinfo->si_mntflags, XINO);
24406 + while (!err && opt->type != Opt_tail)
24407 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
24408 + if (unlikely(err))
24411 + err = au_opts_verify(sb, sb->s_flags, tmp);
24412 + if (unlikely(err))
24415 + /* restore xino */
24416 + if (au_opt_test(tmp, XINO) && !opt_xino) {
24417 + xino.file = au_xino_def(sb);
24418 + err = PTR_ERR(xino.file);
24419 + if (IS_ERR(xino.file))
24422 + err = au_xino_set(sb, &xino, /*remount*/0);
24424 + if (unlikely(err))
24428 + /* restore udba */
24429 + tmp &= AuOptMask_UDBA;
24430 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
24431 + sbinfo->si_mntflags |= tmp;
24432 + bend = au_sbend(sb);
24433 + for (bindex = 0; bindex <= bend; bindex++) {
24434 + br = au_sbr(sb, bindex);
24435 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
24436 + if (unlikely(err))
24437 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
24439 + /* go on even if err */
24441 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
24442 + dir = d_inode(sb->s_root);
24443 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
24450 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
24453 + unsigned char no_dreval;
24454 + struct inode *dir;
24455 + struct au_opt_xino *opt_xino;
24456 + struct au_opt *opt;
24457 + struct au_sbinfo *sbinfo;
24459 + SiMustWriteLock(sb);
24462 + dir = d_inode(sb->s_root);
24463 + sbinfo = au_sbi(sb);
24466 + while (err >= 0 && opt->type != Opt_tail) {
24467 + err = au_opt_simple(sb, opt, opts);
24469 + err = au_opt_br(sb, opt, opts);
24471 + err = au_opt_xino(sb, opt, &opt_xino, opts);
24477 + /* go on even err */
24479 + no_dreval = !!au_ftest_si(sbinfo, NO_DREVAL);
24480 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
24481 + if (unlikely(rerr && !err))
24484 + if (no_dreval != !!au_ftest_si(sbinfo, NO_DREVAL))
24485 + au_fset_opts(opts->flags, REFRESH_DOP);
24487 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
24488 + rerr = au_xib_trunc(sb);
24489 + if (unlikely(rerr && !err))
24493 + /* will be handled by the caller */
24494 + if (!au_ftest_opts(opts->flags, REFRESH)
24495 + && (opts->given_udba
24496 + || au_opt_test(sbinfo->si_mntflags, XINO)
24497 + || au_ftest_opts(opts->flags, REFRESH_DOP)
24499 + au_fset_opts(opts->flags, REFRESH);
24501 + AuDbg("status 0x%x\n", opts->flags);
24505 +/* ---------------------------------------------------------------------- */
24507 +unsigned int au_opt_udba(struct super_block *sb)
24509 + return au_mntflags(sb) & AuOptMask_UDBA;
24511 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
24512 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
24513 +++ linux/fs/aufs/opts.h 2015-11-11 17:21:46.918863802 +0100
24516 + * Copyright (C) 2005-2015 Junjiro R. Okajima
24518 + * This program, aufs is free software; you can redistribute it and/or modify
24519 + * it under the terms of the GNU General Public License as published by
24520 + * the Free Software Foundation; either version 2 of the License, or
24521 + * (at your option) any later version.
24523 + * This program is distributed in the hope that it will be useful,
24524 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24525 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24526 + * GNU General Public License for more details.
24528 + * You should have received a copy of the GNU General Public License
24529 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24533 + * mount options/flags
24536 +#ifndef __AUFS_OPTS_H__
24537 +#define __AUFS_OPTS_H__
24541 +#include <linux/path.h>
24544 +struct super_block;
24546 +/* ---------------------------------------------------------------------- */
24549 +#define AuOpt_XINO 1 /* external inode number bitmap
24550 + and translation table */
24551 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
24552 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
24553 +#define AuOpt_UDBA_REVAL (1 << 3)
24554 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
24555 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
24556 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
24557 +#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
24559 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
24560 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
24561 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
24562 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
24563 +#define AuOpt_VERBOSE (1 << 13) /* busy inode when del-branch */
24564 +#define AuOpt_DIO (1 << 14) /* direct io */
24566 +#ifndef CONFIG_AUFS_HNOTIFY
24567 +#undef AuOpt_UDBA_HNOTIFY
24568 +#define AuOpt_UDBA_HNOTIFY 0
24570 +#ifndef CONFIG_AUFS_SHWH
24572 +#define AuOpt_SHWH 0
24575 +#define AuOpt_Def (AuOpt_XINO \
24576 + | AuOpt_UDBA_REVAL \
24578 + /* | AuOpt_DIRPERM1 */ \
24579 + | AuOpt_WARN_PERM)
24580 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
24581 + | AuOpt_UDBA_REVAL \
24582 + | AuOpt_UDBA_HNOTIFY)
24584 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
24585 +#define au_opt_set(flags, name) do { \
24586 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
24587 + ((flags) |= AuOpt_##name); \
24589 +#define au_opt_set_udba(flags, name) do { \
24590 + (flags) &= ~AuOptMask_UDBA; \
24591 + ((flags) |= AuOpt_##name); \
24593 +#define au_opt_clr(flags, name) do { \
24594 + ((flags) &= ~AuOpt_##name); \
24597 +static inline unsigned int au_opts_plink(unsigned int mntflags)
24599 +#ifdef CONFIG_PROC_FS
24602 + return mntflags & ~AuOpt_PLINK;
24606 +/* ---------------------------------------------------------------------- */
24608 +/* policies to select one among multiple writable branches */
24610 + AuWbrCreate_TDP, /* top down parent */
24611 + AuWbrCreate_RR, /* round robin */
24612 + AuWbrCreate_MFS, /* most free space */
24613 + AuWbrCreate_MFSV, /* mfs with seconds */
24614 + AuWbrCreate_MFSRR, /* mfs then rr */
24615 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
24616 + AuWbrCreate_PMFS, /* parent and mfs */
24617 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
24618 + AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
24619 + AuWbrCreate_PMFSRRV, /* plus seconds */
24621 + AuWbrCreate_Def = AuWbrCreate_TDP
24625 + AuWbrCopyup_TDP, /* top down parent */
24626 + AuWbrCopyup_BUP, /* bottom up parent */
24627 + AuWbrCopyup_BU, /* bottom up */
24629 + AuWbrCopyup_Def = AuWbrCopyup_TDP
24632 +/* ---------------------------------------------------------------------- */
24634 +struct au_opt_add {
24635 + aufs_bindex_t bindex;
24638 + struct path path;
24641 +struct au_opt_del {
24643 + struct path h_path;
24646 +struct au_opt_mod {
24649 + struct dentry *h_root;
24652 +struct au_opt_xino {
24654 + struct file *file;
24657 +struct au_opt_xino_itrunc {
24658 + aufs_bindex_t bindex;
24661 +struct au_opt_wbr_create {
24664 + unsigned long long mfsrr_watermark;
24670 + struct au_opt_xino xino;
24671 + struct au_opt_xino_itrunc xino_itrunc;
24672 + struct au_opt_add add;
24673 + struct au_opt_del del;
24674 + struct au_opt_mod mod;
24677 + unsigned int rdblk;
24678 + unsigned int rdhash;
24680 + struct au_opt_wbr_create wbr_create;
24682 + unsigned int fhsm_second;
24687 +#define AuOpts_REMOUNT 1
24688 +#define AuOpts_REFRESH (1 << 1)
24689 +#define AuOpts_TRUNC_XIB (1 << 2)
24690 +#define AuOpts_REFRESH_DYAOP (1 << 3)
24691 +#define AuOpts_REFRESH_DOP (1 << 4)
24692 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
24693 +#define au_fset_opts(flags, name) \
24694 + do { (flags) |= AuOpts_##name; } while (0)
24695 +#define au_fclr_opts(flags, name) \
24696 + do { (flags) &= ~AuOpts_##name; } while (0)
24699 + struct au_opt *opt;
24702 + unsigned int given_udba;
24703 + unsigned int flags;
24704 + unsigned long sb_flags;
24707 +/* ---------------------------------------------------------------------- */
24710 +void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
24711 +const char *au_optstr_udba(int udba);
24712 +const char *au_optstr_wbr_copyup(int wbr_copyup);
24713 +const char *au_optstr_wbr_create(int wbr_create);
24715 +void au_opts_free(struct au_opts *opts);
24716 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts);
24717 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
24718 + unsigned int pending);
24719 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
24720 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
24722 +unsigned int au_opt_udba(struct super_block *sb);
24724 +#endif /* __KERNEL__ */
24725 +#endif /* __AUFS_OPTS_H__ */
24726 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
24727 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
24728 +++ linux/fs/aufs/plink.c 2015-09-24 10:47:58.254719746 +0200
24731 + * Copyright (C) 2005-2015 Junjiro R. Okajima
24733 + * This program, aufs is free software; you can redistribute it and/or modify
24734 + * it under the terms of the GNU General Public License as published by
24735 + * the Free Software Foundation; either version 2 of the License, or
24736 + * (at your option) any later version.
24738 + * This program is distributed in the hope that it will be useful,
24739 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24740 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24741 + * GNU General Public License for more details.
24743 + * You should have received a copy of the GNU General Public License
24744 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
24754 + * the pseudo-link maintenance mode.
24755 + * during a user process maintains the pseudo-links,
24756 + * prohibit adding a new plink and branch manipulation.
24760 + * For entry functions which will handle plink, and i_mutex is already held
24762 + * They cannot wait and should return an error at once.
24763 + * Callers has to check the error.
24765 + * For entry functions which will handle plink, but i_mutex is not held
24767 + * They can wait the plink maintenance mode to finish.
24769 + * They behave like F_SETLK and F_SETLKW.
24770 + * If the caller never handle plink, then both flags are unnecessary.
24773 +int au_plink_maint(struct super_block *sb, int flags)
24777 + struct au_sbinfo *sbi;
24779 + SiMustAnyLock(sb);
24782 + if (!au_opt_test(au_mntflags(sb), PLINK))
24785 + sbi = au_sbi(sb);
24786 + pid = sbi->si_plink_maint_pid;
24787 + if (!pid || pid == current->pid)
24790 + /* todo: it highly depends upon /sbin/mount.aufs */
24792 + ppid = task_pid_vnr(rcu_dereference(current->real_parent));
24793 + rcu_read_unlock();
24797 + if (au_ftest_lock(flags, NOPLMW)) {
24798 + /* if there is no i_mutex lock in VFS, we don't need to wait */
24799 + /* AuDebugOn(!lockdep_depth(current)); */
24800 + while (sbi->si_plink_maint_pid) {
24801 + si_read_unlock(sb);
24802 + /* gave up wake_up_bit() */
24803 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
24805 + if (au_ftest_lock(flags, FLUSH))
24806 + au_nwt_flush(&sbi->si_nowait);
24807 + si_noflush_read_lock(sb);
24809 + } else if (au_ftest_lock(flags, NOPLM)) {
24810 + AuDbg("ppid %d, pid %d\n", ppid, pid);
24818 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
24820 + spin_lock(&sbinfo->si_plink_maint_lock);
24821 + sbinfo->si_plink_maint_pid = 0;
24822 + spin_unlock(&sbinfo->si_plink_maint_lock);
24823 + wake_up_all(&sbinfo->si_plink_wq);
24826 +int au_plink_maint_enter(struct super_block *sb)
24829 + struct au_sbinfo *sbinfo;
24832 + sbinfo = au_sbi(sb);
24833 + /* make sure i am the only one in this fs */
24834 + si_write_lock(sb, AuLock_FLUSH);
24835 + if (au_opt_test(au_mntflags(sb), PLINK)) {
24836 + spin_lock(&sbinfo->si_plink_maint_lock);
24837 + if (!sbinfo->si_plink_maint_pid)
24838 + sbinfo->si_plink_maint_pid = current->pid;
24841 + spin_unlock(&sbinfo->si_plink_maint_lock);
24843 + si_write_unlock(sb);
24848 +/* ---------------------------------------------------------------------- */
24850 +#ifdef CONFIG_AUFS_DEBUG
24851 +void au_plink_list(struct super_block *sb)
24854 + struct au_sbinfo *sbinfo;
24855 + struct hlist_head *plink_hlist;
24856 + struct pseudo_link *plink;
24858 + SiMustAnyLock(sb);
24860 + sbinfo = au_sbi(sb);
24861 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
24862 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
24864 + for (i = 0; i < AuPlink_NHASH; i++) {
24865 + plink_hlist = &sbinfo->si_plink[i].head;
24867 + hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
24868 + AuDbg("%lu\n", plink->inode->i_ino);
24869 + rcu_read_unlock();
24874 +/* is the inode pseudo-linked? */
24875 +int au_plink_test(struct inode *inode)
24878 + struct au_sbinfo *sbinfo;
24879 + struct hlist_head *plink_hlist;
24880 + struct pseudo_link *plink;
24882 + sbinfo = au_sbi(inode->i_sb);
24883 + AuRwMustAnyLock(&sbinfo->si_rwsem);
24884 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
24885 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
24888 + i = au_plink_hash(inode->i_ino);
24889 + plink_hlist = &sbinfo->si_plink[i].head;
24891 + hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
24892 + if (plink->inode == inode) {
24896 + rcu_read_unlock();
24900 +/* ---------------------------------------------------------------------- */
24903 + * generate a name for plink.
24904 + * the file will be stored under AUFS_WH_PLINKDIR.
24906 +/* 20 is max digits length of ulong 64 */
24907 +#define PLINK_NAME_LEN ((20 + 1) * 2)
24909 +static int plink_name(char *name, int len, struct inode *inode,
24910 + aufs_bindex_t bindex)
24913 + struct inode *h_inode;
24915 + h_inode = au_h_iptr(inode, bindex);
24916 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
24920 +struct au_do_plink_lkup_args {
24921 + struct dentry **errp;
24922 + struct qstr *tgtname;
24923 + struct dentry *h_parent;
24924 + struct au_branch *br;
24927 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
24928 + struct dentry *h_parent,
24929 + struct au_branch *br)
24931 + struct dentry *h_dentry;
24932 + struct mutex *h_mtx;
24934 + h_mtx = &d_inode(h_parent)->i_mutex;
24935 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
24936 + h_dentry = vfsub_lkup_one(tgtname, h_parent);
24937 + mutex_unlock(h_mtx);
24941 +static void au_call_do_plink_lkup(void *args)
24943 + struct au_do_plink_lkup_args *a = args;
24944 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br);
24947 +/* lookup the plink-ed @inode under the branch at @bindex */
24948 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
24950 + struct dentry *h_dentry, *h_parent;
24951 + struct au_branch *br;
24953 + char a[PLINK_NAME_LEN];
24954 + struct qstr tgtname = QSTR_INIT(a, 0);
24956 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
24958 + br = au_sbr(inode->i_sb, bindex);
24959 + h_parent = br->br_wbr->wbr_plink;
24960 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
24962 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
24963 + struct au_do_plink_lkup_args args = {
24964 + .errp = &h_dentry,
24965 + .tgtname = &tgtname,
24966 + .h_parent = h_parent,
24970 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
24971 + if (unlikely(wkq_err))
24972 + h_dentry = ERR_PTR(wkq_err);
24974 + h_dentry = au_do_plink_lkup(&tgtname, h_parent, br);
24979 +/* create a pseudo-link */
24980 +static int do_whplink(struct qstr *tgt, struct dentry *h_parent,
24981 + struct dentry *h_dentry, struct au_branch *br)
24984 + struct path h_path = {
24985 + .mnt = au_br_mnt(br)
24987 + struct inode *h_dir, *delegated;
24989 + h_dir = d_inode(h_parent);
24990 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_CHILD2);
24992 + h_path.dentry = vfsub_lkup_one(tgt, h_parent);
24993 + err = PTR_ERR(h_path.dentry);
24994 + if (IS_ERR(h_path.dentry))
24998 + /* wh.plink dir is not monitored */
24999 + /* todo: is it really safe? */
25000 + if (d_is_positive(h_path.dentry)
25001 + && d_inode(h_path.dentry) != d_inode(h_dentry)) {
25002 + delegated = NULL;
25003 + err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
25004 + if (unlikely(err == -EWOULDBLOCK)) {
25005 + pr_warn("cannot retry for NFSv4 delegation"
25006 + " for an internal unlink\n");
25009 + dput(h_path.dentry);
25010 + h_path.dentry = NULL;
25014 + if (!err && d_is_negative(h_path.dentry)) {
25015 + delegated = NULL;
25016 + err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
25017 + if (unlikely(err == -EWOULDBLOCK)) {
25018 + pr_warn("cannot retry for NFSv4 delegation"
25019 + " for an internal link\n");
25023 + dput(h_path.dentry);
25026 + mutex_unlock(&h_dir->i_mutex);
25030 +struct do_whplink_args {
25032 + struct qstr *tgt;
25033 + struct dentry *h_parent;
25034 + struct dentry *h_dentry;
25035 + struct au_branch *br;
25038 +static void call_do_whplink(void *args)
25040 + struct do_whplink_args *a = args;
25041 + *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br);
25044 +static int whplink(struct dentry *h_dentry, struct inode *inode,
25045 + aufs_bindex_t bindex, struct au_branch *br)
25047 + int err, wkq_err;
25048 + struct au_wbr *wbr;
25049 + struct dentry *h_parent;
25050 + char a[PLINK_NAME_LEN];
25051 + struct qstr tgtname = QSTR_INIT(a, 0);
25053 + wbr = au_sbr(inode->i_sb, bindex)->br_wbr;
25054 + h_parent = wbr->wbr_plink;
25055 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
25057 + /* always superio. */
25058 + if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
25059 + struct do_whplink_args args = {
25062 + .h_parent = h_parent,
25063 + .h_dentry = h_dentry,
25066 + wkq_err = au_wkq_wait(call_do_whplink, &args);
25067 + if (unlikely(wkq_err))
25070 + err = do_whplink(&tgtname, h_parent, h_dentry, br);
25075 +/* free a single plink */
25076 +static void do_put_plink(struct pseudo_link *plink, int do_del)
25079 + hlist_del(&plink->hlist);
25080 + iput(plink->inode);
25084 +static void do_put_plink_rcu(struct rcu_head *rcu)
25086 + struct pseudo_link *plink;
25088 + plink = container_of(rcu, struct pseudo_link, rcu);
25089 + iput(plink->inode);
25094 + * create a new pseudo-link for @h_dentry on @bindex.
25095 + * the linked inode is held in aufs @inode.
25097 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
25098 + struct dentry *h_dentry)
25100 + struct super_block *sb;
25101 + struct au_sbinfo *sbinfo;
25102 + struct hlist_head *plink_hlist;
25103 + struct pseudo_link *plink, *tmp;
25104 + struct au_sphlhead *sphl;
25105 + int found, err, cnt, i;
25107 + sb = inode->i_sb;
25108 + sbinfo = au_sbi(sb);
25109 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
25110 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
25112 + found = au_plink_test(inode);
25116 + i = au_plink_hash(inode->i_ino);
25117 + sphl = sbinfo->si_plink + i;
25118 + plink_hlist = &sphl->head;
25119 + tmp = kmalloc(sizeof(*plink), GFP_NOFS);
25121 + tmp->inode = au_igrab(inode);
25127 + spin_lock(&sphl->spin);
25128 + hlist_for_each_entry(plink, plink_hlist, hlist) {
25129 + if (plink->inode == inode) {
25135 + hlist_add_head_rcu(&tmp->hlist, plink_hlist);
25136 + spin_unlock(&sphl->spin);
25138 + cnt = au_sphl_count(sphl);
25139 +#define msg "unexpectedly unblanced or too many pseudo-links"
25140 + if (cnt > AUFS_PLINK_WARN)
25141 + AuWarn1(msg ", %d\n", cnt);
25143 + err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex));
25145 + do_put_plink(tmp, 0);
25150 + if (unlikely(err)) {
25151 + pr_warn("err %d, damaged pseudo link.\n", err);
25153 + au_sphl_del_rcu(&tmp->hlist, sphl);
25154 + call_rcu(&tmp->rcu, do_put_plink_rcu);
25159 +/* free all plinks */
25160 +void au_plink_put(struct super_block *sb, int verbose)
25163 + struct au_sbinfo *sbinfo;
25164 + struct hlist_head *plink_hlist;
25165 + struct hlist_node *tmp;
25166 + struct pseudo_link *plink;
25168 + SiMustWriteLock(sb);
25170 + sbinfo = au_sbi(sb);
25171 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
25172 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
25174 + /* no spin_lock since sbinfo is write-locked */
25176 + for (i = 0; i < AuPlink_NHASH; i++) {
25177 + plink_hlist = &sbinfo->si_plink[i].head;
25178 + if (!warned && verbose && !hlist_empty(plink_hlist)) {
25179 + pr_warn("pseudo-link is not flushed");
25182 + hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist)
25183 + do_put_plink(plink, 0);
25184 + INIT_HLIST_HEAD(plink_hlist);
25188 +void au_plink_clean(struct super_block *sb, int verbose)
25190 + struct dentry *root;
25192 + root = sb->s_root;
25193 + aufs_write_lock(root);
25194 + if (au_opt_test(au_mntflags(sb), PLINK))
25195 + au_plink_put(sb, verbose);
25196 + aufs_write_unlock(root);
25199 +static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
25202 + aufs_bindex_t bstart, bend, bindex;
25205 + bstart = au_ibstart(inode);
25206 + bend = au_ibend(inode);
25207 + if (bstart >= 0) {
25208 + for (bindex = bstart; bindex <= bend; bindex++) {
25209 + if (!au_h_iptr(inode, bindex)
25210 + || au_ii_br_id(inode, bindex) != br_id)
25212 + au_set_h_iptr(inode, bindex, NULL, 0);
25217 + for (bindex = bstart; bindex <= bend; bindex++)
25218 + if (au_h_iptr(inode, bindex)) {
25228 +/* free the plinks on a branch specified by @br_id */
25229 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
25231 + struct au_sbinfo *sbinfo;
25232 + struct hlist_head *plink_hlist;
25233 + struct hlist_node *tmp;
25234 + struct pseudo_link *plink;
25235 + struct inode *inode;
25238 + SiMustWriteLock(sb);
25240 + sbinfo = au_sbi(sb);
25241 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
25242 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
25244 + /* no spin_lock since sbinfo is write-locked */
25245 + for (i = 0; i < AuPlink_NHASH; i++) {
25246 + plink_hlist = &sbinfo->si_plink[i].head;
25247 + hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist) {
25248 + inode = au_igrab(plink->inode);
25249 + ii_write_lock_child(inode);
25250 + do_put = au_plink_do_half_refresh(inode, br_id);
25252 + do_put_plink(plink, 1);
25253 + ii_write_unlock(inode);
25258 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
25259 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
25260 +++ linux/fs/aufs/poll.c 2015-09-24 10:47:58.254719746 +0200
25263 + * Copyright (C) 2005-2015 Junjiro R. Okajima
25265 + * This program, aufs is free software; you can redistribute it and/or modify
25266 + * it under the terms of the GNU General Public License as published by
25267 + * the Free Software Foundation; either version 2 of the License, or
25268 + * (at your option) any later version.
25270 + * This program is distributed in the hope that it will be useful,
25271 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25272 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25273 + * GNU General Public License for more details.
25275 + * You should have received a copy of the GNU General Public License
25276 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25281 + * There is only one filesystem which implements ->poll operation, currently.
25286 +unsigned int aufs_poll(struct file *file, poll_table *wait)
25288 + unsigned int mask;
25290 + struct file *h_file;
25291 + struct super_block *sb;
25293 + /* We should pretend an error happened. */
25294 + mask = POLLERR /* | POLLIN | POLLOUT */;
25295 + sb = file->f_path.dentry->d_sb;
25296 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
25298 + h_file = au_read_pre(file, /*keep_fi*/0);
25299 + err = PTR_ERR(h_file);
25300 + if (IS_ERR(h_file))
25303 + /* it is not an error if h_file has no operation */
25304 + mask = DEFAULT_POLLMASK;
25305 + if (h_file->f_op->poll)
25306 + mask = h_file->f_op->poll(h_file, wait);
25307 + fput(h_file); /* instead of au_read_post() */
25310 + si_read_unlock(sb);
25311 + AuTraceErr((int)mask);
25314 diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
25315 --- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
25316 +++ linux/fs/aufs/posix_acl.c 2015-09-24 10:47:58.254719746 +0200
25319 + * Copyright (C) 2014-2015 Junjiro R. Okajima
25321 + * This program, aufs is free software; you can redistribute it and/or modify
25322 + * it under the terms of the GNU General Public License as published by
25323 + * the Free Software Foundation; either version 2 of the License, or
25324 + * (at your option) any later version.
25326 + * This program is distributed in the hope that it will be useful,
25327 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25328 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25329 + * GNU General Public License for more details.
25331 + * You should have received a copy of the GNU General Public License
25332 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25336 + * posix acl operations
25339 +#include <linux/fs.h>
25340 +#include <linux/posix_acl.h>
25343 +struct posix_acl *aufs_get_acl(struct inode *inode, int type)
25345 + struct posix_acl *acl;
25347 + aufs_bindex_t bindex;
25348 + struct inode *h_inode;
25349 + struct super_block *sb;
25352 + sb = inode->i_sb;
25353 + si_read_lock(sb, AuLock_FLUSH);
25354 + ii_read_lock_child(inode);
25355 + if (!(sb->s_flags & MS_POSIXACL))
25358 + bindex = au_ibstart(inode);
25359 + h_inode = au_h_iptr(inode, bindex);
25360 + if (unlikely(!h_inode
25361 + || ((h_inode->i_mode & S_IFMT)
25362 + != (inode->i_mode & S_IFMT)))) {
25363 + err = au_busy_or_stale();
25364 + acl = ERR_PTR(err);
25368 + /* always topmost only */
25369 + acl = get_acl(h_inode, type);
25372 + ii_read_unlock(inode);
25373 + si_read_unlock(sb);
25375 + AuTraceErrPtr(acl);
25379 +int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
25383 + struct dentry *dentry;
25384 + struct au_srxattr arg = {
25385 + .type = AU_ACL_SET,
25392 + mutex_lock(&inode->i_mutex);
25393 + if (inode->i_ino == AUFS_ROOT_INO)
25394 + dentry = dget(inode->i_sb->s_root);
25396 + dentry = d_find_alias(inode);
25398 + dentry = d_find_any_alias(inode);
25400 + pr_warn("cannot handle this inode, "
25401 + "please report to aufs-users ML\n");
25407 + ssz = au_srxattr(dentry, &arg);
25414 + mutex_unlock(&inode->i_mutex);
25417 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
25418 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
25419 +++ linux/fs/aufs/procfs.c 2015-09-24 10:47:58.254719746 +0200
25422 + * Copyright (C) 2010-2015 Junjiro R. Okajima
25424 + * This program, aufs is free software; you can redistribute it and/or modify
25425 + * it under the terms of the GNU General Public License as published by
25426 + * the Free Software Foundation; either version 2 of the License, or
25427 + * (at your option) any later version.
25429 + * This program is distributed in the hope that it will be useful,
25430 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25431 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25432 + * GNU General Public License for more details.
25434 + * You should have received a copy of the GNU General Public License
25435 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25439 + * procfs interfaces
25442 +#include <linux/proc_fs.h>
25445 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
25447 + struct au_sbinfo *sbinfo;
25449 + sbinfo = file->private_data;
25451 + au_plink_maint_leave(sbinfo);
25452 + kobject_put(&sbinfo->si_kobj);
25458 +static void au_procfs_plm_write_clean(struct file *file)
25460 + struct au_sbinfo *sbinfo;
25462 + sbinfo = file->private_data;
25464 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
25467 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
25470 + struct super_block *sb;
25471 + struct au_sbinfo *sbinfo;
25474 + if (unlikely(file->private_data))
25478 + /* don't use au_sbilist_lock() here */
25479 + spin_lock(&au_sbilist.spin);
25480 + list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
25481 + if (id == sysaufs_si_id(sbinfo)) {
25482 + kobject_get(&sbinfo->si_kobj);
25483 + sb = sbinfo->si_sb;
25486 + spin_unlock(&au_sbilist.spin);
25489 + if (unlikely(!sb))
25492 + err = au_plink_maint_enter(sb);
25494 + /* keep kobject_get() */
25495 + file->private_data = sbinfo;
25497 + kobject_put(&sbinfo->si_kobj);
25503 + * Accept a valid "si=xxxx" only.
25504 + * Once it is accepted successfully, accept "clean" too.
25506 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
25507 + size_t count, loff_t *ppos)
25510 + unsigned long id;
25511 + /* last newline is allowed */
25512 + char buf[3 + sizeof(unsigned long) * 2 + 1];
25515 + if (unlikely(!capable(CAP_SYS_ADMIN)))
25519 + if (unlikely(count > sizeof(buf)))
25522 + err = copy_from_user(buf, ubuf, count);
25523 + if (unlikely(err)) {
25530 + if (!strcmp("clean", buf)) {
25531 + au_procfs_plm_write_clean(file);
25532 + goto out_success;
25533 + } else if (unlikely(strncmp("si=", buf, 3)))
25536 + err = kstrtoul(buf + 3, 16, &id);
25537 + if (unlikely(err))
25540 + err = au_procfs_plm_write_si(file, id);
25541 + if (unlikely(err))
25545 + err = count; /* success */
25550 +static const struct file_operations au_procfs_plm_fop = {
25551 + .write = au_procfs_plm_write,
25552 + .release = au_procfs_plm_release,
25553 + .owner = THIS_MODULE
25556 +/* ---------------------------------------------------------------------- */
25558 +static struct proc_dir_entry *au_procfs_dir;
25560 +void au_procfs_fin(void)
25562 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
25563 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
25566 +int __init au_procfs_init(void)
25569 + struct proc_dir_entry *entry;
25572 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
25573 + if (unlikely(!au_procfs_dir))
25576 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | S_IWUSR,
25577 + au_procfs_dir, &au_procfs_plm_fop);
25578 + if (unlikely(!entry))
25582 + goto out; /* success */
25586 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
25590 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
25591 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
25592 +++ linux/fs/aufs/rdu.c 2015-09-24 10:47:58.254719746 +0200
25595 + * Copyright (C) 2005-2015 Junjiro R. Okajima
25597 + * This program, aufs is free software; you can redistribute it and/or modify
25598 + * it under the terms of the GNU General Public License as published by
25599 + * the Free Software Foundation; either version 2 of the License, or
25600 + * (at your option) any later version.
25602 + * This program is distributed in the hope that it will be useful,
25603 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25604 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25605 + * GNU General Public License for more details.
25607 + * You should have received a copy of the GNU General Public License
25608 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
25612 + * readdir in userspace.
25615 +#include <linux/compat.h>
25616 +#include <linux/fs_stack.h>
25617 +#include <linux/security.h>
25620 +/* bits for struct aufs_rdu.flags */
25621 +#define AuRdu_CALLED 1
25622 +#define AuRdu_CONT (1 << 1)
25623 +#define AuRdu_FULL (1 << 2)
25624 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
25625 +#define au_fset_rdu(flags, name) \
25626 + do { (flags) |= AuRdu_##name; } while (0)
25627 +#define au_fclr_rdu(flags, name) \
25628 + do { (flags) &= ~AuRdu_##name; } while (0)
25630 +struct au_rdu_arg {
25631 + struct dir_context ctx;
25632 + struct aufs_rdu *rdu;
25633 + union au_rdu_ent_ul ent;
25634 + unsigned long end;
25636 + struct super_block *sb;
25640 +static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
25641 + loff_t offset, u64 h_ino, unsigned int d_type)
25644 + struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
25645 + struct aufs_rdu *rdu = arg->rdu;
25646 + struct au_rdu_ent ent;
25650 + au_fset_rdu(rdu->cookie.flags, CALLED);
25651 + len = au_rdu_len(nlen);
25652 + if (arg->ent.ul + len < arg->end) {
25654 + ent.bindex = rdu->cookie.bindex;
25655 + ent.type = d_type;
25657 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
25658 + ent.type = DT_UNKNOWN;
25660 + /* unnecessary to support mmap_sem since this is a dir */
25662 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
25664 + if (copy_to_user(arg->ent.e->name, name, nlen))
25666 + /* the terminating NULL */
25667 + if (__put_user(0, arg->ent.e->name + nlen))
25670 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
25671 + arg->ent.ul += len;
25675 + au_fset_rdu(rdu->cookie.flags, FULL);
25677 + rdu->tail = arg->ent;
25681 + /* AuTraceErr(err); */
25685 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
25689 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
25691 + /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
25692 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
25694 + if (unlikely(offset != cookie->h_pos))
25700 + au_fclr_rdu(cookie->flags, CALLED);
25702 + err = vfsub_iterate_dir(h_file, &arg->ctx);
25706 + && au_ftest_rdu(cookie->flags, CALLED)
25707 + && !au_ftest_rdu(cookie->flags, FULL));
25708 + cookie->h_pos = h_file->f_pos;
25715 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
25718 + aufs_bindex_t bend;
25719 + struct au_rdu_arg arg = {
25721 + .actor = au_rdu_fill
25724 + struct dentry *dentry;
25725 + struct inode *inode;
25726 + struct file *h_file;
25727 + struct au_rdu_cookie *cookie = &rdu->cookie;
25729 + err = !access_ok(VERIFY_WRITE, rdu->ent.e, rdu->sz);
25730 + if (unlikely(err)) {
25736 + rdu->tail = rdu->ent;
25739 + arg.ent = rdu->ent;
25740 + arg.end = arg.ent.ul;
25741 + arg.end += rdu->sz;
25744 + if (unlikely(!file->f_op->iterate))
25747 + err = security_file_permission(file, MAY_READ);
25749 + if (unlikely(err))
25752 + dentry = file->f_path.dentry;
25753 + inode = d_inode(dentry);
25755 + mutex_lock(&inode->i_mutex);
25757 + err = mutex_lock_killable(&inode->i_mutex);
25759 + if (unlikely(err))
25763 + arg.sb = inode->i_sb;
25764 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
25765 + if (unlikely(err))
25767 + err = au_alive_dir(dentry);
25768 + if (unlikely(err))
25770 + /* todo: reval? */
25771 + fi_read_lock(file);
25774 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
25775 + && cookie->generation != au_figen(file)))
25780 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
25782 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
25784 + bend = au_fbstart(file);
25785 + if (cookie->bindex < bend)
25786 + cookie->bindex = bend;
25787 + bend = au_fbend_dir(file);
25788 + /* AuDbg("b%d, b%d\n", cookie->bindex, bend); */
25789 + for (; !err && cookie->bindex <= bend;
25790 + cookie->bindex++, cookie->h_pos = 0) {
25791 + h_file = au_hf_dir(file, cookie->bindex);
25795 + au_fclr_rdu(cookie->flags, FULL);
25796 + err = au_rdu_do(h_file, &arg);
25798 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
25801 + AuDbg("rent %llu\n", rdu->rent);
25803 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
25804 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
25805 + au_fset_rdu(cookie->flags, CONT);
25806 + cookie->generation = au_figen(file);
25809 + ii_read_lock_child(inode);
25810 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibstart(inode)));
25811 + ii_read_unlock(inode);
25814 + fi_read_unlock(file);
25816 + si_read_unlock(arg.sb);
25818 + mutex_unlock(&inode->i_mutex);
25824 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
25828 + unsigned long long nent;
25829 + union au_rdu_ent_ul *u;
25830 + struct au_rdu_ent ent;
25831 + struct super_block *sb;
25834 + nent = rdu->nent;
25836 + sb = file->f_path.dentry->d_sb;
25837 + si_read_lock(sb, AuLock_FLUSH);
25838 + while (nent-- > 0) {
25839 + /* unnecessary to support mmap_sem since this is a dir */
25840 + err = copy_from_user(&ent, u->e, sizeof(ent));
25842 + err = !access_ok(VERIFY_WRITE, &u->e->ino, sizeof(ino));
25843 + if (unlikely(err)) {
25849 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
25851 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
25853 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
25855 + if (unlikely(err)) {
25860 + err = __put_user(ino, &u->e->ino);
25861 + if (unlikely(err)) {
25866 + u->ul += au_rdu_len(ent.nlen);
25868 + si_read_unlock(sb);
25873 +/* ---------------------------------------------------------------------- */
25875 +static int au_rdu_verify(struct aufs_rdu *rdu)
25877 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
25878 + "%llu, b%d, 0x%x, g%u}\n",
25879 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
25881 + rdu->rent, rdu->shwh, rdu->full,
25882 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
25883 + rdu->cookie.generation);
25885 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
25889 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
25893 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
25896 + struct aufs_rdu rdu;
25897 + void __user *p = (void __user *)arg;
25899 + err = copy_from_user(&rdu, p, sizeof(rdu));
25900 + if (unlikely(err)) {
25905 + err = au_rdu_verify(&rdu);
25906 + if (unlikely(err))
25910 + case AUFS_CTL_RDU:
25911 + err = au_rdu(file, &rdu);
25912 + if (unlikely(err))
25915 + e = copy_to_user(p, &rdu, sizeof(rdu));
25916 + if (unlikely(e)) {
25921 + case AUFS_CTL_RDU_INO:
25922 + err = au_rdu_ino(file, &rdu);
25926 + /* err = -ENOTTY; */
25935 +#ifdef CONFIG_COMPAT
25936 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
25939 + struct aufs_rdu rdu;
25940 + void __user *p = compat_ptr(arg);
25942 + /* todo: get_user()? */
25943 + err = copy_from_user(&rdu, p, sizeof(rdu));
25944 + if (unlikely(err)) {
25949 + rdu.ent.e = compat_ptr(rdu.ent.ul);
25950 + err = au_rdu_verify(&rdu);
25951 + if (unlikely(err))
25955 + case AUFS_CTL_RDU:
25956 + err = au_rdu(file, &rdu);
25957 + if (unlikely(err))
25960 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
25961 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
25962 + e = copy_to_user(p, &rdu, sizeof(rdu));
25963 + if (unlikely(e)) {
25968 + case AUFS_CTL_RDU_INO:
25969 + err = au_rdu_ino(file, &rdu);
25973 + /* err = -ENOTTY; */
25982 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
25983 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
25984 +++ linux/fs/aufs/rwsem.h 2015-09-24 10:47:58.254719746 +0200
25987 + * Copyright (C) 2005-2015 Junjiro R. Okajima
25989 + * This program, aufs is free software; you can redistribute it and/or modify
25990 + * it under the terms of the GNU General Public License as published by
25991 + * the Free Software Foundation; either version 2 of the License, or
25992 + * (at your option) any later version.
25994 + * This program is distributed in the hope that it will be useful,
25995 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25996 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25997 + * GNU General Public License for more details.
25999 + * You should have received a copy of the GNU General Public License
26000 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26004 + * simple read-write semaphore wrappers
26007 +#ifndef __AUFS_RWSEM_H__
26008 +#define __AUFS_RWSEM_H__
26012 +#include "debug.h"
26015 + struct rw_semaphore rwsem;
26016 +#ifdef CONFIG_AUFS_DEBUG
26017 + /* just for debugging, not almighty counter */
26018 + atomic_t rcnt, wcnt;
26022 +#ifdef CONFIG_AUFS_DEBUG
26023 +#define AuDbgCntInit(rw) do { \
26024 + atomic_set(&(rw)->rcnt, 0); \
26025 + atomic_set(&(rw)->wcnt, 0); \
26026 + smp_mb(); /* atomic set */ \
26029 +#define AuDbgRcntInc(rw) atomic_inc(&(rw)->rcnt)
26030 +#define AuDbgRcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->rcnt) < 0)
26031 +#define AuDbgWcntInc(rw) atomic_inc(&(rw)->wcnt)
26032 +#define AuDbgWcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->wcnt) < 0)
26034 +#define AuDbgCntInit(rw) do {} while (0)
26035 +#define AuDbgRcntInc(rw) do {} while (0)
26036 +#define AuDbgRcntDec(rw) do {} while (0)
26037 +#define AuDbgWcntInc(rw) do {} while (0)
26038 +#define AuDbgWcntDec(rw) do {} while (0)
26039 +#endif /* CONFIG_AUFS_DEBUG */
26041 +/* to debug easier, do not make them inlined functions */
26042 +#define AuRwMustNoWaiters(rw) AuDebugOn(!list_empty(&(rw)->rwsem.wait_list))
26043 +/* rwsem_is_locked() is unusable */
26044 +#define AuRwMustReadLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0)
26045 +#define AuRwMustWriteLock(rw) AuDebugOn(atomic_read(&(rw)->wcnt) <= 0)
26046 +#define AuRwMustAnyLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0 \
26047 + && atomic_read(&(rw)->wcnt) <= 0)
26048 +#define AuRwDestroy(rw) AuDebugOn(atomic_read(&(rw)->rcnt) \
26049 + || atomic_read(&(rw)->wcnt))
26051 +#define au_rw_class(rw, key) lockdep_set_class(&(rw)->rwsem, key)
26053 +static inline void au_rw_init(struct au_rwsem *rw)
26055 + AuDbgCntInit(rw);
26056 + init_rwsem(&rw->rwsem);
26059 +static inline void au_rw_init_wlock(struct au_rwsem *rw)
26062 + down_write(&rw->rwsem);
26063 + AuDbgWcntInc(rw);
26066 +static inline void au_rw_init_wlock_nested(struct au_rwsem *rw,
26067 + unsigned int lsc)
26070 + down_write_nested(&rw->rwsem, lsc);
26071 + AuDbgWcntInc(rw);
26074 +static inline void au_rw_read_lock(struct au_rwsem *rw)
26076 + down_read(&rw->rwsem);
26077 + AuDbgRcntInc(rw);
26080 +static inline void au_rw_read_lock_nested(struct au_rwsem *rw, unsigned int lsc)
26082 + down_read_nested(&rw->rwsem, lsc);
26083 + AuDbgRcntInc(rw);
26086 +static inline void au_rw_read_unlock(struct au_rwsem *rw)
26088 + AuRwMustReadLock(rw);
26089 + AuDbgRcntDec(rw);
26090 + up_read(&rw->rwsem);
26093 +static inline void au_rw_dgrade_lock(struct au_rwsem *rw)
26095 + AuRwMustWriteLock(rw);
26096 + AuDbgRcntInc(rw);
26097 + AuDbgWcntDec(rw);
26098 + downgrade_write(&rw->rwsem);
26101 +static inline void au_rw_write_lock(struct au_rwsem *rw)
26103 + down_write(&rw->rwsem);
26104 + AuDbgWcntInc(rw);
26107 +static inline void au_rw_write_lock_nested(struct au_rwsem *rw,
26108 + unsigned int lsc)
26110 + down_write_nested(&rw->rwsem, lsc);
26111 + AuDbgWcntInc(rw);
26114 +static inline void au_rw_write_unlock(struct au_rwsem *rw)
26116 + AuRwMustWriteLock(rw);
26117 + AuDbgWcntDec(rw);
26118 + up_write(&rw->rwsem);
26121 +/* why is not _nested version defined */
26122 +static inline int au_rw_read_trylock(struct au_rwsem *rw)
26126 + ret = down_read_trylock(&rw->rwsem);
26128 + AuDbgRcntInc(rw);
26132 +static inline int au_rw_write_trylock(struct au_rwsem *rw)
26136 + ret = down_write_trylock(&rw->rwsem);
26138 + AuDbgWcntInc(rw);
26142 +#undef AuDbgCntInit
26143 +#undef AuDbgRcntInc
26144 +#undef AuDbgRcntDec
26145 +#undef AuDbgWcntInc
26146 +#undef AuDbgWcntDec
26148 +#define AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
26149 +static inline void prefix##_read_lock(param) \
26150 +{ au_rw_read_lock(rwsem); } \
26151 +static inline void prefix##_write_lock(param) \
26152 +{ au_rw_write_lock(rwsem); } \
26153 +static inline int prefix##_read_trylock(param) \
26154 +{ return au_rw_read_trylock(rwsem); } \
26155 +static inline int prefix##_write_trylock(param) \
26156 +{ return au_rw_write_trylock(rwsem); }
26157 +/* why is not _nested version defined */
26158 +/* static inline void prefix##_read_trylock_nested(param, lsc)
26159 +{ au_rw_read_trylock_nested(rwsem, lsc)); }
26160 +static inline void prefix##_write_trylock_nestd(param, lsc)
26161 +{ au_rw_write_trylock_nested(rwsem, lsc); } */
26163 +#define AuSimpleUnlockRwsemFuncs(prefix, param, rwsem) \
26164 +static inline void prefix##_read_unlock(param) \
26165 +{ au_rw_read_unlock(rwsem); } \
26166 +static inline void prefix##_write_unlock(param) \
26167 +{ au_rw_write_unlock(rwsem); } \
26168 +static inline void prefix##_downgrade_lock(param) \
26169 +{ au_rw_dgrade_lock(rwsem); }
26171 +#define AuSimpleRwsemFuncs(prefix, param, rwsem) \
26172 + AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
26173 + AuSimpleUnlockRwsemFuncs(prefix, param, rwsem)
26175 +#endif /* __KERNEL__ */
26176 +#endif /* __AUFS_RWSEM_H__ */
26177 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
26178 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
26179 +++ linux/fs/aufs/sbinfo.c 2015-11-11 17:21:46.922197217 +0100
26182 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26184 + * This program, aufs is free software; you can redistribute it and/or modify
26185 + * it under the terms of the GNU General Public License as published by
26186 + * the Free Software Foundation; either version 2 of the License, or
26187 + * (at your option) any later version.
26189 + * This program is distributed in the hope that it will be useful,
26190 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26191 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26192 + * GNU General Public License for more details.
26194 + * You should have received a copy of the GNU General Public License
26195 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26199 + * superblock private data
26205 + * they are necessary regardless sysfs is disabled.
26207 +void au_si_free(struct kobject *kobj)
26210 + struct au_sbinfo *sbinfo;
26211 + char *locked __maybe_unused; /* debug only */
26213 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
26214 + for (i = 0; i < AuPlink_NHASH; i++)
26215 + AuDebugOn(!hlist_empty(&sbinfo->si_plink[i].head));
26216 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
26218 + AuDebugOn(!hlist_empty(&sbinfo->si_symlink.head));
26220 + au_rw_write_lock(&sbinfo->si_rwsem);
26221 + au_br_free(sbinfo);
26222 + au_rw_write_unlock(&sbinfo->si_rwsem);
26224 + AuDebugOn(radix_tree_gang_lookup
26225 + (&sbinfo->au_si_pid.tree, (void **)&locked,
26226 + /*first_index*/PID_MAX_DEFAULT - 1,
26227 + /*max_items*/sizeof(locked)/sizeof(*locked)));
26229 + kfree(sbinfo->si_branch);
26230 + kfree(sbinfo->au_si_pid.bitmap);
26231 + mutex_destroy(&sbinfo->si_xib_mtx);
26232 + AuRwDestroy(&sbinfo->si_rwsem);
26237 +int au_si_alloc(struct super_block *sb)
26240 + struct au_sbinfo *sbinfo;
26241 + static struct lock_class_key aufs_si;
26244 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
26245 + if (unlikely(!sbinfo))
26248 + BUILD_BUG_ON(sizeof(unsigned long) !=
26249 + sizeof(*sbinfo->au_si_pid.bitmap));
26250 + sbinfo->au_si_pid.bitmap = kcalloc(BITS_TO_LONGS(PID_MAX_DEFAULT),
26251 + sizeof(*sbinfo->au_si_pid.bitmap),
26253 + if (unlikely(!sbinfo->au_si_pid.bitmap))
26256 + /* will be reallocated separately */
26257 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
26258 + if (unlikely(!sbinfo->si_branch))
26261 + err = sysaufs_si_init(sbinfo);
26262 + if (unlikely(err))
26265 + au_nwt_init(&sbinfo->si_nowait);
26266 + au_rw_init_wlock(&sbinfo->si_rwsem);
26267 + au_rw_class(&sbinfo->si_rwsem, &aufs_si);
26268 + spin_lock_init(&sbinfo->au_si_pid.tree_lock);
26269 + INIT_RADIX_TREE(&sbinfo->au_si_pid.tree, GFP_ATOMIC | __GFP_NOFAIL);
26271 + atomic_long_set(&sbinfo->si_ninodes, 0);
26272 + atomic_long_set(&sbinfo->si_nfiles, 0);
26274 + sbinfo->si_bend = -1;
26275 + sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
26277 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
26278 + sbinfo->si_wbr_create = AuWbrCreate_Def;
26279 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
26280 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
26282 + au_fhsm_init(sbinfo);
26284 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
26286 + au_sphl_init(&sbinfo->si_symlink);
26288 + sbinfo->si_xino_jiffy = jiffies;
26289 + sbinfo->si_xino_expire
26290 + = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
26291 + mutex_init(&sbinfo->si_xib_mtx);
26292 + sbinfo->si_xino_brid = -1;
26293 + /* leave si_xib_last_pindex and si_xib_next_bit */
26295 + au_sphl_init(&sbinfo->si_aopen);
26297 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
26298 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
26299 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
26300 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
26302 + for (i = 0; i < AuPlink_NHASH; i++)
26303 + au_sphl_init(sbinfo->si_plink + i);
26304 + init_waitqueue_head(&sbinfo->si_plink_wq);
26305 + spin_lock_init(&sbinfo->si_plink_maint_lock);
26307 + au_sphl_init(&sbinfo->si_files);
26309 + /* leave other members for sysaufs and si_mnt. */
26310 + sbinfo->si_sb = sb;
26311 + sb->s_fs_info = sbinfo;
26313 + return 0; /* success */
26316 + kfree(sbinfo->si_branch);
26318 + kfree(sbinfo->au_si_pid.bitmap);
26325 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr)
26328 + struct au_branch **brp;
26330 + AuRwMustWriteLock(&sbinfo->si_rwsem);
26333 + sz = sizeof(*brp) * (sbinfo->si_bend + 1);
26334 + if (unlikely(!sz))
26335 + sz = sizeof(*brp);
26336 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS);
26338 + sbinfo->si_branch = brp;
26345 +/* ---------------------------------------------------------------------- */
26347 +unsigned int au_sigen_inc(struct super_block *sb)
26349 + unsigned int gen;
26350 + struct inode *inode;
26352 + SiMustWriteLock(sb);
26354 + gen = ++au_sbi(sb)->si_generation;
26355 + au_update_digen(sb->s_root);
26356 + inode = d_inode(sb->s_root);
26357 + au_update_iigen(inode, /*half*/0);
26358 + inode->i_version++;
26362 +aufs_bindex_t au_new_br_id(struct super_block *sb)
26364 + aufs_bindex_t br_id;
26366 + struct au_sbinfo *sbinfo;
26368 + SiMustWriteLock(sb);
26370 + sbinfo = au_sbi(sb);
26371 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
26372 + br_id = ++sbinfo->si_last_br_id;
26373 + AuDebugOn(br_id < 0);
26374 + if (br_id && au_br_index(sb, br_id) < 0)
26381 +/* ---------------------------------------------------------------------- */
26383 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
26384 +int si_read_lock(struct super_block *sb, int flags)
26389 + if (au_ftest_lock(flags, FLUSH))
26390 + au_nwt_flush(&au_sbi(sb)->si_nowait);
26392 + si_noflush_read_lock(sb);
26393 + err = au_plink_maint(sb, flags);
26394 + if (unlikely(err))
26395 + si_read_unlock(sb);
26400 +int si_write_lock(struct super_block *sb, int flags)
26404 + if (au_ftest_lock(flags, FLUSH))
26405 + au_nwt_flush(&au_sbi(sb)->si_nowait);
26407 + si_noflush_write_lock(sb);
26408 + err = au_plink_maint(sb, flags);
26409 + if (unlikely(err))
26410 + si_write_unlock(sb);
26415 +/* dentry and super_block lock. call at entry point */
26416 +int aufs_read_lock(struct dentry *dentry, int flags)
26419 + struct super_block *sb;
26421 + sb = dentry->d_sb;
26422 + err = si_read_lock(sb, flags);
26423 + if (unlikely(err))
26426 + if (au_ftest_lock(flags, DW))
26427 + di_write_lock_child(dentry);
26429 + di_read_lock_child(dentry, flags);
26431 + if (au_ftest_lock(flags, GEN)) {
26432 + err = au_digen_test(dentry, au_sigen(sb));
26433 + if (!au_opt_test(au_mntflags(sb), UDBA_NONE))
26434 + AuDebugOn(!err && au_dbrange_test(dentry));
26436 + err = au_dbrange_test(dentry);
26437 + if (unlikely(err))
26438 + aufs_read_unlock(dentry, flags);
26445 +void aufs_read_unlock(struct dentry *dentry, int flags)
26447 + if (au_ftest_lock(flags, DW))
26448 + di_write_unlock(dentry);
26450 + di_read_unlock(dentry, flags);
26451 + si_read_unlock(dentry->d_sb);
26454 +void aufs_write_lock(struct dentry *dentry)
26456 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
26457 + di_write_lock_child(dentry);
26460 +void aufs_write_unlock(struct dentry *dentry)
26462 + di_write_unlock(dentry);
26463 + si_write_unlock(dentry->d_sb);
26466 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
26469 + unsigned int sigen;
26470 + struct super_block *sb;
26473 + err = si_read_lock(sb, flags);
26474 + if (unlikely(err))
26477 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIR));
26479 + if (au_ftest_lock(flags, GEN)) {
26480 + sigen = au_sigen(sb);
26481 + err = au_digen_test(d1, sigen);
26482 + AuDebugOn(!err && au_dbrange_test(d1));
26484 + err = au_digen_test(d2, sigen);
26485 + AuDebugOn(!err && au_dbrange_test(d2));
26487 + if (unlikely(err))
26488 + aufs_read_and_write_unlock2(d1, d2);
26495 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
26497 + di_write_unlock2(d1, d2);
26498 + si_read_unlock(d1->d_sb);
26501 +/* ---------------------------------------------------------------------- */
26503 +int si_pid_test_slow(struct super_block *sb)
26508 + p = radix_tree_lookup(&au_sbi(sb)->au_si_pid.tree, current->pid);
26509 + rcu_read_unlock();
26511 + return (long)!!p;
26514 +void si_pid_set_slow(struct super_block *sb)
26517 + struct au_sbinfo *sbinfo;
26519 + AuDebugOn(si_pid_test_slow(sb));
26521 + sbinfo = au_sbi(sb);
26522 + err = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
26524 + spin_lock(&sbinfo->au_si_pid.tree_lock);
26525 + err = radix_tree_insert(&sbinfo->au_si_pid.tree, current->pid,
26526 + /*any valid ptr*/sb);
26527 + spin_unlock(&sbinfo->au_si_pid.tree_lock);
26529 + radix_tree_preload_end();
26532 +void si_pid_clr_slow(struct super_block *sb)
26535 + struct au_sbinfo *sbinfo;
26537 + AuDebugOn(!si_pid_test_slow(sb));
26539 + sbinfo = au_sbi(sb);
26540 + spin_lock(&sbinfo->au_si_pid.tree_lock);
26541 + p = radix_tree_delete(&sbinfo->au_si_pid.tree, current->pid);
26542 + spin_unlock(&sbinfo->au_si_pid.tree_lock);
26544 diff -urN /usr/share/empty/fs/aufs/spl.h linux/fs/aufs/spl.h
26545 --- /usr/share/empty/fs/aufs/spl.h 1970-01-01 01:00:00.000000000 +0100
26546 +++ linux/fs/aufs/spl.h 2015-09-24 10:47:58.254719746 +0200
26549 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26551 + * This program, aufs is free software; you can redistribute it and/or modify
26552 + * it under the terms of the GNU General Public License as published by
26553 + * the Free Software Foundation; either version 2 of the License, or
26554 + * (at your option) any later version.
26556 + * This program is distributed in the hope that it will be useful,
26557 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26558 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26559 + * GNU General Public License for more details.
26561 + * You should have received a copy of the GNU General Public License
26562 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26566 + * simple list protected by a spinlock
26569 +#ifndef __AUFS_SPL_H__
26570 +#define __AUFS_SPL_H__
26574 +struct au_splhead {
26576 + struct list_head head;
26579 +static inline void au_spl_init(struct au_splhead *spl)
26581 + spin_lock_init(&spl->spin);
26582 + INIT_LIST_HEAD(&spl->head);
26585 +static inline void au_spl_add(struct list_head *list, struct au_splhead *spl)
26587 + spin_lock(&spl->spin);
26588 + list_add(list, &spl->head);
26589 + spin_unlock(&spl->spin);
26592 +static inline void au_spl_del(struct list_head *list, struct au_splhead *spl)
26594 + spin_lock(&spl->spin);
26596 + spin_unlock(&spl->spin);
26599 +static inline void au_spl_del_rcu(struct list_head *list,
26600 + struct au_splhead *spl)
26602 + spin_lock(&spl->spin);
26603 + list_del_rcu(list);
26604 + spin_unlock(&spl->spin);
26607 +/* ---------------------------------------------------------------------- */
26609 +struct au_sphlhead {
26611 + struct hlist_head head;
26614 +static inline void au_sphl_init(struct au_sphlhead *sphl)
26616 + spin_lock_init(&sphl->spin);
26617 + INIT_HLIST_HEAD(&sphl->head);
26620 +static inline void au_sphl_add(struct hlist_node *hlist,
26621 + struct au_sphlhead *sphl)
26623 + spin_lock(&sphl->spin);
26624 + hlist_add_head(hlist, &sphl->head);
26625 + spin_unlock(&sphl->spin);
26628 +static inline void au_sphl_del(struct hlist_node *hlist,
26629 + struct au_sphlhead *sphl)
26631 + spin_lock(&sphl->spin);
26632 + hlist_del(hlist);
26633 + spin_unlock(&sphl->spin);
26636 +static inline void au_sphl_del_rcu(struct hlist_node *hlist,
26637 + struct au_sphlhead *sphl)
26639 + spin_lock(&sphl->spin);
26640 + hlist_del_rcu(hlist);
26641 + spin_unlock(&sphl->spin);
26644 +static inline unsigned long au_sphl_count(struct au_sphlhead *sphl)
26646 + unsigned long cnt;
26647 + struct hlist_node *pos;
26650 + spin_lock(&sphl->spin);
26651 + hlist_for_each(pos, &sphl->head)
26653 + spin_unlock(&sphl->spin);
26657 +#endif /* __KERNEL__ */
26658 +#endif /* __AUFS_SPL_H__ */
26659 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
26660 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
26661 +++ linux/fs/aufs/super.c 2015-11-11 17:21:46.922197217 +0100
26664 + * Copyright (C) 2005-2015 Junjiro R. Okajima
26666 + * This program, aufs is free software; you can redistribute it and/or modify
26667 + * it under the terms of the GNU General Public License as published by
26668 + * the Free Software Foundation; either version 2 of the License, or
26669 + * (at your option) any later version.
26671 + * This program is distributed in the hope that it will be useful,
26672 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26673 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26674 + * GNU General Public License for more details.
26676 + * You should have received a copy of the GNU General Public License
26677 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
26681 + * mount and super_block operations
26684 +#include <linux/mm.h>
26685 +#include <linux/seq_file.h>
26686 +#include <linux/statfs.h>
26687 +#include <linux/vmalloc.h>
26691 + * super_operations
26693 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
26695 + struct au_icntnr *c;
26697 + c = au_cache_alloc_icntnr();
26699 + au_icntnr_init(c);
26700 + c->vfs_inode.i_version = 1; /* sigen(sb); */
26701 + c->iinfo.ii_hinode = NULL;
26702 + return &c->vfs_inode;
26707 +static void aufs_destroy_inode_cb(struct rcu_head *head)
26709 + struct inode *inode = container_of(head, struct inode, i_rcu);
26711 + INIT_HLIST_HEAD(&inode->i_dentry);
26712 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
26715 +static void aufs_destroy_inode(struct inode *inode)
26717 + au_iinfo_fin(inode);
26718 + call_rcu(&inode->i_rcu, aufs_destroy_inode_cb);
26721 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
26723 + struct inode *inode;
26726 + inode = iget_locked(sb, ino);
26727 + if (unlikely(!inode)) {
26728 + inode = ERR_PTR(-ENOMEM);
26731 + if (!(inode->i_state & I_NEW))
26734 + err = au_xigen_new(inode);
26736 + err = au_iinfo_init(inode);
26738 + inode->i_version++;
26740 + iget_failed(inode);
26741 + inode = ERR_PTR(err);
26745 + /* never return NULL */
26746 + AuDebugOn(!inode);
26747 + AuTraceErrPtr(inode);
26751 +/* lock free root dinfo */
26752 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
26755 + aufs_bindex_t bindex, bend;
26756 + struct path path;
26757 + struct au_hdentry *hdp;
26758 + struct au_branch *br;
26759 + au_br_perm_str_t perm;
26762 + bend = au_sbend(sb);
26763 + hdp = au_di(sb->s_root)->di_hdentry;
26764 + for (bindex = 0; !err && bindex <= bend; bindex++) {
26765 + br = au_sbr(sb, bindex);
26766 + path.mnt = au_br_mnt(br);
26767 + path.dentry = hdp[bindex].hd_dentry;
26768 + err = au_seq_path(seq, &path);
26770 + au_optstr_br_perm(&perm, br->br_perm);
26771 + seq_printf(seq, "=%s", perm.a);
26772 + if (bindex != bend)
26773 + seq_putc(seq, ':');
26776 + if (unlikely(err || seq_has_overflowed(seq)))
26782 +static void au_show_wbr_create(struct seq_file *m, int v,
26783 + struct au_sbinfo *sbinfo)
26787 + AuRwMustAnyLock(&sbinfo->si_rwsem);
26789 + seq_puts(m, ",create=");
26790 + pat = au_optstr_wbr_create(v);
26792 + case AuWbrCreate_TDP:
26793 + case AuWbrCreate_RR:
26794 + case AuWbrCreate_MFS:
26795 + case AuWbrCreate_PMFS:
26796 + seq_puts(m, pat);
26798 + case AuWbrCreate_MFSV:
26799 + seq_printf(m, /*pat*/"mfs:%lu",
26800 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26803 + case AuWbrCreate_PMFSV:
26804 + seq_printf(m, /*pat*/"pmfs:%lu",
26805 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26808 + case AuWbrCreate_MFSRR:
26809 + seq_printf(m, /*pat*/"mfsrr:%llu",
26810 + sbinfo->si_wbr_mfs.mfsrr_watermark);
26812 + case AuWbrCreate_MFSRRV:
26813 + seq_printf(m, /*pat*/"mfsrr:%llu:%lu",
26814 + sbinfo->si_wbr_mfs.mfsrr_watermark,
26815 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26818 + case AuWbrCreate_PMFSRR:
26819 + seq_printf(m, /*pat*/"pmfsrr:%llu",
26820 + sbinfo->si_wbr_mfs.mfsrr_watermark);
26822 + case AuWbrCreate_PMFSRRV:
26823 + seq_printf(m, /*pat*/"pmfsrr:%llu:%lu",
26824 + sbinfo->si_wbr_mfs.mfsrr_watermark,
26825 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26831 +static int au_show_xino(struct seq_file *seq, struct super_block *sb)
26833 +#ifdef CONFIG_SYSFS
26837 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
26838 + aufs_bindex_t bindex, brid;
26839 + struct qstr *name;
26841 + struct dentry *d, *h_root;
26842 + struct au_hdentry *hdp;
26844 + AuRwMustAnyLock(&sbinfo->si_rwsem);
26847 + f = au_sbi(sb)->si_xib;
26851 + /* stop printing the default xino path on the first writable branch */
26853 + brid = au_xino_brid(sb);
26855 + bindex = au_br_index(sb, brid);
26856 + hdp = au_di(sb->s_root)->di_hdentry;
26857 + h_root = hdp[0 + bindex].hd_dentry;
26859 + d = f->f_path.dentry;
26860 + name = &d->d_name;
26861 + /* safe ->d_parent because the file is unlinked */
26862 + if (d->d_parent == h_root
26863 + && name->len == len
26864 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
26867 + seq_puts(seq, ",xino=");
26868 + err = au_xino_path(seq, f);
26875 +/* seq_file will re-call me in case of too long string */
26876 +static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
26879 + unsigned int mnt_flags, v;
26880 + struct super_block *sb;
26881 + struct au_sbinfo *sbinfo;
26883 +#define AuBool(name, str) do { \
26884 + v = au_opt_test(mnt_flags, name); \
26885 + if (v != au_opt_test(AuOpt_Def, name)) \
26886 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
26889 +#define AuStr(name, str) do { \
26890 + v = mnt_flags & AuOptMask_##name; \
26891 + if (v != (AuOpt_Def & AuOptMask_##name)) \
26892 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
26895 +#define AuUInt(name, str, val) do { \
26896 + if (val != AUFS_##name##_DEF) \
26897 + seq_printf(m, "," #str "=%u", val); \
26900 + sb = dentry->d_sb;
26901 + if (sb->s_flags & MS_POSIXACL)
26902 + seq_puts(m, ",acl");
26904 + /* lock free root dinfo */
26905 + si_noflush_read_lock(sb);
26906 + sbinfo = au_sbi(sb);
26907 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
26909 + mnt_flags = au_mntflags(sb);
26910 + if (au_opt_test(mnt_flags, XINO)) {
26911 + err = au_show_xino(m, sb);
26912 + if (unlikely(err))
26915 + seq_puts(m, ",noxino");
26917 + AuBool(TRUNC_XINO, trunc_xino);
26918 + AuStr(UDBA, udba);
26919 + AuBool(SHWH, shwh);
26920 + AuBool(PLINK, plink);
26921 + AuBool(DIO, dio);
26922 + AuBool(DIRPERM1, dirperm1);
26924 + v = sbinfo->si_wbr_create;
26925 + if (v != AuWbrCreate_Def)
26926 + au_show_wbr_create(m, v, sbinfo);
26928 + v = sbinfo->si_wbr_copyup;
26929 + if (v != AuWbrCopyup_Def)
26930 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
26932 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
26933 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
26934 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
26936 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
26938 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
26939 + AuUInt(RDCACHE, rdcache, v);
26941 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
26942 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
26944 + au_fhsm_show(m, sbinfo);
26946 + AuBool(SUM, sum);
26947 + /* AuBool(SUM_W, wsum); */
26948 + AuBool(WARN_PERM, warn_perm);
26949 + AuBool(VERBOSE, verbose);
26952 + /* be sure to print "br:" last */
26953 + if (!sysaufs_brs) {
26954 + seq_puts(m, ",br:");
26955 + au_show_brs(m, sb);
26957 + si_read_unlock(sb);
26965 +/* ---------------------------------------------------------------------- */
26967 +/* sum mode which returns the summation for statfs(2) */
26969 +static u64 au_add_till_max(u64 a, u64 b)
26977 + return ULLONG_MAX;
26980 +static u64 au_mul_till_max(u64 a, long mul)
26988 + return ULLONG_MAX;
26991 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
26994 + long bsize, factor;
26995 + u64 blocks, bfree, bavail, files, ffree;
26996 + aufs_bindex_t bend, bindex, i;
26997 + unsigned char shared;
26998 + struct path h_path;
26999 + struct super_block *h_sb;
27002 + bsize = LONG_MAX;
27008 + bend = au_sbend(sb);
27009 + for (bindex = 0; bindex <= bend; bindex++) {
27010 + h_path.mnt = au_sbr_mnt(sb, bindex);
27011 + h_sb = h_path.mnt->mnt_sb;
27013 + for (i = 0; !shared && i < bindex; i++)
27014 + shared = (au_sbr_sb(sb, i) == h_sb);
27018 + /* sb->s_root for NFS is unreliable */
27019 + h_path.dentry = h_path.mnt->mnt_root;
27020 + err = vfs_statfs(&h_path, buf);
27021 + if (unlikely(err))
27024 + if (bsize > buf->f_bsize) {
27026 + * we will reduce bsize, so we have to expand blocks
27027 + * etc. to match them again
27029 + factor = (bsize / buf->f_bsize);
27030 + blocks = au_mul_till_max(blocks, factor);
27031 + bfree = au_mul_till_max(bfree, factor);
27032 + bavail = au_mul_till_max(bavail, factor);
27033 + bsize = buf->f_bsize;
27036 + factor = (buf->f_bsize / bsize);
27037 + blocks = au_add_till_max(blocks,
27038 + au_mul_till_max(buf->f_blocks, factor));
27039 + bfree = au_add_till_max(bfree,
27040 + au_mul_till_max(buf->f_bfree, factor));
27041 + bavail = au_add_till_max(bavail,
27042 + au_mul_till_max(buf->f_bavail, factor));
27043 + files = au_add_till_max(files, buf->f_files);
27044 + ffree = au_add_till_max(ffree, buf->f_ffree);
27047 + buf->f_bsize = bsize;
27048 + buf->f_blocks = blocks;
27049 + buf->f_bfree = bfree;
27050 + buf->f_bavail = bavail;
27051 + buf->f_files = files;
27052 + buf->f_ffree = ffree;
27053 + buf->f_frsize = 0;
27059 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
27062 + struct path h_path;
27063 + struct super_block *sb;
27065 + /* lock free root dinfo */
27066 + sb = dentry->d_sb;
27067 + si_noflush_read_lock(sb);
27068 + if (!au_opt_test(au_mntflags(sb), SUM)) {
27069 + /* sb->s_root for NFS is unreliable */
27070 + h_path.mnt = au_sbr_mnt(sb, 0);
27071 + h_path.dentry = h_path.mnt->mnt_root;
27072 + err = vfs_statfs(&h_path, buf);
27074 + err = au_statfs_sum(sb, buf);
27075 + si_read_unlock(sb);
27078 + buf->f_type = AUFS_SUPER_MAGIC;
27079 + buf->f_namelen = AUFS_MAX_NAMELEN;
27080 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
27082 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
27087 +/* ---------------------------------------------------------------------- */
27089 +static int aufs_sync_fs(struct super_block *sb, int wait)
27092 + aufs_bindex_t bend, bindex;
27093 + struct au_branch *br;
27094 + struct super_block *h_sb;
27097 + si_noflush_read_lock(sb);
27098 + bend = au_sbend(sb);
27099 + for (bindex = 0; bindex <= bend; bindex++) {
27100 + br = au_sbr(sb, bindex);
27101 + if (!au_br_writable(br->br_perm))
27104 + h_sb = au_sbr_sb(sb, bindex);
27105 + if (h_sb->s_op->sync_fs) {
27106 + e = h_sb->s_op->sync_fs(h_sb, wait);
27107 + if (unlikely(e && !err))
27109 + /* go on even if an error happens */
27112 + si_read_unlock(sb);
27117 +/* ---------------------------------------------------------------------- */
27119 +/* final actions when unmounting a file system */
27120 +static void aufs_put_super(struct super_block *sb)
27122 + struct au_sbinfo *sbinfo;
27124 + sbinfo = au_sbi(sb);
27128 + dbgaufs_si_fin(sbinfo);
27129 + kobject_put(&sbinfo->si_kobj);
27132 +/* ---------------------------------------------------------------------- */
27134 +void au_array_free(void *array)
27137 + if (!is_vmalloc_addr(array))
27144 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
27145 + struct super_block *sb, void *arg)
27148 + unsigned long long n, sz;
27155 + if (*hint > ULLONG_MAX / sizeof(array)) {
27156 + array = ERR_PTR(-EMFILE);
27157 + pr_err("hint %llu\n", *hint);
27161 + sz = sizeof(array) * *hint;
27162 + array = kzalloc(sz, GFP_NOFS);
27163 + if (unlikely(!array))
27164 + array = vzalloc(sz);
27165 + if (unlikely(!array)) {
27166 + array = ERR_PTR(-ENOMEM);
27170 + n = cb(sb, array, *hint, arg);
27171 + AuDebugOn(n > *hint);
27178 +static unsigned long long au_iarray_cb(struct super_block *sb, void *a,
27179 + unsigned long long max __maybe_unused,
27182 + unsigned long long n;
27183 + struct inode **p, *inode;
27184 + struct list_head *head;
27189 + spin_lock(&sb->s_inode_list_lock);
27190 + list_for_each_entry(inode, head, i_sb_list) {
27191 + if (!is_bad_inode(inode)
27192 + && au_ii(inode)->ii_bstart >= 0) {
27193 + spin_lock(&inode->i_lock);
27194 + if (atomic_read(&inode->i_count)) {
27198 + AuDebugOn(n > max);
27200 + spin_unlock(&inode->i_lock);
27203 + spin_unlock(&sb->s_inode_list_lock);
27208 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
27210 + *max = atomic_long_read(&au_sbi(sb)->si_ninodes);
27211 + return au_array_alloc(max, au_iarray_cb, sb, &sb->s_inodes);
27214 +void au_iarray_free(struct inode **a, unsigned long long max)
27216 + unsigned long long ull;
27218 + for (ull = 0; ull < max; ull++)
27220 + au_array_free(a);
27223 +/* ---------------------------------------------------------------------- */
27226 + * refresh dentry and inode at remount time.
27228 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
27229 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
27230 + struct dentry *parent)
27234 + di_write_lock_child(dentry);
27235 + di_read_lock_parent(parent, AuLock_IR);
27236 + err = au_refresh_dentry(dentry, parent);
27237 + if (!err && dir_flags)
27238 + au_hn_reset(d_inode(dentry), dir_flags);
27239 + di_read_unlock(parent, AuLock_IR);
27240 + di_write_unlock(dentry);
27245 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
27246 + struct au_sbinfo *sbinfo,
27247 + const unsigned int dir_flags, unsigned int do_dop)
27250 + struct dentry *parent;
27253 + parent = dget_parent(dentry);
27254 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
27255 + if (d_really_is_positive(dentry)) {
27256 + if (!d_is_dir(dentry))
27257 + err = au_do_refresh(dentry, /*dir_flags*/0,
27260 + err = au_do_refresh(dentry, dir_flags, parent);
27261 + if (unlikely(err))
27262 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
27265 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
27266 + AuDbgDentry(dentry);
27272 + au_refresh_dop(dentry, /*force_reval*/0);
27274 + au_refresh_dop(dentry, /*force_reval*/1);
27280 +static int au_refresh_d(struct super_block *sb, unsigned int do_dop)
27282 + int err, i, j, ndentry, e;
27283 + unsigned int sigen;
27284 + struct au_dcsub_pages dpages;
27285 + struct au_dpage *dpage;
27286 + struct dentry **dentries, *d;
27287 + struct au_sbinfo *sbinfo;
27288 + struct dentry *root = sb->s_root;
27289 + const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
27292 + au_refresh_dop(root, /*force_reval*/0);
27294 + err = au_dpages_init(&dpages, GFP_NOFS);
27295 + if (unlikely(err))
27297 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
27298 + if (unlikely(err))
27301 + sigen = au_sigen(sb);
27302 + sbinfo = au_sbi(sb);
27303 + for (i = 0; i < dpages.ndpage; i++) {
27304 + dpage = dpages.dpages + i;
27305 + dentries = dpage->dentries;
27306 + ndentry = dpage->ndentry;
27307 + for (j = 0; j < ndentry; j++) {
27309 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags,
27311 + if (unlikely(e && !err))
27313 + /* go on even err */
27318 + au_dpages_free(&dpages);
27323 +static int au_refresh_i(struct super_block *sb)
27326 + unsigned int sigen;
27327 + unsigned long long max, ull;
27328 + struct inode *inode, **array;
27330 + array = au_iarray_alloc(sb, &max);
27331 + err = PTR_ERR(array);
27332 + if (IS_ERR(array))
27336 + sigen = au_sigen(sb);
27337 + for (ull = 0; ull < max; ull++) {
27338 + inode = array[ull];
27339 + if (unlikely(!inode))
27341 + if (au_iigen(inode, NULL) != sigen) {
27342 + ii_write_lock_child(inode);
27343 + e = au_refresh_hinode_self(inode);
27344 + ii_write_unlock(inode);
27345 + if (unlikely(e)) {
27346 + pr_err("error %d, i%lu\n", e, inode->i_ino);
27349 + /* go on even if err */
27354 + au_iarray_free(array, max);
27360 +static void au_remount_refresh(struct super_block *sb, unsigned int do_dop)
27363 + unsigned int udba;
27364 + aufs_bindex_t bindex, bend;
27365 + struct dentry *root;
27366 + struct inode *inode;
27367 + struct au_branch *br;
27368 + struct au_sbinfo *sbi;
27370 + au_sigen_inc(sb);
27371 + sbi = au_sbi(sb);
27372 + au_fclr_si(sbi, FAILED_REFRESH_DIR);
27374 + root = sb->s_root;
27375 + DiMustNoWaiters(root);
27376 + inode = d_inode(root);
27377 + IiMustNoWaiters(inode);
27379 + udba = au_opt_udba(sb);
27380 + bend = au_sbend(sb);
27381 + for (bindex = 0; bindex <= bend; bindex++) {
27382 + br = au_sbr(sb, bindex);
27383 + err = au_hnotify_reset_br(udba, br, br->br_perm);
27384 + if (unlikely(err))
27385 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
27387 + /* go on even if err */
27389 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
27392 + if (au_ftest_si(sbi, NO_DREVAL)) {
27393 + AuDebugOn(sb->s_d_op == &aufs_dop_noreval);
27394 + sb->s_d_op = &aufs_dop_noreval;
27396 + AuDebugOn(sb->s_d_op == &aufs_dop);
27397 + sb->s_d_op = &aufs_dop;
27399 + pr_info("reset to %pf\n", sb->s_d_op);
27402 + di_write_unlock(root);
27403 + err = au_refresh_d(sb, do_dop);
27404 + e = au_refresh_i(sb);
27405 + if (unlikely(e && !err))
27407 + /* aufs_write_lock() calls ..._child() */
27408 + di_write_lock_child(root);
27410 + au_cpup_attr_all(inode, /*force*/1);
27412 + if (unlikely(err))
27413 + AuIOErr("refresh failed, ignored, %d\n", err);
27416 +/* stop extra interpretation of errno in mount(8), and strange error messages */
27417 +static int cvt_err(int err)
27431 +static int aufs_remount_fs(struct super_block *sb, int *flags, char *data)
27434 + unsigned int mntflags;
27435 + struct au_opts opts;
27436 + struct dentry *root;
27437 + struct inode *inode;
27438 + struct au_sbinfo *sbinfo;
27441 + root = sb->s_root;
27442 + if (!data || !*data) {
27443 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
27445 + di_write_lock_child(root);
27446 + err = au_opts_verify(sb, *flags, /*pending*/0);
27447 + aufs_write_unlock(root);
27453 + memset(&opts, 0, sizeof(opts));
27454 + opts.opt = (void *)__get_free_page(GFP_NOFS);
27455 + if (unlikely(!opts.opt))
27457 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
27458 + opts.flags = AuOpts_REMOUNT;
27459 + opts.sb_flags = *flags;
27461 + /* parse it before aufs lock */
27462 + err = au_opts_parse(sb, data, &opts);
27463 + if (unlikely(err))
27466 + sbinfo = au_sbi(sb);
27467 + inode = d_inode(root);
27468 + mutex_lock(&inode->i_mutex);
27469 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
27470 + if (unlikely(err))
27472 + di_write_lock_child(root);
27474 + /* au_opts_remount() may return an error */
27475 + err = au_opts_remount(sb, &opts);
27476 + au_opts_free(&opts);
27478 + if (au_ftest_opts(opts.flags, REFRESH))
27479 + au_remount_refresh(sb, au_ftest_opts(opts.flags, REFRESH_DOP));
27481 + if (au_ftest_opts(opts.flags, REFRESH_DYAOP)) {
27482 + mntflags = au_mntflags(sb);
27483 + do_dx = !!au_opt_test(mntflags, DIO);
27484 + au_dy_arefresh(do_dx);
27487 + au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
27488 + aufs_write_unlock(root);
27491 + mutex_unlock(&inode->i_mutex);
27493 + free_page((unsigned long)opts.opt);
27495 + err = cvt_err(err);
27500 +static const struct super_operations aufs_sop = {
27501 + .alloc_inode = aufs_alloc_inode,
27502 + .destroy_inode = aufs_destroy_inode,
27503 + /* always deleting, no clearing */
27504 + .drop_inode = generic_delete_inode,
27505 + .show_options = aufs_show_options,
27506 + .statfs = aufs_statfs,
27507 + .put_super = aufs_put_super,
27508 + .sync_fs = aufs_sync_fs,
27509 + .remount_fs = aufs_remount_fs
27512 +/* ---------------------------------------------------------------------- */
27514 +static int alloc_root(struct super_block *sb)
27517 + struct inode *inode;
27518 + struct dentry *root;
27521 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
27522 + err = PTR_ERR(inode);
27523 + if (IS_ERR(inode))
27526 + inode->i_op = &aufs_dir_iop;
27527 + inode->i_fop = &aufs_dir_fop;
27528 + inode->i_mode = S_IFDIR;
27529 + set_nlink(inode, 2);
27530 + unlock_new_inode(inode);
27532 + root = d_make_root(inode);
27533 + if (unlikely(!root))
27535 + err = PTR_ERR(root);
27536 + if (IS_ERR(root))
27539 + err = au_di_init(root);
27541 + sb->s_root = root;
27542 + return 0; /* success */
27550 +static int aufs_fill_super(struct super_block *sb, void *raw_data,
27551 + int silent __maybe_unused)
27554 + struct au_opts opts;
27555 + struct au_sbinfo *sbinfo;
27556 + struct dentry *root;
27557 + struct inode *inode;
27558 + char *arg = raw_data;
27560 + if (unlikely(!arg || !*arg)) {
27562 + pr_err("no arg\n");
27567 + memset(&opts, 0, sizeof(opts));
27568 + opts.opt = (void *)__get_free_page(GFP_NOFS);
27569 + if (unlikely(!opts.opt))
27571 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
27572 + opts.sb_flags = sb->s_flags;
27574 + err = au_si_alloc(sb);
27575 + if (unlikely(err))
27577 + sbinfo = au_sbi(sb);
27579 + /* all timestamps always follow the ones on the branch */
27580 + sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
27581 + sb->s_op = &aufs_sop;
27582 + sb->s_d_op = &aufs_dop;
27583 + sb->s_magic = AUFS_SUPER_MAGIC;
27584 + sb->s_maxbytes = 0;
27585 + sb->s_stack_depth = 1;
27586 + au_export_init(sb);
27587 + /* au_xattr_init(sb); */
27589 + err = alloc_root(sb);
27590 + if (unlikely(err)) {
27591 + si_write_unlock(sb);
27594 + root = sb->s_root;
27595 + inode = d_inode(root);
27598 + * actually we can parse options regardless aufs lock here.
27599 + * but at remount time, parsing must be done before aufs lock.
27600 + * so we follow the same rule.
27602 + ii_write_lock_parent(inode);
27603 + aufs_write_unlock(root);
27604 + err = au_opts_parse(sb, arg, &opts);
27605 + if (unlikely(err))
27608 + /* lock vfs_inode first, then aufs. */
27609 + mutex_lock(&inode->i_mutex);
27610 + aufs_write_lock(root);
27611 + err = au_opts_mount(sb, &opts);
27612 + au_opts_free(&opts);
27613 + if (!err && au_ftest_si(sbinfo, NO_DREVAL)) {
27614 + sb->s_d_op = &aufs_dop_noreval;
27615 + pr_info("%pf\n", sb->s_d_op);
27616 + au_refresh_dop(root, /*force_reval*/0);
27618 + aufs_write_unlock(root);
27619 + mutex_unlock(&inode->i_mutex);
27621 + goto out_opts; /* success */
27625 + sb->s_root = NULL;
27627 + dbgaufs_si_fin(sbinfo);
27628 + kobject_put(&sbinfo->si_kobj);
27629 + sb->s_fs_info = NULL;
27631 + free_page((unsigned long)opts.opt);
27634 + err = cvt_err(err);
27639 +/* ---------------------------------------------------------------------- */
27641 +static struct dentry *aufs_mount(struct file_system_type *fs_type, int flags,
27642 + const char *dev_name __maybe_unused,
27645 + struct dentry *root;
27646 + struct super_block *sb;
27648 + /* all timestamps always follow the ones on the branch */
27649 + /* mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; */
27650 + root = mount_nodev(fs_type, flags, raw_data, aufs_fill_super);
27651 + if (IS_ERR(root))
27655 + si_write_lock(sb, !AuLock_FLUSH);
27656 + sysaufs_brs_add(sb, 0);
27657 + si_write_unlock(sb);
27658 + au_sbilist_add(sb);
27664 +static void aufs_kill_sb(struct super_block *sb)
27666 + struct au_sbinfo *sbinfo;
27668 + sbinfo = au_sbi(sb);
27670 + au_sbilist_del(sb);
27671 + aufs_write_lock(sb->s_root);
27673 + if (sbinfo->si_wbr_create_ops->fin)
27674 + sbinfo->si_wbr_create_ops->fin(sb);
27675 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
27676 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
27677 + au_remount_refresh(sb, /*do_dop*/0);
27679 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
27680 + au_plink_put(sb, /*verbose*/1);
27682 + sbinfo->si_sb = NULL;
27683 + aufs_write_unlock(sb->s_root);
27684 + au_nwt_flush(&sbinfo->si_nowait);
27686 + kill_anon_super(sb);
27689 +struct file_system_type aufs_fs_type = {
27690 + .name = AUFS_FSTYPE,
27691 + /* a race between rename and others */
27692 + .fs_flags = FS_RENAME_DOES_D_MOVE,
27693 + .mount = aufs_mount,
27694 + .kill_sb = aufs_kill_sb,
27695 + /* no need to __module_get() and module_put(). */
27696 + .owner = THIS_MODULE,
27698 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
27699 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
27700 +++ linux/fs/aufs/super.h 2015-11-11 17:21:46.922197217 +0100
27703 + * Copyright (C) 2005-2015 Junjiro R. Okajima
27705 + * This program, aufs is free software; you can redistribute it and/or modify
27706 + * it under the terms of the GNU General Public License as published by
27707 + * the Free Software Foundation; either version 2 of the License, or
27708 + * (at your option) any later version.
27710 + * This program is distributed in the hope that it will be useful,
27711 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27712 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27713 + * GNU General Public License for more details.
27715 + * You should have received a copy of the GNU General Public License
27716 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
27720 + * super_block operations
27723 +#ifndef __AUFS_SUPER_H__
27724 +#define __AUFS_SUPER_H__
27728 +#include <linux/fs.h>
27729 +#include <linux/kobject.h>
27730 +#include "rwsem.h"
27734 +/* policies to select one among multiple writable branches */
27735 +struct au_wbr_copyup_operations {
27736 + int (*copyup)(struct dentry *dentry);
27739 +#define AuWbr_DIR 1 /* target is a dir */
27740 +#define AuWbr_PARENT (1 << 1) /* always require a parent */
27742 +#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
27743 +#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
27744 +#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
27746 +struct au_wbr_create_operations {
27747 + int (*create)(struct dentry *dentry, unsigned int flags);
27748 + int (*init)(struct super_block *sb);
27749 + int (*fin)(struct super_block *sb);
27752 +struct au_wbr_mfs {
27753 + struct mutex mfs_lock; /* protect this structure */
27754 + unsigned long mfs_jiffy;
27755 + unsigned long mfs_expire;
27756 + aufs_bindex_t mfs_bindex;
27758 + unsigned long long mfsrr_bytes;
27759 + unsigned long long mfsrr_watermark;
27762 +struct pseudo_link {
27764 + struct hlist_node hlist;
27765 + struct rcu_head rcu;
27767 + struct inode *inode;
27770 +#define AuPlink_NHASH 100
27771 +static inline int au_plink_hash(ino_t ino)
27773 + return ino % AuPlink_NHASH;
27776 +/* File-based Hierarchical Storage Management */
27778 +#ifdef CONFIG_AUFS_FHSM
27779 + /* allow only one process who can receive the notification */
27780 + spinlock_t fhsm_spin;
27782 + wait_queue_head_t fhsm_wqh;
27783 + atomic_t fhsm_readable;
27785 + /* these are protected by si_rwsem */
27786 + unsigned long fhsm_expire;
27787 + aufs_bindex_t fhsm_bottom;
27792 +struct au_sbinfo {
27793 + /* nowait tasks in the system-wide workqueue */
27794 + struct au_nowait_tasks si_nowait;
27797 + * tried sb->s_umount, but failed due to the dependecy between i_mutex.
27798 + * rwsem for au_sbinfo is necessary.
27800 + struct au_rwsem si_rwsem;
27802 + /* prevent recursive locking in deleting inode */
27804 + unsigned long *bitmap;
27805 + spinlock_t tree_lock;
27806 + struct radix_tree_root tree;
27810 + * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
27813 + atomic_long_t si_ninodes, si_nfiles;
27815 + /* branch management */
27816 + unsigned int si_generation;
27818 + /* see AuSi_ flags */
27819 + unsigned char au_si_status;
27821 + aufs_bindex_t si_bend;
27823 + /* dirty trick to keep br_id plus */
27824 + unsigned int si_last_br_id :
27825 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
27826 + struct au_branch **si_branch;
27828 + /* policy to select a writable branch */
27829 + unsigned char si_wbr_copyup;
27830 + unsigned char si_wbr_create;
27831 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
27832 + struct au_wbr_create_operations *si_wbr_create_ops;
27834 + /* round robin */
27835 + atomic_t si_wbr_rr_next;
27837 + /* most free space */
27838 + struct au_wbr_mfs si_wbr_mfs;
27840 + /* File-based Hierarchical Storage Management */
27841 + struct au_fhsm si_fhsm;
27843 + /* mount flags */
27844 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
27845 + unsigned int si_mntflags;
27847 + /* symlink to follow_link() and put_link() */
27848 + struct au_sphlhead si_symlink;
27850 + /* external inode number (bitmap and translation table) */
27851 + vfs_readf_t si_xread;
27852 + vfs_writef_t si_xwrite;
27853 + struct file *si_xib;
27854 + struct mutex si_xib_mtx; /* protect xib members */
27855 + unsigned long *si_xib_buf;
27856 + unsigned long si_xib_last_pindex;
27857 + int si_xib_next_bit;
27858 + aufs_bindex_t si_xino_brid;
27859 + unsigned long si_xino_jiffy;
27860 + unsigned long si_xino_expire;
27861 + /* reserved for future use */
27862 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
27864 +#ifdef CONFIG_AUFS_EXPORT
27865 + /* i_generation */
27866 + struct file *si_xigen;
27867 + atomic_t si_xigen_next;
27870 + /* dirty trick to suppoer atomic_open */
27871 + struct au_sphlhead si_aopen;
27873 + /* vdir parameters */
27874 + unsigned long si_rdcache; /* max cache time in jiffies */
27875 + unsigned int si_rdblk; /* deblk size */
27876 + unsigned int si_rdhash; /* hash size */
27879 + * If the number of whiteouts are larger than si_dirwh, leave all of
27880 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
27881 + * future fsck.aufs or kernel thread will remove them later.
27882 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
27884 + unsigned int si_dirwh;
27886 + /* pseudo_link list */
27887 + struct au_sphlhead si_plink[AuPlink_NHASH];
27888 + wait_queue_head_t si_plink_wq;
27889 + spinlock_t si_plink_maint_lock;
27890 + pid_t si_plink_maint_pid;
27893 + struct au_sphlhead si_files;
27896 + * sysfs and lifetime management.
27897 + * this is not a small structure and it may be a waste of memory in case
27898 + * of sysfs is disabled, particulary when many aufs-es are mounted.
27899 + * but using sysfs is majority.
27901 + struct kobject si_kobj;
27902 +#ifdef CONFIG_DEBUG_FS
27903 + struct dentry *si_dbgaufs;
27904 + struct dentry *si_dbgaufs_plink;
27905 + struct dentry *si_dbgaufs_xib;
27906 +#ifdef CONFIG_AUFS_EXPORT
27907 + struct dentry *si_dbgaufs_xigen;
27911 +#ifdef CONFIG_AUFS_SBILIST
27912 + struct list_head si_list;
27915 + /* dirty, necessary for unmounting, sysfs and sysrq */
27916 + struct super_block *si_sb;
27919 +/* sbinfo status flags */
27921 + * set true when refresh_dirs() failed at remount time.
27922 + * then try refreshing dirs at access time again.
27923 + * if it is false, refreshing dirs at access time is unnecesary
27925 +#define AuSi_FAILED_REFRESH_DIR 1
27927 +#define AuSi_FHSM (1 << 1) /* fhsm is active now */
27928 +#define AuSi_NO_DREVAL (1 << 2) /* disable all d_revalidate */
27930 +#ifndef CONFIG_AUFS_FHSM
27932 +#define AuSi_FHSM 0
27935 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
27936 + unsigned int flag)
27938 + AuRwMustAnyLock(&sbi->si_rwsem);
27939 + return sbi->au_si_status & flag;
27941 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
27942 +#define au_fset_si(sbinfo, name) do { \
27943 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
27944 + (sbinfo)->au_si_status |= AuSi_##name; \
27946 +#define au_fclr_si(sbinfo, name) do { \
27947 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
27948 + (sbinfo)->au_si_status &= ~AuSi_##name; \
27951 +/* ---------------------------------------------------------------------- */
27953 +/* policy to select one among writable branches */
27954 +#define AuWbrCopyup(sbinfo, ...) \
27955 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
27956 +#define AuWbrCreate(sbinfo, ...) \
27957 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
27959 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
27960 +#define AuLock_DW 1 /* write-lock dentry */
27961 +#define AuLock_IR (1 << 1) /* read-lock inode */
27962 +#define AuLock_IW (1 << 2) /* write-lock inode */
27963 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
27964 +#define AuLock_DIR (1 << 4) /* target is a dir */
27965 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
27966 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
27967 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
27968 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
27969 +#define au_fset_lock(flags, name) \
27970 + do { (flags) |= AuLock_##name; } while (0)
27971 +#define au_fclr_lock(flags, name) \
27972 + do { (flags) &= ~AuLock_##name; } while (0)
27974 +/* ---------------------------------------------------------------------- */
27977 +extern struct file_system_type aufs_fs_type;
27978 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
27979 +typedef unsigned long long (*au_arraycb_t)(struct super_block *sb, void *array,
27980 + unsigned long long max, void *arg);
27981 +void au_array_free(void *array);
27982 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb,
27983 + struct super_block *sb, void *arg);
27984 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
27985 +void au_iarray_free(struct inode **a, unsigned long long max);
27988 +void au_si_free(struct kobject *kobj);
27989 +int au_si_alloc(struct super_block *sb);
27990 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr);
27992 +unsigned int au_sigen_inc(struct super_block *sb);
27993 +aufs_bindex_t au_new_br_id(struct super_block *sb);
27995 +int si_read_lock(struct super_block *sb, int flags);
27996 +int si_write_lock(struct super_block *sb, int flags);
27997 +int aufs_read_lock(struct dentry *dentry, int flags);
27998 +void aufs_read_unlock(struct dentry *dentry, int flags);
27999 +void aufs_write_lock(struct dentry *dentry);
28000 +void aufs_write_unlock(struct dentry *dentry);
28001 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
28002 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
28004 +int si_pid_test_slow(struct super_block *sb);
28005 +void si_pid_set_slow(struct super_block *sb);
28006 +void si_pid_clr_slow(struct super_block *sb);
28008 +/* wbr_policy.c */
28009 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
28010 +extern struct au_wbr_create_operations au_wbr_create_ops[];
28011 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
28012 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
28013 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t bstart);
28016 +int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
28018 +#ifdef CONFIG_AUFS_FHSM
28021 +static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
28025 + spin_lock(&fhsm->fhsm_spin);
28026 + pid = fhsm->fhsm_pid;
28027 + spin_unlock(&fhsm->fhsm_spin);
28032 +void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
28033 +void au_fhsm_wrote_all(struct super_block *sb, int force);
28034 +int au_fhsm_fd(struct super_block *sb, int oflags);
28035 +int au_fhsm_br_alloc(struct au_branch *br);
28036 +void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
28037 +void au_fhsm_fin(struct super_block *sb);
28038 +void au_fhsm_init(struct au_sbinfo *sbinfo);
28039 +void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
28040 +void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
28042 +AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
28044 +AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
28045 +AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
28046 +AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
28047 +AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
28048 +AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
28049 +AuStubVoid(au_fhsm_fin, struct super_block *sb)
28050 +AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
28051 +AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
28052 +AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
28055 +/* ---------------------------------------------------------------------- */
28057 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
28059 + return sb->s_fs_info;
28062 +/* ---------------------------------------------------------------------- */
28064 +#ifdef CONFIG_AUFS_EXPORT
28065 +int au_test_nfsd(void);
28066 +void au_export_init(struct super_block *sb);
28067 +void au_xigen_inc(struct inode *inode);
28068 +int au_xigen_new(struct inode *inode);
28069 +int au_xigen_set(struct super_block *sb, struct file *base);
28070 +void au_xigen_clr(struct super_block *sb);
28072 +static inline int au_busy_or_stale(void)
28074 + if (!au_test_nfsd())
28079 +AuStubInt0(au_test_nfsd, void)
28080 +AuStubVoid(au_export_init, struct super_block *sb)
28081 +AuStubVoid(au_xigen_inc, struct inode *inode)
28082 +AuStubInt0(au_xigen_new, struct inode *inode)
28083 +AuStubInt0(au_xigen_set, struct super_block *sb, struct file *base)
28084 +AuStubVoid(au_xigen_clr, struct super_block *sb)
28085 +AuStub(int, au_busy_or_stale, return -EBUSY, void)
28086 +#endif /* CONFIG_AUFS_EXPORT */
28088 +/* ---------------------------------------------------------------------- */
28090 +#ifdef CONFIG_AUFS_SBILIST
28092 +extern struct au_splhead au_sbilist;
28094 +static inline void au_sbilist_init(void)
28096 + au_spl_init(&au_sbilist);
28099 +static inline void au_sbilist_add(struct super_block *sb)
28101 + au_spl_add(&au_sbi(sb)->si_list, &au_sbilist);
28104 +static inline void au_sbilist_del(struct super_block *sb)
28106 + au_spl_del(&au_sbi(sb)->si_list, &au_sbilist);
28109 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
28110 +static inline void au_sbilist_lock(void)
28112 + spin_lock(&au_sbilist.spin);
28115 +static inline void au_sbilist_unlock(void)
28117 + spin_unlock(&au_sbilist.spin);
28119 +#define AuGFP_SBILIST GFP_ATOMIC
28121 +AuStubVoid(au_sbilist_lock, void)
28122 +AuStubVoid(au_sbilist_unlock, void)
28123 +#define AuGFP_SBILIST GFP_NOFS
28124 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
28126 +AuStubVoid(au_sbilist_init, void)
28127 +AuStubVoid(au_sbilist_add, struct super_block *sb)
28128 +AuStubVoid(au_sbilist_del, struct super_block *sb)
28129 +AuStubVoid(au_sbilist_lock, void)
28130 +AuStubVoid(au_sbilist_unlock, void)
28131 +#define AuGFP_SBILIST GFP_NOFS
28134 +/* ---------------------------------------------------------------------- */
28136 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
28139 + * This function is a dynamic '__init' function actually,
28140 + * so the tiny check for si_rwsem is unnecessary.
28142 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
28143 +#ifdef CONFIG_DEBUG_FS
28144 + sbinfo->si_dbgaufs = NULL;
28145 + sbinfo->si_dbgaufs_plink = NULL;
28146 + sbinfo->si_dbgaufs_xib = NULL;
28147 +#ifdef CONFIG_AUFS_EXPORT
28148 + sbinfo->si_dbgaufs_xigen = NULL;
28153 +/* ---------------------------------------------------------------------- */
28155 +static inline pid_t si_pid_bit(void)
28157 + /* the origin of pid is 1, but the bitmap's is 0 */
28158 + return current->pid - 1;
28161 +static inline int si_pid_test(struct super_block *sb)
28165 + bit = si_pid_bit();
28166 + if (bit < PID_MAX_DEFAULT)
28167 + return test_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28168 + return si_pid_test_slow(sb);
28171 +static inline void si_pid_set(struct super_block *sb)
28175 + bit = si_pid_bit();
28176 + if (bit < PID_MAX_DEFAULT) {
28177 + AuDebugOn(test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
28178 + set_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28181 + si_pid_set_slow(sb);
28184 +static inline void si_pid_clr(struct super_block *sb)
28188 + bit = si_pid_bit();
28189 + if (bit < PID_MAX_DEFAULT) {
28190 + AuDebugOn(!test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
28191 + clear_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28194 + si_pid_clr_slow(sb);
28197 +/* ---------------------------------------------------------------------- */
28199 +/* lock superblock. mainly for entry point functions */
28201 + * __si_read_lock, __si_write_lock,
28202 + * __si_read_unlock, __si_write_unlock, __si_downgrade_lock
28204 +AuSimpleRwsemFuncs(__si, struct super_block *sb, &au_sbi(sb)->si_rwsem);
28206 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
28207 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
28208 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
28210 +static inline void si_noflush_read_lock(struct super_block *sb)
28212 + __si_read_lock(sb);
28216 +static inline int si_noflush_read_trylock(struct super_block *sb)
28220 + locked = __si_read_trylock(sb);
28226 +static inline void si_noflush_write_lock(struct super_block *sb)
28228 + __si_write_lock(sb);
28232 +static inline int si_noflush_write_trylock(struct super_block *sb)
28236 + locked = __si_write_trylock(sb);
28242 +#if 0 /* reserved */
28243 +static inline int si_read_trylock(struct super_block *sb, int flags)
28245 + if (au_ftest_lock(flags, FLUSH))
28246 + au_nwt_flush(&au_sbi(sb)->si_nowait);
28247 + return si_noflush_read_trylock(sb);
28251 +static inline void si_read_unlock(struct super_block *sb)
28254 + __si_read_unlock(sb);
28257 +#if 0 /* reserved */
28258 +static inline int si_write_trylock(struct super_block *sb, int flags)
28260 + if (au_ftest_lock(flags, FLUSH))
28261 + au_nwt_flush(&au_sbi(sb)->si_nowait);
28262 + return si_noflush_write_trylock(sb);
28266 +static inline void si_write_unlock(struct super_block *sb)
28269 + __si_write_unlock(sb);
28272 +#if 0 /* reserved */
28273 +static inline void si_downgrade_lock(struct super_block *sb)
28275 + __si_downgrade_lock(sb);
28279 +/* ---------------------------------------------------------------------- */
28281 +static inline aufs_bindex_t au_sbend(struct super_block *sb)
28283 + SiMustAnyLock(sb);
28284 + return au_sbi(sb)->si_bend;
28287 +static inline unsigned int au_mntflags(struct super_block *sb)
28289 + SiMustAnyLock(sb);
28290 + return au_sbi(sb)->si_mntflags;
28293 +static inline unsigned int au_sigen(struct super_block *sb)
28295 + SiMustAnyLock(sb);
28296 + return au_sbi(sb)->si_generation;
28299 +static inline void au_ninodes_inc(struct super_block *sb)
28301 + atomic_long_inc(&au_sbi(sb)->si_ninodes);
28304 +static inline void au_ninodes_dec(struct super_block *sb)
28306 + AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_ninodes));
28307 + atomic_long_dec(&au_sbi(sb)->si_ninodes);
28310 +static inline void au_nfiles_inc(struct super_block *sb)
28312 + atomic_long_inc(&au_sbi(sb)->si_nfiles);
28315 +static inline void au_nfiles_dec(struct super_block *sb)
28317 + AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_nfiles));
28318 + atomic_long_dec(&au_sbi(sb)->si_nfiles);
28321 +static inline struct au_branch *au_sbr(struct super_block *sb,
28322 + aufs_bindex_t bindex)
28324 + SiMustAnyLock(sb);
28325 + return au_sbi(sb)->si_branch[0 + bindex];
28328 +static inline void au_xino_brid_set(struct super_block *sb, aufs_bindex_t brid)
28330 + SiMustWriteLock(sb);
28331 + au_sbi(sb)->si_xino_brid = brid;
28334 +static inline aufs_bindex_t au_xino_brid(struct super_block *sb)
28336 + SiMustAnyLock(sb);
28337 + return au_sbi(sb)->si_xino_brid;
28340 +#endif /* __KERNEL__ */
28341 +#endif /* __AUFS_SUPER_H__ */
28342 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
28343 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
28344 +++ linux/fs/aufs/sysaufs.c 2015-09-24 10:47:58.254719746 +0200
28347 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28349 + * This program, aufs is free software; you can redistribute it and/or modify
28350 + * it under the terms of the GNU General Public License as published by
28351 + * the Free Software Foundation; either version 2 of the License, or
28352 + * (at your option) any later version.
28354 + * This program is distributed in the hope that it will be useful,
28355 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28356 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28357 + * GNU General Public License for more details.
28359 + * You should have received a copy of the GNU General Public License
28360 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28364 + * sysfs interface and lifetime management
28365 + * they are necessary regardless sysfs is disabled.
28368 +#include <linux/random.h>
28371 +unsigned long sysaufs_si_mask;
28372 +struct kset *sysaufs_kset;
28374 +#define AuSiAttr(_name) { \
28375 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
28376 + .show = sysaufs_si_##_name, \
28379 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
28380 +struct attribute *sysaufs_si_attrs[] = {
28381 + &sysaufs_si_attr_xi_path.attr,
28385 +static const struct sysfs_ops au_sbi_ops = {
28386 + .show = sysaufs_si_show
28389 +static struct kobj_type au_sbi_ktype = {
28390 + .release = au_si_free,
28391 + .sysfs_ops = &au_sbi_ops,
28392 + .default_attrs = sysaufs_si_attrs
28395 +/* ---------------------------------------------------------------------- */
28397 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
28401 + sbinfo->si_kobj.kset = sysaufs_kset;
28402 + /* cf. sysaufs_name() */
28403 + err = kobject_init_and_add
28404 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
28405 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
28407 + dbgaufs_si_null(sbinfo);
28409 + err = dbgaufs_si_init(sbinfo);
28410 + if (unlikely(err))
28411 + kobject_put(&sbinfo->si_kobj);
28416 +void sysaufs_fin(void)
28419 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
28420 + kset_unregister(sysaufs_kset);
28423 +int __init sysaufs_init(void)
28428 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
28429 + } while (!sysaufs_si_mask);
28432 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
28433 + if (unlikely(!sysaufs_kset))
28435 + err = PTR_ERR(sysaufs_kset);
28436 + if (IS_ERR(sysaufs_kset))
28438 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
28439 + if (unlikely(err)) {
28440 + kset_unregister(sysaufs_kset);
28444 + err = dbgaufs_init();
28445 + if (unlikely(err))
28450 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
28451 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
28452 +++ linux/fs/aufs/sysaufs.h 2015-09-24 10:47:58.254719746 +0200
28455 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28457 + * This program, aufs is free software; you can redistribute it and/or modify
28458 + * it under the terms of the GNU General Public License as published by
28459 + * the Free Software Foundation; either version 2 of the License, or
28460 + * (at your option) any later version.
28462 + * This program is distributed in the hope that it will be useful,
28463 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28464 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28465 + * GNU General Public License for more details.
28467 + * You should have received a copy of the GNU General Public License
28468 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28472 + * sysfs interface and mount lifetime management
28475 +#ifndef __SYSAUFS_H__
28476 +#define __SYSAUFS_H__
28480 +#include <linux/sysfs.h>
28481 +#include "module.h"
28483 +struct super_block;
28486 +struct sysaufs_si_attr {
28487 + struct attribute attr;
28488 + int (*show)(struct seq_file *seq, struct super_block *sb);
28491 +/* ---------------------------------------------------------------------- */
28494 +extern unsigned long sysaufs_si_mask;
28495 +extern struct kset *sysaufs_kset;
28496 +extern struct attribute *sysaufs_si_attrs[];
28497 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
28498 +int __init sysaufs_init(void);
28499 +void sysaufs_fin(void);
28501 +/* ---------------------------------------------------------------------- */
28503 +/* some people doesn't like to show a pointer in kernel */
28504 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
28506 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
28509 +#define SysaufsSiNamePrefix "si_"
28510 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
28511 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
28513 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
28514 + sysaufs_si_id(sbinfo));
28518 +#ifdef CONFIG_SYSFS
28520 +extern struct attribute_group *sysaufs_attr_group;
28522 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
28523 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
28525 +long au_brinfo_ioctl(struct file *file, unsigned long arg);
28526 +#ifdef CONFIG_COMPAT
28527 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
28530 +void sysaufs_br_init(struct au_branch *br);
28531 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
28532 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
28534 +#define sysaufs_brs_init() do {} while (0)
28537 +#define sysaufs_attr_group NULL
28539 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
28540 +AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
28541 + struct attribute *attr, char *buf)
28542 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
28543 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
28544 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
28546 +static inline void sysaufs_brs_init(void)
28551 +#endif /* CONFIG_SYSFS */
28553 +#endif /* __KERNEL__ */
28554 +#endif /* __SYSAUFS_H__ */
28555 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
28556 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
28557 +++ linux/fs/aufs/sysfs.c 2015-11-11 17:21:46.922197217 +0100
28560 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28562 + * This program, aufs is free software; you can redistribute it and/or modify
28563 + * it under the terms of the GNU General Public License as published by
28564 + * the Free Software Foundation; either version 2 of the License, or
28565 + * (at your option) any later version.
28567 + * This program is distributed in the hope that it will be useful,
28568 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28569 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28570 + * GNU General Public License for more details.
28572 + * You should have received a copy of the GNU General Public License
28573 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28577 + * sysfs interface
28580 +#include <linux/compat.h>
28581 +#include <linux/seq_file.h>
28584 +#ifdef CONFIG_AUFS_FS_MODULE
28585 +/* this entry violates the "one line per file" policy of sysfs */
28586 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
28590 + static char *conf =
28591 +/* this file is generated at compiling */
28592 +#include "conf.str"
28595 + err = snprintf(buf, PAGE_SIZE, conf);
28596 + if (unlikely(err >= PAGE_SIZE))
28601 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
28604 +static struct attribute *au_attr[] = {
28605 +#ifdef CONFIG_AUFS_FS_MODULE
28606 + &au_config_attr.attr,
28608 + NULL, /* need to NULL terminate the list of attributes */
28611 +static struct attribute_group sysaufs_attr_group_body = {
28615 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
28617 +/* ---------------------------------------------------------------------- */
28619 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
28623 + SiMustAnyLock(sb);
28626 + if (au_opt_test(au_mntflags(sb), XINO)) {
28627 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
28628 + seq_putc(seq, '\n');
28634 + * the lifetime of branch is independent from the entry under sysfs.
28635 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
28638 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
28639 + aufs_bindex_t bindex, int idx)
28642 + struct path path;
28643 + struct dentry *root;
28644 + struct au_branch *br;
28645 + au_br_perm_str_t perm;
28647 + AuDbg("b%d\n", bindex);
28650 + root = sb->s_root;
28651 + di_read_lock_parent(root, !AuLock_IR);
28652 + br = au_sbr(sb, bindex);
28655 + case AuBrSysfs_BR:
28656 + path.mnt = au_br_mnt(br);
28657 + path.dentry = au_h_dptr(root, bindex);
28658 + err = au_seq_path(seq, &path);
28660 + au_optstr_br_perm(&perm, br->br_perm);
28661 + seq_printf(seq, "=%s\n", perm.a);
28664 + case AuBrSysfs_BRID:
28665 + seq_printf(seq, "%d\n", br->br_id);
28668 + di_read_unlock(root, !AuLock_IR);
28669 + if (unlikely(err || seq_has_overflowed(seq)))
28675 +/* ---------------------------------------------------------------------- */
28677 +static struct seq_file *au_seq(char *p, ssize_t len)
28679 + struct seq_file *seq;
28681 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
28683 + /* mutex_init(&seq.lock); */
28686 + return seq; /* success */
28689 + seq = ERR_PTR(-ENOMEM);
28693 +#define SysaufsBr_PREFIX "br"
28694 +#define SysaufsBrid_PREFIX "brid"
28696 +/* todo: file size may exceed PAGE_SIZE */
28697 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
28703 + aufs_bindex_t bend;
28704 + struct au_sbinfo *sbinfo;
28705 + struct super_block *sb;
28706 + struct seq_file *seq;
28708 + struct attribute **cattr;
28710 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
28711 + sb = sbinfo->si_sb;
28714 + * prevent a race condition between sysfs and aufs.
28715 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
28716 + * prohibits maintaining the sysfs entries.
28717 + * hew we acquire read lock after sysfs_get_active_two().
28718 + * on the other hand, the remount process may maintain the sysfs/aufs
28719 + * entries after acquiring write lock.
28720 + * it can cause a deadlock.
28721 + * simply we gave up processing read here.
28724 + if (unlikely(!si_noflush_read_trylock(sb)))
28727 + seq = au_seq(buf, PAGE_SIZE);
28728 + err = PTR_ERR(seq);
28732 + name = (void *)attr->name;
28733 + cattr = sysaufs_si_attrs;
28735 + if (!strcmp(name, (*cattr)->name)) {
28736 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
28743 + if (!strncmp(name, SysaufsBrid_PREFIX,
28744 + sizeof(SysaufsBrid_PREFIX) - 1)) {
28745 + idx = AuBrSysfs_BRID;
28746 + name += sizeof(SysaufsBrid_PREFIX) - 1;
28747 + } else if (!strncmp(name, SysaufsBr_PREFIX,
28748 + sizeof(SysaufsBr_PREFIX) - 1)) {
28749 + idx = AuBrSysfs_BR;
28750 + name += sizeof(SysaufsBr_PREFIX) - 1;
28754 + err = kstrtol(name, 10, &l);
28756 + bend = au_sbend(sb);
28758 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
28765 + err = seq->count;
28766 + /* sysfs limit */
28767 + if (unlikely(err == PAGE_SIZE))
28772 + si_read_unlock(sb);
28777 +/* ---------------------------------------------------------------------- */
28779 +static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
28783 + aufs_bindex_t bindex, bend;
28786 + struct seq_file *seq;
28787 + struct au_branch *br;
28789 + si_read_lock(sb, AuLock_FLUSH);
28790 + bend = au_sbend(sb);
28796 + buf = (void *)__get_free_page(GFP_NOFS);
28797 + if (unlikely(!buf))
28800 + seq = au_seq(buf, PAGE_SIZE);
28801 + err = PTR_ERR(seq);
28805 + sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
28806 + for (bindex = 0; bindex <= bend; bindex++, arg++) {
28807 + err = !access_ok(VERIFY_WRITE, arg, sizeof(*arg));
28808 + if (unlikely(err))
28811 + br = au_sbr(sb, bindex);
28812 + brid = br->br_id;
28813 + BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
28814 + err = __put_user(brid, &arg->id);
28815 + if (unlikely(err))
28818 + BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
28819 + err = __put_user(br->br_perm, &arg->perm);
28820 + if (unlikely(err))
28823 + err = au_seq_path(seq, &br->br_path);
28824 + if (unlikely(err))
28826 + seq_putc(seq, '\0');
28827 + if (!seq_has_overflowed(seq)) {
28828 + err = copy_to_user(arg->path, seq->buf, seq->count);
28830 + if (unlikely(err))
28837 + if (unlikely(err))
28843 + free_page((unsigned long)buf);
28845 + si_read_unlock(sb);
28849 +long au_brinfo_ioctl(struct file *file, unsigned long arg)
28851 + return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
28854 +#ifdef CONFIG_COMPAT
28855 +long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
28857 + return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
28861 +/* ---------------------------------------------------------------------- */
28863 +void sysaufs_br_init(struct au_branch *br)
28866 + struct au_brsysfs *br_sysfs;
28867 + struct attribute *attr;
28869 + br_sysfs = br->br_sysfs;
28870 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28871 + attr = &br_sysfs->attr;
28872 + sysfs_attr_init(attr);
28873 + attr->name = br_sysfs->name;
28874 + attr->mode = S_IRUGO;
28879 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
28881 + struct au_branch *br;
28882 + struct kobject *kobj;
28883 + struct au_brsysfs *br_sysfs;
28885 + aufs_bindex_t bend;
28887 + dbgaufs_brs_del(sb, bindex);
28889 + if (!sysaufs_brs)
28892 + kobj = &au_sbi(sb)->si_kobj;
28893 + bend = au_sbend(sb);
28894 + for (; bindex <= bend; bindex++) {
28895 + br = au_sbr(sb, bindex);
28896 + br_sysfs = br->br_sysfs;
28897 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28898 + sysfs_remove_file(kobj, &br_sysfs->attr);
28904 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
28907 + aufs_bindex_t bend;
28908 + struct kobject *kobj;
28909 + struct au_branch *br;
28910 + struct au_brsysfs *br_sysfs;
28912 + dbgaufs_brs_add(sb, bindex);
28914 + if (!sysaufs_brs)
28917 + kobj = &au_sbi(sb)->si_kobj;
28918 + bend = au_sbend(sb);
28919 + for (; bindex <= bend; bindex++) {
28920 + br = au_sbr(sb, bindex);
28921 + br_sysfs = br->br_sysfs;
28922 + snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
28923 + SysaufsBr_PREFIX "%d", bindex);
28924 + snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
28925 + SysaufsBrid_PREFIX "%d", bindex);
28926 + for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28927 + err = sysfs_create_file(kobj, &br_sysfs->attr);
28928 + if (unlikely(err))
28929 + pr_warn("failed %s under sysfs(%d)\n",
28930 + br_sysfs->name, err);
28935 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
28936 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
28937 +++ linux/fs/aufs/sysrq.c 2015-11-11 17:21:46.922197217 +0100
28940 + * Copyright (C) 2005-2015 Junjiro R. Okajima
28942 + * This program, aufs is free software; you can redistribute it and/or modify
28943 + * it under the terms of the GNU General Public License as published by
28944 + * the Free Software Foundation; either version 2 of the License, or
28945 + * (at your option) any later version.
28947 + * This program is distributed in the hope that it will be useful,
28948 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28949 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28950 + * GNU General Public License for more details.
28952 + * You should have received a copy of the GNU General Public License
28953 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
28957 + * magic sysrq hanlder
28960 +/* #include <linux/sysrq.h> */
28961 +#include <linux/writeback.h>
28964 +/* ---------------------------------------------------------------------- */
28966 +static void sysrq_sb(struct super_block *sb)
28969 + struct au_sbinfo *sbinfo;
28970 + struct file *file;
28971 + struct au_sphlhead *files;
28972 + struct au_finfo *finfo;
28974 + plevel = au_plevel;
28975 + au_plevel = KERN_WARNING;
28977 + /* since we define pr_fmt, call printk directly */
28978 +#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
28980 + sbinfo = au_sbi(sb);
28981 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
28982 + pr("superblock\n");
28986 + pr("root dentry\n");
28987 + au_dpri_dentry(sb->s_root);
28988 + pr("root inode\n");
28989 + au_dpri_inode(d_inode(sb->s_root));
28994 + int err, i, j, ndentry;
28995 + struct au_dcsub_pages dpages;
28996 + struct au_dpage *dpage;
28998 + err = au_dpages_init(&dpages, GFP_ATOMIC);
28999 + if (unlikely(err))
29001 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
29003 + for (i = 0; i < dpages.ndpage; i++) {
29004 + dpage = dpages.dpages + i;
29005 + ndentry = dpage->ndentry;
29006 + for (j = 0; j < ndentry; j++)
29007 + au_dpri_dentry(dpage->dentries[j]);
29009 + au_dpages_free(&dpages);
29017 + pr("isolated inode\n");
29018 + spin_lock(&sb->s_inode_list_lock);
29019 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
29020 + spin_lock(&i->i_lock);
29021 + if (1 || hlist_empty(&i->i_dentry))
29022 + au_dpri_inode(i);
29023 + spin_unlock(&i->i_lock);
29025 + spin_unlock(&sb->s_inode_list_lock);
29029 + files = &au_sbi(sb)->si_files;
29030 + spin_lock(&files->spin);
29031 + hlist_for_each_entry(finfo, &files->head, fi_hlist) {
29034 + file = finfo->fi_file;
29035 + mode = file_inode(file)->i_mode;
29036 + if (!special_file(mode))
29037 + au_dpri_file(file);
29039 + spin_unlock(&files->spin);
29043 + au_plevel = plevel;
29046 +/* ---------------------------------------------------------------------- */
29048 +/* module parameter */
29049 +static char *aufs_sysrq_key = "a";
29050 +module_param_named(sysrq, aufs_sysrq_key, charp, S_IRUGO);
29051 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
29053 +static void au_sysrq(int key __maybe_unused)
29055 + struct au_sbinfo *sbinfo;
29058 + au_sbilist_lock();
29059 + list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
29060 + sysrq_sb(sbinfo->si_sb);
29061 + au_sbilist_unlock();
29065 +static struct sysrq_key_op au_sysrq_op = {
29066 + .handler = au_sysrq,
29067 + .help_msg = "Aufs",
29068 + .action_msg = "Aufs",
29069 + .enable_mask = SYSRQ_ENABLE_DUMP
29072 +/* ---------------------------------------------------------------------- */
29074 +int __init au_sysrq_init(void)
29080 + key = *aufs_sysrq_key;
29081 + if ('a' <= key && key <= 'z')
29082 + err = register_sysrq_key(key, &au_sysrq_op);
29083 + if (unlikely(err))
29084 + pr_err("err %d, sysrq=%c\n", err, key);
29088 +void au_sysrq_fin(void)
29092 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
29093 + if (unlikely(err))
29094 + pr_err("err %d (ignored)\n", err);
29096 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
29097 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
29098 +++ linux/fs/aufs/vdir.c 2015-11-11 17:21:46.922197217 +0100
29101 + * Copyright (C) 2005-2015 Junjiro R. Okajima
29103 + * This program, aufs is free software; you can redistribute it and/or modify
29104 + * it under the terms of the GNU General Public License as published by
29105 + * the Free Software Foundation; either version 2 of the License, or
29106 + * (at your option) any later version.
29108 + * This program is distributed in the hope that it will be useful,
29109 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29110 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29111 + * GNU General Public License for more details.
29113 + * You should have received a copy of the GNU General Public License
29114 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
29118 + * virtual or vertical directory
29123 +static unsigned int calc_size(int nlen)
29125 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
29128 +static int set_deblk_end(union au_vdir_deblk_p *p,
29129 + union au_vdir_deblk_p *deblk_end)
29131 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
29132 + p->de->de_str.len = 0;
29136 + return -1; /* error */
29139 +/* returns true or false */
29140 +static int is_deblk_end(union au_vdir_deblk_p *p,
29141 + union au_vdir_deblk_p *deblk_end)
29143 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
29144 + return !p->de->de_str.len;
29148 +static unsigned char *last_deblk(struct au_vdir *vdir)
29150 + return vdir->vd_deblk[vdir->vd_nblk - 1];
29153 +/* ---------------------------------------------------------------------- */
29155 +/* estimate the appropriate size for name hash table */
29156 +unsigned int au_rdhash_est(loff_t sz)
29164 + if (sz < AUFS_RDHASH_DEF)
29165 + n = AUFS_RDHASH_DEF;
29166 + /* pr_info("n %u\n", n); */
29171 + * the allocated memory has to be freed by
29172 + * au_nhash_wh_free() or au_nhash_de_free().
29174 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
29176 + struct hlist_head *head;
29180 + sz = sizeof(*nhash->nh_head) * num_hash;
29181 + head = kmalloc(sz, gfp);
29183 + nhash->nh_num = num_hash;
29184 + nhash->nh_head = head;
29185 + for (u = 0; u < num_hash; u++)
29186 + INIT_HLIST_HEAD(head++);
29187 + return 0; /* success */
29193 +static void nhash_count(struct hlist_head *head)
29197 + struct hlist_node *pos;
29200 + hlist_for_each(pos, head)
29202 + pr_info("%lu\n", n);
29206 +static void au_nhash_wh_do_free(struct hlist_head *head)
29208 + struct au_vdir_wh *pos;
29209 + struct hlist_node *node;
29211 + hlist_for_each_entry_safe(pos, node, head, wh_hash)
29215 +static void au_nhash_de_do_free(struct hlist_head *head)
29217 + struct au_vdir_dehstr *pos;
29218 + struct hlist_node *node;
29220 + hlist_for_each_entry_safe(pos, node, head, hash)
29221 + au_cache_free_vdir_dehstr(pos);
29224 +static void au_nhash_do_free(struct au_nhash *nhash,
29225 + void (*free)(struct hlist_head *head))
29228 + struct hlist_head *head;
29230 + n = nhash->nh_num;
29234 + head = nhash->nh_head;
29235 + while (n-- > 0) {
29236 + nhash_count(head);
29239 + kfree(nhash->nh_head);
29242 +void au_nhash_wh_free(struct au_nhash *whlist)
29244 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
29247 +static void au_nhash_de_free(struct au_nhash *delist)
29249 + au_nhash_do_free(delist, au_nhash_de_do_free);
29252 +/* ---------------------------------------------------------------------- */
29254 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
29258 + unsigned int u, n;
29259 + struct hlist_head *head;
29260 + struct au_vdir_wh *pos;
29263 + n = whlist->nh_num;
29264 + head = whlist->nh_head;
29265 + for (u = 0; u < n; u++, head++)
29266 + hlist_for_each_entry(pos, head, wh_hash)
29267 + if (pos->wh_bindex == btgt && ++num > limit)
29272 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
29273 + unsigned char *name,
29274 + unsigned int len)
29277 + /* const unsigned int magic_bit = 12; */
29279 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
29284 + /* v = hash_long(v, magic_bit); */
29285 + v %= nhash->nh_num;
29286 + return nhash->nh_head + v;
29289 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
29292 + return str->len == nlen && !memcmp(str->name, name, nlen);
29295 +/* returns found or not */
29296 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
29298 + struct hlist_head *head;
29299 + struct au_vdir_wh *pos;
29300 + struct au_vdir_destr *str;
29302 + head = au_name_hash(whlist, name, nlen);
29303 + hlist_for_each_entry(pos, head, wh_hash) {
29304 + str = &pos->wh_str;
29305 + AuDbg("%.*s\n", str->len, str->name);
29306 + if (au_nhash_test_name(str, name, nlen))
29312 +/* returns found(true) or not */
29313 +static int test_known(struct au_nhash *delist, char *name, int nlen)
29315 + struct hlist_head *head;
29316 + struct au_vdir_dehstr *pos;
29317 + struct au_vdir_destr *str;
29319 + head = au_name_hash(delist, name, nlen);
29320 + hlist_for_each_entry(pos, head, hash) {
29322 + AuDbg("%.*s\n", str->len, str->name);
29323 + if (au_nhash_test_name(str, name, nlen))
29329 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
29330 + unsigned char d_type)
29332 +#ifdef CONFIG_AUFS_SHWH
29333 + wh->wh_ino = ino;
29334 + wh->wh_type = d_type;
29338 +/* ---------------------------------------------------------------------- */
29340 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
29341 + unsigned int d_type, aufs_bindex_t bindex,
29342 + unsigned char shwh)
29345 + struct au_vdir_destr *str;
29346 + struct au_vdir_wh *wh;
29348 + AuDbg("%.*s\n", nlen, name);
29349 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
29352 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
29353 + if (unlikely(!wh))
29357 + wh->wh_bindex = bindex;
29359 + au_shwh_init_wh(wh, ino, d_type);
29360 + str = &wh->wh_str;
29362 + memcpy(str->name, name, nlen);
29363 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
29370 +static int append_deblk(struct au_vdir *vdir)
29373 + unsigned long ul;
29374 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
29375 + union au_vdir_deblk_p p, deblk_end;
29376 + unsigned char **o;
29379 + o = krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
29381 + if (unlikely(!o))
29384 + vdir->vd_deblk = o;
29385 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
29387 + ul = vdir->vd_nblk++;
29388 + vdir->vd_deblk[ul] = p.deblk;
29389 + vdir->vd_last.ul = ul;
29390 + vdir->vd_last.p.deblk = p.deblk;
29391 + deblk_end.deblk = p.deblk + deblk_sz;
29392 + err = set_deblk_end(&p, &deblk_end);
29399 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
29400 + unsigned int d_type, struct au_nhash *delist)
29404 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
29405 + union au_vdir_deblk_p p, *room, deblk_end;
29406 + struct au_vdir_dehstr *dehstr;
29408 + p.deblk = last_deblk(vdir);
29409 + deblk_end.deblk = p.deblk + deblk_sz;
29410 + room = &vdir->vd_last.p;
29411 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
29412 + || !is_deblk_end(room, &deblk_end));
29414 + sz = calc_size(nlen);
29415 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
29416 + err = append_deblk(vdir);
29417 + if (unlikely(err))
29420 + p.deblk = last_deblk(vdir);
29421 + deblk_end.deblk = p.deblk + deblk_sz;
29423 + AuDebugOn(room->deblk != p.deblk);
29427 + dehstr = au_cache_alloc_vdir_dehstr();
29428 + if (unlikely(!dehstr))
29431 + dehstr->str = &room->de->de_str;
29432 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
29433 + room->de->de_ino = ino;
29434 + room->de->de_type = d_type;
29435 + room->de->de_str.len = nlen;
29436 + memcpy(room->de->de_str.name, name, nlen);
29439 + room->deblk += sz;
29440 + if (unlikely(set_deblk_end(room, &deblk_end)))
29441 + err = append_deblk(vdir);
29448 +/* ---------------------------------------------------------------------- */
29450 +void au_vdir_free(struct au_vdir *vdir)
29452 + unsigned char **deblk;
29454 + deblk = vdir->vd_deblk;
29455 + while (vdir->vd_nblk--)
29457 + kfree(vdir->vd_deblk);
29458 + au_cache_free_vdir(vdir);
29461 +static struct au_vdir *alloc_vdir(struct file *file)
29463 + struct au_vdir *vdir;
29464 + struct super_block *sb;
29467 + sb = file->f_path.dentry->d_sb;
29468 + SiMustAnyLock(sb);
29471 + vdir = au_cache_alloc_vdir();
29472 + if (unlikely(!vdir))
29475 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
29476 + if (unlikely(!vdir->vd_deblk))
29479 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
29480 + if (!vdir->vd_deblk_sz) {
29481 + /* estimate the appropriate size for deblk */
29482 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
29483 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
29485 + vdir->vd_nblk = 0;
29486 + vdir->vd_version = 0;
29487 + vdir->vd_jiffy = 0;
29488 + err = append_deblk(vdir);
29490 + return vdir; /* success */
29492 + kfree(vdir->vd_deblk);
29495 + au_cache_free_vdir(vdir);
29497 + vdir = ERR_PTR(err);
29501 +static int reinit_vdir(struct au_vdir *vdir)
29504 + union au_vdir_deblk_p p, deblk_end;
29506 + while (vdir->vd_nblk > 1) {
29507 + kfree(vdir->vd_deblk[vdir->vd_nblk - 1]);
29508 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
29511 + p.deblk = vdir->vd_deblk[0];
29512 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
29513 + err = set_deblk_end(&p, &deblk_end);
29514 + /* keep vd_dblk_sz */
29515 + vdir->vd_last.ul = 0;
29516 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
29517 + vdir->vd_version = 0;
29518 + vdir->vd_jiffy = 0;
29523 +/* ---------------------------------------------------------------------- */
29525 +#define AuFillVdir_CALLED 1
29526 +#define AuFillVdir_WHABLE (1 << 1)
29527 +#define AuFillVdir_SHWH (1 << 2)
29528 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
29529 +#define au_fset_fillvdir(flags, name) \
29530 + do { (flags) |= AuFillVdir_##name; } while (0)
29531 +#define au_fclr_fillvdir(flags, name) \
29532 + do { (flags) &= ~AuFillVdir_##name; } while (0)
29534 +#ifndef CONFIG_AUFS_SHWH
29535 +#undef AuFillVdir_SHWH
29536 +#define AuFillVdir_SHWH 0
29539 +struct fillvdir_arg {
29540 + struct dir_context ctx;
29541 + struct file *file;
29542 + struct au_vdir *vdir;
29543 + struct au_nhash delist;
29544 + struct au_nhash whlist;
29545 + aufs_bindex_t bindex;
29546 + unsigned int flags;
29550 +static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
29551 + loff_t offset __maybe_unused, u64 h_ino,
29552 + unsigned int d_type)
29554 + struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
29555 + char *name = (void *)__name;
29556 + struct super_block *sb;
29558 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
29561 + sb = arg->file->f_path.dentry->d_sb;
29562 + au_fset_fillvdir(arg->flags, CALLED);
29564 + if (nlen <= AUFS_WH_PFX_LEN
29565 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
29566 + if (test_known(&arg->delist, name, nlen)
29567 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
29568 + goto out; /* already exists or whiteouted */
29570 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
29572 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
29573 + d_type = DT_UNKNOWN;
29574 + arg->err = append_de(arg->vdir, name, nlen, ino,
29575 + d_type, &arg->delist);
29577 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
29578 + name += AUFS_WH_PFX_LEN;
29579 + nlen -= AUFS_WH_PFX_LEN;
29580 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
29581 + goto out; /* already whiteouted */
29584 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
29587 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
29588 + d_type = DT_UNKNOWN;
29589 + arg->err = au_nhash_append_wh
29590 + (&arg->whlist, name, nlen, ino, d_type,
29591 + arg->bindex, shwh);
29597 + arg->vdir->vd_jiffy = jiffies;
29599 + AuTraceErr(arg->err);
29603 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
29604 + struct au_nhash *whlist, struct au_nhash *delist)
29606 +#ifdef CONFIG_AUFS_SHWH
29608 + unsigned int nh, u;
29609 + struct hlist_head *head;
29610 + struct au_vdir_wh *pos;
29611 + struct hlist_node *n;
29613 + struct au_vdir_destr *destr;
29615 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
29618 + o = p = (void *)__get_free_page(GFP_NOFS);
29619 + if (unlikely(!p))
29623 + nh = whlist->nh_num;
29624 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
29625 + p += AUFS_WH_PFX_LEN;
29626 + for (u = 0; u < nh; u++) {
29627 + head = whlist->nh_head + u;
29628 + hlist_for_each_entry_safe(pos, n, head, wh_hash) {
29629 + destr = &pos->wh_str;
29630 + memcpy(p, destr->name, destr->len);
29631 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
29632 + pos->wh_ino, pos->wh_type, delist);
29633 + if (unlikely(err))
29638 + free_page((unsigned long)o);
29648 +static int au_do_read_vdir(struct fillvdir_arg *arg)
29651 + unsigned int rdhash;
29653 + aufs_bindex_t bend, bindex, bstart;
29654 + unsigned char shwh;
29655 + struct file *hf, *file;
29656 + struct super_block *sb;
29658 + file = arg->file;
29659 + sb = file->f_path.dentry->d_sb;
29660 + SiMustAnyLock(sb);
29662 + rdhash = au_sbi(sb)->si_rdhash;
29664 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
29665 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
29666 + if (unlikely(err))
29668 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
29669 + if (unlikely(err))
29675 + if (au_opt_test(au_mntflags(sb), SHWH)) {
29677 + au_fset_fillvdir(arg->flags, SHWH);
29679 + bstart = au_fbstart(file);
29680 + bend = au_fbend_dir(file);
29681 + for (bindex = bstart; !err && bindex <= bend; bindex++) {
29682 + hf = au_hf_dir(file, bindex);
29686 + offset = vfsub_llseek(hf, 0, SEEK_SET);
29688 + if (unlikely(offset))
29691 + arg->bindex = bindex;
29692 + au_fclr_fillvdir(arg->flags, WHABLE);
29694 + || (bindex != bend
29695 + && au_br_whable(au_sbr_perm(sb, bindex))))
29696 + au_fset_fillvdir(arg->flags, WHABLE);
29699 + au_fclr_fillvdir(arg->flags, CALLED);
29701 + err = vfsub_iterate_dir(hf, &arg->ctx);
29704 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
29707 + * dir_relax() may be good for concurrency, but aufs should not
29708 + * use it since it will cause a lockdep problem.
29712 + if (!err && shwh)
29713 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
29715 + au_nhash_wh_free(&arg->whlist);
29718 + au_nhash_de_free(&arg->delist);
29723 +static int read_vdir(struct file *file, int may_read)
29726 + unsigned long expire;
29727 + unsigned char do_read;
29728 + struct fillvdir_arg arg = {
29730 + .actor = fillvdir
29733 + struct inode *inode;
29734 + struct au_vdir *vdir, *allocated;
29737 + inode = file_inode(file);
29738 + IMustLock(inode);
29739 + SiMustAnyLock(inode->i_sb);
29741 + allocated = NULL;
29743 + expire = au_sbi(inode->i_sb)->si_rdcache;
29744 + vdir = au_ivdir(inode);
29747 + vdir = alloc_vdir(file);
29748 + err = PTR_ERR(vdir);
29749 + if (IS_ERR(vdir))
29752 + allocated = vdir;
29753 + } else if (may_read
29754 + && (inode->i_version != vdir->vd_version
29755 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
29757 + err = reinit_vdir(vdir);
29758 + if (unlikely(err))
29763 + return 0; /* success */
29767 + err = au_do_read_vdir(&arg);
29769 + /* file->f_pos = 0; */ /* todo: ctx->pos? */
29770 + vdir->vd_version = inode->i_version;
29771 + vdir->vd_last.ul = 0;
29772 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
29774 + au_set_ivdir(inode, allocated);
29775 + } else if (allocated)
29776 + au_vdir_free(allocated);
29782 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
29785 + unsigned long ul, n;
29786 + const unsigned int deblk_sz = src->vd_deblk_sz;
29788 + AuDebugOn(tgt->vd_nblk != 1);
29791 + if (tgt->vd_nblk < src->vd_nblk) {
29792 + unsigned char **p;
29794 + p = krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
29796 + if (unlikely(!p))
29798 + tgt->vd_deblk = p;
29801 + if (tgt->vd_deblk_sz != deblk_sz) {
29802 + unsigned char *p;
29804 + tgt->vd_deblk_sz = deblk_sz;
29805 + p = krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS);
29806 + if (unlikely(!p))
29808 + tgt->vd_deblk[0] = p;
29810 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
29811 + tgt->vd_version = src->vd_version;
29812 + tgt->vd_jiffy = src->vd_jiffy;
29814 + n = src->vd_nblk;
29815 + for (ul = 1; ul < n; ul++) {
29816 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
29818 + if (unlikely(!tgt->vd_deblk[ul]))
29822 + tgt->vd_nblk = n;
29823 + tgt->vd_last.ul = tgt->vd_last.ul;
29824 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
29825 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
29826 + - src->vd_deblk[src->vd_last.ul];
29828 + return 0; /* success */
29831 + rerr = reinit_vdir(tgt);
29836 +int au_vdir_init(struct file *file)
29839 + struct inode *inode;
29840 + struct au_vdir *vdir_cache, *allocated;
29842 + /* test file->f_pos here instead of ctx->pos */
29843 + err = read_vdir(file, !file->f_pos);
29844 + if (unlikely(err))
29847 + allocated = NULL;
29848 + vdir_cache = au_fvdir_cache(file);
29849 + if (!vdir_cache) {
29850 + vdir_cache = alloc_vdir(file);
29851 + err = PTR_ERR(vdir_cache);
29852 + if (IS_ERR(vdir_cache))
29854 + allocated = vdir_cache;
29855 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
29856 + /* test file->f_pos here instead of ctx->pos */
29857 + err = reinit_vdir(vdir_cache);
29858 + if (unlikely(err))
29861 + return 0; /* success */
29863 + inode = file_inode(file);
29864 + err = copy_vdir(vdir_cache, au_ivdir(inode));
29866 + file->f_version = inode->i_version;
29868 + au_set_fvdir_cache(file, allocated);
29869 + } else if (allocated)
29870 + au_vdir_free(allocated);
29876 +static loff_t calc_offset(struct au_vdir *vdir)
29879 + union au_vdir_deblk_p p;
29881 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
29882 + offset = vdir->vd_last.p.deblk - p.deblk;
29883 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
29887 +/* returns true or false */
29888 +static int seek_vdir(struct file *file, struct dir_context *ctx)
29891 + unsigned int deblk_sz;
29892 + unsigned long ul, n;
29894 + union au_vdir_deblk_p p, deblk_end;
29895 + struct au_vdir *vdir_cache;
29898 + vdir_cache = au_fvdir_cache(file);
29899 + offset = calc_offset(vdir_cache);
29900 + AuDbg("offset %lld\n", offset);
29901 + if (ctx->pos == offset)
29904 + vdir_cache->vd_last.ul = 0;
29905 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
29910 + deblk_sz = vdir_cache->vd_deblk_sz;
29911 + ul = div64_u64(ctx->pos, deblk_sz);
29912 + AuDbg("ul %lu\n", ul);
29913 + if (ul >= vdir_cache->vd_nblk)
29916 + n = vdir_cache->vd_nblk;
29917 + for (; ul < n; ul++) {
29918 + p.deblk = vdir_cache->vd_deblk[ul];
29919 + deblk_end.deblk = p.deblk + deblk_sz;
29921 + offset *= deblk_sz;
29922 + while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
29925 + l = calc_size(p.de->de_str.len);
29929 + if (!is_deblk_end(&p, &deblk_end)) {
29931 + vdir_cache->vd_last.ul = ul;
29932 + vdir_cache->vd_last.p = p;
29939 + AuTraceErr(!valid);
29943 +int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
29945 + unsigned int l, deblk_sz;
29946 + union au_vdir_deblk_p deblk_end;
29947 + struct au_vdir *vdir_cache;
29948 + struct au_vdir_de *de;
29950 + vdir_cache = au_fvdir_cache(file);
29951 + if (!seek_vdir(file, ctx))
29954 + deblk_sz = vdir_cache->vd_deblk_sz;
29956 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
29957 + deblk_end.deblk += deblk_sz;
29958 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
29959 + de = vdir_cache->vd_last.p.de;
29960 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
29961 + de->de_str.len, de->de_str.name, ctx->pos,
29962 + (unsigned long)de->de_ino, de->de_type);
29963 + if (unlikely(!dir_emit(ctx, de->de_str.name,
29964 + de->de_str.len, de->de_ino,
29966 + /* todo: ignore the error caused by udba? */
29967 + /* return err; */
29971 + l = calc_size(de->de_str.len);
29972 + vdir_cache->vd_last.p.deblk += l;
29975 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
29976 + vdir_cache->vd_last.ul++;
29977 + vdir_cache->vd_last.p.deblk
29978 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
29979 + ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
29988 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
29989 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
29990 +++ linux/fs/aufs/vfsub.c 2015-09-24 10:47:58.258053165 +0200
29993 + * Copyright (C) 2005-2015 Junjiro R. Okajima
29995 + * This program, aufs is free software; you can redistribute it and/or modify
29996 + * it under the terms of the GNU General Public License as published by
29997 + * the Free Software Foundation; either version 2 of the License, or
29998 + * (at your option) any later version.
30000 + * This program is distributed in the hope that it will be useful,
30001 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30002 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30003 + * GNU General Public License for more details.
30005 + * You should have received a copy of the GNU General Public License
30006 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30010 + * sub-routines for VFS
30013 +#include <linux/namei.h>
30014 +#include <linux/security.h>
30015 +#include <linux/splice.h>
30018 +int vfsub_update_h_iattr(struct path *h_path, int *did)
30022 + struct super_block *h_sb;
30024 + /* for remote fs, leave work for its getattr or d_revalidate */
30025 + /* for bad i_attr fs, handle them in aufs_getattr() */
30026 + /* still some fs may acquire i_mutex. we need to skip them */
30030 + h_sb = h_path->dentry->d_sb;
30031 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
30033 + err = vfs_getattr(h_path, &st);
30038 +/* ---------------------------------------------------------------------- */
30040 +struct file *vfsub_dentry_open(struct path *path, int flags)
30042 + struct file *file;
30044 + file = dentry_open(path, flags /* | __FMODE_NONOTIFY */,
30046 + if (!IS_ERR_OR_NULL(file)
30047 + && (file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
30048 + i_readcount_inc(d_inode(path->dentry));
30053 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
30055 + struct file *file;
30058 + file = filp_open(path,
30059 + oflags /* | __FMODE_NONOTIFY */,
30062 + if (IS_ERR(file))
30064 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30071 + * Ideally this function should call VFS:do_last() in order to keep all its
30072 + * checkings. But it is very hard for aufs to regenerate several VFS internal
30073 + * structure such as nameidata. This is a second (or third) best approach.
30074 + * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
30076 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
30077 + struct vfsub_aopen_args *args, struct au_branch *br)
30080 + struct file *file = args->file;
30081 + /* copied from linux/fs/namei.c:atomic_open() */
30082 + struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
30085 + AuDebugOn(!dir->i_op->atomic_open);
30087 + err = au_br_test_oflag(args->open_flag, br);
30088 + if (unlikely(err))
30091 + args->file->f_path.dentry = DENTRY_NOT_SET;
30092 + args->file->f_path.mnt = au_br_mnt(br);
30093 + err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
30094 + args->create_mode, args->opened);
30096 + /* some filesystems don't set FILE_CREATED while succeeded? */
30097 + if (*args->opened & FILE_CREATED)
30098 + fsnotify_create(dir, dentry);
30104 + /* todo: call VFS:may_open() here */
30105 + err = open_check_o_direct(file);
30106 + /* todo: ima_file_check() too? */
30107 + if (!err && (args->open_flag & __FMODE_EXEC))
30108 + err = deny_write_access(file);
30109 + if (unlikely(err))
30110 + /* note that the file is created and still opened */
30114 + atomic_inc(&br->br_count);
30115 + fsnotify_open(file);
30121 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
30125 + err = kern_path(name, flags, path);
30126 + if (!err && d_is_positive(path->dentry))
30127 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
30131 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
30134 + struct path path = {
30138 + /* VFS checks it too, but by WARN_ON_ONCE() */
30139 + IMustLock(d_inode(parent));
30141 + path.dentry = lookup_one_len(name, parent, len);
30142 + if (IS_ERR(path.dentry))
30144 + if (d_is_positive(path.dentry))
30145 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
30148 + AuTraceErrPtr(path.dentry);
30149 + return path.dentry;
30152 +void vfsub_call_lkup_one(void *args)
30154 + struct vfsub_lkup_one_args *a = args;
30155 + *a->errp = vfsub_lkup_one(a->name, a->parent);
30158 +/* ---------------------------------------------------------------------- */
30160 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
30161 + struct dentry *d2, struct au_hinode *hdir2)
30163 + struct dentry *d;
30166 + d = lock_rename(d1, d2);
30168 + au_hn_suspend(hdir1);
30169 + if (hdir1 != hdir2)
30170 + au_hn_suspend(hdir2);
30175 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
30176 + struct dentry *d2, struct au_hinode *hdir2)
30178 + au_hn_resume(hdir1);
30179 + if (hdir1 != hdir2)
30180 + au_hn_resume(hdir2);
30182 + unlock_rename(d1, d2);
30186 +/* ---------------------------------------------------------------------- */
30188 +int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
30191 + struct dentry *d;
30195 + d = path->dentry;
30196 + path->dentry = d->d_parent;
30197 + err = security_path_mknod(path, d, mode, 0);
30198 + path->dentry = d;
30199 + if (unlikely(err))
30203 + err = vfs_create(dir, path->dentry, mode, want_excl);
30206 + struct path tmp = *path;
30209 + vfsub_update_h_iattr(&tmp, &did);
30211 + tmp.dentry = path->dentry->d_parent;
30212 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30221 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
30224 + struct dentry *d;
30228 + d = path->dentry;
30229 + path->dentry = d->d_parent;
30230 + err = security_path_symlink(path, d, symname);
30231 + path->dentry = d;
30232 + if (unlikely(err))
30236 + err = vfs_symlink(dir, path->dentry, symname);
30239 + struct path tmp = *path;
30242 + vfsub_update_h_iattr(&tmp, &did);
30244 + tmp.dentry = path->dentry->d_parent;
30245 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30254 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
30257 + struct dentry *d;
30261 + d = path->dentry;
30262 + path->dentry = d->d_parent;
30263 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
30264 + path->dentry = d;
30265 + if (unlikely(err))
30269 + err = vfs_mknod(dir, path->dentry, mode, dev);
30272 + struct path tmp = *path;
30275 + vfsub_update_h_iattr(&tmp, &did);
30277 + tmp.dentry = path->dentry->d_parent;
30278 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30287 +static int au_test_nlink(struct inode *inode)
30289 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
30291 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
30292 + || inode->i_nlink < link_max)
30297 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
30298 + struct inode **delegated_inode)
30301 + struct dentry *d;
30305 + err = au_test_nlink(d_inode(src_dentry));
30306 + if (unlikely(err))
30309 + /* we don't call may_linkat() */
30310 + d = path->dentry;
30311 + path->dentry = d->d_parent;
30312 + err = security_path_link(src_dentry, path, d);
30313 + path->dentry = d;
30314 + if (unlikely(err))
30318 + err = vfs_link(src_dentry, dir, path->dentry, delegated_inode);
30321 + struct path tmp = *path;
30324 + /* fuse has different memory inode for the same inumber */
30325 + vfsub_update_h_iattr(&tmp, &did);
30327 + tmp.dentry = path->dentry->d_parent;
30328 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30329 + tmp.dentry = src_dentry;
30330 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30339 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
30340 + struct inode *dir, struct path *path,
30341 + struct inode **delegated_inode)
30344 + struct path tmp = {
30347 + struct dentry *d;
30350 + IMustLock(src_dir);
30352 + d = path->dentry;
30353 + path->dentry = d->d_parent;
30354 + tmp.dentry = src_dentry->d_parent;
30355 + err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
30356 + path->dentry = d;
30357 + if (unlikely(err))
30361 + err = vfs_rename(src_dir, src_dentry, dir, path->dentry,
30362 + delegated_inode, /*flags*/0);
30367 + tmp.dentry = d->d_parent;
30368 + vfsub_update_h_iattr(&tmp, &did);
30370 + tmp.dentry = src_dentry;
30371 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30372 + tmp.dentry = src_dentry->d_parent;
30373 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30382 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
30385 + struct dentry *d;
30389 + d = path->dentry;
30390 + path->dentry = d->d_parent;
30391 + err = security_path_mkdir(path, d, mode);
30392 + path->dentry = d;
30393 + if (unlikely(err))
30397 + err = vfs_mkdir(dir, path->dentry, mode);
30400 + struct path tmp = *path;
30403 + vfsub_update_h_iattr(&tmp, &did);
30405 + tmp.dentry = path->dentry->d_parent;
30406 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
30415 +int vfsub_rmdir(struct inode *dir, struct path *path)
30418 + struct dentry *d;
30422 + d = path->dentry;
30423 + path->dentry = d->d_parent;
30424 + err = security_path_rmdir(path, d);
30425 + path->dentry = d;
30426 + if (unlikely(err))
30430 + err = vfs_rmdir(dir, path->dentry);
30433 + struct path tmp = {
30434 + .dentry = path->dentry->d_parent,
30438 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
30445 +/* ---------------------------------------------------------------------- */
30447 +/* todo: support mmap_sem? */
30448 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
30454 + err = vfs_read(file, ubuf, count, ppos);
30457 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30461 +/* todo: kernel_read()? */
30462 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
30466 + mm_segment_t oldfs;
30473 + oldfs = get_fs();
30474 + set_fs(KERNEL_DS);
30475 + err = vfsub_read_u(file, buf.u, count, ppos);
30480 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
30486 + err = vfs_write(file, ubuf, count, ppos);
30489 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30493 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
30496 + mm_segment_t oldfs;
30499 + const char __user *u;
30503 + oldfs = get_fs();
30504 + set_fs(KERNEL_DS);
30505 + err = vfsub_write_u(file, buf.u, count, ppos);
30510 +int vfsub_flush(struct file *file, fl_owner_t id)
30515 + if (file->f_op->flush) {
30516 + if (!au_test_nfs(file->f_path.dentry->d_sb))
30517 + err = file->f_op->flush(file, id);
30520 + err = file->f_op->flush(file, id);
30524 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
30530 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
30534 + AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
30537 + err = iterate_dir(file, ctx);
30540 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30544 +long vfsub_splice_to(struct file *in, loff_t *ppos,
30545 + struct pipe_inode_info *pipe, size_t len,
30546 + unsigned int flags)
30551 + err = do_splice_to(in, ppos, pipe, len, flags);
30553 + file_accessed(in);
30555 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
30559 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
30560 + loff_t *ppos, size_t len, unsigned int flags)
30565 + err = do_splice_from(pipe, out, ppos, len, flags);
30568 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
30572 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
30576 + /* file can be NULL */
30578 + err = vfs_fsync(file, datasync);
30582 + AuDebugOn(!file);
30583 + path = &file->f_path;
30585 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
30590 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
30591 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
30592 + struct file *h_file)
30595 + struct inode *h_inode;
30596 + struct super_block *h_sb;
30599 + err = vfsub_truncate(h_path, length);
30603 + h_inode = d_inode(h_path->dentry);
30604 + h_sb = h_inode->i_sb;
30606 + sb_start_write(h_sb);
30608 + err = locks_verify_truncate(h_inode, h_file, length);
30610 + err = security_path_truncate(h_path);
30613 + err = do_truncate(h_path->dentry, length, attr, h_file);
30617 + sb_end_write(h_sb);
30624 +/* ---------------------------------------------------------------------- */
30626 +struct au_vfsub_mkdir_args {
30628 + struct inode *dir;
30629 + struct path *path;
30633 +static void au_call_vfsub_mkdir(void *args)
30635 + struct au_vfsub_mkdir_args *a = args;
30636 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
30639 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
30641 + int err, do_sio, wkq_err;
30643 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
30646 + err = vfsub_mkdir(dir, path, mode);
30649 + struct au_vfsub_mkdir_args args = {
30655 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
30656 + if (unlikely(wkq_err))
30663 +struct au_vfsub_rmdir_args {
30665 + struct inode *dir;
30666 + struct path *path;
30669 +static void au_call_vfsub_rmdir(void *args)
30671 + struct au_vfsub_rmdir_args *a = args;
30672 + *a->errp = vfsub_rmdir(a->dir, a->path);
30675 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
30677 + int err, do_sio, wkq_err;
30679 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
30682 + err = vfsub_rmdir(dir, path);
30685 + struct au_vfsub_rmdir_args args = {
30690 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
30691 + if (unlikely(wkq_err))
30698 +/* ---------------------------------------------------------------------- */
30700 +struct notify_change_args {
30702 + struct path *path;
30703 + struct iattr *ia;
30704 + struct inode **delegated_inode;
30707 +static void call_notify_change(void *args)
30709 + struct notify_change_args *a = args;
30710 + struct inode *h_inode;
30712 + h_inode = d_inode(a->path->dentry);
30713 + IMustLock(h_inode);
30715 + *a->errp = -EPERM;
30716 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
30718 + *a->errp = notify_change(a->path->dentry, a->ia,
30719 + a->delegated_inode);
30722 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
30724 + AuTraceErr(*a->errp);
30727 +int vfsub_notify_change(struct path *path, struct iattr *ia,
30728 + struct inode **delegated_inode)
30731 + struct notify_change_args args = {
30735 + .delegated_inode = delegated_inode
30738 + call_notify_change(&args);
30743 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
30744 + struct inode **delegated_inode)
30746 + int err, wkq_err;
30747 + struct notify_change_args args = {
30751 + .delegated_inode = delegated_inode
30754 + wkq_err = au_wkq_wait(call_notify_change, &args);
30755 + if (unlikely(wkq_err))
30761 +/* ---------------------------------------------------------------------- */
30763 +struct unlink_args {
30765 + struct inode *dir;
30766 + struct path *path;
30767 + struct inode **delegated_inode;
30770 +static void call_unlink(void *args)
30772 + struct unlink_args *a = args;
30773 + struct dentry *d = a->path->dentry;
30774 + struct inode *h_inode;
30775 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
30776 + && au_dcount(d) == 1);
30778 + IMustLock(a->dir);
30780 + a->path->dentry = d->d_parent;
30781 + *a->errp = security_path_unlink(a->path, d);
30782 + a->path->dentry = d;
30783 + if (unlikely(*a->errp))
30786 + if (!stop_sillyrename)
30789 + if (d_is_positive(d)) {
30790 + h_inode = d_inode(d);
30795 + *a->errp = vfs_unlink(a->dir, d, a->delegated_inode);
30798 + struct path tmp = {
30799 + .dentry = d->d_parent,
30800 + .mnt = a->path->mnt
30802 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
30805 + if (!stop_sillyrename)
30810 + AuTraceErr(*a->errp);
30814 + * @dir: must be locked.
30815 + * @dentry: target dentry.
30817 +int vfsub_unlink(struct inode *dir, struct path *path,
30818 + struct inode **delegated_inode, int force)
30821 + struct unlink_args args = {
30825 + .delegated_inode = delegated_inode
30829 + call_unlink(&args);
30833 + wkq_err = au_wkq_wait(call_unlink, &args);
30834 + if (unlikely(wkq_err))
30840 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
30841 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
30842 +++ linux/fs/aufs/vfsub.h 2015-11-11 17:21:46.922197217 +0100
30845 + * Copyright (C) 2005-2015 Junjiro R. Okajima
30847 + * This program, aufs is free software; you can redistribute it and/or modify
30848 + * it under the terms of the GNU General Public License as published by
30849 + * the Free Software Foundation; either version 2 of the License, or
30850 + * (at your option) any later version.
30852 + * This program is distributed in the hope that it will be useful,
30853 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30854 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30855 + * GNU General Public License for more details.
30857 + * You should have received a copy of the GNU General Public License
30858 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
30862 + * sub-routines for VFS
30865 +#ifndef __AUFS_VFSUB_H__
30866 +#define __AUFS_VFSUB_H__
30870 +#include <linux/fs.h>
30871 +#include <linux/mount.h>
30872 +#include <linux/xattr.h>
30873 +#include "debug.h"
30875 +/* copied from linux/fs/internal.h */
30876 +/* todo: BAD approach!! */
30877 +extern void __mnt_drop_write(struct vfsmount *);
30878 +extern int open_check_o_direct(struct file *f);
30880 +/* ---------------------------------------------------------------------- */
30882 +/* lock subclass for lower inode */
30883 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
30884 +/* reduce? gave up. */
30886 + AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
30887 + AuLsc_I_PARENT, /* lower inode, parent first */
30888 + AuLsc_I_PARENT2, /* copyup dirs */
30889 + AuLsc_I_PARENT3, /* copyup wh */
30895 +/* to debug easier, do not make them inlined functions */
30896 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
30897 +#define IMustLock(i) MtxMustLock(&(i)->i_mutex)
30899 +/* ---------------------------------------------------------------------- */
30901 +static inline void vfsub_drop_nlink(struct inode *inode)
30903 + AuDebugOn(!inode->i_nlink);
30904 + drop_nlink(inode);
30907 +static inline void vfsub_dead_dir(struct inode *inode)
30909 + AuDebugOn(!S_ISDIR(inode->i_mode));
30910 + inode->i_flags |= S_DEAD;
30911 + clear_nlink(inode);
30914 +static inline int vfsub_native_ro(struct inode *inode)
30916 + return (inode->i_sb->s_flags & MS_RDONLY)
30917 + || IS_RDONLY(inode)
30918 + /* || IS_APPEND(inode) */
30919 + || IS_IMMUTABLE(inode);
30922 +/* ---------------------------------------------------------------------- */
30924 +int vfsub_update_h_iattr(struct path *h_path, int *did);
30925 +struct file *vfsub_dentry_open(struct path *path, int flags);
30926 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
30927 +struct vfsub_aopen_args {
30928 + struct file *file;
30929 + unsigned int open_flag;
30930 + umode_t create_mode;
30934 +int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
30935 + struct vfsub_aopen_args *args, struct au_branch *br);
30936 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
30938 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
30941 +struct vfsub_lkup_one_args {
30942 + struct dentry **errp;
30943 + struct qstr *name;
30944 + struct dentry *parent;
30947 +static inline struct dentry *vfsub_lkup_one(struct qstr *name,
30948 + struct dentry *parent)
30950 + return vfsub_lookup_one_len(name->name, parent, name->len);
30953 +void vfsub_call_lkup_one(void *args);
30955 +/* ---------------------------------------------------------------------- */
30957 +static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
30962 + err = mnt_want_write(mnt);
30967 +static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
30970 + mnt_drop_write(mnt);
30974 +#if 0 /* reserved */
30975 +static inline void vfsub_mnt_drop_write_file(struct file *file)
30978 + mnt_drop_write_file(file);
30983 +/* ---------------------------------------------------------------------- */
30986 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
30987 + struct dentry *d2, struct au_hinode *hdir2);
30988 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
30989 + struct dentry *d2, struct au_hinode *hdir2);
30991 +int vfsub_create(struct inode *dir, struct path *path, int mode,
30993 +int vfsub_symlink(struct inode *dir, struct path *path,
30994 + const char *symname);
30995 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
30996 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
30997 + struct path *path, struct inode **delegated_inode);
30998 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
30999 + struct inode *hdir, struct path *path,
31000 + struct inode **delegated_inode);
31001 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
31002 +int vfsub_rmdir(struct inode *dir, struct path *path);
31004 +/* ---------------------------------------------------------------------- */
31006 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
31008 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
31010 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
31012 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
31014 +int vfsub_flush(struct file *file, fl_owner_t id);
31015 +int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
31017 +static inline loff_t vfsub_f_size_read(struct file *file)
31019 + return i_size_read(file_inode(file));
31022 +static inline unsigned int vfsub_file_flags(struct file *file)
31024 + unsigned int flags;
31026 + spin_lock(&file->f_lock);
31027 + flags = file->f_flags;
31028 + spin_unlock(&file->f_lock);
31033 +#if 0 /* reserved */
31034 +static inline void vfsub_file_accessed(struct file *h_file)
31036 + file_accessed(h_file);
31037 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
31041 +#if 0 /* reserved */
31042 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
31043 + struct dentry *h_dentry)
31045 + struct path h_path = {
31046 + .dentry = h_dentry,
31049 + touch_atime(&h_path);
31050 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
31054 +static inline int vfsub_update_time(struct inode *h_inode, struct timespec *ts,
31057 + return generic_update_time(h_inode, ts, flags);
31058 + /* no vfsub_update_h_iattr() since we don't have struct path */
31061 +long vfsub_splice_to(struct file *in, loff_t *ppos,
31062 + struct pipe_inode_info *pipe, size_t len,
31063 + unsigned int flags);
31064 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
31065 + loff_t *ppos, size_t len, unsigned int flags);
31067 +static inline long vfsub_truncate(struct path *path, loff_t length)
31072 + err = vfs_truncate(path, length);
31077 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
31078 + struct file *h_file);
31079 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
31081 +/* ---------------------------------------------------------------------- */
31083 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
31088 + err = vfs_llseek(file, offset, origin);
31093 +/* ---------------------------------------------------------------------- */
31095 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
31096 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
31097 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
31098 + struct inode **delegated_inode);
31099 +int vfsub_notify_change(struct path *path, struct iattr *ia,
31100 + struct inode **delegated_inode);
31101 +int vfsub_unlink(struct inode *dir, struct path *path,
31102 + struct inode **delegated_inode, int force);
31104 +/* ---------------------------------------------------------------------- */
31106 +static inline int vfsub_setxattr(struct dentry *dentry, const char *name,
31107 + const void *value, size_t size, int flags)
31112 + err = vfs_setxattr(dentry, name, value, size, flags);
31118 +static inline int vfsub_removexattr(struct dentry *dentry, const char *name)
31123 + err = vfs_removexattr(dentry, name);
31129 +#endif /* __KERNEL__ */
31130 +#endif /* __AUFS_VFSUB_H__ */
31131 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
31132 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
31133 +++ linux/fs/aufs/wbr_policy.c 2015-09-24 10:47:58.258053165 +0200
31136 + * Copyright (C) 2005-2015 Junjiro R. Okajima
31138 + * This program, aufs is free software; you can redistribute it and/or modify
31139 + * it under the terms of the GNU General Public License as published by
31140 + * the Free Software Foundation; either version 2 of the License, or
31141 + * (at your option) any later version.
31143 + * This program is distributed in the hope that it will be useful,
31144 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31145 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31146 + * GNU General Public License for more details.
31148 + * You should have received a copy of the GNU General Public License
31149 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31153 + * policies for selecting one among multiple writable branches
31156 +#include <linux/statfs.h>
31159 +/* subset of cpup_attr() */
31160 +static noinline_for_stack
31161 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
31165 + struct inode *h_isrc;
31167 + h_isrc = d_inode(h_src);
31168 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
31169 + ia.ia_mode = h_isrc->i_mode;
31170 + ia.ia_uid = h_isrc->i_uid;
31171 + ia.ia_gid = h_isrc->i_gid;
31172 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
31173 + au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
31174 + /* no delegation since it is just created */
31175 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
31177 + /* is this nfs only? */
31178 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
31179 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
31180 + ia.ia_mode = h_isrc->i_mode;
31181 + err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
31187 +#define AuCpdown_PARENT_OPQ 1
31188 +#define AuCpdown_WHED (1 << 1)
31189 +#define AuCpdown_MADE_DIR (1 << 2)
31190 +#define AuCpdown_DIROPQ (1 << 3)
31191 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
31192 +#define au_fset_cpdown(flags, name) \
31193 + do { (flags) |= AuCpdown_##name; } while (0)
31194 +#define au_fclr_cpdown(flags, name) \
31195 + do { (flags) &= ~AuCpdown_##name; } while (0)
31197 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
31198 + unsigned int *flags)
31201 + struct dentry *opq_dentry;
31203 + opq_dentry = au_diropq_create(dentry, bdst);
31204 + err = PTR_ERR(opq_dentry);
31205 + if (IS_ERR(opq_dentry))
31207 + dput(opq_dentry);
31208 + au_fset_cpdown(*flags, DIROPQ);
31214 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
31215 + struct inode *dir, aufs_bindex_t bdst)
31218 + struct path h_path;
31219 + struct au_branch *br;
31221 + br = au_sbr(dentry->d_sb, bdst);
31222 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
31223 + err = PTR_ERR(h_path.dentry);
31224 + if (IS_ERR(h_path.dentry))
31228 + if (d_is_positive(h_path.dentry)) {
31229 + h_path.mnt = au_br_mnt(br);
31230 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
31233 + dput(h_path.dentry);
31239 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
31240 + struct au_pin *pin,
31241 + struct dentry *h_parent, void *arg)
31244 + aufs_bindex_t bopq, bstart;
31245 + struct path h_path;
31246 + struct dentry *parent;
31247 + struct inode *h_dir, *h_inode, *inode, *dir;
31248 + unsigned int *flags = arg;
31250 + bstart = au_dbstart(dentry);
31251 + /* dentry is di-locked */
31252 + parent = dget_parent(dentry);
31253 + dir = d_inode(parent);
31254 + h_dir = d_inode(h_parent);
31255 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
31256 + IMustLock(h_dir);
31258 + err = au_lkup_neg(dentry, bdst, /*wh*/0);
31259 + if (unlikely(err < 0))
31261 + h_path.dentry = au_h_dptr(dentry, bdst);
31262 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
31263 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path,
31264 + S_IRWXU | S_IRUGO | S_IXUGO);
31265 + if (unlikely(err))
31267 + au_fset_cpdown(*flags, MADE_DIR);
31269 + bopq = au_dbdiropq(dentry);
31270 + au_fclr_cpdown(*flags, WHED);
31271 + au_fclr_cpdown(*flags, DIROPQ);
31272 + if (au_dbwh(dentry) == bdst)
31273 + au_fset_cpdown(*flags, WHED);
31274 + if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
31275 + au_fset_cpdown(*flags, PARENT_OPQ);
31276 + h_inode = d_inode(h_path.dentry);
31277 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
31278 + if (au_ftest_cpdown(*flags, WHED)) {
31279 + err = au_cpdown_dir_opq(dentry, bdst, flags);
31280 + if (unlikely(err)) {
31281 + mutex_unlock(&h_inode->i_mutex);
31286 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, bstart));
31287 + mutex_unlock(&h_inode->i_mutex);
31288 + if (unlikely(err))
31291 + if (au_ftest_cpdown(*flags, WHED)) {
31292 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
31293 + if (unlikely(err))
31297 + inode = d_inode(dentry);
31298 + if (au_ibend(inode) < bdst)
31299 + au_set_ibend(inode, bdst);
31300 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
31301 + au_hi_flags(inode, /*isdir*/1));
31302 + au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
31303 + goto out; /* success */
31307 + if (au_ftest_cpdown(*flags, DIROPQ)) {
31308 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
31309 + rerr = au_diropq_remove(dentry, bdst);
31310 + mutex_unlock(&h_inode->i_mutex);
31311 + if (unlikely(rerr)) {
31312 + AuIOErr("failed removing diropq for %pd b%d (%d)\n",
31313 + dentry, bdst, rerr);
31319 + if (au_ftest_cpdown(*flags, MADE_DIR)) {
31320 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
31321 + if (unlikely(rerr)) {
31322 + AuIOErr("failed removing %pd b%d (%d)\n",
31323 + dentry, bdst, rerr);
31328 + au_set_h_dptr(dentry, bdst, NULL);
31329 + if (au_dbend(dentry) == bdst)
31330 + au_update_dbend(dentry);
31336 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
31339 + unsigned int flags;
31342 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
31347 +/* ---------------------------------------------------------------------- */
31349 +/* policies for create */
31351 +int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
31353 + int err, i, j, ndentry;
31354 + aufs_bindex_t bopq;
31355 + struct au_dcsub_pages dpages;
31356 + struct au_dpage *dpage;
31357 + struct dentry **dentries, *parent, *d;
31359 + err = au_dpages_init(&dpages, GFP_NOFS);
31360 + if (unlikely(err))
31362 + parent = dget_parent(dentry);
31363 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
31364 + if (unlikely(err))
31368 + for (i = 0; i < dpages.ndpage; i++) {
31369 + dpage = dpages.dpages + i;
31370 + dentries = dpage->dentries;
31371 + ndentry = dpage->ndentry;
31372 + for (j = 0; j < ndentry; j++) {
31374 + di_read_lock_parent2(d, !AuLock_IR);
31375 + bopq = au_dbdiropq(d);
31376 + di_read_unlock(d, !AuLock_IR);
31377 + if (bopq >= 0 && bopq < err)
31384 + au_dpages_free(&dpages);
31389 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
31391 + for (; bindex >= 0; bindex--)
31392 + if (!au_br_rdonly(au_sbr(sb, bindex)))
31397 +/* top down parent */
31398 +static int au_wbr_create_tdp(struct dentry *dentry,
31399 + unsigned int flags __maybe_unused)
31402 + aufs_bindex_t bstart, bindex;
31403 + struct super_block *sb;
31404 + struct dentry *parent, *h_parent;
31406 + sb = dentry->d_sb;
31407 + bstart = au_dbstart(dentry);
31409 + if (!au_br_rdonly(au_sbr(sb, bstart)))
31413 + parent = dget_parent(dentry);
31414 + for (bindex = au_dbstart(parent); bindex < bstart; bindex++) {
31415 + h_parent = au_h_dptr(parent, bindex);
31416 + if (!h_parent || d_is_negative(h_parent))
31419 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
31426 + /* bottom up here */
31427 + if (unlikely(err < 0)) {
31428 + err = au_wbr_bu(sb, bstart - 1);
31430 + err = au_wbr_nonopq(dentry, err);
31434 + AuDbg("b%d\n", err);
31438 +/* ---------------------------------------------------------------------- */
31440 +/* an exception for the policy other than tdp */
31441 +static int au_wbr_create_exp(struct dentry *dentry)
31444 + aufs_bindex_t bwh, bdiropq;
31445 + struct dentry *parent;
31448 + bwh = au_dbwh(dentry);
31449 + parent = dget_parent(dentry);
31450 + bdiropq = au_dbdiropq(parent);
31452 + if (bdiropq >= 0)
31453 + err = min(bdiropq, bwh);
31456 + AuDbg("%d\n", err);
31457 + } else if (bdiropq >= 0) {
31459 + AuDbg("%d\n", err);
31464 + err = au_wbr_nonopq(dentry, err);
31466 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
31469 + AuDbg("%d\n", err);
31473 +/* ---------------------------------------------------------------------- */
31476 +static int au_wbr_create_init_rr(struct super_block *sb)
31480 + err = au_wbr_bu(sb, au_sbend(sb));
31481 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
31484 + AuDbg("b%d\n", err);
31488 +static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
31492 + aufs_bindex_t bindex, bend;
31493 + struct super_block *sb;
31496 + err = au_wbr_create_exp(dentry);
31500 + sb = dentry->d_sb;
31501 + next = &au_sbi(sb)->si_wbr_rr_next;
31502 + bend = au_sbend(sb);
31504 + for (bindex = 0; bindex <= bend; bindex++) {
31505 + if (!au_ftest_wbr(flags, DIR)) {
31506 + err = atomic_dec_return(next) + 1;
31507 + /* modulo for 0 is meaningless */
31508 + if (unlikely(!err))
31509 + err = atomic_dec_return(next) + 1;
31511 + err = atomic_read(next);
31512 + AuDbg("%d\n", err);
31515 + AuDbg("%d\n", err);
31516 + if (!au_br_rdonly(au_sbr(sb, err)))
31522 + err = au_wbr_nonopq(dentry, err);
31525 + AuDbg("%d\n", err);
31529 +/* ---------------------------------------------------------------------- */
31531 +/* most free space */
31532 +static void au_mfs(struct dentry *dentry, struct dentry *parent)
31534 + struct super_block *sb;
31535 + struct au_branch *br;
31536 + struct au_wbr_mfs *mfs;
31537 + struct dentry *h_parent;
31538 + aufs_bindex_t bindex, bend;
31540 + unsigned long long b, bavail;
31541 + struct path h_path;
31542 + /* reduce the stack usage */
31543 + struct kstatfs *st;
31545 + st = kmalloc(sizeof(*st), GFP_NOFS);
31546 + if (unlikely(!st)) {
31547 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
31552 + sb = dentry->d_sb;
31553 + mfs = &au_sbi(sb)->si_wbr_mfs;
31554 + MtxMustLock(&mfs->mfs_lock);
31555 + mfs->mfs_bindex = -EROFS;
31556 + mfs->mfsrr_bytes = 0;
31559 + bend = au_sbend(sb);
31561 + bindex = au_dbstart(parent);
31562 + bend = au_dbtaildir(parent);
31565 + for (; bindex <= bend; bindex++) {
31567 + h_parent = au_h_dptr(parent, bindex);
31568 + if (!h_parent || d_is_negative(h_parent))
31571 + br = au_sbr(sb, bindex);
31572 + if (au_br_rdonly(br))
31575 + /* sb->s_root for NFS is unreliable */
31576 + h_path.mnt = au_br_mnt(br);
31577 + h_path.dentry = h_path.mnt->mnt_root;
31578 + err = vfs_statfs(&h_path, st);
31579 + if (unlikely(err)) {
31580 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
31584 + /* when the available size is equal, select the lower one */
31585 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
31586 + || sizeof(b) < sizeof(st->f_bsize));
31587 + b = st->f_bavail * st->f_bsize;
31588 + br->br_wbr->wbr_bytes = b;
31589 + if (b >= bavail) {
31591 + mfs->mfs_bindex = bindex;
31592 + mfs->mfs_jiffy = jiffies;
31596 + mfs->mfsrr_bytes = bavail;
31597 + AuDbg("b%d\n", mfs->mfs_bindex);
31601 +static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
31604 + struct dentry *parent;
31605 + struct super_block *sb;
31606 + struct au_wbr_mfs *mfs;
31608 + err = au_wbr_create_exp(dentry);
31612 + sb = dentry->d_sb;
31614 + if (au_ftest_wbr(flags, PARENT))
31615 + parent = dget_parent(dentry);
31616 + mfs = &au_sbi(sb)->si_wbr_mfs;
31617 + mutex_lock(&mfs->mfs_lock);
31618 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
31619 + || mfs->mfs_bindex < 0
31620 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
31621 + au_mfs(dentry, parent);
31622 + mutex_unlock(&mfs->mfs_lock);
31623 + err = mfs->mfs_bindex;
31627 + err = au_wbr_nonopq(dentry, err);
31630 + AuDbg("b%d\n", err);
31634 +static int au_wbr_create_init_mfs(struct super_block *sb)
31636 + struct au_wbr_mfs *mfs;
31638 + mfs = &au_sbi(sb)->si_wbr_mfs;
31639 + mutex_init(&mfs->mfs_lock);
31640 + mfs->mfs_jiffy = 0;
31641 + mfs->mfs_bindex = -EROFS;
31646 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
31648 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
31652 +/* ---------------------------------------------------------------------- */
31654 +/* most free space and then round robin */
31655 +static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
31658 + struct au_wbr_mfs *mfs;
31660 + err = au_wbr_create_mfs(dentry, flags);
31662 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
31663 + mutex_lock(&mfs->mfs_lock);
31664 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
31665 + err = au_wbr_create_rr(dentry, flags);
31666 + mutex_unlock(&mfs->mfs_lock);
31669 + AuDbg("b%d\n", err);
31673 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
31677 + au_wbr_create_init_mfs(sb); /* ignore */
31678 + err = au_wbr_create_init_rr(sb);
31683 +/* ---------------------------------------------------------------------- */
31685 +/* top down parent and most free space */
31686 +static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
31689 + unsigned long long b;
31690 + aufs_bindex_t bindex, bstart, bend;
31691 + struct super_block *sb;
31692 + struct dentry *parent, *h_parent;
31693 + struct au_branch *br;
31695 + err = au_wbr_create_tdp(dentry, flags);
31696 + if (unlikely(err < 0))
31698 + parent = dget_parent(dentry);
31699 + bstart = au_dbstart(parent);
31700 + bend = au_dbtaildir(parent);
31701 + if (bstart == bend)
31702 + goto out_parent; /* success */
31704 + e2 = au_wbr_create_mfs(dentry, flags);
31706 + goto out_parent; /* success */
31708 + /* when the available size is equal, select upper one */
31709 + sb = dentry->d_sb;
31710 + br = au_sbr(sb, err);
31711 + b = br->br_wbr->wbr_bytes;
31712 + AuDbg("b%d, %llu\n", err, b);
31714 + for (bindex = bstart; bindex <= bend; bindex++) {
31715 + h_parent = au_h_dptr(parent, bindex);
31716 + if (!h_parent || d_is_negative(h_parent))
31719 + br = au_sbr(sb, bindex);
31720 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
31721 + b = br->br_wbr->wbr_bytes;
31723 + AuDbg("b%d, %llu\n", err, b);
31728 + err = au_wbr_nonopq(dentry, err);
31733 + AuDbg("b%d\n", err);
31737 +/* ---------------------------------------------------------------------- */
31740 + * - top down parent
31741 + * - most free space with parent
31742 + * - most free space round-robin regardless parent
31744 +static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
31747 + unsigned long long watermark;
31748 + struct super_block *sb;
31749 + struct au_branch *br;
31750 + struct au_wbr_mfs *mfs;
31752 + err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
31753 + if (unlikely(err < 0))
31756 + sb = dentry->d_sb;
31757 + br = au_sbr(sb, err);
31758 + mfs = &au_sbi(sb)->si_wbr_mfs;
31759 + mutex_lock(&mfs->mfs_lock);
31760 + watermark = mfs->mfsrr_watermark;
31761 + mutex_unlock(&mfs->mfs_lock);
31762 + if (br->br_wbr->wbr_bytes < watermark)
31763 + /* regardless the parent dir */
31764 + err = au_wbr_create_mfsrr(dentry, flags);
31767 + AuDbg("b%d\n", err);
31771 +/* ---------------------------------------------------------------------- */
31773 +/* policies for copyup */
31775 +/* top down parent */
31776 +static int au_wbr_copyup_tdp(struct dentry *dentry)
31778 + return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
31781 +/* bottom up parent */
31782 +static int au_wbr_copyup_bup(struct dentry *dentry)
31785 + aufs_bindex_t bindex, bstart;
31786 + struct dentry *parent, *h_parent;
31787 + struct super_block *sb;
31790 + sb = dentry->d_sb;
31791 + parent = dget_parent(dentry);
31792 + bstart = au_dbstart(parent);
31793 + for (bindex = au_dbstart(dentry); bindex >= bstart; bindex--) {
31794 + h_parent = au_h_dptr(parent, bindex);
31795 + if (!h_parent || d_is_negative(h_parent))
31798 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
31805 + /* bottom up here */
31806 + if (unlikely(err < 0))
31807 + err = au_wbr_bu(sb, bstart - 1);
31809 + AuDbg("b%d\n", err);
31814 +int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t bstart)
31818 + err = au_wbr_bu(dentry->d_sb, bstart);
31819 + AuDbg("b%d\n", err);
31820 + if (err > bstart)
31821 + err = au_wbr_nonopq(dentry, err);
31823 + AuDbg("b%d\n", err);
31827 +static int au_wbr_copyup_bu(struct dentry *dentry)
31830 + aufs_bindex_t bstart;
31832 + bstart = au_dbstart(dentry);
31833 + err = au_wbr_do_copyup_bu(dentry, bstart);
31837 +/* ---------------------------------------------------------------------- */
31839 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
31840 + [AuWbrCopyup_TDP] = {
31841 + .copyup = au_wbr_copyup_tdp
31843 + [AuWbrCopyup_BUP] = {
31844 + .copyup = au_wbr_copyup_bup
31846 + [AuWbrCopyup_BU] = {
31847 + .copyup = au_wbr_copyup_bu
31851 +struct au_wbr_create_operations au_wbr_create_ops[] = {
31852 + [AuWbrCreate_TDP] = {
31853 + .create = au_wbr_create_tdp
31855 + [AuWbrCreate_RR] = {
31856 + .create = au_wbr_create_rr,
31857 + .init = au_wbr_create_init_rr
31859 + [AuWbrCreate_MFS] = {
31860 + .create = au_wbr_create_mfs,
31861 + .init = au_wbr_create_init_mfs,
31862 + .fin = au_wbr_create_fin_mfs
31864 + [AuWbrCreate_MFSV] = {
31865 + .create = au_wbr_create_mfs,
31866 + .init = au_wbr_create_init_mfs,
31867 + .fin = au_wbr_create_fin_mfs
31869 + [AuWbrCreate_MFSRR] = {
31870 + .create = au_wbr_create_mfsrr,
31871 + .init = au_wbr_create_init_mfsrr,
31872 + .fin = au_wbr_create_fin_mfs
31874 + [AuWbrCreate_MFSRRV] = {
31875 + .create = au_wbr_create_mfsrr,
31876 + .init = au_wbr_create_init_mfsrr,
31877 + .fin = au_wbr_create_fin_mfs
31879 + [AuWbrCreate_PMFS] = {
31880 + .create = au_wbr_create_pmfs,
31881 + .init = au_wbr_create_init_mfs,
31882 + .fin = au_wbr_create_fin_mfs
31884 + [AuWbrCreate_PMFSV] = {
31885 + .create = au_wbr_create_pmfs,
31886 + .init = au_wbr_create_init_mfs,
31887 + .fin = au_wbr_create_fin_mfs
31889 + [AuWbrCreate_PMFSRR] = {
31890 + .create = au_wbr_create_pmfsrr,
31891 + .init = au_wbr_create_init_mfsrr,
31892 + .fin = au_wbr_create_fin_mfs
31894 + [AuWbrCreate_PMFSRRV] = {
31895 + .create = au_wbr_create_pmfsrr,
31896 + .init = au_wbr_create_init_mfsrr,
31897 + .fin = au_wbr_create_fin_mfs
31900 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
31901 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
31902 +++ linux/fs/aufs/whout.c 2015-09-24 10:47:58.258053165 +0200
31905 + * Copyright (C) 2005-2015 Junjiro R. Okajima
31907 + * This program, aufs is free software; you can redistribute it and/or modify
31908 + * it under the terms of the GNU General Public License as published by
31909 + * the Free Software Foundation; either version 2 of the License, or
31910 + * (at your option) any later version.
31912 + * This program is distributed in the hope that it will be useful,
31913 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
31914 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31915 + * GNU General Public License for more details.
31917 + * You should have received a copy of the GNU General Public License
31918 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
31922 + * whiteout for logical deletion and opaque directory
31927 +#define WH_MASK S_IRUGO
31930 + * If a directory contains this file, then it is opaque. We start with the
31931 + * .wh. flag so that it is blocked by lookup.
31933 +static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
31934 + sizeof(AUFS_WH_DIROPQ) - 1);
31937 + * generate whiteout name, which is NOT terminated by NULL.
31938 + * @name: original d_name.name
31939 + * @len: original d_name.len
31940 + * @wh: whiteout qstr
31941 + * returns zero when succeeds, otherwise error.
31942 + * succeeded value as wh->name should be freed by kfree().
31944 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
31948 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
31949 + return -ENAMETOOLONG;
31951 + wh->len = name->len + AUFS_WH_PFX_LEN;
31952 + p = kmalloc(wh->len, GFP_NOFS);
31955 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
31956 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
31963 +/* ---------------------------------------------------------------------- */
31966 + * test if the @wh_name exists under @h_parent.
31967 + * @try_sio specifies the necessary of super-io.
31969 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio)
31972 + struct dentry *wh_dentry;
31975 + wh_dentry = vfsub_lkup_one(wh_name, h_parent);
31977 + wh_dentry = au_sio_lkup_one(wh_name, h_parent);
31978 + err = PTR_ERR(wh_dentry);
31979 + if (IS_ERR(wh_dentry)) {
31980 + if (err == -ENAMETOOLONG)
31986 + if (d_is_negative(wh_dentry))
31987 + goto out_wh; /* success */
31990 + if (d_is_reg(wh_dentry))
31991 + goto out_wh; /* success */
31994 + AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
31995 + wh_dentry, d_inode(wh_dentry)->i_mode);
32004 + * test if the @h_dentry sets opaque or not.
32006 +int au_diropq_test(struct dentry *h_dentry)
32009 + struct inode *h_dir;
32011 + h_dir = d_inode(h_dentry);
32012 + err = au_wh_test(h_dentry, &diropq_name,
32013 + au_test_h_perm_sio(h_dir, MAY_EXEC));
32018 + * returns a negative dentry whose name is unique and temporary.
32020 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
32021 + struct qstr *prefix)
32023 + struct dentry *dentry;
32025 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
32027 + /* strict atomic_t is unnecessary here */
32028 + static unsigned short cnt;
32031 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
32034 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
32035 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
32036 + dentry = ERR_PTR(-ENAMETOOLONG);
32037 + if (unlikely(qs.len > NAME_MAX))
32039 + dentry = ERR_PTR(-ENOMEM);
32040 + name = kmalloc(qs.len + 1, GFP_NOFS);
32041 + if (unlikely(!name))
32045 + /* doubly whiteout-ed */
32046 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
32047 + p = name + AUFS_WH_PFX_LEN * 2;
32048 + memcpy(p, prefix->name, prefix->len);
32049 + p += prefix->len;
32051 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
32054 + for (i = 0; i < 3; i++) {
32055 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
32056 + dentry = au_sio_lkup_one(&qs, h_parent);
32057 + if (IS_ERR(dentry) || d_is_negative(dentry))
32061 + /* pr_warn("could not get random name\n"); */
32062 + dentry = ERR_PTR(-EEXIST);
32063 + AuDbg("%.*s\n", AuLNPair(&qs));
32067 + if (name != defname)
32070 + AuTraceErrPtr(dentry);
32075 + * rename the @h_dentry on @br to the whiteouted temporary name.
32077 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
32080 + struct path h_path = {
32081 + .mnt = au_br_mnt(br)
32083 + struct inode *h_dir, *delegated;
32084 + struct dentry *h_parent;
32086 + h_parent = h_dentry->d_parent; /* dir inode is locked */
32087 + h_dir = d_inode(h_parent);
32088 + IMustLock(h_dir);
32090 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
32091 + err = PTR_ERR(h_path.dentry);
32092 + if (IS_ERR(h_path.dentry))
32095 + /* under the same dir, no need to lock_rename() */
32096 + delegated = NULL;
32097 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated);
32099 + if (unlikely(err == -EWOULDBLOCK)) {
32100 + pr_warn("cannot retry for NFSv4 delegation"
32101 + " for an internal rename\n");
32104 + dput(h_path.dentry);
32111 +/* ---------------------------------------------------------------------- */
32113 + * functions for removing a whiteout
32116 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
32119 + struct inode *delegated;
32122 + * forces superio when the dir has a sticky bit.
32123 + * this may be a violation of unix fs semantics.
32125 + force = (h_dir->i_mode & S_ISVTX)
32126 + && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
32127 + delegated = NULL;
32128 + err = vfsub_unlink(h_dir, h_path, &delegated, force);
32129 + if (unlikely(err == -EWOULDBLOCK)) {
32130 + pr_warn("cannot retry for NFSv4 delegation"
32131 + " for an internal unlink\n");
32137 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
32138 + struct dentry *dentry)
32142 + err = do_unlink_wh(h_dir, h_path);
32143 + if (!err && dentry)
32144 + au_set_dbwh(dentry, -1);
32149 +static int unlink_wh_name(struct dentry *h_parent, struct qstr *wh,
32150 + struct au_branch *br)
32153 + struct path h_path = {
32154 + .mnt = au_br_mnt(br)
32158 + h_path.dentry = vfsub_lkup_one(wh, h_parent);
32159 + if (IS_ERR(h_path.dentry))
32160 + err = PTR_ERR(h_path.dentry);
32162 + if (d_is_reg(h_path.dentry))
32163 + err = do_unlink_wh(d_inode(h_parent), &h_path);
32164 + dput(h_path.dentry);
32170 +/* ---------------------------------------------------------------------- */
32172 + * initialize/clean whiteout for a branch
32175 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
32179 + struct inode *delegated;
32181 + if (d_is_negative(whpath->dentry))
32185 + err = vfsub_rmdir(h_dir, whpath);
32187 + delegated = NULL;
32188 + err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
32189 + if (unlikely(err == -EWOULDBLOCK)) {
32190 + pr_warn("cannot retry for NFSv4 delegation"
32191 + " for an internal unlink\n");
32195 + if (unlikely(err))
32196 + pr_warn("failed removing %pd (%d), ignored.\n",
32197 + whpath->dentry, err);
32200 +static int test_linkable(struct dentry *h_root)
32202 + struct inode *h_dir = d_inode(h_root);
32204 + if (h_dir->i_op->link)
32207 + pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
32208 + h_root, au_sbtype(h_root->d_sb));
32212 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
32213 +static int au_whdir(struct inode *h_dir, struct path *path)
32218 + if (d_is_negative(path->dentry)) {
32219 + int mode = S_IRWXU;
32221 + if (au_test_nfs(path->dentry->d_sb))
32223 + err = vfsub_mkdir(h_dir, path, mode);
32224 + } else if (d_is_dir(path->dentry))
32227 + pr_err("unknown %pd exists\n", path->dentry);
32232 +struct au_wh_base {
32233 + const struct qstr *name;
32234 + struct dentry *dentry;
32237 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
32238 + struct path *h_path)
32240 + h_path->dentry = base[AuBrWh_BASE].dentry;
32241 + au_wh_clean(h_dir, h_path, /*isdir*/0);
32242 + h_path->dentry = base[AuBrWh_PLINK].dentry;
32243 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32244 + h_path->dentry = base[AuBrWh_ORPH].dentry;
32245 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32249 + * returns tri-state,
32250 + * minus: error, caller should print the message
32252 + * plus: error, caller should NOT print the message
32254 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
32255 + int do_plink, struct au_wh_base base[],
32256 + struct path *h_path)
32259 + struct inode *h_dir;
32261 + h_dir = d_inode(h_root);
32262 + h_path->dentry = base[AuBrWh_BASE].dentry;
32263 + au_wh_clean(h_dir, h_path, /*isdir*/0);
32264 + h_path->dentry = base[AuBrWh_PLINK].dentry;
32266 + err = test_linkable(h_root);
32267 + if (unlikely(err)) {
32272 + err = au_whdir(h_dir, h_path);
32273 + if (unlikely(err))
32275 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
32277 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32278 + h_path->dentry = base[AuBrWh_ORPH].dentry;
32279 + err = au_whdir(h_dir, h_path);
32280 + if (unlikely(err))
32282 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
32289 + * for the moment, aufs supports the branch filesystem which does not support
32290 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
32291 + * copyup failed. finally, such filesystem will not be used as the writable
32294 + * returns tri-state, see above.
32296 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
32297 + int do_plink, struct au_wh_base base[],
32298 + struct path *h_path)
32301 + struct inode *h_dir;
32303 + WbrWhMustWriteLock(wbr);
32305 + err = test_linkable(h_root);
32306 + if (unlikely(err)) {
32312 + * todo: should this create be done in /sbin/mount.aufs helper?
32315 + h_dir = d_inode(h_root);
32316 + if (d_is_negative(base[AuBrWh_BASE].dentry)) {
32317 + h_path->dentry = base[AuBrWh_BASE].dentry;
32318 + err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
32319 + } else if (d_is_reg(base[AuBrWh_BASE].dentry))
32322 + pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
32323 + if (unlikely(err))
32326 + h_path->dentry = base[AuBrWh_PLINK].dentry;
32328 + err = au_whdir(h_dir, h_path);
32329 + if (unlikely(err))
32331 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
32333 + au_wh_clean(h_dir, h_path, /*isdir*/1);
32334 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
32336 + h_path->dentry = base[AuBrWh_ORPH].dentry;
32337 + err = au_whdir(h_dir, h_path);
32338 + if (unlikely(err))
32340 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
32347 + * initialize the whiteout base file/dir for @br.
32349 +int au_wh_init(struct au_branch *br, struct super_block *sb)
32352 + const unsigned char do_plink
32353 + = !!au_opt_test(au_mntflags(sb), PLINK);
32354 + struct inode *h_dir;
32355 + struct path path = br->br_path;
32356 + struct dentry *h_root = path.dentry;
32357 + struct au_wbr *wbr = br->br_wbr;
32358 + static const struct qstr base_name[] = {
32359 + [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
32360 + sizeof(AUFS_BASE_NAME) - 1),
32361 + [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
32362 + sizeof(AUFS_PLINKDIR_NAME) - 1),
32363 + [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
32364 + sizeof(AUFS_ORPHDIR_NAME) - 1)
32366 + struct au_wh_base base[] = {
32367 + [AuBrWh_BASE] = {
32368 + .name = base_name + AuBrWh_BASE,
32371 + [AuBrWh_PLINK] = {
32372 + .name = base_name + AuBrWh_PLINK,
32375 + [AuBrWh_ORPH] = {
32376 + .name = base_name + AuBrWh_ORPH,
32382 + WbrWhMustWriteLock(wbr);
32384 + for (i = 0; i < AuBrWh_Last; i++) {
32385 + /* doubly whiteouted */
32386 + struct dentry *d;
32388 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
32389 + err = PTR_ERR(d);
32393 + base[i].dentry = d;
32395 + && wbr->wbr_wh[i]
32396 + && wbr->wbr_wh[i] != base[i].dentry);
32400 + for (i = 0; i < AuBrWh_Last; i++) {
32401 + dput(wbr->wbr_wh[i]);
32402 + wbr->wbr_wh[i] = NULL;
32406 + if (!au_br_writable(br->br_perm)) {
32407 + h_dir = d_inode(h_root);
32408 + au_wh_init_ro(h_dir, base, &path);
32409 + } else if (!au_br_wh_linkable(br->br_perm)) {
32410 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
32416 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
32422 + goto out; /* success */
32425 + pr_err("an error(%d) on the writable branch %pd(%s)\n",
32426 + err, h_root, au_sbtype(h_root->d_sb));
32428 + for (i = 0; i < AuBrWh_Last; i++)
32429 + dput(base[i].dentry);
32433 +/* ---------------------------------------------------------------------- */
32435 + * whiteouts are all hard-linked usually.
32436 + * when its link count reaches a ceiling, we create a new whiteout base
32437 + * asynchronously.
32440 +struct reinit_br_wh {
32441 + struct super_block *sb;
32442 + struct au_branch *br;
32445 +static void reinit_br_wh(void *arg)
32448 + aufs_bindex_t bindex;
32449 + struct path h_path;
32450 + struct reinit_br_wh *a = arg;
32451 + struct au_wbr *wbr;
32452 + struct inode *dir, *delegated;
32453 + struct dentry *h_root;
32454 + struct au_hinode *hdir;
32457 + wbr = a->br->br_wbr;
32458 + /* big aufs lock */
32459 + si_noflush_write_lock(a->sb);
32460 + if (!au_br_writable(a->br->br_perm))
32462 + bindex = au_br_index(a->sb, a->br->br_id);
32463 + if (unlikely(bindex < 0))
32466 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
32467 + dir = d_inode(a->sb->s_root);
32468 + hdir = au_hi(dir, bindex);
32469 + h_root = au_h_dptr(a->sb->s_root, bindex);
32470 + AuDebugOn(h_root != au_br_dentry(a->br));
32472 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
32473 + wbr_wh_write_lock(wbr);
32474 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
32477 + h_path.dentry = wbr->wbr_whbase;
32478 + h_path.mnt = au_br_mnt(a->br);
32479 + delegated = NULL;
32480 + err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
32482 + if (unlikely(err == -EWOULDBLOCK)) {
32483 + pr_warn("cannot retry for NFSv4 delegation"
32484 + " for an internal unlink\n");
32488 + pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
32491 + dput(wbr->wbr_whbase);
32492 + wbr->wbr_whbase = NULL;
32494 + err = au_wh_init(a->br, a->sb);
32495 + wbr_wh_write_unlock(wbr);
32496 + au_hn_imtx_unlock(hdir);
32497 + di_read_unlock(a->sb->s_root, AuLock_IR);
32499 + au_fhsm_wrote(a->sb, bindex, /*force*/0);
32503 + atomic_dec(&wbr->wbr_wh_running);
32504 + atomic_dec(&a->br->br_count);
32505 + si_write_unlock(a->sb);
32506 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
32508 + if (unlikely(err))
32509 + AuIOErr("err %d\n", err);
32512 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
32514 + int do_dec, wkq_err;
32515 + struct reinit_br_wh *arg;
32518 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
32521 + /* ignore ENOMEM */
32522 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
32525 + * dec(wh_running), kfree(arg) and dec(br_count)
32526 + * in reinit function
32530 + atomic_inc(&br->br_count);
32531 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
32532 + if (unlikely(wkq_err)) {
32533 + atomic_dec(&br->br_wbr->wbr_wh_running);
32534 + atomic_dec(&br->br_count);
32542 + atomic_dec(&br->br_wbr->wbr_wh_running);
32545 +/* ---------------------------------------------------------------------- */
32548 + * create the whiteout @wh.
32550 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
32551 + struct dentry *wh)
32554 + struct path h_path = {
32557 + struct au_branch *br;
32558 + struct au_wbr *wbr;
32559 + struct dentry *h_parent;
32560 + struct inode *h_dir, *delegated;
32562 + h_parent = wh->d_parent; /* dir inode is locked */
32563 + h_dir = d_inode(h_parent);
32564 + IMustLock(h_dir);
32566 + br = au_sbr(sb, bindex);
32567 + h_path.mnt = au_br_mnt(br);
32568 + wbr = br->br_wbr;
32569 + wbr_wh_read_lock(wbr);
32570 + if (wbr->wbr_whbase) {
32571 + delegated = NULL;
32572 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
32573 + if (unlikely(err == -EWOULDBLOCK)) {
32574 + pr_warn("cannot retry for NFSv4 delegation"
32575 + " for an internal link\n");
32578 + if (!err || err != -EMLINK)
32581 + /* link count full. re-initialize br_whbase. */
32582 + kick_reinit_br_wh(sb, br);
32585 + /* return this error in this context */
32586 + err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
32588 + au_fhsm_wrote(sb, bindex, /*force*/0);
32591 + wbr_wh_read_unlock(wbr);
32595 +/* ---------------------------------------------------------------------- */
32598 + * create or remove the diropq.
32600 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
32601 + unsigned int flags)
32603 + struct dentry *opq_dentry, *h_dentry;
32604 + struct super_block *sb;
32605 + struct au_branch *br;
32608 + sb = dentry->d_sb;
32609 + br = au_sbr(sb, bindex);
32610 + h_dentry = au_h_dptr(dentry, bindex);
32611 + opq_dentry = vfsub_lkup_one(&diropq_name, h_dentry);
32612 + if (IS_ERR(opq_dentry))
32615 + if (au_ftest_diropq(flags, CREATE)) {
32616 + err = link_or_create_wh(sb, bindex, opq_dentry);
32618 + au_set_dbdiropq(dentry, bindex);
32619 + goto out; /* success */
32622 + struct path tmp = {
32623 + .dentry = opq_dentry,
32624 + .mnt = au_br_mnt(br)
32626 + err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &tmp);
32628 + au_set_dbdiropq(dentry, -1);
32630 + dput(opq_dentry);
32631 + opq_dentry = ERR_PTR(err);
32634 + return opq_dentry;
32637 +struct do_diropq_args {
32638 + struct dentry **errp;
32639 + struct dentry *dentry;
32640 + aufs_bindex_t bindex;
32641 + unsigned int flags;
32644 +static void call_do_diropq(void *args)
32646 + struct do_diropq_args *a = args;
32647 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
32650 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
32651 + unsigned int flags)
32653 + struct dentry *diropq, *h_dentry;
32655 + h_dentry = au_h_dptr(dentry, bindex);
32656 + if (!au_test_h_perm_sio(d_inode(h_dentry), MAY_EXEC | MAY_WRITE))
32657 + diropq = do_diropq(dentry, bindex, flags);
32660 + struct do_diropq_args args = {
32662 + .dentry = dentry,
32663 + .bindex = bindex,
32667 + wkq_err = au_wkq_wait(call_do_diropq, &args);
32668 + if (unlikely(wkq_err))
32669 + diropq = ERR_PTR(wkq_err);
32675 +/* ---------------------------------------------------------------------- */
32678 + * lookup whiteout dentry.
32679 + * @h_parent: lower parent dentry which must exist and be locked
32680 + * @base_name: name of dentry which will be whiteouted
32681 + * returns dentry for whiteout.
32683 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
32684 + struct au_branch *br)
32687 + struct qstr wh_name;
32688 + struct dentry *wh_dentry;
32690 + err = au_wh_name_alloc(&wh_name, base_name);
32691 + wh_dentry = ERR_PTR(err);
32693 + wh_dentry = vfsub_lkup_one(&wh_name, h_parent);
32694 + kfree(wh_name.name);
32696 + return wh_dentry;
32700 + * link/create a whiteout for @dentry on @bindex.
32702 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
32703 + struct dentry *h_parent)
32705 + struct dentry *wh_dentry;
32706 + struct super_block *sb;
32709 + sb = dentry->d_sb;
32710 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
32711 + if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
32712 + err = link_or_create_wh(sb, bindex, wh_dentry);
32714 + au_set_dbwh(dentry, bindex);
32715 + au_fhsm_wrote(sb, bindex, /*force*/0);
32718 + wh_dentry = ERR_PTR(err);
32722 + return wh_dentry;
32725 +/* ---------------------------------------------------------------------- */
32727 +/* Delete all whiteouts in this directory on branch bindex. */
32728 +static int del_wh_children(struct dentry *h_dentry, struct au_nhash *whlist,
32729 + aufs_bindex_t bindex, struct au_branch *br)
32732 + unsigned long ul, n;
32733 + struct qstr wh_name;
32735 + struct hlist_head *head;
32736 + struct au_vdir_wh *pos;
32737 + struct au_vdir_destr *str;
32740 + p = (void *)__get_free_page(GFP_NOFS);
32741 + wh_name.name = p;
32742 + if (unlikely(!wh_name.name))
32746 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32747 + p += AUFS_WH_PFX_LEN;
32748 + n = whlist->nh_num;
32749 + head = whlist->nh_head;
32750 + for (ul = 0; !err && ul < n; ul++, head++) {
32751 + hlist_for_each_entry(pos, head, wh_hash) {
32752 + if (pos->wh_bindex != bindex)
32755 + str = &pos->wh_str;
32756 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
32757 + memcpy(p, str->name, str->len);
32758 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
32759 + err = unlink_wh_name(h_dentry, &wh_name, br);
32764 + AuIOErr("whiteout name too long %.*s\n",
32765 + str->len, str->name);
32770 + free_page((unsigned long)wh_name.name);
32776 +struct del_wh_children_args {
32778 + struct dentry *h_dentry;
32779 + struct au_nhash *whlist;
32780 + aufs_bindex_t bindex;
32781 + struct au_branch *br;
32784 +static void call_del_wh_children(void *args)
32786 + struct del_wh_children_args *a = args;
32787 + *a->errp = del_wh_children(a->h_dentry, a->whlist, a->bindex, a->br);
32790 +/* ---------------------------------------------------------------------- */
32792 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
32794 + struct au_whtmp_rmdir *whtmp;
32796 + unsigned int rdhash;
32798 + SiMustAnyLock(sb);
32800 + whtmp = kmalloc(sizeof(*whtmp), gfp);
32801 + if (unlikely(!whtmp)) {
32802 + whtmp = ERR_PTR(-ENOMEM);
32806 + whtmp->dir = NULL;
32807 + whtmp->br = NULL;
32808 + whtmp->wh_dentry = NULL;
32809 + /* no estimation for dir size */
32810 + rdhash = au_sbi(sb)->si_rdhash;
32812 + rdhash = AUFS_RDHASH_DEF;
32813 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
32814 + if (unlikely(err)) {
32816 + whtmp = ERR_PTR(err);
32823 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
32826 + atomic_dec(&whtmp->br->br_count);
32827 + dput(whtmp->wh_dentry);
32828 + iput(whtmp->dir);
32829 + au_nhash_wh_free(&whtmp->whlist);
32834 + * rmdir the whiteouted temporary named dir @h_dentry.
32835 + * @whlist: whiteouted children.
32837 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
32838 + struct dentry *wh_dentry, struct au_nhash *whlist)
32841 + unsigned int h_nlink;
32842 + struct path h_tmp;
32843 + struct inode *wh_inode, *h_dir;
32844 + struct au_branch *br;
32846 + h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
32847 + IMustLock(h_dir);
32849 + br = au_sbr(dir->i_sb, bindex);
32850 + wh_inode = d_inode(wh_dentry);
32851 + mutex_lock_nested(&wh_inode->i_mutex, AuLsc_I_CHILD);
32854 + * someone else might change some whiteouts while we were sleeping.
32855 + * it means this whlist may have an obsoleted entry.
32857 + if (!au_test_h_perm_sio(wh_inode, MAY_EXEC | MAY_WRITE))
32858 + err = del_wh_children(wh_dentry, whlist, bindex, br);
32861 + struct del_wh_children_args args = {
32863 + .h_dentry = wh_dentry,
32864 + .whlist = whlist,
32865 + .bindex = bindex,
32869 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
32870 + if (unlikely(wkq_err))
32873 + mutex_unlock(&wh_inode->i_mutex);
32876 + h_tmp.dentry = wh_dentry;
32877 + h_tmp.mnt = au_br_mnt(br);
32878 + h_nlink = h_dir->i_nlink;
32879 + err = vfsub_rmdir(h_dir, &h_tmp);
32880 + /* some fs doesn't change the parent nlink in some cases */
32881 + h_nlink -= h_dir->i_nlink;
32885 + if (au_ibstart(dir) == bindex) {
32886 + /* todo: dir->i_mutex is necessary */
32887 + au_cpup_attr_timesizes(dir);
32889 + vfsub_drop_nlink(dir);
32891 + return 0; /* success */
32894 + pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
32898 +static void call_rmdir_whtmp(void *args)
32901 + aufs_bindex_t bindex;
32902 + struct au_whtmp_rmdir *a = args;
32903 + struct super_block *sb;
32904 + struct dentry *h_parent;
32905 + struct inode *h_dir;
32906 + struct au_hinode *hdir;
32908 + /* rmdir by nfsd may cause deadlock with this i_mutex */
32909 + /* mutex_lock(&a->dir->i_mutex); */
32911 + sb = a->dir->i_sb;
32912 + si_read_lock(sb, !AuLock_FLUSH);
32913 + if (!au_br_writable(a->br->br_perm))
32915 + bindex = au_br_index(sb, a->br->br_id);
32916 + if (unlikely(bindex < 0))
32920 + ii_write_lock_parent(a->dir);
32921 + h_parent = dget_parent(a->wh_dentry);
32922 + h_dir = d_inode(h_parent);
32923 + hdir = au_hi(a->dir, bindex);
32924 + err = vfsub_mnt_want_write(au_br_mnt(a->br));
32925 + if (unlikely(err))
32927 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
32928 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
32931 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
32932 + au_hn_imtx_unlock(hdir);
32933 + vfsub_mnt_drop_write(au_br_mnt(a->br));
32937 + ii_write_unlock(a->dir);
32939 + /* mutex_unlock(&a->dir->i_mutex); */
32940 + au_whtmp_rmdir_free(a);
32941 + si_read_unlock(sb);
32942 + au_nwt_done(&au_sbi(sb)->si_nowait);
32943 + if (unlikely(err))
32944 + AuIOErr("err %d\n", err);
32947 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
32948 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
32951 + struct super_block *sb;
32955 + /* all post-process will be done in do_rmdir_whtmp(). */
32957 + args->dir = au_igrab(dir);
32958 + args->br = au_sbr(sb, bindex);
32959 + atomic_inc(&args->br->br_count);
32960 + args->wh_dentry = dget(wh_dentry);
32961 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
32962 + if (unlikely(wkq_err)) {
32963 + pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
32964 + au_whtmp_rmdir_free(args);
32967 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
32968 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
32969 +++ linux/fs/aufs/whout.h 2015-09-24 10:47:58.258053165 +0200
32972 + * Copyright (C) 2005-2015 Junjiro R. Okajima
32974 + * This program, aufs is free software; you can redistribute it and/or modify
32975 + * it under the terms of the GNU General Public License as published by
32976 + * the Free Software Foundation; either version 2 of the License, or
32977 + * (at your option) any later version.
32979 + * This program is distributed in the hope that it will be useful,
32980 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
32981 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32982 + * GNU General Public License for more details.
32984 + * You should have received a copy of the GNU General Public License
32985 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
32989 + * whiteout for logical deletion and opaque directory
32992 +#ifndef __AUFS_WHOUT_H__
32993 +#define __AUFS_WHOUT_H__
33000 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
33001 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio);
33002 +int au_diropq_test(struct dentry *h_dentry);
33004 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
33005 + struct qstr *prefix);
33006 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
33007 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
33008 + struct dentry *dentry);
33009 +int au_wh_init(struct au_branch *br, struct super_block *sb);
33011 +/* diropq flags */
33012 +#define AuDiropq_CREATE 1
33013 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
33014 +#define au_fset_diropq(flags, name) \
33015 + do { (flags) |= AuDiropq_##name; } while (0)
33016 +#define au_fclr_diropq(flags, name) \
33017 + do { (flags) &= ~AuDiropq_##name; } while (0)
33019 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
33020 + unsigned int flags);
33021 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
33022 + struct au_branch *br);
33023 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
33024 + struct dentry *h_parent);
33026 +/* real rmdir for the whiteout-ed dir */
33027 +struct au_whtmp_rmdir {
33028 + struct inode *dir;
33029 + struct au_branch *br;
33030 + struct dentry *wh_dentry;
33031 + struct au_nhash whlist;
33034 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
33035 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
33036 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
33037 + struct dentry *wh_dentry, struct au_nhash *whlist);
33038 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
33039 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
33041 +/* ---------------------------------------------------------------------- */
33043 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
33044 + aufs_bindex_t bindex)
33046 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
33049 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
33051 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
33054 +#endif /* __KERNEL__ */
33055 +#endif /* __AUFS_WHOUT_H__ */
33056 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
33057 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
33058 +++ linux/fs/aufs/wkq.c 2015-09-24 10:47:58.258053165 +0200
33061 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33063 + * This program, aufs is free software; you can redistribute it and/or modify
33064 + * it under the terms of the GNU General Public License as published by
33065 + * the Free Software Foundation; either version 2 of the License, or
33066 + * (at your option) any later version.
33068 + * This program is distributed in the hope that it will be useful,
33069 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33070 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33071 + * GNU General Public License for more details.
33073 + * You should have received a copy of the GNU General Public License
33074 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33078 + * workqueue for asynchronous/super-io operations
33079 + * todo: try new dredential scheme
33082 +#include <linux/module.h>
33085 +/* internal workqueue named AUFS_WKQ_NAME */
33087 +static struct workqueue_struct *au_wkq;
33089 +struct au_wkinfo {
33090 + struct work_struct wk;
33091 + struct kobject *kobj;
33093 + unsigned int flags; /* see wkq.h */
33095 + au_wkq_func_t func;
33098 + struct completion *comp;
33101 +/* ---------------------------------------------------------------------- */
33103 +static void wkq_func(struct work_struct *wk)
33105 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
33107 + AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
33108 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
33110 + wkinfo->func(wkinfo->args);
33111 + if (au_ftest_wkq(wkinfo->flags, WAIT))
33112 + complete(wkinfo->comp);
33114 + kobject_put(wkinfo->kobj);
33115 + module_put(THIS_MODULE); /* todo: ?? */
33121 + * Since struct completion is large, try allocating it dynamically.
33123 +#if 1 /* defined(CONFIG_4KSTACKS) || defined(AuTest4KSTACKS) */
33124 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
33126 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
33128 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
33130 + init_completion(*comp);
33131 + wkinfo->comp = *comp;
33137 +static void au_wkq_comp_free(struct completion *comp)
33145 +#define AuWkqCompDeclare(name) \
33146 + DECLARE_COMPLETION_ONSTACK(_ ## name); \
33147 + struct completion *comp = &_ ## name
33149 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
33151 + wkinfo->comp = *comp;
33155 +static void au_wkq_comp_free(struct completion *comp __maybe_unused)
33159 +#endif /* 4KSTACKS */
33161 +static void au_wkq_run(struct au_wkinfo *wkinfo)
33163 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
33164 + if (au_wkq_test()) {
33165 + AuWarn1("wkq from wkq, unless silly-rename on NFS,"
33166 + " due to a dead dir by UDBA?\n");
33167 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
33170 + au_dbg_verify_kthread();
33172 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
33173 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
33174 + queue_work(au_wkq, &wkinfo->wk);
33176 + INIT_WORK(&wkinfo->wk, wkq_func);
33177 + schedule_work(&wkinfo->wk);
33182 + * Be careful. It is easy to make deadlock happen.
33183 + * processA: lock, wkq and wait
33184 + * processB: wkq and wait, lock in wkq
33187 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
33190 + AuWkqCompDeclare(comp);
33191 + struct au_wkinfo wkinfo = {
33197 + err = au_wkq_comp_alloc(&wkinfo, &comp);
33199 + au_wkq_run(&wkinfo);
33200 + /* no timeout, no interrupt */
33201 + wait_for_completion(wkinfo.comp);
33202 + au_wkq_comp_free(comp);
33203 + destroy_work_on_stack(&wkinfo.wk);
33211 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
33212 + * problem in a concurrent umounting.
33214 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
33215 + unsigned int flags)
33218 + struct au_wkinfo *wkinfo;
33220 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
33223 + * wkq_func() must free this wkinfo.
33224 + * it highly depends upon the implementation of workqueue.
33227 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
33229 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
33230 + wkinfo->flags = flags & ~AuWkq_WAIT;
33231 + wkinfo->func = func;
33232 + wkinfo->args = args;
33233 + wkinfo->comp = NULL;
33234 + kobject_get(wkinfo->kobj);
33235 + __module_get(THIS_MODULE); /* todo: ?? */
33237 + au_wkq_run(wkinfo);
33240 + au_nwt_done(&au_sbi(sb)->si_nowait);
33246 +/* ---------------------------------------------------------------------- */
33248 +void au_nwt_init(struct au_nowait_tasks *nwt)
33250 + atomic_set(&nwt->nw_len, 0);
33251 + /* smp_mb(); */ /* atomic_set */
33252 + init_waitqueue_head(&nwt->nw_wq);
33255 +void au_wkq_fin(void)
33257 + destroy_workqueue(au_wkq);
33260 +int __init au_wkq_init(void)
33265 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
33266 + if (IS_ERR(au_wkq))
33267 + err = PTR_ERR(au_wkq);
33268 + else if (!au_wkq)
33273 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
33274 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
33275 +++ linux/fs/aufs/wkq.h 2015-09-24 10:47:58.258053165 +0200
33278 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33280 + * This program, aufs is free software; you can redistribute it and/or modify
33281 + * it under the terms of the GNU General Public License as published by
33282 + * the Free Software Foundation; either version 2 of the License, or
33283 + * (at your option) any later version.
33285 + * This program is distributed in the hope that it will be useful,
33286 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33287 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33288 + * GNU General Public License for more details.
33290 + * You should have received a copy of the GNU General Public License
33291 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33295 + * workqueue for asynchronous/super-io operations
33296 + * todo: try new credentials management scheme
33299 +#ifndef __AUFS_WKQ_H__
33300 +#define __AUFS_WKQ_H__
33304 +struct super_block;
33306 +/* ---------------------------------------------------------------------- */
33309 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
33311 +struct au_nowait_tasks {
33313 + wait_queue_head_t nw_wq;
33316 +/* ---------------------------------------------------------------------- */
33318 +typedef void (*au_wkq_func_t)(void *args);
33321 +#define AuWkq_WAIT 1
33322 +#define AuWkq_NEST (1 << 1)
33323 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
33324 +#define au_fset_wkq(flags, name) \
33325 + do { (flags) |= AuWkq_##name; } while (0)
33326 +#define au_fclr_wkq(flags, name) \
33327 + do { (flags) &= ~AuWkq_##name; } while (0)
33329 +#ifndef CONFIG_AUFS_HNOTIFY
33331 +#define AuWkq_NEST 0
33335 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
33336 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
33337 + unsigned int flags);
33338 +void au_nwt_init(struct au_nowait_tasks *nwt);
33339 +int __init au_wkq_init(void);
33340 +void au_wkq_fin(void);
33342 +/* ---------------------------------------------------------------------- */
33344 +static inline int au_wkq_test(void)
33346 + return current->flags & PF_WQ_WORKER;
33349 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
33351 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
33354 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
33356 + if (atomic_dec_and_test(&nwt->nw_len))
33357 + wake_up_all(&nwt->nw_wq);
33360 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
33362 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
33366 +#endif /* __KERNEL__ */
33367 +#endif /* __AUFS_WKQ_H__ */
33368 diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
33369 --- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
33370 +++ linux/fs/aufs/xattr.c 2015-09-24 10:47:58.258053165 +0200
33373 + * Copyright (C) 2014-2015 Junjiro R. Okajima
33375 + * This program, aufs is free software; you can redistribute it and/or modify
33376 + * it under the terms of the GNU General Public License as published by
33377 + * the Free Software Foundation; either version 2 of the License, or
33378 + * (at your option) any later version.
33380 + * This program is distributed in the hope that it will be useful,
33381 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33382 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33383 + * GNU General Public License for more details.
33385 + * You should have received a copy of the GNU General Public License
33386 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33390 + * handling xattr functions
33393 +#include <linux/xattr.h>
33396 +static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
33398 + if (!ignore_flags)
33406 + if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
33411 +#define cmp(brattr, prefix) do { \
33412 + if (!strncmp(name, XATTR_##prefix##_PREFIX, \
33413 + XATTR_##prefix##_PREFIX_LEN)) { \
33414 + if (ignore_flags & AuBrAttr_ICEX_##brattr) \
33420 + cmp(SEC, SECURITY);
33421 + cmp(SYS, SYSTEM);
33422 + cmp(TR, TRUSTED);
33426 + if (ignore_flags & AuBrAttr_ICEX_OTH)
33433 +static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
33435 +static int au_do_cpup_xattr(struct dentry *h_dst, struct dentry *h_src,
33436 + char *name, char **buf, unsigned int ignore_flags,
33437 + unsigned int verbose)
33441 + struct inode *h_idst;
33443 + ssz = vfs_getxattr_alloc(h_src, name, buf, 0, GFP_NOFS);
33445 + if (unlikely(err <= 0)) {
33446 + if (err == -ENODATA
33447 + || (err == -EOPNOTSUPP
33448 + && ((ignore_flags & au_xattr_out_of_list)
33449 + || (au_test_nfs_noacl(d_inode(h_src))
33450 + && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
33452 + XATTR_NAME_POSIX_ACL_DEFAULT))))
33455 + if (err && (verbose || au_debug_test()))
33456 + pr_err("%s, err %d\n", name, err);
33460 + /* unlock it temporary */
33461 + h_idst = d_inode(h_dst);
33462 + mutex_unlock(&h_idst->i_mutex);
33463 + err = vfsub_setxattr(h_dst, name, *buf, ssz, /*flags*/0);
33464 + mutex_lock_nested(&h_idst->i_mutex, AuLsc_I_CHILD2);
33465 + if (unlikely(err)) {
33466 + if (verbose || au_debug_test())
33467 + pr_err("%s, err %d\n", name, err);
33468 + err = au_xattr_ignore(err, name, ignore_flags);
33475 +int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
33476 + unsigned int verbose)
33478 + int err, unlocked, acl_access, acl_default;
33480 + struct inode *h_isrc, *h_idst;
33481 + char *value, *p, *o, *e;
33483 + /* try stopping to update the source inode while we are referencing */
33484 + /* there should not be the parent-child relationship between them */
33485 + h_isrc = d_inode(h_src);
33486 + h_idst = d_inode(h_dst);
33487 + mutex_unlock(&h_idst->i_mutex);
33488 + mutex_lock_nested(&h_isrc->i_mutex, AuLsc_I_CHILD);
33489 + mutex_lock_nested(&h_idst->i_mutex, AuLsc_I_CHILD2);
33492 + /* some filesystems don't list POSIX ACL, for example tmpfs */
33493 + ssz = vfs_listxattr(h_src, NULL, 0);
33495 + if (unlikely(err < 0)) {
33497 + if (err == -ENODATA
33498 + || err == -EOPNOTSUPP)
33499 + err = 0; /* ignore */
33508 + p = kmalloc(ssz, GFP_NOFS);
33510 + if (unlikely(!p))
33512 + err = vfs_listxattr(h_src, p, ssz);
33514 + mutex_unlock(&h_isrc->i_mutex);
33516 + AuDbg("err %d, ssz %zd\n", err, ssz);
33517 + if (unlikely(err < 0))
33525 + while (!err && p < e) {
33526 + acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
33527 + sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
33528 + acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
33529 + sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
33531 + err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
33533 + p += strlen(p) + 1;
33536 + ignore_flags |= au_xattr_out_of_list;
33537 + if (!err && !acl_access) {
33538 + err = au_do_cpup_xattr(h_dst, h_src,
33539 + XATTR_NAME_POSIX_ACL_ACCESS, &value,
33540 + ignore_flags, verbose);
33543 + if (!err && !acl_default) {
33544 + err = au_do_cpup_xattr(h_dst, h_src,
33545 + XATTR_NAME_POSIX_ACL_DEFAULT, &value,
33546 + ignore_flags, verbose);
33556 + mutex_unlock(&h_isrc->i_mutex);
33561 +/* ---------------------------------------------------------------------- */
33568 +struct au_lgxattr {
33576 + const char *name;
33583 +static ssize_t au_lgxattr(struct dentry *dentry, struct au_lgxattr *arg)
33586 + struct path h_path;
33587 + struct super_block *sb;
33589 + sb = dentry->d_sb;
33590 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
33591 + if (unlikely(err))
33593 + err = au_h_path_getattr(dentry, /*force*/1, &h_path);
33594 + if (unlikely(err))
33596 + if (unlikely(!h_path.dentry))
33597 + /* illegally overlapped or something */
33598 + goto out_di; /* pretending success */
33600 + /* always topmost entry only */
33601 + switch (arg->type) {
33602 + case AU_XATTR_LIST:
33603 + err = vfs_listxattr(h_path.dentry,
33604 + arg->u.list.list, arg->u.list.size);
33606 + case AU_XATTR_GET:
33607 + err = vfs_getxattr(h_path.dentry,
33608 + arg->u.get.name, arg->u.get.value,
33609 + arg->u.get.size);
33614 + di_read_unlock(dentry, AuLock_IR);
33616 + si_read_unlock(sb);
33622 +ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
33624 + struct au_lgxattr arg = {
33625 + .type = AU_XATTR_LIST,
33632 + return au_lgxattr(dentry, &arg);
33635 +ssize_t aufs_getxattr(struct dentry *dentry, const char *name, void *value,
33638 + struct au_lgxattr arg = {
33639 + .type = AU_XATTR_GET,
33647 + return au_lgxattr(dentry, &arg);
33650 +int aufs_setxattr(struct dentry *dentry, const char *name, const void *value,
33651 + size_t size, int flags)
33653 + struct au_srxattr arg = {
33654 + .type = AU_XATTR_SET,
33663 + return au_srxattr(dentry, &arg);
33666 +int aufs_removexattr(struct dentry *dentry, const char *name)
33668 + struct au_srxattr arg = {
33669 + .type = AU_XATTR_REMOVE,
33675 + return au_srxattr(dentry, &arg);
33678 +/* ---------------------------------------------------------------------- */
33681 +static size_t au_xattr_list(struct dentry *dentry, char *list, size_t list_size,
33682 + const char *name, size_t name_len, int type)
33684 + return aufs_listxattr(dentry, list, list_size);
33687 +static int au_xattr_get(struct dentry *dentry, const char *name, void *buffer,
33688 + size_t size, int type)
33690 + return aufs_getxattr(dentry, name, buffer, size);
33693 +static int au_xattr_set(struct dentry *dentry, const char *name,
33694 + const void *value, size_t size, int flags, int type)
33696 + return aufs_setxattr(dentry, name, value, size, flags);
33699 +static const struct xattr_handler au_xattr_handler = {
33700 + /* no prefix, no flags */
33701 + .list = au_xattr_list,
33702 + .get = au_xattr_get,
33703 + .set = au_xattr_set
33704 + /* why no remove? */
33707 +static const struct xattr_handler *au_xattr_handlers[] = {
33708 + &au_xattr_handler
33711 +void au_xattr_init(struct super_block *sb)
33713 + /* sb->s_xattr = au_xattr_handlers; */
33716 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
33717 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
33718 +++ linux/fs/aufs/xino.c 2015-11-11 17:21:46.922197217 +0100
33721 + * Copyright (C) 2005-2015 Junjiro R. Okajima
33723 + * This program, aufs is free software; you can redistribute it and/or modify
33724 + * it under the terms of the GNU General Public License as published by
33725 + * the Free Software Foundation; either version 2 of the License, or
33726 + * (at your option) any later version.
33728 + * This program is distributed in the hope that it will be useful,
33729 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
33730 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33731 + * GNU General Public License for more details.
33733 + * You should have received a copy of the GNU General Public License
33734 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
33738 + * external inode number translation table and bitmap
33741 +#include <linux/seq_file.h>
33742 +#include <linux/statfs.h>
33745 +/* todo: unnecessary to support mmap_sem since kernel-space? */
33746 +ssize_t xino_fread(vfs_readf_t func, struct file *file, void *kbuf, size_t size,
33750 + mm_segment_t oldfs;
33757 + oldfs = get_fs();
33758 + set_fs(KERNEL_DS);
33760 + /* todo: signal_pending? */
33761 + err = func(file, buf.u, size, pos);
33762 + } while (err == -EAGAIN || err == -EINTR);
33765 +#if 0 /* reserved for future use */
33767 + fsnotify_access(file->f_path.dentry);
33773 +/* ---------------------------------------------------------------------- */
33775 +static ssize_t do_xino_fwrite(vfs_writef_t func, struct file *file, void *kbuf,
33776 + size_t size, loff_t *pos)
33779 + mm_segment_t oldfs;
33782 + const char __user *u;
33786 + oldfs = get_fs();
33787 + set_fs(KERNEL_DS);
33789 + /* todo: signal_pending? */
33790 + err = func(file, buf.u, size, pos);
33791 + } while (err == -EAGAIN || err == -EINTR);
33794 +#if 0 /* reserved for future use */
33796 + fsnotify_modify(file->f_path.dentry);
33802 +struct do_xino_fwrite_args {
33804 + vfs_writef_t func;
33805 + struct file *file;
33811 +static void call_do_xino_fwrite(void *args)
33813 + struct do_xino_fwrite_args *a = args;
33814 + *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos);
33817 +ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
33818 + size_t size, loff_t *pos)
33822 + /* todo: signal block and no wkq? */
33823 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
33825 + err = do_xino_fwrite(func, file, buf, size, pos);
33829 + * it breaks RLIMIT_FSIZE and normal user's limit,
33830 + * users should care about quota and real 'filesystem full.'
33833 + struct do_xino_fwrite_args args = {
33842 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
33843 + if (unlikely(wkq_err))
33850 +/* ---------------------------------------------------------------------- */
33853 + * create a new xinofile at the same place/path as @base_file.
33855 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src)
33857 + struct file *file;
33858 + struct dentry *base, *parent;
33859 + struct inode *dir, *delegated;
33860 + struct qstr *name;
33861 + struct path path;
33864 + base = base_file->f_path.dentry;
33865 + parent = base->d_parent; /* dir inode is locked */
33866 + dir = d_inode(parent);
33869 + file = ERR_PTR(-EINVAL);
33870 + name = &base->d_name;
33871 + path.dentry = vfsub_lookup_one_len(name->name, parent, name->len);
33872 + if (IS_ERR(path.dentry)) {
33873 + file = (void *)path.dentry;
33874 + pr_err("%pd lookup err %ld\n",
33875 + base, PTR_ERR(path.dentry));
33879 + /* no need to mnt_want_write() since we call dentry_open() later */
33880 + err = vfs_create(dir, path.dentry, S_IRUGO | S_IWUGO, NULL);
33881 + if (unlikely(err)) {
33882 + file = ERR_PTR(err);
33883 + pr_err("%pd create err %d\n", base, err);
33887 + path.mnt = base_file->f_path.mnt;
33888 + file = vfsub_dentry_open(&path,
33889 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
33890 + /* | __FMODE_NONOTIFY */);
33891 + if (IS_ERR(file)) {
33892 + pr_err("%pd open err %ld\n", base, PTR_ERR(file));
33896 + delegated = NULL;
33897 + err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
33898 + if (unlikely(err == -EWOULDBLOCK)) {
33899 + pr_warn("cannot retry for NFSv4 delegation"
33900 + " for an internal unlink\n");
33903 + if (unlikely(err)) {
33904 + pr_err("%pd unlink err %d\n", base, err);
33909 + /* no one can touch copy_src xino */
33910 + err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
33911 + if (unlikely(err)) {
33912 + pr_err("%pd copy err %d\n", base, err);
33916 + goto out_dput; /* success */
33920 + file = ERR_PTR(err);
33922 + dput(path.dentry);
33927 +struct au_xino_lock_dir {
33928 + struct au_hinode *hdir;
33929 + struct dentry *parent;
33930 + struct mutex *mtx;
33933 +static void au_xino_lock_dir(struct super_block *sb, struct file *xino,
33934 + struct au_xino_lock_dir *ldir)
33936 + aufs_bindex_t brid, bindex;
33938 + ldir->hdir = NULL;
33940 + brid = au_xino_brid(sb);
33942 + bindex = au_br_index(sb, brid);
33943 + if (bindex >= 0) {
33944 + ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
33945 + au_hn_imtx_lock_nested(ldir->hdir, AuLsc_I_PARENT);
33947 + ldir->parent = dget_parent(xino->f_path.dentry);
33948 + ldir->mtx = &d_inode(ldir->parent)->i_mutex;
33949 + mutex_lock_nested(ldir->mtx, AuLsc_I_PARENT);
33953 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
33956 + au_hn_imtx_unlock(ldir->hdir);
33958 + mutex_unlock(ldir->mtx);
33959 + dput(ldir->parent);
33963 +/* ---------------------------------------------------------------------- */
33965 +/* trucate xino files asynchronously */
33967 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex)
33970 + unsigned long jiffy;
33972 + aufs_bindex_t bi, bend;
33973 + struct kstatfs *st;
33974 + struct au_branch *br;
33975 + struct file *new_xino, *file;
33976 + struct super_block *h_sb;
33977 + struct au_xino_lock_dir ldir;
33980 + st = kzalloc(sizeof(*st), GFP_NOFS);
33981 + if (unlikely(!st))
33985 + bend = au_sbend(sb);
33986 + if (unlikely(bindex < 0 || bend < bindex))
33988 + br = au_sbr(sb, bindex);
33989 + file = br->br_xino.xi_file;
33993 + err = vfs_statfs(&file->f_path, st);
33994 + if (unlikely(err))
33995 + AuErr1("statfs err %d, ignored\n", err);
33997 + blocks = file_inode(file)->i_blocks;
33998 + pr_info("begin truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
33999 + bindex, (u64)blocks, st->f_bfree, st->f_blocks);
34001 + au_xino_lock_dir(sb, file, &ldir);
34002 + /* mnt_want_write() is unnecessary here */
34003 + new_xino = au_xino_create2(file, file);
34004 + au_xino_unlock_dir(&ldir);
34005 + err = PTR_ERR(new_xino);
34006 + if (IS_ERR(new_xino)) {
34007 + pr_err("err %d, ignored\n", err);
34012 + br->br_xino.xi_file = new_xino;
34014 + h_sb = au_br_sb(br);
34015 + for (bi = 0; bi <= bend; bi++) {
34016 + if (unlikely(bi == bindex))
34018 + br = au_sbr(sb, bi);
34019 + if (au_br_sb(br) != h_sb)
34022 + fput(br->br_xino.xi_file);
34023 + br->br_xino.xi_file = new_xino;
34024 + get_file(new_xino);
34027 + err = vfs_statfs(&new_xino->f_path, st);
34029 + pr_info("end truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
34030 + bindex, (u64)file_inode(new_xino)->i_blocks,
34031 + st->f_bfree, st->f_blocks);
34032 + if (file_inode(new_xino)->i_blocks < blocks)
34033 + au_sbi(sb)->si_xino_jiffy = jiffy;
34035 + AuErr1("statfs err %d, ignored\n", err);
34043 +struct xino_do_trunc_args {
34044 + struct super_block *sb;
34045 + struct au_branch *br;
34048 +static void xino_do_trunc(void *_args)
34050 + struct xino_do_trunc_args *args = _args;
34051 + struct super_block *sb;
34052 + struct au_branch *br;
34053 + struct inode *dir;
34055 + aufs_bindex_t bindex;
34059 + dir = d_inode(sb->s_root);
34062 + si_noflush_write_lock(sb);
34063 + ii_read_lock_parent(dir);
34064 + bindex = au_br_index(sb, br->br_id);
34065 + err = au_xino_trunc(sb, bindex);
34066 + ii_read_unlock(dir);
34067 + if (unlikely(err))
34068 + pr_warn("err b%d, (%d)\n", bindex, err);
34069 + atomic_dec(&br->br_xino_running);
34070 + atomic_dec(&br->br_count);
34071 + si_write_unlock(sb);
34072 + au_nwt_done(&au_sbi(sb)->si_nowait);
34076 +static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
34079 + struct kstatfs st;
34080 + struct au_sbinfo *sbinfo;
34082 + /* todo: si_xino_expire and the ratio should be customizable */
34083 + sbinfo = au_sbi(sb);
34084 + if (time_before(jiffies,
34085 + sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
34088 + /* truncation border */
34089 + err = vfs_statfs(&br->br_xino.xi_file->f_path, &st);
34090 + if (unlikely(err)) {
34091 + AuErr1("statfs err %d, ignored\n", err);
34094 + if (div64_u64(st.f_bfree * 100, st.f_blocks) >= AUFS_XINO_DEF_TRUNC)
34100 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
34102 + struct xino_do_trunc_args *args;
34105 + if (!xino_trunc_test(sb, br))
34108 + if (atomic_inc_return(&br->br_xino_running) > 1)
34111 + /* lock and kfree() will be called in trunc_xino() */
34112 + args = kmalloc(sizeof(*args), GFP_NOFS);
34113 + if (unlikely(!args)) {
34114 + AuErr1("no memory\n");
34118 + atomic_inc(&br->br_count);
34121 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
34123 + return; /* success */
34125 + pr_err("wkq %d\n", wkq_err);
34126 + atomic_dec(&br->br_count);
34131 + atomic_dec(&br->br_xino_running);
34134 +/* ---------------------------------------------------------------------- */
34136 +static int au_xino_do_write(vfs_writef_t write, struct file *file,
34137 + ino_t h_ino, ino_t ino)
34143 + if (unlikely(au_loff_max / sizeof(ino) - 1 < pos)) {
34144 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
34147 + pos *= sizeof(ino);
34148 + sz = xino_fwrite(write, file, &ino, sizeof(ino), &pos);
34149 + if (sz == sizeof(ino))
34150 + return 0; /* success */
34152 + AuIOErr("write failed (%zd)\n", sz);
34157 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
34158 + * at the position of @h_ino.
34159 + * even if @ino is zero, it is written to the xinofile and means no entry.
34160 + * if the size of the xino file on a specific filesystem exceeds the watermark,
34161 + * try truncating it.
34163 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
34167 + unsigned int mnt_flags;
34168 + struct au_branch *br;
34170 + BUILD_BUG_ON(sizeof(long long) != sizeof(au_loff_max)
34171 + || ((loff_t)-1) > 0);
34172 + SiMustAnyLock(sb);
34174 + mnt_flags = au_mntflags(sb);
34175 + if (!au_opt_test(mnt_flags, XINO))
34178 + br = au_sbr(sb, bindex);
34179 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
34182 + if (au_opt_test(mnt_flags, TRUNC_XINO)
34183 + && au_test_fs_trunc_xino(au_br_sb(br)))
34184 + xino_try_trunc(sb, br);
34185 + return 0; /* success */
34188 + AuIOErr("write failed (%d)\n", err);
34192 +/* ---------------------------------------------------------------------- */
34194 +/* aufs inode number bitmap */
34196 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
34197 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
34201 + AuDebugOn(bit < 0 || page_bits <= bit);
34202 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
34206 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
34208 + AuDebugOn(ino < AUFS_FIRST_INO);
34209 + ino -= AUFS_FIRST_INO;
34210 + *pindex = ino / page_bits;
34211 + *bit = ino % page_bits;
34214 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
34219 + struct au_sbinfo *sbinfo;
34220 + struct file *xib;
34221 + unsigned long *p;
34223 + sbinfo = au_sbi(sb);
34224 + MtxMustLock(&sbinfo->si_xib_mtx);
34225 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
34226 + || !au_opt_test(sbinfo->si_mntflags, XINO));
34228 + if (pindex == sbinfo->si_xib_last_pindex)
34231 + xib = sbinfo->si_xib;
34232 + p = sbinfo->si_xib_buf;
34233 + pos = sbinfo->si_xib_last_pindex;
34234 + pos *= PAGE_SIZE;
34235 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
34236 + if (unlikely(sz != PAGE_SIZE))
34240 + pos *= PAGE_SIZE;
34241 + if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
34242 + sz = xino_fread(sbinfo->si_xread, xib, p, PAGE_SIZE, &pos);
34244 + memset(p, 0, PAGE_SIZE);
34245 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
34247 + if (sz == PAGE_SIZE) {
34248 + sbinfo->si_xib_last_pindex = pindex;
34249 + return 0; /* success */
34253 + AuIOErr1("write failed (%zd)\n", sz);
34260 +/* ---------------------------------------------------------------------- */
34262 +static void au_xib_clear_bit(struct inode *inode)
34265 + unsigned long pindex;
34266 + struct super_block *sb;
34267 + struct au_sbinfo *sbinfo;
34269 + AuDebugOn(inode->i_nlink);
34271 + sb = inode->i_sb;
34272 + xib_calc_bit(inode->i_ino, &pindex, &bit);
34273 + AuDebugOn(page_bits <= bit);
34274 + sbinfo = au_sbi(sb);
34275 + mutex_lock(&sbinfo->si_xib_mtx);
34276 + err = xib_pindex(sb, pindex);
34278 + clear_bit(bit, sbinfo->si_xib_buf);
34279 + sbinfo->si_xib_next_bit = bit;
34281 + mutex_unlock(&sbinfo->si_xib_mtx);
34284 +/* for s_op->delete_inode() */
34285 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
34288 + unsigned int mnt_flags;
34289 + aufs_bindex_t bindex, bend, bi;
34290 + unsigned char try_trunc;
34291 + struct au_iinfo *iinfo;
34292 + struct super_block *sb;
34293 + struct au_hinode *hi;
34294 + struct inode *h_inode;
34295 + struct au_branch *br;
34296 + vfs_writef_t xwrite;
34298 + sb = inode->i_sb;
34299 + mnt_flags = au_mntflags(sb);
34300 + if (!au_opt_test(mnt_flags, XINO)
34301 + || inode->i_ino == AUFS_ROOT_INO)
34305 + au_xigen_inc(inode);
34306 + au_xib_clear_bit(inode);
34309 + iinfo = au_ii(inode);
34313 + bindex = iinfo->ii_bstart;
34317 + xwrite = au_sbi(sb)->si_xwrite;
34318 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
34319 + hi = iinfo->ii_hinode + bindex;
34320 + bend = iinfo->ii_bend;
34321 + for (; bindex <= bend; bindex++, hi++) {
34322 + h_inode = hi->hi_inode;
34324 + || (!unlinked && h_inode->i_nlink))
34327 + /* inode may not be revalidated */
34328 + bi = au_br_index(sb, hi->hi_id);
34332 + br = au_sbr(sb, bi);
34333 + err = au_xino_do_write(xwrite, br->br_xino.xi_file,
34334 + h_inode->i_ino, /*ino*/0);
34335 + if (!err && try_trunc
34336 + && au_test_fs_trunc_xino(au_br_sb(br)))
34337 + xino_try_trunc(sb, br);
34341 +/* get an unused inode number from bitmap */
34342 +ino_t au_xino_new_ino(struct super_block *sb)
34345 + unsigned long *p, pindex, ul, pend;
34346 + struct au_sbinfo *sbinfo;
34347 + struct file *file;
34348 + int free_bit, err;
34350 + if (!au_opt_test(au_mntflags(sb), XINO))
34351 + return iunique(sb, AUFS_FIRST_INO);
34353 + sbinfo = au_sbi(sb);
34354 + mutex_lock(&sbinfo->si_xib_mtx);
34355 + p = sbinfo->si_xib_buf;
34356 + free_bit = sbinfo->si_xib_next_bit;
34357 + if (free_bit < page_bits && !test_bit(free_bit, p))
34358 + goto out; /* success */
34359 + free_bit = find_first_zero_bit(p, page_bits);
34360 + if (free_bit < page_bits)
34361 + goto out; /* success */
34363 + pindex = sbinfo->si_xib_last_pindex;
34364 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
34365 + err = xib_pindex(sb, ul);
34366 + if (unlikely(err))
34368 + free_bit = find_first_zero_bit(p, page_bits);
34369 + if (free_bit < page_bits)
34370 + goto out; /* success */
34373 + file = sbinfo->si_xib;
34374 + pend = vfsub_f_size_read(file) / PAGE_SIZE;
34375 + for (ul = pindex + 1; ul <= pend; ul++) {
34376 + err = xib_pindex(sb, ul);
34377 + if (unlikely(err))
34379 + free_bit = find_first_zero_bit(p, page_bits);
34380 + if (free_bit < page_bits)
34381 + goto out; /* success */
34386 + set_bit(free_bit, p);
34387 + sbinfo->si_xib_next_bit = free_bit + 1;
34388 + pindex = sbinfo->si_xib_last_pindex;
34389 + mutex_unlock(&sbinfo->si_xib_mtx);
34390 + ino = xib_calc_ino(pindex, free_bit);
34391 + AuDbg("i%lu\n", (unsigned long)ino);
34394 + mutex_unlock(&sbinfo->si_xib_mtx);
34400 + * read @ino from xinofile for the specified branch{@sb, @bindex}
34401 + * at the position of @h_ino.
34402 + * if @ino does not exist and @do_new is true, get new one.
34404 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
34410 + struct file *file;
34411 + struct au_sbinfo *sbinfo;
34414 + if (!au_opt_test(au_mntflags(sb), XINO))
34415 + return 0; /* no xino */
34418 + sbinfo = au_sbi(sb);
34420 + if (unlikely(au_loff_max / sizeof(*ino) - 1 < pos)) {
34421 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
34424 + pos *= sizeof(*ino);
34426 + file = au_sbr(sb, bindex)->br_xino.xi_file;
34427 + if (vfsub_f_size_read(file) < pos + sizeof(*ino))
34428 + return 0; /* no ino */
34430 + sz = xino_fread(sbinfo->si_xread, file, ino, sizeof(*ino), &pos);
34431 + if (sz == sizeof(*ino))
34432 + return 0; /* success */
34435 + if (unlikely(sz >= 0)) {
34437 + AuIOErr("xino read error (%zd)\n", sz);
34443 +/* ---------------------------------------------------------------------- */
34445 +/* create and set a new xino file */
34447 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent)
34449 + struct file *file;
34450 + struct dentry *h_parent, *d;
34451 + struct inode *h_dir, *inode;
34455 + * at mount-time, and the xino file is the default path,
34456 + * hnotify is disabled so we have no notify events to ignore.
34457 + * when a user specified the xino, we cannot get au_hdir to be ignored.
34459 + file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
34460 + /* | __FMODE_NONOTIFY */,
34461 + S_IRUGO | S_IWUGO);
34462 + if (IS_ERR(file)) {
34464 + pr_err("open %s(%ld)\n", fname, PTR_ERR(file));
34468 + /* keep file count */
34470 + inode = file_inode(file);
34471 + h_parent = dget_parent(file->f_path.dentry);
34472 + h_dir = d_inode(h_parent);
34473 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
34474 + /* mnt_want_write() is unnecessary here */
34475 + /* no delegation since it is just created */
34476 + if (inode->i_nlink)
34477 + err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
34479 + mutex_unlock(&h_dir->i_mutex);
34481 + if (unlikely(err)) {
34483 + pr_err("unlink %s(%d)\n", fname, err);
34488 + d = file->f_path.dentry;
34489 + if (unlikely(sb == d->d_sb)) {
34491 + pr_err("%s must be outside\n", fname);
34494 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
34496 + pr_err("xino doesn't support %s(%s)\n",
34497 + fname, au_sbtype(d->d_sb));
34500 + return file; /* success */
34504 + file = ERR_PTR(err);
34509 + * find another branch who is on the same filesystem of the specified
34510 + * branch{@btgt}. search until @bend.
34512 +static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
34513 + aufs_bindex_t bend)
34515 + aufs_bindex_t bindex;
34516 + struct super_block *tgt_sb = au_sbr_sb(sb, btgt);
34518 + for (bindex = 0; bindex < btgt; bindex++)
34519 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
34521 + for (bindex++; bindex <= bend; bindex++)
34522 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
34527 +/* ---------------------------------------------------------------------- */
34530 + * initialize the xinofile for the specified branch @br
34531 + * at the place/path where @base_file indicates.
34532 + * test whether another branch is on the same filesystem or not,
34533 + * if @do_test is true.
34535 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
34536 + struct file *base_file, int do_test)
34540 + aufs_bindex_t bend, bindex;
34541 + struct au_branch *shared_br, *b;
34542 + struct file *file;
34543 + struct super_block *tgt_sb;
34545 + shared_br = NULL;
34546 + bend = au_sbend(sb);
34548 + tgt_sb = au_br_sb(br);
34549 + for (bindex = 0; bindex <= bend; bindex++) {
34550 + b = au_sbr(sb, bindex);
34551 + if (tgt_sb == au_br_sb(b)) {
34558 + if (!shared_br || !shared_br->br_xino.xi_file) {
34559 + struct au_xino_lock_dir ldir;
34561 + au_xino_lock_dir(sb, base_file, &ldir);
34562 + /* mnt_want_write() is unnecessary here */
34563 + file = au_xino_create2(base_file, NULL);
34564 + au_xino_unlock_dir(&ldir);
34565 + err = PTR_ERR(file);
34566 + if (IS_ERR(file))
34568 + br->br_xino.xi_file = file;
34570 + br->br_xino.xi_file = shared_br->br_xino.xi_file;
34571 + get_file(br->br_xino.xi_file);
34574 + ino = AUFS_ROOT_INO;
34575 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
34577 + if (unlikely(err)) {
34578 + fput(br->br_xino.xi_file);
34579 + br->br_xino.xi_file = NULL;
34586 +/* ---------------------------------------------------------------------- */
34588 +/* trucate a xino bitmap file */
34591 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
34595 + unsigned long pindex;
34596 + loff_t pos, pend;
34597 + struct au_sbinfo *sbinfo;
34598 + vfs_readf_t func;
34600 + unsigned long *p;
34603 + sbinfo = au_sbi(sb);
34604 + MtxMustLock(&sbinfo->si_xib_mtx);
34605 + p = sbinfo->si_xib_buf;
34606 + func = sbinfo->si_xread;
34607 + pend = vfsub_f_size_read(file);
34609 + while (pos < pend) {
34610 + sz = xino_fread(func, file, page, PAGE_SIZE, &pos);
34612 + if (unlikely(sz <= 0))
34616 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
34617 + if (unlikely(*ino < AUFS_FIRST_INO))
34620 + xib_calc_bit(*ino, &pindex, &bit);
34621 + AuDebugOn(page_bits <= bit);
34622 + err = xib_pindex(sb, pindex);
34634 +static int xib_restore(struct super_block *sb)
34637 + aufs_bindex_t bindex, bend;
34641 + page = (void *)__get_free_page(GFP_NOFS);
34642 + if (unlikely(!page))
34646 + bend = au_sbend(sb);
34647 + for (bindex = 0; !err && bindex <= bend; bindex++)
34648 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0)
34649 + err = do_xib_restore
34650 + (sb, au_sbr(sb, bindex)->br_xino.xi_file, page);
34652 + AuDbg("b%d\n", bindex);
34653 + free_page((unsigned long)page);
34659 +int au_xib_trunc(struct super_block *sb)
34664 + struct au_xino_lock_dir ldir;
34665 + struct au_sbinfo *sbinfo;
34666 + unsigned long *p;
34667 + struct file *file;
34669 + SiMustWriteLock(sb);
34672 + sbinfo = au_sbi(sb);
34673 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
34676 + file = sbinfo->si_xib;
34677 + if (vfsub_f_size_read(file) <= PAGE_SIZE)
34680 + au_xino_lock_dir(sb, file, &ldir);
34681 + /* mnt_want_write() is unnecessary here */
34682 + file = au_xino_create2(sbinfo->si_xib, NULL);
34683 + au_xino_unlock_dir(&ldir);
34684 + err = PTR_ERR(file);
34685 + if (IS_ERR(file))
34687 + fput(sbinfo->si_xib);
34688 + sbinfo->si_xib = file;
34690 + p = sbinfo->si_xib_buf;
34691 + memset(p, 0, PAGE_SIZE);
34693 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xib, p, PAGE_SIZE, &pos);
34694 + if (unlikely(sz != PAGE_SIZE)) {
34696 + AuIOErr("err %d\n", err);
34702 + mutex_lock(&sbinfo->si_xib_mtx);
34703 + /* mnt_want_write() is unnecessary here */
34704 + err = xib_restore(sb);
34705 + mutex_unlock(&sbinfo->si_xib_mtx);
34711 +/* ---------------------------------------------------------------------- */
34714 + * xino mount option handlers
34718 +static void xino_clear_xib(struct super_block *sb)
34720 + struct au_sbinfo *sbinfo;
34722 + SiMustWriteLock(sb);
34724 + sbinfo = au_sbi(sb);
34725 + sbinfo->si_xread = NULL;
34726 + sbinfo->si_xwrite = NULL;
34727 + if (sbinfo->si_xib)
34728 + fput(sbinfo->si_xib);
34729 + sbinfo->si_xib = NULL;
34730 + free_page((unsigned long)sbinfo->si_xib_buf);
34731 + sbinfo->si_xib_buf = NULL;
34734 +static int au_xino_set_xib(struct super_block *sb, struct file *base)
34738 + struct au_sbinfo *sbinfo;
34739 + struct file *file;
34741 + SiMustWriteLock(sb);
34743 + sbinfo = au_sbi(sb);
34744 + file = au_xino_create2(base, sbinfo->si_xib);
34745 + err = PTR_ERR(file);
34746 + if (IS_ERR(file))
34748 + if (sbinfo->si_xib)
34749 + fput(sbinfo->si_xib);
34750 + sbinfo->si_xib = file;
34751 + sbinfo->si_xread = vfs_readf(file);
34752 + sbinfo->si_xwrite = vfs_writef(file);
34755 + if (!sbinfo->si_xib_buf)
34756 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
34757 + if (unlikely(!sbinfo->si_xib_buf))
34760 + sbinfo->si_xib_last_pindex = 0;
34761 + sbinfo->si_xib_next_bit = 0;
34762 + if (vfsub_f_size_read(file) < PAGE_SIZE) {
34764 + err = xino_fwrite(sbinfo->si_xwrite, file, sbinfo->si_xib_buf,
34765 + PAGE_SIZE, &pos);
34766 + if (unlikely(err != PAGE_SIZE))
34770 + goto out; /* success */
34773 + free_page((unsigned long)sbinfo->si_xib_buf);
34774 + sbinfo->si_xib_buf = NULL;
34778 + fput(sbinfo->si_xib);
34779 + sbinfo->si_xib = NULL;
34780 + sbinfo->si_xread = NULL;
34781 + sbinfo->si_xwrite = NULL;
34786 +/* xino for each branch */
34787 +static void xino_clear_br(struct super_block *sb)
34789 + aufs_bindex_t bindex, bend;
34790 + struct au_branch *br;
34792 + bend = au_sbend(sb);
34793 + for (bindex = 0; bindex <= bend; bindex++) {
34794 + br = au_sbr(sb, bindex);
34795 + if (!br || !br->br_xino.xi_file)
34798 + fput(br->br_xino.xi_file);
34799 + br->br_xino.xi_file = NULL;
34803 +static int au_xino_set_br(struct super_block *sb, struct file *base)
34807 + aufs_bindex_t bindex, bend, bshared;
34809 + struct file *old, *new;
34811 + struct au_branch *br;
34812 + struct inode *inode;
34813 + vfs_writef_t writef;
34815 + SiMustWriteLock(sb);
34818 + bend = au_sbend(sb);
34819 + fpair = kcalloc(bend + 1, sizeof(*fpair), GFP_NOFS);
34820 + if (unlikely(!fpair))
34823 + inode = d_inode(sb->s_root);
34824 + ino = AUFS_ROOT_INO;
34825 + writef = au_sbi(sb)->si_xwrite;
34826 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
34827 + br = au_sbr(sb, bindex);
34828 + bshared = is_sb_shared(sb, bindex, bindex - 1);
34829 + if (bshared >= 0) {
34830 + /* shared xino */
34831 + *p = fpair[bshared];
34832 + get_file(p->new);
34837 + p->old = br->br_xino.xi_file;
34838 + p->new = au_xino_create2(base, br->br_xino.xi_file);
34839 + err = PTR_ERR(p->new);
34840 + if (IS_ERR(p->new)) {
34846 + err = au_xino_do_write(writef, p->new,
34847 + au_h_iptr(inode, bindex)->i_ino, ino);
34848 + if (unlikely(err))
34852 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
34853 + br = au_sbr(sb, bindex);
34854 + if (br->br_xino.xi_file)
34855 + fput(br->br_xino.xi_file);
34856 + get_file(p->new);
34857 + br->br_xino.xi_file = p->new;
34861 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++)
34871 +void au_xino_clr(struct super_block *sb)
34873 + struct au_sbinfo *sbinfo;
34875 + au_xigen_clr(sb);
34876 + xino_clear_xib(sb);
34877 + xino_clear_br(sb);
34878 + sbinfo = au_sbi(sb);
34879 + /* lvalue, do not call au_mntflags() */
34880 + au_opt_clr(sbinfo->si_mntflags, XINO);
34883 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount)
34886 + struct dentry *parent, *cur_parent;
34887 + struct qstr *dname, *cur_name;
34888 + struct file *cur_xino;
34889 + struct inode *dir;
34890 + struct au_sbinfo *sbinfo;
34892 + SiMustWriteLock(sb);
34895 + sbinfo = au_sbi(sb);
34896 + parent = dget_parent(xino->file->f_path.dentry);
34899 + dname = &xino->file->f_path.dentry->d_name;
34900 + cur_xino = sbinfo->si_xib;
34902 + cur_parent = dget_parent(cur_xino->f_path.dentry);
34903 + cur_name = &cur_xino->f_path.dentry->d_name;
34904 + skip = (cur_parent == parent
34905 + && au_qstreq(dname, cur_name));
34906 + dput(cur_parent);
34912 + au_opt_set(sbinfo->si_mntflags, XINO);
34913 + dir = d_inode(parent);
34914 + mutex_lock_nested(&dir->i_mutex, AuLsc_I_PARENT);
34915 + /* mnt_want_write() is unnecessary here */
34916 + err = au_xino_set_xib(sb, xino->file);
34918 + err = au_xigen_set(sb, xino->file);
34920 + err = au_xino_set_br(sb, xino->file);
34921 + mutex_unlock(&dir->i_mutex);
34923 + goto out; /* success */
34926 + AuIOErr("failed creating xino(%d).\n", err);
34927 + au_xigen_clr(sb);
34928 + xino_clear_xib(sb);
34935 +/* ---------------------------------------------------------------------- */
34938 + * create a xinofile at the default place/path.
34940 +struct file *au_xino_def(struct super_block *sb)
34942 + struct file *file;
34944 + struct au_branch *br;
34945 + struct super_block *h_sb;
34946 + struct path path;
34947 + aufs_bindex_t bend, bindex, bwr;
34950 + bend = au_sbend(sb);
34952 + for (bindex = 0; bindex <= bend; bindex++) {
34953 + br = au_sbr(sb, bindex);
34954 + if (au_br_writable(br->br_perm)
34955 + && !au_test_fs_bad_xino(au_br_sb(br))) {
34962 + file = ERR_PTR(-ENOMEM);
34963 + page = (void *)__get_free_page(GFP_NOFS);
34964 + if (unlikely(!page))
34966 + path.mnt = au_br_mnt(br);
34967 + path.dentry = au_h_dptr(sb->s_root, bwr);
34968 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
34969 + file = (void *)p;
34970 + if (!IS_ERR(p)) {
34971 + strcat(p, "/" AUFS_XINO_FNAME);
34972 + AuDbg("%s\n", p);
34973 + file = au_xino_create(sb, p, /*silent*/0);
34974 + if (!IS_ERR(file))
34975 + au_xino_brid_set(sb, br->br_id);
34977 + free_page((unsigned long)page);
34979 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0);
34980 + if (IS_ERR(file))
34982 + h_sb = file->f_path.dentry->d_sb;
34983 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
34984 + pr_err("xino doesn't support %s(%s)\n",
34985 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
34987 + file = ERR_PTR(-EINVAL);
34989 + if (!IS_ERR(file))
34990 + au_xino_brid_set(sb, -1);
34997 +/* ---------------------------------------------------------------------- */
34999 +int au_xino_path(struct seq_file *seq, struct file *file)
35003 + err = au_seq_path(seq, &file->f_path);
35004 + if (unlikely(err))
35007 +#define Deleted "\\040(deleted)"
35008 + seq->count -= sizeof(Deleted) - 1;
35009 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
35010 + sizeof(Deleted) - 1));
35016 diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
35017 --- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
35018 +++ linux/include/uapi/linux/aufs_type.h 2015-11-11 17:22:06.926020273 +0100
35021 + * Copyright (C) 2005-2015 Junjiro R. Okajima
35023 + * This program, aufs is free software; you can redistribute it and/or modify
35024 + * it under the terms of the GNU General Public License as published by
35025 + * the Free Software Foundation; either version 2 of the License, or
35026 + * (at your option) any later version.
35028 + * This program is distributed in the hope that it will be useful,
35029 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
35030 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
35031 + * GNU General Public License for more details.
35033 + * You should have received a copy of the GNU General Public License
35034 + * along with this program. If not, see <http://www.gnu.org/licenses/>.
35037 +#ifndef __AUFS_TYPE_H__
35038 +#define __AUFS_TYPE_H__
35040 +#define AUFS_NAME "aufs"
35044 + * define it before including all other headers.
35045 + * sched.h may use pr_* macros before defining "current", so define the
35046 + * no-current version first, and re-define later.
35048 +#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
35049 +#include <linux/sched.h>
35051 +#define pr_fmt(fmt) \
35052 + AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
35053 + (int)sizeof(current->comm), current->comm, current->pid
35055 +#include <stdint.h>
35056 +#include <sys/types.h>
35057 +#endif /* __KERNEL__ */
35059 +#include <linux/limits.h>
35061 +#define AUFS_VERSION "4.3-20151109"
35063 +/* todo? move this to linux-2.6.19/include/magic.h */
35064 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
35066 +/* ---------------------------------------------------------------------- */
35068 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
35069 +typedef int8_t aufs_bindex_t;
35070 +#define AUFS_BRANCH_MAX 127
35072 +typedef int16_t aufs_bindex_t;
35073 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
35074 +#define AUFS_BRANCH_MAX 511
35075 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
35076 +#define AUFS_BRANCH_MAX 1023
35077 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
35078 +#define AUFS_BRANCH_MAX 32767
35083 +#ifndef AUFS_BRANCH_MAX
35084 +#error unknown CONFIG_AUFS_BRANCH_MAX value
35086 +#endif /* __KERNEL__ */
35088 +/* ---------------------------------------------------------------------- */
35090 +#define AUFS_FSTYPE AUFS_NAME
35092 +#define AUFS_ROOT_INO 2
35093 +#define AUFS_FIRST_INO 11
35095 +#define AUFS_WH_PFX ".wh."
35096 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
35097 +#define AUFS_WH_TMP_LEN 4
35098 +/* a limit for rmdir/rename a dir and copyup */
35099 +#define AUFS_MAX_NAMELEN (NAME_MAX \
35100 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
35102 + - AUFS_WH_TMP_LEN) /* hex */
35103 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
35104 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
35105 +#define AUFS_XINO_DEF_SEC 30 /* seconds */
35106 +#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
35107 +#define AUFS_DIRWH_DEF 3
35108 +#define AUFS_RDCACHE_DEF 10 /* seconds */
35109 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
35110 +#define AUFS_RDBLK_DEF 512 /* bytes */
35111 +#define AUFS_RDHASH_DEF 32
35112 +#define AUFS_WKQ_NAME AUFS_NAME "d"
35113 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
35114 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
35115 +#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
35116 +#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
35118 +/* pseudo-link maintenace under /proc */
35119 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
35120 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
35121 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
35123 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
35124 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
35126 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
35127 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
35128 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
35130 +/* doubly whiteouted */
35131 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
35132 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
35133 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
35135 +/* branch permissions and attributes */
35136 +#define AUFS_BRPERM_RW "rw"
35137 +#define AUFS_BRPERM_RO "ro"
35138 +#define AUFS_BRPERM_RR "rr"
35139 +#define AUFS_BRATTR_COO_REG "coo_reg"
35140 +#define AUFS_BRATTR_COO_ALL "coo_all"
35141 +#define AUFS_BRATTR_FHSM "fhsm"
35142 +#define AUFS_BRATTR_UNPIN "unpin"
35143 +#define AUFS_BRATTR_ICEX "icex"
35144 +#define AUFS_BRATTR_ICEX_SEC "icexsec"
35145 +#define AUFS_BRATTR_ICEX_SYS "icexsys"
35146 +#define AUFS_BRATTR_ICEX_TR "icextr"
35147 +#define AUFS_BRATTR_ICEX_USR "icexusr"
35148 +#define AUFS_BRATTR_ICEX_OTH "icexoth"
35149 +#define AUFS_BRRATTR_WH "wh"
35150 +#define AUFS_BRWATTR_NLWH "nolwh"
35151 +#define AUFS_BRWATTR_MOO "moo"
35153 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
35154 +#define AuBrPerm_RO (1 << 1) /* readonly */
35155 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
35156 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
35158 +#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
35159 +#define AuBrAttr_COO_ALL (1 << 4)
35160 +#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
35162 +#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
35163 +#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
35164 + branch. meaningless since
35165 + linux-3.18-rc1 */
35167 +/* ignore error in copying XATTR */
35168 +#define AuBrAttr_ICEX_SEC (1 << 7)
35169 +#define AuBrAttr_ICEX_SYS (1 << 8)
35170 +#define AuBrAttr_ICEX_TR (1 << 9)
35171 +#define AuBrAttr_ICEX_USR (1 << 10)
35172 +#define AuBrAttr_ICEX_OTH (1 << 11)
35173 +#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
35174 + | AuBrAttr_ICEX_SYS \
35175 + | AuBrAttr_ICEX_TR \
35176 + | AuBrAttr_ICEX_USR \
35177 + | AuBrAttr_ICEX_OTH)
35179 +#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
35180 +#define AuBrRAttr_Mask AuBrRAttr_WH
35182 +#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
35183 +#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
35184 +#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
35186 +#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
35188 +/* #warning test userspace */
35190 +#ifndef CONFIG_AUFS_FHSM
35191 +#undef AuBrAttr_FHSM
35192 +#define AuBrAttr_FHSM 0
35194 +#ifndef CONFIG_AUFS_XATTR
35195 +#undef AuBrAttr_ICEX
35196 +#define AuBrAttr_ICEX 0
35197 +#undef AuBrAttr_ICEX_SEC
35198 +#define AuBrAttr_ICEX_SEC 0
35199 +#undef AuBrAttr_ICEX_SYS
35200 +#define AuBrAttr_ICEX_SYS 0
35201 +#undef AuBrAttr_ICEX_TR
35202 +#define AuBrAttr_ICEX_TR 0
35203 +#undef AuBrAttr_ICEX_USR
35204 +#define AuBrAttr_ICEX_USR 0
35205 +#undef AuBrAttr_ICEX_OTH
35206 +#define AuBrAttr_ICEX_OTH 0
35210 +/* the longest combination */
35211 +/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
35212 +#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
35213 + "+" AUFS_BRATTR_COO_REG \
35214 + "+" AUFS_BRATTR_FHSM \
35215 + "+" AUFS_BRATTR_UNPIN \
35216 + "+" AUFS_BRATTR_ICEX_SEC \
35217 + "+" AUFS_BRATTR_ICEX_SYS \
35218 + "+" AUFS_BRATTR_ICEX_USR \
35219 + "+" AUFS_BRATTR_ICEX_OTH \
35220 + "+" AUFS_BRWATTR_NLWH)
35223 + char a[AuBrPermStrSz];
35224 +} au_br_perm_str_t;
35226 +static inline int au_br_writable(int brperm)
35228 + return brperm & AuBrPerm_RW;
35231 +static inline int au_br_whable(int brperm)
35233 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
35236 +static inline int au_br_wh_linkable(int brperm)
35238 + return !(brperm & AuBrWAttr_NoLinkWH);
35241 +static inline int au_br_cmoo(int brperm)
35243 + return brperm & AuBrAttr_CMOO_Mask;
35246 +static inline int au_br_fhsm(int brperm)
35248 + return brperm & AuBrAttr_FHSM;
35251 +/* ---------------------------------------------------------------------- */
35255 + /* readdir in userspace */
35259 + AuCtl_WBR_FD, /* pathconf wrapper */
35260 + AuCtl_IBUSY, /* busy inode */
35261 + AuCtl_MVDOWN, /* move-down */
35262 + AuCtl_BR, /* info about branches */
35263 + AuCtl_FHSM_FD /* connection for fhsm */
35266 +/* borrowed from linux/include/linux/kernel.h */
35268 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
35269 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
35272 +/* borrowed from linux/include/linux/compiler-gcc3.h */
35274 +#define __aligned(x) __attribute__((aligned(x)))
35279 +#define __packed __attribute__((packed))
35283 +struct au_rdu_cookie {
35288 + uint32_t generation;
35291 +struct au_rdu_ent {
35300 +static inline int au_rdu_len(int nlen)
35302 + /* include the terminating NULL */
35303 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
35304 + sizeof(uint64_t));
35307 +union au_rdu_ent_ul {
35308 + struct au_rdu_ent __user *e;
35320 + uint64_t sz; /* AuCtl_RDU */
35321 + uint64_t nent; /* AuCtl_RDU_INO */
35323 + union au_rdu_ent_ul ent;
35324 + uint16_t verify[AufsCtlRduV_End];
35326 + /* input/output */
35330 + union au_rdu_ent_ul tail;
35331 + /* number of entries which were added in a single call */
35336 + struct au_rdu_cookie cookie;
35339 +/* ---------------------------------------------------------------------- */
35341 +struct aufs_wbr_fd {
35346 +/* ---------------------------------------------------------------------- */
35348 +struct aufs_ibusy {
35349 + uint64_t ino, h_ino;
35353 +/* ---------------------------------------------------------------------- */
35355 +/* error code for move-down */
35356 +/* the actual message strings are implemented in aufs-util.git */
35358 + EAU_MVDOWN_OPAQUE = 1,
35359 + EAU_MVDOWN_WHITEOUT,
35360 + EAU_MVDOWN_UPPER,
35361 + EAU_MVDOWN_BOTTOM,
35362 + EAU_MVDOWN_NOUPPER,
35363 + EAU_MVDOWN_NOLOWERBR,
35367 +/* flags for move-down */
35368 +#define AUFS_MVDOWN_DMSG 1
35369 +#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
35370 +#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
35371 +#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
35372 +#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
35373 +#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
35374 +#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
35375 +#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
35376 +#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
35377 +#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
35378 +#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
35379 +#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
35380 +#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
35382 +/* index for move-down */
35384 + AUFS_MVDOWN_UPPER,
35385 + AUFS_MVDOWN_LOWER,
35386 + AUFS_MVDOWN_NARRAY
35390 + * additional info of move-down
35391 + * number of free blocks and inodes.
35392 + * subset of struct kstatfs, but smaller and always 64bit.
35394 +struct aufs_stfs {
35395 + uint64_t f_blocks;
35396 + uint64_t f_bavail;
35397 + uint64_t f_files;
35398 + uint64_t f_ffree;
35401 +struct aufs_stbr {
35402 + int16_t brid; /* optional input */
35403 + int16_t bindex; /* output */
35404 + struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
35407 +struct aufs_mvdown {
35408 + uint32_t flags; /* input/output */
35409 + struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
35410 + int8_t au_errno; /* output */
35413 +/* ---------------------------------------------------------------------- */
35415 +union aufs_brinfo {
35416 + /* PATH_MAX may differ between kernel-space and user-space */
35417 + char _spacer[4096];
35425 +/* ---------------------------------------------------------------------- */
35427 +#define AuCtlType 'A'
35428 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
35429 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
35430 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
35431 + struct aufs_wbr_fd)
35432 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
35433 +#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
35434 + struct aufs_mvdown)
35435 +#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
35436 +#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
35438 +#endif /* __AUFS_TYPE_H__ */
35439 aufs4.3 loopback patch
35441 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
35442 index 291ec9e..1b8190d 100644
35443 --- a/drivers/block/loop.c
35444 +++ b/drivers/block/loop.c
35445 @@ -417,7 +417,7 @@ static int do_req_filebacked(struct loop_device *lo, struct request *rq)
35448 struct switch_request {
35449 - struct file *file;
35450 + struct file *file, *virt_file;
35451 struct completion wait;
35454 @@ -437,6 +437,7 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
35455 mapping = file->f_mapping;
35456 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
35457 lo->lo_backing_file = file;
35458 + lo->lo_backing_virt_file = p->virt_file;
35459 lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ?
35460 mapping->host->i_bdev->bd_block_size : PAGE_SIZE;
35461 lo->old_gfp_mask = mapping_gfp_mask(mapping);
35462 @@ -448,11 +449,13 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
35463 * First it needs to flush existing IO, it does this by sending a magic
35464 * BIO down the pipe. The completion of this BIO does the actual switch.
35466 -static int loop_switch(struct loop_device *lo, struct file *file)
35467 +static int loop_switch(struct loop_device *lo, struct file *file,
35468 + struct file *virt_file)
35470 struct switch_request w;
35473 + w.virt_file = virt_file;
35475 /* freeze queue and wait for completion of scheduled requests */
35476 blk_mq_freeze_queue(lo->lo_queue);
35477 @@ -471,7 +474,16 @@ static int loop_switch(struct loop_device *lo, struct file *file)
35479 static int loop_flush(struct loop_device *lo)
35481 - return loop_switch(lo, NULL);
35482 + return loop_switch(lo, NULL, NULL);
35485 +static struct file *loop_real_file(struct file *file)
35487 + struct file *f = NULL;
35489 + if (file->f_path.dentry->d_sb->s_op->real_loop)
35490 + f = file->f_path.dentry->d_sb->s_op->real_loop(file);
35494 static void loop_reread_partitions(struct loop_device *lo,
35495 @@ -508,6 +520,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
35498 struct file *file, *old_file;
35499 + struct file *f, *virt_file = NULL, *old_virt_file;
35500 struct inode *inode;
35503 @@ -524,9 +537,16 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
35507 + f = loop_real_file(file);
35509 + virt_file = file;
35514 inode = file->f_mapping->host;
35515 old_file = lo->lo_backing_file;
35516 + old_virt_file = lo->lo_backing_virt_file;
35520 @@ -538,17 +558,21 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
35523 /* and ... switch */
35524 - error = loop_switch(lo, file);
35525 + error = loop_switch(lo, file, virt_file);
35530 + if (old_virt_file)
35531 + fput(old_virt_file);
35532 if (lo->lo_flags & LO_FLAGS_PARTSCAN)
35533 loop_reread_partitions(lo, bdev);
35543 @@ -709,7 +733,7 @@ static void loop_config_discard(struct loop_device *lo)
35544 static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35545 struct block_device *bdev, unsigned int arg)
35547 - struct file *file, *f;
35548 + struct file *file, *f, *virt_file = NULL;
35549 struct inode *inode;
35550 struct address_space *mapping;
35551 unsigned lo_blocksize;
35552 @@ -724,6 +748,12 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35556 + f = loop_real_file(file);
35558 + virt_file = file;
35564 if (lo->lo_state != Lo_unbound)
35565 @@ -778,6 +808,7 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35566 lo->lo_device = bdev;
35567 lo->lo_flags = lo_flags;
35568 lo->lo_backing_file = file;
35569 + lo->lo_backing_virt_file = virt_file;
35570 lo->transfer = NULL;
35572 lo->lo_sizelimit = 0;
35573 @@ -809,6 +840,8 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35580 /* This is safe: open() is still holding a reference. */
35581 module_put(THIS_MODULE);
35582 @@ -855,6 +888,7 @@ loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
35583 static int loop_clr_fd(struct loop_device *lo)
35585 struct file *filp = lo->lo_backing_file;
35586 + struct file *virt_filp = lo->lo_backing_virt_file;
35587 gfp_t gfp = lo->old_gfp_mask;
35588 struct block_device *bdev = lo->lo_device;
35590 @@ -886,6 +920,7 @@ static int loop_clr_fd(struct loop_device *lo)
35591 spin_lock_irq(&lo->lo_lock);
35592 lo->lo_state = Lo_rundown;
35593 lo->lo_backing_file = NULL;
35594 + lo->lo_backing_virt_file = NULL;
35595 spin_unlock_irq(&lo->lo_lock);
35597 loop_release_xfer(lo);
35598 @@ -931,6 +966,8 @@ static int loop_clr_fd(struct loop_device *lo)
35599 * bd_mutex which is usually taken before lo_ctl_mutex.
35607 diff --git a/drivers/block/loop.h b/drivers/block/loop.h
35608 index 25e8997..93b6fce 100644
35609 --- a/drivers/block/loop.h
35610 +++ b/drivers/block/loop.h
35611 @@ -46,7 +46,7 @@ struct loop_device {
35612 int (*ioctl)(struct loop_device *, int cmd,
35613 unsigned long arg);
35615 - struct file * lo_backing_file;
35616 + struct file * lo_backing_file, *lo_backing_virt_file;
35617 struct block_device *lo_device;
35618 unsigned lo_blocksize;
35620 diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
35621 index 91c2ce7..d4ee5a7 100644
35622 --- a/fs/aufs/f_op.c
35623 +++ b/fs/aufs/f_op.c
35624 @@ -389,7 +389,7 @@ static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
35625 if (IS_ERR(h_file))
35628 - if (au_test_loopback_kthread()) {
35629 + if (0 && au_test_loopback_kthread()) {
35630 au_warn_loopback(h_file->f_path.dentry->d_sb);
35631 if (file->f_mapping != h_file->f_mapping) {
35632 file->f_mapping = h_file->f_mapping;
35633 diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
35634 index f324758..4555e7b 100644
35635 --- a/fs/aufs/loop.c
35636 +++ b/fs/aufs/loop.c
35637 @@ -131,3 +131,19 @@ void au_loopback_fin(void)
35638 symbol_put(loop_backing_file);
35639 kfree(au_warn_loopback_array);
35642 +/* ---------------------------------------------------------------------- */
35644 +/* support the loopback block device insude aufs */
35646 +struct file *aufs_real_loop(struct file *file)
35650 + BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
35651 + fi_read_lock(file);
35652 + f = au_hf_top(file);
35653 + fi_read_unlock(file);
35657 diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
35658 index 6d9864d..3322557 100644
35659 --- a/fs/aufs/loop.h
35660 +++ b/fs/aufs/loop.h
35661 @@ -25,7 +25,11 @@ void au_warn_loopback(struct super_block *h_sb);
35663 int au_loopback_init(void);
35664 void au_loopback_fin(void);
35666 +struct file *aufs_real_loop(struct file *file);
35668 +AuStub(struct file *, loop_backing_file, return NULL)
35670 AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
35671 struct dentry *h_adding)
35672 AuStubInt0(au_test_loopback_kthread, void)
35673 @@ -33,6 +37,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
35675 AuStubInt0(au_loopback_init, void)
35676 AuStubVoid(au_loopback_fin, void)
35678 +AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
35679 #endif /* BLK_DEV_LOOP */
35681 #endif /* __KERNEL__ */
35682 diff --git a/fs/aufs/super.c b/fs/aufs/super.c
35683 index 1b25343..3b0c6c5 100644
35684 --- a/fs/aufs/super.c
35685 +++ b/fs/aufs/super.c
35686 @@ -831,7 +831,10 @@ static const struct super_operations aufs_sop = {
35687 .statfs = aufs_statfs,
35688 .put_super = aufs_put_super,
35689 .sync_fs = aufs_sync_fs,
35690 - .remount_fs = aufs_remount_fs
35691 + .remount_fs = aufs_remount_fs,
35692 +#ifdef CONFIG_AUFS_BDEV_LOOP
35693 + .real_loop = aufs_real_loop
35697 /* ---------------------------------------------------------------------- */
35698 diff --git a/include/linux/fs.h b/include/linux/fs.h
35699 index fabd9d7a..90174cf 100644
35700 --- a/include/linux/fs.h
35701 +++ b/include/linux/fs.h
35702 @@ -1734,6 +1734,10 @@ struct super_operations {
35703 struct shrink_control *);
35704 long (*free_cached_objects)(struct super_block *,
35705 struct shrink_control *);
35706 +#if defined(CONFIG_BLK_DEV_LOOP) || defined(CONFIG_BLK_DEV_LOOP_MODULE)
35708 + struct file *(*real_loop)(struct file *);