]> git.pld-linux.org Git - packages/kernel.git/blame - kernel-aufs3.patch
- xfs hang fix
[packages/kernel.git] / kernel-aufs3.patch
CommitLineData
38d290e6 1aufs3.15 kbuild patch
7f207e10
AM
2
3diff --git a/fs/Kconfig b/fs/Kconfig
38d290e6 4index 312393f..78632ed 100644
7f207e10
AM
5--- a/fs/Kconfig
6+++ b/fs/Kconfig
38d290e6 7@@ -209,6 +209,7 @@ source "fs/ufs/Kconfig"
7f207e10 8 source "fs/exofs/Kconfig"
1716fcea 9 source "fs/f2fs/Kconfig"
c06a8ce3 10 source "fs/efivarfs/Kconfig"
7f207e10
AM
11+source "fs/aufs/Kconfig"
12
13 endif # MISC_FILESYSTEMS
14
15diff --git a/fs/Makefile b/fs/Makefile
38d290e6 16index f9cb987..a0c580f 100644
7f207e10
AM
17--- a/fs/Makefile
18+++ b/fs/Makefile
38d290e6 19@@ -126,3 +126,4 @@ obj-y += exofs/ # Multiple modules
7f207e10 20 obj-$(CONFIG_CEPH_FS) += ceph/
bf0370f2 21 obj-$(CONFIG_PSTORE) += pstore/
c06a8ce3 22 obj-$(CONFIG_EFIVAR_FS) += efivarfs/
86dc4139 23+obj-$(CONFIG_AUFS_FS) += aufs/
c06a8ce3 24diff --git a/include/uapi/linux/Kbuild b/include/uapi/linux/Kbuild
38d290e6 25index 6929571..351ca48 100644
c06a8ce3
AM
26--- a/include/uapi/linux/Kbuild
27+++ b/include/uapi/linux/Kbuild
28@@ -56,6 +56,7 @@ header-y += atmppp.h
29 header-y += atmsap.h
30 header-y += atmsvc.h
31 header-y += audit.h
32+header-y += aufs_type.h
33 header-y += auto_fs.h
34 header-y += auto_fs4.h
35 header-y += auxvec.h
38d290e6 36aufs3.15 base patch
7f207e10 37
392086de 38diff --git a/drivers/block/loop.c b/drivers/block/loop.c
38d290e6 39index f70a230..138104b 100644
392086de
AM
40--- a/drivers/block/loop.c
41+++ b/drivers/block/loop.c
fb47a38f 42@@ -692,6 +692,24 @@ static inline int is_loop_device(struct file *file)
392086de
AM
43 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
44 }
45
46+/*
47+ * for AUFS
48+ * no get/put for file.
49+ */
50+struct file *loop_backing_file(struct super_block *sb)
51+{
52+ struct file *ret;
53+ struct loop_device *l;
54+
55+ ret = NULL;
56+ if (MAJOR(sb->s_dev) == LOOP_MAJOR) {
57+ l = sb->s_bdev->bd_disk->private_data;
58+ ret = l->lo_backing_file;
59+ }
60+ return ret;
61+}
62+EXPORT_SYMBOL(loop_backing_file);
63+
64 /* loop sysfs attributes */
65
66 static ssize_t loop_attr_show(struct device *dev, char *page,
0c3ec466 67diff --git a/fs/inode.c b/fs/inode.c
38d290e6 68index f96d2a6..2d72083 100644
0c3ec466
AM
69--- a/fs/inode.c
70+++ b/fs/inode.c
38d290e6 71@@ -1496,7 +1496,7 @@ static int relatime_need_update(struct vfsmount *mnt, struct inode *inode,
0c3ec466
AM
72 * This does the actual work of updating an inodes time or version. Must have
73 * had called mnt_want_write() before calling this.
74 */
75-static int update_time(struct inode *inode, struct timespec *time, int flags)
76+int update_time(struct inode *inode, struct timespec *time, int flags)
77 {
78 if (inode->i_op->update_time)
79 return inode->i_op->update_time(inode, time, flags);
7f207e10 80diff --git a/fs/splice.c b/fs/splice.c
38d290e6 81index e246954..4797013 100644
7f207e10
AM
82--- a/fs/splice.c
83+++ b/fs/splice.c
38d290e6 84@@ -1103,8 +1103,8 @@ EXPORT_SYMBOL(generic_splice_sendpage);
7f207e10
AM
85 /*
86 * Attempt to initiate a splice from pipe to file.
87 */
88-static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
89- loff_t *ppos, size_t len, unsigned int flags)
90+long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
91+ loff_t *ppos, size_t len, unsigned int flags)
92 {
93 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
94 loff_t *, size_t, unsigned int);
38d290e6 95@@ -1120,9 +1120,9 @@ static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
7f207e10
AM
96 /*
97 * Attempt to initiate a splice from a file to a pipe.
98 */
99-static long do_splice_to(struct file *in, loff_t *ppos,
100- struct pipe_inode_info *pipe, size_t len,
101- unsigned int flags)
102+long do_splice_to(struct file *in, loff_t *ppos,
103+ struct pipe_inode_info *pipe, size_t len,
104+ unsigned int flags)
105 {
106 ssize_t (*splice_read)(struct file *, loff_t *,
107 struct pipe_inode_info *, size_t, unsigned int);
0c3ec466 108diff --git a/include/linux/fs.h b/include/linux/fs.h
38d290e6 109index 8780312..3ff3e89 100644
0c3ec466
AM
110--- a/include/linux/fs.h
111+++ b/include/linux/fs.h
38d290e6 112@@ -2606,6 +2606,7 @@ extern int inode_change_ok(const struct inode *, struct iattr *);
0c3ec466
AM
113 extern int inode_newsize_ok(const struct inode *, loff_t offset);
114 extern void setattr_copy(struct inode *inode, const struct iattr *attr);
115
116+extern int update_time(struct inode *, struct timespec *, int);
117 extern int file_update_time(struct file *file);
118
119 extern int generic_show_options(struct seq_file *m, struct dentry *root);
1e00d052 120diff --git a/include/linux/splice.h b/include/linux/splice.h
fb47a38f 121index 0e43906..304169e 100644
1e00d052
AM
122--- a/include/linux/splice.h
123+++ b/include/linux/splice.h
fb47a38f 124@@ -93,4 +93,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
4b3da204
AM
125 extern void spd_release_page(struct splice_pipe_desc *, unsigned int);
126
127 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
1e00d052
AM
128+
129+extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
130+ loff_t *ppos, size_t len, unsigned int flags);
131+extern long do_splice_to(struct file *in, loff_t *ppos,
132+ struct pipe_inode_info *pipe, size_t len,
133+ unsigned int flags);
134 #endif
38d290e6 135aufs3.15 mmap patch
fb47a38f
JR
136
137diff --git a/fs/buffer.c b/fs/buffer.c
38d290e6 138index 9ddb9fc..1059a0b 100644
fb47a38f
JR
139--- a/fs/buffer.c
140+++ b/fs/buffer.c
141@@ -2448,7 +2448,7 @@ int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
142 * Update file times before taking page lock. We may end up failing the
143 * fault so this update may be superfluous but who really cares...
144 */
145- file_update_time(vma->vm_file);
146+ vma_file_update_time(vma);
147
148 ret = __block_page_mkwrite(vma, vmf, get_block);
149 sb_end_pagefault(sb);
150diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
151index d4a3574..e44a744 100644
152--- a/fs/proc/nommu.c
153+++ b/fs/proc/nommu.c
154@@ -45,7 +45,9 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
155 file = region->vm_file;
156
157 if (file) {
158- struct inode *inode = file_inode(region->vm_file);
159+ struct inode *inode;
160+ file = vmr_pr_or_file(region);
161+ inode = file_inode(file);
162 dev = inode->i_sb->s_dev;
163 ino = inode->i_ino;
164 }
165diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
38d290e6 166index c4b2646..dc3c0dc 100644
fb47a38f
JR
167--- a/fs/proc/task_mmu.c
168+++ b/fs/proc/task_mmu.c
38d290e6 169@@ -265,7 +265,9 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid)
fb47a38f
JR
170 const char *name = NULL;
171
172 if (file) {
173- struct inode *inode = file_inode(vma->vm_file);
174+ struct inode *inode;
175+ file = vma_pr_or_file(vma);
176+ inode = file_inode(file);
177 dev = inode->i_sb->s_dev;
178 ino = inode->i_ino;
179 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
38d290e6 180@@ -1408,6 +1410,7 @@ static int show_numa_map(struct seq_file *m, void *v, int is_pid)
fb47a38f
JR
181 seq_printf(m, "%08lx %s", vma->vm_start, buffer);
182
183 if (file) {
184+ file = vma_pr_or_file(vma);
185 seq_printf(m, " file=");
186 seq_path(m, &file->f_path, "\n\t= ");
187 } else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) {
188diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
189index 678455d..ad0ce45 100644
190--- a/fs/proc/task_nommu.c
191+++ b/fs/proc/task_nommu.c
192@@ -141,7 +141,9 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma,
193 file = vma->vm_file;
194
195 if (file) {
196- struct inode *inode = file_inode(vma->vm_file);
197+ struct inode *inode;
198+ file = vma_pr_or_file(file);
199+ inode = file_inode(file);
200 dev = inode->i_sb->s_dev;
201 ino = inode->i_ino;
202 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
203diff --git a/include/linux/mm.h b/include/linux/mm.h
38d290e6 204index d677706..192e7ac4 100644
fb47a38f
JR
205--- a/include/linux/mm.h
206+++ b/include/linux/mm.h
207@@ -18,6 +18,9 @@
208 #include <linux/pfn.h>
209 #include <linux/bit_spinlock.h>
210 #include <linux/shrinker.h>
211+#include <linux/dcache.h>
212+#include <linux/file.h>
213+#include <linux/fs.h>
214
215 struct mempolicy;
216 struct anon_vma;
38d290e6 217@@ -1173,6 +1176,87 @@ static inline int fixup_user_fault(struct task_struct *tsk,
fb47a38f
JR
218 }
219 #endif
220
221+/*
222+ * Mainly for aufs which mmap(2) diffrent file and wants to print different path
223+ * in /proc/PID/maps.
224+ */
225+/* #define AUFS_DEBUG_MMAP */
226+static inline void aufs_trace(struct file *f, struct file *pr,
227+ const char func[], int line, const char func2[])
228+{
229+#ifdef AUFS_DEBUG_MMAP
230+ if (pr)
231+ pr_info("%s:%d: %s, %p\n", func, line, func2,
232+ f ? (char *)f->f_dentry->d_name.name : "(null)");
233+#endif
234+}
235+
236+static inline struct file *vmr_do_pr_or_file(struct vm_region *region,
237+ const char func[], int line)
238+{
239+ struct file *f = region->vm_file, *pr = region->vm_prfile;
240+ aufs_trace(f, pr, func, line, __func__);
241+ return (f && pr) ? pr : f;
242+}
243+
244+static inline void vmr_do_fput(struct vm_region *region,
245+ const char func[], int line)
246+{
247+ struct file *f = region->vm_file, *pr = region->vm_prfile;
248+ aufs_trace(f, pr, func, line, __func__);
249+ fput(f);
250+ if (f && pr)
251+ fput(pr);
252+}
253+
254+static inline void vma_do_file_update_time(struct vm_area_struct *vma,
255+ const char func[], int line)
256+{
257+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
258+ aufs_trace(f, pr, func, line, __func__);
259+ file_update_time(f);
260+ if (f && pr)
261+ file_update_time(pr);
262+}
263+
264+static inline struct file *vma_do_pr_or_file(struct vm_area_struct *vma,
265+ const char func[], int line)
266+{
267+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
268+ aufs_trace(f, pr, func, line, __func__);
269+ return (f && pr) ? pr : f;
270+}
271+
272+static inline void vma_do_get_file(struct vm_area_struct *vma,
273+ const char func[], int line)
274+{
275+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
276+ aufs_trace(f, pr, func, line, __func__);
277+ get_file(f);
278+ if (f && pr)
279+ get_file(pr);
280+}
281+
282+static inline void vma_do_fput(struct vm_area_struct *vma,
283+ const char func[], int line)
284+{
285+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
286+ aufs_trace(f, pr, func, line, __func__);
287+ fput(f);
288+ if (f && pr)
289+ fput(pr);
290+}
291+
292+#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
293+ __LINE__)
294+#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
295+#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
296+ __LINE__)
297+#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
298+ __LINE__)
299+#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
300+#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
301+
302 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
303 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
304 void *buf, int len, int write);
305diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
38d290e6 306index 8967e20..a57b589 100644
fb47a38f
JR
307--- a/include/linux/mm_types.h
308+++ b/include/linux/mm_types.h
38d290e6 309@@ -232,6 +232,7 @@ struct vm_region {
fb47a38f
JR
310 unsigned long vm_top; /* region allocated to here */
311 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
312 struct file *vm_file; /* the backing file or NULL */
313+ struct file *vm_prfile; /* the virtual backing file or NULL */
314
315 int vm_usage; /* region usage count (access under nommu_region_sem) */
316 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
38d290e6 317@@ -300,6 +301,7 @@ struct vm_area_struct {
fb47a38f
JR
318 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
319 units, *not* PAGE_CACHE_SIZE */
320 struct file * vm_file; /* File we map to (can be NULL). */
321+ struct file *vm_prfile; /* shadow of vm_file */
322 void * vm_private_data; /* was vm_pte (shared mem) */
323
324 #ifndef CONFIG_MMU
325diff --git a/kernel/fork.c b/kernel/fork.c
38d290e6 326index 54a8d26..dcf08b6 100644
fb47a38f
JR
327--- a/kernel/fork.c
328+++ b/kernel/fork.c
38d290e6 329@@ -416,7 +416,7 @@ static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
fb47a38f
JR
330 struct inode *inode = file_inode(file);
331 struct address_space *mapping = file->f_mapping;
332
333- get_file(file);
334+ vma_get_file(tmp);
335 if (tmp->vm_flags & VM_DENYWRITE)
336 atomic_dec(&inode->i_writecount);
337 mutex_lock(&mapping->i_mmap_mutex);
338diff --git a/mm/filemap.c b/mm/filemap.c
38d290e6 339index 088358c..138a88c 100644
fb47a38f
JR
340--- a/mm/filemap.c
341+++ b/mm/filemap.c
38d290e6 342@@ -2091,7 +2091,7 @@ int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
fb47a38f
JR
343 int ret = VM_FAULT_LOCKED;
344
345 sb_start_pagefault(inode->i_sb);
346- file_update_time(vma->vm_file);
347+ vma_file_update_time(vma);
348 lock_page(page);
349 if (page->mapping != inode->i_mapping) {
350 unlock_page(page);
351diff --git a/mm/fremap.c b/mm/fremap.c
38d290e6 352index 34feba6..8d6c451 100644
fb47a38f
JR
353--- a/mm/fremap.c
354+++ b/mm/fremap.c
38d290e6
JR
355@@ -223,16 +223,27 @@ get_write_lock:
356 */
357 if (mapping_cap_account_dirty(mapping)) {
358 unsigned long addr;
359- struct file *file = get_file(vma->vm_file);
360+ struct file *file = vma->vm_file,
361+ *prfile = vma->vm_prfile;
362+
fb47a38f
JR
363 /* mmap_region may free vma; grab the info now */
364 vm_flags = vma->vm_flags;
365
366+ vma_get_file(vma);
367 addr = mmap_region(file, start, size, vm_flags, pgoff);
38d290e6 368- fput(file);
fb47a38f 369+ vma_fput(vma);
fb47a38f
JR
370 if (IS_ERR_VALUE(addr)) {
371 err = addr;
38d290e6
JR
372 } else {
373 BUG_ON(addr != start);
374+ if (prfile) {
375+ struct vm_area_struct *new_vma;
376+ new_vma = find_vma(mm, addr);
377+ if (!new_vma->vm_prfile)
378+ new_vma->vm_prfile = prfile;
379+ if (new_vma != vma)
380+ get_file(prfile);
381+ }
382 err = 0;
383 }
384 goto out_freed;
fb47a38f 385diff --git a/mm/madvise.c b/mm/madvise.c
38d290e6 386index a402f8f..134e15d 100644
fb47a38f
JR
387--- a/mm/madvise.c
388+++ b/mm/madvise.c
389@@ -327,12 +327,12 @@ static long madvise_remove(struct vm_area_struct *vma,
390 * vma's reference to the file) can go away as soon as we drop
391 * mmap_sem.
392 */
393- get_file(f);
394+ vma_get_file(vma);
395 up_read(&current->mm->mmap_sem);
396 error = do_fallocate(f,
397 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
398 offset, end - start);
399- fput(f);
400+ vma_fput(vma);
401 down_read(&current->mm->mmap_sem);
402 return error;
403 }
404diff --git a/mm/memory.c b/mm/memory.c
38d290e6 405index 037b812..6e7d241 100644
fb47a38f
JR
406--- a/mm/memory.c
407+++ b/mm/memory.c
38d290e6
JR
408@@ -2805,7 +2805,7 @@ reuse:
409 set_page_dirty_balance(dirty_page);
fb47a38f
JR
410 /* file_update_time outside page_lock */
411 if (vma->vm_file)
412- file_update_time(vma->vm_file);
413+ vma_file_update_time(vma);
414 }
415 put_page(dirty_page);
416 if (page_mkwrite) {
fb47a38f 417diff --git a/mm/mmap.c b/mm/mmap.c
38d290e6 418index b1202cf..40dd067 100644
fb47a38f
JR
419--- a/mm/mmap.c
420+++ b/mm/mmap.c
38d290e6 421@@ -250,7 +250,7 @@ static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
fb47a38f
JR
422 if (vma->vm_ops && vma->vm_ops->close)
423 vma->vm_ops->close(vma);
424 if (vma->vm_file)
425- fput(vma->vm_file);
426+ vma_fput(vma);
427 mpol_put(vma_policy(vma));
428 kmem_cache_free(vm_area_cachep, vma);
429 return next;
38d290e6 430@@ -861,7 +861,7 @@ again: remove_next = 1 + (end > next->vm_end);
fb47a38f
JR
431 if (remove_next) {
432 if (file) {
433 uprobe_munmap(next, next->vm_start, next->vm_end);
434- fput(file);
435+ vma_fput(vma);
436 }
437 if (next->anon_vma)
438 anon_vma_merge(vma, next);
38d290e6 439@@ -1641,8 +1641,8 @@ out:
fb47a38f
JR
440 unmap_and_free_vma:
441 if (vm_flags & VM_DENYWRITE)
442 allow_write_access(file);
443+ vma_fput(vma);
444 vma->vm_file = NULL;
445- fput(file);
446
447 /* Undo any partial mapping done by a device driver. */
448 unmap_region(mm, vma, prev, vma->vm_start, vma->vm_end);
38d290e6 449@@ -2432,7 +2432,7 @@ static int __split_vma(struct mm_struct * mm, struct vm_area_struct * vma,
fb47a38f
JR
450 goto out_free_mpol;
451
452 if (new->vm_file)
453- get_file(new->vm_file);
454+ vma_get_file(new);
455
456 if (new->vm_ops && new->vm_ops->open)
457 new->vm_ops->open(new);
38d290e6 458@@ -2451,7 +2451,7 @@ static int __split_vma(struct mm_struct * mm, struct vm_area_struct * vma,
fb47a38f
JR
459 if (new->vm_ops && new->vm_ops->close)
460 new->vm_ops->close(new);
461 if (new->vm_file)
462- fput(new->vm_file);
463+ vma_fput(new);
464 unlink_anon_vmas(new);
465 out_free_mpol:
466 mpol_put(vma_policy(new));
38d290e6 467@@ -2840,7 +2840,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
fb47a38f
JR
468 if (anon_vma_clone(new_vma, vma))
469 goto out_free_mempol;
470 if (new_vma->vm_file)
471- get_file(new_vma->vm_file);
472+ vma_get_file(new_vma);
473 if (new_vma->vm_ops && new_vma->vm_ops->open)
474 new_vma->vm_ops->open(new_vma);
475 vma_link(mm, new_vma, prev, rb_link, rb_parent);
476diff --git a/mm/msync.c b/mm/msync.c
477index 632df45..02d770e 100644
478--- a/mm/msync.c
479+++ b/mm/msync.c
480@@ -80,10 +80,10 @@ SYSCALL_DEFINE3(msync, unsigned long, start, size_t, len, int, flags)
481 start = vma->vm_end;
482 if ((flags & MS_SYNC) && file &&
483 (vma->vm_flags & VM_SHARED)) {
484- get_file(file);
485+ vma_get_file(vma);
486 up_read(&mm->mmap_sem);
487 error = vfs_fsync(file, 0);
488- fput(file);
489+ vma_fput(vma);
490 if (error || start >= end)
491 goto out;
492 down_read(&mm->mmap_sem);
493diff --git a/mm/nommu.c b/mm/nommu.c
38d290e6 494index 85f8d66..9f471fa 100644
fb47a38f
JR
495--- a/mm/nommu.c
496+++ b/mm/nommu.c
38d290e6 497@@ -655,7 +655,7 @@ static void __put_nommu_region(struct vm_region *region)
fb47a38f
JR
498 up_write(&nommu_region_sem);
499
500 if (region->vm_file)
501- fput(region->vm_file);
502+ vmr_fput(region);
503
504 /* IO memory and memory shared directly out of the pagecache
505 * from ramfs/tmpfs mustn't be released here */
38d290e6 506@@ -820,7 +820,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
fb47a38f
JR
507 if (vma->vm_ops && vma->vm_ops->close)
508 vma->vm_ops->close(vma);
509 if (vma->vm_file)
510- fput(vma->vm_file);
511+ vma_fput(vma);
512 put_nommu_region(vma->vm_region);
513 kmem_cache_free(vm_area_cachep, vma);
514 }
38d290e6 515@@ -1382,7 +1382,7 @@ unsigned long do_mmap_pgoff(struct file *file,
fb47a38f
JR
516 goto error_just_free;
517 }
518 }
519- fput(region->vm_file);
520+ vmr_fput(region);
521 kmem_cache_free(vm_region_jar, region);
522 region = pregion;
523 result = start;
38d290e6 524@@ -1458,10 +1458,10 @@ error_just_free:
fb47a38f
JR
525 up_write(&nommu_region_sem);
526 error:
527 if (region->vm_file)
528- fput(region->vm_file);
529+ vmr_fput(region);
530 kmem_cache_free(vm_region_jar, region);
531 if (vma->vm_file)
532- fput(vma->vm_file);
533+ vma_fput(vma);
534 kmem_cache_free(vm_area_cachep, vma);
535 kleave(" = %d", ret);
536 return ret;
38d290e6 537aufs3.15 standalone patch
7f207e10 538
1e00d052 539diff --git a/fs/inode.c b/fs/inode.c
38d290e6 540index 2d72083..30b69da 100644
1e00d052
AM
541--- a/fs/inode.c
542+++ b/fs/inode.c
392086de 543@@ -57,6 +57,7 @@ static struct hlist_head *inode_hashtable __read_mostly;
4b3da204 544 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
2cbb1c4b
JR
545
546 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
2cbb1c4b 547+EXPORT_SYMBOL(inode_sb_list_lock);
7f207e10
AM
548
549 /*
4b3da204 550 * Empty aops. Can be used for the cases where the user does not
38d290e6 551@@ -1512,6 +1513,7 @@ int update_time(struct inode *inode, struct timespec *time, int flags)
0c3ec466
AM
552 mark_inode_dirty_sync(inode);
553 return 0;
554 }
555+EXPORT_SYMBOL(update_time);
556
557 /**
558 * touch_atime - update the access time
7f207e10 559diff --git a/fs/namespace.c b/fs/namespace.c
38d290e6 560index 182bc41..c88e101 100644
7f207e10
AM
561--- a/fs/namespace.c
562+++ b/fs/namespace.c
38d290e6 563@@ -453,6 +453,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
c06a8ce3
AM
564 mnt_dec_writers(real_mount(mnt));
565 preempt_enable();
566 }
567+EXPORT_SYMBOL_GPL(__mnt_drop_write);
568
569 /**
570 * mnt_drop_write - give up write access to a mount
38d290e6 571@@ -1564,6 +1565,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
7f207e10
AM
572 }
573 return 0;
574 }
575+EXPORT_SYMBOL(iterate_mounts);
576
7eafdf33 577 static void cleanup_group_ids(struct mount *mnt, struct mount *end)
7f207e10
AM
578 {
579diff --git a/fs/notify/group.c b/fs/notify/group.c
fb47a38f 580index ad19959..adf290d 100644
7f207e10
AM
581--- a/fs/notify/group.c
582+++ b/fs/notify/group.c
583@@ -22,6 +22,7 @@
584 #include <linux/srcu.h>
585 #include <linux/rculist.h>
586 #include <linux/wait.h>
587+#include <linux/module.h>
588
589 #include <linux/fsnotify_backend.h>
590 #include "fsnotify.h"
fb47a38f 591@@ -72,6 +73,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
1716fcea
AM
592 {
593 atomic_inc(&group->refcnt);
594 }
595+EXPORT_SYMBOL(fsnotify_get_group);
596
597 /*
598 * Drop a reference to a group. Free it if it's through.
fb47a38f 599@@ -81,6 +83,7 @@ void fsnotify_put_group(struct fsnotify_group *group)
7f207e10 600 if (atomic_dec_and_test(&group->refcnt))
1716fcea 601 fsnotify_final_destroy_group(group);
7f207e10
AM
602 }
603+EXPORT_SYMBOL(fsnotify_put_group);
604
605 /*
606 * Create a new fsnotify_group and hold a reference for the group returned.
fb47a38f 607@@ -109,6 +112,7 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
7f207e10
AM
608
609 return group;
610 }
611+EXPORT_SYMBOL(fsnotify_alloc_group);
1716fcea
AM
612
613 int fsnotify_fasync(int fd, struct file *file, int on)
614 {
7f207e10 615diff --git a/fs/notify/mark.c b/fs/notify/mark.c
392086de 616index 923fe4a..176b435 100644
7f207e10
AM
617--- a/fs/notify/mark.c
618+++ b/fs/notify/mark.c
392086de 619@@ -109,6 +109,7 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
7f207e10 620 mark->free_mark(mark);
1716fcea 621 }
7f207e10
AM
622 }
623+EXPORT_SYMBOL(fsnotify_put_mark);
624
625 /*
626 * Any time a mark is getting freed we end up here.
392086de 627@@ -191,6 +192,7 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark,
1716fcea
AM
628 fsnotify_destroy_mark_locked(mark, group);
629 mutex_unlock(&group->mark_mutex);
7f207e10
AM
630 }
631+EXPORT_SYMBOL(fsnotify_destroy_mark);
632
633 void fsnotify_set_mark_mask_locked(struct fsnotify_mark *mark, __u32 mask)
634 {
392086de 635@@ -275,6 +277,7 @@ err:
7f207e10
AM
636
637 return ret;
638 }
639+EXPORT_SYMBOL(fsnotify_add_mark);
640
1716fcea
AM
641 int fsnotify_add_mark(struct fsnotify_mark *mark, struct fsnotify_group *group,
642 struct inode *inode, struct vfsmount *mnt, int allow_dups)
392086de 643@@ -336,6 +339,7 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
7f207e10
AM
644 atomic_set(&mark->refcnt, 1);
645 mark->free_mark = free_mark;
646 }
647+EXPORT_SYMBOL(fsnotify_init_mark);
648
649 static int fsnotify_mark_destroy(void *ignored)
650 {
651diff --git a/fs/open.c b/fs/open.c
38d290e6 652index 9d64679..1aa876a 100644
7f207e10
AM
653--- a/fs/open.c
654+++ b/fs/open.c
523b37e3 655@@ -62,6 +62,7 @@ int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
7f207e10
AM
656 mutex_unlock(&dentry->d_inode->i_mutex);
657 return ret;
658 }
659+EXPORT_SYMBOL(do_truncate);
660
1716fcea 661 long vfs_truncate(struct path *path, loff_t length)
7f207e10 662 {
38d290e6
JR
663@@ -299,6 +300,7 @@ int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
664 sb_end_write(inode->i_sb);
665 return ret;
666 }
667+EXPORT_SYMBOL(do_fallocate);
668
669 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
670 {
7f207e10 671diff --git a/fs/splice.c b/fs/splice.c
38d290e6 672index 4797013..44b603a 100644
7f207e10
AM
673--- a/fs/splice.c
674+++ b/fs/splice.c
38d290e6 675@@ -1116,6 +1116,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
392086de
AM
676
677 return splice_write(pipe, out, ppos, len, flags);
7f207e10
AM
678 }
679+EXPORT_SYMBOL(do_splice_from);
680
681 /*
682 * Attempt to initiate a splice from a file to a pipe.
38d290e6 683@@ -1142,6 +1143,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
7f207e10
AM
684
685 return splice_read(in, ppos, pipe, len, flags);
686 }
687+EXPORT_SYMBOL(do_splice_to);
688
689 /**
690 * splice_direct_to_actor - splices data directly between two non-pipes
691diff --git a/security/commoncap.c b/security/commoncap.c
392086de 692index b9d613e..ba3b618 100644
7f207e10
AM
693--- a/security/commoncap.c
694+++ b/security/commoncap.c
1716fcea 695@@ -988,9 +988,11 @@ int cap_mmap_addr(unsigned long addr)
94337f0d 696 }
7f207e10
AM
697 return ret;
698 }
0c3ec466
AM
699+EXPORT_SYMBOL(cap_mmap_addr);
700
701 int cap_mmap_file(struct file *file, unsigned long reqprot,
702 unsigned long prot, unsigned long flags)
703 {
704 return 0;
705 }
706+EXPORT_SYMBOL(cap_mmap_file);
7f207e10 707diff --git a/security/device_cgroup.c b/security/device_cgroup.c
38d290e6 708index 9134dbf..4d4217a 100644
7f207e10
AM
709--- a/security/device_cgroup.c
710+++ b/security/device_cgroup.c
f6c5ef8b
AM
711@@ -7,6 +7,7 @@
712 #include <linux/device_cgroup.h>
713 #include <linux/cgroup.h>
714 #include <linux/ctype.h>
715+#include <linux/export.h>
716 #include <linux/list.h>
717 #include <linux/uaccess.h>
718 #include <linux/seq_file.h>
38d290e6 719@@ -856,6 +857,7 @@ int __devcgroup_inode_permission(struct inode *inode, int mask)
537831f9
AM
720 return __devcgroup_check_permission(type, imajor(inode), iminor(inode),
721 access);
7f207e10 722 }
2cbb1c4b 723+EXPORT_SYMBOL(__devcgroup_inode_permission);
7f207e10
AM
724
725 int devcgroup_inode_mknod(int mode, dev_t dev)
726 {
727diff --git a/security/security.c b/security/security.c
38d290e6 728index 8b774f3..e3190b9 100644
7f207e10
AM
729--- a/security/security.c
730+++ b/security/security.c
392086de 731@@ -407,6 +407,7 @@ int security_path_rmdir(struct path *dir, struct dentry *dentry)
7f207e10
AM
732 return 0;
733 return security_ops->path_rmdir(dir, dentry);
734 }
735+EXPORT_SYMBOL(security_path_rmdir);
736
737 int security_path_unlink(struct path *dir, struct dentry *dentry)
738 {
392086de 739@@ -423,6 +424,7 @@ int security_path_symlink(struct path *dir, struct dentry *dentry,
7f207e10
AM
740 return 0;
741 return security_ops->path_symlink(dir, dentry, old_name);
742 }
743+EXPORT_SYMBOL(security_path_symlink);
744
745 int security_path_link(struct dentry *old_dentry, struct path *new_dir,
746 struct dentry *new_dentry)
392086de 747@@ -431,6 +433,7 @@ int security_path_link(struct dentry *old_dentry, struct path *new_dir,
7f207e10
AM
748 return 0;
749 return security_ops->path_link(old_dentry, new_dir, new_dentry);
750 }
751+EXPORT_SYMBOL(security_path_link);
752
753 int security_path_rename(struct path *old_dir, struct dentry *old_dentry,
38d290e6
JR
754 struct path *new_dir, struct dentry *new_dentry,
755@@ -458,6 +461,7 @@ int security_path_truncate(struct path *path)
7f207e10
AM
756 return 0;
757 return security_ops->path_truncate(path);
758 }
759+EXPORT_SYMBOL(security_path_truncate);
760
7eafdf33
AM
761 int security_path_chmod(struct path *path, umode_t mode)
762 {
38d290e6 763@@ -465,6 +469,7 @@ int security_path_chmod(struct path *path, umode_t mode)
7f207e10 764 return 0;
7eafdf33 765 return security_ops->path_chmod(path, mode);
7f207e10
AM
766 }
767+EXPORT_SYMBOL(security_path_chmod);
768
537831f9 769 int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
7f207e10 770 {
38d290e6 771@@ -472,6 +477,7 @@ int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
7f207e10
AM
772 return 0;
773 return security_ops->path_chown(path, uid, gid);
774 }
775+EXPORT_SYMBOL(security_path_chown);
776
777 int security_path_chroot(struct path *path)
778 {
38d290e6 779@@ -557,6 +563,7 @@ int security_inode_readlink(struct dentry *dentry)
7f207e10
AM
780 return 0;
781 return security_ops->inode_readlink(dentry);
782 }
783+EXPORT_SYMBOL(security_inode_readlink);
784
785 int security_inode_follow_link(struct dentry *dentry, struct nameidata *nd)
786 {
38d290e6 787@@ -571,6 +578,7 @@ int security_inode_permission(struct inode *inode, int mask)
7f207e10 788 return 0;
1e00d052 789 return security_ops->inode_permission(inode, mask);
7f207e10
AM
790 }
791+EXPORT_SYMBOL(security_inode_permission);
792
1e00d052 793 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
7f207e10 794 {
38d290e6 795@@ -693,6 +701,7 @@ int security_file_permission(struct file *file, int mask)
7f207e10
AM
796
797 return fsnotify_perm(file, mask);
798 }
799+EXPORT_SYMBOL(security_file_permission);
800
801 int security_file_alloc(struct file *file)
802 {
38d290e6 803@@ -753,6 +762,7 @@ int security_mmap_file(struct file *file, unsigned long prot,
7f207e10
AM
804 return ret;
805 return ima_file_mmap(file, prot);
806 }
0c3ec466 807+EXPORT_SYMBOL(security_mmap_file);
7f207e10 808
0c3ec466
AM
809 int security_mmap_addr(unsigned long addr)
810 {
7f207e10
AM
811diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
812--- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
38d290e6 813+++ linux/Documentation/ABI/testing/debugfs-aufs 2014-06-24 00:14:06.769385243 +0200
86dc4139 814@@ -0,0 +1,50 @@
7f207e10
AM
815+What: /debug/aufs/si_<id>/
816+Date: March 2009
f6b6e03d 817+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
818+Description:
819+ Under /debug/aufs, a directory named si_<id> is created
820+ per aufs mount, where <id> is a unique id generated
821+ internally.
1facf9fc 822+
86dc4139
AM
823+What: /debug/aufs/si_<id>/plink
824+Date: Apr 2013
f6b6e03d 825+Contact: J. R. Okajima <hooanon05g@gmail.com>
86dc4139
AM
826+Description:
827+ It has three lines and shows the information about the
828+ pseudo-link. The first line is a single number
829+ representing a number of buckets. The second line is a
830+ number of pseudo-links per buckets (separated by a
831+ blank). The last line is a single number representing a
832+ total number of psedo-links.
833+ When the aufs mount option 'noplink' is specified, it
834+ will show "1\n0\n0\n".
835+
7f207e10
AM
836+What: /debug/aufs/si_<id>/xib
837+Date: March 2009
f6b6e03d 838+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
839+Description:
840+ It shows the consumed blocks by xib (External Inode Number
841+ Bitmap), its block size and file size.
842+ When the aufs mount option 'noxino' is specified, it
843+ will be empty. About XINO files, see the aufs manual.
844+
845+What: /debug/aufs/si_<id>/xino0, xino1 ... xinoN
846+Date: March 2009
f6b6e03d 847+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
848+Description:
849+ It shows the consumed blocks by xino (External Inode Number
850+ Translation Table), its link count, block size and file
851+ size.
852+ When the aufs mount option 'noxino' is specified, it
853+ will be empty. About XINO files, see the aufs manual.
854+
855+What: /debug/aufs/si_<id>/xigen
856+Date: March 2009
f6b6e03d 857+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
858+Description:
859+ It shows the consumed blocks by xigen (External Inode
860+ Generation Table), its block size and file size.
861+ If CONFIG_AUFS_EXPORT is disabled, this entry will not
862+ be created.
863+ When the aufs mount option 'noxino' is specified, it
864+ will be empty. About XINO files, see the aufs manual.
865diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
866--- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
38d290e6 867+++ linux/Documentation/ABI/testing/sysfs-aufs 2014-06-24 00:14:06.769385243 +0200
392086de 868@@ -0,0 +1,31 @@
7f207e10
AM
869+What: /sys/fs/aufs/si_<id>/
870+Date: March 2009
f6b6e03d 871+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
872+Description:
873+ Under /sys/fs/aufs, a directory named si_<id> is created
874+ per aufs mount, where <id> is a unique id generated
875+ internally.
876+
877+What: /sys/fs/aufs/si_<id>/br0, br1 ... brN
878+Date: March 2009
f6b6e03d 879+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
880+Description:
881+ It shows the abolute path of a member directory (which
882+ is called branch) in aufs, and its permission.
883+
392086de
AM
884+What: /sys/fs/aufs/si_<id>/brid0, brid1 ... bridN
885+Date: July 2013
f6b6e03d 886+Contact: J. R. Okajima <hooanon05g@gmail.com>
392086de
AM
887+Description:
888+ It shows the id of a member directory (which is called
889+ branch) in aufs.
890+
7f207e10
AM
891+What: /sys/fs/aufs/si_<id>/xi_path
892+Date: March 2009
f6b6e03d 893+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
894+Description:
895+ It shows the abolute path of XINO (External Inode Number
896+ Bitmap, Translation Table and Generation Table) file
897+ even if it is the default path.
898+ When the aufs mount option 'noxino' is specified, it
899+ will be empty. About XINO files, see the aufs manual.
53392da6
AM
900diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
901--- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 902+++ linux/Documentation/filesystems/aufs/design/01intro.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 903@@ -0,0 +1,161 @@
53392da6 904+
523b37e3 905+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
906+#
907+# This program is free software; you can redistribute it and/or modify
908+# it under the terms of the GNU General Public License as published by
909+# the Free Software Foundation; either version 2 of the License, or
910+# (at your option) any later version.
911+#
912+# This program is distributed in the hope that it will be useful,
913+# but WITHOUT ANY WARRANTY; without even the implied warranty of
914+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
915+# GNU General Public License for more details.
916+#
917+# You should have received a copy of the GNU General Public License
523b37e3 918+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
919+
920+Introduction
921+----------------------------------------
922+
923+aufs [ei ju: ef es] | [a u f s]
924+1. abbrev. for "advanced multi-layered unification filesystem".
925+2. abbrev. for "another unionfs".
926+3. abbrev. for "auf das" in German which means "on the" in English.
927+ Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
928+ But "Filesystem aufs Filesystem" is hard to understand.
929+
930+AUFS is a filesystem with features:
931+- multi layered stackable unification filesystem, the member directory
932+ is called as a branch.
933+- branch permission and attribute, 'readonly', 'real-readonly',
934+ 'readwrite', 'whiteout-able', 'link-able whiteout' and their
935+ combination.
936+- internal "file copy-on-write".
937+- logical deletion, whiteout.
938+- dynamic branch manipulation, adding, deleting and changing permission.
939+- allow bypassing aufs, user's direct branch access.
940+- external inode number translation table and bitmap which maintains the
941+ persistent aufs inode number.
942+- seekable directory, including NFS readdir.
943+- file mapping, mmap and sharing pages.
944+- pseudo-link, hardlink over branches.
945+- loopback mounted filesystem as a branch.
946+- several policies to select one among multiple writable branches.
947+- revert a single systemcall when an error occurs in aufs.
948+- and more...
949+
950+
951+Multi Layered Stackable Unification Filesystem
952+----------------------------------------------------------------------
953+Most people already knows what it is.
954+It is a filesystem which unifies several directories and provides a
955+merged single directory. When users access a file, the access will be
956+passed/re-directed/converted (sorry, I am not sure which English word is
957+correct) to the real file on the member filesystem. The member
958+filesystem is called 'lower filesystem' or 'branch' and has a mode
959+'readonly' and 'readwrite.' And the deletion for a file on the lower
960+readonly branch is handled by creating 'whiteout' on the upper writable
961+branch.
962+
963+On LKML, there have been discussions about UnionMount (Jan Blunck,
964+Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
965+different approaches to implement the merged-view.
966+The former tries putting it into VFS, and the latter implements as a
967+separate filesystem.
968+(If I misunderstand about these implementations, please let me know and
969+I shall correct it. Because it is a long time ago when I read their
970+source files last time).
971+
972+UnionMount's approach will be able to small, but may be hard to share
973+branches between several UnionMount since the whiteout in it is
974+implemented in the inode on branch filesystem and always
975+shared. According to Bharata's post, readdir does not seems to be
976+finished yet.
977+There are several missing features known in this implementations such as
978+- for users, the inode number may change silently. eg. copy-up.
979+- link(2) may break by copy-up.
980+- read(2) may get an obsoleted filedata (fstat(2) too).
981+- fcntl(F_SETLK) may be broken by copy-up.
982+- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
983+ open(O_RDWR).
984+
985+Unionfs has a longer history. When I started implementing a stacking filesystem
986+(Aug 2005), it already existed. It has virtual super_block, inode,
987+dentry and file objects and they have an array pointing lower same kind
988+objects. After contributing many patches for Unionfs, I re-started my
989+project AUFS (Jun 2006).
990+
991+In AUFS, the structure of filesystem resembles to Unionfs, but I
992+implemented my own ideas, approaches and enhancements and it became
993+totally different one.
994+
995+Comparing DM snapshot and fs based implementation
996+- the number of bytes to be copied between devices is much smaller.
997+- the type of filesystem must be one and only.
998+- the fs must be writable, no readonly fs, even for the lower original
999+ device. so the compression fs will not be usable. but if we use
1000+ loopback mount, we may address this issue.
1001+ for instance,
1002+ mount /cdrom/squashfs.img /sq
1003+ losetup /sq/ext2.img
1004+ losetup /somewhere/cow
1005+ dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1006+- it will be difficult (or needs more operations) to extract the
1007+ difference between the original device and COW.
1008+- DM snapshot-merge may help a lot when users try merging. in the
1009+ fs-layer union, users will use rsync(1).
1010+
1011+
1012+Several characters/aspects of aufs
1013+----------------------------------------------------------------------
1014+
1015+Aufs has several characters or aspects.
1016+1. a filesystem, callee of VFS helper
1017+2. sub-VFS, caller of VFS helper for branches
1018+3. a virtual filesystem which maintains persistent inode number
1019+4. reader/writer of files on branches such like an application
1020+
1021+1. Callee of VFS Helper
1022+As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1023+unlink(2) from an application reaches sys_unlink() kernel function and
1024+then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1025+calls filesystem specific unlink operation. Actually aufs implements the
1026+unlink operation but it behaves like a redirector.
1027+
1028+2. Caller of VFS Helper for Branches
1029+aufs_unlink() passes the unlink request to the branch filesystem as if
1030+it were called from VFS. So the called unlink operation of the branch
1031+filesystem acts as usual. As a caller of VFS helper, aufs should handle
1032+every necessary pre/post operation for the branch filesystem.
1033+- acquire the lock for the parent dir on a branch
1034+- lookup in a branch
1035+- revalidate dentry on a branch
1036+- mnt_want_write() for a branch
1037+- vfs_unlink() for a branch
1038+- mnt_drop_write() for a branch
1039+- release the lock on a branch
1040+
1041+3. Persistent Inode Number
1042+One of the most important issue for a filesystem is to maintain inode
1043+numbers. This is particularly important to support exporting a
1044+filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1045+backend block device for its own. But some storage is necessary to
1046+maintain inode number. It may be a large space and may not suit to keep
1047+in memory. Aufs rents some space from its first writable branch
1048+filesystem (by default) and creates file(s) on it. These files are
1049+created by aufs internally and removed soon (currently) keeping opened.
1050+Note: Because these files are removed, they are totally gone after
1051+ unmounting aufs. It means the inode numbers are not persistent
1052+ across unmount or reboot. I have a plan to make them really
1053+ persistent which will be important for aufs on NFS server.
1054+
1055+4. Read/Write Files Internally (copy-on-write)
1056+Because a branch can be readonly, when you write a file on it, aufs will
1057+"copy-up" it to the upper writable branch internally. And then write the
1058+originally requested thing to the file. Generally kernel doesn't
1059+open/read/write file actively. In aufs, even a single write may cause a
1060+internal "file copy". This behaviour is very similar to cp(1) command.
1061+
1062+Some people may think it is better to pass such work to user space
1063+helper, instead of doing in kernel space. Actually I am still thinking
1064+about it. But currently I have implemented it in kernel space.
1065diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1066--- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1067+++ linux/Documentation/filesystems/aufs/design/02struct.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1068@@ -0,0 +1,242 @@
53392da6 1069+
523b37e3 1070+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1071+#
1072+# This program is free software; you can redistribute it and/or modify
1073+# it under the terms of the GNU General Public License as published by
1074+# the Free Software Foundation; either version 2 of the License, or
1075+# (at your option) any later version.
1076+#
1077+# This program is distributed in the hope that it will be useful,
1078+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1079+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1080+# GNU General Public License for more details.
1081+#
1082+# You should have received a copy of the GNU General Public License
523b37e3 1083+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1084+
1085+Basic Aufs Internal Structure
1086+
1087+Superblock/Inode/Dentry/File Objects
1088+----------------------------------------------------------------------
1089+As like an ordinary filesystem, aufs has its own
1090+superblock/inode/dentry/file objects. All these objects have a
1091+dynamically allocated array and store the same kind of pointers to the
1092+lower filesystem, branch.
1093+For example, when you build a union with one readwrite branch and one
1094+readonly, mounted /au, /rw and /ro respectively.
1095+- /au = /rw + /ro
1096+- /ro/fileA exists but /rw/fileA
1097+
1098+Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1099+pointers are stored in a aufs dentry. The array in aufs dentry will be,
1100+- [0] = NULL
1101+- [1] = /ro/fileA
1102+
1103+This style of an array is essentially same to the aufs
1104+superblock/inode/dentry/file objects.
1105+
1106+Because aufs supports manipulating branches, ie. add/delete/change
1107+dynamically, these objects has its own generation. When branches are
1108+changed, the generation in aufs superblock is incremented. And a
1109+generation in other object are compared when it is accessed.
1110+When a generation in other objects are obsoleted, aufs refreshes the
1111+internal array.
1112+
1113+
1114+Superblock
1115+----------------------------------------------------------------------
1116+Additionally aufs superblock has some data for policies to select one
1117+among multiple writable branches, XIB files, pseudo-links and kobject.
1118+See below in detail.
1119+About the policies which supports copy-down a directory, see policy.txt
1120+too.
1121+
1122+
1123+Branch and XINO(External Inode Number Translation Table)
1124+----------------------------------------------------------------------
1125+Every branch has its own xino (external inode number translation table)
1126+file. The xino file is created and unlinked by aufs internally. When two
1127+members of a union exist on the same filesystem, they share the single
1128+xino file.
1129+The struct of a xino file is simple, just a sequence of aufs inode
1130+numbers which is indexed by the lower inode number.
1131+In the above sample, assume the inode number of /ro/fileA is i111 and
1132+aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1133+4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1134+
1135+When the inode numbers are not contiguous, the xino file will be sparse
1136+which has a hole in it and doesn't consume as much disk space as it
1137+might appear. If your branch filesystem consumes disk space for such
1138+holes, then you should specify 'xino=' option at mounting aufs.
1139+
1140+Also a writable branch has three kinds of "whiteout bases". All these
1141+are existed when the branch is joined to aufs and the names are
1142+whiteout-ed doubly, so that users will never see their names in aufs
1143+hierarchy.
1144+1. a regular file which will be linked to all whiteouts.
1145+2. a directory to store a pseudo-link.
1146+3. a directory to store an "orphan-ed" file temporary.
1147+
1148+1. Whiteout Base
1149+ When you remove a file on a readonly branch, aufs handles it as a
1150+ logical deletion and creates a whiteout on the upper writable branch
1151+ as a hardlink of this file in order not to consume inode on the
1152+ writable branch.
1153+2. Pseudo-link Dir
1154+ See below, Pseudo-link.
1155+3. Step-Parent Dir
1156+ When "fileC" exists on the lower readonly branch only and it is
1157+ opened and removed with its parent dir, and then user writes
1158+ something into it, then aufs copies-up fileC to this
1159+ directory. Because there is no other dir to store fileC. After
1160+ creating a file under this dir, the file is unlinked.
1161+
1162+Because aufs supports manipulating branches, ie. add/delete/change
1163+dynamically, a branch has its own id. When the branch order changes, aufs
1164+finds the new index by searching the branch id.
1165+
1166+
1167+Pseudo-link
1168+----------------------------------------------------------------------
1169+Assume "fileA" exists on the lower readonly branch only and it is
1170+hardlinked to "fileB" on the branch. When you write something to fileA,
1171+aufs copies-up it to the upper writable branch. Additionally aufs
1172+creates a hardlink under the Pseudo-link Directory of the writable
1173+branch. The inode of a pseudo-link is kept in aufs super_block as a
1174+simple list. If fileB is read after unlinking fileA, aufs returns
1175+filedata from the pseudo-link instead of the lower readonly
1176+branch. Because the pseudo-link is based upon the inode, to keep the
1177+inode number by xino (see above) is important.
1178+
1179+All the hardlinks under the Pseudo-link Directory of the writable branch
1180+should be restored in a proper location later. Aufs provides a utility
1181+to do this. The userspace helpers executed at remounting and unmounting
1182+aufs by default.
1183+During this utility is running, it puts aufs into the pseudo-link
1184+maintenance mode. In this mode, only the process which began the
1185+maintenance mode (and its child processes) is allowed to operate in
1186+aufs. Some other processes which are not related to the pseudo-link will
1187+be allowed to run too, but the rest have to return an error or wait
1188+until the maintenance mode ends. If a process already acquires an inode
1189+mutex (in VFS), it has to return an error.
1190+
1191+
1192+XIB(external inode number bitmap)
1193+----------------------------------------------------------------------
1194+Addition to the xino file per a branch, aufs has an external inode number
1195+bitmap in a superblock object. It is also a file such like a xino file.
1196+It is a simple bitmap to mark whether the aufs inode number is in-use or
1197+not.
1198+To reduce the file I/O, aufs prepares a single memory page to cache xib.
1199+
1200+Aufs implements a feature to truncate/refresh both of xino and xib to
1201+reduce the number of consumed disk blocks for these files.
1202+
1203+
1204+Virtual or Vertical Dir, and Readdir in Userspace
1205+----------------------------------------------------------------------
1206+In order to support multiple layers (branches), aufs readdir operation
1207+constructs a virtual dir block on memory. For readdir, aufs calls
1208+vfs_readdir() internally for each dir on branches, merges their entries
1209+with eliminating the whiteout-ed ones, and sets it to file (dir)
1210+object. So the file object has its entry list until it is closed. The
1211+entry list will be updated when the file position is zero and becomes
1212+old. This decision is made in aufs automatically.
1213+
1214+The dynamically allocated memory block for the name of entries has a
1215+unit of 512 bytes (by default) and stores the names contiguously (no
1216+padding). Another block for each entry is handled by kmem_cache too.
1217+During building dir blocks, aufs creates hash list and judging whether
1218+the entry is whiteouted by its upper branch or already listed.
1219+The merged result is cached in the corresponding inode object and
1220+maintained by a customizable life-time option.
1221+
1222+Some people may call it can be a security hole or invite DoS attack
1223+since the opened and once readdir-ed dir (file object) holds its entry
1224+list and becomes a pressure for system memory. But I'd say it is similar
1225+to files under /proc or /sys. The virtual files in them also holds a
1226+memory page (generally) while they are opened. When an idea to reduce
1227+memory for them is introduced, it will be applied to aufs too.
1228+For those who really hate this situation, I've developed readdir(3)
1229+library which operates this merging in userspace. You just need to set
1230+LD_PRELOAD environment variable, and aufs will not consume no memory in
1231+kernel space for readdir(3).
1232+
1233+
1234+Workqueue
1235+----------------------------------------------------------------------
1236+Aufs sometimes requires privilege access to a branch. For instance,
1237+in copy-up/down operation. When a user process is going to make changes
1238+to a file which exists in the lower readonly branch only, and the mode
1239+of one of ancestor directories may not be writable by a user
1240+process. Here aufs copy-up the file with its ancestors and they may
1241+require privilege to set its owner/group/mode/etc.
1242+This is a typical case of a application character of aufs (see
1243+Introduction).
1244+
1245+Aufs uses workqueue synchronously for this case. It creates its own
1246+workqueue. The workqueue is a kernel thread and has privilege. Aufs
1247+passes the request to call mkdir or write (for example), and wait for
1248+its completion. This approach solves a problem of a signal handler
1249+simply.
1250+If aufs didn't adopt the workqueue and changed the privilege of the
1251+process, and if the mkdir/write call arises SIGXFSZ or other signal,
1252+then the user process might gain a privilege or the generated core file
1253+was owned by a superuser.
1254+
1255+Also aufs uses the system global workqueue ("events" kernel thread) too
1256+for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1257+whiteout base and etc. This is unrelated to a privilege.
1258+Most of aufs operation tries acquiring a rw_semaphore for aufs
1259+superblock at the beginning, at the same time waits for the completion
1260+of all queued asynchronous tasks.
1261+
1262+
1263+Whiteout
1264+----------------------------------------------------------------------
1265+The whiteout in aufs is very similar to Unionfs's. That is represented
1266+by its filename. UnionMount takes an approach of a file mode, but I am
1267+afraid several utilities (find(1) or something) will have to support it.
1268+
1269+Basically the whiteout represents "logical deletion" which stops aufs to
1270+lookup further, but also it represents "dir is opaque" which also stop
1271+lookup.
1272+
1273+In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1274+In order to make several functions in a single systemcall to be
1275+revertible, aufs adopts an approach to rename a directory to a temporary
1276+unique whiteouted name.
1277+For example, in rename(2) dir where the target dir already existed, aufs
1278+renames the target dir to a temporary unique whiteouted name before the
1279+actual rename on a branch and then handles other actions (make it opaque,
1280+update the attributes, etc). If an error happens in these actions, aufs
1281+simply renames the whiteouted name back and returns an error. If all are
1282+succeeded, aufs registers a function to remove the whiteouted unique
1283+temporary name completely and asynchronously to the system global
1284+workqueue.
1285+
1286+
1287+Copy-up
1288+----------------------------------------------------------------------
1289+It is a well-known feature or concept.
1290+When user modifies a file on a readonly branch, aufs operate "copy-up"
1291+internally and makes change to the new file on the upper writable branch.
1292+When the trigger systemcall does not update the timestamps of the parent
1293+dir, aufs reverts it after copy-up.
c2b27bf2
AM
1294+
1295+
1296+Move-down (aufs3.9 and later)
1297+----------------------------------------------------------------------
1298+"Copy-up" is one of the essential feature in aufs. It copies a file from
1299+the lower readonly branch to the upper writable branch when a user
1300+changes something about the file.
1301+"Move-down" is an opposite action of copy-up. Basically this action is
1302+ran manually instead of automatically and internally.
1303+
1304+Sometimes users want to move-down a file from the upper writable branch
1305+to the lower readonly or writable branch. For instance,
1306+- the free space of the upper writable branch is going to run out.
1307+- create a new intermediate branch between the upper and lower branch.
1308+- etc.
1309+
1310+For this purpose, use "aumvdown" command in aufs-util.git.
53392da6
AM
1311diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1312--- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1313+++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1314@@ -0,0 +1,105 @@
53392da6 1315+
523b37e3 1316+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1317+#
1318+# This program is free software; you can redistribute it and/or modify
1319+# it under the terms of the GNU General Public License as published by
1320+# the Free Software Foundation; either version 2 of the License, or
1321+# (at your option) any later version.
1322+#
1323+# This program is distributed in the hope that it will be useful,
1324+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1325+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1326+# GNU General Public License for more details.
1327+#
1328+# You should have received a copy of the GNU General Public License
523b37e3 1329+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1330+
1331+Lookup in a Branch
1332+----------------------------------------------------------------------
1333+Since aufs has a character of sub-VFS (see Introduction), it operates
1334+lookup for branches as VFS does. It may be a heavy work. Generally
1335+speaking struct nameidata is a bigger structure and includes many
1336+information. But almost all lookup operation in aufs is the simplest
1337+case, ie. lookup only an entry directly connected to its parent. Digging
1338+down the directory hierarchy is unnecessary.
1339+
1340+VFS has a function lookup_one_len() for that use, but it is not usable
1341+for a branch filesystem which requires struct nameidata. So aufs
1342+implements a simple lookup wrapper function. When a branch filesystem
1343+allows NULL as nameidata, it calls lookup_one_len(). Otherwise it builds
1344+a simplest nameidata and calls lookup_hash().
1345+Here aufs applies "a principle in NFSD", ie. if the filesystem supports
1346+NFS-export, then it has to support NULL as a nameidata parameter for
1347+->create(), ->lookup() and ->d_revalidate(). So the lookup wrapper in
1348+aufs tests if ->s_export_op in the branch is NULL or not.
1349+
1350+When a branch is a remote filesystem, aufs basically trusts its
1351+->d_revalidate(), also aufs forces the hardest revalidate tests for
1352+them.
1353+For d_revalidate, aufs implements three levels of revalidate tests. See
1354+"Revalidate Dentry and UDBA" in detail.
1355+
1356+
1357+Loopback Mount
1358+----------------------------------------------------------------------
1359+Basically aufs supports any type of filesystem and block device for a
1360+branch (actually there are some exceptions). But it is prohibited to add
1361+a loopback mounted one whose backend file exists in a filesystem which is
1362+already added to aufs. The reason is to protect aufs from a recursive
1363+lookup. If it was allowed, the aufs lookup operation might re-enter a
1364+lookup for the loopback mounted branch in the same context, and will
1365+cause a deadlock.
1366+
1367+
1368+Revalidate Dentry and UDBA (User's Direct Branch Access)
1369+----------------------------------------------------------------------
1370+Generally VFS helpers re-validate a dentry as a part of lookup.
1371+0. digging down the directory hierarchy.
1372+1. lock the parent dir by its i_mutex.
1373+2. lookup the final (child) entry.
1374+3. revalidate it.
1375+4. call the actual operation (create, unlink, etc.)
1376+5. unlock the parent dir
1377+
1378+If the filesystem implements its ->d_revalidate() (step 3), then it is
1379+called. Actually aufs implements it and checks the dentry on a branch is
1380+still valid.
1381+But it is not enough. Because aufs has to release the lock for the
1382+parent dir on a branch at the end of ->lookup() (step 2) and
1383+->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1384+held by VFS.
1385+If the file on a branch is changed directly, eg. bypassing aufs, after
1386+aufs released the lock, then the subsequent operation may cause
1387+something unpleasant result.
1388+
1389+This situation is a result of VFS architecture, ->lookup() and
1390+->d_revalidate() is separated. But I never say it is wrong. It is a good
1391+design from VFS's point of view. It is just not suitable for sub-VFS
1392+character in aufs.
1393+
1394+Aufs supports such case by three level of revalidation which is
1395+selectable by user.
1396+1. Simple Revalidate
1397+ Addition to the native flow in VFS's, confirm the child-parent
1398+ relationship on the branch just after locking the parent dir on the
1399+ branch in the "actual operation" (step 4). When this validation
1400+ fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1401+ checks the validation of the dentry on branches.
1402+2. Monitor Changes Internally by Inotify/Fsnotify
1403+ Addition to above, in the "actual operation" (step 4) aufs re-lookup
1404+ the dentry on the branch, and returns EBUSY if it finds different
1405+ dentry.
1406+ Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1407+ during it is in cache. When the event is notified, aufs registers a
1408+ function to kernel 'events' thread by schedule_work(). And the
1409+ function sets some special status to the cached aufs dentry and inode
1410+ private data. If they are not cached, then aufs has nothing to
1411+ do. When the same file is accessed through aufs (step 0-3) later,
1412+ aufs will detect the status and refresh all necessary data.
1413+ In this mode, aufs has to ignore the event which is fired by aufs
1414+ itself.
1415+3. No Extra Validation
1416+ This is the simplest test and doesn't add any additional revalidation
1417+ test, and skip therevalidatin in step 4. It is useful and improves
1418+ aufs performance when system surely hide the aufs branches from user,
1419+ by over-mounting something (or another method).
1420diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1421--- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1422+++ linux/Documentation/filesystems/aufs/design/04branch.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1423@@ -0,0 +1,75 @@
53392da6 1424+
523b37e3 1425+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1426+#
1427+# This program is free software; you can redistribute it and/or modify
1428+# it under the terms of the GNU General Public License as published by
1429+# the Free Software Foundation; either version 2 of the License, or
1430+# (at your option) any later version.
1431+#
1432+# This program is distributed in the hope that it will be useful,
1433+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1434+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1435+# GNU General Public License for more details.
1436+#
1437+# You should have received a copy of the GNU General Public License
523b37e3 1438+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1439+
1440+Branch Manipulation
1441+
1442+Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1443+and changing its permission/attribute, there are a lot of works to do.
1444+
1445+
1446+Add a Branch
1447+----------------------------------------------------------------------
1448+o Confirm the adding dir exists outside of aufs, including loopback
1449+ mount.
1450+- and other various attributes...
1451+o Initialize the xino file and whiteout bases if necessary.
1452+ See struct.txt.
1453+
1454+o Check the owner/group/mode of the directory
1455+ When the owner/group/mode of the adding directory differs from the
1456+ existing branch, aufs issues a warning because it may impose a
1457+ security risk.
1458+ For example, when a upper writable branch has a world writable empty
1459+ top directory, a malicious user can create any files on the writable
1460+ branch directly, like copy-up and modify manually. If something like
1461+ /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1462+ writable branch, and the writable branch is world-writable, then a
1463+ malicious guy may create /etc/passwd on the writable branch directly
1464+ and the infected file will be valid in aufs.
1465+ I am afraid it can be a security issue, but nothing to do except
1466+ producing a warning.
1467+
1468+
1469+Delete a Branch
1470+----------------------------------------------------------------------
1471+o Confirm the deleting branch is not busy
1472+ To be general, there is one merit to adopt "remount" interface to
1473+ manipulate branches. It is to discard caches. At deleting a branch,
1474+ aufs checks the still cached (and connected) dentries and inodes. If
1475+ there are any, then they are all in-use. An inode without its
1476+ corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1477+
1478+ For the cached one, aufs checks whether the same named entry exists on
1479+ other branches.
1480+ If the cached one is a directory, because aufs provides a merged view
1481+ to users, as long as one dir is left on any branch aufs can show the
1482+ dir to users. In this case, the branch can be removed from aufs.
1483+ Otherwise aufs rejects deleting the branch.
1484+
1485+ If any file on the deleting branch is opened by aufs, then aufs
1486+ rejects deleting.
1487+
1488+
1489+Modify the Permission of a Branch
1490+----------------------------------------------------------------------
1491+o Re-initialize or remove the xino file and whiteout bases if necessary.
1492+ See struct.txt.
1493+
1494+o rw --> ro: Confirm the modifying branch is not busy
1495+ Aufs rejects the request if any of these conditions are true.
1496+ - a file on the branch is mmap-ed.
1497+ - a regular file on the branch is opened for write and there is no
1498+ same named entry on the upper branch.
1499diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1500--- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1501+++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1502@@ -0,0 +1,64 @@
53392da6 1503+
523b37e3 1504+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1505+#
1506+# This program is free software; you can redistribute it and/or modify
1507+# it under the terms of the GNU General Public License as published by
1508+# the Free Software Foundation; either version 2 of the License, or
1509+# (at your option) any later version.
1510+#
1511+# This program is distributed in the hope that it will be useful,
1512+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1513+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1514+# GNU General Public License for more details.
1515+#
1516+# You should have received a copy of the GNU General Public License
523b37e3 1517+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1518+
1519+Policies to Select One among Multiple Writable Branches
1520+----------------------------------------------------------------------
1521+When the number of writable branch is more than one, aufs has to decide
1522+the target branch for file creation or copy-up. By default, the highest
1523+writable branch which has the parent (or ancestor) dir of the target
1524+file is chosen (top-down-parent policy).
1525+By user's request, aufs implements some other policies to select the
1526+writable branch, for file creation two policies, round-robin and
1527+most-free-space policies. For copy-up three policies, top-down-parent,
1528+bottom-up-parent and bottom-up policies.
1529+
1530+As expected, the round-robin policy selects the branch in circular. When
1531+you have two writable branches and creates 10 new files, 5 files will be
1532+created for each branch. mkdir(2) systemcall is an exception. When you
1533+create 10 new directories, all will be created on the same branch.
1534+And the most-free-space policy selects the one which has most free
1535+space among the writable branches. The amount of free space will be
1536+checked by aufs internally, and users can specify its time interval.
1537+
1538+The policies for copy-up is more simple,
1539+top-down-parent is equivalent to the same named on in create policy,
1540+bottom-up-parent selects the writable branch where the parent dir
1541+exists and the nearest upper one from the copyup-source,
1542+bottom-up selects the nearest upper writable branch from the
1543+copyup-source, regardless the existence of the parent dir.
1544+
1545+There are some rules or exceptions to apply these policies.
1546+- If there is a readonly branch above the policy-selected branch and
1547+ the parent dir is marked as opaque (a variation of whiteout), or the
1548+ target (creating) file is whiteout-ed on the upper readonly branch,
1549+ then the result of the policy is ignored and the target file will be
1550+ created on the nearest upper writable branch than the readonly branch.
1551+- If there is a writable branch above the policy-selected branch and
1552+ the parent dir is marked as opaque or the target file is whiteouted
1553+ on the branch, then the result of the policy is ignored and the target
1554+ file will be created on the highest one among the upper writable
1555+ branches who has diropq or whiteout. In case of whiteout, aufs removes
1556+ it as usual.
1557+- link(2) and rename(2) systemcalls are exceptions in every policy.
1558+ They try selecting the branch where the source exists as possible
1559+ since copyup a large file will take long time. If it can't be,
1560+ ie. the branch where the source exists is readonly, then they will
1561+ follow the copyup policy.
1562+- There is an exception for rename(2) when the target exists.
1563+ If the rename target exists, aufs compares the index of the branches
1564+ where the source and the target exists and selects the higher
1565+ one. If the selected branch is readonly, then aufs follows the
1566+ copyup policy.
1567diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
1568--- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1569+++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1570@@ -0,0 +1,46 @@
53392da6 1571+
523b37e3 1572+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1573+#
1574+# This program is free software; you can redistribute it and/or modify
1575+# it under the terms of the GNU General Public License as published by
1576+# the Free Software Foundation; either version 2 of the License, or
1577+# (at your option) any later version.
1578+#
1579+# This program is distributed in the hope that it will be useful,
1580+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1581+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1582+# GNU General Public License for more details.
1583+#
1584+# You should have received a copy of the GNU General Public License
523b37e3 1585+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1586+
1587+mmap(2) -- File Memory Mapping
1588+----------------------------------------------------------------------
1589+In aufs, the file-mapped pages are handled by a branch fs directly, no
1590+interaction with aufs. It means aufs_mmap() calls the branch fs's
1591+->mmap().
1592+This approach is simple and good, but there is one problem.
1593+Under /proc, several entries show the mmap-ped files by its path (with
1594+device and inode number), and the printed path will be the path on the
1595+branch fs's instead of virtual aufs's.
1596+This is not a problem in most cases, but some utilities lsof(1) (and its
1597+user) may expect the path on aufs.
1598+
1599+To address this issue, aufs adds a new member called vm_prfile in struct
1600+vm_area_struct (and struct vm_region). The original vm_file points to
1601+the file on the branch fs in order to handle everything correctly as
1602+usual. The new vm_prfile points to a virtual file in aufs, and the
1603+show-functions in procfs refers to vm_prfile if it is set.
1604+Also we need to maintain several other places where touching vm_file
1605+such like
1606+- fork()/clone() copies vma and the reference count of vm_file is
1607+ incremented.
1608+- merging vma maintains the ref count too.
1609+
1610+This is not a good approach. It just faking the printed path. But it
1611+leaves all behaviour around f_mapping unchanged. This is surely an
1612+advantage.
1613+Actually aufs had adopted another complicated approach which calls
1614+generic_file_mmap() and handles struct vm_operations_struct. In this
1615+approach, aufs met a hard problem and I could not solve it without
1616+switching the approach.
1617diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
1618--- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1619+++ linux/Documentation/filesystems/aufs/design/07export.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1620@@ -0,0 +1,58 @@
53392da6 1621+
523b37e3 1622+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1623+#
1624+# This program is free software; you can redistribute it and/or modify
1625+# it under the terms of the GNU General Public License as published by
1626+# the Free Software Foundation; either version 2 of the License, or
1627+# (at your option) any later version.
1628+#
1629+# This program is distributed in the hope that it will be useful,
1630+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1631+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1632+# GNU General Public License for more details.
1633+#
1634+# You should have received a copy of the GNU General Public License
523b37e3 1635+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1636+
1637+Export Aufs via NFS
1638+----------------------------------------------------------------------
1639+Here is an approach.
1640+- like xino/xib, add a new file 'xigen' which stores aufs inode
1641+ generation.
1642+- iget_locked(): initialize aufs inode generation for a new inode, and
1643+ store it in xigen file.
1644+- destroy_inode(): increment aufs inode generation and store it in xigen
1645+ file. it is necessary even if it is not unlinked, because any data of
1646+ inode may be changed by UDBA.
1647+- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
1648+ build file handle by
1649+ + branch id (4 bytes)
1650+ + superblock generation (4 bytes)
1651+ + inode number (4 or 8 bytes)
1652+ + parent dir inode number (4 or 8 bytes)
1653+ + inode generation (4 bytes))
1654+ + return value of exportfs_encode_fh() for the parent on a branch (4
1655+ bytes)
1656+ + file handle for a branch (by exportfs_encode_fh())
1657+- fh_to_dentry():
1658+ + find the index of a branch from its id in handle, and check it is
1659+ still exist in aufs.
1660+ + 1st level: get the inode number from handle and search it in cache.
1661+ + 2nd level: if not found, get the parent inode number from handle and
1662+ search it in cache. and then open the parent dir, find the matching
1663+ inode number by vfs_readdir() and get its name, and call
1664+ lookup_one_len() for the target dentry.
1665+ + 3rd level: if the parent dir is not cached, call
1666+ exportfs_decode_fh() for a branch and get the parent on a branch,
1667+ build a pathname of it, convert it a pathname in aufs, call
1668+ path_lookup(). now aufs gets a parent dir dentry, then handle it as
1669+ the 2nd level.
1670+ + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
1671+ for every branch, but not itself. to get this, (currently) aufs
1672+ searches in current->nsproxy->mnt_ns list. it may not be a good
1673+ idea, but I didn't get other approach.
1674+ + test the generation of the gotten inode.
1675+- every inode operation: they may get EBUSY due to UDBA. in this case,
1676+ convert it into ESTALE for NFSD.
1677+- readdir(): call lockdep_on/off() because filldir in NFSD calls
1678+ lookup_one_len(), vfs_getattr(), encode_fh() and others.
1679diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
1680--- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1681+++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1682@@ -0,0 +1,52 @@
53392da6 1683+
523b37e3 1684+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1685+#
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.
1690+#
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.
1695+#
1696+# You should have received a copy of the GNU General Public License
523b37e3 1697+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1698+
1699+Show Whiteout Mode (shwh)
1700+----------------------------------------------------------------------
1701+Generally aufs hides the name of whiteouts. But in some cases, to show
1702+them is very useful for users. For instance, creating a new middle layer
1703+(branch) by merging existing layers.
1704+
1705+(borrowing aufs1 HOW-TO from a user, Michael Towers)
1706+When you have three branches,
1707+- Bottom: 'system', squashfs (underlying base system), read-only
1708+- Middle: 'mods', squashfs, read-only
1709+- Top: 'overlay', ram (tmpfs), read-write
1710+
1711+The top layer is loaded at boot time and saved at shutdown, to preserve
1712+the changes made to the system during the session.
1713+When larger changes have been made, or smaller changes have accumulated,
1714+the size of the saved top layer data grows. At this point, it would be
1715+nice to be able to merge the two overlay branches ('mods' and 'overlay')
1716+and rewrite the 'mods' squashfs, clearing the top layer and thus
1717+restoring save and load speed.
1718+
1719+This merging is simplified by the use of another aufs mount, of just the
1720+two overlay branches using the 'shwh' option.
1721+# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
1722+ aufs /livesys/merge_union
1723+
1724+A merged view of these two branches is then available at
1725+/livesys/merge_union, and the new feature is that the whiteouts are
1726+visible!
1727+Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
1728+writing to all branches. Also the default mode for all branches is 'ro'.
1729+It is now possible to save the combined contents of the two overlay
1730+branches to a new squashfs, e.g.:
1731+# mksquashfs /livesys/merge_union /path/to/newmods.squash
1732+
1733+This new squashfs archive can be stored on the boot device and the
1734+initramfs will use it to replace the old one at the next boot.
1735diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
1736--- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1737+++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1738@@ -0,0 +1,46 @@
53392da6 1739+
523b37e3 1740+# Copyright (C) 2010-2014 Junjiro R. Okajima
53392da6
AM
1741+#
1742+# This program is free software; you can redistribute it and/or modify
1743+# it under the terms of the GNU General Public License as published by
1744+# the Free Software Foundation; either version 2 of the License, or
1745+# (at your option) any later version.
1746+#
1747+# This program is distributed in the hope that it will be useful,
1748+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1749+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1750+# GNU General Public License for more details.
1751+#
1752+# You should have received a copy of the GNU General Public License
523b37e3 1753+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1754+
1755+Dynamically customizable FS operations
1756+----------------------------------------------------------------------
1757+Generally FS operations (struct inode_operations, struct
1758+address_space_operations, struct file_operations, etc.) are defined as
1759+"static const", but it never means that FS have only one set of
1760+operation. Some FS have multiple sets of them. For instance, ext2 has
1761+three sets, one for XIP, for NOBH, and for normal.
1762+Since aufs overrides and redirects these operations, sometimes aufs has
1763+to change its behaviour according to the branch FS type. More imporantly
1764+VFS acts differently if a function (member in the struct) is set or
1765+not. It means aufs should have several sets of operations and select one
1766+among them according to the branch FS definition.
1767+
1768+In order to solve this problem and not to affect the behavour of VFS,
1769+aufs defines these operations dynamically. For instance, aufs defines
1770+aio_read function for struct file_operations, but it may not be set to
1771+the file_operations. When the branch FS doesn't have it, aufs doesn't
1772+set it to its file_operations while the function definition itself is
1773+still alive. So the behaviour of io_submit(2) will not change, and it
1774+will return an error when aio_read is not defined.
1775+
1776+The lifetime of these dynamically generated operation object is
1777+maintained by aufs branch object. When the branch is removed from aufs,
1778+the reference counter of the object is decremented. When it reaches
1779+zero, the dynamically generated operation object will be freed.
1780+
1781+This approach is designed to support AIO (io_submit), Direcit I/O and
1782+XIP mainly.
1783+Currently this approach is applied to file_operations and
1784+vm_operations_struct for regular files only.
1785diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/99plan.txt linux/Documentation/filesystems/aufs/design/99plan.txt
1786--- /usr/share/empty/Documentation/filesystems/aufs/design/99plan.txt 1970-01-01 01:00:00.000000000 +0100
38d290e6 1787+++ linux/Documentation/filesystems/aufs/design/99plan.txt 2014-06-24 00:14:06.772718759 +0200
523b37e3 1788@@ -0,0 +1,95 @@
53392da6 1789+
523b37e3 1790+# Copyright (C) 2005-2014 Junjiro R. Okajima
53392da6
AM
1791+#
1792+# This program is free software; you can redistribute it and/or modify
1793+# it under the terms of the GNU General Public License as published by
1794+# the Free Software Foundation; either version 2 of the License, or
1795+# (at your option) any later version.
1796+#
1797+# This program is distributed in the hope that it will be useful,
1798+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1799+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1800+# GNU General Public License for more details.
1801+#
1802+# You should have received a copy of the GNU General Public License
523b37e3 1803+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1804+
1805+Plan
1806+
1807+Restoring some features which was implemented in aufs1.
1808+They were dropped in aufs2 in order to make source files simpler and
1809+easier to be reviewed.
1810+
1811+
1812+Test Only the Highest One for the Directory Permission (dirperm1 option)
1813+----------------------------------------------------------------------
1814+Let's try case study.
1815+- aufs has two branches, upper readwrite and lower readonly.
1816+ /au = /rw + /ro
1817+- "dirA" exists under /ro, but /rw. and its mode is 0700.
1818+- user invoked "chmod a+rx /au/dirA"
1819+- then "dirA" becomes world readable?
1820+
1821+In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1822+or it may be a natively readonly filesystem. If aufs respects the lower
1823+branch, it should not respond readdir request from other users. But user
1824+allowed it by chmod. Should really aufs rejects showing the entries
1825+under /ro/dirA?
1826+
1827+To be honest, I don't have a best solution for this case. So I
1828+implemented 'dirperm1' and 'nodirperm1' option in aufs1, and leave it to
1829+users.
1830+When dirperm1 is specified, aufs checks only the highest one for the
1831+directory permission, and shows the entries. Otherwise, as usual, checks
1832+every dir existing on all branches and rejects the request.
1833+
1834+As a side effect, dirperm1 option improves the performance of aufs
1835+because the number of permission check is reduced.
1836+
1837+
1838+Being Another Aufs's Readonly Branch (robr)
1839+----------------------------------------------------------------------
1840+Aufs1 allows aufs to be another aufs's readonly branch.
1841+This feature was developed by a user's request. But it may not be used
1842+currecnly.
1843+
1844+
1845+Copy-up on Open (coo=)
1846+----------------------------------------------------------------------
1847+By default the internal copy-up is executed when it is really necessary.
1848+It is not done when a file is opened for writing, but when write(2) is
1849+done. Users who have many (over 100) branches want to know and analyse
1850+when and what file is copied-up. To insert a new upper branch which
1851+contains such files only may improve the performance of aufs.
1852+
1853+Aufs1 implemented "coo=none | leaf | all" option.
1854+
1855+
1856+Refresh the Opened File (refrof)
1857+----------------------------------------------------------------------
1858+This option is implemented in aufs1 but incomplete.
1859+
1860+When user reads from a file, he expects to get its latest filedata
1861+generally. If the file is removed and a new same named file is created,
1862+the content he gets is unchanged, ie. the unlinked filedata.
1863+
1864+Let's try case study again.
1865+- aufs has two branches.
1866+ /au = /rw + /ro
1867+- "fileA" exists under /ro, but /rw.
1868+- user opened "/au/fileA".
1869+- he or someone else inserts a branch (/new) between /rw and /ro.
1870+ /au = /rw + /new + /ro
1871+- the new branch has "fileA".
1872+- user reads from the opened "fileA"
1873+- which filedata should aufs return, from /ro or /new?
1874+
1875+Some people says it has to be "from /ro" and it is a semantics of Unix.
1876+The others say it should be "from /new" because the file is not removed
1877+and it is equivalent to the case of someone else modifies the file.
1878+
1879+Here again I don't have a best and final answer. I got an idea to
1880+implement 'refrof' and 'norefrof' option. When 'refrof' (REFResh the
1881+Opened File) is specified (by default), aufs returns the filedata from
1882+/new.
1883+Otherwise from /new.
1884diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
1885--- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
1886+++ linux/Documentation/filesystems/aufs/README 2014-06-24 00:14:06.769385243 +0200
1887@@ -0,0 +1,368 @@
53392da6
AM
1888+
1889+Aufs3 -- advanced multi layered unification filesystem version 3.x
1890+http://aufs.sf.net
1891+Junjiro R. Okajima
1892+
1893+
1894+0. Introduction
1895+----------------------------------------
1896+In the early days, aufs was entirely re-designed and re-implemented
1897+Unionfs Version 1.x series. After many original ideas, approaches,
1898+improvements and implementations, it becomes totally different from
1899+Unionfs while keeping the basic features.
1900+Recently, Unionfs Version 2.x series begin taking some of the same
1901+approaches to aufs1's.
1902+Unionfs is being developed by Professor Erez Zadok at Stony Brook
1903+University and his team.
1904+
1905+Aufs3 supports linux-3.0 and later.
1906+If you want older kernel version support, try aufs2-2.6.git or
1907+aufs2-standalone.git repository, aufs1 from CVS on SourceForge.
1908+
1909+Note: it becomes clear that "Aufs was rejected. Let's give it up."
38d290e6
JR
1910+ According to Christoph Hellwig, linux rejects all union-type
1911+ filesystems but UnionMount.
53392da6
AM
1912+<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
1913+
38d290e6
JR
1914+PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
1915+ UnionMount, and he pointed out an issue around a directory mutex
1916+ lock and aufs addressed it. But it is still unsure whether aufs will
1917+ be merged (or any other union solution).
1918+<http://lkml.org/lkml/2013/3/12/637>
1919+
53392da6
AM
1920+
1921+1. Features
1922+----------------------------------------
1923+- unite several directories into a single virtual filesystem. The member
1924+ directory is called as a branch.
1925+- you can specify the permission flags to the branch, which are 'readonly',
1926+ 'readwrite' and 'whiteout-able.'
1927+- by upper writable branch, internal copyup and whiteout, files/dirs on
1928+ readonly branch are modifiable logically.
1929+- dynamic branch manipulation, add, del.
1930+- etc...
1931+
1932+Also there are many enhancements in aufs1, such as:
1933+- readdir(3) in userspace.
1934+- keep inode number by external inode number table
1935+- keep the timestamps of file/dir in internal copyup operation
1936+- seekable directory, supporting NFS readdir.
1937+- whiteout is hardlinked in order to reduce the consumption of inodes
1938+ on branch
1939+- do not copyup, nor create a whiteout when it is unnecessary
1940+- revert a single systemcall when an error occurs in aufs
1941+- remount interface instead of ioctl
1942+- maintain /etc/mtab by an external command, /sbin/mount.aufs.
1943+- loopback mounted filesystem as a branch
1944+- kernel thread for removing the dir who has a plenty of whiteouts
1945+- support copyup sparse file (a file which has a 'hole' in it)
1946+- default permission flags for branches
1947+- selectable permission flags for ro branch, whether whiteout can
1948+ exist or not
1949+- export via NFS.
1950+- support <sysfs>/fs/aufs and <debugfs>/aufs.
1951+- support multiple writable branches, some policies to select one
1952+ among multiple writable branches.
1953+- a new semantics for link(2) and rename(2) to support multiple
1954+ writable branches.
1955+- no glibc changes are required.
1956+- pseudo hardlink (hardlink over branches)
1957+- allow a direct access manually to a file on branch, e.g. bypassing aufs.
1958+ including NFS or remote filesystem branch.
1959+- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
1960+- and more...
1961+
1962+Currently these features are dropped temporary from aufs3.
1963+See design/08plan.txt in detail.
1964+- test only the highest one for the directory permission (dirperm1)
1965+- copyup on open (coo=)
1966+- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
1967+ (robr)
1968+- statistics of aufs thread (/sys/fs/aufs/stat)
1969+- delegation mode (dlgt)
1970+ a delegation of the internal branch access to support task I/O
1971+ accounting, which also supports Linux Security Modules (LSM) mainly
1972+ for Suse AppArmor.
1973+- intent.open/create (file open in a single lookup)
1974+
1975+Features or just an idea in the future (see also design/*.txt),
1976+- reorder the branch index without del/re-add.
1977+- permanent xino files for NFSD
1978+- an option for refreshing the opened files after add/del branches
1979+- 'move' policy for copy-up between two writable branches, after
1980+ checking free space.
1981+- light version, without branch manipulation. (unnecessary?)
1982+- copyup in userspace
1983+- inotify in userspace
1984+- readv/writev
1985+- xattr, acl
1986+
1987+
1988+2. Download
1989+----------------------------------------
1e00d052
AM
1990+There were three GIT trees for aufs3, aufs3-linux.git,
1991+aufs3-standalone.git, and aufs-util.git. Note that there is no "3" in
1992+"aufs-util.git."
1993+While the aufs-util is always necessary, you need either of aufs3-linux
1994+or aufs3-standalone.
1995+
1996+The aufs3-linux tree includes the whole linux mainline GIT tree,
1997+git://git.kernel.org/.../torvalds/linux.git.
1998+And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
b4510431 1999+build aufs3 as an external kernel module.
1e00d052
AM
2000+
2001+On the other hand, the aufs3-standalone tree has only aufs source files
53392da6
AM
2002+and necessary patches, and you can select CONFIG_AUFS_FS=m.
2003+
2004+You will find GIT branches whose name is in form of "aufs3.x" where "x"
2005+represents the linux kernel version, "linux-3.x". For instance,
1e00d052
AM
2006+"aufs3.0" is for linux-3.0. For latest "linux-3.x-rcN", use
2007+"aufs3.x-rcN" branch.
2008+
2009+o aufs3-linux tree
2010+$ git clone --reference /your/linux/git/tree \
86dc4139 2011+ git://git.code.sf.net/p/aufs/aufs3-linux aufs-aufs3-linux \
1e00d052
AM
2012+ aufs3-linux.git
2013+- if you don't have linux GIT tree, then remove "--reference ..."
2014+$ cd aufs3-linux.git
2015+$ git checkout origin/aufs3.0
53392da6
AM
2016+
2017+o aufs3-standalone tree
86dc4139 2018+$ git clone git://git.code.sf.net/p/aufs/aufs3-standalone \
53392da6
AM
2019+ aufs3-standalone.git
2020+$ cd aufs3-standalone.git
2021+$ git checkout origin/aufs3.0
2022+
2023+o aufs-util tree
86dc4139 2024+$ git clone git://git.code.sf.net/p/aufs/aufs-util \
53392da6
AM
2025+ aufs-util.git
2026+$ cd aufs-util.git
2027+$ git checkout origin/aufs3.0
2028+
9dbd164d
AM
2029+Note: The 3.x-rcN branch is to be used with `rc' kernel versions ONLY.
2030+The minor version number, 'x' in '3.x', of aufs may not always
2031+follow the minor version number of the kernel.
2032+Because changes in the kernel that cause the use of a new
2033+minor version number do not always require changes to aufs-util.
2034+
2035+Since aufs-util has its own minor version number, you may not be
2036+able to find a GIT branch in aufs-util for your kernel's
2037+exact minor version number.
2038+In this case, you should git-checkout the branch for the
53392da6 2039+nearest lower number.
9dbd164d
AM
2040+
2041+For (an unreleased) example:
2042+If you are using "linux-3.10" and the "aufs3.10" branch
7eafdf33 2043+does not exist in aufs-util repository, then "aufs3.9", "aufs3.8"
9dbd164d
AM
2044+or something numerically smaller is the branch for your kernel.
2045+
53392da6
AM
2046+Also you can view all branches by
2047+ $ git branch -a
2048+
2049+
2050+3. Configuration and Compilation
2051+----------------------------------------
2052+Make sure you have git-checkout'ed the correct branch.
2053+
1e00d052 2054+For aufs3-linux tree,
c06a8ce3 2055+- enable CONFIG_AUFS_FS.
1e00d052
AM
2056+- set other aufs configurations if necessary.
2057+
53392da6
AM
2058+For aufs3-standalone tree,
2059+There are several ways to build.
2060+
2061+1.
2062+- apply ./aufs3-kbuild.patch to your kernel source files.
2063+- apply ./aufs3-base.patch too.
523b37e3 2064+- apply ./aufs3-mmap.patch too.
53392da6
AM
2065+- apply ./aufs3-standalone.patch too, if you have a plan to set
2066+ CONFIG_AUFS_FS=m. otherwise you don't need ./aufs3-standalone.patch.
537831f9
AM
2067+- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2068+ kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
c06a8ce3 2069+- enable CONFIG_AUFS_FS, you can select either
53392da6
AM
2070+ =m or =y.
2071+- and build your kernel as usual.
2072+- install the built kernel.
c06a8ce3
AM
2073+ Note: Since linux-3.9, every filesystem module requires an alias
2074+ "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2075+ modules.aliases file if you set CONFIG_AUFS_FS=m.
7eafdf33
AM
2076+- install the header files too by "make headers_install" to the
2077+ directory where you specify. By default, it is $PWD/usr.
b4510431 2078+ "make help" shows a brief note for headers_install.
53392da6
AM
2079+- and reboot your system.
2080+
2081+2.
2082+- module only (CONFIG_AUFS_FS=m).
2083+- apply ./aufs3-base.patch to your kernel source files.
523b37e3 2084+- apply ./aufs3-mmap.patch too.
53392da6
AM
2085+- apply ./aufs3-standalone.patch too.
2086+- build your kernel, don't forget "make headers_install", and reboot.
2087+- edit ./config.mk and set other aufs configurations if necessary.
b4510431 2088+ Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
53392da6
AM
2089+ every aufs configurations.
2090+- build the module by simple "make".
c06a8ce3
AM
2091+ Note: Since linux-3.9, every filesystem module requires an alias
2092+ "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2093+ modules.aliases file.
53392da6
AM
2094+- you can specify ${KDIR} make variable which points to your kernel
2095+ source tree.
2096+- install the files
2097+ + run "make install" to install the aufs module, or copy the built
b4510431
AM
2098+ $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2099+ + run "make install_headers" (instead of headers_install) to install
2100+ the modified aufs header file (you can specify DESTDIR which is
2101+ available in aufs standalone version's Makefile only), or copy
2102+ $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2103+ you like manually. By default, the target directory is $PWD/usr.
53392da6
AM
2104+- no need to apply aufs3-kbuild.patch, nor copying source files to your
2105+ kernel source tree.
2106+
b4510431 2107+Note: The header file aufs_type.h is necessary to build aufs-util
53392da6
AM
2108+ as well as "make headers_install" in the kernel source tree.
2109+ headers_install is subject to be forgotten, but it is essentially
2110+ necessary, not only for building aufs-util.
2111+ You may not meet problems without headers_install in some older
2112+ version though.
2113+
2114+And then,
2115+- read README in aufs-util, build and install it
9dbd164d
AM
2116+- note that your distribution may contain an obsoleted version of
2117+ aufs_type.h in /usr/include/linux or something. When you build aufs
2118+ utilities, make sure that your compiler refers the correct aufs header
2119+ file which is built by "make headers_install."
53392da6
AM
2120+- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2121+ then run "make install_ulib" too. And refer to the aufs manual in
2122+ detail.
2123+
38d290e6
JR
2124+There several other patches in aufs3-standalone.git. They are all
2125+optional. When you meet some problems, they will help you.
2126+- aufs3-loopback.patch
2127+ Supports a nested loopback mount in a branch-fs. This patch is
2128+ unnecessary until aufs produces a message like "you may want to try
2129+ another patch for loopback file".
2130+- vfs-ino.patch
2131+ Modifies a system global kernel internal function get_next_ino() in
2132+ order to stop assigning 0 for an inode-number. Not directly related to
2133+ aufs, but recommended generally.
2134+- tmpfs-idr.patch
2135+ Keeps the tmpfs inode number as the lowest value. Effective to reduce
2136+ the size of aufs XINO files for tmpfs branch. Also it prevents the
2137+ duplication of inode number, which is important for backup tools and
2138+ other utilities. When you find aufs XINO files for tmpfs branch
2139+ growing too much, try this patch.
2140+
53392da6
AM
2141+
2142+4. Usage
2143+----------------------------------------
2144+At first, make sure aufs-util are installed, and please read the aufs
2145+manual, aufs.5 in aufs-util.git tree.
2146+$ man -l aufs.5
2147+
2148+And then,
2149+$ mkdir /tmp/rw /tmp/aufs
2150+# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2151+
2152+Here is another example. The result is equivalent.
2153+# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2154+ Or
2155+# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2156+# mount -o remount,append:${HOME} /tmp/aufs
2157+
2158+Then, you can see whole tree of your home dir through /tmp/aufs. If
2159+you modify a file under /tmp/aufs, the one on your home directory is
2160+not affected, instead the same named file will be newly created under
2161+/tmp/rw. And all of your modification to a file will be applied to
2162+the one under /tmp/rw. This is called the file based Copy on Write
2163+(COW) method.
2164+Aufs mount options are described in aufs.5.
2165+If you run chroot or something and make your aufs as a root directory,
2166+then you need to customize the shutdown script. See the aufs manual in
2167+detail.
2168+
2169+Additionally, there are some sample usages of aufs which are a
2170+diskless system with network booting, and LiveCD over NFS.
2171+See sample dir in CVS tree on SourceForge.
2172+
2173+
2174+5. Contact
2175+----------------------------------------
2176+When you have any problems or strange behaviour in aufs, please let me
2177+know with:
2178+- /proc/mounts (instead of the output of mount(8))
2179+- /sys/module/aufs/*
2180+- /sys/fs/aufs/* (if you have them)
2181+- /debug/aufs/* (if you have them)
2182+- linux kernel version
2183+ if your kernel is not plain, for example modified by distributor,
2184+ the url where i can download its source is necessary too.
2185+- aufs version which was printed at loading the module or booting the
2186+ system, instead of the date you downloaded.
2187+- configuration (define/undefine CONFIG_AUFS_xxx)
2188+- kernel configuration or /proc/config.gz (if you have it)
2189+- behaviour which you think to be incorrect
2190+- actual operation, reproducible one is better
2191+- mailto: aufs-users at lists.sourceforge.net
2192+
2193+Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2194+and Feature Requests) on SourceForge. Please join and write to
2195+aufs-users ML.
2196+
2197+
2198+6. Acknowledgements
2199+----------------------------------------
2200+Thanks to everyone who have tried and are using aufs, whoever
2201+have reported a bug or any feedback.
2202+
2203+Especially donators:
2204+Tomas Matejicek(slax.org) made a donation (much more than once).
2205+ Since Apr 2010, Tomas M (the author of Slax and Linux Live
2206+ scripts) is making "doubling" donations.
2207+ Unfortunately I cannot list all of the donators, but I really
b4510431 2208+ appreciate.
53392da6
AM
2209+ It ends Aug 2010, but the ordinary donation URL is still available.
2210+ <http://sourceforge.net/donate/index.php?group_id=167503>
2211+Dai Itasaka made a donation (2007/8).
2212+Chuck Smith made a donation (2008/4, 10 and 12).
2213+Henk Schoneveld made a donation (2008/9).
2214+Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2215+Francois Dupoux made a donation (2008/11).
2216+Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2217+ aufs2 GIT tree (2009/2).
2218+William Grant made a donation (2009/3).
2219+Patrick Lane made a donation (2009/4).
2220+The Mail Archive (mail-archive.com) made donations (2009/5).
2221+Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2222+New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2223+Pavel Pronskiy made a donation (2011/2).
2224+Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2225+ Networks (Ed Wildgoose) made a donation for hardware (2011/3).
537831f9
AM
2226+Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2227+11).
1e00d052 2228+Sam Liddicott made a donation (2011/9).
86dc4139
AM
2229+Era Scarecrow made a donation (2013/4).
2230+Bor Ratajc made a donation (2013/4).
2231+Alessandro Gorreta made a donation (2013/4).
2232+POIRETTE Marc made a donation (2013/4).
2233+Alessandro Gorreta made a donation (2013/4).
2234+lauri kasvandik made a donation (2013/5).
392086de 2235+"pemasu from Finland" made a donation (2013/7).
523b37e3
AM
2236+The Parted Magic Project made a donation (2013/9 and 11).
2237+Pavel Barta made a donation (2013/10).
38d290e6 2238+Nikolay Pertsev made a donation (2014/5).
53392da6
AM
2239+
2240+Thank you very much.
2241+Donations are always, including future donations, very important and
2242+helpful for me to keep on developing aufs.
2243+
2244+
2245+7.
2246+----------------------------------------
2247+If you are an experienced user, no explanation is needed. Aufs is
2248+just a linux filesystem.
2249+
2250+
2251+Enjoy!
2252+
2253+# Local variables: ;
2254+# mode: text;
2255+# End: ;
7f207e10
AM
2256diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2257--- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 2258+++ linux/fs/aufs/aufs.h 2014-06-24 00:14:06.776052275 +0200
523b37e3 2259@@ -0,0 +1,59 @@
7f207e10 2260+/*
523b37e3 2261+ * Copyright (C) 2005-2014 Junjiro R. Okajima
7f207e10
AM
2262+ *
2263+ * This program, aufs is free software; you can redistribute it and/or modify
2264+ * it under the terms of the GNU General Public License as published by
2265+ * the Free Software Foundation; either version 2 of the License, or
2266+ * (at your option) any later version.
2267+ *
2268+ * This program is distributed in the hope that it will be useful,
2269+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
2270+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2271+ * GNU General Public License for more details.
2272+ *
2273+ * You should have received a copy of the GNU General Public License
523b37e3 2274+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
7f207e10
AM
2275+ */
2276+
2277+/*
2278+ * all header files
2279+ */
2280+
2281+#ifndef __AUFS_H__
2282+#define __AUFS_H__
2283+
2284+#ifdef __KERNEL__
2285+
2286+#define AuStub(type, name, body, ...) \
2287+ static inline type name(__VA_ARGS__) { body; }
2288+
2289+#define AuStubVoid(name, ...) \
2290+ AuStub(void, name, , __VA_ARGS__)
2291+#define AuStubInt0(name, ...) \
2292+ AuStub(int, name, return 0, __VA_ARGS__)
2293+
2294+#include "debug.h"
2295+
2296+#include "branch.h"
2297+#include "cpup.h"
2298+#include "dcsub.h"
2299+#include "dbgaufs.h"
2300+#include "dentry.h"
2301+#include "dir.h"
2302+#include "dynop.h"
2303+#include "file.h"
2304+#include "fstype.h"
2305+#include "inode.h"
2306+#include "loop.h"
2307+#include "module.h"
7f207e10
AM
2308+#include "opts.h"
2309+#include "rwsem.h"
2310+#include "spl.h"
2311+#include "super.h"
2312+#include "sysaufs.h"
2313+#include "vfsub.h"
2314+#include "whout.h"
2315+#include "wkq.h"
2316+
2317+#endif /* __KERNEL__ */
2318+#endif /* __AUFS_H__ */
2319diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2320--- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
2321+++ linux/fs/aufs/branch.c 2014-06-24 00:14:06.776052275 +0200
2322@@ -0,0 +1,1220 @@
7f207e10 2323+/*
523b37e3 2324+ * Copyright (C) 2005-2014 Junjiro R. Okajima
7f207e10
AM
2325+ *
2326+ * This program, aufs is free software; you can redistribute it and/or modify
2327+ * it under the terms of the GNU General Public License as published by
2328+ * the Free Software Foundation; either version 2 of the License, or
2329+ * (at your option) any later version.
2330+ *
2331+ * This program is distributed in the hope that it will be useful,
2332+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
2333+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2334+ * GNU General Public License for more details.
2335+ *
2336+ * You should have received a copy of the GNU General Public License
523b37e3 2337+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
7f207e10
AM
2338+ */
2339+
2340+/*
2341+ * branch management
2342+ */
2343+
027c5e7a 2344+#include <linux/compat.h>
7f207e10
AM
2345+#include <linux/statfs.h>
2346+#include "aufs.h"
2347+
2348+/*
2349+ * free a single branch
1facf9fc 2350+ */
86dc4139
AM
2351+
2352+/* prohibit rmdir to the root of the branch */
2353+/* todo: another new flag? */
2354+static void au_br_dflags_force(struct au_branch *br)
2355+{
2356+ struct dentry *h_dentry;
2357+
2358+ h_dentry = au_br_dentry(br);
2359+ spin_lock(&h_dentry->d_lock);
2360+ br->br_dflags = h_dentry->d_flags & DCACHE_MOUNTED;
2361+ h_dentry->d_flags |= DCACHE_MOUNTED;
2362+ spin_unlock(&h_dentry->d_lock);
2363+}
2364+
2365+/* restore its d_flags */
2366+static void au_br_dflags_restore(struct au_branch *br)
2367+{
2368+ struct dentry *h_dentry;
2369+
2370+ if (br->br_dflags)
2371+ return;
2372+
2373+ h_dentry = au_br_dentry(br);
2374+ spin_lock(&h_dentry->d_lock);
2375+ h_dentry->d_flags &= ~DCACHE_MOUNTED;
2376+ spin_unlock(&h_dentry->d_lock);
2377+}
2378+
1facf9fc 2379+static void au_br_do_free(struct au_branch *br)
2380+{
2381+ int i;
2382+ struct au_wbr *wbr;
4a4d8108 2383+ struct au_dykey **key;
1facf9fc 2384+
027c5e7a
AM
2385+ au_hnotify_fin_br(br);
2386+
1facf9fc 2387+ if (br->br_xino.xi_file)
2388+ fput(br->br_xino.xi_file);
2389+ mutex_destroy(&br->br_xino.xi_nondir_mtx);
2390+
2391+ AuDebugOn(atomic_read(&br->br_count));
2392+
2393+ wbr = br->br_wbr;
2394+ if (wbr) {
2395+ for (i = 0; i < AuBrWh_Last; i++)
2396+ dput(wbr->wbr_wh[i]);
2397+ AuDebugOn(atomic_read(&wbr->wbr_wh_running));
dece6358 2398+ AuRwDestroy(&wbr->wbr_wh_rwsem);
1facf9fc 2399+ }
2400+
4a4d8108
AM
2401+ key = br->br_dykey;
2402+ for (i = 0; i < AuBrDynOp; i++, key++)
2403+ if (*key)
2404+ au_dy_put(*key);
2405+ else
2406+ break;
2407+
86dc4139
AM
2408+ au_br_dflags_restore(br);
2409+
537831f9
AM
2410+ /* recursive lock, s_umount of branch's */
2411+ lockdep_off();
86dc4139 2412+ path_put(&br->br_path);
537831f9 2413+ lockdep_on();
1facf9fc 2414+ kfree(wbr);
2415+ kfree(br);
2416+}
2417+
2418+/*
2419+ * frees all branches
2420+ */
2421+void au_br_free(struct au_sbinfo *sbinfo)
2422+{
2423+ aufs_bindex_t bmax;
2424+ struct au_branch **br;
2425+
dece6358
AM
2426+ AuRwMustWriteLock(&sbinfo->si_rwsem);
2427+
1facf9fc 2428+ bmax = sbinfo->si_bend + 1;
2429+ br = sbinfo->si_branch;
2430+ while (bmax--)
2431+ au_br_do_free(*br++);
2432+}
2433+
2434+/*
2435+ * find the index of a branch which is specified by @br_id.
2436+ */
2437+int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2438+{
2439+ aufs_bindex_t bindex, bend;
2440+
2441+ bend = au_sbend(sb);
2442+ for (bindex = 0; bindex <= bend; bindex++)
2443+ if (au_sbr_id(sb, bindex) == br_id)
2444+ return bindex;
2445+ return -1;
2446+}
2447+
2448+/* ---------------------------------------------------------------------- */
2449+
2450+/*
2451+ * add a branch
2452+ */
2453+
b752ccd1
AM
2454+static int test_overlap(struct super_block *sb, struct dentry *h_adding,
2455+ struct dentry *h_root)
1facf9fc 2456+{
b752ccd1
AM
2457+ if (unlikely(h_adding == h_root
2458+ || au_test_loopback_overlap(sb, h_adding)))
1facf9fc 2459+ return 1;
b752ccd1
AM
2460+ if (h_adding->d_sb != h_root->d_sb)
2461+ return 0;
2462+ return au_test_subdir(h_adding, h_root)
2463+ || au_test_subdir(h_root, h_adding);
1facf9fc 2464+}
2465+
2466+/*
2467+ * returns a newly allocated branch. @new_nbranch is a number of branches
2468+ * after adding a branch.
2469+ */
2470+static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
2471+ int perm)
2472+{
2473+ struct au_branch *add_branch;
2474+ struct dentry *root;
4a4d8108 2475+ int err;
1facf9fc 2476+
4a4d8108 2477+ err = -ENOMEM;
1facf9fc 2478+ root = sb->s_root;
2479+ add_branch = kmalloc(sizeof(*add_branch), GFP_NOFS);
2480+ if (unlikely(!add_branch))
2481+ goto out;
2482+
027c5e7a
AM
2483+ err = au_hnotify_init_br(add_branch, perm);
2484+ if (unlikely(err))
2485+ goto out_br;
2486+
1facf9fc 2487+ add_branch->br_wbr = NULL;
2488+ if (au_br_writable(perm)) {
2489+ /* may be freed separately at changing the branch permission */
2490+ add_branch->br_wbr = kmalloc(sizeof(*add_branch->br_wbr),
2491+ GFP_NOFS);
2492+ if (unlikely(!add_branch->br_wbr))
027c5e7a 2493+ goto out_hnotify;
1facf9fc 2494+ }
2495+
4a4d8108
AM
2496+ err = au_sbr_realloc(au_sbi(sb), new_nbranch);
2497+ if (!err)
2498+ err = au_di_realloc(au_di(root), new_nbranch);
2499+ if (!err)
2500+ err = au_ii_realloc(au_ii(root->d_inode), new_nbranch);
2501+ if (!err)
2502+ return add_branch; /* success */
1facf9fc 2503+
1facf9fc 2504+ kfree(add_branch->br_wbr);
4a4d8108 2505+
027c5e7a
AM
2506+out_hnotify:
2507+ au_hnotify_fin_br(add_branch);
4f0767ce 2508+out_br:
1facf9fc 2509+ kfree(add_branch);
4f0767ce 2510+out:
4a4d8108 2511+ return ERR_PTR(err);
1facf9fc 2512+}
2513+
2514+/*
2515+ * test if the branch permission is legal or not.
2516+ */
2517+static int test_br(struct inode *inode, int brperm, char *path)
2518+{
2519+ int err;
2520+
4a4d8108
AM
2521+ err = (au_br_writable(brperm) && IS_RDONLY(inode));
2522+ if (!err)
2523+ goto out;
1facf9fc 2524+
4a4d8108
AM
2525+ err = -EINVAL;
2526+ pr_err("write permission for readonly mount or inode, %s\n", path);
2527+
4f0767ce 2528+out:
1facf9fc 2529+ return err;
2530+}
2531+
2532+/*
2533+ * returns:
2534+ * 0: success, the caller will add it
2535+ * plus: success, it is already unified, the caller should ignore it
2536+ * minus: error
2537+ */
2538+static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
2539+{
2540+ int err;
2541+ aufs_bindex_t bend, bindex;
2542+ struct dentry *root;
2543+ struct inode *inode, *h_inode;
2544+
2545+ root = sb->s_root;
2546+ bend = au_sbend(sb);
2547+ if (unlikely(bend >= 0
2548+ && au_find_dbindex(root, add->path.dentry) >= 0)) {
2549+ err = 1;
2550+ if (!remount) {
2551+ err = -EINVAL;
4a4d8108 2552+ pr_err("%s duplicated\n", add->pathname);
1facf9fc 2553+ }
2554+ goto out;
2555+ }
2556+
2557+ err = -ENOSPC; /* -E2BIG; */
2558+ if (unlikely(AUFS_BRANCH_MAX <= add->bindex
2559+ || AUFS_BRANCH_MAX - 1 <= bend)) {
4a4d8108 2560+ pr_err("number of branches exceeded %s\n", add->pathname);
1facf9fc 2561+ goto out;
2562+ }
2563+
2564+ err = -EDOM;
2565+ if (unlikely(add->bindex < 0 || bend + 1 < add->bindex)) {
4a4d8108 2566+ pr_err("bad index %d\n", add->bindex);
1facf9fc 2567+ goto out;
2568+ }
2569+
2570+ inode = add->path.dentry->d_inode;
2571+ err = -ENOENT;
2572+ if (unlikely(!inode->i_nlink)) {
4a4d8108 2573+ pr_err("no existence %s\n", add->pathname);
1facf9fc 2574+ goto out;
2575+ }
2576+
2577+ err = -EINVAL;
2578+ if (unlikely(inode->i_sb == sb)) {
4a4d8108 2579+ pr_err("%s must be outside\n", add->pathname);
1facf9fc 2580+ goto out;
2581+ }
2582+
2583+ if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
4a4d8108
AM
2584+ pr_err("unsupported filesystem, %s (%s)\n",
2585+ add->pathname, au_sbtype(inode->i_sb));
1facf9fc 2586+ goto out;
2587+ }
2588+
2589+ err = test_br(add->path.dentry->d_inode, add->perm, add->pathname);
2590+ if (unlikely(err))
2591+ goto out;
2592+
2593+ if (bend < 0)
2594+ return 0; /* success */
2595+
2596+ err = -EINVAL;
2597+ for (bindex = 0; bindex <= bend; bindex++)
2598+ if (unlikely(test_overlap(sb, add->path.dentry,
2599+ au_h_dptr(root, bindex)))) {
4a4d8108 2600+ pr_err("%s is overlapped\n", add->pathname);
1facf9fc 2601+ goto out;
2602+ }
2603+
2604+ err = 0;
2605+ if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
2606+ h_inode = au_h_dptr(root, 0)->d_inode;
2607+ if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
0c3ec466
AM
2608+ || !uid_eq(h_inode->i_uid, inode->i_uid)
2609+ || !gid_eq(h_inode->i_gid, inode->i_gid))
2610+ pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
2611+ add->pathname,
2612+ i_uid_read(inode), i_gid_read(inode),
2613+ (inode->i_mode & S_IALLUGO),
2614+ i_uid_read(h_inode), i_gid_read(h_inode),
2615+ (h_inode->i_mode & S_IALLUGO));
1facf9fc 2616+ }
2617+
4f0767ce 2618+out:
1facf9fc 2619+ return err;
2620+}
2621+
2622+/*
2623+ * initialize or clean the whiteouts for an adding branch
2624+ */
2625+static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
86dc4139 2626+ int new_perm)
1facf9fc 2627+{
2628+ int err, old_perm;
2629+ aufs_bindex_t bindex;
2630+ struct mutex *h_mtx;
2631+ struct au_wbr *wbr;
2632+ struct au_hinode *hdir;
2633+
86dc4139
AM
2634+ err = vfsub_mnt_want_write(au_br_mnt(br));
2635+ if (unlikely(err))
2636+ goto out;
2637+
1facf9fc 2638+ wbr = br->br_wbr;
2639+ old_perm = br->br_perm;
2640+ br->br_perm = new_perm;
2641+ hdir = NULL;
2642+ h_mtx = NULL;
2643+ bindex = au_br_index(sb, br->br_id);
2644+ if (0 <= bindex) {
2645+ hdir = au_hi(sb->s_root->d_inode, bindex);
4a4d8108 2646+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
1facf9fc 2647+ } else {
86dc4139 2648+ h_mtx = &au_br_dentry(br)->d_inode->i_mutex;
1facf9fc 2649+ mutex_lock_nested(h_mtx, AuLsc_I_PARENT);
2650+ }
2651+ if (!wbr)
86dc4139 2652+ err = au_wh_init(br, sb);
1facf9fc 2653+ else {
2654+ wbr_wh_write_lock(wbr);
86dc4139 2655+ err = au_wh_init(br, sb);
1facf9fc 2656+ wbr_wh_write_unlock(wbr);
2657+ }
2658+ if (hdir)
4a4d8108 2659+ au_hn_imtx_unlock(hdir);
1facf9fc 2660+ else
2661+ mutex_unlock(h_mtx);
86dc4139 2662+ vfsub_mnt_drop_write(au_br_mnt(br));
1facf9fc 2663+ br->br_perm = old_perm;
2664+
2665+ if (!err && wbr && !au_br_writable(new_perm)) {
2666+ kfree(wbr);
2667+ br->br_wbr = NULL;
2668+ }
2669+
86dc4139 2670+out:
1facf9fc 2671+ return err;
2672+}
2673+
2674+static int au_wbr_init(struct au_branch *br, struct super_block *sb,
86dc4139 2675+ int perm)
1facf9fc 2676+{
2677+ int err;
4a4d8108 2678+ struct kstatfs kst;
1facf9fc 2679+ struct au_wbr *wbr;
2680+
2681+ wbr = br->br_wbr;
dece6358 2682+ au_rw_init(&wbr->wbr_wh_rwsem);
1facf9fc 2683+ memset(wbr->wbr_wh, 0, sizeof(wbr->wbr_wh));
2684+ atomic_set(&wbr->wbr_wh_running, 0);
2685+ wbr->wbr_bytes = 0;
2686+
4a4d8108
AM
2687+ /*
2688+ * a limit for rmdir/rename a dir
523b37e3 2689+ * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
4a4d8108 2690+ */
86dc4139 2691+ err = vfs_statfs(&br->br_path, &kst);
4a4d8108
AM
2692+ if (unlikely(err))
2693+ goto out;
2694+ err = -EINVAL;
2695+ if (kst.f_namelen >= NAME_MAX)
86dc4139 2696+ err = au_br_init_wh(sb, br, perm);
4a4d8108 2697+ else
523b37e3
AM
2698+ pr_err("%pd(%s), unsupported namelen %ld\n",
2699+ au_br_dentry(br),
86dc4139 2700+ au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
1facf9fc 2701+
4f0767ce 2702+out:
1facf9fc 2703+ return err;
2704+}
2705+
2706+/* intialize a new branch */
2707+static int au_br_init(struct au_branch *br, struct super_block *sb,
2708+ struct au_opt_add *add)
2709+{
2710+ int err;
2711+
2712+ err = 0;
2713+ memset(&br->br_xino, 0, sizeof(br->br_xino));
2714+ mutex_init(&br->br_xino.xi_nondir_mtx);
2715+ br->br_perm = add->perm;
86dc4139
AM
2716+ BUILD_BUG_ON(sizeof(br->br_dflags)
2717+ != sizeof(br->br_path.dentry->d_flags));
2718+ br->br_dflags = DCACHE_MOUNTED;
2719+ br->br_path = add->path; /* set first, path_get() later */
4a4d8108
AM
2720+ spin_lock_init(&br->br_dykey_lock);
2721+ memset(br->br_dykey, 0, sizeof(br->br_dykey));
1facf9fc 2722+ atomic_set(&br->br_count, 0);
1facf9fc 2723+ atomic_set(&br->br_xino_running, 0);
2724+ br->br_id = au_new_br_id(sb);
7f207e10 2725+ AuDebugOn(br->br_id < 0);
1facf9fc 2726+
2727+ if (au_br_writable(add->perm)) {
86dc4139 2728+ err = au_wbr_init(br, sb, add->perm);
1facf9fc 2729+ if (unlikely(err))
b752ccd1 2730+ goto out_err;
1facf9fc 2731+ }
2732+
2733+ if (au_opt_test(au_mntflags(sb), XINO)) {
2734+ err = au_xino_br(sb, br, add->path.dentry->d_inode->i_ino,
2735+ au_sbr(sb, 0)->br_xino.xi_file, /*do_test*/1);
2736+ if (unlikely(err)) {
2737+ AuDebugOn(br->br_xino.xi_file);
b752ccd1 2738+ goto out_err;
1facf9fc 2739+ }
2740+ }
2741+
2742+ sysaufs_br_init(br);
86dc4139 2743+ path_get(&br->br_path);
b752ccd1 2744+ goto out; /* success */
1facf9fc 2745+
4f0767ce 2746+out_err:
86dc4139 2747+ memset(&br->br_path, 0, sizeof(br->br_path));
4f0767ce 2748+out:
1facf9fc 2749+ return err;
2750+}
2751+
2752+static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
2753+ struct au_branch *br, aufs_bindex_t bend,
2754+ aufs_bindex_t amount)
2755+{
2756+ struct au_branch **brp;
2757+
dece6358
AM
2758+ AuRwMustWriteLock(&sbinfo->si_rwsem);
2759+
1facf9fc 2760+ brp = sbinfo->si_branch + bindex;
2761+ memmove(brp + 1, brp, sizeof(*brp) * amount);
2762+ *brp = br;
2763+ sbinfo->si_bend++;
2764+ if (unlikely(bend < 0))
2765+ sbinfo->si_bend = 0;
2766+}
2767+
2768+static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
2769+ aufs_bindex_t bend, aufs_bindex_t amount)
2770+{
2771+ struct au_hdentry *hdp;
2772+
1308ab2a 2773+ AuRwMustWriteLock(&dinfo->di_rwsem);
2774+
1facf9fc 2775+ hdp = dinfo->di_hdentry + bindex;
2776+ memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
2777+ au_h_dentry_init(hdp);
2778+ dinfo->di_bend++;
2779+ if (unlikely(bend < 0))
2780+ dinfo->di_bstart = 0;
2781+}
2782+
2783+static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
2784+ aufs_bindex_t bend, aufs_bindex_t amount)
2785+{
2786+ struct au_hinode *hip;
2787+
1308ab2a 2788+ AuRwMustWriteLock(&iinfo->ii_rwsem);
2789+
1facf9fc 2790+ hip = iinfo->ii_hinode + bindex;
2791+ memmove(hip + 1, hip, sizeof(*hip) * amount);
2792+ hip->hi_inode = NULL;
4a4d8108 2793+ au_hn_init(hip);
1facf9fc 2794+ iinfo->ii_bend++;
2795+ if (unlikely(bend < 0))
2796+ iinfo->ii_bstart = 0;
2797+}
2798+
86dc4139
AM
2799+static void au_br_do_add(struct super_block *sb, struct au_branch *br,
2800+ aufs_bindex_t bindex)
1facf9fc 2801+{
86dc4139 2802+ struct dentry *root, *h_dentry;
1facf9fc 2803+ struct inode *root_inode;
2804+ aufs_bindex_t bend, amount;
2805+
86dc4139
AM
2806+ au_br_dflags_force(br);
2807+
1facf9fc 2808+ root = sb->s_root;
2809+ root_inode = root->d_inode;
1facf9fc 2810+ bend = au_sbend(sb);
2811+ amount = bend + 1 - bindex;
86dc4139 2812+ h_dentry = au_br_dentry(br);
53392da6 2813+ au_sbilist_lock();
1facf9fc 2814+ au_br_do_add_brp(au_sbi(sb), bindex, br, bend, amount);
2815+ au_br_do_add_hdp(au_di(root), bindex, bend, amount);
2816+ au_br_do_add_hip(au_ii(root_inode), bindex, bend, amount);
2817+ au_set_h_dptr(root, bindex, dget(h_dentry));
2818+ au_set_h_iptr(root_inode, bindex, au_igrab(h_dentry->d_inode),
2819+ /*flags*/0);
53392da6 2820+ au_sbilist_unlock();
1facf9fc 2821+}
2822+
2823+int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
2824+{
2825+ int err;
1facf9fc 2826+ aufs_bindex_t bend, add_bindex;
2827+ struct dentry *root, *h_dentry;
2828+ struct inode *root_inode;
2829+ struct au_branch *add_branch;
2830+
2831+ root = sb->s_root;
2832+ root_inode = root->d_inode;
2833+ IMustLock(root_inode);
2834+ err = test_add(sb, add, remount);
2835+ if (unlikely(err < 0))
2836+ goto out;
2837+ if (err) {
2838+ err = 0;
2839+ goto out; /* success */
2840+ }
2841+
2842+ bend = au_sbend(sb);
2843+ add_branch = au_br_alloc(sb, bend + 2, add->perm);
2844+ err = PTR_ERR(add_branch);
2845+ if (IS_ERR(add_branch))
2846+ goto out;
2847+
2848+ err = au_br_init(add_branch, sb, add);
2849+ if (unlikely(err)) {
2850+ au_br_do_free(add_branch);
2851+ goto out;
2852+ }
2853+
2854+ add_bindex = add->bindex;
1facf9fc 2855+ if (!remount)
86dc4139 2856+ au_br_do_add(sb, add_branch, add_bindex);
1facf9fc 2857+ else {
2858+ sysaufs_brs_del(sb, add_bindex);
86dc4139 2859+ au_br_do_add(sb, add_branch, add_bindex);
1facf9fc 2860+ sysaufs_brs_add(sb, add_bindex);
2861+ }
2862+
86dc4139 2863+ h_dentry = add->path.dentry;
1308ab2a 2864+ if (!add_bindex) {
1facf9fc 2865+ au_cpup_attr_all(root_inode, /*force*/1);
1308ab2a 2866+ sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
2867+ } else
1facf9fc 2868+ au_add_nlink(root_inode, h_dentry->d_inode);
1facf9fc 2869+
2870+ /*
4a4d8108 2871+ * this test/set prevents aufs from handling unnecesary notify events
027c5e7a 2872+ * of xino files, in case of re-adding a writable branch which was
1facf9fc 2873+ * once detached from aufs.
2874+ */
2875+ if (au_xino_brid(sb) < 0
2876+ && au_br_writable(add_branch->br_perm)
2877+ && !au_test_fs_bad_xino(h_dentry->d_sb)
2878+ && add_branch->br_xino.xi_file
2879+ && add_branch->br_xino.xi_file->f_dentry->d_parent == h_dentry)
2880+ au_xino_brid_set(sb, add_branch->br_id);
2881+
4f0767ce 2882+out:
1facf9fc 2883+ return err;
2884+}
2885+
2886+/* ---------------------------------------------------------------------- */
2887+
2888+/*
2889+ * delete a branch
2890+ */
2891+
2892+/* to show the line number, do not make it inlined function */
4a4d8108 2893+#define AuVerbose(do_info, fmt, ...) do { \
1facf9fc 2894+ if (do_info) \
4a4d8108 2895+ pr_info(fmt, ##__VA_ARGS__); \
1facf9fc 2896+} while (0)
2897+
027c5e7a
AM
2898+static int au_test_ibusy(struct inode *inode, aufs_bindex_t bstart,
2899+ aufs_bindex_t bend)
2900+{
2901+ return (inode && !S_ISDIR(inode->i_mode)) || bstart == bend;
2902+}
2903+
2904+static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t bstart,
2905+ aufs_bindex_t bend)
2906+{
2907+ return au_test_ibusy(dentry->d_inode, bstart, bend);
2908+}
2909+
1facf9fc 2910+/*
2911+ * test if the branch is deletable or not.
2912+ */
2913+static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
b752ccd1 2914+ unsigned int sigen, const unsigned int verbose)
1facf9fc 2915+{
2916+ int err, i, j, ndentry;
2917+ aufs_bindex_t bstart, bend;
1facf9fc 2918+ struct au_dcsub_pages dpages;
2919+ struct au_dpage *dpage;
2920+ struct dentry *d;
1facf9fc 2921+
2922+ err = au_dpages_init(&dpages, GFP_NOFS);
2923+ if (unlikely(err))
2924+ goto out;
2925+ err = au_dcsub_pages(&dpages, root, NULL, NULL);
2926+ if (unlikely(err))
2927+ goto out_dpages;
2928+
1facf9fc 2929+ for (i = 0; !err && i < dpages.ndpage; i++) {
2930+ dpage = dpages.dpages + i;
2931+ ndentry = dpage->ndentry;
2932+ for (j = 0; !err && j < ndentry; j++) {
2933+ d = dpage->dentries[j];
392086de 2934+ AuDebugOn(!d_count(d));
027c5e7a 2935+ if (!au_digen_test(d, sigen)) {
1facf9fc 2936+ di_read_lock_child(d, AuLock_IR);
027c5e7a
AM
2937+ if (unlikely(au_dbrange_test(d))) {
2938+ di_read_unlock(d, AuLock_IR);
2939+ continue;
2940+ }
2941+ } else {
1facf9fc 2942+ di_write_lock_child(d);
027c5e7a
AM
2943+ if (unlikely(au_dbrange_test(d))) {
2944+ di_write_unlock(d);
2945+ continue;
2946+ }
1facf9fc 2947+ err = au_reval_dpath(d, sigen);
2948+ if (!err)
2949+ di_downgrade_lock(d, AuLock_IR);
2950+ else {
2951+ di_write_unlock(d);
2952+ break;
2953+ }
2954+ }
2955+
027c5e7a 2956+ /* AuDbgDentry(d); */
1facf9fc 2957+ bstart = au_dbstart(d);
2958+ bend = au_dbend(d);
2959+ if (bstart <= bindex
2960+ && bindex <= bend
2961+ && au_h_dptr(d, bindex)
027c5e7a 2962+ && au_test_dbusy(d, bstart, bend)) {
1facf9fc 2963+ err = -EBUSY;
523b37e3 2964+ AuVerbose(verbose, "busy %pd\n", d);
027c5e7a 2965+ AuDbgDentry(d);
1facf9fc 2966+ }
2967+ di_read_unlock(d, AuLock_IR);
2968+ }
2969+ }
2970+
4f0767ce 2971+out_dpages:
1facf9fc 2972+ au_dpages_free(&dpages);
4f0767ce 2973+out:
1facf9fc 2974+ return err;
2975+}
2976+
2977+static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
b752ccd1 2978+ unsigned int sigen, const unsigned int verbose)
1facf9fc 2979+{
2980+ int err;
7f207e10
AM
2981+ unsigned long long max, ull;
2982+ struct inode *i, **array;
1facf9fc 2983+ aufs_bindex_t bstart, bend;
1facf9fc 2984+
7f207e10
AM
2985+ array = au_iarray_alloc(sb, &max);
2986+ err = PTR_ERR(array);
2987+ if (IS_ERR(array))
2988+ goto out;
2989+
1facf9fc 2990+ err = 0;
7f207e10
AM
2991+ AuDbg("b%d\n", bindex);
2992+ for (ull = 0; !err && ull < max; ull++) {
2993+ i = array[ull];
2994+ if (i->i_ino == AUFS_ROOT_INO)
1facf9fc 2995+ continue;
2996+
7f207e10 2997+ /* AuDbgInode(i); */
537831f9 2998+ if (au_iigen(i, NULL) == sigen)
1facf9fc 2999+ ii_read_lock_child(i);
3000+ else {
3001+ ii_write_lock_child(i);
027c5e7a
AM
3002+ err = au_refresh_hinode_self(i);
3003+ au_iigen_dec(i);
1facf9fc 3004+ if (!err)
3005+ ii_downgrade_lock(i);
3006+ else {
3007+ ii_write_unlock(i);
3008+ break;
3009+ }
3010+ }
3011+
3012+ bstart = au_ibstart(i);
3013+ bend = au_ibend(i);
3014+ if (bstart <= bindex
3015+ && bindex <= bend
3016+ && au_h_iptr(i, bindex)
027c5e7a 3017+ && au_test_ibusy(i, bstart, bend)) {
1facf9fc 3018+ err = -EBUSY;
3019+ AuVerbose(verbose, "busy i%lu\n", i->i_ino);
7f207e10 3020+ AuDbgInode(i);
1facf9fc 3021+ }
3022+ ii_read_unlock(i);
3023+ }
7f207e10 3024+ au_iarray_free(array, max);
1facf9fc 3025+
7f207e10 3026+out:
1facf9fc 3027+ return err;
3028+}
3029+
b752ccd1
AM
3030+static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3031+ const unsigned int verbose)
1facf9fc 3032+{
3033+ int err;
3034+ unsigned int sigen;
3035+
3036+ sigen = au_sigen(root->d_sb);
3037+ DiMustNoWaiters(root);
3038+ IiMustNoWaiters(root->d_inode);
3039+ di_write_unlock(root);
b752ccd1 3040+ err = test_dentry_busy(root, bindex, sigen, verbose);
1facf9fc 3041+ if (!err)
b752ccd1 3042+ err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
1facf9fc 3043+ di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3044+
3045+ return err;
3046+}
3047+
3048+static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3049+ const aufs_bindex_t bindex,
3050+ const aufs_bindex_t bend)
3051+{
3052+ struct au_branch **brp, **p;
3053+
dece6358
AM
3054+ AuRwMustWriteLock(&sbinfo->si_rwsem);
3055+
1facf9fc 3056+ brp = sbinfo->si_branch + bindex;
3057+ if (bindex < bend)
3058+ memmove(brp, brp + 1, sizeof(*brp) * (bend - bindex));
3059+ sbinfo->si_branch[0 + bend] = NULL;
3060+ sbinfo->si_bend--;
3061+
53392da6 3062+ p = krealloc(sbinfo->si_branch, sizeof(*p) * bend, AuGFP_SBILIST);
1facf9fc 3063+ if (p)
3064+ sbinfo->si_branch = p;
4a4d8108 3065+ /* harmless error */
1facf9fc 3066+}
3067+
3068+static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3069+ const aufs_bindex_t bend)
3070+{
3071+ struct au_hdentry *hdp, *p;
3072+
1308ab2a 3073+ AuRwMustWriteLock(&dinfo->di_rwsem);
3074+
4a4d8108 3075+ hdp = dinfo->di_hdentry;
1facf9fc 3076+ if (bindex < bend)
4a4d8108
AM
3077+ memmove(hdp + bindex, hdp + bindex + 1,
3078+ sizeof(*hdp) * (bend - bindex));
3079+ hdp[0 + bend].hd_dentry = NULL;
1facf9fc 3080+ dinfo->di_bend--;
3081+
53392da6 3082+ p = krealloc(hdp, sizeof(*p) * bend, AuGFP_SBILIST);
1facf9fc 3083+ if (p)
3084+ dinfo->di_hdentry = p;
4a4d8108 3085+ /* harmless error */
1facf9fc 3086+}
3087+
3088+static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3089+ const aufs_bindex_t bend)
3090+{
3091+ struct au_hinode *hip, *p;
3092+
1308ab2a 3093+ AuRwMustWriteLock(&iinfo->ii_rwsem);
3094+
1facf9fc 3095+ hip = iinfo->ii_hinode + bindex;
3096+ if (bindex < bend)
3097+ memmove(hip, hip + 1, sizeof(*hip) * (bend - bindex));
3098+ iinfo->ii_hinode[0 + bend].hi_inode = NULL;
4a4d8108 3099+ au_hn_init(iinfo->ii_hinode + bend);
1facf9fc 3100+ iinfo->ii_bend--;
3101+
53392da6 3102+ p = krealloc(iinfo->ii_hinode, sizeof(*p) * bend, AuGFP_SBILIST);
1facf9fc 3103+ if (p)
3104+ iinfo->ii_hinode = p;
4a4d8108 3105+ /* harmless error */
1facf9fc 3106+}
3107+
3108+static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3109+ struct au_branch *br)
3110+{
3111+ aufs_bindex_t bend;
3112+ struct au_sbinfo *sbinfo;
53392da6
AM
3113+ struct dentry *root, *h_root;
3114+ struct inode *inode, *h_inode;
3115+ struct au_hinode *hinode;
1facf9fc 3116+
dece6358
AM
3117+ SiMustWriteLock(sb);
3118+
1facf9fc 3119+ root = sb->s_root;
3120+ inode = root->d_inode;
1facf9fc 3121+ sbinfo = au_sbi(sb);
3122+ bend = sbinfo->si_bend;
3123+
53392da6
AM
3124+ h_root = au_h_dptr(root, bindex);
3125+ hinode = au_hi(inode, bindex);
3126+ h_inode = au_igrab(hinode->hi_inode);
3127+ au_hiput(hinode);
1facf9fc 3128+
53392da6 3129+ au_sbilist_lock();
1facf9fc 3130+ au_br_do_del_brp(sbinfo, bindex, bend);
3131+ au_br_do_del_hdp(au_di(root), bindex, bend);
3132+ au_br_do_del_hip(au_ii(inode), bindex, bend);
53392da6
AM
3133+ au_sbilist_unlock();
3134+
3135+ dput(h_root);
3136+ iput(h_inode);
3137+ au_br_do_free(br);
1facf9fc 3138+}
3139+
3140+int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3141+{
3142+ int err, rerr, i;
3143+ unsigned int mnt_flags;
3144+ aufs_bindex_t bindex, bend, br_id;
3145+ unsigned char do_wh, verbose;
3146+ struct au_branch *br;
3147+ struct au_wbr *wbr;
3148+
3149+ err = 0;
3150+ bindex = au_find_dbindex(sb->s_root, del->h_path.dentry);
3151+ if (bindex < 0) {
3152+ if (remount)
3153+ goto out; /* success */
3154+ err = -ENOENT;
4a4d8108 3155+ pr_err("%s no such branch\n", del->pathname);
1facf9fc 3156+ goto out;
3157+ }
3158+ AuDbg("bindex b%d\n", bindex);
3159+
3160+ err = -EBUSY;
3161+ mnt_flags = au_mntflags(sb);
3162+ verbose = !!au_opt_test(mnt_flags, VERBOSE);
3163+ bend = au_sbend(sb);
3164+ if (unlikely(!bend)) {
3165+ AuVerbose(verbose, "no more branches left\n");
3166+ goto out;
3167+ }
3168+ br = au_sbr(sb, bindex);
86dc4139 3169+ AuDebugOn(!path_equal(&br->br_path, &del->h_path));
1facf9fc 3170+ i = atomic_read(&br->br_count);
3171+ if (unlikely(i)) {
3172+ AuVerbose(verbose, "%d file(s) opened\n", i);
e49829fe 3173+ goto out;
1facf9fc 3174+ }
3175+
3176+ wbr = br->br_wbr;
3177+ do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3178+ if (do_wh) {
1308ab2a 3179+ /* instead of WbrWhMustWriteLock(wbr) */
3180+ SiMustWriteLock(sb);
1facf9fc 3181+ for (i = 0; i < AuBrWh_Last; i++) {
3182+ dput(wbr->wbr_wh[i]);
3183+ wbr->wbr_wh[i] = NULL;
3184+ }
3185+ }
3186+
b752ccd1 3187+ err = test_children_busy(sb->s_root, bindex, verbose);
1facf9fc 3188+ if (unlikely(err)) {
3189+ if (do_wh)
3190+ goto out_wh;
3191+ goto out;
3192+ }
3193+
3194+ err = 0;
3195+ br_id = br->br_id;
3196+ if (!remount)
3197+ au_br_do_del(sb, bindex, br);
3198+ else {
3199+ sysaufs_brs_del(sb, bindex);
3200+ au_br_do_del(sb, bindex, br);
3201+ sysaufs_brs_add(sb, bindex);
3202+ }
3203+
1308ab2a 3204+ if (!bindex) {
1facf9fc 3205+ au_cpup_attr_all(sb->s_root->d_inode, /*force*/1);
1308ab2a 3206+ sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3207+ } else
1facf9fc 3208+ au_sub_nlink(sb->s_root->d_inode, del->h_path.dentry->d_inode);
3209+ if (au_opt_test(mnt_flags, PLINK))
3210+ au_plink_half_refresh(sb, br_id);
3211+
b752ccd1 3212+ if (au_xino_brid(sb) == br_id)
1facf9fc 3213+ au_xino_brid_set(sb, -1);
3214+ goto out; /* success */
3215+
4f0767ce 3216+out_wh:
1facf9fc 3217+ /* revert */
86dc4139 3218+ rerr = au_br_init_wh(sb, br, br->br_perm);
1facf9fc 3219+ if (rerr)
0c3ec466
AM
3220+ pr_warn("failed re-creating base whiteout, %s. (%d)\n",
3221+ del->pathname, rerr);
4f0767ce 3222+out:
1facf9fc 3223+ return err;
3224+}
3225+
3226+/* ---------------------------------------------------------------------- */
3227+
027c5e7a
AM
3228+static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
3229+{
3230+ int err;
3231+ aufs_bindex_t bstart, bend;
3232+ struct aufs_ibusy ibusy;
3233+ struct inode *inode, *h_inode;
3234+
3235+ err = -EPERM;
3236+ if (unlikely(!capable(CAP_SYS_ADMIN)))
3237+ goto out;
3238+
3239+ err = copy_from_user(&ibusy, arg, sizeof(ibusy));
3240+ if (!err)
3241+ err = !access_ok(VERIFY_WRITE, &arg->h_ino, sizeof(arg->h_ino));
3242+ if (unlikely(err)) {
3243+ err = -EFAULT;
3244+ AuTraceErr(err);
3245+ goto out;
3246+ }
3247+
3248+ err = -EINVAL;
3249+ si_read_lock(sb, AuLock_FLUSH);
3250+ if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbend(sb)))
3251+ goto out_unlock;
3252+
3253+ err = 0;
3254+ ibusy.h_ino = 0; /* invalid */
3255+ inode = ilookup(sb, ibusy.ino);
3256+ if (!inode
3257+ || inode->i_ino == AUFS_ROOT_INO
3258+ || is_bad_inode(inode))
3259+ goto out_unlock;
3260+
3261+ ii_read_lock_child(inode);
3262+ bstart = au_ibstart(inode);
3263+ bend = au_ibend(inode);
3264+ if (bstart <= ibusy.bindex && ibusy.bindex <= bend) {
3265+ h_inode = au_h_iptr(inode, ibusy.bindex);
3266+ if (h_inode && au_test_ibusy(inode, bstart, bend))
3267+ ibusy.h_ino = h_inode->i_ino;
3268+ }
3269+ ii_read_unlock(inode);
3270+ iput(inode);
3271+
3272+out_unlock:
3273+ si_read_unlock(sb);
3274+ if (!err) {
3275+ err = __put_user(ibusy.h_ino, &arg->h_ino);
3276+ if (unlikely(err)) {
3277+ err = -EFAULT;
3278+ AuTraceErr(err);
3279+ }
3280+ }
3281+out:
3282+ return err;
3283+}
3284+
3285+long au_ibusy_ioctl(struct file *file, unsigned long arg)
3286+{
3287+ return au_ibusy(file->f_dentry->d_sb, (void __user *)arg);
3288+}
3289+
3290+#ifdef CONFIG_COMPAT
3291+long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
3292+{
3293+ return au_ibusy(file->f_dentry->d_sb, compat_ptr(arg));
3294+}
3295+#endif
3296+
3297+/* ---------------------------------------------------------------------- */
3298+
1facf9fc 3299+/*
3300+ * change a branch permission
3301+ */
3302+
dece6358
AM
3303+static void au_warn_ima(void)
3304+{
3305+#ifdef CONFIG_IMA
1308ab2a 3306+ /* since it doesn't support mark_files_ro() */
027c5e7a 3307+ AuWarn1("RW -> RO makes IMA to produce wrong message\n");
dece6358
AM
3308+#endif
3309+}
3310+
1facf9fc 3311+static int do_need_sigen_inc(int a, int b)
3312+{
3313+ return au_br_whable(a) && !au_br_whable(b);
3314+}
3315+
3316+static int need_sigen_inc(int old, int new)
3317+{
3318+ return do_need_sigen_inc(old, new)
3319+ || do_need_sigen_inc(new, old);
3320+}
3321+
7f207e10
AM
3322+static unsigned long long au_farray_cb(void *a,
3323+ unsigned long long max __maybe_unused,
3324+ void *arg)
3325+{
3326+ unsigned long long n;
3327+ struct file **p, *f;
523b37e3
AM
3328+ struct au_sphlhead *files;
3329+ struct au_finfo *finfo;
7f207e10
AM
3330+ struct super_block *sb = arg;
3331+
3332+ n = 0;
3333+ p = a;
523b37e3
AM
3334+ files = &au_sbi(sb)->si_files;
3335+ spin_lock(&files->spin);
3336+ hlist_for_each_entry(finfo, &files->head, fi_hlist) {
3337+ f = finfo->fi_file;
3338+ if (file_count(f)
c06a8ce3 3339+ && !special_file(file_inode(f)->i_mode)) {
7f207e10
AM
3340+ get_file(f);
3341+ *p++ = f;
3342+ n++;
3343+ AuDebugOn(n > max);
3344+ }
523b37e3
AM
3345+ }
3346+ spin_unlock(&files->spin);
7f207e10
AM
3347+
3348+ return n;
3349+}
3350+
3351+static struct file **au_farray_alloc(struct super_block *sb,
3352+ unsigned long long *max)
3353+{
3354+ *max = atomic_long_read(&au_sbi(sb)->si_nfiles);
3355+ return au_array_alloc(max, au_farray_cb, sb);
3356+}
3357+
3358+static void au_farray_free(struct file **a, unsigned long long max)
3359+{
3360+ unsigned long long ull;
3361+
3362+ for (ull = 0; ull < max; ull++)
3363+ if (a[ull])
3364+ fput(a[ull]);
3365+ au_array_free(a);
3366+}
3367+
1facf9fc 3368+static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
3369+{
7f207e10 3370+ int err, do_warn;
027c5e7a 3371+ unsigned int mnt_flags;
7f207e10 3372+ unsigned long long ull, max;
e49829fe 3373+ aufs_bindex_t br_id;
38d290e6 3374+ unsigned char verbose, writer;
7f207e10 3375+ struct file *file, *hf, **array;
e49829fe
JR
3376+ struct inode *inode;
3377+ struct au_hfile *hfile;
1facf9fc 3378+
027c5e7a
AM
3379+ mnt_flags = au_mntflags(sb);
3380+ verbose = !!au_opt_test(mnt_flags, VERBOSE);
3381+
7f207e10
AM
3382+ array = au_farray_alloc(sb, &max);
3383+ err = PTR_ERR(array);
3384+ if (IS_ERR(array))
1facf9fc 3385+ goto out;
3386+
7f207e10 3387+ do_warn = 0;
e49829fe 3388+ br_id = au_sbr_id(sb, bindex);
7f207e10
AM
3389+ for (ull = 0; ull < max; ull++) {
3390+ file = array[ull];
1facf9fc 3391+
523b37e3 3392+ /* AuDbg("%pD\n", file); */
1facf9fc 3393+ fi_read_lock(file);
3394+ if (unlikely(au_test_mmapped(file))) {
3395+ err = -EBUSY;
523b37e3 3396+ AuVerbose(verbose, "mmapped %pD\n", file);
7f207e10 3397+ AuDbgFile(file);
1facf9fc 3398+ FiMustNoWaiters(file);
3399+ fi_read_unlock(file);
7f207e10 3400+ goto out_array;
1facf9fc 3401+ }
3402+
c06a8ce3 3403+ inode = file_inode(file);
e49829fe
JR
3404+ hfile = &au_fi(file)->fi_htop;
3405+ hf = hfile->hf_file;
3406+ if (!S_ISREG(inode->i_mode)
1facf9fc 3407+ || !(file->f_mode & FMODE_WRITE)
e49829fe 3408+ || hfile->hf_br->br_id != br_id
7f207e10
AM
3409+ || !(hf->f_mode & FMODE_WRITE))
3410+ array[ull] = NULL;
3411+ else {
3412+ do_warn = 1;
3413+ get_file(file);
1facf9fc 3414+ }
3415+
1facf9fc 3416+ FiMustNoWaiters(file);
3417+ fi_read_unlock(file);
7f207e10
AM
3418+ fput(file);
3419+ }
1facf9fc 3420+
3421+ err = 0;
7f207e10 3422+ if (do_warn)
dece6358 3423+ au_warn_ima();
7f207e10
AM
3424+
3425+ for (ull = 0; ull < max; ull++) {
3426+ file = array[ull];
3427+ if (!file)
3428+ continue;
3429+
1facf9fc 3430+ /* todo: already flushed? */
523b37e3
AM
3431+ /*
3432+ * fs/super.c:mark_files_ro() is gone, but aufs keeps its
3433+ * approach which resets f_mode and calls mnt_drop_write() and
3434+ * file_release_write() for each file, because the branch
3435+ * attribute in aufs world is totally different from the native
3436+ * fs rw/ro mode.
3437+ */
7f207e10
AM
3438+ /* fi_read_lock(file); */
3439+ hfile = &au_fi(file)->fi_htop;
3440+ hf = hfile->hf_file;
3441+ /* fi_read_unlock(file); */
027c5e7a 3442+ spin_lock(&hf->f_lock);
38d290e6
JR
3443+ writer = !!(hf->f_mode & FMODE_WRITER);
3444+ hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
027c5e7a 3445+ spin_unlock(&hf->f_lock);
38d290e6
JR
3446+ if (writer) {
3447+ put_write_access(file_inode(hf));
c06a8ce3 3448+ __mnt_drop_write(hf->f_path.mnt);
1facf9fc 3449+ }
3450+ }
3451+
7f207e10
AM
3452+out_array:
3453+ au_farray_free(array, max);
4f0767ce 3454+out:
7f207e10 3455+ AuTraceErr(err);
1facf9fc 3456+ return err;
3457+}
3458+
3459+int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
7f207e10 3460+ int *do_refresh)
1facf9fc 3461+{
3462+ int err, rerr;
3463+ aufs_bindex_t bindex;
3464+ struct dentry *root;
3465+ struct au_branch *br;
3466+
3467+ root = sb->s_root;
1facf9fc 3468+ bindex = au_find_dbindex(root, mod->h_root);
3469+ if (bindex < 0) {
3470+ if (remount)
3471+ return 0; /* success */
3472+ err = -ENOENT;
4a4d8108 3473+ pr_err("%s no such branch\n", mod->path);
1facf9fc 3474+ goto out;
3475+ }
3476+ AuDbg("bindex b%d\n", bindex);
3477+
3478+ err = test_br(mod->h_root->d_inode, mod->perm, mod->path);
3479+ if (unlikely(err))
3480+ goto out;
3481+
3482+ br = au_sbr(sb, bindex);
86dc4139 3483+ AuDebugOn(mod->h_root != au_br_dentry(br));
1facf9fc 3484+ if (br->br_perm == mod->perm)
3485+ return 0; /* success */
3486+
3487+ if (au_br_writable(br->br_perm)) {
3488+ /* remove whiteout base */
86dc4139 3489+ err = au_br_init_wh(sb, br, mod->perm);
1facf9fc 3490+ if (unlikely(err))
3491+ goto out;
3492+
3493+ if (!au_br_writable(mod->perm)) {
3494+ /* rw --> ro, file might be mmapped */
3495+ DiMustNoWaiters(root);
3496+ IiMustNoWaiters(root->d_inode);
3497+ di_write_unlock(root);
3498+ err = au_br_mod_files_ro(sb, bindex);
3499+ /* aufs_write_lock() calls ..._child() */
3500+ di_write_lock_child(root);
3501+
3502+ if (unlikely(err)) {
3503+ rerr = -ENOMEM;
3504+ br->br_wbr = kmalloc(sizeof(*br->br_wbr),
3505+ GFP_NOFS);
86dc4139
AM
3506+ if (br->br_wbr)
3507+ rerr = au_wbr_init(br, sb, br->br_perm);
1facf9fc 3508+ if (unlikely(rerr)) {
3509+ AuIOErr("nested error %d (%d)\n",
3510+ rerr, err);
3511+ br->br_perm = mod->perm;
3512+ }
3513+ }
3514+ }
3515+ } else if (au_br_writable(mod->perm)) {
3516+ /* ro --> rw */
3517+ err = -ENOMEM;
3518+ br->br_wbr = kmalloc(sizeof(*br->br_wbr), GFP_NOFS);
3519+ if (br->br_wbr) {
86dc4139 3520+ err = au_wbr_init(br, sb, mod->perm);
1facf9fc 3521+ if (unlikely(err)) {
3522+ kfree(br->br_wbr);
3523+ br->br_wbr = NULL;
3524+ }
3525+ }
3526+ }
3527+
3528+ if (!err) {
86dc4139
AM
3529+ if ((br->br_perm & AuBrAttr_UNPIN)
3530+ && !(mod->perm & AuBrAttr_UNPIN))
3531+ au_br_dflags_force(br);
3532+ else if (!(br->br_perm & AuBrAttr_UNPIN)
3533+ && (mod->perm & AuBrAttr_UNPIN))
3534+ au_br_dflags_restore(br);
7f207e10 3535+ *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
1facf9fc 3536+ br->br_perm = mod->perm;
3537+ }
3538+
4f0767ce 3539+out:
7f207e10 3540+ AuTraceErr(err);
1facf9fc 3541+ return err;
3542+}
7f207e10
AM
3543diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
3544--- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 3545+++ linux/fs/aufs/branch.h 2014-06-24 00:14:06.776052275 +0200
523b37e3 3546@@ -0,0 +1,264 @@
1facf9fc 3547+/*
523b37e3 3548+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 3549+ *
3550+ * This program, aufs is free software; you can redistribute it and/or modify
3551+ * it under the terms of the GNU General Public License as published by
3552+ * the Free Software Foundation; either version 2 of the License, or
3553+ * (at your option) any later version.
dece6358
AM
3554+ *
3555+ * This program is distributed in the hope that it will be useful,
3556+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
3557+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
3558+ * GNU General Public License for more details.
3559+ *
3560+ * You should have received a copy of the GNU General Public License
523b37e3 3561+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 3562+ */
3563+
3564+/*
3565+ * branch filesystems and xino for them
3566+ */
3567+
3568+#ifndef __AUFS_BRANCH_H__
3569+#define __AUFS_BRANCH_H__
3570+
3571+#ifdef __KERNEL__
3572+
1facf9fc 3573+#include <linux/mount.h>
4a4d8108 3574+#include "dynop.h"
1facf9fc 3575+#include "rwsem.h"
3576+#include "super.h"
3577+
3578+/* ---------------------------------------------------------------------- */
3579+
3580+/* a xino file */
3581+struct au_xino_file {
3582+ struct file *xi_file;
3583+ struct mutex xi_nondir_mtx;
3584+
3585+ /* todo: make xino files an array to support huge inode number */
3586+
3587+#ifdef CONFIG_DEBUG_FS
3588+ struct dentry *xi_dbgaufs;
3589+#endif
3590+};
3591+
3592+/* members for writable branch only */
3593+enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
3594+struct au_wbr {
dece6358 3595+ struct au_rwsem wbr_wh_rwsem;
1facf9fc 3596+ struct dentry *wbr_wh[AuBrWh_Last];
4a4d8108 3597+ atomic_t wbr_wh_running;
1facf9fc 3598+#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
3599+#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
3600+#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
3601+
3602+ /* mfs mode */
3603+ unsigned long long wbr_bytes;
3604+};
3605+
4a4d8108
AM
3606+/* ext2 has 3 types of operations at least, ext3 has 4 */
3607+#define AuBrDynOp (AuDyLast * 4)
3608+
1716fcea
AM
3609+#ifdef CONFIG_AUFS_HFSNOTIFY
3610+/* support for asynchronous destruction */
3611+struct au_br_hfsnotify {
3612+ struct fsnotify_group *hfsn_group;
3613+};
3614+#endif
3615+
392086de
AM
3616+/* sysfs entries */
3617+struct au_brsysfs {
3618+ char name[16];
3619+ struct attribute attr;
3620+};
3621+
3622+enum {
3623+ AuBrSysfs_BR,
3624+ AuBrSysfs_BRID,
3625+ AuBrSysfs_Last
3626+};
3627+
1facf9fc 3628+/* protected by superblock rwsem */
3629+struct au_branch {
3630+ struct au_xino_file br_xino;
3631+
3632+ aufs_bindex_t br_id;
3633+
3634+ int br_perm;
86dc4139
AM
3635+ unsigned int br_dflags;
3636+ struct path br_path;
4a4d8108
AM
3637+ spinlock_t br_dykey_lock;
3638+ struct au_dykey *br_dykey[AuBrDynOp];
1facf9fc 3639+ atomic_t br_count;
3640+
3641+ struct au_wbr *br_wbr;
3642+
3643+ /* xino truncation */
1facf9fc 3644+ atomic_t br_xino_running;
3645+
027c5e7a 3646+#ifdef CONFIG_AUFS_HFSNOTIFY
1716fcea 3647+ struct au_br_hfsnotify *br_hfsn;
027c5e7a
AM
3648+#endif
3649+
1facf9fc 3650+#ifdef CONFIG_SYSFS
392086de
AM
3651+ /* entries under sysfs per mount-point */
3652+ struct au_brsysfs br_sysfs[AuBrSysfs_Last];
1facf9fc 3653+#endif
3654+};
3655+
3656+/* ---------------------------------------------------------------------- */
3657+
86dc4139
AM
3658+static inline struct vfsmount *au_br_mnt(struct au_branch *br)
3659+{
3660+ return br->br_path.mnt;
3661+}
3662+
3663+static inline struct dentry *au_br_dentry(struct au_branch *br)
3664+{
3665+ return br->br_path.dentry;
3666+}
3667+
3668+static inline struct super_block *au_br_sb(struct au_branch *br)
3669+{
3670+ return au_br_mnt(br)->mnt_sb;
3671+}
3672+
1e00d052
AM
3673+/* branch permissions and attributes */
3674+#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
3675+#define AuBrPerm_RO (1 << 1) /* readonly */
3676+#define AuBrPerm_RR (1 << 2) /* natively readonly */
3677+#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
1facf9fc 3678+
1e00d052 3679+#define AuBrRAttr_WH (1 << 3) /* whiteout-able */
1facf9fc 3680+
1e00d052 3681+#define AuBrWAttr_NoLinkWH (1 << 4) /* un-hardlinkable whiteouts */
1facf9fc 3682+
86dc4139
AM
3683+#define AuBrAttr_UNPIN (1 << 5) /* rename-able top dir of
3684+ branch */
3685+
1facf9fc 3686+static inline int au_br_writable(int brperm)
3687+{
1e00d052 3688+ return brperm & AuBrPerm_RW;
1facf9fc 3689+}
3690+
3691+static inline int au_br_whable(int brperm)
3692+{
1e00d052
AM
3693+ return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
3694+}
3695+
3696+static inline int au_br_wh_linkable(int brperm)
3697+{
3698+ return !(brperm & AuBrWAttr_NoLinkWH);
1facf9fc 3699+}
3700+
3701+static inline int au_br_rdonly(struct au_branch *br)
3702+{
86dc4139 3703+ return ((au_br_sb(br)->s_flags & MS_RDONLY)
1facf9fc 3704+ || !au_br_writable(br->br_perm))
3705+ ? -EROFS : 0;
3706+}
3707+
4a4d8108 3708+static inline int au_br_hnotifyable(int brperm __maybe_unused)
1facf9fc 3709+{
4a4d8108 3710+#ifdef CONFIG_AUFS_HNOTIFY
1e00d052 3711+ return !(brperm & AuBrPerm_RR);
1facf9fc 3712+#else
3713+ return 0;
3714+#endif
3715+}
3716+
3717+/* ---------------------------------------------------------------------- */
3718+
3719+/* branch.c */
3720+struct au_sbinfo;
3721+void au_br_free(struct au_sbinfo *sinfo);
3722+int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
3723+struct au_opt_add;
3724+int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
3725+struct au_opt_del;
3726+int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
027c5e7a
AM
3727+long au_ibusy_ioctl(struct file *file, unsigned long arg);
3728+#ifdef CONFIG_COMPAT
3729+long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
3730+#endif
1facf9fc 3731+struct au_opt_mod;
3732+int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
7f207e10 3733+ int *do_refresh);
1facf9fc 3734+
3735+/* xino.c */
3736+static const loff_t au_loff_max = LLONG_MAX;
3737+
3738+int au_xib_trunc(struct super_block *sb);
3739+ssize_t xino_fread(au_readf_t func, struct file *file, void *buf, size_t size,
3740+ loff_t *pos);
3741+ssize_t xino_fwrite(au_writef_t func, struct file *file, void *buf, size_t size,
3742+ loff_t *pos);
3743+struct file *au_xino_create2(struct file *base_file, struct file *copy_src);
3744+struct file *au_xino_create(struct super_block *sb, char *fname, int silent);
3745+ino_t au_xino_new_ino(struct super_block *sb);
b752ccd1 3746+void au_xino_delete_inode(struct inode *inode, const int unlinked);
1facf9fc 3747+int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
3748+ ino_t ino);
3749+int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
3750+ ino_t *ino);
3751+int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t hino,
3752+ struct file *base_file, int do_test);
3753+int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex);
3754+
3755+struct au_opt_xino;
3756+int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount);
3757+void au_xino_clr(struct super_block *sb);
3758+struct file *au_xino_def(struct super_block *sb);
3759+int au_xino_path(struct seq_file *seq, struct file *file);
3760+
3761+/* ---------------------------------------------------------------------- */
3762+
3763+/* Superblock to branch */
3764+static inline
3765+aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
3766+{
3767+ return au_sbr(sb, bindex)->br_id;
3768+}
3769+
3770+static inline
3771+struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
3772+{
86dc4139 3773+ return au_br_mnt(au_sbr(sb, bindex));
1facf9fc 3774+}
3775+
3776+static inline
3777+struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
3778+{
86dc4139 3779+ return au_br_sb(au_sbr(sb, bindex));
1facf9fc 3780+}
3781+
3782+static inline void au_sbr_put(struct super_block *sb, aufs_bindex_t bindex)
3783+{
e49829fe 3784+ atomic_dec(&au_sbr(sb, bindex)->br_count);
1facf9fc 3785+}
3786+
3787+static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
3788+{
3789+ return au_sbr(sb, bindex)->br_perm;
3790+}
3791+
3792+static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
3793+{
3794+ return au_br_whable(au_sbr_perm(sb, bindex));
3795+}
3796+
3797+/* ---------------------------------------------------------------------- */
3798+
3799+/*
3800+ * wbr_wh_read_lock, wbr_wh_write_lock
3801+ * wbr_wh_read_unlock, wbr_wh_write_unlock, wbr_wh_downgrade_lock
3802+ */
3803+AuSimpleRwsemFuncs(wbr_wh, struct au_wbr *wbr, &wbr->wbr_wh_rwsem);
3804+
dece6358
AM
3805+#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&wbr->wbr_wh_rwsem)
3806+#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&wbr->wbr_wh_rwsem)
3807+#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&wbr->wbr_wh_rwsem)
3808+
1facf9fc 3809+#endif /* __KERNEL__ */
3810+#endif /* __AUFS_BRANCH_H__ */
7f207e10
AM
3811diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
3812--- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
3813+++ linux/fs/aufs/conf.mk 2014-06-24 00:14:06.776052275 +0200
3814@@ -0,0 +1,36 @@
4a4d8108
AM
3815+
3816+AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
3817+
3818+define AuConf
3819+ifdef ${1}
3820+AuConfStr += ${1}=${${1}}
3821+endif
3822+endef
3823+
b752ccd1 3824+AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
e49829fe 3825+ SBILIST \
7f207e10 3826+ HNOTIFY HFSNOTIFY \
4a4d8108
AM
3827+ EXPORT INO_T_64 \
3828+ RDU \
4a4d8108
AM
3829+ SHWH \
3830+ BR_RAMFS \
3831+ BR_FUSE POLL \
3832+ BR_HFSPLUS \
3833+ BDEV_LOOP \
b752ccd1
AM
3834+ DEBUG MAGIC_SYSRQ
3835+$(foreach i, ${AuConfAll}, \
4a4d8108
AM
3836+ $(eval $(call AuConf,CONFIG_AUFS_${i})))
3837+
3838+AuConfName = ${obj}/conf.str
3839+${AuConfName}.tmp: FORCE
3840+ @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
3841+${AuConfName}: ${AuConfName}.tmp
3842+ @diff -q $< $@ > /dev/null 2>&1 || { \
3843+ echo ' GEN ' $@; \
3844+ cp -p $< $@; \
3845+ }
3846+FORCE:
3847+clean-files += ${AuConfName} ${AuConfName}.tmp
3848+${obj}/sysfs.o: ${AuConfName}
b752ccd1
AM
3849+
3850+-include ${srctree}/${src}/conf_priv.mk
7f207e10
AM
3851diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
3852--- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
3853+++ linux/fs/aufs/cpup.c 2014-06-24 00:14:06.779385790 +0200
3854@@ -0,0 +1,1289 @@
1facf9fc 3855+/*
523b37e3 3856+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 3857+ *
3858+ * This program, aufs is free software; you can redistribute it and/or modify
3859+ * it under the terms of the GNU General Public License as published by
3860+ * the Free Software Foundation; either version 2 of the License, or
3861+ * (at your option) any later version.
dece6358
AM
3862+ *
3863+ * This program is distributed in the hope that it will be useful,
3864+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
3865+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
3866+ * GNU General Public License for more details.
3867+ *
3868+ * You should have received a copy of the GNU General Public License
523b37e3 3869+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 3870+ */
3871+
3872+/*
3873+ * copy-up functions, see wbr_policy.c for copy-down
3874+ */
3875+
3876+#include <linux/fs_stack.h>
dece6358 3877+#include <linux/mm.h>
1facf9fc 3878+#include "aufs.h"
3879+
86dc4139 3880+void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
1facf9fc 3881+{
3882+ const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
367653fa 3883+ | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
1facf9fc 3884+
86dc4139
AM
3885+ BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
3886+
3887+ dst->i_flags |= iflags & ~mask;
1facf9fc 3888+ if (au_test_fs_notime(dst->i_sb))
3889+ dst->i_flags |= S_NOATIME | S_NOCMTIME;
3890+}
3891+
3892+void au_cpup_attr_timesizes(struct inode *inode)
3893+{
3894+ struct inode *h_inode;
3895+
3896+ h_inode = au_h_iptr(inode, au_ibstart(inode));
3897+ fsstack_copy_attr_times(inode, h_inode);
4a4d8108 3898+ fsstack_copy_inode_size(inode, h_inode);
1facf9fc 3899+}
3900+
3901+void au_cpup_attr_nlink(struct inode *inode, int force)
3902+{
3903+ struct inode *h_inode;
3904+ struct super_block *sb;
3905+ aufs_bindex_t bindex, bend;
3906+
3907+ sb = inode->i_sb;
3908+ bindex = au_ibstart(inode);
3909+ h_inode = au_h_iptr(inode, bindex);
3910+ if (!force
3911+ && !S_ISDIR(h_inode->i_mode)
3912+ && au_opt_test(au_mntflags(sb), PLINK)
3913+ && au_plink_test(inode))
3914+ return;
3915+
7eafdf33
AM
3916+ /*
3917+ * 0 can happen in revalidating.
38d290e6
JR
3918+ * h_inode->i_mutex may not be held here, but it is harmless since once
3919+ * i_nlink reaches 0, it will never become positive except O_TMPFILE
3920+ * case.
3921+ * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
3922+ * the incorrect link count.
7eafdf33 3923+ */
92d182d2 3924+ set_nlink(inode, h_inode->i_nlink);
1facf9fc 3925+
3926+ /*
3927+ * fewer nlink makes find(1) noisy, but larger nlink doesn't.
3928+ * it may includes whplink directory.
3929+ */
3930+ if (S_ISDIR(h_inode->i_mode)) {
3931+ bend = au_ibend(inode);
3932+ for (bindex++; bindex <= bend; bindex++) {
3933+ h_inode = au_h_iptr(inode, bindex);
3934+ if (h_inode)
3935+ au_add_nlink(inode, h_inode);
3936+ }
3937+ }
3938+}
3939+
3940+void au_cpup_attr_changeable(struct inode *inode)
3941+{
3942+ struct inode *h_inode;
3943+
3944+ h_inode = au_h_iptr(inode, au_ibstart(inode));
3945+ inode->i_mode = h_inode->i_mode;
3946+ inode->i_uid = h_inode->i_uid;
3947+ inode->i_gid = h_inode->i_gid;
3948+ au_cpup_attr_timesizes(inode);
86dc4139 3949+ au_cpup_attr_flags(inode, h_inode->i_flags);
1facf9fc 3950+}
3951+
3952+void au_cpup_igen(struct inode *inode, struct inode *h_inode)
3953+{
3954+ struct au_iinfo *iinfo = au_ii(inode);
3955+
1308ab2a 3956+ IiMustWriteLock(inode);
3957+
1facf9fc 3958+ iinfo->ii_higen = h_inode->i_generation;
3959+ iinfo->ii_hsb1 = h_inode->i_sb;
3960+}
3961+
3962+void au_cpup_attr_all(struct inode *inode, int force)
3963+{
3964+ struct inode *h_inode;
3965+
3966+ h_inode = au_h_iptr(inode, au_ibstart(inode));
3967+ au_cpup_attr_changeable(inode);
3968+ if (inode->i_nlink > 0)
3969+ au_cpup_attr_nlink(inode, force);
3970+ inode->i_rdev = h_inode->i_rdev;
3971+ inode->i_blkbits = h_inode->i_blkbits;
3972+ au_cpup_igen(inode, h_inode);
3973+}
3974+
3975+/* ---------------------------------------------------------------------- */
3976+
3977+/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
3978+
3979+/* keep the timestamps of the parent dir when cpup */
3980+void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
3981+ struct path *h_path)
3982+{
3983+ struct inode *h_inode;
3984+
3985+ dt->dt_dentry = dentry;
3986+ dt->dt_h_path = *h_path;
3987+ h_inode = h_path->dentry->d_inode;
3988+ dt->dt_atime = h_inode->i_atime;
3989+ dt->dt_mtime = h_inode->i_mtime;
3990+ /* smp_mb(); */
3991+}
3992+
3993+void au_dtime_revert(struct au_dtime *dt)
3994+{
3995+ struct iattr attr;
3996+ int err;
3997+
3998+ attr.ia_atime = dt->dt_atime;
3999+ attr.ia_mtime = dt->dt_mtime;
4000+ attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4001+ | ATTR_ATIME | ATTR_ATIME_SET;
4002+
523b37e3
AM
4003+ /* no delegation since this is a directory */
4004+ err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
1facf9fc 4005+ if (unlikely(err))
0c3ec466 4006+ pr_warn("restoring timestamps failed(%d). ignored\n", err);
1facf9fc 4007+}
4008+
4009+/* ---------------------------------------------------------------------- */
4010+
86dc4139
AM
4011+/* internal use only */
4012+struct au_cpup_reg_attr {
4013+ int valid;
4014+ struct kstat st;
4015+ unsigned int iflags; /* inode->i_flags */
4016+};
4017+
1facf9fc 4018+static noinline_for_stack
86dc4139
AM
4019+int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct dentry *h_src,
4020+ struct au_cpup_reg_attr *h_src_attr)
1facf9fc 4021+{
4022+ int err, sbits;
4023+ struct iattr ia;
4024+ struct path h_path;
1308ab2a 4025+ struct inode *h_isrc, *h_idst;
86dc4139 4026+ struct kstat *h_st;
1facf9fc 4027+
4028+ h_path.dentry = au_h_dptr(dst, bindex);
1308ab2a 4029+ h_idst = h_path.dentry->d_inode;
1facf9fc 4030+ h_path.mnt = au_sbr_mnt(dst->d_sb, bindex);
4031+ h_isrc = h_src->d_inode;
1308ab2a 4032+ ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
1facf9fc 4033+ | ATTR_ATIME | ATTR_MTIME
4034+ | ATTR_ATIME_SET | ATTR_MTIME_SET;
86dc4139
AM
4035+ if (h_src_attr && h_src_attr->valid) {
4036+ h_st = &h_src_attr->st;
4037+ ia.ia_uid = h_st->uid;
4038+ ia.ia_gid = h_st->gid;
4039+ ia.ia_atime = h_st->atime;
4040+ ia.ia_mtime = h_st->mtime;
4041+ if (h_idst->i_mode != h_st->mode
4042+ && !S_ISLNK(h_idst->i_mode)) {
4043+ ia.ia_valid |= ATTR_MODE;
4044+ ia.ia_mode = h_st->mode;
4045+ }
4046+ sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4047+ au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4048+ } else {
4049+ ia.ia_uid = h_isrc->i_uid;
4050+ ia.ia_gid = h_isrc->i_gid;
4051+ ia.ia_atime = h_isrc->i_atime;
4052+ ia.ia_mtime = h_isrc->i_mtime;
4053+ if (h_idst->i_mode != h_isrc->i_mode
4054+ && !S_ISLNK(h_idst->i_mode)) {
4055+ ia.ia_valid |= ATTR_MODE;
4056+ ia.ia_mode = h_isrc->i_mode;
4057+ }
4058+ sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4059+ au_cpup_attr_flags(h_idst, h_isrc->i_flags);
1308ab2a 4060+ }
523b37e3
AM
4061+ /* no delegation since it is just created */
4062+ err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
1facf9fc 4063+
4064+ /* is this nfs only? */
4065+ if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4066+ ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4067+ ia.ia_mode = h_isrc->i_mode;
523b37e3 4068+ err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
1facf9fc 4069+ }
4070+
4071+ return err;
4072+}
4073+
4074+/* ---------------------------------------------------------------------- */
4075+
4076+static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4077+ char *buf, unsigned long blksize)
4078+{
4079+ int err;
4080+ size_t sz, rbytes, wbytes;
4081+ unsigned char all_zero;
4082+ char *p, *zp;
4083+ struct mutex *h_mtx;
4084+ /* reduce stack usage */
4085+ struct iattr *ia;
4086+
4087+ zp = page_address(ZERO_PAGE(0));
4088+ if (unlikely(!zp))
4089+ return -ENOMEM; /* possible? */
4090+
4091+ err = 0;
4092+ all_zero = 0;
4093+ while (len) {
4094+ AuDbg("len %lld\n", len);
4095+ sz = blksize;
4096+ if (len < blksize)
4097+ sz = len;
4098+
4099+ rbytes = 0;
4100+ /* todo: signal_pending? */
4101+ while (!rbytes || err == -EAGAIN || err == -EINTR) {
4102+ rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
4103+ err = rbytes;
4104+ }
4105+ if (unlikely(err < 0))
4106+ break;
4107+
4108+ all_zero = 0;
4109+ if (len >= rbytes && rbytes == blksize)
4110+ all_zero = !memcmp(buf, zp, rbytes);
4111+ if (!all_zero) {
4112+ wbytes = rbytes;
4113+ p = buf;
4114+ while (wbytes) {
4115+ size_t b;
4116+
4117+ b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
4118+ err = b;
4119+ /* todo: signal_pending? */
4120+ if (unlikely(err == -EAGAIN || err == -EINTR))
4121+ continue;
4122+ if (unlikely(err < 0))
4123+ break;
4124+ wbytes -= b;
4125+ p += b;
4126+ }
392086de
AM
4127+ if (unlikely(err < 0))
4128+ break;
1facf9fc 4129+ } else {
4130+ loff_t res;
4131+
4132+ AuLabel(hole);
4133+ res = vfsub_llseek(dst, rbytes, SEEK_CUR);
4134+ err = res;
4135+ if (unlikely(res < 0))
4136+ break;
4137+ }
4138+ len -= rbytes;
4139+ err = 0;
4140+ }
4141+
4142+ /* the last block may be a hole */
4143+ if (!err && all_zero) {
4144+ AuLabel(last hole);
4145+
4146+ err = 1;
4147+ if (au_test_nfs(dst->f_dentry->d_sb)) {
4148+ /* nfs requires this step to make last hole */
4149+ /* is this only nfs? */
4150+ do {
4151+ /* todo: signal_pending? */
4152+ err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
4153+ } while (err == -EAGAIN || err == -EINTR);
4154+ if (err == 1)
4155+ dst->f_pos--;
4156+ }
4157+
4158+ if (err == 1) {
4159+ ia = (void *)buf;
4160+ ia->ia_size = dst->f_pos;
4161+ ia->ia_valid = ATTR_SIZE | ATTR_FILE;
4162+ ia->ia_file = dst;
c06a8ce3 4163+ h_mtx = &file_inode(dst)->i_mutex;
1facf9fc 4164+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
523b37e3
AM
4165+ /* no delegation since it is just created */
4166+ err = vfsub_notify_change(&dst->f_path, ia,
4167+ /*delegated*/NULL);
1facf9fc 4168+ mutex_unlock(h_mtx);
4169+ }
4170+ }
4171+
4172+ return err;
4173+}
4174+
4175+int au_copy_file(struct file *dst, struct file *src, loff_t len)
4176+{
4177+ int err;
4178+ unsigned long blksize;
4179+ unsigned char do_kfree;
4180+ char *buf;
4181+
4182+ err = -ENOMEM;
4183+ blksize = dst->f_dentry->d_sb->s_blocksize;
4184+ if (!blksize || PAGE_SIZE < blksize)
4185+ blksize = PAGE_SIZE;
4186+ AuDbg("blksize %lu\n", blksize);
4187+ do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
4188+ if (do_kfree)
4189+ buf = kmalloc(blksize, GFP_NOFS);
4190+ else
4191+ buf = (void *)__get_free_page(GFP_NOFS);
4192+ if (unlikely(!buf))
4193+ goto out;
4194+
4195+ if (len > (1 << 22))
4196+ AuDbg("copying a large file %lld\n", (long long)len);
4197+
4198+ src->f_pos = 0;
4199+ dst->f_pos = 0;
4200+ err = au_do_copy_file(dst, src, len, buf, blksize);
4201+ if (do_kfree)
4202+ kfree(buf);
4203+ else
4204+ free_page((unsigned long)buf);
4205+
4f0767ce 4206+out:
1facf9fc 4207+ return err;
4208+}
4209+
4210+/*
4211+ * to support a sparse file which is opened with O_APPEND,
4212+ * we need to close the file.
4213+ */
c2b27bf2 4214+static int au_cp_regular(struct au_cp_generic *cpg)
1facf9fc 4215+{
4216+ int err, i;
4217+ enum { SRC, DST };
4218+ struct {
4219+ aufs_bindex_t bindex;
4220+ unsigned int flags;
4221+ struct dentry *dentry;
392086de 4222+ int force_wr;
1facf9fc 4223+ struct file *file;
523b37e3 4224+ void *label;
1facf9fc 4225+ } *f, file[] = {
4226+ {
c2b27bf2 4227+ .bindex = cpg->bsrc,
1facf9fc 4228+ .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
523b37e3 4229+ .label = &&out
1facf9fc 4230+ },
4231+ {
c2b27bf2 4232+ .bindex = cpg->bdst,
1facf9fc 4233+ .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
392086de 4234+ .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
523b37e3 4235+ .label = &&out_src
1facf9fc 4236+ }
4237+ };
4238+ struct super_block *sb;
4239+
4240+ /* bsrc branch can be ro/rw. */
c2b27bf2 4241+ sb = cpg->dentry->d_sb;
1facf9fc 4242+ f = file;
4243+ for (i = 0; i < 2; i++, f++) {
c2b27bf2
AM
4244+ f->dentry = au_h_dptr(cpg->dentry, f->bindex);
4245+ f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
392086de 4246+ /*file*/NULL, f->force_wr);
1facf9fc 4247+ err = PTR_ERR(f->file);
4248+ if (IS_ERR(f->file))
4249+ goto *f->label;
1facf9fc 4250+ }
4251+
4252+ /* try stopping to update while we copyup */
4253+ IMustLock(file[SRC].dentry->d_inode);
c2b27bf2 4254+ err = au_copy_file(file[DST].file, file[SRC].file, cpg->len);
1facf9fc 4255+
1facf9fc 4256+ fput(file[DST].file);
4257+ au_sbr_put(sb, file[DST].bindex);
523b37e3 4258+
4f0767ce 4259+out_src:
1facf9fc 4260+ fput(file[SRC].file);
4261+ au_sbr_put(sb, file[SRC].bindex);
4f0767ce 4262+out:
1facf9fc 4263+ return err;
4264+}
4265+
c2b27bf2 4266+static int au_do_cpup_regular(struct au_cp_generic *cpg,
86dc4139 4267+ struct au_cpup_reg_attr *h_src_attr)
1facf9fc 4268+{
4269+ int err, rerr;
4270+ loff_t l;
86dc4139 4271+ struct path h_path;
38d290e6 4272+ struct inode *h_src_inode, *h_dst_inode;
1facf9fc 4273+
4274+ err = 0;
c2b27bf2 4275+ h_src_inode = au_h_iptr(cpg->dentry->d_inode, cpg->bsrc);
86dc4139 4276+ l = i_size_read(h_src_inode);
c2b27bf2
AM
4277+ if (cpg->len == -1 || l < cpg->len)
4278+ cpg->len = l;
4279+ if (cpg->len) {
86dc4139
AM
4280+ /* try stopping to update while we are referencing */
4281+ mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
c2b27bf2 4282+ au_pin_hdir_unlock(cpg->pin);
1facf9fc 4283+
c2b27bf2
AM
4284+ h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
4285+ h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
86dc4139
AM
4286+ h_src_attr->iflags = h_src_inode->i_flags;
4287+ err = vfs_getattr(&h_path, &h_src_attr->st);
4288+ if (unlikely(err)) {
4289+ mutex_unlock(&h_src_inode->i_mutex);
4290+ goto out;
4291+ }
4292+ h_src_attr->valid = 1;
c2b27bf2 4293+ err = au_cp_regular(cpg);
86dc4139 4294+ mutex_unlock(&h_src_inode->i_mutex);
c2b27bf2 4295+ rerr = au_pin_hdir_relock(cpg->pin);
86dc4139
AM
4296+ if (!err && rerr)
4297+ err = rerr;
1facf9fc 4298+ }
38d290e6
JR
4299+ if (!err && (h_src_inode->i_state & I_LINKABLE)) {
4300+ h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
4301+ h_dst_inode = h_path.dentry->d_inode;
4302+ spin_lock(&h_dst_inode->i_lock);
4303+ h_dst_inode->i_state |= I_LINKABLE;
4304+ spin_unlock(&h_dst_inode->i_lock);
4305+ }
1facf9fc 4306+
4f0767ce 4307+out:
1facf9fc 4308+ return err;
4309+}
4310+
4311+static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
4312+ struct inode *h_dir)
4313+{
4314+ int err, symlen;
4315+ mm_segment_t old_fs;
b752ccd1
AM
4316+ union {
4317+ char *k;
4318+ char __user *u;
4319+ } sym;
1facf9fc 4320+
4321+ err = -ENOSYS;
4322+ if (unlikely(!h_src->d_inode->i_op->readlink))
4323+ goto out;
4324+
4325+ err = -ENOMEM;
537831f9 4326+ sym.k = (void *)__get_free_page(GFP_NOFS);
b752ccd1 4327+ if (unlikely(!sym.k))
1facf9fc 4328+ goto out;
4329+
9dbd164d 4330+ /* unnecessary to support mmap_sem since symlink is not mmap-able */
1facf9fc 4331+ old_fs = get_fs();
4332+ set_fs(KERNEL_DS);
b752ccd1 4333+ symlen = h_src->d_inode->i_op->readlink(h_src, sym.u, PATH_MAX);
1facf9fc 4334+ err = symlen;
4335+ set_fs(old_fs);
4336+
4337+ if (symlen > 0) {
b752ccd1
AM
4338+ sym.k[symlen] = 0;
4339+ err = vfsub_symlink(h_dir, h_path, sym.k);
1facf9fc 4340+ }
537831f9 4341+ free_page((unsigned long)sym.k);
1facf9fc 4342+
4f0767ce 4343+out:
1facf9fc 4344+ return err;
4345+}
4346+
1facf9fc 4347+static noinline_for_stack
c2b27bf2 4348+int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
86dc4139 4349+ struct au_cpup_reg_attr *h_src_attr)
1facf9fc 4350+{
4351+ int err;
4352+ umode_t mode;
4353+ unsigned int mnt_flags;
4354+ unsigned char isdir;
c2b27bf2 4355+ const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
1facf9fc 4356+ struct au_dtime dt;
4357+ struct path h_path;
4358+ struct dentry *h_src, *h_dst, *h_parent;
4359+ struct inode *h_inode, *h_dir;
4360+ struct super_block *sb;
4361+
4362+ /* bsrc branch can be ro/rw. */
c2b27bf2 4363+ h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
1facf9fc 4364+ h_inode = h_src->d_inode;
c2b27bf2 4365+ AuDebugOn(h_inode != au_h_iptr(cpg->dentry->d_inode, cpg->bsrc));
1facf9fc 4366+
4367+ /* try stopping to be referenced while we are creating */
c2b27bf2
AM
4368+ h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
4369+ if (au_ftest_cpup(cpg->flags, RENAME))
86dc4139
AM
4370+ AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
4371+ AUFS_WH_PFX_LEN));
1facf9fc 4372+ h_parent = h_dst->d_parent; /* dir inode is locked */
4373+ h_dir = h_parent->d_inode;
4374+ IMustLock(h_dir);
4375+ AuDebugOn(h_parent != h_dst->d_parent);
4376+
c2b27bf2
AM
4377+ sb = cpg->dentry->d_sb;
4378+ h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
1facf9fc 4379+ if (do_dt) {
4380+ h_path.dentry = h_parent;
4381+ au_dtime_store(&dt, dst_parent, &h_path);
4382+ }
4383+ h_path.dentry = h_dst;
4384+
4385+ isdir = 0;
4386+ mode = h_inode->i_mode;
4387+ switch (mode & S_IFMT) {
4388+ case S_IFREG:
b4510431
AM
4389+ err = vfsub_create(h_dir, &h_path, mode | S_IWUSR,
4390+ /*want_excl*/true);
1facf9fc 4391+ if (!err)
c2b27bf2 4392+ err = au_do_cpup_regular(cpg, h_src_attr);
1facf9fc 4393+ break;
4394+ case S_IFDIR:
4395+ isdir = 1;
4396+ err = vfsub_mkdir(h_dir, &h_path, mode);
4397+ if (!err) {
4398+ /*
4399+ * strange behaviour from the users view,
4400+ * particularry setattr case
4401+ */
c2b27bf2 4402+ if (au_ibstart(dst_parent->d_inode) == cpg->bdst)
1facf9fc 4403+ au_cpup_attr_nlink(dst_parent->d_inode,
4404+ /*force*/1);
c2b27bf2 4405+ au_cpup_attr_nlink(cpg->dentry->d_inode, /*force*/1);
1facf9fc 4406+ }
4407+ break;
4408+ case S_IFLNK:
4409+ err = au_do_cpup_symlink(&h_path, h_src, h_dir);
4410+ break;
4411+ case S_IFCHR:
4412+ case S_IFBLK:
4413+ AuDebugOn(!capable(CAP_MKNOD));
4414+ /*FALLTHROUGH*/
4415+ case S_IFIFO:
4416+ case S_IFSOCK:
4417+ err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
4418+ break;
4419+ default:
4420+ AuIOErr("Unknown inode type 0%o\n", mode);
4421+ err = -EIO;
4422+ }
4423+
4424+ mnt_flags = au_mntflags(sb);
4425+ if (!au_opt_test(mnt_flags, UDBA_NONE)
4426+ && !isdir
4427+ && au_opt_test(mnt_flags, XINO)
38d290e6
JR
4428+ && (h_inode->i_nlink == 1
4429+ || (h_inode->i_state & I_LINKABLE))
1facf9fc 4430+ /* todo: unnecessary? */
c2b27bf2
AM
4431+ /* && cpg->dentry->d_inode->i_nlink == 1 */
4432+ && cpg->bdst < cpg->bsrc
4433+ && !au_ftest_cpup(cpg->flags, KEEPLINO))
4434+ au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
1facf9fc 4435+ /* ignore this error */
4436+
4437+ if (do_dt)
4438+ au_dtime_revert(&dt);
4439+ return err;
4440+}
4441+
392086de 4442+static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
86dc4139
AM
4443+{
4444+ int err;
392086de 4445+ struct dentry *dentry, *h_dentry, *h_parent, *parent;
86dc4139 4446+ struct inode *h_dir;
392086de 4447+ aufs_bindex_t bdst;
86dc4139 4448+
392086de
AM
4449+ dentry = cpg->dentry;
4450+ bdst = cpg->bdst;
4451+ h_dentry = au_h_dptr(dentry, bdst);
4452+ if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
4453+ dget(h_dentry);
4454+ au_set_h_dptr(dentry, bdst, NULL);
4455+ err = au_lkup_neg(dentry, bdst, /*wh*/0);
4456+ if (!err)
4457+ h_path->dentry = dget(au_h_dptr(dentry, bdst));
86dc4139 4458+ au_set_h_dptr(dentry, bdst, h_dentry);
392086de
AM
4459+ } else {
4460+ err = 0;
4461+ parent = dget_parent(dentry);
4462+ h_parent = au_h_dptr(parent, bdst);
4463+ dput(parent);
4464+ h_path->dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
4465+ if (IS_ERR(h_path->dentry))
4466+ err = PTR_ERR(h_path->dentry);
86dc4139 4467+ }
392086de
AM
4468+ if (unlikely(err))
4469+ goto out;
86dc4139 4470+
86dc4139
AM
4471+ h_parent = h_dentry->d_parent; /* dir inode is locked */
4472+ h_dir = h_parent->d_inode;
4473+ IMustLock(h_dir);
523b37e3
AM
4474+ AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
4475+ /* no delegation since it is just created */
4476+ err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL);
86dc4139
AM
4477+ dput(h_path->dentry);
4478+
4479+out:
4480+ return err;
4481+}
4482+
1facf9fc 4483+/*
4484+ * copyup the @dentry from @bsrc to @bdst.
4485+ * the caller must set the both of lower dentries.
4486+ * @len is for truncating when it is -1 copyup the entire file.
4487+ * in link/rename cases, @dst_parent may be different from the real one.
c2b27bf2 4488+ * basic->bsrc can be larger than basic->bdst.
1facf9fc 4489+ */
c2b27bf2 4490+static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
1facf9fc 4491+{
4492+ int err, rerr;
4493+ aufs_bindex_t old_ibstart;
4494+ unsigned char isdir, plink;
1facf9fc 4495+ struct dentry *h_src, *h_dst, *h_parent;
523b37e3 4496+ struct inode *dst_inode, *h_dir, *inode, *delegated;
1facf9fc 4497+ struct super_block *sb;
86dc4139 4498+ struct au_branch *br;
c2b27bf2
AM
4499+ /* to reuduce stack size */
4500+ struct {
4501+ struct au_dtime dt;
4502+ struct path h_path;
4503+ struct au_cpup_reg_attr h_src_attr;
4504+ } *a;
1facf9fc 4505+
c2b27bf2
AM
4506+ err = -ENOMEM;
4507+ a = kmalloc(sizeof(*a), GFP_NOFS);
4508+ if (unlikely(!a))
4509+ goto out;
4510+ a->h_src_attr.valid = 0;
1facf9fc 4511+
c2b27bf2
AM
4512+ sb = cpg->dentry->d_sb;
4513+ br = au_sbr(sb, cpg->bdst);
4514+ a->h_path.mnt = au_br_mnt(br);
4515+ h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
1facf9fc 4516+ h_parent = h_dst->d_parent; /* dir inode is locked */
4517+ h_dir = h_parent->d_inode;
4518+ IMustLock(h_dir);
4519+
c2b27bf2
AM
4520+ h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
4521+ inode = cpg->dentry->d_inode;
1facf9fc 4522+
4523+ if (!dst_parent)
c2b27bf2 4524+ dst_parent = dget_parent(cpg->dentry);
1facf9fc 4525+ else
4526+ dget(dst_parent);
4527+
4528+ plink = !!au_opt_test(au_mntflags(sb), PLINK);
c2b27bf2 4529+ dst_inode = au_h_iptr(inode, cpg->bdst);
1facf9fc 4530+ if (dst_inode) {
4531+ if (unlikely(!plink)) {
4532+ err = -EIO;
027c5e7a
AM
4533+ AuIOErr("hi%lu(i%lu) exists on b%d "
4534+ "but plink is disabled\n",
c2b27bf2
AM
4535+ dst_inode->i_ino, inode->i_ino, cpg->bdst);
4536+ goto out_parent;
1facf9fc 4537+ }
4538+
4539+ if (dst_inode->i_nlink) {
c2b27bf2 4540+ const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
1facf9fc 4541+
c2b27bf2 4542+ h_src = au_plink_lkup(inode, cpg->bdst);
1facf9fc 4543+ err = PTR_ERR(h_src);
4544+ if (IS_ERR(h_src))
c2b27bf2 4545+ goto out_parent;
1facf9fc 4546+ if (unlikely(!h_src->d_inode)) {
4547+ err = -EIO;
4548+ AuIOErr("i%lu exists on a upper branch "
027c5e7a
AM
4549+ "but not pseudo-linked\n",
4550+ inode->i_ino);
1facf9fc 4551+ dput(h_src);
c2b27bf2 4552+ goto out_parent;
1facf9fc 4553+ }
4554+
4555+ if (do_dt) {
c2b27bf2
AM
4556+ a->h_path.dentry = h_parent;
4557+ au_dtime_store(&a->dt, dst_parent, &a->h_path);
1facf9fc 4558+ }
86dc4139 4559+
c2b27bf2 4560+ a->h_path.dentry = h_dst;
523b37e3
AM
4561+ delegated = NULL;
4562+ err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
c2b27bf2 4563+ if (!err && au_ftest_cpup(cpg->flags, RENAME))
392086de 4564+ err = au_do_ren_after_cpup(cpg, &a->h_path);
1facf9fc 4565+ if (do_dt)
c2b27bf2 4566+ au_dtime_revert(&a->dt);
523b37e3
AM
4567+ if (unlikely(err == -EWOULDBLOCK)) {
4568+ pr_warn("cannot retry for NFSv4 delegation"
4569+ " for an internal link\n");
4570+ iput(delegated);
4571+ }
1facf9fc 4572+ dput(h_src);
c2b27bf2 4573+ goto out_parent;
1facf9fc 4574+ } else
4575+ /* todo: cpup_wh_file? */
4576+ /* udba work */
4a4d8108 4577+ au_update_ibrange(inode, /*do_put_zero*/1);
1facf9fc 4578+ }
4579+
86dc4139 4580+ isdir = S_ISDIR(inode->i_mode);
1facf9fc 4581+ old_ibstart = au_ibstart(inode);
c2b27bf2 4582+ err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
1facf9fc 4583+ if (unlikely(err))
86dc4139 4584+ goto out_rev;
1facf9fc 4585+ dst_inode = h_dst->d_inode;
4586+ mutex_lock_nested(&dst_inode->i_mutex, AuLsc_I_CHILD2);
86dc4139 4587+ /* todo: necessary? */
c2b27bf2 4588+ /* au_pin_hdir_unlock(cpg->pin); */
1facf9fc 4589+
c2b27bf2 4590+ err = cpup_iattr(cpg->dentry, cpg->bdst, h_src, &a->h_src_attr);
86dc4139
AM
4591+ if (unlikely(err)) {
4592+ /* todo: necessary? */
c2b27bf2 4593+ /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
86dc4139
AM
4594+ mutex_unlock(&dst_inode->i_mutex);
4595+ goto out_rev;
4596+ }
4597+
c2b27bf2 4598+ if (cpg->bdst < old_ibstart) {
86dc4139 4599+ if (S_ISREG(inode->i_mode)) {
c2b27bf2 4600+ err = au_dy_iaop(inode, cpg->bdst, dst_inode);
86dc4139 4601+ if (unlikely(err)) {
c2b27bf2
AM
4602+ /* ignore an error */
4603+ /* au_pin_hdir_relock(cpg->pin); */
86dc4139
AM
4604+ mutex_unlock(&dst_inode->i_mutex);
4605+ goto out_rev;
4a4d8108 4606+ }
4a4d8108 4607+ }
c2b27bf2
AM
4608+ au_set_ibstart(inode, cpg->bdst);
4609+ } else
4610+ au_set_ibend(inode, cpg->bdst);
4611+ au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
86dc4139
AM
4612+ au_hi_flags(inode, isdir));
4613+
4614+ /* todo: necessary? */
c2b27bf2 4615+ /* err = au_pin_hdir_relock(cpg->pin); */
86dc4139
AM
4616+ mutex_unlock(&dst_inode->i_mutex);
4617+ if (unlikely(err))
4618+ goto out_rev;
4619+
4620+ if (!isdir
38d290e6
JR
4621+ && (h_src->d_inode->i_nlink > 1
4622+ || h_src->d_inode->i_state & I_LINKABLE)
86dc4139 4623+ && plink)
c2b27bf2 4624+ au_plink_append(inode, cpg->bdst, h_dst);
86dc4139 4625+
c2b27bf2
AM
4626+ if (au_ftest_cpup(cpg->flags, RENAME)) {
4627+ a->h_path.dentry = h_dst;
392086de 4628+ err = au_do_ren_after_cpup(cpg, &a->h_path);
86dc4139
AM
4629+ }
4630+ if (!err)
c2b27bf2 4631+ goto out_parent; /* success */
1facf9fc 4632+
4633+ /* revert */
4a4d8108 4634+out_rev:
c2b27bf2
AM
4635+ a->h_path.dentry = h_parent;
4636+ au_dtime_store(&a->dt, dst_parent, &a->h_path);
4637+ a->h_path.dentry = h_dst;
86dc4139
AM
4638+ rerr = 0;
4639+ if (h_dst->d_inode) {
523b37e3
AM
4640+ if (!isdir) {
4641+ /* no delegation since it is just created */
4642+ rerr = vfsub_unlink(h_dir, &a->h_path,
4643+ /*delegated*/NULL, /*force*/0);
4644+ } else
c2b27bf2 4645+ rerr = vfsub_rmdir(h_dir, &a->h_path);
86dc4139 4646+ }
c2b27bf2 4647+ au_dtime_revert(&a->dt);
1facf9fc 4648+ if (rerr) {
4649+ AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
4650+ err = -EIO;
4651+ }
c2b27bf2 4652+out_parent:
1facf9fc 4653+ dput(dst_parent);
c2b27bf2
AM
4654+ kfree(a);
4655+out:
1facf9fc 4656+ return err;
4657+}
4658+
c2b27bf2 4659+#if 0 /* unused */
1facf9fc 4660+struct au_cpup_single_args {
4661+ int *errp;
c2b27bf2 4662+ struct au_cp_generic *cpg;
1facf9fc 4663+ struct dentry *dst_parent;
4664+};
4665+
4666+static void au_call_cpup_single(void *args)
4667+{
4668+ struct au_cpup_single_args *a = args;
86dc4139 4669+
c2b27bf2
AM
4670+ au_pin_hdir_acquire_nest(a->cpg->pin);
4671+ *a->errp = au_cpup_single(a->cpg, a->dst_parent);
4672+ au_pin_hdir_release(a->cpg->pin);
1facf9fc 4673+}
c2b27bf2 4674+#endif
1facf9fc 4675+
53392da6
AM
4676+/*
4677+ * prevent SIGXFSZ in copy-up.
4678+ * testing CAP_MKNOD is for generic fs,
4679+ * but CAP_FSETID is for xfs only, currently.
4680+ */
86dc4139 4681+static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
53392da6
AM
4682+{
4683+ int do_sio;
86dc4139
AM
4684+ struct super_block *sb;
4685+ struct inode *h_dir;
53392da6
AM
4686+
4687+ do_sio = 0;
86dc4139 4688+ sb = au_pinned_parent(pin)->d_sb;
53392da6
AM
4689+ if (!au_wkq_test()
4690+ && (!au_sbi(sb)->si_plink_maint_pid
4691+ || au_plink_maint(sb, AuLock_NOPLM))) {
4692+ switch (mode & S_IFMT) {
4693+ case S_IFREG:
4694+ /* no condition about RLIMIT_FSIZE and the file size */
4695+ do_sio = 1;
4696+ break;
4697+ case S_IFCHR:
4698+ case S_IFBLK:
4699+ do_sio = !capable(CAP_MKNOD);
4700+ break;
4701+ }
4702+ if (!do_sio)
4703+ do_sio = ((mode & (S_ISUID | S_ISGID))
4704+ && !capable(CAP_FSETID));
86dc4139
AM
4705+ /* this workaround may be removed in the future */
4706+ if (!do_sio) {
4707+ h_dir = au_pinned_h_dir(pin);
4708+ do_sio = h_dir->i_mode & S_ISVTX;
4709+ }
53392da6
AM
4710+ }
4711+
4712+ return do_sio;
4713+}
4714+
c2b27bf2
AM
4715+#if 0 /* unused */
4716+int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
1facf9fc 4717+{
4718+ int err, wkq_err;
1facf9fc 4719+ struct dentry *h_dentry;
4720+
c2b27bf2 4721+ h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
86dc4139 4722+ if (!au_cpup_sio_test(pin, h_dentry->d_inode->i_mode))
c2b27bf2 4723+ err = au_cpup_single(cpg, dst_parent);
1facf9fc 4724+ else {
4725+ struct au_cpup_single_args args = {
4726+ .errp = &err,
c2b27bf2
AM
4727+ .cpg = cpg,
4728+ .dst_parent = dst_parent
1facf9fc 4729+ };
4730+ wkq_err = au_wkq_wait(au_call_cpup_single, &args);
4731+ if (unlikely(wkq_err))
4732+ err = wkq_err;
4733+ }
4734+
4735+ return err;
4736+}
c2b27bf2 4737+#endif
1facf9fc 4738+
4739+/*
4740+ * copyup the @dentry from the first active lower branch to @bdst,
4741+ * using au_cpup_single().
4742+ */
c2b27bf2 4743+static int au_cpup_simple(struct au_cp_generic *cpg)
1facf9fc 4744+{
4745+ int err;
c2b27bf2
AM
4746+ unsigned int flags_orig;
4747+ struct dentry *dentry;
4748+
4749+ AuDebugOn(cpg->bsrc < 0);
1facf9fc 4750+
c2b27bf2 4751+ dentry = cpg->dentry;
86dc4139 4752+ DiMustWriteLock(dentry);
1facf9fc 4753+
c2b27bf2 4754+ err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
1facf9fc 4755+ if (!err) {
c2b27bf2
AM
4756+ flags_orig = cpg->flags;
4757+ au_fset_cpup(cpg->flags, RENAME);
4758+ err = au_cpup_single(cpg, NULL);
4759+ cpg->flags = flags_orig;
1facf9fc 4760+ if (!err)
4761+ return 0; /* success */
4762+
4763+ /* revert */
c2b27bf2
AM
4764+ au_set_h_dptr(dentry, cpg->bdst, NULL);
4765+ au_set_dbstart(dentry, cpg->bsrc);
1facf9fc 4766+ }
4767+
4768+ return err;
4769+}
4770+
4771+struct au_cpup_simple_args {
4772+ int *errp;
c2b27bf2 4773+ struct au_cp_generic *cpg;
1facf9fc 4774+};
4775+
4776+static void au_call_cpup_simple(void *args)
4777+{
4778+ struct au_cpup_simple_args *a = args;
86dc4139 4779+
c2b27bf2
AM
4780+ au_pin_hdir_acquire_nest(a->cpg->pin);
4781+ *a->errp = au_cpup_simple(a->cpg);
4782+ au_pin_hdir_release(a->cpg->pin);
1facf9fc 4783+}
4784+
c2b27bf2 4785+static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
1facf9fc 4786+{
4787+ int err, wkq_err;
c2b27bf2
AM
4788+ struct dentry *dentry, *parent;
4789+ struct file *h_file;
1facf9fc 4790+ struct inode *h_dir;
4791+
c2b27bf2
AM
4792+ dentry = cpg->dentry;
4793+ h_file = NULL;
4794+ if (au_ftest_cpup(cpg->flags, HOPEN)) {
4795+ AuDebugOn(cpg->bsrc < 0);
392086de 4796+ h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
c2b27bf2
AM
4797+ err = PTR_ERR(h_file);
4798+ if (IS_ERR(h_file))
4799+ goto out;
4800+ }
4801+
1facf9fc 4802+ parent = dget_parent(dentry);
c2b27bf2 4803+ h_dir = au_h_iptr(parent->d_inode, cpg->bdst);
53392da6 4804+ if (!au_test_h_perm_sio(h_dir, MAY_EXEC | MAY_WRITE)
c2b27bf2
AM
4805+ && !au_cpup_sio_test(cpg->pin, dentry->d_inode->i_mode))
4806+ err = au_cpup_simple(cpg);
1facf9fc 4807+ else {
4808+ struct au_cpup_simple_args args = {
4809+ .errp = &err,
c2b27bf2 4810+ .cpg = cpg
1facf9fc 4811+ };
4812+ wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
4813+ if (unlikely(wkq_err))
4814+ err = wkq_err;
4815+ }
4816+
4817+ dput(parent);
c2b27bf2
AM
4818+ if (h_file)
4819+ au_h_open_post(dentry, cpg->bsrc, h_file);
4820+
4821+out:
1facf9fc 4822+ return err;
4823+}
4824+
c2b27bf2 4825+int au_sio_cpup_simple(struct au_cp_generic *cpg)
367653fa 4826+{
c2b27bf2
AM
4827+ aufs_bindex_t bsrc, bend;
4828+ struct dentry *dentry, *h_dentry;
367653fa 4829+
c2b27bf2
AM
4830+ if (cpg->bsrc < 0) {
4831+ dentry = cpg->dentry;
4832+ bend = au_dbend(dentry);
4833+ for (bsrc = cpg->bdst + 1; bsrc <= bend; bsrc++) {
4834+ h_dentry = au_h_dptr(dentry, bsrc);
4835+ if (h_dentry) {
4836+ AuDebugOn(!h_dentry->d_inode);
4837+ break;
4838+ }
4839+ }
4840+ AuDebugOn(bsrc > bend);
4841+ cpg->bsrc = bsrc;
367653fa 4842+ }
c2b27bf2
AM
4843+ AuDebugOn(cpg->bsrc <= cpg->bdst);
4844+ return au_do_sio_cpup_simple(cpg);
4845+}
367653fa 4846+
c2b27bf2
AM
4847+int au_sio_cpdown_simple(struct au_cp_generic *cpg)
4848+{
4849+ AuDebugOn(cpg->bdst <= cpg->bsrc);
4850+ return au_do_sio_cpup_simple(cpg);
367653fa
AM
4851+}
4852+
1facf9fc 4853+/* ---------------------------------------------------------------------- */
4854+
4855+/*
4856+ * copyup the deleted file for writing.
4857+ */
c2b27bf2
AM
4858+static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
4859+ struct file *file)
1facf9fc 4860+{
4861+ int err;
c2b27bf2
AM
4862+ unsigned int flags_orig;
4863+ aufs_bindex_t bsrc_orig;
1facf9fc 4864+ struct dentry *h_d_dst, *h_d_start;
c2b27bf2 4865+ struct au_dinfo *dinfo;
4a4d8108 4866+ struct au_hdentry *hdp;
1facf9fc 4867+
c2b27bf2 4868+ dinfo = au_di(cpg->dentry);
1308ab2a 4869+ AuRwMustWriteLock(&dinfo->di_rwsem);
4870+
c2b27bf2
AM
4871+ bsrc_orig = cpg->bsrc;
4872+ cpg->bsrc = dinfo->di_bstart;
4a4d8108 4873+ hdp = dinfo->di_hdentry;
c2b27bf2
AM
4874+ h_d_dst = hdp[0 + cpg->bdst].hd_dentry;
4875+ dinfo->di_bstart = cpg->bdst;
4876+ hdp[0 + cpg->bdst].hd_dentry = wh_dentry;
86dc4139 4877+ h_d_start = NULL;
027c5e7a 4878+ if (file) {
c2b27bf2
AM
4879+ h_d_start = hdp[0 + cpg->bsrc].hd_dentry;
4880+ hdp[0 + cpg->bsrc].hd_dentry = au_hf_top(file)->f_dentry;
027c5e7a 4881+ }
c2b27bf2
AM
4882+ flags_orig = cpg->flags;
4883+ cpg->flags = !AuCpup_DTIME;
4884+ err = au_cpup_single(cpg, /*h_parent*/NULL);
4885+ cpg->flags = flags_orig;
027c5e7a
AM
4886+ if (file) {
4887+ if (!err)
4888+ err = au_reopen_nondir(file);
c2b27bf2 4889+ hdp[0 + cpg->bsrc].hd_dentry = h_d_start;
1facf9fc 4890+ }
c2b27bf2
AM
4891+ hdp[0 + cpg->bdst].hd_dentry = h_d_dst;
4892+ dinfo->di_bstart = cpg->bsrc;
4893+ cpg->bsrc = bsrc_orig;
1facf9fc 4894+
4895+ return err;
4896+}
4897+
c2b27bf2 4898+static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
1facf9fc 4899+{
4900+ int err;
c2b27bf2 4901+ aufs_bindex_t bdst;
1facf9fc 4902+ struct au_dtime dt;
c2b27bf2 4903+ struct dentry *dentry, *parent, *h_parent, *wh_dentry;
1facf9fc 4904+ struct au_branch *br;
4905+ struct path h_path;
4906+
c2b27bf2
AM
4907+ dentry = cpg->dentry;
4908+ bdst = cpg->bdst;
1facf9fc 4909+ br = au_sbr(dentry->d_sb, bdst);
4910+ parent = dget_parent(dentry);
4911+ h_parent = au_h_dptr(parent, bdst);
4912+ wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
4913+ err = PTR_ERR(wh_dentry);
4914+ if (IS_ERR(wh_dentry))
4915+ goto out;
4916+
4917+ h_path.dentry = h_parent;
86dc4139 4918+ h_path.mnt = au_br_mnt(br);
1facf9fc 4919+ au_dtime_store(&dt, parent, &h_path);
c2b27bf2 4920+ err = au_do_cpup_wh(cpg, wh_dentry, file);
1facf9fc 4921+ if (unlikely(err))
4922+ goto out_wh;
4923+
4924+ dget(wh_dentry);
4925+ h_path.dentry = wh_dentry;
523b37e3
AM
4926+ if (!S_ISDIR(wh_dentry->d_inode->i_mode)) {
4927+ /* no delegation since it is just created */
4928+ err = vfsub_unlink(h_parent->d_inode, &h_path,
4929+ /*delegated*/NULL, /*force*/0);
4930+ } else
4a4d8108 4931+ err = vfsub_rmdir(h_parent->d_inode, &h_path);
1facf9fc 4932+ if (unlikely(err)) {
523b37e3
AM
4933+ AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
4934+ wh_dentry, err);
1facf9fc 4935+ err = -EIO;
4936+ }
4937+ au_dtime_revert(&dt);
4938+ au_set_hi_wh(dentry->d_inode, bdst, wh_dentry);
4939+
4f0767ce 4940+out_wh:
1facf9fc 4941+ dput(wh_dentry);
4f0767ce 4942+out:
1facf9fc 4943+ dput(parent);
4944+ return err;
4945+}
4946+
4947+struct au_cpup_wh_args {
4948+ int *errp;
c2b27bf2 4949+ struct au_cp_generic *cpg;
1facf9fc 4950+ struct file *file;
4951+};
4952+
4953+static void au_call_cpup_wh(void *args)
4954+{
4955+ struct au_cpup_wh_args *a = args;
86dc4139 4956+
c2b27bf2
AM
4957+ au_pin_hdir_acquire_nest(a->cpg->pin);
4958+ *a->errp = au_cpup_wh(a->cpg, a->file);
4959+ au_pin_hdir_release(a->cpg->pin);
1facf9fc 4960+}
4961+
c2b27bf2 4962+int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
1facf9fc 4963+{
4964+ int err, wkq_err;
c2b27bf2
AM
4965+ aufs_bindex_t bdst;
4966+ struct dentry *dentry, *parent, *h_orph, *h_parent, *h_dentry;
86dc4139 4967+ struct inode *dir, *h_dir, *h_tmpdir;
1facf9fc 4968+ struct au_wbr *wbr;
c2b27bf2 4969+ struct au_pin wh_pin, *pin_orig;
1facf9fc 4970+
c2b27bf2
AM
4971+ dentry = cpg->dentry;
4972+ bdst = cpg->bdst;
1facf9fc 4973+ parent = dget_parent(dentry);
4974+ dir = parent->d_inode;
4975+ h_orph = NULL;
4976+ h_parent = NULL;
4977+ h_dir = au_igrab(au_h_iptr(dir, bdst));
4978+ h_tmpdir = h_dir;
c2b27bf2 4979+ pin_orig = NULL;
1facf9fc 4980+ if (!h_dir->i_nlink) {
4981+ wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
4982+ h_orph = wbr->wbr_orph;
4983+
4984+ h_parent = dget(au_h_dptr(parent, bdst));
1facf9fc 4985+ au_set_h_dptr(parent, bdst, dget(h_orph));
4986+ h_tmpdir = h_orph->d_inode;
1facf9fc 4987+ au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
4988+
1facf9fc 4989+ if (file)
4a4d8108 4990+ h_dentry = au_hf_top(file)->f_dentry;
1facf9fc 4991+ else
4992+ h_dentry = au_h_dptr(dentry, au_dbstart(dentry));
dece6358 4993+ mutex_lock_nested(&h_tmpdir->i_mutex, AuLsc_I_PARENT3);
4a4d8108 4994+ /* todo: au_h_open_pre()? */
86dc4139 4995+
c2b27bf2 4996+ pin_orig = cpg->pin;
86dc4139 4997+ au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
c2b27bf2
AM
4998+ AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
4999+ cpg->pin = &wh_pin;
1facf9fc 5000+ }
5001+
53392da6 5002+ if (!au_test_h_perm_sio(h_tmpdir, MAY_EXEC | MAY_WRITE)
c2b27bf2
AM
5003+ && !au_cpup_sio_test(cpg->pin, dentry->d_inode->i_mode))
5004+ err = au_cpup_wh(cpg, file);
1facf9fc 5005+ else {
5006+ struct au_cpup_wh_args args = {
5007+ .errp = &err,
c2b27bf2
AM
5008+ .cpg = cpg,
5009+ .file = file
1facf9fc 5010+ };
5011+ wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
5012+ if (unlikely(wkq_err))
5013+ err = wkq_err;
5014+ }
5015+
5016+ if (h_orph) {
5017+ mutex_unlock(&h_tmpdir->i_mutex);
4a4d8108 5018+ /* todo: au_h_open_post()? */
1facf9fc 5019+ au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
1facf9fc 5020+ au_set_h_dptr(parent, bdst, h_parent);
c2b27bf2
AM
5021+ AuDebugOn(!pin_orig);
5022+ cpg->pin = pin_orig;
1facf9fc 5023+ }
5024+ iput(h_dir);
5025+ dput(parent);
5026+
5027+ return err;
5028+}
5029+
5030+/* ---------------------------------------------------------------------- */
5031+
5032+/*
5033+ * generic routine for both of copy-up and copy-down.
5034+ */
5035+/* cf. revalidate function in file.c */
5036+int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5037+ int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 5038+ struct au_pin *pin,
1facf9fc 5039+ struct dentry *h_parent, void *arg),
5040+ void *arg)
5041+{
5042+ int err;
5043+ struct au_pin pin;
5044+ struct dentry *d, *parent, *h_parent, *real_parent;
5045+
5046+ err = 0;
5047+ parent = dget_parent(dentry);
5048+ if (IS_ROOT(parent))
5049+ goto out;
5050+
5051+ au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
5052+ au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
5053+
5054+ /* do not use au_dpage */
5055+ real_parent = parent;
5056+ while (1) {
5057+ dput(parent);
5058+ parent = dget_parent(dentry);
5059+ h_parent = au_h_dptr(parent, bdst);
5060+ if (h_parent)
5061+ goto out; /* success */
5062+
5063+ /* find top dir which is necessary to cpup */
5064+ do {
5065+ d = parent;
5066+ dput(parent);
5067+ parent = dget_parent(d);
5068+ di_read_lock_parent3(parent, !AuLock_IR);
5069+ h_parent = au_h_dptr(parent, bdst);
5070+ di_read_unlock(parent, !AuLock_IR);
5071+ } while (!h_parent);
5072+
5073+ if (d != real_parent)
5074+ di_write_lock_child3(d);
5075+
5076+ /* somebody else might create while we were sleeping */
5077+ if (!au_h_dptr(d, bdst) || !au_h_dptr(d, bdst)->d_inode) {
5078+ if (au_h_dptr(d, bdst))
5079+ au_update_dbstart(d);
5080+
5081+ au_pin_set_dentry(&pin, d);
5082+ err = au_do_pin(&pin);
5083+ if (!err) {
86dc4139 5084+ err = cp(d, bdst, &pin, h_parent, arg);
1facf9fc 5085+ au_unpin(&pin);
5086+ }
5087+ }
5088+
5089+ if (d != real_parent)
5090+ di_write_unlock(d);
5091+ if (unlikely(err))
5092+ break;
5093+ }
5094+
4f0767ce 5095+out:
1facf9fc 5096+ dput(parent);
5097+ return err;
5098+}
5099+
5100+static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 5101+ struct au_pin *pin,
1facf9fc 5102+ struct dentry *h_parent __maybe_unused ,
5103+ void *arg __maybe_unused)
5104+{
c2b27bf2
AM
5105+ struct au_cp_generic cpg = {
5106+ .dentry = dentry,
5107+ .bdst = bdst,
5108+ .bsrc = -1,
5109+ .len = 0,
5110+ .pin = pin,
5111+ .flags = AuCpup_DTIME
5112+ };
5113+ return au_sio_cpup_simple(&cpg);
1facf9fc 5114+}
5115+
5116+int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5117+{
5118+ return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
5119+}
5120+
5121+int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5122+{
5123+ int err;
5124+ struct dentry *parent;
5125+ struct inode *dir;
5126+
5127+ parent = dget_parent(dentry);
5128+ dir = parent->d_inode;
5129+ err = 0;
5130+ if (au_h_iptr(dir, bdst))
5131+ goto out;
5132+
5133+ di_read_unlock(parent, AuLock_IR);
5134+ di_write_lock_parent(parent);
5135+ /* someone else might change our inode while we were sleeping */
5136+ if (!au_h_iptr(dir, bdst))
5137+ err = au_cpup_dirs(dentry, bdst);
5138+ di_downgrade_lock(parent, AuLock_IR);
5139+
4f0767ce 5140+out:
1facf9fc 5141+ dput(parent);
5142+ return err;
5143+}
7f207e10
AM
5144diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
5145--- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 5146+++ linux/fs/aufs/cpup.h 2014-06-24 00:14:06.779385790 +0200
523b37e3 5147@@ -0,0 +1,94 @@
1facf9fc 5148+/*
523b37e3 5149+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 5150+ *
5151+ * This program, aufs is free software; you can redistribute it and/or modify
5152+ * it under the terms of the GNU General Public License as published by
5153+ * the Free Software Foundation; either version 2 of the License, or
5154+ * (at your option) any later version.
dece6358
AM
5155+ *
5156+ * This program is distributed in the hope that it will be useful,
5157+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5158+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5159+ * GNU General Public License for more details.
5160+ *
5161+ * You should have received a copy of the GNU General Public License
523b37e3 5162+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5163+ */
5164+
5165+/*
5166+ * copy-up/down functions
5167+ */
5168+
5169+#ifndef __AUFS_CPUP_H__
5170+#define __AUFS_CPUP_H__
5171+
5172+#ifdef __KERNEL__
5173+
dece6358 5174+#include <linux/path.h>
1facf9fc 5175+
dece6358
AM
5176+struct inode;
5177+struct file;
86dc4139 5178+struct au_pin;
dece6358 5179+
86dc4139 5180+void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
1facf9fc 5181+void au_cpup_attr_timesizes(struct inode *inode);
5182+void au_cpup_attr_nlink(struct inode *inode, int force);
5183+void au_cpup_attr_changeable(struct inode *inode);
5184+void au_cpup_igen(struct inode *inode, struct inode *h_inode);
5185+void au_cpup_attr_all(struct inode *inode, int force);
5186+
5187+/* ---------------------------------------------------------------------- */
5188+
c2b27bf2
AM
5189+struct au_cp_generic {
5190+ struct dentry *dentry;
5191+ aufs_bindex_t bdst, bsrc;
5192+ loff_t len;
5193+ struct au_pin *pin;
5194+ unsigned int flags;
5195+};
5196+
1facf9fc 5197+/* cpup flags */
392086de
AM
5198+#define AuCpup_DTIME 1 /* do dtime_store/revert */
5199+#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
5200+ for link(2) */
5201+#define AuCpup_RENAME (1 << 2) /* rename after cpup */
5202+#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
5203+ cpup */
5204+#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
5205+ existing entry */
5206+#define AuCpup_RWDST (1 << 5) /* force write target even if
5207+ the branch is marked as RO */
c2b27bf2 5208+
1facf9fc 5209+#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
7f207e10
AM
5210+#define au_fset_cpup(flags, name) \
5211+ do { (flags) |= AuCpup_##name; } while (0)
5212+#define au_fclr_cpup(flags, name) \
5213+ do { (flags) &= ~AuCpup_##name; } while (0)
1facf9fc 5214+
5215+int au_copy_file(struct file *dst, struct file *src, loff_t len);
c2b27bf2
AM
5216+int au_sio_cpup_simple(struct au_cp_generic *cpg);
5217+int au_sio_cpdown_simple(struct au_cp_generic *cpg);
5218+int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
1facf9fc 5219+
5220+int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5221+ int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 5222+ struct au_pin *pin,
1facf9fc 5223+ struct dentry *h_parent, void *arg),
5224+ void *arg);
5225+int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5226+int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5227+
5228+/* ---------------------------------------------------------------------- */
5229+
5230+/* keep timestamps when copyup */
5231+struct au_dtime {
5232+ struct dentry *dt_dentry;
5233+ struct path dt_h_path;
5234+ struct timespec dt_atime, dt_mtime;
5235+};
5236+void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
5237+ struct path *h_path);
5238+void au_dtime_revert(struct au_dtime *dt);
5239+
5240+#endif /* __KERNEL__ */
5241+#endif /* __AUFS_CPUP_H__ */
7f207e10
AM
5242diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
5243--- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 5244+++ linux/fs/aufs/dbgaufs.c 2014-06-24 00:14:06.779385790 +0200
523b37e3 5245@@ -0,0 +1,432 @@
1facf9fc 5246+/*
523b37e3 5247+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 5248+ *
5249+ * This program, aufs is free software; you can redistribute it and/or modify
5250+ * it under the terms of the GNU General Public License as published by
5251+ * the Free Software Foundation; either version 2 of the License, or
5252+ * (at your option) any later version.
dece6358
AM
5253+ *
5254+ * This program is distributed in the hope that it will be useful,
5255+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5256+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5257+ * GNU General Public License for more details.
5258+ *
5259+ * You should have received a copy of the GNU General Public License
523b37e3 5260+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5261+ */
5262+
5263+/*
5264+ * debugfs interface
5265+ */
5266+
5267+#include <linux/debugfs.h>
5268+#include "aufs.h"
5269+
5270+#ifndef CONFIG_SYSFS
5271+#error DEBUG_FS depends upon SYSFS
5272+#endif
5273+
5274+static struct dentry *dbgaufs;
5275+static const mode_t dbgaufs_mode = S_IRUSR | S_IRGRP | S_IROTH;
5276+
5277+/* 20 is max digits length of ulong 64 */
5278+struct dbgaufs_arg {
5279+ int n;
5280+ char a[20 * 4];
5281+};
5282+
5283+/*
5284+ * common function for all XINO files
5285+ */
5286+static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
5287+ struct file *file)
5288+{
5289+ kfree(file->private_data);
5290+ return 0;
5291+}
5292+
5293+static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt)
5294+{
5295+ int err;
5296+ struct kstat st;
5297+ struct dbgaufs_arg *p;
5298+
5299+ err = -ENOMEM;
5300+ p = kmalloc(sizeof(*p), GFP_NOFS);
5301+ if (unlikely(!p))
5302+ goto out;
5303+
5304+ err = 0;
5305+ p->n = 0;
5306+ file->private_data = p;
5307+ if (!xf)
5308+ goto out;
5309+
c06a8ce3 5310+ err = vfs_getattr(&xf->f_path, &st);
1facf9fc 5311+ if (!err) {
5312+ if (do_fcnt)
5313+ p->n = snprintf
5314+ (p->a, sizeof(p->a), "%ld, %llux%lu %lld\n",
5315+ (long)file_count(xf), st.blocks, st.blksize,
5316+ (long long)st.size);
5317+ else
5318+ p->n = snprintf(p->a, sizeof(p->a), "%llux%lu %lld\n",
5319+ st.blocks, st.blksize,
5320+ (long long)st.size);
5321+ AuDebugOn(p->n >= sizeof(p->a));
5322+ } else {
5323+ p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
5324+ err = 0;
5325+ }
5326+
4f0767ce 5327+out:
1facf9fc 5328+ return err;
5329+
5330+}
5331+
5332+static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
5333+ size_t count, loff_t *ppos)
5334+{
5335+ struct dbgaufs_arg *p;
5336+
5337+ p = file->private_data;
5338+ return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
5339+}
5340+
5341+/* ---------------------------------------------------------------------- */
5342+
86dc4139
AM
5343+struct dbgaufs_plink_arg {
5344+ int n;
5345+ char a[];
5346+};
5347+
5348+static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
5349+ struct file *file)
5350+{
5351+ free_page((unsigned long)file->private_data);
5352+ return 0;
5353+}
5354+
5355+static int dbgaufs_plink_open(struct inode *inode, struct file *file)
5356+{
5357+ int err, i, limit;
5358+ unsigned long n, sum;
5359+ struct dbgaufs_plink_arg *p;
5360+ struct au_sbinfo *sbinfo;
5361+ struct super_block *sb;
5362+ struct au_sphlhead *sphl;
5363+
5364+ err = -ENOMEM;
5365+ p = (void *)get_zeroed_page(GFP_NOFS);
5366+ if (unlikely(!p))
5367+ goto out;
5368+
5369+ err = -EFBIG;
5370+ sbinfo = inode->i_private;
5371+ sb = sbinfo->si_sb;
5372+ si_noflush_read_lock(sb);
5373+ if (au_opt_test(au_mntflags(sb), PLINK)) {
5374+ limit = PAGE_SIZE - sizeof(p->n);
5375+
5376+ /* the number of buckets */
5377+ n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
5378+ p->n += n;
5379+ limit -= n;
5380+
5381+ sum = 0;
5382+ for (i = 0, sphl = sbinfo->si_plink;
5383+ i < AuPlink_NHASH;
5384+ i++, sphl++) {
5385+ n = au_sphl_count(sphl);
5386+ sum += n;
5387+
5388+ n = snprintf(p->a + p->n, limit, "%lu ", n);
5389+ p->n += n;
5390+ limit -= n;
5391+ if (unlikely(limit <= 0))
5392+ goto out_free;
5393+ }
5394+ p->a[p->n - 1] = '\n';
5395+
5396+ /* the sum of plinks */
5397+ n = snprintf(p->a + p->n, limit, "%lu\n", sum);
5398+ p->n += n;
5399+ limit -= n;
5400+ if (unlikely(limit <= 0))
5401+ goto out_free;
5402+ } else {
5403+#define str "1\n0\n0\n"
5404+ p->n = sizeof(str) - 1;
5405+ strcpy(p->a, str);
5406+#undef str
5407+ }
5408+ si_read_unlock(sb);
5409+
5410+ err = 0;
5411+ file->private_data = p;
5412+ goto out; /* success */
5413+
5414+out_free:
5415+ free_page((unsigned long)p);
5416+out:
5417+ return err;
5418+}
5419+
5420+static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
5421+ size_t count, loff_t *ppos)
5422+{
5423+ struct dbgaufs_plink_arg *p;
5424+
5425+ p = file->private_data;
5426+ return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
5427+}
5428+
5429+static const struct file_operations dbgaufs_plink_fop = {
5430+ .owner = THIS_MODULE,
5431+ .open = dbgaufs_plink_open,
5432+ .release = dbgaufs_plink_release,
5433+ .read = dbgaufs_plink_read
5434+};
5435+
5436+/* ---------------------------------------------------------------------- */
5437+
1facf9fc 5438+static int dbgaufs_xib_open(struct inode *inode, struct file *file)
5439+{
5440+ int err;
5441+ struct au_sbinfo *sbinfo;
5442+ struct super_block *sb;
5443+
5444+ sbinfo = inode->i_private;
5445+ sb = sbinfo->si_sb;
5446+ si_noflush_read_lock(sb);
5447+ err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0);
5448+ si_read_unlock(sb);
5449+ return err;
5450+}
5451+
5452+static const struct file_operations dbgaufs_xib_fop = {
4a4d8108 5453+ .owner = THIS_MODULE,
1facf9fc 5454+ .open = dbgaufs_xib_open,
5455+ .release = dbgaufs_xi_release,
5456+ .read = dbgaufs_xi_read
5457+};
5458+
5459+/* ---------------------------------------------------------------------- */
5460+
5461+#define DbgaufsXi_PREFIX "xi"
5462+
5463+static int dbgaufs_xino_open(struct inode *inode, struct file *file)
5464+{
5465+ int err;
5466+ long l;
5467+ struct au_sbinfo *sbinfo;
5468+ struct super_block *sb;
5469+ struct file *xf;
5470+ struct qstr *name;
5471+
5472+ err = -ENOENT;
5473+ xf = NULL;
5474+ name = &file->f_dentry->d_name;
5475+ if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
5476+ || memcmp(name->name, DbgaufsXi_PREFIX,
5477+ sizeof(DbgaufsXi_PREFIX) - 1)))
5478+ goto out;
9dbd164d 5479+ err = kstrtol(name->name + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
1facf9fc 5480+ if (unlikely(err))
5481+ goto out;
5482+
5483+ sbinfo = inode->i_private;
5484+ sb = sbinfo->si_sb;
5485+ si_noflush_read_lock(sb);
5486+ if (l <= au_sbend(sb)) {
5487+ xf = au_sbr(sb, (aufs_bindex_t)l)->br_xino.xi_file;
5488+ err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1);
5489+ } else
5490+ err = -ENOENT;
5491+ si_read_unlock(sb);
5492+
4f0767ce 5493+out:
1facf9fc 5494+ return err;
5495+}
5496+
5497+static const struct file_operations dbgaufs_xino_fop = {
4a4d8108 5498+ .owner = THIS_MODULE,
1facf9fc 5499+ .open = dbgaufs_xino_open,
5500+ .release = dbgaufs_xi_release,
5501+ .read = dbgaufs_xi_read
5502+};
5503+
5504+void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
5505+{
5506+ aufs_bindex_t bend;
5507+ struct au_branch *br;
5508+ struct au_xino_file *xi;
5509+
5510+ if (!au_sbi(sb)->si_dbgaufs)
5511+ return;
5512+
5513+ bend = au_sbend(sb);
5514+ for (; bindex <= bend; bindex++) {
5515+ br = au_sbr(sb, bindex);
5516+ xi = &br->br_xino;
c06a8ce3
AM
5517+ debugfs_remove(xi->xi_dbgaufs);
5518+ xi->xi_dbgaufs = NULL;
1facf9fc 5519+ }
5520+}
5521+
5522+void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
5523+{
5524+ struct au_sbinfo *sbinfo;
5525+ struct dentry *parent;
5526+ struct au_branch *br;
5527+ struct au_xino_file *xi;
5528+ aufs_bindex_t bend;
5529+ char name[sizeof(DbgaufsXi_PREFIX) + 5]; /* "xi" bindex NULL */
5530+
5531+ sbinfo = au_sbi(sb);
5532+ parent = sbinfo->si_dbgaufs;
5533+ if (!parent)
5534+ return;
5535+
5536+ bend = au_sbend(sb);
5537+ for (; bindex <= bend; bindex++) {
5538+ snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
5539+ br = au_sbr(sb, bindex);
5540+ xi = &br->br_xino;
5541+ AuDebugOn(xi->xi_dbgaufs);
5542+ xi->xi_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
5543+ sbinfo, &dbgaufs_xino_fop);
5544+ /* ignore an error */
5545+ if (unlikely(!xi->xi_dbgaufs))
5546+ AuWarn1("failed %s under debugfs\n", name);
5547+ }
5548+}
5549+
5550+/* ---------------------------------------------------------------------- */
5551+
5552+#ifdef CONFIG_AUFS_EXPORT
5553+static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
5554+{
5555+ int err;
5556+ struct au_sbinfo *sbinfo;
5557+ struct super_block *sb;
5558+
5559+ sbinfo = inode->i_private;
5560+ sb = sbinfo->si_sb;
5561+ si_noflush_read_lock(sb);
5562+ err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0);
5563+ si_read_unlock(sb);
5564+ return err;
5565+}
5566+
5567+static const struct file_operations dbgaufs_xigen_fop = {
4a4d8108 5568+ .owner = THIS_MODULE,
1facf9fc 5569+ .open = dbgaufs_xigen_open,
5570+ .release = dbgaufs_xi_release,
5571+ .read = dbgaufs_xi_read
5572+};
5573+
5574+static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
5575+{
5576+ int err;
5577+
dece6358
AM
5578+ /*
5579+ * This function is a dynamic '__init' fucntion actually,
5580+ * so the tiny check for si_rwsem is unnecessary.
5581+ */
5582+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
5583+
1facf9fc 5584+ err = -EIO;
5585+ sbinfo->si_dbgaufs_xigen = debugfs_create_file
5586+ ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
5587+ &dbgaufs_xigen_fop);
5588+ if (sbinfo->si_dbgaufs_xigen)
5589+ err = 0;
5590+
5591+ return err;
5592+}
5593+#else
5594+static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
5595+{
5596+ return 0;
5597+}
5598+#endif /* CONFIG_AUFS_EXPORT */
5599+
5600+/* ---------------------------------------------------------------------- */
5601+
5602+void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
5603+{
dece6358
AM
5604+ /*
5605+ * This function is a dynamic '__init' fucntion actually,
5606+ * so the tiny check for si_rwsem is unnecessary.
5607+ */
5608+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
5609+
1facf9fc 5610+ debugfs_remove_recursive(sbinfo->si_dbgaufs);
5611+ sbinfo->si_dbgaufs = NULL;
5612+ kobject_put(&sbinfo->si_kobj);
5613+}
5614+
5615+int dbgaufs_si_init(struct au_sbinfo *sbinfo)
5616+{
5617+ int err;
5618+ char name[SysaufsSiNameLen];
5619+
dece6358
AM
5620+ /*
5621+ * This function is a dynamic '__init' fucntion actually,
5622+ * so the tiny check for si_rwsem is unnecessary.
5623+ */
5624+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
5625+
1facf9fc 5626+ err = -ENOENT;
5627+ if (!dbgaufs) {
5628+ AuErr1("/debug/aufs is uninitialized\n");
5629+ goto out;
5630+ }
5631+
5632+ err = -EIO;
5633+ sysaufs_name(sbinfo, name);
5634+ sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
5635+ if (unlikely(!sbinfo->si_dbgaufs))
5636+ goto out;
5637+ kobject_get(&sbinfo->si_kobj);
5638+
5639+ sbinfo->si_dbgaufs_xib = debugfs_create_file
5640+ ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
5641+ &dbgaufs_xib_fop);
5642+ if (unlikely(!sbinfo->si_dbgaufs_xib))
5643+ goto out_dir;
5644+
86dc4139
AM
5645+ sbinfo->si_dbgaufs_plink = debugfs_create_file
5646+ ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
5647+ &dbgaufs_plink_fop);
5648+ if (unlikely(!sbinfo->si_dbgaufs_plink))
5649+ goto out_dir;
5650+
1facf9fc 5651+ err = dbgaufs_xigen_init(sbinfo);
5652+ if (!err)
5653+ goto out; /* success */
5654+
4f0767ce 5655+out_dir:
1facf9fc 5656+ dbgaufs_si_fin(sbinfo);
4f0767ce 5657+out:
1facf9fc 5658+ return err;
5659+}
5660+
5661+/* ---------------------------------------------------------------------- */
5662+
5663+void dbgaufs_fin(void)
5664+{
5665+ debugfs_remove(dbgaufs);
5666+}
5667+
5668+int __init dbgaufs_init(void)
5669+{
5670+ int err;
5671+
5672+ err = -EIO;
5673+ dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
5674+ if (dbgaufs)
5675+ err = 0;
5676+ return err;
5677+}
7f207e10
AM
5678diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
5679--- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 5680+++ linux/fs/aufs/dbgaufs.h 2014-06-24 00:14:06.779385790 +0200
523b37e3 5681@@ -0,0 +1,48 @@
1facf9fc 5682+/*
523b37e3 5683+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 5684+ *
5685+ * This program, aufs is free software; you can redistribute it and/or modify
5686+ * it under the terms of the GNU General Public License as published by
5687+ * the Free Software Foundation; either version 2 of the License, or
5688+ * (at your option) any later version.
dece6358
AM
5689+ *
5690+ * This program is distributed in the hope that it will be useful,
5691+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5692+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5693+ * GNU General Public License for more details.
5694+ *
5695+ * You should have received a copy of the GNU General Public License
523b37e3 5696+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5697+ */
5698+
5699+/*
5700+ * debugfs interface
5701+ */
5702+
5703+#ifndef __DBGAUFS_H__
5704+#define __DBGAUFS_H__
5705+
5706+#ifdef __KERNEL__
5707+
dece6358 5708+struct super_block;
1facf9fc 5709+struct au_sbinfo;
dece6358 5710+
1facf9fc 5711+#ifdef CONFIG_DEBUG_FS
5712+/* dbgaufs.c */
5713+void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
5714+void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
5715+void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
5716+int dbgaufs_si_init(struct au_sbinfo *sbinfo);
5717+void dbgaufs_fin(void);
5718+int __init dbgaufs_init(void);
1facf9fc 5719+#else
4a4d8108
AM
5720+AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
5721+AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
5722+AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
5723+AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
5724+AuStubVoid(dbgaufs_fin, void)
5725+AuStubInt0(__init dbgaufs_init, void)
1facf9fc 5726+#endif /* CONFIG_DEBUG_FS */
5727+
5728+#endif /* __KERNEL__ */
5729+#endif /* __DBGAUFS_H__ */
7f207e10
AM
5730diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
5731--- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 5732+++ linux/fs/aufs/dcsub.c 2014-06-24 00:14:06.779385790 +0200
027c5e7a 5733@@ -0,0 +1,243 @@
1facf9fc 5734+/*
523b37e3 5735+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 5736+ *
5737+ * This program, aufs is free software; you can redistribute it and/or modify
5738+ * it under the terms of the GNU General Public License as published by
5739+ * the Free Software Foundation; either version 2 of the License, or
5740+ * (at your option) any later version.
dece6358
AM
5741+ *
5742+ * This program is distributed in the hope that it will be useful,
5743+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5744+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5745+ * GNU General Public License for more details.
5746+ *
5747+ * You should have received a copy of the GNU General Public License
523b37e3 5748+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5749+ */
5750+
5751+/*
5752+ * sub-routines for dentry cache
5753+ */
5754+
5755+#include "aufs.h"
5756+
5757+static void au_dpage_free(struct au_dpage *dpage)
5758+{
5759+ int i;
5760+ struct dentry **p;
5761+
5762+ p = dpage->dentries;
5763+ for (i = 0; i < dpage->ndentry; i++)
5764+ dput(*p++);
5765+ free_page((unsigned long)dpage->dentries);
5766+}
5767+
5768+int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
5769+{
5770+ int err;
5771+ void *p;
5772+
5773+ err = -ENOMEM;
5774+ dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
5775+ if (unlikely(!dpages->dpages))
5776+ goto out;
5777+
5778+ p = (void *)__get_free_page(gfp);
5779+ if (unlikely(!p))
5780+ goto out_dpages;
5781+
5782+ dpages->dpages[0].ndentry = 0;
5783+ dpages->dpages[0].dentries = p;
5784+ dpages->ndpage = 1;
5785+ return 0; /* success */
5786+
4f0767ce 5787+out_dpages:
1facf9fc 5788+ kfree(dpages->dpages);
4f0767ce 5789+out:
1facf9fc 5790+ return err;
5791+}
5792+
5793+void au_dpages_free(struct au_dcsub_pages *dpages)
5794+{
5795+ int i;
5796+ struct au_dpage *p;
5797+
5798+ p = dpages->dpages;
5799+ for (i = 0; i < dpages->ndpage; i++)
5800+ au_dpage_free(p++);
5801+ kfree(dpages->dpages);
5802+}
5803+
5804+static int au_dpages_append(struct au_dcsub_pages *dpages,
5805+ struct dentry *dentry, gfp_t gfp)
5806+{
5807+ int err, sz;
5808+ struct au_dpage *dpage;
5809+ void *p;
5810+
5811+ dpage = dpages->dpages + dpages->ndpage - 1;
5812+ sz = PAGE_SIZE / sizeof(dentry);
5813+ if (unlikely(dpage->ndentry >= sz)) {
5814+ AuLabel(new dpage);
5815+ err = -ENOMEM;
5816+ sz = dpages->ndpage * sizeof(*dpages->dpages);
5817+ p = au_kzrealloc(dpages->dpages, sz,
5818+ sz + sizeof(*dpages->dpages), gfp);
5819+ if (unlikely(!p))
5820+ goto out;
5821+
5822+ dpages->dpages = p;
5823+ dpage = dpages->dpages + dpages->ndpage;
5824+ p = (void *)__get_free_page(gfp);
5825+ if (unlikely(!p))
5826+ goto out;
5827+
5828+ dpage->ndentry = 0;
5829+ dpage->dentries = p;
5830+ dpages->ndpage++;
5831+ }
5832+
392086de 5833+ AuDebugOn(!d_count(dentry));
027c5e7a 5834+ dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
1facf9fc 5835+ return 0; /* success */
5836+
4f0767ce 5837+out:
1facf9fc 5838+ return err;
5839+}
5840+
523b37e3 5841+/* try d_walk() in linux/fs/dcache.c */
1facf9fc 5842+int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
5843+ au_dpages_test test, void *arg)
5844+{
5845+ int err;
027c5e7a 5846+ struct dentry *this_parent;
1facf9fc 5847+ struct list_head *next;
5848+ struct super_block *sb = root->d_sb;
5849+
5850+ err = 0;
027c5e7a
AM
5851+ write_seqlock(&rename_lock);
5852+ this_parent = root;
5853+ spin_lock(&this_parent->d_lock);
4f0767ce 5854+repeat:
1facf9fc 5855+ next = this_parent->d_subdirs.next;
4f0767ce 5856+resume:
1facf9fc 5857+ if (this_parent->d_sb == sb
5858+ && !IS_ROOT(this_parent)
027c5e7a 5859+ && au_di(this_parent)
392086de 5860+ && d_count(this_parent)
1facf9fc 5861+ && (!test || test(this_parent, arg))) {
5862+ err = au_dpages_append(dpages, this_parent, GFP_ATOMIC);
5863+ if (unlikely(err))
5864+ goto out;
5865+ }
5866+
5867+ while (next != &this_parent->d_subdirs) {
5868+ struct list_head *tmp = next;
5869+ struct dentry *dentry = list_entry(tmp, struct dentry,
5870+ d_u.d_child);
027c5e7a 5871+
1facf9fc 5872+ next = tmp->next;
027c5e7a 5873+ spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
392086de 5874+ if (d_count(dentry)) {
027c5e7a
AM
5875+ if (!list_empty(&dentry->d_subdirs)) {
5876+ spin_unlock(&this_parent->d_lock);
5877+ spin_release(&dentry->d_lock.dep_map, 1,
5878+ _RET_IP_);
5879+ this_parent = dentry;
5880+ spin_acquire(&this_parent->d_lock.dep_map, 0, 1,
5881+ _RET_IP_);
5882+ goto repeat;
5883+ }
5884+ if (dentry->d_sb == sb
5885+ && au_di(dentry)
5886+ && (!test || test(dentry, arg)))
5887+ err = au_dpages_append(dpages, dentry,
5888+ GFP_ATOMIC);
1facf9fc 5889+ }
027c5e7a
AM
5890+ spin_unlock(&dentry->d_lock);
5891+ if (unlikely(err))
5892+ goto out;
1facf9fc 5893+ }
5894+
5895+ if (this_parent != root) {
027c5e7a
AM
5896+ struct dentry *tmp;
5897+ struct dentry *child;
5898+
5899+ tmp = this_parent->d_parent;
5900+ rcu_read_lock();
5901+ spin_unlock(&this_parent->d_lock);
5902+ child = this_parent;
5903+ this_parent = tmp;
5904+ spin_lock(&this_parent->d_lock);
5905+ rcu_read_unlock();
5906+ next = child->d_u.d_child.next;
1facf9fc 5907+ goto resume;
5908+ }
027c5e7a 5909+
4f0767ce 5910+out:
027c5e7a
AM
5911+ spin_unlock(&this_parent->d_lock);
5912+ write_sequnlock(&rename_lock);
1facf9fc 5913+ return err;
5914+}
5915+
5916+int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
5917+ int do_include, au_dpages_test test, void *arg)
5918+{
5919+ int err;
5920+
5921+ err = 0;
027c5e7a
AM
5922+ write_seqlock(&rename_lock);
5923+ spin_lock(&dentry->d_lock);
5924+ if (do_include
392086de 5925+ && d_count(dentry)
027c5e7a 5926+ && (!test || test(dentry, arg)))
1facf9fc 5927+ err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
027c5e7a
AM
5928+ spin_unlock(&dentry->d_lock);
5929+ if (unlikely(err))
5930+ goto out;
5931+
5932+ /*
523b37e3 5933+ * RCU for vfsmount is unnecessary since this is a traverse in a single
027c5e7a
AM
5934+ * mount
5935+ */
1facf9fc 5936+ while (!IS_ROOT(dentry)) {
027c5e7a
AM
5937+ dentry = dentry->d_parent; /* rename_lock is locked */
5938+ spin_lock(&dentry->d_lock);
392086de 5939+ if (d_count(dentry)
027c5e7a 5940+ && (!test || test(dentry, arg)))
1facf9fc 5941+ err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
027c5e7a
AM
5942+ spin_unlock(&dentry->d_lock);
5943+ if (unlikely(err))
5944+ break;
1facf9fc 5945+ }
5946+
4f0767ce 5947+out:
027c5e7a 5948+ write_sequnlock(&rename_lock);
1facf9fc 5949+ return err;
5950+}
5951+
027c5e7a
AM
5952+static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
5953+{
5954+ return au_di(dentry) && dentry->d_sb == arg;
5955+}
5956+
5957+int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
5958+ struct dentry *dentry, int do_include)
5959+{
5960+ return au_dcsub_pages_rev(dpages, dentry, do_include,
5961+ au_dcsub_dpages_aufs, dentry->d_sb);
5962+}
5963+
4a4d8108 5964+int au_test_subdir(struct dentry *d1, struct dentry *d2)
1facf9fc 5965+{
4a4d8108
AM
5966+ struct path path[2] = {
5967+ {
5968+ .dentry = d1
5969+ },
5970+ {
5971+ .dentry = d2
5972+ }
5973+ };
1facf9fc 5974+
4a4d8108 5975+ return path_is_under(path + 0, path + 1);
1facf9fc 5976+}
7f207e10
AM
5977diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
5978--- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
5979+++ linux/fs/aufs/dcsub.h 2014-06-24 00:14:06.779385790 +0200
5980@@ -0,0 +1,116 @@
1facf9fc 5981+/*
523b37e3 5982+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 5983+ *
5984+ * This program, aufs is free software; you can redistribute it and/or modify
5985+ * it under the terms of the GNU General Public License as published by
5986+ * the Free Software Foundation; either version 2 of the License, or
5987+ * (at your option) any later version.
dece6358
AM
5988+ *
5989+ * This program is distributed in the hope that it will be useful,
5990+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5991+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5992+ * GNU General Public License for more details.
5993+ *
5994+ * You should have received a copy of the GNU General Public License
523b37e3 5995+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5996+ */
5997+
5998+/*
5999+ * sub-routines for dentry cache
6000+ */
6001+
6002+#ifndef __AUFS_DCSUB_H__
6003+#define __AUFS_DCSUB_H__
6004+
6005+#ifdef __KERNEL__
6006+
7f207e10 6007+#include <linux/dcache.h>
027c5e7a 6008+#include <linux/fs.h>
dece6358
AM
6009+
6010+struct dentry;
1facf9fc 6011+
6012+struct au_dpage {
6013+ int ndentry;
6014+ struct dentry **dentries;
6015+};
6016+
6017+struct au_dcsub_pages {
6018+ int ndpage;
6019+ struct au_dpage *dpages;
6020+};
6021+
6022+/* ---------------------------------------------------------------------- */
6023+
7f207e10 6024+/* dcsub.c */
1facf9fc 6025+int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
6026+void au_dpages_free(struct au_dcsub_pages *dpages);
6027+typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
6028+int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6029+ au_dpages_test test, void *arg);
6030+int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6031+ int do_include, au_dpages_test test, void *arg);
027c5e7a
AM
6032+int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6033+ struct dentry *dentry, int do_include);
4a4d8108 6034+int au_test_subdir(struct dentry *d1, struct dentry *d2);
1facf9fc 6035+
7f207e10
AM
6036+/* ---------------------------------------------------------------------- */
6037+
523b37e3
AM
6038+/*
6039+ * todo: in linux-3.13, several similar (but faster) helpers are added to
6040+ * include/linux/dcache.h. Try them (in the future).
6041+ */
6042+
027c5e7a
AM
6043+static inline int au_d_hashed_positive(struct dentry *d)
6044+{
6045+ int err;
6046+ struct inode *inode = d->d_inode;
6047+ err = 0;
6048+ if (unlikely(d_unhashed(d) || !inode || !inode->i_nlink))
6049+ err = -ENOENT;
6050+ return err;
6051+}
6052+
38d290e6
JR
6053+static inline int au_d_linkable(struct dentry *d)
6054+{
6055+ int err;
6056+ struct inode *inode = d->d_inode;
6057+ err = au_d_hashed_positive(d);
6058+ if (err
6059+ && inode
6060+ && (inode->i_state & I_LINKABLE))
6061+ err = 0;
6062+ return err;
6063+}
6064+
027c5e7a
AM
6065+static inline int au_d_alive(struct dentry *d)
6066+{
6067+ int err;
6068+ struct inode *inode;
6069+ err = 0;
6070+ if (!IS_ROOT(d))
6071+ err = au_d_hashed_positive(d);
6072+ else {
6073+ inode = d->d_inode;
6074+ if (unlikely(d_unlinked(d) || !inode || !inode->i_nlink))
6075+ err = -ENOENT;
6076+ }
6077+ return err;
6078+}
6079+
6080+static inline int au_alive_dir(struct dentry *d)
7f207e10 6081+{
027c5e7a
AM
6082+ int err;
6083+ err = au_d_alive(d);
6084+ if (unlikely(err || IS_DEADDIR(d->d_inode)))
6085+ err = -ENOENT;
6086+ return err;
7f207e10
AM
6087+}
6088+
38d290e6
JR
6089+static inline int au_qstreq(struct qstr *a, struct qstr *b)
6090+{
6091+ return a->len == b->len
6092+ && !memcmp(a->name, b->name, a->len);
6093+}
6094+
1facf9fc 6095+#endif /* __KERNEL__ */
6096+#endif /* __AUFS_DCSUB_H__ */
7f207e10
AM
6097diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
6098--- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
6099+++ linux/fs/aufs/debug.c 2014-06-24 00:14:06.779385790 +0200
6100@@ -0,0 +1,518 @@
1facf9fc 6101+/*
523b37e3 6102+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 6103+ *
6104+ * This program, aufs is free software; you can redistribute it and/or modify
6105+ * it under the terms of the GNU General Public License as published by
6106+ * the Free Software Foundation; either version 2 of the License, or
6107+ * (at your option) any later version.
dece6358
AM
6108+ *
6109+ * This program is distributed in the hope that it will be useful,
6110+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6111+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6112+ * GNU General Public License for more details.
6113+ *
6114+ * You should have received a copy of the GNU General Public License
523b37e3 6115+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6116+ */
6117+
6118+/*
6119+ * debug print functions
6120+ */
6121+
7f207e10 6122+#include <linux/vt_kern.h>
1facf9fc 6123+#include "aufs.h"
6124+
392086de
AM
6125+/* Returns 0, or -errno. arg is in kp->arg. */
6126+static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
6127+{
6128+ int err, n;
6129+
6130+ err = kstrtoint(val, 0, &n);
6131+ if (!err) {
6132+ if (n > 0)
6133+ au_debug_on();
6134+ else
6135+ au_debug_off();
6136+ }
6137+ return err;
6138+}
6139+
6140+/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
6141+static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
6142+{
6143+ atomic_t *a;
6144+
6145+ a = kp->arg;
6146+ return sprintf(buffer, "%d", atomic_read(a));
6147+}
6148+
6149+static struct kernel_param_ops param_ops_atomic_t = {
6150+ .set = param_atomic_t_set,
6151+ .get = param_atomic_t_get
6152+ /* void (*free)(void *arg) */
6153+};
6154+
6155+atomic_t aufs_debug = ATOMIC_INIT(0);
1facf9fc 6156+MODULE_PARM_DESC(debug, "debug print");
392086de 6157+module_param_named(debug, aufs_debug, atomic_t, S_IRUGO | S_IWUSR | S_IWGRP);
1facf9fc 6158+
6159+char *au_plevel = KERN_DEBUG;
e49829fe
JR
6160+#define dpri(fmt, ...) do { \
6161+ if ((au_plevel \
6162+ && strcmp(au_plevel, KERN_DEBUG)) \
6163+ || au_debug_test()) \
6164+ printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
1facf9fc 6165+} while (0)
6166+
6167+/* ---------------------------------------------------------------------- */
6168+
6169+void au_dpri_whlist(struct au_nhash *whlist)
6170+{
6171+ unsigned long ul, n;
6172+ struct hlist_head *head;
c06a8ce3 6173+ struct au_vdir_wh *pos;
1facf9fc 6174+
6175+ n = whlist->nh_num;
6176+ head = whlist->nh_head;
6177+ for (ul = 0; ul < n; ul++) {
c06a8ce3 6178+ hlist_for_each_entry(pos, head, wh_hash)
1facf9fc 6179+ dpri("b%d, %.*s, %d\n",
c06a8ce3
AM
6180+ pos->wh_bindex,
6181+ pos->wh_str.len, pos->wh_str.name,
6182+ pos->wh_str.len);
1facf9fc 6183+ head++;
6184+ }
6185+}
6186+
6187+void au_dpri_vdir(struct au_vdir *vdir)
6188+{
6189+ unsigned long ul;
6190+ union au_vdir_deblk_p p;
6191+ unsigned char *o;
6192+
6193+ if (!vdir || IS_ERR(vdir)) {
6194+ dpri("err %ld\n", PTR_ERR(vdir));
6195+ return;
6196+ }
6197+
6198+ dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %lu\n",
6199+ vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
6200+ vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
6201+ for (ul = 0; ul < vdir->vd_nblk; ul++) {
6202+ p.deblk = vdir->vd_deblk[ul];
6203+ o = p.deblk;
6204+ dpri("[%lu]: %p\n", ul, o);
6205+ }
6206+}
6207+
53392da6 6208+static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
1facf9fc 6209+ struct dentry *wh)
6210+{
6211+ char *n = NULL;
6212+ int l = 0;
6213+
6214+ if (!inode || IS_ERR(inode)) {
6215+ dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
6216+ return -1;
6217+ }
6218+
c2b27bf2 6219+ /* the type of i_blocks depends upon CONFIG_LBDAF */
1facf9fc 6220+ BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
6221+ && sizeof(inode->i_blocks) != sizeof(u64));
6222+ if (wh) {
6223+ n = (void *)wh->d_name.name;
6224+ l = wh->d_name.len;
6225+ }
6226+
53392da6
AM
6227+ dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
6228+ " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
6229+ bindex, inode,
1facf9fc 6230+ inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
6231+ atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
6232+ i_size_read(inode), (unsigned long long)inode->i_blocks,
53392da6 6233+ hn, (long long)timespec_to_ns(&inode->i_ctime) & 0x0ffff,
1facf9fc 6234+ inode->i_mapping ? inode->i_mapping->nrpages : 0,
b752ccd1
AM
6235+ inode->i_state, inode->i_flags, inode->i_version,
6236+ inode->i_generation,
1facf9fc 6237+ l ? ", wh " : "", l, n);
6238+ return 0;
6239+}
6240+
6241+void au_dpri_inode(struct inode *inode)
6242+{
6243+ struct au_iinfo *iinfo;
6244+ aufs_bindex_t bindex;
53392da6 6245+ int err, hn;
1facf9fc 6246+
53392da6 6247+ err = do_pri_inode(-1, inode, -1, NULL);
1facf9fc 6248+ if (err || !au_test_aufs(inode->i_sb))
6249+ return;
6250+
6251+ iinfo = au_ii(inode);
6252+ if (!iinfo)
6253+ return;
6254+ dpri("i-1: bstart %d, bend %d, gen %d\n",
537831f9 6255+ iinfo->ii_bstart, iinfo->ii_bend, au_iigen(inode, NULL));
1facf9fc 6256+ if (iinfo->ii_bstart < 0)
6257+ return;
53392da6
AM
6258+ hn = 0;
6259+ for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; bindex++) {
6260+ hn = !!au_hn(iinfo->ii_hinode + bindex);
6261+ do_pri_inode(bindex, iinfo->ii_hinode[0 + bindex].hi_inode, hn,
1facf9fc 6262+ iinfo->ii_hinode[0 + bindex].hi_whdentry);
53392da6 6263+ }
1facf9fc 6264+}
6265+
2cbb1c4b
JR
6266+void au_dpri_dalias(struct inode *inode)
6267+{
6268+ struct dentry *d;
6269+
6270+ spin_lock(&inode->i_lock);
c06a8ce3 6271+ hlist_for_each_entry(d, &inode->i_dentry, d_alias)
2cbb1c4b
JR
6272+ au_dpri_dentry(d);
6273+ spin_unlock(&inode->i_lock);
6274+}
6275+
1facf9fc 6276+static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
6277+{
6278+ struct dentry *wh = NULL;
53392da6 6279+ int hn;
1facf9fc 6280+
6281+ if (!dentry || IS_ERR(dentry)) {
6282+ dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
6283+ return -1;
6284+ }
6285+ /* do not call dget_parent() here */
027c5e7a 6286+ /* note: access d_xxx without d_lock */
523b37e3
AM
6287+ dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
6288+ bindex, dentry, dentry,
1facf9fc 6289+ dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
523b37e3
AM
6290+ d_count(dentry), dentry->d_flags,
6291+ d_unhashed(dentry) ? "un" : "");
53392da6 6292+ hn = -1;
1facf9fc 6293+ if (bindex >= 0 && dentry->d_inode && au_test_aufs(dentry->d_sb)) {
6294+ struct au_iinfo *iinfo = au_ii(dentry->d_inode);
53392da6
AM
6295+ if (iinfo) {
6296+ hn = !!au_hn(iinfo->ii_hinode + bindex);
1facf9fc 6297+ wh = iinfo->ii_hinode[0 + bindex].hi_whdentry;
53392da6 6298+ }
1facf9fc 6299+ }
53392da6 6300+ do_pri_inode(bindex, dentry->d_inode, hn, wh);
1facf9fc 6301+ return 0;
6302+}
6303+
6304+void au_dpri_dentry(struct dentry *dentry)
6305+{
6306+ struct au_dinfo *dinfo;
6307+ aufs_bindex_t bindex;
6308+ int err;
4a4d8108 6309+ struct au_hdentry *hdp;
1facf9fc 6310+
6311+ err = do_pri_dentry(-1, dentry);
6312+ if (err || !au_test_aufs(dentry->d_sb))
6313+ return;
6314+
6315+ dinfo = au_di(dentry);
6316+ if (!dinfo)
6317+ return;
38d290e6 6318+ dpri("d-1: bstart %d, bend %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
1facf9fc 6319+ dinfo->di_bstart, dinfo->di_bend,
38d290e6
JR
6320+ dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
6321+ dinfo->di_tmpfile);
1facf9fc 6322+ if (dinfo->di_bstart < 0)
6323+ return;
4a4d8108 6324+ hdp = dinfo->di_hdentry;
1facf9fc 6325+ for (bindex = dinfo->di_bstart; bindex <= dinfo->di_bend; bindex++)
4a4d8108 6326+ do_pri_dentry(bindex, hdp[0 + bindex].hd_dentry);
1facf9fc 6327+}
6328+
6329+static int do_pri_file(aufs_bindex_t bindex, struct file *file)
6330+{
6331+ char a[32];
6332+
6333+ if (!file || IS_ERR(file)) {
6334+ dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
6335+ return -1;
6336+ }
6337+ a[0] = 0;
6338+ if (bindex < 0
6339+ && file->f_dentry
6340+ && au_test_aufs(file->f_dentry->d_sb)
6341+ && au_fi(file))
e49829fe 6342+ snprintf(a, sizeof(a), ", gen %d, mmapped %d",
2cbb1c4b 6343+ au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
b752ccd1 6344+ dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
1facf9fc 6345+ bindex, file->f_mode, file->f_flags, (long)file_count(file),
b752ccd1 6346+ file->f_version, file->f_pos, a);
1facf9fc 6347+ if (file->f_dentry)
6348+ do_pri_dentry(bindex, file->f_dentry);
6349+ return 0;
6350+}
6351+
6352+void au_dpri_file(struct file *file)
6353+{
6354+ struct au_finfo *finfo;
4a4d8108
AM
6355+ struct au_fidir *fidir;
6356+ struct au_hfile *hfile;
1facf9fc 6357+ aufs_bindex_t bindex;
6358+ int err;
6359+
6360+ err = do_pri_file(-1, file);
6361+ if (err || !file->f_dentry || !au_test_aufs(file->f_dentry->d_sb))
6362+ return;
6363+
6364+ finfo = au_fi(file);
6365+ if (!finfo)
6366+ return;
4a4d8108 6367+ if (finfo->fi_btop < 0)
1facf9fc 6368+ return;
4a4d8108
AM
6369+ fidir = finfo->fi_hdir;
6370+ if (!fidir)
6371+ do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
6372+ else
e49829fe
JR
6373+ for (bindex = finfo->fi_btop;
6374+ bindex >= 0 && bindex <= fidir->fd_bbot;
4a4d8108
AM
6375+ bindex++) {
6376+ hfile = fidir->fd_hfile + bindex;
6377+ do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
6378+ }
1facf9fc 6379+}
6380+
6381+static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
6382+{
6383+ struct vfsmount *mnt;
6384+ struct super_block *sb;
6385+
6386+ if (!br || IS_ERR(br))
6387+ goto out;
86dc4139 6388+ mnt = au_br_mnt(br);
1facf9fc 6389+ if (!mnt || IS_ERR(mnt))
6390+ goto out;
6391+ sb = mnt->mnt_sb;
6392+ if (!sb || IS_ERR(sb))
6393+ goto out;
6394+
1e00d052 6395+ dpri("s%d: {perm 0x%x, id %d, cnt %d, wbr %p}, "
b752ccd1 6396+ "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
1facf9fc 6397+ "xino %d\n",
1e00d052
AM
6398+ bindex, br->br_perm, br->br_id, atomic_read(&br->br_count),
6399+ br->br_wbr, au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
b752ccd1 6400+ sb->s_flags, sb->s_count,
1facf9fc 6401+ atomic_read(&sb->s_active), !!br->br_xino.xi_file);
6402+ return 0;
6403+
4f0767ce 6404+out:
1facf9fc 6405+ dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
6406+ return -1;
6407+}
6408+
6409+void au_dpri_sb(struct super_block *sb)
6410+{
6411+ struct au_sbinfo *sbinfo;
6412+ aufs_bindex_t bindex;
6413+ int err;
6414+ /* to reuduce stack size */
6415+ struct {
6416+ struct vfsmount mnt;
6417+ struct au_branch fake;
6418+ } *a;
6419+
6420+ /* this function can be called from magic sysrq */
6421+ a = kzalloc(sizeof(*a), GFP_ATOMIC);
6422+ if (unlikely(!a)) {
6423+ dpri("no memory\n");
6424+ return;
6425+ }
6426+
6427+ a->mnt.mnt_sb = sb;
6428+ a->fake.br_perm = 0;
86dc4139 6429+ a->fake.br_path.mnt = &a->mnt;
1facf9fc 6430+ a->fake.br_xino.xi_file = NULL;
6431+ atomic_set(&a->fake.br_count, 0);
6432+ smp_mb(); /* atomic_set */
6433+ err = do_pri_br(-1, &a->fake);
6434+ kfree(a);
6435+ dpri("dev 0x%x\n", sb->s_dev);
6436+ if (err || !au_test_aufs(sb))
6437+ return;
6438+
6439+ sbinfo = au_sbi(sb);
6440+ if (!sbinfo)
6441+ return;
6442+ dpri("nw %d, gen %u, kobj %d\n",
6443+ atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
6444+ atomic_read(&sbinfo->si_kobj.kref.refcount));
6445+ for (bindex = 0; bindex <= sbinfo->si_bend; bindex++)
6446+ do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
6447+}
6448+
6449+/* ---------------------------------------------------------------------- */
6450+
6451+void au_dbg_sleep_jiffy(int jiffy)
6452+{
6453+ while (jiffy)
6454+ jiffy = schedule_timeout_uninterruptible(jiffy);
6455+}
6456+
6457+void au_dbg_iattr(struct iattr *ia)
6458+{
c06a8ce3
AM
6459+#define AuBit(name) \
6460+ do { \
6461+ if (ia->ia_valid & ATTR_ ## name) \
6462+ dpri(#name "\n"); \
6463+ } while (0)
1facf9fc 6464+ AuBit(MODE);
6465+ AuBit(UID);
6466+ AuBit(GID);
6467+ AuBit(SIZE);
6468+ AuBit(ATIME);
6469+ AuBit(MTIME);
6470+ AuBit(CTIME);
6471+ AuBit(ATIME_SET);
6472+ AuBit(MTIME_SET);
6473+ AuBit(FORCE);
6474+ AuBit(ATTR_FLAG);
6475+ AuBit(KILL_SUID);
6476+ AuBit(KILL_SGID);
6477+ AuBit(FILE);
6478+ AuBit(KILL_PRIV);
6479+ AuBit(OPEN);
6480+ AuBit(TIMES_SET);
6481+#undef AuBit
6482+ dpri("ia_file %p\n", ia->ia_file);
6483+}
6484+
6485+/* ---------------------------------------------------------------------- */
6486+
027c5e7a
AM
6487+void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
6488+{
6489+ struct inode *h_inode, *inode = dentry->d_inode;
6490+ struct dentry *h_dentry;
6491+ aufs_bindex_t bindex, bend, bi;
6492+
6493+ if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
6494+ return;
6495+
6496+ bend = au_dbend(dentry);
6497+ bi = au_ibend(inode);
6498+ if (bi < bend)
6499+ bend = bi;
6500+ bindex = au_dbstart(dentry);
6501+ bi = au_ibstart(inode);
6502+ if (bi > bindex)
6503+ bindex = bi;
6504+
6505+ for (; bindex <= bend; bindex++) {
6506+ h_dentry = au_h_dptr(dentry, bindex);
6507+ if (!h_dentry)
6508+ continue;
6509+ h_inode = au_h_iptr(inode, bindex);
6510+ if (unlikely(h_inode != h_dentry->d_inode)) {
392086de 6511+ au_debug_on();
027c5e7a
AM
6512+ AuDbg("b%d, %s:%d\n", bindex, func, line);
6513+ AuDbgDentry(dentry);
6514+ AuDbgInode(inode);
392086de 6515+ au_debug_off();
027c5e7a
AM
6516+ BUG();
6517+ }
6518+ }
6519+}
6520+
1facf9fc 6521+void au_dbg_verify_dir_parent(struct dentry *dentry, unsigned int sigen)
6522+{
6523+ struct dentry *parent;
6524+
6525+ parent = dget_parent(dentry);
027c5e7a
AM
6526+ AuDebugOn(!S_ISDIR(dentry->d_inode->i_mode));
6527+ AuDebugOn(IS_ROOT(dentry));
6528+ AuDebugOn(au_digen_test(parent, sigen));
1facf9fc 6529+ dput(parent);
6530+}
6531+
6532+void au_dbg_verify_nondir_parent(struct dentry *dentry, unsigned int sigen)
6533+{
6534+ struct dentry *parent;
027c5e7a 6535+ struct inode *inode;
1facf9fc 6536+
6537+ parent = dget_parent(dentry);
027c5e7a
AM
6538+ inode = dentry->d_inode;
6539+ AuDebugOn(inode && S_ISDIR(dentry->d_inode->i_mode));
6540+ AuDebugOn(au_digen_test(parent, sigen));
1facf9fc 6541+ dput(parent);
6542+}
6543+
6544+void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
6545+{
6546+ int err, i, j;
6547+ struct au_dcsub_pages dpages;
6548+ struct au_dpage *dpage;
6549+ struct dentry **dentries;
6550+
6551+ err = au_dpages_init(&dpages, GFP_NOFS);
6552+ AuDebugOn(err);
027c5e7a 6553+ err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
1facf9fc 6554+ AuDebugOn(err);
6555+ for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
6556+ dpage = dpages.dpages + i;
6557+ dentries = dpage->dentries;
6558+ for (j = dpage->ndentry - 1; !err && j >= 0; j--)
027c5e7a 6559+ AuDebugOn(au_digen_test(dentries[j], sigen));
1facf9fc 6560+ }
6561+ au_dpages_free(&dpages);
6562+}
6563+
1facf9fc 6564+void au_dbg_verify_kthread(void)
6565+{
53392da6 6566+ if (au_wkq_test()) {
1facf9fc 6567+ au_dbg_blocked();
1e00d052
AM
6568+ /*
6569+ * It may be recursive, but udba=notify between two aufs mounts,
6570+ * where a single ro branch is shared, is not a problem.
6571+ */
6572+ /* WARN_ON(1); */
1facf9fc 6573+ }
6574+}
6575+
6576+/* ---------------------------------------------------------------------- */
6577+
6578+void au_debug_sbinfo_init(struct au_sbinfo *sbinfo __maybe_unused)
6579+{
6580+#ifdef AuForceNoPlink
6581+ au_opt_clr(sbinfo->si_mntflags, PLINK);
6582+#endif
6583+#ifdef AuForceNoXino
6584+ au_opt_clr(sbinfo->si_mntflags, XINO);
6585+#endif
6586+#ifdef AuForceNoRefrof
6587+ au_opt_clr(sbinfo->si_mntflags, REFROF);
6588+#endif
4a4d8108
AM
6589+#ifdef AuForceHnotify
6590+ au_opt_set_udba(sbinfo->si_mntflags, UDBA_HNOTIFY);
1facf9fc 6591+#endif
1308ab2a 6592+#ifdef AuForceRd0
6593+ sbinfo->si_rdblk = 0;
6594+ sbinfo->si_rdhash = 0;
6595+#endif
1facf9fc 6596+}
6597+
6598+int __init au_debug_init(void)
6599+{
6600+ aufs_bindex_t bindex;
6601+ struct au_vdir_destr destr;
6602+
6603+ bindex = -1;
6604+ AuDebugOn(bindex >= 0);
6605+
6606+ destr.len = -1;
6607+ AuDebugOn(destr.len < NAME_MAX);
6608+
6609+#ifdef CONFIG_4KSTACKS
0c3ec466 6610+ pr_warn("CONFIG_4KSTACKS is defined.\n");
1facf9fc 6611+#endif
6612+
6613+#ifdef AuForceNoBrs
6614+ sysaufs_brs = 0;
6615+#endif
6616+
6617+ return 0;
6618+}
7f207e10
AM
6619diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
6620--- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 6621+++ linux/fs/aufs/debug.h 2014-06-24 00:14:06.779385790 +0200
523b37e3 6622@@ -0,0 +1,247 @@
1facf9fc 6623+/*
523b37e3 6624+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 6625+ *
6626+ * This program, aufs is free software; you can redistribute it and/or modify
6627+ * it under the terms of the GNU General Public License as published by
6628+ * the Free Software Foundation; either version 2 of the License, or
6629+ * (at your option) any later version.
dece6358
AM
6630+ *
6631+ * This program is distributed in the hope that it will be useful,
6632+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6633+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6634+ * GNU General Public License for more details.
6635+ *
6636+ * You should have received a copy of the GNU General Public License
523b37e3 6637+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6638+ */
6639+
6640+/*
6641+ * debug print functions
6642+ */
6643+
6644+#ifndef __AUFS_DEBUG_H__
6645+#define __AUFS_DEBUG_H__
6646+
6647+#ifdef __KERNEL__
6648+
392086de 6649+#include <linux/atomic.h>
4a4d8108
AM
6650+#include <linux/module.h>
6651+#include <linux/kallsyms.h>
1facf9fc 6652+#include <linux/sysrq.h>
4a4d8108 6653+
1facf9fc 6654+#ifdef CONFIG_AUFS_DEBUG
6655+#define AuDebugOn(a) BUG_ON(a)
6656+
6657+/* module parameter */
392086de
AM
6658+extern atomic_t aufs_debug;
6659+static inline void au_debug_on(void)
1facf9fc 6660+{
392086de
AM
6661+ atomic_inc(&aufs_debug);
6662+}
6663+static inline void au_debug_off(void)
6664+{
6665+ atomic_dec_if_positive(&aufs_debug);
1facf9fc 6666+}
6667+
6668+static inline int au_debug_test(void)
6669+{
392086de 6670+ return atomic_read(&aufs_debug) > 0;
1facf9fc 6671+}
6672+#else
6673+#define AuDebugOn(a) do {} while (0)
392086de
AM
6674+AuStubVoid(au_debug_on, void)
6675+AuStubVoid(au_debug_off, void)
4a4d8108 6676+AuStubInt0(au_debug_test, void)
1facf9fc 6677+#endif /* CONFIG_AUFS_DEBUG */
6678+
392086de
AM
6679+#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
6680+
1facf9fc 6681+/* ---------------------------------------------------------------------- */
6682+
6683+/* debug print */
6684+
4a4d8108 6685+#define AuDbg(fmt, ...) do { \
1facf9fc 6686+ if (au_debug_test()) \
4a4d8108 6687+ pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
1facf9fc 6688+} while (0)
4a4d8108
AM
6689+#define AuLabel(l) AuDbg(#l "\n")
6690+#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
6691+#define AuWarn1(fmt, ...) do { \
1facf9fc 6692+ static unsigned char _c; \
6693+ if (!_c++) \
0c3ec466 6694+ pr_warn(fmt, ##__VA_ARGS__); \
1facf9fc 6695+} while (0)
6696+
4a4d8108 6697+#define AuErr1(fmt, ...) do { \
1facf9fc 6698+ static unsigned char _c; \
6699+ if (!_c++) \
4a4d8108 6700+ pr_err(fmt, ##__VA_ARGS__); \
1facf9fc 6701+} while (0)
6702+
4a4d8108 6703+#define AuIOErr1(fmt, ...) do { \
1facf9fc 6704+ static unsigned char _c; \
6705+ if (!_c++) \
4a4d8108 6706+ AuIOErr(fmt, ##__VA_ARGS__); \
1facf9fc 6707+} while (0)
6708+
6709+#define AuUnsupportMsg "This operation is not supported." \
6710+ " Please report this application to aufs-users ML."
4a4d8108
AM
6711+#define AuUnsupport(fmt, ...) do { \
6712+ pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
1facf9fc 6713+ dump_stack(); \
6714+} while (0)
6715+
6716+#define AuTraceErr(e) do { \
6717+ if (unlikely((e) < 0)) \
6718+ AuDbg("err %d\n", (int)(e)); \
6719+} while (0)
6720+
6721+#define AuTraceErrPtr(p) do { \
6722+ if (IS_ERR(p)) \
6723+ AuDbg("err %ld\n", PTR_ERR(p)); \
6724+} while (0)
6725+
6726+/* dirty macros for debug print, use with "%.*s" and caution */
6727+#define AuLNPair(qstr) (qstr)->len, (qstr)->name
1facf9fc 6728+
6729+/* ---------------------------------------------------------------------- */
6730+
6731+struct au_sbinfo;
6732+struct au_finfo;
dece6358 6733+struct dentry;
1facf9fc 6734+#ifdef CONFIG_AUFS_DEBUG
6735+extern char *au_plevel;
6736+struct au_nhash;
6737+void au_dpri_whlist(struct au_nhash *whlist);
6738+struct au_vdir;
6739+void au_dpri_vdir(struct au_vdir *vdir);
dece6358 6740+struct inode;
1facf9fc 6741+void au_dpri_inode(struct inode *inode);
2cbb1c4b 6742+void au_dpri_dalias(struct inode *inode);
1facf9fc 6743+void au_dpri_dentry(struct dentry *dentry);
dece6358 6744+struct file;
1facf9fc 6745+void au_dpri_file(struct file *filp);
dece6358 6746+struct super_block;
1facf9fc 6747+void au_dpri_sb(struct super_block *sb);
6748+
6749+void au_dbg_sleep_jiffy(int jiffy);
dece6358 6750+struct iattr;
1facf9fc 6751+void au_dbg_iattr(struct iattr *ia);
6752+
027c5e7a
AM
6753+#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
6754+void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
1facf9fc 6755+void au_dbg_verify_dir_parent(struct dentry *dentry, unsigned int sigen);
6756+void au_dbg_verify_nondir_parent(struct dentry *dentry, unsigned int sigen);
6757+void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
1facf9fc 6758+void au_dbg_verify_kthread(void);
6759+
6760+int __init au_debug_init(void);
6761+void au_debug_sbinfo_init(struct au_sbinfo *sbinfo);
6762+#define AuDbgWhlist(w) do { \
6763+ AuDbg(#w "\n"); \
6764+ au_dpri_whlist(w); \
6765+} while (0)
6766+
6767+#define AuDbgVdir(v) do { \
6768+ AuDbg(#v "\n"); \
6769+ au_dpri_vdir(v); \
6770+} while (0)
6771+
6772+#define AuDbgInode(i) do { \
6773+ AuDbg(#i "\n"); \
6774+ au_dpri_inode(i); \
6775+} while (0)
6776+
2cbb1c4b
JR
6777+#define AuDbgDAlias(i) do { \
6778+ AuDbg(#i "\n"); \
6779+ au_dpri_dalias(i); \
6780+} while (0)
6781+
1facf9fc 6782+#define AuDbgDentry(d) do { \
6783+ AuDbg(#d "\n"); \
6784+ au_dpri_dentry(d); \
6785+} while (0)
6786+
6787+#define AuDbgFile(f) do { \
6788+ AuDbg(#f "\n"); \
6789+ au_dpri_file(f); \
6790+} while (0)
6791+
6792+#define AuDbgSb(sb) do { \
6793+ AuDbg(#sb "\n"); \
6794+ au_dpri_sb(sb); \
6795+} while (0)
6796+
6797+#define AuDbgSleep(sec) do { \
6798+ AuDbg("sleep %d sec\n", sec); \
6799+ ssleep(sec); \
6800+} while (0)
6801+
6802+#define AuDbgSleepJiffy(jiffy) do { \
6803+ AuDbg("sleep %d jiffies\n", jiffy); \
6804+ au_dbg_sleep_jiffy(jiffy); \
6805+} while (0)
6806+
6807+#define AuDbgIAttr(ia) do { \
6808+ AuDbg("ia_valid 0x%x\n", (ia)->ia_valid); \
6809+ au_dbg_iattr(ia); \
6810+} while (0)
4a4d8108
AM
6811+
6812+#define AuDbgSym(addr) do { \
6813+ char sym[KSYM_SYMBOL_LEN]; \
6814+ sprint_symbol(sym, (unsigned long)addr); \
6815+ AuDbg("%s\n", sym); \
6816+} while (0)
6817+
6818+#define AuInfoSym(addr) do { \
6819+ char sym[KSYM_SYMBOL_LEN]; \
6820+ sprint_symbol(sym, (unsigned long)addr); \
6821+ AuInfo("%s\n", sym); \
6822+} while (0)
1facf9fc 6823+#else
027c5e7a 6824+AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
4a4d8108
AM
6825+AuStubVoid(au_dbg_verify_dir_parent, struct dentry *dentry, unsigned int sigen)
6826+AuStubVoid(au_dbg_verify_nondir_parent, struct dentry *dentry,
6827+ unsigned int sigen)
6828+AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
6829+AuStubVoid(au_dbg_verify_kthread, void)
6830+AuStubInt0(__init au_debug_init, void)
6831+AuStubVoid(au_debug_sbinfo_init, struct au_sbinfo *sbinfo)
1facf9fc 6832+
1facf9fc 6833+#define AuDbgWhlist(w) do {} while (0)
6834+#define AuDbgVdir(v) do {} while (0)
6835+#define AuDbgInode(i) do {} while (0)
2cbb1c4b 6836+#define AuDbgDAlias(i) do {} while (0)
1facf9fc 6837+#define AuDbgDentry(d) do {} while (0)
6838+#define AuDbgFile(f) do {} while (0)
6839+#define AuDbgSb(sb) do {} while (0)
6840+#define AuDbgSleep(sec) do {} while (0)
6841+#define AuDbgSleepJiffy(jiffy) do {} while (0)
6842+#define AuDbgIAttr(ia) do {} while (0)
4a4d8108
AM
6843+#define AuDbgSym(addr) do {} while (0)
6844+#define AuInfoSym(addr) do {} while (0)
1facf9fc 6845+#endif /* CONFIG_AUFS_DEBUG */
6846+
6847+/* ---------------------------------------------------------------------- */
6848+
6849+#ifdef CONFIG_AUFS_MAGIC_SYSRQ
6850+int __init au_sysrq_init(void);
6851+void au_sysrq_fin(void);
6852+
6853+#ifdef CONFIG_HW_CONSOLE
6854+#define au_dbg_blocked() do { \
6855+ WARN_ON(1); \
0c5527e5 6856+ handle_sysrq('w'); \
1facf9fc 6857+} while (0)
6858+#else
4a4d8108 6859+AuStubVoid(au_dbg_blocked, void)
1facf9fc 6860+#endif
6861+
6862+#else
4a4d8108
AM
6863+AuStubInt0(__init au_sysrq_init, void)
6864+AuStubVoid(au_sysrq_fin, void)
6865+AuStubVoid(au_dbg_blocked, void)
1facf9fc 6866+#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
6867+
6868+#endif /* __KERNEL__ */
6869+#endif /* __AUFS_DEBUG_H__ */
7f207e10
AM
6870diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
6871--- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
6872+++ linux/fs/aufs/dentry.c 2014-06-24 00:14:06.779385790 +0200
6873@@ -0,0 +1,1084 @@
1facf9fc 6874+/*
523b37e3 6875+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 6876+ *
6877+ * This program, aufs is free software; you can redistribute it and/or modify
6878+ * it under the terms of the GNU General Public License as published by
6879+ * the Free Software Foundation; either version 2 of the License, or
6880+ * (at your option) any later version.
dece6358
AM
6881+ *
6882+ * This program is distributed in the hope that it will be useful,
6883+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6884+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6885+ * GNU General Public License for more details.
6886+ *
6887+ * You should have received a copy of the GNU General Public License
523b37e3 6888+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6889+ */
6890+
6891+/*
6892+ * lookup and dentry operations
6893+ */
6894+
dece6358 6895+#include <linux/namei.h>
1facf9fc 6896+#include "aufs.h"
6897+
1facf9fc 6898+#define AuLkup_ALLOW_NEG 1
6899+#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
7f207e10
AM
6900+#define au_fset_lkup(flags, name) \
6901+ do { (flags) |= AuLkup_##name; } while (0)
6902+#define au_fclr_lkup(flags, name) \
6903+ do { (flags) &= ~AuLkup_##name; } while (0)
1facf9fc 6904+
6905+struct au_do_lookup_args {
6906+ unsigned int flags;
6907+ mode_t type;
1facf9fc 6908+};
6909+
6910+/*
6911+ * returns positive/negative dentry, NULL or an error.
6912+ * NULL means whiteout-ed or not-found.
6913+ */
6914+static struct dentry*
6915+au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
6916+ aufs_bindex_t bindex, struct qstr *wh_name,
6917+ struct au_do_lookup_args *args)
6918+{
6919+ struct dentry *h_dentry;
6920+ struct inode *h_inode, *inode;
1facf9fc 6921+ struct au_branch *br;
6922+ int wh_found, opq;
6923+ unsigned char wh_able;
6924+ const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
6925+
1facf9fc 6926+ wh_found = 0;
6927+ br = au_sbr(dentry->d_sb, bindex);
6928+ wh_able = !!au_br_whable(br->br_perm);
6929+ if (wh_able)
6930+ wh_found = au_wh_test(h_parent, wh_name, br, /*try_sio*/0);
6931+ h_dentry = ERR_PTR(wh_found);
6932+ if (!wh_found)
6933+ goto real_lookup;
6934+ if (unlikely(wh_found < 0))
6935+ goto out;
6936+
6937+ /* We found a whiteout */
6938+ /* au_set_dbend(dentry, bindex); */
6939+ au_set_dbwh(dentry, bindex);
6940+ if (!allow_neg)
6941+ return NULL; /* success */
6942+
4f0767ce 6943+real_lookup:
b4510431 6944+ h_dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
1facf9fc 6945+ if (IS_ERR(h_dentry))
6946+ goto out;
6947+
6948+ h_inode = h_dentry->d_inode;
6949+ if (!h_inode) {
6950+ if (!allow_neg)
6951+ goto out_neg;
6952+ } else if (wh_found
6953+ || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
6954+ goto out_neg;
6955+
6956+ if (au_dbend(dentry) <= bindex)
6957+ au_set_dbend(dentry, bindex);
6958+ if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
6959+ au_set_dbstart(dentry, bindex);
6960+ au_set_h_dptr(dentry, bindex, h_dentry);
6961+
6962+ inode = dentry->d_inode;
6963+ if (!h_inode || !S_ISDIR(h_inode->i_mode) || !wh_able
6964+ || (inode && !S_ISDIR(inode->i_mode)))
6965+ goto out; /* success */
6966+
6967+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
6968+ opq = au_diropq_test(h_dentry, br);
6969+ mutex_unlock(&h_inode->i_mutex);
6970+ if (opq > 0)
6971+ au_set_dbdiropq(dentry, bindex);
6972+ else if (unlikely(opq < 0)) {
6973+ au_set_h_dptr(dentry, bindex, NULL);
6974+ h_dentry = ERR_PTR(opq);
6975+ }
6976+ goto out;
6977+
4f0767ce 6978+out_neg:
1facf9fc 6979+ dput(h_dentry);
6980+ h_dentry = NULL;
4f0767ce 6981+out:
1facf9fc 6982+ return h_dentry;
6983+}
6984+
dece6358
AM
6985+static int au_test_shwh(struct super_block *sb, const struct qstr *name)
6986+{
6987+ if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
6988+ && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
6989+ return -EPERM;
6990+ return 0;
6991+}
6992+
1facf9fc 6993+/*
6994+ * returns the number of lower positive dentries,
6995+ * otherwise an error.
6996+ * can be called at unlinking with @type is zero.
6997+ */
537831f9 6998+int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type)
1facf9fc 6999+{
7000+ int npositive, err;
7001+ aufs_bindex_t bindex, btail, bdiropq;
7002+ unsigned char isdir;
7003+ struct qstr whname;
7004+ struct au_do_lookup_args args = {
b4510431 7005+ .flags = 0,
537831f9 7006+ .type = type
1facf9fc 7007+ };
7008+ const struct qstr *name = &dentry->d_name;
7009+ struct dentry *parent;
7010+ struct inode *inode;
7011+
dece6358
AM
7012+ err = au_test_shwh(dentry->d_sb, name);
7013+ if (unlikely(err))
1facf9fc 7014+ goto out;
7015+
7016+ err = au_wh_name_alloc(&whname, name);
7017+ if (unlikely(err))
7018+ goto out;
7019+
7020+ inode = dentry->d_inode;
7021+ isdir = !!(inode && S_ISDIR(inode->i_mode));
7022+ if (!type)
7023+ au_fset_lkup(args.flags, ALLOW_NEG);
7024+
7025+ npositive = 0;
4a4d8108 7026+ parent = dget_parent(dentry);
1facf9fc 7027+ btail = au_dbtaildir(parent);
7028+ for (bindex = bstart; bindex <= btail; bindex++) {
7029+ struct dentry *h_parent, *h_dentry;
7030+ struct inode *h_inode, *h_dir;
7031+
7032+ h_dentry = au_h_dptr(dentry, bindex);
7033+ if (h_dentry) {
7034+ if (h_dentry->d_inode)
7035+ npositive++;
7036+ if (type != S_IFDIR)
7037+ break;
7038+ continue;
7039+ }
7040+ h_parent = au_h_dptr(parent, bindex);
7041+ if (!h_parent)
7042+ continue;
7043+ h_dir = h_parent->d_inode;
7044+ if (!h_dir || !S_ISDIR(h_dir->i_mode))
7045+ continue;
7046+
7047+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
7048+ h_dentry = au_do_lookup(h_parent, dentry, bindex, &whname,
7049+ &args);
7050+ mutex_unlock(&h_dir->i_mutex);
7051+ err = PTR_ERR(h_dentry);
7052+ if (IS_ERR(h_dentry))
4a4d8108 7053+ goto out_parent;
1facf9fc 7054+ au_fclr_lkup(args.flags, ALLOW_NEG);
7055+
7056+ if (au_dbwh(dentry) >= 0)
7057+ break;
7058+ if (!h_dentry)
7059+ continue;
7060+ h_inode = h_dentry->d_inode;
7061+ if (!h_inode)
7062+ continue;
7063+ npositive++;
7064+ if (!args.type)
7065+ args.type = h_inode->i_mode & S_IFMT;
7066+ if (args.type != S_IFDIR)
7067+ break;
7068+ else if (isdir) {
7069+ /* the type of lower may be different */
7070+ bdiropq = au_dbdiropq(dentry);
7071+ if (bdiropq >= 0 && bdiropq <= bindex)
7072+ break;
7073+ }
7074+ }
7075+
7076+ if (npositive) {
7077+ AuLabel(positive);
7078+ au_update_dbstart(dentry);
7079+ }
7080+ err = npositive;
7081+ if (unlikely(!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
027c5e7a 7082+ && au_dbstart(dentry) < 0)) {
1facf9fc 7083+ err = -EIO;
523b37e3
AM
7084+ AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
7085+ dentry, err);
027c5e7a 7086+ }
1facf9fc 7087+
4f0767ce 7088+out_parent:
4a4d8108 7089+ dput(parent);
1facf9fc 7090+ kfree(whname.name);
4f0767ce 7091+out:
1facf9fc 7092+ return err;
7093+}
7094+
7095+struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent,
7096+ struct au_branch *br)
7097+{
7098+ struct dentry *dentry;
7099+ int wkq_err;
7100+
7101+ if (!au_test_h_perm_sio(parent->d_inode, MAY_EXEC))
b4510431 7102+ dentry = vfsub_lkup_one(name, parent);
1facf9fc 7103+ else {
b4510431
AM
7104+ struct vfsub_lkup_one_args args = {
7105+ .errp = &dentry,
7106+ .name = name,
7107+ .parent = parent
1facf9fc 7108+ };
7109+
b4510431 7110+ wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
1facf9fc 7111+ if (unlikely(wkq_err))
7112+ dentry = ERR_PTR(wkq_err);
7113+ }
7114+
7115+ return dentry;
7116+}
7117+
7118+/*
7119+ * lookup @dentry on @bindex which should be negative.
7120+ */
86dc4139 7121+int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
1facf9fc 7122+{
7123+ int err;
7124+ struct dentry *parent, *h_parent, *h_dentry;
86dc4139 7125+ struct au_branch *br;
1facf9fc 7126+
1facf9fc 7127+ parent = dget_parent(dentry);
7128+ h_parent = au_h_dptr(parent, bindex);
86dc4139
AM
7129+ br = au_sbr(dentry->d_sb, bindex);
7130+ if (wh)
7131+ h_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
7132+ else
7133+ h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent, br);
1facf9fc 7134+ err = PTR_ERR(h_dentry);
7135+ if (IS_ERR(h_dentry))
7136+ goto out;
7137+ if (unlikely(h_dentry->d_inode)) {
7138+ err = -EIO;
523b37e3 7139+ AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
1facf9fc 7140+ dput(h_dentry);
7141+ goto out;
7142+ }
7143+
4a4d8108 7144+ err = 0;
1facf9fc 7145+ if (bindex < au_dbstart(dentry))
7146+ au_set_dbstart(dentry, bindex);
7147+ if (au_dbend(dentry) < bindex)
7148+ au_set_dbend(dentry, bindex);
7149+ au_set_h_dptr(dentry, bindex, h_dentry);
1facf9fc 7150+
4f0767ce 7151+out:
1facf9fc 7152+ dput(parent);
7153+ return err;
7154+}
7155+
7156+/* ---------------------------------------------------------------------- */
7157+
7158+/* subset of struct inode */
7159+struct au_iattr {
7160+ unsigned long i_ino;
7161+ /* unsigned int i_nlink; */
0c3ec466
AM
7162+ kuid_t i_uid;
7163+ kgid_t i_gid;
1facf9fc 7164+ u64 i_version;
7165+/*
7166+ loff_t i_size;
7167+ blkcnt_t i_blocks;
7168+*/
7169+ umode_t i_mode;
7170+};
7171+
7172+static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
7173+{
7174+ ia->i_ino = h_inode->i_ino;
7175+ /* ia->i_nlink = h_inode->i_nlink; */
7176+ ia->i_uid = h_inode->i_uid;
7177+ ia->i_gid = h_inode->i_gid;
7178+ ia->i_version = h_inode->i_version;
7179+/*
7180+ ia->i_size = h_inode->i_size;
7181+ ia->i_blocks = h_inode->i_blocks;
7182+*/
7183+ ia->i_mode = (h_inode->i_mode & S_IFMT);
7184+}
7185+
7186+static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
7187+{
7188+ return ia->i_ino != h_inode->i_ino
7189+ /* || ia->i_nlink != h_inode->i_nlink */
0c3ec466 7190+ || !uid_eq(ia->i_uid, h_inode->i_uid)
2dfbb274 7191+ || !gid_eq(ia->i_gid, h_inode->i_gid)
1facf9fc 7192+ || ia->i_version != h_inode->i_version
7193+/*
7194+ || ia->i_size != h_inode->i_size
7195+ || ia->i_blocks != h_inode->i_blocks
7196+*/
7197+ || ia->i_mode != (h_inode->i_mode & S_IFMT);
7198+}
7199+
7200+static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
7201+ struct au_branch *br)
7202+{
7203+ int err;
7204+ struct au_iattr ia;
7205+ struct inode *h_inode;
7206+ struct dentry *h_d;
7207+ struct super_block *h_sb;
7208+
7209+ err = 0;
7210+ memset(&ia, -1, sizeof(ia));
7211+ h_sb = h_dentry->d_sb;
7212+ h_inode = h_dentry->d_inode;
7213+ if (h_inode)
7214+ au_iattr_save(&ia, h_inode);
7215+ else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
7216+ /* nfs d_revalidate may return 0 for negative dentry */
7217+ /* fuse d_revalidate always return 0 for negative dentry */
7218+ goto out;
7219+
7220+ /* main purpose is namei.c:cached_lookup() and d_revalidate */
b4510431 7221+ h_d = vfsub_lkup_one(&h_dentry->d_name, h_parent);
1facf9fc 7222+ err = PTR_ERR(h_d);
7223+ if (IS_ERR(h_d))
7224+ goto out;
7225+
7226+ err = 0;
7227+ if (unlikely(h_d != h_dentry
7228+ || h_d->d_inode != h_inode
7229+ || (h_inode && au_iattr_test(&ia, h_inode))))
7230+ err = au_busy_or_stale();
7231+ dput(h_d);
7232+
4f0767ce 7233+out:
1facf9fc 7234+ AuTraceErr(err);
7235+ return err;
7236+}
7237+
7238+int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
7239+ struct dentry *h_parent, struct au_branch *br)
7240+{
7241+ int err;
7242+
7243+ err = 0;
027c5e7a
AM
7244+ if (udba == AuOpt_UDBA_REVAL
7245+ && !au_test_fs_remote(h_dentry->d_sb)) {
1facf9fc 7246+ IMustLock(h_dir);
7247+ err = (h_dentry->d_parent->d_inode != h_dir);
027c5e7a 7248+ } else if (udba != AuOpt_UDBA_NONE)
1facf9fc 7249+ err = au_h_verify_dentry(h_dentry, h_parent, br);
7250+
7251+ return err;
7252+}
7253+
7254+/* ---------------------------------------------------------------------- */
7255+
027c5e7a 7256+static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
1facf9fc 7257+{
027c5e7a 7258+ int err;
1facf9fc 7259+ aufs_bindex_t new_bindex, bindex, bend, bwh, bdiropq;
027c5e7a
AM
7260+ struct au_hdentry tmp, *p, *q;
7261+ struct au_dinfo *dinfo;
7262+ struct super_block *sb;
1facf9fc 7263+
027c5e7a 7264+ DiMustWriteLock(dentry);
1308ab2a 7265+
027c5e7a
AM
7266+ sb = dentry->d_sb;
7267+ dinfo = au_di(dentry);
1facf9fc 7268+ bend = dinfo->di_bend;
7269+ bwh = dinfo->di_bwh;
7270+ bdiropq = dinfo->di_bdiropq;
027c5e7a 7271+ p = dinfo->di_hdentry + dinfo->di_bstart;
1facf9fc 7272+ for (bindex = dinfo->di_bstart; bindex <= bend; bindex++, p++) {
027c5e7a 7273+ if (!p->hd_dentry)
1facf9fc 7274+ continue;
7275+
027c5e7a
AM
7276+ new_bindex = au_br_index(sb, p->hd_id);
7277+ if (new_bindex == bindex)
1facf9fc 7278+ continue;
1facf9fc 7279+
1facf9fc 7280+ if (dinfo->di_bwh == bindex)
7281+ bwh = new_bindex;
7282+ if (dinfo->di_bdiropq == bindex)
7283+ bdiropq = new_bindex;
7284+ if (new_bindex < 0) {
7285+ au_hdput(p);
7286+ p->hd_dentry = NULL;
7287+ continue;
7288+ }
7289+
7290+ /* swap two lower dentries, and loop again */
7291+ q = dinfo->di_hdentry + new_bindex;
7292+ tmp = *q;
7293+ *q = *p;
7294+ *p = tmp;
7295+ if (tmp.hd_dentry) {
7296+ bindex--;
7297+ p--;
7298+ }
7299+ }
7300+
1facf9fc 7301+ dinfo->di_bwh = -1;
7302+ if (bwh >= 0 && bwh <= au_sbend(sb) && au_sbr_whable(sb, bwh))
7303+ dinfo->di_bwh = bwh;
7304+
7305+ dinfo->di_bdiropq = -1;
7306+ if (bdiropq >= 0
7307+ && bdiropq <= au_sbend(sb)
7308+ && au_sbr_whable(sb, bdiropq))
7309+ dinfo->di_bdiropq = bdiropq;
7310+
027c5e7a
AM
7311+ err = -EIO;
7312+ dinfo->di_bstart = -1;
7313+ dinfo->di_bend = -1;
1facf9fc 7314+ bend = au_dbend(parent);
7315+ p = dinfo->di_hdentry;
7316+ for (bindex = 0; bindex <= bend; bindex++, p++)
7317+ if (p->hd_dentry) {
7318+ dinfo->di_bstart = bindex;
7319+ break;
7320+ }
7321+
027c5e7a
AM
7322+ if (dinfo->di_bstart >= 0) {
7323+ p = dinfo->di_hdentry + bend;
7324+ for (bindex = bend; bindex >= 0; bindex--, p--)
7325+ if (p->hd_dentry) {
7326+ dinfo->di_bend = bindex;
7327+ err = 0;
7328+ break;
7329+ }
7330+ }
7331+
7332+ return err;
1facf9fc 7333+}
7334+
027c5e7a 7335+static void au_do_hide(struct dentry *dentry)
1facf9fc 7336+{
027c5e7a 7337+ struct inode *inode;
1facf9fc 7338+
027c5e7a
AM
7339+ inode = dentry->d_inode;
7340+ if (inode) {
7341+ if (!S_ISDIR(inode->i_mode)) {
7342+ if (inode->i_nlink && !d_unhashed(dentry))
7343+ drop_nlink(inode);
7344+ } else {
7345+ clear_nlink(inode);
7346+ /* stop next lookup */
7347+ inode->i_flags |= S_DEAD;
7348+ }
7349+ smp_mb(); /* necessary? */
7350+ }
7351+ d_drop(dentry);
7352+}
1308ab2a 7353+
027c5e7a
AM
7354+static int au_hide_children(struct dentry *parent)
7355+{
7356+ int err, i, j, ndentry;
7357+ struct au_dcsub_pages dpages;
7358+ struct au_dpage *dpage;
7359+ struct dentry *dentry;
1facf9fc 7360+
027c5e7a 7361+ err = au_dpages_init(&dpages, GFP_NOFS);
1facf9fc 7362+ if (unlikely(err))
7363+ goto out;
027c5e7a
AM
7364+ err = au_dcsub_pages(&dpages, parent, NULL, NULL);
7365+ if (unlikely(err))
7366+ goto out_dpages;
1facf9fc 7367+
027c5e7a
AM
7368+ /* in reverse order */
7369+ for (i = dpages.ndpage - 1; i >= 0; i--) {
7370+ dpage = dpages.dpages + i;
7371+ ndentry = dpage->ndentry;
7372+ for (j = ndentry - 1; j >= 0; j--) {
7373+ dentry = dpage->dentries[j];
7374+ if (dentry != parent)
7375+ au_do_hide(dentry);
7376+ }
7377+ }
1facf9fc 7378+
027c5e7a
AM
7379+out_dpages:
7380+ au_dpages_free(&dpages);
4f0767ce 7381+out:
027c5e7a 7382+ return err;
1facf9fc 7383+}
7384+
027c5e7a 7385+static void au_hide(struct dentry *dentry)
1facf9fc 7386+{
027c5e7a
AM
7387+ int err;
7388+ struct inode *inode;
1facf9fc 7389+
027c5e7a
AM
7390+ AuDbgDentry(dentry);
7391+ inode = dentry->d_inode;
7392+ if (inode && S_ISDIR(inode->i_mode)) {
7393+ /* shrink_dcache_parent(dentry); */
7394+ err = au_hide_children(dentry);
7395+ if (unlikely(err))
523b37e3
AM
7396+ AuIOErr("%pd, failed hiding children, ignored %d\n",
7397+ dentry, err);
027c5e7a
AM
7398+ }
7399+ au_do_hide(dentry);
7400+}
1facf9fc 7401+
027c5e7a
AM
7402+/*
7403+ * By adding a dirty branch, a cached dentry may be affected in various ways.
7404+ *
7405+ * a dirty branch is added
7406+ * - on the top of layers
7407+ * - in the middle of layers
7408+ * - to the bottom of layers
7409+ *
7410+ * on the added branch there exists
7411+ * - a whiteout
7412+ * - a diropq
7413+ * - a same named entry
7414+ * + exist
7415+ * * negative --> positive
7416+ * * positive --> positive
7417+ * - type is unchanged
7418+ * - type is changed
7419+ * + doesn't exist
7420+ * * negative --> negative
7421+ * * positive --> negative (rejected by au_br_del() for non-dir case)
7422+ * - none
7423+ */
7424+static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
7425+ struct au_dinfo *tmp)
7426+{
7427+ int err;
7428+ aufs_bindex_t bindex, bend;
7429+ struct {
7430+ struct dentry *dentry;
7431+ struct inode *inode;
7432+ mode_t mode;
7433+ } orig_h, tmp_h;
7434+ struct au_hdentry *hd;
7435+ struct inode *inode, *h_inode;
7436+ struct dentry *h_dentry;
7437+
7438+ err = 0;
7439+ AuDebugOn(dinfo->di_bstart < 0);
7440+ orig_h.dentry = dinfo->di_hdentry[dinfo->di_bstart].hd_dentry;
7441+ orig_h.inode = orig_h.dentry->d_inode;
7442+ orig_h.mode = 0;
7443+ if (orig_h.inode)
7444+ orig_h.mode = orig_h.inode->i_mode & S_IFMT;
7445+ memset(&tmp_h, 0, sizeof(tmp_h));
7446+ if (tmp->di_bstart >= 0) {
7447+ tmp_h.dentry = tmp->di_hdentry[tmp->di_bstart].hd_dentry;
7448+ tmp_h.inode = tmp_h.dentry->d_inode;
7449+ if (tmp_h.inode)
7450+ tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
7451+ }
7452+
7453+ inode = dentry->d_inode;
7454+ if (!orig_h.inode) {
7455+ AuDbg("nagative originally\n");
7456+ if (inode) {
7457+ au_hide(dentry);
7458+ goto out;
7459+ }
7460+ AuDebugOn(inode);
7461+ AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
7462+ AuDebugOn(dinfo->di_bdiropq != -1);
7463+
7464+ if (!tmp_h.inode) {
7465+ AuDbg("negative --> negative\n");
7466+ /* should have only one negative lower */
7467+ if (tmp->di_bstart >= 0
7468+ && tmp->di_bstart < dinfo->di_bstart) {
7469+ AuDebugOn(tmp->di_bstart != tmp->di_bend);
7470+ AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
7471+ au_set_h_dptr(dentry, dinfo->di_bstart, NULL);
7472+ au_di_cp(dinfo, tmp);
7473+ hd = tmp->di_hdentry + tmp->di_bstart;
7474+ au_set_h_dptr(dentry, tmp->di_bstart,
7475+ dget(hd->hd_dentry));
7476+ }
7477+ au_dbg_verify_dinode(dentry);
7478+ } else {
7479+ AuDbg("negative --> positive\n");
7480+ /*
7481+ * similar to the behaviour of creating with bypassing
7482+ * aufs.
7483+ * unhash it in order to force an error in the
7484+ * succeeding create operation.
7485+ * we should not set S_DEAD here.
7486+ */
7487+ d_drop(dentry);
7488+ /* au_di_swap(tmp, dinfo); */
7489+ au_dbg_verify_dinode(dentry);
7490+ }
7491+ } else {
7492+ AuDbg("positive originally\n");
7493+ /* inode may be NULL */
7494+ AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
7495+ if (!tmp_h.inode) {
7496+ AuDbg("positive --> negative\n");
7497+ /* or bypassing aufs */
7498+ au_hide(dentry);
7499+ if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_bstart)
7500+ dinfo->di_bwh = tmp->di_bwh;
7501+ if (inode)
7502+ err = au_refresh_hinode_self(inode);
7503+ au_dbg_verify_dinode(dentry);
7504+ } else if (orig_h.mode == tmp_h.mode) {
7505+ AuDbg("positive --> positive, same type\n");
7506+ if (!S_ISDIR(orig_h.mode)
7507+ && dinfo->di_bstart > tmp->di_bstart) {
7508+ /*
7509+ * similar to the behaviour of removing and
7510+ * creating.
7511+ */
7512+ au_hide(dentry);
7513+ if (inode)
7514+ err = au_refresh_hinode_self(inode);
7515+ au_dbg_verify_dinode(dentry);
7516+ } else {
7517+ /* fill empty slots */
7518+ if (dinfo->di_bstart > tmp->di_bstart)
7519+ dinfo->di_bstart = tmp->di_bstart;
7520+ if (dinfo->di_bend < tmp->di_bend)
7521+ dinfo->di_bend = tmp->di_bend;
7522+ dinfo->di_bwh = tmp->di_bwh;
7523+ dinfo->di_bdiropq = tmp->di_bdiropq;
7524+ hd = tmp->di_hdentry;
7525+ bend = dinfo->di_bend;
7526+ for (bindex = tmp->di_bstart; bindex <= bend;
7527+ bindex++) {
7528+ if (au_h_dptr(dentry, bindex))
7529+ continue;
7530+ h_dentry = hd[bindex].hd_dentry;
7531+ if (!h_dentry)
7532+ continue;
7533+ h_inode = h_dentry->d_inode;
7534+ AuDebugOn(!h_inode);
7535+ AuDebugOn(orig_h.mode
7536+ != (h_inode->i_mode
7537+ & S_IFMT));
7538+ au_set_h_dptr(dentry, bindex,
7539+ dget(h_dentry));
7540+ }
7541+ err = au_refresh_hinode(inode, dentry);
7542+ au_dbg_verify_dinode(dentry);
7543+ }
7544+ } else {
7545+ AuDbg("positive --> positive, different type\n");
7546+ /* similar to the behaviour of removing and creating */
7547+ au_hide(dentry);
7548+ if (inode)
7549+ err = au_refresh_hinode_self(inode);
7550+ au_dbg_verify_dinode(dentry);
7551+ }
7552+ }
7553+
7554+out:
7555+ return err;
7556+}
7557+
7558+int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
7559+{
7560+ int err, ebrange;
7561+ unsigned int sigen;
7562+ struct au_dinfo *dinfo, *tmp;
7563+ struct super_block *sb;
7564+ struct inode *inode;
7565+
7566+ DiMustWriteLock(dentry);
7567+ AuDebugOn(IS_ROOT(dentry));
7568+ AuDebugOn(!parent->d_inode);
7569+
7570+ sb = dentry->d_sb;
7571+ inode = dentry->d_inode;
7572+ sigen = au_sigen(sb);
7573+ err = au_digen_test(parent, sigen);
7574+ if (unlikely(err))
7575+ goto out;
7576+
7577+ dinfo = au_di(dentry);
7578+ err = au_di_realloc(dinfo, au_sbend(sb) + 1);
7579+ if (unlikely(err))
7580+ goto out;
7581+ ebrange = au_dbrange_test(dentry);
7582+ if (!ebrange)
7583+ ebrange = au_do_refresh_hdentry(dentry, parent);
7584+
38d290e6 7585+ if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
027c5e7a
AM
7586+ AuDebugOn(au_dbstart(dentry) < 0 && au_dbend(dentry) >= 0);
7587+ if (inode)
7588+ err = au_refresh_hinode_self(inode);
7589+ au_dbg_verify_dinode(dentry);
7590+ if (!err)
7591+ goto out_dgen; /* success */
7592+ goto out;
7593+ }
7594+
7595+ /* temporary dinfo */
7596+ AuDbgDentry(dentry);
7597+ err = -ENOMEM;
7598+ tmp = au_di_alloc(sb, AuLsc_DI_TMP);
7599+ if (unlikely(!tmp))
7600+ goto out;
7601+ au_di_swap(tmp, dinfo);
7602+ /* returns the number of positive dentries */
7603+ /*
7604+ * if current working dir is removed, it returns an error.
7605+ * but the dentry is legal.
7606+ */
537831f9 7607+ err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
027c5e7a
AM
7608+ AuDbgDentry(dentry);
7609+ au_di_swap(tmp, dinfo);
7610+ if (err == -ENOENT)
7611+ err = 0;
7612+ if (err >= 0) {
7613+ /* compare/refresh by dinfo */
7614+ AuDbgDentry(dentry);
7615+ err = au_refresh_by_dinfo(dentry, dinfo, tmp);
7616+ au_dbg_verify_dinode(dentry);
7617+ AuTraceErr(err);
7618+ }
7619+ au_rw_write_unlock(&tmp->di_rwsem);
7620+ au_di_free(tmp);
7621+ if (unlikely(err))
7622+ goto out;
7623+
7624+out_dgen:
7625+ au_update_digen(dentry);
7626+out:
7627+ if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
523b37e3 7628+ AuIOErr("failed refreshing %pd, %d\n", dentry, err);
027c5e7a
AM
7629+ AuDbgDentry(dentry);
7630+ }
7631+ AuTraceErr(err);
7632+ return err;
7633+}
7634+
b4510431
AM
7635+static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
7636+ struct dentry *dentry, aufs_bindex_t bindex)
027c5e7a
AM
7637+{
7638+ int err, valid;
027c5e7a
AM
7639+
7640+ err = 0;
7641+ if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
7642+ goto out;
027c5e7a
AM
7643+
7644+ AuDbg("b%d\n", bindex);
b4510431
AM
7645+ /*
7646+ * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
7647+ * due to whiteout and branch permission.
7648+ */
7649+ flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
7650+ | LOOKUP_FOLLOW | LOOKUP_EXCL);
7651+ /* it may return tri-state */
7652+ valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
1facf9fc 7653+
7654+ if (unlikely(valid < 0))
7655+ err = valid;
7656+ else if (!valid)
7657+ err = -EINVAL;
7658+
4f0767ce 7659+out:
1facf9fc 7660+ AuTraceErr(err);
7661+ return err;
7662+}
7663+
7664+/* todo: remove this */
7665+static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
b4510431 7666+ unsigned int flags, int do_udba)
1facf9fc 7667+{
7668+ int err;
7669+ umode_t mode, h_mode;
7670+ aufs_bindex_t bindex, btail, bstart, ibs, ibe;
38d290e6 7671+ unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
4a4d8108 7672+ struct inode *h_inode, *h_cached_inode;
1facf9fc 7673+ struct dentry *h_dentry;
7674+ struct qstr *name, *h_name;
7675+
7676+ err = 0;
7677+ plus = 0;
7678+ mode = 0;
1facf9fc 7679+ ibs = -1;
7680+ ibe = -1;
7681+ unhashed = !!d_unhashed(dentry);
7682+ is_root = !!IS_ROOT(dentry);
7683+ name = &dentry->d_name;
38d290e6 7684+ tmpfile = au_di(dentry)->di_tmpfile;
1facf9fc 7685+
7686+ /*
7f207e10
AM
7687+ * Theoretically, REVAL test should be unnecessary in case of
7688+ * {FS,I}NOTIFY.
7689+ * But {fs,i}notify doesn't fire some necessary events,
1facf9fc 7690+ * IN_ATTRIB for atime/nlink/pageio
1facf9fc 7691+ * Let's do REVAL test too.
7692+ */
7693+ if (do_udba && inode) {
7694+ mode = (inode->i_mode & S_IFMT);
7695+ plus = (inode->i_nlink > 0);
1facf9fc 7696+ ibs = au_ibstart(inode);
7697+ ibe = au_ibend(inode);
7698+ }
7699+
7700+ bstart = au_dbstart(dentry);
7701+ btail = bstart;
7702+ if (inode && S_ISDIR(inode->i_mode))
7703+ btail = au_dbtaildir(dentry);
7704+ for (bindex = bstart; bindex <= btail; bindex++) {
7705+ h_dentry = au_h_dptr(dentry, bindex);
7706+ if (!h_dentry)
7707+ continue;
7708+
523b37e3
AM
7709+ AuDbg("b%d, %pd\n", bindex, h_dentry);
7710+ h_nfs = !!au_test_nfs(h_dentry->d_sb);
027c5e7a 7711+ spin_lock(&h_dentry->d_lock);
1facf9fc 7712+ h_name = &h_dentry->d_name;
7713+ if (unlikely(do_udba
7714+ && !is_root
523b37e3
AM
7715+ && ((!h_nfs
7716+ && (unhashed != !!d_unhashed(h_dentry)
38d290e6
JR
7717+ || (!tmpfile
7718+ && !au_qstreq(name, h_name))
7719+ ))
523b37e3
AM
7720+ || (h_nfs
7721+ && !(flags & LOOKUP_OPEN)
7722+ && (h_dentry->d_flags
7723+ & DCACHE_NFSFS_RENAMED)))
1facf9fc 7724+ )) {
38d290e6
JR
7725+ int h_unhashed;
7726+
7727+ h_unhashed = d_unhashed(h_dentry);
027c5e7a 7728+ spin_unlock(&h_dentry->d_lock);
38d290e6
JR
7729+ AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
7730+ unhashed, h_unhashed, dentry, h_dentry);
1facf9fc 7731+ goto err;
7732+ }
027c5e7a 7733+ spin_unlock(&h_dentry->d_lock);
1facf9fc 7734+
b4510431 7735+ err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
1facf9fc 7736+ if (unlikely(err))
7737+ /* do not goto err, to keep the errno */
7738+ break;
7739+
7740+ /* todo: plink too? */
7741+ if (!do_udba)
7742+ continue;
7743+
7744+ /* UDBA tests */
7745+ h_inode = h_dentry->d_inode;
7746+ if (unlikely(!!inode != !!h_inode))
7747+ goto err;
7748+
7749+ h_plus = plus;
7750+ h_mode = mode;
7751+ h_cached_inode = h_inode;
7752+ if (h_inode) {
7753+ h_mode = (h_inode->i_mode & S_IFMT);
7754+ h_plus = (h_inode->i_nlink > 0);
7755+ }
7756+ if (inode && ibs <= bindex && bindex <= ibe)
7757+ h_cached_inode = au_h_iptr(inode, bindex);
7758+
523b37e3 7759+ if (!h_nfs) {
38d290e6 7760+ if (unlikely(plus != h_plus && !tmpfile))
523b37e3
AM
7761+ goto err;
7762+ } else {
7763+ if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
7764+ && !is_root
7765+ && !IS_ROOT(h_dentry)
7766+ && unhashed != d_unhashed(h_dentry)))
7767+ goto err;
7768+ }
7769+ if (unlikely(mode != h_mode
1facf9fc 7770+ || h_cached_inode != h_inode))
7771+ goto err;
7772+ continue;
7773+
f6b6e03d 7774+err:
1facf9fc 7775+ err = -EINVAL;
7776+ break;
7777+ }
7778+
523b37e3 7779+ AuTraceErr(err);
1facf9fc 7780+ return err;
7781+}
7782+
027c5e7a 7783+/* todo: consolidate with do_refresh() and au_reval_for_attr() */
1facf9fc 7784+static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
7785+{
7786+ int err;
7787+ struct dentry *parent;
1facf9fc 7788+
027c5e7a 7789+ if (!au_digen_test(dentry, sigen))
1facf9fc 7790+ return 0;
7791+
7792+ parent = dget_parent(dentry);
7793+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a 7794+ AuDebugOn(au_digen_test(parent, sigen));
1facf9fc 7795+ au_dbg_verify_gen(parent, sigen);
027c5e7a 7796+ err = au_refresh_dentry(dentry, parent);
1facf9fc 7797+ di_read_unlock(parent, AuLock_IR);
7798+ dput(parent);
027c5e7a 7799+ AuTraceErr(err);
1facf9fc 7800+ return err;
7801+}
7802+
7803+int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
7804+{
7805+ int err;
7806+ struct dentry *d, *parent;
7807+ struct inode *inode;
7808+
027c5e7a 7809+ if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
1facf9fc 7810+ return simple_reval_dpath(dentry, sigen);
7811+
7812+ /* slow loop, keep it simple and stupid */
7813+ /* cf: au_cpup_dirs() */
7814+ err = 0;
7815+ parent = NULL;
027c5e7a 7816+ while (au_digen_test(dentry, sigen)) {
1facf9fc 7817+ d = dentry;
7818+ while (1) {
7819+ dput(parent);
7820+ parent = dget_parent(d);
027c5e7a 7821+ if (!au_digen_test(parent, sigen))
1facf9fc 7822+ break;
7823+ d = parent;
7824+ }
7825+
7826+ inode = d->d_inode;
7827+ if (d != dentry)
027c5e7a 7828+ di_write_lock_child2(d);
1facf9fc 7829+
7830+ /* someone might update our dentry while we were sleeping */
027c5e7a
AM
7831+ if (au_digen_test(d, sigen)) {
7832+ /*
7833+ * todo: consolidate with simple_reval_dpath(),
7834+ * do_refresh() and au_reval_for_attr().
7835+ */
1facf9fc 7836+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a 7837+ err = au_refresh_dentry(d, parent);
1facf9fc 7838+ di_read_unlock(parent, AuLock_IR);
7839+ }
7840+
7841+ if (d != dentry)
7842+ di_write_unlock(d);
7843+ dput(parent);
7844+ if (unlikely(err))
7845+ break;
7846+ }
7847+
7848+ return err;
7849+}
7850+
7851+/*
7852+ * if valid returns 1, otherwise 0.
7853+ */
b4510431 7854+static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
1facf9fc 7855+{
7856+ int valid, err;
7857+ unsigned int sigen;
7858+ unsigned char do_udba;
7859+ struct super_block *sb;
7860+ struct inode *inode;
7861+
027c5e7a 7862+ /* todo: support rcu-walk? */
b4510431 7863+ if (flags & LOOKUP_RCU)
027c5e7a
AM
7864+ return -ECHILD;
7865+
7866+ valid = 0;
7867+ if (unlikely(!au_di(dentry)))
7868+ goto out;
7869+
7870+ inode = dentry->d_inode;
7871+ if (inode && is_bad_inode(inode))
7872+ goto out;
7873+
e49829fe 7874+ valid = 1;
1facf9fc 7875+ sb = dentry->d_sb;
e49829fe
JR
7876+ /*
7877+ * todo: very ugly
7878+ * i_mutex of parent dir may be held,
7879+ * but we should not return 'invalid' due to busy.
7880+ */
7881+ err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
7882+ if (unlikely(err)) {
7883+ valid = err;
027c5e7a 7884+ AuTraceErr(err);
e49829fe
JR
7885+ goto out;
7886+ }
027c5e7a
AM
7887+ if (unlikely(au_dbrange_test(dentry))) {
7888+ err = -EINVAL;
7889+ AuTraceErr(err);
7890+ goto out_dgrade;
1facf9fc 7891+ }
027c5e7a
AM
7892+
7893+ sigen = au_sigen(sb);
7894+ if (au_digen_test(dentry, sigen)) {
1facf9fc 7895+ AuDebugOn(IS_ROOT(dentry));
027c5e7a
AM
7896+ err = au_reval_dpath(dentry, sigen);
7897+ if (unlikely(err)) {
7898+ AuTraceErr(err);
1facf9fc 7899+ goto out_dgrade;
027c5e7a 7900+ }
1facf9fc 7901+ }
7902+ di_downgrade_lock(dentry, AuLock_IR);
7903+
1facf9fc 7904+ err = -EINVAL;
523b37e3
AM
7905+ if (!(flags & LOOKUP_OPEN)
7906+ && inode
38d290e6 7907+ && !(inode->i_state && I_LINKABLE)
523b37e3 7908+ && (IS_DEADDIR(inode) || !inode->i_nlink))
027c5e7a
AM
7909+ goto out_inval;
7910+
1facf9fc 7911+ do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
7912+ if (do_udba && inode) {
7913+ aufs_bindex_t bstart = au_ibstart(inode);
027c5e7a 7914+ struct inode *h_inode;
1facf9fc 7915+
027c5e7a
AM
7916+ if (bstart >= 0) {
7917+ h_inode = au_h_iptr(inode, bstart);
7918+ if (h_inode && au_test_higen(inode, h_inode))
7919+ goto out_inval;
7920+ }
1facf9fc 7921+ }
7922+
b4510431 7923+ err = h_d_revalidate(dentry, inode, flags, do_udba);
027c5e7a 7924+ if (unlikely(!err && do_udba && au_dbstart(dentry) < 0)) {
1facf9fc 7925+ err = -EIO;
523b37e3
AM
7926+ AuDbg("both of real entry and whiteout found, %p, err %d\n",
7927+ dentry, err);
027c5e7a 7928+ }
e49829fe 7929+ goto out_inval;
1facf9fc 7930+
4f0767ce 7931+out_dgrade:
1facf9fc 7932+ di_downgrade_lock(dentry, AuLock_IR);
e49829fe 7933+out_inval:
1facf9fc 7934+ aufs_read_unlock(dentry, AuLock_IR);
7935+ AuTraceErr(err);
7936+ valid = !err;
e49829fe 7937+out:
027c5e7a 7938+ if (!valid) {
523b37e3 7939+ AuDbg("%pd invalid, %d\n", dentry, valid);
027c5e7a
AM
7940+ d_drop(dentry);
7941+ }
1facf9fc 7942+ return valid;
7943+}
7944+
7945+static void aufs_d_release(struct dentry *dentry)
7946+{
027c5e7a 7947+ if (au_di(dentry)) {
4a4d8108
AM
7948+ au_di_fin(dentry);
7949+ au_hn_di_reinit(dentry);
1facf9fc 7950+ }
1facf9fc 7951+}
7952+
4a4d8108 7953+const struct dentry_operations aufs_dop = {
c06a8ce3
AM
7954+ .d_revalidate = aufs_d_revalidate,
7955+ .d_weak_revalidate = aufs_d_revalidate,
7956+ .d_release = aufs_d_release
1facf9fc 7957+};
7f207e10
AM
7958diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
7959--- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
7960+++ linux/fs/aufs/dentry.h 2014-06-24 00:14:06.779385790 +0200
7961@@ -0,0 +1,234 @@
1facf9fc 7962+/*
523b37e3 7963+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 7964+ *
7965+ * This program, aufs is free software; you can redistribute it and/or modify
7966+ * it under the terms of the GNU General Public License as published by
7967+ * the Free Software Foundation; either version 2 of the License, or
7968+ * (at your option) any later version.
dece6358
AM
7969+ *
7970+ * This program is distributed in the hope that it will be useful,
7971+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
7972+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7973+ * GNU General Public License for more details.
7974+ *
7975+ * You should have received a copy of the GNU General Public License
523b37e3 7976+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 7977+ */
7978+
7979+/*
7980+ * lookup and dentry operations
7981+ */
7982+
7983+#ifndef __AUFS_DENTRY_H__
7984+#define __AUFS_DENTRY_H__
7985+
7986+#ifdef __KERNEL__
7987+
dece6358 7988+#include <linux/dcache.h>
1facf9fc 7989+#include "rwsem.h"
7990+
1facf9fc 7991+struct au_hdentry {
7992+ struct dentry *hd_dentry;
027c5e7a 7993+ aufs_bindex_t hd_id;
1facf9fc 7994+};
7995+
7996+struct au_dinfo {
7997+ atomic_t di_generation;
7998+
dece6358 7999+ struct au_rwsem di_rwsem;
1facf9fc 8000+ aufs_bindex_t di_bstart, di_bend, di_bwh, di_bdiropq;
38d290e6 8001+ unsigned char di_tmpfile; /* to allow the different name */
1facf9fc 8002+ struct au_hdentry *di_hdentry;
4a4d8108 8003+} ____cacheline_aligned_in_smp;
1facf9fc 8004+
8005+/* ---------------------------------------------------------------------- */
8006+
8007+/* dentry.c */
4a4d8108 8008+extern const struct dentry_operations aufs_dop;
1facf9fc 8009+struct au_branch;
1facf9fc 8010+struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent,
8011+ struct au_branch *br);
8012+int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8013+ struct dentry *h_parent, struct au_branch *br);
8014+
537831f9 8015+int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type);
86dc4139 8016+int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
027c5e7a 8017+int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
1facf9fc 8018+int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
8019+
8020+/* dinfo.c */
4a4d8108 8021+void au_di_init_once(void *_di);
027c5e7a
AM
8022+struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
8023+void au_di_free(struct au_dinfo *dinfo);
8024+void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
8025+void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
4a4d8108
AM
8026+int au_di_init(struct dentry *dentry);
8027+void au_di_fin(struct dentry *dentry);
1facf9fc 8028+int au_di_realloc(struct au_dinfo *dinfo, int nbr);
8029+
8030+void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
8031+void di_read_unlock(struct dentry *d, int flags);
8032+void di_downgrade_lock(struct dentry *d, int flags);
8033+void di_write_lock(struct dentry *d, unsigned int lsc);
8034+void di_write_unlock(struct dentry *d);
8035+void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
8036+void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
8037+void di_write_unlock2(struct dentry *d1, struct dentry *d2);
8038+
8039+struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
2cbb1c4b 8040+struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
1facf9fc 8041+aufs_bindex_t au_dbtail(struct dentry *dentry);
8042+aufs_bindex_t au_dbtaildir(struct dentry *dentry);
8043+
8044+void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
8045+ struct dentry *h_dentry);
027c5e7a
AM
8046+int au_digen_test(struct dentry *dentry, unsigned int sigen);
8047+int au_dbrange_test(struct dentry *dentry);
1facf9fc 8048+void au_update_digen(struct dentry *dentry);
8049+void au_update_dbrange(struct dentry *dentry, int do_put_zero);
8050+void au_update_dbstart(struct dentry *dentry);
8051+void au_update_dbend(struct dentry *dentry);
8052+int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
8053+
8054+/* ---------------------------------------------------------------------- */
8055+
8056+static inline struct au_dinfo *au_di(struct dentry *dentry)
8057+{
8058+ return dentry->d_fsdata;
8059+}
8060+
8061+/* ---------------------------------------------------------------------- */
8062+
8063+/* lock subclass for dinfo */
8064+enum {
8065+ AuLsc_DI_CHILD, /* child first */
4a4d8108 8066+ AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
1facf9fc 8067+ AuLsc_DI_CHILD3, /* copyup dirs */
8068+ AuLsc_DI_PARENT,
8069+ AuLsc_DI_PARENT2,
027c5e7a
AM
8070+ AuLsc_DI_PARENT3,
8071+ AuLsc_DI_TMP /* temp for replacing dinfo */
1facf9fc 8072+};
8073+
8074+/*
8075+ * di_read_lock_child, di_write_lock_child,
8076+ * di_read_lock_child2, di_write_lock_child2,
8077+ * di_read_lock_child3, di_write_lock_child3,
8078+ * di_read_lock_parent, di_write_lock_parent,
8079+ * di_read_lock_parent2, di_write_lock_parent2,
8080+ * di_read_lock_parent3, di_write_lock_parent3,
8081+ */
8082+#define AuReadLockFunc(name, lsc) \
8083+static inline void di_read_lock_##name(struct dentry *d, int flags) \
8084+{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
8085+
8086+#define AuWriteLockFunc(name, lsc) \
8087+static inline void di_write_lock_##name(struct dentry *d) \
8088+{ di_write_lock(d, AuLsc_DI_##lsc); }
8089+
8090+#define AuRWLockFuncs(name, lsc) \
8091+ AuReadLockFunc(name, lsc) \
8092+ AuWriteLockFunc(name, lsc)
8093+
8094+AuRWLockFuncs(child, CHILD);
8095+AuRWLockFuncs(child2, CHILD2);
8096+AuRWLockFuncs(child3, CHILD3);
8097+AuRWLockFuncs(parent, PARENT);
8098+AuRWLockFuncs(parent2, PARENT2);
8099+AuRWLockFuncs(parent3, PARENT3);
8100+
8101+#undef AuReadLockFunc
8102+#undef AuWriteLockFunc
8103+#undef AuRWLockFuncs
8104+
8105+#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
dece6358
AM
8106+#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
8107+#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
1facf9fc 8108+
8109+/* ---------------------------------------------------------------------- */
8110+
8111+/* todo: memory barrier? */
8112+static inline unsigned int au_digen(struct dentry *d)
8113+{
8114+ return atomic_read(&au_di(d)->di_generation);
8115+}
8116+
8117+static inline void au_h_dentry_init(struct au_hdentry *hdentry)
8118+{
8119+ hdentry->hd_dentry = NULL;
8120+}
8121+
8122+static inline void au_hdput(struct au_hdentry *hd)
8123+{
4a4d8108
AM
8124+ if (hd)
8125+ dput(hd->hd_dentry);
1facf9fc 8126+}
8127+
8128+static inline aufs_bindex_t au_dbstart(struct dentry *dentry)
8129+{
1308ab2a 8130+ DiMustAnyLock(dentry);
1facf9fc 8131+ return au_di(dentry)->di_bstart;
8132+}
8133+
8134+static inline aufs_bindex_t au_dbend(struct dentry *dentry)
8135+{
1308ab2a 8136+ DiMustAnyLock(dentry);
1facf9fc 8137+ return au_di(dentry)->di_bend;
8138+}
8139+
8140+static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
8141+{
1308ab2a 8142+ DiMustAnyLock(dentry);
1facf9fc 8143+ return au_di(dentry)->di_bwh;
8144+}
8145+
8146+static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
8147+{
1308ab2a 8148+ DiMustAnyLock(dentry);
1facf9fc 8149+ return au_di(dentry)->di_bdiropq;
8150+}
8151+
8152+/* todo: hard/soft set? */
8153+static inline void au_set_dbstart(struct dentry *dentry, aufs_bindex_t bindex)
8154+{
1308ab2a 8155+ DiMustWriteLock(dentry);
1facf9fc 8156+ au_di(dentry)->di_bstart = bindex;
8157+}
8158+
8159+static inline void au_set_dbend(struct dentry *dentry, aufs_bindex_t bindex)
8160+{
1308ab2a 8161+ DiMustWriteLock(dentry);
1facf9fc 8162+ au_di(dentry)->di_bend = bindex;
8163+}
8164+
8165+static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
8166+{
1308ab2a 8167+ DiMustWriteLock(dentry);
1facf9fc 8168+ /* dbwh can be outside of bstart - bend range */
8169+ au_di(dentry)->di_bwh = bindex;
8170+}
8171+
8172+static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
8173+{
1308ab2a 8174+ DiMustWriteLock(dentry);
1facf9fc 8175+ au_di(dentry)->di_bdiropq = bindex;
8176+}
8177+
8178+/* ---------------------------------------------------------------------- */
8179+
4a4d8108 8180+#ifdef CONFIG_AUFS_HNOTIFY
1facf9fc 8181+static inline void au_digen_dec(struct dentry *d)
8182+{
e49829fe 8183+ atomic_dec(&au_di(d)->di_generation);
1facf9fc 8184+}
8185+
4a4d8108 8186+static inline void au_hn_di_reinit(struct dentry *dentry)
1facf9fc 8187+{
8188+ dentry->d_fsdata = NULL;
8189+}
8190+#else
4a4d8108
AM
8191+AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
8192+#endif /* CONFIG_AUFS_HNOTIFY */
1facf9fc 8193+
8194+#endif /* __KERNEL__ */
8195+#endif /* __AUFS_DENTRY_H__ */
7f207e10
AM
8196diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
8197--- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
8198+++ linux/fs/aufs/dinfo.c 2014-06-24 00:14:06.779385790 +0200
8199@@ -0,0 +1,544 @@
1facf9fc 8200+/*
523b37e3 8201+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 8202+ *
8203+ * This program, aufs is free software; you can redistribute it and/or modify
8204+ * it under the terms of the GNU General Public License as published by
8205+ * the Free Software Foundation; either version 2 of the License, or
8206+ * (at your option) any later version.
dece6358
AM
8207+ *
8208+ * This program is distributed in the hope that it will be useful,
8209+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
8210+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8211+ * GNU General Public License for more details.
8212+ *
8213+ * You should have received a copy of the GNU General Public License
523b37e3 8214+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 8215+ */
8216+
8217+/*
8218+ * dentry private data
8219+ */
8220+
8221+#include "aufs.h"
8222+
e49829fe 8223+void au_di_init_once(void *_dinfo)
4a4d8108 8224+{
e49829fe
JR
8225+ struct au_dinfo *dinfo = _dinfo;
8226+ static struct lock_class_key aufs_di;
4a4d8108 8227+
e49829fe
JR
8228+ au_rw_init(&dinfo->di_rwsem);
8229+ au_rw_class(&dinfo->di_rwsem, &aufs_di);
4a4d8108
AM
8230+}
8231+
027c5e7a 8232+struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
1facf9fc 8233+{
8234+ struct au_dinfo *dinfo;
027c5e7a 8235+ int nbr, i;
1facf9fc 8236+
8237+ dinfo = au_cache_alloc_dinfo();
8238+ if (unlikely(!dinfo))
8239+ goto out;
8240+
1facf9fc 8241+ nbr = au_sbend(sb) + 1;
8242+ if (nbr <= 0)
8243+ nbr = 1;
8244+ dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
027c5e7a
AM
8245+ if (dinfo->di_hdentry) {
8246+ au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
8247+ dinfo->di_bstart = -1;
8248+ dinfo->di_bend = -1;
8249+ dinfo->di_bwh = -1;
8250+ dinfo->di_bdiropq = -1;
38d290e6 8251+ dinfo->di_tmpfile = 0;
027c5e7a
AM
8252+ for (i = 0; i < nbr; i++)
8253+ dinfo->di_hdentry[i].hd_id = -1;
8254+ goto out;
8255+ }
1facf9fc 8256+
1facf9fc 8257+ au_cache_free_dinfo(dinfo);
027c5e7a
AM
8258+ dinfo = NULL;
8259+
4f0767ce 8260+out:
027c5e7a 8261+ return dinfo;
1facf9fc 8262+}
8263+
027c5e7a 8264+void au_di_free(struct au_dinfo *dinfo)
4a4d8108 8265+{
4a4d8108
AM
8266+ struct au_hdentry *p;
8267+ aufs_bindex_t bend, bindex;
8268+
8269+ /* dentry may not be revalidated */
027c5e7a 8270+ bindex = dinfo->di_bstart;
4a4d8108 8271+ if (bindex >= 0) {
027c5e7a
AM
8272+ bend = dinfo->di_bend;
8273+ p = dinfo->di_hdentry + bindex;
4a4d8108
AM
8274+ while (bindex++ <= bend)
8275+ au_hdput(p++);
8276+ }
027c5e7a
AM
8277+ kfree(dinfo->di_hdentry);
8278+ au_cache_free_dinfo(dinfo);
8279+}
8280+
8281+void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
8282+{
8283+ struct au_hdentry *p;
8284+ aufs_bindex_t bi;
8285+
8286+ AuRwMustWriteLock(&a->di_rwsem);
8287+ AuRwMustWriteLock(&b->di_rwsem);
8288+
8289+#define DiSwap(v, name) \
8290+ do { \
8291+ v = a->di_##name; \
8292+ a->di_##name = b->di_##name; \
8293+ b->di_##name = v; \
8294+ } while (0)
8295+
8296+ DiSwap(p, hdentry);
8297+ DiSwap(bi, bstart);
8298+ DiSwap(bi, bend);
8299+ DiSwap(bi, bwh);
8300+ DiSwap(bi, bdiropq);
8301+ /* smp_mb(); */
8302+
8303+#undef DiSwap
8304+}
8305+
8306+void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
8307+{
8308+ AuRwMustWriteLock(&dst->di_rwsem);
8309+ AuRwMustWriteLock(&src->di_rwsem);
8310+
8311+ dst->di_bstart = src->di_bstart;
8312+ dst->di_bend = src->di_bend;
8313+ dst->di_bwh = src->di_bwh;
8314+ dst->di_bdiropq = src->di_bdiropq;
8315+ /* smp_mb(); */
8316+}
8317+
8318+int au_di_init(struct dentry *dentry)
8319+{
8320+ int err;
8321+ struct super_block *sb;
8322+ struct au_dinfo *dinfo;
8323+
8324+ err = 0;
8325+ sb = dentry->d_sb;
8326+ dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
8327+ if (dinfo) {
8328+ atomic_set(&dinfo->di_generation, au_sigen(sb));
8329+ /* smp_mb(); */ /* atomic_set */
8330+ dentry->d_fsdata = dinfo;
8331+ } else
8332+ err = -ENOMEM;
8333+
8334+ return err;
8335+}
8336+
8337+void au_di_fin(struct dentry *dentry)
8338+{
8339+ struct au_dinfo *dinfo;
8340+
8341+ dinfo = au_di(dentry);
8342+ AuRwDestroy(&dinfo->di_rwsem);
8343+ au_di_free(dinfo);
4a4d8108
AM
8344+}
8345+
1facf9fc 8346+int au_di_realloc(struct au_dinfo *dinfo, int nbr)
8347+{
8348+ int err, sz;
8349+ struct au_hdentry *hdp;
8350+
1308ab2a 8351+ AuRwMustWriteLock(&dinfo->di_rwsem);
8352+
1facf9fc 8353+ err = -ENOMEM;
8354+ sz = sizeof(*hdp) * (dinfo->di_bend + 1);
8355+ if (!sz)
8356+ sz = sizeof(*hdp);
8357+ hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS);
8358+ if (hdp) {
8359+ dinfo->di_hdentry = hdp;
8360+ err = 0;
8361+ }
8362+
8363+ return err;
8364+}
8365+
8366+/* ---------------------------------------------------------------------- */
8367+
8368+static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
8369+{
8370+ switch (lsc) {
8371+ case AuLsc_DI_CHILD:
8372+ ii_write_lock_child(inode);
8373+ break;
8374+ case AuLsc_DI_CHILD2:
8375+ ii_write_lock_child2(inode);
8376+ break;
8377+ case AuLsc_DI_CHILD3:
8378+ ii_write_lock_child3(inode);
8379+ break;
8380+ case AuLsc_DI_PARENT:
8381+ ii_write_lock_parent(inode);
8382+ break;
8383+ case AuLsc_DI_PARENT2:
8384+ ii_write_lock_parent2(inode);
8385+ break;
8386+ case AuLsc_DI_PARENT3:
8387+ ii_write_lock_parent3(inode);
8388+ break;
8389+ default:
8390+ BUG();
8391+ }
8392+}
8393+
8394+static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
8395+{
8396+ switch (lsc) {
8397+ case AuLsc_DI_CHILD:
8398+ ii_read_lock_child(inode);
8399+ break;
8400+ case AuLsc_DI_CHILD2:
8401+ ii_read_lock_child2(inode);
8402+ break;
8403+ case AuLsc_DI_CHILD3:
8404+ ii_read_lock_child3(inode);
8405+ break;
8406+ case AuLsc_DI_PARENT:
8407+ ii_read_lock_parent(inode);
8408+ break;
8409+ case AuLsc_DI_PARENT2:
8410+ ii_read_lock_parent2(inode);
8411+ break;
8412+ case AuLsc_DI_PARENT3:
8413+ ii_read_lock_parent3(inode);
8414+ break;
8415+ default:
8416+ BUG();
8417+ }
8418+}
8419+
8420+void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
8421+{
dece6358 8422+ au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
1facf9fc 8423+ if (d->d_inode) {
8424+ if (au_ftest_lock(flags, IW))
8425+ do_ii_write_lock(d->d_inode, lsc);
8426+ else if (au_ftest_lock(flags, IR))
8427+ do_ii_read_lock(d->d_inode, lsc);
8428+ }
8429+}
8430+
8431+void di_read_unlock(struct dentry *d, int flags)
8432+{
8433+ if (d->d_inode) {
027c5e7a
AM
8434+ if (au_ftest_lock(flags, IW)) {
8435+ au_dbg_verify_dinode(d);
1facf9fc 8436+ ii_write_unlock(d->d_inode);
027c5e7a
AM
8437+ } else if (au_ftest_lock(flags, IR)) {
8438+ au_dbg_verify_dinode(d);
1facf9fc 8439+ ii_read_unlock(d->d_inode);
027c5e7a 8440+ }
1facf9fc 8441+ }
dece6358 8442+ au_rw_read_unlock(&au_di(d)->di_rwsem);
1facf9fc 8443+}
8444+
8445+void di_downgrade_lock(struct dentry *d, int flags)
8446+{
1facf9fc 8447+ if (d->d_inode && au_ftest_lock(flags, IR))
8448+ ii_downgrade_lock(d->d_inode);
dece6358 8449+ au_rw_dgrade_lock(&au_di(d)->di_rwsem);
1facf9fc 8450+}
8451+
8452+void di_write_lock(struct dentry *d, unsigned int lsc)
8453+{
dece6358 8454+ au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
1facf9fc 8455+ if (d->d_inode)
8456+ do_ii_write_lock(d->d_inode, lsc);
8457+}
8458+
8459+void di_write_unlock(struct dentry *d)
8460+{
027c5e7a 8461+ au_dbg_verify_dinode(d);
1facf9fc 8462+ if (d->d_inode)
8463+ ii_write_unlock(d->d_inode);
dece6358 8464+ au_rw_write_unlock(&au_di(d)->di_rwsem);
1facf9fc 8465+}
8466+
8467+void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
8468+{
8469+ AuDebugOn(d1 == d2
8470+ || d1->d_inode == d2->d_inode
8471+ || d1->d_sb != d2->d_sb);
8472+
8473+ if (isdir && au_test_subdir(d1, d2)) {
8474+ di_write_lock_child(d1);
8475+ di_write_lock_child2(d2);
8476+ } else {
8477+ /* there should be no races */
8478+ di_write_lock_child(d2);
8479+ di_write_lock_child2(d1);
8480+ }
8481+}
8482+
8483+void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
8484+{
8485+ AuDebugOn(d1 == d2
8486+ || d1->d_inode == d2->d_inode
8487+ || d1->d_sb != d2->d_sb);
8488+
8489+ if (isdir && au_test_subdir(d1, d2)) {
8490+ di_write_lock_parent(d1);
8491+ di_write_lock_parent2(d2);
8492+ } else {
8493+ /* there should be no races */
8494+ di_write_lock_parent(d2);
8495+ di_write_lock_parent2(d1);
8496+ }
8497+}
8498+
8499+void di_write_unlock2(struct dentry *d1, struct dentry *d2)
8500+{
8501+ di_write_unlock(d1);
8502+ if (d1->d_inode == d2->d_inode)
dece6358 8503+ au_rw_write_unlock(&au_di(d2)->di_rwsem);
1facf9fc 8504+ else
8505+ di_write_unlock(d2);
8506+}
8507+
8508+/* ---------------------------------------------------------------------- */
8509+
8510+struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
8511+{
8512+ struct dentry *d;
8513+
1308ab2a 8514+ DiMustAnyLock(dentry);
8515+
1facf9fc 8516+ if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
8517+ return NULL;
8518+ AuDebugOn(bindex < 0);
8519+ d = au_di(dentry)->di_hdentry[0 + bindex].hd_dentry;
392086de 8520+ AuDebugOn(d && d_count(d) <= 0);
1facf9fc 8521+ return d;
8522+}
8523+
2cbb1c4b
JR
8524+/*
8525+ * extended version of au_h_dptr().
38d290e6
JR
8526+ * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
8527+ * error.
2cbb1c4b
JR
8528+ */
8529+struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
8530+{
8531+ struct dentry *h_dentry;
8532+ struct inode *inode, *h_inode;
8533+
8534+ inode = dentry->d_inode;
8535+ AuDebugOn(!inode);
8536+
8537+ h_dentry = NULL;
8538+ if (au_dbstart(dentry) <= bindex
8539+ && bindex <= au_dbend(dentry))
8540+ h_dentry = au_h_dptr(dentry, bindex);
38d290e6 8541+ if (h_dentry && !au_d_linkable(h_dentry)) {
2cbb1c4b
JR
8542+ dget(h_dentry);
8543+ goto out; /* success */
8544+ }
8545+
8546+ AuDebugOn(bindex < au_ibstart(inode));
8547+ AuDebugOn(au_ibend(inode) < bindex);
8548+ h_inode = au_h_iptr(inode, bindex);
8549+ h_dentry = d_find_alias(h_inode);
8550+ if (h_dentry) {
8551+ if (!IS_ERR(h_dentry)) {
38d290e6 8552+ if (!au_d_linkable(h_dentry))
2cbb1c4b
JR
8553+ goto out; /* success */
8554+ dput(h_dentry);
8555+ } else
8556+ goto out;
8557+ }
8558+
8559+ if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
8560+ h_dentry = au_plink_lkup(inode, bindex);
8561+ AuDebugOn(!h_dentry);
8562+ if (!IS_ERR(h_dentry)) {
8563+ if (!au_d_hashed_positive(h_dentry))
8564+ goto out; /* success */
8565+ dput(h_dentry);
8566+ h_dentry = NULL;
8567+ }
8568+ }
8569+
8570+out:
8571+ AuDbgDentry(h_dentry);
8572+ return h_dentry;
8573+}
8574+
1facf9fc 8575+aufs_bindex_t au_dbtail(struct dentry *dentry)
8576+{
8577+ aufs_bindex_t bend, bwh;
8578+
8579+ bend = au_dbend(dentry);
8580+ if (0 <= bend) {
8581+ bwh = au_dbwh(dentry);
8582+ if (!bwh)
8583+ return bwh;
8584+ if (0 < bwh && bwh < bend)
8585+ return bwh - 1;
8586+ }
8587+ return bend;
8588+}
8589+
8590+aufs_bindex_t au_dbtaildir(struct dentry *dentry)
8591+{
8592+ aufs_bindex_t bend, bopq;
8593+
8594+ bend = au_dbtail(dentry);
8595+ if (0 <= bend) {
8596+ bopq = au_dbdiropq(dentry);
8597+ if (0 <= bopq && bopq < bend)
8598+ bend = bopq;
8599+ }
8600+ return bend;
8601+}
8602+
8603+/* ---------------------------------------------------------------------- */
8604+
8605+void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
8606+ struct dentry *h_dentry)
8607+{
8608+ struct au_hdentry *hd = au_di(dentry)->di_hdentry + bindex;
027c5e7a 8609+ struct au_branch *br;
1facf9fc 8610+
1308ab2a 8611+ DiMustWriteLock(dentry);
8612+
4a4d8108 8613+ au_hdput(hd);
1facf9fc 8614+ hd->hd_dentry = h_dentry;
027c5e7a
AM
8615+ if (h_dentry) {
8616+ br = au_sbr(dentry->d_sb, bindex);
8617+ hd->hd_id = br->br_id;
8618+ }
8619+}
8620+
8621+int au_dbrange_test(struct dentry *dentry)
8622+{
8623+ int err;
8624+ aufs_bindex_t bstart, bend;
8625+
8626+ err = 0;
8627+ bstart = au_dbstart(dentry);
8628+ bend = au_dbend(dentry);
8629+ if (bstart >= 0)
8630+ AuDebugOn(bend < 0 && bstart > bend);
8631+ else {
8632+ err = -EIO;
8633+ AuDebugOn(bend >= 0);
8634+ }
8635+
8636+ return err;
8637+}
8638+
8639+int au_digen_test(struct dentry *dentry, unsigned int sigen)
8640+{
8641+ int err;
8642+
8643+ err = 0;
8644+ if (unlikely(au_digen(dentry) != sigen
8645+ || au_iigen_test(dentry->d_inode, sigen)))
8646+ err = -EIO;
8647+
8648+ return err;
1facf9fc 8649+}
8650+
8651+void au_update_digen(struct dentry *dentry)
8652+{
8653+ atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
8654+ /* smp_mb(); */ /* atomic_set */
8655+}
8656+
8657+void au_update_dbrange(struct dentry *dentry, int do_put_zero)
8658+{
8659+ struct au_dinfo *dinfo;
8660+ struct dentry *h_d;
4a4d8108 8661+ struct au_hdentry *hdp;
1facf9fc 8662+
1308ab2a 8663+ DiMustWriteLock(dentry);
8664+
1facf9fc 8665+ dinfo = au_di(dentry);
8666+ if (!dinfo || dinfo->di_bstart < 0)
8667+ return;
8668+
4a4d8108 8669+ hdp = dinfo->di_hdentry;
1facf9fc 8670+ if (do_put_zero) {
8671+ aufs_bindex_t bindex, bend;
8672+
8673+ bend = dinfo->di_bend;
8674+ for (bindex = dinfo->di_bstart; bindex <= bend; bindex++) {
4a4d8108 8675+ h_d = hdp[0 + bindex].hd_dentry;
1facf9fc 8676+ if (h_d && !h_d->d_inode)
8677+ au_set_h_dptr(dentry, bindex, NULL);
8678+ }
8679+ }
8680+
8681+ dinfo->di_bstart = -1;
8682+ while (++dinfo->di_bstart <= dinfo->di_bend)
4a4d8108 8683+ if (hdp[0 + dinfo->di_bstart].hd_dentry)
1facf9fc 8684+ break;
8685+ if (dinfo->di_bstart > dinfo->di_bend) {
8686+ dinfo->di_bstart = -1;
8687+ dinfo->di_bend = -1;
8688+ return;
8689+ }
8690+
8691+ dinfo->di_bend++;
8692+ while (0 <= --dinfo->di_bend)
4a4d8108 8693+ if (hdp[0 + dinfo->di_bend].hd_dentry)
1facf9fc 8694+ break;
8695+ AuDebugOn(dinfo->di_bstart > dinfo->di_bend || dinfo->di_bend < 0);
8696+}
8697+
8698+void au_update_dbstart(struct dentry *dentry)
8699+{
8700+ aufs_bindex_t bindex, bend;
8701+ struct dentry *h_dentry;
8702+
8703+ bend = au_dbend(dentry);
8704+ for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
8705+ h_dentry = au_h_dptr(dentry, bindex);
8706+ if (!h_dentry)
8707+ continue;
8708+ if (h_dentry->d_inode) {
8709+ au_set_dbstart(dentry, bindex);
8710+ return;
8711+ }
8712+ au_set_h_dptr(dentry, bindex, NULL);
8713+ }
8714+}
8715+
8716+void au_update_dbend(struct dentry *dentry)
8717+{
8718+ aufs_bindex_t bindex, bstart;
8719+ struct dentry *h_dentry;
8720+
8721+ bstart = au_dbstart(dentry);
7f207e10 8722+ for (bindex = au_dbend(dentry); bindex >= bstart; bindex--) {
1facf9fc 8723+ h_dentry = au_h_dptr(dentry, bindex);
8724+ if (!h_dentry)
8725+ continue;
8726+ if (h_dentry->d_inode) {
8727+ au_set_dbend(dentry, bindex);
8728+ return;
8729+ }
8730+ au_set_h_dptr(dentry, bindex, NULL);
8731+ }
8732+}
8733+
8734+int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
8735+{
8736+ aufs_bindex_t bindex, bend;
8737+
8738+ bend = au_dbend(dentry);
8739+ for (bindex = au_dbstart(dentry); bindex <= bend; bindex++)
8740+ if (au_h_dptr(dentry, bindex) == h_dentry)
8741+ return bindex;
8742+ return -1;
8743+}
7f207e10
AM
8744diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
8745--- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 8746+++ linux/fs/aufs/dir.c 2014-06-24 00:14:06.779385790 +0200
523b37e3 8747@@ -0,0 +1,639 @@
1facf9fc 8748+/*
523b37e3 8749+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 8750+ *
8751+ * This program, aufs is free software; you can redistribute it and/or modify
8752+ * it under the terms of the GNU General Public License as published by
8753+ * the Free Software Foundation; either version 2 of the License, or
8754+ * (at your option) any later version.
dece6358
AM
8755+ *
8756+ * This program is distributed in the hope that it will be useful,
8757+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
8758+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8759+ * GNU General Public License for more details.
8760+ *
8761+ * You should have received a copy of the GNU General Public License
523b37e3 8762+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 8763+ */
8764+
8765+/*
8766+ * directory operations
8767+ */
8768+
8769+#include <linux/fs_stack.h>
8770+#include "aufs.h"
8771+
8772+void au_add_nlink(struct inode *dir, struct inode *h_dir)
8773+{
9dbd164d
AM
8774+ unsigned int nlink;
8775+
1facf9fc 8776+ AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
8777+
9dbd164d
AM
8778+ nlink = dir->i_nlink;
8779+ nlink += h_dir->i_nlink - 2;
1facf9fc 8780+ if (h_dir->i_nlink < 2)
9dbd164d 8781+ nlink += 2;
f6b6e03d 8782+ smp_mb(); /* for i_nlink */
7eafdf33 8783+ /* 0 can happen in revaliding */
92d182d2 8784+ set_nlink(dir, nlink);
1facf9fc 8785+}
8786+
8787+void au_sub_nlink(struct inode *dir, struct inode *h_dir)
8788+{
9dbd164d
AM
8789+ unsigned int nlink;
8790+
1facf9fc 8791+ AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
8792+
9dbd164d
AM
8793+ nlink = dir->i_nlink;
8794+ nlink -= h_dir->i_nlink - 2;
1facf9fc 8795+ if (h_dir->i_nlink < 2)
9dbd164d 8796+ nlink -= 2;
f6b6e03d 8797+ smp_mb(); /* for i_nlink */
92d182d2 8798+ /* nlink == 0 means the branch-fs is broken */
9dbd164d 8799+ set_nlink(dir, nlink);
1facf9fc 8800+}
8801+
1308ab2a 8802+loff_t au_dir_size(struct file *file, struct dentry *dentry)
8803+{
8804+ loff_t sz;
8805+ aufs_bindex_t bindex, bend;
8806+ struct file *h_file;
8807+ struct dentry *h_dentry;
8808+
8809+ sz = 0;
8810+ if (file) {
c06a8ce3
AM
8811+ AuDebugOn(!file_inode(file));
8812+ AuDebugOn(!S_ISDIR(file_inode(file)->i_mode));
1308ab2a 8813+
4a4d8108 8814+ bend = au_fbend_dir(file);
1308ab2a 8815+ for (bindex = au_fbstart(file);
8816+ bindex <= bend && sz < KMALLOC_MAX_SIZE;
8817+ bindex++) {
4a4d8108 8818+ h_file = au_hf_dir(file, bindex);
c06a8ce3
AM
8819+ if (h_file && file_inode(h_file))
8820+ sz += vfsub_f_size_read(h_file);
1308ab2a 8821+ }
8822+ } else {
8823+ AuDebugOn(!dentry);
8824+ AuDebugOn(!dentry->d_inode);
8825+ AuDebugOn(!S_ISDIR(dentry->d_inode->i_mode));
8826+
8827+ bend = au_dbtaildir(dentry);
8828+ for (bindex = au_dbstart(dentry);
8829+ bindex <= bend && sz < KMALLOC_MAX_SIZE;
8830+ bindex++) {
8831+ h_dentry = au_h_dptr(dentry, bindex);
8832+ if (h_dentry && h_dentry->d_inode)
8833+ sz += i_size_read(h_dentry->d_inode);
8834+ }
8835+ }
8836+ if (sz < KMALLOC_MAX_SIZE)
8837+ sz = roundup_pow_of_two(sz);
8838+ if (sz > KMALLOC_MAX_SIZE)
8839+ sz = KMALLOC_MAX_SIZE;
8840+ else if (sz < NAME_MAX) {
8841+ BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
8842+ sz = AUFS_RDBLK_DEF;
8843+ }
8844+ return sz;
8845+}
8846+
1facf9fc 8847+/* ---------------------------------------------------------------------- */
8848+
8849+static int reopen_dir(struct file *file)
8850+{
8851+ int err;
8852+ unsigned int flags;
8853+ aufs_bindex_t bindex, btail, bstart;
8854+ struct dentry *dentry, *h_dentry;
8855+ struct file *h_file;
8856+
8857+ /* open all lower dirs */
8858+ dentry = file->f_dentry;
8859+ bstart = au_dbstart(dentry);
8860+ for (bindex = au_fbstart(file); bindex < bstart; bindex++)
8861+ au_set_h_fptr(file, bindex, NULL);
8862+ au_set_fbstart(file, bstart);
8863+
8864+ btail = au_dbtaildir(dentry);
4a4d8108 8865+ for (bindex = au_fbend_dir(file); btail < bindex; bindex--)
1facf9fc 8866+ au_set_h_fptr(file, bindex, NULL);
4a4d8108 8867+ au_set_fbend_dir(file, btail);
1facf9fc 8868+
4a4d8108 8869+ flags = vfsub_file_flags(file);
1facf9fc 8870+ for (bindex = bstart; bindex <= btail; bindex++) {
8871+ h_dentry = au_h_dptr(dentry, bindex);
8872+ if (!h_dentry)
8873+ continue;
4a4d8108 8874+ h_file = au_hf_dir(file, bindex);
1facf9fc 8875+ if (h_file)
8876+ continue;
8877+
392086de 8878+ h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
1facf9fc 8879+ err = PTR_ERR(h_file);
8880+ if (IS_ERR(h_file))
8881+ goto out; /* close all? */
8882+ au_set_h_fptr(file, bindex, h_file);
8883+ }
8884+ au_update_figen(file);
8885+ /* todo: necessary? */
8886+ /* file->f_ra = h_file->f_ra; */
8887+ err = 0;
8888+
4f0767ce 8889+out:
1facf9fc 8890+ return err;
8891+}
8892+
8893+static int do_open_dir(struct file *file, int flags)
8894+{
8895+ int err;
8896+ aufs_bindex_t bindex, btail;
8897+ struct dentry *dentry, *h_dentry;
8898+ struct file *h_file;
8899+
1308ab2a 8900+ FiMustWriteLock(file);
8901+
523b37e3 8902+ err = 0;
1facf9fc 8903+ dentry = file->f_dentry;
1facf9fc 8904+ file->f_version = dentry->d_inode->i_version;
8905+ bindex = au_dbstart(dentry);
8906+ au_set_fbstart(file, bindex);
8907+ btail = au_dbtaildir(dentry);
4a4d8108 8908+ au_set_fbend_dir(file, btail);
1facf9fc 8909+ for (; !err && bindex <= btail; bindex++) {
8910+ h_dentry = au_h_dptr(dentry, bindex);
8911+ if (!h_dentry)
8912+ continue;
8913+
392086de 8914+ h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
1facf9fc 8915+ if (IS_ERR(h_file)) {
8916+ err = PTR_ERR(h_file);
8917+ break;
8918+ }
8919+ au_set_h_fptr(file, bindex, h_file);
8920+ }
8921+ au_update_figen(file);
8922+ /* todo: necessary? */
8923+ /* file->f_ra = h_file->f_ra; */
8924+ if (!err)
8925+ return 0; /* success */
8926+
8927+ /* close all */
8928+ for (bindex = au_fbstart(file); bindex <= btail; bindex++)
8929+ au_set_h_fptr(file, bindex, NULL);
8930+ au_set_fbstart(file, -1);
4a4d8108
AM
8931+ au_set_fbend_dir(file, -1);
8932+
1facf9fc 8933+ return err;
8934+}
8935+
8936+static int aufs_open_dir(struct inode *inode __maybe_unused,
8937+ struct file *file)
8938+{
4a4d8108
AM
8939+ int err;
8940+ struct super_block *sb;
8941+ struct au_fidir *fidir;
8942+
8943+ err = -ENOMEM;
8944+ sb = file->f_dentry->d_sb;
8945+ si_read_lock(sb, AuLock_FLUSH);
e49829fe 8946+ fidir = au_fidir_alloc(sb);
4a4d8108
AM
8947+ if (fidir) {
8948+ err = au_do_open(file, do_open_dir, fidir);
8949+ if (unlikely(err))
8950+ kfree(fidir);
8951+ }
8952+ si_read_unlock(sb);
8953+ return err;
1facf9fc 8954+}
8955+
8956+static int aufs_release_dir(struct inode *inode __maybe_unused,
8957+ struct file *file)
8958+{
8959+ struct au_vdir *vdir_cache;
4a4d8108
AM
8960+ struct au_finfo *finfo;
8961+ struct au_fidir *fidir;
8962+ aufs_bindex_t bindex, bend;
1facf9fc 8963+
4a4d8108
AM
8964+ finfo = au_fi(file);
8965+ fidir = finfo->fi_hdir;
8966+ if (fidir) {
8967+ vdir_cache = fidir->fd_vdir_cache; /* lock-free */
8968+ if (vdir_cache)
8969+ au_vdir_free(vdir_cache);
8970+
8971+ bindex = finfo->fi_btop;
8972+ if (bindex >= 0) {
8973+ /*
8974+ * calls fput() instead of filp_close(),
8975+ * since no dnotify or lock for the lower file.
8976+ */
8977+ bend = fidir->fd_bbot;
8978+ for (; bindex <= bend; bindex++)
8979+ au_set_h_fptr(file, bindex, NULL);
8980+ }
8981+ kfree(fidir);
8982+ finfo->fi_hdir = NULL;
1facf9fc 8983+ }
1facf9fc 8984+ au_finfo_fin(file);
1facf9fc 8985+ return 0;
8986+}
8987+
8988+/* ---------------------------------------------------------------------- */
8989+
4a4d8108
AM
8990+static int au_do_flush_dir(struct file *file, fl_owner_t id)
8991+{
8992+ int err;
8993+ aufs_bindex_t bindex, bend;
8994+ struct file *h_file;
8995+
8996+ err = 0;
8997+ bend = au_fbend_dir(file);
8998+ for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
8999+ h_file = au_hf_dir(file, bindex);
9000+ if (h_file)
9001+ err = vfsub_flush(h_file, id);
9002+ }
9003+ return err;
9004+}
9005+
9006+static int aufs_flush_dir(struct file *file, fl_owner_t id)
9007+{
9008+ return au_do_flush(file, id, au_do_flush_dir);
9009+}
9010+
9011+/* ---------------------------------------------------------------------- */
9012+
1facf9fc 9013+static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
9014+{
9015+ int err;
9016+ aufs_bindex_t bend, bindex;
9017+ struct inode *inode;
9018+ struct super_block *sb;
9019+
9020+ err = 0;
9021+ sb = dentry->d_sb;
9022+ inode = dentry->d_inode;
9023+ IMustLock(inode);
9024+ bend = au_dbend(dentry);
9025+ for (bindex = au_dbstart(dentry); !err && bindex <= bend; bindex++) {
9026+ struct path h_path;
1facf9fc 9027+
9028+ if (au_test_ro(sb, bindex, inode))
9029+ continue;
9030+ h_path.dentry = au_h_dptr(dentry, bindex);
9031+ if (!h_path.dentry)
9032+ continue;
1facf9fc 9033+
1facf9fc 9034+ h_path.mnt = au_sbr_mnt(sb, bindex);
53392da6 9035+ err = vfsub_fsync(NULL, &h_path, datasync);
1facf9fc 9036+ }
9037+
9038+ return err;
9039+}
9040+
9041+static int au_do_fsync_dir(struct file *file, int datasync)
9042+{
9043+ int err;
9044+ aufs_bindex_t bend, bindex;
9045+ struct file *h_file;
9046+ struct super_block *sb;
9047+ struct inode *inode;
1facf9fc 9048+
9049+ err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9050+ if (unlikely(err))
9051+ goto out;
9052+
9053+ sb = file->f_dentry->d_sb;
c06a8ce3 9054+ inode = file_inode(file);
4a4d8108 9055+ bend = au_fbend_dir(file);
1facf9fc 9056+ for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
4a4d8108 9057+ h_file = au_hf_dir(file, bindex);
1facf9fc 9058+ if (!h_file || au_test_ro(sb, bindex, inode))
9059+ continue;
9060+
53392da6 9061+ err = vfsub_fsync(h_file, &h_file->f_path, datasync);
1facf9fc 9062+ }
9063+
4f0767ce 9064+out:
1facf9fc 9065+ return err;
9066+}
9067+
9068+/*
9069+ * @file may be NULL
9070+ */
1e00d052
AM
9071+static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
9072+ int datasync)
1facf9fc 9073+{
9074+ int err;
b752ccd1 9075+ struct dentry *dentry;
1facf9fc 9076+ struct super_block *sb;
1e00d052 9077+ struct mutex *mtx;
1facf9fc 9078+
9079+ err = 0;
1e00d052
AM
9080+ dentry = file->f_dentry;
9081+ mtx = &dentry->d_inode->i_mutex;
9082+ mutex_lock(mtx);
1facf9fc 9083+ sb = dentry->d_sb;
9084+ si_noflush_read_lock(sb);
9085+ if (file)
9086+ err = au_do_fsync_dir(file, datasync);
9087+ else {
9088+ di_write_lock_child(dentry);
9089+ err = au_do_fsync_dir_no_file(dentry, datasync);
9090+ }
9091+ au_cpup_attr_timesizes(dentry->d_inode);
9092+ di_write_unlock(dentry);
9093+ if (file)
9094+ fi_write_unlock(file);
9095+
9096+ si_read_unlock(sb);
1e00d052 9097+ mutex_unlock(mtx);
1facf9fc 9098+ return err;
9099+}
9100+
9101+/* ---------------------------------------------------------------------- */
9102+
392086de 9103+static int aufs_iterate(struct file *file, struct dir_context *ctx)
1facf9fc 9104+{
9105+ int err;
9106+ struct dentry *dentry;
9dbd164d 9107+ struct inode *inode, *h_inode;
1facf9fc 9108+ struct super_block *sb;
9109+
523b37e3 9110+ AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
392086de 9111+
1facf9fc 9112+ dentry = file->f_dentry;
9113+ inode = dentry->d_inode;
9114+ IMustLock(inode);
9115+
9116+ sb = dentry->d_sb;
9117+ si_read_lock(sb, AuLock_FLUSH);
9118+ err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9119+ if (unlikely(err))
9120+ goto out;
027c5e7a
AM
9121+ err = au_alive_dir(dentry);
9122+ if (!err)
9123+ err = au_vdir_init(file);
1facf9fc 9124+ di_downgrade_lock(dentry, AuLock_IR);
9125+ if (unlikely(err))
9126+ goto out_unlock;
9127+
9dbd164d 9128+ h_inode = au_h_iptr(inode, au_ibstart(inode));
b752ccd1 9129+ if (!au_test_nfsd()) {
392086de 9130+ err = au_vdir_fill_de(file, ctx);
9dbd164d 9131+ fsstack_copy_attr_atime(inode, h_inode);
1facf9fc 9132+ } else {
9133+ /*
9134+ * nfsd filldir may call lookup_one_len(), vfs_getattr(),
9135+ * encode_fh() and others.
9136+ */
9dbd164d 9137+ atomic_inc(&h_inode->i_count);
1facf9fc 9138+ di_read_unlock(dentry, AuLock_IR);
9139+ si_read_unlock(sb);
392086de 9140+ err = au_vdir_fill_de(file, ctx);
1facf9fc 9141+ fsstack_copy_attr_atime(inode, h_inode);
9142+ fi_write_unlock(file);
9dbd164d 9143+ iput(h_inode);
1facf9fc 9144+
9145+ AuTraceErr(err);
9146+ return err;
9147+ }
9148+
4f0767ce 9149+out_unlock:
1facf9fc 9150+ di_read_unlock(dentry, AuLock_IR);
9151+ fi_write_unlock(file);
4f0767ce 9152+out:
1facf9fc 9153+ si_read_unlock(sb);
9154+ return err;
9155+}
9156+
9157+/* ---------------------------------------------------------------------- */
9158+
9159+#define AuTestEmpty_WHONLY 1
dece6358
AM
9160+#define AuTestEmpty_CALLED (1 << 1)
9161+#define AuTestEmpty_SHWH (1 << 2)
1facf9fc 9162+#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
7f207e10
AM
9163+#define au_fset_testempty(flags, name) \
9164+ do { (flags) |= AuTestEmpty_##name; } while (0)
9165+#define au_fclr_testempty(flags, name) \
9166+ do { (flags) &= ~AuTestEmpty_##name; } while (0)
1facf9fc 9167+
dece6358
AM
9168+#ifndef CONFIG_AUFS_SHWH
9169+#undef AuTestEmpty_SHWH
9170+#define AuTestEmpty_SHWH 0
9171+#endif
9172+
1facf9fc 9173+struct test_empty_arg {
392086de 9174+ struct dir_context ctx;
1308ab2a 9175+ struct au_nhash *whlist;
1facf9fc 9176+ unsigned int flags;
9177+ int err;
9178+ aufs_bindex_t bindex;
9179+};
9180+
392086de
AM
9181+static int test_empty_cb(struct dir_context *ctx, const char *__name,
9182+ int namelen, loff_t offset __maybe_unused, u64 ino,
dece6358 9183+ unsigned int d_type)
1facf9fc 9184+{
392086de
AM
9185+ struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
9186+ ctx);
1facf9fc 9187+ char *name = (void *)__name;
9188+
9189+ arg->err = 0;
9190+ au_fset_testempty(arg->flags, CALLED);
9191+ /* smp_mb(); */
9192+ if (name[0] == '.'
9193+ && (namelen == 1 || (name[1] == '.' && namelen == 2)))
9194+ goto out; /* success */
9195+
9196+ if (namelen <= AUFS_WH_PFX_LEN
9197+ || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
9198+ if (au_ftest_testempty(arg->flags, WHONLY)
1308ab2a 9199+ && !au_nhash_test_known_wh(arg->whlist, name, namelen))
1facf9fc 9200+ arg->err = -ENOTEMPTY;
9201+ goto out;
9202+ }
9203+
9204+ name += AUFS_WH_PFX_LEN;
9205+ namelen -= AUFS_WH_PFX_LEN;
1308ab2a 9206+ if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
1facf9fc 9207+ arg->err = au_nhash_append_wh
1308ab2a 9208+ (arg->whlist, name, namelen, ino, d_type, arg->bindex,
dece6358 9209+ au_ftest_testempty(arg->flags, SHWH));
1facf9fc 9210+
4f0767ce 9211+out:
1facf9fc 9212+ /* smp_mb(); */
9213+ AuTraceErr(arg->err);
9214+ return arg->err;
9215+}
9216+
9217+static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9218+{
9219+ int err;
9220+ struct file *h_file;
9221+
9222+ h_file = au_h_open(dentry, arg->bindex,
9223+ O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
392086de 9224+ /*file*/NULL, /*force_wr*/0);
1facf9fc 9225+ err = PTR_ERR(h_file);
9226+ if (IS_ERR(h_file))
9227+ goto out;
9228+
9229+ err = 0;
9230+ if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
c06a8ce3 9231+ && !file_inode(h_file)->i_nlink)
1facf9fc 9232+ goto out_put;
9233+
9234+ do {
9235+ arg->err = 0;
9236+ au_fclr_testempty(arg->flags, CALLED);
9237+ /* smp_mb(); */
392086de 9238+ err = vfsub_iterate_dir(h_file, &arg->ctx);
1facf9fc 9239+ if (err >= 0)
9240+ err = arg->err;
9241+ } while (!err && au_ftest_testempty(arg->flags, CALLED));
9242+
4f0767ce 9243+out_put:
1facf9fc 9244+ fput(h_file);
9245+ au_sbr_put(dentry->d_sb, arg->bindex);
4f0767ce 9246+out:
1facf9fc 9247+ return err;
9248+}
9249+
9250+struct do_test_empty_args {
9251+ int *errp;
9252+ struct dentry *dentry;
9253+ struct test_empty_arg *arg;
9254+};
9255+
9256+static void call_do_test_empty(void *args)
9257+{
9258+ struct do_test_empty_args *a = args;
9259+ *a->errp = do_test_empty(a->dentry, a->arg);
9260+}
9261+
9262+static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9263+{
9264+ int err, wkq_err;
9265+ struct dentry *h_dentry;
9266+ struct inode *h_inode;
9267+
9268+ h_dentry = au_h_dptr(dentry, arg->bindex);
9269+ h_inode = h_dentry->d_inode;
53392da6 9270+ /* todo: i_mode changes anytime? */
1facf9fc 9271+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
9272+ err = au_test_h_perm_sio(h_inode, MAY_EXEC | MAY_READ);
9273+ mutex_unlock(&h_inode->i_mutex);
9274+ if (!err)
9275+ err = do_test_empty(dentry, arg);
9276+ else {
9277+ struct do_test_empty_args args = {
9278+ .errp = &err,
9279+ .dentry = dentry,
9280+ .arg = arg
9281+ };
9282+ unsigned int flags = arg->flags;
9283+
9284+ wkq_err = au_wkq_wait(call_do_test_empty, &args);
9285+ if (unlikely(wkq_err))
9286+ err = wkq_err;
9287+ arg->flags = flags;
9288+ }
9289+
9290+ return err;
9291+}
9292+
9293+int au_test_empty_lower(struct dentry *dentry)
9294+{
9295+ int err;
1308ab2a 9296+ unsigned int rdhash;
1facf9fc 9297+ aufs_bindex_t bindex, bstart, btail;
1308ab2a 9298+ struct au_nhash whlist;
392086de
AM
9299+ struct test_empty_arg arg = {
9300+ .ctx = {
9301+ .actor = au_diractor(test_empty_cb)
9302+ }
9303+ };
1facf9fc 9304+
dece6358
AM
9305+ SiMustAnyLock(dentry->d_sb);
9306+
1308ab2a 9307+ rdhash = au_sbi(dentry->d_sb)->si_rdhash;
9308+ if (!rdhash)
9309+ rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
9310+ err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
dece6358 9311+ if (unlikely(err))
1facf9fc 9312+ goto out;
9313+
1facf9fc 9314+ arg.flags = 0;
1308ab2a 9315+ arg.whlist = &whlist;
9316+ bstart = au_dbstart(dentry);
dece6358
AM
9317+ if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
9318+ au_fset_testempty(arg.flags, SHWH);
1facf9fc 9319+ arg.bindex = bstart;
9320+ err = do_test_empty(dentry, &arg);
9321+ if (unlikely(err))
9322+ goto out_whlist;
9323+
9324+ au_fset_testempty(arg.flags, WHONLY);
9325+ btail = au_dbtaildir(dentry);
9326+ for (bindex = bstart + 1; !err && bindex <= btail; bindex++) {
9327+ struct dentry *h_dentry;
9328+
9329+ h_dentry = au_h_dptr(dentry, bindex);
9330+ if (h_dentry && h_dentry->d_inode) {
9331+ arg.bindex = bindex;
9332+ err = do_test_empty(dentry, &arg);
9333+ }
9334+ }
9335+
4f0767ce 9336+out_whlist:
1308ab2a 9337+ au_nhash_wh_free(&whlist);
4f0767ce 9338+out:
1facf9fc 9339+ return err;
9340+}
9341+
9342+int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
9343+{
9344+ int err;
392086de
AM
9345+ struct test_empty_arg arg = {
9346+ .ctx = {
9347+ .actor = au_diractor(test_empty_cb)
9348+ }
9349+ };
1facf9fc 9350+ aufs_bindex_t bindex, btail;
9351+
9352+ err = 0;
1308ab2a 9353+ arg.whlist = whlist;
1facf9fc 9354+ arg.flags = AuTestEmpty_WHONLY;
dece6358
AM
9355+ if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
9356+ au_fset_testempty(arg.flags, SHWH);
1facf9fc 9357+ btail = au_dbtaildir(dentry);
9358+ for (bindex = au_dbstart(dentry); !err && bindex <= btail; bindex++) {
9359+ struct dentry *h_dentry;
9360+
9361+ h_dentry = au_h_dptr(dentry, bindex);
9362+ if (h_dentry && h_dentry->d_inode) {
9363+ arg.bindex = bindex;
9364+ err = sio_test_empty(dentry, &arg);
9365+ }
9366+ }
9367+
9368+ return err;
9369+}
9370+
9371+/* ---------------------------------------------------------------------- */
9372+
9373+const struct file_operations aufs_dir_fop = {
4a4d8108 9374+ .owner = THIS_MODULE,
027c5e7a 9375+ .llseek = default_llseek,
1facf9fc 9376+ .read = generic_read_dir,
392086de 9377+ .iterate = aufs_iterate,
1facf9fc 9378+ .unlocked_ioctl = aufs_ioctl_dir,
b752ccd1
AM
9379+#ifdef CONFIG_COMPAT
9380+ .compat_ioctl = aufs_compat_ioctl_dir,
9381+#endif
1facf9fc 9382+ .open = aufs_open_dir,
9383+ .release = aufs_release_dir,
4a4d8108 9384+ .flush = aufs_flush_dir,
1facf9fc 9385+ .fsync = aufs_fsync_dir
9386+};
7f207e10
AM
9387diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
9388--- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 9389+++ linux/fs/aufs/dir.h 2014-06-24 00:14:06.779385790 +0200
523b37e3 9390@@ -0,0 +1,136 @@
1facf9fc 9391+/*
523b37e3 9392+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 9393+ *
9394+ * This program, aufs is free software; you can redistribute it and/or modify
9395+ * it under the terms of the GNU General Public License as published by
9396+ * the Free Software Foundation; either version 2 of the License, or
9397+ * (at your option) any later version.
dece6358
AM
9398+ *
9399+ * This program is distributed in the hope that it will be useful,
9400+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
9401+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9402+ * GNU General Public License for more details.
9403+ *
9404+ * You should have received a copy of the GNU General Public License
523b37e3 9405+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 9406+ */
9407+
9408+/*
9409+ * directory operations
9410+ */
9411+
9412+#ifndef __AUFS_DIR_H__
9413+#define __AUFS_DIR_H__
9414+
9415+#ifdef __KERNEL__
9416+
9417+#include <linux/fs.h>
1facf9fc 9418+
9419+/* ---------------------------------------------------------------------- */
9420+
9421+/* need to be faster and smaller */
9422+
9423+struct au_nhash {
dece6358
AM
9424+ unsigned int nh_num;
9425+ struct hlist_head *nh_head;
1facf9fc 9426+};
9427+
9428+struct au_vdir_destr {
9429+ unsigned char len;
9430+ unsigned char name[0];
9431+} __packed;
9432+
9433+struct au_vdir_dehstr {
9434+ struct hlist_node hash;
9435+ struct au_vdir_destr *str;
4a4d8108 9436+} ____cacheline_aligned_in_smp;
1facf9fc 9437+
9438+struct au_vdir_de {
9439+ ino_t de_ino;
9440+ unsigned char de_type;
9441+ /* caution: packed */
9442+ struct au_vdir_destr de_str;
9443+} __packed;
9444+
9445+struct au_vdir_wh {
9446+ struct hlist_node wh_hash;
dece6358
AM
9447+#ifdef CONFIG_AUFS_SHWH
9448+ ino_t wh_ino;
1facf9fc 9449+ aufs_bindex_t wh_bindex;
dece6358
AM
9450+ unsigned char wh_type;
9451+#else
9452+ aufs_bindex_t wh_bindex;
9453+#endif
9454+ /* caution: packed */
1facf9fc 9455+ struct au_vdir_destr wh_str;
9456+} __packed;
9457+
9458+union au_vdir_deblk_p {
9459+ unsigned char *deblk;
9460+ struct au_vdir_de *de;
9461+};
9462+
9463+struct au_vdir {
9464+ unsigned char **vd_deblk;
9465+ unsigned long vd_nblk;
1facf9fc 9466+ struct {
9467+ unsigned long ul;
9468+ union au_vdir_deblk_p p;
9469+ } vd_last;
9470+
9471+ unsigned long vd_version;
dece6358 9472+ unsigned int vd_deblk_sz;
1facf9fc 9473+ unsigned long vd_jiffy;
4a4d8108 9474+} ____cacheline_aligned_in_smp;
1facf9fc 9475+
9476+/* ---------------------------------------------------------------------- */
9477+
9478+/* dir.c */
9479+extern const struct file_operations aufs_dir_fop;
9480+void au_add_nlink(struct inode *dir, struct inode *h_dir);
9481+void au_sub_nlink(struct inode *dir, struct inode *h_dir);
1308ab2a 9482+loff_t au_dir_size(struct file *file, struct dentry *dentry);
1facf9fc 9483+int au_test_empty_lower(struct dentry *dentry);
9484+int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
9485+
9486+/* vdir.c */
1308ab2a 9487+unsigned int au_rdhash_est(loff_t sz);
dece6358
AM
9488+int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
9489+void au_nhash_wh_free(struct au_nhash *whlist);
1facf9fc 9490+int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
9491+ int limit);
dece6358
AM
9492+int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
9493+int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
9494+ unsigned int d_type, aufs_bindex_t bindex,
9495+ unsigned char shwh);
1facf9fc 9496+void au_vdir_free(struct au_vdir *vdir);
9497+int au_vdir_init(struct file *file);
392086de 9498+int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
1facf9fc 9499+
9500+/* ioctl.c */
9501+long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
9502+
1308ab2a 9503+#ifdef CONFIG_AUFS_RDU
9504+/* rdu.c */
9505+long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
b752ccd1
AM
9506+#ifdef CONFIG_COMPAT
9507+long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
9508+ unsigned long arg);
9509+#endif
1308ab2a 9510+#else
9511+static inline long au_rdu_ioctl(struct file *file, unsigned int cmd,
9512+ unsigned long arg)
9513+{
9514+ return -EINVAL;
9515+}
b752ccd1
AM
9516+#ifdef CONFIG_COMPAT
9517+static inline long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
9518+ unsigned long arg)
9519+{
9520+ return -EINVAL;
9521+}
9522+#endif
1308ab2a 9523+#endif
9524+
1facf9fc 9525+#endif /* __KERNEL__ */
9526+#endif /* __AUFS_DIR_H__ */
7f207e10
AM
9527diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
9528--- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 9529+++ linux/fs/aufs/dynop.c 2014-06-24 00:14:06.782719306 +0200
523b37e3 9530@@ -0,0 +1,379 @@
1facf9fc 9531+/*
523b37e3 9532+ * Copyright (C) 2010-2014 Junjiro R. Okajima
1facf9fc 9533+ *
9534+ * This program, aufs is free software; you can redistribute it and/or modify
9535+ * it under the terms of the GNU General Public License as published by
9536+ * the Free Software Foundation; either version 2 of the License, or
9537+ * (at your option) any later version.
dece6358
AM
9538+ *
9539+ * This program is distributed in the hope that it will be useful,
9540+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
9541+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9542+ * GNU General Public License for more details.
9543+ *
9544+ * You should have received a copy of the GNU General Public License
523b37e3 9545+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 9546+ */
9547+
9548+/*
4a4d8108 9549+ * dynamically customizable operations for regular files
1facf9fc 9550+ */
9551+
1facf9fc 9552+#include "aufs.h"
9553+
4a4d8108 9554+#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
1facf9fc 9555+
4a4d8108
AM
9556+/*
9557+ * How large will these lists be?
9558+ * Usually just a few elements, 20-30 at most for each, I guess.
9559+ */
9560+static struct au_splhead dynop[AuDyLast];
9561+
9562+static struct au_dykey *dy_gfind_get(struct au_splhead *spl, const void *h_op)
1facf9fc 9563+{
4a4d8108
AM
9564+ struct au_dykey *key, *tmp;
9565+ struct list_head *head;
1facf9fc 9566+
4a4d8108
AM
9567+ key = NULL;
9568+ head = &spl->head;
9569+ rcu_read_lock();
9570+ list_for_each_entry_rcu(tmp, head, dk_list)
9571+ if (tmp->dk_op.dy_hop == h_op) {
9572+ key = tmp;
9573+ kref_get(&key->dk_kref);
9574+ break;
9575+ }
9576+ rcu_read_unlock();
9577+
9578+ return key;
1facf9fc 9579+}
9580+
4a4d8108 9581+static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
1facf9fc 9582+{
4a4d8108
AM
9583+ struct au_dykey **k, *found;
9584+ const void *h_op = key->dk_op.dy_hop;
9585+ int i;
1facf9fc 9586+
4a4d8108
AM
9587+ found = NULL;
9588+ k = br->br_dykey;
9589+ for (i = 0; i < AuBrDynOp; i++)
9590+ if (k[i]) {
9591+ if (k[i]->dk_op.dy_hop == h_op) {
9592+ found = k[i];
9593+ break;
9594+ }
9595+ } else
9596+ break;
9597+ if (!found) {
9598+ spin_lock(&br->br_dykey_lock);
9599+ for (; i < AuBrDynOp; i++)
9600+ if (k[i]) {
9601+ if (k[i]->dk_op.dy_hop == h_op) {
9602+ found = k[i];
9603+ break;
9604+ }
9605+ } else {
9606+ k[i] = key;
9607+ break;
9608+ }
9609+ spin_unlock(&br->br_dykey_lock);
9610+ BUG_ON(i == AuBrDynOp); /* expand the array */
9611+ }
9612+
9613+ return found;
1facf9fc 9614+}
9615+
4a4d8108
AM
9616+/* kref_get() if @key is already added */
9617+static struct au_dykey *dy_gadd(struct au_splhead *spl, struct au_dykey *key)
9618+{
9619+ struct au_dykey *tmp, *found;
9620+ struct list_head *head;
9621+ const void *h_op = key->dk_op.dy_hop;
1facf9fc 9622+
4a4d8108
AM
9623+ found = NULL;
9624+ head = &spl->head;
9625+ spin_lock(&spl->spin);
9626+ list_for_each_entry(tmp, head, dk_list)
9627+ if (tmp->dk_op.dy_hop == h_op) {
9628+ kref_get(&tmp->dk_kref);
9629+ found = tmp;
9630+ break;
9631+ }
9632+ if (!found)
9633+ list_add_rcu(&key->dk_list, head);
9634+ spin_unlock(&spl->spin);
1facf9fc 9635+
4a4d8108
AM
9636+ if (!found)
9637+ DyPrSym(key);
9638+ return found;
9639+}
9640+
9641+static void dy_free_rcu(struct rcu_head *rcu)
1facf9fc 9642+{
4a4d8108
AM
9643+ struct au_dykey *key;
9644+
9645+ key = container_of(rcu, struct au_dykey, dk_rcu);
9646+ DyPrSym(key);
9647+ kfree(key);
1facf9fc 9648+}
9649+
4a4d8108
AM
9650+static void dy_free(struct kref *kref)
9651+{
9652+ struct au_dykey *key;
9653+ struct au_splhead *spl;
1facf9fc 9654+
4a4d8108
AM
9655+ key = container_of(kref, struct au_dykey, dk_kref);
9656+ spl = dynop + key->dk_op.dy_type;
9657+ au_spl_del_rcu(&key->dk_list, spl);
9658+ call_rcu(&key->dk_rcu, dy_free_rcu);
9659+}
9660+
9661+void au_dy_put(struct au_dykey *key)
1facf9fc 9662+{
4a4d8108
AM
9663+ kref_put(&key->dk_kref, dy_free);
9664+}
1facf9fc 9665+
4a4d8108
AM
9666+/* ---------------------------------------------------------------------- */
9667+
9668+#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
9669+
9670+#ifdef CONFIG_AUFS_DEBUG
9671+#define DyDbgDeclare(cnt) unsigned int cnt = 0
4f0767ce 9672+#define DyDbgInc(cnt) do { cnt++; } while (0)
4a4d8108
AM
9673+#else
9674+#define DyDbgDeclare(cnt) do {} while (0)
9675+#define DyDbgInc(cnt) do {} while (0)
9676+#endif
9677+
9678+#define DySet(func, dst, src, h_op, h_sb) do { \
9679+ DyDbgInc(cnt); \
9680+ if (h_op->func) { \
9681+ if (src.func) \
9682+ dst.func = src.func; \
9683+ else \
9684+ AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
9685+ } \
9686+} while (0)
9687+
9688+#define DySetForce(func, dst, src) do { \
9689+ AuDebugOn(!src.func); \
9690+ DyDbgInc(cnt); \
9691+ dst.func = src.func; \
9692+} while (0)
9693+
9694+#define DySetAop(func) \
9695+ DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
9696+#define DySetAopForce(func) \
9697+ DySetForce(func, dyaop->da_op, aufs_aop)
9698+
9699+static void dy_aop(struct au_dykey *key, const void *h_op,
9700+ struct super_block *h_sb __maybe_unused)
9701+{
9702+ struct au_dyaop *dyaop = (void *)key;
9703+ const struct address_space_operations *h_aop = h_op;
9704+ DyDbgDeclare(cnt);
9705+
9706+ AuDbg("%s\n", au_sbtype(h_sb));
9707+
9708+ DySetAop(writepage);
9709+ DySetAopForce(readpage); /* force */
4a4d8108
AM
9710+ DySetAop(writepages);
9711+ DySetAop(set_page_dirty);
9712+ DySetAop(readpages);
9713+ DySetAop(write_begin);
9714+ DySetAop(write_end);
9715+ DySetAop(bmap);
9716+ DySetAop(invalidatepage);
9717+ DySetAop(releasepage);
027c5e7a 9718+ DySetAop(freepage);
4a4d8108
AM
9719+ /* these two will be changed according to an aufs mount option */
9720+ DySetAop(direct_IO);
9721+ DySetAop(get_xip_mem);
9722+ DySetAop(migratepage);
9723+ DySetAop(launder_page);
9724+ DySetAop(is_partially_uptodate);
392086de 9725+ DySetAop(is_dirty_writeback);
4a4d8108 9726+ DySetAop(error_remove_page);
b4510431
AM
9727+ DySetAop(swap_activate);
9728+ DySetAop(swap_deactivate);
4a4d8108
AM
9729+
9730+ DyDbgSize(cnt, *h_aop);
9731+ dyaop->da_get_xip_mem = h_aop->get_xip_mem;
9732+}
9733+
4a4d8108
AM
9734+/* ---------------------------------------------------------------------- */
9735+
9736+static void dy_bug(struct kref *kref)
9737+{
9738+ BUG();
9739+}
9740+
9741+static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
9742+{
9743+ struct au_dykey *key, *old;
9744+ struct au_splhead *spl;
b752ccd1 9745+ struct op {
4a4d8108 9746+ unsigned int sz;
b752ccd1
AM
9747+ void (*set)(struct au_dykey *key, const void *h_op,
9748+ struct super_block *h_sb __maybe_unused);
9749+ };
9750+ static const struct op a[] = {
4a4d8108
AM
9751+ [AuDy_AOP] = {
9752+ .sz = sizeof(struct au_dyaop),
b752ccd1 9753+ .set = dy_aop
4a4d8108 9754+ }
b752ccd1
AM
9755+ };
9756+ const struct op *p;
4a4d8108
AM
9757+
9758+ spl = dynop + op->dy_type;
9759+ key = dy_gfind_get(spl, op->dy_hop);
9760+ if (key)
9761+ goto out_add; /* success */
9762+
9763+ p = a + op->dy_type;
9764+ key = kzalloc(p->sz, GFP_NOFS);
9765+ if (unlikely(!key)) {
9766+ key = ERR_PTR(-ENOMEM);
9767+ goto out;
9768+ }
9769+
9770+ key->dk_op.dy_hop = op->dy_hop;
9771+ kref_init(&key->dk_kref);
86dc4139 9772+ p->set(key, op->dy_hop, au_br_sb(br));
4a4d8108
AM
9773+ old = dy_gadd(spl, key);
9774+ if (old) {
9775+ kfree(key);
9776+ key = old;
9777+ }
9778+
9779+out_add:
9780+ old = dy_bradd(br, key);
9781+ if (old)
9782+ /* its ref-count should never be zero here */
9783+ kref_put(&key->dk_kref, dy_bug);
9784+out:
9785+ return key;
9786+}
9787+
9788+/* ---------------------------------------------------------------------- */
9789+/*
9790+ * Aufs prohibits O_DIRECT by defaut even if the branch supports it.
9791+ * This behaviour is neccessary to return an error from open(O_DIRECT) instead
9792+ * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
9793+ * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
9794+ * See the aufs manual in detail.
9795+ *
9796+ * To keep this behaviour, aufs has to set NULL to ->get_xip_mem too, and the
9797+ * performance of fadvise() and madvise() may be affected.
9798+ */
9799+static void dy_adx(struct au_dyaop *dyaop, int do_dx)
9800+{
9801+ if (!do_dx) {
9802+ dyaop->da_op.direct_IO = NULL;
9803+ dyaop->da_op.get_xip_mem = NULL;
9804+ } else {
9805+ dyaop->da_op.direct_IO = aufs_aop.direct_IO;
9806+ dyaop->da_op.get_xip_mem = aufs_aop.get_xip_mem;
9807+ if (!dyaop->da_get_xip_mem)
9808+ dyaop->da_op.get_xip_mem = NULL;
9809+ }
9810+}
9811+
9812+static struct au_dyaop *dy_aget(struct au_branch *br,
9813+ const struct address_space_operations *h_aop,
9814+ int do_dx)
9815+{
9816+ struct au_dyaop *dyaop;
9817+ struct au_dynop op;
9818+
9819+ op.dy_type = AuDy_AOP;
9820+ op.dy_haop = h_aop;
9821+ dyaop = (void *)dy_get(&op, br);
9822+ if (IS_ERR(dyaop))
9823+ goto out;
9824+ dy_adx(dyaop, do_dx);
9825+
9826+out:
9827+ return dyaop;
9828+}
9829+
9830+int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
9831+ struct inode *h_inode)
9832+{
9833+ int err, do_dx;
9834+ struct super_block *sb;
9835+ struct au_branch *br;
9836+ struct au_dyaop *dyaop;
9837+
9838+ AuDebugOn(!S_ISREG(h_inode->i_mode));
9839+ IiMustWriteLock(inode);
9840+
9841+ sb = inode->i_sb;
9842+ br = au_sbr(sb, bindex);
9843+ do_dx = !!au_opt_test(au_mntflags(sb), DIO);
9844+ dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
9845+ err = PTR_ERR(dyaop);
9846+ if (IS_ERR(dyaop))
9847+ /* unnecessary to call dy_fput() */
9848+ goto out;
9849+
9850+ err = 0;
9851+ inode->i_mapping->a_ops = &dyaop->da_op;
9852+
9853+out:
9854+ return err;
9855+}
9856+
b752ccd1
AM
9857+/*
9858+ * Is it safe to replace a_ops during the inode/file is in operation?
9859+ * Yes, I hope so.
9860+ */
9861+int au_dy_irefresh(struct inode *inode)
9862+{
9863+ int err;
9864+ aufs_bindex_t bstart;
9865+ struct inode *h_inode;
9866+
9867+ err = 0;
9868+ if (S_ISREG(inode->i_mode)) {
9869+ bstart = au_ibstart(inode);
9870+ h_inode = au_h_iptr(inode, bstart);
9871+ err = au_dy_iaop(inode, bstart, h_inode);
9872+ }
9873+ return err;
9874+}
9875+
4a4d8108
AM
9876+void au_dy_arefresh(int do_dx)
9877+{
9878+ struct au_splhead *spl;
9879+ struct list_head *head;
9880+ struct au_dykey *key;
9881+
9882+ spl = dynop + AuDy_AOP;
9883+ head = &spl->head;
9884+ spin_lock(&spl->spin);
9885+ list_for_each_entry(key, head, dk_list)
9886+ dy_adx((void *)key, do_dx);
9887+ spin_unlock(&spl->spin);
9888+}
9889+
4a4d8108
AM
9890+/* ---------------------------------------------------------------------- */
9891+
9892+void __init au_dy_init(void)
9893+{
9894+ int i;
9895+
9896+ /* make sure that 'struct au_dykey *' can be any type */
9897+ BUILD_BUG_ON(offsetof(struct au_dyaop, da_key));
4a4d8108
AM
9898+
9899+ for (i = 0; i < AuDyLast; i++)
9900+ au_spl_init(dynop + i);
9901+}
9902+
9903+void au_dy_fin(void)
9904+{
9905+ int i;
9906+
9907+ for (i = 0; i < AuDyLast; i++)
9908+ WARN_ON(!list_empty(&dynop[i].head));
9909+}
7f207e10
AM
9910diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
9911--- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 9912+++ linux/fs/aufs/dynop.h 2014-06-24 00:14:06.782719306 +0200
523b37e3 9913@@ -0,0 +1,75 @@
4a4d8108 9914+/*
523b37e3 9915+ * Copyright (C) 2010-2014 Junjiro R. Okajima
4a4d8108
AM
9916+ *
9917+ * This program, aufs is free software; you can redistribute it and/or modify
9918+ * it under the terms of the GNU General Public License as published by
9919+ * the Free Software Foundation; either version 2 of the License, or
9920+ * (at your option) any later version.
9921+ *
9922+ * This program is distributed in the hope that it will be useful,
9923+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
9924+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9925+ * GNU General Public License for more details.
9926+ *
9927+ * You should have received a copy of the GNU General Public License
523b37e3 9928+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
9929+ */
9930+
9931+/*
9932+ * dynamically customizable operations (for regular files only)
9933+ */
9934+
9935+#ifndef __AUFS_DYNOP_H__
9936+#define __AUFS_DYNOP_H__
9937+
9938+#ifdef __KERNEL__
9939+
4a4d8108
AM
9940+#include "inode.h"
9941+
2cbb1c4b 9942+enum {AuDy_AOP, AuDyLast};
4a4d8108
AM
9943+
9944+struct au_dynop {
9945+ int dy_type;
9946+ union {
9947+ const void *dy_hop;
9948+ const struct address_space_operations *dy_haop;
4a4d8108
AM
9949+ };
9950+};
9951+
9952+struct au_dykey {
9953+ union {
9954+ struct list_head dk_list;
9955+ struct rcu_head dk_rcu;
9956+ };
9957+ struct au_dynop dk_op;
9958+
9959+ /*
9960+ * during I am in the branch local array, kref is gotten. when the
9961+ * branch is removed, kref is put.
9962+ */
9963+ struct kref dk_kref;
9964+};
9965+
9966+/* stop unioning since their sizes are very different from each other */
9967+struct au_dyaop {
9968+ struct au_dykey da_key;
9969+ struct address_space_operations da_op; /* not const */
9970+ int (*da_get_xip_mem)(struct address_space *, pgoff_t, int,
9971+ void **, unsigned long *);
9972+};
9973+
4a4d8108
AM
9974+/* ---------------------------------------------------------------------- */
9975+
9976+/* dynop.c */
9977+struct au_branch;
9978+void au_dy_put(struct au_dykey *key);
9979+int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
9980+ struct inode *h_inode);
b752ccd1 9981+int au_dy_irefresh(struct inode *inode);
4a4d8108 9982+void au_dy_arefresh(int do_dio);
4a4d8108
AM
9983+
9984+void __init au_dy_init(void);
9985+void au_dy_fin(void);
9986+
4a4d8108
AM
9987+#endif /* __KERNEL__ */
9988+#endif /* __AUFS_DYNOP_H__ */
7f207e10
AM
9989diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
9990--- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 9991+++ linux/fs/aufs/export.c 2014-06-24 00:14:06.782719306 +0200
523b37e3 9992@@ -0,0 +1,831 @@
4a4d8108 9993+/*
523b37e3 9994+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
9995+ *
9996+ * This program, aufs is free software; you can redistribute it and/or modify
9997+ * it under the terms of the GNU General Public License as published by
9998+ * the Free Software Foundation; either version 2 of the License, or
9999+ * (at your option) any later version.
10000+ *
10001+ * This program is distributed in the hope that it will be useful,
10002+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10003+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10004+ * GNU General Public License for more details.
10005+ *
10006+ * You should have received a copy of the GNU General Public License
523b37e3 10007+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
10008+ */
10009+
10010+/*
10011+ * export via nfs
10012+ */
10013+
10014+#include <linux/exportfs.h>
7eafdf33 10015+#include <linux/fs_struct.h>
4a4d8108
AM
10016+#include <linux/namei.h>
10017+#include <linux/nsproxy.h>
10018+#include <linux/random.h>
10019+#include <linux/writeback.h>
7eafdf33 10020+#include "../fs/mount.h"
4a4d8108
AM
10021+#include "aufs.h"
10022+
10023+union conv {
10024+#ifdef CONFIG_AUFS_INO_T_64
10025+ __u32 a[2];
10026+#else
10027+ __u32 a[1];
10028+#endif
10029+ ino_t ino;
10030+};
10031+
10032+static ino_t decode_ino(__u32 *a)
10033+{
10034+ union conv u;
10035+
10036+ BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
10037+ u.a[0] = a[0];
10038+#ifdef CONFIG_AUFS_INO_T_64
10039+ u.a[1] = a[1];
10040+#endif
10041+ return u.ino;
10042+}
10043+
10044+static void encode_ino(__u32 *a, ino_t ino)
10045+{
10046+ union conv u;
10047+
10048+ u.ino = ino;
10049+ a[0] = u.a[0];
10050+#ifdef CONFIG_AUFS_INO_T_64
10051+ a[1] = u.a[1];
10052+#endif
10053+}
10054+
10055+/* NFS file handle */
10056+enum {
10057+ Fh_br_id,
10058+ Fh_sigen,
10059+#ifdef CONFIG_AUFS_INO_T_64
10060+ /* support 64bit inode number */
10061+ Fh_ino1,
10062+ Fh_ino2,
10063+ Fh_dir_ino1,
10064+ Fh_dir_ino2,
10065+#else
10066+ Fh_ino1,
10067+ Fh_dir_ino1,
10068+#endif
10069+ Fh_igen,
10070+ Fh_h_type,
10071+ Fh_tail,
10072+
10073+ Fh_ino = Fh_ino1,
10074+ Fh_dir_ino = Fh_dir_ino1
10075+};
10076+
10077+static int au_test_anon(struct dentry *dentry)
10078+{
027c5e7a 10079+ /* note: read d_flags without d_lock */
4a4d8108
AM
10080+ return !!(dentry->d_flags & DCACHE_DISCONNECTED);
10081+}
10082+
a2a7ad62
AM
10083+int au_test_nfsd(void)
10084+{
10085+ int ret;
10086+ struct task_struct *tsk = current;
10087+ char comm[sizeof(tsk->comm)];
10088+
10089+ ret = 0;
10090+ if (tsk->flags & PF_KTHREAD) {
10091+ get_task_comm(comm, tsk);
10092+ ret = !strcmp(comm, "nfsd");
10093+ }
10094+
10095+ return ret;
10096+}
10097+
4a4d8108
AM
10098+/* ---------------------------------------------------------------------- */
10099+/* inode generation external table */
10100+
b752ccd1 10101+void au_xigen_inc(struct inode *inode)
4a4d8108 10102+{
4a4d8108
AM
10103+ loff_t pos;
10104+ ssize_t sz;
10105+ __u32 igen;
10106+ struct super_block *sb;
10107+ struct au_sbinfo *sbinfo;
10108+
4a4d8108 10109+ sb = inode->i_sb;
b752ccd1 10110+ AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
1facf9fc 10111+
b752ccd1 10112+ sbinfo = au_sbi(sb);
1facf9fc 10113+ pos = inode->i_ino;
10114+ pos *= sizeof(igen);
10115+ igen = inode->i_generation + 1;
1facf9fc 10116+ sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xigen, &igen,
10117+ sizeof(igen), &pos);
10118+ if (sz == sizeof(igen))
b752ccd1 10119+ return; /* success */
1facf9fc 10120+
b752ccd1 10121+ if (unlikely(sz >= 0))
1facf9fc 10122+ AuIOErr("xigen error (%zd)\n", sz);
1facf9fc 10123+}
10124+
10125+int au_xigen_new(struct inode *inode)
10126+{
10127+ int err;
10128+ loff_t pos;
10129+ ssize_t sz;
10130+ struct super_block *sb;
10131+ struct au_sbinfo *sbinfo;
10132+ struct file *file;
10133+
10134+ err = 0;
10135+ /* todo: dirty, at mount time */
10136+ if (inode->i_ino == AUFS_ROOT_INO)
10137+ goto out;
10138+ sb = inode->i_sb;
dece6358 10139+ SiMustAnyLock(sb);
1facf9fc 10140+ if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
10141+ goto out;
10142+
10143+ err = -EFBIG;
10144+ pos = inode->i_ino;
10145+ if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
10146+ AuIOErr1("too large i%lld\n", pos);
10147+ goto out;
10148+ }
10149+ pos *= sizeof(inode->i_generation);
10150+
10151+ err = 0;
10152+ sbinfo = au_sbi(sb);
10153+ file = sbinfo->si_xigen;
10154+ BUG_ON(!file);
10155+
c06a8ce3 10156+ if (vfsub_f_size_read(file)
1facf9fc 10157+ < pos + sizeof(inode->i_generation)) {
10158+ inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
10159+ sz = xino_fwrite(sbinfo->si_xwrite, file, &inode->i_generation,
10160+ sizeof(inode->i_generation), &pos);
10161+ } else
10162+ sz = xino_fread(sbinfo->si_xread, file, &inode->i_generation,
10163+ sizeof(inode->i_generation), &pos);
10164+ if (sz == sizeof(inode->i_generation))
10165+ goto out; /* success */
10166+
10167+ err = sz;
10168+ if (unlikely(sz >= 0)) {
10169+ err = -EIO;
10170+ AuIOErr("xigen error (%zd)\n", sz);
10171+ }
10172+
4f0767ce 10173+out:
1facf9fc 10174+ return err;
10175+}
10176+
10177+int au_xigen_set(struct super_block *sb, struct file *base)
10178+{
10179+ int err;
10180+ struct au_sbinfo *sbinfo;
10181+ struct file *file;
10182+
dece6358
AM
10183+ SiMustWriteLock(sb);
10184+
1facf9fc 10185+ sbinfo = au_sbi(sb);
10186+ file = au_xino_create2(base, sbinfo->si_xigen);
10187+ err = PTR_ERR(file);
10188+ if (IS_ERR(file))
10189+ goto out;
10190+ err = 0;
10191+ if (sbinfo->si_xigen)
10192+ fput(sbinfo->si_xigen);
10193+ sbinfo->si_xigen = file;
10194+
4f0767ce 10195+out:
1facf9fc 10196+ return err;
10197+}
10198+
10199+void au_xigen_clr(struct super_block *sb)
10200+{
10201+ struct au_sbinfo *sbinfo;
10202+
dece6358
AM
10203+ SiMustWriteLock(sb);
10204+
1facf9fc 10205+ sbinfo = au_sbi(sb);
10206+ if (sbinfo->si_xigen) {
10207+ fput(sbinfo->si_xigen);
10208+ sbinfo->si_xigen = NULL;
10209+ }
10210+}
10211+
10212+/* ---------------------------------------------------------------------- */
10213+
10214+static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
10215+ ino_t dir_ino)
10216+{
10217+ struct dentry *dentry, *d;
10218+ struct inode *inode;
10219+ unsigned int sigen;
10220+
10221+ dentry = NULL;
10222+ inode = ilookup(sb, ino);
10223+ if (!inode)
10224+ goto out;
10225+
10226+ dentry = ERR_PTR(-ESTALE);
10227+ sigen = au_sigen(sb);
10228+ if (unlikely(is_bad_inode(inode)
10229+ || IS_DEADDIR(inode)
537831f9 10230+ || sigen != au_iigen(inode, NULL)))
1facf9fc 10231+ goto out_iput;
10232+
10233+ dentry = NULL;
10234+ if (!dir_ino || S_ISDIR(inode->i_mode))
10235+ dentry = d_find_alias(inode);
10236+ else {
027c5e7a 10237+ spin_lock(&inode->i_lock);
c06a8ce3 10238+ hlist_for_each_entry(d, &inode->i_dentry, d_alias) {
027c5e7a 10239+ spin_lock(&d->d_lock);
1facf9fc 10240+ if (!au_test_anon(d)
10241+ && d->d_parent->d_inode->i_ino == dir_ino) {
027c5e7a
AM
10242+ dentry = dget_dlock(d);
10243+ spin_unlock(&d->d_lock);
1facf9fc 10244+ break;
10245+ }
027c5e7a
AM
10246+ spin_unlock(&d->d_lock);
10247+ }
10248+ spin_unlock(&inode->i_lock);
1facf9fc 10249+ }
027c5e7a 10250+ if (unlikely(dentry && au_digen_test(dentry, sigen))) {
2cbb1c4b 10251+ /* need to refresh */
1facf9fc 10252+ dput(dentry);
2cbb1c4b 10253+ dentry = NULL;
1facf9fc 10254+ }
10255+
4f0767ce 10256+out_iput:
1facf9fc 10257+ iput(inode);
4f0767ce 10258+out:
2cbb1c4b 10259+ AuTraceErrPtr(dentry);
1facf9fc 10260+ return dentry;
10261+}
10262+
10263+/* ---------------------------------------------------------------------- */
10264+
10265+/* todo: dirty? */
10266+/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
4a4d8108
AM
10267+
10268+struct au_compare_mnt_args {
10269+ /* input */
10270+ struct super_block *sb;
10271+
10272+ /* output */
10273+ struct vfsmount *mnt;
10274+};
10275+
10276+static int au_compare_mnt(struct vfsmount *mnt, void *arg)
10277+{
10278+ struct au_compare_mnt_args *a = arg;
10279+
10280+ if (mnt->mnt_sb != a->sb)
10281+ return 0;
10282+ a->mnt = mntget(mnt);
10283+ return 1;
10284+}
10285+
1facf9fc 10286+static struct vfsmount *au_mnt_get(struct super_block *sb)
10287+{
4a4d8108 10288+ int err;
7eafdf33 10289+ struct path root;
4a4d8108
AM
10290+ struct au_compare_mnt_args args = {
10291+ .sb = sb
10292+ };
1facf9fc 10293+
7eafdf33 10294+ get_fs_root(current->fs, &root);
523b37e3 10295+ rcu_read_lock();
7eafdf33 10296+ err = iterate_mounts(au_compare_mnt, &args, root.mnt);
523b37e3 10297+ rcu_read_unlock();
7eafdf33 10298+ path_put(&root);
4a4d8108
AM
10299+ AuDebugOn(!err);
10300+ AuDebugOn(!args.mnt);
10301+ return args.mnt;
1facf9fc 10302+}
10303+
10304+struct au_nfsd_si_lock {
4a4d8108 10305+ unsigned int sigen;
027c5e7a 10306+ aufs_bindex_t bindex, br_id;
1facf9fc 10307+ unsigned char force_lock;
10308+};
10309+
027c5e7a
AM
10310+static int si_nfsd_read_lock(struct super_block *sb,
10311+ struct au_nfsd_si_lock *nsi_lock)
1facf9fc 10312+{
027c5e7a 10313+ int err;
1facf9fc 10314+ aufs_bindex_t bindex;
10315+
10316+ si_read_lock(sb, AuLock_FLUSH);
10317+
10318+ /* branch id may be wrapped around */
027c5e7a 10319+ err = 0;
1facf9fc 10320+ bindex = au_br_index(sb, nsi_lock->br_id);
10321+ if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
10322+ goto out; /* success */
10323+
027c5e7a
AM
10324+ err = -ESTALE;
10325+ bindex = -1;
1facf9fc 10326+ if (!nsi_lock->force_lock)
10327+ si_read_unlock(sb);
1facf9fc 10328+
4f0767ce 10329+out:
027c5e7a
AM
10330+ nsi_lock->bindex = bindex;
10331+ return err;
1facf9fc 10332+}
10333+
10334+struct find_name_by_ino {
392086de 10335+ struct dir_context ctx;
1facf9fc 10336+ int called, found;
10337+ ino_t ino;
10338+ char *name;
10339+ int namelen;
10340+};
10341+
10342+static int
392086de
AM
10343+find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
10344+ loff_t offset, u64 ino, unsigned int d_type)
1facf9fc 10345+{
392086de
AM
10346+ struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
10347+ ctx);
1facf9fc 10348+
10349+ a->called++;
10350+ if (a->ino != ino)
10351+ return 0;
10352+
10353+ memcpy(a->name, name, namelen);
10354+ a->namelen = namelen;
10355+ a->found = 1;
10356+ return 1;
10357+}
10358+
10359+static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
10360+ struct au_nfsd_si_lock *nsi_lock)
10361+{
10362+ struct dentry *dentry, *parent;
10363+ struct file *file;
10364+ struct inode *dir;
392086de
AM
10365+ struct find_name_by_ino arg = {
10366+ .ctx = {
10367+ .actor = au_diractor(find_name_by_ino)
10368+ }
10369+ };
1facf9fc 10370+ int err;
10371+
10372+ parent = path->dentry;
10373+ if (nsi_lock)
10374+ si_read_unlock(parent->d_sb);
4a4d8108 10375+ file = vfsub_dentry_open(path, au_dir_roflags);
1facf9fc 10376+ dentry = (void *)file;
10377+ if (IS_ERR(file))
10378+ goto out;
10379+
10380+ dentry = ERR_PTR(-ENOMEM);
537831f9 10381+ arg.name = (void *)__get_free_page(GFP_NOFS);
1facf9fc 10382+ if (unlikely(!arg.name))
10383+ goto out_file;
10384+ arg.ino = ino;
10385+ arg.found = 0;
10386+ do {
10387+ arg.called = 0;
10388+ /* smp_mb(); */
392086de 10389+ err = vfsub_iterate_dir(file, &arg.ctx);
1facf9fc 10390+ } while (!err && !arg.found && arg.called);
10391+ dentry = ERR_PTR(err);
10392+ if (unlikely(err))
10393+ goto out_name;
1716fcea
AM
10394+ /* instead of ENOENT */
10395+ dentry = ERR_PTR(-ESTALE);
1facf9fc 10396+ if (!arg.found)
10397+ goto out_name;
10398+
b4510431 10399+ /* do not call vfsub_lkup_one() */
1facf9fc 10400+ dir = parent->d_inode;
10401+ mutex_lock(&dir->i_mutex);
10402+ dentry = vfsub_lookup_one_len(arg.name, parent, arg.namelen);
10403+ mutex_unlock(&dir->i_mutex);
10404+ AuTraceErrPtr(dentry);
10405+ if (IS_ERR(dentry))
10406+ goto out_name;
10407+ AuDebugOn(au_test_anon(dentry));
10408+ if (unlikely(!dentry->d_inode)) {
10409+ dput(dentry);
10410+ dentry = ERR_PTR(-ENOENT);
10411+ }
10412+
4f0767ce 10413+out_name:
537831f9 10414+ free_page((unsigned long)arg.name);
4f0767ce 10415+out_file:
1facf9fc 10416+ fput(file);
4f0767ce 10417+out:
1facf9fc 10418+ if (unlikely(nsi_lock
10419+ && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
10420+ if (!IS_ERR(dentry)) {
10421+ dput(dentry);
10422+ dentry = ERR_PTR(-ESTALE);
10423+ }
10424+ AuTraceErrPtr(dentry);
10425+ return dentry;
10426+}
10427+
10428+static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
10429+ ino_t dir_ino,
10430+ struct au_nfsd_si_lock *nsi_lock)
10431+{
10432+ struct dentry *dentry;
10433+ struct path path;
10434+
10435+ if (dir_ino != AUFS_ROOT_INO) {
10436+ path.dentry = decode_by_ino(sb, dir_ino, 0);
10437+ dentry = path.dentry;
10438+ if (!path.dentry || IS_ERR(path.dentry))
10439+ goto out;
10440+ AuDebugOn(au_test_anon(path.dentry));
10441+ } else
10442+ path.dentry = dget(sb->s_root);
10443+
10444+ path.mnt = au_mnt_get(sb);
10445+ dentry = au_lkup_by_ino(&path, ino, nsi_lock);
10446+ path_put(&path);
10447+
4f0767ce 10448+out:
1facf9fc 10449+ AuTraceErrPtr(dentry);
10450+ return dentry;
10451+}
10452+
10453+/* ---------------------------------------------------------------------- */
10454+
10455+static int h_acceptable(void *expv, struct dentry *dentry)
10456+{
10457+ return 1;
10458+}
10459+
10460+static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
10461+ char *buf, int len, struct super_block *sb)
10462+{
10463+ char *p;
10464+ int n;
10465+ struct path path;
10466+
10467+ p = d_path(h_rootpath, buf, len);
10468+ if (IS_ERR(p))
10469+ goto out;
10470+ n = strlen(p);
10471+
10472+ path.mnt = h_rootpath->mnt;
10473+ path.dentry = h_parent;
10474+ p = d_path(&path, buf, len);
10475+ if (IS_ERR(p))
10476+ goto out;
10477+ if (n != 1)
10478+ p += n;
10479+
10480+ path.mnt = au_mnt_get(sb);
10481+ path.dentry = sb->s_root;
10482+ p = d_path(&path, buf, len - strlen(p));
10483+ mntput(path.mnt);
10484+ if (IS_ERR(p))
10485+ goto out;
10486+ if (n != 1)
10487+ p[strlen(p)] = '/';
10488+
4f0767ce 10489+out:
1facf9fc 10490+ AuTraceErrPtr(p);
10491+ return p;
10492+}
10493+
10494+static
027c5e7a
AM
10495+struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
10496+ int fh_len, struct au_nfsd_si_lock *nsi_lock)
1facf9fc 10497+{
10498+ struct dentry *dentry, *h_parent, *root;
10499+ struct super_block *h_sb;
10500+ char *pathname, *p;
10501+ struct vfsmount *h_mnt;
10502+ struct au_branch *br;
10503+ int err;
10504+ struct path path;
10505+
027c5e7a 10506+ br = au_sbr(sb, nsi_lock->bindex);
86dc4139 10507+ h_mnt = au_br_mnt(br);
1facf9fc 10508+ h_sb = h_mnt->mnt_sb;
10509+ /* todo: call lower fh_to_dentry()? fh_to_parent()? */
10510+ h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
10511+ fh_len - Fh_tail, fh[Fh_h_type],
10512+ h_acceptable, /*context*/NULL);
10513+ dentry = h_parent;
10514+ if (unlikely(!h_parent || IS_ERR(h_parent))) {
10515+ AuWarn1("%s decode_fh failed, %ld\n",
10516+ au_sbtype(h_sb), PTR_ERR(h_parent));
10517+ goto out;
10518+ }
10519+ dentry = NULL;
10520+ if (unlikely(au_test_anon(h_parent))) {
10521+ AuWarn1("%s decode_fh returned a disconnected dentry\n",
10522+ au_sbtype(h_sb));
10523+ goto out_h_parent;
10524+ }
10525+
10526+ dentry = ERR_PTR(-ENOMEM);
10527+ pathname = (void *)__get_free_page(GFP_NOFS);
10528+ if (unlikely(!pathname))
10529+ goto out_h_parent;
10530+
10531+ root = sb->s_root;
10532+ path.mnt = h_mnt;
10533+ di_read_lock_parent(root, !AuLock_IR);
027c5e7a 10534+ path.dentry = au_h_dptr(root, nsi_lock->bindex);
1facf9fc 10535+ di_read_unlock(root, !AuLock_IR);
10536+ p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
10537+ dentry = (void *)p;
10538+ if (IS_ERR(p))
10539+ goto out_pathname;
10540+
10541+ si_read_unlock(sb);
10542+ err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
10543+ dentry = ERR_PTR(err);
10544+ if (unlikely(err))
10545+ goto out_relock;
10546+
10547+ dentry = ERR_PTR(-ENOENT);
10548+ AuDebugOn(au_test_anon(path.dentry));
10549+ if (unlikely(!path.dentry->d_inode))
10550+ goto out_path;
10551+
10552+ if (ino != path.dentry->d_inode->i_ino)
10553+ dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
10554+ else
10555+ dentry = dget(path.dentry);
10556+
4f0767ce 10557+out_path:
1facf9fc 10558+ path_put(&path);
4f0767ce 10559+out_relock:
1facf9fc 10560+ if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
10561+ if (!IS_ERR(dentry)) {
10562+ dput(dentry);
10563+ dentry = ERR_PTR(-ESTALE);
10564+ }
4f0767ce 10565+out_pathname:
1facf9fc 10566+ free_page((unsigned long)pathname);
4f0767ce 10567+out_h_parent:
1facf9fc 10568+ dput(h_parent);
4f0767ce 10569+out:
1facf9fc 10570+ AuTraceErrPtr(dentry);
10571+ return dentry;
10572+}
10573+
10574+/* ---------------------------------------------------------------------- */
10575+
10576+static struct dentry *
10577+aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
10578+ int fh_type)
10579+{
10580+ struct dentry *dentry;
10581+ __u32 *fh = fid->raw;
027c5e7a 10582+ struct au_branch *br;
1facf9fc 10583+ ino_t ino, dir_ino;
1facf9fc 10584+ struct au_nfsd_si_lock nsi_lock = {
1facf9fc 10585+ .force_lock = 0
10586+ };
10587+
1facf9fc 10588+ dentry = ERR_PTR(-ESTALE);
4a4d8108
AM
10589+ /* it should never happen, but the file handle is unreliable */
10590+ if (unlikely(fh_len < Fh_tail))
10591+ goto out;
10592+ nsi_lock.sigen = fh[Fh_sigen];
10593+ nsi_lock.br_id = fh[Fh_br_id];
10594+
1facf9fc 10595+ /* branch id may be wrapped around */
027c5e7a
AM
10596+ br = NULL;
10597+ if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
1facf9fc 10598+ goto out;
10599+ nsi_lock.force_lock = 1;
10600+
10601+ /* is this inode still cached? */
10602+ ino = decode_ino(fh + Fh_ino);
4a4d8108
AM
10603+ /* it should never happen */
10604+ if (unlikely(ino == AUFS_ROOT_INO))
10605+ goto out;
10606+
1facf9fc 10607+ dir_ino = decode_ino(fh + Fh_dir_ino);
10608+ dentry = decode_by_ino(sb, ino, dir_ino);
10609+ if (IS_ERR(dentry))
10610+ goto out_unlock;
10611+ if (dentry)
10612+ goto accept;
10613+
10614+ /* is the parent dir cached? */
027c5e7a
AM
10615+ br = au_sbr(sb, nsi_lock.bindex);
10616+ atomic_inc(&br->br_count);
1facf9fc 10617+ dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
10618+ if (IS_ERR(dentry))
10619+ goto out_unlock;
10620+ if (dentry)
10621+ goto accept;
10622+
10623+ /* lookup path */
027c5e7a 10624+ dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
1facf9fc 10625+ if (IS_ERR(dentry))
10626+ goto out_unlock;
10627+ if (unlikely(!dentry))
10628+ /* todo?: make it ESTALE */
10629+ goto out_unlock;
10630+
4f0767ce 10631+accept:
027c5e7a
AM
10632+ if (!au_digen_test(dentry, au_sigen(sb))
10633+ && dentry->d_inode->i_generation == fh[Fh_igen])
1facf9fc 10634+ goto out_unlock; /* success */
10635+
10636+ dput(dentry);
10637+ dentry = ERR_PTR(-ESTALE);
4f0767ce 10638+out_unlock:
027c5e7a
AM
10639+ if (br)
10640+ atomic_dec(&br->br_count);
1facf9fc 10641+ si_read_unlock(sb);
4f0767ce 10642+out:
1facf9fc 10643+ AuTraceErrPtr(dentry);
10644+ return dentry;
10645+}
10646+
10647+#if 0 /* reserved for future use */
10648+/* support subtreecheck option */
10649+static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
10650+ int fh_len, int fh_type)
10651+{
10652+ struct dentry *parent;
10653+ __u32 *fh = fid->raw;
10654+ ino_t dir_ino;
10655+
10656+ dir_ino = decode_ino(fh + Fh_dir_ino);
10657+ parent = decode_by_ino(sb, dir_ino, 0);
10658+ if (IS_ERR(parent))
10659+ goto out;
10660+ if (!parent)
10661+ parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
10662+ dir_ino, fh, fh_len);
10663+
4f0767ce 10664+out:
1facf9fc 10665+ AuTraceErrPtr(parent);
10666+ return parent;
10667+}
10668+#endif
10669+
10670+/* ---------------------------------------------------------------------- */
10671+
0c3ec466
AM
10672+static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
10673+ struct inode *dir)
1facf9fc 10674+{
10675+ int err;
0c3ec466 10676+ aufs_bindex_t bindex;
1facf9fc 10677+ struct super_block *sb, *h_sb;
0c3ec466
AM
10678+ struct dentry *dentry, *parent, *h_parent;
10679+ struct inode *h_dir;
1facf9fc 10680+ struct au_branch *br;
10681+
1facf9fc 10682+ err = -ENOSPC;
10683+ if (unlikely(*max_len <= Fh_tail)) {
10684+ AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
10685+ goto out;
10686+ }
10687+
10688+ err = FILEID_ROOT;
0c3ec466
AM
10689+ if (inode->i_ino == AUFS_ROOT_INO) {
10690+ AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
1facf9fc 10691+ goto out;
10692+ }
10693+
1facf9fc 10694+ h_parent = NULL;
0c3ec466
AM
10695+ sb = inode->i_sb;
10696+ err = si_read_lock(sb, AuLock_FLUSH);
027c5e7a
AM
10697+ if (unlikely(err))
10698+ goto out;
10699+
1facf9fc 10700+#ifdef CONFIG_AUFS_DEBUG
10701+ if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
10702+ AuWarn1("NFS-exporting requires xino\n");
10703+#endif
027c5e7a 10704+ err = -EIO;
0c3ec466
AM
10705+ parent = NULL;
10706+ ii_read_lock_child(inode);
10707+ bindex = au_ibstart(inode);
10708+ if (!dir) {
10709+ dentry = d_find_alias(inode);
10710+ if (unlikely(!dentry))
10711+ goto out_unlock;
10712+ AuDebugOn(au_test_anon(dentry));
10713+ parent = dget_parent(dentry);
10714+ dput(dentry);
10715+ if (unlikely(!parent))
10716+ goto out_unlock;
10717+ dir = parent->d_inode;
1facf9fc 10718+ }
0c3ec466
AM
10719+
10720+ ii_read_lock_parent(dir);
10721+ h_dir = au_h_iptr(dir, bindex);
10722+ ii_read_unlock(dir);
10723+ if (unlikely(!h_dir))
10724+ goto out_parent;
10725+ h_parent = d_find_alias(h_dir);
1facf9fc 10726+ if (unlikely(!h_parent))
0c3ec466 10727+ goto out_hparent;
1facf9fc 10728+
10729+ err = -EPERM;
10730+ br = au_sbr(sb, bindex);
86dc4139 10731+ h_sb = au_br_sb(br);
1facf9fc 10732+ if (unlikely(!h_sb->s_export_op)) {
10733+ AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
0c3ec466 10734+ goto out_hparent;
1facf9fc 10735+ }
10736+
10737+ fh[Fh_br_id] = br->br_id;
10738+ fh[Fh_sigen] = au_sigen(sb);
10739+ encode_ino(fh + Fh_ino, inode->i_ino);
0c3ec466 10740+ encode_ino(fh + Fh_dir_ino, dir->i_ino);
1facf9fc 10741+ fh[Fh_igen] = inode->i_generation;
10742+
10743+ *max_len -= Fh_tail;
10744+ fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
10745+ max_len,
10746+ /*connectable or subtreecheck*/0);
10747+ err = fh[Fh_h_type];
10748+ *max_len += Fh_tail;
10749+ /* todo: macros? */
1716fcea 10750+ if (err != FILEID_INVALID)
1facf9fc 10751+ err = 99;
10752+ else
10753+ AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
10754+
0c3ec466 10755+out_hparent:
1facf9fc 10756+ dput(h_parent);
0c3ec466 10757+out_parent:
1facf9fc 10758+ dput(parent);
0c3ec466
AM
10759+out_unlock:
10760+ ii_read_unlock(inode);
10761+ si_read_unlock(sb);
4f0767ce 10762+out:
1facf9fc 10763+ if (unlikely(err < 0))
1716fcea 10764+ err = FILEID_INVALID;
1facf9fc 10765+ return err;
10766+}
10767+
10768+/* ---------------------------------------------------------------------- */
10769+
4a4d8108
AM
10770+static int aufs_commit_metadata(struct inode *inode)
10771+{
10772+ int err;
10773+ aufs_bindex_t bindex;
10774+ struct super_block *sb;
10775+ struct inode *h_inode;
10776+ int (*f)(struct inode *inode);
10777+
10778+ sb = inode->i_sb;
e49829fe 10779+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108
AM
10780+ ii_write_lock_child(inode);
10781+ bindex = au_ibstart(inode);
10782+ AuDebugOn(bindex < 0);
10783+ h_inode = au_h_iptr(inode, bindex);
10784+
10785+ f = h_inode->i_sb->s_export_op->commit_metadata;
10786+ if (f)
10787+ err = f(h_inode);
10788+ else {
10789+ struct writeback_control wbc = {
10790+ .sync_mode = WB_SYNC_ALL,
10791+ .nr_to_write = 0 /* metadata only */
10792+ };
10793+
10794+ err = sync_inode(h_inode, &wbc);
10795+ }
10796+
10797+ au_cpup_attr_timesizes(inode);
10798+ ii_write_unlock(inode);
10799+ si_read_unlock(sb);
10800+ return err;
10801+}
10802+
10803+/* ---------------------------------------------------------------------- */
10804+
1facf9fc 10805+static struct export_operations aufs_export_op = {
4a4d8108 10806+ .fh_to_dentry = aufs_fh_to_dentry,
1facf9fc 10807+ /* .fh_to_parent = aufs_fh_to_parent, */
4a4d8108
AM
10808+ .encode_fh = aufs_encode_fh,
10809+ .commit_metadata = aufs_commit_metadata
1facf9fc 10810+};
10811+
10812+void au_export_init(struct super_block *sb)
10813+{
10814+ struct au_sbinfo *sbinfo;
10815+ __u32 u;
10816+
10817+ sb->s_export_op = &aufs_export_op;
10818+ sbinfo = au_sbi(sb);
10819+ sbinfo->si_xigen = NULL;
10820+ get_random_bytes(&u, sizeof(u));
10821+ BUILD_BUG_ON(sizeof(u) != sizeof(int));
10822+ atomic_set(&sbinfo->si_xigen_next, u);
10823+}
7f207e10
AM
10824diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
10825--- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
10826+++ linux/fs/aufs/file.c 2014-06-24 00:14:06.782719306 +0200
10827@@ -0,0 +1,711 @@
1facf9fc 10828+/*
523b37e3 10829+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 10830+ *
10831+ * This program, aufs is free software; you can redistribute it and/or modify
10832+ * it under the terms of the GNU General Public License as published by
10833+ * the Free Software Foundation; either version 2 of the License, or
10834+ * (at your option) any later version.
dece6358
AM
10835+ *
10836+ * This program is distributed in the hope that it will be useful,
10837+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10838+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10839+ * GNU General Public License for more details.
10840+ *
10841+ * You should have received a copy of the GNU General Public License
523b37e3 10842+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 10843+ */
10844+
10845+/*
4a4d8108 10846+ * handling file/dir, and address_space operation
1facf9fc 10847+ */
10848+
7eafdf33
AM
10849+#ifdef CONFIG_AUFS_DEBUG
10850+#include <linux/migrate.h>
10851+#endif
4a4d8108 10852+#include <linux/pagemap.h>
1facf9fc 10853+#include "aufs.h"
10854+
4a4d8108
AM
10855+/* drop flags for writing */
10856+unsigned int au_file_roflags(unsigned int flags)
10857+{
10858+ flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
10859+ flags |= O_RDONLY | O_NOATIME;
10860+ return flags;
10861+}
10862+
10863+/* common functions to regular file and dir */
10864+struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
392086de 10865+ struct file *file, int force_wr)
1facf9fc 10866+{
1308ab2a 10867+ struct file *h_file;
4a4d8108
AM
10868+ struct dentry *h_dentry;
10869+ struct inode *h_inode;
10870+ struct super_block *sb;
10871+ struct au_branch *br;
10872+ struct path h_path;
10873+ int err, exec_flag;
1facf9fc 10874+
4a4d8108
AM
10875+ /* a race condition can happen between open and unlink/rmdir */
10876+ h_file = ERR_PTR(-ENOENT);
10877+ h_dentry = au_h_dptr(dentry, bindex);
b752ccd1 10878+ if (au_test_nfsd() && !h_dentry)
4a4d8108
AM
10879+ goto out;
10880+ h_inode = h_dentry->d_inode;
b752ccd1 10881+ if (au_test_nfsd() && !h_inode)
4a4d8108 10882+ goto out;
027c5e7a
AM
10883+ spin_lock(&h_dentry->d_lock);
10884+ err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
10885+ || !h_inode
10886+ /* || !dentry->d_inode->i_nlink */
10887+ ;
10888+ spin_unlock(&h_dentry->d_lock);
10889+ if (unlikely(err))
4a4d8108 10890+ goto out;
1facf9fc 10891+
4a4d8108
AM
10892+ sb = dentry->d_sb;
10893+ br = au_sbr(sb, bindex);
10894+ h_file = ERR_PTR(-EACCES);
2cbb1c4b 10895+ exec_flag = flags & __FMODE_EXEC;
86dc4139 10896+ if (exec_flag && (au_br_mnt(br)->mnt_flags & MNT_NOEXEC))
027c5e7a 10897+ goto out;
1facf9fc 10898+
4a4d8108 10899+ /* drop flags for writing */
392086de
AM
10900+ if (au_test_ro(sb, bindex, dentry->d_inode)) {
10901+ if (force_wr && !(flags & O_WRONLY))
10902+ force_wr = 0;
4a4d8108 10903+ flags = au_file_roflags(flags);
392086de
AM
10904+ if (force_wr) {
10905+ h_file = ERR_PTR(-EROFS);
10906+ flags = au_file_roflags(flags);
10907+ if (unlikely(vfsub_native_ro(h_inode)
10908+ || IS_APPEND(h_inode)))
10909+ goto out;
10910+ flags &= ~O_ACCMODE;
10911+ flags |= O_WRONLY;
10912+ }
10913+ }
4a4d8108
AM
10914+ flags &= ~O_CREAT;
10915+ atomic_inc(&br->br_count);
10916+ h_path.dentry = h_dentry;
86dc4139 10917+ h_path.mnt = au_br_mnt(br);
38d290e6 10918+ h_file = vfsub_dentry_open(&h_path, flags);
4a4d8108
AM
10919+ if (IS_ERR(h_file))
10920+ goto out_br;
dece6358 10921+
4a4d8108
AM
10922+ if (exec_flag) {
10923+ err = deny_write_access(h_file);
10924+ if (unlikely(err)) {
10925+ fput(h_file);
10926+ h_file = ERR_PTR(err);
10927+ goto out_br;
10928+ }
10929+ }
953406b4 10930+ fsnotify_open(h_file);
4a4d8108 10931+ goto out; /* success */
1facf9fc 10932+
4f0767ce 10933+out_br:
4a4d8108 10934+ atomic_dec(&br->br_count);
4f0767ce 10935+out:
4a4d8108
AM
10936+ return h_file;
10937+}
1308ab2a 10938+
4a4d8108
AM
10939+int au_do_open(struct file *file, int (*open)(struct file *file, int flags),
10940+ struct au_fidir *fidir)
1facf9fc 10941+{
dece6358 10942+ int err;
1facf9fc 10943+ struct dentry *dentry;
1308ab2a 10944+
4a4d8108
AM
10945+ err = au_finfo_init(file, fidir);
10946+ if (unlikely(err))
10947+ goto out;
1facf9fc 10948+
10949+ dentry = file->f_dentry;
4a4d8108
AM
10950+ di_read_lock_child(dentry, AuLock_IR);
10951+ err = open(file, vfsub_file_flags(file));
10952+ di_read_unlock(dentry, AuLock_IR);
1facf9fc 10953+
4a4d8108
AM
10954+ fi_write_unlock(file);
10955+ if (unlikely(err)) {
10956+ au_fi(file)->fi_hdir = NULL;
10957+ au_finfo_fin(file);
1308ab2a 10958+ }
4a4d8108 10959+
4f0767ce 10960+out:
1308ab2a 10961+ return err;
10962+}
dece6358 10963+
4a4d8108 10964+int au_reopen_nondir(struct file *file)
1308ab2a 10965+{
4a4d8108
AM
10966+ int err;
10967+ aufs_bindex_t bstart;
10968+ struct dentry *dentry;
10969+ struct file *h_file, *h_file_tmp;
1308ab2a 10970+
4a4d8108 10971+ dentry = file->f_dentry;
4a4d8108
AM
10972+ bstart = au_dbstart(dentry);
10973+ h_file_tmp = NULL;
10974+ if (au_fbstart(file) == bstart) {
10975+ h_file = au_hf_top(file);
10976+ if (file->f_mode == h_file->f_mode)
10977+ return 0; /* success */
10978+ h_file_tmp = h_file;
10979+ get_file(h_file_tmp);
10980+ au_set_h_fptr(file, bstart, NULL);
10981+ }
10982+ AuDebugOn(au_fi(file)->fi_hdir);
86dc4139
AM
10983+ /*
10984+ * it can happen
10985+ * file exists on both of rw and ro
10986+ * open --> dbstart and fbstart are both 0
10987+ * prepend a branch as rw, "rw" become ro
10988+ * remove rw/file
10989+ * delete the top branch, "rw" becomes rw again
10990+ * --> dbstart is 1, fbstart is still 0
10991+ * write --> fbstart is 0 but dbstart is 1
10992+ */
10993+ /* AuDebugOn(au_fbstart(file) < bstart); */
1308ab2a 10994+
4a4d8108 10995+ h_file = au_h_open(dentry, bstart, vfsub_file_flags(file) & ~O_TRUNC,
392086de 10996+ file, /*force_wr*/0);
4a4d8108 10997+ err = PTR_ERR(h_file);
86dc4139
AM
10998+ if (IS_ERR(h_file)) {
10999+ if (h_file_tmp) {
11000+ atomic_inc(&au_sbr(dentry->d_sb, bstart)->br_count);
11001+ au_set_h_fptr(file, bstart, h_file_tmp);
11002+ h_file_tmp = NULL;
11003+ }
4a4d8108 11004+ goto out; /* todo: close all? */
86dc4139 11005+ }
4a4d8108
AM
11006+
11007+ err = 0;
11008+ au_set_fbstart(file, bstart);
11009+ au_set_h_fptr(file, bstart, h_file);
11010+ au_update_figen(file);
11011+ /* todo: necessary? */
11012+ /* file->f_ra = h_file->f_ra; */
11013+
4f0767ce 11014+out:
4a4d8108
AM
11015+ if (h_file_tmp)
11016+ fput(h_file_tmp);
11017+ return err;
1facf9fc 11018+}
11019+
1308ab2a 11020+/* ---------------------------------------------------------------------- */
11021+
4a4d8108
AM
11022+static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
11023+ struct dentry *hi_wh)
1facf9fc 11024+{
4a4d8108
AM
11025+ int err;
11026+ aufs_bindex_t bstart;
11027+ struct au_dinfo *dinfo;
11028+ struct dentry *h_dentry;
11029+ struct au_hdentry *hdp;
1facf9fc 11030+
4a4d8108
AM
11031+ dinfo = au_di(file->f_dentry);
11032+ AuRwMustWriteLock(&dinfo->di_rwsem);
dece6358 11033+
4a4d8108
AM
11034+ bstart = dinfo->di_bstart;
11035+ dinfo->di_bstart = btgt;
11036+ hdp = dinfo->di_hdentry;
11037+ h_dentry = hdp[0 + btgt].hd_dentry;
11038+ hdp[0 + btgt].hd_dentry = hi_wh;
11039+ err = au_reopen_nondir(file);
11040+ hdp[0 + btgt].hd_dentry = h_dentry;
11041+ dinfo->di_bstart = bstart;
1facf9fc 11042+
1facf9fc 11043+ return err;
11044+}
11045+
4a4d8108 11046+static int au_ready_to_write_wh(struct file *file, loff_t len,
86dc4139 11047+ aufs_bindex_t bcpup, struct au_pin *pin)
1facf9fc 11048+{
4a4d8108 11049+ int err;
027c5e7a 11050+ struct inode *inode, *h_inode;
c2b27bf2
AM
11051+ struct dentry *h_dentry, *hi_wh;
11052+ struct au_cp_generic cpg = {
11053+ .dentry = file->f_dentry,
11054+ .bdst = bcpup,
11055+ .bsrc = -1,
11056+ .len = len,
11057+ .pin = pin
11058+ };
1facf9fc 11059+
c2b27bf2
AM
11060+ au_update_dbstart(cpg.dentry);
11061+ inode = cpg.dentry->d_inode;
027c5e7a 11062+ h_inode = NULL;
c2b27bf2
AM
11063+ if (au_dbstart(cpg.dentry) <= bcpup
11064+ && au_dbend(cpg.dentry) >= bcpup) {
11065+ h_dentry = au_h_dptr(cpg.dentry, bcpup);
027c5e7a
AM
11066+ if (h_dentry)
11067+ h_inode = h_dentry->d_inode;
11068+ }
4a4d8108 11069+ hi_wh = au_hi_wh(inode, bcpup);
027c5e7a 11070+ if (!hi_wh && !h_inode)
c2b27bf2 11071+ err = au_sio_cpup_wh(&cpg, file);
4a4d8108
AM
11072+ else
11073+ /* already copied-up after unlink */
11074+ err = au_reopen_wh(file, bcpup, hi_wh);
1facf9fc 11075+
4a4d8108 11076+ if (!err
38d290e6
JR
11077+ && (inode->i_nlink > 1
11078+ || (inode->i_state & I_LINKABLE))
c2b27bf2
AM
11079+ && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
11080+ au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
1308ab2a 11081+
dece6358 11082+ return err;
1facf9fc 11083+}
11084+
4a4d8108
AM
11085+/*
11086+ * prepare the @file for writing.
11087+ */
11088+int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
1facf9fc 11089+{
4a4d8108 11090+ int err;
c2b27bf2
AM
11091+ aufs_bindex_t dbstart;
11092+ struct dentry *parent, *h_dentry;
86dc4139 11093+ struct inode *inode;
1facf9fc 11094+ struct super_block *sb;
4a4d8108 11095+ struct file *h_file;
c2b27bf2
AM
11096+ struct au_cp_generic cpg = {
11097+ .dentry = file->f_dentry,
11098+ .bdst = -1,
11099+ .bsrc = -1,
11100+ .len = len,
11101+ .pin = pin,
11102+ .flags = AuCpup_DTIME
11103+ };
1facf9fc 11104+
c2b27bf2
AM
11105+ sb = cpg.dentry->d_sb;
11106+ inode = cpg.dentry->d_inode;
c2b27bf2
AM
11107+ cpg.bsrc = au_fbstart(file);
11108+ err = au_test_ro(sb, cpg.bsrc, inode);
4a4d8108 11109+ if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
c2b27bf2
AM
11110+ err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
11111+ /*flags*/0);
1facf9fc 11112+ goto out;
4a4d8108 11113+ }
1facf9fc 11114+
027c5e7a 11115+ /* need to cpup or reopen */
c2b27bf2 11116+ parent = dget_parent(cpg.dentry);
4a4d8108 11117+ di_write_lock_parent(parent);
c2b27bf2
AM
11118+ err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
11119+ cpg.bdst = err;
4a4d8108
AM
11120+ if (unlikely(err < 0))
11121+ goto out_dgrade;
11122+ err = 0;
11123+
c2b27bf2
AM
11124+ if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
11125+ err = au_cpup_dirs(cpg.dentry, cpg.bdst);
1facf9fc 11126+ if (unlikely(err))
4a4d8108
AM
11127+ goto out_dgrade;
11128+ }
11129+
c2b27bf2 11130+ err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
4a4d8108
AM
11131+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
11132+ if (unlikely(err))
11133+ goto out_dgrade;
11134+
11135+ h_dentry = au_hf_top(file)->f_dentry;
c2b27bf2
AM
11136+ dbstart = au_dbstart(cpg.dentry);
11137+ if (dbstart <= cpg.bdst) {
11138+ h_dentry = au_h_dptr(cpg.dentry, cpg.bdst);
027c5e7a 11139+ AuDebugOn(!h_dentry);
c2b27bf2 11140+ cpg.bsrc = cpg.bdst;
027c5e7a
AM
11141+ }
11142+
c2b27bf2
AM
11143+ if (dbstart <= cpg.bdst /* just reopen */
11144+ || !d_unhashed(cpg.dentry) /* copyup and reopen */
027c5e7a 11145+ ) {
392086de 11146+ h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
86dc4139 11147+ if (IS_ERR(h_file))
027c5e7a 11148+ err = PTR_ERR(h_file);
86dc4139 11149+ else {
027c5e7a 11150+ di_downgrade_lock(parent, AuLock_IR);
c2b27bf2
AM
11151+ if (dbstart > cpg.bdst)
11152+ err = au_sio_cpup_simple(&cpg);
027c5e7a
AM
11153+ if (!err)
11154+ err = au_reopen_nondir(file);
c2b27bf2 11155+ au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
027c5e7a 11156+ }
027c5e7a
AM
11157+ } else { /* copyup as wh and reopen */
11158+ /*
11159+ * since writable hfsplus branch is not supported,
11160+ * h_open_pre/post() are unnecessary.
11161+ */
c2b27bf2 11162+ err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
4a4d8108 11163+ di_downgrade_lock(parent, AuLock_IR);
4a4d8108 11164+ }
4a4d8108
AM
11165+
11166+ if (!err) {
11167+ au_pin_set_parent_lflag(pin, /*lflag*/0);
11168+ goto out_dput; /* success */
11169+ }
11170+ au_unpin(pin);
11171+ goto out_unlock;
1facf9fc 11172+
4f0767ce 11173+out_dgrade:
4a4d8108 11174+ di_downgrade_lock(parent, AuLock_IR);
4f0767ce 11175+out_unlock:
4a4d8108 11176+ di_read_unlock(parent, AuLock_IR);
4f0767ce 11177+out_dput:
4a4d8108 11178+ dput(parent);
4f0767ce 11179+out:
1facf9fc 11180+ return err;
11181+}
11182+
4a4d8108
AM
11183+/* ---------------------------------------------------------------------- */
11184+
11185+int au_do_flush(struct file *file, fl_owner_t id,
11186+ int (*flush)(struct file *file, fl_owner_t id))
1facf9fc 11187+{
4a4d8108 11188+ int err;
1facf9fc 11189+ struct super_block *sb;
4a4d8108 11190+ struct inode *inode;
1facf9fc 11191+
c06a8ce3
AM
11192+ inode = file_inode(file);
11193+ sb = inode->i_sb;
4a4d8108
AM
11194+ si_noflush_read_lock(sb);
11195+ fi_read_lock(file);
b752ccd1 11196+ ii_read_lock_child(inode);
1facf9fc 11197+
4a4d8108
AM
11198+ err = flush(file, id);
11199+ au_cpup_attr_timesizes(inode);
1facf9fc 11200+
b752ccd1 11201+ ii_read_unlock(inode);
4a4d8108 11202+ fi_read_unlock(file);
1308ab2a 11203+ si_read_unlock(sb);
dece6358 11204+ return err;
1facf9fc 11205+}
11206+
4a4d8108
AM
11207+/* ---------------------------------------------------------------------- */
11208+
11209+static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
1facf9fc 11210+{
4a4d8108 11211+ int err;
4a4d8108
AM
11212+ struct au_pin pin;
11213+ struct au_finfo *finfo;
c2b27bf2 11214+ struct dentry *parent, *hi_wh;
4a4d8108 11215+ struct inode *inode;
1facf9fc 11216+ struct super_block *sb;
c2b27bf2
AM
11217+ struct au_cp_generic cpg = {
11218+ .dentry = file->f_dentry,
11219+ .bdst = -1,
11220+ .bsrc = -1,
11221+ .len = -1,
11222+ .pin = &pin,
11223+ .flags = AuCpup_DTIME
11224+ };
1facf9fc 11225+
4a4d8108
AM
11226+ FiMustWriteLock(file);
11227+
11228+ err = 0;
11229+ finfo = au_fi(file);
c2b27bf2
AM
11230+ sb = cpg.dentry->d_sb;
11231+ inode = cpg.dentry->d_inode;
11232+ cpg.bdst = au_ibstart(inode);
11233+ if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
1308ab2a 11234+ goto out;
dece6358 11235+
c2b27bf2
AM
11236+ parent = dget_parent(cpg.dentry);
11237+ if (au_test_ro(sb, cpg.bdst, inode)) {
4a4d8108 11238+ di_read_lock_parent(parent, !AuLock_IR);
c2b27bf2
AM
11239+ err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
11240+ cpg.bdst = err;
4a4d8108
AM
11241+ di_read_unlock(parent, !AuLock_IR);
11242+ if (unlikely(err < 0))
11243+ goto out_parent;
11244+ err = 0;
1facf9fc 11245+ }
1facf9fc 11246+
4a4d8108 11247+ di_read_lock_parent(parent, AuLock_IR);
c2b27bf2 11248+ hi_wh = au_hi_wh(inode, cpg.bdst);
7f207e10
AM
11249+ if (!S_ISDIR(inode->i_mode)
11250+ && au_opt_test(au_mntflags(sb), PLINK)
4a4d8108 11251+ && au_plink_test(inode)
c2b27bf2
AM
11252+ && !d_unhashed(cpg.dentry)
11253+ && cpg.bdst < au_dbstart(cpg.dentry)) {
11254+ err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
4a4d8108
AM
11255+ if (unlikely(err))
11256+ goto out_unlock;
11257+
11258+ /* always superio. */
c2b27bf2 11259+ err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
4a4d8108 11260+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
367653fa 11261+ if (!err) {
c2b27bf2 11262+ err = au_sio_cpup_simple(&cpg);
367653fa
AM
11263+ au_unpin(&pin);
11264+ }
4a4d8108
AM
11265+ } else if (hi_wh) {
11266+ /* already copied-up after unlink */
c2b27bf2 11267+ err = au_reopen_wh(file, cpg.bdst, hi_wh);
4a4d8108
AM
11268+ *need_reopen = 0;
11269+ }
1facf9fc 11270+
4f0767ce 11271+out_unlock:
4a4d8108 11272+ di_read_unlock(parent, AuLock_IR);
4f0767ce 11273+out_parent:
4a4d8108 11274+ dput(parent);
4f0767ce 11275+out:
1308ab2a 11276+ return err;
dece6358 11277+}
1facf9fc 11278+
4a4d8108 11279+static void au_do_refresh_dir(struct file *file)
dece6358 11280+{
4a4d8108
AM
11281+ aufs_bindex_t bindex, bend, new_bindex, brid;
11282+ struct au_hfile *p, tmp, *q;
11283+ struct au_finfo *finfo;
1308ab2a 11284+ struct super_block *sb;
4a4d8108 11285+ struct au_fidir *fidir;
1facf9fc 11286+
4a4d8108 11287+ FiMustWriteLock(file);
1facf9fc 11288+
4a4d8108
AM
11289+ sb = file->f_dentry->d_sb;
11290+ finfo = au_fi(file);
11291+ fidir = finfo->fi_hdir;
11292+ AuDebugOn(!fidir);
11293+ p = fidir->fd_hfile + finfo->fi_btop;
11294+ brid = p->hf_br->br_id;
11295+ bend = fidir->fd_bbot;
11296+ for (bindex = finfo->fi_btop; bindex <= bend; bindex++, p++) {
11297+ if (!p->hf_file)
11298+ continue;
1308ab2a 11299+
4a4d8108
AM
11300+ new_bindex = au_br_index(sb, p->hf_br->br_id);
11301+ if (new_bindex == bindex)
11302+ continue;
11303+ if (new_bindex < 0) {
11304+ au_set_h_fptr(file, bindex, NULL);
11305+ continue;
11306+ }
1308ab2a 11307+
4a4d8108
AM
11308+ /* swap two lower inode, and loop again */
11309+ q = fidir->fd_hfile + new_bindex;
11310+ tmp = *q;
11311+ *q = *p;
11312+ *p = tmp;
11313+ if (tmp.hf_file) {
11314+ bindex--;
11315+ p--;
11316+ }
11317+ }
1308ab2a 11318+
4a4d8108 11319+ p = fidir->fd_hfile;
027c5e7a 11320+ if (!au_test_mmapped(file) && !d_unlinked(file->f_dentry)) {
4a4d8108
AM
11321+ bend = au_sbend(sb);
11322+ for (finfo->fi_btop = 0; finfo->fi_btop <= bend;
11323+ finfo->fi_btop++, p++)
11324+ if (p->hf_file) {
c06a8ce3 11325+ if (file_inode(p->hf_file))
4a4d8108
AM
11326+ break;
11327+ else
11328+ au_hfput(p, file);
11329+ }
11330+ } else {
11331+ bend = au_br_index(sb, brid);
11332+ for (finfo->fi_btop = 0; finfo->fi_btop < bend;
11333+ finfo->fi_btop++, p++)
11334+ if (p->hf_file)
11335+ au_hfput(p, file);
11336+ bend = au_sbend(sb);
11337+ }
1308ab2a 11338+
4a4d8108
AM
11339+ p = fidir->fd_hfile + bend;
11340+ for (fidir->fd_bbot = bend; fidir->fd_bbot >= finfo->fi_btop;
11341+ fidir->fd_bbot--, p--)
11342+ if (p->hf_file) {
c06a8ce3 11343+ if (file_inode(p->hf_file))
4a4d8108
AM
11344+ break;
11345+ else
11346+ au_hfput(p, file);
11347+ }
11348+ AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
1308ab2a 11349+}
11350+
4a4d8108
AM
11351+/*
11352+ * after branch manipulating, refresh the file.
11353+ */
11354+static int refresh_file(struct file *file, int (*reopen)(struct file *file))
1facf9fc 11355+{
4a4d8108
AM
11356+ int err, need_reopen;
11357+ aufs_bindex_t bend, bindex;
11358+ struct dentry *dentry;
1308ab2a 11359+ struct au_finfo *finfo;
4a4d8108 11360+ struct au_hfile *hfile;
1facf9fc 11361+
4a4d8108 11362+ dentry = file->f_dentry;
1308ab2a 11363+ finfo = au_fi(file);
4a4d8108
AM
11364+ if (!finfo->fi_hdir) {
11365+ hfile = &finfo->fi_htop;
11366+ AuDebugOn(!hfile->hf_file);
11367+ bindex = au_br_index(dentry->d_sb, hfile->hf_br->br_id);
11368+ AuDebugOn(bindex < 0);
11369+ if (bindex != finfo->fi_btop)
11370+ au_set_fbstart(file, bindex);
11371+ } else {
11372+ err = au_fidir_realloc(finfo, au_sbend(dentry->d_sb) + 1);
11373+ if (unlikely(err))
11374+ goto out;
11375+ au_do_refresh_dir(file);
11376+ }
1facf9fc 11377+
4a4d8108
AM
11378+ err = 0;
11379+ need_reopen = 1;
11380+ if (!au_test_mmapped(file))
11381+ err = au_file_refresh_by_inode(file, &need_reopen);
027c5e7a 11382+ if (!err && need_reopen && !d_unlinked(dentry))
4a4d8108
AM
11383+ err = reopen(file);
11384+ if (!err) {
11385+ au_update_figen(file);
11386+ goto out; /* success */
11387+ }
11388+
11389+ /* error, close all lower files */
11390+ if (finfo->fi_hdir) {
11391+ bend = au_fbend_dir(file);
11392+ for (bindex = au_fbstart(file); bindex <= bend; bindex++)
11393+ au_set_h_fptr(file, bindex, NULL);
11394+ }
1facf9fc 11395+
4f0767ce 11396+out:
1facf9fc 11397+ return err;
11398+}
11399+
4a4d8108
AM
11400+/* common function to regular file and dir */
11401+int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
11402+ int wlock)
dece6358 11403+{
1308ab2a 11404+ int err;
4a4d8108
AM
11405+ unsigned int sigen, figen;
11406+ aufs_bindex_t bstart;
11407+ unsigned char pseudo_link;
11408+ struct dentry *dentry;
11409+ struct inode *inode;
1facf9fc 11410+
4a4d8108
AM
11411+ err = 0;
11412+ dentry = file->f_dentry;
11413+ inode = dentry->d_inode;
4a4d8108
AM
11414+ sigen = au_sigen(dentry->d_sb);
11415+ fi_write_lock(file);
11416+ figen = au_figen(file);
11417+ di_write_lock_child(dentry);
11418+ bstart = au_dbstart(dentry);
11419+ pseudo_link = (bstart != au_ibstart(inode));
11420+ if (sigen == figen && !pseudo_link && au_fbstart(file) == bstart) {
11421+ if (!wlock) {
11422+ di_downgrade_lock(dentry, AuLock_IR);
11423+ fi_downgrade_lock(file);
11424+ }
11425+ goto out; /* success */
11426+ }
dece6358 11427+
4a4d8108 11428+ AuDbg("sigen %d, figen %d\n", sigen, figen);
027c5e7a 11429+ if (au_digen_test(dentry, sigen)) {
4a4d8108 11430+ err = au_reval_dpath(dentry, sigen);
027c5e7a 11431+ AuDebugOn(!err && au_digen_test(dentry, sigen));
4a4d8108 11432+ }
dece6358 11433+
027c5e7a
AM
11434+ if (!err)
11435+ err = refresh_file(file, reopen);
4a4d8108
AM
11436+ if (!err) {
11437+ if (!wlock) {
11438+ di_downgrade_lock(dentry, AuLock_IR);
11439+ fi_downgrade_lock(file);
11440+ }
11441+ } else {
11442+ di_write_unlock(dentry);
11443+ fi_write_unlock(file);
11444+ }
1facf9fc 11445+
4f0767ce 11446+out:
1308ab2a 11447+ return err;
11448+}
1facf9fc 11449+
4a4d8108
AM
11450+/* ---------------------------------------------------------------------- */
11451+
11452+/* cf. aufs_nopage() */
11453+/* for madvise(2) */
11454+static int aufs_readpage(struct file *file __maybe_unused, struct page *page)
1308ab2a 11455+{
4a4d8108
AM
11456+ unlock_page(page);
11457+ return 0;
11458+}
1facf9fc 11459+
4a4d8108
AM
11460+/* it will never be called, but necessary to support O_DIRECT */
11461+static ssize_t aufs_direct_IO(int rw, struct kiocb *iocb,
11462+ const struct iovec *iov, loff_t offset,
11463+ unsigned long nr_segs)
11464+{ BUG(); return 0; }
1facf9fc 11465+
4a4d8108
AM
11466+/*
11467+ * it will never be called, but madvise and fadvise behaves differently
11468+ * when get_xip_mem is defined
11469+ */
11470+static int aufs_get_xip_mem(struct address_space *mapping, pgoff_t pgoff,
11471+ int create, void **kmem, unsigned long *pfn)
11472+{ BUG(); return 0; }
1facf9fc 11473+
4a4d8108
AM
11474+/* they will never be called. */
11475+#ifdef CONFIG_AUFS_DEBUG
11476+static int aufs_write_begin(struct file *file, struct address_space *mapping,
11477+ loff_t pos, unsigned len, unsigned flags,
11478+ struct page **pagep, void **fsdata)
11479+{ AuUnsupport(); return 0; }
11480+static int aufs_write_end(struct file *file, struct address_space *mapping,
11481+ loff_t pos, unsigned len, unsigned copied,
11482+ struct page *page, void *fsdata)
11483+{ AuUnsupport(); return 0; }
11484+static int aufs_writepage(struct page *page, struct writeback_control *wbc)
11485+{ AuUnsupport(); return 0; }
1308ab2a 11486+
4a4d8108
AM
11487+static int aufs_set_page_dirty(struct page *page)
11488+{ AuUnsupport(); return 0; }
392086de
AM
11489+static void aufs_invalidatepage(struct page *page, unsigned int offset,
11490+ unsigned int length)
4a4d8108
AM
11491+{ AuUnsupport(); }
11492+static int aufs_releasepage(struct page *page, gfp_t gfp)
11493+{ AuUnsupport(); return 0; }
11494+static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
7eafdf33 11495+ struct page *page, enum migrate_mode mode)
4a4d8108
AM
11496+{ AuUnsupport(); return 0; }
11497+static int aufs_launder_page(struct page *page)
11498+{ AuUnsupport(); return 0; }
11499+static int aufs_is_partially_uptodate(struct page *page,
38d290e6
JR
11500+ unsigned long from,
11501+ unsigned long count)
4a4d8108 11502+{ AuUnsupport(); return 0; }
392086de
AM
11503+static void aufs_is_dirty_writeback(struct page *page, bool *dirty,
11504+ bool *writeback)
11505+{ AuUnsupport(); }
4a4d8108
AM
11506+static int aufs_error_remove_page(struct address_space *mapping,
11507+ struct page *page)
11508+{ AuUnsupport(); return 0; }
b4510431
AM
11509+static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
11510+ sector_t *span)
11511+{ AuUnsupport(); return 0; }
11512+static void aufs_swap_deactivate(struct file *file)
11513+{ AuUnsupport(); }
4a4d8108
AM
11514+#endif /* CONFIG_AUFS_DEBUG */
11515+
11516+const struct address_space_operations aufs_aop = {
11517+ .readpage = aufs_readpage,
11518+ .direct_IO = aufs_direct_IO,
11519+ .get_xip_mem = aufs_get_xip_mem,
11520+#ifdef CONFIG_AUFS_DEBUG
11521+ .writepage = aufs_writepage,
4a4d8108
AM
11522+ /* no writepages, because of writepage */
11523+ .set_page_dirty = aufs_set_page_dirty,
11524+ /* no readpages, because of readpage */
11525+ .write_begin = aufs_write_begin,
11526+ .write_end = aufs_write_end,
11527+ /* no bmap, no block device */
11528+ .invalidatepage = aufs_invalidatepage,
11529+ .releasepage = aufs_releasepage,
11530+ .migratepage = aufs_migratepage,
11531+ .launder_page = aufs_launder_page,
11532+ .is_partially_uptodate = aufs_is_partially_uptodate,
392086de 11533+ .is_dirty_writeback = aufs_is_dirty_writeback,
b4510431
AM
11534+ .error_remove_page = aufs_error_remove_page,
11535+ .swap_activate = aufs_swap_activate,
11536+ .swap_deactivate = aufs_swap_deactivate
4a4d8108 11537+#endif /* CONFIG_AUFS_DEBUG */
dece6358 11538+};
7f207e10
AM
11539diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
11540--- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
11541+++ linux/fs/aufs/file.h 2014-06-24 00:14:06.782719306 +0200
11542@@ -0,0 +1,289 @@
4a4d8108 11543+/*
523b37e3 11544+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
11545+ *
11546+ * This program, aufs is free software; you can redistribute it and/or modify
11547+ * it under the terms of the GNU General Public License as published by
11548+ * the Free Software Foundation; either version 2 of the License, or
11549+ * (at your option) any later version.
11550+ *
11551+ * This program is distributed in the hope that it will be useful,
11552+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
11553+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11554+ * GNU General Public License for more details.
11555+ *
11556+ * You should have received a copy of the GNU General Public License
523b37e3 11557+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 11558+ */
1facf9fc 11559+
4a4d8108
AM
11560+/*
11561+ * file operations
11562+ */
1facf9fc 11563+
4a4d8108
AM
11564+#ifndef __AUFS_FILE_H__
11565+#define __AUFS_FILE_H__
1facf9fc 11566+
4a4d8108 11567+#ifdef __KERNEL__
1facf9fc 11568+
2cbb1c4b 11569+#include <linux/file.h>
4a4d8108
AM
11570+#include <linux/fs.h>
11571+#include <linux/poll.h>
4a4d8108 11572+#include "rwsem.h"
1facf9fc 11573+
4a4d8108
AM
11574+struct au_branch;
11575+struct au_hfile {
11576+ struct file *hf_file;
11577+ struct au_branch *hf_br;
11578+};
1facf9fc 11579+
4a4d8108
AM
11580+struct au_vdir;
11581+struct au_fidir {
11582+ aufs_bindex_t fd_bbot;
11583+ aufs_bindex_t fd_nent;
11584+ struct au_vdir *fd_vdir_cache;
11585+ struct au_hfile fd_hfile[];
11586+};
1facf9fc 11587+
4a4d8108 11588+static inline int au_fidir_sz(int nent)
dece6358 11589+{
4f0767ce
JR
11590+ AuDebugOn(nent < 0);
11591+ return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
4a4d8108 11592+}
1facf9fc 11593+
4a4d8108
AM
11594+struct au_finfo {
11595+ atomic_t fi_generation;
dece6358 11596+
4a4d8108
AM
11597+ struct au_rwsem fi_rwsem;
11598+ aufs_bindex_t fi_btop;
11599+
11600+ /* do not union them */
11601+ struct { /* for non-dir */
11602+ struct au_hfile fi_htop;
2cbb1c4b 11603+ atomic_t fi_mmapped;
4a4d8108
AM
11604+ };
11605+ struct au_fidir *fi_hdir; /* for dir only */
523b37e3
AM
11606+
11607+ struct hlist_node fi_hlist;
11608+ struct file *fi_file; /* very ugly */
4a4d8108 11609+} ____cacheline_aligned_in_smp;
1facf9fc 11610+
4a4d8108 11611+/* ---------------------------------------------------------------------- */
1facf9fc 11612+
4a4d8108
AM
11613+/* file.c */
11614+extern const struct address_space_operations aufs_aop;
11615+unsigned int au_file_roflags(unsigned int flags);
11616+struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
392086de 11617+ struct file *file, int force_wr);
4a4d8108
AM
11618+int au_do_open(struct file *file, int (*open)(struct file *file, int flags),
11619+ struct au_fidir *fidir);
11620+int au_reopen_nondir(struct file *file);
11621+struct au_pin;
11622+int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
11623+int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
11624+ int wlock);
11625+int au_do_flush(struct file *file, fl_owner_t id,
11626+ int (*flush)(struct file *file, fl_owner_t id));
1facf9fc 11627+
4a4d8108
AM
11628+/* poll.c */
11629+#ifdef CONFIG_AUFS_POLL
11630+unsigned int aufs_poll(struct file *file, poll_table *wait);
11631+#endif
1facf9fc 11632+
4a4d8108
AM
11633+#ifdef CONFIG_AUFS_BR_HFSPLUS
11634+/* hfsplus.c */
392086de
AM
11635+struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
11636+ int force_wr);
4a4d8108
AM
11637+void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
11638+ struct file *h_file);
11639+#else
11640+static inline
392086de
AM
11641+struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
11642+ int force_wr)
dece6358 11643+{
4a4d8108
AM
11644+ return NULL;
11645+}
1facf9fc 11646+
4a4d8108
AM
11647+AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
11648+ struct file *h_file);
11649+#endif
1facf9fc 11650+
4a4d8108
AM
11651+/* f_op.c */
11652+extern const struct file_operations aufs_file_fop;
4a4d8108
AM
11653+int au_do_open_nondir(struct file *file, int flags);
11654+int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
11655+
4a4d8108
AM
11656+/* finfo.c */
11657+void au_hfput(struct au_hfile *hf, struct file *file);
11658+void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
11659+ struct file *h_file);
1facf9fc 11660+
4a4d8108 11661+void au_update_figen(struct file *file);
4a4d8108
AM
11662+struct au_fidir *au_fidir_alloc(struct super_block *sb);
11663+int au_fidir_realloc(struct au_finfo *finfo, int nbr);
1facf9fc 11664+
4a4d8108
AM
11665+void au_fi_init_once(void *_fi);
11666+void au_finfo_fin(struct file *file);
11667+int au_finfo_init(struct file *file, struct au_fidir *fidir);
1facf9fc 11668+
4a4d8108
AM
11669+/* ioctl.c */
11670+long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
b752ccd1
AM
11671+#ifdef CONFIG_COMPAT
11672+long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
11673+ unsigned long arg);
c2b27bf2
AM
11674+long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
11675+ unsigned long arg);
b752ccd1 11676+#endif
1facf9fc 11677+
4a4d8108 11678+/* ---------------------------------------------------------------------- */
1facf9fc 11679+
4a4d8108
AM
11680+static inline struct au_finfo *au_fi(struct file *file)
11681+{
38d290e6 11682+ return file->private_data;
4a4d8108 11683+}
1facf9fc 11684+
4a4d8108 11685+/* ---------------------------------------------------------------------- */
1facf9fc 11686+
4a4d8108
AM
11687+/*
11688+ * fi_read_lock, fi_write_lock,
11689+ * fi_read_unlock, fi_write_unlock, fi_downgrade_lock
11690+ */
11691+AuSimpleRwsemFuncs(fi, struct file *f, &au_fi(f)->fi_rwsem);
1308ab2a 11692+
4a4d8108
AM
11693+#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
11694+#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
11695+#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
1facf9fc 11696+
1308ab2a 11697+/* ---------------------------------------------------------------------- */
11698+
4a4d8108
AM
11699+/* todo: hard/soft set? */
11700+static inline aufs_bindex_t au_fbstart(struct file *file)
dece6358 11701+{
4a4d8108
AM
11702+ FiMustAnyLock(file);
11703+ return au_fi(file)->fi_btop;
11704+}
dece6358 11705+
4a4d8108
AM
11706+static inline aufs_bindex_t au_fbend_dir(struct file *file)
11707+{
11708+ FiMustAnyLock(file);
11709+ AuDebugOn(!au_fi(file)->fi_hdir);
11710+ return au_fi(file)->fi_hdir->fd_bbot;
11711+}
1facf9fc 11712+
4a4d8108
AM
11713+static inline struct au_vdir *au_fvdir_cache(struct file *file)
11714+{
11715+ FiMustAnyLock(file);
11716+ AuDebugOn(!au_fi(file)->fi_hdir);
11717+ return au_fi(file)->fi_hdir->fd_vdir_cache;
11718+}
1facf9fc 11719+
4a4d8108
AM
11720+static inline void au_set_fbstart(struct file *file, aufs_bindex_t bindex)
11721+{
11722+ FiMustWriteLock(file);
11723+ au_fi(file)->fi_btop = bindex;
11724+}
1facf9fc 11725+
4a4d8108
AM
11726+static inline void au_set_fbend_dir(struct file *file, aufs_bindex_t bindex)
11727+{
11728+ FiMustWriteLock(file);
11729+ AuDebugOn(!au_fi(file)->fi_hdir);
11730+ au_fi(file)->fi_hdir->fd_bbot = bindex;
11731+}
1308ab2a 11732+
4a4d8108
AM
11733+static inline void au_set_fvdir_cache(struct file *file,
11734+ struct au_vdir *vdir_cache)
11735+{
11736+ FiMustWriteLock(file);
11737+ AuDebugOn(!au_fi(file)->fi_hdir);
11738+ au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
11739+}
dece6358 11740+
4a4d8108
AM
11741+static inline struct file *au_hf_top(struct file *file)
11742+{
11743+ FiMustAnyLock(file);
11744+ AuDebugOn(au_fi(file)->fi_hdir);
11745+ return au_fi(file)->fi_htop.hf_file;
11746+}
1facf9fc 11747+
4a4d8108
AM
11748+static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
11749+{
11750+ FiMustAnyLock(file);
11751+ AuDebugOn(!au_fi(file)->fi_hdir);
11752+ return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
dece6358
AM
11753+}
11754+
4a4d8108
AM
11755+/* todo: memory barrier? */
11756+static inline unsigned int au_figen(struct file *f)
dece6358 11757+{
4a4d8108
AM
11758+ return atomic_read(&au_fi(f)->fi_generation);
11759+}
dece6358 11760+
2cbb1c4b
JR
11761+static inline void au_set_mmapped(struct file *f)
11762+{
11763+ if (atomic_inc_return(&au_fi(f)->fi_mmapped))
11764+ return;
0c3ec466 11765+ pr_warn("fi_mmapped wrapped around\n");
2cbb1c4b
JR
11766+ while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
11767+ ;
11768+}
11769+
11770+static inline void au_unset_mmapped(struct file *f)
11771+{
11772+ atomic_dec(&au_fi(f)->fi_mmapped);
11773+}
11774+
4a4d8108
AM
11775+static inline int au_test_mmapped(struct file *f)
11776+{
2cbb1c4b
JR
11777+ return atomic_read(&au_fi(f)->fi_mmapped);
11778+}
11779+
11780+/* customize vma->vm_file */
11781+
11782+static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
11783+ struct file *file)
11784+{
53392da6
AM
11785+ struct file *f;
11786+
11787+ f = vma->vm_file;
2cbb1c4b
JR
11788+ get_file(file);
11789+ vma->vm_file = file;
53392da6 11790+ fput(f);
2cbb1c4b
JR
11791+}
11792+
11793+#ifdef CONFIG_MMU
11794+#define AuDbgVmRegion(file, vma) do {} while (0)
11795+
11796+static inline void au_vm_file_reset(struct vm_area_struct *vma,
11797+ struct file *file)
11798+{
11799+ au_do_vm_file_reset(vma, file);
11800+}
11801+#else
11802+#define AuDbgVmRegion(file, vma) \
11803+ AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
11804+
11805+static inline void au_vm_file_reset(struct vm_area_struct *vma,
11806+ struct file *file)
11807+{
53392da6
AM
11808+ struct file *f;
11809+
2cbb1c4b 11810+ au_do_vm_file_reset(vma, file);
53392da6 11811+ f = vma->vm_region->vm_file;
2cbb1c4b
JR
11812+ get_file(file);
11813+ vma->vm_region->vm_file = file;
53392da6 11814+ fput(f);
2cbb1c4b
JR
11815+}
11816+#endif /* CONFIG_MMU */
11817+
11818+/* handle vma->vm_prfile */
fb47a38f 11819+static inline void au_vm_prfile_set(struct vm_area_struct *vma,
2cbb1c4b
JR
11820+ struct file *file)
11821+{
2cbb1c4b
JR
11822+ get_file(file);
11823+ vma->vm_prfile = file;
11824+#ifndef CONFIG_MMU
11825+ get_file(file);
11826+ vma->vm_region->vm_prfile = file;
11827+#endif
fb47a38f 11828+}
1308ab2a 11829+
4a4d8108
AM
11830+#endif /* __KERNEL__ */
11831+#endif /* __AUFS_FILE_H__ */
7f207e10
AM
11832diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
11833--- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 11834+++ linux/fs/aufs/finfo.c 2014-06-24 00:14:06.782719306 +0200
523b37e3 11835@@ -0,0 +1,156 @@
4a4d8108 11836+/*
523b37e3 11837+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
11838+ *
11839+ * This program, aufs is free software; you can redistribute it and/or modify
11840+ * it under the terms of the GNU General Public License as published by
11841+ * the Free Software Foundation; either version 2 of the License, or
11842+ * (at your option) any later version.
11843+ *
11844+ * This program is distributed in the hope that it will be useful,
11845+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
11846+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11847+ * GNU General Public License for more details.
11848+ *
11849+ * You should have received a copy of the GNU General Public License
523b37e3 11850+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 11851+ */
1308ab2a 11852+
4a4d8108
AM
11853+/*
11854+ * file private data
11855+ */
1facf9fc 11856+
4a4d8108 11857+#include "aufs.h"
1facf9fc 11858+
4a4d8108
AM
11859+void au_hfput(struct au_hfile *hf, struct file *file)
11860+{
11861+ /* todo: direct access f_flags */
2cbb1c4b 11862+ if (vfsub_file_flags(file) & __FMODE_EXEC)
4a4d8108
AM
11863+ allow_write_access(hf->hf_file);
11864+ fput(hf->hf_file);
11865+ hf->hf_file = NULL;
e49829fe 11866+ atomic_dec(&hf->hf_br->br_count);
4a4d8108
AM
11867+ hf->hf_br = NULL;
11868+}
1facf9fc 11869+
4a4d8108
AM
11870+void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
11871+{
11872+ struct au_finfo *finfo = au_fi(file);
11873+ struct au_hfile *hf;
11874+ struct au_fidir *fidir;
11875+
11876+ fidir = finfo->fi_hdir;
11877+ if (!fidir) {
11878+ AuDebugOn(finfo->fi_btop != bindex);
11879+ hf = &finfo->fi_htop;
11880+ } else
11881+ hf = fidir->fd_hfile + bindex;
11882+
11883+ if (hf && hf->hf_file)
11884+ au_hfput(hf, file);
11885+ if (val) {
11886+ FiMustWriteLock(file);
11887+ hf->hf_file = val;
11888+ hf->hf_br = au_sbr(file->f_dentry->d_sb, bindex);
1308ab2a 11889+ }
4a4d8108 11890+}
1facf9fc 11891+
4a4d8108
AM
11892+void au_update_figen(struct file *file)
11893+{
11894+ atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_dentry));
11895+ /* smp_mb(); */ /* atomic_set */
1facf9fc 11896+}
11897+
4a4d8108
AM
11898+/* ---------------------------------------------------------------------- */
11899+
4a4d8108
AM
11900+struct au_fidir *au_fidir_alloc(struct super_block *sb)
11901+{
11902+ struct au_fidir *fidir;
11903+ int nbr;
11904+
11905+ nbr = au_sbend(sb) + 1;
11906+ if (nbr < 2)
11907+ nbr = 2; /* initial allocate for 2 branches */
11908+ fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
11909+ if (fidir) {
11910+ fidir->fd_bbot = -1;
11911+ fidir->fd_nent = nbr;
11912+ fidir->fd_vdir_cache = NULL;
11913+ }
11914+
11915+ return fidir;
11916+}
11917+
11918+int au_fidir_realloc(struct au_finfo *finfo, int nbr)
11919+{
11920+ int err;
11921+ struct au_fidir *fidir, *p;
11922+
11923+ AuRwMustWriteLock(&finfo->fi_rwsem);
11924+ fidir = finfo->fi_hdir;
11925+ AuDebugOn(!fidir);
11926+
11927+ err = -ENOMEM;
11928+ p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
11929+ GFP_NOFS);
11930+ if (p) {
11931+ p->fd_nent = nbr;
11932+ finfo->fi_hdir = p;
11933+ err = 0;
11934+ }
1facf9fc 11935+
dece6358 11936+ return err;
1facf9fc 11937+}
1308ab2a 11938+
11939+/* ---------------------------------------------------------------------- */
11940+
4a4d8108 11941+void au_finfo_fin(struct file *file)
1308ab2a 11942+{
4a4d8108
AM
11943+ struct au_finfo *finfo;
11944+
7f207e10
AM
11945+ au_nfiles_dec(file->f_dentry->d_sb);
11946+
4a4d8108
AM
11947+ finfo = au_fi(file);
11948+ AuDebugOn(finfo->fi_hdir);
11949+ AuRwDestroy(&finfo->fi_rwsem);
11950+ au_cache_free_finfo(finfo);
1308ab2a 11951+}
1308ab2a 11952+
e49829fe 11953+void au_fi_init_once(void *_finfo)
4a4d8108 11954+{
e49829fe 11955+ struct au_finfo *finfo = _finfo;
2cbb1c4b 11956+ static struct lock_class_key aufs_fi;
1308ab2a 11957+
e49829fe
JR
11958+ au_rw_init(&finfo->fi_rwsem);
11959+ au_rw_class(&finfo->fi_rwsem, &aufs_fi);
4a4d8108 11960+}
1308ab2a 11961+
4a4d8108
AM
11962+int au_finfo_init(struct file *file, struct au_fidir *fidir)
11963+{
1716fcea 11964+ int err;
4a4d8108
AM
11965+ struct au_finfo *finfo;
11966+ struct dentry *dentry;
11967+
11968+ err = -ENOMEM;
11969+ dentry = file->f_dentry;
11970+ finfo = au_cache_alloc_finfo();
11971+ if (unlikely(!finfo))
11972+ goto out;
11973+
11974+ err = 0;
7f207e10 11975+ au_nfiles_inc(dentry->d_sb);
1716fcea
AM
11976+ /* verbose coding for lock class name */
11977+ if (!fidir)
11978+ au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcNonDir_FIINFO);
11979+ else
11980+ au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcDir_FIINFO);
4a4d8108
AM
11981+ au_rw_write_lock(&finfo->fi_rwsem);
11982+ finfo->fi_btop = -1;
11983+ finfo->fi_hdir = fidir;
11984+ atomic_set(&finfo->fi_generation, au_digen(dentry));
11985+ /* smp_mb(); */ /* atomic_set */
11986+
11987+ file->private_data = finfo;
11988+
11989+out:
11990+ return err;
11991+}
7f207e10
AM
11992diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
11993--- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
11994+++ linux/fs/aufs/f_op.c 2014-06-24 00:14:06.782719306 +0200
11995@@ -0,0 +1,782 @@
dece6358 11996+/*
523b37e3 11997+ * Copyright (C) 2005-2014 Junjiro R. Okajima
dece6358
AM
11998+ *
11999+ * This program, aufs is free software; you can redistribute it and/or modify
12000+ * it under the terms of the GNU General Public License as published by
12001+ * the Free Software Foundation; either version 2 of the License, or
12002+ * (at your option) any later version.
12003+ *
12004+ * This program is distributed in the hope that it will be useful,
12005+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12006+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12007+ * GNU General Public License for more details.
12008+ *
12009+ * You should have received a copy of the GNU General Public License
523b37e3 12010+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
dece6358 12011+ */
1facf9fc 12012+
12013+/*
4a4d8108 12014+ * file and vm operations
1facf9fc 12015+ */
dece6358 12016+
86dc4139 12017+#include <linux/aio.h>
4a4d8108
AM
12018+#include <linux/fs_stack.h>
12019+#include <linux/mman.h>
4a4d8108 12020+#include <linux/security.h>
dece6358
AM
12021+#include "aufs.h"
12022+
4a4d8108 12023+int au_do_open_nondir(struct file *file, int flags)
1facf9fc 12024+{
4a4d8108
AM
12025+ int err;
12026+ aufs_bindex_t bindex;
12027+ struct file *h_file;
12028+ struct dentry *dentry;
12029+ struct au_finfo *finfo;
38d290e6 12030+ struct inode *h_inode;
4a4d8108
AM
12031+
12032+ FiMustWriteLock(file);
12033+
523b37e3 12034+ err = 0;
4a4d8108
AM
12035+ dentry = file->f_dentry;
12036+ finfo = au_fi(file);
12037+ memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
2cbb1c4b 12038+ atomic_set(&finfo->fi_mmapped, 0);
4a4d8108 12039+ bindex = au_dbstart(dentry);
392086de 12040+ h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
4a4d8108
AM
12041+ if (IS_ERR(h_file))
12042+ err = PTR_ERR(h_file);
12043+ else {
38d290e6
JR
12044+ if ((flags & __O_TMPFILE)
12045+ && !(flags & O_EXCL)) {
12046+ h_inode = file_inode(h_file);
12047+ spin_lock(&h_inode->i_lock);
12048+ h_inode->i_state |= I_LINKABLE;
12049+ spin_unlock(&h_inode->i_lock);
12050+ }
4a4d8108
AM
12051+ au_set_fbstart(file, bindex);
12052+ au_set_h_fptr(file, bindex, h_file);
12053+ au_update_figen(file);
523b37e3
AM
12054+ finfo->fi_file = file;
12055+ au_sphl_add(&finfo->fi_hlist, &au_sbi(dentry->d_sb)->si_files);
4a4d8108
AM
12056+ /* todo: necessary? */
12057+ /* file->f_ra = h_file->f_ra; */
12058+ }
027c5e7a 12059+
4a4d8108 12060+ return err;
1facf9fc 12061+}
12062+
4a4d8108
AM
12063+static int aufs_open_nondir(struct inode *inode __maybe_unused,
12064+ struct file *file)
1facf9fc 12065+{
4a4d8108 12066+ int err;
1308ab2a 12067+ struct super_block *sb;
1facf9fc 12068+
523b37e3
AM
12069+ AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
12070+ file, vfsub_file_flags(file), file->f_mode);
1facf9fc 12071+
4a4d8108
AM
12072+ sb = file->f_dentry->d_sb;
12073+ si_read_lock(sb, AuLock_FLUSH);
12074+ err = au_do_open(file, au_do_open_nondir, /*fidir*/NULL);
12075+ si_read_unlock(sb);
12076+ return err;
12077+}
1facf9fc 12078+
4a4d8108
AM
12079+int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
12080+{
12081+ struct au_finfo *finfo;
12082+ aufs_bindex_t bindex;
1facf9fc 12083+
4a4d8108 12084+ finfo = au_fi(file);
523b37e3 12085+ au_sphl_del(&finfo->fi_hlist, &au_sbi(file->f_dentry->d_sb)->si_files);
4a4d8108 12086+ bindex = finfo->fi_btop;
b4510431 12087+ if (bindex >= 0)
4a4d8108 12088+ au_set_h_fptr(file, bindex, NULL);
7f207e10 12089+
4a4d8108
AM
12090+ au_finfo_fin(file);
12091+ return 0;
1facf9fc 12092+}
12093+
4a4d8108
AM
12094+/* ---------------------------------------------------------------------- */
12095+
12096+static int au_do_flush_nondir(struct file *file, fl_owner_t id)
dece6358 12097+{
1308ab2a 12098+ int err;
4a4d8108
AM
12099+ struct file *h_file;
12100+
12101+ err = 0;
12102+ h_file = au_hf_top(file);
12103+ if (h_file)
12104+ err = vfsub_flush(h_file, id);
12105+ return err;
12106+}
12107+
12108+static int aufs_flush_nondir(struct file *file, fl_owner_t id)
12109+{
12110+ return au_do_flush(file, id, au_do_flush_nondir);
12111+}
12112+
12113+/* ---------------------------------------------------------------------- */
9dbd164d
AM
12114+/*
12115+ * read and write functions acquire [fdi]_rwsem once, but release before
12116+ * mmap_sem. This is because to stop a race condition between mmap(2).
12117+ * Releasing these aufs-rwsem should be safe, no branch-mamagement (by keeping
12118+ * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
12119+ * read functions after [fdi]_rwsem are released, but it should be harmless.
12120+ */
4a4d8108
AM
12121+
12122+static ssize_t aufs_read(struct file *file, char __user *buf, size_t count,
12123+ loff_t *ppos)
12124+{
12125+ ssize_t err;
dece6358 12126+ struct dentry *dentry;
4a4d8108 12127+ struct file *h_file;
dece6358 12128+ struct super_block *sb;
1facf9fc 12129+
dece6358
AM
12130+ dentry = file->f_dentry;
12131+ sb = dentry->d_sb;
e49829fe 12132+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108 12133+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
dece6358
AM
12134+ if (unlikely(err))
12135+ goto out;
1facf9fc 12136+
4a4d8108 12137+ h_file = au_hf_top(file);
9dbd164d
AM
12138+ get_file(h_file);
12139+ di_read_unlock(dentry, AuLock_IR);
12140+ fi_read_unlock(file);
12141+
12142+ /* filedata may be obsoleted by concurrent copyup, but no problem */
4a4d8108
AM
12143+ err = vfsub_read_u(h_file, buf, count, ppos);
12144+ /* todo: necessary? */
12145+ /* file->f_ra = h_file->f_ra; */
9dbd164d 12146+ /* update without lock, I don't think it a problem */
c06a8ce3 12147+ fsstack_copy_attr_atime(dentry->d_inode, file_inode(h_file));
9dbd164d 12148+ fput(h_file);
1308ab2a 12149+
4f0767ce 12150+out:
dece6358
AM
12151+ si_read_unlock(sb);
12152+ return err;
12153+}
1facf9fc 12154+
e49829fe
JR
12155+/*
12156+ * todo: very ugly
12157+ * it locks both of i_mutex and si_rwsem for read in safe.
12158+ * if the plink maintenance mode continues forever (that is the problem),
12159+ * may loop forever.
12160+ */
12161+static void au_mtx_and_read_lock(struct inode *inode)
12162+{
12163+ int err;
12164+ struct super_block *sb = inode->i_sb;
12165+
12166+ while (1) {
12167+ mutex_lock(&inode->i_mutex);
12168+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
12169+ if (!err)
12170+ break;
12171+ mutex_unlock(&inode->i_mutex);
12172+ si_read_lock(sb, AuLock_NOPLMW);
12173+ si_read_unlock(sb);
12174+ }
12175+}
12176+
4a4d8108
AM
12177+static ssize_t aufs_write(struct file *file, const char __user *ubuf,
12178+ size_t count, loff_t *ppos)
dece6358 12179+{
4a4d8108
AM
12180+ ssize_t err;
12181+ struct au_pin pin;
dece6358 12182+ struct dentry *dentry;
9dbd164d 12183+ struct super_block *sb;
4a4d8108 12184+ struct inode *inode;
4a4d8108
AM
12185+ struct file *h_file;
12186+ char __user *buf = (char __user *)ubuf;
1facf9fc 12187+
dece6358 12188+ dentry = file->f_dentry;
9dbd164d 12189+ sb = dentry->d_sb;
4a4d8108 12190+ inode = dentry->d_inode;
e49829fe 12191+ au_mtx_and_read_lock(inode);
1facf9fc 12192+
4a4d8108
AM
12193+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12194+ if (unlikely(err))
12195+ goto out;
1facf9fc 12196+
4a4d8108
AM
12197+ err = au_ready_to_write(file, -1, &pin);
12198+ di_downgrade_lock(dentry, AuLock_IR);
9dbd164d
AM
12199+ if (unlikely(err)) {
12200+ di_read_unlock(dentry, AuLock_IR);
12201+ fi_write_unlock(file);
12202+ goto out;
12203+ }
1facf9fc 12204+
4a4d8108 12205+ h_file = au_hf_top(file);
9dbd164d 12206+ get_file(h_file);
4a4d8108 12207+ au_unpin(&pin);
9dbd164d
AM
12208+ di_read_unlock(dentry, AuLock_IR);
12209+ fi_write_unlock(file);
12210+
4a4d8108 12211+ err = vfsub_write_u(h_file, buf, count, ppos);
9dbd164d 12212+ ii_write_lock_child(inode);
4a4d8108 12213+ au_cpup_attr_timesizes(inode);
c06a8ce3 12214+ inode->i_mode = file_inode(h_file)->i_mode;
9dbd164d
AM
12215+ ii_write_unlock(inode);
12216+ fput(h_file);
1facf9fc 12217+
4f0767ce 12218+out:
9dbd164d 12219+ si_read_unlock(sb);
4a4d8108 12220+ mutex_unlock(&inode->i_mutex);
dece6358
AM
12221+ return err;
12222+}
1facf9fc 12223+
4a4d8108
AM
12224+static ssize_t au_do_aio(struct file *h_file, int rw, struct kiocb *kio,
12225+ const struct iovec *iov, unsigned long nv, loff_t pos)
dece6358 12226+{
4a4d8108
AM
12227+ ssize_t err;
12228+ struct file *file;
12229+ ssize_t (*func)(struct kiocb *, const struct iovec *, unsigned long,
12230+ loff_t);
1facf9fc 12231+
4a4d8108
AM
12232+ err = security_file_permission(h_file, rw);
12233+ if (unlikely(err))
12234+ goto out;
1facf9fc 12235+
4a4d8108
AM
12236+ err = -ENOSYS;
12237+ func = NULL;
12238+ if (rw == MAY_READ)
12239+ func = h_file->f_op->aio_read;
12240+ else if (rw == MAY_WRITE)
12241+ func = h_file->f_op->aio_write;
12242+ if (func) {
12243+ file = kio->ki_filp;
12244+ kio->ki_filp = h_file;
2cbb1c4b 12245+ lockdep_off();
4a4d8108 12246+ err = func(kio, iov, nv, pos);
2cbb1c4b 12247+ lockdep_on();
4a4d8108
AM
12248+ kio->ki_filp = file;
12249+ } else
12250+ /* currently there is no such fs */
12251+ WARN_ON_ONCE(1);
1facf9fc 12252+
4f0767ce 12253+out:
dece6358
AM
12254+ return err;
12255+}
1facf9fc 12256+
4a4d8108
AM
12257+static ssize_t aufs_aio_read(struct kiocb *kio, const struct iovec *iov,
12258+ unsigned long nv, loff_t pos)
1facf9fc 12259+{
4a4d8108
AM
12260+ ssize_t err;
12261+ struct file *file, *h_file;
12262+ struct dentry *dentry;
dece6358 12263+ struct super_block *sb;
1facf9fc 12264+
4a4d8108 12265+ file = kio->ki_filp;
dece6358 12266+ dentry = file->f_dentry;
1308ab2a 12267+ sb = dentry->d_sb;
e49829fe 12268+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108
AM
12269+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
12270+ if (unlikely(err))
12271+ goto out;
12272+
12273+ h_file = au_hf_top(file);
9dbd164d
AM
12274+ get_file(h_file);
12275+ di_read_unlock(dentry, AuLock_IR);
12276+ fi_read_unlock(file);
12277+
4a4d8108
AM
12278+ err = au_do_aio(h_file, MAY_READ, kio, iov, nv, pos);
12279+ /* todo: necessary? */
12280+ /* file->f_ra = h_file->f_ra; */
9dbd164d 12281+ /* update without lock, I don't think it a problem */
c06a8ce3 12282+ fsstack_copy_attr_atime(dentry->d_inode, file_inode(h_file));
9dbd164d 12283+ fput(h_file);
1facf9fc 12284+
4f0767ce 12285+out:
4a4d8108 12286+ si_read_unlock(sb);
1308ab2a 12287+ return err;
12288+}
1facf9fc 12289+
4a4d8108
AM
12290+static ssize_t aufs_aio_write(struct kiocb *kio, const struct iovec *iov,
12291+ unsigned long nv, loff_t pos)
1308ab2a 12292+{
4a4d8108
AM
12293+ ssize_t err;
12294+ struct au_pin pin;
12295+ struct dentry *dentry;
12296+ struct inode *inode;
4a4d8108 12297+ struct file *file, *h_file;
9dbd164d 12298+ struct super_block *sb;
1308ab2a 12299+
4a4d8108 12300+ file = kio->ki_filp;
1308ab2a 12301+ dentry = file->f_dentry;
9dbd164d 12302+ sb = dentry->d_sb;
1308ab2a 12303+ inode = dentry->d_inode;
e49829fe
JR
12304+ au_mtx_and_read_lock(inode);
12305+
4a4d8108
AM
12306+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12307+ if (unlikely(err))
1308ab2a 12308+ goto out;
1facf9fc 12309+
4a4d8108
AM
12310+ err = au_ready_to_write(file, -1, &pin);
12311+ di_downgrade_lock(dentry, AuLock_IR);
9dbd164d
AM
12312+ if (unlikely(err)) {
12313+ di_read_unlock(dentry, AuLock_IR);
12314+ fi_write_unlock(file);
12315+ goto out;
12316+ }
1facf9fc 12317+
4a4d8108 12318+ h_file = au_hf_top(file);
9dbd164d
AM
12319+ get_file(h_file);
12320+ au_unpin(&pin);
12321+ di_read_unlock(dentry, AuLock_IR);
12322+ fi_write_unlock(file);
12323+
4a4d8108 12324+ err = au_do_aio(h_file, MAY_WRITE, kio, iov, nv, pos);
9dbd164d 12325+ ii_write_lock_child(inode);
4a4d8108 12326+ au_cpup_attr_timesizes(inode);
c06a8ce3 12327+ inode->i_mode = file_inode(h_file)->i_mode;
9dbd164d
AM
12328+ ii_write_unlock(inode);
12329+ fput(h_file);
1facf9fc 12330+
4f0767ce 12331+out:
9dbd164d 12332+ si_read_unlock(sb);
4a4d8108 12333+ mutex_unlock(&inode->i_mutex);
dece6358 12334+ return err;
1facf9fc 12335+}
12336+
4a4d8108
AM
12337+static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
12338+ struct pipe_inode_info *pipe, size_t len,
12339+ unsigned int flags)
1facf9fc 12340+{
4a4d8108
AM
12341+ ssize_t err;
12342+ struct file *h_file;
12343+ struct dentry *dentry;
dece6358 12344+ struct super_block *sb;
1facf9fc 12345+
dece6358 12346+ dentry = file->f_dentry;
dece6358 12347+ sb = dentry->d_sb;
e49829fe 12348+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108
AM
12349+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
12350+ if (unlikely(err))
dece6358 12351+ goto out;
1facf9fc 12352+
4a4d8108
AM
12353+ err = -EINVAL;
12354+ h_file = au_hf_top(file);
9dbd164d 12355+ get_file(h_file);
4a4d8108 12356+ if (au_test_loopback_kthread()) {
87a755f4
AM
12357+ au_warn_loopback(h_file->f_dentry->d_sb);
12358+ if (file->f_mapping != h_file->f_mapping) {
12359+ file->f_mapping = h_file->f_mapping;
12360+ smp_mb(); /* unnecessary? */
12361+ }
1308ab2a 12362+ }
9dbd164d
AM
12363+ di_read_unlock(dentry, AuLock_IR);
12364+ fi_read_unlock(file);
12365+
4a4d8108
AM
12366+ err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
12367+ /* todo: necessasry? */
12368+ /* file->f_ra = h_file->f_ra; */
9dbd164d 12369+ /* update without lock, I don't think it a problem */
c06a8ce3 12370+ fsstack_copy_attr_atime(dentry->d_inode, file_inode(h_file));
9dbd164d 12371+ fput(h_file);
1facf9fc 12372+
4f0767ce 12373+out:
4a4d8108 12374+ si_read_unlock(sb);
dece6358 12375+ return err;
1facf9fc 12376+}
12377+
4a4d8108
AM
12378+static ssize_t
12379+aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
12380+ size_t len, unsigned int flags)
1facf9fc 12381+{
4a4d8108
AM
12382+ ssize_t err;
12383+ struct au_pin pin;
12384+ struct dentry *dentry;
12385+ struct inode *inode;
4a4d8108 12386+ struct file *h_file;
9dbd164d 12387+ struct super_block *sb;
1facf9fc 12388+
4a4d8108 12389+ dentry = file->f_dentry;
9dbd164d 12390+ sb = dentry->d_sb;
4a4d8108 12391+ inode = dentry->d_inode;
e49829fe 12392+ au_mtx_and_read_lock(inode);
9dbd164d 12393+
4a4d8108
AM
12394+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12395+ if (unlikely(err))
12396+ goto out;
1facf9fc 12397+
4a4d8108
AM
12398+ err = au_ready_to_write(file, -1, &pin);
12399+ di_downgrade_lock(dentry, AuLock_IR);
9dbd164d
AM
12400+ if (unlikely(err)) {
12401+ di_read_unlock(dentry, AuLock_IR);
12402+ fi_write_unlock(file);
12403+ goto out;
12404+ }
1facf9fc 12405+
4a4d8108 12406+ h_file = au_hf_top(file);
9dbd164d 12407+ get_file(h_file);
4a4d8108 12408+ au_unpin(&pin);
9dbd164d
AM
12409+ di_read_unlock(dentry, AuLock_IR);
12410+ fi_write_unlock(file);
12411+
4a4d8108 12412+ err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
9dbd164d 12413+ ii_write_lock_child(inode);
4a4d8108 12414+ au_cpup_attr_timesizes(inode);
c06a8ce3 12415+ inode->i_mode = file_inode(h_file)->i_mode;
9dbd164d
AM
12416+ ii_write_unlock(inode);
12417+ fput(h_file);
1facf9fc 12418+
4f0767ce 12419+out:
9dbd164d 12420+ si_read_unlock(sb);
4a4d8108
AM
12421+ mutex_unlock(&inode->i_mutex);
12422+ return err;
12423+}
1facf9fc 12424+
38d290e6
JR
12425+static long aufs_fallocate(struct file *file, int mode, loff_t offset,
12426+ loff_t len)
12427+{
12428+ long err;
12429+ struct au_pin pin;
12430+ struct dentry *dentry;
12431+ struct super_block *sb;
12432+ struct inode *inode;
12433+ struct file *h_file;
12434+
12435+ dentry = file->f_dentry;
12436+ sb = dentry->d_sb;
12437+ inode = dentry->d_inode;
12438+ au_mtx_and_read_lock(inode);
12439+
12440+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12441+ if (unlikely(err))
12442+ goto out;
12443+
12444+ err = au_ready_to_write(file, -1, &pin);
12445+ di_downgrade_lock(dentry, AuLock_IR);
12446+ if (unlikely(err)) {
12447+ di_read_unlock(dentry, AuLock_IR);
12448+ fi_write_unlock(file);
12449+ goto out;
12450+ }
12451+
12452+ h_file = au_hf_top(file);
12453+ get_file(h_file);
12454+ au_unpin(&pin);
12455+ di_read_unlock(dentry, AuLock_IR);
12456+ fi_write_unlock(file);
12457+
12458+ lockdep_off();
12459+ err = do_fallocate(h_file, mode, offset, len);
12460+ lockdep_on();
12461+ ii_write_lock_child(inode);
12462+ au_cpup_attr_timesizes(inode);
12463+ inode->i_mode = file_inode(h_file)->i_mode;
12464+ ii_write_unlock(inode);
12465+ fput(h_file);
12466+
12467+out:
12468+ si_read_unlock(sb);
12469+ mutex_unlock(&inode->i_mutex);
12470+ return err;
12471+}
12472+
4a4d8108
AM
12473+/* ---------------------------------------------------------------------- */
12474+
9dbd164d
AM
12475+/*
12476+ * The locking order around current->mmap_sem.
12477+ * - in most and regular cases
12478+ * file I/O syscall -- aufs_read() or something
12479+ * -- si_rwsem for read -- mmap_sem
12480+ * (Note that [fdi]i_rwsem are released before mmap_sem).
12481+ * - in mmap case
12482+ * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
12483+ * This AB-BA order is definitly bad, but is not a problem since "si_rwsem for
12484+ * read" allows muliple processes to acquire it and [fdi]i_rwsem are not held in
12485+ * file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
12486+ * It means that when aufs acquires si_rwsem for write, the process should never
12487+ * acquire mmap_sem.
12488+ *
392086de 12489+ * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
9dbd164d
AM
12490+ * problem either since any directory is not able to be mmap-ed.
12491+ * The similar scenario is applied to aufs_readlink() too.
12492+ */
12493+
38d290e6 12494+#if 0 /* stop calling security_file_mmap() */
2dfbb274
AM
12495+/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
12496+#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
12497+
12498+static unsigned long au_arch_prot_conv(unsigned long flags)
12499+{
12500+ /* currently ppc64 only */
12501+#ifdef CONFIG_PPC64
12502+ /* cf. linux/arch/powerpc/include/asm/mman.h */
12503+ AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
12504+ return AuConv_VM_PROT(flags, SAO);
12505+#else
12506+ AuDebugOn(arch_calc_vm_prot_bits(-1));
12507+ return 0;
12508+#endif
12509+}
12510+
12511+static unsigned long au_prot_conv(unsigned long flags)
12512+{
12513+ return AuConv_VM_PROT(flags, READ)
12514+ | AuConv_VM_PROT(flags, WRITE)
12515+ | AuConv_VM_PROT(flags, EXEC)
12516+ | au_arch_prot_conv(flags);
12517+}
12518+
12519+/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
12520+#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
12521+
12522+static unsigned long au_flag_conv(unsigned long flags)
12523+{
12524+ return AuConv_VM_MAP(flags, GROWSDOWN)
12525+ | AuConv_VM_MAP(flags, DENYWRITE)
2dfbb274
AM
12526+ | AuConv_VM_MAP(flags, LOCKED);
12527+}
38d290e6 12528+#endif
2dfbb274 12529+
9dbd164d 12530+static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
dece6358 12531+{
4a4d8108
AM
12532+ int err;
12533+ aufs_bindex_t bstart;
12534+ const unsigned char wlock
9dbd164d 12535+ = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
4a4d8108
AM
12536+ struct dentry *dentry;
12537+ struct super_block *sb;
9dbd164d
AM
12538+ struct file *h_file;
12539+ struct au_branch *br;
12540+ struct au_pin pin;
12541+
12542+ AuDbgVmRegion(file, vma);
1308ab2a 12543+
4a4d8108
AM
12544+ dentry = file->f_dentry;
12545+ sb = dentry->d_sb;
9dbd164d 12546+ lockdep_off();
e49829fe 12547+ si_read_lock(sb, AuLock_NOPLMW);
4a4d8108
AM
12548+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12549+ if (unlikely(err))
12550+ goto out;
12551+
4a4d8108 12552+ if (wlock) {
4a4d8108
AM
12553+ err = au_ready_to_write(file, -1, &pin);
12554+ di_write_unlock(dentry);
9dbd164d
AM
12555+ if (unlikely(err)) {
12556+ fi_write_unlock(file);
12557+ goto out;
12558+ }
4a4d8108
AM
12559+ au_unpin(&pin);
12560+ } else
12561+ di_write_unlock(dentry);
9dbd164d 12562+
4a4d8108 12563+ bstart = au_fbstart(file);
9dbd164d
AM
12564+ br = au_sbr(sb, bstart);
12565+ h_file = au_hf_top(file);
12566+ get_file(h_file);
2cbb1c4b 12567+ au_set_mmapped(file);
4a4d8108 12568+ fi_write_unlock(file);
9dbd164d 12569+ lockdep_on();
1308ab2a 12570+
9dbd164d 12571+ au_vm_file_reset(vma, h_file);
38d290e6
JR
12572+ /*
12573+ * we cannot call security_mmap_file() here since it may acquire
12574+ * mmap_sem or i_mutex.
12575+ *
12576+ * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
12577+ * au_flag_conv(vma->vm_flags));
12578+ */
9dbd164d
AM
12579+ if (!err)
12580+ err = h_file->f_op->mmap(h_file, vma);
2cbb1c4b
JR
12581+ if (unlikely(err))
12582+ goto out_reset;
4a4d8108 12583+
fb47a38f 12584+ au_vm_prfile_set(vma, file);
4a4d8108 12585+ /* update without lock, I don't think it a problem */
c06a8ce3 12586+ fsstack_copy_attr_atime(file_inode(file), file_inode(h_file));
2cbb1c4b 12587+ goto out_fput; /* success */
4a4d8108 12588+
2cbb1c4b
JR
12589+out_reset:
12590+ au_unset_mmapped(file);
12591+ au_vm_file_reset(vma, file);
12592+out_fput:
9dbd164d
AM
12593+ fput(h_file);
12594+ lockdep_off();
4f0767ce 12595+out:
9dbd164d
AM
12596+ si_read_unlock(sb);
12597+ lockdep_on();
12598+ AuTraceErr(err);
4a4d8108
AM
12599+ return err;
12600+}
12601+
12602+/* ---------------------------------------------------------------------- */
12603+
1e00d052
AM
12604+static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
12605+ int datasync)
4a4d8108
AM
12606+{
12607+ int err;
12608+ struct au_pin pin;
b752ccd1 12609+ struct dentry *dentry;
4a4d8108
AM
12610+ struct inode *inode;
12611+ struct file *h_file;
12612+ struct super_block *sb;
12613+
b752ccd1 12614+ dentry = file->f_dentry;
4a4d8108 12615+ inode = dentry->d_inode;
4a4d8108 12616+ sb = dentry->d_sb;
1e00d052 12617+ mutex_lock(&inode->i_mutex);
e49829fe
JR
12618+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
12619+ if (unlikely(err))
12620+ goto out;
4a4d8108
AM
12621+
12622+ err = 0; /* -EBADF; */ /* posix? */
12623+ if (unlikely(!(file->f_mode & FMODE_WRITE)))
e49829fe 12624+ goto out_si;
4a4d8108
AM
12625+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12626+ if (unlikely(err))
e49829fe 12627+ goto out_si;
4a4d8108
AM
12628+
12629+ err = au_ready_to_write(file, -1, &pin);
12630+ di_downgrade_lock(dentry, AuLock_IR);
12631+ if (unlikely(err))
12632+ goto out_unlock;
12633+ au_unpin(&pin);
12634+
12635+ err = -EINVAL;
12636+ h_file = au_hf_top(file);
53392da6
AM
12637+ err = vfsub_fsync(h_file, &h_file->f_path, datasync);
12638+ au_cpup_attr_timesizes(inode);
4a4d8108 12639+
4f0767ce 12640+out_unlock:
4a4d8108 12641+ di_read_unlock(dentry, AuLock_IR);
1308ab2a 12642+ fi_write_unlock(file);
e49829fe 12643+out_si:
953406b4 12644+ si_read_unlock(sb);
e49829fe 12645+out:
1e00d052 12646+ mutex_unlock(&inode->i_mutex);
4a4d8108 12647+ return err;
dece6358
AM
12648+}
12649+
4a4d8108
AM
12650+/* no one supports this operation, currently */
12651+#if 0
12652+static int aufs_aio_fsync_nondir(struct kiocb *kio, int datasync)
dece6358 12653+{
4a4d8108
AM
12654+ int err;
12655+ struct au_pin pin;
1308ab2a 12656+ struct dentry *dentry;
4a4d8108
AM
12657+ struct inode *inode;
12658+ struct file *file, *h_file;
1308ab2a 12659+
4a4d8108 12660+ file = kio->ki_filp;
1308ab2a 12661+ dentry = file->f_dentry;
4a4d8108 12662+ inode = dentry->d_inode;
e49829fe 12663+ au_mtx_and_read_lock(inode);
4a4d8108
AM
12664+
12665+ err = 0; /* -EBADF; */ /* posix? */
12666+ if (unlikely(!(file->f_mode & FMODE_WRITE)))
12667+ goto out;
12668+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
12669+ if (unlikely(err))
1308ab2a 12670+ goto out;
12671+
4a4d8108
AM
12672+ err = au_ready_to_write(file, -1, &pin);
12673+ di_downgrade_lock(dentry, AuLock_IR);
12674+ if (unlikely(err))
12675+ goto out_unlock;
12676+ au_unpin(&pin);
1308ab2a 12677+
4a4d8108
AM
12678+ err = -ENOSYS;
12679+ h_file = au_hf_top(file);
523b37e3 12680+ if (h_file->f_op->aio_fsync) {
4a4d8108 12681+ struct mutex *h_mtx;
1308ab2a 12682+
c06a8ce3 12683+ h_mtx = &file_inode(h_file)->i_mutex;
4a4d8108
AM
12684+ if (!is_sync_kiocb(kio)) {
12685+ get_file(h_file);
12686+ fput(file);
12687+ }
12688+ kio->ki_filp = h_file;
12689+ err = h_file->f_op->aio_fsync(kio, datasync);
12690+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
12691+ if (!err)
12692+ vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL);
12693+ /*ignore*/
12694+ au_cpup_attr_timesizes(inode);
12695+ mutex_unlock(h_mtx);
12696+ }
1308ab2a 12697+
4f0767ce 12698+out_unlock:
4a4d8108
AM
12699+ di_read_unlock(dentry, AuLock_IR);
12700+ fi_write_unlock(file);
4f0767ce 12701+out:
e49829fe 12702+ si_read_unlock(inode->sb);
4a4d8108
AM
12703+ mutex_unlock(&inode->i_mutex);
12704+ return err;
dece6358 12705+}
4a4d8108 12706+#endif
dece6358 12707+
4a4d8108 12708+static int aufs_fasync(int fd, struct file *file, int flag)
dece6358 12709+{
4a4d8108
AM
12710+ int err;
12711+ struct file *h_file;
12712+ struct dentry *dentry;
12713+ struct super_block *sb;
1308ab2a 12714+
4a4d8108
AM
12715+ dentry = file->f_dentry;
12716+ sb = dentry->d_sb;
e49829fe 12717+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108
AM
12718+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
12719+ if (unlikely(err))
12720+ goto out;
12721+
12722+ h_file = au_hf_top(file);
523b37e3 12723+ if (h_file->f_op->fasync)
4a4d8108
AM
12724+ err = h_file->f_op->fasync(fd, h_file, flag);
12725+
12726+ di_read_unlock(dentry, AuLock_IR);
12727+ fi_read_unlock(file);
1308ab2a 12728+
4f0767ce 12729+out:
4a4d8108 12730+ si_read_unlock(sb);
1308ab2a 12731+ return err;
dece6358 12732+}
4a4d8108
AM
12733+
12734+/* ---------------------------------------------------------------------- */
12735+
12736+/* no one supports this operation, currently */
12737+#if 0
12738+static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
12739+ size_t len, loff_t *pos , int more)
12740+{
12741+}
12742+#endif
12743+
12744+/* ---------------------------------------------------------------------- */
12745+
12746+const struct file_operations aufs_file_fop = {
12747+ .owner = THIS_MODULE,
2cbb1c4b 12748+
027c5e7a 12749+ .llseek = default_llseek,
4a4d8108
AM
12750+
12751+ .read = aufs_read,
12752+ .write = aufs_write,
12753+ .aio_read = aufs_aio_read,
12754+ .aio_write = aufs_aio_write,
12755+#ifdef CONFIG_AUFS_POLL
12756+ .poll = aufs_poll,
12757+#endif
12758+ .unlocked_ioctl = aufs_ioctl_nondir,
b752ccd1 12759+#ifdef CONFIG_COMPAT
c2b27bf2 12760+ .compat_ioctl = aufs_compat_ioctl_nondir,
b752ccd1 12761+#endif
4a4d8108
AM
12762+ .mmap = aufs_mmap,
12763+ .open = aufs_open_nondir,
12764+ .flush = aufs_flush_nondir,
12765+ .release = aufs_release_nondir,
12766+ .fsync = aufs_fsync_nondir,
12767+ /* .aio_fsync = aufs_aio_fsync_nondir, */
12768+ .fasync = aufs_fasync,
12769+ /* .sendpage = aufs_sendpage, */
12770+ .splice_write = aufs_splice_write,
12771+ .splice_read = aufs_splice_read,
12772+#if 0
12773+ .aio_splice_write = aufs_aio_splice_write,
38d290e6 12774+ .aio_splice_read = aufs_aio_splice_read,
4a4d8108 12775+#endif
38d290e6 12776+ .fallocate = aufs_fallocate
4a4d8108 12777+};
7f207e10
AM
12778diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
12779--- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 12780+++ linux/fs/aufs/fstype.h 2014-06-24 00:14:06.782719306 +0200
523b37e3 12781@@ -0,0 +1,469 @@
4a4d8108 12782+/*
523b37e3 12783+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
12784+ *
12785+ * This program, aufs is free software; you can redistribute it and/or modify
12786+ * it under the terms of the GNU General Public License as published by
12787+ * the Free Software Foundation; either version 2 of the License, or
12788+ * (at your option) any later version.
12789+ *
12790+ * This program is distributed in the hope that it will be useful,
12791+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12792+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12793+ * GNU General Public License for more details.
12794+ *
12795+ * You should have received a copy of the GNU General Public License
523b37e3 12796+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
12797+ */
12798+
12799+/*
12800+ * judging filesystem type
12801+ */
12802+
12803+#ifndef __AUFS_FSTYPE_H__
12804+#define __AUFS_FSTYPE_H__
12805+
12806+#ifdef __KERNEL__
12807+
12808+#include <linux/fs.h>
12809+#include <linux/magic.h>
12810+#include <linux/romfs_fs.h>
4a4d8108
AM
12811+
12812+static inline int au_test_aufs(struct super_block *sb)
12813+{
12814+ return sb->s_magic == AUFS_SUPER_MAGIC;
12815+}
12816+
12817+static inline const char *au_sbtype(struct super_block *sb)
12818+{
12819+ return sb->s_type->name;
12820+}
1308ab2a 12821+
12822+static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
12823+{
12824+#if defined(CONFIG_ROMFS_FS) || defined(CONFIG_ROMFS_FS_MODULE)
12825+ return sb->s_magic == ROMFS_MAGIC;
dece6358
AM
12826+#else
12827+ return 0;
12828+#endif
12829+}
12830+
1308ab2a 12831+static inline int au_test_romfs(struct super_block *sb __maybe_unused)
dece6358 12832+{
1308ab2a 12833+#if defined(CONFIG_ISO9660_FS) || defined(CONFIG_ISO9660_FS_MODULE)
12834+ return sb->s_magic == ISOFS_SUPER_MAGIC;
dece6358
AM
12835+#else
12836+ return 0;
12837+#endif
12838+}
12839+
1308ab2a 12840+static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
dece6358 12841+{
1308ab2a 12842+#if defined(CONFIG_CRAMFS) || defined(CONFIG_CRAMFS_MODULE)
12843+ return sb->s_magic == CRAMFS_MAGIC;
12844+#endif
12845+ return 0;
12846+}
12847+
12848+static inline int au_test_nfs(struct super_block *sb __maybe_unused)
12849+{
12850+#if defined(CONFIG_NFS_FS) || defined(CONFIG_NFS_FS_MODULE)
12851+ return sb->s_magic == NFS_SUPER_MAGIC;
dece6358
AM
12852+#else
12853+ return 0;
12854+#endif
12855+}
12856+
1308ab2a 12857+static inline int au_test_fuse(struct super_block *sb __maybe_unused)
dece6358 12858+{
1308ab2a 12859+#if defined(CONFIG_FUSE_FS) || defined(CONFIG_FUSE_FS_MODULE)
12860+ return sb->s_magic == FUSE_SUPER_MAGIC;
dece6358
AM
12861+#else
12862+ return 0;
12863+#endif
12864+}
12865+
1308ab2a 12866+static inline int au_test_xfs(struct super_block *sb __maybe_unused)
dece6358 12867+{
1308ab2a 12868+#if defined(CONFIG_XFS_FS) || defined(CONFIG_XFS_FS_MODULE)
12869+ return sb->s_magic == XFS_SB_MAGIC;
dece6358
AM
12870+#else
12871+ return 0;
12872+#endif
12873+}
12874+
1308ab2a 12875+static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
dece6358 12876+{
1308ab2a 12877+#ifdef CONFIG_TMPFS
12878+ return sb->s_magic == TMPFS_MAGIC;
12879+#else
12880+ return 0;
dece6358 12881+#endif
dece6358
AM
12882+}
12883+
1308ab2a 12884+static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
1facf9fc 12885+{
1308ab2a 12886+#if defined(CONFIG_ECRYPT_FS) || defined(CONFIG_ECRYPT_FS_MODULE)
12887+ return !strcmp(au_sbtype(sb), "ecryptfs");
12888+#else
12889+ return 0;
12890+#endif
1facf9fc 12891+}
12892+
1308ab2a 12893+static inline int au_test_ocfs2(struct super_block *sb __maybe_unused)
1facf9fc 12894+{
1308ab2a 12895+#if defined(CONFIG_OCFS2_FS) || defined(CONFIG_OCFS2_FS_MODULE)
12896+ return sb->s_magic == OCFS2_SUPER_MAGIC;
12897+#else
12898+ return 0;
12899+#endif
1facf9fc 12900+}
12901+
1308ab2a 12902+static inline int au_test_ocfs2_dlmfs(struct super_block *sb __maybe_unused)
1facf9fc 12903+{
1308ab2a 12904+#if defined(CONFIG_OCFS2_FS_O2CB) || defined(CONFIG_OCFS2_FS_O2CB_MODULE)
12905+ return sb->s_magic == DLMFS_MAGIC;
12906+#else
12907+ return 0;
12908+#endif
1facf9fc 12909+}
12910+
1308ab2a 12911+static inline int au_test_coda(struct super_block *sb __maybe_unused)
1facf9fc 12912+{
1308ab2a 12913+#if defined(CONFIG_CODA_FS) || defined(CONFIG_CODA_FS_MODULE)
12914+ return sb->s_magic == CODA_SUPER_MAGIC;
12915+#else
12916+ return 0;
12917+#endif
12918+}
12919+
12920+static inline int au_test_v9fs(struct super_block *sb __maybe_unused)
12921+{
12922+#if defined(CONFIG_9P_FS) || defined(CONFIG_9P_FS_MODULE)
12923+ return sb->s_magic == V9FS_MAGIC;
12924+#else
12925+ return 0;
12926+#endif
12927+}
12928+
12929+static inline int au_test_ext4(struct super_block *sb __maybe_unused)
12930+{
c2b27bf2 12931+#if defined(CONFIG_EXT4_FS) || defined(CONFIG_EXT4_FS_MODULE)
1308ab2a 12932+ return sb->s_magic == EXT4_SUPER_MAGIC;
12933+#else
12934+ return 0;
12935+#endif
12936+}
12937+
12938+static inline int au_test_sysv(struct super_block *sb __maybe_unused)
12939+{
12940+#if defined(CONFIG_SYSV_FS) || defined(CONFIG_SYSV_FS_MODULE)
12941+ return !strcmp(au_sbtype(sb), "sysv");
12942+#else
12943+ return 0;
12944+#endif
12945+}
12946+
12947+static inline int au_test_ramfs(struct super_block *sb)
12948+{
12949+ return sb->s_magic == RAMFS_MAGIC;
12950+}
12951+
12952+static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
12953+{
12954+#if defined(CONFIG_UBIFS_FS) || defined(CONFIG_UBIFS_FS_MODULE)
12955+ return sb->s_magic == UBIFS_SUPER_MAGIC;
12956+#else
12957+ return 0;
12958+#endif
12959+}
12960+
12961+static inline int au_test_procfs(struct super_block *sb __maybe_unused)
12962+{
12963+#ifdef CONFIG_PROC_FS
12964+ return sb->s_magic == PROC_SUPER_MAGIC;
12965+#else
12966+ return 0;
12967+#endif
12968+}
12969+
12970+static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
12971+{
12972+#ifdef CONFIG_SYSFS
12973+ return sb->s_magic == SYSFS_MAGIC;
12974+#else
12975+ return 0;
12976+#endif
12977+}
12978+
12979+static inline int au_test_configfs(struct super_block *sb __maybe_unused)
12980+{
12981+#if defined(CONFIG_CONFIGFS_FS) || defined(CONFIG_CONFIGFS_FS_MODULE)
12982+ return sb->s_magic == CONFIGFS_MAGIC;
12983+#else
12984+ return 0;
12985+#endif
12986+}
12987+
12988+static inline int au_test_minix(struct super_block *sb __maybe_unused)
12989+{
12990+#if defined(CONFIG_MINIX_FS) || defined(CONFIG_MINIX_FS_MODULE)
12991+ return sb->s_magic == MINIX3_SUPER_MAGIC
12992+ || sb->s_magic == MINIX2_SUPER_MAGIC
12993+ || sb->s_magic == MINIX2_SUPER_MAGIC2
12994+ || sb->s_magic == MINIX_SUPER_MAGIC
12995+ || sb->s_magic == MINIX_SUPER_MAGIC2;
12996+#else
12997+ return 0;
12998+#endif
12999+}
13000+
13001+static inline int au_test_cifs(struct super_block *sb __maybe_unused)
13002+{
13003+#if defined(CONFIG_CIFS_FS) || defined(CONFIGCIFS_FS_MODULE)
13004+ return sb->s_magic == CIFS_MAGIC_NUMBER;
13005+#else
13006+ return 0;
13007+#endif
13008+}
13009+
13010+static inline int au_test_fat(struct super_block *sb __maybe_unused)
13011+{
13012+#if defined(CONFIG_FAT_FS) || defined(CONFIG_FAT_FS_MODULE)
13013+ return sb->s_magic == MSDOS_SUPER_MAGIC;
13014+#else
13015+ return 0;
13016+#endif
13017+}
13018+
13019+static inline int au_test_msdos(struct super_block *sb)
13020+{
13021+ return au_test_fat(sb);
13022+}
13023+
13024+static inline int au_test_vfat(struct super_block *sb)
13025+{
13026+ return au_test_fat(sb);
13027+}
13028+
13029+static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
13030+{
13031+#ifdef CONFIG_SECURITYFS
13032+ return sb->s_magic == SECURITYFS_MAGIC;
13033+#else
13034+ return 0;
13035+#endif
13036+}
13037+
13038+static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
13039+{
13040+#if defined(CONFIG_SQUASHFS) || defined(CONFIG_SQUASHFS_MODULE)
13041+ return sb->s_magic == SQUASHFS_MAGIC;
13042+#else
13043+ return 0;
13044+#endif
13045+}
13046+
13047+static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
13048+{
13049+#if defined(CONFIG_BTRFS_FS) || defined(CONFIG_BTRFS_FS_MODULE)
13050+ return sb->s_magic == BTRFS_SUPER_MAGIC;
13051+#else
13052+ return 0;
13053+#endif
13054+}
13055+
13056+static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
13057+{
13058+#if defined(CONFIG_XENFS) || defined(CONFIG_XENFS_MODULE)
13059+ return sb->s_magic == XENFS_SUPER_MAGIC;
13060+#else
13061+ return 0;
13062+#endif
13063+}
13064+
13065+static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
13066+{
13067+#ifdef CONFIG_DEBUG_FS
13068+ return sb->s_magic == DEBUGFS_MAGIC;
13069+#else
13070+ return 0;
13071+#endif
13072+}
13073+
13074+static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
13075+{
13076+#if defined(CONFIG_NILFS) || defined(CONFIG_NILFS_MODULE)
13077+ return sb->s_magic == NILFS_SUPER_MAGIC;
13078+#else
13079+ return 0;
13080+#endif
13081+}
13082+
4a4d8108
AM
13083+static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
13084+{
13085+#if defined(CONFIG_HFSPLUS_FS) || defined(CONFIG_HFSPLUS_FS_MODULE)
13086+ return sb->s_magic == HFSPLUS_SUPER_MAGIC;
13087+#else
13088+ return 0;
13089+#endif
13090+}
13091+
1308ab2a 13092+/* ---------------------------------------------------------------------- */
13093+/*
13094+ * they can't be an aufs branch.
13095+ */
13096+static inline int au_test_fs_unsuppoted(struct super_block *sb)
13097+{
13098+ return
13099+#ifndef CONFIG_AUFS_BR_RAMFS
13100+ au_test_ramfs(sb) ||
13101+#endif
13102+ au_test_procfs(sb)
13103+ || au_test_sysfs(sb)
13104+ || au_test_configfs(sb)
13105+ || au_test_debugfs(sb)
13106+ || au_test_securityfs(sb)
13107+ || au_test_xenfs(sb)
13108+ || au_test_ecryptfs(sb)
13109+ /* || !strcmp(au_sbtype(sb), "unionfs") */
13110+ || au_test_aufs(sb); /* will be supported in next version */
13111+}
13112+
1308ab2a 13113+static inline int au_test_fs_remote(struct super_block *sb)
13114+{
13115+ return !au_test_tmpfs(sb)
13116+#ifdef CONFIG_AUFS_BR_RAMFS
13117+ && !au_test_ramfs(sb)
13118+#endif
13119+ && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
13120+}
13121+
13122+/* ---------------------------------------------------------------------- */
13123+
13124+/*
13125+ * Note: these functions (below) are created after reading ->getattr() in all
13126+ * filesystems under linux/fs. it means we have to do so in every update...
13127+ */
13128+
13129+/*
13130+ * some filesystems require getattr to refresh the inode attributes before
13131+ * referencing.
13132+ * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
13133+ * and leave the work for d_revalidate()
13134+ */
13135+static inline int au_test_fs_refresh_iattr(struct super_block *sb)
13136+{
13137+ return au_test_nfs(sb)
13138+ || au_test_fuse(sb)
1308ab2a 13139+ /* || au_test_ocfs2(sb) */ /* untested */
13140+ /* || au_test_btrfs(sb) */ /* untested */
13141+ /* || au_test_coda(sb) */ /* untested */
13142+ /* || au_test_v9fs(sb) */ /* untested */
13143+ ;
13144+}
13145+
13146+/*
13147+ * filesystems which don't maintain i_size or i_blocks.
13148+ */
13149+static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
13150+{
13151+ return au_test_xfs(sb)
4a4d8108
AM
13152+ || au_test_btrfs(sb)
13153+ || au_test_ubifs(sb)
13154+ || au_test_hfsplus(sb) /* maintained, but incorrect */
1308ab2a 13155+ /* || au_test_ext4(sb) */ /* untested */
13156+ /* || au_test_ocfs2(sb) */ /* untested */
13157+ /* || au_test_ocfs2_dlmfs(sb) */ /* untested */
13158+ /* || au_test_sysv(sb) */ /* untested */
1308ab2a 13159+ /* || au_test_minix(sb) */ /* untested */
13160+ ;
13161+}
13162+
13163+/*
13164+ * filesystems which don't store the correct value in some of their inode
13165+ * attributes.
13166+ */
13167+static inline int au_test_fs_bad_iattr(struct super_block *sb)
13168+{
13169+ return au_test_fs_bad_iattr_size(sb)
13170+ /* || au_test_cifs(sb) */ /* untested */
13171+ || au_test_fat(sb)
13172+ || au_test_msdos(sb)
13173+ || au_test_vfat(sb);
1facf9fc 13174+}
13175+
13176+/* they don't check i_nlink in link(2) */
13177+static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
13178+{
13179+ return au_test_tmpfs(sb)
13180+#ifdef CONFIG_AUFS_BR_RAMFS
13181+ || au_test_ramfs(sb)
13182+#endif
4a4d8108 13183+ || au_test_ubifs(sb)
4a4d8108 13184+ || au_test_hfsplus(sb);
1facf9fc 13185+}
13186+
13187+/*
13188+ * filesystems which sets S_NOATIME and S_NOCMTIME.
13189+ */
13190+static inline int au_test_fs_notime(struct super_block *sb)
13191+{
13192+ return au_test_nfs(sb)
13193+ || au_test_fuse(sb)
dece6358 13194+ || au_test_ubifs(sb)
1facf9fc 13195+ /* || au_test_cifs(sb) */ /* untested */
1facf9fc 13196+ ;
13197+}
13198+
13199+/*
13200+ * filesystems which requires replacing i_mapping.
13201+ */
13202+static inline int au_test_fs_bad_mapping(struct super_block *sb)
13203+{
dece6358
AM
13204+ return au_test_fuse(sb)
13205+ || au_test_ubifs(sb);
1facf9fc 13206+}
13207+
13208+/* temporary support for i#1 in cramfs */
13209+static inline int au_test_fs_unique_ino(struct inode *inode)
13210+{
13211+ if (au_test_cramfs(inode->i_sb))
13212+ return inode->i_ino != 1;
13213+ return 1;
13214+}
13215+
13216+/* ---------------------------------------------------------------------- */
13217+
13218+/*
13219+ * the filesystem where the xino files placed must support i/o after unlink and
13220+ * maintain i_size and i_blocks.
13221+ */
13222+static inline int au_test_fs_bad_xino(struct super_block *sb)
13223+{
13224+ return au_test_fs_remote(sb)
13225+ || au_test_fs_bad_iattr_size(sb)
1facf9fc 13226+ /* don't want unnecessary work for xino */
13227+ || au_test_aufs(sb)
1308ab2a 13228+ || au_test_ecryptfs(sb)
13229+ || au_test_nilfs(sb);
1facf9fc 13230+}
13231+
13232+static inline int au_test_fs_trunc_xino(struct super_block *sb)
13233+{
13234+ return au_test_tmpfs(sb)
13235+ || au_test_ramfs(sb);
13236+}
13237+
13238+/*
13239+ * test if the @sb is real-readonly.
13240+ */
13241+static inline int au_test_fs_rr(struct super_block *sb)
13242+{
13243+ return au_test_squashfs(sb)
13244+ || au_test_iso9660(sb)
13245+ || au_test_cramfs(sb)
13246+ || au_test_romfs(sb);
13247+}
13248+
13249+#endif /* __KERNEL__ */
13250+#endif /* __AUFS_FSTYPE_H__ */
7f207e10
AM
13251diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
13252--- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 13253+++ linux/fs/aufs/hfsnotify.c 2014-06-24 00:14:06.782719306 +0200
fb47a38f 13254@@ -0,0 +1,281 @@
1facf9fc 13255+/*
523b37e3 13256+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 13257+ *
13258+ * This program, aufs is free software; you can redistribute it and/or modify
13259+ * it under the terms of the GNU General Public License as published by
13260+ * the Free Software Foundation; either version 2 of the License, or
13261+ * (at your option) any later version.
dece6358
AM
13262+ *
13263+ * This program is distributed in the hope that it will be useful,
13264+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
13265+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13266+ * GNU General Public License for more details.
13267+ *
13268+ * You should have received a copy of the GNU General Public License
523b37e3 13269+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 13270+ */
13271+
13272+/*
4a4d8108 13273+ * fsnotify for the lower directories
1facf9fc 13274+ */
13275+
13276+#include "aufs.h"
13277+
4a4d8108
AM
13278+/* FS_IN_IGNORED is unnecessary */
13279+static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
13280+ | FS_CREATE | FS_EVENT_ON_CHILD);
7f207e10 13281+static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
7eafdf33 13282+static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
1facf9fc 13283+
0c5527e5 13284+static void au_hfsn_free_mark(struct fsnotify_mark *mark)
1facf9fc 13285+{
0c5527e5
AM
13286+ struct au_hnotify *hn = container_of(mark, struct au_hnotify,
13287+ hn_mark);
4a4d8108 13288+ AuDbg("here\n");
7eafdf33
AM
13289+ au_cache_free_hnotify(hn);
13290+ smp_mb__before_atomic_dec();
1716fcea
AM
13291+ if (atomic64_dec_and_test(&au_hfsn_ifree))
13292+ wake_up(&au_hfsn_wq);
4a4d8108 13293+}
1facf9fc 13294+
027c5e7a 13295+static int au_hfsn_alloc(struct au_hinode *hinode)
4a4d8108 13296+{
1716fcea 13297+ int err;
027c5e7a
AM
13298+ struct au_hnotify *hn;
13299+ struct super_block *sb;
13300+ struct au_branch *br;
0c5527e5 13301+ struct fsnotify_mark *mark;
027c5e7a 13302+ aufs_bindex_t bindex;
1facf9fc 13303+
027c5e7a
AM
13304+ hn = hinode->hi_notify;
13305+ sb = hn->hn_aufs_inode->i_sb;
13306+ bindex = au_br_index(sb, hinode->hi_id);
13307+ br = au_sbr(sb, bindex);
1716fcea
AM
13308+ AuDebugOn(!br->br_hfsn);
13309+
0c5527e5
AM
13310+ mark = &hn->hn_mark;
13311+ fsnotify_init_mark(mark, au_hfsn_free_mark);
13312+ mark->mask = AuHfsnMask;
7f207e10
AM
13313+ /*
13314+ * by udba rename or rmdir, aufs assign a new inode to the known
13315+ * h_inode, so specify 1 to allow dups.
13316+ */
1716fcea 13317+ err = fsnotify_add_mark(mark, br->br_hfsn->hfsn_group, hinode->hi_inode,
027c5e7a 13318+ /*mnt*/NULL, /*allow_dups*/1);
1716fcea
AM
13319+ /* even if err */
13320+ fsnotify_put_mark(mark);
13321+
13322+ return err;
1facf9fc 13323+}
13324+
7eafdf33 13325+static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
1facf9fc 13326+{
0c5527e5 13327+ struct fsnotify_mark *mark;
7eafdf33 13328+ unsigned long long ull;
1716fcea 13329+ struct fsnotify_group *group;
7eafdf33
AM
13330+
13331+ ull = atomic64_inc_return(&au_hfsn_ifree);
13332+ BUG_ON(!ull);
953406b4 13333+
0c5527e5 13334+ mark = &hn->hn_mark;
1716fcea
AM
13335+ spin_lock(&mark->lock);
13336+ group = mark->group;
13337+ fsnotify_get_group(group);
13338+ spin_unlock(&mark->lock);
13339+ fsnotify_destroy_mark(mark, group);
13340+ fsnotify_put_group(group);
7f207e10 13341+
7eafdf33
AM
13342+ /* free hn by myself */
13343+ return 0;
1facf9fc 13344+}
13345+
13346+/* ---------------------------------------------------------------------- */
13347+
4a4d8108 13348+static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
1facf9fc 13349+{
0c5527e5 13350+ struct fsnotify_mark *mark;
1facf9fc 13351+
0c5527e5
AM
13352+ mark = &hinode->hi_notify->hn_mark;
13353+ spin_lock(&mark->lock);
1facf9fc 13354+ if (do_set) {
0c5527e5
AM
13355+ AuDebugOn(mark->mask & AuHfsnMask);
13356+ mark->mask |= AuHfsnMask;
1facf9fc 13357+ } else {
0c5527e5
AM
13358+ AuDebugOn(!(mark->mask & AuHfsnMask));
13359+ mark->mask &= ~AuHfsnMask;
1facf9fc 13360+ }
0c5527e5 13361+ spin_unlock(&mark->lock);
4a4d8108 13362+ /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
1facf9fc 13363+}
13364+
4a4d8108 13365+/* ---------------------------------------------------------------------- */
1facf9fc 13366+
4a4d8108
AM
13367+/* #define AuDbgHnotify */
13368+#ifdef AuDbgHnotify
13369+static char *au_hfsn_name(u32 mask)
13370+{
13371+#ifdef CONFIG_AUFS_DEBUG
c06a8ce3
AM
13372+#define test_ret(flag) \
13373+ do { \
13374+ if (mask & flag) \
13375+ return #flag; \
13376+ } while (0)
4a4d8108
AM
13377+ test_ret(FS_ACCESS);
13378+ test_ret(FS_MODIFY);
13379+ test_ret(FS_ATTRIB);
13380+ test_ret(FS_CLOSE_WRITE);
13381+ test_ret(FS_CLOSE_NOWRITE);
13382+ test_ret(FS_OPEN);
13383+ test_ret(FS_MOVED_FROM);
13384+ test_ret(FS_MOVED_TO);
13385+ test_ret(FS_CREATE);
13386+ test_ret(FS_DELETE);
13387+ test_ret(FS_DELETE_SELF);
13388+ test_ret(FS_MOVE_SELF);
13389+ test_ret(FS_UNMOUNT);
13390+ test_ret(FS_Q_OVERFLOW);
13391+ test_ret(FS_IN_IGNORED);
13392+ test_ret(FS_IN_ISDIR);
13393+ test_ret(FS_IN_ONESHOT);
13394+ test_ret(FS_EVENT_ON_CHILD);
13395+ return "";
13396+#undef test_ret
13397+#else
13398+ return "??";
13399+#endif
1facf9fc 13400+}
4a4d8108 13401+#endif
1facf9fc 13402+
13403+/* ---------------------------------------------------------------------- */
13404+
1716fcea
AM
13405+static void au_hfsn_free_group(struct fsnotify_group *group)
13406+{
13407+ struct au_br_hfsnotify *hfsn = group->private;
13408+
13409+ AuDbg("here\n");
13410+ kfree(hfsn);
13411+}
13412+
4a4d8108 13413+static int au_hfsn_handle_event(struct fsnotify_group *group,
fb47a38f 13414+ struct inode *inode,
0c5527e5
AM
13415+ struct fsnotify_mark *inode_mark,
13416+ struct fsnotify_mark *vfsmount_mark,
fb47a38f
JR
13417+ u32 mask, void *data, int data_type,
13418+ const unsigned char *file_name, u32 cookie)
1facf9fc 13419+{
13420+ int err;
4a4d8108
AM
13421+ struct au_hnotify *hnotify;
13422+ struct inode *h_dir, *h_inode;
fb47a38f 13423+ struct qstr h_child_qstr = QSTR_INIT(file_name, strlen(file_name));
4a4d8108 13424+
fb47a38f 13425+ AuDebugOn(data_type != FSNOTIFY_EVENT_INODE);
1facf9fc 13426+
13427+ err = 0;
0c5527e5 13428+ /* if FS_UNMOUNT happens, there must be another bug */
4a4d8108 13429+ AuDebugOn(mask & FS_UNMOUNT);
0c5527e5 13430+ if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
1facf9fc 13431+ goto out;
1facf9fc 13432+
fb47a38f
JR
13433+ h_dir = inode;
13434+ h_inode = NULL;
4a4d8108 13435+#ifdef AuDbgHnotify
392086de 13436+ au_debug_on();
4a4d8108
AM
13437+ if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
13438+ || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
13439+ AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
13440+ h_dir->i_ino, mask, au_hfsn_name(mask),
13441+ AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
13442+ /* WARN_ON(1); */
1facf9fc 13443+ }
392086de 13444+ au_debug_off();
1facf9fc 13445+#endif
4a4d8108 13446+
0c5527e5
AM
13447+ AuDebugOn(!inode_mark);
13448+ hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
13449+ err = au_hnotify(h_dir, hnotify, mask, &h_child_qstr, h_inode);
1facf9fc 13450+
4a4d8108
AM
13451+out:
13452+ return err;
13453+}
1facf9fc 13454+
4a4d8108 13455+static struct fsnotify_ops au_hfsn_ops = {
1716fcea
AM
13456+ .handle_event = au_hfsn_handle_event,
13457+ .free_group_priv = au_hfsn_free_group
4a4d8108
AM
13458+};
13459+
13460+/* ---------------------------------------------------------------------- */
13461+
027c5e7a
AM
13462+static void au_hfsn_fin_br(struct au_branch *br)
13463+{
1716fcea 13464+ struct au_br_hfsnotify *hfsn;
027c5e7a 13465+
1716fcea
AM
13466+ hfsn = br->br_hfsn;
13467+ if (hfsn)
13468+ fsnotify_put_group(hfsn->hfsn_group);
027c5e7a
AM
13469+}
13470+
1716fcea 13471+static int au_hfsn_init_br(struct au_branch *br, int perm)
4a4d8108
AM
13472+{
13473+ int err;
1716fcea
AM
13474+ struct fsnotify_group *group;
13475+ struct au_br_hfsnotify *hfsn;
1facf9fc 13476+
4a4d8108 13477+ err = 0;
1716fcea
AM
13478+ br->br_hfsn = NULL;
13479+ if (!au_br_hnotifyable(perm))
027c5e7a 13480+ goto out;
027c5e7a 13481+
1716fcea
AM
13482+ err = -ENOMEM;
13483+ hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
13484+ if (unlikely(!hfsn))
027c5e7a
AM
13485+ goto out;
13486+
1716fcea
AM
13487+ err = 0;
13488+ group = fsnotify_alloc_group(&au_hfsn_ops);
13489+ if (IS_ERR(group)) {
13490+ err = PTR_ERR(group);
0c5527e5 13491+ pr_err("fsnotify_alloc_group() failed, %d\n", err);
1716fcea 13492+ goto out_hfsn;
4a4d8108 13493+ }
1facf9fc 13494+
1716fcea
AM
13495+ group->private = hfsn;
13496+ hfsn->hfsn_group = group;
13497+ br->br_hfsn = hfsn;
13498+ goto out; /* success */
13499+
13500+out_hfsn:
13501+ kfree(hfsn);
027c5e7a 13502+out:
1716fcea
AM
13503+ return err;
13504+}
13505+
13506+static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
13507+{
13508+ int err;
13509+
13510+ err = 0;
13511+ if (!br->br_hfsn)
13512+ err = au_hfsn_init_br(br, perm);
13513+
1facf9fc 13514+ return err;
13515+}
13516+
7eafdf33
AM
13517+/* ---------------------------------------------------------------------- */
13518+
13519+static void au_hfsn_fin(void)
13520+{
13521+ AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
13522+ wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
13523+}
13524+
4a4d8108
AM
13525+const struct au_hnotify_op au_hnotify_op = {
13526+ .ctl = au_hfsn_ctl,
13527+ .alloc = au_hfsn_alloc,
13528+ .free = au_hfsn_free,
1facf9fc 13529+
7eafdf33
AM
13530+ .fin = au_hfsn_fin,
13531+
027c5e7a
AM
13532+ .reset_br = au_hfsn_reset_br,
13533+ .fin_br = au_hfsn_fin_br,
13534+ .init_br = au_hfsn_init_br
4a4d8108 13535+};
7f207e10
AM
13536diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
13537--- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 13538+++ linux/fs/aufs/hfsplus.c 2014-06-24 00:14:06.782719306 +0200
523b37e3 13539@@ -0,0 +1,56 @@
4a4d8108 13540+/*
523b37e3 13541+ * Copyright (C) 2010-2014 Junjiro R. Okajima
4a4d8108
AM
13542+ *
13543+ * This program, aufs is free software; you can redistribute it and/or modify
13544+ * it under the terms of the GNU General Public License as published by
13545+ * the Free Software Foundation; either version 2 of the License, or
13546+ * (at your option) any later version.
13547+ *
13548+ * This program is distributed in the hope that it will be useful,
13549+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
13550+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13551+ * GNU General Public License for more details.
13552+ *
13553+ * You should have received a copy of the GNU General Public License
523b37e3 13554+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 13555+ */
1facf9fc 13556+
4a4d8108
AM
13557+/*
13558+ * special support for filesystems which aqucires an inode mutex
13559+ * at final closing a file, eg, hfsplus.
13560+ *
13561+ * This trick is very simple and stupid, just to open the file before really
13562+ * neceeary open to tell hfsplus that this is not the final closing.
13563+ * The caller should call au_h_open_pre() after acquiring the inode mutex,
13564+ * and au_h_open_post() after releasing it.
13565+ */
1facf9fc 13566+
4a4d8108 13567+#include "aufs.h"
1facf9fc 13568+
392086de
AM
13569+struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
13570+ int force_wr)
4a4d8108
AM
13571+{
13572+ struct file *h_file;
13573+ struct dentry *h_dentry;
1facf9fc 13574+
4a4d8108
AM
13575+ h_dentry = au_h_dptr(dentry, bindex);
13576+ AuDebugOn(!h_dentry);
13577+ AuDebugOn(!h_dentry->d_inode);
4a4d8108
AM
13578+
13579+ h_file = NULL;
13580+ if (au_test_hfsplus(h_dentry->d_sb)
13581+ && S_ISREG(h_dentry->d_inode->i_mode))
13582+ h_file = au_h_open(dentry, bindex,
13583+ O_RDONLY | O_NOATIME | O_LARGEFILE,
392086de 13584+ /*file*/NULL, force_wr);
4a4d8108 13585+ return h_file;
1facf9fc 13586+}
13587+
4a4d8108
AM
13588+void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
13589+ struct file *h_file)
13590+{
13591+ if (h_file) {
13592+ fput(h_file);
13593+ au_sbr_put(dentry->d_sb, bindex);
13594+ }
13595+}
7f207e10
AM
13596diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
13597--- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
13598+++ linux/fs/aufs/hnotify.c 2014-06-24 00:14:06.782719306 +0200
13599@@ -0,0 +1,713 @@
e49829fe 13600+/*
523b37e3 13601+ * Copyright (C) 2005-2014 Junjiro R. Okajima
e49829fe
JR
13602+ *
13603+ * This program, aufs is free software; you can redistribute it and/or modify
13604+ * it under the terms of the GNU General Public License as published by
13605+ * the Free Software Foundation; either version 2 of the License, or
13606+ * (at your option) any later version.
13607+ *
13608+ * This program is distributed in the hope that it will be useful,
13609+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
13610+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13611+ * GNU General Public License for more details.
13612+ *
13613+ * You should have received a copy of the GNU General Public License
523b37e3 13614+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
e49829fe
JR
13615+ */
13616+
13617+/*
7f207e10 13618+ * abstraction to notify the direct changes on lower directories
e49829fe
JR
13619+ */
13620+
13621+#include "aufs.h"
13622+
027c5e7a 13623+int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
e49829fe
JR
13624+{
13625+ int err;
7f207e10 13626+ struct au_hnotify *hn;
1facf9fc 13627+
4a4d8108
AM
13628+ err = -ENOMEM;
13629+ hn = au_cache_alloc_hnotify();
13630+ if (hn) {
13631+ hn->hn_aufs_inode = inode;
027c5e7a
AM
13632+ hinode->hi_notify = hn;
13633+ err = au_hnotify_op.alloc(hinode);
13634+ AuTraceErr(err);
13635+ if (unlikely(err)) {
13636+ hinode->hi_notify = NULL;
4a4d8108
AM
13637+ au_cache_free_hnotify(hn);
13638+ /*
13639+ * The upper dir was removed by udba, but the same named
13640+ * dir left. In this case, aufs assignes a new inode
13641+ * number and set the monitor again.
13642+ * For the lower dir, the old monitnor is still left.
13643+ */
13644+ if (err == -EEXIST)
13645+ err = 0;
13646+ }
1308ab2a 13647+ }
1308ab2a 13648+
027c5e7a 13649+ AuTraceErr(err);
1308ab2a 13650+ return err;
dece6358 13651+}
1facf9fc 13652+
4a4d8108 13653+void au_hn_free(struct au_hinode *hinode)
dece6358 13654+{
4a4d8108 13655+ struct au_hnotify *hn;
1facf9fc 13656+
4a4d8108
AM
13657+ hn = hinode->hi_notify;
13658+ if (hn) {
4a4d8108 13659+ hinode->hi_notify = NULL;
7eafdf33
AM
13660+ if (au_hnotify_op.free(hinode, hn))
13661+ au_cache_free_hnotify(hn);
4a4d8108
AM
13662+ }
13663+}
dece6358 13664+
4a4d8108 13665+/* ---------------------------------------------------------------------- */
dece6358 13666+
4a4d8108
AM
13667+void au_hn_ctl(struct au_hinode *hinode, int do_set)
13668+{
13669+ if (hinode->hi_notify)
13670+ au_hnotify_op.ctl(hinode, do_set);
13671+}
13672+
13673+void au_hn_reset(struct inode *inode, unsigned int flags)
13674+{
13675+ aufs_bindex_t bindex, bend;
13676+ struct inode *hi;
13677+ struct dentry *iwhdentry;
1facf9fc 13678+
1308ab2a 13679+ bend = au_ibend(inode);
4a4d8108
AM
13680+ for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
13681+ hi = au_h_iptr(inode, bindex);
13682+ if (!hi)
13683+ continue;
1308ab2a 13684+
4a4d8108
AM
13685+ /* mutex_lock_nested(&hi->i_mutex, AuLsc_I_CHILD); */
13686+ iwhdentry = au_hi_wh(inode, bindex);
13687+ if (iwhdentry)
13688+ dget(iwhdentry);
13689+ au_igrab(hi);
13690+ au_set_h_iptr(inode, bindex, NULL, 0);
13691+ au_set_h_iptr(inode, bindex, au_igrab(hi),
13692+ flags & ~AuHi_XINO);
13693+ iput(hi);
13694+ dput(iwhdentry);
13695+ /* mutex_unlock(&hi->i_mutex); */
1facf9fc 13696+ }
1facf9fc 13697+}
13698+
1308ab2a 13699+/* ---------------------------------------------------------------------- */
1facf9fc 13700+
4a4d8108 13701+static int hn_xino(struct inode *inode, struct inode *h_inode)
1facf9fc 13702+{
4a4d8108
AM
13703+ int err;
13704+ aufs_bindex_t bindex, bend, bfound, bstart;
13705+ struct inode *h_i;
1facf9fc 13706+
4a4d8108
AM
13707+ err = 0;
13708+ if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
0c3ec466 13709+ pr_warn("branch root dir was changed\n");
4a4d8108
AM
13710+ goto out;
13711+ }
1facf9fc 13712+
4a4d8108
AM
13713+ bfound = -1;
13714+ bend = au_ibend(inode);
13715+ bstart = au_ibstart(inode);
13716+#if 0 /* reserved for future use */
13717+ if (bindex == bend) {
13718+ /* keep this ino in rename case */
13719+ goto out;
13720+ }
13721+#endif
13722+ for (bindex = bstart; bindex <= bend; bindex++)
13723+ if (au_h_iptr(inode, bindex) == h_inode) {
13724+ bfound = bindex;
13725+ break;
13726+ }
13727+ if (bfound < 0)
1308ab2a 13728+ goto out;
1facf9fc 13729+
4a4d8108
AM
13730+ for (bindex = bstart; bindex <= bend; bindex++) {
13731+ h_i = au_h_iptr(inode, bindex);
13732+ if (!h_i)
13733+ continue;
1facf9fc 13734+
4a4d8108
AM
13735+ err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
13736+ /* ignore this error */
13737+ /* bad action? */
1facf9fc 13738+ }
1facf9fc 13739+
4a4d8108 13740+ /* children inode number will be broken */
1facf9fc 13741+
4f0767ce 13742+out:
4a4d8108
AM
13743+ AuTraceErr(err);
13744+ return err;
1facf9fc 13745+}
13746+
4a4d8108 13747+static int hn_gen_tree(struct dentry *dentry)
1facf9fc 13748+{
4a4d8108
AM
13749+ int err, i, j, ndentry;
13750+ struct au_dcsub_pages dpages;
13751+ struct au_dpage *dpage;
13752+ struct dentry **dentries;
1facf9fc 13753+
4a4d8108
AM
13754+ err = au_dpages_init(&dpages, GFP_NOFS);
13755+ if (unlikely(err))
13756+ goto out;
13757+ err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
13758+ if (unlikely(err))
13759+ goto out_dpages;
1facf9fc 13760+
4a4d8108
AM
13761+ for (i = 0; i < dpages.ndpage; i++) {
13762+ dpage = dpages.dpages + i;
13763+ dentries = dpage->dentries;
13764+ ndentry = dpage->ndentry;
13765+ for (j = 0; j < ndentry; j++) {
13766+ struct dentry *d;
13767+
13768+ d = dentries[j];
13769+ if (IS_ROOT(d))
13770+ continue;
13771+
4a4d8108
AM
13772+ au_digen_dec(d);
13773+ if (d->d_inode)
13774+ /* todo: reset children xino?
13775+ cached children only? */
13776+ au_iigen_dec(d->d_inode);
1308ab2a 13777+ }
dece6358 13778+ }
1facf9fc 13779+
4f0767ce 13780+out_dpages:
4a4d8108 13781+ au_dpages_free(&dpages);
dece6358 13782+
027c5e7a 13783+#if 0
4a4d8108
AM
13784+ /* discard children */
13785+ dentry_unhash(dentry);
13786+ dput(dentry);
027c5e7a 13787+#endif
4f0767ce 13788+out:
dece6358
AM
13789+ return err;
13790+}
13791+
1308ab2a 13792+/*
4a4d8108 13793+ * return 0 if processed.
1308ab2a 13794+ */
4a4d8108
AM
13795+static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
13796+ const unsigned int isdir)
dece6358 13797+{
1308ab2a 13798+ int err;
4a4d8108
AM
13799+ struct dentry *d;
13800+ struct qstr *dname;
1facf9fc 13801+
4a4d8108
AM
13802+ err = 1;
13803+ if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
0c3ec466 13804+ pr_warn("branch root dir was changed\n");
4a4d8108
AM
13805+ err = 0;
13806+ goto out;
13807+ }
dece6358 13808+
4a4d8108
AM
13809+ if (!isdir) {
13810+ AuDebugOn(!name);
13811+ au_iigen_dec(inode);
027c5e7a 13812+ spin_lock(&inode->i_lock);
c06a8ce3 13813+ hlist_for_each_entry(d, &inode->i_dentry, d_alias) {
027c5e7a 13814+ spin_lock(&d->d_lock);
4a4d8108
AM
13815+ dname = &d->d_name;
13816+ if (dname->len != nlen
027c5e7a
AM
13817+ && memcmp(dname->name, name, nlen)) {
13818+ spin_unlock(&d->d_lock);
4a4d8108 13819+ continue;
027c5e7a 13820+ }
4a4d8108 13821+ err = 0;
4a4d8108
AM
13822+ au_digen_dec(d);
13823+ spin_unlock(&d->d_lock);
13824+ break;
1facf9fc 13825+ }
027c5e7a 13826+ spin_unlock(&inode->i_lock);
1308ab2a 13827+ } else {
027c5e7a 13828+ au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
4a4d8108
AM
13829+ d = d_find_alias(inode);
13830+ if (!d) {
13831+ au_iigen_dec(inode);
13832+ goto out;
13833+ }
1facf9fc 13834+
027c5e7a 13835+ spin_lock(&d->d_lock);
4a4d8108 13836+ dname = &d->d_name;
027c5e7a
AM
13837+ if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
13838+ spin_unlock(&d->d_lock);
4a4d8108 13839+ err = hn_gen_tree(d);
027c5e7a
AM
13840+ spin_lock(&d->d_lock);
13841+ }
13842+ spin_unlock(&d->d_lock);
4a4d8108
AM
13843+ dput(d);
13844+ }
1facf9fc 13845+
4f0767ce 13846+out:
4a4d8108 13847+ AuTraceErr(err);
1308ab2a 13848+ return err;
13849+}
dece6358 13850+
4a4d8108 13851+static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
1facf9fc 13852+{
4a4d8108
AM
13853+ int err;
13854+ struct inode *inode;
1facf9fc 13855+
4a4d8108
AM
13856+ inode = dentry->d_inode;
13857+ if (IS_ROOT(dentry)
13858+ /* || (inode && inode->i_ino == AUFS_ROOT_INO) */
13859+ ) {
0c3ec466 13860+ pr_warn("branch root dir was changed\n");
4a4d8108
AM
13861+ return 0;
13862+ }
1308ab2a 13863+
4a4d8108
AM
13864+ err = 0;
13865+ if (!isdir) {
4a4d8108
AM
13866+ au_digen_dec(dentry);
13867+ if (inode)
13868+ au_iigen_dec(inode);
13869+ } else {
027c5e7a 13870+ au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
4a4d8108
AM
13871+ if (inode)
13872+ err = hn_gen_tree(dentry);
13873+ }
13874+
13875+ AuTraceErr(err);
13876+ return err;
1facf9fc 13877+}
13878+
4a4d8108 13879+/* ---------------------------------------------------------------------- */
1facf9fc 13880+
4a4d8108
AM
13881+/* hnotify job flags */
13882+#define AuHnJob_XINO0 1
13883+#define AuHnJob_GEN (1 << 1)
13884+#define AuHnJob_DIRENT (1 << 2)
13885+#define AuHnJob_ISDIR (1 << 3)
13886+#define AuHnJob_TRYXINO0 (1 << 4)
13887+#define AuHnJob_MNTPNT (1 << 5)
13888+#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
7f207e10
AM
13889+#define au_fset_hnjob(flags, name) \
13890+ do { (flags) |= AuHnJob_##name; } while (0)
13891+#define au_fclr_hnjob(flags, name) \
13892+ do { (flags) &= ~AuHnJob_##name; } while (0)
1facf9fc 13893+
4a4d8108
AM
13894+enum {
13895+ AuHn_CHILD,
13896+ AuHn_PARENT,
13897+ AuHnLast
13898+};
1facf9fc 13899+
4a4d8108
AM
13900+struct au_hnotify_args {
13901+ struct inode *h_dir, *dir, *h_child_inode;
13902+ u32 mask;
13903+ unsigned int flags[AuHnLast];
13904+ unsigned int h_child_nlen;
13905+ char h_child_name[];
13906+};
1facf9fc 13907+
4a4d8108
AM
13908+struct hn_job_args {
13909+ unsigned int flags;
13910+ struct inode *inode, *h_inode, *dir, *h_dir;
13911+ struct dentry *dentry;
13912+ char *h_name;
13913+ int h_nlen;
13914+};
1308ab2a 13915+
4a4d8108
AM
13916+static int hn_job(struct hn_job_args *a)
13917+{
13918+ const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
1308ab2a 13919+
4a4d8108
AM
13920+ /* reset xino */
13921+ if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
13922+ hn_xino(a->inode, a->h_inode); /* ignore this error */
1308ab2a 13923+
4a4d8108
AM
13924+ if (au_ftest_hnjob(a->flags, TRYXINO0)
13925+ && a->inode
13926+ && a->h_inode) {
13927+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
38d290e6
JR
13928+ if (!a->h_inode->i_nlink
13929+ && !(a->h_inode->i_state & I_LINKABLE))
4a4d8108
AM
13930+ hn_xino(a->inode, a->h_inode); /* ignore this error */
13931+ mutex_unlock(&a->h_inode->i_mutex);
1308ab2a 13932+ }
1facf9fc 13933+
4a4d8108
AM
13934+ /* make the generation obsolete */
13935+ if (au_ftest_hnjob(a->flags, GEN)) {
13936+ int err = -1;
13937+ if (a->inode)
13938+ err = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
13939+ isdir);
13940+ if (err && a->dentry)
13941+ hn_gen_by_name(a->dentry, isdir);
13942+ /* ignore this error */
1facf9fc 13943+ }
1facf9fc 13944+
4a4d8108
AM
13945+ /* make dir entries obsolete */
13946+ if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
13947+ struct au_vdir *vdir;
1facf9fc 13948+
4a4d8108
AM
13949+ vdir = au_ivdir(a->inode);
13950+ if (vdir)
13951+ vdir->vd_jiffy = 0;
13952+ /* IMustLock(a->inode); */
13953+ /* a->inode->i_version++; */
13954+ }
1facf9fc 13955+
4a4d8108
AM
13956+ /* can do nothing but warn */
13957+ if (au_ftest_hnjob(a->flags, MNTPNT)
13958+ && a->dentry
13959+ && d_mountpoint(a->dentry))
523b37e3 13960+ pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
1facf9fc 13961+
4a4d8108 13962+ return 0;
1308ab2a 13963+}
1facf9fc 13964+
1308ab2a 13965+/* ---------------------------------------------------------------------- */
1facf9fc 13966+
4a4d8108
AM
13967+static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
13968+ struct inode *dir)
1308ab2a 13969+{
4a4d8108
AM
13970+ struct dentry *dentry, *d, *parent;
13971+ struct qstr *dname;
1308ab2a 13972+
4a4d8108
AM
13973+ parent = d_find_alias(dir);
13974+ if (!parent)
13975+ return NULL;
1308ab2a 13976+
4a4d8108 13977+ dentry = NULL;
027c5e7a 13978+ spin_lock(&parent->d_lock);
4a4d8108 13979+ list_for_each_entry(d, &parent->d_subdirs, d_u.d_child) {
523b37e3 13980+ /* AuDbg("%pd\n", d); */
027c5e7a 13981+ spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
4a4d8108
AM
13982+ dname = &d->d_name;
13983+ if (dname->len != nlen || memcmp(dname->name, name, nlen))
027c5e7a
AM
13984+ goto cont_unlock;
13985+ if (au_di(d))
13986+ au_digen_dec(d);
13987+ else
13988+ goto cont_unlock;
392086de 13989+ if (d_count(d)) {
027c5e7a 13990+ dentry = dget_dlock(d);
4a4d8108 13991+ spin_unlock(&d->d_lock);
027c5e7a 13992+ break;
dece6358 13993+ }
1facf9fc 13994+
f6b6e03d 13995+cont_unlock:
027c5e7a 13996+ spin_unlock(&d->d_lock);
1308ab2a 13997+ }
027c5e7a 13998+ spin_unlock(&parent->d_lock);
4a4d8108 13999+ dput(parent);
1facf9fc 14000+
4a4d8108
AM
14001+ if (dentry)
14002+ di_write_lock_child(dentry);
1308ab2a 14003+
4a4d8108
AM
14004+ return dentry;
14005+}
dece6358 14006+
4a4d8108
AM
14007+static struct inode *lookup_wlock_by_ino(struct super_block *sb,
14008+ aufs_bindex_t bindex, ino_t h_ino)
14009+{
14010+ struct inode *inode;
14011+ ino_t ino;
14012+ int err;
14013+
14014+ inode = NULL;
14015+ err = au_xino_read(sb, bindex, h_ino, &ino);
14016+ if (!err && ino)
14017+ inode = ilookup(sb, ino);
14018+ if (!inode)
14019+ goto out;
14020+
14021+ if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
0c3ec466 14022+ pr_warn("wrong root branch\n");
4a4d8108
AM
14023+ iput(inode);
14024+ inode = NULL;
14025+ goto out;
1308ab2a 14026+ }
14027+
4a4d8108 14028+ ii_write_lock_child(inode);
1308ab2a 14029+
4f0767ce 14030+out:
4a4d8108 14031+ return inode;
dece6358
AM
14032+}
14033+
4a4d8108 14034+static void au_hn_bh(void *_args)
1facf9fc 14035+{
4a4d8108
AM
14036+ struct au_hnotify_args *a = _args;
14037+ struct super_block *sb;
14038+ aufs_bindex_t bindex, bend, bfound;
14039+ unsigned char xino, try_iput;
1facf9fc 14040+ int err;
1308ab2a 14041+ struct inode *inode;
4a4d8108
AM
14042+ ino_t h_ino;
14043+ struct hn_job_args args;
14044+ struct dentry *dentry;
14045+ struct au_sbinfo *sbinfo;
1facf9fc 14046+
4a4d8108
AM
14047+ AuDebugOn(!_args);
14048+ AuDebugOn(!a->h_dir);
14049+ AuDebugOn(!a->dir);
14050+ AuDebugOn(!a->mask);
14051+ AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
14052+ a->mask, a->dir->i_ino, a->h_dir->i_ino,
14053+ a->h_child_inode ? a->h_child_inode->i_ino : 0);
1facf9fc 14054+
4a4d8108
AM
14055+ inode = NULL;
14056+ dentry = NULL;
14057+ /*
14058+ * do not lock a->dir->i_mutex here
14059+ * because of d_revalidate() may cause a deadlock.
14060+ */
14061+ sb = a->dir->i_sb;
14062+ AuDebugOn(!sb);
14063+ sbinfo = au_sbi(sb);
14064+ AuDebugOn(!sbinfo);
7f207e10 14065+ si_write_lock(sb, AuLock_NOPLMW);
1facf9fc 14066+
4a4d8108
AM
14067+ ii_read_lock_parent(a->dir);
14068+ bfound = -1;
14069+ bend = au_ibend(a->dir);
14070+ for (bindex = au_ibstart(a->dir); bindex <= bend; bindex++)
14071+ if (au_h_iptr(a->dir, bindex) == a->h_dir) {
14072+ bfound = bindex;
14073+ break;
14074+ }
14075+ ii_read_unlock(a->dir);
14076+ if (unlikely(bfound < 0))
14077+ goto out;
1facf9fc 14078+
4a4d8108
AM
14079+ xino = !!au_opt_test(au_mntflags(sb), XINO);
14080+ h_ino = 0;
14081+ if (a->h_child_inode)
14082+ h_ino = a->h_child_inode->i_ino;
1facf9fc 14083+
4a4d8108
AM
14084+ if (a->h_child_nlen
14085+ && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
14086+ || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
14087+ dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
14088+ a->dir);
14089+ try_iput = 0;
14090+ if (dentry)
14091+ inode = dentry->d_inode;
14092+ if (xino && !inode && h_ino
14093+ && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
14094+ || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
14095+ || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
14096+ inode = lookup_wlock_by_ino(sb, bfound, h_ino);
14097+ try_iput = 1;
14098+ }
1facf9fc 14099+
4a4d8108
AM
14100+ args.flags = a->flags[AuHn_CHILD];
14101+ args.dentry = dentry;
14102+ args.inode = inode;
14103+ args.h_inode = a->h_child_inode;
14104+ args.dir = a->dir;
14105+ args.h_dir = a->h_dir;
14106+ args.h_name = a->h_child_name;
14107+ args.h_nlen = a->h_child_nlen;
14108+ err = hn_job(&args);
14109+ if (dentry) {
027c5e7a 14110+ if (au_di(dentry))
4a4d8108
AM
14111+ di_write_unlock(dentry);
14112+ dput(dentry);
14113+ }
14114+ if (inode && try_iput) {
14115+ ii_write_unlock(inode);
14116+ iput(inode);
14117+ }
1facf9fc 14118+
4a4d8108
AM
14119+ ii_write_lock_parent(a->dir);
14120+ args.flags = a->flags[AuHn_PARENT];
14121+ args.dentry = NULL;
14122+ args.inode = a->dir;
14123+ args.h_inode = a->h_dir;
14124+ args.dir = NULL;
14125+ args.h_dir = NULL;
14126+ args.h_name = NULL;
14127+ args.h_nlen = 0;
14128+ err = hn_job(&args);
14129+ ii_write_unlock(a->dir);
1facf9fc 14130+
4f0767ce 14131+out:
4a4d8108
AM
14132+ iput(a->h_child_inode);
14133+ iput(a->h_dir);
14134+ iput(a->dir);
027c5e7a
AM
14135+ si_write_unlock(sb);
14136+ au_nwt_done(&sbinfo->si_nowait);
1308ab2a 14137+ kfree(a);
dece6358 14138+}
1facf9fc 14139+
4a4d8108
AM
14140+/* ---------------------------------------------------------------------- */
14141+
14142+int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
14143+ struct qstr *h_child_qstr, struct inode *h_child_inode)
dece6358 14144+{
4a4d8108 14145+ int err, len;
53392da6 14146+ unsigned int flags[AuHnLast], f;
4a4d8108
AM
14147+ unsigned char isdir, isroot, wh;
14148+ struct inode *dir;
14149+ struct au_hnotify_args *args;
14150+ char *p, *h_child_name;
dece6358 14151+
1308ab2a 14152+ err = 0;
4a4d8108
AM
14153+ AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
14154+ dir = igrab(hnotify->hn_aufs_inode);
14155+ if (!dir)
14156+ goto out;
1facf9fc 14157+
4a4d8108
AM
14158+ isroot = (dir->i_ino == AUFS_ROOT_INO);
14159+ wh = 0;
14160+ h_child_name = (void *)h_child_qstr->name;
14161+ len = h_child_qstr->len;
14162+ if (h_child_name) {
14163+ if (len > AUFS_WH_PFX_LEN
14164+ && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
14165+ h_child_name += AUFS_WH_PFX_LEN;
14166+ len -= AUFS_WH_PFX_LEN;
14167+ wh = 1;
14168+ }
1facf9fc 14169+ }
dece6358 14170+
4a4d8108
AM
14171+ isdir = 0;
14172+ if (h_child_inode)
14173+ isdir = !!S_ISDIR(h_child_inode->i_mode);
14174+ flags[AuHn_PARENT] = AuHnJob_ISDIR;
14175+ flags[AuHn_CHILD] = 0;
14176+ if (isdir)
14177+ flags[AuHn_CHILD] = AuHnJob_ISDIR;
14178+ au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
14179+ au_fset_hnjob(flags[AuHn_CHILD], GEN);
14180+ switch (mask & FS_EVENTS_POSS_ON_CHILD) {
14181+ case FS_MOVED_FROM:
14182+ case FS_MOVED_TO:
14183+ au_fset_hnjob(flags[AuHn_CHILD], XINO0);
14184+ au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
14185+ /*FALLTHROUGH*/
14186+ case FS_CREATE:
fb47a38f 14187+ AuDebugOn(!h_child_name);
4a4d8108 14188+ break;
1facf9fc 14189+
4a4d8108
AM
14190+ case FS_DELETE:
14191+ /*
14192+ * aufs never be able to get this child inode.
14193+ * revalidation should be in d_revalidate()
14194+ * by checking i_nlink, i_generation or d_unhashed().
14195+ */
14196+ AuDebugOn(!h_child_name);
14197+ au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
14198+ au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
14199+ break;
dece6358 14200+
4a4d8108
AM
14201+ default:
14202+ AuDebugOn(1);
14203+ }
1308ab2a 14204+
4a4d8108
AM
14205+ if (wh)
14206+ h_child_inode = NULL;
1308ab2a 14207+
4a4d8108
AM
14208+ err = -ENOMEM;
14209+ /* iput() and kfree() will be called in au_hnotify() */
4a4d8108 14210+ args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
4a4d8108
AM
14211+ if (unlikely(!args)) {
14212+ AuErr1("no memory\n");
14213+ iput(dir);
14214+ goto out;
14215+ }
14216+ args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
14217+ args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
14218+ args->mask = mask;
14219+ args->dir = dir;
14220+ args->h_dir = igrab(h_dir);
14221+ if (h_child_inode)
14222+ h_child_inode = igrab(h_child_inode); /* can be NULL */
14223+ args->h_child_inode = h_child_inode;
14224+ args->h_child_nlen = len;
14225+ if (len) {
14226+ p = (void *)args;
14227+ p += sizeof(*args);
14228+ memcpy(p, h_child_name, len);
14229+ p[len] = 0;
1308ab2a 14230+ }
1308ab2a 14231+
38d290e6 14232+ /* NFS fires the event for silly-renamed one from kworker */
53392da6 14233+ f = 0;
38d290e6
JR
14234+ if (!dir->i_nlink
14235+ || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
53392da6
AM
14236+ f = AuWkq_NEST;
14237+ err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
4a4d8108
AM
14238+ if (unlikely(err)) {
14239+ pr_err("wkq %d\n", err);
14240+ iput(args->h_child_inode);
14241+ iput(args->h_dir);
14242+ iput(args->dir);
14243+ kfree(args);
1facf9fc 14244+ }
1facf9fc 14245+
4a4d8108 14246+out:
1facf9fc 14247+ return err;
14248+}
14249+
027c5e7a
AM
14250+/* ---------------------------------------------------------------------- */
14251+
14252+int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
14253+{
14254+ int err;
14255+
14256+ AuDebugOn(!(udba & AuOptMask_UDBA));
14257+
14258+ err = 0;
14259+ if (au_hnotify_op.reset_br)
14260+ err = au_hnotify_op.reset_br(udba, br, perm);
14261+
14262+ return err;
14263+}
14264+
14265+int au_hnotify_init_br(struct au_branch *br, int perm)
14266+{
14267+ int err;
14268+
14269+ err = 0;
14270+ if (au_hnotify_op.init_br)
14271+ err = au_hnotify_op.init_br(br, perm);
14272+
14273+ return err;
14274+}
14275+
14276+void au_hnotify_fin_br(struct au_branch *br)
14277+{
14278+ if (au_hnotify_op.fin_br)
14279+ au_hnotify_op.fin_br(br);
14280+}
14281+
4a4d8108
AM
14282+static void au_hn_destroy_cache(void)
14283+{
14284+ kmem_cache_destroy(au_cachep[AuCache_HNOTIFY]);
14285+ au_cachep[AuCache_HNOTIFY] = NULL;
14286+}
1308ab2a 14287+
4a4d8108 14288+int __init au_hnotify_init(void)
1facf9fc 14289+{
1308ab2a 14290+ int err;
1308ab2a 14291+
4a4d8108
AM
14292+ err = -ENOMEM;
14293+ au_cachep[AuCache_HNOTIFY] = AuCache(au_hnotify);
14294+ if (au_cachep[AuCache_HNOTIFY]) {
027c5e7a
AM
14295+ err = 0;
14296+ if (au_hnotify_op.init)
14297+ err = au_hnotify_op.init();
4a4d8108
AM
14298+ if (unlikely(err))
14299+ au_hn_destroy_cache();
1308ab2a 14300+ }
1308ab2a 14301+ AuTraceErr(err);
4a4d8108 14302+ return err;
1308ab2a 14303+}
14304+
4a4d8108 14305+void au_hnotify_fin(void)
1308ab2a 14306+{
027c5e7a
AM
14307+ if (au_hnotify_op.fin)
14308+ au_hnotify_op.fin();
4a4d8108
AM
14309+ /* cf. au_cache_fin() */
14310+ if (au_cachep[AuCache_HNOTIFY])
14311+ au_hn_destroy_cache();
dece6358 14312+}
7f207e10
AM
14313diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
14314--- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
14315+++ linux/fs/aufs/iinfo.c 2014-06-24 00:14:06.786052822 +0200
14316@@ -0,0 +1,277 @@
dece6358 14317+/*
523b37e3 14318+ * Copyright (C) 2005-2014 Junjiro R. Okajima
dece6358
AM
14319+ *
14320+ * This program, aufs is free software; you can redistribute it and/or modify
14321+ * it under the terms of the GNU General Public License as published by
14322+ * the Free Software Foundation; either version 2 of the License, or
14323+ * (at your option) any later version.
14324+ *
14325+ * This program is distributed in the hope that it will be useful,
14326+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14327+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14328+ * GNU General Public License for more details.
14329+ *
14330+ * You should have received a copy of the GNU General Public License
523b37e3 14331+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
dece6358 14332+ */
1facf9fc 14333+
dece6358 14334+/*
4a4d8108 14335+ * inode private data
dece6358 14336+ */
1facf9fc 14337+
1308ab2a 14338+#include "aufs.h"
1facf9fc 14339+
4a4d8108 14340+struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
1308ab2a 14341+{
4a4d8108 14342+ struct inode *h_inode;
1facf9fc 14343+
4a4d8108 14344+ IiMustAnyLock(inode);
1facf9fc 14345+
4a4d8108
AM
14346+ h_inode = au_ii(inode)->ii_hinode[0 + bindex].hi_inode;
14347+ AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
14348+ return h_inode;
14349+}
1facf9fc 14350+
4a4d8108
AM
14351+/* todo: hard/soft set? */
14352+void au_hiput(struct au_hinode *hinode)
14353+{
14354+ au_hn_free(hinode);
14355+ dput(hinode->hi_whdentry);
14356+ iput(hinode->hi_inode);
14357+}
1facf9fc 14358+
4a4d8108
AM
14359+unsigned int au_hi_flags(struct inode *inode, int isdir)
14360+{
14361+ unsigned int flags;
14362+ const unsigned int mnt_flags = au_mntflags(inode->i_sb);
1facf9fc 14363+
4a4d8108
AM
14364+ flags = 0;
14365+ if (au_opt_test(mnt_flags, XINO))
14366+ au_fset_hi(flags, XINO);
14367+ if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
14368+ au_fset_hi(flags, HNOTIFY);
14369+ return flags;
1facf9fc 14370+}
14371+
4a4d8108
AM
14372+void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
14373+ struct inode *h_inode, unsigned int flags)
1308ab2a 14374+{
4a4d8108
AM
14375+ struct au_hinode *hinode;
14376+ struct inode *hi;
14377+ struct au_iinfo *iinfo = au_ii(inode);
1facf9fc 14378+
4a4d8108 14379+ IiMustWriteLock(inode);
dece6358 14380+
4a4d8108
AM
14381+ hinode = iinfo->ii_hinode + bindex;
14382+ hi = hinode->hi_inode;
14383+ AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
14384+
14385+ if (hi)
14386+ au_hiput(hinode);
14387+ hinode->hi_inode = h_inode;
14388+ if (h_inode) {
14389+ int err;
14390+ struct super_block *sb = inode->i_sb;
14391+ struct au_branch *br;
14392+
027c5e7a
AM
14393+ AuDebugOn(inode->i_mode
14394+ && (h_inode->i_mode & S_IFMT)
14395+ != (inode->i_mode & S_IFMT));
4a4d8108
AM
14396+ if (bindex == iinfo->ii_bstart)
14397+ au_cpup_igen(inode, h_inode);
14398+ br = au_sbr(sb, bindex);
14399+ hinode->hi_id = br->br_id;
14400+ if (au_ftest_hi(flags, XINO)) {
14401+ err = au_xino_write(sb, bindex, h_inode->i_ino,
14402+ inode->i_ino);
14403+ if (unlikely(err))
14404+ AuIOErr1("failed au_xino_write() %d\n", err);
14405+ }
14406+
14407+ if (au_ftest_hi(flags, HNOTIFY)
14408+ && au_br_hnotifyable(br->br_perm)) {
027c5e7a 14409+ err = au_hn_alloc(hinode, inode);
4a4d8108
AM
14410+ if (unlikely(err))
14411+ AuIOErr1("au_hn_alloc() %d\n", err);
1308ab2a 14412+ }
14413+ }
4a4d8108 14414+}
dece6358 14415+
4a4d8108
AM
14416+void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
14417+ struct dentry *h_wh)
14418+{
14419+ struct au_hinode *hinode;
dece6358 14420+
4a4d8108
AM
14421+ IiMustWriteLock(inode);
14422+
14423+ hinode = au_ii(inode)->ii_hinode + bindex;
14424+ AuDebugOn(hinode->hi_whdentry);
14425+ hinode->hi_whdentry = h_wh;
1facf9fc 14426+}
14427+
537831f9 14428+void au_update_iigen(struct inode *inode, int half)
1308ab2a 14429+{
537831f9
AM
14430+ struct au_iinfo *iinfo;
14431+ struct au_iigen *iigen;
14432+ unsigned int sigen;
14433+
14434+ sigen = au_sigen(inode->i_sb);
14435+ iinfo = au_ii(inode);
14436+ iigen = &iinfo->ii_generation;
14437+ spin_lock(&iinfo->ii_genspin);
14438+ iigen->ig_generation = sigen;
14439+ if (half)
14440+ au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
14441+ else
14442+ au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
14443+ spin_unlock(&iinfo->ii_genspin);
4a4d8108 14444+}
1facf9fc 14445+
4a4d8108
AM
14446+/* it may be called at remount time, too */
14447+void au_update_ibrange(struct inode *inode, int do_put_zero)
14448+{
14449+ struct au_iinfo *iinfo;
027c5e7a 14450+ aufs_bindex_t bindex, bend;
1facf9fc 14451+
4a4d8108 14452+ iinfo = au_ii(inode);
027c5e7a 14453+ if (!iinfo)
4a4d8108 14454+ return;
1facf9fc 14455+
4a4d8108 14456+ IiMustWriteLock(inode);
1facf9fc 14457+
027c5e7a 14458+ if (do_put_zero && iinfo->ii_bstart >= 0) {
4a4d8108
AM
14459+ for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
14460+ bindex++) {
14461+ struct inode *h_i;
1facf9fc 14462+
4a4d8108 14463+ h_i = iinfo->ii_hinode[0 + bindex].hi_inode;
38d290e6
JR
14464+ if (h_i
14465+ && !h_i->i_nlink
14466+ && !(h_i->i_state & I_LINKABLE))
027c5e7a
AM
14467+ au_set_h_iptr(inode, bindex, NULL, 0);
14468+ }
4a4d8108
AM
14469+ }
14470+
027c5e7a
AM
14471+ iinfo->ii_bstart = -1;
14472+ iinfo->ii_bend = -1;
14473+ bend = au_sbend(inode->i_sb);
14474+ for (bindex = 0; bindex <= bend; bindex++)
14475+ if (iinfo->ii_hinode[0 + bindex].hi_inode) {
14476+ iinfo->ii_bstart = bindex;
4a4d8108 14477+ break;
027c5e7a
AM
14478+ }
14479+ if (iinfo->ii_bstart >= 0)
14480+ for (bindex = bend; bindex >= iinfo->ii_bstart; bindex--)
14481+ if (iinfo->ii_hinode[0 + bindex].hi_inode) {
14482+ iinfo->ii_bend = bindex;
14483+ break;
14484+ }
14485+ AuDebugOn(iinfo->ii_bstart > iinfo->ii_bend);
1308ab2a 14486+}
1facf9fc 14487+
dece6358 14488+/* ---------------------------------------------------------------------- */
1facf9fc 14489+
4a4d8108 14490+void au_icntnr_init_once(void *_c)
dece6358 14491+{
4a4d8108
AM
14492+ struct au_icntnr *c = _c;
14493+ struct au_iinfo *iinfo = &c->iinfo;
e49829fe 14494+ static struct lock_class_key aufs_ii;
1facf9fc 14495+
537831f9 14496+ spin_lock_init(&iinfo->ii_genspin);
4a4d8108 14497+ au_rw_init(&iinfo->ii_rwsem);
e49829fe 14498+ au_rw_class(&iinfo->ii_rwsem, &aufs_ii);
4a4d8108
AM
14499+ inode_init_once(&c->vfs_inode);
14500+}
1facf9fc 14501+
4a4d8108
AM
14502+int au_iinfo_init(struct inode *inode)
14503+{
14504+ struct au_iinfo *iinfo;
14505+ struct super_block *sb;
14506+ int nbr, i;
1facf9fc 14507+
4a4d8108
AM
14508+ sb = inode->i_sb;
14509+ iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
14510+ nbr = au_sbend(sb) + 1;
14511+ if (unlikely(nbr <= 0))
14512+ nbr = 1;
14513+ iinfo->ii_hinode = kcalloc(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
14514+ if (iinfo->ii_hinode) {
7f207e10 14515+ au_ninodes_inc(sb);
4a4d8108
AM
14516+ for (i = 0; i < nbr; i++)
14517+ iinfo->ii_hinode[i].hi_id = -1;
1facf9fc 14518+
537831f9 14519+ iinfo->ii_generation.ig_generation = au_sigen(sb);
4a4d8108
AM
14520+ iinfo->ii_bstart = -1;
14521+ iinfo->ii_bend = -1;
14522+ iinfo->ii_vdir = NULL;
14523+ return 0;
1308ab2a 14524+ }
4a4d8108
AM
14525+ return -ENOMEM;
14526+}
1facf9fc 14527+
4a4d8108
AM
14528+int au_ii_realloc(struct au_iinfo *iinfo, int nbr)
14529+{
14530+ int err, sz;
14531+ struct au_hinode *hip;
1facf9fc 14532+
4a4d8108
AM
14533+ AuRwMustWriteLock(&iinfo->ii_rwsem);
14534+
14535+ err = -ENOMEM;
14536+ sz = sizeof(*hip) * (iinfo->ii_bend + 1);
14537+ if (!sz)
14538+ sz = sizeof(*hip);
14539+ hip = au_kzrealloc(iinfo->ii_hinode, sz, sizeof(*hip) * nbr, GFP_NOFS);
14540+ if (hip) {
14541+ iinfo->ii_hinode = hip;
14542+ err = 0;
1308ab2a 14543+ }
4a4d8108 14544+
1308ab2a 14545+ return err;
1facf9fc 14546+}
14547+
4a4d8108 14548+void au_iinfo_fin(struct inode *inode)
1facf9fc 14549+{
4a4d8108
AM
14550+ struct au_iinfo *iinfo;
14551+ struct au_hinode *hi;
14552+ struct super_block *sb;
b752ccd1
AM
14553+ aufs_bindex_t bindex, bend;
14554+ const unsigned char unlinked = !inode->i_nlink;
1308ab2a 14555+
4a4d8108
AM
14556+ iinfo = au_ii(inode);
14557+ /* bad_inode case */
14558+ if (!iinfo)
14559+ return;
1308ab2a 14560+
b752ccd1 14561+ sb = inode->i_sb;
7f207e10 14562+ au_ninodes_dec(sb);
b752ccd1
AM
14563+ if (si_pid_test(sb))
14564+ au_xino_delete_inode(inode, unlinked);
14565+ else {
14566+ /*
14567+ * it is safe to hide the dependency between sbinfo and
14568+ * sb->s_umount.
14569+ */
14570+ lockdep_off();
14571+ si_noflush_read_lock(sb);
14572+ au_xino_delete_inode(inode, unlinked);
14573+ si_read_unlock(sb);
14574+ lockdep_on();
14575+ }
14576+
4a4d8108
AM
14577+ if (iinfo->ii_vdir)
14578+ au_vdir_free(iinfo->ii_vdir);
1308ab2a 14579+
b752ccd1
AM
14580+ bindex = iinfo->ii_bstart;
14581+ if (bindex >= 0) {
14582+ hi = iinfo->ii_hinode + bindex;
4a4d8108 14583+ bend = iinfo->ii_bend;
b752ccd1
AM
14584+ while (bindex++ <= bend) {
14585+ if (hi->hi_inode)
4a4d8108 14586+ au_hiput(hi);
4a4d8108
AM
14587+ hi++;
14588+ }
14589+ }
4a4d8108 14590+ kfree(iinfo->ii_hinode);
027c5e7a 14591+ iinfo->ii_hinode = NULL;
4a4d8108 14592+ AuRwDestroy(&iinfo->ii_rwsem);
dece6358 14593+}
7f207e10
AM
14594diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
14595--- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 14596+++ linux/fs/aufs/inode.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 14597@@ -0,0 +1,491 @@
4a4d8108 14598+/*
523b37e3 14599+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
14600+ *
14601+ * This program, aufs is free software; you can redistribute it and/or modify
14602+ * it under the terms of the GNU General Public License as published by
14603+ * the Free Software Foundation; either version 2 of the License, or
14604+ * (at your option) any later version.
14605+ *
14606+ * This program is distributed in the hope that it will be useful,
14607+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14608+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14609+ * GNU General Public License for more details.
14610+ *
14611+ * You should have received a copy of the GNU General Public License
523b37e3 14612+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 14613+ */
1facf9fc 14614+
4a4d8108
AM
14615+/*
14616+ * inode functions
14617+ */
1facf9fc 14618+
4a4d8108 14619+#include "aufs.h"
1308ab2a 14620+
4a4d8108
AM
14621+struct inode *au_igrab(struct inode *inode)
14622+{
14623+ if (inode) {
14624+ AuDebugOn(!atomic_read(&inode->i_count));
027c5e7a 14625+ ihold(inode);
1facf9fc 14626+ }
4a4d8108
AM
14627+ return inode;
14628+}
1facf9fc 14629+
4a4d8108
AM
14630+static void au_refresh_hinode_attr(struct inode *inode, int do_version)
14631+{
14632+ au_cpup_attr_all(inode, /*force*/0);
537831f9 14633+ au_update_iigen(inode, /*half*/1);
4a4d8108
AM
14634+ if (do_version)
14635+ inode->i_version++;
dece6358 14636+}
1facf9fc 14637+
027c5e7a 14638+static int au_ii_refresh(struct inode *inode, int *update)
dece6358 14639+{
4a4d8108 14640+ int err, e;
027c5e7a 14641+ umode_t type;
4a4d8108 14642+ aufs_bindex_t bindex, new_bindex;
1308ab2a 14643+ struct super_block *sb;
4a4d8108 14644+ struct au_iinfo *iinfo;
027c5e7a 14645+ struct au_hinode *p, *q, tmp;
1facf9fc 14646+
4a4d8108 14647+ IiMustWriteLock(inode);
1facf9fc 14648+
027c5e7a 14649+ *update = 0;
4a4d8108 14650+ sb = inode->i_sb;
027c5e7a 14651+ type = inode->i_mode & S_IFMT;
4a4d8108
AM
14652+ iinfo = au_ii(inode);
14653+ err = au_ii_realloc(iinfo, au_sbend(sb) + 1);
14654+ if (unlikely(err))
1308ab2a 14655+ goto out;
1facf9fc 14656+
027c5e7a 14657+ AuDebugOn(iinfo->ii_bstart < 0);
4a4d8108 14658+ p = iinfo->ii_hinode + iinfo->ii_bstart;
4a4d8108
AM
14659+ for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
14660+ bindex++, p++) {
14661+ if (!p->hi_inode)
14662+ continue;
1facf9fc 14663+
027c5e7a 14664+ AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
4a4d8108
AM
14665+ new_bindex = au_br_index(sb, p->hi_id);
14666+ if (new_bindex == bindex)
14667+ continue;
1facf9fc 14668+
4a4d8108 14669+ if (new_bindex < 0) {
027c5e7a 14670+ *update = 1;
4a4d8108
AM
14671+ au_hiput(p);
14672+ p->hi_inode = NULL;
14673+ continue;
1308ab2a 14674+ }
4a4d8108
AM
14675+
14676+ if (new_bindex < iinfo->ii_bstart)
14677+ iinfo->ii_bstart = new_bindex;
14678+ if (iinfo->ii_bend < new_bindex)
14679+ iinfo->ii_bend = new_bindex;
14680+ /* swap two lower inode, and loop again */
14681+ q = iinfo->ii_hinode + new_bindex;
14682+ tmp = *q;
14683+ *q = *p;
14684+ *p = tmp;
14685+ if (tmp.hi_inode) {
14686+ bindex--;
14687+ p--;
1308ab2a 14688+ }
14689+ }
4a4d8108
AM
14690+ au_update_ibrange(inode, /*do_put_zero*/0);
14691+ e = au_dy_irefresh(inode);
14692+ if (unlikely(e && !err))
14693+ err = e;
1facf9fc 14694+
4f0767ce 14695+out:
027c5e7a
AM
14696+ AuTraceErr(err);
14697+ return err;
14698+}
14699+
14700+int au_refresh_hinode_self(struct inode *inode)
14701+{
14702+ int err, update;
14703+
14704+ err = au_ii_refresh(inode, &update);
14705+ if (!err)
14706+ au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
14707+
14708+ AuTraceErr(err);
4a4d8108
AM
14709+ return err;
14710+}
1facf9fc 14711+
4a4d8108
AM
14712+int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
14713+{
027c5e7a 14714+ int err, e, update;
4a4d8108 14715+ unsigned int flags;
027c5e7a 14716+ umode_t mode;
4a4d8108 14717+ aufs_bindex_t bindex, bend;
027c5e7a 14718+ unsigned char isdir;
4a4d8108
AM
14719+ struct au_hinode *p;
14720+ struct au_iinfo *iinfo;
1facf9fc 14721+
027c5e7a 14722+ err = au_ii_refresh(inode, &update);
4a4d8108
AM
14723+ if (unlikely(err))
14724+ goto out;
14725+
14726+ update = 0;
14727+ iinfo = au_ii(inode);
14728+ p = iinfo->ii_hinode + iinfo->ii_bstart;
027c5e7a
AM
14729+ mode = (inode->i_mode & S_IFMT);
14730+ isdir = S_ISDIR(mode);
4a4d8108
AM
14731+ flags = au_hi_flags(inode, isdir);
14732+ bend = au_dbend(dentry);
14733+ for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
14734+ struct inode *h_i;
14735+ struct dentry *h_d;
14736+
14737+ h_d = au_h_dptr(dentry, bindex);
14738+ if (!h_d || !h_d->d_inode)
14739+ continue;
14740+
027c5e7a 14741+ AuDebugOn(mode != (h_d->d_inode->i_mode & S_IFMT));
4a4d8108
AM
14742+ if (iinfo->ii_bstart <= bindex && bindex <= iinfo->ii_bend) {
14743+ h_i = au_h_iptr(inode, bindex);
14744+ if (h_i) {
14745+ if (h_i == h_d->d_inode)
14746+ continue;
14747+ err = -EIO;
14748+ break;
14749+ }
14750+ }
14751+ if (bindex < iinfo->ii_bstart)
14752+ iinfo->ii_bstart = bindex;
14753+ if (iinfo->ii_bend < bindex)
14754+ iinfo->ii_bend = bindex;
14755+ au_set_h_iptr(inode, bindex, au_igrab(h_d->d_inode), flags);
14756+ update = 1;
1308ab2a 14757+ }
4a4d8108
AM
14758+ au_update_ibrange(inode, /*do_put_zero*/0);
14759+ e = au_dy_irefresh(inode);
14760+ if (unlikely(e && !err))
14761+ err = e;
027c5e7a
AM
14762+ if (!err)
14763+ au_refresh_hinode_attr(inode, update && isdir);
4a4d8108 14764+
4f0767ce 14765+out:
4a4d8108 14766+ AuTraceErr(err);
1308ab2a 14767+ return err;
dece6358
AM
14768+}
14769+
4a4d8108 14770+static int set_inode(struct inode *inode, struct dentry *dentry)
dece6358 14771+{
4a4d8108
AM
14772+ int err;
14773+ unsigned int flags;
14774+ umode_t mode;
14775+ aufs_bindex_t bindex, bstart, btail;
14776+ unsigned char isdir;
14777+ struct dentry *h_dentry;
14778+ struct inode *h_inode;
14779+ struct au_iinfo *iinfo;
dece6358 14780+
4a4d8108 14781+ IiMustWriteLock(inode);
dece6358 14782+
4a4d8108
AM
14783+ err = 0;
14784+ isdir = 0;
14785+ bstart = au_dbstart(dentry);
14786+ h_inode = au_h_dptr(dentry, bstart)->d_inode;
14787+ mode = h_inode->i_mode;
14788+ switch (mode & S_IFMT) {
14789+ case S_IFREG:
14790+ btail = au_dbtail(dentry);
14791+ inode->i_op = &aufs_iop;
14792+ inode->i_fop = &aufs_file_fop;
14793+ err = au_dy_iaop(inode, bstart, h_inode);
14794+ if (unlikely(err))
14795+ goto out;
14796+ break;
14797+ case S_IFDIR:
14798+ isdir = 1;
14799+ btail = au_dbtaildir(dentry);
14800+ inode->i_op = &aufs_dir_iop;
14801+ inode->i_fop = &aufs_dir_fop;
14802+ break;
14803+ case S_IFLNK:
14804+ btail = au_dbtail(dentry);
14805+ inode->i_op = &aufs_symlink_iop;
14806+ break;
14807+ case S_IFBLK:
14808+ case S_IFCHR:
14809+ case S_IFIFO:
14810+ case S_IFSOCK:
14811+ btail = au_dbtail(dentry);
14812+ inode->i_op = &aufs_iop;
38d290e6 14813+ init_special_inode(inode, mode, h_inode->i_rdev);
4a4d8108
AM
14814+ break;
14815+ default:
14816+ AuIOErr("Unknown file type 0%o\n", mode);
14817+ err = -EIO;
1308ab2a 14818+ goto out;
4a4d8108 14819+ }
dece6358 14820+
4a4d8108
AM
14821+ /* do not set hnotify for whiteouted dirs (SHWH mode) */
14822+ flags = au_hi_flags(inode, isdir);
14823+ if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
14824+ && au_ftest_hi(flags, HNOTIFY)
14825+ && dentry->d_name.len > AUFS_WH_PFX_LEN
14826+ && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
14827+ au_fclr_hi(flags, HNOTIFY);
14828+ iinfo = au_ii(inode);
14829+ iinfo->ii_bstart = bstart;
14830+ iinfo->ii_bend = btail;
14831+ for (bindex = bstart; bindex <= btail; bindex++) {
14832+ h_dentry = au_h_dptr(dentry, bindex);
14833+ if (h_dentry)
14834+ au_set_h_iptr(inode, bindex,
14835+ au_igrab(h_dentry->d_inode), flags);
14836+ }
14837+ au_cpup_attr_all(inode, /*force*/1);
dece6358 14838+
4f0767ce 14839+out:
4a4d8108
AM
14840+ return err;
14841+}
dece6358 14842+
027c5e7a
AM
14843+/*
14844+ * successful returns with iinfo write_locked
14845+ * minus: errno
14846+ * zero: success, matched
14847+ * plus: no error, but unmatched
14848+ */
14849+static int reval_inode(struct inode *inode, struct dentry *dentry)
4a4d8108
AM
14850+{
14851+ int err;
537831f9
AM
14852+ unsigned int gen;
14853+ struct au_iigen iigen;
4a4d8108
AM
14854+ aufs_bindex_t bindex, bend;
14855+ struct inode *h_inode, *h_dinode;
dece6358 14856+
4a4d8108
AM
14857+ /*
14858+ * before this function, if aufs got any iinfo lock, it must be only
14859+ * one, the parent dir.
14860+ * it can happen by UDBA and the obsoleted inode number.
14861+ */
14862+ err = -EIO;
14863+ if (unlikely(inode->i_ino == parent_ino(dentry)))
14864+ goto out;
14865+
027c5e7a 14866+ err = 1;
4a4d8108
AM
14867+ ii_write_lock_new_child(inode);
14868+ h_dinode = au_h_dptr(dentry, au_dbstart(dentry))->d_inode;
14869+ bend = au_ibend(inode);
14870+ for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
14871+ h_inode = au_h_iptr(inode, bindex);
537831f9
AM
14872+ if (!h_inode || h_inode != h_dinode)
14873+ continue;
14874+
14875+ err = 0;
14876+ gen = au_iigen(inode, &iigen);
14877+ if (gen == au_digen(dentry)
14878+ && !au_ig_ftest(iigen.ig_flags, HALF_REFRESHED))
4a4d8108 14879+ break;
537831f9
AM
14880+
14881+ /* fully refresh inode using dentry */
14882+ err = au_refresh_hinode(inode, dentry);
14883+ if (!err)
14884+ au_update_iigen(inode, /*half*/0);
14885+ break;
1facf9fc 14886+ }
dece6358 14887+
4a4d8108
AM
14888+ if (unlikely(err))
14889+ ii_write_unlock(inode);
4f0767ce 14890+out:
1facf9fc 14891+ return err;
14892+}
1facf9fc 14893+
4a4d8108
AM
14894+int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
14895+ unsigned int d_type, ino_t *ino)
1facf9fc 14896+{
4a4d8108
AM
14897+ int err;
14898+ struct mutex *mtx;
1facf9fc 14899+
b752ccd1 14900+ /* prevent hardlinked inode number from race condition */
4a4d8108 14901+ mtx = NULL;
b752ccd1 14902+ if (d_type != DT_DIR) {
4a4d8108
AM
14903+ mtx = &au_sbr(sb, bindex)->br_xino.xi_nondir_mtx;
14904+ mutex_lock(mtx);
14905+ }
14906+ err = au_xino_read(sb, bindex, h_ino, ino);
14907+ if (unlikely(err))
14908+ goto out;
1308ab2a 14909+
4a4d8108
AM
14910+ if (!*ino) {
14911+ err = -EIO;
14912+ *ino = au_xino_new_ino(sb);
14913+ if (unlikely(!*ino))
1facf9fc 14914+ goto out;
4a4d8108
AM
14915+ err = au_xino_write(sb, bindex, h_ino, *ino);
14916+ if (unlikely(err))
1308ab2a 14917+ goto out;
1308ab2a 14918+ }
1facf9fc 14919+
4f0767ce 14920+out:
b752ccd1 14921+ if (mtx)
4a4d8108 14922+ mutex_unlock(mtx);
1facf9fc 14923+ return err;
14924+}
14925+
4a4d8108
AM
14926+/* successful returns with iinfo write_locked */
14927+/* todo: return with unlocked? */
14928+struct inode *au_new_inode(struct dentry *dentry, int must_new)
1facf9fc 14929+{
b752ccd1 14930+ struct inode *inode, *h_inode;
4a4d8108
AM
14931+ struct dentry *h_dentry;
14932+ struct super_block *sb;
b752ccd1 14933+ struct mutex *mtx;
4a4d8108 14934+ ino_t h_ino, ino;
1716fcea 14935+ int err;
4a4d8108 14936+ aufs_bindex_t bstart;
1facf9fc 14937+
4a4d8108
AM
14938+ sb = dentry->d_sb;
14939+ bstart = au_dbstart(dentry);
14940+ h_dentry = au_h_dptr(dentry, bstart);
b752ccd1
AM
14941+ h_inode = h_dentry->d_inode;
14942+ h_ino = h_inode->i_ino;
14943+
14944+ /*
14945+ * stop 'race'-ing between hardlinks under different
14946+ * parents.
14947+ */
14948+ mtx = NULL;
14949+ if (!S_ISDIR(h_inode->i_mode))
14950+ mtx = &au_sbr(sb, bstart)->br_xino.xi_nondir_mtx;
14951+
4f0767ce 14952+new_ino:
b752ccd1
AM
14953+ if (mtx)
14954+ mutex_lock(mtx);
4a4d8108
AM
14955+ err = au_xino_read(sb, bstart, h_ino, &ino);
14956+ inode = ERR_PTR(err);
14957+ if (unlikely(err))
14958+ goto out;
b752ccd1 14959+
4a4d8108
AM
14960+ if (!ino) {
14961+ ino = au_xino_new_ino(sb);
14962+ if (unlikely(!ino)) {
14963+ inode = ERR_PTR(-EIO);
dece6358
AM
14964+ goto out;
14965+ }
14966+ }
1facf9fc 14967+
4a4d8108
AM
14968+ AuDbg("i%lu\n", (unsigned long)ino);
14969+ inode = au_iget_locked(sb, ino);
14970+ err = PTR_ERR(inode);
14971+ if (IS_ERR(inode))
1facf9fc 14972+ goto out;
1facf9fc 14973+
4a4d8108
AM
14974+ AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
14975+ if (inode->i_state & I_NEW) {
1716fcea
AM
14976+ /* verbose coding for lock class name */
14977+ if (unlikely(S_ISLNK(h_inode->i_mode)))
14978+ au_rw_class(&au_ii(inode)->ii_rwsem,
14979+ au_lc_key + AuLcSymlink_IIINFO);
14980+ else if (unlikely(S_ISDIR(h_inode->i_mode)))
14981+ au_rw_class(&au_ii(inode)->ii_rwsem,
14982+ au_lc_key + AuLcDir_IIINFO);
14983+ else /* likely */
14984+ au_rw_class(&au_ii(inode)->ii_rwsem,
14985+ au_lc_key + AuLcNonDir_IIINFO);
2dfbb274 14986+
4a4d8108
AM
14987+ ii_write_lock_new_child(inode);
14988+ err = set_inode(inode, dentry);
14989+ if (!err) {
14990+ unlock_new_inode(inode);
14991+ goto out; /* success */
14992+ }
1308ab2a 14993+
027c5e7a
AM
14994+ /*
14995+ * iget_failed() calls iput(), but we need to call
14996+ * ii_write_unlock() after iget_failed(). so dirty hack for
14997+ * i_count.
14998+ */
14999+ atomic_inc(&inode->i_count);
4a4d8108 15000+ iget_failed(inode);
027c5e7a
AM
15001+ ii_write_unlock(inode);
15002+ au_xino_write(sb, bstart, h_ino, /*ino*/0);
15003+ /* ignore this error */
15004+ goto out_iput;
15005+ } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
b752ccd1
AM
15006+ /*
15007+ * horrible race condition between lookup, readdir and copyup
15008+ * (or something).
15009+ */
15010+ if (mtx)
15011+ mutex_unlock(mtx);
027c5e7a
AM
15012+ err = reval_inode(inode, dentry);
15013+ if (unlikely(err < 0)) {
15014+ mtx = NULL;
15015+ goto out_iput;
15016+ }
15017+
b752ccd1
AM
15018+ if (!err) {
15019+ mtx = NULL;
4a4d8108 15020+ goto out; /* success */
b752ccd1
AM
15021+ } else if (mtx)
15022+ mutex_lock(mtx);
4a4d8108
AM
15023+ }
15024+
15025+ if (unlikely(au_test_fs_unique_ino(h_dentry->d_inode)))
15026+ AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
523b37e3
AM
15027+ " b%d, %s, %pd, hi%lu, i%lu.\n",
15028+ bstart, au_sbtype(h_dentry->d_sb), dentry,
4a4d8108
AM
15029+ (unsigned long)h_ino, (unsigned long)ino);
15030+ ino = 0;
15031+ err = au_xino_write(sb, bstart, h_ino, /*ino*/0);
15032+ if (!err) {
15033+ iput(inode);
b752ccd1
AM
15034+ if (mtx)
15035+ mutex_unlock(mtx);
4a4d8108
AM
15036+ goto new_ino;
15037+ }
1308ab2a 15038+
4f0767ce 15039+out_iput:
4a4d8108 15040+ iput(inode);
4a4d8108 15041+ inode = ERR_PTR(err);
4f0767ce 15042+out:
b752ccd1
AM
15043+ if (mtx)
15044+ mutex_unlock(mtx);
4a4d8108 15045+ return inode;
1facf9fc 15046+}
15047+
4a4d8108 15048+/* ---------------------------------------------------------------------- */
1facf9fc 15049+
4a4d8108
AM
15050+int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
15051+ struct inode *inode)
15052+{
15053+ int err;
1facf9fc 15054+
4a4d8108 15055+ err = au_br_rdonly(au_sbr(sb, bindex));
1facf9fc 15056+
4a4d8108
AM
15057+ /* pseudo-link after flushed may happen out of bounds */
15058+ if (!err
15059+ && inode
15060+ && au_ibstart(inode) <= bindex
15061+ && bindex <= au_ibend(inode)) {
15062+ /*
15063+ * permission check is unnecessary since vfsub routine
15064+ * will be called later
15065+ */
15066+ struct inode *hi = au_h_iptr(inode, bindex);
15067+ if (hi)
15068+ err = IS_IMMUTABLE(hi) ? -EROFS : 0;
1facf9fc 15069+ }
15070+
4a4d8108
AM
15071+ return err;
15072+}
dece6358 15073+
4a4d8108
AM
15074+int au_test_h_perm(struct inode *h_inode, int mask)
15075+{
2dfbb274 15076+ if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
4a4d8108
AM
15077+ return 0;
15078+ return inode_permission(h_inode, mask);
15079+}
1facf9fc 15080+
4a4d8108
AM
15081+int au_test_h_perm_sio(struct inode *h_inode, int mask)
15082+{
15083+ if (au_test_nfs(h_inode->i_sb)
15084+ && (mask & MAY_WRITE)
15085+ && S_ISDIR(h_inode->i_mode))
15086+ mask |= MAY_READ; /* force permission check */
15087+ return au_test_h_perm(h_inode, mask);
1facf9fc 15088+}
7f207e10
AM
15089diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
15090--- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
15091+++ linux/fs/aufs/inode.h 2014-06-24 00:14:06.786052822 +0200
15092@@ -0,0 +1,601 @@
4a4d8108 15093+/*
523b37e3 15094+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
15095+ *
15096+ * This program, aufs is free software; you can redistribute it and/or modify
15097+ * it under the terms of the GNU General Public License as published by
15098+ * the Free Software Foundation; either version 2 of the License, or
15099+ * (at your option) any later version.
15100+ *
15101+ * This program is distributed in the hope that it will be useful,
15102+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
15103+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15104+ * GNU General Public License for more details.
15105+ *
15106+ * You should have received a copy of the GNU General Public License
523b37e3 15107+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 15108+ */
1facf9fc 15109+
1308ab2a 15110+/*
4a4d8108 15111+ * inode operations
1308ab2a 15112+ */
dece6358 15113+
4a4d8108
AM
15114+#ifndef __AUFS_INODE_H__
15115+#define __AUFS_INODE_H__
dece6358 15116+
4a4d8108 15117+#ifdef __KERNEL__
1308ab2a 15118+
4a4d8108 15119+#include <linux/fsnotify.h>
4a4d8108 15120+#include "rwsem.h"
1308ab2a 15121+
4a4d8108 15122+struct vfsmount;
1facf9fc 15123+
4a4d8108
AM
15124+struct au_hnotify {
15125+#ifdef CONFIG_AUFS_HNOTIFY
15126+#ifdef CONFIG_AUFS_HFSNOTIFY
7f207e10 15127+ /* never use fsnotify_add_vfsmount_mark() */
0c5527e5 15128+ struct fsnotify_mark hn_mark;
4a4d8108 15129+#endif
7f207e10 15130+ struct inode *hn_aufs_inode; /* no get/put */
4a4d8108
AM
15131+#endif
15132+} ____cacheline_aligned_in_smp;
1facf9fc 15133+
4a4d8108
AM
15134+struct au_hinode {
15135+ struct inode *hi_inode;
15136+ aufs_bindex_t hi_id;
15137+#ifdef CONFIG_AUFS_HNOTIFY
15138+ struct au_hnotify *hi_notify;
15139+#endif
dece6358 15140+
4a4d8108
AM
15141+ /* reference to the copied-up whiteout with get/put */
15142+ struct dentry *hi_whdentry;
15143+};
dece6358 15144+
537831f9
AM
15145+/* ig_flags */
15146+#define AuIG_HALF_REFRESHED 1
15147+#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
15148+#define au_ig_fset(flags, name) \
15149+ do { (flags) |= AuIG_##name; } while (0)
15150+#define au_ig_fclr(flags, name) \
15151+ do { (flags) &= ~AuIG_##name; } while (0)
15152+
15153+struct au_iigen {
15154+ __u32 ig_generation, ig_flags;
15155+};
15156+
4a4d8108
AM
15157+struct au_vdir;
15158+struct au_iinfo {
537831f9 15159+ spinlock_t ii_genspin;
7a9e40b8 15160+ struct au_iigen ii_generation;
4a4d8108 15161+ struct super_block *ii_hsb1; /* no get/put */
1facf9fc 15162+
4a4d8108
AM
15163+ struct au_rwsem ii_rwsem;
15164+ aufs_bindex_t ii_bstart, ii_bend;
15165+ __u32 ii_higen;
15166+ struct au_hinode *ii_hinode;
15167+ struct au_vdir *ii_vdir;
15168+};
1facf9fc 15169+
4a4d8108
AM
15170+struct au_icntnr {
15171+ struct au_iinfo iinfo;
15172+ struct inode vfs_inode;
15173+} ____cacheline_aligned_in_smp;
1308ab2a 15174+
4a4d8108
AM
15175+/* au_pin flags */
15176+#define AuPin_DI_LOCKED 1
15177+#define AuPin_MNT_WRITE (1 << 1)
15178+#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
7f207e10
AM
15179+#define au_fset_pin(flags, name) \
15180+ do { (flags) |= AuPin_##name; } while (0)
15181+#define au_fclr_pin(flags, name) \
15182+ do { (flags) &= ~AuPin_##name; } while (0)
4a4d8108
AM
15183+
15184+struct au_pin {
15185+ /* input */
15186+ struct dentry *dentry;
15187+ unsigned int udba;
15188+ unsigned char lsc_di, lsc_hi, flags;
15189+ aufs_bindex_t bindex;
15190+
15191+ /* output */
15192+ struct dentry *parent;
15193+ struct au_hinode *hdir;
15194+ struct vfsmount *h_mnt;
86dc4139
AM
15195+
15196+ /* temporary unlock/relock for copyup */
15197+ struct dentry *h_dentry, *h_parent;
15198+ struct au_branch *br;
15199+ struct task_struct *task;
4a4d8108 15200+};
1facf9fc 15201+
86dc4139
AM
15202+void au_pin_hdir_unlock(struct au_pin *p);
15203+int au_pin_hdir_relock(struct au_pin *p);
15204+void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task);
15205+void au_pin_hdir_acquire_nest(struct au_pin *p);
15206+void au_pin_hdir_release(struct au_pin *p);
15207+
1308ab2a 15208+/* ---------------------------------------------------------------------- */
15209+
4a4d8108 15210+static inline struct au_iinfo *au_ii(struct inode *inode)
1facf9fc 15211+{
4a4d8108 15212+ struct au_iinfo *iinfo;
1facf9fc 15213+
4a4d8108
AM
15214+ iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
15215+ if (iinfo->ii_hinode)
15216+ return iinfo;
15217+ return NULL; /* debugging bad_inode case */
15218+}
1facf9fc 15219+
4a4d8108 15220+/* ---------------------------------------------------------------------- */
1facf9fc 15221+
4a4d8108
AM
15222+/* inode.c */
15223+struct inode *au_igrab(struct inode *inode);
027c5e7a 15224+int au_refresh_hinode_self(struct inode *inode);
4a4d8108
AM
15225+int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
15226+int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
15227+ unsigned int d_type, ino_t *ino);
15228+struct inode *au_new_inode(struct dentry *dentry, int must_new);
15229+int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
15230+ struct inode *inode);
15231+int au_test_h_perm(struct inode *h_inode, int mask);
15232+int au_test_h_perm_sio(struct inode *h_inode, int mask);
1facf9fc 15233+
4a4d8108
AM
15234+static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
15235+ ino_t h_ino, unsigned int d_type, ino_t *ino)
15236+{
15237+#ifdef CONFIG_AUFS_SHWH
15238+ return au_ino(sb, bindex, h_ino, d_type, ino);
15239+#else
15240+ return 0;
15241+#endif
15242+}
1facf9fc 15243+
4a4d8108
AM
15244+/* i_op.c */
15245+extern struct inode_operations aufs_iop, aufs_symlink_iop, aufs_dir_iop;
1308ab2a 15246+
4a4d8108
AM
15247+/* au_wr_dir flags */
15248+#define AuWrDir_ADD_ENTRY 1
86dc4139
AM
15249+#define AuWrDir_TMP_WHENTRY (1 << 1)
15250+#define AuWrDir_ISDIR (1 << 2)
38d290e6 15251+#define AuWrDir_TMPFILE (1 << 3)
4a4d8108 15252+#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
7f207e10
AM
15253+#define au_fset_wrdir(flags, name) \
15254+ do { (flags) |= AuWrDir_##name; } while (0)
15255+#define au_fclr_wrdir(flags, name) \
15256+ do { (flags) &= ~AuWrDir_##name; } while (0)
1facf9fc 15257+
4a4d8108
AM
15258+struct au_wr_dir_args {
15259+ aufs_bindex_t force_btgt;
15260+ unsigned char flags;
15261+};
15262+int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
15263+ struct au_wr_dir_args *args);
dece6358 15264+
4a4d8108
AM
15265+struct dentry *au_pinned_h_parent(struct au_pin *pin);
15266+void au_pin_init(struct au_pin *pin, struct dentry *dentry,
15267+ aufs_bindex_t bindex, int lsc_di, int lsc_hi,
15268+ unsigned int udba, unsigned char flags);
15269+int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
15270+ unsigned int udba, unsigned char flags) __must_check;
15271+int au_do_pin(struct au_pin *pin) __must_check;
15272+void au_unpin(struct au_pin *pin);
1facf9fc 15273+
4a4d8108
AM
15274+/* i_op_add.c */
15275+int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
15276+ struct dentry *h_parent, int isdir);
7eafdf33
AM
15277+int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
15278+ dev_t dev);
4a4d8108 15279+int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname);
7eafdf33 15280+int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
b4510431 15281+ bool want_excl);
38d290e6 15282+int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode);
4a4d8108
AM
15283+int aufs_link(struct dentry *src_dentry, struct inode *dir,
15284+ struct dentry *dentry);
7eafdf33 15285+int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
1facf9fc 15286+
4a4d8108
AM
15287+/* i_op_del.c */
15288+int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
15289+int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
15290+ struct dentry *h_parent, int isdir);
15291+int aufs_unlink(struct inode *dir, struct dentry *dentry);
15292+int aufs_rmdir(struct inode *dir, struct dentry *dentry);
1308ab2a 15293+
4a4d8108
AM
15294+/* i_op_ren.c */
15295+int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
15296+int aufs_rename(struct inode *src_dir, struct dentry *src_dentry,
15297+ struct inode *dir, struct dentry *dentry);
1facf9fc 15298+
4a4d8108
AM
15299+/* iinfo.c */
15300+struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
15301+void au_hiput(struct au_hinode *hinode);
15302+void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
15303+ struct dentry *h_wh);
15304+unsigned int au_hi_flags(struct inode *inode, int isdir);
1308ab2a 15305+
4a4d8108
AM
15306+/* hinode flags */
15307+#define AuHi_XINO 1
15308+#define AuHi_HNOTIFY (1 << 1)
15309+#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
7f207e10
AM
15310+#define au_fset_hi(flags, name) \
15311+ do { (flags) |= AuHi_##name; } while (0)
15312+#define au_fclr_hi(flags, name) \
15313+ do { (flags) &= ~AuHi_##name; } while (0)
1facf9fc 15314+
4a4d8108
AM
15315+#ifndef CONFIG_AUFS_HNOTIFY
15316+#undef AuHi_HNOTIFY
15317+#define AuHi_HNOTIFY 0
15318+#endif
1facf9fc 15319+
4a4d8108
AM
15320+void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
15321+ struct inode *h_inode, unsigned int flags);
1facf9fc 15322+
537831f9 15323+void au_update_iigen(struct inode *inode, int half);
4a4d8108 15324+void au_update_ibrange(struct inode *inode, int do_put_zero);
1facf9fc 15325+
4a4d8108
AM
15326+void au_icntnr_init_once(void *_c);
15327+int au_iinfo_init(struct inode *inode);
15328+void au_iinfo_fin(struct inode *inode);
15329+int au_ii_realloc(struct au_iinfo *iinfo, int nbr);
1308ab2a 15330+
e49829fe 15331+#ifdef CONFIG_PROC_FS
4a4d8108 15332+/* plink.c */
e49829fe
JR
15333+int au_plink_maint(struct super_block *sb, int flags);
15334+void au_plink_maint_leave(struct au_sbinfo *sbinfo);
15335+int au_plink_maint_enter(struct super_block *sb);
4a4d8108
AM
15336+#ifdef CONFIG_AUFS_DEBUG
15337+void au_plink_list(struct super_block *sb);
15338+#else
15339+AuStubVoid(au_plink_list, struct super_block *sb)
15340+#endif
15341+int au_plink_test(struct inode *inode);
15342+struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
15343+void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
15344+ struct dentry *h_dentry);
e49829fe
JR
15345+void au_plink_put(struct super_block *sb, int verbose);
15346+void au_plink_clean(struct super_block *sb, int verbose);
4a4d8108 15347+void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
e49829fe
JR
15348+#else
15349+AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
15350+AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
15351+AuStubInt0(au_plink_maint_enter, struct super_block *sb);
15352+AuStubVoid(au_plink_list, struct super_block *sb);
15353+AuStubInt0(au_plink_test, struct inode *inode);
15354+AuStub(struct dentry *, au_plink_lkup, return NULL,
15355+ struct inode *inode, aufs_bindex_t bindex);
15356+AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
15357+ struct dentry *h_dentry);
15358+AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
15359+AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
15360+AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
15361+#endif /* CONFIG_PROC_FS */
1facf9fc 15362+
4a4d8108 15363+/* ---------------------------------------------------------------------- */
1308ab2a 15364+
4a4d8108
AM
15365+/* lock subclass for iinfo */
15366+enum {
15367+ AuLsc_II_CHILD, /* child first */
15368+ AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
15369+ AuLsc_II_CHILD3, /* copyup dirs */
15370+ AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
15371+ AuLsc_II_PARENT2,
15372+ AuLsc_II_PARENT3, /* copyup dirs */
15373+ AuLsc_II_NEW_CHILD
15374+};
1308ab2a 15375+
1facf9fc 15376+/*
4a4d8108
AM
15377+ * ii_read_lock_child, ii_write_lock_child,
15378+ * ii_read_lock_child2, ii_write_lock_child2,
15379+ * ii_read_lock_child3, ii_write_lock_child3,
15380+ * ii_read_lock_parent, ii_write_lock_parent,
15381+ * ii_read_lock_parent2, ii_write_lock_parent2,
15382+ * ii_read_lock_parent3, ii_write_lock_parent3,
15383+ * ii_read_lock_new_child, ii_write_lock_new_child,
1facf9fc 15384+ */
4a4d8108
AM
15385+#define AuReadLockFunc(name, lsc) \
15386+static inline void ii_read_lock_##name(struct inode *i) \
15387+{ \
15388+ au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
15389+}
15390+
15391+#define AuWriteLockFunc(name, lsc) \
15392+static inline void ii_write_lock_##name(struct inode *i) \
15393+{ \
15394+ au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
15395+}
15396+
15397+#define AuRWLockFuncs(name, lsc) \
15398+ AuReadLockFunc(name, lsc) \
15399+ AuWriteLockFunc(name, lsc)
15400+
15401+AuRWLockFuncs(child, CHILD);
15402+AuRWLockFuncs(child2, CHILD2);
15403+AuRWLockFuncs(child3, CHILD3);
15404+AuRWLockFuncs(parent, PARENT);
15405+AuRWLockFuncs(parent2, PARENT2);
15406+AuRWLockFuncs(parent3, PARENT3);
15407+AuRWLockFuncs(new_child, NEW_CHILD);
15408+
15409+#undef AuReadLockFunc
15410+#undef AuWriteLockFunc
15411+#undef AuRWLockFuncs
1facf9fc 15412+
15413+/*
4a4d8108 15414+ * ii_read_unlock, ii_write_unlock, ii_downgrade_lock
1facf9fc 15415+ */
4a4d8108 15416+AuSimpleUnlockRwsemFuncs(ii, struct inode *i, &au_ii(i)->ii_rwsem);
1facf9fc 15417+
4a4d8108
AM
15418+#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
15419+#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
15420+#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
1facf9fc 15421+
4a4d8108 15422+/* ---------------------------------------------------------------------- */
1308ab2a 15423+
027c5e7a
AM
15424+static inline void au_icntnr_init(struct au_icntnr *c)
15425+{
15426+#ifdef CONFIG_AUFS_DEBUG
15427+ c->vfs_inode.i_mode = 0;
15428+#endif
15429+}
15430+
537831f9 15431+static inline unsigned int au_iigen(struct inode *inode, struct au_iigen *iigen)
4a4d8108 15432+{
537831f9
AM
15433+ unsigned int gen;
15434+ struct au_iinfo *iinfo;
15435+
15436+ iinfo = au_ii(inode);
15437+ spin_lock(&iinfo->ii_genspin);
15438+ if (iigen)
15439+ *iigen = iinfo->ii_generation;
15440+ gen = iinfo->ii_generation.ig_generation;
15441+ spin_unlock(&iinfo->ii_genspin);
15442+
15443+ return gen;
4a4d8108 15444+}
1308ab2a 15445+
4a4d8108
AM
15446+/* tiny test for inode number */
15447+/* tmpfs generation is too rough */
15448+static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
15449+{
15450+ struct au_iinfo *iinfo;
1308ab2a 15451+
4a4d8108
AM
15452+ iinfo = au_ii(inode);
15453+ AuRwMustAnyLock(&iinfo->ii_rwsem);
15454+ return !(iinfo->ii_hsb1 == h_inode->i_sb
15455+ && iinfo->ii_higen == h_inode->i_generation);
15456+}
1308ab2a 15457+
4a4d8108
AM
15458+static inline void au_iigen_dec(struct inode *inode)
15459+{
537831f9
AM
15460+ struct au_iinfo *iinfo;
15461+
15462+ iinfo = au_ii(inode);
15463+ spin_lock(&iinfo->ii_genspin);
15464+ iinfo->ii_generation.ig_generation--;
15465+ spin_unlock(&iinfo->ii_genspin);
027c5e7a
AM
15466+}
15467+
15468+static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
15469+{
15470+ int err;
15471+
15472+ err = 0;
537831f9 15473+ if (unlikely(inode && au_iigen(inode, NULL) != sigen))
027c5e7a
AM
15474+ err = -EIO;
15475+
15476+ return err;
4a4d8108 15477+}
1308ab2a 15478+
4a4d8108 15479+/* ---------------------------------------------------------------------- */
1308ab2a 15480+
4a4d8108
AM
15481+static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
15482+ aufs_bindex_t bindex)
15483+{
15484+ IiMustAnyLock(inode);
15485+ return au_ii(inode)->ii_hinode[0 + bindex].hi_id;
15486+}
1308ab2a 15487+
4a4d8108
AM
15488+static inline aufs_bindex_t au_ibstart(struct inode *inode)
15489+{
15490+ IiMustAnyLock(inode);
15491+ return au_ii(inode)->ii_bstart;
15492+}
1308ab2a 15493+
4a4d8108
AM
15494+static inline aufs_bindex_t au_ibend(struct inode *inode)
15495+{
15496+ IiMustAnyLock(inode);
15497+ return au_ii(inode)->ii_bend;
15498+}
1308ab2a 15499+
4a4d8108
AM
15500+static inline struct au_vdir *au_ivdir(struct inode *inode)
15501+{
15502+ IiMustAnyLock(inode);
15503+ return au_ii(inode)->ii_vdir;
15504+}
1308ab2a 15505+
4a4d8108
AM
15506+static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
15507+{
15508+ IiMustAnyLock(inode);
15509+ return au_ii(inode)->ii_hinode[0 + bindex].hi_whdentry;
15510+}
1308ab2a 15511+
4a4d8108 15512+static inline void au_set_ibstart(struct inode *inode, aufs_bindex_t bindex)
1308ab2a 15513+{
4a4d8108
AM
15514+ IiMustWriteLock(inode);
15515+ au_ii(inode)->ii_bstart = bindex;
15516+}
1308ab2a 15517+
4a4d8108
AM
15518+static inline void au_set_ibend(struct inode *inode, aufs_bindex_t bindex)
15519+{
15520+ IiMustWriteLock(inode);
15521+ au_ii(inode)->ii_bend = bindex;
1308ab2a 15522+}
15523+
4a4d8108
AM
15524+static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
15525+{
15526+ IiMustWriteLock(inode);
15527+ au_ii(inode)->ii_vdir = vdir;
15528+}
1facf9fc 15529+
4a4d8108 15530+static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
1308ab2a 15531+{
4a4d8108
AM
15532+ IiMustAnyLock(inode);
15533+ return au_ii(inode)->ii_hinode + bindex;
15534+}
dece6358 15535+
4a4d8108 15536+/* ---------------------------------------------------------------------- */
1facf9fc 15537+
4a4d8108
AM
15538+static inline struct dentry *au_pinned_parent(struct au_pin *pin)
15539+{
15540+ if (pin)
15541+ return pin->parent;
15542+ return NULL;
1facf9fc 15543+}
15544+
4a4d8108 15545+static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
1facf9fc 15546+{
4a4d8108
AM
15547+ if (pin && pin->hdir)
15548+ return pin->hdir->hi_inode;
15549+ return NULL;
1308ab2a 15550+}
1facf9fc 15551+
4a4d8108
AM
15552+static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
15553+{
15554+ if (pin)
15555+ return pin->hdir;
15556+ return NULL;
15557+}
1facf9fc 15558+
4a4d8108 15559+static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
1308ab2a 15560+{
4a4d8108
AM
15561+ if (pin)
15562+ pin->dentry = dentry;
15563+}
1308ab2a 15564+
4a4d8108
AM
15565+static inline void au_pin_set_parent_lflag(struct au_pin *pin,
15566+ unsigned char lflag)
15567+{
15568+ if (pin) {
7f207e10 15569+ if (lflag)
4a4d8108 15570+ au_fset_pin(pin->flags, DI_LOCKED);
7f207e10 15571+ else
4a4d8108 15572+ au_fclr_pin(pin->flags, DI_LOCKED);
1308ab2a 15573+ }
4a4d8108
AM
15574+}
15575+
15576+static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
15577+{
15578+ if (pin) {
15579+ dput(pin->parent);
15580+ pin->parent = dget(parent);
1facf9fc 15581+ }
4a4d8108 15582+}
1facf9fc 15583+
4a4d8108
AM
15584+/* ---------------------------------------------------------------------- */
15585+
027c5e7a 15586+struct au_branch;
4a4d8108
AM
15587+#ifdef CONFIG_AUFS_HNOTIFY
15588+struct au_hnotify_op {
15589+ void (*ctl)(struct au_hinode *hinode, int do_set);
027c5e7a 15590+ int (*alloc)(struct au_hinode *hinode);
7eafdf33
AM
15591+
15592+ /*
15593+ * if it returns true, the the caller should free hinode->hi_notify,
15594+ * otherwise ->free() frees it.
15595+ */
15596+ int (*free)(struct au_hinode *hinode,
15597+ struct au_hnotify *hn) __must_check;
4a4d8108
AM
15598+
15599+ void (*fin)(void);
15600+ int (*init)(void);
027c5e7a
AM
15601+
15602+ int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
15603+ void (*fin_br)(struct au_branch *br);
15604+ int (*init_br)(struct au_branch *br, int perm);
4a4d8108
AM
15605+};
15606+
15607+/* hnotify.c */
027c5e7a 15608+int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
4a4d8108
AM
15609+void au_hn_free(struct au_hinode *hinode);
15610+void au_hn_ctl(struct au_hinode *hinode, int do_set);
15611+void au_hn_reset(struct inode *inode, unsigned int flags);
15612+int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
15613+ struct qstr *h_child_qstr, struct inode *h_child_inode);
027c5e7a
AM
15614+int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
15615+int au_hnotify_init_br(struct au_branch *br, int perm);
15616+void au_hnotify_fin_br(struct au_branch *br);
4a4d8108
AM
15617+int __init au_hnotify_init(void);
15618+void au_hnotify_fin(void);
15619+
7f207e10 15620+/* hfsnotify.c */
4a4d8108
AM
15621+extern const struct au_hnotify_op au_hnotify_op;
15622+
15623+static inline
15624+void au_hn_init(struct au_hinode *hinode)
15625+{
15626+ hinode->hi_notify = NULL;
1308ab2a 15627+}
15628+
53392da6
AM
15629+static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
15630+{
15631+ return hinode->hi_notify;
15632+}
15633+
4a4d8108
AM
15634+#else
15635+static inline
15636+int au_hn_alloc(struct au_hinode *hinode __maybe_unused,
027c5e7a 15637+ struct inode *inode __maybe_unused)
1308ab2a 15638+{
4a4d8108
AM
15639+ return -EOPNOTSUPP;
15640+}
1308ab2a 15641+
53392da6
AM
15642+static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
15643+{
15644+ return NULL;
15645+}
15646+
4a4d8108
AM
15647+AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
15648+AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
15649+ int do_set __maybe_unused)
15650+AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
15651+ unsigned int flags __maybe_unused)
027c5e7a
AM
15652+AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
15653+ struct au_branch *br __maybe_unused,
15654+ int perm __maybe_unused)
15655+AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
15656+ int perm __maybe_unused)
15657+AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
4a4d8108
AM
15658+AuStubInt0(__init au_hnotify_init, void)
15659+AuStubVoid(au_hnotify_fin, void)
15660+AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
15661+#endif /* CONFIG_AUFS_HNOTIFY */
15662+
15663+static inline void au_hn_suspend(struct au_hinode *hdir)
15664+{
15665+ au_hn_ctl(hdir, /*do_set*/0);
1308ab2a 15666+}
15667+
4a4d8108 15668+static inline void au_hn_resume(struct au_hinode *hdir)
1308ab2a 15669+{
4a4d8108
AM
15670+ au_hn_ctl(hdir, /*do_set*/1);
15671+}
1308ab2a 15672+
4a4d8108
AM
15673+static inline void au_hn_imtx_lock(struct au_hinode *hdir)
15674+{
15675+ mutex_lock(&hdir->hi_inode->i_mutex);
15676+ au_hn_suspend(hdir);
15677+}
dece6358 15678+
4a4d8108
AM
15679+static inline void au_hn_imtx_lock_nested(struct au_hinode *hdir,
15680+ unsigned int sc __maybe_unused)
15681+{
15682+ mutex_lock_nested(&hdir->hi_inode->i_mutex, sc);
15683+ au_hn_suspend(hdir);
1facf9fc 15684+}
1facf9fc 15685+
4a4d8108
AM
15686+static inline void au_hn_imtx_unlock(struct au_hinode *hdir)
15687+{
15688+ au_hn_resume(hdir);
15689+ mutex_unlock(&hdir->hi_inode->i_mutex);
15690+}
15691+
15692+#endif /* __KERNEL__ */
15693+#endif /* __AUFS_INODE_H__ */
7f207e10
AM
15694diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
15695--- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 15696+++ linux/fs/aufs/ioctl.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 15697@@ -0,0 +1,201 @@
4a4d8108 15698+/*
523b37e3 15699+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
15700+ *
15701+ * This program, aufs is free software; you can redistribute it and/or modify
15702+ * it under the terms of the GNU General Public License as published by
15703+ * the Free Software Foundation; either version 2 of the License, or
15704+ * (at your option) any later version.
15705+ *
15706+ * This program is distributed in the hope that it will be useful,
15707+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
15708+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15709+ * GNU General Public License for more details.
15710+ *
15711+ * You should have received a copy of the GNU General Public License
523b37e3 15712+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
15713+ */
15714+
15715+/*
15716+ * ioctl
15717+ * plink-management and readdir in userspace.
15718+ * assist the pathconf(3) wrapper library.
c2b27bf2 15719+ * move-down
4a4d8108
AM
15720+ */
15721+
c2b27bf2
AM
15722+#include <linux/compat.h>
15723+#include <linux/file.h>
4a4d8108
AM
15724+#include "aufs.h"
15725+
1e00d052 15726+static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
4a4d8108
AM
15727+{
15728+ int err, fd;
15729+ aufs_bindex_t wbi, bindex, bend;
15730+ struct file *h_file;
15731+ struct super_block *sb;
15732+ struct dentry *root;
1e00d052
AM
15733+ struct au_branch *br;
15734+ struct aufs_wbr_fd wbrfd = {
15735+ .oflags = au_dir_roflags,
15736+ .brid = -1
15737+ };
15738+ const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
15739+ | O_NOATIME | O_CLOEXEC;
4a4d8108 15740+
1e00d052
AM
15741+ AuDebugOn(wbrfd.oflags & ~valid);
15742+
15743+ if (arg) {
15744+ err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
15745+ if (unlikely(err)) {
15746+ err = -EFAULT;
15747+ goto out;
15748+ }
15749+
15750+ err = -EINVAL;
15751+ AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
15752+ wbrfd.oflags |= au_dir_roflags;
15753+ AuDbg("0%o\n", wbrfd.oflags);
15754+ if (unlikely(wbrfd.oflags & ~valid))
15755+ goto out;
15756+ }
15757+
15758+ fd = get_unused_fd();
15759+ err = fd;
15760+ if (unlikely(fd < 0))
4a4d8108 15761+ goto out;
4a4d8108 15762+
1e00d052 15763+ h_file = ERR_PTR(-EINVAL);
4a4d8108 15764+ wbi = 0;
1e00d052 15765+ br = NULL;
4a4d8108
AM
15766+ sb = path->dentry->d_sb;
15767+ root = sb->s_root;
15768+ aufs_read_lock(root, AuLock_IR);
1e00d052
AM
15769+ bend = au_sbend(sb);
15770+ if (wbrfd.brid >= 0) {
15771+ wbi = au_br_index(sb, wbrfd.brid);
15772+ if (unlikely(wbi < 0 || wbi > bend))
15773+ goto out_unlock;
15774+ }
15775+
15776+ h_file = ERR_PTR(-ENOENT);
15777+ br = au_sbr(sb, wbi);
15778+ if (!au_br_writable(br->br_perm)) {
15779+ if (arg)
15780+ goto out_unlock;
15781+
15782+ bindex = wbi + 1;
15783+ wbi = -1;
15784+ for (; bindex <= bend; bindex++) {
15785+ br = au_sbr(sb, bindex);
15786+ if (au_br_writable(br->br_perm)) {
4a4d8108 15787+ wbi = bindex;
1e00d052 15788+ br = au_sbr(sb, wbi);
4a4d8108
AM
15789+ break;
15790+ }
15791+ }
4a4d8108
AM
15792+ }
15793+ AuDbg("wbi %d\n", wbi);
1e00d052 15794+ if (wbi >= 0)
392086de
AM
15795+ h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
15796+ /*force_wr*/0);
1e00d052
AM
15797+
15798+out_unlock:
4a4d8108
AM
15799+ aufs_read_unlock(root, AuLock_IR);
15800+ err = PTR_ERR(h_file);
15801+ if (IS_ERR(h_file))
15802+ goto out_fd;
15803+
1e00d052 15804+ atomic_dec(&br->br_count); /* cf. au_h_open() */
4a4d8108
AM
15805+ fd_install(fd, h_file);
15806+ err = fd;
15807+ goto out; /* success */
15808+
4f0767ce 15809+out_fd:
4a4d8108 15810+ put_unused_fd(fd);
4f0767ce 15811+out:
1e00d052 15812+ AuTraceErr(err);
4a4d8108
AM
15813+ return err;
15814+}
15815+
15816+/* ---------------------------------------------------------------------- */
15817+
15818+long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
15819+{
15820+ long err;
15821+
15822+ switch (cmd) {
4a4d8108
AM
15823+ case AUFS_CTL_RDU:
15824+ case AUFS_CTL_RDU_INO:
15825+ err = au_rdu_ioctl(file, cmd, arg);
15826+ break;
15827+
15828+ case AUFS_CTL_WBR_FD:
1e00d052 15829+ err = au_wbr_fd(&file->f_path, (void __user *)arg);
4a4d8108
AM
15830+ break;
15831+
027c5e7a
AM
15832+ case AUFS_CTL_IBUSY:
15833+ err = au_ibusy_ioctl(file, arg);
15834+ break;
15835+
4a4d8108
AM
15836+ default:
15837+ /* do not call the lower */
15838+ AuDbg("0x%x\n", cmd);
15839+ err = -ENOTTY;
15840+ }
15841+
15842+ AuTraceErr(err);
15843+ return err;
15844+}
15845+
15846+long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
15847+{
15848+ long err;
15849+
15850+ switch (cmd) {
c2b27bf2 15851+ case AUFS_CTL_MVDOWN:
c2b27bf2
AM
15852+ err = au_mvdown(file->f_dentry, (void __user *)arg);
15853+ break;
15854+
4a4d8108 15855+ case AUFS_CTL_WBR_FD:
1e00d052 15856+ err = au_wbr_fd(&file->f_path, (void __user *)arg);
4a4d8108
AM
15857+ break;
15858+
15859+ default:
15860+ /* do not call the lower */
15861+ AuDbg("0x%x\n", cmd);
15862+ err = -ENOTTY;
15863+ }
15864+
15865+ AuTraceErr(err);
15866+ return err;
15867+}
b752ccd1
AM
15868+
15869+#ifdef CONFIG_COMPAT
15870+long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15871+ unsigned long arg)
15872+{
15873+ long err;
15874+
15875+ switch (cmd) {
15876+ case AUFS_CTL_RDU:
15877+ case AUFS_CTL_RDU_INO:
15878+ err = au_rdu_compat_ioctl(file, cmd, arg);
15879+ break;
15880+
027c5e7a
AM
15881+ case AUFS_CTL_IBUSY:
15882+ err = au_ibusy_compat_ioctl(file, arg);
15883+ break;
15884+
b752ccd1
AM
15885+ default:
15886+ err = aufs_ioctl_dir(file, cmd, arg);
15887+ }
15888+
15889+ AuTraceErr(err);
15890+ return err;
15891+}
15892+
b752ccd1
AM
15893+long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15894+ unsigned long arg)
15895+{
15896+ return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
15897+}
15898+#endif
7f207e10
AM
15899diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
15900--- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
15901+++ linux/fs/aufs/i_op_add.c 2014-06-24 00:14:06.782719306 +0200
15902@@ -0,0 +1,881 @@
4a4d8108 15903+/*
523b37e3 15904+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
15905+ *
15906+ * This program, aufs is free software; you can redistribute it and/or modify
15907+ * it under the terms of the GNU General Public License as published by
15908+ * the Free Software Foundation; either version 2 of the License, or
15909+ * (at your option) any later version.
15910+ *
15911+ * This program is distributed in the hope that it will be useful,
15912+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
15913+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15914+ * GNU General Public License for more details.
15915+ *
15916+ * You should have received a copy of the GNU General Public License
523b37e3 15917+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
15918+ */
15919+
15920+/*
15921+ * inode operations (add entry)
15922+ */
15923+
15924+#include "aufs.h"
15925+
15926+/*
15927+ * final procedure of adding a new entry, except link(2).
15928+ * remove whiteout, instantiate, copyup the parent dir's times and size
15929+ * and update version.
15930+ * if it failed, re-create the removed whiteout.
15931+ */
15932+static int epilog(struct inode *dir, aufs_bindex_t bindex,
15933+ struct dentry *wh_dentry, struct dentry *dentry)
15934+{
15935+ int err, rerr;
15936+ aufs_bindex_t bwh;
15937+ struct path h_path;
15938+ struct inode *inode, *h_dir;
15939+ struct dentry *wh;
15940+
15941+ bwh = -1;
15942+ if (wh_dentry) {
15943+ h_dir = wh_dentry->d_parent->d_inode; /* dir inode is locked */
15944+ IMustLock(h_dir);
15945+ AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
15946+ bwh = au_dbwh(dentry);
15947+ h_path.dentry = wh_dentry;
15948+ h_path.mnt = au_sbr_mnt(dir->i_sb, bindex);
15949+ err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
15950+ dentry);
15951+ if (unlikely(err))
15952+ goto out;
15953+ }
15954+
15955+ inode = au_new_inode(dentry, /*must_new*/1);
15956+ if (!IS_ERR(inode)) {
15957+ d_instantiate(dentry, inode);
15958+ dir = dentry->d_parent->d_inode; /* dir inode is locked */
15959+ IMustLock(dir);
15960+ if (au_ibstart(dir) == au_dbstart(dentry))
15961+ au_cpup_attr_timesizes(dir);
15962+ dir->i_version++;
15963+ return 0; /* success */
15964+ }
15965+
15966+ err = PTR_ERR(inode);
15967+ if (!wh_dentry)
15968+ goto out;
15969+
15970+ /* revert */
15971+ /* dir inode is locked */
15972+ wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
15973+ rerr = PTR_ERR(wh);
15974+ if (IS_ERR(wh)) {
523b37e3
AM
15975+ AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
15976+ dentry, err, rerr);
4a4d8108
AM
15977+ err = -EIO;
15978+ } else
15979+ dput(wh);
15980+
4f0767ce 15981+out:
4a4d8108
AM
15982+ return err;
15983+}
15984+
027c5e7a
AM
15985+static int au_d_may_add(struct dentry *dentry)
15986+{
15987+ int err;
15988+
15989+ err = 0;
15990+ if (unlikely(d_unhashed(dentry)))
15991+ err = -ENOENT;
15992+ if (unlikely(dentry->d_inode))
15993+ err = -EEXIST;
15994+ return err;
15995+}
15996+
4a4d8108
AM
15997+/*
15998+ * simple tests for the adding inode operations.
15999+ * following the checks in vfs, plus the parent-child relationship.
16000+ */
16001+int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
16002+ struct dentry *h_parent, int isdir)
16003+{
16004+ int err;
16005+ umode_t h_mode;
16006+ struct dentry *h_dentry;
16007+ struct inode *h_inode;
16008+
16009+ err = -ENAMETOOLONG;
16010+ if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
16011+ goto out;
16012+
16013+ h_dentry = au_h_dptr(dentry, bindex);
16014+ h_inode = h_dentry->d_inode;
16015+ if (!dentry->d_inode) {
16016+ err = -EEXIST;
16017+ if (unlikely(h_inode))
16018+ goto out;
16019+ } else {
16020+ /* rename(2) case */
16021+ err = -EIO;
16022+ if (unlikely(!h_inode || !h_inode->i_nlink))
16023+ goto out;
16024+
16025+ h_mode = h_inode->i_mode;
16026+ if (!isdir) {
16027+ err = -EISDIR;
16028+ if (unlikely(S_ISDIR(h_mode)))
16029+ goto out;
16030+ } else if (unlikely(!S_ISDIR(h_mode))) {
16031+ err = -ENOTDIR;
16032+ goto out;
16033+ }
16034+ }
16035+
16036+ err = 0;
16037+ /* expected parent dir is locked */
16038+ if (unlikely(h_parent != h_dentry->d_parent))
16039+ err = -EIO;
16040+
4f0767ce 16041+out:
4a4d8108
AM
16042+ AuTraceErr(err);
16043+ return err;
16044+}
16045+
16046+/*
16047+ * initial procedure of adding a new entry.
16048+ * prepare writable branch and the parent dir, lock it,
16049+ * and lookup whiteout for the new entry.
16050+ */
16051+static struct dentry*
16052+lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
16053+ struct dentry *src_dentry, struct au_pin *pin,
16054+ struct au_wr_dir_args *wr_dir_args)
16055+{
16056+ struct dentry *wh_dentry, *h_parent;
16057+ struct super_block *sb;
16058+ struct au_branch *br;
16059+ int err;
16060+ unsigned int udba;
16061+ aufs_bindex_t bcpup;
16062+
523b37e3 16063+ AuDbg("%pd\n", dentry);
4a4d8108
AM
16064+
16065+ err = au_wr_dir(dentry, src_dentry, wr_dir_args);
16066+ bcpup = err;
16067+ wh_dentry = ERR_PTR(err);
16068+ if (unlikely(err < 0))
16069+ goto out;
16070+
16071+ sb = dentry->d_sb;
16072+ udba = au_opt_udba(sb);
16073+ err = au_pin(pin, dentry, bcpup, udba,
16074+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
16075+ wh_dentry = ERR_PTR(err);
16076+ if (unlikely(err))
16077+ goto out;
16078+
16079+ h_parent = au_pinned_h_parent(pin);
16080+ if (udba != AuOpt_UDBA_NONE
16081+ && au_dbstart(dentry) == bcpup)
16082+ err = au_may_add(dentry, bcpup, h_parent,
16083+ au_ftest_wrdir(wr_dir_args->flags, ISDIR));
16084+ else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
16085+ err = -ENAMETOOLONG;
16086+ wh_dentry = ERR_PTR(err);
16087+ if (unlikely(err))
16088+ goto out_unpin;
16089+
16090+ br = au_sbr(sb, bcpup);
16091+ if (dt) {
16092+ struct path tmp = {
16093+ .dentry = h_parent,
86dc4139 16094+ .mnt = au_br_mnt(br)
4a4d8108
AM
16095+ };
16096+ au_dtime_store(dt, au_pinned_parent(pin), &tmp);
16097+ }
16098+
16099+ wh_dentry = NULL;
16100+ if (bcpup != au_dbwh(dentry))
16101+ goto out; /* success */
16102+
16103+ wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
16104+
4f0767ce 16105+out_unpin:
4a4d8108
AM
16106+ if (IS_ERR(wh_dentry))
16107+ au_unpin(pin);
4f0767ce 16108+out:
4a4d8108
AM
16109+ return wh_dentry;
16110+}
16111+
16112+/* ---------------------------------------------------------------------- */
16113+
16114+enum { Mknod, Symlink, Creat };
16115+struct simple_arg {
16116+ int type;
16117+ union {
16118+ struct {
7eafdf33 16119+ umode_t mode;
b4510431 16120+ bool want_excl;
4a4d8108
AM
16121+ } c;
16122+ struct {
16123+ const char *symname;
16124+ } s;
16125+ struct {
7eafdf33 16126+ umode_t mode;
4a4d8108
AM
16127+ dev_t dev;
16128+ } m;
16129+ } u;
16130+};
16131+
16132+static int add_simple(struct inode *dir, struct dentry *dentry,
16133+ struct simple_arg *arg)
16134+{
16135+ int err;
16136+ aufs_bindex_t bstart;
16137+ unsigned char created;
4a4d8108
AM
16138+ struct dentry *wh_dentry, *parent;
16139+ struct inode *h_dir;
c2b27bf2
AM
16140+ /* to reuduce stack size */
16141+ struct {
16142+ struct au_dtime dt;
16143+ struct au_pin pin;
16144+ struct path h_path;
16145+ struct au_wr_dir_args wr_dir_args;
16146+ } *a;
4a4d8108 16147+
523b37e3 16148+ AuDbg("%pd\n", dentry);
4a4d8108
AM
16149+ IMustLock(dir);
16150+
c2b27bf2
AM
16151+ err = -ENOMEM;
16152+ a = kmalloc(sizeof(*a), GFP_NOFS);
16153+ if (unlikely(!a))
16154+ goto out;
16155+ a->wr_dir_args.force_btgt = -1;
16156+ a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
16157+
4a4d8108 16158+ parent = dentry->d_parent; /* dir inode is locked */
027c5e7a
AM
16159+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
16160+ if (unlikely(err))
c2b27bf2 16161+ goto out_free;
027c5e7a
AM
16162+ err = au_d_may_add(dentry);
16163+ if (unlikely(err))
16164+ goto out_unlock;
4a4d8108 16165+ di_write_lock_parent(parent);
c2b27bf2
AM
16166+ wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
16167+ &a->pin, &a->wr_dir_args);
4a4d8108
AM
16168+ err = PTR_ERR(wh_dentry);
16169+ if (IS_ERR(wh_dentry))
027c5e7a 16170+ goto out_parent;
4a4d8108
AM
16171+
16172+ bstart = au_dbstart(dentry);
c2b27bf2
AM
16173+ a->h_path.dentry = au_h_dptr(dentry, bstart);
16174+ a->h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
16175+ h_dir = au_pinned_h_dir(&a->pin);
4a4d8108
AM
16176+ switch (arg->type) {
16177+ case Creat:
c2b27bf2 16178+ err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
537831f9 16179+ arg->u.c.want_excl);
4a4d8108
AM
16180+ break;
16181+ case Symlink:
c2b27bf2 16182+ err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
4a4d8108
AM
16183+ break;
16184+ case Mknod:
c2b27bf2
AM
16185+ err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
16186+ arg->u.m.dev);
4a4d8108
AM
16187+ break;
16188+ default:
16189+ BUG();
16190+ }
16191+ created = !err;
16192+ if (!err)
16193+ err = epilog(dir, bstart, wh_dentry, dentry);
16194+
16195+ /* revert */
c2b27bf2 16196+ if (unlikely(created && err && a->h_path.dentry->d_inode)) {
4a4d8108 16197+ int rerr;
523b37e3
AM
16198+ /* no delegation since it is just created */
16199+ rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
16200+ /*force*/0);
4a4d8108 16201+ if (rerr) {
523b37e3
AM
16202+ AuIOErr("%pd revert failure(%d, %d)\n",
16203+ dentry, err, rerr);
4a4d8108
AM
16204+ err = -EIO;
16205+ }
c2b27bf2 16206+ au_dtime_revert(&a->dt);
4a4d8108
AM
16207+ }
16208+
c2b27bf2 16209+ au_unpin(&a->pin);
4a4d8108
AM
16210+ dput(wh_dentry);
16211+
027c5e7a
AM
16212+out_parent:
16213+ di_write_unlock(parent);
16214+out_unlock:
4a4d8108
AM
16215+ if (unlikely(err)) {
16216+ au_update_dbstart(dentry);
16217+ d_drop(dentry);
16218+ }
4a4d8108 16219+ aufs_read_unlock(dentry, AuLock_DW);
c2b27bf2
AM
16220+out_free:
16221+ kfree(a);
027c5e7a 16222+out:
4a4d8108
AM
16223+ return err;
16224+}
16225+
7eafdf33
AM
16226+int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
16227+ dev_t dev)
4a4d8108
AM
16228+{
16229+ struct simple_arg arg = {
16230+ .type = Mknod,
16231+ .u.m = {
16232+ .mode = mode,
16233+ .dev = dev
16234+ }
16235+ };
16236+ return add_simple(dir, dentry, &arg);
16237+}
16238+
16239+int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
16240+{
16241+ struct simple_arg arg = {
16242+ .type = Symlink,
16243+ .u.s.symname = symname
16244+ };
16245+ return add_simple(dir, dentry, &arg);
16246+}
16247+
7eafdf33 16248+int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
b4510431 16249+ bool want_excl)
4a4d8108
AM
16250+{
16251+ struct simple_arg arg = {
16252+ .type = Creat,
16253+ .u.c = {
b4510431
AM
16254+ .mode = mode,
16255+ .want_excl = want_excl
4a4d8108
AM
16256+ }
16257+ };
16258+ return add_simple(dir, dentry, &arg);
16259+}
16260+
38d290e6
JR
16261+int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
16262+{
16263+ int err;
16264+ aufs_bindex_t bindex;
16265+ struct super_block *sb;
16266+ struct dentry *parent, *h_parent, *h_dentry;
16267+ struct inode *h_dir, *inode;
16268+ struct vfsmount *h_mnt;
16269+ struct au_wr_dir_args wr_dir_args = {
16270+ .force_btgt = -1,
16271+ .flags = AuWrDir_TMPFILE
16272+ };
16273+
16274+ /* copy-up may happen */
16275+ mutex_lock(&dir->i_mutex);
16276+
16277+ sb = dir->i_sb;
16278+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16279+ if (unlikely(err))
16280+ goto out;
16281+
16282+ err = au_di_init(dentry);
16283+ if (unlikely(err))
16284+ goto out_si;
16285+
16286+ err = -EBUSY;
16287+ parent = d_find_any_alias(dir);
16288+ AuDebugOn(!parent);
16289+ di_write_lock_parent(parent);
16290+ if (unlikely(parent->d_inode != dir))
16291+ goto out_parent;
16292+
16293+ err = au_digen_test(parent, au_sigen(sb));
16294+ if (unlikely(err))
16295+ goto out_parent;
16296+
16297+ bindex = au_dbstart(parent);
16298+ au_set_dbstart(dentry, bindex);
16299+ au_set_dbend(dentry, bindex);
16300+ err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
16301+ bindex = err;
16302+ if (unlikely(err < 0))
16303+ goto out_parent;
16304+
16305+ err = -EOPNOTSUPP;
16306+ h_dir = au_h_iptr(dir, bindex);
16307+ if (unlikely(!h_dir->i_op->tmpfile))
16308+ goto out_parent;
16309+
16310+ h_mnt = au_sbr_mnt(sb, bindex);
16311+ err = vfsub_mnt_want_write(h_mnt);
16312+ if (unlikely(err))
16313+ goto out_parent;
16314+
16315+ h_parent = au_h_dptr(parent, bindex);
16316+ err = inode_permission(h_parent->d_inode, MAY_WRITE | MAY_EXEC);
16317+ if (unlikely(err))
16318+ goto out_mnt;
16319+
16320+ err = -ENOMEM;
16321+ h_dentry = d_alloc(h_parent, &dentry->d_name);
16322+ if (unlikely(!h_dentry))
16323+ goto out_mnt;
16324+
16325+ err = h_dir->i_op->tmpfile(h_dir, h_dentry, mode);
16326+ if (unlikely(err))
16327+ goto out_dentry;
16328+
16329+ au_set_dbstart(dentry, bindex);
16330+ au_set_dbend(dentry, bindex);
16331+ au_set_h_dptr(dentry, bindex, dget(h_dentry));
16332+ inode = au_new_inode(dentry, /*must_new*/1);
16333+ if (IS_ERR(inode)) {
16334+ err = PTR_ERR(inode);
16335+ au_set_h_dptr(dentry, bindex, NULL);
16336+ au_set_dbstart(dentry, -1);
16337+ au_set_dbend(dentry, -1);
16338+ } else {
16339+ if (!inode->i_nlink)
16340+ set_nlink(inode, 1);
16341+ d_tmpfile(dentry, inode);
16342+ au_di(dentry)->di_tmpfile = 1;
16343+
16344+ /* update without i_mutex */
16345+ if (au_ibstart(dir) == au_dbstart(dentry))
16346+ au_cpup_attr_timesizes(dir);
16347+ }
16348+
16349+out_dentry:
16350+ dput(h_dentry);
16351+out_mnt:
16352+ vfsub_mnt_drop_write(h_mnt);
16353+out_parent:
16354+ di_write_unlock(parent);
16355+ dput(parent);
16356+ di_write_unlock(dentry);
16357+ if (!err)
16358+#if 0
16359+ /* verbose coding for lock class name */
16360+ au_rw_class(&au_di(dentry)->di_rwsem,
16361+ au_lc_key + AuLcNonDir_DIINFO);
16362+#else
16363+ ;
16364+#endif
16365+ else {
16366+ au_di_fin(dentry);
16367+ dentry->d_fsdata = NULL;
16368+ }
16369+out_si:
16370+ si_read_unlock(sb);
16371+out:
16372+ mutex_unlock(&dir->i_mutex);
16373+ return err;
16374+}
16375+
4a4d8108
AM
16376+/* ---------------------------------------------------------------------- */
16377+
16378+struct au_link_args {
16379+ aufs_bindex_t bdst, bsrc;
16380+ struct au_pin pin;
16381+ struct path h_path;
16382+ struct dentry *src_parent, *parent;
16383+};
16384+
16385+static int au_cpup_before_link(struct dentry *src_dentry,
16386+ struct au_link_args *a)
16387+{
16388+ int err;
16389+ struct dentry *h_src_dentry;
c2b27bf2
AM
16390+ struct au_cp_generic cpg = {
16391+ .dentry = src_dentry,
16392+ .bdst = a->bdst,
16393+ .bsrc = a->bsrc,
16394+ .len = -1,
16395+ .pin = &a->pin,
16396+ .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
16397+ };
4a4d8108
AM
16398+
16399+ di_read_lock_parent(a->src_parent, AuLock_IR);
16400+ err = au_test_and_cpup_dirs(src_dentry, a->bdst);
16401+ if (unlikely(err))
16402+ goto out;
16403+
16404+ h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
4a4d8108
AM
16405+ err = au_pin(&a->pin, src_dentry, a->bdst,
16406+ au_opt_udba(src_dentry->d_sb),
16407+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
16408+ if (unlikely(err))
16409+ goto out;
367653fa 16410+
c2b27bf2 16411+ err = au_sio_cpup_simple(&cpg);
4a4d8108
AM
16412+ au_unpin(&a->pin);
16413+
4f0767ce 16414+out:
4a4d8108
AM
16415+ di_read_unlock(a->src_parent, AuLock_IR);
16416+ return err;
16417+}
16418+
86dc4139
AM
16419+static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
16420+ struct au_link_args *a)
4a4d8108
AM
16421+{
16422+ int err;
16423+ unsigned char plink;
86dc4139 16424+ aufs_bindex_t bend;
4a4d8108 16425+ struct dentry *h_src_dentry;
523b37e3 16426+ struct inode *h_inode, *inode, *delegated;
4a4d8108
AM
16427+ struct super_block *sb;
16428+ struct file *h_file;
16429+
16430+ plink = 0;
16431+ h_inode = NULL;
16432+ sb = src_dentry->d_sb;
16433+ inode = src_dentry->d_inode;
16434+ if (au_ibstart(inode) <= a->bdst)
16435+ h_inode = au_h_iptr(inode, a->bdst);
16436+ if (!h_inode || !h_inode->i_nlink) {
16437+ /* copyup src_dentry as the name of dentry. */
86dc4139
AM
16438+ bend = au_dbend(dentry);
16439+ if (bend < a->bsrc)
16440+ au_set_dbend(dentry, a->bsrc);
16441+ au_set_h_dptr(dentry, a->bsrc,
16442+ dget(au_h_dptr(src_dentry, a->bsrc)));
16443+ dget(a->h_path.dentry);
16444+ au_set_h_dptr(dentry, a->bdst, NULL);
16445+ dentry->d_inode = src_dentry->d_inode; /* tmp */
392086de 16446+ h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
86dc4139 16447+ if (IS_ERR(h_file))
4a4d8108 16448+ err = PTR_ERR(h_file);
86dc4139 16449+ else {
c2b27bf2
AM
16450+ struct au_cp_generic cpg = {
16451+ .dentry = dentry,
16452+ .bdst = a->bdst,
16453+ .bsrc = -1,
16454+ .len = -1,
16455+ .pin = &a->pin,
16456+ .flags = AuCpup_KEEPLINO
16457+ };
16458+ err = au_sio_cpup_simple(&cpg);
86dc4139
AM
16459+ au_h_open_post(dentry, a->bsrc, h_file);
16460+ if (!err) {
16461+ dput(a->h_path.dentry);
16462+ a->h_path.dentry = au_h_dptr(dentry, a->bdst);
16463+ } else
16464+ au_set_h_dptr(dentry, a->bdst,
16465+ a->h_path.dentry);
16466+ }
16467+ dentry->d_inode = NULL; /* restore */
16468+ au_set_h_dptr(dentry, a->bsrc, NULL);
16469+ au_set_dbend(dentry, bend);
4a4d8108
AM
16470+ } else {
16471+ /* the inode of src_dentry already exists on a.bdst branch */
16472+ h_src_dentry = d_find_alias(h_inode);
16473+ if (!h_src_dentry && au_plink_test(inode)) {
16474+ plink = 1;
16475+ h_src_dentry = au_plink_lkup(inode, a->bdst);
16476+ err = PTR_ERR(h_src_dentry);
16477+ if (IS_ERR(h_src_dentry))
16478+ goto out;
16479+
16480+ if (unlikely(!h_src_dentry->d_inode)) {
16481+ dput(h_src_dentry);
16482+ h_src_dentry = NULL;
16483+ }
16484+
16485+ }
16486+ if (h_src_dentry) {
523b37e3 16487+ delegated = NULL;
4a4d8108 16488+ err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
523b37e3
AM
16489+ &a->h_path, &delegated);
16490+ if (unlikely(err == -EWOULDBLOCK)) {
16491+ pr_warn("cannot retry for NFSv4 delegation"
16492+ " for an internal link\n");
16493+ iput(delegated);
16494+ }
4a4d8108
AM
16495+ dput(h_src_dentry);
16496+ } else {
16497+ AuIOErr("no dentry found for hi%lu on b%d\n",
16498+ h_inode->i_ino, a->bdst);
16499+ err = -EIO;
16500+ }
16501+ }
16502+
16503+ if (!err && !plink)
16504+ au_plink_append(inode, a->bdst, a->h_path.dentry);
16505+
16506+out:
2cbb1c4b 16507+ AuTraceErr(err);
4a4d8108
AM
16508+ return err;
16509+}
16510+
16511+int aufs_link(struct dentry *src_dentry, struct inode *dir,
16512+ struct dentry *dentry)
16513+{
16514+ int err, rerr;
16515+ struct au_dtime dt;
16516+ struct au_link_args *a;
16517+ struct dentry *wh_dentry, *h_src_dentry;
523b37e3 16518+ struct inode *inode, *delegated;
4a4d8108
AM
16519+ struct super_block *sb;
16520+ struct au_wr_dir_args wr_dir_args = {
16521+ /* .force_btgt = -1, */
16522+ .flags = AuWrDir_ADD_ENTRY
16523+ };
16524+
16525+ IMustLock(dir);
16526+ inode = src_dentry->d_inode;
16527+ IMustLock(inode);
16528+
4a4d8108
AM
16529+ err = -ENOMEM;
16530+ a = kzalloc(sizeof(*a), GFP_NOFS);
16531+ if (unlikely(!a))
16532+ goto out;
16533+
16534+ a->parent = dentry->d_parent; /* dir inode is locked */
027c5e7a
AM
16535+ err = aufs_read_and_write_lock2(dentry, src_dentry,
16536+ AuLock_NOPLM | AuLock_GEN);
e49829fe
JR
16537+ if (unlikely(err))
16538+ goto out_kfree;
38d290e6 16539+ err = au_d_linkable(src_dentry);
027c5e7a
AM
16540+ if (unlikely(err))
16541+ goto out_unlock;
16542+ err = au_d_may_add(dentry);
16543+ if (unlikely(err))
16544+ goto out_unlock;
e49829fe 16545+
4a4d8108 16546+ a->src_parent = dget_parent(src_dentry);
2cbb1c4b 16547+ wr_dir_args.force_btgt = au_ibstart(inode);
4a4d8108
AM
16548+
16549+ di_write_lock_parent(a->parent);
16550+ wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
16551+ wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
16552+ &wr_dir_args);
16553+ err = PTR_ERR(wh_dentry);
16554+ if (IS_ERR(wh_dentry))
027c5e7a 16555+ goto out_parent;
4a4d8108
AM
16556+
16557+ err = 0;
16558+ sb = dentry->d_sb;
16559+ a->bdst = au_dbstart(dentry);
16560+ a->h_path.dentry = au_h_dptr(dentry, a->bdst);
16561+ a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
2cbb1c4b
JR
16562+ a->bsrc = au_ibstart(inode);
16563+ h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
38d290e6
JR
16564+ if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
16565+ h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
2cbb1c4b
JR
16566+ if (!h_src_dentry) {
16567+ a->bsrc = au_dbstart(src_dentry);
16568+ h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
16569+ AuDebugOn(!h_src_dentry);
38d290e6
JR
16570+ } else if (IS_ERR(h_src_dentry)) {
16571+ err = PTR_ERR(h_src_dentry);
2cbb1c4b 16572+ goto out_parent;
38d290e6 16573+ }
2cbb1c4b 16574+
4a4d8108
AM
16575+ if (au_opt_test(au_mntflags(sb), PLINK)) {
16576+ if (a->bdst < a->bsrc
16577+ /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
86dc4139 16578+ err = au_cpup_or_link(src_dentry, dentry, a);
523b37e3
AM
16579+ else {
16580+ delegated = NULL;
4a4d8108 16581+ err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
523b37e3
AM
16582+ &a->h_path, &delegated);
16583+ if (unlikely(err == -EWOULDBLOCK)) {
16584+ pr_warn("cannot retry for NFSv4 delegation"
16585+ " for an internal link\n");
16586+ iput(delegated);
16587+ }
16588+ }
2cbb1c4b 16589+ dput(h_src_dentry);
4a4d8108
AM
16590+ } else {
16591+ /*
16592+ * copyup src_dentry to the branch we process,
16593+ * and then link(2) to it.
16594+ */
2cbb1c4b 16595+ dput(h_src_dentry);
4a4d8108
AM
16596+ if (a->bdst < a->bsrc
16597+ /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
16598+ au_unpin(&a->pin);
16599+ di_write_unlock(a->parent);
16600+ err = au_cpup_before_link(src_dentry, a);
16601+ di_write_lock_parent(a->parent);
16602+ if (!err)
16603+ err = au_pin(&a->pin, dentry, a->bdst,
16604+ au_opt_udba(sb),
16605+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
16606+ if (unlikely(err))
16607+ goto out_wh;
16608+ }
16609+ if (!err) {
16610+ h_src_dentry = au_h_dptr(src_dentry, a->bdst);
16611+ err = -ENOENT;
523b37e3
AM
16612+ if (h_src_dentry && h_src_dentry->d_inode) {
16613+ delegated = NULL;
4a4d8108
AM
16614+ err = vfsub_link(h_src_dentry,
16615+ au_pinned_h_dir(&a->pin),
523b37e3
AM
16616+ &a->h_path, &delegated);
16617+ if (unlikely(err == -EWOULDBLOCK)) {
16618+ pr_warn("cannot retry"
16619+ " for NFSv4 delegation"
16620+ " for an internal link\n");
16621+ iput(delegated);
16622+ }
16623+ }
4a4d8108
AM
16624+ }
16625+ }
16626+ if (unlikely(err))
16627+ goto out_unpin;
16628+
16629+ if (wh_dentry) {
16630+ a->h_path.dentry = wh_dentry;
16631+ err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
16632+ dentry);
16633+ if (unlikely(err))
16634+ goto out_revert;
16635+ }
16636+
16637+ dir->i_version++;
16638+ if (au_ibstart(dir) == au_dbstart(dentry))
16639+ au_cpup_attr_timesizes(dir);
16640+ inc_nlink(inode);
16641+ inode->i_ctime = dir->i_ctime;
027c5e7a
AM
16642+ d_instantiate(dentry, au_igrab(inode));
16643+ if (d_unhashed(a->h_path.dentry))
4a4d8108
AM
16644+ /* some filesystem calls d_drop() */
16645+ d_drop(dentry);
16646+ goto out_unpin; /* success */
16647+
4f0767ce 16648+out_revert:
523b37e3
AM
16649+ /* no delegation since it is just created */
16650+ rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
16651+ /*delegated*/NULL, /*force*/0);
027c5e7a 16652+ if (unlikely(rerr)) {
523b37e3 16653+ AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
027c5e7a
AM
16654+ err = -EIO;
16655+ }
4a4d8108 16656+ au_dtime_revert(&dt);
4f0767ce 16657+out_unpin:
4a4d8108 16658+ au_unpin(&a->pin);
4f0767ce 16659+out_wh:
4a4d8108 16660+ dput(wh_dentry);
027c5e7a
AM
16661+out_parent:
16662+ di_write_unlock(a->parent);
16663+ dput(a->src_parent);
4f0767ce 16664+out_unlock:
4a4d8108
AM
16665+ if (unlikely(err)) {
16666+ au_update_dbstart(dentry);
16667+ d_drop(dentry);
16668+ }
4a4d8108 16669+ aufs_read_and_write_unlock2(dentry, src_dentry);
e49829fe 16670+out_kfree:
4a4d8108 16671+ kfree(a);
4f0767ce 16672+out:
86dc4139 16673+ AuTraceErr(err);
4a4d8108
AM
16674+ return err;
16675+}
16676+
7eafdf33 16677+int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4a4d8108
AM
16678+{
16679+ int err, rerr;
16680+ aufs_bindex_t bindex;
16681+ unsigned char diropq;
16682+ struct path h_path;
16683+ struct dentry *wh_dentry, *parent, *opq_dentry;
16684+ struct mutex *h_mtx;
16685+ struct super_block *sb;
16686+ struct {
16687+ struct au_pin pin;
16688+ struct au_dtime dt;
16689+ } *a; /* reduce the stack usage */
16690+ struct au_wr_dir_args wr_dir_args = {
16691+ .force_btgt = -1,
16692+ .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
16693+ };
16694+
16695+ IMustLock(dir);
16696+
16697+ err = -ENOMEM;
16698+ a = kmalloc(sizeof(*a), GFP_NOFS);
16699+ if (unlikely(!a))
16700+ goto out;
16701+
027c5e7a
AM
16702+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
16703+ if (unlikely(err))
16704+ goto out_free;
16705+ err = au_d_may_add(dentry);
16706+ if (unlikely(err))
16707+ goto out_unlock;
16708+
4a4d8108
AM
16709+ parent = dentry->d_parent; /* dir inode is locked */
16710+ di_write_lock_parent(parent);
16711+ wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
16712+ &a->pin, &wr_dir_args);
16713+ err = PTR_ERR(wh_dentry);
16714+ if (IS_ERR(wh_dentry))
027c5e7a 16715+ goto out_parent;
4a4d8108
AM
16716+
16717+ sb = dentry->d_sb;
16718+ bindex = au_dbstart(dentry);
16719+ h_path.dentry = au_h_dptr(dentry, bindex);
16720+ h_path.mnt = au_sbr_mnt(sb, bindex);
16721+ err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
16722+ if (unlikely(err))
027c5e7a 16723+ goto out_unpin;
4a4d8108
AM
16724+
16725+ /* make the dir opaque */
16726+ diropq = 0;
16727+ h_mtx = &h_path.dentry->d_inode->i_mutex;
16728+ if (wh_dentry
16729+ || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
16730+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
16731+ opq_dentry = au_diropq_create(dentry, bindex);
16732+ mutex_unlock(h_mtx);
16733+ err = PTR_ERR(opq_dentry);
16734+ if (IS_ERR(opq_dentry))
16735+ goto out_dir;
16736+ dput(opq_dentry);
16737+ diropq = 1;
16738+ }
16739+
16740+ err = epilog(dir, bindex, wh_dentry, dentry);
16741+ if (!err) {
16742+ inc_nlink(dir);
027c5e7a 16743+ goto out_unpin; /* success */
4a4d8108
AM
16744+ }
16745+
16746+ /* revert */
16747+ if (diropq) {
16748+ AuLabel(revert opq);
16749+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
16750+ rerr = au_diropq_remove(dentry, bindex);
16751+ mutex_unlock(h_mtx);
16752+ if (rerr) {
523b37e3
AM
16753+ AuIOErr("%pd reverting diropq failed(%d, %d)\n",
16754+ dentry, err, rerr);
4a4d8108
AM
16755+ err = -EIO;
16756+ }
16757+ }
16758+
4f0767ce 16759+out_dir:
4a4d8108
AM
16760+ AuLabel(revert dir);
16761+ rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
16762+ if (rerr) {
523b37e3
AM
16763+ AuIOErr("%pd reverting dir failed(%d, %d)\n",
16764+ dentry, err, rerr);
4a4d8108
AM
16765+ err = -EIO;
16766+ }
4a4d8108 16767+ au_dtime_revert(&a->dt);
027c5e7a 16768+out_unpin:
4a4d8108
AM
16769+ au_unpin(&a->pin);
16770+ dput(wh_dentry);
027c5e7a
AM
16771+out_parent:
16772+ di_write_unlock(parent);
16773+out_unlock:
4a4d8108
AM
16774+ if (unlikely(err)) {
16775+ au_update_dbstart(dentry);
16776+ d_drop(dentry);
16777+ }
4a4d8108 16778+ aufs_read_unlock(dentry, AuLock_DW);
027c5e7a 16779+out_free:
4a4d8108 16780+ kfree(a);
4f0767ce 16781+out:
4a4d8108
AM
16782+ return err;
16783+}
7f207e10
AM
16784diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
16785--- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
16786+++ linux/fs/aufs/i_op.c 2014-06-24 00:14:06.782719306 +0200
16787@@ -0,0 +1,1138 @@
4a4d8108 16788+/*
523b37e3 16789+ * Copyright (C) 2005-2014 Junjiro R. Okajima
4a4d8108
AM
16790+ *
16791+ * This program, aufs is free software; you can redistribute it and/or modify
16792+ * it under the terms of the GNU General Public License as published by
16793+ * the Free Software Foundation; either version 2 of the License, or
16794+ * (at your option) any later version.
16795+ *
16796+ * This program is distributed in the hope that it will be useful,
16797+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
16798+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16799+ * GNU General Public License for more details.
16800+ *
16801+ * You should have received a copy of the GNU General Public License
523b37e3 16802+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 16803+ */
1facf9fc 16804+
1308ab2a 16805+/*
4a4d8108 16806+ * inode operations (except add/del/rename)
1308ab2a 16807+ */
4a4d8108
AM
16808+
16809+#include <linux/device_cgroup.h>
16810+#include <linux/fs_stack.h>
92d182d2 16811+#include <linux/mm.h>
4a4d8108
AM
16812+#include <linux/namei.h>
16813+#include <linux/security.h>
4a4d8108
AM
16814+#include "aufs.h"
16815+
1e00d052 16816+static int h_permission(struct inode *h_inode, int mask,
4a4d8108 16817+ struct vfsmount *h_mnt, int brperm)
1facf9fc 16818+{
1308ab2a 16819+ int err;
4a4d8108 16820+ const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
1facf9fc 16821+
4a4d8108
AM
16822+ err = -EACCES;
16823+ if ((write_mask && IS_IMMUTABLE(h_inode))
16824+ || ((mask & MAY_EXEC)
16825+ && S_ISREG(h_inode->i_mode)
16826+ && ((h_mnt->mnt_flags & MNT_NOEXEC)
16827+ || !(h_inode->i_mode & S_IXUGO))))
16828+ goto out;
16829+
16830+ /*
16831+ * - skip the lower fs test in the case of write to ro branch.
16832+ * - nfs dir permission write check is optimized, but a policy for
16833+ * link/rename requires a real check.
16834+ */
16835+ if ((write_mask && !au_br_writable(brperm))
16836+ || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
16837+ && write_mask && !(mask & MAY_READ))
16838+ || !h_inode->i_op->permission) {
16839+ /* AuLabel(generic_permission); */
1e00d052 16840+ err = generic_permission(h_inode, mask);
1308ab2a 16841+ } else {
4a4d8108 16842+ /* AuLabel(h_inode->permission); */
1e00d052 16843+ err = h_inode->i_op->permission(h_inode, mask);
4a4d8108
AM
16844+ AuTraceErr(err);
16845+ }
1facf9fc 16846+
4a4d8108
AM
16847+ if (!err)
16848+ err = devcgroup_inode_permission(h_inode, mask);
7f207e10 16849+ if (!err)
4a4d8108 16850+ err = security_inode_permission(h_inode, mask);
4a4d8108
AM
16851+
16852+#if 0
16853+ if (!err) {
16854+ /* todo: do we need to call ima_path_check()? */
16855+ struct path h_path = {
16856+ .dentry =
16857+ .mnt = h_mnt
16858+ };
16859+ err = ima_path_check(&h_path,
16860+ mask & (MAY_READ | MAY_WRITE | MAY_EXEC),
16861+ IMA_COUNT_LEAVE);
1308ab2a 16862+ }
4a4d8108 16863+#endif
dece6358 16864+
4f0767ce 16865+out:
1308ab2a 16866+ return err;
16867+}
dece6358 16868+
1e00d052 16869+static int aufs_permission(struct inode *inode, int mask)
1308ab2a 16870+{
16871+ int err;
4a4d8108
AM
16872+ aufs_bindex_t bindex, bend;
16873+ const unsigned char isdir = !!S_ISDIR(inode->i_mode),
16874+ write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
16875+ struct inode *h_inode;
16876+ struct super_block *sb;
16877+ struct au_branch *br;
1facf9fc 16878+
027c5e7a 16879+ /* todo: support rcu-walk? */
1e00d052 16880+ if (mask & MAY_NOT_BLOCK)
027c5e7a
AM
16881+ return -ECHILD;
16882+
4a4d8108
AM
16883+ sb = inode->i_sb;
16884+ si_read_lock(sb, AuLock_FLUSH);
16885+ ii_read_lock_child(inode);
027c5e7a
AM
16886+#if 0
16887+ err = au_iigen_test(inode, au_sigen(sb));
16888+ if (unlikely(err))
16889+ goto out;
16890+#endif
dece6358 16891+
4a4d8108
AM
16892+ if (!isdir || write_mask) {
16893+ err = au_busy_or_stale();
16894+ h_inode = au_h_iptr(inode, au_ibstart(inode));
16895+ if (unlikely(!h_inode
16896+ || (h_inode->i_mode & S_IFMT)
16897+ != (inode->i_mode & S_IFMT)))
16898+ goto out;
1facf9fc 16899+
4a4d8108
AM
16900+ err = 0;
16901+ bindex = au_ibstart(inode);
16902+ br = au_sbr(sb, bindex);
86dc4139 16903+ err = h_permission(h_inode, mask, au_br_mnt(br), br->br_perm);
4a4d8108
AM
16904+ if (write_mask
16905+ && !err
16906+ && !special_file(h_inode->i_mode)) {
16907+ /* test whether the upper writable branch exists */
16908+ err = -EROFS;
16909+ for (; bindex >= 0; bindex--)
16910+ if (!au_br_rdonly(au_sbr(sb, bindex))) {
16911+ err = 0;
16912+ break;
16913+ }
16914+ }
16915+ goto out;
16916+ }
dece6358 16917+
4a4d8108 16918+ /* non-write to dir */
1308ab2a 16919+ err = 0;
4a4d8108
AM
16920+ bend = au_ibend(inode);
16921+ for (bindex = au_ibstart(inode); !err && bindex <= bend; bindex++) {
16922+ h_inode = au_h_iptr(inode, bindex);
16923+ if (h_inode) {
16924+ err = au_busy_or_stale();
16925+ if (unlikely(!S_ISDIR(h_inode->i_mode)))
16926+ break;
16927+
16928+ br = au_sbr(sb, bindex);
86dc4139 16929+ err = h_permission(h_inode, mask, au_br_mnt(br),
4a4d8108
AM
16930+ br->br_perm);
16931+ }
16932+ }
1308ab2a 16933+
4f0767ce 16934+out:
4a4d8108
AM
16935+ ii_read_unlock(inode);
16936+ si_read_unlock(sb);
1308ab2a 16937+ return err;
16938+}
16939+
4a4d8108 16940+/* ---------------------------------------------------------------------- */
1facf9fc 16941+
4a4d8108 16942+static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
b4510431 16943+ unsigned int flags)
4a4d8108
AM
16944+{
16945+ struct dentry *ret, *parent;
b752ccd1 16946+ struct inode *inode;
4a4d8108 16947+ struct super_block *sb;
1716fcea 16948+ int err, npositive;
dece6358 16949+
4a4d8108 16950+ IMustLock(dir);
1308ab2a 16951+
537831f9
AM
16952+ /* todo: support rcu-walk? */
16953+ ret = ERR_PTR(-ECHILD);
16954+ if (flags & LOOKUP_RCU)
16955+ goto out;
16956+
16957+ ret = ERR_PTR(-ENAMETOOLONG);
16958+ if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
16959+ goto out;
16960+
4a4d8108 16961+ sb = dir->i_sb;
7f207e10
AM
16962+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16963+ ret = ERR_PTR(err);
16964+ if (unlikely(err))
16965+ goto out;
16966+
4a4d8108
AM
16967+ err = au_di_init(dentry);
16968+ ret = ERR_PTR(err);
16969+ if (unlikely(err))
7f207e10 16970+ goto out_si;
1308ab2a 16971+
9dbd164d 16972+ inode = NULL;
027c5e7a 16973+ npositive = 0; /* suppress a warning */
4a4d8108
AM
16974+ parent = dentry->d_parent; /* dir inode is locked */
16975+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a
AM
16976+ err = au_alive_dir(parent);
16977+ if (!err)
16978+ err = au_digen_test(parent, au_sigen(sb));
16979+ if (!err) {
16980+ npositive = au_lkup_dentry(dentry, au_dbstart(parent),
537831f9 16981+ /*type*/0);
027c5e7a
AM
16982+ err = npositive;
16983+ }
4a4d8108 16984+ di_read_unlock(parent, AuLock_IR);
4a4d8108
AM
16985+ ret = ERR_PTR(err);
16986+ if (unlikely(err < 0))
16987+ goto out_unlock;
1308ab2a 16988+
4a4d8108 16989+ if (npositive) {
b752ccd1 16990+ inode = au_new_inode(dentry, /*must_new*/0);
4a4d8108 16991+ ret = (void *)inode;
1facf9fc 16992+ }
9dbd164d
AM
16993+ if (IS_ERR(inode)) {
16994+ inode = NULL;
4a4d8108 16995+ goto out_unlock;
9dbd164d 16996+ }
4a4d8108
AM
16997+
16998+ ret = d_splice_alias(inode, dentry);
537831f9
AM
16999+#if 0
17000+ if (unlikely(d_need_lookup(dentry))) {
17001+ spin_lock(&dentry->d_lock);
17002+ dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
17003+ spin_unlock(&dentry->d_lock);
17004+ } else
17005+#endif
7f207e10 17006+ if (unlikely(IS_ERR(ret) && inode)) {
4a4d8108 17007+ ii_write_unlock(inode);
7f207e10 17008+ iput(inode);
2dfbb274 17009+ inode = NULL;
7f207e10 17010+ }
1facf9fc 17011+
4f0767ce 17012+out_unlock:
4a4d8108 17013+ di_write_unlock(dentry);
2dfbb274 17014+ if (inode) {
1716fcea
AM
17015+ /* verbose coding for lock class name */
17016+ if (unlikely(S_ISLNK(inode->i_mode)))
17017+ au_rw_class(&au_di(dentry)->di_rwsem,
17018+ au_lc_key + AuLcSymlink_DIINFO);
17019+ else if (unlikely(S_ISDIR(inode->i_mode)))
17020+ au_rw_class(&au_di(dentry)->di_rwsem,
17021+ au_lc_key + AuLcDir_DIINFO);
17022+ else /* likely */
17023+ au_rw_class(&au_di(dentry)->di_rwsem,
17024+ au_lc_key + AuLcNonDir_DIINFO);
9dbd164d 17025+ }
7f207e10 17026+out_si:
4a4d8108 17027+ si_read_unlock(sb);
7f207e10 17028+out:
4a4d8108
AM
17029+ return ret;
17030+}
1facf9fc 17031+
4a4d8108 17032+/* ---------------------------------------------------------------------- */
1facf9fc 17033+
4a4d8108
AM
17034+static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
17035+ const unsigned char add_entry, aufs_bindex_t bcpup,
17036+ aufs_bindex_t bstart)
17037+{
17038+ int err;
17039+ struct dentry *h_parent;
17040+ struct inode *h_dir;
1facf9fc 17041+
027c5e7a 17042+ if (add_entry)
4a4d8108 17043+ IMustLock(parent->d_inode);
027c5e7a 17044+ else
4a4d8108
AM
17045+ di_write_lock_parent(parent);
17046+
17047+ err = 0;
17048+ if (!au_h_dptr(parent, bcpup)) {
c2b27bf2
AM
17049+ if (bstart > bcpup)
17050+ err = au_cpup_dirs(dentry, bcpup);
17051+ else if (bstart < bcpup)
4a4d8108
AM
17052+ err = au_cpdown_dirs(dentry, bcpup);
17053+ else
c2b27bf2 17054+ BUG();
4a4d8108 17055+ }
38d290e6 17056+ if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
4a4d8108
AM
17057+ h_parent = au_h_dptr(parent, bcpup);
17058+ h_dir = h_parent->d_inode;
17059+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
86dc4139
AM
17060+ err = au_lkup_neg(dentry, bcpup,
17061+ au_ftest_wrdir(add_entry, TMP_WHENTRY));
4a4d8108
AM
17062+ /* todo: no unlock here */
17063+ mutex_unlock(&h_dir->i_mutex);
027c5e7a
AM
17064+
17065+ AuDbg("bcpup %d\n", bcpup);
17066+ if (!err) {
17067+ if (!dentry->d_inode)
17068+ au_set_h_dptr(dentry, bstart, NULL);
4a4d8108
AM
17069+ au_update_dbrange(dentry, /*do_put_zero*/0);
17070+ }
1308ab2a 17071+ }
1facf9fc 17072+
4a4d8108
AM
17073+ if (!add_entry)
17074+ di_write_unlock(parent);
17075+ if (!err)
17076+ err = bcpup; /* success */
1308ab2a 17077+
027c5e7a 17078+ AuTraceErr(err);
4a4d8108
AM
17079+ return err;
17080+}
1facf9fc 17081+
4a4d8108
AM
17082+/*
17083+ * decide the branch and the parent dir where we will create a new entry.
17084+ * returns new bindex or an error.
17085+ * copyup the parent dir if needed.
17086+ */
17087+int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
17088+ struct au_wr_dir_args *args)
17089+{
17090+ int err;
392086de 17091+ unsigned int flags;
4a4d8108 17092+ aufs_bindex_t bcpup, bstart, src_bstart;
86dc4139
AM
17093+ const unsigned char add_entry
17094+ = au_ftest_wrdir(args->flags, ADD_ENTRY)
38d290e6
JR
17095+ | au_ftest_wrdir(args->flags, TMP_WHENTRY)
17096+ | au_ftest_wrdir(args->flags, TMPFILE);
4a4d8108
AM
17097+ struct super_block *sb;
17098+ struct dentry *parent;
17099+ struct au_sbinfo *sbinfo;
1facf9fc 17100+
4a4d8108
AM
17101+ sb = dentry->d_sb;
17102+ sbinfo = au_sbi(sb);
17103+ parent = dget_parent(dentry);
17104+ bstart = au_dbstart(dentry);
17105+ bcpup = bstart;
17106+ if (args->force_btgt < 0) {
17107+ if (src_dentry) {
17108+ src_bstart = au_dbstart(src_dentry);
17109+ if (src_bstart < bstart)
17110+ bcpup = src_bstart;
17111+ } else if (add_entry) {
392086de
AM
17112+ flags = 0;
17113+ if (au_ftest_wrdir(args->flags, ISDIR))
17114+ au_fset_wbr(flags, DIR);
17115+ err = AuWbrCreate(sbinfo, dentry, flags);
4a4d8108
AM
17116+ bcpup = err;
17117+ }
1facf9fc 17118+
4a4d8108
AM
17119+ if (bcpup < 0 || au_test_ro(sb, bcpup, dentry->d_inode)) {
17120+ if (add_entry)
17121+ err = AuWbrCopyup(sbinfo, dentry);
17122+ else {
17123+ if (!IS_ROOT(dentry)) {
17124+ di_read_lock_parent(parent, !AuLock_IR);
17125+ err = AuWbrCopyup(sbinfo, dentry);
17126+ di_read_unlock(parent, !AuLock_IR);
17127+ } else
17128+ err = AuWbrCopyup(sbinfo, dentry);
17129+ }
17130+ bcpup = err;
17131+ if (unlikely(err < 0))
17132+ goto out;
17133+ }
17134+ } else {
17135+ bcpup = args->force_btgt;
17136+ AuDebugOn(au_test_ro(sb, bcpup, dentry->d_inode));
1308ab2a 17137+ }
027c5e7a 17138+
4a4d8108
AM
17139+ AuDbg("bstart %d, bcpup %d\n", bstart, bcpup);
17140+ err = bcpup;
17141+ if (bcpup == bstart)
17142+ goto out; /* success */
4a4d8108
AM
17143+
17144+ /* copyup the new parent into the branch we process */
17145+ err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, bstart);
027c5e7a
AM
17146+ if (err >= 0) {
17147+ if (!dentry->d_inode) {
17148+ au_set_h_dptr(dentry, bstart, NULL);
17149+ au_set_dbstart(dentry, bcpup);
17150+ au_set_dbend(dentry, bcpup);
17151+ }
38d290e6
JR
17152+ AuDebugOn(add_entry
17153+ && !au_ftest_wrdir(args->flags, TMPFILE)
17154+ && !au_h_dptr(dentry, bcpup));
027c5e7a 17155+ }
86dc4139
AM
17156+
17157+out:
17158+ dput(parent);
17159+ return err;
17160+}
17161+
17162+/* ---------------------------------------------------------------------- */
17163+
17164+void au_pin_hdir_unlock(struct au_pin *p)
17165+{
17166+ if (p->hdir)
17167+ au_hn_imtx_unlock(p->hdir);
17168+}
17169+
17170+static int au_pin_hdir_lock(struct au_pin *p)
17171+{
17172+ int err;
17173+
17174+ err = 0;
17175+ if (!p->hdir)
17176+ goto out;
17177+
17178+ /* even if an error happens later, keep this lock */
17179+ au_hn_imtx_lock_nested(p->hdir, p->lsc_hi);
17180+
17181+ err = -EBUSY;
17182+ if (unlikely(p->hdir->hi_inode != p->h_parent->d_inode))
17183+ goto out;
17184+
17185+ err = 0;
17186+ if (p->h_dentry)
17187+ err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
17188+ p->h_parent, p->br);
17189+
17190+out:
17191+ return err;
17192+}
17193+
17194+int au_pin_hdir_relock(struct au_pin *p)
17195+{
17196+ int err, i;
17197+ struct inode *h_i;
17198+ struct dentry *h_d[] = {
17199+ p->h_dentry,
17200+ p->h_parent
17201+ };
17202+
17203+ err = au_pin_hdir_lock(p);
17204+ if (unlikely(err))
17205+ goto out;
17206+
17207+ for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
17208+ if (!h_d[i])
17209+ continue;
17210+ h_i = h_d[i]->d_inode;
17211+ if (h_i)
17212+ err = !h_i->i_nlink;
17213+ }
17214+
17215+out:
17216+ return err;
17217+}
17218+
17219+void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
17220+{
17221+#if defined(CONFIG_DEBUG_MUTEXES) || defined(CONFIG_SMP)
17222+ p->hdir->hi_inode->i_mutex.owner = task;
17223+#endif
17224+}
17225+
17226+void au_pin_hdir_acquire_nest(struct au_pin *p)
17227+{
17228+ if (p->hdir) {
17229+ mutex_acquire_nest(&p->hdir->hi_inode->i_mutex.dep_map,
17230+ p->lsc_hi, 0, NULL, _RET_IP_);
17231+ au_pin_hdir_set_owner(p, current);
17232+ }
dece6358 17233+}
1facf9fc 17234+
86dc4139
AM
17235+void au_pin_hdir_release(struct au_pin *p)
17236+{
17237+ if (p->hdir) {
17238+ au_pin_hdir_set_owner(p, p->task);
17239+ mutex_release(&p->hdir->hi_inode->i_mutex.dep_map, 1, _RET_IP_);
17240+ }
17241+}
1308ab2a 17242+
4a4d8108 17243+struct dentry *au_pinned_h_parent(struct au_pin *pin)
1308ab2a 17244+{
4a4d8108
AM
17245+ if (pin && pin->parent)
17246+ return au_h_dptr(pin->parent, pin->bindex);
17247+ return NULL;
dece6358 17248+}
1facf9fc 17249+
4a4d8108 17250+void au_unpin(struct au_pin *p)
dece6358 17251+{
86dc4139
AM
17252+ if (p->hdir)
17253+ au_pin_hdir_unlock(p);
e49829fe 17254+ if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
b4510431 17255+ vfsub_mnt_drop_write(p->h_mnt);
4a4d8108
AM
17256+ if (!p->hdir)
17257+ return;
1facf9fc 17258+
4a4d8108
AM
17259+ if (!au_ftest_pin(p->flags, DI_LOCKED))
17260+ di_read_unlock(p->parent, AuLock_IR);
17261+ iput(p->hdir->hi_inode);
17262+ dput(p->parent);
17263+ p->parent = NULL;
17264+ p->hdir = NULL;
17265+ p->h_mnt = NULL;
86dc4139 17266+ /* do not clear p->task */
4a4d8108 17267+}
1308ab2a 17268+
4a4d8108
AM
17269+int au_do_pin(struct au_pin *p)
17270+{
17271+ int err;
17272+ struct super_block *sb;
4a4d8108
AM
17273+ struct inode *h_dir;
17274+
17275+ err = 0;
17276+ sb = p->dentry->d_sb;
86dc4139 17277+ p->br = au_sbr(sb, p->bindex);
4a4d8108
AM
17278+ if (IS_ROOT(p->dentry)) {
17279+ if (au_ftest_pin(p->flags, MNT_WRITE)) {
86dc4139 17280+ p->h_mnt = au_br_mnt(p->br);
b4510431 17281+ err = vfsub_mnt_want_write(p->h_mnt);
4a4d8108
AM
17282+ if (unlikely(err)) {
17283+ au_fclr_pin(p->flags, MNT_WRITE);
17284+ goto out_err;
17285+ }
17286+ }
dece6358 17287+ goto out;
1facf9fc 17288+ }
17289+
86dc4139 17290+ p->h_dentry = NULL;
4a4d8108 17291+ if (p->bindex <= au_dbend(p->dentry))
86dc4139 17292+ p->h_dentry = au_h_dptr(p->dentry, p->bindex);
dece6358 17293+
4a4d8108
AM
17294+ p->parent = dget_parent(p->dentry);
17295+ if (!au_ftest_pin(p->flags, DI_LOCKED))
17296+ di_read_lock(p->parent, AuLock_IR, p->lsc_di);
dece6358 17297+
4a4d8108 17298+ h_dir = NULL;
86dc4139 17299+ p->h_parent = au_h_dptr(p->parent, p->bindex);
4a4d8108
AM
17300+ p->hdir = au_hi(p->parent->d_inode, p->bindex);
17301+ if (p->hdir)
17302+ h_dir = p->hdir->hi_inode;
dece6358 17303+
b752ccd1
AM
17304+ /*
17305+ * udba case, or
17306+ * if DI_LOCKED is not set, then p->parent may be different
17307+ * and h_parent can be NULL.
17308+ */
86dc4139 17309+ if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
e49829fe 17310+ err = -EBUSY;
4a4d8108
AM
17311+ if (!au_ftest_pin(p->flags, DI_LOCKED))
17312+ di_read_unlock(p->parent, AuLock_IR);
17313+ dput(p->parent);
17314+ p->parent = NULL;
17315+ goto out_err;
17316+ }
1308ab2a 17317+
4a4d8108 17318+ if (au_ftest_pin(p->flags, MNT_WRITE)) {
86dc4139 17319+ p->h_mnt = au_br_mnt(p->br);
b4510431 17320+ err = vfsub_mnt_want_write(p->h_mnt);
dece6358 17321+ if (unlikely(err)) {
4a4d8108 17322+ au_fclr_pin(p->flags, MNT_WRITE);
86dc4139
AM
17323+ if (!au_ftest_pin(p->flags, DI_LOCKED))
17324+ di_read_unlock(p->parent, AuLock_IR);
17325+ dput(p->parent);
17326+ p->parent = NULL;
17327+ goto out_err;
dece6358
AM
17328+ }
17329+ }
4a4d8108 17330+
86dc4139
AM
17331+ au_igrab(h_dir);
17332+ err = au_pin_hdir_lock(p);
17333+ if (!err)
17334+ goto out; /* success */
17335+
4f0767ce 17336+out_err:
4a4d8108
AM
17337+ pr_err("err %d\n", err);
17338+ err = au_busy_or_stale();
4f0767ce 17339+out:
1facf9fc 17340+ return err;
17341+}
17342+
4a4d8108
AM
17343+void au_pin_init(struct au_pin *p, struct dentry *dentry,
17344+ aufs_bindex_t bindex, int lsc_di, int lsc_hi,
17345+ unsigned int udba, unsigned char flags)
17346+{
17347+ p->dentry = dentry;
17348+ p->udba = udba;
17349+ p->lsc_di = lsc_di;
17350+ p->lsc_hi = lsc_hi;
17351+ p->flags = flags;
17352+ p->bindex = bindex;
17353+
17354+ p->parent = NULL;
17355+ p->hdir = NULL;
17356+ p->h_mnt = NULL;
86dc4139
AM
17357+
17358+ p->h_dentry = NULL;
17359+ p->h_parent = NULL;
17360+ p->br = NULL;
17361+ p->task = current;
4a4d8108
AM
17362+}
17363+
17364+int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
17365+ unsigned int udba, unsigned char flags)
17366+{
17367+ au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
17368+ udba, flags);
17369+ return au_do_pin(pin);
17370+}
17371+
dece6358
AM
17372+/* ---------------------------------------------------------------------- */
17373+
1308ab2a 17374+/*
4a4d8108
AM
17375+ * ->setattr() and ->getattr() are called in various cases.
17376+ * chmod, stat: dentry is revalidated.
17377+ * fchmod, fstat: file and dentry are not revalidated, additionally they may be
17378+ * unhashed.
17379+ * for ->setattr(), ia->ia_file is passed from ftruncate only.
1308ab2a 17380+ */
027c5e7a 17381+/* todo: consolidate with do_refresh() and simple_reval_dpath() */
4a4d8108 17382+static int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
1facf9fc 17383+{
4a4d8108
AM
17384+ int err;
17385+ struct inode *inode;
17386+ struct dentry *parent;
1facf9fc 17387+
1308ab2a 17388+ err = 0;
4a4d8108 17389+ inode = dentry->d_inode;
027c5e7a 17390+ if (au_digen_test(dentry, sigen)) {
4a4d8108
AM
17391+ parent = dget_parent(dentry);
17392+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a 17393+ err = au_refresh_dentry(dentry, parent);
4a4d8108
AM
17394+ di_read_unlock(parent, AuLock_IR);
17395+ dput(parent);
dece6358 17396+ }
1facf9fc 17397+
4a4d8108 17398+ AuTraceErr(err);
1308ab2a 17399+ return err;
17400+}
dece6358 17401+
4a4d8108
AM
17402+#define AuIcpup_DID_CPUP 1
17403+#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
7f207e10
AM
17404+#define au_fset_icpup(flags, name) \
17405+ do { (flags) |= AuIcpup_##name; } while (0)
17406+#define au_fclr_icpup(flags, name) \
17407+ do { (flags) &= ~AuIcpup_##name; } while (0)
1308ab2a 17408+
4a4d8108
AM
17409+struct au_icpup_args {
17410+ unsigned char flags;
17411+ unsigned char pin_flags;
17412+ aufs_bindex_t btgt;
17413+ unsigned int udba;
17414+ struct au_pin pin;
17415+ struct path h_path;
17416+ struct inode *h_inode;
17417+};
1308ab2a 17418+
4a4d8108
AM
17419+static int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
17420+ struct au_icpup_args *a)
1308ab2a 17421+{
17422+ int err;
4a4d8108 17423+ loff_t sz;
e49829fe 17424+ aufs_bindex_t bstart, ibstart;
4a4d8108
AM
17425+ struct dentry *hi_wh, *parent;
17426+ struct inode *inode;
4a4d8108
AM
17427+ struct au_wr_dir_args wr_dir_args = {
17428+ .force_btgt = -1,
17429+ .flags = 0
17430+ };
17431+
17432+ bstart = au_dbstart(dentry);
17433+ inode = dentry->d_inode;
17434+ if (S_ISDIR(inode->i_mode))
17435+ au_fset_wrdir(wr_dir_args.flags, ISDIR);
17436+ /* plink or hi_wh() case */
e49829fe 17437+ ibstart = au_ibstart(inode);
027c5e7a 17438+ if (bstart != ibstart && !au_test_ro(inode->i_sb, ibstart, inode))
e49829fe 17439+ wr_dir_args.force_btgt = ibstart;
4a4d8108
AM
17440+ err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
17441+ if (unlikely(err < 0))
17442+ goto out;
17443+ a->btgt = err;
17444+ if (err != bstart)
17445+ au_fset_icpup(a->flags, DID_CPUP);
17446+
17447+ err = 0;
17448+ a->pin_flags = AuPin_MNT_WRITE;
17449+ parent = NULL;
17450+ if (!IS_ROOT(dentry)) {
17451+ au_fset_pin(a->pin_flags, DI_LOCKED);
17452+ parent = dget_parent(dentry);
17453+ di_write_lock_parent(parent);
17454+ }
17455+
17456+ err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
17457+ if (unlikely(err))
17458+ goto out_parent;
17459+
17460+ a->h_path.dentry = au_h_dptr(dentry, bstart);
17461+ a->h_inode = a->h_path.dentry->d_inode;
17462+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
17463+ sz = -1;
17464+ if ((ia->ia_valid & ATTR_SIZE) && ia->ia_size < i_size_read(a->h_inode))
17465+ sz = ia->ia_size;
86dc4139 17466+ mutex_unlock(&a->h_inode->i_mutex);
4a4d8108 17467+
4a4d8108 17468+ hi_wh = NULL;
027c5e7a 17469+ if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
4a4d8108
AM
17470+ hi_wh = au_hi_wh(inode, a->btgt);
17471+ if (!hi_wh) {
c2b27bf2
AM
17472+ struct au_cp_generic cpg = {
17473+ .dentry = dentry,
17474+ .bdst = a->btgt,
17475+ .bsrc = -1,
17476+ .len = sz,
17477+ .pin = &a->pin
17478+ };
17479+ err = au_sio_cpup_wh(&cpg, /*file*/NULL);
4a4d8108
AM
17480+ if (unlikely(err))
17481+ goto out_unlock;
17482+ hi_wh = au_hi_wh(inode, a->btgt);
17483+ /* todo: revalidate hi_wh? */
17484+ }
17485+ }
17486+
17487+ if (parent) {
17488+ au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
17489+ di_downgrade_lock(parent, AuLock_IR);
17490+ dput(parent);
17491+ parent = NULL;
17492+ }
17493+ if (!au_ftest_icpup(a->flags, DID_CPUP))
17494+ goto out; /* success */
17495+
17496+ if (!d_unhashed(dentry)) {
c2b27bf2
AM
17497+ struct au_cp_generic cpg = {
17498+ .dentry = dentry,
17499+ .bdst = a->btgt,
17500+ .bsrc = bstart,
17501+ .len = sz,
17502+ .pin = &a->pin,
17503+ .flags = AuCpup_DTIME | AuCpup_HOPEN
17504+ };
17505+ err = au_sio_cpup_simple(&cpg);
4a4d8108
AM
17506+ if (!err)
17507+ a->h_path.dentry = au_h_dptr(dentry, a->btgt);
17508+ } else if (!hi_wh)
17509+ a->h_path.dentry = au_h_dptr(dentry, a->btgt);
17510+ else
17511+ a->h_path.dentry = hi_wh; /* do not dget here */
1308ab2a 17512+
4f0767ce 17513+out_unlock:
4a4d8108 17514+ a->h_inode = a->h_path.dentry->d_inode;
86dc4139 17515+ if (!err)
dece6358 17516+ goto out; /* success */
4a4d8108 17517+ au_unpin(&a->pin);
4f0767ce 17518+out_parent:
4a4d8108
AM
17519+ if (parent) {
17520+ di_write_unlock(parent);
17521+ dput(parent);
17522+ }
4f0767ce 17523+out:
86dc4139
AM
17524+ if (!err)
17525+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
1facf9fc 17526+ return err;
17527+}
17528+
4a4d8108 17529+static int aufs_setattr(struct dentry *dentry, struct iattr *ia)
1facf9fc 17530+{
4a4d8108 17531+ int err;
523b37e3 17532+ struct inode *inode, *delegated;
4a4d8108
AM
17533+ struct super_block *sb;
17534+ struct file *file;
17535+ struct au_icpup_args *a;
1facf9fc 17536+
4a4d8108
AM
17537+ inode = dentry->d_inode;
17538+ IMustLock(inode);
dece6358 17539+
4a4d8108
AM
17540+ err = -ENOMEM;
17541+ a = kzalloc(sizeof(*a), GFP_NOFS);
17542+ if (unlikely(!a))
17543+ goto out;
1facf9fc 17544+
4a4d8108
AM
17545+ if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
17546+ ia->ia_valid &= ~ATTR_MODE;
dece6358 17547+
4a4d8108
AM
17548+ file = NULL;
17549+ sb = dentry->d_sb;
e49829fe
JR
17550+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
17551+ if (unlikely(err))
17552+ goto out_kfree;
17553+
4a4d8108
AM
17554+ if (ia->ia_valid & ATTR_FILE) {
17555+ /* currently ftruncate(2) only */
17556+ AuDebugOn(!S_ISREG(inode->i_mode));
17557+ file = ia->ia_file;
17558+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
17559+ if (unlikely(err))
17560+ goto out_si;
17561+ ia->ia_file = au_hf_top(file);
17562+ a->udba = AuOpt_UDBA_NONE;
17563+ } else {
17564+ /* fchmod() doesn't pass ia_file */
17565+ a->udba = au_opt_udba(sb);
027c5e7a
AM
17566+ di_write_lock_child(dentry);
17567+ /* no d_unlinked(), to set UDBA_NONE for root */
4a4d8108
AM
17568+ if (d_unhashed(dentry))
17569+ a->udba = AuOpt_UDBA_NONE;
4a4d8108
AM
17570+ if (a->udba != AuOpt_UDBA_NONE) {
17571+ AuDebugOn(IS_ROOT(dentry));
17572+ err = au_reval_for_attr(dentry, au_sigen(sb));
17573+ if (unlikely(err))
17574+ goto out_dentry;
17575+ }
dece6358 17576+ }
dece6358 17577+
4a4d8108
AM
17578+ err = au_pin_and_icpup(dentry, ia, a);
17579+ if (unlikely(err < 0))
17580+ goto out_dentry;
17581+ if (au_ftest_icpup(a->flags, DID_CPUP)) {
17582+ ia->ia_file = NULL;
17583+ ia->ia_valid &= ~ATTR_FILE;
1308ab2a 17584+ }
dece6358 17585+
4a4d8108
AM
17586+ a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
17587+ if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
17588+ == (ATTR_MODE | ATTR_CTIME)) {
7eafdf33 17589+ err = security_path_chmod(&a->h_path, ia->ia_mode);
4a4d8108
AM
17590+ if (unlikely(err))
17591+ goto out_unlock;
17592+ } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
17593+ && (ia->ia_valid & ATTR_CTIME)) {
86dc4139 17594+ err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
4a4d8108
AM
17595+ if (unlikely(err))
17596+ goto out_unlock;
17597+ }
dece6358 17598+
4a4d8108
AM
17599+ if (ia->ia_valid & ATTR_SIZE) {
17600+ struct file *f;
1308ab2a 17601+
953406b4 17602+ if (ia->ia_size < i_size_read(inode))
4a4d8108 17603+ /* unmap only */
953406b4 17604+ truncate_setsize(inode, ia->ia_size);
1308ab2a 17605+
4a4d8108
AM
17606+ f = NULL;
17607+ if (ia->ia_valid & ATTR_FILE)
17608+ f = ia->ia_file;
17609+ mutex_unlock(&a->h_inode->i_mutex);
17610+ err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
17611+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
523b37e3
AM
17612+ } else {
17613+ delegated = NULL;
17614+ while (1) {
17615+ err = vfsub_notify_change(&a->h_path, ia, &delegated);
17616+ if (delegated) {
17617+ err = break_deleg_wait(&delegated);
17618+ if (!err)
17619+ continue;
17620+ }
17621+ break;
17622+ }
17623+ }
4a4d8108
AM
17624+ if (!err)
17625+ au_cpup_attr_changeable(inode);
1308ab2a 17626+
4f0767ce 17627+out_unlock:
4a4d8108
AM
17628+ mutex_unlock(&a->h_inode->i_mutex);
17629+ au_unpin(&a->pin);
027c5e7a
AM
17630+ if (unlikely(err))
17631+ au_update_dbstart(dentry);
4f0767ce 17632+out_dentry:
4a4d8108
AM
17633+ di_write_unlock(dentry);
17634+ if (file) {
17635+ fi_write_unlock(file);
17636+ ia->ia_file = file;
17637+ ia->ia_valid |= ATTR_FILE;
17638+ }
4f0767ce 17639+out_si:
4a4d8108 17640+ si_read_unlock(sb);
e49829fe 17641+out_kfree:
4a4d8108 17642+ kfree(a);
4f0767ce 17643+out:
4a4d8108
AM
17644+ AuTraceErr(err);
17645+ return err;
1facf9fc 17646+}
17647+
4a4d8108
AM
17648+static void au_refresh_iattr(struct inode *inode, struct kstat *st,
17649+ unsigned int nlink)
1facf9fc 17650+{
9dbd164d
AM
17651+ unsigned int n;
17652+
4a4d8108 17653+ inode->i_mode = st->mode;
86dc4139
AM
17654+ /* don't i_[ug]id_write() here */
17655+ inode->i_uid = st->uid;
17656+ inode->i_gid = st->gid;
4a4d8108
AM
17657+ inode->i_atime = st->atime;
17658+ inode->i_mtime = st->mtime;
17659+ inode->i_ctime = st->ctime;
1facf9fc 17660+
4a4d8108
AM
17661+ au_cpup_attr_nlink(inode, /*force*/0);
17662+ if (S_ISDIR(inode->i_mode)) {
9dbd164d
AM
17663+ n = inode->i_nlink;
17664+ n -= nlink;
17665+ n += st->nlink;
f6b6e03d 17666+ smp_mb(); /* for i_nlink */
7eafdf33 17667+ /* 0 can happen */
92d182d2 17668+ set_nlink(inode, n);
4a4d8108 17669+ }
1facf9fc 17670+
4a4d8108
AM
17671+ spin_lock(&inode->i_lock);
17672+ inode->i_blocks = st->blocks;
17673+ i_size_write(inode, st->size);
17674+ spin_unlock(&inode->i_lock);
1facf9fc 17675+}
17676+
4a4d8108
AM
17677+static int aufs_getattr(struct vfsmount *mnt __maybe_unused,
17678+ struct dentry *dentry, struct kstat *st)
1facf9fc 17679+{
4a4d8108
AM
17680+ int err;
17681+ unsigned int mnt_flags;
17682+ aufs_bindex_t bindex;
17683+ unsigned char udba_none, positive;
17684+ struct super_block *sb, *h_sb;
17685+ struct inode *inode;
c06a8ce3 17686+ struct path h_path;
1facf9fc 17687+
4a4d8108
AM
17688+ sb = dentry->d_sb;
17689+ inode = dentry->d_inode;
7f207e10
AM
17690+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
17691+ if (unlikely(err))
17692+ goto out;
4a4d8108
AM
17693+ mnt_flags = au_mntflags(sb);
17694+ udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
1facf9fc 17695+
4a4d8108 17696+ /* support fstat(2) */
027c5e7a 17697+ if (!d_unlinked(dentry) && !udba_none) {
4a4d8108 17698+ unsigned int sigen = au_sigen(sb);
027c5e7a
AM
17699+ err = au_digen_test(dentry, sigen);
17700+ if (!err) {
4a4d8108 17701+ di_read_lock_child(dentry, AuLock_IR);
027c5e7a
AM
17702+ err = au_dbrange_test(dentry);
17703+ if (unlikely(err))
17704+ goto out_unlock;
17705+ } else {
4a4d8108
AM
17706+ AuDebugOn(IS_ROOT(dentry));
17707+ di_write_lock_child(dentry);
027c5e7a
AM
17708+ err = au_dbrange_test(dentry);
17709+ if (!err)
17710+ err = au_reval_for_attr(dentry, sigen);
4a4d8108
AM
17711+ di_downgrade_lock(dentry, AuLock_IR);
17712+ if (unlikely(err))
7f207e10 17713+ goto out_unlock;
4a4d8108
AM
17714+ }
17715+ } else
17716+ di_read_lock_child(dentry, AuLock_IR);
1facf9fc 17717+
4a4d8108 17718+ bindex = au_ibstart(inode);
c06a8ce3
AM
17719+ h_path.mnt = au_sbr_mnt(sb, bindex);
17720+ h_sb = h_path.mnt->mnt_sb;
4a4d8108
AM
17721+ if (!au_test_fs_bad_iattr(h_sb) && udba_none)
17722+ goto out_fill; /* success */
1facf9fc 17723+
c06a8ce3 17724+ h_path.dentry = NULL;
4a4d8108 17725+ if (au_dbstart(dentry) == bindex)
c06a8ce3 17726+ h_path.dentry = dget(au_h_dptr(dentry, bindex));
4a4d8108 17727+ else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
c06a8ce3
AM
17728+ h_path.dentry = au_plink_lkup(inode, bindex);
17729+ if (IS_ERR(h_path.dentry))
4a4d8108
AM
17730+ goto out_fill; /* pretending success */
17731+ }
17732+ /* illegally overlapped or something */
c06a8ce3 17733+ if (unlikely(!h_path.dentry))
4a4d8108
AM
17734+ goto out_fill; /* pretending success */
17735+
c06a8ce3 17736+ positive = !!h_path.dentry->d_inode;
4a4d8108 17737+ if (positive)
c06a8ce3
AM
17738+ err = vfs_getattr(&h_path, st);
17739+ dput(h_path.dentry);
4a4d8108
AM
17740+ if (!err) {
17741+ if (positive)
c06a8ce3
AM
17742+ au_refresh_iattr(inode, st,
17743+ h_path.dentry->d_inode->i_nlink);
4a4d8108 17744+ goto out_fill; /* success */
1facf9fc 17745+ }
7f207e10
AM
17746+ AuTraceErr(err);
17747+ goto out_unlock;
4a4d8108 17748+
4f0767ce 17749+out_fill:
4a4d8108 17750+ generic_fillattr(inode, st);
7f207e10 17751+out_unlock:
4a4d8108
AM
17752+ di_read_unlock(dentry, AuLock_IR);
17753+ si_read_unlock(sb);
7f207e10
AM
17754+out:
17755+ AuTraceErr(err);
4a4d8108 17756+ return err;
1facf9fc 17757+}
17758+
17759+/* ---------------------------------------------------------------------- */
17760+
4a4d8108
AM
17761+static int h_readlink(struct dentry *dentry, int bindex, char __user *buf,
17762+ int bufsiz)
1facf9fc 17763+{
17764+ int err;
4a4d8108
AM
17765+ struct super_block *sb;
17766+ struct dentry *h_dentry;
1facf9fc 17767+
4a4d8108
AM
17768+ err = -EINVAL;
17769+ h_dentry = au_h_dptr(dentry, bindex);
17770+ if (unlikely(!h_dentry->d_inode->i_op->readlink))
17771+ goto out;
1facf9fc 17772+
4a4d8108
AM
17773+ err = security_inode_readlink(h_dentry);
17774+ if (unlikely(err))
dece6358 17775+ goto out;
1facf9fc 17776+
4a4d8108
AM
17777+ sb = dentry->d_sb;
17778+ if (!au_test_ro(sb, bindex, dentry->d_inode)) {
17779+ vfsub_touch_atime(au_sbr_mnt(sb, bindex), h_dentry);
17780+ fsstack_copy_attr_atime(dentry->d_inode, h_dentry->d_inode);
1facf9fc 17781+ }
4a4d8108 17782+ err = h_dentry->d_inode->i_op->readlink(h_dentry, buf, bufsiz);
1facf9fc 17783+
4f0767ce 17784+out:
4a4d8108
AM
17785+ return err;
17786+}
1facf9fc 17787+
4a4d8108
AM
17788+static int aufs_readlink(struct dentry *dentry, char __user *buf, int bufsiz)
17789+{
17790+ int err;
1facf9fc 17791+
027c5e7a
AM
17792+ err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
17793+ if (unlikely(err))
17794+ goto out;
17795+ err = au_d_hashed_positive(dentry);
17796+ if (!err)
17797+ err = h_readlink(dentry, au_dbstart(dentry), buf, bufsiz);
4a4d8108 17798+ aufs_read_unlock(dentry, AuLock_IR);
1facf9fc 17799+
027c5e7a 17800+out:
4a4d8108
AM
17801+ return err;
17802+}
1facf9fc 17803+
4a4d8108
AM
17804+static void *aufs_follow_link(struct dentry *dentry, struct nameidata *nd)
17805+{
17806+ int err;
4a4d8108 17807+ mm_segment_t old_fs;
b752ccd1
AM
17808+ union {
17809+ char *k;
17810+ char __user *u;
17811+ } buf;
1facf9fc 17812+
4a4d8108 17813+ err = -ENOMEM;
537831f9 17814+ buf.k = (void *)__get_free_page(GFP_NOFS);
b752ccd1 17815+ if (unlikely(!buf.k))
4a4d8108 17816+ goto out;
1facf9fc 17817+
027c5e7a
AM
17818+ err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
17819+ if (unlikely(err))
17820+ goto out_name;
17821+
17822+ err = au_d_hashed_positive(dentry);
17823+ if (!err) {
17824+ old_fs = get_fs();
17825+ set_fs(KERNEL_DS);
17826+ err = h_readlink(dentry, au_dbstart(dentry), buf.u, PATH_MAX);
17827+ set_fs(old_fs);
17828+ }
4a4d8108 17829+ aufs_read_unlock(dentry, AuLock_IR);
1facf9fc 17830+
4a4d8108 17831+ if (err >= 0) {
b752ccd1 17832+ buf.k[err] = 0;
4a4d8108 17833+ /* will be freed by put_link */
b752ccd1 17834+ nd_set_link(nd, buf.k);
4a4d8108 17835+ return NULL; /* success */
1308ab2a 17836+ }
1facf9fc 17837+
027c5e7a 17838+out_name:
537831f9 17839+ free_page((unsigned long)buf.k);
4f0767ce 17840+out:
4a4d8108
AM
17841+ AuTraceErr(err);
17842+ return ERR_PTR(err);
17843+}
1facf9fc 17844+
4a4d8108
AM
17845+static void aufs_put_link(struct dentry *dentry __maybe_unused,
17846+ struct nameidata *nd, void *cookie __maybe_unused)
17847+{
537831f9
AM
17848+ char *p;
17849+
17850+ p = nd_get_link(nd);
17851+ if (!IS_ERR_OR_NULL(p))
17852+ free_page((unsigned long)p);
4a4d8108 17853+}
1facf9fc 17854+
4a4d8108 17855+/* ---------------------------------------------------------------------- */
1facf9fc 17856+
0c3ec466 17857+static int aufs_update_time(struct inode *inode, struct timespec *ts, int flags)
4a4d8108 17858+{
0c3ec466
AM
17859+ int err;
17860+ struct super_block *sb;
17861+ struct inode *h_inode;
17862+
17863+ sb = inode->i_sb;
17864+ /* mmap_sem might be acquired already, cf. aufs_mmap() */
17865+ lockdep_off();
17866+ si_read_lock(sb, AuLock_FLUSH);
17867+ ii_write_lock_child(inode);
17868+ lockdep_on();
17869+ h_inode = au_h_iptr(inode, au_ibstart(inode));
17870+ err = vfsub_update_time(h_inode, ts, flags);
17871+ lockdep_off();
38d290e6
JR
17872+ if (!err)
17873+ au_cpup_attr_timesizes(inode);
0c3ec466
AM
17874+ ii_write_unlock(inode);
17875+ si_read_unlock(sb);
17876+ lockdep_on();
38d290e6
JR
17877+
17878+ if (!err && (flags & S_VERSION))
17879+ inode_inc_iversion(inode);
17880+
0c3ec466 17881+ return err;
4a4d8108 17882+}
1facf9fc 17883+
4a4d8108 17884+/* ---------------------------------------------------------------------- */
1308ab2a 17885+
4a4d8108
AM
17886+struct inode_operations aufs_symlink_iop = {
17887+ .permission = aufs_permission,
17888+ .setattr = aufs_setattr,
17889+ .getattr = aufs_getattr,
0c3ec466 17890+
4a4d8108
AM
17891+ .readlink = aufs_readlink,
17892+ .follow_link = aufs_follow_link,
0c3ec466
AM
17893+ .put_link = aufs_put_link,
17894+
17895+ /* .update_time = aufs_update_time */
4a4d8108
AM
17896+};
17897+
17898+struct inode_operations aufs_dir_iop = {
17899+ .create = aufs_create,
17900+ .lookup = aufs_lookup,
17901+ .link = aufs_link,
17902+ .unlink = aufs_unlink,
17903+ .symlink = aufs_symlink,
17904+ .mkdir = aufs_mkdir,
17905+ .rmdir = aufs_rmdir,
17906+ .mknod = aufs_mknod,
17907+ .rename = aufs_rename,
17908+
17909+ .permission = aufs_permission,
17910+ .setattr = aufs_setattr,
0c3ec466
AM
17911+ .getattr = aufs_getattr,
17912+
38d290e6 17913+ .update_time = aufs_update_time,
b4510431 17914+ /* no support for atomic_open() */
38d290e6
JR
17915+
17916+ .tmpfile = aufs_tmpfile
4a4d8108
AM
17917+};
17918+
17919+struct inode_operations aufs_iop = {
17920+ .permission = aufs_permission,
17921+ .setattr = aufs_setattr,
17922+ .getattr = aufs_getattr,
0c3ec466
AM
17923+
17924+ .update_time = aufs_update_time
4a4d8108 17925+};
7f207e10
AM
17926diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
17927--- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 17928+++ linux/fs/aufs/i_op_del.c 2014-06-24 00:14:06.782719306 +0200
523b37e3 17929@@ -0,0 +1,506 @@
1facf9fc 17930+/*
523b37e3 17931+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 17932+ *
17933+ * This program, aufs is free software; you can redistribute it and/or modify
17934+ * it under the terms of the GNU General Public License as published by
17935+ * the Free Software Foundation; either version 2 of the License, or
17936+ * (at your option) any later version.
dece6358
AM
17937+ *
17938+ * This program is distributed in the hope that it will be useful,
17939+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
17940+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17941+ * GNU General Public License for more details.
17942+ *
17943+ * You should have received a copy of the GNU General Public License
523b37e3 17944+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 17945+ */
17946+
17947+/*
4a4d8108 17948+ * inode operations (del entry)
1308ab2a 17949+ */
dece6358 17950+
1308ab2a 17951+#include "aufs.h"
dece6358 17952+
4a4d8108
AM
17953+/*
17954+ * decide if a new whiteout for @dentry is necessary or not.
17955+ * when it is necessary, prepare the parent dir for the upper branch whose
17956+ * branch index is @bcpup for creation. the actual creation of the whiteout will
17957+ * be done by caller.
17958+ * return value:
17959+ * 0: wh is unnecessary
17960+ * plus: wh is necessary
17961+ * minus: error
17962+ */
17963+int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
1308ab2a 17964+{
4a4d8108
AM
17965+ int need_wh, err;
17966+ aufs_bindex_t bstart;
17967+ struct super_block *sb;
dece6358 17968+
4a4d8108
AM
17969+ sb = dentry->d_sb;
17970+ bstart = au_dbstart(dentry);
17971+ if (*bcpup < 0) {
17972+ *bcpup = bstart;
17973+ if (au_test_ro(sb, bstart, dentry->d_inode)) {
17974+ err = AuWbrCopyup(au_sbi(sb), dentry);
17975+ *bcpup = err;
17976+ if (unlikely(err < 0))
17977+ goto out;
17978+ }
17979+ } else
17980+ AuDebugOn(bstart < *bcpup
17981+ || au_test_ro(sb, *bcpup, dentry->d_inode));
17982+ AuDbg("bcpup %d, bstart %d\n", *bcpup, bstart);
1308ab2a 17983+
4a4d8108
AM
17984+ if (*bcpup != bstart) {
17985+ err = au_cpup_dirs(dentry, *bcpup);
17986+ if (unlikely(err))
17987+ goto out;
17988+ need_wh = 1;
17989+ } else {
027c5e7a 17990+ struct au_dinfo *dinfo, *tmp;
4a4d8108 17991+
027c5e7a
AM
17992+ need_wh = -ENOMEM;
17993+ dinfo = au_di(dentry);
17994+ tmp = au_di_alloc(sb, AuLsc_DI_TMP);
17995+ if (tmp) {
17996+ au_di_cp(tmp, dinfo);
17997+ au_di_swap(tmp, dinfo);
17998+ /* returns the number of positive dentries */
537831f9 17999+ need_wh = au_lkup_dentry(dentry, bstart + 1, /*type*/0);
027c5e7a
AM
18000+ au_di_swap(tmp, dinfo);
18001+ au_rw_write_unlock(&tmp->di_rwsem);
18002+ au_di_free(tmp);
4a4d8108
AM
18003+ }
18004+ }
18005+ AuDbg("need_wh %d\n", need_wh);
18006+ err = need_wh;
18007+
4f0767ce 18008+out:
4a4d8108 18009+ return err;
1facf9fc 18010+}
18011+
4a4d8108
AM
18012+/*
18013+ * simple tests for the del-entry operations.
18014+ * following the checks in vfs, plus the parent-child relationship.
18015+ */
18016+int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
18017+ struct dentry *h_parent, int isdir)
1facf9fc 18018+{
4a4d8108
AM
18019+ int err;
18020+ umode_t h_mode;
18021+ struct dentry *h_dentry, *h_latest;
1308ab2a 18022+ struct inode *h_inode;
1facf9fc 18023+
4a4d8108
AM
18024+ h_dentry = au_h_dptr(dentry, bindex);
18025+ h_inode = h_dentry->d_inode;
18026+ if (dentry->d_inode) {
18027+ err = -ENOENT;
18028+ if (unlikely(!h_inode || !h_inode->i_nlink))
18029+ goto out;
1facf9fc 18030+
4a4d8108
AM
18031+ h_mode = h_inode->i_mode;
18032+ if (!isdir) {
18033+ err = -EISDIR;
18034+ if (unlikely(S_ISDIR(h_mode)))
18035+ goto out;
18036+ } else if (unlikely(!S_ISDIR(h_mode))) {
18037+ err = -ENOTDIR;
18038+ goto out;
18039+ }
18040+ } else {
18041+ /* rename(2) case */
18042+ err = -EIO;
18043+ if (unlikely(h_inode))
18044+ goto out;
18045+ }
1facf9fc 18046+
4a4d8108
AM
18047+ err = -ENOENT;
18048+ /* expected parent dir is locked */
18049+ if (unlikely(h_parent != h_dentry->d_parent))
18050+ goto out;
18051+ err = 0;
18052+
18053+ /*
18054+ * rmdir a dir may break the consistency on some filesystem.
18055+ * let's try heavy test.
18056+ */
18057+ err = -EACCES;
18058+ if (unlikely(au_test_h_perm(h_parent->d_inode, MAY_EXEC | MAY_WRITE)))
18059+ goto out;
18060+
18061+ h_latest = au_sio_lkup_one(&dentry->d_name, h_parent,
18062+ au_sbr(dentry->d_sb, bindex));
18063+ err = -EIO;
18064+ if (IS_ERR(h_latest))
18065+ goto out;
18066+ if (h_latest == h_dentry)
18067+ err = 0;
18068+ dput(h_latest);
18069+
4f0767ce 18070+out:
4a4d8108 18071+ return err;
1308ab2a 18072+}
1facf9fc 18073+
4a4d8108
AM
18074+/*
18075+ * decide the branch where we operate for @dentry. the branch index will be set
18076+ * @rbcpup. after diciding it, 'pin' it and store the timestamps of the parent
18077+ * dir for reverting.
18078+ * when a new whiteout is necessary, create it.
18079+ */
18080+static struct dentry*
18081+lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
18082+ struct au_dtime *dt, struct au_pin *pin)
1308ab2a 18083+{
4a4d8108
AM
18084+ struct dentry *wh_dentry;
18085+ struct super_block *sb;
18086+ struct path h_path;
18087+ int err, need_wh;
18088+ unsigned int udba;
18089+ aufs_bindex_t bcpup;
dece6358 18090+
4a4d8108
AM
18091+ need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
18092+ wh_dentry = ERR_PTR(need_wh);
18093+ if (unlikely(need_wh < 0))
18094+ goto out;
18095+
18096+ sb = dentry->d_sb;
18097+ udba = au_opt_udba(sb);
18098+ bcpup = *rbcpup;
18099+ err = au_pin(pin, dentry, bcpup, udba,
18100+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
18101+ wh_dentry = ERR_PTR(err);
18102+ if (unlikely(err))
18103+ goto out;
18104+
18105+ h_path.dentry = au_pinned_h_parent(pin);
18106+ if (udba != AuOpt_UDBA_NONE
18107+ && au_dbstart(dentry) == bcpup) {
18108+ err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
18109+ wh_dentry = ERR_PTR(err);
18110+ if (unlikely(err))
18111+ goto out_unpin;
18112+ }
18113+
18114+ h_path.mnt = au_sbr_mnt(sb, bcpup);
18115+ au_dtime_store(dt, au_pinned_parent(pin), &h_path);
18116+ wh_dentry = NULL;
18117+ if (!need_wh)
18118+ goto out; /* success, no need to create whiteout */
18119+
18120+ wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
18121+ if (IS_ERR(wh_dentry))
18122+ goto out_unpin;
18123+
18124+ /* returns with the parent is locked and wh_dentry is dget-ed */
18125+ goto out; /* success */
18126+
4f0767ce 18127+out_unpin:
4a4d8108 18128+ au_unpin(pin);
4f0767ce 18129+out:
4a4d8108 18130+ return wh_dentry;
1facf9fc 18131+}
18132+
4a4d8108
AM
18133+/*
18134+ * when removing a dir, rename it to a unique temporary whiteout-ed name first
18135+ * in order to be revertible and save time for removing many child whiteouts
18136+ * under the dir.
18137+ * returns 1 when there are too many child whiteout and caller should remove
18138+ * them asynchronously. returns 0 when the number of children is enough small to
18139+ * remove now or the branch fs is a remote fs.
18140+ * otherwise return an error.
18141+ */
18142+static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
18143+ struct au_nhash *whlist, struct inode *dir)
1facf9fc 18144+{
4a4d8108
AM
18145+ int rmdir_later, err, dirwh;
18146+ struct dentry *h_dentry;
18147+ struct super_block *sb;
18148+
18149+ sb = dentry->d_sb;
18150+ SiMustAnyLock(sb);
18151+ h_dentry = au_h_dptr(dentry, bindex);
18152+ err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
18153+ if (unlikely(err))
18154+ goto out;
18155+
18156+ /* stop monitoring */
18157+ au_hn_free(au_hi(dentry->d_inode, bindex));
18158+
18159+ if (!au_test_fs_remote(h_dentry->d_sb)) {
18160+ dirwh = au_sbi(sb)->si_dirwh;
18161+ rmdir_later = (dirwh <= 1);
18162+ if (!rmdir_later)
18163+ rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
18164+ dirwh);
18165+ if (rmdir_later)
18166+ return rmdir_later;
18167+ }
1facf9fc 18168+
4a4d8108
AM
18169+ err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
18170+ if (unlikely(err)) {
523b37e3
AM
18171+ AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
18172+ h_dentry, bindex, err);
4a4d8108
AM
18173+ err = 0;
18174+ }
dece6358 18175+
4f0767ce 18176+out:
4a4d8108
AM
18177+ AuTraceErr(err);
18178+ return err;
18179+}
1308ab2a 18180+
4a4d8108
AM
18181+/*
18182+ * final procedure for deleting a entry.
18183+ * maintain dentry and iattr.
18184+ */
18185+static void epilog(struct inode *dir, struct dentry *dentry,
18186+ aufs_bindex_t bindex)
18187+{
18188+ struct inode *inode;
1308ab2a 18189+
4a4d8108
AM
18190+ inode = dentry->d_inode;
18191+ d_drop(dentry);
18192+ inode->i_ctime = dir->i_ctime;
1308ab2a 18193+
4a4d8108
AM
18194+ if (au_ibstart(dir) == bindex)
18195+ au_cpup_attr_timesizes(dir);
18196+ dir->i_version++;
1facf9fc 18197+}
18198+
4a4d8108
AM
18199+/*
18200+ * when an error happened, remove the created whiteout and revert everything.
18201+ */
7f207e10
AM
18202+static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
18203+ aufs_bindex_t bwh, struct dentry *wh_dentry,
18204+ struct dentry *dentry, struct au_dtime *dt)
1facf9fc 18205+{
4a4d8108
AM
18206+ int rerr;
18207+ struct path h_path = {
18208+ .dentry = wh_dentry,
7f207e10 18209+ .mnt = au_sbr_mnt(dir->i_sb, bindex)
4a4d8108 18210+ };
dece6358 18211+
7f207e10 18212+ rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
4a4d8108
AM
18213+ if (!rerr) {
18214+ au_set_dbwh(dentry, bwh);
18215+ au_dtime_revert(dt);
18216+ return 0;
18217+ }
dece6358 18218+
523b37e3 18219+ AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
4a4d8108 18220+ return -EIO;
1facf9fc 18221+}
18222+
4a4d8108 18223+/* ---------------------------------------------------------------------- */
1facf9fc 18224+
4a4d8108 18225+int aufs_unlink(struct inode *dir, struct dentry *dentry)
1308ab2a 18226+{
4a4d8108
AM
18227+ int err;
18228+ aufs_bindex_t bwh, bindex, bstart;
523b37e3 18229+ struct inode *inode, *h_dir, *delegated;
4a4d8108 18230+ struct dentry *parent, *wh_dentry;
c2b27bf2
AM
18231+ /* to reuduce stack size */
18232+ struct {
18233+ struct au_dtime dt;
18234+ struct au_pin pin;
18235+ struct path h_path;
18236+ } *a;
1facf9fc 18237+
4a4d8108 18238+ IMustLock(dir);
027c5e7a 18239+
c2b27bf2
AM
18240+ err = -ENOMEM;
18241+ a = kmalloc(sizeof(*a), GFP_NOFS);
18242+ if (unlikely(!a))
18243+ goto out;
18244+
027c5e7a
AM
18245+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
18246+ if (unlikely(err))
c2b27bf2 18247+ goto out_free;
027c5e7a
AM
18248+ err = au_d_hashed_positive(dentry);
18249+ if (unlikely(err))
18250+ goto out_unlock;
4a4d8108 18251+ inode = dentry->d_inode;
4a4d8108 18252+ IMustLock(inode);
027c5e7a
AM
18253+ err = -EISDIR;
18254+ if (unlikely(S_ISDIR(inode->i_mode)))
18255+ goto out_unlock; /* possible? */
1facf9fc 18256+
4a4d8108
AM
18257+ bstart = au_dbstart(dentry);
18258+ bwh = au_dbwh(dentry);
18259+ bindex = -1;
027c5e7a
AM
18260+ parent = dentry->d_parent; /* dir inode is locked */
18261+ di_write_lock_parent(parent);
c2b27bf2
AM
18262+ wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
18263+ &a->pin);
4a4d8108
AM
18264+ err = PTR_ERR(wh_dentry);
18265+ if (IS_ERR(wh_dentry))
027c5e7a 18266+ goto out_parent;
1facf9fc 18267+
c2b27bf2
AM
18268+ a->h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
18269+ a->h_path.dentry = au_h_dptr(dentry, bstart);
18270+ dget(a->h_path.dentry);
4a4d8108 18271+ if (bindex == bstart) {
c2b27bf2 18272+ h_dir = au_pinned_h_dir(&a->pin);
523b37e3
AM
18273+ delegated = NULL;
18274+ err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
18275+ if (unlikely(err == -EWOULDBLOCK)) {
18276+ pr_warn("cannot retry for NFSv4 delegation"
18277+ " for an internal unlink\n");
18278+ iput(delegated);
18279+ }
4a4d8108
AM
18280+ } else {
18281+ /* dir inode is locked */
18282+ h_dir = wh_dentry->d_parent->d_inode;
18283+ IMustLock(h_dir);
18284+ err = 0;
18285+ }
dece6358 18286+
4a4d8108 18287+ if (!err) {
7f207e10 18288+ vfsub_drop_nlink(inode);
4a4d8108
AM
18289+ epilog(dir, dentry, bindex);
18290+
18291+ /* update target timestamps */
18292+ if (bindex == bstart) {
c2b27bf2
AM
18293+ vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
18294+ /*ignore*/
18295+ inode->i_ctime = a->h_path.dentry->d_inode->i_ctime;
4a4d8108
AM
18296+ } else
18297+ /* todo: this timestamp may be reverted later */
18298+ inode->i_ctime = h_dir->i_ctime;
027c5e7a 18299+ goto out_unpin; /* success */
1facf9fc 18300+ }
18301+
4a4d8108
AM
18302+ /* revert */
18303+ if (wh_dentry) {
18304+ int rerr;
18305+
c2b27bf2
AM
18306+ rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
18307+ &a->dt);
4a4d8108
AM
18308+ if (rerr)
18309+ err = rerr;
dece6358 18310+ }
1facf9fc 18311+
027c5e7a 18312+out_unpin:
c2b27bf2 18313+ au_unpin(&a->pin);
4a4d8108 18314+ dput(wh_dentry);
c2b27bf2 18315+ dput(a->h_path.dentry);
027c5e7a 18316+out_parent:
4a4d8108 18317+ di_write_unlock(parent);
027c5e7a 18318+out_unlock:
4a4d8108 18319+ aufs_read_unlock(dentry, AuLock_DW);
c2b27bf2
AM
18320+out_free:
18321+ kfree(a);
027c5e7a 18322+out:
4a4d8108 18323+ return err;
dece6358
AM
18324+}
18325+
4a4d8108 18326+int aufs_rmdir(struct inode *dir, struct dentry *dentry)
1308ab2a 18327+{
4a4d8108
AM
18328+ int err, rmdir_later;
18329+ aufs_bindex_t bwh, bindex, bstart;
4a4d8108
AM
18330+ struct inode *inode;
18331+ struct dentry *parent, *wh_dentry, *h_dentry;
18332+ struct au_whtmp_rmdir *args;
c2b27bf2
AM
18333+ /* to reuduce stack size */
18334+ struct {
18335+ struct au_dtime dt;
18336+ struct au_pin pin;
18337+ } *a;
1facf9fc 18338+
4a4d8108 18339+ IMustLock(dir);
027c5e7a 18340+
c2b27bf2
AM
18341+ err = -ENOMEM;
18342+ a = kmalloc(sizeof(*a), GFP_NOFS);
18343+ if (unlikely(!a))
18344+ goto out;
18345+
027c5e7a
AM
18346+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
18347+ if (unlikely(err))
c2b27bf2 18348+ goto out_free;
53392da6
AM
18349+ err = au_alive_dir(dentry);
18350+ if (unlikely(err))
027c5e7a 18351+ goto out_unlock;
53392da6 18352+ inode = dentry->d_inode;
4a4d8108 18353+ IMustLock(inode);
027c5e7a
AM
18354+ err = -ENOTDIR;
18355+ if (unlikely(!S_ISDIR(inode->i_mode)))
18356+ goto out_unlock; /* possible? */
dece6358 18357+
4a4d8108
AM
18358+ err = -ENOMEM;
18359+ args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
18360+ if (unlikely(!args))
18361+ goto out_unlock;
dece6358 18362+
4a4d8108
AM
18363+ parent = dentry->d_parent; /* dir inode is locked */
18364+ di_write_lock_parent(parent);
18365+ err = au_test_empty(dentry, &args->whlist);
18366+ if (unlikely(err))
027c5e7a 18367+ goto out_parent;
1facf9fc 18368+
4a4d8108
AM
18369+ bstart = au_dbstart(dentry);
18370+ bwh = au_dbwh(dentry);
18371+ bindex = -1;
c2b27bf2
AM
18372+ wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
18373+ &a->pin);
4a4d8108
AM
18374+ err = PTR_ERR(wh_dentry);
18375+ if (IS_ERR(wh_dentry))
027c5e7a 18376+ goto out_parent;
1facf9fc 18377+
4a4d8108
AM
18378+ h_dentry = au_h_dptr(dentry, bstart);
18379+ dget(h_dentry);
18380+ rmdir_later = 0;
18381+ if (bindex == bstart) {
18382+ err = renwh_and_rmdir(dentry, bstart, &args->whlist, dir);
18383+ if (err > 0) {
18384+ rmdir_later = err;
18385+ err = 0;
18386+ }
18387+ } else {
18388+ /* stop monitoring */
18389+ au_hn_free(au_hi(inode, bstart));
18390+
18391+ /* dir inode is locked */
18392+ IMustLock(wh_dentry->d_parent->d_inode);
1facf9fc 18393+ err = 0;
18394+ }
18395+
4a4d8108 18396+ if (!err) {
027c5e7a 18397+ vfsub_dead_dir(inode);
4a4d8108
AM
18398+ au_set_dbdiropq(dentry, -1);
18399+ epilog(dir, dentry, bindex);
1308ab2a 18400+
4a4d8108
AM
18401+ if (rmdir_later) {
18402+ au_whtmp_kick_rmdir(dir, bstart, h_dentry, args);
18403+ args = NULL;
18404+ }
1308ab2a 18405+
4a4d8108 18406+ goto out_unpin; /* success */
1facf9fc 18407+ }
18408+
4a4d8108
AM
18409+ /* revert */
18410+ AuLabel(revert);
18411+ if (wh_dentry) {
18412+ int rerr;
1308ab2a 18413+
c2b27bf2
AM
18414+ rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
18415+ &a->dt);
4a4d8108
AM
18416+ if (rerr)
18417+ err = rerr;
1facf9fc 18418+ }
18419+
4f0767ce 18420+out_unpin:
c2b27bf2 18421+ au_unpin(&a->pin);
4a4d8108
AM
18422+ dput(wh_dentry);
18423+ dput(h_dentry);
027c5e7a 18424+out_parent:
4a4d8108
AM
18425+ di_write_unlock(parent);
18426+ if (args)
18427+ au_whtmp_rmdir_free(args);
4f0767ce 18428+out_unlock:
4a4d8108 18429+ aufs_read_unlock(dentry, AuLock_DW);
c2b27bf2
AM
18430+out_free:
18431+ kfree(a);
4f0767ce 18432+out:
4a4d8108
AM
18433+ AuTraceErr(err);
18434+ return err;
dece6358 18435+}
7f207e10
AM
18436diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
18437--- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 18438+++ linux/fs/aufs/i_op_ren.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 18439@@ -0,0 +1,1032 @@
1facf9fc 18440+/*
523b37e3 18441+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 18442+ *
18443+ * This program, aufs is free software; you can redistribute it and/or modify
18444+ * it under the terms of the GNU General Public License as published by
18445+ * the Free Software Foundation; either version 2 of the License, or
18446+ * (at your option) any later version.
dece6358
AM
18447+ *
18448+ * This program is distributed in the hope that it will be useful,
18449+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
18450+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18451+ * GNU General Public License for more details.
18452+ *
18453+ * You should have received a copy of the GNU General Public License
523b37e3 18454+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 18455+ */
18456+
18457+/*
4a4d8108
AM
18458+ * inode operation (rename entry)
18459+ * todo: this is crazy monster
1facf9fc 18460+ */
18461+
18462+#include "aufs.h"
18463+
4a4d8108
AM
18464+enum { AuSRC, AuDST, AuSrcDst };
18465+enum { AuPARENT, AuCHILD, AuParentChild };
1facf9fc 18466+
4a4d8108
AM
18467+#define AuRen_ISDIR 1
18468+#define AuRen_ISSAMEDIR (1 << 1)
18469+#define AuRen_WHSRC (1 << 2)
18470+#define AuRen_WHDST (1 << 3)
18471+#define AuRen_MNT_WRITE (1 << 4)
18472+#define AuRen_DT_DSTDIR (1 << 5)
18473+#define AuRen_DIROPQ (1 << 6)
18474+#define AuRen_CPUP (1 << 7)
18475+#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
7f207e10
AM
18476+#define au_fset_ren(flags, name) \
18477+ do { (flags) |= AuRen_##name; } while (0)
18478+#define au_fclr_ren(flags, name) \
18479+ do { (flags) &= ~AuRen_##name; } while (0)
1facf9fc 18480+
4a4d8108
AM
18481+struct au_ren_args {
18482+ struct {
18483+ struct dentry *dentry, *h_dentry, *parent, *h_parent,
18484+ *wh_dentry;
18485+ struct inode *dir, *inode;
18486+ struct au_hinode *hdir;
18487+ struct au_dtime dt[AuParentChild];
18488+ aufs_bindex_t bstart;
18489+ } sd[AuSrcDst];
1facf9fc 18490+
4a4d8108
AM
18491+#define src_dentry sd[AuSRC].dentry
18492+#define src_dir sd[AuSRC].dir
18493+#define src_inode sd[AuSRC].inode
18494+#define src_h_dentry sd[AuSRC].h_dentry
18495+#define src_parent sd[AuSRC].parent
18496+#define src_h_parent sd[AuSRC].h_parent
18497+#define src_wh_dentry sd[AuSRC].wh_dentry
18498+#define src_hdir sd[AuSRC].hdir
18499+#define src_h_dir sd[AuSRC].hdir->hi_inode
18500+#define src_dt sd[AuSRC].dt
18501+#define src_bstart sd[AuSRC].bstart
1facf9fc 18502+
4a4d8108
AM
18503+#define dst_dentry sd[AuDST].dentry
18504+#define dst_dir sd[AuDST].dir
18505+#define dst_inode sd[AuDST].inode
18506+#define dst_h_dentry sd[AuDST].h_dentry
18507+#define dst_parent sd[AuDST].parent
18508+#define dst_h_parent sd[AuDST].h_parent
18509+#define dst_wh_dentry sd[AuDST].wh_dentry
18510+#define dst_hdir sd[AuDST].hdir
18511+#define dst_h_dir sd[AuDST].hdir->hi_inode
18512+#define dst_dt sd[AuDST].dt
18513+#define dst_bstart sd[AuDST].bstart
18514+
18515+ struct dentry *h_trap;
18516+ struct au_branch *br;
18517+ struct au_hinode *src_hinode;
18518+ struct path h_path;
18519+ struct au_nhash whlist;
027c5e7a 18520+ aufs_bindex_t btgt, src_bwh, src_bdiropq;
1facf9fc 18521+
1308ab2a 18522+ unsigned int flags;
1facf9fc 18523+
4a4d8108
AM
18524+ struct au_whtmp_rmdir *thargs;
18525+ struct dentry *h_dst;
18526+};
1308ab2a 18527+
4a4d8108 18528+/* ---------------------------------------------------------------------- */
1308ab2a 18529+
4a4d8108
AM
18530+/*
18531+ * functions for reverting.
18532+ * when an error happened in a single rename systemcall, we should revert
18533+ * everything as if nothing happend.
18534+ * we don't need to revert the copied-up/down the parent dir since they are
18535+ * harmless.
18536+ */
1facf9fc 18537+
4a4d8108
AM
18538+#define RevertFailure(fmt, ...) do { \
18539+ AuIOErr("revert failure: " fmt " (%d, %d)\n", \
18540+ ##__VA_ARGS__, err, rerr); \
18541+ err = -EIO; \
18542+} while (0)
1facf9fc 18543+
4a4d8108 18544+static void au_ren_rev_diropq(int err, struct au_ren_args *a)
1facf9fc 18545+{
4a4d8108 18546+ int rerr;
1facf9fc 18547+
4a4d8108
AM
18548+ au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
18549+ rerr = au_diropq_remove(a->src_dentry, a->btgt);
18550+ au_hn_imtx_unlock(a->src_hinode);
027c5e7a 18551+ au_set_dbdiropq(a->src_dentry, a->src_bdiropq);
4a4d8108 18552+ if (rerr)
523b37e3 18553+ RevertFailure("remove diropq %pd", a->src_dentry);
4a4d8108 18554+}
1facf9fc 18555+
4a4d8108
AM
18556+static void au_ren_rev_rename(int err, struct au_ren_args *a)
18557+{
18558+ int rerr;
523b37e3 18559+ struct inode *delegated;
1facf9fc 18560+
b4510431
AM
18561+ a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name,
18562+ a->src_h_parent);
4a4d8108
AM
18563+ rerr = PTR_ERR(a->h_path.dentry);
18564+ if (IS_ERR(a->h_path.dentry)) {
523b37e3 18565+ RevertFailure("lkup one %pd", a->src_dentry);
4a4d8108 18566+ return;
1facf9fc 18567+ }
18568+
523b37e3 18569+ delegated = NULL;
4a4d8108
AM
18570+ rerr = vfsub_rename(a->dst_h_dir,
18571+ au_h_dptr(a->src_dentry, a->btgt),
523b37e3
AM
18572+ a->src_h_dir, &a->h_path, &delegated);
18573+ if (unlikely(rerr == -EWOULDBLOCK)) {
18574+ pr_warn("cannot retry for NFSv4 delegation"
18575+ " for an internal rename\n");
18576+ iput(delegated);
18577+ }
4a4d8108
AM
18578+ d_drop(a->h_path.dentry);
18579+ dput(a->h_path.dentry);
18580+ /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
18581+ if (rerr)
523b37e3 18582+ RevertFailure("rename %pd", a->src_dentry);
1facf9fc 18583+}
18584+
4a4d8108 18585+static void au_ren_rev_cpup(int err, struct au_ren_args *a)
1facf9fc 18586+{
4a4d8108 18587+ int rerr;
1facf9fc 18588+
4a4d8108 18589+ a->h_path.dentry = a->dst_h_dentry;
523b37e3
AM
18590+ /* no delegation since it is just created */
18591+ rerr = vfsub_unlink(a->dst_h_dir, &a->h_path, /*delegated*/NULL,
18592+ /*force*/0);
4a4d8108
AM
18593+ au_set_h_dptr(a->src_dentry, a->btgt, NULL);
18594+ au_set_dbstart(a->src_dentry, a->src_bstart);
18595+ if (rerr)
523b37e3 18596+ RevertFailure("unlink %pd", a->dst_h_dentry);
1facf9fc 18597+}
18598+
4a4d8108 18599+static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
1facf9fc 18600+{
4a4d8108 18601+ int rerr;
523b37e3 18602+ struct inode *delegated;
dece6358 18603+
b4510431
AM
18604+ a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name,
18605+ a->dst_h_parent);
4a4d8108
AM
18606+ rerr = PTR_ERR(a->h_path.dentry);
18607+ if (IS_ERR(a->h_path.dentry)) {
523b37e3 18608+ RevertFailure("lkup one %pd", a->dst_dentry);
4a4d8108
AM
18609+ return;
18610+ }
18611+ if (a->h_path.dentry->d_inode) {
18612+ d_drop(a->h_path.dentry);
18613+ dput(a->h_path.dentry);
18614+ return;
dece6358
AM
18615+ }
18616+
523b37e3
AM
18617+ delegated = NULL;
18618+ rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
18619+ &delegated);
18620+ if (unlikely(rerr == -EWOULDBLOCK)) {
18621+ pr_warn("cannot retry for NFSv4 delegation"
18622+ " for an internal rename\n");
18623+ iput(delegated);
18624+ }
4a4d8108
AM
18625+ d_drop(a->h_path.dentry);
18626+ dput(a->h_path.dentry);
18627+ if (!rerr)
18628+ au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
18629+ else
523b37e3 18630+ RevertFailure("rename %pd", a->h_dst);
4a4d8108 18631+}
1308ab2a 18632+
4a4d8108
AM
18633+static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
18634+{
18635+ int rerr;
1308ab2a 18636+
4a4d8108
AM
18637+ a->h_path.dentry = a->src_wh_dentry;
18638+ rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
027c5e7a 18639+ au_set_dbwh(a->src_dentry, a->src_bwh);
4a4d8108 18640+ if (rerr)
523b37e3 18641+ RevertFailure("unlink %pd", a->src_wh_dentry);
4a4d8108 18642+}
4a4d8108 18643+#undef RevertFailure
1facf9fc 18644+
1308ab2a 18645+/* ---------------------------------------------------------------------- */
18646+
4a4d8108
AM
18647+/*
18648+ * when we have to copyup the renaming entry, do it with the rename-target name
18649+ * in order to minimize the cost (the later actual rename is unnecessary).
18650+ * otherwise rename it on the target branch.
18651+ */
18652+static int au_ren_or_cpup(struct au_ren_args *a)
1facf9fc 18653+{
dece6358 18654+ int err;
4a4d8108 18655+ struct dentry *d;
523b37e3 18656+ struct inode *delegated;
1facf9fc 18657+
4a4d8108
AM
18658+ d = a->src_dentry;
18659+ if (au_dbstart(d) == a->btgt) {
18660+ a->h_path.dentry = a->dst_h_dentry;
18661+ if (au_ftest_ren(a->flags, DIROPQ)
18662+ && au_dbdiropq(d) == a->btgt)
18663+ au_fclr_ren(a->flags, DIROPQ);
18664+ AuDebugOn(au_dbstart(d) != a->btgt);
523b37e3 18665+ delegated = NULL;
4a4d8108 18666+ err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
523b37e3
AM
18667+ a->dst_h_dir, &a->h_path, &delegated);
18668+ if (unlikely(err == -EWOULDBLOCK)) {
18669+ pr_warn("cannot retry for NFSv4 delegation"
18670+ " for an internal rename\n");
18671+ iput(delegated);
18672+ }
c2b27bf2 18673+ } else
86dc4139 18674+ BUG();
1308ab2a 18675+
027c5e7a
AM
18676+ if (!err && a->h_dst)
18677+ /* it will be set to dinfo later */
18678+ dget(a->h_dst);
1facf9fc 18679+
dece6358
AM
18680+ return err;
18681+}
1facf9fc 18682+
4a4d8108
AM
18683+/* cf. aufs_rmdir() */
18684+static int au_ren_del_whtmp(struct au_ren_args *a)
dece6358 18685+{
4a4d8108
AM
18686+ int err;
18687+ struct inode *dir;
1facf9fc 18688+
4a4d8108
AM
18689+ dir = a->dst_dir;
18690+ SiMustAnyLock(dir->i_sb);
18691+ if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
18692+ au_sbi(dir->i_sb)->si_dirwh)
18693+ || au_test_fs_remote(a->h_dst->d_sb)) {
18694+ err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
18695+ if (unlikely(err))
523b37e3
AM
18696+ pr_warn("failed removing whtmp dir %pd (%d), "
18697+ "ignored.\n", a->h_dst, err);
4a4d8108
AM
18698+ } else {
18699+ au_nhash_wh_free(&a->thargs->whlist);
18700+ a->thargs->whlist = a->whlist;
18701+ a->whlist.nh_num = 0;
18702+ au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
18703+ dput(a->h_dst);
18704+ a->thargs = NULL;
18705+ }
18706+
18707+ return 0;
1308ab2a 18708+}
1facf9fc 18709+
4a4d8108
AM
18710+/* make it 'opaque' dir. */
18711+static int au_ren_diropq(struct au_ren_args *a)
18712+{
18713+ int err;
18714+ struct dentry *diropq;
1facf9fc 18715+
4a4d8108 18716+ err = 0;
027c5e7a 18717+ a->src_bdiropq = au_dbdiropq(a->src_dentry);
4a4d8108
AM
18718+ a->src_hinode = au_hi(a->src_inode, a->btgt);
18719+ au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
18720+ diropq = au_diropq_create(a->src_dentry, a->btgt);
18721+ au_hn_imtx_unlock(a->src_hinode);
18722+ if (IS_ERR(diropq))
18723+ err = PTR_ERR(diropq);
18724+ dput(diropq);
1facf9fc 18725+
4a4d8108
AM
18726+ return err;
18727+}
1facf9fc 18728+
4a4d8108
AM
18729+static int do_rename(struct au_ren_args *a)
18730+{
18731+ int err;
18732+ struct dentry *d, *h_d;
1facf9fc 18733+
4a4d8108
AM
18734+ /* prepare workqueue args for asynchronous rmdir */
18735+ h_d = a->dst_h_dentry;
18736+ if (au_ftest_ren(a->flags, ISDIR) && h_d->d_inode) {
18737+ err = -ENOMEM;
18738+ a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb, GFP_NOFS);
18739+ if (unlikely(!a->thargs))
18740+ goto out;
18741+ a->h_dst = dget(h_d);
18742+ }
1facf9fc 18743+
4a4d8108
AM
18744+ /* create whiteout for src_dentry */
18745+ if (au_ftest_ren(a->flags, WHSRC)) {
027c5e7a
AM
18746+ a->src_bwh = au_dbwh(a->src_dentry);
18747+ AuDebugOn(a->src_bwh >= 0);
4a4d8108
AM
18748+ a->src_wh_dentry
18749+ = au_wh_create(a->src_dentry, a->btgt, a->src_h_parent);
18750+ err = PTR_ERR(a->src_wh_dentry);
18751+ if (IS_ERR(a->src_wh_dentry))
18752+ goto out_thargs;
18753+ }
1facf9fc 18754+
4a4d8108
AM
18755+ /* lookup whiteout for dentry */
18756+ if (au_ftest_ren(a->flags, WHDST)) {
18757+ h_d = au_wh_lkup(a->dst_h_parent, &a->dst_dentry->d_name,
18758+ a->br);
18759+ err = PTR_ERR(h_d);
18760+ if (IS_ERR(h_d))
18761+ goto out_whsrc;
18762+ if (!h_d->d_inode)
18763+ dput(h_d);
18764+ else
18765+ a->dst_wh_dentry = h_d;
18766+ }
1facf9fc 18767+
4a4d8108
AM
18768+ /* rename dentry to tmpwh */
18769+ if (a->thargs) {
18770+ err = au_whtmp_ren(a->dst_h_dentry, a->br);
18771+ if (unlikely(err))
18772+ goto out_whdst;
dece6358 18773+
4a4d8108
AM
18774+ d = a->dst_dentry;
18775+ au_set_h_dptr(d, a->btgt, NULL);
86dc4139 18776+ err = au_lkup_neg(d, a->btgt, /*wh*/0);
4a4d8108
AM
18777+ if (unlikely(err))
18778+ goto out_whtmp;
18779+ a->dst_h_dentry = au_h_dptr(d, a->btgt);
18780+ }
1facf9fc 18781+
c2b27bf2 18782+ BUG_ON(a->dst_h_dentry->d_inode && a->src_bstart != a->btgt);
1facf9fc 18783+
4a4d8108
AM
18784+ /* rename by vfs_rename or cpup */
18785+ d = a->dst_dentry;
18786+ if (au_ftest_ren(a->flags, ISDIR)
18787+ && (a->dst_wh_dentry
18788+ || au_dbdiropq(d) == a->btgt
18789+ /* hide the lower to keep xino */
18790+ || a->btgt < au_dbend(d)
18791+ || au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ)))
18792+ au_fset_ren(a->flags, DIROPQ);
18793+ err = au_ren_or_cpup(a);
18794+ if (unlikely(err))
18795+ /* leave the copied-up one */
18796+ goto out_whtmp;
1308ab2a 18797+
4a4d8108
AM
18798+ /* make dir opaque */
18799+ if (au_ftest_ren(a->flags, DIROPQ)) {
18800+ err = au_ren_diropq(a);
18801+ if (unlikely(err))
18802+ goto out_rename;
18803+ }
1308ab2a 18804+
4a4d8108
AM
18805+ /* update target timestamps */
18806+ AuDebugOn(au_dbstart(a->src_dentry) != a->btgt);
18807+ a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
18808+ vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
18809+ a->src_inode->i_ctime = a->h_path.dentry->d_inode->i_ctime;
1facf9fc 18810+
4a4d8108
AM
18811+ /* remove whiteout for dentry */
18812+ if (a->dst_wh_dentry) {
18813+ a->h_path.dentry = a->dst_wh_dentry;
18814+ err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
18815+ a->dst_dentry);
18816+ if (unlikely(err))
18817+ goto out_diropq;
18818+ }
1facf9fc 18819+
4a4d8108
AM
18820+ /* remove whtmp */
18821+ if (a->thargs)
18822+ au_ren_del_whtmp(a); /* ignore this error */
1308ab2a 18823+
4a4d8108
AM
18824+ err = 0;
18825+ goto out_success;
18826+
4f0767ce 18827+out_diropq:
4a4d8108
AM
18828+ if (au_ftest_ren(a->flags, DIROPQ))
18829+ au_ren_rev_diropq(err, a);
4f0767ce 18830+out_rename:
4a4d8108
AM
18831+ if (!au_ftest_ren(a->flags, CPUP))
18832+ au_ren_rev_rename(err, a);
18833+ else
18834+ au_ren_rev_cpup(err, a);
027c5e7a 18835+ dput(a->h_dst);
4f0767ce 18836+out_whtmp:
4a4d8108
AM
18837+ if (a->thargs)
18838+ au_ren_rev_whtmp(err, a);
4f0767ce 18839+out_whdst:
4a4d8108
AM
18840+ dput(a->dst_wh_dentry);
18841+ a->dst_wh_dentry = NULL;
4f0767ce 18842+out_whsrc:
4a4d8108
AM
18843+ if (a->src_wh_dentry)
18844+ au_ren_rev_whsrc(err, a);
4f0767ce 18845+out_success:
4a4d8108
AM
18846+ dput(a->src_wh_dentry);
18847+ dput(a->dst_wh_dentry);
4f0767ce 18848+out_thargs:
4a4d8108
AM
18849+ if (a->thargs) {
18850+ dput(a->h_dst);
18851+ au_whtmp_rmdir_free(a->thargs);
18852+ a->thargs = NULL;
18853+ }
4f0767ce 18854+out:
4a4d8108 18855+ return err;
dece6358 18856+}
1facf9fc 18857+
1308ab2a 18858+/* ---------------------------------------------------------------------- */
1facf9fc 18859+
4a4d8108
AM
18860+/*
18861+ * test if @dentry dir can be rename destination or not.
18862+ * success means, it is a logically empty dir.
18863+ */
18864+static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
1308ab2a 18865+{
4a4d8108 18866+ return au_test_empty(dentry, whlist);
1308ab2a 18867+}
1facf9fc 18868+
4a4d8108
AM
18869+/*
18870+ * test if @dentry dir can be rename source or not.
18871+ * if it can, return 0 and @children is filled.
18872+ * success means,
18873+ * - it is a logically empty dir.
18874+ * - or, it exists on writable branch and has no children including whiteouts
18875+ * on the lower branch.
18876+ */
18877+static int may_rename_srcdir(struct dentry *dentry, aufs_bindex_t btgt)
18878+{
18879+ int err;
18880+ unsigned int rdhash;
18881+ aufs_bindex_t bstart;
1facf9fc 18882+
4a4d8108
AM
18883+ bstart = au_dbstart(dentry);
18884+ if (bstart != btgt) {
18885+ struct au_nhash whlist;
dece6358 18886+
4a4d8108
AM
18887+ SiMustAnyLock(dentry->d_sb);
18888+ rdhash = au_sbi(dentry->d_sb)->si_rdhash;
18889+ if (!rdhash)
18890+ rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
18891+ dentry));
18892+ err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
18893+ if (unlikely(err))
18894+ goto out;
18895+ err = au_test_empty(dentry, &whlist);
18896+ au_nhash_wh_free(&whlist);
18897+ goto out;
18898+ }
dece6358 18899+
4a4d8108
AM
18900+ if (bstart == au_dbtaildir(dentry))
18901+ return 0; /* success */
dece6358 18902+
4a4d8108 18903+ err = au_test_empty_lower(dentry);
1facf9fc 18904+
4f0767ce 18905+out:
4a4d8108
AM
18906+ if (err == -ENOTEMPTY) {
18907+ AuWarn1("renaming dir who has child(ren) on multiple branches,"
18908+ " is not supported\n");
18909+ err = -EXDEV;
18910+ }
18911+ return err;
18912+}
1308ab2a 18913+
4a4d8108
AM
18914+/* side effect: sets whlist and h_dentry */
18915+static int au_ren_may_dir(struct au_ren_args *a)
1308ab2a 18916+{
4a4d8108
AM
18917+ int err;
18918+ unsigned int rdhash;
18919+ struct dentry *d;
1facf9fc 18920+
4a4d8108
AM
18921+ d = a->dst_dentry;
18922+ SiMustAnyLock(d->d_sb);
1facf9fc 18923+
4a4d8108
AM
18924+ err = 0;
18925+ if (au_ftest_ren(a->flags, ISDIR) && a->dst_inode) {
18926+ rdhash = au_sbi(d->d_sb)->si_rdhash;
18927+ if (!rdhash)
18928+ rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
18929+ err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
18930+ if (unlikely(err))
18931+ goto out;
1308ab2a 18932+
4a4d8108
AM
18933+ au_set_dbstart(d, a->dst_bstart);
18934+ err = may_rename_dstdir(d, &a->whlist);
18935+ au_set_dbstart(d, a->btgt);
18936+ }
18937+ a->dst_h_dentry = au_h_dptr(d, au_dbstart(d));
18938+ if (unlikely(err))
18939+ goto out;
18940+
18941+ d = a->src_dentry;
18942+ a->src_h_dentry = au_h_dptr(d, au_dbstart(d));
18943+ if (au_ftest_ren(a->flags, ISDIR)) {
18944+ err = may_rename_srcdir(d, a->btgt);
18945+ if (unlikely(err)) {
18946+ au_nhash_wh_free(&a->whlist);
18947+ a->whlist.nh_num = 0;
18948+ }
18949+ }
4f0767ce 18950+out:
4a4d8108 18951+ return err;
1facf9fc 18952+}
18953+
4a4d8108 18954+/* ---------------------------------------------------------------------- */
1facf9fc 18955+
4a4d8108
AM
18956+/*
18957+ * simple tests for rename.
18958+ * following the checks in vfs, plus the parent-child relationship.
18959+ */
18960+static int au_may_ren(struct au_ren_args *a)
18961+{
18962+ int err, isdir;
18963+ struct inode *h_inode;
1facf9fc 18964+
4a4d8108
AM
18965+ if (a->src_bstart == a->btgt) {
18966+ err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
18967+ au_ftest_ren(a->flags, ISDIR));
18968+ if (unlikely(err))
18969+ goto out;
18970+ err = -EINVAL;
18971+ if (unlikely(a->src_h_dentry == a->h_trap))
18972+ goto out;
18973+ }
1facf9fc 18974+
4a4d8108
AM
18975+ err = 0;
18976+ if (a->dst_bstart != a->btgt)
18977+ goto out;
1facf9fc 18978+
027c5e7a
AM
18979+ err = -ENOTEMPTY;
18980+ if (unlikely(a->dst_h_dentry == a->h_trap))
18981+ goto out;
18982+
4a4d8108
AM
18983+ err = -EIO;
18984+ h_inode = a->dst_h_dentry->d_inode;
18985+ isdir = !!au_ftest_ren(a->flags, ISDIR);
18986+ if (!a->dst_dentry->d_inode) {
18987+ if (unlikely(h_inode))
18988+ goto out;
18989+ err = au_may_add(a->dst_dentry, a->btgt, a->dst_h_parent,
18990+ isdir);
18991+ } else {
18992+ if (unlikely(!h_inode || !h_inode->i_nlink))
18993+ goto out;
18994+ err = au_may_del(a->dst_dentry, a->btgt, a->dst_h_parent,
18995+ isdir);
18996+ if (unlikely(err))
18997+ goto out;
4a4d8108 18998+ }
1facf9fc 18999+
4f0767ce 19000+out:
4a4d8108
AM
19001+ if (unlikely(err == -ENOENT || err == -EEXIST))
19002+ err = -EIO;
19003+ AuTraceErr(err);
19004+ return err;
19005+}
1facf9fc 19006+
1308ab2a 19007+/* ---------------------------------------------------------------------- */
1facf9fc 19008+
4a4d8108
AM
19009+/*
19010+ * locking order
19011+ * (VFS)
19012+ * - src_dir and dir by lock_rename()
19013+ * - inode if exitsts
19014+ * (aufs)
19015+ * - lock all
19016+ * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
19017+ * + si_read_lock
19018+ * + di_write_lock2_child()
19019+ * + di_write_lock_child()
19020+ * + ii_write_lock_child()
19021+ * + di_write_lock_child2()
19022+ * + ii_write_lock_child2()
19023+ * + src_parent and parent
19024+ * + di_write_lock_parent()
19025+ * + ii_write_lock_parent()
19026+ * + di_write_lock_parent2()
19027+ * + ii_write_lock_parent2()
19028+ * + lower src_dir and dir by vfsub_lock_rename()
19029+ * + verify the every relationships between child and parent. if any
19030+ * of them failed, unlock all and return -EBUSY.
19031+ */
19032+static void au_ren_unlock(struct au_ren_args *a)
1308ab2a 19033+{
4a4d8108
AM
19034+ vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
19035+ a->dst_h_parent, a->dst_hdir);
86dc4139
AM
19036+ if (au_ftest_ren(a->flags, MNT_WRITE))
19037+ vfsub_mnt_drop_write(au_br_mnt(a->br));
1308ab2a 19038+}
19039+
4a4d8108 19040+static int au_ren_lock(struct au_ren_args *a)
1308ab2a 19041+{
4a4d8108
AM
19042+ int err;
19043+ unsigned int udba;
1308ab2a 19044+
4a4d8108
AM
19045+ err = 0;
19046+ a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
19047+ a->src_hdir = au_hi(a->src_dir, a->btgt);
19048+ a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
19049+ a->dst_hdir = au_hi(a->dst_dir, a->btgt);
86dc4139
AM
19050+
19051+ err = vfsub_mnt_want_write(au_br_mnt(a->br));
19052+ if (unlikely(err))
19053+ goto out;
19054+ au_fset_ren(a->flags, MNT_WRITE);
4a4d8108
AM
19055+ a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
19056+ a->dst_h_parent, a->dst_hdir);
19057+ udba = au_opt_udba(a->src_dentry->d_sb);
19058+ if (unlikely(a->src_hdir->hi_inode != a->src_h_parent->d_inode
19059+ || a->dst_hdir->hi_inode != a->dst_h_parent->d_inode))
19060+ err = au_busy_or_stale();
19061+ if (!err && au_dbstart(a->src_dentry) == a->btgt)
19062+ err = au_h_verify(a->src_h_dentry, udba,
19063+ a->src_h_parent->d_inode, a->src_h_parent,
19064+ a->br);
19065+ if (!err && au_dbstart(a->dst_dentry) == a->btgt)
19066+ err = au_h_verify(a->dst_h_dentry, udba,
19067+ a->dst_h_parent->d_inode, a->dst_h_parent,
19068+ a->br);
86dc4139 19069+ if (!err)
4a4d8108 19070+ goto out; /* success */
4a4d8108
AM
19071+
19072+ err = au_busy_or_stale();
4a4d8108 19073+ au_ren_unlock(a);
86dc4139 19074+
4f0767ce 19075+out:
4a4d8108 19076+ return err;
1facf9fc 19077+}
19078+
19079+/* ---------------------------------------------------------------------- */
19080+
4a4d8108 19081+static void au_ren_refresh_dir(struct au_ren_args *a)
1facf9fc 19082+{
4a4d8108 19083+ struct inode *dir;
dece6358 19084+
4a4d8108
AM
19085+ dir = a->dst_dir;
19086+ dir->i_version++;
19087+ if (au_ftest_ren(a->flags, ISDIR)) {
19088+ /* is this updating defined in POSIX? */
19089+ au_cpup_attr_timesizes(a->src_inode);
19090+ au_cpup_attr_nlink(dir, /*force*/1);
4a4d8108 19091+ }
027c5e7a 19092+
4a4d8108
AM
19093+ if (au_ibstart(dir) == a->btgt)
19094+ au_cpup_attr_timesizes(dir);
dece6358 19095+
4a4d8108
AM
19096+ if (au_ftest_ren(a->flags, ISSAMEDIR))
19097+ return;
dece6358 19098+
4a4d8108
AM
19099+ dir = a->src_dir;
19100+ dir->i_version++;
19101+ if (au_ftest_ren(a->flags, ISDIR))
19102+ au_cpup_attr_nlink(dir, /*force*/1);
19103+ if (au_ibstart(dir) == a->btgt)
19104+ au_cpup_attr_timesizes(dir);
1facf9fc 19105+}
19106+
4a4d8108 19107+static void au_ren_refresh(struct au_ren_args *a)
1facf9fc 19108+{
4a4d8108
AM
19109+ aufs_bindex_t bend, bindex;
19110+ struct dentry *d, *h_d;
19111+ struct inode *i, *h_i;
19112+ struct super_block *sb;
dece6358 19113+
027c5e7a
AM
19114+ d = a->dst_dentry;
19115+ d_drop(d);
19116+ if (a->h_dst)
19117+ /* already dget-ed by au_ren_or_cpup() */
19118+ au_set_h_dptr(d, a->btgt, a->h_dst);
19119+
19120+ i = a->dst_inode;
19121+ if (i) {
19122+ if (!au_ftest_ren(a->flags, ISDIR))
19123+ vfsub_drop_nlink(i);
19124+ else {
19125+ vfsub_dead_dir(i);
19126+ au_cpup_attr_timesizes(i);
19127+ }
19128+ au_update_dbrange(d, /*do_put_zero*/1);
19129+ } else {
19130+ bend = a->btgt;
19131+ for (bindex = au_dbstart(d); bindex < bend; bindex++)
19132+ au_set_h_dptr(d, bindex, NULL);
19133+ bend = au_dbend(d);
19134+ for (bindex = a->btgt + 1; bindex <= bend; bindex++)
19135+ au_set_h_dptr(d, bindex, NULL);
19136+ au_update_dbrange(d, /*do_put_zero*/0);
19137+ }
19138+
4a4d8108
AM
19139+ d = a->src_dentry;
19140+ au_set_dbwh(d, -1);
19141+ bend = au_dbend(d);
19142+ for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
19143+ h_d = au_h_dptr(d, bindex);
19144+ if (h_d)
19145+ au_set_h_dptr(d, bindex, NULL);
19146+ }
19147+ au_set_dbend(d, a->btgt);
19148+
19149+ sb = d->d_sb;
19150+ i = a->src_inode;
19151+ if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
19152+ return; /* success */
19153+
19154+ bend = au_ibend(i);
19155+ for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
19156+ h_i = au_h_iptr(i, bindex);
19157+ if (h_i) {
19158+ au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
19159+ /* ignore this error */
19160+ au_set_h_iptr(i, bindex, NULL, 0);
19161+ }
19162+ }
19163+ au_set_ibend(i, a->btgt);
1308ab2a 19164+}
dece6358 19165+
4a4d8108
AM
19166+/* ---------------------------------------------------------------------- */
19167+
19168+/* mainly for link(2) and rename(2) */
19169+int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
1308ab2a 19170+{
4a4d8108
AM
19171+ aufs_bindex_t bdiropq, bwh;
19172+ struct dentry *parent;
19173+ struct au_branch *br;
19174+
19175+ parent = dentry->d_parent;
19176+ IMustLock(parent->d_inode); /* dir is locked */
19177+
19178+ bdiropq = au_dbdiropq(parent);
19179+ bwh = au_dbwh(dentry);
19180+ br = au_sbr(dentry->d_sb, btgt);
19181+ if (au_br_rdonly(br)
19182+ || (0 <= bdiropq && bdiropq < btgt)
19183+ || (0 <= bwh && bwh < btgt))
19184+ btgt = -1;
19185+
19186+ AuDbg("btgt %d\n", btgt);
19187+ return btgt;
1facf9fc 19188+}
19189+
4a4d8108
AM
19190+/* sets src_bstart, dst_bstart and btgt */
19191+static int au_ren_wbr(struct au_ren_args *a)
1facf9fc 19192+{
4a4d8108
AM
19193+ int err;
19194+ struct au_wr_dir_args wr_dir_args = {
19195+ /* .force_btgt = -1, */
19196+ .flags = AuWrDir_ADD_ENTRY
19197+ };
dece6358 19198+
4a4d8108
AM
19199+ a->src_bstart = au_dbstart(a->src_dentry);
19200+ a->dst_bstart = au_dbstart(a->dst_dentry);
19201+ if (au_ftest_ren(a->flags, ISDIR))
19202+ au_fset_wrdir(wr_dir_args.flags, ISDIR);
19203+ wr_dir_args.force_btgt = a->src_bstart;
19204+ if (a->dst_inode && a->dst_bstart < a->src_bstart)
19205+ wr_dir_args.force_btgt = a->dst_bstart;
19206+ wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
19207+ err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
19208+ a->btgt = err;
dece6358 19209+
4a4d8108 19210+ return err;
1facf9fc 19211+}
19212+
4a4d8108 19213+static void au_ren_dt(struct au_ren_args *a)
1facf9fc 19214+{
4a4d8108
AM
19215+ a->h_path.dentry = a->src_h_parent;
19216+ au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
19217+ if (!au_ftest_ren(a->flags, ISSAMEDIR)) {
19218+ a->h_path.dentry = a->dst_h_parent;
19219+ au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
19220+ }
1facf9fc 19221+
4a4d8108
AM
19222+ au_fclr_ren(a->flags, DT_DSTDIR);
19223+ if (!au_ftest_ren(a->flags, ISDIR))
19224+ return;
dece6358 19225+
4a4d8108
AM
19226+ a->h_path.dentry = a->src_h_dentry;
19227+ au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
19228+ if (a->dst_h_dentry->d_inode) {
19229+ au_fset_ren(a->flags, DT_DSTDIR);
19230+ a->h_path.dentry = a->dst_h_dentry;
19231+ au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
19232+ }
1308ab2a 19233+}
dece6358 19234+
4a4d8108 19235+static void au_ren_rev_dt(int err, struct au_ren_args *a)
1308ab2a 19236+{
4a4d8108
AM
19237+ struct dentry *h_d;
19238+ struct mutex *h_mtx;
19239+
19240+ au_dtime_revert(a->src_dt + AuPARENT);
19241+ if (!au_ftest_ren(a->flags, ISSAMEDIR))
19242+ au_dtime_revert(a->dst_dt + AuPARENT);
19243+
19244+ if (au_ftest_ren(a->flags, ISDIR) && err != -EIO) {
19245+ h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
19246+ h_mtx = &h_d->d_inode->i_mutex;
19247+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
19248+ au_dtime_revert(a->src_dt + AuCHILD);
19249+ mutex_unlock(h_mtx);
19250+
19251+ if (au_ftest_ren(a->flags, DT_DSTDIR)) {
19252+ h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
19253+ h_mtx = &h_d->d_inode->i_mutex;
19254+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
19255+ au_dtime_revert(a->dst_dt + AuCHILD);
19256+ mutex_unlock(h_mtx);
1facf9fc 19257+ }
19258+ }
19259+}
19260+
4a4d8108
AM
19261+/* ---------------------------------------------------------------------- */
19262+
19263+int aufs_rename(struct inode *_src_dir, struct dentry *_src_dentry,
19264+ struct inode *_dst_dir, struct dentry *_dst_dentry)
1facf9fc 19265+{
e49829fe 19266+ int err, flags;
4a4d8108
AM
19267+ /* reduce stack space */
19268+ struct au_ren_args *a;
19269+
523b37e3 19270+ AuDbg("%pd, %pd\n", _src_dentry, _dst_dentry);
4a4d8108
AM
19271+ IMustLock(_src_dir);
19272+ IMustLock(_dst_dir);
19273+
19274+ err = -ENOMEM;
19275+ BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
19276+ a = kzalloc(sizeof(*a), GFP_NOFS);
19277+ if (unlikely(!a))
19278+ goto out;
19279+
19280+ a->src_dir = _src_dir;
19281+ a->src_dentry = _src_dentry;
19282+ a->src_inode = a->src_dentry->d_inode;
19283+ a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
19284+ a->dst_dir = _dst_dir;
19285+ a->dst_dentry = _dst_dentry;
19286+ a->dst_inode = a->dst_dentry->d_inode;
19287+ a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
19288+ if (a->dst_inode) {
19289+ IMustLock(a->dst_inode);
19290+ au_igrab(a->dst_inode);
1facf9fc 19291+ }
1facf9fc 19292+
4a4d8108 19293+ err = -ENOTDIR;
027c5e7a 19294+ flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
4a4d8108
AM
19295+ if (S_ISDIR(a->src_inode->i_mode)) {
19296+ au_fset_ren(a->flags, ISDIR);
19297+ if (unlikely(a->dst_inode && !S_ISDIR(a->dst_inode->i_mode)))
19298+ goto out_free;
e49829fe
JR
19299+ err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
19300+ AuLock_DIR | flags);
4a4d8108 19301+ } else
e49829fe
JR
19302+ err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
19303+ flags);
19304+ if (unlikely(err))
19305+ goto out_free;
1facf9fc 19306+
027c5e7a
AM
19307+ err = au_d_hashed_positive(a->src_dentry);
19308+ if (unlikely(err))
19309+ goto out_unlock;
19310+ err = -ENOENT;
19311+ if (a->dst_inode) {
19312+ /*
19313+ * If it is a dir, VFS unhash dst_dentry before this
19314+ * function. It means we cannot rely upon d_unhashed().
19315+ */
19316+ if (unlikely(!a->dst_inode->i_nlink))
19317+ goto out_unlock;
19318+ if (!S_ISDIR(a->dst_inode->i_mode)) {
19319+ err = au_d_hashed_positive(a->dst_dentry);
19320+ if (unlikely(err))
19321+ goto out_unlock;
19322+ } else if (unlikely(IS_DEADDIR(a->dst_inode)))
19323+ goto out_unlock;
19324+ } else if (unlikely(d_unhashed(a->dst_dentry)))
19325+ goto out_unlock;
19326+
7eafdf33
AM
19327+ /*
19328+ * is it possible?
19329+ * yes, it happend (in linux-3.3-rcN) but I don't know why.
19330+ * there may exist a problem somewhere else.
19331+ */
19332+ err = -EINVAL;
19333+ if (unlikely(a->dst_parent->d_inode == a->src_dentry->d_inode))
19334+ goto out_unlock;
19335+
4a4d8108
AM
19336+ au_fset_ren(a->flags, ISSAMEDIR); /* temporary */
19337+ di_write_lock_parent(a->dst_parent);
1facf9fc 19338+
4a4d8108
AM
19339+ /* which branch we process */
19340+ err = au_ren_wbr(a);
19341+ if (unlikely(err < 0))
027c5e7a 19342+ goto out_parent;
4a4d8108 19343+ a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
86dc4139 19344+ a->h_path.mnt = au_br_mnt(a->br);
1facf9fc 19345+
4a4d8108
AM
19346+ /* are they available to be renamed */
19347+ err = au_ren_may_dir(a);
19348+ if (unlikely(err))
19349+ goto out_children;
1facf9fc 19350+
4a4d8108
AM
19351+ /* prepare the writable parent dir on the same branch */
19352+ if (a->dst_bstart == a->btgt) {
19353+ au_fset_ren(a->flags, WHDST);
19354+ } else {
19355+ err = au_cpup_dirs(a->dst_dentry, a->btgt);
19356+ if (unlikely(err))
19357+ goto out_children;
19358+ }
1facf9fc 19359+
4a4d8108
AM
19360+ if (a->src_dir != a->dst_dir) {
19361+ /*
19362+ * this temporary unlock is safe,
19363+ * because both dir->i_mutex are locked.
19364+ */
19365+ di_write_unlock(a->dst_parent);
19366+ di_write_lock_parent(a->src_parent);
19367+ err = au_wr_dir_need_wh(a->src_dentry,
19368+ au_ftest_ren(a->flags, ISDIR),
19369+ &a->btgt);
19370+ di_write_unlock(a->src_parent);
19371+ di_write_lock2_parent(a->src_parent, a->dst_parent, /*isdir*/1);
19372+ au_fclr_ren(a->flags, ISSAMEDIR);
19373+ } else
19374+ err = au_wr_dir_need_wh(a->src_dentry,
19375+ au_ftest_ren(a->flags, ISDIR),
19376+ &a->btgt);
19377+ if (unlikely(err < 0))
19378+ goto out_children;
19379+ if (err)
19380+ au_fset_ren(a->flags, WHSRC);
1facf9fc 19381+
86dc4139
AM
19382+ /* cpup src */
19383+ if (a->src_bstart != a->btgt) {
86dc4139
AM
19384+ struct au_pin pin;
19385+
19386+ err = au_pin(&pin, a->src_dentry, a->btgt,
19387+ au_opt_udba(a->src_dentry->d_sb),
19388+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
367653fa 19389+ if (!err) {
c2b27bf2
AM
19390+ struct au_cp_generic cpg = {
19391+ .dentry = a->src_dentry,
19392+ .bdst = a->btgt,
19393+ .bsrc = a->src_bstart,
19394+ .len = -1,
19395+ .pin = &pin,
19396+ .flags = AuCpup_DTIME | AuCpup_HOPEN
19397+ };
367653fa 19398+ AuDebugOn(au_dbstart(a->src_dentry) != a->src_bstart);
c2b27bf2 19399+ err = au_sio_cpup_simple(&cpg);
367653fa 19400+ au_unpin(&pin);
86dc4139 19401+ }
86dc4139
AM
19402+ if (unlikely(err))
19403+ goto out_children;
19404+ a->src_bstart = a->btgt;
19405+ a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
19406+ au_fset_ren(a->flags, WHSRC);
19407+ }
19408+
4a4d8108
AM
19409+ /* lock them all */
19410+ err = au_ren_lock(a);
19411+ if (unlikely(err))
86dc4139 19412+ /* leave the copied-up one */
4a4d8108 19413+ goto out_children;
1facf9fc 19414+
4a4d8108
AM
19415+ if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
19416+ err = au_may_ren(a);
19417+ else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
19418+ err = -ENAMETOOLONG;
19419+ if (unlikely(err))
19420+ goto out_hdir;
1facf9fc 19421+
4a4d8108
AM
19422+ /* store timestamps to be revertible */
19423+ au_ren_dt(a);
1facf9fc 19424+
4a4d8108
AM
19425+ /* here we go */
19426+ err = do_rename(a);
19427+ if (unlikely(err))
19428+ goto out_dt;
19429+
19430+ /* update dir attributes */
19431+ au_ren_refresh_dir(a);
19432+
19433+ /* dput/iput all lower dentries */
19434+ au_ren_refresh(a);
19435+
19436+ goto out_hdir; /* success */
19437+
4f0767ce 19438+out_dt:
4a4d8108 19439+ au_ren_rev_dt(err, a);
4f0767ce 19440+out_hdir:
4a4d8108 19441+ au_ren_unlock(a);
4f0767ce 19442+out_children:
4a4d8108 19443+ au_nhash_wh_free(&a->whlist);
027c5e7a
AM
19444+ if (err && a->dst_inode && a->dst_bstart != a->btgt) {
19445+ AuDbg("bstart %d, btgt %d\n", a->dst_bstart, a->btgt);
19446+ au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
19447+ au_set_dbstart(a->dst_dentry, a->dst_bstart);
4a4d8108 19448+ }
027c5e7a 19449+out_parent:
4a4d8108
AM
19450+ if (!err)
19451+ d_move(a->src_dentry, a->dst_dentry);
027c5e7a
AM
19452+ else {
19453+ au_update_dbstart(a->dst_dentry);
19454+ if (!a->dst_inode)
19455+ d_drop(a->dst_dentry);
19456+ }
4a4d8108
AM
19457+ if (au_ftest_ren(a->flags, ISSAMEDIR))
19458+ di_write_unlock(a->dst_parent);
19459+ else
19460+ di_write_unlock2(a->src_parent, a->dst_parent);
027c5e7a 19461+out_unlock:
4a4d8108 19462+ aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
4f0767ce 19463+out_free:
4a4d8108
AM
19464+ iput(a->dst_inode);
19465+ if (a->thargs)
19466+ au_whtmp_rmdir_free(a->thargs);
19467+ kfree(a);
4f0767ce 19468+out:
4a4d8108
AM
19469+ AuTraceErr(err);
19470+ return err;
1308ab2a 19471+}
7f207e10
AM
19472diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
19473--- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
19474+++ linux/fs/aufs/Kconfig 2014-06-24 00:14:06.776052275 +0200
19475@@ -0,0 +1,168 @@
4a4d8108
AM
19476+config AUFS_FS
19477+ tristate "Aufs (Advanced multi layered unification filesystem) support"
4a4d8108
AM
19478+ help
19479+ Aufs is a stackable unification filesystem such as Unionfs,
19480+ which unifies several directories and provides a merged single
19481+ directory.
19482+ In the early days, aufs was entirely re-designed and
19483+ re-implemented Unionfs Version 1.x series. Introducing many
19484+ original ideas, approaches and improvements, it becomes totally
19485+ different from Unionfs while keeping the basic features.
1facf9fc 19486+
4a4d8108
AM
19487+if AUFS_FS
19488+choice
19489+ prompt "Maximum number of branches"
19490+ default AUFS_BRANCH_MAX_127
19491+ help
19492+ Specifies the maximum number of branches (or member directories)
19493+ in a single aufs. The larger value consumes more system
19494+ resources and has a minor impact to performance.
19495+config AUFS_BRANCH_MAX_127
19496+ bool "127"
19497+ help
19498+ Specifies the maximum number of branches (or member directories)
19499+ in a single aufs. The larger value consumes more system
19500+ resources and has a minor impact to performance.
19501+config AUFS_BRANCH_MAX_511
19502+ bool "511"
19503+ help
19504+ Specifies the maximum number of branches (or member directories)
19505+ in a single aufs. The larger value consumes more system
19506+ resources and has a minor impact to performance.
19507+config AUFS_BRANCH_MAX_1023
19508+ bool "1023"
19509+ help
19510+ Specifies the maximum number of branches (or member directories)
19511+ in a single aufs. The larger value consumes more system
19512+ resources and has a minor impact to performance.
19513+config AUFS_BRANCH_MAX_32767
19514+ bool "32767"
19515+ help
19516+ Specifies the maximum number of branches (or member directories)
19517+ in a single aufs. The larger value consumes more system
19518+ resources and has a minor impact to performance.
19519+endchoice
1facf9fc 19520+
e49829fe
JR
19521+config AUFS_SBILIST
19522+ bool
19523+ depends on AUFS_MAGIC_SYSRQ || PROC_FS
19524+ default y
19525+ help
19526+ Automatic configuration for internal use.
19527+ When aufs supports Magic SysRq or /proc, enabled automatically.
19528+
4a4d8108
AM
19529+config AUFS_HNOTIFY
19530+ bool "Detect direct branch access (bypassing aufs)"
19531+ help
19532+ If you want to modify files on branches directly, eg. bypassing aufs,
19533+ and want aufs to detect the changes of them fully, then enable this
19534+ option and use 'udba=notify' mount option.
7f207e10 19535+ Currently there is only one available configuration, "fsnotify".
4a4d8108
AM
19536+ It will have a negative impact to the performance.
19537+ See detail in aufs.5.
dece6358 19538+
4a4d8108
AM
19539+choice
19540+ prompt "method" if AUFS_HNOTIFY
19541+ default AUFS_HFSNOTIFY
19542+config AUFS_HFSNOTIFY
19543+ bool "fsnotify"
19544+ select FSNOTIFY
4a4d8108 19545+endchoice
1facf9fc 19546+
4a4d8108
AM
19547+config AUFS_EXPORT
19548+ bool "NFS-exportable aufs"
2cbb1c4b 19549+ depends on EXPORTFS
4a4d8108
AM
19550+ help
19551+ If you want to export your mounted aufs via NFS, then enable this
19552+ option. There are several requirements for this configuration.
19553+ See detail in aufs.5.
1facf9fc 19554+
4a4d8108
AM
19555+config AUFS_INO_T_64
19556+ bool
19557+ depends on AUFS_EXPORT
19558+ depends on 64BIT && !(ALPHA || S390)
19559+ default y
19560+ help
19561+ Automatic configuration for internal use.
19562+ /* typedef unsigned long/int __kernel_ino_t */
19563+ /* alpha and s390x are int */
1facf9fc 19564+
4a4d8108
AM
19565+config AUFS_RDU
19566+ bool "Readdir in userspace"
19567+ help
19568+ Aufs has two methods to provide a merged view for a directory,
19569+ by a user-space library and by kernel-space natively. The latter
19570+ is always enabled but sometimes large and slow.
19571+ If you enable this option, install the library in aufs2-util
19572+ package, and set some environment variables for your readdir(3),
19573+ then the work will be handled in user-space which generally
19574+ shows better performance in most cases.
19575+ See detail in aufs.5.
1facf9fc 19576+
4a4d8108
AM
19577+config AUFS_SHWH
19578+ bool "Show whiteouts"
19579+ help
19580+ If you want to make the whiteouts in aufs visible, then enable
19581+ this option and specify 'shwh' mount option. Although it may
19582+ sounds like philosophy or something, but in technically it
19583+ simply shows the name of whiteout with keeping its behaviour.
1facf9fc 19584+
4a4d8108
AM
19585+config AUFS_BR_RAMFS
19586+ bool "Ramfs (initramfs/rootfs) as an aufs branch"
19587+ help
19588+ If you want to use ramfs as an aufs branch fs, then enable this
19589+ option. Generally tmpfs is recommended.
19590+ Aufs prohibited them to be a branch fs by default, because
19591+ initramfs becomes unusable after switch_root or something
19592+ generally. If you sets initramfs as an aufs branch and boot your
19593+ system by switch_root, you will meet a problem easily since the
19594+ files in initramfs may be inaccessible.
19595+ Unless you are going to use ramfs as an aufs branch fs without
19596+ switch_root or something, leave it N.
1facf9fc 19597+
4a4d8108
AM
19598+config AUFS_BR_FUSE
19599+ bool "Fuse fs as an aufs branch"
19600+ depends on FUSE_FS
19601+ select AUFS_POLL
19602+ help
19603+ If you want to use fuse-based userspace filesystem as an aufs
19604+ branch fs, then enable this option.
19605+ It implements the internal poll(2) operation which is
19606+ implemented by fuse only (curretnly).
1facf9fc 19607+
4a4d8108
AM
19608+config AUFS_POLL
19609+ bool
19610+ help
19611+ Automatic configuration for internal use.
1facf9fc 19612+
4a4d8108
AM
19613+config AUFS_BR_HFSPLUS
19614+ bool "Hfsplus as an aufs branch"
19615+ depends on HFSPLUS_FS
19616+ default y
19617+ help
19618+ If you want to use hfsplus fs as an aufs branch fs, then enable
19619+ this option. This option introduces a small overhead at
19620+ copying-up a file on hfsplus.
1facf9fc 19621+
4a4d8108
AM
19622+config AUFS_BDEV_LOOP
19623+ bool
19624+ depends on BLK_DEV_LOOP
19625+ default y
19626+ help
19627+ Automatic configuration for internal use.
19628+ Convert =[ym] into =y.
1308ab2a 19629+
4a4d8108
AM
19630+config AUFS_DEBUG
19631+ bool "Debug aufs"
19632+ help
19633+ Enable this to compile aufs internal debug code.
19634+ It will have a negative impact to the performance.
19635+
19636+config AUFS_MAGIC_SYSRQ
19637+ bool
19638+ depends on AUFS_DEBUG && MAGIC_SYSRQ
19639+ default y
19640+ help
19641+ Automatic configuration for internal use.
19642+ When aufs supports Magic SysRq, enabled automatically.
19643+endif
7f207e10
AM
19644diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
19645--- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 19646+++ linux/fs/aufs/loop.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 19647@@ -0,0 +1,145 @@
1facf9fc 19648+/*
523b37e3 19649+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 19650+ *
19651+ * This program, aufs is free software; you can redistribute it and/or modify
19652+ * it under the terms of the GNU General Public License as published by
19653+ * the Free Software Foundation; either version 2 of the License, or
19654+ * (at your option) any later version.
dece6358
AM
19655+ *
19656+ * This program is distributed in the hope that it will be useful,
19657+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
19658+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19659+ * GNU General Public License for more details.
19660+ *
19661+ * You should have received a copy of the GNU General Public License
523b37e3 19662+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 19663+ */
19664+
19665+/*
19666+ * support for loopback block device as a branch
19667+ */
19668+
1facf9fc 19669+#include "aufs.h"
19670+
392086de
AM
19671+/* added into drivers/block/loop.c */
19672+static struct file *(*backing_file_func)(struct super_block *sb);
19673+
1facf9fc 19674+/*
19675+ * test if two lower dentries have overlapping branches.
19676+ */
b752ccd1 19677+int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
1facf9fc 19678+{
b752ccd1 19679+ struct super_block *h_sb;
392086de
AM
19680+ struct file *backing_file;
19681+
19682+ if (unlikely(!backing_file_func)) {
19683+ /* don't load "loop" module here */
19684+ backing_file_func = symbol_get(loop_backing_file);
19685+ if (unlikely(!backing_file_func))
19686+ /* "loop" module is not loaded */
19687+ return 0;
19688+ }
1facf9fc 19689+
b752ccd1 19690+ h_sb = h_adding->d_sb;
392086de
AM
19691+ backing_file = backing_file_func(h_sb);
19692+ if (!backing_file)
1facf9fc 19693+ return 0;
19694+
392086de 19695+ h_adding = backing_file->f_dentry;
b752ccd1
AM
19696+ /*
19697+ * h_adding can be local NFS.
19698+ * in this case aufs cannot detect the loop.
19699+ */
19700+ if (unlikely(h_adding->d_sb == sb))
1facf9fc 19701+ return 1;
b752ccd1 19702+ return !!au_test_subdir(h_adding, sb->s_root);
1facf9fc 19703+}
19704+
19705+/* true if a kernel thread named 'loop[0-9].*' accesses a file */
19706+int au_test_loopback_kthread(void)
19707+{
b752ccd1
AM
19708+ int ret;
19709+ struct task_struct *tsk = current;
a2a7ad62 19710+ char c, comm[sizeof(tsk->comm)];
b752ccd1
AM
19711+
19712+ ret = 0;
19713+ if (tsk->flags & PF_KTHREAD) {
a2a7ad62
AM
19714+ get_task_comm(comm, tsk);
19715+ c = comm[4];
b752ccd1 19716+ ret = ('0' <= c && c <= '9'
a2a7ad62 19717+ && !strncmp(comm, "loop", 4));
b752ccd1 19718+ }
1facf9fc 19719+
b752ccd1 19720+ return ret;
1facf9fc 19721+}
87a755f4
AM
19722+
19723+/* ---------------------------------------------------------------------- */
19724+
19725+#define au_warn_loopback_step 16
19726+static int au_warn_loopback_nelem = au_warn_loopback_step;
19727+static unsigned long *au_warn_loopback_array;
19728+
19729+void au_warn_loopback(struct super_block *h_sb)
19730+{
19731+ int i, new_nelem;
19732+ unsigned long *a, magic;
19733+ static DEFINE_SPINLOCK(spin);
19734+
19735+ magic = h_sb->s_magic;
19736+ spin_lock(&spin);
19737+ a = au_warn_loopback_array;
19738+ for (i = 0; i < au_warn_loopback_nelem && *a; i++)
19739+ if (a[i] == magic) {
19740+ spin_unlock(&spin);
19741+ return;
19742+ }
19743+
19744+ /* h_sb is new to us, print it */
19745+ if (i < au_warn_loopback_nelem) {
19746+ a[i] = magic;
19747+ goto pr;
19748+ }
19749+
19750+ /* expand the array */
19751+ new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
19752+ a = au_kzrealloc(au_warn_loopback_array,
19753+ au_warn_loopback_nelem * sizeof(unsigned long),
19754+ new_nelem * sizeof(unsigned long), GFP_ATOMIC);
19755+ if (a) {
19756+ au_warn_loopback_nelem = new_nelem;
19757+ au_warn_loopback_array = a;
19758+ a[i] = magic;
19759+ goto pr;
19760+ }
19761+
19762+ spin_unlock(&spin);
19763+ AuWarn1("realloc failed, ignored\n");
19764+ return;
19765+
19766+pr:
19767+ spin_unlock(&spin);
0c3ec466
AM
19768+ pr_warn("you may want to try another patch for loopback file "
19769+ "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
87a755f4
AM
19770+}
19771+
19772+int au_loopback_init(void)
19773+{
19774+ int err;
19775+ struct super_block *sb __maybe_unused;
19776+
19777+ AuDebugOn(sizeof(sb->s_magic) != sizeof(unsigned long));
19778+
19779+ err = 0;
19780+ au_warn_loopback_array = kcalloc(au_warn_loopback_step,
19781+ sizeof(unsigned long), GFP_NOFS);
19782+ if (unlikely(!au_warn_loopback_array))
19783+ err = -ENOMEM;
19784+
19785+ return err;
19786+}
19787+
19788+void au_loopback_fin(void)
19789+{
392086de 19790+ symbol_put(loop_backing_file);
87a755f4
AM
19791+ kfree(au_warn_loopback_array);
19792+}
7f207e10
AM
19793diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
19794--- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 19795+++ linux/fs/aufs/loop.h 2014-06-24 00:14:06.786052822 +0200
523b37e3 19796@@ -0,0 +1,52 @@
1facf9fc 19797+/*
523b37e3 19798+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 19799+ *
19800+ * This program, aufs is free software; you can redistribute it and/or modify
19801+ * it under the terms of the GNU General Public License as published by
19802+ * the Free Software Foundation; either version 2 of the License, or
19803+ * (at your option) any later version.
dece6358
AM
19804+ *
19805+ * This program is distributed in the hope that it will be useful,
19806+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
19807+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19808+ * GNU General Public License for more details.
19809+ *
19810+ * You should have received a copy of the GNU General Public License
523b37e3 19811+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 19812+ */
19813+
19814+/*
19815+ * support for loopback mount as a branch
19816+ */
19817+
19818+#ifndef __AUFS_LOOP_H__
19819+#define __AUFS_LOOP_H__
19820+
19821+#ifdef __KERNEL__
19822+
dece6358
AM
19823+struct dentry;
19824+struct super_block;
1facf9fc 19825+
19826+#ifdef CONFIG_AUFS_BDEV_LOOP
392086de
AM
19827+/* drivers/block/loop.c */
19828+struct file *loop_backing_file(struct super_block *sb);
19829+
1facf9fc 19830+/* loop.c */
b752ccd1 19831+int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
1facf9fc 19832+int au_test_loopback_kthread(void);
87a755f4
AM
19833+void au_warn_loopback(struct super_block *h_sb);
19834+
19835+int au_loopback_init(void);
19836+void au_loopback_fin(void);
1facf9fc 19837+#else
4a4d8108 19838+AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
b752ccd1 19839+ struct dentry *h_adding)
4a4d8108 19840+AuStubInt0(au_test_loopback_kthread, void)
87a755f4
AM
19841+AuStubVoid(au_warn_loopback, struct super_block *h_sb)
19842+
19843+AuStubInt0(au_loopback_init, void)
19844+AuStubVoid(au_loopback_fin, void)
1facf9fc 19845+#endif /* BLK_DEV_LOOP */
19846+
19847+#endif /* __KERNEL__ */
19848+#endif /* __AUFS_LOOP_H__ */
7f207e10
AM
19849diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
19850--- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
38d290e6 19851+++ linux/fs/aufs/magic.mk 2014-06-24 00:14:06.786052822 +0200
4a4d8108 19852@@ -0,0 +1,54 @@
1facf9fc 19853+
19854+# defined in ${srctree}/fs/fuse/inode.c
19855+# tristate
19856+ifdef CONFIG_FUSE_FS
19857+ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
19858+endif
19859+
19860+# defined in ${srctree}/fs/ocfs2/ocfs2_fs.h
19861+# tristate
19862+ifdef CONFIG_OCFS2_FS
19863+ccflags-y += -DOCFS2_SUPER_MAGIC=0x7461636f
19864+endif
19865+
19866+# defined in ${srctree}/fs/ocfs2/dlm/userdlm.h
19867+# tristate
19868+ifdef CONFIG_OCFS2_FS_O2CB
19869+ccflags-y += -DDLMFS_MAGIC=0x76a9f425
19870+endif
19871+
1facf9fc 19872+# defined in ${srctree}/fs/cifs/cifsfs.c
19873+# tristate
19874+ifdef CONFIG_CIFS_FS
19875+ccflags-y += -DCIFS_MAGIC_NUMBER=0xFF534D42
19876+endif
19877+
19878+# defined in ${srctree}/fs/xfs/xfs_sb.h
19879+# tristate
19880+ifdef CONFIG_XFS_FS
19881+ccflags-y += -DXFS_SB_MAGIC=0x58465342
19882+endif
19883+
19884+# defined in ${srctree}/fs/configfs/mount.c
19885+# tristate
19886+ifdef CONFIG_CONFIGFS_FS
19887+ccflags-y += -DCONFIGFS_MAGIC=0x62656570
19888+endif
19889+
19890+# defined in ${srctree}/fs/9p/v9fs.h
19891+# tristate
19892+ifdef CONFIG_9P_FS
19893+ccflags-y += -DV9FS_MAGIC=0x01021997
19894+endif
19895+
19896+# defined in ${srctree}/fs/ubifs/ubifs.h
19897+# tristate
19898+ifdef CONFIG_UBIFS_FS
19899+ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
19900+endif
4a4d8108
AM
19901+
19902+# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
19903+# tristate
19904+ifdef CONFIG_HFSPLUS_FS
19905+ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
19906+endif
7f207e10
AM
19907diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
19908--- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
19909+++ linux/fs/aufs/Makefile 2014-06-24 00:14:06.776052275 +0200
19910@@ -0,0 +1,41 @@
4a4d8108
AM
19911+
19912+include ${src}/magic.mk
19913+ifeq (${CONFIG_AUFS_FS},m)
19914+include ${src}/conf.mk
19915+endif
19916+-include ${src}/priv_def.mk
19917+
19918+# cf. include/linux/kernel.h
19919+# enable pr_debug
19920+ccflags-y += -DDEBUG
f6c5ef8b
AM
19921+# sparse requires the full pathname
19922+ifdef M
523b37e3 19923+ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
f6c5ef8b 19924+else
523b37e3 19925+ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
f6c5ef8b 19926+endif
4a4d8108
AM
19927+
19928+obj-$(CONFIG_AUFS_FS) += aufs.o
19929+aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o \
19930+ wkq.o vfsub.o dcsub.o \
e49829fe 19931+ cpup.o whout.o wbr_policy.o \
4a4d8108
AM
19932+ dinfo.o dentry.o \
19933+ dynop.o \
19934+ finfo.o file.o f_op.o \
19935+ dir.o vdir.o \
19936+ iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
c2b27bf2 19937+ mvdown.o ioctl.o
4a4d8108
AM
19938+
19939+# all are boolean
e49829fe 19940+aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
4a4d8108
AM
19941+aufs-$(CONFIG_SYSFS) += sysfs.o
19942+aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
19943+aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
19944+aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
19945+aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
4a4d8108
AM
19946+aufs-$(CONFIG_AUFS_EXPORT) += export.o
19947+aufs-$(CONFIG_AUFS_POLL) += poll.o
19948+aufs-$(CONFIG_AUFS_RDU) += rdu.o
4a4d8108
AM
19949+aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
19950+aufs-$(CONFIG_AUFS_DEBUG) += debug.o
19951+aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
7f207e10
AM
19952diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
19953--- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 19954+++ linux/fs/aufs/module.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 19955@@ -0,0 +1,202 @@
1facf9fc 19956+/*
523b37e3 19957+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 19958+ *
19959+ * This program, aufs is free software; you can redistribute it and/or modify
19960+ * it under the terms of the GNU General Public License as published by
19961+ * the Free Software Foundation; either version 2 of the License, or
19962+ * (at your option) any later version.
dece6358
AM
19963+ *
19964+ * This program is distributed in the hope that it will be useful,
19965+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
19966+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19967+ * GNU General Public License for more details.
19968+ *
19969+ * You should have received a copy of the GNU General Public License
523b37e3 19970+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 19971+ */
19972+
19973+/*
19974+ * module global variables and operations
19975+ */
19976+
19977+#include <linux/module.h>
19978+#include <linux/seq_file.h>
19979+#include "aufs.h"
19980+
19981+void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp)
19982+{
19983+ if (new_sz <= nused)
19984+ return p;
19985+
19986+ p = krealloc(p, new_sz, gfp);
19987+ if (p)
19988+ memset(p + nused, 0, new_sz - nused);
19989+ return p;
19990+}
19991+
19992+/* ---------------------------------------------------------------------- */
19993+
19994+/*
19995+ * aufs caches
19996+ */
19997+struct kmem_cache *au_cachep[AuCache_Last];
19998+static int __init au_cache_init(void)
19999+{
4a4d8108 20000+ au_cachep[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
1facf9fc 20001+ if (au_cachep[AuCache_DINFO])
027c5e7a 20002+ /* SLAB_DESTROY_BY_RCU */
4a4d8108
AM
20003+ au_cachep[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
20004+ au_icntnr_init_once);
1facf9fc 20005+ if (au_cachep[AuCache_ICNTNR])
4a4d8108
AM
20006+ au_cachep[AuCache_FINFO] = AuCacheCtor(au_finfo,
20007+ au_fi_init_once);
1facf9fc 20008+ if (au_cachep[AuCache_FINFO])
20009+ au_cachep[AuCache_VDIR] = AuCache(au_vdir);
20010+ if (au_cachep[AuCache_VDIR])
20011+ au_cachep[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
20012+ if (au_cachep[AuCache_DEHSTR])
20013+ return 0;
20014+
20015+ return -ENOMEM;
20016+}
20017+
20018+static void au_cache_fin(void)
20019+{
20020+ int i;
4a4d8108 20021+
537831f9
AM
20022+ /*
20023+ * Make sure all delayed rcu free inodes are flushed before we
20024+ * destroy cache.
20025+ */
20026+ rcu_barrier();
20027+
7eafdf33
AM
20028+ /* excluding AuCache_HNOTIFY */
20029+ BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
20030+ for (i = 0; i < AuCache_HNOTIFY; i++)
1facf9fc 20031+ if (au_cachep[i]) {
20032+ kmem_cache_destroy(au_cachep[i]);
20033+ au_cachep[i] = NULL;
20034+ }
20035+}
20036+
20037+/* ---------------------------------------------------------------------- */
20038+
20039+int au_dir_roflags;
20040+
e49829fe 20041+#ifdef CONFIG_AUFS_SBILIST
1e00d052
AM
20042+/*
20043+ * iterate_supers_type() doesn't protect us from
20044+ * remounting (branch management)
20045+ */
e49829fe
JR
20046+struct au_splhead au_sbilist;
20047+#endif
20048+
9dbd164d
AM
20049+struct lock_class_key au_lc_key[AuLcKey_Last];
20050+
1facf9fc 20051+/*
20052+ * functions for module interface.
20053+ */
20054+MODULE_LICENSE("GPL");
20055+/* MODULE_LICENSE("GPL v2"); */
dece6358 20056+MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
1facf9fc 20057+MODULE_DESCRIPTION(AUFS_NAME
20058+ " -- Advanced multi layered unification filesystem");
20059+MODULE_VERSION(AUFS_VERSION);
c06a8ce3 20060+MODULE_ALIAS_FS(AUFS_NAME);
1facf9fc 20061+
1facf9fc 20062+/* this module parameter has no meaning when SYSFS is disabled */
20063+int sysaufs_brs = 1;
20064+MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
20065+module_param_named(brs, sysaufs_brs, int, S_IRUGO);
20066+
20067+/* ---------------------------------------------------------------------- */
20068+
20069+static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
20070+
20071+int au_seq_path(struct seq_file *seq, struct path *path)
20072+{
20073+ return seq_path(seq, path, au_esc_chars);
20074+}
20075+
20076+/* ---------------------------------------------------------------------- */
20077+
20078+static int __init aufs_init(void)
20079+{
20080+ int err, i;
20081+ char *p;
20082+
20083+ p = au_esc_chars;
20084+ for (i = 1; i <= ' '; i++)
20085+ *p++ = i;
20086+ *p++ = '\\';
20087+ *p++ = '\x7f';
20088+ *p = 0;
20089+
20090+ au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
20091+
e49829fe 20092+ au_sbilist_init();
1facf9fc 20093+ sysaufs_brs_init();
20094+ au_debug_init();
4a4d8108 20095+ au_dy_init();
1facf9fc 20096+ err = sysaufs_init();
20097+ if (unlikely(err))
20098+ goto out;
e49829fe 20099+ err = au_procfs_init();
4f0767ce 20100+ if (unlikely(err))
953406b4 20101+ goto out_sysaufs;
e49829fe
JR
20102+ err = au_wkq_init();
20103+ if (unlikely(err))
20104+ goto out_procfs;
87a755f4 20105+ err = au_loopback_init();
1facf9fc 20106+ if (unlikely(err))
20107+ goto out_wkq;
87a755f4
AM
20108+ err = au_hnotify_init();
20109+ if (unlikely(err))
20110+ goto out_loopback;
1facf9fc 20111+ err = au_sysrq_init();
20112+ if (unlikely(err))
20113+ goto out_hin;
20114+ err = au_cache_init();
20115+ if (unlikely(err))
20116+ goto out_sysrq;
20117+ err = register_filesystem(&aufs_fs_type);
20118+ if (unlikely(err))
20119+ goto out_cache;
4a4d8108
AM
20120+ /* since we define pr_fmt, call printk directly */
20121+ printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
1facf9fc 20122+ goto out; /* success */
20123+
4f0767ce 20124+out_cache:
1facf9fc 20125+ au_cache_fin();
4f0767ce 20126+out_sysrq:
1facf9fc 20127+ au_sysrq_fin();
4f0767ce 20128+out_hin:
4a4d8108 20129+ au_hnotify_fin();
87a755f4
AM
20130+out_loopback:
20131+ au_loopback_fin();
4f0767ce 20132+out_wkq:
1facf9fc 20133+ au_wkq_fin();
e49829fe
JR
20134+out_procfs:
20135+ au_procfs_fin();
4f0767ce 20136+out_sysaufs:
1facf9fc 20137+ sysaufs_fin();
4a4d8108 20138+ au_dy_fin();
4f0767ce 20139+out:
1facf9fc 20140+ return err;
20141+}
20142+
20143+static void __exit aufs_exit(void)
20144+{
20145+ unregister_filesystem(&aufs_fs_type);
20146+ au_cache_fin();
20147+ au_sysrq_fin();
4a4d8108 20148+ au_hnotify_fin();
87a755f4 20149+ au_loopback_fin();
1facf9fc 20150+ au_wkq_fin();
e49829fe 20151+ au_procfs_fin();
1facf9fc 20152+ sysaufs_fin();
4a4d8108 20153+ au_dy_fin();
1facf9fc 20154+}
20155+
20156+module_init(aufs_init);
20157+module_exit(aufs_exit);
7f207e10
AM
20158diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
20159--- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 20160+++ linux/fs/aufs/module.h 2014-06-24 00:14:06.786052822 +0200
523b37e3 20161@@ -0,0 +1,104 @@
1facf9fc 20162+/*
523b37e3 20163+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 20164+ *
20165+ * This program, aufs is free software; you can redistribute it and/or modify
20166+ * it under the terms of the GNU General Public License as published by
20167+ * the Free Software Foundation; either version 2 of the License, or
20168+ * (at your option) any later version.
dece6358
AM
20169+ *
20170+ * This program is distributed in the hope that it will be useful,
20171+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
20172+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20173+ * GNU General Public License for more details.
20174+ *
20175+ * You should have received a copy of the GNU General Public License
523b37e3 20176+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 20177+ */
20178+
20179+/*
20180+ * module initialization and module-global
20181+ */
20182+
20183+#ifndef __AUFS_MODULE_H__
20184+#define __AUFS_MODULE_H__
20185+
20186+#ifdef __KERNEL__
20187+
20188+#include <linux/slab.h>
20189+
dece6358
AM
20190+struct path;
20191+struct seq_file;
20192+
1facf9fc 20193+/* module parameters */
1facf9fc 20194+extern int sysaufs_brs;
20195+
20196+/* ---------------------------------------------------------------------- */
20197+
20198+extern int au_dir_roflags;
20199+
9dbd164d
AM
20200+enum {
20201+ AuLcNonDir_FIINFO,
20202+ AuLcNonDir_DIINFO,
20203+ AuLcNonDir_IIINFO,
20204+
20205+ AuLcDir_FIINFO,
20206+ AuLcDir_DIINFO,
20207+ AuLcDir_IIINFO,
20208+
20209+ AuLcSymlink_DIINFO,
20210+ AuLcSymlink_IIINFO,
20211+
20212+ AuLcKey_Last
20213+};
20214+extern struct lock_class_key au_lc_key[AuLcKey_Last];
20215+
1facf9fc 20216+void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp);
20217+int au_seq_path(struct seq_file *seq, struct path *path);
20218+
e49829fe
JR
20219+#ifdef CONFIG_PROC_FS
20220+/* procfs.c */
20221+int __init au_procfs_init(void);
20222+void au_procfs_fin(void);
20223+#else
20224+AuStubInt0(au_procfs_init, void);
20225+AuStubVoid(au_procfs_fin, void);
20226+#endif
20227+
4f0767ce
JR
20228+/* ---------------------------------------------------------------------- */
20229+
20230+/* kmem cache */
1facf9fc 20231+enum {
20232+ AuCache_DINFO,
20233+ AuCache_ICNTNR,
20234+ AuCache_FINFO,
20235+ AuCache_VDIR,
20236+ AuCache_DEHSTR,
7eafdf33 20237+ AuCache_HNOTIFY, /* must be last */
1facf9fc 20238+ AuCache_Last
20239+};
20240+
4a4d8108
AM
20241+#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
20242+#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
20243+#define AuCacheCtor(type, ctor) \
20244+ kmem_cache_create(#type, sizeof(struct type), \
20245+ __alignof__(struct type), AuCacheFlags, ctor)
1facf9fc 20246+
20247+extern struct kmem_cache *au_cachep[];
20248+
20249+#define AuCacheFuncs(name, index) \
4a4d8108 20250+static inline struct au_##name *au_cache_alloc_##name(void) \
1facf9fc 20251+{ return kmem_cache_alloc(au_cachep[AuCache_##index], GFP_NOFS); } \
4a4d8108 20252+static inline void au_cache_free_##name(struct au_##name *p) \
1facf9fc 20253+{ kmem_cache_free(au_cachep[AuCache_##index], p); }
20254+
20255+AuCacheFuncs(dinfo, DINFO);
20256+AuCacheFuncs(icntnr, ICNTNR);
20257+AuCacheFuncs(finfo, FINFO);
20258+AuCacheFuncs(vdir, VDIR);
4a4d8108
AM
20259+AuCacheFuncs(vdir_dehstr, DEHSTR);
20260+#ifdef CONFIG_AUFS_HNOTIFY
20261+AuCacheFuncs(hnotify, HNOTIFY);
20262+#endif
1facf9fc 20263+
4a4d8108
AM
20264+#endif /* __KERNEL__ */
20265+#endif /* __AUFS_MODULE_H__ */
c2b27bf2
AM
20266diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
20267--- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 20268+++ linux/fs/aufs/mvdown.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 20269@@ -0,0 +1,627 @@
c2b27bf2 20270+/*
523b37e3 20271+ * Copyright (C) 2011-2014 Junjiro R. Okajima
c2b27bf2
AM
20272+ *
20273+ * This program, aufs is free software; you can redistribute it and/or modify
20274+ * it under the terms of the GNU General Public License as published by
20275+ * the Free Software Foundation; either version 2 of the License, or
20276+ * (at your option) any later version.
20277+ *
20278+ * This program is distributed in the hope that it will be useful,
20279+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
20280+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20281+ * GNU General Public License for more details.
20282+ *
20283+ * You should have received a copy of the GNU General Public License
523b37e3
AM
20284+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
20285+ */
20286+
20287+/*
20288+ * move-down, opposite of copy-up
c2b27bf2
AM
20289+ */
20290+
20291+#include "aufs.h"
20292+
c2b27bf2
AM
20293+struct au_mvd_args {
20294+ struct {
c2b27bf2
AM
20295+ struct super_block *h_sb;
20296+ struct dentry *h_parent;
20297+ struct au_hinode *hdir;
392086de 20298+ struct inode *h_dir, *h_inode;
c2b27bf2
AM
20299+ } info[AUFS_MVDOWN_NARRAY];
20300+
20301+ struct aufs_mvdown mvdown;
20302+ struct dentry *dentry, *parent;
20303+ struct inode *inode, *dir;
20304+ struct super_block *sb;
20305+ aufs_bindex_t bopq, bwh, bfound;
20306+ unsigned char rename_lock;
20307+ struct au_pin pin;
20308+};
20309+
392086de
AM
20310+#define mvd_errno mvdown.au_errno
20311+#define mvd_bsrc mvdown.a[AUFS_MVDOWN_UPPER].bindex
20312+#define mvd_src_brid mvdown.a[AUFS_MVDOWN_UPPER].brid
20313+#define mvd_bdst mvdown.a[AUFS_MVDOWN_LOWER].bindex
20314+#define mvd_dst_brid mvdown.a[AUFS_MVDOWN_LOWER].brid
c2b27bf2 20315+
392086de
AM
20316+#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
20317+#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
20318+#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
20319+#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
20320+#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
20321+
20322+#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
20323+#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
20324+#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
20325+#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
20326+#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
c2b27bf2
AM
20327+
20328+#define AU_MVD_PR(flag, ...) do { \
20329+ if (flag) \
20330+ pr_err(__VA_ARGS__); \
20331+ } while (0)
20332+
20333+/* make the parent dir on bdst */
392086de 20334+static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20335+{
20336+ int err;
20337+
20338+ err = 0;
20339+ a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
20340+ a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
20341+ a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
20342+ a->mvd_h_dst_parent = NULL;
20343+ if (au_dbend(a->parent) >= a->mvd_bdst)
20344+ a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
20345+ if (!a->mvd_h_dst_parent) {
20346+ err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
20347+ if (unlikely(err)) {
392086de 20348+ AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
c2b27bf2
AM
20349+ goto out;
20350+ }
20351+ a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
20352+ }
20353+
20354+out:
20355+ AuTraceErr(err);
20356+ return err;
20357+}
20358+
20359+/* lock them all */
392086de 20360+static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20361+{
20362+ int err;
20363+ struct dentry *h_trap;
20364+
20365+ a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
20366+ a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
20367+ if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
20368+ a->rename_lock = 0;
392086de
AM
20369+ err = au_pin(&a->pin, a->dentry, a->mvd_bdst,
20370+ au_opt_udba(a->sb),
c2b27bf2
AM
20371+ AuPin_MNT_WRITE | AuPin_DI_LOCKED);
20372+ if (!err) {
20373+ a->mvd_h_src_dir = a->mvd_h_src_parent->d_inode;
20374+ mutex_lock_nested(&a->mvd_h_src_dir->i_mutex,
20375+ AuLsc_I_PARENT3);
20376+ } else
392086de 20377+ AU_MVD_PR(dmsg, "pin failed\n");
c2b27bf2
AM
20378+ goto out;
20379+ }
20380+
20381+ err = 0;
20382+ a->rename_lock = 1;
20383+ h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
20384+ a->mvd_h_dst_parent, a->mvd_hdir_dst);
20385+ if (h_trap) {
20386+ err = (h_trap != a->mvd_h_src_parent);
20387+ if (err)
20388+ err = (h_trap != a->mvd_h_dst_parent);
20389+ }
20390+ BUG_ON(err); /* it should never happen */
20391+
20392+out:
20393+ AuTraceErr(err);
20394+ return err;
20395+}
20396+
392086de 20397+static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20398+{
20399+ if (!a->rename_lock) {
20400+ mutex_unlock(&a->mvd_h_src_dir->i_mutex);
20401+ au_unpin(&a->pin);
20402+ } else
20403+ vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
20404+ a->mvd_h_dst_parent, a->mvd_hdir_dst);
20405+}
20406+
20407+/* copy-down the file */
392086de 20408+static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20409+{
20410+ int err;
20411+ struct au_cp_generic cpg = {
20412+ .dentry = a->dentry,
20413+ .bdst = a->mvd_bdst,
20414+ .bsrc = a->mvd_bsrc,
20415+ .len = -1,
20416+ .pin = &a->pin,
20417+ .flags = AuCpup_DTIME | AuCpup_HOPEN
20418+ };
20419+
20420+ AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
392086de
AM
20421+ if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
20422+ au_fset_cpup(cpg.flags, OVERWRITE);
20423+ if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
20424+ au_fset_cpup(cpg.flags, RWDST);
c2b27bf2
AM
20425+ err = au_sio_cpdown_simple(&cpg);
20426+ if (unlikely(err))
392086de 20427+ AU_MVD_PR(dmsg, "cpdown failed\n");
c2b27bf2
AM
20428+
20429+ AuTraceErr(err);
20430+ return err;
20431+}
20432+
20433+/*
20434+ * unlink the whiteout on bdst if exist which may be created by UDBA while we
20435+ * were sleeping
20436+ */
392086de 20437+static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20438+{
20439+ int err;
20440+ struct path h_path;
20441+ struct au_branch *br;
523b37e3 20442+ struct inode *delegated;
c2b27bf2
AM
20443+
20444+ br = au_sbr(a->sb, a->mvd_bdst);
20445+ h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
20446+ err = PTR_ERR(h_path.dentry);
20447+ if (IS_ERR(h_path.dentry)) {
392086de 20448+ AU_MVD_PR(dmsg, "wh_lkup failed\n");
c2b27bf2
AM
20449+ goto out;
20450+ }
20451+
20452+ err = 0;
20453+ if (h_path.dentry->d_inode) {
20454+ h_path.mnt = au_br_mnt(br);
523b37e3 20455+ delegated = NULL;
c2b27bf2 20456+ err = vfsub_unlink(a->mvd_h_dst_parent->d_inode, &h_path,
523b37e3
AM
20457+ &delegated, /*force*/0);
20458+ if (unlikely(err == -EWOULDBLOCK)) {
20459+ pr_warn("cannot retry for NFSv4 delegation"
20460+ " for an internal unlink\n");
20461+ iput(delegated);
20462+ }
c2b27bf2 20463+ if (unlikely(err))
392086de 20464+ AU_MVD_PR(dmsg, "wh_unlink failed\n");
c2b27bf2
AM
20465+ }
20466+ dput(h_path.dentry);
20467+
20468+out:
20469+ AuTraceErr(err);
20470+ return err;
20471+}
20472+
20473+/*
20474+ * unlink the topmost h_dentry
20475+ * Note: the target file MAY be modified by UDBA between this mutex_unlock() and
20476+ * mutex_lock() in vfs_unlink(). in this case, such changes may be lost.
20477+ */
392086de 20478+static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20479+{
20480+ int err;
20481+ struct path h_path;
523b37e3 20482+ struct inode *delegated;
c2b27bf2
AM
20483+
20484+ h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
20485+ h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
523b37e3
AM
20486+ delegated = NULL;
20487+ err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
20488+ if (unlikely(err == -EWOULDBLOCK)) {
20489+ pr_warn("cannot retry for NFSv4 delegation"
20490+ " for an internal unlink\n");
20491+ iput(delegated);
20492+ }
c2b27bf2 20493+ if (unlikely(err))
392086de 20494+ AU_MVD_PR(dmsg, "unlink failed\n");
c2b27bf2
AM
20495+
20496+ AuTraceErr(err);
20497+ return err;
20498+}
20499+
20500+/*
20501+ * copy-down the file and unlink the bsrc file.
20502+ * - unlink the bdst whout if exist
20503+ * - copy-down the file (with whtmp name and rename)
20504+ * - unlink the bsrc file
20505+ */
392086de 20506+static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20507+{
20508+ int err;
20509+
392086de 20510+ err = au_do_mkdir(dmsg, a);
c2b27bf2 20511+ if (!err)
392086de 20512+ err = au_do_lock(dmsg, a);
c2b27bf2
AM
20513+ if (unlikely(err))
20514+ goto out;
20515+
20516+ /*
20517+ * do not revert the activities we made on bdst since they should be
20518+ * harmless in aufs.
20519+ */
20520+
392086de 20521+ err = au_do_cpdown(dmsg, a);
c2b27bf2 20522+ if (!err)
392086de
AM
20523+ err = au_do_unlink_wh(dmsg, a);
20524+ if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
20525+ err = au_do_unlink(dmsg, a);
c2b27bf2
AM
20526+ if (unlikely(err))
20527+ goto out_unlock;
20528+
20529+ /* maintain internal array */
392086de
AM
20530+ if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
20531+ au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
20532+ au_set_dbstart(a->dentry, a->mvd_bdst);
20533+ au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
20534+ au_set_ibstart(a->inode, a->mvd_bdst);
20535+ }
c2b27bf2
AM
20536+ if (au_dbend(a->dentry) < a->mvd_bdst)
20537+ au_set_dbend(a->dentry, a->mvd_bdst);
c2b27bf2
AM
20538+ if (au_ibend(a->inode) < a->mvd_bdst)
20539+ au_set_ibend(a->inode, a->mvd_bdst);
20540+
20541+out_unlock:
392086de 20542+ au_do_unlock(dmsg, a);
c2b27bf2
AM
20543+out:
20544+ AuTraceErr(err);
20545+ return err;
20546+}
20547+
20548+/* ---------------------------------------------------------------------- */
20549+
392086de 20550+static int find_lower_writable(struct au_mvd_args *a)
c2b27bf2 20551+{
392086de
AM
20552+ struct super_block *sb;
20553+ aufs_bindex_t bindex, bend;
c2b27bf2
AM
20554+ struct au_branch *br;
20555+
392086de
AM
20556+ sb = a->sb;
20557+ bindex = a->mvd_bsrc;
c2b27bf2 20558+ bend = au_sbend(sb);
392086de
AM
20559+ if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER)) {
20560+ for (bindex++; bindex <= bend; bindex++) {
20561+ br = au_sbr(sb, bindex);
20562+ if (!au_br_rdonly(br))
20563+ return bindex;
20564+ }
20565+ } else {
20566+ for (bindex++; bindex <= bend; bindex++) {
20567+ br = au_sbr(sb, bindex);
20568+ if (!(au_br_sb(br)->s_flags & MS_RDONLY)) {
20569+ if (au_br_rdonly(br))
20570+ a->mvdown.flags
20571+ |= AUFS_MVDOWN_ROLOWER_R;
20572+ return bindex;
20573+ }
20574+ }
c2b27bf2
AM
20575+ }
20576+
20577+ return -1;
20578+}
20579+
20580+/* make sure the file is idle */
392086de 20581+static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20582+{
20583+ int err, plinked;
c2b27bf2
AM
20584+
20585+ err = 0;
c2b27bf2
AM
20586+ plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
20587+ if (au_dbstart(a->dentry) == a->mvd_bsrc
392086de 20588+ && d_count(a->dentry) == 1
c2b27bf2 20589+ && atomic_read(&a->inode->i_count) == 1
392086de 20590+ /* && a->mvd_h_src_inode->i_nlink == 1 */
c2b27bf2
AM
20591+ && (!plinked || !au_plink_test(a->inode))
20592+ && a->inode->i_nlink == 1)
20593+ goto out;
20594+
20595+ err = -EBUSY;
392086de 20596+ AU_MVD_PR(dmsg,
c2b27bf2 20597+ "b%d, d{b%d, c%u?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
392086de 20598+ a->mvd_bsrc, au_dbstart(a->dentry), d_count(a->dentry),
c2b27bf2 20599+ atomic_read(&a->inode->i_count), a->inode->i_nlink,
392086de 20600+ a->mvd_h_src_inode->i_nlink,
c2b27bf2
AM
20601+ plinked, plinked ? au_plink_test(a->inode) : 0);
20602+
20603+out:
20604+ AuTraceErr(err);
20605+ return err;
20606+}
20607+
20608+/* make sure the parent dir is fine */
392086de 20609+static int au_mvd_args_parent(const unsigned char dmsg,
c2b27bf2
AM
20610+ struct au_mvd_args *a)
20611+{
20612+ int err;
20613+ aufs_bindex_t bindex;
20614+
20615+ err = 0;
20616+ if (unlikely(au_alive_dir(a->parent))) {
20617+ err = -ENOENT;
392086de 20618+ AU_MVD_PR(dmsg, "parent dir is dead\n");
c2b27bf2
AM
20619+ goto out;
20620+ }
20621+
20622+ a->bopq = au_dbdiropq(a->parent);
20623+ bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
20624+ AuDbg("b%d\n", bindex);
20625+ if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
20626+ || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
20627+ err = -EINVAL;
392086de
AM
20628+ a->mvd_errno = EAU_MVDOWN_OPAQUE;
20629+ AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
c2b27bf2
AM
20630+ a->bopq, a->mvd_bdst);
20631+ }
20632+
20633+out:
20634+ AuTraceErr(err);
20635+ return err;
20636+}
20637+
392086de 20638+static int au_mvd_args_intermediate(const unsigned char dmsg,
c2b27bf2
AM
20639+ struct au_mvd_args *a)
20640+{
20641+ int err;
20642+ struct au_dinfo *dinfo, *tmp;
20643+
20644+ /* lookup the next lower positive entry */
20645+ err = -ENOMEM;
20646+ tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
20647+ if (unlikely(!tmp))
20648+ goto out;
20649+
20650+ a->bfound = -1;
20651+ a->bwh = -1;
20652+ dinfo = au_di(a->dentry);
20653+ au_di_cp(tmp, dinfo);
20654+ au_di_swap(tmp, dinfo);
20655+
20656+ /* returns the number of positive dentries */
20657+ err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1, /*type*/0);
20658+ if (!err)
20659+ a->bwh = au_dbwh(a->dentry);
20660+ else if (err > 0)
20661+ a->bfound = au_dbstart(a->dentry);
20662+
20663+ au_di_swap(tmp, dinfo);
20664+ au_rw_write_unlock(&tmp->di_rwsem);
20665+ au_di_free(tmp);
20666+ if (unlikely(err < 0))
392086de 20667+ AU_MVD_PR(dmsg, "failed look-up lower\n");
c2b27bf2
AM
20668+
20669+ /*
20670+ * here, we have these cases.
20671+ * bfound == -1
20672+ * no positive dentry under bsrc. there are more sub-cases.
20673+ * bwh < 0
20674+ * there no whiteout, we can safely move-down.
20675+ * bwh <= bsrc
20676+ * impossible
20677+ * bsrc < bwh && bwh < bdst
20678+ * there is a whiteout on RO branch. cannot proceed.
20679+ * bwh == bdst
20680+ * there is a whiteout on the RW target branch. it should
20681+ * be removed.
20682+ * bdst < bwh
20683+ * there is a whiteout somewhere unrelated branch.
20684+ * -1 < bfound && bfound <= bsrc
20685+ * impossible.
20686+ * bfound < bdst
20687+ * found, but it is on RO branch between bsrc and bdst. cannot
20688+ * proceed.
20689+ * bfound == bdst
20690+ * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
20691+ * error.
20692+ * bdst < bfound
20693+ * found, after we create the file on bdst, it will be hidden.
20694+ */
20695+
20696+ AuDebugOn(a->bfound == -1
20697+ && a->bwh != -1
20698+ && a->bwh <= a->mvd_bsrc);
20699+ AuDebugOn(-1 < a->bfound
20700+ && a->bfound <= a->mvd_bsrc);
20701+
20702+ err = -EINVAL;
20703+ if (a->bfound == -1
20704+ && a->mvd_bsrc < a->bwh
20705+ && a->bwh != -1
20706+ && a->bwh < a->mvd_bdst) {
392086de
AM
20707+ a->mvd_errno = EAU_MVDOWN_WHITEOUT;
20708+ AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
c2b27bf2
AM
20709+ a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
20710+ goto out;
20711+ } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
392086de
AM
20712+ a->mvd_errno = EAU_MVDOWN_UPPER;
20713+ AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
c2b27bf2
AM
20714+ a->mvd_bdst, a->bfound);
20715+ goto out;
20716+ }
20717+
20718+ err = 0; /* success */
20719+
20720+out:
20721+ AuTraceErr(err);
20722+ return err;
20723+}
20724+
392086de 20725+static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20726+{
20727+ int err;
20728+
392086de
AM
20729+ err = 0;
20730+ if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
20731+ && a->bfound == a->mvd_bdst)
20732+ err = -EEXIST;
c2b27bf2
AM
20733+ AuTraceErr(err);
20734+ return err;
20735+}
20736+
392086de 20737+static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
20738+{
20739+ int err;
20740+ struct au_branch *br;
20741+
20742+ err = -EISDIR;
20743+ if (unlikely(S_ISDIR(a->inode->i_mode)))
20744+ goto out;
20745+
20746+ err = -EINVAL;
392086de
AM
20747+ if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
20748+ a->mvd_bsrc = au_ibstart(a->inode);
20749+ else {
20750+ a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
20751+ if (unlikely(a->mvd_bsrc < 0
20752+ || (a->mvd_bsrc < au_dbstart(a->dentry)
20753+ || au_dbend(a->dentry) < a->mvd_bsrc
20754+ || !au_h_dptr(a->dentry, a->mvd_bsrc))
20755+ || (a->mvd_bsrc < au_ibstart(a->inode)
20756+ || au_ibend(a->inode) < a->mvd_bsrc
20757+ || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
20758+ a->mvd_errno = EAU_MVDOWN_NOUPPER;
20759+ AU_MVD_PR(dmsg, "no upper\n");
20760+ goto out;
20761+ }
20762+ }
c2b27bf2 20763+ if (unlikely(a->mvd_bsrc == au_sbend(a->sb))) {
392086de
AM
20764+ a->mvd_errno = EAU_MVDOWN_BOTTOM;
20765+ AU_MVD_PR(dmsg, "on the bottom\n");
c2b27bf2
AM
20766+ goto out;
20767+ }
392086de 20768+ a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
c2b27bf2
AM
20769+ br = au_sbr(a->sb, a->mvd_bsrc);
20770+ err = au_br_rdonly(br);
392086de
AM
20771+ if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
20772+ if (unlikely(err))
20773+ goto out;
20774+ } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
20775+ || IS_APPEND(a->mvd_h_src_inode))) {
20776+ if (err)
20777+ a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
20778+ /* go on */
20779+ } else
c2b27bf2
AM
20780+ goto out;
20781+
20782+ err = -EINVAL;
392086de
AM
20783+ if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
20784+ a->mvd_bdst = find_lower_writable(a);
20785+ if (unlikely(a->mvd_bdst < 0)) {
20786+ a->mvd_errno = EAU_MVDOWN_BOTTOM;
20787+ AU_MVD_PR(dmsg, "no writable lower branch\n");
20788+ goto out;
20789+ }
20790+ } else {
20791+ a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
20792+ if (unlikely(a->mvd_bdst < 0
20793+ || au_sbend(a->sb) < a->mvd_bdst)) {
20794+ a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
20795+ AU_MVD_PR(dmsg, "no lower brid\n");
20796+ goto out;
20797+ }
c2b27bf2
AM
20798+ }
20799+
392086de 20800+ err = au_mvd_args_busy(dmsg, a);
c2b27bf2 20801+ if (!err)
392086de 20802+ err = au_mvd_args_parent(dmsg, a);
c2b27bf2 20803+ if (!err)
392086de 20804+ err = au_mvd_args_intermediate(dmsg, a);
c2b27bf2 20805+ if (!err)
392086de 20806+ err = au_mvd_args_exist(dmsg, a);
c2b27bf2
AM
20807+ if (!err)
20808+ AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
20809+
20810+out:
20811+ AuTraceErr(err);
20812+ return err;
20813+}
20814+
20815+int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
20816+{
392086de
AM
20817+ int err, e;
20818+ unsigned char dmsg;
20819+ struct au_mvd_args *args;
c2b27bf2
AM
20820+
20821+ err = -EPERM;
20822+ if (unlikely(!capable(CAP_SYS_ADMIN)))
20823+ goto out;
20824+
392086de
AM
20825+ WARN_ONCE(1, "move-down is still testing...\n");
20826+
20827+ err = -ENOMEM;
20828+ args = kmalloc(sizeof(*args), GFP_NOFS);
20829+ if (unlikely(!args))
20830+ goto out;
20831+
20832+ err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
20833+ if (!err)
20834+ err = !access_ok(VERIFY_WRITE, uarg, sizeof(*uarg));
c2b27bf2
AM
20835+ if (unlikely(err)) {
20836+ err = -EFAULT;
392086de
AM
20837+ AuTraceErr(err);
20838+ goto out_free;
c2b27bf2 20839+ }
392086de
AM
20840+ AuDbg("flags 0x%x\n", args->mvdown.flags);
20841+ args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
20842+ args->mvdown.au_errno = 0;
20843+ args->dentry = dentry;
20844+ args->inode = dentry->d_inode;
20845+ args->sb = dentry->d_sb;
c2b27bf2 20846+
392086de
AM
20847+ err = -ENOENT;
20848+ dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
20849+ args->parent = dget_parent(dentry);
20850+ args->dir = args->parent->d_inode;
20851+ mutex_lock_nested(&args->dir->i_mutex, I_MUTEX_PARENT);
20852+ dput(args->parent);
20853+ if (unlikely(args->parent != dentry->d_parent)) {
20854+ AU_MVD_PR(dmsg, "parent dir is moved\n");
c2b27bf2
AM
20855+ goto out_dir;
20856+ }
20857+
392086de 20858+ mutex_lock_nested(&args->inode->i_mutex, I_MUTEX_CHILD);
c2b27bf2
AM
20859+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH);
20860+ if (unlikely(err))
20861+ goto out_inode;
20862+
392086de
AM
20863+ di_write_lock_parent(args->parent);
20864+ err = au_mvd_args(dmsg, args);
c2b27bf2
AM
20865+ if (unlikely(err))
20866+ goto out_parent;
20867+
20868+ AuDbgDentry(dentry);
392086de
AM
20869+ AuDbgInode(args->inode);
20870+ err = au_do_mvdown(dmsg, args);
c2b27bf2
AM
20871+ if (unlikely(err))
20872+ goto out_parent;
20873+ AuDbgDentry(dentry);
392086de 20874+ AuDbgInode(args->inode);
c2b27bf2 20875+
392086de
AM
20876+ au_cpup_attr_timesizes(args->dir);
20877+ au_cpup_attr_timesizes(args->inode);
20878+ au_cpup_igen(args->inode, au_h_iptr(args->inode, args->mvd_bdst));
c2b27bf2
AM
20879+ /* au_digen_dec(dentry); */
20880+
20881+out_parent:
392086de 20882+ di_write_unlock(args->parent);
c2b27bf2
AM
20883+ aufs_read_unlock(dentry, AuLock_DW);
20884+out_inode:
392086de 20885+ mutex_unlock(&args->inode->i_mutex);
c2b27bf2 20886+out_dir:
392086de
AM
20887+ mutex_unlock(&args->dir->i_mutex);
20888+out_free:
20889+ e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
20890+ if (unlikely(e))
20891+ err = -EFAULT;
20892+ kfree(args);
c2b27bf2
AM
20893+out:
20894+ AuTraceErr(err);
20895+ return err;
20896+}
20897diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
20898--- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 20899+++ linux/fs/aufs/opts.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 20900@@ -0,0 +1,1701 @@
1facf9fc 20901+/*
523b37e3 20902+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 20903+ *
20904+ * This program, aufs is free software; you can redistribute it and/or modify
20905+ * it under the terms of the GNU General Public License as published by
20906+ * the Free Software Foundation; either version 2 of the License, or
20907+ * (at your option) any later version.
dece6358
AM
20908+ *
20909+ * This program is distributed in the hope that it will be useful,
20910+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
20911+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20912+ * GNU General Public License for more details.
20913+ *
20914+ * You should have received a copy of the GNU General Public License
523b37e3 20915+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 20916+ */
20917+
20918+/*
20919+ * mount options/flags
20920+ */
20921+
dece6358 20922+#include <linux/namei.h>
1facf9fc 20923+#include <linux/types.h> /* a distribution requires */
20924+#include <linux/parser.h>
20925+#include "aufs.h"
20926+
20927+/* ---------------------------------------------------------------------- */
20928+
20929+enum {
20930+ Opt_br,
20931+ Opt_add, Opt_del, Opt_mod, Opt_reorder, Opt_append, Opt_prepend,
20932+ Opt_idel, Opt_imod, Opt_ireorder,
20933+ Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash, Opt_rendir,
dece6358 20934+ Opt_rdblk_def, Opt_rdhash_def,
1facf9fc 20935+ Opt_xino, Opt_zxino, Opt_noxino,
20936+ Opt_trunc_xino, Opt_trunc_xino_v, Opt_notrunc_xino,
20937+ Opt_trunc_xino_path, Opt_itrunc_xino,
20938+ Opt_trunc_xib, Opt_notrunc_xib,
dece6358 20939+ Opt_shwh, Opt_noshwh,
1facf9fc 20940+ Opt_plink, Opt_noplink, Opt_list_plink,
20941+ Opt_udba,
4a4d8108 20942+ Opt_dio, Opt_nodio,
1facf9fc 20943+ /* Opt_lock, Opt_unlock, */
20944+ Opt_cmd, Opt_cmd_args,
20945+ Opt_diropq_a, Opt_diropq_w,
20946+ Opt_warn_perm, Opt_nowarn_perm,
20947+ Opt_wbr_copyup, Opt_wbr_create,
20948+ Opt_refrof, Opt_norefrof,
20949+ Opt_verbose, Opt_noverbose,
20950+ Opt_sum, Opt_nosum, Opt_wsum,
20951+ Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
20952+};
20953+
20954+static match_table_t options = {
20955+ {Opt_br, "br=%s"},
20956+ {Opt_br, "br:%s"},
20957+
20958+ {Opt_add, "add=%d:%s"},
20959+ {Opt_add, "add:%d:%s"},
20960+ {Opt_add, "ins=%d:%s"},
20961+ {Opt_add, "ins:%d:%s"},
20962+ {Opt_append, "append=%s"},
20963+ {Opt_append, "append:%s"},
20964+ {Opt_prepend, "prepend=%s"},
20965+ {Opt_prepend, "prepend:%s"},
20966+
20967+ {Opt_del, "del=%s"},
20968+ {Opt_del, "del:%s"},
20969+ /* {Opt_idel, "idel:%d"}, */
20970+ {Opt_mod, "mod=%s"},
20971+ {Opt_mod, "mod:%s"},
20972+ /* {Opt_imod, "imod:%d:%s"}, */
20973+
20974+ {Opt_dirwh, "dirwh=%d"},
20975+
20976+ {Opt_xino, "xino=%s"},
20977+ {Opt_noxino, "noxino"},
20978+ {Opt_trunc_xino, "trunc_xino"},
20979+ {Opt_trunc_xino_v, "trunc_xino_v=%d:%d"},
20980+ {Opt_notrunc_xino, "notrunc_xino"},
20981+ {Opt_trunc_xino_path, "trunc_xino=%s"},
20982+ {Opt_itrunc_xino, "itrunc_xino=%d"},
20983+ /* {Opt_zxino, "zxino=%s"}, */
20984+ {Opt_trunc_xib, "trunc_xib"},
20985+ {Opt_notrunc_xib, "notrunc_xib"},
20986+
e49829fe 20987+#ifdef CONFIG_PROC_FS
1facf9fc 20988+ {Opt_plink, "plink"},
e49829fe
JR
20989+#else
20990+ {Opt_ignore_silent, "plink"},
20991+#endif
20992+
1facf9fc 20993+ {Opt_noplink, "noplink"},
e49829fe 20994+
1facf9fc 20995+#ifdef CONFIG_AUFS_DEBUG
20996+ {Opt_list_plink, "list_plink"},
20997+#endif
20998+
20999+ {Opt_udba, "udba=%s"},
21000+
4a4d8108
AM
21001+ {Opt_dio, "dio"},
21002+ {Opt_nodio, "nodio"},
21003+
1facf9fc 21004+ {Opt_diropq_a, "diropq=always"},
21005+ {Opt_diropq_a, "diropq=a"},
21006+ {Opt_diropq_w, "diropq=whiteouted"},
21007+ {Opt_diropq_w, "diropq=w"},
21008+
21009+ {Opt_warn_perm, "warn_perm"},
21010+ {Opt_nowarn_perm, "nowarn_perm"},
21011+
21012+ /* keep them temporary */
21013+ {Opt_ignore_silent, "coo=%s"},
21014+ {Opt_ignore_silent, "nodlgt"},
21015+ {Opt_ignore_silent, "nodirperm1"},
1facf9fc 21016+ {Opt_ignore_silent, "clean_plink"},
21017+
dece6358
AM
21018+#ifdef CONFIG_AUFS_SHWH
21019+ {Opt_shwh, "shwh"},
21020+#endif
21021+ {Opt_noshwh, "noshwh"},
21022+
1facf9fc 21023+ {Opt_rendir, "rendir=%d"},
21024+
21025+ {Opt_refrof, "refrof"},
21026+ {Opt_norefrof, "norefrof"},
21027+
21028+ {Opt_verbose, "verbose"},
21029+ {Opt_verbose, "v"},
21030+ {Opt_noverbose, "noverbose"},
21031+ {Opt_noverbose, "quiet"},
21032+ {Opt_noverbose, "q"},
21033+ {Opt_noverbose, "silent"},
21034+
21035+ {Opt_sum, "sum"},
21036+ {Opt_nosum, "nosum"},
21037+ {Opt_wsum, "wsum"},
21038+
21039+ {Opt_rdcache, "rdcache=%d"},
21040+ {Opt_rdblk, "rdblk=%d"},
dece6358 21041+ {Opt_rdblk_def, "rdblk=def"},
1facf9fc 21042+ {Opt_rdhash, "rdhash=%d"},
dece6358 21043+ {Opt_rdhash_def, "rdhash=def"},
1facf9fc 21044+
21045+ {Opt_wbr_create, "create=%s"},
21046+ {Opt_wbr_create, "create_policy=%s"},
21047+ {Opt_wbr_copyup, "cpup=%s"},
21048+ {Opt_wbr_copyup, "copyup=%s"},
21049+ {Opt_wbr_copyup, "copyup_policy=%s"},
21050+
21051+ /* internal use for the scripts */
21052+ {Opt_ignore_silent, "si=%s"},
21053+
21054+ {Opt_br, "dirs=%s"},
21055+ {Opt_ignore, "debug=%d"},
21056+ {Opt_ignore, "delete=whiteout"},
21057+ {Opt_ignore, "delete=all"},
21058+ {Opt_ignore, "imap=%s"},
21059+
1308ab2a 21060+ /* temporary workaround, due to old mount(8)? */
21061+ {Opt_ignore_silent, "relatime"},
21062+
1facf9fc 21063+ {Opt_err, NULL}
21064+};
21065+
21066+/* ---------------------------------------------------------------------- */
21067+
21068+static const char *au_parser_pattern(int val, struct match_token *token)
21069+{
21070+ while (token->pattern) {
21071+ if (token->token == val)
21072+ return token->pattern;
21073+ token++;
21074+ }
21075+ BUG();
21076+ return "??";
21077+}
21078+
21079+/* ---------------------------------------------------------------------- */
21080+
1e00d052 21081+static match_table_t brperm = {
1facf9fc 21082+ {AuBrPerm_RO, AUFS_BRPERM_RO},
21083+ {AuBrPerm_RR, AUFS_BRPERM_RR},
21084+ {AuBrPerm_RW, AUFS_BRPERM_RW},
1e00d052
AM
21085+ {0, NULL}
21086+};
1facf9fc 21087+
86dc4139
AM
21088+static match_table_t brattr = {
21089+ {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
1e00d052 21090+ {AuBrRAttr_WH, AUFS_BRRATTR_WH},
1e00d052
AM
21091+ {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
21092+ {0, NULL}
1facf9fc 21093+};
21094+
86dc4139
AM
21095+#define AuBrStr_LONGEST AUFS_BRPERM_RW \
21096+ "+" AUFS_BRATTR_UNPIN \
21097+ "+" AUFS_BRWATTR_NLWH
1e00d052
AM
21098+
21099+static int br_attr_val(char *str, match_table_t table, substring_t args[])
21100+{
21101+ int attr, v;
21102+ char *p;
21103+
21104+ attr = 0;
21105+ do {
21106+ p = strchr(str, '+');
21107+ if (p)
21108+ *p = 0;
21109+ v = match_token(str, table, args);
21110+ if (v)
21111+ attr |= v;
21112+ else {
21113+ if (p)
21114+ *p = '+';
0c3ec466 21115+ pr_warn("ignored branch attribute %s\n", str);
1e00d052
AM
21116+ break;
21117+ }
21118+ if (p)
21119+ str = p + 1;
21120+ } while (p);
21121+
21122+ return attr;
21123+}
21124+
4a4d8108 21125+static int noinline_for_stack br_perm_val(char *perm)
1facf9fc 21126+{
21127+ int val;
86dc4139 21128+ char *p, *q;
1facf9fc 21129+ substring_t args[MAX_OPT_ARGS];
21130+
1e00d052
AM
21131+ p = strchr(perm, '+');
21132+ if (p)
21133+ *p = 0;
21134+ val = match_token(perm, brperm, args);
21135+ if (!val) {
21136+ if (p)
21137+ *p = '+';
0c3ec466 21138+ pr_warn("ignored branch permission %s\n", perm);
1e00d052
AM
21139+ val = AuBrPerm_RO;
21140+ goto out;
21141+ }
21142+ if (!p)
21143+ goto out;
21144+
86dc4139
AM
21145+ p++;
21146+ while (1) {
21147+ q = strchr(p, '+');
21148+ if (q)
21149+ *q = 0;
21150+ val |= br_attr_val(p, brattr, args);
21151+ if (q) {
21152+ *q = '+';
21153+ p = q + 1;
21154+ } else
21155+ break;
21156+ }
21157+ switch (val & AuBrPerm_Mask) {
1e00d052
AM
21158+ case AuBrPerm_RO:
21159+ case AuBrPerm_RR:
86dc4139
AM
21160+ if (unlikely(val & AuBrWAttr_NoLinkWH)) {
21161+ pr_warn("ignored branch attribute %s\n",
21162+ AUFS_BRWATTR_NLWH);
21163+ val &= ~AuBrWAttr_NoLinkWH;
21164+ }
1e00d052
AM
21165+ break;
21166+ case AuBrPerm_RW:
86dc4139
AM
21167+ if (unlikely(val & AuBrRAttr_WH)) {
21168+ pr_warn("ignored branch attribute %s\n",
21169+ AUFS_BRRATTR_WH);
21170+ val &= ~AuBrRAttr_WH;
21171+ }
1e00d052
AM
21172+ break;
21173+ }
21174+
21175+out:
1facf9fc 21176+ return val;
21177+}
21178+
1e00d052
AM
21179+/* Caller should free the return value */
21180+char *au_optstr_br_perm(int brperm)
1facf9fc 21181+{
1e00d052
AM
21182+ char *p, a[sizeof(AuBrStr_LONGEST)];
21183+ int sz;
21184+
21185+#define SetPerm(str) do { \
21186+ sz = sizeof(str); \
21187+ memcpy(a, str, sz); \
21188+ p = a + sz - 1; \
21189+ } while (0)
21190+
21191+#define AppendAttr(flag, str) do { \
21192+ if (brperm & flag) { \
21193+ sz = sizeof(str); \
21194+ *p++ = '+'; \
21195+ memcpy(p, str, sz); \
21196+ p += sz - 1; \
21197+ } \
21198+ } while (0)
21199+
21200+ switch (brperm & AuBrPerm_Mask) {
21201+ case AuBrPerm_RO:
21202+ SetPerm(AUFS_BRPERM_RO);
21203+ break;
21204+ case AuBrPerm_RR:
21205+ SetPerm(AUFS_BRPERM_RR);
21206+ break;
21207+ case AuBrPerm_RW:
21208+ SetPerm(AUFS_BRPERM_RW);
21209+ break;
21210+ default:
21211+ AuDebugOn(1);
21212+ }
21213+
86dc4139 21214+ AppendAttr(AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN);
1e00d052
AM
21215+ AppendAttr(AuBrRAttr_WH, AUFS_BRRATTR_WH);
21216+ AppendAttr(AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH);
21217+
21218+ AuDebugOn(strlen(a) >= sizeof(a));
21219+ return kstrdup(a, GFP_NOFS);
21220+#undef SetPerm
21221+#undef AppendAttr
1facf9fc 21222+}
21223+
21224+/* ---------------------------------------------------------------------- */
21225+
21226+static match_table_t udbalevel = {
21227+ {AuOpt_UDBA_REVAL, "reval"},
21228+ {AuOpt_UDBA_NONE, "none"},
4a4d8108
AM
21229+#ifdef CONFIG_AUFS_HNOTIFY
21230+ {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
21231+#ifdef CONFIG_AUFS_HFSNOTIFY
21232+ {AuOpt_UDBA_HNOTIFY, "fsnotify"},
4a4d8108 21233+#endif
1facf9fc 21234+#endif
21235+ {-1, NULL}
21236+};
21237+
4a4d8108 21238+static int noinline_for_stack udba_val(char *str)
1facf9fc 21239+{
21240+ substring_t args[MAX_OPT_ARGS];
21241+
7f207e10 21242+ return match_token(str, udbalevel, args);
1facf9fc 21243+}
21244+
21245+const char *au_optstr_udba(int udba)
21246+{
21247+ return au_parser_pattern(udba, (void *)udbalevel);
21248+}
21249+
21250+/* ---------------------------------------------------------------------- */
21251+
21252+static match_table_t au_wbr_create_policy = {
21253+ {AuWbrCreate_TDP, "tdp"},
21254+ {AuWbrCreate_TDP, "top-down-parent"},
21255+ {AuWbrCreate_RR, "rr"},
21256+ {AuWbrCreate_RR, "round-robin"},
21257+ {AuWbrCreate_MFS, "mfs"},
21258+ {AuWbrCreate_MFS, "most-free-space"},
21259+ {AuWbrCreate_MFSV, "mfs:%d"},
21260+ {AuWbrCreate_MFSV, "most-free-space:%d"},
21261+
21262+ {AuWbrCreate_MFSRR, "mfsrr:%d"},
21263+ {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
21264+ {AuWbrCreate_PMFS, "pmfs"},
21265+ {AuWbrCreate_PMFSV, "pmfs:%d"},
392086de
AM
21266+ {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
21267+ {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
1facf9fc 21268+
21269+ {-1, NULL}
21270+};
21271+
dece6358
AM
21272+/*
21273+ * cf. linux/lib/parser.c and cmdline.c
21274+ * gave up calling memparse() since it uses simple_strtoull() instead of
9dbd164d 21275+ * kstrto...().
dece6358 21276+ */
4a4d8108
AM
21277+static int noinline_for_stack
21278+au_match_ull(substring_t *s, unsigned long long *result)
1facf9fc 21279+{
21280+ int err;
21281+ unsigned int len;
21282+ char a[32];
21283+
21284+ err = -ERANGE;
21285+ len = s->to - s->from;
21286+ if (len + 1 <= sizeof(a)) {
21287+ memcpy(a, s->from, len);
21288+ a[len] = '\0';
9dbd164d 21289+ err = kstrtoull(a, 0, result);
1facf9fc 21290+ }
21291+ return err;
21292+}
21293+
21294+static int au_wbr_mfs_wmark(substring_t *arg, char *str,
21295+ struct au_opt_wbr_create *create)
21296+{
21297+ int err;
21298+ unsigned long long ull;
21299+
21300+ err = 0;
21301+ if (!au_match_ull(arg, &ull))
21302+ create->mfsrr_watermark = ull;
21303+ else {
4a4d8108 21304+ pr_err("bad integer in %s\n", str);
1facf9fc 21305+ err = -EINVAL;
21306+ }
21307+
21308+ return err;
21309+}
21310+
21311+static int au_wbr_mfs_sec(substring_t *arg, char *str,
21312+ struct au_opt_wbr_create *create)
21313+{
21314+ int n, err;
21315+
21316+ err = 0;
027c5e7a 21317+ if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
1facf9fc 21318+ create->mfs_second = n;
21319+ else {
4a4d8108 21320+ pr_err("bad integer in %s\n", str);
1facf9fc 21321+ err = -EINVAL;
21322+ }
21323+
21324+ return err;
21325+}
21326+
4a4d8108
AM
21327+static int noinline_for_stack
21328+au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
1facf9fc 21329+{
21330+ int err, e;
21331+ substring_t args[MAX_OPT_ARGS];
21332+
21333+ err = match_token(str, au_wbr_create_policy, args);
21334+ create->wbr_create = err;
21335+ switch (err) {
21336+ case AuWbrCreate_MFSRRV:
392086de 21337+ case AuWbrCreate_PMFSRRV:
1facf9fc 21338+ e = au_wbr_mfs_wmark(&args[0], str, create);
21339+ if (!e)
21340+ e = au_wbr_mfs_sec(&args[1], str, create);
21341+ if (unlikely(e))
21342+ err = e;
21343+ break;
21344+ case AuWbrCreate_MFSRR:
392086de 21345+ case AuWbrCreate_PMFSRR:
1facf9fc 21346+ e = au_wbr_mfs_wmark(&args[0], str, create);
21347+ if (unlikely(e)) {
21348+ err = e;
21349+ break;
21350+ }
21351+ /*FALLTHROUGH*/
21352+ case AuWbrCreate_MFS:
21353+ case AuWbrCreate_PMFS:
027c5e7a 21354+ create->mfs_second = AUFS_MFS_DEF_SEC;
1facf9fc 21355+ break;
21356+ case AuWbrCreate_MFSV:
21357+ case AuWbrCreate_PMFSV:
21358+ e = au_wbr_mfs_sec(&args[0], str, create);
21359+ if (unlikely(e))
21360+ err = e;
21361+ break;
21362+ }
21363+
21364+ return err;
21365+}
21366+
21367+const char *au_optstr_wbr_create(int wbr_create)
21368+{
21369+ return au_parser_pattern(wbr_create, (void *)au_wbr_create_policy);
21370+}
21371+
21372+static match_table_t au_wbr_copyup_policy = {
21373+ {AuWbrCopyup_TDP, "tdp"},
21374+ {AuWbrCopyup_TDP, "top-down-parent"},
21375+ {AuWbrCopyup_BUP, "bup"},
21376+ {AuWbrCopyup_BUP, "bottom-up-parent"},
21377+ {AuWbrCopyup_BU, "bu"},
21378+ {AuWbrCopyup_BU, "bottom-up"},
21379+ {-1, NULL}
21380+};
21381+
4a4d8108 21382+static int noinline_for_stack au_wbr_copyup_val(char *str)
1facf9fc 21383+{
21384+ substring_t args[MAX_OPT_ARGS];
21385+
21386+ return match_token(str, au_wbr_copyup_policy, args);
21387+}
21388+
21389+const char *au_optstr_wbr_copyup(int wbr_copyup)
21390+{
21391+ return au_parser_pattern(wbr_copyup, (void *)au_wbr_copyup_policy);
21392+}
21393+
21394+/* ---------------------------------------------------------------------- */
21395+
21396+static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
21397+
21398+static void dump_opts(struct au_opts *opts)
21399+{
21400+#ifdef CONFIG_AUFS_DEBUG
21401+ /* reduce stack space */
21402+ union {
21403+ struct au_opt_add *add;
21404+ struct au_opt_del *del;
21405+ struct au_opt_mod *mod;
21406+ struct au_opt_xino *xino;
21407+ struct au_opt_xino_itrunc *xino_itrunc;
21408+ struct au_opt_wbr_create *create;
21409+ } u;
21410+ struct au_opt *opt;
21411+
21412+ opt = opts->opt;
21413+ while (opt->type != Opt_tail) {
21414+ switch (opt->type) {
21415+ case Opt_add:
21416+ u.add = &opt->add;
21417+ AuDbg("add {b%d, %s, 0x%x, %p}\n",
21418+ u.add->bindex, u.add->pathname, u.add->perm,
21419+ u.add->path.dentry);
21420+ break;
21421+ case Opt_del:
21422+ case Opt_idel:
21423+ u.del = &opt->del;
21424+ AuDbg("del {%s, %p}\n",
21425+ u.del->pathname, u.del->h_path.dentry);
21426+ break;
21427+ case Opt_mod:
21428+ case Opt_imod:
21429+ u.mod = &opt->mod;
21430+ AuDbg("mod {%s, 0x%x, %p}\n",
21431+ u.mod->path, u.mod->perm, u.mod->h_root);
21432+ break;
21433+ case Opt_append:
21434+ u.add = &opt->add;
21435+ AuDbg("append {b%d, %s, 0x%x, %p}\n",
21436+ u.add->bindex, u.add->pathname, u.add->perm,
21437+ u.add->path.dentry);
21438+ break;
21439+ case Opt_prepend:
21440+ u.add = &opt->add;
21441+ AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
21442+ u.add->bindex, u.add->pathname, u.add->perm,
21443+ u.add->path.dentry);
21444+ break;
21445+ case Opt_dirwh:
21446+ AuDbg("dirwh %d\n", opt->dirwh);
21447+ break;
21448+ case Opt_rdcache:
21449+ AuDbg("rdcache %d\n", opt->rdcache);
21450+ break;
21451+ case Opt_rdblk:
21452+ AuDbg("rdblk %u\n", opt->rdblk);
21453+ break;
dece6358
AM
21454+ case Opt_rdblk_def:
21455+ AuDbg("rdblk_def\n");
21456+ break;
1facf9fc 21457+ case Opt_rdhash:
21458+ AuDbg("rdhash %u\n", opt->rdhash);
21459+ break;
dece6358
AM
21460+ case Opt_rdhash_def:
21461+ AuDbg("rdhash_def\n");
21462+ break;
1facf9fc 21463+ case Opt_xino:
21464+ u.xino = &opt->xino;
523b37e3 21465+ AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
1facf9fc 21466+ break;
21467+ case Opt_trunc_xino:
21468+ AuLabel(trunc_xino);
21469+ break;
21470+ case Opt_notrunc_xino:
21471+ AuLabel(notrunc_xino);
21472+ break;
21473+ case Opt_trunc_xino_path:
21474+ case Opt_itrunc_xino:
21475+ u.xino_itrunc = &opt->xino_itrunc;
21476+ AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
21477+ break;
21478+
21479+ case Opt_noxino:
21480+ AuLabel(noxino);
21481+ break;
21482+ case Opt_trunc_xib:
21483+ AuLabel(trunc_xib);
21484+ break;
21485+ case Opt_notrunc_xib:
21486+ AuLabel(notrunc_xib);
21487+ break;
dece6358
AM
21488+ case Opt_shwh:
21489+ AuLabel(shwh);
21490+ break;
21491+ case Opt_noshwh:
21492+ AuLabel(noshwh);
21493+ break;
1facf9fc 21494+ case Opt_plink:
21495+ AuLabel(plink);
21496+ break;
21497+ case Opt_noplink:
21498+ AuLabel(noplink);
21499+ break;
21500+ case Opt_list_plink:
21501+ AuLabel(list_plink);
21502+ break;
21503+ case Opt_udba:
21504+ AuDbg("udba %d, %s\n",
21505+ opt->udba, au_optstr_udba(opt->udba));
21506+ break;
4a4d8108
AM
21507+ case Opt_dio:
21508+ AuLabel(dio);
21509+ break;
21510+ case Opt_nodio:
21511+ AuLabel(nodio);
21512+ break;
1facf9fc 21513+ case Opt_diropq_a:
21514+ AuLabel(diropq_a);
21515+ break;
21516+ case Opt_diropq_w:
21517+ AuLabel(diropq_w);
21518+ break;
21519+ case Opt_warn_perm:
21520+ AuLabel(warn_perm);
21521+ break;
21522+ case Opt_nowarn_perm:
21523+ AuLabel(nowarn_perm);
21524+ break;
21525+ case Opt_refrof:
21526+ AuLabel(refrof);
21527+ break;
21528+ case Opt_norefrof:
21529+ AuLabel(norefrof);
21530+ break;
21531+ case Opt_verbose:
21532+ AuLabel(verbose);
21533+ break;
21534+ case Opt_noverbose:
21535+ AuLabel(noverbose);
21536+ break;
21537+ case Opt_sum:
21538+ AuLabel(sum);
21539+ break;
21540+ case Opt_nosum:
21541+ AuLabel(nosum);
21542+ break;
21543+ case Opt_wsum:
21544+ AuLabel(wsum);
21545+ break;
21546+ case Opt_wbr_create:
21547+ u.create = &opt->wbr_create;
21548+ AuDbg("create %d, %s\n", u.create->wbr_create,
21549+ au_optstr_wbr_create(u.create->wbr_create));
21550+ switch (u.create->wbr_create) {
21551+ case AuWbrCreate_MFSV:
21552+ case AuWbrCreate_PMFSV:
21553+ AuDbg("%d sec\n", u.create->mfs_second);
21554+ break;
21555+ case AuWbrCreate_MFSRR:
21556+ AuDbg("%llu watermark\n",
21557+ u.create->mfsrr_watermark);
21558+ break;
21559+ case AuWbrCreate_MFSRRV:
392086de 21560+ case AuWbrCreate_PMFSRRV:
1facf9fc 21561+ AuDbg("%llu watermark, %d sec\n",
21562+ u.create->mfsrr_watermark,
21563+ u.create->mfs_second);
21564+ break;
21565+ }
21566+ break;
21567+ case Opt_wbr_copyup:
21568+ AuDbg("copyup %d, %s\n", opt->wbr_copyup,
21569+ au_optstr_wbr_copyup(opt->wbr_copyup));
21570+ break;
21571+ default:
21572+ BUG();
21573+ }
21574+ opt++;
21575+ }
21576+#endif
21577+}
21578+
21579+void au_opts_free(struct au_opts *opts)
21580+{
21581+ struct au_opt *opt;
21582+
21583+ opt = opts->opt;
21584+ while (opt->type != Opt_tail) {
21585+ switch (opt->type) {
21586+ case Opt_add:
21587+ case Opt_append:
21588+ case Opt_prepend:
21589+ path_put(&opt->add.path);
21590+ break;
21591+ case Opt_del:
21592+ case Opt_idel:
21593+ path_put(&opt->del.h_path);
21594+ break;
21595+ case Opt_mod:
21596+ case Opt_imod:
21597+ dput(opt->mod.h_root);
21598+ break;
21599+ case Opt_xino:
21600+ fput(opt->xino.file);
21601+ break;
21602+ }
21603+ opt++;
21604+ }
21605+}
21606+
21607+static int opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
21608+ aufs_bindex_t bindex)
21609+{
21610+ int err;
21611+ struct au_opt_add *add = &opt->add;
21612+ char *p;
21613+
21614+ add->bindex = bindex;
1e00d052 21615+ add->perm = AuBrPerm_RO;
1facf9fc 21616+ add->pathname = opt_str;
21617+ p = strchr(opt_str, '=');
21618+ if (p) {
21619+ *p++ = 0;
21620+ if (*p)
21621+ add->perm = br_perm_val(p);
21622+ }
21623+
21624+ err = vfsub_kern_path(add->pathname, lkup_dirflags, &add->path);
21625+ if (!err) {
21626+ if (!p) {
21627+ add->perm = AuBrPerm_RO;
21628+ if (au_test_fs_rr(add->path.dentry->d_sb))
21629+ add->perm = AuBrPerm_RR;
21630+ else if (!bindex && !(sb_flags & MS_RDONLY))
21631+ add->perm = AuBrPerm_RW;
21632+ }
21633+ opt->type = Opt_add;
21634+ goto out;
21635+ }
4a4d8108 21636+ pr_err("lookup failed %s (%d)\n", add->pathname, err);
1facf9fc 21637+ err = -EINVAL;
21638+
4f0767ce 21639+out:
1facf9fc 21640+ return err;
21641+}
21642+
21643+static int au_opts_parse_del(struct au_opt_del *del, substring_t args[])
21644+{
21645+ int err;
21646+
21647+ del->pathname = args[0].from;
21648+ AuDbg("del path %s\n", del->pathname);
21649+
21650+ err = vfsub_kern_path(del->pathname, lkup_dirflags, &del->h_path);
21651+ if (unlikely(err))
4a4d8108 21652+ pr_err("lookup failed %s (%d)\n", del->pathname, err);
1facf9fc 21653+
21654+ return err;
21655+}
21656+
21657+#if 0 /* reserved for future use */
21658+static int au_opts_parse_idel(struct super_block *sb, aufs_bindex_t bindex,
21659+ struct au_opt_del *del, substring_t args[])
21660+{
21661+ int err;
21662+ struct dentry *root;
21663+
21664+ err = -EINVAL;
21665+ root = sb->s_root;
21666+ aufs_read_lock(root, AuLock_FLUSH);
21667+ if (bindex < 0 || au_sbend(sb) < bindex) {
4a4d8108 21668+ pr_err("out of bounds, %d\n", bindex);
1facf9fc 21669+ goto out;
21670+ }
21671+
21672+ err = 0;
21673+ del->h_path.dentry = dget(au_h_dptr(root, bindex));
21674+ del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
21675+
4f0767ce 21676+out:
1facf9fc 21677+ aufs_read_unlock(root, !AuLock_IR);
21678+ return err;
21679+}
21680+#endif
21681+
4a4d8108
AM
21682+static int noinline_for_stack
21683+au_opts_parse_mod(struct au_opt_mod *mod, substring_t args[])
1facf9fc 21684+{
21685+ int err;
21686+ struct path path;
21687+ char *p;
21688+
21689+ err = -EINVAL;
21690+ mod->path = args[0].from;
21691+ p = strchr(mod->path, '=');
21692+ if (unlikely(!p)) {
4a4d8108 21693+ pr_err("no permssion %s\n", args[0].from);
1facf9fc 21694+ goto out;
21695+ }
21696+
21697+ *p++ = 0;
21698+ err = vfsub_kern_path(mod->path, lkup_dirflags, &path);
21699+ if (unlikely(err)) {
4a4d8108 21700+ pr_err("lookup failed %s (%d)\n", mod->path, err);
1facf9fc 21701+ goto out;
21702+ }
21703+
21704+ mod->perm = br_perm_val(p);
21705+ AuDbg("mod path %s, perm 0x%x, %s\n", mod->path, mod->perm, p);
21706+ mod->h_root = dget(path.dentry);
21707+ path_put(&path);
21708+
4f0767ce 21709+out:
1facf9fc 21710+ return err;
21711+}
21712+
21713+#if 0 /* reserved for future use */
21714+static int au_opts_parse_imod(struct super_block *sb, aufs_bindex_t bindex,
21715+ struct au_opt_mod *mod, substring_t args[])
21716+{
21717+ int err;
21718+ struct dentry *root;
21719+
21720+ err = -EINVAL;
21721+ root = sb->s_root;
21722+ aufs_read_lock(root, AuLock_FLUSH);
21723+ if (bindex < 0 || au_sbend(sb) < bindex) {
4a4d8108 21724+ pr_err("out of bounds, %d\n", bindex);
1facf9fc 21725+ goto out;
21726+ }
21727+
21728+ err = 0;
21729+ mod->perm = br_perm_val(args[1].from);
21730+ AuDbg("mod path %s, perm 0x%x, %s\n",
21731+ mod->path, mod->perm, args[1].from);
21732+ mod->h_root = dget(au_h_dptr(root, bindex));
21733+
4f0767ce 21734+out:
1facf9fc 21735+ aufs_read_unlock(root, !AuLock_IR);
21736+ return err;
21737+}
21738+#endif
21739+
21740+static int au_opts_parse_xino(struct super_block *sb, struct au_opt_xino *xino,
21741+ substring_t args[])
21742+{
21743+ int err;
21744+ struct file *file;
21745+
21746+ file = au_xino_create(sb, args[0].from, /*silent*/0);
21747+ err = PTR_ERR(file);
21748+ if (IS_ERR(file))
21749+ goto out;
21750+
21751+ err = -EINVAL;
21752+ if (unlikely(file->f_dentry->d_sb == sb)) {
21753+ fput(file);
4a4d8108 21754+ pr_err("%s must be outside\n", args[0].from);
1facf9fc 21755+ goto out;
21756+ }
21757+
21758+ err = 0;
21759+ xino->file = file;
21760+ xino->path = args[0].from;
21761+
4f0767ce 21762+out:
1facf9fc 21763+ return err;
21764+}
21765+
4a4d8108
AM
21766+static int noinline_for_stack
21767+au_opts_parse_xino_itrunc_path(struct super_block *sb,
21768+ struct au_opt_xino_itrunc *xino_itrunc,
21769+ substring_t args[])
1facf9fc 21770+{
21771+ int err;
21772+ aufs_bindex_t bend, bindex;
21773+ struct path path;
21774+ struct dentry *root;
21775+
21776+ err = vfsub_kern_path(args[0].from, lkup_dirflags, &path);
21777+ if (unlikely(err)) {
4a4d8108 21778+ pr_err("lookup failed %s (%d)\n", args[0].from, err);
1facf9fc 21779+ goto out;
21780+ }
21781+
21782+ xino_itrunc->bindex = -1;
21783+ root = sb->s_root;
21784+ aufs_read_lock(root, AuLock_FLUSH);
21785+ bend = au_sbend(sb);
21786+ for (bindex = 0; bindex <= bend; bindex++) {
21787+ if (au_h_dptr(root, bindex) == path.dentry) {
21788+ xino_itrunc->bindex = bindex;
21789+ break;
21790+ }
21791+ }
21792+ aufs_read_unlock(root, !AuLock_IR);
21793+ path_put(&path);
21794+
21795+ if (unlikely(xino_itrunc->bindex < 0)) {
4a4d8108 21796+ pr_err("no such branch %s\n", args[0].from);
1facf9fc 21797+ err = -EINVAL;
21798+ }
21799+
4f0767ce 21800+out:
1facf9fc 21801+ return err;
21802+}
21803+
21804+/* called without aufs lock */
21805+int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts)
21806+{
21807+ int err, n, token;
21808+ aufs_bindex_t bindex;
21809+ unsigned char skipped;
21810+ struct dentry *root;
21811+ struct au_opt *opt, *opt_tail;
21812+ char *opt_str;
21813+ /* reduce the stack space */
21814+ union {
21815+ struct au_opt_xino_itrunc *xino_itrunc;
21816+ struct au_opt_wbr_create *create;
21817+ } u;
21818+ struct {
21819+ substring_t args[MAX_OPT_ARGS];
21820+ } *a;
21821+
21822+ err = -ENOMEM;
21823+ a = kmalloc(sizeof(*a), GFP_NOFS);
21824+ if (unlikely(!a))
21825+ goto out;
21826+
21827+ root = sb->s_root;
21828+ err = 0;
21829+ bindex = 0;
21830+ opt = opts->opt;
21831+ opt_tail = opt + opts->max_opt - 1;
21832+ opt->type = Opt_tail;
21833+ while (!err && (opt_str = strsep(&str, ",")) && *opt_str) {
21834+ err = -EINVAL;
21835+ skipped = 0;
21836+ token = match_token(opt_str, options, a->args);
21837+ switch (token) {
21838+ case Opt_br:
21839+ err = 0;
21840+ while (!err && (opt_str = strsep(&a->args[0].from, ":"))
21841+ && *opt_str) {
21842+ err = opt_add(opt, opt_str, opts->sb_flags,
21843+ bindex++);
21844+ if (unlikely(!err && ++opt > opt_tail)) {
21845+ err = -E2BIG;
21846+ break;
21847+ }
21848+ opt->type = Opt_tail;
21849+ skipped = 1;
21850+ }
21851+ break;
21852+ case Opt_add:
21853+ if (unlikely(match_int(&a->args[0], &n))) {
4a4d8108 21854+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21855+ break;
21856+ }
21857+ bindex = n;
21858+ err = opt_add(opt, a->args[1].from, opts->sb_flags,
21859+ bindex);
21860+ if (!err)
21861+ opt->type = token;
21862+ break;
21863+ case Opt_append:
21864+ err = opt_add(opt, a->args[0].from, opts->sb_flags,
21865+ /*dummy bindex*/1);
21866+ if (!err)
21867+ opt->type = token;
21868+ break;
21869+ case Opt_prepend:
21870+ err = opt_add(opt, a->args[0].from, opts->sb_flags,
21871+ /*bindex*/0);
21872+ if (!err)
21873+ opt->type = token;
21874+ break;
21875+ case Opt_del:
21876+ err = au_opts_parse_del(&opt->del, a->args);
21877+ if (!err)
21878+ opt->type = token;
21879+ break;
21880+#if 0 /* reserved for future use */
21881+ case Opt_idel:
21882+ del->pathname = "(indexed)";
21883+ if (unlikely(match_int(&args[0], &n))) {
4a4d8108 21884+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21885+ break;
21886+ }
21887+ err = au_opts_parse_idel(sb, n, &opt->del, a->args);
21888+ if (!err)
21889+ opt->type = token;
21890+ break;
21891+#endif
21892+ case Opt_mod:
21893+ err = au_opts_parse_mod(&opt->mod, a->args);
21894+ if (!err)
21895+ opt->type = token;
21896+ break;
21897+#ifdef IMOD /* reserved for future use */
21898+ case Opt_imod:
21899+ u.mod->path = "(indexed)";
21900+ if (unlikely(match_int(&a->args[0], &n))) {
4a4d8108 21901+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21902+ break;
21903+ }
21904+ err = au_opts_parse_imod(sb, n, &opt->mod, a->args);
21905+ if (!err)
21906+ opt->type = token;
21907+ break;
21908+#endif
21909+ case Opt_xino:
21910+ err = au_opts_parse_xino(sb, &opt->xino, a->args);
21911+ if (!err)
21912+ opt->type = token;
21913+ break;
21914+
21915+ case Opt_trunc_xino_path:
21916+ err = au_opts_parse_xino_itrunc_path
21917+ (sb, &opt->xino_itrunc, a->args);
21918+ if (!err)
21919+ opt->type = token;
21920+ break;
21921+
21922+ case Opt_itrunc_xino:
21923+ u.xino_itrunc = &opt->xino_itrunc;
21924+ if (unlikely(match_int(&a->args[0], &n))) {
4a4d8108 21925+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21926+ break;
21927+ }
21928+ u.xino_itrunc->bindex = n;
21929+ aufs_read_lock(root, AuLock_FLUSH);
21930+ if (n < 0 || au_sbend(sb) < n) {
4a4d8108 21931+ pr_err("out of bounds, %d\n", n);
1facf9fc 21932+ aufs_read_unlock(root, !AuLock_IR);
21933+ break;
21934+ }
21935+ aufs_read_unlock(root, !AuLock_IR);
21936+ err = 0;
21937+ opt->type = token;
21938+ break;
21939+
21940+ case Opt_dirwh:
21941+ if (unlikely(match_int(&a->args[0], &opt->dirwh)))
21942+ break;
21943+ err = 0;
21944+ opt->type = token;
21945+ break;
21946+
21947+ case Opt_rdcache:
027c5e7a
AM
21948+ if (unlikely(match_int(&a->args[0], &n))) {
21949+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21950+ break;
027c5e7a
AM
21951+ }
21952+ if (unlikely(n > AUFS_RDCACHE_MAX)) {
21953+ pr_err("rdcache must be smaller than %d\n",
21954+ AUFS_RDCACHE_MAX);
21955+ break;
21956+ }
21957+ opt->rdcache = n;
1facf9fc 21958+ err = 0;
21959+ opt->type = token;
21960+ break;
21961+ case Opt_rdblk:
21962+ if (unlikely(match_int(&a->args[0], &n)
1308ab2a 21963+ || n < 0
1facf9fc 21964+ || n > KMALLOC_MAX_SIZE)) {
4a4d8108 21965+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21966+ break;
21967+ }
1308ab2a 21968+ if (unlikely(n && n < NAME_MAX)) {
4a4d8108
AM
21969+ pr_err("rdblk must be larger than %d\n",
21970+ NAME_MAX);
1facf9fc 21971+ break;
21972+ }
21973+ opt->rdblk = n;
21974+ err = 0;
21975+ opt->type = token;
21976+ break;
21977+ case Opt_rdhash:
21978+ if (unlikely(match_int(&a->args[0], &n)
1308ab2a 21979+ || n < 0
1facf9fc 21980+ || n * sizeof(struct hlist_head)
21981+ > KMALLOC_MAX_SIZE)) {
4a4d8108 21982+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 21983+ break;
21984+ }
21985+ opt->rdhash = n;
21986+ err = 0;
21987+ opt->type = token;
21988+ break;
21989+
21990+ case Opt_trunc_xino:
21991+ case Opt_notrunc_xino:
21992+ case Opt_noxino:
21993+ case Opt_trunc_xib:
21994+ case Opt_notrunc_xib:
dece6358
AM
21995+ case Opt_shwh:
21996+ case Opt_noshwh:
1facf9fc 21997+ case Opt_plink:
21998+ case Opt_noplink:
21999+ case Opt_list_plink:
4a4d8108
AM
22000+ case Opt_dio:
22001+ case Opt_nodio:
1facf9fc 22002+ case Opt_diropq_a:
22003+ case Opt_diropq_w:
22004+ case Opt_warn_perm:
22005+ case Opt_nowarn_perm:
22006+ case Opt_refrof:
22007+ case Opt_norefrof:
22008+ case Opt_verbose:
22009+ case Opt_noverbose:
22010+ case Opt_sum:
22011+ case Opt_nosum:
22012+ case Opt_wsum:
dece6358
AM
22013+ case Opt_rdblk_def:
22014+ case Opt_rdhash_def:
1facf9fc 22015+ err = 0;
22016+ opt->type = token;
22017+ break;
22018+
22019+ case Opt_udba:
22020+ opt->udba = udba_val(a->args[0].from);
22021+ if (opt->udba >= 0) {
22022+ err = 0;
22023+ opt->type = token;
22024+ } else
4a4d8108 22025+ pr_err("wrong value, %s\n", opt_str);
1facf9fc 22026+ break;
22027+
22028+ case Opt_wbr_create:
22029+ u.create = &opt->wbr_create;
22030+ u.create->wbr_create
22031+ = au_wbr_create_val(a->args[0].from, u.create);
22032+ if (u.create->wbr_create >= 0) {
22033+ err = 0;
22034+ opt->type = token;
22035+ } else
4a4d8108 22036+ pr_err("wrong value, %s\n", opt_str);
1facf9fc 22037+ break;
22038+ case Opt_wbr_copyup:
22039+ opt->wbr_copyup = au_wbr_copyup_val(a->args[0].from);
22040+ if (opt->wbr_copyup >= 0) {
22041+ err = 0;
22042+ opt->type = token;
22043+ } else
4a4d8108 22044+ pr_err("wrong value, %s\n", opt_str);
1facf9fc 22045+ break;
22046+
22047+ case Opt_ignore:
0c3ec466 22048+ pr_warn("ignored %s\n", opt_str);
1facf9fc 22049+ /*FALLTHROUGH*/
22050+ case Opt_ignore_silent:
22051+ skipped = 1;
22052+ err = 0;
22053+ break;
22054+ case Opt_err:
4a4d8108 22055+ pr_err("unknown option %s\n", opt_str);
1facf9fc 22056+ break;
22057+ }
22058+
22059+ if (!err && !skipped) {
22060+ if (unlikely(++opt > opt_tail)) {
22061+ err = -E2BIG;
22062+ opt--;
22063+ opt->type = Opt_tail;
22064+ break;
22065+ }
22066+ opt->type = Opt_tail;
22067+ }
22068+ }
22069+
22070+ kfree(a);
22071+ dump_opts(opts);
22072+ if (unlikely(err))
22073+ au_opts_free(opts);
22074+
4f0767ce 22075+out:
1facf9fc 22076+ return err;
22077+}
22078+
22079+static int au_opt_wbr_create(struct super_block *sb,
22080+ struct au_opt_wbr_create *create)
22081+{
22082+ int err;
22083+ struct au_sbinfo *sbinfo;
22084+
dece6358
AM
22085+ SiMustWriteLock(sb);
22086+
1facf9fc 22087+ err = 1; /* handled */
22088+ sbinfo = au_sbi(sb);
22089+ if (sbinfo->si_wbr_create_ops->fin) {
22090+ err = sbinfo->si_wbr_create_ops->fin(sb);
22091+ if (!err)
22092+ err = 1;
22093+ }
22094+
22095+ sbinfo->si_wbr_create = create->wbr_create;
22096+ sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
22097+ switch (create->wbr_create) {
22098+ case AuWbrCreate_MFSRRV:
22099+ case AuWbrCreate_MFSRR:
392086de
AM
22100+ case AuWbrCreate_PMFSRR:
22101+ case AuWbrCreate_PMFSRRV:
1facf9fc 22102+ sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
22103+ /*FALLTHROUGH*/
22104+ case AuWbrCreate_MFS:
22105+ case AuWbrCreate_MFSV:
22106+ case AuWbrCreate_PMFS:
22107+ case AuWbrCreate_PMFSV:
e49829fe
JR
22108+ sbinfo->si_wbr_mfs.mfs_expire
22109+ = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
1facf9fc 22110+ break;
22111+ }
22112+
22113+ if (sbinfo->si_wbr_create_ops->init)
22114+ sbinfo->si_wbr_create_ops->init(sb); /* ignore */
22115+
22116+ return err;
22117+}
22118+
22119+/*
22120+ * returns,
22121+ * plus: processed without an error
22122+ * zero: unprocessed
22123+ */
22124+static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
22125+ struct au_opts *opts)
22126+{
22127+ int err;
22128+ struct au_sbinfo *sbinfo;
22129+
dece6358
AM
22130+ SiMustWriteLock(sb);
22131+
1facf9fc 22132+ err = 1; /* handled */
22133+ sbinfo = au_sbi(sb);
22134+ switch (opt->type) {
22135+ case Opt_udba:
22136+ sbinfo->si_mntflags &= ~AuOptMask_UDBA;
22137+ sbinfo->si_mntflags |= opt->udba;
22138+ opts->given_udba |= opt->udba;
22139+ break;
22140+
22141+ case Opt_plink:
22142+ au_opt_set(sbinfo->si_mntflags, PLINK);
22143+ break;
22144+ case Opt_noplink:
22145+ if (au_opt_test(sbinfo->si_mntflags, PLINK))
e49829fe 22146+ au_plink_put(sb, /*verbose*/1);
1facf9fc 22147+ au_opt_clr(sbinfo->si_mntflags, PLINK);
22148+ break;
22149+ case Opt_list_plink:
22150+ if (au_opt_test(sbinfo->si_mntflags, PLINK))
22151+ au_plink_list(sb);
22152+ break;
22153+
4a4d8108
AM
22154+ case Opt_dio:
22155+ au_opt_set(sbinfo->si_mntflags, DIO);
22156+ au_fset_opts(opts->flags, REFRESH_DYAOP);
22157+ break;
22158+ case Opt_nodio:
22159+ au_opt_clr(sbinfo->si_mntflags, DIO);
22160+ au_fset_opts(opts->flags, REFRESH_DYAOP);
22161+ break;
22162+
1facf9fc 22163+ case Opt_diropq_a:
22164+ au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
22165+ break;
22166+ case Opt_diropq_w:
22167+ au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
22168+ break;
22169+
22170+ case Opt_warn_perm:
22171+ au_opt_set(sbinfo->si_mntflags, WARN_PERM);
22172+ break;
22173+ case Opt_nowarn_perm:
22174+ au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
22175+ break;
22176+
22177+ case Opt_refrof:
22178+ au_opt_set(sbinfo->si_mntflags, REFROF);
22179+ break;
22180+ case Opt_norefrof:
22181+ au_opt_clr(sbinfo->si_mntflags, REFROF);
22182+ break;
22183+
22184+ case Opt_verbose:
22185+ au_opt_set(sbinfo->si_mntflags, VERBOSE);
22186+ break;
22187+ case Opt_noverbose:
22188+ au_opt_clr(sbinfo->si_mntflags, VERBOSE);
22189+ break;
22190+
22191+ case Opt_sum:
22192+ au_opt_set(sbinfo->si_mntflags, SUM);
22193+ break;
22194+ case Opt_wsum:
22195+ au_opt_clr(sbinfo->si_mntflags, SUM);
22196+ au_opt_set(sbinfo->si_mntflags, SUM_W);
22197+ case Opt_nosum:
22198+ au_opt_clr(sbinfo->si_mntflags, SUM);
22199+ au_opt_clr(sbinfo->si_mntflags, SUM_W);
22200+ break;
22201+
22202+ case Opt_wbr_create:
22203+ err = au_opt_wbr_create(sb, &opt->wbr_create);
22204+ break;
22205+ case Opt_wbr_copyup:
22206+ sbinfo->si_wbr_copyup = opt->wbr_copyup;
22207+ sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
22208+ break;
22209+
22210+ case Opt_dirwh:
22211+ sbinfo->si_dirwh = opt->dirwh;
22212+ break;
22213+
22214+ case Opt_rdcache:
e49829fe
JR
22215+ sbinfo->si_rdcache
22216+ = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
1facf9fc 22217+ break;
22218+ case Opt_rdblk:
22219+ sbinfo->si_rdblk = opt->rdblk;
22220+ break;
dece6358
AM
22221+ case Opt_rdblk_def:
22222+ sbinfo->si_rdblk = AUFS_RDBLK_DEF;
22223+ break;
1facf9fc 22224+ case Opt_rdhash:
22225+ sbinfo->si_rdhash = opt->rdhash;
22226+ break;
dece6358
AM
22227+ case Opt_rdhash_def:
22228+ sbinfo->si_rdhash = AUFS_RDHASH_DEF;
22229+ break;
22230+
22231+ case Opt_shwh:
22232+ au_opt_set(sbinfo->si_mntflags, SHWH);
22233+ break;
22234+ case Opt_noshwh:
22235+ au_opt_clr(sbinfo->si_mntflags, SHWH);
22236+ break;
1facf9fc 22237+
22238+ case Opt_trunc_xino:
22239+ au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
22240+ break;
22241+ case Opt_notrunc_xino:
22242+ au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
22243+ break;
22244+
22245+ case Opt_trunc_xino_path:
22246+ case Opt_itrunc_xino:
22247+ err = au_xino_trunc(sb, opt->xino_itrunc.bindex);
22248+ if (!err)
22249+ err = 1;
22250+ break;
22251+
22252+ case Opt_trunc_xib:
22253+ au_fset_opts(opts->flags, TRUNC_XIB);
22254+ break;
22255+ case Opt_notrunc_xib:
22256+ au_fclr_opts(opts->flags, TRUNC_XIB);
22257+ break;
22258+
22259+ default:
22260+ err = 0;
22261+ break;
22262+ }
22263+
22264+ return err;
22265+}
22266+
22267+/*
22268+ * returns tri-state.
22269+ * plus: processed without an error
22270+ * zero: unprocessed
22271+ * minus: error
22272+ */
22273+static int au_opt_br(struct super_block *sb, struct au_opt *opt,
22274+ struct au_opts *opts)
22275+{
22276+ int err, do_refresh;
22277+
22278+ err = 0;
22279+ switch (opt->type) {
22280+ case Opt_append:
22281+ opt->add.bindex = au_sbend(sb) + 1;
22282+ if (opt->add.bindex < 0)
22283+ opt->add.bindex = 0;
22284+ goto add;
22285+ case Opt_prepend:
22286+ opt->add.bindex = 0;
f6b6e03d 22287+ add: /* indented label */
1facf9fc 22288+ case Opt_add:
22289+ err = au_br_add(sb, &opt->add,
22290+ au_ftest_opts(opts->flags, REMOUNT));
22291+ if (!err) {
22292+ err = 1;
027c5e7a 22293+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 22294+ }
22295+ break;
22296+
22297+ case Opt_del:
22298+ case Opt_idel:
22299+ err = au_br_del(sb, &opt->del,
22300+ au_ftest_opts(opts->flags, REMOUNT));
22301+ if (!err) {
22302+ err = 1;
22303+ au_fset_opts(opts->flags, TRUNC_XIB);
027c5e7a 22304+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 22305+ }
22306+ break;
22307+
22308+ case Opt_mod:
22309+ case Opt_imod:
22310+ err = au_br_mod(sb, &opt->mod,
22311+ au_ftest_opts(opts->flags, REMOUNT),
22312+ &do_refresh);
22313+ if (!err) {
22314+ err = 1;
027c5e7a
AM
22315+ if (do_refresh)
22316+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 22317+ }
22318+ break;
22319+ }
22320+
22321+ return err;
22322+}
22323+
22324+static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
22325+ struct au_opt_xino **opt_xino,
22326+ struct au_opts *opts)
22327+{
22328+ int err;
22329+ aufs_bindex_t bend, bindex;
22330+ struct dentry *root, *parent, *h_root;
22331+
22332+ err = 0;
22333+ switch (opt->type) {
22334+ case Opt_xino:
22335+ err = au_xino_set(sb, &opt->xino,
22336+ !!au_ftest_opts(opts->flags, REMOUNT));
22337+ if (unlikely(err))
22338+ break;
22339+
22340+ *opt_xino = &opt->xino;
22341+ au_xino_brid_set(sb, -1);
22342+
22343+ /* safe d_parent access */
22344+ parent = opt->xino.file->f_dentry->d_parent;
22345+ root = sb->s_root;
22346+ bend = au_sbend(sb);
22347+ for (bindex = 0; bindex <= bend; bindex++) {
22348+ h_root = au_h_dptr(root, bindex);
22349+ if (h_root == parent) {
22350+ au_xino_brid_set(sb, au_sbr_id(sb, bindex));
22351+ break;
22352+ }
22353+ }
22354+ break;
22355+
22356+ case Opt_noxino:
22357+ au_xino_clr(sb);
22358+ au_xino_brid_set(sb, -1);
22359+ *opt_xino = (void *)-1;
22360+ break;
22361+ }
22362+
22363+ return err;
22364+}
22365+
22366+int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
22367+ unsigned int pending)
22368+{
22369+ int err;
22370+ aufs_bindex_t bindex, bend;
22371+ unsigned char do_plink, skip, do_free;
22372+ struct au_branch *br;
22373+ struct au_wbr *wbr;
22374+ struct dentry *root;
22375+ struct inode *dir, *h_dir;
22376+ struct au_sbinfo *sbinfo;
22377+ struct au_hinode *hdir;
22378+
dece6358
AM
22379+ SiMustAnyLock(sb);
22380+
1facf9fc 22381+ sbinfo = au_sbi(sb);
22382+ AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
22383+
dece6358
AM
22384+ if (!(sb_flags & MS_RDONLY)) {
22385+ if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
0c3ec466 22386+ pr_warn("first branch should be rw\n");
dece6358 22387+ if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
0c3ec466 22388+ pr_warn("shwh should be used with ro\n");
dece6358 22389+ }
1facf9fc 22390+
4a4d8108 22391+ if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
1facf9fc 22392+ && !au_opt_test(sbinfo->si_mntflags, XINO))
0c3ec466 22393+ pr_warn("udba=*notify requires xino\n");
1facf9fc 22394+
22395+ err = 0;
22396+ root = sb->s_root;
4a4d8108 22397+ dir = root->d_inode;
1facf9fc 22398+ do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
22399+ bend = au_sbend(sb);
22400+ for (bindex = 0; !err && bindex <= bend; bindex++) {
22401+ skip = 0;
22402+ h_dir = au_h_iptr(dir, bindex);
22403+ br = au_sbr(sb, bindex);
22404+ do_free = 0;
22405+
22406+ wbr = br->br_wbr;
22407+ if (wbr)
22408+ wbr_wh_read_lock(wbr);
22409+
1e00d052 22410+ if (!au_br_writable(br->br_perm)) {
1facf9fc 22411+ do_free = !!wbr;
22412+ skip = (!wbr
22413+ || (!wbr->wbr_whbase
22414+ && !wbr->wbr_plink
22415+ && !wbr->wbr_orph));
1e00d052 22416+ } else if (!au_br_wh_linkable(br->br_perm)) {
1facf9fc 22417+ /* skip = (!br->br_whbase && !br->br_orph); */
22418+ skip = (!wbr || !wbr->wbr_whbase);
22419+ if (skip && wbr) {
22420+ if (do_plink)
22421+ skip = !!wbr->wbr_plink;
22422+ else
22423+ skip = !wbr->wbr_plink;
22424+ }
1e00d052 22425+ } else {
1facf9fc 22426+ /* skip = (br->br_whbase && br->br_ohph); */
22427+ skip = (wbr && wbr->wbr_whbase);
22428+ if (skip) {
22429+ if (do_plink)
22430+ skip = !!wbr->wbr_plink;
22431+ else
22432+ skip = !wbr->wbr_plink;
22433+ }
1facf9fc 22434+ }
22435+ if (wbr)
22436+ wbr_wh_read_unlock(wbr);
22437+
22438+ if (skip)
22439+ continue;
22440+
22441+ hdir = au_hi(dir, bindex);
4a4d8108 22442+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
1facf9fc 22443+ if (wbr)
22444+ wbr_wh_write_lock(wbr);
86dc4139 22445+ err = au_wh_init(br, sb);
1facf9fc 22446+ if (wbr)
22447+ wbr_wh_write_unlock(wbr);
4a4d8108 22448+ au_hn_imtx_unlock(hdir);
1facf9fc 22449+
22450+ if (!err && do_free) {
22451+ kfree(wbr);
22452+ br->br_wbr = NULL;
22453+ }
22454+ }
22455+
22456+ return err;
22457+}
22458+
22459+int au_opts_mount(struct super_block *sb, struct au_opts *opts)
22460+{
22461+ int err;
22462+ unsigned int tmp;
027c5e7a 22463+ aufs_bindex_t bindex, bend;
1facf9fc 22464+ struct au_opt *opt;
22465+ struct au_opt_xino *opt_xino, xino;
22466+ struct au_sbinfo *sbinfo;
027c5e7a 22467+ struct au_branch *br;
1facf9fc 22468+
dece6358
AM
22469+ SiMustWriteLock(sb);
22470+
1facf9fc 22471+ err = 0;
22472+ opt_xino = NULL;
22473+ opt = opts->opt;
22474+ while (err >= 0 && opt->type != Opt_tail)
22475+ err = au_opt_simple(sb, opt++, opts);
22476+ if (err > 0)
22477+ err = 0;
22478+ else if (unlikely(err < 0))
22479+ goto out;
22480+
22481+ /* disable xino and udba temporary */
22482+ sbinfo = au_sbi(sb);
22483+ tmp = sbinfo->si_mntflags;
22484+ au_opt_clr(sbinfo->si_mntflags, XINO);
22485+ au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
22486+
22487+ opt = opts->opt;
22488+ while (err >= 0 && opt->type != Opt_tail)
22489+ err = au_opt_br(sb, opt++, opts);
22490+ if (err > 0)
22491+ err = 0;
22492+ else if (unlikely(err < 0))
22493+ goto out;
22494+
22495+ bend = au_sbend(sb);
22496+ if (unlikely(bend < 0)) {
22497+ err = -EINVAL;
4a4d8108 22498+ pr_err("no branches\n");
1facf9fc 22499+ goto out;
22500+ }
22501+
22502+ if (au_opt_test(tmp, XINO))
22503+ au_opt_set(sbinfo->si_mntflags, XINO);
22504+ opt = opts->opt;
22505+ while (!err && opt->type != Opt_tail)
22506+ err = au_opt_xino(sb, opt++, &opt_xino, opts);
22507+ if (unlikely(err))
22508+ goto out;
22509+
22510+ err = au_opts_verify(sb, sb->s_flags, tmp);
22511+ if (unlikely(err))
22512+ goto out;
22513+
22514+ /* restore xino */
22515+ if (au_opt_test(tmp, XINO) && !opt_xino) {
22516+ xino.file = au_xino_def(sb);
22517+ err = PTR_ERR(xino.file);
22518+ if (IS_ERR(xino.file))
22519+ goto out;
22520+
22521+ err = au_xino_set(sb, &xino, /*remount*/0);
22522+ fput(xino.file);
22523+ if (unlikely(err))
22524+ goto out;
22525+ }
22526+
22527+ /* restore udba */
027c5e7a 22528+ tmp &= AuOptMask_UDBA;
1facf9fc 22529+ sbinfo->si_mntflags &= ~AuOptMask_UDBA;
027c5e7a
AM
22530+ sbinfo->si_mntflags |= tmp;
22531+ bend = au_sbend(sb);
22532+ for (bindex = 0; bindex <= bend; bindex++) {
22533+ br = au_sbr(sb, bindex);
22534+ err = au_hnotify_reset_br(tmp, br, br->br_perm);
22535+ if (unlikely(err))
22536+ AuIOErr("hnotify failed on br %d, %d, ignored\n",
22537+ bindex, err);
22538+ /* go on even if err */
22539+ }
4a4d8108 22540+ if (au_opt_test(tmp, UDBA_HNOTIFY)) {
1facf9fc 22541+ struct inode *dir = sb->s_root->d_inode;
4a4d8108 22542+ au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
1facf9fc 22543+ }
22544+
4f0767ce 22545+out:
1facf9fc 22546+ return err;
22547+}
22548+
22549+int au_opts_remount(struct super_block *sb, struct au_opts *opts)
22550+{
22551+ int err, rerr;
22552+ struct inode *dir;
22553+ struct au_opt_xino *opt_xino;
22554+ struct au_opt *opt;
22555+ struct au_sbinfo *sbinfo;
22556+
dece6358
AM
22557+ SiMustWriteLock(sb);
22558+
1facf9fc 22559+ dir = sb->s_root->d_inode;
22560+ sbinfo = au_sbi(sb);
22561+ err = 0;
22562+ opt_xino = NULL;
22563+ opt = opts->opt;
22564+ while (err >= 0 && opt->type != Opt_tail) {
22565+ err = au_opt_simple(sb, opt, opts);
22566+ if (!err)
22567+ err = au_opt_br(sb, opt, opts);
22568+ if (!err)
22569+ err = au_opt_xino(sb, opt, &opt_xino, opts);
22570+ opt++;
22571+ }
22572+ if (err > 0)
22573+ err = 0;
22574+ AuTraceErr(err);
22575+ /* go on even err */
22576+
22577+ rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
22578+ if (unlikely(rerr && !err))
22579+ err = rerr;
22580+
22581+ if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
22582+ rerr = au_xib_trunc(sb);
22583+ if (unlikely(rerr && !err))
22584+ err = rerr;
22585+ }
22586+
22587+ /* will be handled by the caller */
027c5e7a 22588+ if (!au_ftest_opts(opts->flags, REFRESH)
1facf9fc 22589+ && (opts->given_udba || au_opt_test(sbinfo->si_mntflags, XINO)))
027c5e7a 22590+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 22591+
22592+ AuDbg("status 0x%x\n", opts->flags);
22593+ return err;
22594+}
22595+
22596+/* ---------------------------------------------------------------------- */
22597+
22598+unsigned int au_opt_udba(struct super_block *sb)
22599+{
22600+ return au_mntflags(sb) & AuOptMask_UDBA;
22601+}
7f207e10
AM
22602diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
22603--- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 22604+++ linux/fs/aufs/opts.h 2014-06-24 00:14:06.786052822 +0200
523b37e3 22605@@ -0,0 +1,210 @@
1facf9fc 22606+/*
523b37e3 22607+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 22608+ *
22609+ * This program, aufs is free software; you can redistribute it and/or modify
22610+ * it under the terms of the GNU General Public License as published by
22611+ * the Free Software Foundation; either version 2 of the License, or
22612+ * (at your option) any later version.
dece6358
AM
22613+ *
22614+ * This program is distributed in the hope that it will be useful,
22615+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
22616+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22617+ * GNU General Public License for more details.
22618+ *
22619+ * You should have received a copy of the GNU General Public License
523b37e3 22620+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 22621+ */
22622+
22623+/*
22624+ * mount options/flags
22625+ */
22626+
22627+#ifndef __AUFS_OPTS_H__
22628+#define __AUFS_OPTS_H__
22629+
22630+#ifdef __KERNEL__
22631+
dece6358 22632+#include <linux/path.h>
1facf9fc 22633+
dece6358
AM
22634+struct file;
22635+struct super_block;
22636+
1facf9fc 22637+/* ---------------------------------------------------------------------- */
22638+
22639+/* mount flags */
22640+#define AuOpt_XINO 1 /* external inode number bitmap
22641+ and translation table */
22642+#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
22643+#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
22644+#define AuOpt_UDBA_REVAL (1 << 3)
4a4d8108 22645+#define AuOpt_UDBA_HNOTIFY (1 << 4)
dece6358
AM
22646+#define AuOpt_SHWH (1 << 5) /* show whiteout */
22647+#define AuOpt_PLINK (1 << 6) /* pseudo-link */
22648+#define AuOpt_DIRPERM1 (1 << 7) /* unimplemented */
22649+#define AuOpt_REFROF (1 << 8) /* unimplemented */
22650+#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
22651+#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
22652+#define AuOpt_SUM_W (1 << 11) /* unimplemented */
22653+#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
22654+#define AuOpt_VERBOSE (1 << 13) /* busy inode when del-branch */
4a4d8108 22655+#define AuOpt_DIO (1 << 14) /* direct io */
1facf9fc 22656+
4a4d8108
AM
22657+#ifndef CONFIG_AUFS_HNOTIFY
22658+#undef AuOpt_UDBA_HNOTIFY
22659+#define AuOpt_UDBA_HNOTIFY 0
1facf9fc 22660+#endif
dece6358
AM
22661+#ifndef CONFIG_AUFS_SHWH
22662+#undef AuOpt_SHWH
22663+#define AuOpt_SHWH 0
22664+#endif
1facf9fc 22665+
22666+#define AuOpt_Def (AuOpt_XINO \
22667+ | AuOpt_UDBA_REVAL \
22668+ | AuOpt_PLINK \
22669+ /* | AuOpt_DIRPERM1 */ \
22670+ | AuOpt_WARN_PERM)
22671+#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
22672+ | AuOpt_UDBA_REVAL \
4a4d8108 22673+ | AuOpt_UDBA_HNOTIFY)
1facf9fc 22674+
22675+#define au_opt_test(flags, name) (flags & AuOpt_##name)
22676+#define au_opt_set(flags, name) do { \
22677+ BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
22678+ ((flags) |= AuOpt_##name); \
22679+} while (0)
22680+#define au_opt_set_udba(flags, name) do { \
22681+ (flags) &= ~AuOptMask_UDBA; \
22682+ ((flags) |= AuOpt_##name); \
22683+} while (0)
7f207e10
AM
22684+#define au_opt_clr(flags, name) do { \
22685+ ((flags) &= ~AuOpt_##name); \
22686+} while (0)
1facf9fc 22687+
e49829fe
JR
22688+static inline unsigned int au_opts_plink(unsigned int mntflags)
22689+{
22690+#ifdef CONFIG_PROC_FS
22691+ return mntflags;
22692+#else
22693+ return mntflags & ~AuOpt_PLINK;
22694+#endif
22695+}
22696+
1facf9fc 22697+/* ---------------------------------------------------------------------- */
22698+
22699+/* policies to select one among multiple writable branches */
22700+enum {
22701+ AuWbrCreate_TDP, /* top down parent */
22702+ AuWbrCreate_RR, /* round robin */
22703+ AuWbrCreate_MFS, /* most free space */
22704+ AuWbrCreate_MFSV, /* mfs with seconds */
22705+ AuWbrCreate_MFSRR, /* mfs then rr */
22706+ AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
22707+ AuWbrCreate_PMFS, /* parent and mfs */
22708+ AuWbrCreate_PMFSV, /* parent and mfs with seconds */
392086de
AM
22709+ AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
22710+ AuWbrCreate_PMFSRRV, /* plus seconds */
1facf9fc 22711+
22712+ AuWbrCreate_Def = AuWbrCreate_TDP
22713+};
22714+
22715+enum {
22716+ AuWbrCopyup_TDP, /* top down parent */
22717+ AuWbrCopyup_BUP, /* bottom up parent */
22718+ AuWbrCopyup_BU, /* bottom up */
22719+
22720+ AuWbrCopyup_Def = AuWbrCopyup_TDP
22721+};
22722+
22723+/* ---------------------------------------------------------------------- */
22724+
22725+struct au_opt_add {
22726+ aufs_bindex_t bindex;
22727+ char *pathname;
22728+ int perm;
22729+ struct path path;
22730+};
22731+
22732+struct au_opt_del {
22733+ char *pathname;
22734+ struct path h_path;
22735+};
22736+
22737+struct au_opt_mod {
22738+ char *path;
22739+ int perm;
22740+ struct dentry *h_root;
22741+};
22742+
22743+struct au_opt_xino {
22744+ char *path;
22745+ struct file *file;
22746+};
22747+
22748+struct au_opt_xino_itrunc {
22749+ aufs_bindex_t bindex;
22750+};
22751+
22752+struct au_opt_wbr_create {
22753+ int wbr_create;
22754+ int mfs_second;
22755+ unsigned long long mfsrr_watermark;
22756+};
22757+
22758+struct au_opt {
22759+ int type;
22760+ union {
22761+ struct au_opt_xino xino;
22762+ struct au_opt_xino_itrunc xino_itrunc;
22763+ struct au_opt_add add;
22764+ struct au_opt_del del;
22765+ struct au_opt_mod mod;
22766+ int dirwh;
22767+ int rdcache;
22768+ unsigned int rdblk;
22769+ unsigned int rdhash;
22770+ int udba;
22771+ struct au_opt_wbr_create wbr_create;
22772+ int wbr_copyup;
22773+ };
22774+};
22775+
22776+/* opts flags */
22777+#define AuOpts_REMOUNT 1
027c5e7a
AM
22778+#define AuOpts_REFRESH (1 << 1)
22779+#define AuOpts_TRUNC_XIB (1 << 2)
22780+#define AuOpts_REFRESH_DYAOP (1 << 3)
1facf9fc 22781+#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
7f207e10
AM
22782+#define au_fset_opts(flags, name) \
22783+ do { (flags) |= AuOpts_##name; } while (0)
22784+#define au_fclr_opts(flags, name) \
22785+ do { (flags) &= ~AuOpts_##name; } while (0)
1facf9fc 22786+
22787+struct au_opts {
22788+ struct au_opt *opt;
22789+ int max_opt;
22790+
22791+ unsigned int given_udba;
22792+ unsigned int flags;
22793+ unsigned long sb_flags;
22794+};
22795+
22796+/* ---------------------------------------------------------------------- */
22797+
1e00d052 22798+char *au_optstr_br_perm(int brperm);
1facf9fc 22799+const char *au_optstr_udba(int udba);
22800+const char *au_optstr_wbr_copyup(int wbr_copyup);
22801+const char *au_optstr_wbr_create(int wbr_create);
22802+
22803+void au_opts_free(struct au_opts *opts);
22804+int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts);
22805+int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
22806+ unsigned int pending);
22807+int au_opts_mount(struct super_block *sb, struct au_opts *opts);
22808+int au_opts_remount(struct super_block *sb, struct au_opts *opts);
22809+
22810+unsigned int au_opt_udba(struct super_block *sb);
22811+
22812+/* ---------------------------------------------------------------------- */
22813+
22814+#endif /* __KERNEL__ */
22815+#endif /* __AUFS_OPTS_H__ */
7f207e10
AM
22816diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
22817--- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 22818+++ linux/fs/aufs/plink.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 22819@@ -0,0 +1,532 @@
1facf9fc 22820+/*
523b37e3 22821+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 22822+ *
22823+ * This program, aufs is free software; you can redistribute it and/or modify
22824+ * it under the terms of the GNU General Public License as published by
22825+ * the Free Software Foundation; either version 2 of the License, or
22826+ * (at your option) any later version.
dece6358
AM
22827+ *
22828+ * This program is distributed in the hope that it will be useful,
22829+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
22830+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22831+ * GNU General Public License for more details.
22832+ *
22833+ * You should have received a copy of the GNU General Public License
523b37e3 22834+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 22835+ */
22836+
22837+/*
22838+ * pseudo-link
22839+ */
22840+
22841+#include "aufs.h"
22842+
22843+/*
e49829fe 22844+ * the pseudo-link maintenance mode.
1facf9fc 22845+ * during a user process maintains the pseudo-links,
22846+ * prohibit adding a new plink and branch manipulation.
e49829fe
JR
22847+ *
22848+ * Flags
22849+ * NOPLM:
22850+ * For entry functions which will handle plink, and i_mutex is already held
22851+ * in VFS.
22852+ * They cannot wait and should return an error at once.
22853+ * Callers has to check the error.
22854+ * NOPLMW:
22855+ * For entry functions which will handle plink, but i_mutex is not held
22856+ * in VFS.
22857+ * They can wait the plink maintenance mode to finish.
22858+ *
22859+ * They behave like F_SETLK and F_SETLKW.
22860+ * If the caller never handle plink, then both flags are unnecessary.
1facf9fc 22861+ */
e49829fe
JR
22862+
22863+int au_plink_maint(struct super_block *sb, int flags)
1facf9fc 22864+{
e49829fe
JR
22865+ int err;
22866+ pid_t pid, ppid;
22867+ struct au_sbinfo *sbi;
dece6358
AM
22868+
22869+ SiMustAnyLock(sb);
22870+
e49829fe
JR
22871+ err = 0;
22872+ if (!au_opt_test(au_mntflags(sb), PLINK))
22873+ goto out;
22874+
22875+ sbi = au_sbi(sb);
22876+ pid = sbi->si_plink_maint_pid;
22877+ if (!pid || pid == current->pid)
22878+ goto out;
22879+
22880+ /* todo: it highly depends upon /sbin/mount.aufs */
22881+ rcu_read_lock();
22882+ ppid = task_pid_vnr(rcu_dereference(current->real_parent));
22883+ rcu_read_unlock();
22884+ if (pid == ppid)
22885+ goto out;
22886+
22887+ if (au_ftest_lock(flags, NOPLMW)) {
027c5e7a
AM
22888+ /* if there is no i_mutex lock in VFS, we don't need to wait */
22889+ /* AuDebugOn(!lockdep_depth(current)); */
e49829fe
JR
22890+ while (sbi->si_plink_maint_pid) {
22891+ si_read_unlock(sb);
22892+ /* gave up wake_up_bit() */
22893+ wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
22894+
22895+ if (au_ftest_lock(flags, FLUSH))
22896+ au_nwt_flush(&sbi->si_nowait);
22897+ si_noflush_read_lock(sb);
22898+ }
22899+ } else if (au_ftest_lock(flags, NOPLM)) {
22900+ AuDbg("ppid %d, pid %d\n", ppid, pid);
22901+ err = -EAGAIN;
22902+ }
22903+
22904+out:
22905+ return err;
4a4d8108
AM
22906+}
22907+
e49829fe 22908+void au_plink_maint_leave(struct au_sbinfo *sbinfo)
4a4d8108 22909+{
4a4d8108 22910+ spin_lock(&sbinfo->si_plink_maint_lock);
027c5e7a 22911+ sbinfo->si_plink_maint_pid = 0;
4a4d8108 22912+ spin_unlock(&sbinfo->si_plink_maint_lock);
027c5e7a 22913+ wake_up_all(&sbinfo->si_plink_wq);
4a4d8108
AM
22914+}
22915+
e49829fe 22916+int au_plink_maint_enter(struct super_block *sb)
4a4d8108
AM
22917+{
22918+ int err;
4a4d8108
AM
22919+ struct au_sbinfo *sbinfo;
22920+
22921+ err = 0;
4a4d8108
AM
22922+ sbinfo = au_sbi(sb);
22923+ /* make sure i am the only one in this fs */
e49829fe
JR
22924+ si_write_lock(sb, AuLock_FLUSH);
22925+ if (au_opt_test(au_mntflags(sb), PLINK)) {
22926+ spin_lock(&sbinfo->si_plink_maint_lock);
22927+ if (!sbinfo->si_plink_maint_pid)
22928+ sbinfo->si_plink_maint_pid = current->pid;
22929+ else
22930+ err = -EBUSY;
22931+ spin_unlock(&sbinfo->si_plink_maint_lock);
22932+ }
4a4d8108
AM
22933+ si_write_unlock(sb);
22934+
22935+ return err;
1facf9fc 22936+}
22937+
22938+/* ---------------------------------------------------------------------- */
22939+
1facf9fc 22940+#ifdef CONFIG_AUFS_DEBUG
22941+void au_plink_list(struct super_block *sb)
22942+{
86dc4139 22943+ int i;
1facf9fc 22944+ struct au_sbinfo *sbinfo;
86dc4139 22945+ struct hlist_head *plink_hlist;
1facf9fc 22946+ struct pseudo_link *plink;
22947+
dece6358
AM
22948+ SiMustAnyLock(sb);
22949+
1facf9fc 22950+ sbinfo = au_sbi(sb);
22951+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 22952+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 22953+
86dc4139
AM
22954+ for (i = 0; i < AuPlink_NHASH; i++) {
22955+ plink_hlist = &sbinfo->si_plink[i].head;
22956+ rcu_read_lock();
22957+ hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
22958+ AuDbg("%lu\n", plink->inode->i_ino);
22959+ rcu_read_unlock();
22960+ }
1facf9fc 22961+}
22962+#endif
22963+
22964+/* is the inode pseudo-linked? */
22965+int au_plink_test(struct inode *inode)
22966+{
86dc4139 22967+ int found, i;
1facf9fc 22968+ struct au_sbinfo *sbinfo;
86dc4139 22969+ struct hlist_head *plink_hlist;
1facf9fc 22970+ struct pseudo_link *plink;
22971+
22972+ sbinfo = au_sbi(inode->i_sb);
dece6358 22973+ AuRwMustAnyLock(&sbinfo->si_rwsem);
1facf9fc 22974+ AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
e49829fe 22975+ AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
1facf9fc 22976+
22977+ found = 0;
86dc4139
AM
22978+ i = au_plink_hash(inode->i_ino);
22979+ plink_hlist = &sbinfo->si_plink[i].head;
4a4d8108 22980+ rcu_read_lock();
86dc4139 22981+ hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
1facf9fc 22982+ if (plink->inode == inode) {
22983+ found = 1;
22984+ break;
22985+ }
4a4d8108 22986+ rcu_read_unlock();
1facf9fc 22987+ return found;
22988+}
22989+
22990+/* ---------------------------------------------------------------------- */
22991+
22992+/*
22993+ * generate a name for plink.
22994+ * the file will be stored under AUFS_WH_PLINKDIR.
22995+ */
22996+/* 20 is max digits length of ulong 64 */
22997+#define PLINK_NAME_LEN ((20 + 1) * 2)
22998+
22999+static int plink_name(char *name, int len, struct inode *inode,
23000+ aufs_bindex_t bindex)
23001+{
23002+ int rlen;
23003+ struct inode *h_inode;
23004+
23005+ h_inode = au_h_iptr(inode, bindex);
23006+ rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
23007+ return rlen;
23008+}
23009+
7f207e10
AM
23010+struct au_do_plink_lkup_args {
23011+ struct dentry **errp;
23012+ struct qstr *tgtname;
23013+ struct dentry *h_parent;
23014+ struct au_branch *br;
23015+};
23016+
23017+static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
23018+ struct dentry *h_parent,
23019+ struct au_branch *br)
23020+{
23021+ struct dentry *h_dentry;
23022+ struct mutex *h_mtx;
23023+
23024+ h_mtx = &h_parent->d_inode->i_mutex;
23025+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
b4510431 23026+ h_dentry = vfsub_lkup_one(tgtname, h_parent);
7f207e10
AM
23027+ mutex_unlock(h_mtx);
23028+ return h_dentry;
23029+}
23030+
23031+static void au_call_do_plink_lkup(void *args)
23032+{
23033+ struct au_do_plink_lkup_args *a = args;
23034+ *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br);
23035+}
23036+
1facf9fc 23037+/* lookup the plink-ed @inode under the branch at @bindex */
23038+struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
23039+{
23040+ struct dentry *h_dentry, *h_parent;
23041+ struct au_branch *br;
23042+ struct inode *h_dir;
7f207e10 23043+ int wkq_err;
1facf9fc 23044+ char a[PLINK_NAME_LEN];
0c3ec466 23045+ struct qstr tgtname = QSTR_INIT(a, 0);
1facf9fc 23046+
e49829fe
JR
23047+ AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
23048+
1facf9fc 23049+ br = au_sbr(inode->i_sb, bindex);
23050+ h_parent = br->br_wbr->wbr_plink;
23051+ h_dir = h_parent->d_inode;
23052+ tgtname.len = plink_name(a, sizeof(a), inode, bindex);
23053+
2dfbb274 23054+ if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
7f207e10
AM
23055+ struct au_do_plink_lkup_args args = {
23056+ .errp = &h_dentry,
23057+ .tgtname = &tgtname,
23058+ .h_parent = h_parent,
23059+ .br = br
23060+ };
23061+
23062+ wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
23063+ if (unlikely(wkq_err))
23064+ h_dentry = ERR_PTR(wkq_err);
23065+ } else
23066+ h_dentry = au_do_plink_lkup(&tgtname, h_parent, br);
23067+
1facf9fc 23068+ return h_dentry;
23069+}
23070+
23071+/* create a pseudo-link */
23072+static int do_whplink(struct qstr *tgt, struct dentry *h_parent,
23073+ struct dentry *h_dentry, struct au_branch *br)
23074+{
23075+ int err;
23076+ struct path h_path = {
86dc4139 23077+ .mnt = au_br_mnt(br)
1facf9fc 23078+ };
523b37e3 23079+ struct inode *h_dir, *delegated;
1facf9fc 23080+
23081+ h_dir = h_parent->d_inode;
7f207e10 23082+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_CHILD2);
4f0767ce 23083+again:
b4510431 23084+ h_path.dentry = vfsub_lkup_one(tgt, h_parent);
1facf9fc 23085+ err = PTR_ERR(h_path.dentry);
23086+ if (IS_ERR(h_path.dentry))
23087+ goto out;
23088+
23089+ err = 0;
23090+ /* wh.plink dir is not monitored */
7f207e10 23091+ /* todo: is it really safe? */
1facf9fc 23092+ if (h_path.dentry->d_inode
23093+ && h_path.dentry->d_inode != h_dentry->d_inode) {
523b37e3
AM
23094+ delegated = NULL;
23095+ err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
23096+ if (unlikely(err == -EWOULDBLOCK)) {
23097+ pr_warn("cannot retry for NFSv4 delegation"
23098+ " for an internal unlink\n");
23099+ iput(delegated);
23100+ }
1facf9fc 23101+ dput(h_path.dentry);
23102+ h_path.dentry = NULL;
23103+ if (!err)
23104+ goto again;
23105+ }
523b37e3
AM
23106+ if (!err && !h_path.dentry->d_inode) {
23107+ delegated = NULL;
23108+ err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
23109+ if (unlikely(err == -EWOULDBLOCK)) {
23110+ pr_warn("cannot retry for NFSv4 delegation"
23111+ " for an internal link\n");
23112+ iput(delegated);
23113+ }
23114+ }
1facf9fc 23115+ dput(h_path.dentry);
23116+
4f0767ce 23117+out:
7f207e10 23118+ mutex_unlock(&h_dir->i_mutex);
1facf9fc 23119+ return err;
23120+}
23121+
23122+struct do_whplink_args {
23123+ int *errp;
23124+ struct qstr *tgt;
23125+ struct dentry *h_parent;
23126+ struct dentry *h_dentry;
23127+ struct au_branch *br;
23128+};
23129+
23130+static void call_do_whplink(void *args)
23131+{
23132+ struct do_whplink_args *a = args;
23133+ *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br);
23134+}
23135+
23136+static int whplink(struct dentry *h_dentry, struct inode *inode,
23137+ aufs_bindex_t bindex, struct au_branch *br)
23138+{
23139+ int err, wkq_err;
23140+ struct au_wbr *wbr;
23141+ struct dentry *h_parent;
23142+ struct inode *h_dir;
23143+ char a[PLINK_NAME_LEN];
0c3ec466 23144+ struct qstr tgtname = QSTR_INIT(a, 0);
1facf9fc 23145+
23146+ wbr = au_sbr(inode->i_sb, bindex)->br_wbr;
23147+ h_parent = wbr->wbr_plink;
23148+ h_dir = h_parent->d_inode;
23149+ tgtname.len = plink_name(a, sizeof(a), inode, bindex);
23150+
23151+ /* always superio. */
2dfbb274 23152+ if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
1facf9fc 23153+ struct do_whplink_args args = {
23154+ .errp = &err,
23155+ .tgt = &tgtname,
23156+ .h_parent = h_parent,
23157+ .h_dentry = h_dentry,
23158+ .br = br
23159+ };
23160+ wkq_err = au_wkq_wait(call_do_whplink, &args);
23161+ if (unlikely(wkq_err))
23162+ err = wkq_err;
23163+ } else
23164+ err = do_whplink(&tgtname, h_parent, h_dentry, br);
1facf9fc 23165+
23166+ return err;
23167+}
23168+
23169+/* free a single plink */
23170+static void do_put_plink(struct pseudo_link *plink, int do_del)
23171+{
1facf9fc 23172+ if (do_del)
86dc4139 23173+ hlist_del(&plink->hlist);
4a4d8108
AM
23174+ iput(plink->inode);
23175+ kfree(plink);
23176+}
23177+
23178+static void do_put_plink_rcu(struct rcu_head *rcu)
23179+{
23180+ struct pseudo_link *plink;
23181+
23182+ plink = container_of(rcu, struct pseudo_link, rcu);
23183+ iput(plink->inode);
1facf9fc 23184+ kfree(plink);
23185+}
23186+
23187+/*
23188+ * create a new pseudo-link for @h_dentry on @bindex.
23189+ * the linked inode is held in aufs @inode.
23190+ */
23191+void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
23192+ struct dentry *h_dentry)
23193+{
23194+ struct super_block *sb;
23195+ struct au_sbinfo *sbinfo;
86dc4139 23196+ struct hlist_head *plink_hlist;
4a4d8108 23197+ struct pseudo_link *plink, *tmp;
86dc4139
AM
23198+ struct au_sphlhead *sphl;
23199+ int found, err, cnt, i;
1facf9fc 23200+
23201+ sb = inode->i_sb;
23202+ sbinfo = au_sbi(sb);
23203+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 23204+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 23205+
86dc4139 23206+ found = au_plink_test(inode);
4a4d8108 23207+ if (found)
1facf9fc 23208+ return;
4a4d8108 23209+
86dc4139
AM
23210+ i = au_plink_hash(inode->i_ino);
23211+ sphl = sbinfo->si_plink + i;
23212+ plink_hlist = &sphl->head;
4a4d8108
AM
23213+ tmp = kmalloc(sizeof(*plink), GFP_NOFS);
23214+ if (tmp)
23215+ tmp->inode = au_igrab(inode);
23216+ else {
23217+ err = -ENOMEM;
23218+ goto out;
1facf9fc 23219+ }
23220+
86dc4139
AM
23221+ spin_lock(&sphl->spin);
23222+ hlist_for_each_entry(plink, plink_hlist, hlist) {
4a4d8108
AM
23223+ if (plink->inode == inode) {
23224+ found = 1;
23225+ break;
23226+ }
1facf9fc 23227+ }
4a4d8108 23228+ if (!found)
86dc4139
AM
23229+ hlist_add_head_rcu(&tmp->hlist, plink_hlist);
23230+ spin_unlock(&sphl->spin);
4a4d8108 23231+ if (!found) {
86dc4139
AM
23232+ cnt = au_sphl_count(sphl);
23233+#define msg "unexpectedly unblanced or too many pseudo-links"
23234+ if (cnt > AUFS_PLINK_WARN)
23235+ AuWarn1(msg ", %d\n", cnt);
23236+#undef msg
1facf9fc 23237+ err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex));
4a4d8108
AM
23238+ } else {
23239+ do_put_plink(tmp, 0);
23240+ return;
1facf9fc 23241+ }
23242+
4a4d8108 23243+out:
1facf9fc 23244+ if (unlikely(err)) {
0c3ec466 23245+ pr_warn("err %d, damaged pseudo link.\n", err);
4a4d8108 23246+ if (tmp) {
86dc4139 23247+ au_sphl_del_rcu(&tmp->hlist, sphl);
4a4d8108
AM
23248+ call_rcu(&tmp->rcu, do_put_plink_rcu);
23249+ }
1facf9fc 23250+ }
23251+}
23252+
23253+/* free all plinks */
e49829fe 23254+void au_plink_put(struct super_block *sb, int verbose)
1facf9fc 23255+{
86dc4139 23256+ int i, warned;
1facf9fc 23257+ struct au_sbinfo *sbinfo;
86dc4139
AM
23258+ struct hlist_head *plink_hlist;
23259+ struct hlist_node *tmp;
23260+ struct pseudo_link *plink;
1facf9fc 23261+
dece6358
AM
23262+ SiMustWriteLock(sb);
23263+
1facf9fc 23264+ sbinfo = au_sbi(sb);
23265+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 23266+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 23267+
1facf9fc 23268+ /* no spin_lock since sbinfo is write-locked */
86dc4139
AM
23269+ warned = 0;
23270+ for (i = 0; i < AuPlink_NHASH; i++) {
23271+ plink_hlist = &sbinfo->si_plink[i].head;
23272+ if (!warned && verbose && !hlist_empty(plink_hlist)) {
23273+ pr_warn("pseudo-link is not flushed");
23274+ warned = 1;
23275+ }
23276+ hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist)
23277+ do_put_plink(plink, 0);
23278+ INIT_HLIST_HEAD(plink_hlist);
23279+ }
1facf9fc 23280+}
23281+
e49829fe
JR
23282+void au_plink_clean(struct super_block *sb, int verbose)
23283+{
23284+ struct dentry *root;
23285+
23286+ root = sb->s_root;
23287+ aufs_write_lock(root);
23288+ if (au_opt_test(au_mntflags(sb), PLINK))
23289+ au_plink_put(sb, verbose);
23290+ aufs_write_unlock(root);
23291+}
23292+
86dc4139
AM
23293+static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
23294+{
23295+ int do_put;
23296+ aufs_bindex_t bstart, bend, bindex;
23297+
23298+ do_put = 0;
23299+ bstart = au_ibstart(inode);
23300+ bend = au_ibend(inode);
23301+ if (bstart >= 0) {
23302+ for (bindex = bstart; bindex <= bend; bindex++) {
23303+ if (!au_h_iptr(inode, bindex)
23304+ || au_ii_br_id(inode, bindex) != br_id)
23305+ continue;
23306+ au_set_h_iptr(inode, bindex, NULL, 0);
23307+ do_put = 1;
23308+ break;
23309+ }
23310+ if (do_put)
23311+ for (bindex = bstart; bindex <= bend; bindex++)
23312+ if (au_h_iptr(inode, bindex)) {
23313+ do_put = 0;
23314+ break;
23315+ }
23316+ } else
23317+ do_put = 1;
23318+
23319+ return do_put;
23320+}
23321+
1facf9fc 23322+/* free the plinks on a branch specified by @br_id */
23323+void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
23324+{
23325+ struct au_sbinfo *sbinfo;
86dc4139
AM
23326+ struct hlist_head *plink_hlist;
23327+ struct hlist_node *tmp;
23328+ struct pseudo_link *plink;
1facf9fc 23329+ struct inode *inode;
86dc4139 23330+ int i, do_put;
1facf9fc 23331+
dece6358
AM
23332+ SiMustWriteLock(sb);
23333+
1facf9fc 23334+ sbinfo = au_sbi(sb);
23335+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 23336+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 23337+
1facf9fc 23338+ /* no spin_lock since sbinfo is write-locked */
86dc4139
AM
23339+ for (i = 0; i < AuPlink_NHASH; i++) {
23340+ plink_hlist = &sbinfo->si_plink[i].head;
23341+ hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist) {
23342+ inode = au_igrab(plink->inode);
23343+ ii_write_lock_child(inode);
23344+ do_put = au_plink_do_half_refresh(inode, br_id);
dece6358
AM
23345+ if (do_put)
23346+ do_put_plink(plink, 1);
86dc4139
AM
23347+ ii_write_unlock(inode);
23348+ iput(inode);
dece6358 23349+ }
dece6358
AM
23350+ }
23351+}
7f207e10
AM
23352diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
23353--- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 23354+++ linux/fs/aufs/poll.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 23355@@ -0,0 +1,55 @@
dece6358 23356+/*
523b37e3 23357+ * Copyright (C) 2005-2014 Junjiro R. Okajima
dece6358
AM
23358+ *
23359+ * This program, aufs is free software; you can redistribute it and/or modify
23360+ * it under the terms of the GNU General Public License as published by
23361+ * the Free Software Foundation; either version 2 of the License, or
23362+ * (at your option) any later version.
23363+ *
23364+ * This program is distributed in the hope that it will be useful,
23365+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
23366+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23367+ * GNU General Public License for more details.
23368+ *
23369+ * You should have received a copy of the GNU General Public License
523b37e3 23370+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
dece6358
AM
23371+ */
23372+
1308ab2a 23373+/*
23374+ * poll operation
23375+ * There is only one filesystem which implements ->poll operation, currently.
23376+ */
23377+
23378+#include "aufs.h"
23379+
23380+unsigned int aufs_poll(struct file *file, poll_table *wait)
23381+{
23382+ unsigned int mask;
23383+ int err;
23384+ struct file *h_file;
23385+ struct dentry *dentry;
23386+ struct super_block *sb;
23387+
23388+ /* We should pretend an error happened. */
23389+ mask = POLLERR /* | POLLIN | POLLOUT */;
23390+ dentry = file->f_dentry;
23391+ sb = dentry->d_sb;
e49829fe 23392+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
1308ab2a 23393+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
23394+ if (unlikely(err))
23395+ goto out;
23396+
23397+ /* it is not an error if h_file has no operation */
23398+ mask = DEFAULT_POLLMASK;
4a4d8108 23399+ h_file = au_hf_top(file);
523b37e3 23400+ if (h_file->f_op->poll)
1308ab2a 23401+ mask = h_file->f_op->poll(h_file, wait);
23402+
23403+ di_read_unlock(dentry, AuLock_IR);
23404+ fi_read_unlock(file);
23405+
4f0767ce 23406+out:
1308ab2a 23407+ si_read_unlock(sb);
23408+ AuTraceErr((int)mask);
23409+ return mask;
23410+}
7f207e10
AM
23411diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
23412--- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 23413+++ linux/fs/aufs/procfs.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 23414@@ -0,0 +1,169 @@
e49829fe 23415+/*
523b37e3 23416+ * Copyright (C) 2010-2014 Junjiro R. Okajima
e49829fe
JR
23417+ *
23418+ * This program, aufs is free software; you can redistribute it and/or modify
23419+ * it under the terms of the GNU General Public License as published by
23420+ * the Free Software Foundation; either version 2 of the License, or
23421+ * (at your option) any later version.
23422+ *
23423+ * This program is distributed in the hope that it will be useful,
23424+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
23425+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23426+ * GNU General Public License for more details.
23427+ *
23428+ * You should have received a copy of the GNU General Public License
523b37e3 23429+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
e49829fe
JR
23430+ */
23431+
23432+/*
23433+ * procfs interfaces
23434+ */
23435+
23436+#include <linux/proc_fs.h>
23437+#include "aufs.h"
23438+
23439+static int au_procfs_plm_release(struct inode *inode, struct file *file)
23440+{
23441+ struct au_sbinfo *sbinfo;
23442+
23443+ sbinfo = file->private_data;
23444+ if (sbinfo) {
23445+ au_plink_maint_leave(sbinfo);
23446+ kobject_put(&sbinfo->si_kobj);
23447+ }
23448+
23449+ return 0;
23450+}
23451+
23452+static void au_procfs_plm_write_clean(struct file *file)
23453+{
23454+ struct au_sbinfo *sbinfo;
23455+
23456+ sbinfo = file->private_data;
23457+ if (sbinfo)
23458+ au_plink_clean(sbinfo->si_sb, /*verbose*/0);
23459+}
23460+
23461+static int au_procfs_plm_write_si(struct file *file, unsigned long id)
23462+{
23463+ int err;
23464+ struct super_block *sb;
23465+ struct au_sbinfo *sbinfo;
23466+
23467+ err = -EBUSY;
23468+ if (unlikely(file->private_data))
23469+ goto out;
23470+
23471+ sb = NULL;
53392da6 23472+ /* don't use au_sbilist_lock() here */
e49829fe
JR
23473+ spin_lock(&au_sbilist.spin);
23474+ list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
23475+ if (id == sysaufs_si_id(sbinfo)) {
23476+ kobject_get(&sbinfo->si_kobj);
23477+ sb = sbinfo->si_sb;
23478+ break;
23479+ }
23480+ spin_unlock(&au_sbilist.spin);
23481+
23482+ err = -EINVAL;
23483+ if (unlikely(!sb))
23484+ goto out;
23485+
23486+ err = au_plink_maint_enter(sb);
23487+ if (!err)
23488+ /* keep kobject_get() */
23489+ file->private_data = sbinfo;
23490+ else
23491+ kobject_put(&sbinfo->si_kobj);
23492+out:
23493+ return err;
23494+}
23495+
23496+/*
23497+ * Accept a valid "si=xxxx" only.
23498+ * Once it is accepted successfully, accept "clean" too.
23499+ */
23500+static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
23501+ size_t count, loff_t *ppos)
23502+{
23503+ ssize_t err;
23504+ unsigned long id;
23505+ /* last newline is allowed */
23506+ char buf[3 + sizeof(unsigned long) * 2 + 1];
23507+
23508+ err = -EACCES;
23509+ if (unlikely(!capable(CAP_SYS_ADMIN)))
23510+ goto out;
23511+
23512+ err = -EINVAL;
23513+ if (unlikely(count > sizeof(buf)))
23514+ goto out;
23515+
23516+ err = copy_from_user(buf, ubuf, count);
23517+ if (unlikely(err)) {
23518+ err = -EFAULT;
23519+ goto out;
23520+ }
23521+ buf[count] = 0;
23522+
23523+ err = -EINVAL;
23524+ if (!strcmp("clean", buf)) {
23525+ au_procfs_plm_write_clean(file);
23526+ goto out_success;
23527+ } else if (unlikely(strncmp("si=", buf, 3)))
23528+ goto out;
23529+
9dbd164d 23530+ err = kstrtoul(buf + 3, 16, &id);
e49829fe
JR
23531+ if (unlikely(err))
23532+ goto out;
23533+
23534+ err = au_procfs_plm_write_si(file, id);
23535+ if (unlikely(err))
23536+ goto out;
23537+
23538+out_success:
23539+ err = count; /* success */
23540+out:
23541+ return err;
23542+}
23543+
23544+static const struct file_operations au_procfs_plm_fop = {
23545+ .write = au_procfs_plm_write,
23546+ .release = au_procfs_plm_release,
23547+ .owner = THIS_MODULE
23548+};
23549+
23550+/* ---------------------------------------------------------------------- */
23551+
23552+static struct proc_dir_entry *au_procfs_dir;
23553+
23554+void au_procfs_fin(void)
23555+{
23556+ remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
23557+ remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
23558+}
23559+
23560+int __init au_procfs_init(void)
23561+{
23562+ int err;
23563+ struct proc_dir_entry *entry;
23564+
23565+ err = -ENOMEM;
23566+ au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
23567+ if (unlikely(!au_procfs_dir))
23568+ goto out;
23569+
23570+ entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | S_IWUSR,
23571+ au_procfs_dir, &au_procfs_plm_fop);
23572+ if (unlikely(!entry))
23573+ goto out_dir;
23574+
23575+ err = 0;
23576+ goto out; /* success */
23577+
23578+
23579+out_dir:
23580+ remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
23581+out:
23582+ return err;
23583+}
7f207e10
AM
23584diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
23585--- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 23586+++ linux/fs/aufs/rdu.c 2014-06-24 00:14:06.786052822 +0200
523b37e3 23587@@ -0,0 +1,388 @@
1308ab2a 23588+/*
523b37e3 23589+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1308ab2a 23590+ *
23591+ * This program, aufs is free software; you can redistribute it and/or modify
23592+ * it under the terms of the GNU General Public License as published by
23593+ * the Free Software Foundation; either version 2 of the License, or
23594+ * (at your option) any later version.
23595+ *
23596+ * This program is distributed in the hope that it will be useful,
23597+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
23598+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23599+ * GNU General Public License for more details.
23600+ *
23601+ * You should have received a copy of the GNU General Public License
523b37e3 23602+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1308ab2a 23603+ */
23604+
23605+/*
23606+ * readdir in userspace.
23607+ */
23608+
b752ccd1 23609+#include <linux/compat.h>
4a4d8108 23610+#include <linux/fs_stack.h>
1308ab2a 23611+#include <linux/security.h>
1308ab2a 23612+#include "aufs.h"
23613+
23614+/* bits for struct aufs_rdu.flags */
23615+#define AuRdu_CALLED 1
23616+#define AuRdu_CONT (1 << 1)
23617+#define AuRdu_FULL (1 << 2)
23618+#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
7f207e10
AM
23619+#define au_fset_rdu(flags, name) \
23620+ do { (flags) |= AuRdu_##name; } while (0)
23621+#define au_fclr_rdu(flags, name) \
23622+ do { (flags) &= ~AuRdu_##name; } while (0)
1308ab2a 23623+
23624+struct au_rdu_arg {
392086de 23625+ struct dir_context ctx;
1308ab2a 23626+ struct aufs_rdu *rdu;
23627+ union au_rdu_ent_ul ent;
23628+ unsigned long end;
23629+
23630+ struct super_block *sb;
23631+ int err;
23632+};
23633+
392086de 23634+static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
1308ab2a 23635+ loff_t offset, u64 h_ino, unsigned int d_type)
23636+{
23637+ int err, len;
392086de 23638+ struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
1308ab2a 23639+ struct aufs_rdu *rdu = arg->rdu;
23640+ struct au_rdu_ent ent;
23641+
23642+ err = 0;
23643+ arg->err = 0;
23644+ au_fset_rdu(rdu->cookie.flags, CALLED);
23645+ len = au_rdu_len(nlen);
23646+ if (arg->ent.ul + len < arg->end) {
23647+ ent.ino = h_ino;
23648+ ent.bindex = rdu->cookie.bindex;
23649+ ent.type = d_type;
23650+ ent.nlen = nlen;
4a4d8108
AM
23651+ if (unlikely(nlen > AUFS_MAX_NAMELEN))
23652+ ent.type = DT_UNKNOWN;
1308ab2a 23653+
9dbd164d 23654+ /* unnecessary to support mmap_sem since this is a dir */
1308ab2a 23655+ err = -EFAULT;
23656+ if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
23657+ goto out;
23658+ if (copy_to_user(arg->ent.e->name, name, nlen))
23659+ goto out;
23660+ /* the terminating NULL */
23661+ if (__put_user(0, arg->ent.e->name + nlen))
23662+ goto out;
23663+ err = 0;
23664+ /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
23665+ arg->ent.ul += len;
23666+ rdu->rent++;
23667+ } else {
23668+ err = -EFAULT;
23669+ au_fset_rdu(rdu->cookie.flags, FULL);
23670+ rdu->full = 1;
23671+ rdu->tail = arg->ent;
23672+ }
23673+
4f0767ce 23674+out:
1308ab2a 23675+ /* AuTraceErr(err); */
23676+ return err;
23677+}
23678+
23679+static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
23680+{
23681+ int err;
23682+ loff_t offset;
23683+ struct au_rdu_cookie *cookie = &arg->rdu->cookie;
23684+
92d182d2 23685+ /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
1308ab2a 23686+ offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
23687+ err = offset;
23688+ if (unlikely(offset != cookie->h_pos))
23689+ goto out;
23690+
23691+ err = 0;
23692+ do {
23693+ arg->err = 0;
23694+ au_fclr_rdu(cookie->flags, CALLED);
23695+ /* smp_mb(); */
392086de 23696+ err = vfsub_iterate_dir(h_file, &arg->ctx);
1308ab2a 23697+ if (err >= 0)
23698+ err = arg->err;
23699+ } while (!err
23700+ && au_ftest_rdu(cookie->flags, CALLED)
23701+ && !au_ftest_rdu(cookie->flags, FULL));
23702+ cookie->h_pos = h_file->f_pos;
23703+
4f0767ce 23704+out:
1308ab2a 23705+ AuTraceErr(err);
23706+ return err;
23707+}
23708+
23709+static int au_rdu(struct file *file, struct aufs_rdu *rdu)
23710+{
23711+ int err;
23712+ aufs_bindex_t bend;
392086de
AM
23713+ struct au_rdu_arg arg = {
23714+ .ctx = {
23715+ .actor = au_diractor(au_rdu_fill)
23716+ }
23717+ };
1308ab2a 23718+ struct dentry *dentry;
23719+ struct inode *inode;
23720+ struct file *h_file;
23721+ struct au_rdu_cookie *cookie = &rdu->cookie;
23722+
23723+ err = !access_ok(VERIFY_WRITE, rdu->ent.e, rdu->sz);
23724+ if (unlikely(err)) {
23725+ err = -EFAULT;
23726+ AuTraceErr(err);
23727+ goto out;
23728+ }
23729+ rdu->rent = 0;
23730+ rdu->tail = rdu->ent;
23731+ rdu->full = 0;
23732+ arg.rdu = rdu;
23733+ arg.ent = rdu->ent;
23734+ arg.end = arg.ent.ul;
23735+ arg.end += rdu->sz;
23736+
23737+ err = -ENOTDIR;
523b37e3 23738+ if (unlikely(!file->f_op->iterate))
1308ab2a 23739+ goto out;
23740+
23741+ err = security_file_permission(file, MAY_READ);
23742+ AuTraceErr(err);
23743+ if (unlikely(err))
23744+ goto out;
23745+
23746+ dentry = file->f_dentry;
23747+ inode = dentry->d_inode;
23748+#if 1
23749+ mutex_lock(&inode->i_mutex);
23750+#else
23751+ err = mutex_lock_killable(&inode->i_mutex);
23752+ AuTraceErr(err);
23753+ if (unlikely(err))
23754+ goto out;
23755+#endif
1308ab2a 23756+
23757+ arg.sb = inode->i_sb;
e49829fe
JR
23758+ err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
23759+ if (unlikely(err))
23760+ goto out_mtx;
027c5e7a
AM
23761+ err = au_alive_dir(dentry);
23762+ if (unlikely(err))
23763+ goto out_si;
e49829fe 23764+ /* todo: reval? */
1308ab2a 23765+ fi_read_lock(file);
23766+
23767+ err = -EAGAIN;
23768+ if (unlikely(au_ftest_rdu(cookie->flags, CONT)
23769+ && cookie->generation != au_figen(file)))
23770+ goto out_unlock;
23771+
23772+ err = 0;
23773+ if (!rdu->blk) {
23774+ rdu->blk = au_sbi(arg.sb)->si_rdblk;
23775+ if (!rdu->blk)
23776+ rdu->blk = au_dir_size(file, /*dentry*/NULL);
23777+ }
23778+ bend = au_fbstart(file);
23779+ if (cookie->bindex < bend)
23780+ cookie->bindex = bend;
4a4d8108 23781+ bend = au_fbend_dir(file);
1308ab2a 23782+ /* AuDbg("b%d, b%d\n", cookie->bindex, bend); */
23783+ for (; !err && cookie->bindex <= bend;
23784+ cookie->bindex++, cookie->h_pos = 0) {
4a4d8108 23785+ h_file = au_hf_dir(file, cookie->bindex);
1308ab2a 23786+ if (!h_file)
23787+ continue;
23788+
23789+ au_fclr_rdu(cookie->flags, FULL);
23790+ err = au_rdu_do(h_file, &arg);
23791+ AuTraceErr(err);
23792+ if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
23793+ break;
23794+ }
23795+ AuDbg("rent %llu\n", rdu->rent);
23796+
23797+ if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
23798+ rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
23799+ au_fset_rdu(cookie->flags, CONT);
23800+ cookie->generation = au_figen(file);
23801+ }
23802+
23803+ ii_read_lock_child(inode);
23804+ fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibstart(inode)));
23805+ ii_read_unlock(inode);
23806+
4f0767ce 23807+out_unlock:
1308ab2a 23808+ fi_read_unlock(file);
027c5e7a 23809+out_si:
1308ab2a 23810+ si_read_unlock(arg.sb);
4f0767ce 23811+out_mtx:
1308ab2a 23812+ mutex_unlock(&inode->i_mutex);
4f0767ce 23813+out:
1308ab2a 23814+ AuTraceErr(err);
23815+ return err;
23816+}
23817+
23818+static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
23819+{
23820+ int err;
23821+ ino_t ino;
23822+ unsigned long long nent;
23823+ union au_rdu_ent_ul *u;
23824+ struct au_rdu_ent ent;
23825+ struct super_block *sb;
23826+
23827+ err = 0;
23828+ nent = rdu->nent;
23829+ u = &rdu->ent;
23830+ sb = file->f_dentry->d_sb;
23831+ si_read_lock(sb, AuLock_FLUSH);
23832+ while (nent-- > 0) {
9dbd164d 23833+ /* unnecessary to support mmap_sem since this is a dir */
1308ab2a 23834+ err = copy_from_user(&ent, u->e, sizeof(ent));
4a4d8108
AM
23835+ if (!err)
23836+ err = !access_ok(VERIFY_WRITE, &u->e->ino, sizeof(ino));
1308ab2a 23837+ if (unlikely(err)) {
23838+ err = -EFAULT;
23839+ AuTraceErr(err);
23840+ break;
23841+ }
23842+
23843+ /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
23844+ if (!ent.wh)
23845+ err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
23846+ else
23847+ err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
23848+ &ino);
23849+ if (unlikely(err)) {
23850+ AuTraceErr(err);
23851+ break;
23852+ }
23853+
23854+ err = __put_user(ino, &u->e->ino);
23855+ if (unlikely(err)) {
23856+ err = -EFAULT;
23857+ AuTraceErr(err);
23858+ break;
23859+ }
23860+ u->ul += au_rdu_len(ent.nlen);
23861+ }
23862+ si_read_unlock(sb);
23863+
23864+ return err;
23865+}
23866+
23867+/* ---------------------------------------------------------------------- */
23868+
23869+static int au_rdu_verify(struct aufs_rdu *rdu)
23870+{
b752ccd1 23871+ AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
1308ab2a 23872+ "%llu, b%d, 0x%x, g%u}\n",
b752ccd1 23873+ rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
1308ab2a 23874+ rdu->blk,
23875+ rdu->rent, rdu->shwh, rdu->full,
23876+ rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
23877+ rdu->cookie.generation);
dece6358 23878+
b752ccd1 23879+ if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
1308ab2a 23880+ return 0;
dece6358 23881+
b752ccd1
AM
23882+ AuDbg("%u:%u\n",
23883+ rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
1308ab2a 23884+ return -EINVAL;
23885+}
23886+
23887+long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
dece6358 23888+{
1308ab2a 23889+ long err, e;
23890+ struct aufs_rdu rdu;
23891+ void __user *p = (void __user *)arg;
dece6358 23892+
1308ab2a 23893+ err = copy_from_user(&rdu, p, sizeof(rdu));
23894+ if (unlikely(err)) {
23895+ err = -EFAULT;
23896+ AuTraceErr(err);
23897+ goto out;
23898+ }
23899+ err = au_rdu_verify(&rdu);
dece6358
AM
23900+ if (unlikely(err))
23901+ goto out;
23902+
1308ab2a 23903+ switch (cmd) {
23904+ case AUFS_CTL_RDU:
23905+ err = au_rdu(file, &rdu);
23906+ if (unlikely(err))
23907+ break;
dece6358 23908+
1308ab2a 23909+ e = copy_to_user(p, &rdu, sizeof(rdu));
23910+ if (unlikely(e)) {
23911+ err = -EFAULT;
23912+ AuTraceErr(err);
23913+ }
23914+ break;
23915+ case AUFS_CTL_RDU_INO:
23916+ err = au_rdu_ino(file, &rdu);
23917+ break;
23918+
23919+ default:
4a4d8108 23920+ /* err = -ENOTTY; */
1308ab2a 23921+ err = -EINVAL;
23922+ }
dece6358 23923+
4f0767ce 23924+out:
1308ab2a 23925+ AuTraceErr(err);
23926+ return err;
1facf9fc 23927+}
b752ccd1
AM
23928+
23929+#ifdef CONFIG_COMPAT
23930+long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
23931+{
23932+ long err, e;
23933+ struct aufs_rdu rdu;
23934+ void __user *p = compat_ptr(arg);
23935+
23936+ /* todo: get_user()? */
23937+ err = copy_from_user(&rdu, p, sizeof(rdu));
23938+ if (unlikely(err)) {
23939+ err = -EFAULT;
23940+ AuTraceErr(err);
23941+ goto out;
23942+ }
23943+ rdu.ent.e = compat_ptr(rdu.ent.ul);
23944+ err = au_rdu_verify(&rdu);
23945+ if (unlikely(err))
23946+ goto out;
23947+
23948+ switch (cmd) {
23949+ case AUFS_CTL_RDU:
23950+ err = au_rdu(file, &rdu);
23951+ if (unlikely(err))
23952+ break;
23953+
23954+ rdu.ent.ul = ptr_to_compat(rdu.ent.e);
23955+ rdu.tail.ul = ptr_to_compat(rdu.tail.e);
23956+ e = copy_to_user(p, &rdu, sizeof(rdu));
23957+ if (unlikely(e)) {
23958+ err = -EFAULT;
23959+ AuTraceErr(err);
23960+ }
23961+ break;
23962+ case AUFS_CTL_RDU_INO:
23963+ err = au_rdu_ino(file, &rdu);
23964+ break;
23965+
23966+ default:
23967+ /* err = -ENOTTY; */
23968+ err = -EINVAL;
23969+ }
23970+
4f0767ce 23971+out:
b752ccd1
AM
23972+ AuTraceErr(err);
23973+ return err;
23974+}
23975+#endif
7f207e10
AM
23976diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
23977--- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 23978+++ linux/fs/aufs/rwsem.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 23979@@ -0,0 +1,187 @@
1facf9fc 23980+/*
523b37e3 23981+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 23982+ *
23983+ * This program, aufs is free software; you can redistribute it and/or modify
23984+ * it under the terms of the GNU General Public License as published by
23985+ * the Free Software Foundation; either version 2 of the License, or
23986+ * (at your option) any later version.
dece6358
AM
23987+ *
23988+ * This program is distributed in the hope that it will be useful,
23989+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
23990+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23991+ * GNU General Public License for more details.
23992+ *
23993+ * You should have received a copy of the GNU General Public License
523b37e3 23994+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 23995+ */
23996+
23997+/*
23998+ * simple read-write semaphore wrappers
23999+ */
24000+
24001+#ifndef __AUFS_RWSEM_H__
24002+#define __AUFS_RWSEM_H__
24003+
24004+#ifdef __KERNEL__
24005+
4a4d8108 24006+#include "debug.h"
dece6358
AM
24007+
24008+struct au_rwsem {
24009+ struct rw_semaphore rwsem;
24010+#ifdef CONFIG_AUFS_DEBUG
24011+ /* just for debugging, not almighty counter */
24012+ atomic_t rcnt, wcnt;
24013+#endif
24014+};
24015+
24016+#ifdef CONFIG_AUFS_DEBUG
24017+#define AuDbgCntInit(rw) do { \
24018+ atomic_set(&(rw)->rcnt, 0); \
24019+ atomic_set(&(rw)->wcnt, 0); \
24020+ smp_mb(); /* atomic set */ \
24021+} while (0)
24022+
e49829fe 24023+#define AuDbgRcntInc(rw) atomic_inc(&(rw)->rcnt)
dece6358 24024+#define AuDbgRcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->rcnt) < 0)
e49829fe 24025+#define AuDbgWcntInc(rw) atomic_inc(&(rw)->wcnt)
dece6358
AM
24026+#define AuDbgWcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->wcnt) < 0)
24027+#else
24028+#define AuDbgCntInit(rw) do {} while (0)
24029+#define AuDbgRcntInc(rw) do {} while (0)
24030+#define AuDbgRcntDec(rw) do {} while (0)
24031+#define AuDbgWcntInc(rw) do {} while (0)
24032+#define AuDbgWcntDec(rw) do {} while (0)
24033+#endif /* CONFIG_AUFS_DEBUG */
24034+
24035+/* to debug easier, do not make them inlined functions */
24036+#define AuRwMustNoWaiters(rw) AuDebugOn(!list_empty(&(rw)->rwsem.wait_list))
24037+/* rwsem_is_locked() is unusable */
24038+#define AuRwMustReadLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0)
24039+#define AuRwMustWriteLock(rw) AuDebugOn(atomic_read(&(rw)->wcnt) <= 0)
24040+#define AuRwMustAnyLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0 \
24041+ && atomic_read(&(rw)->wcnt) <= 0)
24042+#define AuRwDestroy(rw) AuDebugOn(atomic_read(&(rw)->rcnt) \
24043+ || atomic_read(&(rw)->wcnt))
24044+
e49829fe
JR
24045+#define au_rw_class(rw, key) lockdep_set_class(&(rw)->rwsem, key)
24046+
dece6358
AM
24047+static inline void au_rw_init(struct au_rwsem *rw)
24048+{
24049+ AuDbgCntInit(rw);
24050+ init_rwsem(&rw->rwsem);
24051+}
24052+
24053+static inline void au_rw_init_wlock(struct au_rwsem *rw)
24054+{
24055+ au_rw_init(rw);
24056+ down_write(&rw->rwsem);
24057+ AuDbgWcntInc(rw);
24058+}
24059+
24060+static inline void au_rw_init_wlock_nested(struct au_rwsem *rw,
24061+ unsigned int lsc)
24062+{
24063+ au_rw_init(rw);
24064+ down_write_nested(&rw->rwsem, lsc);
24065+ AuDbgWcntInc(rw);
24066+}
24067+
24068+static inline void au_rw_read_lock(struct au_rwsem *rw)
24069+{
24070+ down_read(&rw->rwsem);
24071+ AuDbgRcntInc(rw);
24072+}
24073+
24074+static inline void au_rw_read_lock_nested(struct au_rwsem *rw, unsigned int lsc)
24075+{
24076+ down_read_nested(&rw->rwsem, lsc);
24077+ AuDbgRcntInc(rw);
24078+}
24079+
24080+static inline void au_rw_read_unlock(struct au_rwsem *rw)
24081+{
24082+ AuRwMustReadLock(rw);
24083+ AuDbgRcntDec(rw);
24084+ up_read(&rw->rwsem);
24085+}
24086+
24087+static inline void au_rw_dgrade_lock(struct au_rwsem *rw)
24088+{
24089+ AuRwMustWriteLock(rw);
24090+ AuDbgRcntInc(rw);
24091+ AuDbgWcntDec(rw);
24092+ downgrade_write(&rw->rwsem);
24093+}
24094+
24095+static inline void au_rw_write_lock(struct au_rwsem *rw)
24096+{
24097+ down_write(&rw->rwsem);
24098+ AuDbgWcntInc(rw);
24099+}
24100+
24101+static inline void au_rw_write_lock_nested(struct au_rwsem *rw,
24102+ unsigned int lsc)
24103+{
24104+ down_write_nested(&rw->rwsem, lsc);
24105+ AuDbgWcntInc(rw);
24106+}
1facf9fc 24107+
dece6358
AM
24108+static inline void au_rw_write_unlock(struct au_rwsem *rw)
24109+{
24110+ AuRwMustWriteLock(rw);
24111+ AuDbgWcntDec(rw);
24112+ up_write(&rw->rwsem);
24113+}
24114+
24115+/* why is not _nested version defined */
24116+static inline int au_rw_read_trylock(struct au_rwsem *rw)
24117+{
24118+ int ret = down_read_trylock(&rw->rwsem);
24119+ if (ret)
24120+ AuDbgRcntInc(rw);
24121+ return ret;
24122+}
24123+
24124+static inline int au_rw_write_trylock(struct au_rwsem *rw)
24125+{
24126+ int ret = down_write_trylock(&rw->rwsem);
24127+ if (ret)
24128+ AuDbgWcntInc(rw);
24129+ return ret;
24130+}
24131+
24132+#undef AuDbgCntInit
24133+#undef AuDbgRcntInc
24134+#undef AuDbgRcntDec
24135+#undef AuDbgWcntInc
24136+#undef AuDbgWcntDec
1facf9fc 24137+
24138+#define AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
24139+static inline void prefix##_read_lock(param) \
dece6358 24140+{ au_rw_read_lock(rwsem); } \
1facf9fc 24141+static inline void prefix##_write_lock(param) \
dece6358 24142+{ au_rw_write_lock(rwsem); } \
1facf9fc 24143+static inline int prefix##_read_trylock(param) \
dece6358 24144+{ return au_rw_read_trylock(rwsem); } \
1facf9fc 24145+static inline int prefix##_write_trylock(param) \
dece6358 24146+{ return au_rw_write_trylock(rwsem); }
1facf9fc 24147+/* why is not _nested version defined */
24148+/* static inline void prefix##_read_trylock_nested(param, lsc)
dece6358 24149+{ au_rw_read_trylock_nested(rwsem, lsc)); }
1facf9fc 24150+static inline void prefix##_write_trylock_nestd(param, lsc)
dece6358 24151+{ au_rw_write_trylock_nested(rwsem, lsc); } */
1facf9fc 24152+
24153+#define AuSimpleUnlockRwsemFuncs(prefix, param, rwsem) \
24154+static inline void prefix##_read_unlock(param) \
dece6358 24155+{ au_rw_read_unlock(rwsem); } \
1facf9fc 24156+static inline void prefix##_write_unlock(param) \
dece6358 24157+{ au_rw_write_unlock(rwsem); } \
1facf9fc 24158+static inline void prefix##_downgrade_lock(param) \
dece6358 24159+{ au_rw_dgrade_lock(rwsem); }
1facf9fc 24160+
24161+#define AuSimpleRwsemFuncs(prefix, param, rwsem) \
24162+ AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
24163+ AuSimpleUnlockRwsemFuncs(prefix, param, rwsem)
24164+
24165+#endif /* __KERNEL__ */
24166+#endif /* __AUFS_RWSEM_H__ */
7f207e10
AM
24167diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
24168--- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 24169+++ linux/fs/aufs/sbinfo.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 24170@@ -0,0 +1,351 @@
1facf9fc 24171+/*
523b37e3 24172+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 24173+ *
24174+ * This program, aufs is free software; you can redistribute it and/or modify
24175+ * it under the terms of the GNU General Public License as published by
24176+ * the Free Software Foundation; either version 2 of the License, or
24177+ * (at your option) any later version.
dece6358
AM
24178+ *
24179+ * This program is distributed in the hope that it will be useful,
24180+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
24181+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24182+ * GNU General Public License for more details.
24183+ *
24184+ * You should have received a copy of the GNU General Public License
523b37e3 24185+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 24186+ */
24187+
24188+/*
24189+ * superblock private data
24190+ */
24191+
24192+#include "aufs.h"
24193+
24194+/*
24195+ * they are necessary regardless sysfs is disabled.
24196+ */
24197+void au_si_free(struct kobject *kobj)
24198+{
86dc4139 24199+ int i;
1facf9fc 24200+ struct au_sbinfo *sbinfo;
b752ccd1 24201+ char *locked __maybe_unused; /* debug only */
1facf9fc 24202+
24203+ sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
86dc4139
AM
24204+ for (i = 0; i < AuPlink_NHASH; i++)
24205+ AuDebugOn(!hlist_empty(&sbinfo->si_plink[i].head));
e49829fe 24206+ AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
1facf9fc 24207+
e49829fe 24208+ au_rw_write_lock(&sbinfo->si_rwsem);
1facf9fc 24209+ au_br_free(sbinfo);
e49829fe 24210+ au_rw_write_unlock(&sbinfo->si_rwsem);
b752ccd1
AM
24211+
24212+ AuDebugOn(radix_tree_gang_lookup
24213+ (&sbinfo->au_si_pid.tree, (void **)&locked,
24214+ /*first_index*/PID_MAX_DEFAULT - 1,
24215+ /*max_items*/sizeof(locked)/sizeof(*locked)));
24216+
1facf9fc 24217+ kfree(sbinfo->si_branch);
b752ccd1 24218+ kfree(sbinfo->au_si_pid.bitmap);
1facf9fc 24219+ mutex_destroy(&sbinfo->si_xib_mtx);
dece6358 24220+ AuRwDestroy(&sbinfo->si_rwsem);
1facf9fc 24221+
24222+ kfree(sbinfo);
24223+}
24224+
24225+int au_si_alloc(struct super_block *sb)
24226+{
86dc4139 24227+ int err, i;
1facf9fc 24228+ struct au_sbinfo *sbinfo;
e49829fe 24229+ static struct lock_class_key aufs_si;
1facf9fc 24230+
24231+ err = -ENOMEM;
4a4d8108 24232+ sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
1facf9fc 24233+ if (unlikely(!sbinfo))
24234+ goto out;
24235+
b752ccd1
AM
24236+ BUILD_BUG_ON(sizeof(unsigned long) !=
24237+ sizeof(*sbinfo->au_si_pid.bitmap));
24238+ sbinfo->au_si_pid.bitmap = kcalloc(BITS_TO_LONGS(PID_MAX_DEFAULT),
24239+ sizeof(*sbinfo->au_si_pid.bitmap),
24240+ GFP_NOFS);
24241+ if (unlikely(!sbinfo->au_si_pid.bitmap))
24242+ goto out_sbinfo;
24243+
1facf9fc 24244+ /* will be reallocated separately */
24245+ sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
24246+ if (unlikely(!sbinfo->si_branch))
b752ccd1 24247+ goto out_pidmap;
1facf9fc 24248+
1facf9fc 24249+ err = sysaufs_si_init(sbinfo);
24250+ if (unlikely(err))
24251+ goto out_br;
24252+
24253+ au_nwt_init(&sbinfo->si_nowait);
dece6358 24254+ au_rw_init_wlock(&sbinfo->si_rwsem);
e49829fe 24255+ au_rw_class(&sbinfo->si_rwsem, &aufs_si);
b752ccd1
AM
24256+ spin_lock_init(&sbinfo->au_si_pid.tree_lock);
24257+ INIT_RADIX_TREE(&sbinfo->au_si_pid.tree, GFP_ATOMIC | __GFP_NOFAIL);
24258+
7f207e10 24259+ atomic_long_set(&sbinfo->si_ninodes, 0);
7f207e10
AM
24260+ atomic_long_set(&sbinfo->si_nfiles, 0);
24261+
1facf9fc 24262+ sbinfo->si_bend = -1;
392086de 24263+ sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
1facf9fc 24264+
24265+ sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
24266+ sbinfo->si_wbr_create = AuWbrCreate_Def;
4a4d8108
AM
24267+ sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
24268+ sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
1facf9fc 24269+
e49829fe 24270+ sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
1facf9fc 24271+
392086de
AM
24272+ sbinfo->si_xino_jiffy = jiffies;
24273+ sbinfo->si_xino_expire
24274+ = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
1facf9fc 24275+ mutex_init(&sbinfo->si_xib_mtx);
1facf9fc 24276+ sbinfo->si_xino_brid = -1;
24277+ /* leave si_xib_last_pindex and si_xib_next_bit */
24278+
e49829fe 24279+ sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
1facf9fc 24280+ sbinfo->si_rdblk = AUFS_RDBLK_DEF;
24281+ sbinfo->si_rdhash = AUFS_RDHASH_DEF;
24282+ sbinfo->si_dirwh = AUFS_DIRWH_DEF;
24283+
86dc4139
AM
24284+ for (i = 0; i < AuPlink_NHASH; i++)
24285+ au_sphl_init(sbinfo->si_plink + i);
1facf9fc 24286+ init_waitqueue_head(&sbinfo->si_plink_wq);
4a4d8108 24287+ spin_lock_init(&sbinfo->si_plink_maint_lock);
1facf9fc 24288+
523b37e3
AM
24289+ au_sphl_init(&sbinfo->si_files);
24290+
1facf9fc 24291+ /* leave other members for sysaufs and si_mnt. */
24292+ sbinfo->si_sb = sb;
24293+ sb->s_fs_info = sbinfo;
b752ccd1 24294+ si_pid_set(sb);
1facf9fc 24295+ au_debug_sbinfo_init(sbinfo);
24296+ return 0; /* success */
24297+
4f0767ce 24298+out_br:
1facf9fc 24299+ kfree(sbinfo->si_branch);
4f0767ce 24300+out_pidmap:
b752ccd1 24301+ kfree(sbinfo->au_si_pid.bitmap);
4f0767ce 24302+out_sbinfo:
1facf9fc 24303+ kfree(sbinfo);
4f0767ce 24304+out:
1facf9fc 24305+ return err;
24306+}
24307+
24308+int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr)
24309+{
24310+ int err, sz;
24311+ struct au_branch **brp;
24312+
dece6358
AM
24313+ AuRwMustWriteLock(&sbinfo->si_rwsem);
24314+
1facf9fc 24315+ err = -ENOMEM;
24316+ sz = sizeof(*brp) * (sbinfo->si_bend + 1);
24317+ if (unlikely(!sz))
24318+ sz = sizeof(*brp);
24319+ brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS);
24320+ if (brp) {
24321+ sbinfo->si_branch = brp;
24322+ err = 0;
24323+ }
24324+
24325+ return err;
24326+}
24327+
24328+/* ---------------------------------------------------------------------- */
24329+
24330+unsigned int au_sigen_inc(struct super_block *sb)
24331+{
24332+ unsigned int gen;
24333+
dece6358
AM
24334+ SiMustWriteLock(sb);
24335+
1facf9fc 24336+ gen = ++au_sbi(sb)->si_generation;
24337+ au_update_digen(sb->s_root);
537831f9 24338+ au_update_iigen(sb->s_root->d_inode, /*half*/0);
1facf9fc 24339+ sb->s_root->d_inode->i_version++;
24340+ return gen;
24341+}
24342+
24343+aufs_bindex_t au_new_br_id(struct super_block *sb)
24344+{
24345+ aufs_bindex_t br_id;
24346+ int i;
24347+ struct au_sbinfo *sbinfo;
24348+
dece6358
AM
24349+ SiMustWriteLock(sb);
24350+
1facf9fc 24351+ sbinfo = au_sbi(sb);
24352+ for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
24353+ br_id = ++sbinfo->si_last_br_id;
7f207e10 24354+ AuDebugOn(br_id < 0);
1facf9fc 24355+ if (br_id && au_br_index(sb, br_id) < 0)
24356+ return br_id;
24357+ }
24358+
24359+ return -1;
24360+}
24361+
24362+/* ---------------------------------------------------------------------- */
24363+
e49829fe
JR
24364+/* it is ok that new 'nwt' tasks are appended while we are sleeping */
24365+int si_read_lock(struct super_block *sb, int flags)
24366+{
24367+ int err;
24368+
24369+ err = 0;
24370+ if (au_ftest_lock(flags, FLUSH))
24371+ au_nwt_flush(&au_sbi(sb)->si_nowait);
24372+
24373+ si_noflush_read_lock(sb);
24374+ err = au_plink_maint(sb, flags);
24375+ if (unlikely(err))
24376+ si_read_unlock(sb);
24377+
24378+ return err;
24379+}
24380+
24381+int si_write_lock(struct super_block *sb, int flags)
24382+{
24383+ int err;
24384+
24385+ if (au_ftest_lock(flags, FLUSH))
24386+ au_nwt_flush(&au_sbi(sb)->si_nowait);
24387+
24388+ si_noflush_write_lock(sb);
24389+ err = au_plink_maint(sb, flags);
24390+ if (unlikely(err))
24391+ si_write_unlock(sb);
24392+
24393+ return err;
24394+}
24395+
1facf9fc 24396+/* dentry and super_block lock. call at entry point */
e49829fe 24397+int aufs_read_lock(struct dentry *dentry, int flags)
1facf9fc 24398+{
e49829fe 24399+ int err;
027c5e7a 24400+ struct super_block *sb;
e49829fe 24401+
027c5e7a
AM
24402+ sb = dentry->d_sb;
24403+ err = si_read_lock(sb, flags);
24404+ if (unlikely(err))
24405+ goto out;
24406+
24407+ if (au_ftest_lock(flags, DW))
24408+ di_write_lock_child(dentry);
24409+ else
24410+ di_read_lock_child(dentry, flags);
24411+
24412+ if (au_ftest_lock(flags, GEN)) {
24413+ err = au_digen_test(dentry, au_sigen(sb));
24414+ AuDebugOn(!err && au_dbrange_test(dentry));
24415+ if (unlikely(err))
24416+ aufs_read_unlock(dentry, flags);
e49829fe
JR
24417+ }
24418+
027c5e7a 24419+out:
e49829fe 24420+ return err;
1facf9fc 24421+}
24422+
24423+void aufs_read_unlock(struct dentry *dentry, int flags)
24424+{
24425+ if (au_ftest_lock(flags, DW))
24426+ di_write_unlock(dentry);
24427+ else
24428+ di_read_unlock(dentry, flags);
24429+ si_read_unlock(dentry->d_sb);
24430+}
24431+
24432+void aufs_write_lock(struct dentry *dentry)
24433+{
e49829fe 24434+ si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
1facf9fc 24435+ di_write_lock_child(dentry);
24436+}
24437+
24438+void aufs_write_unlock(struct dentry *dentry)
24439+{
24440+ di_write_unlock(dentry);
24441+ si_write_unlock(dentry->d_sb);
24442+}
24443+
e49829fe 24444+int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
1facf9fc 24445+{
e49829fe 24446+ int err;
027c5e7a
AM
24447+ unsigned int sigen;
24448+ struct super_block *sb;
e49829fe 24449+
027c5e7a
AM
24450+ sb = d1->d_sb;
24451+ err = si_read_lock(sb, flags);
24452+ if (unlikely(err))
24453+ goto out;
24454+
24455+ di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIR));
24456+
24457+ if (au_ftest_lock(flags, GEN)) {
24458+ sigen = au_sigen(sb);
24459+ err = au_digen_test(d1, sigen);
24460+ AuDebugOn(!err && au_dbrange_test(d1));
24461+ if (!err) {
24462+ err = au_digen_test(d2, sigen);
24463+ AuDebugOn(!err && au_dbrange_test(d2));
24464+ }
24465+ if (unlikely(err))
24466+ aufs_read_and_write_unlock2(d1, d2);
24467+ }
24468+
24469+out:
e49829fe 24470+ return err;
1facf9fc 24471+}
24472+
24473+void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
24474+{
24475+ di_write_unlock2(d1, d2);
24476+ si_read_unlock(d1->d_sb);
24477+}
b752ccd1
AM
24478+
24479+/* ---------------------------------------------------------------------- */
24480+
24481+int si_pid_test_slow(struct super_block *sb)
24482+{
24483+ void *p;
24484+
24485+ rcu_read_lock();
24486+ p = radix_tree_lookup(&au_sbi(sb)->au_si_pid.tree, current->pid);
24487+ rcu_read_unlock();
24488+
027c5e7a 24489+ return (long)!!p;
b752ccd1
AM
24490+}
24491+
24492+void si_pid_set_slow(struct super_block *sb)
24493+{
24494+ int err;
24495+ struct au_sbinfo *sbinfo;
24496+
24497+ AuDebugOn(si_pid_test_slow(sb));
24498+
24499+ sbinfo = au_sbi(sb);
24500+ err = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
24501+ AuDebugOn(err);
24502+ spin_lock(&sbinfo->au_si_pid.tree_lock);
24503+ err = radix_tree_insert(&sbinfo->au_si_pid.tree, current->pid,
027c5e7a 24504+ /*any valid ptr*/sb);
b752ccd1
AM
24505+ spin_unlock(&sbinfo->au_si_pid.tree_lock);
24506+ AuDebugOn(err);
24507+ radix_tree_preload_end();
24508+}
24509+
24510+void si_pid_clr_slow(struct super_block *sb)
24511+{
24512+ void *p;
24513+ struct au_sbinfo *sbinfo;
24514+
24515+ AuDebugOn(!si_pid_test_slow(sb));
24516+
24517+ sbinfo = au_sbi(sb);
24518+ spin_lock(&sbinfo->au_si_pid.tree_lock);
24519+ p = radix_tree_delete(&sbinfo->au_si_pid.tree, current->pid);
24520+ spin_unlock(&sbinfo->au_si_pid.tree_lock);
b752ccd1 24521+}
7f207e10
AM
24522diff -urN /usr/share/empty/fs/aufs/spl.h linux/fs/aufs/spl.h
24523--- /usr/share/empty/fs/aufs/spl.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 24524+++ linux/fs/aufs/spl.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 24525@@ -0,0 +1,111 @@
1facf9fc 24526+/*
523b37e3 24527+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 24528+ *
24529+ * This program, aufs is free software; you can redistribute it and/or modify
24530+ * it under the terms of the GNU General Public License as published by
24531+ * the Free Software Foundation; either version 2 of the License, or
24532+ * (at your option) any later version.
dece6358
AM
24533+ *
24534+ * This program is distributed in the hope that it will be useful,
24535+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
24536+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24537+ * GNU General Public License for more details.
24538+ *
24539+ * You should have received a copy of the GNU General Public License
523b37e3 24540+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 24541+ */
24542+
24543+/*
24544+ * simple list protected by a spinlock
24545+ */
24546+
24547+#ifndef __AUFS_SPL_H__
24548+#define __AUFS_SPL_H__
24549+
24550+#ifdef __KERNEL__
24551+
1facf9fc 24552+struct au_splhead {
24553+ spinlock_t spin;
24554+ struct list_head head;
24555+};
24556+
24557+static inline void au_spl_init(struct au_splhead *spl)
24558+{
24559+ spin_lock_init(&spl->spin);
24560+ INIT_LIST_HEAD(&spl->head);
24561+}
24562+
24563+static inline void au_spl_add(struct list_head *list, struct au_splhead *spl)
24564+{
24565+ spin_lock(&spl->spin);
24566+ list_add(list, &spl->head);
24567+ spin_unlock(&spl->spin);
24568+}
24569+
24570+static inline void au_spl_del(struct list_head *list, struct au_splhead *spl)
24571+{
24572+ spin_lock(&spl->spin);
24573+ list_del(list);
24574+ spin_unlock(&spl->spin);
24575+}
24576+
4a4d8108
AM
24577+static inline void au_spl_del_rcu(struct list_head *list,
24578+ struct au_splhead *spl)
24579+{
24580+ spin_lock(&spl->spin);
24581+ list_del_rcu(list);
24582+ spin_unlock(&spl->spin);
24583+}
24584+
86dc4139
AM
24585+/* ---------------------------------------------------------------------- */
24586+
24587+struct au_sphlhead {
24588+ spinlock_t spin;
24589+ struct hlist_head head;
24590+};
24591+
24592+static inline void au_sphl_init(struct au_sphlhead *sphl)
24593+{
24594+ spin_lock_init(&sphl->spin);
24595+ INIT_HLIST_HEAD(&sphl->head);
24596+}
24597+
24598+static inline void au_sphl_add(struct hlist_node *hlist,
24599+ struct au_sphlhead *sphl)
24600+{
24601+ spin_lock(&sphl->spin);
24602+ hlist_add_head(hlist, &sphl->head);
24603+ spin_unlock(&sphl->spin);
24604+}
24605+
24606+static inline void au_sphl_del(struct hlist_node *hlist,
24607+ struct au_sphlhead *sphl)
24608+{
24609+ spin_lock(&sphl->spin);
24610+ hlist_del(hlist);
24611+ spin_unlock(&sphl->spin);
24612+}
24613+
24614+static inline void au_sphl_del_rcu(struct hlist_node *hlist,
24615+ struct au_sphlhead *sphl)
24616+{
24617+ spin_lock(&sphl->spin);
24618+ hlist_del_rcu(hlist);
24619+ spin_unlock(&sphl->spin);
24620+}
24621+
24622+static inline unsigned long au_sphl_count(struct au_sphlhead *sphl)
24623+{
24624+ unsigned long cnt;
24625+ struct hlist_node *pos;
24626+
24627+ cnt = 0;
24628+ spin_lock(&sphl->spin);
24629+ hlist_for_each(pos, &sphl->head)
24630+ cnt++;
24631+ spin_unlock(&sphl->spin);
24632+ return cnt;
24633+}
24634+
1facf9fc 24635+#endif /* __KERNEL__ */
24636+#endif /* __AUFS_SPL_H__ */
7f207e10
AM
24637diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
24638--- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 24639+++ linux/fs/aufs/super.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 24640@@ -0,0 +1,1001 @@
1facf9fc 24641+/*
523b37e3 24642+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 24643+ *
24644+ * This program, aufs is free software; you can redistribute it and/or modify
24645+ * it under the terms of the GNU General Public License as published by
24646+ * the Free Software Foundation; either version 2 of the License, or
24647+ * (at your option) any later version.
dece6358
AM
24648+ *
24649+ * This program is distributed in the hope that it will be useful,
24650+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
24651+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24652+ * GNU General Public License for more details.
24653+ *
24654+ * You should have received a copy of the GNU General Public License
523b37e3 24655+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 24656+ */
24657+
24658+/*
24659+ * mount and super_block operations
24660+ */
24661+
f6c5ef8b 24662+#include <linux/mm.h>
dece6358 24663+#include <linux/module.h>
1facf9fc 24664+#include <linux/seq_file.h>
24665+#include <linux/statfs.h>
7f207e10
AM
24666+#include <linux/vmalloc.h>
24667+#include <linux/writeback.h>
1facf9fc 24668+#include "aufs.h"
24669+
24670+/*
24671+ * super_operations
24672+ */
24673+static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
24674+{
24675+ struct au_icntnr *c;
24676+
24677+ c = au_cache_alloc_icntnr();
24678+ if (c) {
027c5e7a 24679+ au_icntnr_init(c);
1facf9fc 24680+ c->vfs_inode.i_version = 1; /* sigen(sb); */
24681+ c->iinfo.ii_hinode = NULL;
24682+ return &c->vfs_inode;
24683+ }
24684+ return NULL;
24685+}
24686+
027c5e7a
AM
24687+static void aufs_destroy_inode_cb(struct rcu_head *head)
24688+{
24689+ struct inode *inode = container_of(head, struct inode, i_rcu);
24690+
b4510431 24691+ INIT_HLIST_HEAD(&inode->i_dentry);
027c5e7a
AM
24692+ au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
24693+}
24694+
1facf9fc 24695+static void aufs_destroy_inode(struct inode *inode)
24696+{
24697+ au_iinfo_fin(inode);
027c5e7a 24698+ call_rcu(&inode->i_rcu, aufs_destroy_inode_cb);
1facf9fc 24699+}
24700+
24701+struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
24702+{
24703+ struct inode *inode;
24704+ int err;
24705+
24706+ inode = iget_locked(sb, ino);
24707+ if (unlikely(!inode)) {
24708+ inode = ERR_PTR(-ENOMEM);
24709+ goto out;
24710+ }
24711+ if (!(inode->i_state & I_NEW))
24712+ goto out;
24713+
24714+ err = au_xigen_new(inode);
24715+ if (!err)
24716+ err = au_iinfo_init(inode);
24717+ if (!err)
24718+ inode->i_version++;
24719+ else {
24720+ iget_failed(inode);
24721+ inode = ERR_PTR(err);
24722+ }
24723+
4f0767ce 24724+out:
1facf9fc 24725+ /* never return NULL */
24726+ AuDebugOn(!inode);
24727+ AuTraceErrPtr(inode);
24728+ return inode;
24729+}
24730+
24731+/* lock free root dinfo */
24732+static int au_show_brs(struct seq_file *seq, struct super_block *sb)
24733+{
24734+ int err;
24735+ aufs_bindex_t bindex, bend;
24736+ struct path path;
4a4d8108 24737+ struct au_hdentry *hdp;
1facf9fc 24738+ struct au_branch *br;
1e00d052 24739+ char *perm;
1facf9fc 24740+
24741+ err = 0;
24742+ bend = au_sbend(sb);
4a4d8108 24743+ hdp = au_di(sb->s_root)->di_hdentry;
1facf9fc 24744+ for (bindex = 0; !err && bindex <= bend; bindex++) {
24745+ br = au_sbr(sb, bindex);
86dc4139 24746+ path.mnt = au_br_mnt(br);
4a4d8108 24747+ path.dentry = hdp[bindex].hd_dentry;
1facf9fc 24748+ err = au_seq_path(seq, &path);
1e00d052
AM
24749+ if (err > 0) {
24750+ perm = au_optstr_br_perm(br->br_perm);
24751+ if (perm) {
24752+ err = seq_printf(seq, "=%s", perm);
24753+ kfree(perm);
24754+ if (err == -1)
24755+ err = -E2BIG;
24756+ } else
24757+ err = -ENOMEM;
24758+ }
1facf9fc 24759+ if (!err && bindex != bend)
24760+ err = seq_putc(seq, ':');
24761+ }
24762+
24763+ return err;
24764+}
24765+
24766+static void au_show_wbr_create(struct seq_file *m, int v,
24767+ struct au_sbinfo *sbinfo)
24768+{
24769+ const char *pat;
24770+
dece6358
AM
24771+ AuRwMustAnyLock(&sbinfo->si_rwsem);
24772+
c2b27bf2 24773+ seq_puts(m, ",create=");
1facf9fc 24774+ pat = au_optstr_wbr_create(v);
24775+ switch (v) {
24776+ case AuWbrCreate_TDP:
24777+ case AuWbrCreate_RR:
24778+ case AuWbrCreate_MFS:
24779+ case AuWbrCreate_PMFS:
c2b27bf2 24780+ seq_puts(m, pat);
1facf9fc 24781+ break;
24782+ case AuWbrCreate_MFSV:
24783+ seq_printf(m, /*pat*/"mfs:%lu",
e49829fe
JR
24784+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
24785+ / MSEC_PER_SEC);
1facf9fc 24786+ break;
24787+ case AuWbrCreate_PMFSV:
24788+ seq_printf(m, /*pat*/"pmfs:%lu",
e49829fe
JR
24789+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
24790+ / MSEC_PER_SEC);
1facf9fc 24791+ break;
24792+ case AuWbrCreate_MFSRR:
24793+ seq_printf(m, /*pat*/"mfsrr:%llu",
24794+ sbinfo->si_wbr_mfs.mfsrr_watermark);
24795+ break;
24796+ case AuWbrCreate_MFSRRV:
24797+ seq_printf(m, /*pat*/"mfsrr:%llu:%lu",
24798+ sbinfo->si_wbr_mfs.mfsrr_watermark,
e49829fe
JR
24799+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
24800+ / MSEC_PER_SEC);
1facf9fc 24801+ break;
392086de
AM
24802+ case AuWbrCreate_PMFSRR:
24803+ seq_printf(m, /*pat*/"pmfsrr:%llu",
24804+ sbinfo->si_wbr_mfs.mfsrr_watermark);
24805+ break;
24806+ case AuWbrCreate_PMFSRRV:
24807+ seq_printf(m, /*pat*/"pmfsrr:%llu:%lu",
24808+ sbinfo->si_wbr_mfs.mfsrr_watermark,
24809+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
24810+ / MSEC_PER_SEC);
24811+ break;
1facf9fc 24812+ }
24813+}
24814+
7eafdf33 24815+static int au_show_xino(struct seq_file *seq, struct super_block *sb)
1facf9fc 24816+{
24817+#ifdef CONFIG_SYSFS
24818+ return 0;
24819+#else
24820+ int err;
24821+ const int len = sizeof(AUFS_XINO_FNAME) - 1;
24822+ aufs_bindex_t bindex, brid;
1facf9fc 24823+ struct qstr *name;
24824+ struct file *f;
24825+ struct dentry *d, *h_root;
4a4d8108 24826+ struct au_hdentry *hdp;
1facf9fc 24827+
dece6358
AM
24828+ AuRwMustAnyLock(&sbinfo->si_rwsem);
24829+
1facf9fc 24830+ err = 0;
1facf9fc 24831+ f = au_sbi(sb)->si_xib;
24832+ if (!f)
24833+ goto out;
24834+
24835+ /* stop printing the default xino path on the first writable branch */
24836+ h_root = NULL;
24837+ brid = au_xino_brid(sb);
24838+ if (brid >= 0) {
24839+ bindex = au_br_index(sb, brid);
4a4d8108
AM
24840+ hdp = au_di(sb->s_root)->di_hdentry;
24841+ h_root = hdp[0 + bindex].hd_dentry;
1facf9fc 24842+ }
24843+ d = f->f_dentry;
24844+ name = &d->d_name;
24845+ /* safe ->d_parent because the file is unlinked */
24846+ if (d->d_parent == h_root
24847+ && name->len == len
24848+ && !memcmp(name->name, AUFS_XINO_FNAME, len))
24849+ goto out;
24850+
24851+ seq_puts(seq, ",xino=");
24852+ err = au_xino_path(seq, f);
24853+
4f0767ce 24854+out:
1facf9fc 24855+ return err;
24856+#endif
24857+}
24858+
24859+/* seq_file will re-call me in case of too long string */
7eafdf33 24860+static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
1facf9fc 24861+{
027c5e7a 24862+ int err;
1facf9fc 24863+ unsigned int mnt_flags, v;
24864+ struct super_block *sb;
24865+ struct au_sbinfo *sbinfo;
24866+
24867+#define AuBool(name, str) do { \
24868+ v = au_opt_test(mnt_flags, name); \
24869+ if (v != au_opt_test(AuOpt_Def, name)) \
24870+ seq_printf(m, ",%s" #str, v ? "" : "no"); \
24871+} while (0)
24872+
24873+#define AuStr(name, str) do { \
24874+ v = mnt_flags & AuOptMask_##name; \
24875+ if (v != (AuOpt_Def & AuOptMask_##name)) \
24876+ seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
24877+} while (0)
24878+
24879+#define AuUInt(name, str, val) do { \
24880+ if (val != AUFS_##name##_DEF) \
24881+ seq_printf(m, "," #str "=%u", val); \
24882+} while (0)
24883+
24884+ /* lock free root dinfo */
7eafdf33 24885+ sb = dentry->d_sb;
1facf9fc 24886+ si_noflush_read_lock(sb);
24887+ sbinfo = au_sbi(sb);
24888+ seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
24889+
24890+ mnt_flags = au_mntflags(sb);
24891+ if (au_opt_test(mnt_flags, XINO)) {
7eafdf33 24892+ err = au_show_xino(m, sb);
1facf9fc 24893+ if (unlikely(err))
24894+ goto out;
24895+ } else
24896+ seq_puts(m, ",noxino");
24897+
24898+ AuBool(TRUNC_XINO, trunc_xino);
24899+ AuStr(UDBA, udba);
dece6358 24900+ AuBool(SHWH, shwh);
1facf9fc 24901+ AuBool(PLINK, plink);
4a4d8108 24902+ AuBool(DIO, dio);
1facf9fc 24903+ /* AuBool(DIRPERM1, dirperm1); */
24904+ /* AuBool(REFROF, refrof); */
24905+
24906+ v = sbinfo->si_wbr_create;
24907+ if (v != AuWbrCreate_Def)
24908+ au_show_wbr_create(m, v, sbinfo);
24909+
24910+ v = sbinfo->si_wbr_copyup;
24911+ if (v != AuWbrCopyup_Def)
24912+ seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
24913+
24914+ v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
24915+ if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
24916+ seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
24917+
24918+ AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
24919+
027c5e7a
AM
24920+ v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
24921+ AuUInt(RDCACHE, rdcache, v);
1facf9fc 24922+
24923+ AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
24924+ AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
24925+
24926+ AuBool(SUM, sum);
24927+ /* AuBool(SUM_W, wsum); */
24928+ AuBool(WARN_PERM, warn_perm);
24929+ AuBool(VERBOSE, verbose);
24930+
4f0767ce 24931+out:
1facf9fc 24932+ /* be sure to print "br:" last */
24933+ if (!sysaufs_brs) {
24934+ seq_puts(m, ",br:");
24935+ au_show_brs(m, sb);
24936+ }
24937+ si_read_unlock(sb);
24938+ return 0;
24939+
1facf9fc 24940+#undef AuBool
24941+#undef AuStr
4a4d8108 24942+#undef AuUInt
1facf9fc 24943+}
24944+
24945+/* ---------------------------------------------------------------------- */
24946+
24947+/* sum mode which returns the summation for statfs(2) */
24948+
24949+static u64 au_add_till_max(u64 a, u64 b)
24950+{
24951+ u64 old;
24952+
24953+ old = a;
24954+ a += b;
92d182d2
AM
24955+ if (old <= a)
24956+ return a;
24957+ return ULLONG_MAX;
24958+}
24959+
24960+static u64 au_mul_till_max(u64 a, long mul)
24961+{
24962+ u64 old;
24963+
24964+ old = a;
24965+ a *= mul;
24966+ if (old <= a)
1facf9fc 24967+ return a;
24968+ return ULLONG_MAX;
24969+}
24970+
24971+static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
24972+{
24973+ int err;
92d182d2 24974+ long bsize, factor;
1facf9fc 24975+ u64 blocks, bfree, bavail, files, ffree;
24976+ aufs_bindex_t bend, bindex, i;
24977+ unsigned char shared;
7f207e10 24978+ struct path h_path;
1facf9fc 24979+ struct super_block *h_sb;
24980+
92d182d2
AM
24981+ err = 0;
24982+ bsize = LONG_MAX;
24983+ files = 0;
24984+ ffree = 0;
1facf9fc 24985+ blocks = 0;
24986+ bfree = 0;
24987+ bavail = 0;
1facf9fc 24988+ bend = au_sbend(sb);
92d182d2 24989+ for (bindex = 0; bindex <= bend; bindex++) {
7f207e10
AM
24990+ h_path.mnt = au_sbr_mnt(sb, bindex);
24991+ h_sb = h_path.mnt->mnt_sb;
1facf9fc 24992+ shared = 0;
92d182d2 24993+ for (i = 0; !shared && i < bindex; i++)
1facf9fc 24994+ shared = (au_sbr_sb(sb, i) == h_sb);
24995+ if (shared)
24996+ continue;
24997+
24998+ /* sb->s_root for NFS is unreliable */
7f207e10
AM
24999+ h_path.dentry = h_path.mnt->mnt_root;
25000+ err = vfs_statfs(&h_path, buf);
1facf9fc 25001+ if (unlikely(err))
25002+ goto out;
25003+
92d182d2
AM
25004+ if (bsize > buf->f_bsize) {
25005+ /*
25006+ * we will reduce bsize, so we have to expand blocks
25007+ * etc. to match them again
25008+ */
25009+ factor = (bsize / buf->f_bsize);
25010+ blocks = au_mul_till_max(blocks, factor);
25011+ bfree = au_mul_till_max(bfree, factor);
25012+ bavail = au_mul_till_max(bavail, factor);
25013+ bsize = buf->f_bsize;
25014+ }
25015+
25016+ factor = (buf->f_bsize / bsize);
25017+ blocks = au_add_till_max(blocks,
25018+ au_mul_till_max(buf->f_blocks, factor));
25019+ bfree = au_add_till_max(bfree,
25020+ au_mul_till_max(buf->f_bfree, factor));
25021+ bavail = au_add_till_max(bavail,
25022+ au_mul_till_max(buf->f_bavail, factor));
1facf9fc 25023+ files = au_add_till_max(files, buf->f_files);
25024+ ffree = au_add_till_max(ffree, buf->f_ffree);
25025+ }
25026+
92d182d2 25027+ buf->f_bsize = bsize;
1facf9fc 25028+ buf->f_blocks = blocks;
25029+ buf->f_bfree = bfree;
25030+ buf->f_bavail = bavail;
25031+ buf->f_files = files;
25032+ buf->f_ffree = ffree;
92d182d2 25033+ buf->f_frsize = 0;
1facf9fc 25034+
4f0767ce 25035+out:
1facf9fc 25036+ return err;
25037+}
25038+
25039+static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
25040+{
25041+ int err;
7f207e10 25042+ struct path h_path;
1facf9fc 25043+ struct super_block *sb;
25044+
25045+ /* lock free root dinfo */
25046+ sb = dentry->d_sb;
25047+ si_noflush_read_lock(sb);
7f207e10 25048+ if (!au_opt_test(au_mntflags(sb), SUM)) {
1facf9fc 25049+ /* sb->s_root for NFS is unreliable */
7f207e10
AM
25050+ h_path.mnt = au_sbr_mnt(sb, 0);
25051+ h_path.dentry = h_path.mnt->mnt_root;
25052+ err = vfs_statfs(&h_path, buf);
25053+ } else
1facf9fc 25054+ err = au_statfs_sum(sb, buf);
25055+ si_read_unlock(sb);
25056+
25057+ if (!err) {
25058+ buf->f_type = AUFS_SUPER_MAGIC;
4a4d8108 25059+ buf->f_namelen = AUFS_MAX_NAMELEN;
1facf9fc 25060+ memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
25061+ }
25062+ /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
25063+
25064+ return err;
25065+}
25066+
25067+/* ---------------------------------------------------------------------- */
25068+
537831f9
AM
25069+static int aufs_sync_fs(struct super_block *sb, int wait)
25070+{
25071+ int err, e;
25072+ aufs_bindex_t bend, bindex;
25073+ struct au_branch *br;
25074+ struct super_block *h_sb;
25075+
25076+ err = 0;
25077+ si_noflush_read_lock(sb);
25078+ bend = au_sbend(sb);
25079+ for (bindex = 0; bindex <= bend; bindex++) {
25080+ br = au_sbr(sb, bindex);
25081+ if (!au_br_writable(br->br_perm))
25082+ continue;
25083+
25084+ h_sb = au_sbr_sb(sb, bindex);
25085+ if (h_sb->s_op->sync_fs) {
25086+ e = h_sb->s_op->sync_fs(h_sb, wait);
25087+ if (unlikely(e && !err))
25088+ err = e;
25089+ /* go on even if an error happens */
25090+ }
25091+ }
25092+ si_read_unlock(sb);
25093+
25094+ return err;
25095+}
25096+
25097+/* ---------------------------------------------------------------------- */
25098+
1facf9fc 25099+/* final actions when unmounting a file system */
25100+static void aufs_put_super(struct super_block *sb)
25101+{
25102+ struct au_sbinfo *sbinfo;
25103+
25104+ sbinfo = au_sbi(sb);
25105+ if (!sbinfo)
25106+ return;
25107+
1facf9fc 25108+ dbgaufs_si_fin(sbinfo);
25109+ kobject_put(&sbinfo->si_kobj);
25110+}
25111+
25112+/* ---------------------------------------------------------------------- */
25113+
7f207e10
AM
25114+void au_array_free(void *array)
25115+{
25116+ if (array) {
25117+ if (!is_vmalloc_addr(array))
25118+ kfree(array);
25119+ else
25120+ vfree(array);
25121+ }
25122+}
25123+
25124+void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb, void *arg)
25125+{
25126+ void *array;
25127+ unsigned long long n;
25128+
25129+ array = NULL;
25130+ n = 0;
25131+ if (!*hint)
25132+ goto out;
25133+
25134+ if (*hint > ULLONG_MAX / sizeof(array)) {
25135+ array = ERR_PTR(-EMFILE);
25136+ pr_err("hint %llu\n", *hint);
25137+ goto out;
25138+ }
25139+
25140+ array = kmalloc(sizeof(array) * *hint, GFP_NOFS);
25141+ if (unlikely(!array))
25142+ array = vmalloc(sizeof(array) * *hint);
25143+ if (unlikely(!array)) {
25144+ array = ERR_PTR(-ENOMEM);
25145+ goto out;
25146+ }
25147+
25148+ n = cb(array, *hint, arg);
25149+ AuDebugOn(n > *hint);
25150+
25151+out:
25152+ *hint = n;
25153+ return array;
25154+}
25155+
25156+static unsigned long long au_iarray_cb(void *a,
25157+ unsigned long long max __maybe_unused,
25158+ void *arg)
25159+{
25160+ unsigned long long n;
25161+ struct inode **p, *inode;
25162+ struct list_head *head;
25163+
25164+ n = 0;
25165+ p = a;
25166+ head = arg;
2cbb1c4b 25167+ spin_lock(&inode_sb_list_lock);
7f207e10
AM
25168+ list_for_each_entry(inode, head, i_sb_list) {
25169+ if (!is_bad_inode(inode)
25170+ && au_ii(inode)->ii_bstart >= 0) {
2cbb1c4b
JR
25171+ spin_lock(&inode->i_lock);
25172+ if (atomic_read(&inode->i_count)) {
25173+ au_igrab(inode);
25174+ *p++ = inode;
25175+ n++;
25176+ AuDebugOn(n > max);
25177+ }
25178+ spin_unlock(&inode->i_lock);
7f207e10
AM
25179+ }
25180+ }
2cbb1c4b 25181+ spin_unlock(&inode_sb_list_lock);
7f207e10
AM
25182+
25183+ return n;
25184+}
25185+
25186+struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
25187+{
25188+ *max = atomic_long_read(&au_sbi(sb)->si_ninodes);
25189+ return au_array_alloc(max, au_iarray_cb, &sb->s_inodes);
25190+}
25191+
25192+void au_iarray_free(struct inode **a, unsigned long long max)
25193+{
25194+ unsigned long long ull;
25195+
25196+ for (ull = 0; ull < max; ull++)
25197+ iput(a[ull]);
25198+ au_array_free(a);
25199+}
25200+
25201+/* ---------------------------------------------------------------------- */
25202+
1facf9fc 25203+/*
25204+ * refresh dentry and inode at remount time.
25205+ */
027c5e7a
AM
25206+/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
25207+static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
25208+ struct dentry *parent)
1facf9fc 25209+{
25210+ int err;
1facf9fc 25211+
25212+ di_write_lock_child(dentry);
1facf9fc 25213+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a
AM
25214+ err = au_refresh_dentry(dentry, parent);
25215+ if (!err && dir_flags)
25216+ au_hn_reset(dentry->d_inode, dir_flags);
1facf9fc 25217+ di_read_unlock(parent, AuLock_IR);
1facf9fc 25218+ di_write_unlock(dentry);
25219+
25220+ return err;
25221+}
25222+
027c5e7a
AM
25223+static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
25224+ struct au_sbinfo *sbinfo,
25225+ const unsigned int dir_flags)
1facf9fc 25226+{
027c5e7a
AM
25227+ int err;
25228+ struct dentry *parent;
25229+ struct inode *inode;
25230+
25231+ err = 0;
25232+ parent = dget_parent(dentry);
25233+ if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
25234+ inode = dentry->d_inode;
25235+ if (inode) {
25236+ if (!S_ISDIR(inode->i_mode))
25237+ err = au_do_refresh(dentry, /*dir_flags*/0,
25238+ parent);
25239+ else {
25240+ err = au_do_refresh(dentry, dir_flags, parent);
25241+ if (unlikely(err))
25242+ au_fset_si(sbinfo, FAILED_REFRESH_DIR);
25243+ }
25244+ } else
25245+ err = au_do_refresh(dentry, /*dir_flags*/0, parent);
25246+ AuDbgDentry(dentry);
25247+ }
25248+ dput(parent);
25249+
25250+ AuTraceErr(err);
25251+ return err;
1facf9fc 25252+}
25253+
027c5e7a 25254+static int au_refresh_d(struct super_block *sb)
1facf9fc 25255+{
25256+ int err, i, j, ndentry, e;
027c5e7a 25257+ unsigned int sigen;
1facf9fc 25258+ struct au_dcsub_pages dpages;
25259+ struct au_dpage *dpage;
027c5e7a
AM
25260+ struct dentry **dentries, *d;
25261+ struct au_sbinfo *sbinfo;
25262+ struct dentry *root = sb->s_root;
25263+ const unsigned int dir_flags = au_hi_flags(root->d_inode, /*isdir*/1);
1facf9fc 25264+
027c5e7a
AM
25265+ err = au_dpages_init(&dpages, GFP_NOFS);
25266+ if (unlikely(err))
1facf9fc 25267+ goto out;
027c5e7a
AM
25268+ err = au_dcsub_pages(&dpages, root, NULL, NULL);
25269+ if (unlikely(err))
1facf9fc 25270+ goto out_dpages;
1facf9fc 25271+
027c5e7a
AM
25272+ sigen = au_sigen(sb);
25273+ sbinfo = au_sbi(sb);
25274+ for (i = 0; i < dpages.ndpage; i++) {
1facf9fc 25275+ dpage = dpages.dpages + i;
25276+ dentries = dpage->dentries;
25277+ ndentry = dpage->ndentry;
027c5e7a 25278+ for (j = 0; j < ndentry; j++) {
1facf9fc 25279+ d = dentries[j];
027c5e7a
AM
25280+ e = au_do_refresh_d(d, sigen, sbinfo, dir_flags);
25281+ if (unlikely(e && !err))
25282+ err = e;
25283+ /* go on even err */
1facf9fc 25284+ }
25285+ }
25286+
4f0767ce 25287+out_dpages:
1facf9fc 25288+ au_dpages_free(&dpages);
4f0767ce 25289+out:
1facf9fc 25290+ return err;
25291+}
25292+
027c5e7a 25293+static int au_refresh_i(struct super_block *sb)
1facf9fc 25294+{
027c5e7a
AM
25295+ int err, e;
25296+ unsigned int sigen;
25297+ unsigned long long max, ull;
25298+ struct inode *inode, **array;
1facf9fc 25299+
027c5e7a
AM
25300+ array = au_iarray_alloc(sb, &max);
25301+ err = PTR_ERR(array);
25302+ if (IS_ERR(array))
25303+ goto out;
1facf9fc 25304+
25305+ err = 0;
027c5e7a
AM
25306+ sigen = au_sigen(sb);
25307+ for (ull = 0; ull < max; ull++) {
25308+ inode = array[ull];
537831f9 25309+ if (au_iigen(inode, NULL) != sigen) {
1facf9fc 25310+ ii_write_lock_child(inode);
027c5e7a 25311+ e = au_refresh_hinode_self(inode);
1facf9fc 25312+ ii_write_unlock(inode);
25313+ if (unlikely(e)) {
027c5e7a 25314+ pr_err("error %d, i%lu\n", e, inode->i_ino);
1facf9fc 25315+ if (!err)
25316+ err = e;
25317+ /* go on even if err */
25318+ }
25319+ }
1facf9fc 25320+ }
25321+
027c5e7a 25322+ au_iarray_free(array, max);
1facf9fc 25323+
4f0767ce 25324+out:
1facf9fc 25325+ return err;
25326+}
25327+
027c5e7a 25328+static void au_remount_refresh(struct super_block *sb)
1facf9fc 25329+{
027c5e7a
AM
25330+ int err, e;
25331+ unsigned int udba;
25332+ aufs_bindex_t bindex, bend;
1facf9fc 25333+ struct dentry *root;
25334+ struct inode *inode;
027c5e7a 25335+ struct au_branch *br;
1facf9fc 25336+
25337+ au_sigen_inc(sb);
027c5e7a 25338+ au_fclr_si(au_sbi(sb), FAILED_REFRESH_DIR);
1facf9fc 25339+
25340+ root = sb->s_root;
25341+ DiMustNoWaiters(root);
25342+ inode = root->d_inode;
25343+ IiMustNoWaiters(inode);
1facf9fc 25344+
027c5e7a
AM
25345+ udba = au_opt_udba(sb);
25346+ bend = au_sbend(sb);
25347+ for (bindex = 0; bindex <= bend; bindex++) {
25348+ br = au_sbr(sb, bindex);
25349+ err = au_hnotify_reset_br(udba, br, br->br_perm);
1facf9fc 25350+ if (unlikely(err))
027c5e7a
AM
25351+ AuIOErr("hnotify failed on br %d, %d, ignored\n",
25352+ bindex, err);
25353+ /* go on even if err */
1facf9fc 25354+ }
027c5e7a 25355+ au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
1facf9fc 25356+
027c5e7a
AM
25357+ di_write_unlock(root);
25358+ err = au_refresh_d(sb);
25359+ e = au_refresh_i(sb);
25360+ if (unlikely(e && !err))
25361+ err = e;
1facf9fc 25362+ /* aufs_write_lock() calls ..._child() */
25363+ di_write_lock_child(root);
027c5e7a
AM
25364+
25365+ au_cpup_attr_all(inode, /*force*/1);
25366+
25367+ if (unlikely(err))
25368+ AuIOErr("refresh failed, ignored, %d\n", err);
1facf9fc 25369+}
25370+
25371+/* stop extra interpretation of errno in mount(8), and strange error messages */
25372+static int cvt_err(int err)
25373+{
25374+ AuTraceErr(err);
25375+
25376+ switch (err) {
25377+ case -ENOENT:
25378+ case -ENOTDIR:
25379+ case -EEXIST:
25380+ case -EIO:
25381+ err = -EINVAL;
25382+ }
25383+ return err;
25384+}
25385+
25386+static int aufs_remount_fs(struct super_block *sb, int *flags, char *data)
25387+{
4a4d8108
AM
25388+ int err, do_dx;
25389+ unsigned int mntflags;
1facf9fc 25390+ struct au_opts opts;
25391+ struct dentry *root;
25392+ struct inode *inode;
25393+ struct au_sbinfo *sbinfo;
25394+
25395+ err = 0;
25396+ root = sb->s_root;
25397+ if (!data || !*data) {
e49829fe
JR
25398+ err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
25399+ if (!err) {
25400+ di_write_lock_child(root);
25401+ err = au_opts_verify(sb, *flags, /*pending*/0);
25402+ aufs_write_unlock(root);
25403+ }
1facf9fc 25404+ goto out;
25405+ }
25406+
25407+ err = -ENOMEM;
25408+ memset(&opts, 0, sizeof(opts));
25409+ opts.opt = (void *)__get_free_page(GFP_NOFS);
25410+ if (unlikely(!opts.opt))
25411+ goto out;
25412+ opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
25413+ opts.flags = AuOpts_REMOUNT;
25414+ opts.sb_flags = *flags;
25415+
25416+ /* parse it before aufs lock */
25417+ err = au_opts_parse(sb, data, &opts);
25418+ if (unlikely(err))
25419+ goto out_opts;
25420+
25421+ sbinfo = au_sbi(sb);
25422+ inode = root->d_inode;
25423+ mutex_lock(&inode->i_mutex);
e49829fe
JR
25424+ err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
25425+ if (unlikely(err))
25426+ goto out_mtx;
25427+ di_write_lock_child(root);
1facf9fc 25428+
25429+ /* au_opts_remount() may return an error */
25430+ err = au_opts_remount(sb, &opts);
25431+ au_opts_free(&opts);
25432+
027c5e7a
AM
25433+ if (au_ftest_opts(opts.flags, REFRESH))
25434+ au_remount_refresh(sb);
1facf9fc 25435+
4a4d8108
AM
25436+ if (au_ftest_opts(opts.flags, REFRESH_DYAOP)) {
25437+ mntflags = au_mntflags(sb);
25438+ do_dx = !!au_opt_test(mntflags, DIO);
25439+ au_dy_arefresh(do_dx);
25440+ }
25441+
1facf9fc 25442+ aufs_write_unlock(root);
953406b4 25443+
e49829fe
JR
25444+out_mtx:
25445+ mutex_unlock(&inode->i_mutex);
4f0767ce 25446+out_opts:
1facf9fc 25447+ free_page((unsigned long)opts.opt);
4f0767ce 25448+out:
1facf9fc 25449+ err = cvt_err(err);
25450+ AuTraceErr(err);
25451+ return err;
25452+}
25453+
4a4d8108 25454+static const struct super_operations aufs_sop = {
1facf9fc 25455+ .alloc_inode = aufs_alloc_inode,
25456+ .destroy_inode = aufs_destroy_inode,
b752ccd1 25457+ /* always deleting, no clearing */
1facf9fc 25458+ .drop_inode = generic_delete_inode,
25459+ .show_options = aufs_show_options,
25460+ .statfs = aufs_statfs,
25461+ .put_super = aufs_put_super,
537831f9 25462+ .sync_fs = aufs_sync_fs,
1facf9fc 25463+ .remount_fs = aufs_remount_fs
25464+};
25465+
25466+/* ---------------------------------------------------------------------- */
25467+
25468+static int alloc_root(struct super_block *sb)
25469+{
25470+ int err;
25471+ struct inode *inode;
25472+ struct dentry *root;
25473+
25474+ err = -ENOMEM;
25475+ inode = au_iget_locked(sb, AUFS_ROOT_INO);
25476+ err = PTR_ERR(inode);
25477+ if (IS_ERR(inode))
25478+ goto out;
25479+
25480+ inode->i_op = &aufs_dir_iop;
25481+ inode->i_fop = &aufs_dir_fop;
25482+ inode->i_mode = S_IFDIR;
9dbd164d 25483+ set_nlink(inode, 2);
1facf9fc 25484+ unlock_new_inode(inode);
25485+
92d182d2 25486+ root = d_make_root(inode);
1facf9fc 25487+ if (unlikely(!root))
92d182d2 25488+ goto out;
1facf9fc 25489+ err = PTR_ERR(root);
25490+ if (IS_ERR(root))
92d182d2 25491+ goto out;
1facf9fc 25492+
4a4d8108 25493+ err = au_di_init(root);
1facf9fc 25494+ if (!err) {
25495+ sb->s_root = root;
25496+ return 0; /* success */
25497+ }
25498+ dput(root);
1facf9fc 25499+
4f0767ce 25500+out:
1facf9fc 25501+ return err;
1facf9fc 25502+}
25503+
25504+static int aufs_fill_super(struct super_block *sb, void *raw_data,
25505+ int silent __maybe_unused)
25506+{
25507+ int err;
25508+ struct au_opts opts;
25509+ struct dentry *root;
25510+ struct inode *inode;
25511+ char *arg = raw_data;
25512+
25513+ if (unlikely(!arg || !*arg)) {
25514+ err = -EINVAL;
4a4d8108 25515+ pr_err("no arg\n");
1facf9fc 25516+ goto out;
25517+ }
25518+
25519+ err = -ENOMEM;
25520+ memset(&opts, 0, sizeof(opts));
25521+ opts.opt = (void *)__get_free_page(GFP_NOFS);
25522+ if (unlikely(!opts.opt))
25523+ goto out;
25524+ opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
25525+ opts.sb_flags = sb->s_flags;
25526+
25527+ err = au_si_alloc(sb);
25528+ if (unlikely(err))
25529+ goto out_opts;
25530+
25531+ /* all timestamps always follow the ones on the branch */
25532+ sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
25533+ sb->s_op = &aufs_sop;
027c5e7a 25534+ sb->s_d_op = &aufs_dop;
1facf9fc 25535+ sb->s_magic = AUFS_SUPER_MAGIC;
25536+ sb->s_maxbytes = 0;
25537+ au_export_init(sb);
25538+
25539+ err = alloc_root(sb);
25540+ if (unlikely(err)) {
25541+ si_write_unlock(sb);
25542+ goto out_info;
25543+ }
25544+ root = sb->s_root;
25545+ inode = root->d_inode;
25546+
25547+ /*
25548+ * actually we can parse options regardless aufs lock here.
25549+ * but at remount time, parsing must be done before aufs lock.
25550+ * so we follow the same rule.
25551+ */
25552+ ii_write_lock_parent(inode);
25553+ aufs_write_unlock(root);
25554+ err = au_opts_parse(sb, arg, &opts);
25555+ if (unlikely(err))
25556+ goto out_root;
25557+
25558+ /* lock vfs_inode first, then aufs. */
25559+ mutex_lock(&inode->i_mutex);
1facf9fc 25560+ aufs_write_lock(root);
25561+ err = au_opts_mount(sb, &opts);
25562+ au_opts_free(&opts);
1facf9fc 25563+ aufs_write_unlock(root);
25564+ mutex_unlock(&inode->i_mutex);
4a4d8108
AM
25565+ if (!err)
25566+ goto out_opts; /* success */
1facf9fc 25567+
4f0767ce 25568+out_root:
1facf9fc 25569+ dput(root);
25570+ sb->s_root = NULL;
4f0767ce 25571+out_info:
2cbb1c4b 25572+ dbgaufs_si_fin(au_sbi(sb));
1facf9fc 25573+ kobject_put(&au_sbi(sb)->si_kobj);
25574+ sb->s_fs_info = NULL;
4f0767ce 25575+out_opts:
1facf9fc 25576+ free_page((unsigned long)opts.opt);
4f0767ce 25577+out:
1facf9fc 25578+ AuTraceErr(err);
25579+ err = cvt_err(err);
25580+ AuTraceErr(err);
25581+ return err;
25582+}
25583+
25584+/* ---------------------------------------------------------------------- */
25585+
027c5e7a
AM
25586+static struct dentry *aufs_mount(struct file_system_type *fs_type, int flags,
25587+ const char *dev_name __maybe_unused,
25588+ void *raw_data)
1facf9fc 25589+{
027c5e7a 25590+ struct dentry *root;
1facf9fc 25591+ struct super_block *sb;
25592+
25593+ /* all timestamps always follow the ones on the branch */
25594+ /* mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; */
027c5e7a
AM
25595+ root = mount_nodev(fs_type, flags, raw_data, aufs_fill_super);
25596+ if (IS_ERR(root))
25597+ goto out;
25598+
25599+ sb = root->d_sb;
25600+ si_write_lock(sb, !AuLock_FLUSH);
25601+ sysaufs_brs_add(sb, 0);
25602+ si_write_unlock(sb);
25603+ au_sbilist_add(sb);
25604+
25605+out:
25606+ return root;
1facf9fc 25607+}
25608+
e49829fe
JR
25609+static void aufs_kill_sb(struct super_block *sb)
25610+{
25611+ struct au_sbinfo *sbinfo;
25612+
25613+ sbinfo = au_sbi(sb);
25614+ if (sbinfo) {
25615+ au_sbilist_del(sb);
25616+ aufs_write_lock(sb->s_root);
25617+ if (sbinfo->si_wbr_create_ops->fin)
25618+ sbinfo->si_wbr_create_ops->fin(sb);
25619+ if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
25620+ au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
027c5e7a 25621+ au_remount_refresh(sb);
e49829fe
JR
25622+ }
25623+ if (au_opt_test(sbinfo->si_mntflags, PLINK))
25624+ au_plink_put(sb, /*verbose*/1);
25625+ au_xino_clr(sb);
1e00d052 25626+ sbinfo->si_sb = NULL;
e49829fe 25627+ aufs_write_unlock(sb->s_root);
e49829fe
JR
25628+ au_nwt_flush(&sbinfo->si_nowait);
25629+ }
98d9a5b1 25630+ kill_anon_super(sb);
e49829fe
JR
25631+}
25632+
1facf9fc 25633+struct file_system_type aufs_fs_type = {
25634+ .name = AUFS_FSTYPE,
c06a8ce3
AM
25635+ /* a race between rename and others */
25636+ .fs_flags = FS_RENAME_DOES_D_MOVE,
027c5e7a 25637+ .mount = aufs_mount,
e49829fe 25638+ .kill_sb = aufs_kill_sb,
1facf9fc 25639+ /* no need to __module_get() and module_put(). */
25640+ .owner = THIS_MODULE,
25641+};
7f207e10
AM
25642diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
25643--- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 25644+++ linux/fs/aufs/super.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 25645@@ -0,0 +1,571 @@
1facf9fc 25646+/*
523b37e3 25647+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 25648+ *
25649+ * This program, aufs is free software; you can redistribute it and/or modify
25650+ * it under the terms of the GNU General Public License as published by
25651+ * the Free Software Foundation; either version 2 of the License, or
25652+ * (at your option) any later version.
dece6358
AM
25653+ *
25654+ * This program is distributed in the hope that it will be useful,
25655+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
25656+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25657+ * GNU General Public License for more details.
25658+ *
25659+ * You should have received a copy of the GNU General Public License
523b37e3 25660+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 25661+ */
25662+
25663+/*
25664+ * super_block operations
25665+ */
25666+
25667+#ifndef __AUFS_SUPER_H__
25668+#define __AUFS_SUPER_H__
25669+
25670+#ifdef __KERNEL__
25671+
25672+#include <linux/fs.h>
1facf9fc 25673+#include "rwsem.h"
25674+#include "spl.h"
25675+#include "wkq.h"
25676+
25677+typedef ssize_t (*au_readf_t)(struct file *, char __user *, size_t, loff_t *);
25678+typedef ssize_t (*au_writef_t)(struct file *, const char __user *, size_t,
25679+ loff_t *);
25680+
25681+/* policies to select one among multiple writable branches */
25682+struct au_wbr_copyup_operations {
25683+ int (*copyup)(struct dentry *dentry);
25684+};
25685+
392086de
AM
25686+#define AuWbr_DIR 1 /* target is a dir */
25687+#define AuWbr_PARENT (1 << 1) /* always require a parent */
25688+
25689+#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
25690+#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
25691+#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
25692+
1facf9fc 25693+struct au_wbr_create_operations {
392086de 25694+ int (*create)(struct dentry *dentry, unsigned int flags);
1facf9fc 25695+ int (*init)(struct super_block *sb);
25696+ int (*fin)(struct super_block *sb);
25697+};
25698+
25699+struct au_wbr_mfs {
25700+ struct mutex mfs_lock; /* protect this structure */
25701+ unsigned long mfs_jiffy;
25702+ unsigned long mfs_expire;
25703+ aufs_bindex_t mfs_bindex;
25704+
25705+ unsigned long long mfsrr_bytes;
25706+ unsigned long long mfsrr_watermark;
25707+};
25708+
86dc4139
AM
25709+struct pseudo_link {
25710+ union {
25711+ struct hlist_node hlist;
25712+ struct rcu_head rcu;
25713+ };
25714+ struct inode *inode;
25715+};
25716+
25717+#define AuPlink_NHASH 100
25718+static inline int au_plink_hash(ino_t ino)
25719+{
25720+ return ino % AuPlink_NHASH;
25721+}
25722+
1facf9fc 25723+struct au_branch;
25724+struct au_sbinfo {
25725+ /* nowait tasks in the system-wide workqueue */
25726+ struct au_nowait_tasks si_nowait;
25727+
b752ccd1
AM
25728+ /*
25729+ * tried sb->s_umount, but failed due to the dependecy between i_mutex.
25730+ * rwsem for au_sbinfo is necessary.
25731+ */
dece6358 25732+ struct au_rwsem si_rwsem;
1facf9fc 25733+
b752ccd1
AM
25734+ /* prevent recursive locking in deleting inode */
25735+ struct {
25736+ unsigned long *bitmap;
25737+ spinlock_t tree_lock;
25738+ struct radix_tree_root tree;
25739+ } au_si_pid;
25740+
7f207e10 25741+ /*
523b37e3
AM
25742+ * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
25743+ * remount.
7f207e10
AM
25744+ */
25745+ atomic_long_t si_ninodes, si_nfiles;
25746+
1facf9fc 25747+ /* branch management */
25748+ unsigned int si_generation;
25749+
25750+ /* see above flags */
25751+ unsigned char au_si_status;
25752+
25753+ aufs_bindex_t si_bend;
7f207e10
AM
25754+
25755+ /* dirty trick to keep br_id plus */
25756+ unsigned int si_last_br_id :
25757+ sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
1facf9fc 25758+ struct au_branch **si_branch;
25759+
25760+ /* policy to select a writable branch */
25761+ unsigned char si_wbr_copyup;
25762+ unsigned char si_wbr_create;
25763+ struct au_wbr_copyup_operations *si_wbr_copyup_ops;
25764+ struct au_wbr_create_operations *si_wbr_create_ops;
25765+
25766+ /* round robin */
25767+ atomic_t si_wbr_rr_next;
25768+
25769+ /* most free space */
25770+ struct au_wbr_mfs si_wbr_mfs;
25771+
25772+ /* mount flags */
25773+ /* include/asm-ia64/siginfo.h defines a macro named si_flags */
25774+ unsigned int si_mntflags;
25775+
25776+ /* external inode number (bitmap and translation table) */
25777+ au_readf_t si_xread;
25778+ au_writef_t si_xwrite;
25779+ struct file *si_xib;
25780+ struct mutex si_xib_mtx; /* protect xib members */
25781+ unsigned long *si_xib_buf;
25782+ unsigned long si_xib_last_pindex;
25783+ int si_xib_next_bit;
25784+ aufs_bindex_t si_xino_brid;
392086de
AM
25785+ unsigned long si_xino_jiffy;
25786+ unsigned long si_xino_expire;
1facf9fc 25787+ /* reserved for future use */
25788+ /* unsigned long long si_xib_limit; */ /* Max xib file size */
25789+
25790+#ifdef CONFIG_AUFS_EXPORT
25791+ /* i_generation */
25792+ struct file *si_xigen;
25793+ atomic_t si_xigen_next;
25794+#endif
25795+
25796+ /* vdir parameters */
e49829fe 25797+ unsigned long si_rdcache; /* max cache time in jiffies */
1facf9fc 25798+ unsigned int si_rdblk; /* deblk size */
25799+ unsigned int si_rdhash; /* hash size */
25800+
25801+ /*
25802+ * If the number of whiteouts are larger than si_dirwh, leave all of
25803+ * them after au_whtmp_ren to reduce the cost of rmdir(2).
25804+ * future fsck.aufs or kernel thread will remove them later.
25805+ * Otherwise, remove all whiteouts and the dir in rmdir(2).
25806+ */
25807+ unsigned int si_dirwh;
25808+
25809+ /*
25810+ * rename(2) a directory with all children.
25811+ */
25812+ /* reserved for future use */
25813+ /* int si_rendir; */
25814+
25815+ /* pseudo_link list */
86dc4139 25816+ struct au_sphlhead si_plink[AuPlink_NHASH];
1facf9fc 25817+ wait_queue_head_t si_plink_wq;
4a4d8108 25818+ spinlock_t si_plink_maint_lock;
e49829fe 25819+ pid_t si_plink_maint_pid;
1facf9fc 25820+
523b37e3
AM
25821+ /* file list */
25822+ struct au_sphlhead si_files;
25823+
1facf9fc 25824+ /*
25825+ * sysfs and lifetime management.
25826+ * this is not a small structure and it may be a waste of memory in case
25827+ * of sysfs is disabled, particulary when many aufs-es are mounted.
25828+ * but using sysfs is majority.
25829+ */
25830+ struct kobject si_kobj;
25831+#ifdef CONFIG_DEBUG_FS
86dc4139
AM
25832+ struct dentry *si_dbgaufs;
25833+ struct dentry *si_dbgaufs_plink;
25834+ struct dentry *si_dbgaufs_xib;
1facf9fc 25835+#ifdef CONFIG_AUFS_EXPORT
25836+ struct dentry *si_dbgaufs_xigen;
25837+#endif
25838+#endif
25839+
e49829fe
JR
25840+#ifdef CONFIG_AUFS_SBILIST
25841+ struct list_head si_list;
25842+#endif
25843+
1facf9fc 25844+ /* dirty, necessary for unmounting, sysfs and sysrq */
25845+ struct super_block *si_sb;
25846+};
25847+
dece6358
AM
25848+/* sbinfo status flags */
25849+/*
25850+ * set true when refresh_dirs() failed at remount time.
25851+ * then try refreshing dirs at access time again.
25852+ * if it is false, refreshing dirs at access time is unnecesary
25853+ */
027c5e7a 25854+#define AuSi_FAILED_REFRESH_DIR 1
dece6358
AM
25855+static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
25856+ unsigned int flag)
25857+{
25858+ AuRwMustAnyLock(&sbi->si_rwsem);
25859+ return sbi->au_si_status & flag;
25860+}
25861+#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
25862+#define au_fset_si(sbinfo, name) do { \
25863+ AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
25864+ (sbinfo)->au_si_status |= AuSi_##name; \
25865+} while (0)
25866+#define au_fclr_si(sbinfo, name) do { \
25867+ AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
25868+ (sbinfo)->au_si_status &= ~AuSi_##name; \
25869+} while (0)
25870+
1facf9fc 25871+/* ---------------------------------------------------------------------- */
25872+
25873+/* policy to select one among writable branches */
4a4d8108
AM
25874+#define AuWbrCopyup(sbinfo, ...) \
25875+ ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
25876+#define AuWbrCreate(sbinfo, ...) \
25877+ ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
1facf9fc 25878+
25879+/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
25880+#define AuLock_DW 1 /* write-lock dentry */
25881+#define AuLock_IR (1 << 1) /* read-lock inode */
25882+#define AuLock_IW (1 << 2) /* write-lock inode */
25883+#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
25884+#define AuLock_DIR (1 << 4) /* target is a dir */
e49829fe
JR
25885+#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
25886+#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
027c5e7a 25887+#define AuLock_GEN (1 << 7) /* test digen/iigen */
1facf9fc 25888+#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
7f207e10
AM
25889+#define au_fset_lock(flags, name) \
25890+ do { (flags) |= AuLock_##name; } while (0)
25891+#define au_fclr_lock(flags, name) \
25892+ do { (flags) &= ~AuLock_##name; } while (0)
1facf9fc 25893+
25894+/* ---------------------------------------------------------------------- */
25895+
25896+/* super.c */
25897+extern struct file_system_type aufs_fs_type;
25898+struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
7f207e10
AM
25899+typedef unsigned long long (*au_arraycb_t)(void *array, unsigned long long max,
25900+ void *arg);
25901+void au_array_free(void *array);
25902+void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb, void *arg);
25903+struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
25904+void au_iarray_free(struct inode **a, unsigned long long max);
1facf9fc 25905+
25906+/* sbinfo.c */
25907+void au_si_free(struct kobject *kobj);
25908+int au_si_alloc(struct super_block *sb);
25909+int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr);
25910+
25911+unsigned int au_sigen_inc(struct super_block *sb);
25912+aufs_bindex_t au_new_br_id(struct super_block *sb);
25913+
e49829fe
JR
25914+int si_read_lock(struct super_block *sb, int flags);
25915+int si_write_lock(struct super_block *sb, int flags);
25916+int aufs_read_lock(struct dentry *dentry, int flags);
1facf9fc 25917+void aufs_read_unlock(struct dentry *dentry, int flags);
25918+void aufs_write_lock(struct dentry *dentry);
25919+void aufs_write_unlock(struct dentry *dentry);
e49829fe 25920+int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
1facf9fc 25921+void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
25922+
b752ccd1
AM
25923+int si_pid_test_slow(struct super_block *sb);
25924+void si_pid_set_slow(struct super_block *sb);
25925+void si_pid_clr_slow(struct super_block *sb);
25926+
1facf9fc 25927+/* wbr_policy.c */
25928+extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
25929+extern struct au_wbr_create_operations au_wbr_create_ops[];
25930+int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
c2b27bf2
AM
25931+int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
25932+
25933+/* mvdown.c */
25934+int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
1facf9fc 25935+
25936+/* ---------------------------------------------------------------------- */
25937+
25938+static inline struct au_sbinfo *au_sbi(struct super_block *sb)
25939+{
25940+ return sb->s_fs_info;
25941+}
25942+
25943+/* ---------------------------------------------------------------------- */
25944+
25945+#ifdef CONFIG_AUFS_EXPORT
a2a7ad62 25946+int au_test_nfsd(void);
1facf9fc 25947+void au_export_init(struct super_block *sb);
b752ccd1 25948+void au_xigen_inc(struct inode *inode);
1facf9fc 25949+int au_xigen_new(struct inode *inode);
25950+int au_xigen_set(struct super_block *sb, struct file *base);
25951+void au_xigen_clr(struct super_block *sb);
25952+
25953+static inline int au_busy_or_stale(void)
25954+{
b752ccd1 25955+ if (!au_test_nfsd())
1facf9fc 25956+ return -EBUSY;
25957+ return -ESTALE;
25958+}
25959+#else
b752ccd1 25960+AuStubInt0(au_test_nfsd, void)
a2a7ad62 25961+AuStubVoid(au_export_init, struct super_block *sb)
b752ccd1 25962+AuStubVoid(au_xigen_inc, struct inode *inode)
4a4d8108
AM
25963+AuStubInt0(au_xigen_new, struct inode *inode)
25964+AuStubInt0(au_xigen_set, struct super_block *sb, struct file *base)
25965+AuStubVoid(au_xigen_clr, struct super_block *sb)
1facf9fc 25966+static inline int au_busy_or_stale(void)
25967+{
25968+ return -EBUSY;
25969+}
25970+#endif /* CONFIG_AUFS_EXPORT */
25971+
25972+/* ---------------------------------------------------------------------- */
25973+
e49829fe
JR
25974+#ifdef CONFIG_AUFS_SBILIST
25975+/* module.c */
25976+extern struct au_splhead au_sbilist;
25977+
25978+static inline void au_sbilist_init(void)
25979+{
25980+ au_spl_init(&au_sbilist);
25981+}
25982+
25983+static inline void au_sbilist_add(struct super_block *sb)
25984+{
25985+ au_spl_add(&au_sbi(sb)->si_list, &au_sbilist);
25986+}
25987+
25988+static inline void au_sbilist_del(struct super_block *sb)
25989+{
25990+ au_spl_del(&au_sbi(sb)->si_list, &au_sbilist);
25991+}
53392da6
AM
25992+
25993+#ifdef CONFIG_AUFS_MAGIC_SYSRQ
25994+static inline void au_sbilist_lock(void)
25995+{
25996+ spin_lock(&au_sbilist.spin);
25997+}
25998+
25999+static inline void au_sbilist_unlock(void)
26000+{
26001+ spin_unlock(&au_sbilist.spin);
26002+}
26003+#define AuGFP_SBILIST GFP_ATOMIC
26004+#else
26005+AuStubVoid(au_sbilist_lock, void)
26006+AuStubVoid(au_sbilist_unlock, void)
26007+#define AuGFP_SBILIST GFP_NOFS
26008+#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
e49829fe
JR
26009+#else
26010+AuStubVoid(au_sbilist_init, void)
26011+AuStubVoid(au_sbilist_add, struct super_block*)
26012+AuStubVoid(au_sbilist_del, struct super_block*)
53392da6
AM
26013+AuStubVoid(au_sbilist_lock, void)
26014+AuStubVoid(au_sbilist_unlock, void)
26015+#define AuGFP_SBILIST GFP_NOFS
e49829fe
JR
26016+#endif
26017+
26018+/* ---------------------------------------------------------------------- */
26019+
1facf9fc 26020+static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
26021+{
dece6358
AM
26022+ /*
26023+ * This function is a dynamic '__init' fucntion actually,
26024+ * so the tiny check for si_rwsem is unnecessary.
26025+ */
26026+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
1facf9fc 26027+#ifdef CONFIG_DEBUG_FS
26028+ sbinfo->si_dbgaufs = NULL;
86dc4139 26029+ sbinfo->si_dbgaufs_plink = NULL;
1facf9fc 26030+ sbinfo->si_dbgaufs_xib = NULL;
26031+#ifdef CONFIG_AUFS_EXPORT
26032+ sbinfo->si_dbgaufs_xigen = NULL;
26033+#endif
26034+#endif
26035+}
26036+
26037+/* ---------------------------------------------------------------------- */
26038+
b752ccd1
AM
26039+static inline pid_t si_pid_bit(void)
26040+{
26041+ /* the origin of pid is 1, but the bitmap's is 0 */
26042+ return current->pid - 1;
26043+}
26044+
26045+static inline int si_pid_test(struct super_block *sb)
26046+{
26047+ pid_t bit = si_pid_bit();
26048+ if (bit < PID_MAX_DEFAULT)
26049+ return test_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
26050+ else
26051+ return si_pid_test_slow(sb);
26052+}
26053+
26054+static inline void si_pid_set(struct super_block *sb)
26055+{
26056+ pid_t bit = si_pid_bit();
26057+ if (bit < PID_MAX_DEFAULT) {
26058+ AuDebugOn(test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
26059+ set_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
26060+ /* smp_mb(); */
26061+ } else
26062+ si_pid_set_slow(sb);
26063+}
26064+
26065+static inline void si_pid_clr(struct super_block *sb)
26066+{
26067+ pid_t bit = si_pid_bit();
26068+ if (bit < PID_MAX_DEFAULT) {
26069+ AuDebugOn(!test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
26070+ clear_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
26071+ /* smp_mb(); */
26072+ } else
26073+ si_pid_clr_slow(sb);
26074+}
26075+
26076+/* ---------------------------------------------------------------------- */
26077+
1facf9fc 26078+/* lock superblock. mainly for entry point functions */
26079+/*
b752ccd1
AM
26080+ * __si_read_lock, __si_write_lock,
26081+ * __si_read_unlock, __si_write_unlock, __si_downgrade_lock
1facf9fc 26082+ */
b752ccd1 26083+AuSimpleRwsemFuncs(__si, struct super_block *sb, &au_sbi(sb)->si_rwsem);
1facf9fc 26084+
dece6358
AM
26085+#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
26086+#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
26087+#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
26088+
b752ccd1
AM
26089+static inline void si_noflush_read_lock(struct super_block *sb)
26090+{
26091+ __si_read_lock(sb);
26092+ si_pid_set(sb);
26093+}
26094+
26095+static inline int si_noflush_read_trylock(struct super_block *sb)
26096+{
26097+ int locked = __si_read_trylock(sb);
26098+ if (locked)
26099+ si_pid_set(sb);
26100+ return locked;
26101+}
26102+
26103+static inline void si_noflush_write_lock(struct super_block *sb)
26104+{
26105+ __si_write_lock(sb);
26106+ si_pid_set(sb);
26107+}
26108+
26109+static inline int si_noflush_write_trylock(struct super_block *sb)
26110+{
26111+ int locked = __si_write_trylock(sb);
26112+ if (locked)
26113+ si_pid_set(sb);
26114+ return locked;
26115+}
26116+
e49829fe 26117+#if 0 /* unused */
1facf9fc 26118+static inline int si_read_trylock(struct super_block *sb, int flags)
26119+{
26120+ if (au_ftest_lock(flags, FLUSH))
26121+ au_nwt_flush(&au_sbi(sb)->si_nowait);
26122+ return si_noflush_read_trylock(sb);
26123+}
e49829fe 26124+#endif
1facf9fc 26125+
b752ccd1
AM
26126+static inline void si_read_unlock(struct super_block *sb)
26127+{
26128+ si_pid_clr(sb);
26129+ __si_read_unlock(sb);
26130+}
26131+
b752ccd1 26132+#if 0 /* unused */
1facf9fc 26133+static inline int si_write_trylock(struct super_block *sb, int flags)
26134+{
26135+ if (au_ftest_lock(flags, FLUSH))
26136+ au_nwt_flush(&au_sbi(sb)->si_nowait);
26137+ return si_noflush_write_trylock(sb);
26138+}
b752ccd1
AM
26139+#endif
26140+
26141+static inline void si_write_unlock(struct super_block *sb)
26142+{
26143+ si_pid_clr(sb);
26144+ __si_write_unlock(sb);
26145+}
26146+
26147+#if 0 /* unused */
26148+static inline void si_downgrade_lock(struct super_block *sb)
26149+{
26150+ __si_downgrade_lock(sb);
26151+}
26152+#endif
1facf9fc 26153+
26154+/* ---------------------------------------------------------------------- */
26155+
26156+static inline aufs_bindex_t au_sbend(struct super_block *sb)
26157+{
dece6358 26158+ SiMustAnyLock(sb);
1facf9fc 26159+ return au_sbi(sb)->si_bend;
26160+}
26161+
26162+static inline unsigned int au_mntflags(struct super_block *sb)
26163+{
dece6358 26164+ SiMustAnyLock(sb);
1facf9fc 26165+ return au_sbi(sb)->si_mntflags;
26166+}
26167+
26168+static inline unsigned int au_sigen(struct super_block *sb)
26169+{
dece6358 26170+ SiMustAnyLock(sb);
1facf9fc 26171+ return au_sbi(sb)->si_generation;
26172+}
26173+
7f207e10
AM
26174+static inline void au_ninodes_inc(struct super_block *sb)
26175+{
26176+ atomic_long_inc(&au_sbi(sb)->si_ninodes);
26177+}
26178+
26179+static inline void au_ninodes_dec(struct super_block *sb)
26180+{
26181+ AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_ninodes));
26182+ atomic_long_dec(&au_sbi(sb)->si_ninodes);
26183+}
26184+
26185+static inline void au_nfiles_inc(struct super_block *sb)
26186+{
26187+ atomic_long_inc(&au_sbi(sb)->si_nfiles);
26188+}
26189+
26190+static inline void au_nfiles_dec(struct super_block *sb)
26191+{
26192+ AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_nfiles));
26193+ atomic_long_dec(&au_sbi(sb)->si_nfiles);
26194+}
26195+
1facf9fc 26196+static inline struct au_branch *au_sbr(struct super_block *sb,
26197+ aufs_bindex_t bindex)
26198+{
dece6358 26199+ SiMustAnyLock(sb);
1facf9fc 26200+ return au_sbi(sb)->si_branch[0 + bindex];
26201+}
26202+
26203+static inline void au_xino_brid_set(struct super_block *sb, aufs_bindex_t brid)
26204+{
dece6358 26205+ SiMustWriteLock(sb);
1facf9fc 26206+ au_sbi(sb)->si_xino_brid = brid;
26207+}
26208+
26209+static inline aufs_bindex_t au_xino_brid(struct super_block *sb)
26210+{
dece6358 26211+ SiMustAnyLock(sb);
1facf9fc 26212+ return au_sbi(sb)->si_xino_brid;
26213+}
26214+
26215+#endif /* __KERNEL__ */
26216+#endif /* __AUFS_SUPER_H__ */
7f207e10
AM
26217diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
26218--- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 26219+++ linux/fs/aufs/sysaufs.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 26220@@ -0,0 +1,104 @@
1facf9fc 26221+/*
523b37e3 26222+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 26223+ *
26224+ * This program, aufs is free software; you can redistribute it and/or modify
26225+ * it under the terms of the GNU General Public License as published by
26226+ * the Free Software Foundation; either version 2 of the License, or
26227+ * (at your option) any later version.
dece6358
AM
26228+ *
26229+ * This program is distributed in the hope that it will be useful,
26230+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26231+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26232+ * GNU General Public License for more details.
26233+ *
26234+ * You should have received a copy of the GNU General Public License
523b37e3 26235+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26236+ */
26237+
26238+/*
26239+ * sysfs interface and lifetime management
26240+ * they are necessary regardless sysfs is disabled.
26241+ */
26242+
1facf9fc 26243+#include <linux/random.h>
1facf9fc 26244+#include "aufs.h"
26245+
26246+unsigned long sysaufs_si_mask;
e49829fe 26247+struct kset *sysaufs_kset;
1facf9fc 26248+
26249+#define AuSiAttr(_name) { \
26250+ .attr = { .name = __stringify(_name), .mode = 0444 }, \
26251+ .show = sysaufs_si_##_name, \
26252+}
26253+
26254+static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
26255+struct attribute *sysaufs_si_attrs[] = {
26256+ &sysaufs_si_attr_xi_path.attr,
26257+ NULL,
26258+};
26259+
4a4d8108 26260+static const struct sysfs_ops au_sbi_ops = {
1facf9fc 26261+ .show = sysaufs_si_show
26262+};
26263+
26264+static struct kobj_type au_sbi_ktype = {
26265+ .release = au_si_free,
26266+ .sysfs_ops = &au_sbi_ops,
26267+ .default_attrs = sysaufs_si_attrs
26268+};
26269+
26270+/* ---------------------------------------------------------------------- */
26271+
26272+int sysaufs_si_init(struct au_sbinfo *sbinfo)
26273+{
26274+ int err;
26275+
e49829fe 26276+ sbinfo->si_kobj.kset = sysaufs_kset;
1facf9fc 26277+ /* cf. sysaufs_name() */
26278+ err = kobject_init_and_add
e49829fe 26279+ (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
1facf9fc 26280+ SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
26281+
26282+ dbgaufs_si_null(sbinfo);
26283+ if (!err) {
26284+ err = dbgaufs_si_init(sbinfo);
26285+ if (unlikely(err))
26286+ kobject_put(&sbinfo->si_kobj);
26287+ }
26288+ return err;
26289+}
26290+
26291+void sysaufs_fin(void)
26292+{
26293+ dbgaufs_fin();
e49829fe
JR
26294+ sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
26295+ kset_unregister(sysaufs_kset);
1facf9fc 26296+}
26297+
26298+int __init sysaufs_init(void)
26299+{
26300+ int err;
26301+
26302+ do {
26303+ get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
26304+ } while (!sysaufs_si_mask);
26305+
4a4d8108 26306+ err = -EINVAL;
e49829fe
JR
26307+ sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
26308+ if (unlikely(!sysaufs_kset))
4a4d8108 26309+ goto out;
e49829fe
JR
26310+ err = PTR_ERR(sysaufs_kset);
26311+ if (IS_ERR(sysaufs_kset))
1facf9fc 26312+ goto out;
e49829fe 26313+ err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
1facf9fc 26314+ if (unlikely(err)) {
e49829fe 26315+ kset_unregister(sysaufs_kset);
1facf9fc 26316+ goto out;
26317+ }
26318+
26319+ err = dbgaufs_init();
26320+ if (unlikely(err))
26321+ sysaufs_fin();
4f0767ce 26322+out:
1facf9fc 26323+ return err;
26324+}
7f207e10
AM
26325diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
26326--- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 26327+++ linux/fs/aufs/sysaufs.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 26328@@ -0,0 +1,103 @@
1facf9fc 26329+/*
523b37e3 26330+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 26331+ *
26332+ * This program, aufs is free software; you can redistribute it and/or modify
26333+ * it under the terms of the GNU General Public License as published by
26334+ * the Free Software Foundation; either version 2 of the License, or
26335+ * (at your option) any later version.
dece6358
AM
26336+ *
26337+ * This program is distributed in the hope that it will be useful,
26338+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26339+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26340+ * GNU General Public License for more details.
26341+ *
26342+ * You should have received a copy of the GNU General Public License
523b37e3 26343+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26344+ */
26345+
26346+/*
26347+ * sysfs interface and mount lifetime management
26348+ */
26349+
26350+#ifndef __SYSAUFS_H__
26351+#define __SYSAUFS_H__
26352+
26353+#ifdef __KERNEL__
26354+
1facf9fc 26355+#include <linux/sysfs.h>
1facf9fc 26356+#include "module.h"
26357+
dece6358
AM
26358+struct super_block;
26359+struct au_sbinfo;
26360+
1facf9fc 26361+struct sysaufs_si_attr {
26362+ struct attribute attr;
26363+ int (*show)(struct seq_file *seq, struct super_block *sb);
26364+};
26365+
26366+/* ---------------------------------------------------------------------- */
26367+
26368+/* sysaufs.c */
26369+extern unsigned long sysaufs_si_mask;
e49829fe 26370+extern struct kset *sysaufs_kset;
1facf9fc 26371+extern struct attribute *sysaufs_si_attrs[];
26372+int sysaufs_si_init(struct au_sbinfo *sbinfo);
26373+int __init sysaufs_init(void);
26374+void sysaufs_fin(void);
26375+
26376+/* ---------------------------------------------------------------------- */
26377+
26378+/* some people doesn't like to show a pointer in kernel */
26379+static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
26380+{
26381+ return sysaufs_si_mask ^ (unsigned long)sbinfo;
26382+}
26383+
26384+#define SysaufsSiNamePrefix "si_"
26385+#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
26386+static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
26387+{
26388+ snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
26389+ sysaufs_si_id(sbinfo));
26390+}
26391+
26392+struct au_branch;
26393+#ifdef CONFIG_SYSFS
26394+/* sysfs.c */
26395+extern struct attribute_group *sysaufs_attr_group;
26396+
26397+int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
26398+ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
26399+ char *buf);
26400+
26401+void sysaufs_br_init(struct au_branch *br);
26402+void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
26403+void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
26404+
26405+#define sysaufs_brs_init() do {} while (0)
26406+
26407+#else
26408+#define sysaufs_attr_group NULL
26409+
4a4d8108 26410+AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
1facf9fc 26411+
26412+static inline
26413+ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
26414+ char *buf)
26415+{
26416+ return 0;
26417+}
26418+
4a4d8108
AM
26419+AuStubVoid(sysaufs_br_init, struct au_branch *br)
26420+AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
26421+AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
1facf9fc 26422+
26423+static inline void sysaufs_brs_init(void)
26424+{
26425+ sysaufs_brs = 0;
26426+}
26427+
26428+#endif /* CONFIG_SYSFS */
26429+
26430+#endif /* __KERNEL__ */
26431+#endif /* __SYSAUFS_H__ */
7f207e10
AM
26432diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
26433--- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 26434+++ linux/fs/aufs/sysfs.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 26435@@ -0,0 +1,296 @@
1facf9fc 26436+/*
523b37e3 26437+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 26438+ *
26439+ * This program, aufs is free software; you can redistribute it and/or modify
26440+ * it under the terms of the GNU General Public License as published by
26441+ * the Free Software Foundation; either version 2 of the License, or
26442+ * (at your option) any later version.
dece6358
AM
26443+ *
26444+ * This program is distributed in the hope that it will be useful,
26445+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26446+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26447+ * GNU General Public License for more details.
26448+ *
26449+ * You should have received a copy of the GNU General Public License
523b37e3 26450+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26451+ */
26452+
26453+/*
26454+ * sysfs interface
26455+ */
26456+
1facf9fc 26457+#include <linux/seq_file.h>
1facf9fc 26458+#include "aufs.h"
26459+
4a4d8108
AM
26460+#ifdef CONFIG_AUFS_FS_MODULE
26461+/* this entry violates the "one line per file" policy of sysfs */
26462+static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
26463+ char *buf)
26464+{
26465+ ssize_t err;
26466+ static char *conf =
26467+/* this file is generated at compiling */
26468+#include "conf.str"
26469+ ;
26470+
26471+ err = snprintf(buf, PAGE_SIZE, conf);
26472+ if (unlikely(err >= PAGE_SIZE))
26473+ err = -EFBIG;
26474+ return err;
26475+}
26476+
26477+static struct kobj_attribute au_config_attr = __ATTR_RO(config);
26478+#endif
26479+
1facf9fc 26480+static struct attribute *au_attr[] = {
4a4d8108
AM
26481+#ifdef CONFIG_AUFS_FS_MODULE
26482+ &au_config_attr.attr,
26483+#endif
1facf9fc 26484+ NULL, /* need to NULL terminate the list of attributes */
26485+};
26486+
26487+static struct attribute_group sysaufs_attr_group_body = {
26488+ .attrs = au_attr
26489+};
26490+
26491+struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
26492+
26493+/* ---------------------------------------------------------------------- */
26494+
26495+int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
26496+{
26497+ int err;
26498+
dece6358
AM
26499+ SiMustAnyLock(sb);
26500+
1facf9fc 26501+ err = 0;
26502+ if (au_opt_test(au_mntflags(sb), XINO)) {
26503+ err = au_xino_path(seq, au_sbi(sb)->si_xib);
26504+ seq_putc(seq, '\n');
26505+ }
26506+ return err;
26507+}
26508+
26509+/*
26510+ * the lifetime of branch is independent from the entry under sysfs.
26511+ * sysfs handles the lifetime of the entry, and never call ->show() after it is
26512+ * unlinked.
26513+ */
26514+static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
392086de 26515+ aufs_bindex_t bindex, int idx)
1facf9fc 26516+{
1e00d052 26517+ int err;
1facf9fc 26518+ struct path path;
26519+ struct dentry *root;
26520+ struct au_branch *br;
1e00d052 26521+ char *perm;
1facf9fc 26522+
26523+ AuDbg("b%d\n", bindex);
26524+
1e00d052 26525+ err = 0;
1facf9fc 26526+ root = sb->s_root;
26527+ di_read_lock_parent(root, !AuLock_IR);
26528+ br = au_sbr(sb, bindex);
392086de
AM
26529+
26530+ switch (idx) {
26531+ case AuBrSysfs_BR:
26532+ path.mnt = au_br_mnt(br);
26533+ path.dentry = au_h_dptr(root, bindex);
26534+ au_seq_path(seq, &path);
26535+ di_read_unlock(root, !AuLock_IR);
26536+ perm = au_optstr_br_perm(br->br_perm);
26537+ if (perm) {
26538+ err = seq_printf(seq, "=%s\n", perm);
26539+ kfree(perm);
26540+ if (err == -1)
26541+ err = -E2BIG;
26542+ } else
26543+ err = -ENOMEM;
26544+ break;
26545+ case AuBrSysfs_BRID:
26546+ err = seq_printf(seq, "%d\n", br->br_id);
26547+ di_read_unlock(root, !AuLock_IR);
1e00d052
AM
26548+ if (err == -1)
26549+ err = -E2BIG;
392086de
AM
26550+ break;
26551+ }
26552+
1e00d052 26553+ return err;
1facf9fc 26554+}
26555+
26556+/* ---------------------------------------------------------------------- */
26557+
26558+static struct seq_file *au_seq(char *p, ssize_t len)
26559+{
26560+ struct seq_file *seq;
26561+
26562+ seq = kzalloc(sizeof(*seq), GFP_NOFS);
26563+ if (seq) {
26564+ /* mutex_init(&seq.lock); */
26565+ seq->buf = p;
26566+ seq->size = len;
26567+ return seq; /* success */
26568+ }
26569+
26570+ seq = ERR_PTR(-ENOMEM);
26571+ return seq;
26572+}
26573+
392086de
AM
26574+#define SysaufsBr_PREFIX "br"
26575+#define SysaufsBrid_PREFIX "brid"
1facf9fc 26576+
26577+/* todo: file size may exceed PAGE_SIZE */
26578+ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
1308ab2a 26579+ char *buf)
1facf9fc 26580+{
26581+ ssize_t err;
392086de 26582+ int idx;
1facf9fc 26583+ long l;
26584+ aufs_bindex_t bend;
26585+ struct au_sbinfo *sbinfo;
26586+ struct super_block *sb;
26587+ struct seq_file *seq;
26588+ char *name;
26589+ struct attribute **cattr;
26590+
26591+ sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
26592+ sb = sbinfo->si_sb;
1308ab2a 26593+
26594+ /*
26595+ * prevent a race condition between sysfs and aufs.
26596+ * for instance, sysfs_file_read() calls sysfs_get_active_two() which
26597+ * prohibits maintaining the sysfs entries.
26598+ * hew we acquire read lock after sysfs_get_active_two().
26599+ * on the other hand, the remount process may maintain the sysfs/aufs
26600+ * entries after acquiring write lock.
26601+ * it can cause a deadlock.
26602+ * simply we gave up processing read here.
26603+ */
26604+ err = -EBUSY;
26605+ if (unlikely(!si_noflush_read_trylock(sb)))
26606+ goto out;
1facf9fc 26607+
26608+ seq = au_seq(buf, PAGE_SIZE);
26609+ err = PTR_ERR(seq);
26610+ if (IS_ERR(seq))
1308ab2a 26611+ goto out_unlock;
1facf9fc 26612+
26613+ name = (void *)attr->name;
26614+ cattr = sysaufs_si_attrs;
26615+ while (*cattr) {
26616+ if (!strcmp(name, (*cattr)->name)) {
26617+ err = container_of(*cattr, struct sysaufs_si_attr, attr)
26618+ ->show(seq, sb);
26619+ goto out_seq;
26620+ }
26621+ cattr++;
26622+ }
26623+
392086de
AM
26624+ if (!strncmp(name, SysaufsBrid_PREFIX,
26625+ sizeof(SysaufsBrid_PREFIX) - 1)) {
26626+ idx = AuBrSysfs_BRID;
26627+ name += sizeof(SysaufsBrid_PREFIX) - 1;
26628+ } else if (!strncmp(name, SysaufsBr_PREFIX,
26629+ sizeof(SysaufsBr_PREFIX) - 1)) {
26630+ idx = AuBrSysfs_BR;
1facf9fc 26631+ name += sizeof(SysaufsBr_PREFIX) - 1;
392086de
AM
26632+ } else
26633+ BUG();
26634+
26635+ err = kstrtol(name, 10, &l);
26636+ if (!err) {
26637+ bend = au_sbend(sb);
26638+ if (l <= bend)
26639+ err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
26640+ else
26641+ err = -ENOENT;
1facf9fc 26642+ }
1facf9fc 26643+
4f0767ce 26644+out_seq:
1facf9fc 26645+ if (!err) {
26646+ err = seq->count;
26647+ /* sysfs limit */
26648+ if (unlikely(err == PAGE_SIZE))
26649+ err = -EFBIG;
26650+ }
26651+ kfree(seq);
4f0767ce 26652+out_unlock:
1facf9fc 26653+ si_read_unlock(sb);
4f0767ce 26654+out:
1facf9fc 26655+ return err;
26656+}
26657+
26658+/* ---------------------------------------------------------------------- */
26659+
26660+void sysaufs_br_init(struct au_branch *br)
26661+{
392086de
AM
26662+ int i;
26663+ struct au_brsysfs *br_sysfs;
26664+ struct attribute *attr;
4a4d8108 26665+
392086de
AM
26666+ br_sysfs = br->br_sysfs;
26667+ for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
26668+ attr = &br_sysfs->attr;
26669+ sysfs_attr_init(attr);
26670+ attr->name = br_sysfs->name;
26671+ attr->mode = S_IRUGO;
26672+ br_sysfs++;
26673+ }
1facf9fc 26674+}
26675+
26676+void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
26677+{
26678+ struct au_branch *br;
26679+ struct kobject *kobj;
392086de
AM
26680+ struct au_brsysfs *br_sysfs;
26681+ int i;
1facf9fc 26682+ aufs_bindex_t bend;
26683+
26684+ dbgaufs_brs_del(sb, bindex);
26685+
26686+ if (!sysaufs_brs)
26687+ return;
26688+
26689+ kobj = &au_sbi(sb)->si_kobj;
26690+ bend = au_sbend(sb);
26691+ for (; bindex <= bend; bindex++) {
26692+ br = au_sbr(sb, bindex);
392086de
AM
26693+ br_sysfs = br->br_sysfs;
26694+ for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
26695+ sysfs_remove_file(kobj, &br_sysfs->attr);
26696+ br_sysfs++;
26697+ }
1facf9fc 26698+ }
26699+}
26700+
26701+void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
26702+{
392086de 26703+ int err, i;
1facf9fc 26704+ aufs_bindex_t bend;
26705+ struct kobject *kobj;
26706+ struct au_branch *br;
392086de 26707+ struct au_brsysfs *br_sysfs;
1facf9fc 26708+
26709+ dbgaufs_brs_add(sb, bindex);
26710+
26711+ if (!sysaufs_brs)
26712+ return;
26713+
26714+ kobj = &au_sbi(sb)->si_kobj;
26715+ bend = au_sbend(sb);
26716+ for (; bindex <= bend; bindex++) {
26717+ br = au_sbr(sb, bindex);
392086de
AM
26718+ br_sysfs = br->br_sysfs;
26719+ snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
26720+ SysaufsBr_PREFIX "%d", bindex);
26721+ snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
26722+ SysaufsBrid_PREFIX "%d", bindex);
26723+ for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
26724+ err = sysfs_create_file(kobj, &br_sysfs->attr);
26725+ if (unlikely(err))
26726+ pr_warn("failed %s under sysfs(%d)\n",
26727+ br_sysfs->name, err);
26728+ br_sysfs++;
26729+ }
1facf9fc 26730+ }
26731+}
7f207e10
AM
26732diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
26733--- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 26734+++ linux/fs/aufs/sysrq.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 26735@@ -0,0 +1,154 @@
1facf9fc 26736+/*
523b37e3 26737+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 26738+ *
26739+ * This program, aufs is free software; you can redistribute it and/or modify
26740+ * it under the terms of the GNU General Public License as published by
26741+ * the Free Software Foundation; either version 2 of the License, or
26742+ * (at your option) any later version.
dece6358
AM
26743+ *
26744+ * This program is distributed in the hope that it will be useful,
26745+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26746+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26747+ * GNU General Public License for more details.
26748+ *
26749+ * You should have received a copy of the GNU General Public License
523b37e3 26750+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26751+ */
26752+
26753+/*
26754+ * magic sysrq hanlder
26755+ */
26756+
1facf9fc 26757+/* #include <linux/sysrq.h> */
027c5e7a 26758+#include <linux/writeback.h>
1facf9fc 26759+#include "aufs.h"
26760+
26761+/* ---------------------------------------------------------------------- */
26762+
26763+static void sysrq_sb(struct super_block *sb)
26764+{
26765+ char *plevel;
26766+ struct au_sbinfo *sbinfo;
26767+ struct file *file;
523b37e3
AM
26768+ struct au_sphlhead *files;
26769+ struct au_finfo *finfo;
1facf9fc 26770+
26771+ plevel = au_plevel;
26772+ au_plevel = KERN_WARNING;
1facf9fc 26773+
4a4d8108 26774+ /* since we define pr_fmt, call printk directly */
c06a8ce3
AM
26775+#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
26776+
26777+ sbinfo = au_sbi(sb);
4a4d8108 26778+ printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
c06a8ce3 26779+ pr("superblock\n");
1facf9fc 26780+ au_dpri_sb(sb);
027c5e7a
AM
26781+
26782+#if 0
c06a8ce3 26783+ pr("root dentry\n");
1facf9fc 26784+ au_dpri_dentry(sb->s_root);
c06a8ce3 26785+ pr("root inode\n");
1facf9fc 26786+ au_dpri_inode(sb->s_root->d_inode);
027c5e7a
AM
26787+#endif
26788+
1facf9fc 26789+#if 0
027c5e7a
AM
26790+ do {
26791+ int err, i, j, ndentry;
26792+ struct au_dcsub_pages dpages;
26793+ struct au_dpage *dpage;
26794+
26795+ err = au_dpages_init(&dpages, GFP_ATOMIC);
26796+ if (unlikely(err))
26797+ break;
26798+ err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
26799+ if (!err)
26800+ for (i = 0; i < dpages.ndpage; i++) {
26801+ dpage = dpages.dpages + i;
26802+ ndentry = dpage->ndentry;
26803+ for (j = 0; j < ndentry; j++)
26804+ au_dpri_dentry(dpage->dentries[j]);
26805+ }
26806+ au_dpages_free(&dpages);
26807+ } while (0);
26808+#endif
26809+
26810+#if 1
26811+ {
26812+ struct inode *i;
c06a8ce3 26813+ pr("isolated inode\n");
2cbb1c4b
JR
26814+ spin_lock(&inode_sb_list_lock);
26815+ list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
26816+ spin_lock(&i->i_lock);
b4510431 26817+ if (1 || hlist_empty(&i->i_dentry))
027c5e7a 26818+ au_dpri_inode(i);
2cbb1c4b
JR
26819+ spin_unlock(&i->i_lock);
26820+ }
26821+ spin_unlock(&inode_sb_list_lock);
027c5e7a 26822+ }
1facf9fc 26823+#endif
c06a8ce3 26824+ pr("files\n");
523b37e3
AM
26825+ files = &au_sbi(sb)->si_files;
26826+ spin_lock(&files->spin);
26827+ hlist_for_each_entry(finfo, &files->head, fi_hlist) {
4a4d8108 26828+ umode_t mode;
523b37e3 26829+ file = finfo->fi_file;
c06a8ce3 26830+ mode = file_inode(file)->i_mode;
38d290e6 26831+ if (!special_file(mode))
1facf9fc 26832+ au_dpri_file(file);
523b37e3
AM
26833+ }
26834+ spin_unlock(&files->spin);
c06a8ce3 26835+ pr("done\n");
1facf9fc 26836+
c06a8ce3 26837+#undef pr
1facf9fc 26838+ au_plevel = plevel;
1facf9fc 26839+}
26840+
26841+/* ---------------------------------------------------------------------- */
26842+
26843+/* module parameter */
26844+static char *aufs_sysrq_key = "a";
26845+module_param_named(sysrq, aufs_sysrq_key, charp, S_IRUGO);
26846+MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
26847+
0c5527e5 26848+static void au_sysrq(int key __maybe_unused)
1facf9fc 26849+{
1facf9fc 26850+ struct au_sbinfo *sbinfo;
26851+
027c5e7a 26852+ lockdep_off();
53392da6 26853+ au_sbilist_lock();
e49829fe 26854+ list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
1facf9fc 26855+ sysrq_sb(sbinfo->si_sb);
53392da6 26856+ au_sbilist_unlock();
027c5e7a 26857+ lockdep_on();
1facf9fc 26858+}
26859+
26860+static struct sysrq_key_op au_sysrq_op = {
26861+ .handler = au_sysrq,
26862+ .help_msg = "Aufs",
26863+ .action_msg = "Aufs",
26864+ .enable_mask = SYSRQ_ENABLE_DUMP
26865+};
26866+
26867+/* ---------------------------------------------------------------------- */
26868+
26869+int __init au_sysrq_init(void)
26870+{
26871+ int err;
26872+ char key;
26873+
26874+ err = -1;
26875+ key = *aufs_sysrq_key;
26876+ if ('a' <= key && key <= 'z')
26877+ err = register_sysrq_key(key, &au_sysrq_op);
26878+ if (unlikely(err))
4a4d8108 26879+ pr_err("err %d, sysrq=%c\n", err, key);
1facf9fc 26880+ return err;
26881+}
26882+
26883+void au_sysrq_fin(void)
26884+{
26885+ int err;
26886+ err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
26887+ if (unlikely(err))
4a4d8108 26888+ pr_err("err %d (ignored)\n", err);
1facf9fc 26889+}
7f207e10
AM
26890diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
26891--- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 26892+++ linux/fs/aufs/vdir.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 26893@@ -0,0 +1,887 @@
1facf9fc 26894+/*
523b37e3 26895+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 26896+ *
26897+ * This program, aufs is free software; you can redistribute it and/or modify
26898+ * it under the terms of the GNU General Public License as published by
26899+ * the Free Software Foundation; either version 2 of the License, or
26900+ * (at your option) any later version.
dece6358
AM
26901+ *
26902+ * This program is distributed in the hope that it will be useful,
26903+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26904+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26905+ * GNU General Public License for more details.
26906+ *
26907+ * You should have received a copy of the GNU General Public License
523b37e3 26908+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26909+ */
26910+
26911+/*
26912+ * virtual or vertical directory
26913+ */
26914+
26915+#include "aufs.h"
26916+
dece6358 26917+static unsigned int calc_size(int nlen)
1facf9fc 26918+{
dece6358 26919+ return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
1facf9fc 26920+}
26921+
26922+static int set_deblk_end(union au_vdir_deblk_p *p,
26923+ union au_vdir_deblk_p *deblk_end)
26924+{
26925+ if (calc_size(0) <= deblk_end->deblk - p->deblk) {
26926+ p->de->de_str.len = 0;
26927+ /* smp_mb(); */
26928+ return 0;
26929+ }
26930+ return -1; /* error */
26931+}
26932+
26933+/* returns true or false */
26934+static int is_deblk_end(union au_vdir_deblk_p *p,
26935+ union au_vdir_deblk_p *deblk_end)
26936+{
26937+ if (calc_size(0) <= deblk_end->deblk - p->deblk)
26938+ return !p->de->de_str.len;
26939+ return 1;
26940+}
26941+
26942+static unsigned char *last_deblk(struct au_vdir *vdir)
26943+{
26944+ return vdir->vd_deblk[vdir->vd_nblk - 1];
26945+}
26946+
26947+/* ---------------------------------------------------------------------- */
26948+
1308ab2a 26949+/* estimate the apropriate size for name hash table */
26950+unsigned int au_rdhash_est(loff_t sz)
26951+{
26952+ unsigned int n;
26953+
26954+ n = UINT_MAX;
26955+ sz >>= 10;
26956+ if (sz < n)
26957+ n = sz;
26958+ if (sz < AUFS_RDHASH_DEF)
26959+ n = AUFS_RDHASH_DEF;
4a4d8108 26960+ /* pr_info("n %u\n", n); */
1308ab2a 26961+ return n;
26962+}
26963+
1facf9fc 26964+/*
26965+ * the allocated memory has to be freed by
dece6358 26966+ * au_nhash_wh_free() or au_nhash_de_free().
1facf9fc 26967+ */
dece6358 26968+int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
1facf9fc 26969+{
1facf9fc 26970+ struct hlist_head *head;
dece6358 26971+ unsigned int u;
1facf9fc 26972+
dece6358
AM
26973+ head = kmalloc(sizeof(*nhash->nh_head) * num_hash, gfp);
26974+ if (head) {
26975+ nhash->nh_num = num_hash;
26976+ nhash->nh_head = head;
26977+ for (u = 0; u < num_hash; u++)
1facf9fc 26978+ INIT_HLIST_HEAD(head++);
dece6358 26979+ return 0; /* success */
1facf9fc 26980+ }
1facf9fc 26981+
dece6358 26982+ return -ENOMEM;
1facf9fc 26983+}
26984+
dece6358
AM
26985+static void nhash_count(struct hlist_head *head)
26986+{
26987+#if 0
26988+ unsigned long n;
26989+ struct hlist_node *pos;
26990+
26991+ n = 0;
26992+ hlist_for_each(pos, head)
26993+ n++;
4a4d8108 26994+ pr_info("%lu\n", n);
dece6358
AM
26995+#endif
26996+}
26997+
26998+static void au_nhash_wh_do_free(struct hlist_head *head)
1facf9fc 26999+{
c06a8ce3
AM
27000+ struct au_vdir_wh *pos;
27001+ struct hlist_node *node;
1facf9fc 27002+
c06a8ce3
AM
27003+ hlist_for_each_entry_safe(pos, node, head, wh_hash)
27004+ kfree(pos);
1facf9fc 27005+}
27006+
dece6358 27007+static void au_nhash_de_do_free(struct hlist_head *head)
1facf9fc 27008+{
c06a8ce3
AM
27009+ struct au_vdir_dehstr *pos;
27010+ struct hlist_node *node;
1facf9fc 27011+
c06a8ce3
AM
27012+ hlist_for_each_entry_safe(pos, node, head, hash)
27013+ au_cache_free_vdir_dehstr(pos);
1facf9fc 27014+}
27015+
dece6358
AM
27016+static void au_nhash_do_free(struct au_nhash *nhash,
27017+ void (*free)(struct hlist_head *head))
1facf9fc 27018+{
1308ab2a 27019+ unsigned int n;
1facf9fc 27020+ struct hlist_head *head;
1facf9fc 27021+
dece6358 27022+ n = nhash->nh_num;
1308ab2a 27023+ if (!n)
27024+ return;
27025+
dece6358 27026+ head = nhash->nh_head;
1308ab2a 27027+ while (n-- > 0) {
dece6358
AM
27028+ nhash_count(head);
27029+ free(head++);
1facf9fc 27030+ }
dece6358 27031+ kfree(nhash->nh_head);
1facf9fc 27032+}
27033+
dece6358 27034+void au_nhash_wh_free(struct au_nhash *whlist)
1facf9fc 27035+{
dece6358
AM
27036+ au_nhash_do_free(whlist, au_nhash_wh_do_free);
27037+}
1facf9fc 27038+
dece6358
AM
27039+static void au_nhash_de_free(struct au_nhash *delist)
27040+{
27041+ au_nhash_do_free(delist, au_nhash_de_do_free);
1facf9fc 27042+}
27043+
27044+/* ---------------------------------------------------------------------- */
27045+
27046+int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
27047+ int limit)
27048+{
27049+ int num;
27050+ unsigned int u, n;
27051+ struct hlist_head *head;
c06a8ce3 27052+ struct au_vdir_wh *pos;
1facf9fc 27053+
27054+ num = 0;
27055+ n = whlist->nh_num;
27056+ head = whlist->nh_head;
1308ab2a 27057+ for (u = 0; u < n; u++, head++)
c06a8ce3
AM
27058+ hlist_for_each_entry(pos, head, wh_hash)
27059+ if (pos->wh_bindex == btgt && ++num > limit)
1facf9fc 27060+ return 1;
1facf9fc 27061+ return 0;
27062+}
27063+
27064+static struct hlist_head *au_name_hash(struct au_nhash *nhash,
dece6358 27065+ unsigned char *name,
1facf9fc 27066+ unsigned int len)
27067+{
dece6358
AM
27068+ unsigned int v;
27069+ /* const unsigned int magic_bit = 12; */
27070+
1308ab2a 27071+ AuDebugOn(!nhash->nh_num || !nhash->nh_head);
27072+
dece6358
AM
27073+ v = 0;
27074+ while (len--)
27075+ v += *name++;
27076+ /* v = hash_long(v, magic_bit); */
27077+ v %= nhash->nh_num;
27078+ return nhash->nh_head + v;
27079+}
27080+
27081+static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
27082+ int nlen)
27083+{
27084+ return str->len == nlen && !memcmp(str->name, name, nlen);
1facf9fc 27085+}
27086+
27087+/* returns found or not */
dece6358 27088+int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
1facf9fc 27089+{
27090+ struct hlist_head *head;
c06a8ce3 27091+ struct au_vdir_wh *pos;
1facf9fc 27092+ struct au_vdir_destr *str;
27093+
dece6358 27094+ head = au_name_hash(whlist, name, nlen);
c06a8ce3
AM
27095+ hlist_for_each_entry(pos, head, wh_hash) {
27096+ str = &pos->wh_str;
1facf9fc 27097+ AuDbg("%.*s\n", str->len, str->name);
dece6358
AM
27098+ if (au_nhash_test_name(str, name, nlen))
27099+ return 1;
27100+ }
27101+ return 0;
27102+}
27103+
27104+/* returns found(true) or not */
27105+static int test_known(struct au_nhash *delist, char *name, int nlen)
27106+{
27107+ struct hlist_head *head;
c06a8ce3 27108+ struct au_vdir_dehstr *pos;
dece6358
AM
27109+ struct au_vdir_destr *str;
27110+
27111+ head = au_name_hash(delist, name, nlen);
c06a8ce3
AM
27112+ hlist_for_each_entry(pos, head, hash) {
27113+ str = pos->str;
dece6358
AM
27114+ AuDbg("%.*s\n", str->len, str->name);
27115+ if (au_nhash_test_name(str, name, nlen))
1facf9fc 27116+ return 1;
27117+ }
27118+ return 0;
27119+}
27120+
dece6358
AM
27121+static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
27122+ unsigned char d_type)
27123+{
27124+#ifdef CONFIG_AUFS_SHWH
27125+ wh->wh_ino = ino;
27126+ wh->wh_type = d_type;
27127+#endif
27128+}
27129+
27130+/* ---------------------------------------------------------------------- */
27131+
27132+int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
27133+ unsigned int d_type, aufs_bindex_t bindex,
27134+ unsigned char shwh)
1facf9fc 27135+{
27136+ int err;
27137+ struct au_vdir_destr *str;
27138+ struct au_vdir_wh *wh;
27139+
dece6358 27140+ AuDbg("%.*s\n", nlen, name);
1308ab2a 27141+ AuDebugOn(!whlist->nh_num || !whlist->nh_head);
27142+
1facf9fc 27143+ err = -ENOMEM;
dece6358 27144+ wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
1facf9fc 27145+ if (unlikely(!wh))
27146+ goto out;
27147+
27148+ err = 0;
27149+ wh->wh_bindex = bindex;
dece6358
AM
27150+ if (shwh)
27151+ au_shwh_init_wh(wh, ino, d_type);
1facf9fc 27152+ str = &wh->wh_str;
dece6358
AM
27153+ str->len = nlen;
27154+ memcpy(str->name, name, nlen);
27155+ hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
1facf9fc 27156+ /* smp_mb(); */
27157+
4f0767ce 27158+out:
1facf9fc 27159+ return err;
27160+}
27161+
1facf9fc 27162+static int append_deblk(struct au_vdir *vdir)
27163+{
27164+ int err;
dece6358 27165+ unsigned long ul;
1facf9fc 27166+ const unsigned int deblk_sz = vdir->vd_deblk_sz;
27167+ union au_vdir_deblk_p p, deblk_end;
27168+ unsigned char **o;
27169+
27170+ err = -ENOMEM;
dece6358
AM
27171+ o = krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
27172+ GFP_NOFS);
1facf9fc 27173+ if (unlikely(!o))
27174+ goto out;
27175+
27176+ vdir->vd_deblk = o;
27177+ p.deblk = kmalloc(deblk_sz, GFP_NOFS);
27178+ if (p.deblk) {
27179+ ul = vdir->vd_nblk++;
27180+ vdir->vd_deblk[ul] = p.deblk;
27181+ vdir->vd_last.ul = ul;
27182+ vdir->vd_last.p.deblk = p.deblk;
27183+ deblk_end.deblk = p.deblk + deblk_sz;
27184+ err = set_deblk_end(&p, &deblk_end);
27185+ }
27186+
4f0767ce 27187+out:
1facf9fc 27188+ return err;
27189+}
27190+
dece6358
AM
27191+static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
27192+ unsigned int d_type, struct au_nhash *delist)
27193+{
27194+ int err;
27195+ unsigned int sz;
27196+ const unsigned int deblk_sz = vdir->vd_deblk_sz;
27197+ union au_vdir_deblk_p p, *room, deblk_end;
27198+ struct au_vdir_dehstr *dehstr;
27199+
27200+ p.deblk = last_deblk(vdir);
27201+ deblk_end.deblk = p.deblk + deblk_sz;
27202+ room = &vdir->vd_last.p;
27203+ AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
27204+ || !is_deblk_end(room, &deblk_end));
27205+
27206+ sz = calc_size(nlen);
27207+ if (unlikely(sz > deblk_end.deblk - room->deblk)) {
27208+ err = append_deblk(vdir);
27209+ if (unlikely(err))
27210+ goto out;
27211+
27212+ p.deblk = last_deblk(vdir);
27213+ deblk_end.deblk = p.deblk + deblk_sz;
27214+ /* smp_mb(); */
27215+ AuDebugOn(room->deblk != p.deblk);
27216+ }
27217+
27218+ err = -ENOMEM;
4a4d8108 27219+ dehstr = au_cache_alloc_vdir_dehstr();
dece6358
AM
27220+ if (unlikely(!dehstr))
27221+ goto out;
27222+
27223+ dehstr->str = &room->de->de_str;
27224+ hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
27225+ room->de->de_ino = ino;
27226+ room->de->de_type = d_type;
27227+ room->de->de_str.len = nlen;
27228+ memcpy(room->de->de_str.name, name, nlen);
27229+
27230+ err = 0;
27231+ room->deblk += sz;
27232+ if (unlikely(set_deblk_end(room, &deblk_end)))
27233+ err = append_deblk(vdir);
27234+ /* smp_mb(); */
27235+
4f0767ce 27236+out:
dece6358
AM
27237+ return err;
27238+}
27239+
27240+/* ---------------------------------------------------------------------- */
27241+
27242+void au_vdir_free(struct au_vdir *vdir)
27243+{
27244+ unsigned char **deblk;
27245+
27246+ deblk = vdir->vd_deblk;
27247+ while (vdir->vd_nblk--)
27248+ kfree(*deblk++);
27249+ kfree(vdir->vd_deblk);
27250+ au_cache_free_vdir(vdir);
27251+}
27252+
1308ab2a 27253+static struct au_vdir *alloc_vdir(struct file *file)
1facf9fc 27254+{
27255+ struct au_vdir *vdir;
1308ab2a 27256+ struct super_block *sb;
1facf9fc 27257+ int err;
27258+
1308ab2a 27259+ sb = file->f_dentry->d_sb;
dece6358
AM
27260+ SiMustAnyLock(sb);
27261+
1facf9fc 27262+ err = -ENOMEM;
27263+ vdir = au_cache_alloc_vdir();
27264+ if (unlikely(!vdir))
27265+ goto out;
27266+
27267+ vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
27268+ if (unlikely(!vdir->vd_deblk))
27269+ goto out_free;
27270+
27271+ vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
1308ab2a 27272+ if (!vdir->vd_deblk_sz) {
27273+ /* estimate the apropriate size for deblk */
27274+ vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
4a4d8108 27275+ /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
1308ab2a 27276+ }
1facf9fc 27277+ vdir->vd_nblk = 0;
27278+ vdir->vd_version = 0;
27279+ vdir->vd_jiffy = 0;
27280+ err = append_deblk(vdir);
27281+ if (!err)
27282+ return vdir; /* success */
27283+
27284+ kfree(vdir->vd_deblk);
27285+
4f0767ce 27286+out_free:
1facf9fc 27287+ au_cache_free_vdir(vdir);
4f0767ce 27288+out:
1facf9fc 27289+ vdir = ERR_PTR(err);
27290+ return vdir;
27291+}
27292+
27293+static int reinit_vdir(struct au_vdir *vdir)
27294+{
27295+ int err;
27296+ union au_vdir_deblk_p p, deblk_end;
27297+
27298+ while (vdir->vd_nblk > 1) {
27299+ kfree(vdir->vd_deblk[vdir->vd_nblk - 1]);
27300+ /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
27301+ vdir->vd_nblk--;
27302+ }
27303+ p.deblk = vdir->vd_deblk[0];
27304+ deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
27305+ err = set_deblk_end(&p, &deblk_end);
27306+ /* keep vd_dblk_sz */
27307+ vdir->vd_last.ul = 0;
27308+ vdir->vd_last.p.deblk = vdir->vd_deblk[0];
27309+ vdir->vd_version = 0;
27310+ vdir->vd_jiffy = 0;
27311+ /* smp_mb(); */
27312+ return err;
27313+}
27314+
27315+/* ---------------------------------------------------------------------- */
27316+
1facf9fc 27317+#define AuFillVdir_CALLED 1
27318+#define AuFillVdir_WHABLE (1 << 1)
dece6358 27319+#define AuFillVdir_SHWH (1 << 2)
1facf9fc 27320+#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
7f207e10
AM
27321+#define au_fset_fillvdir(flags, name) \
27322+ do { (flags) |= AuFillVdir_##name; } while (0)
27323+#define au_fclr_fillvdir(flags, name) \
27324+ do { (flags) &= ~AuFillVdir_##name; } while (0)
1facf9fc 27325+
dece6358
AM
27326+#ifndef CONFIG_AUFS_SHWH
27327+#undef AuFillVdir_SHWH
27328+#define AuFillVdir_SHWH 0
27329+#endif
27330+
1facf9fc 27331+struct fillvdir_arg {
392086de 27332+ struct dir_context ctx;
1facf9fc 27333+ struct file *file;
27334+ struct au_vdir *vdir;
dece6358
AM
27335+ struct au_nhash delist;
27336+ struct au_nhash whlist;
1facf9fc 27337+ aufs_bindex_t bindex;
27338+ unsigned int flags;
27339+ int err;
27340+};
27341+
392086de 27342+static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
1facf9fc 27343+ loff_t offset __maybe_unused, u64 h_ino,
27344+ unsigned int d_type)
27345+{
392086de 27346+ struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
1facf9fc 27347+ char *name = (void *)__name;
27348+ struct super_block *sb;
1facf9fc 27349+ ino_t ino;
dece6358 27350+ const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
1facf9fc 27351+
1facf9fc 27352+ arg->err = 0;
dece6358 27353+ sb = arg->file->f_dentry->d_sb;
1facf9fc 27354+ au_fset_fillvdir(arg->flags, CALLED);
27355+ /* smp_mb(); */
dece6358 27356+ if (nlen <= AUFS_WH_PFX_LEN
1facf9fc 27357+ || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
dece6358
AM
27358+ if (test_known(&arg->delist, name, nlen)
27359+ || au_nhash_test_known_wh(&arg->whlist, name, nlen))
27360+ goto out; /* already exists or whiteouted */
1facf9fc 27361+
27362+ sb = arg->file->f_dentry->d_sb;
dece6358 27363+ arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
4a4d8108
AM
27364+ if (!arg->err) {
27365+ if (unlikely(nlen > AUFS_MAX_NAMELEN))
27366+ d_type = DT_UNKNOWN;
dece6358
AM
27367+ arg->err = append_de(arg->vdir, name, nlen, ino,
27368+ d_type, &arg->delist);
4a4d8108 27369+ }
1facf9fc 27370+ } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
27371+ name += AUFS_WH_PFX_LEN;
dece6358
AM
27372+ nlen -= AUFS_WH_PFX_LEN;
27373+ if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
27374+ goto out; /* already whiteouted */
1facf9fc 27375+
dece6358
AM
27376+ if (shwh)
27377+ arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
27378+ &ino);
4a4d8108
AM
27379+ if (!arg->err) {
27380+ if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
27381+ d_type = DT_UNKNOWN;
1facf9fc 27382+ arg->err = au_nhash_append_wh
dece6358
AM
27383+ (&arg->whlist, name, nlen, ino, d_type,
27384+ arg->bindex, shwh);
4a4d8108 27385+ }
1facf9fc 27386+ }
27387+
4f0767ce 27388+out:
1facf9fc 27389+ if (!arg->err)
27390+ arg->vdir->vd_jiffy = jiffies;
27391+ /* smp_mb(); */
27392+ AuTraceErr(arg->err);
27393+ return arg->err;
27394+}
27395+
dece6358
AM
27396+static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
27397+ struct au_nhash *whlist, struct au_nhash *delist)
27398+{
27399+#ifdef CONFIG_AUFS_SHWH
27400+ int err;
27401+ unsigned int nh, u;
27402+ struct hlist_head *head;
c06a8ce3
AM
27403+ struct au_vdir_wh *pos;
27404+ struct hlist_node *n;
dece6358
AM
27405+ char *p, *o;
27406+ struct au_vdir_destr *destr;
27407+
27408+ AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
27409+
27410+ err = -ENOMEM;
537831f9 27411+ o = p = (void *)__get_free_page(GFP_NOFS);
dece6358
AM
27412+ if (unlikely(!p))
27413+ goto out;
27414+
27415+ err = 0;
27416+ nh = whlist->nh_num;
27417+ memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
27418+ p += AUFS_WH_PFX_LEN;
27419+ for (u = 0; u < nh; u++) {
27420+ head = whlist->nh_head + u;
c06a8ce3
AM
27421+ hlist_for_each_entry_safe(pos, n, head, wh_hash) {
27422+ destr = &pos->wh_str;
dece6358
AM
27423+ memcpy(p, destr->name, destr->len);
27424+ err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
c06a8ce3 27425+ pos->wh_ino, pos->wh_type, delist);
dece6358
AM
27426+ if (unlikely(err))
27427+ break;
27428+ }
27429+ }
27430+
537831f9 27431+ free_page((unsigned long)o);
dece6358 27432+
4f0767ce 27433+out:
dece6358
AM
27434+ AuTraceErr(err);
27435+ return err;
27436+#else
27437+ return 0;
27438+#endif
27439+}
27440+
1facf9fc 27441+static int au_do_read_vdir(struct fillvdir_arg *arg)
27442+{
27443+ int err;
dece6358 27444+ unsigned int rdhash;
1facf9fc 27445+ loff_t offset;
dece6358
AM
27446+ aufs_bindex_t bend, bindex, bstart;
27447+ unsigned char shwh;
1facf9fc 27448+ struct file *hf, *file;
27449+ struct super_block *sb;
27450+
1facf9fc 27451+ file = arg->file;
27452+ sb = file->f_dentry->d_sb;
dece6358
AM
27453+ SiMustAnyLock(sb);
27454+
27455+ rdhash = au_sbi(sb)->si_rdhash;
1308ab2a 27456+ if (!rdhash)
27457+ rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
dece6358
AM
27458+ err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
27459+ if (unlikely(err))
1facf9fc 27460+ goto out;
dece6358
AM
27461+ err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
27462+ if (unlikely(err))
1facf9fc 27463+ goto out_delist;
27464+
27465+ err = 0;
27466+ arg->flags = 0;
dece6358
AM
27467+ shwh = 0;
27468+ if (au_opt_test(au_mntflags(sb), SHWH)) {
27469+ shwh = 1;
27470+ au_fset_fillvdir(arg->flags, SHWH);
27471+ }
27472+ bstart = au_fbstart(file);
4a4d8108 27473+ bend = au_fbend_dir(file);
dece6358 27474+ for (bindex = bstart; !err && bindex <= bend; bindex++) {
4a4d8108 27475+ hf = au_hf_dir(file, bindex);
1facf9fc 27476+ if (!hf)
27477+ continue;
27478+
27479+ offset = vfsub_llseek(hf, 0, SEEK_SET);
27480+ err = offset;
27481+ if (unlikely(offset))
27482+ break;
27483+
27484+ arg->bindex = bindex;
27485+ au_fclr_fillvdir(arg->flags, WHABLE);
dece6358
AM
27486+ if (shwh
27487+ || (bindex != bend
27488+ && au_br_whable(au_sbr_perm(sb, bindex))))
1facf9fc 27489+ au_fset_fillvdir(arg->flags, WHABLE);
27490+ do {
27491+ arg->err = 0;
27492+ au_fclr_fillvdir(arg->flags, CALLED);
27493+ /* smp_mb(); */
392086de 27494+ err = vfsub_iterate_dir(hf, &arg->ctx);
1facf9fc 27495+ if (err >= 0)
27496+ err = arg->err;
27497+ } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
392086de
AM
27498+
27499+ /*
27500+ * dir_relax() may be good for concurrency, but aufs should not
27501+ * use it since it will cause a lockdep problem.
27502+ */
1facf9fc 27503+ }
dece6358
AM
27504+
27505+ if (!err && shwh)
27506+ err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
27507+
27508+ au_nhash_wh_free(&arg->whlist);
1facf9fc 27509+
4f0767ce 27510+out_delist:
dece6358 27511+ au_nhash_de_free(&arg->delist);
4f0767ce 27512+out:
1facf9fc 27513+ return err;
27514+}
27515+
27516+static int read_vdir(struct file *file, int may_read)
27517+{
27518+ int err;
27519+ unsigned long expire;
27520+ unsigned char do_read;
392086de
AM
27521+ struct fillvdir_arg arg = {
27522+ .ctx = {
27523+ .actor = au_diractor(fillvdir)
27524+ }
27525+ };
1facf9fc 27526+ struct inode *inode;
27527+ struct au_vdir *vdir, *allocated;
27528+
27529+ err = 0;
c06a8ce3 27530+ inode = file_inode(file);
1facf9fc 27531+ IMustLock(inode);
dece6358
AM
27532+ SiMustAnyLock(inode->i_sb);
27533+
1facf9fc 27534+ allocated = NULL;
27535+ do_read = 0;
27536+ expire = au_sbi(inode->i_sb)->si_rdcache;
27537+ vdir = au_ivdir(inode);
27538+ if (!vdir) {
27539+ do_read = 1;
1308ab2a 27540+ vdir = alloc_vdir(file);
1facf9fc 27541+ err = PTR_ERR(vdir);
27542+ if (IS_ERR(vdir))
27543+ goto out;
27544+ err = 0;
27545+ allocated = vdir;
27546+ } else if (may_read
27547+ && (inode->i_version != vdir->vd_version
27548+ || time_after(jiffies, vdir->vd_jiffy + expire))) {
27549+ do_read = 1;
27550+ err = reinit_vdir(vdir);
27551+ if (unlikely(err))
27552+ goto out;
27553+ }
27554+
27555+ if (!do_read)
27556+ return 0; /* success */
27557+
27558+ arg.file = file;
27559+ arg.vdir = vdir;
27560+ err = au_do_read_vdir(&arg);
27561+ if (!err) {
392086de 27562+ /* file->f_pos = 0; */ /* todo: ctx->pos? */
1facf9fc 27563+ vdir->vd_version = inode->i_version;
27564+ vdir->vd_last.ul = 0;
27565+ vdir->vd_last.p.deblk = vdir->vd_deblk[0];
27566+ if (allocated)
27567+ au_set_ivdir(inode, allocated);
27568+ } else if (allocated)
27569+ au_vdir_free(allocated);
27570+
4f0767ce 27571+out:
1facf9fc 27572+ return err;
27573+}
27574+
27575+static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
27576+{
27577+ int err, rerr;
27578+ unsigned long ul, n;
27579+ const unsigned int deblk_sz = src->vd_deblk_sz;
27580+
27581+ AuDebugOn(tgt->vd_nblk != 1);
27582+
27583+ err = -ENOMEM;
27584+ if (tgt->vd_nblk < src->vd_nblk) {
27585+ unsigned char **p;
27586+
dece6358
AM
27587+ p = krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
27588+ GFP_NOFS);
1facf9fc 27589+ if (unlikely(!p))
27590+ goto out;
27591+ tgt->vd_deblk = p;
27592+ }
27593+
1308ab2a 27594+ if (tgt->vd_deblk_sz != deblk_sz) {
27595+ unsigned char *p;
27596+
27597+ tgt->vd_deblk_sz = deblk_sz;
27598+ p = krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS);
27599+ if (unlikely(!p))
27600+ goto out;
27601+ tgt->vd_deblk[0] = p;
27602+ }
1facf9fc 27603+ memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
1facf9fc 27604+ tgt->vd_version = src->vd_version;
27605+ tgt->vd_jiffy = src->vd_jiffy;
27606+
27607+ n = src->vd_nblk;
27608+ for (ul = 1; ul < n; ul++) {
dece6358
AM
27609+ tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
27610+ GFP_NOFS);
27611+ if (unlikely(!tgt->vd_deblk[ul]))
1facf9fc 27612+ goto out;
1308ab2a 27613+ tgt->vd_nblk++;
1facf9fc 27614+ }
1308ab2a 27615+ tgt->vd_nblk = n;
27616+ tgt->vd_last.ul = tgt->vd_last.ul;
27617+ tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
27618+ tgt->vd_last.p.deblk += src->vd_last.p.deblk
27619+ - src->vd_deblk[src->vd_last.ul];
1facf9fc 27620+ /* smp_mb(); */
27621+ return 0; /* success */
27622+
4f0767ce 27623+out:
1facf9fc 27624+ rerr = reinit_vdir(tgt);
27625+ BUG_ON(rerr);
27626+ return err;
27627+}
27628+
27629+int au_vdir_init(struct file *file)
27630+{
27631+ int err;
27632+ struct inode *inode;
27633+ struct au_vdir *vdir_cache, *allocated;
27634+
392086de 27635+ /* test file->f_pos here instead of ctx->pos */
1facf9fc 27636+ err = read_vdir(file, !file->f_pos);
27637+ if (unlikely(err))
27638+ goto out;
27639+
27640+ allocated = NULL;
27641+ vdir_cache = au_fvdir_cache(file);
27642+ if (!vdir_cache) {
1308ab2a 27643+ vdir_cache = alloc_vdir(file);
1facf9fc 27644+ err = PTR_ERR(vdir_cache);
27645+ if (IS_ERR(vdir_cache))
27646+ goto out;
27647+ allocated = vdir_cache;
27648+ } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
392086de 27649+ /* test file->f_pos here instead of ctx->pos */
1facf9fc 27650+ err = reinit_vdir(vdir_cache);
27651+ if (unlikely(err))
27652+ goto out;
27653+ } else
27654+ return 0; /* success */
27655+
c06a8ce3 27656+ inode = file_inode(file);
1facf9fc 27657+ err = copy_vdir(vdir_cache, au_ivdir(inode));
27658+ if (!err) {
27659+ file->f_version = inode->i_version;
27660+ if (allocated)
27661+ au_set_fvdir_cache(file, allocated);
27662+ } else if (allocated)
27663+ au_vdir_free(allocated);
27664+
4f0767ce 27665+out:
1facf9fc 27666+ return err;
27667+}
27668+
27669+static loff_t calc_offset(struct au_vdir *vdir)
27670+{
27671+ loff_t offset;
27672+ union au_vdir_deblk_p p;
27673+
27674+ p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
27675+ offset = vdir->vd_last.p.deblk - p.deblk;
27676+ offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
27677+ return offset;
27678+}
27679+
27680+/* returns true or false */
392086de 27681+static int seek_vdir(struct file *file, struct dir_context *ctx)
1facf9fc 27682+{
27683+ int valid;
27684+ unsigned int deblk_sz;
27685+ unsigned long ul, n;
27686+ loff_t offset;
27687+ union au_vdir_deblk_p p, deblk_end;
27688+ struct au_vdir *vdir_cache;
27689+
27690+ valid = 1;
27691+ vdir_cache = au_fvdir_cache(file);
27692+ offset = calc_offset(vdir_cache);
27693+ AuDbg("offset %lld\n", offset);
392086de 27694+ if (ctx->pos == offset)
1facf9fc 27695+ goto out;
27696+
27697+ vdir_cache->vd_last.ul = 0;
27698+ vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
392086de 27699+ if (!ctx->pos)
1facf9fc 27700+ goto out;
27701+
27702+ valid = 0;
27703+ deblk_sz = vdir_cache->vd_deblk_sz;
392086de 27704+ ul = div64_u64(ctx->pos, deblk_sz);
1facf9fc 27705+ AuDbg("ul %lu\n", ul);
27706+ if (ul >= vdir_cache->vd_nblk)
27707+ goto out;
27708+
27709+ n = vdir_cache->vd_nblk;
27710+ for (; ul < n; ul++) {
27711+ p.deblk = vdir_cache->vd_deblk[ul];
27712+ deblk_end.deblk = p.deblk + deblk_sz;
27713+ offset = ul;
27714+ offset *= deblk_sz;
392086de 27715+ while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
1facf9fc 27716+ unsigned int l;
27717+
27718+ l = calc_size(p.de->de_str.len);
27719+ offset += l;
27720+ p.deblk += l;
27721+ }
27722+ if (!is_deblk_end(&p, &deblk_end)) {
27723+ valid = 1;
27724+ vdir_cache->vd_last.ul = ul;
27725+ vdir_cache->vd_last.p = p;
27726+ break;
27727+ }
27728+ }
27729+
4f0767ce 27730+out:
1facf9fc 27731+ /* smp_mb(); */
27732+ AuTraceErr(!valid);
27733+ return valid;
27734+}
27735+
392086de 27736+int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
1facf9fc 27737+{
1facf9fc 27738+ unsigned int l, deblk_sz;
27739+ union au_vdir_deblk_p deblk_end;
27740+ struct au_vdir *vdir_cache;
27741+ struct au_vdir_de *de;
27742+
27743+ vdir_cache = au_fvdir_cache(file);
392086de 27744+ if (!seek_vdir(file, ctx))
1facf9fc 27745+ return 0;
27746+
27747+ deblk_sz = vdir_cache->vd_deblk_sz;
27748+ while (1) {
27749+ deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
27750+ deblk_end.deblk += deblk_sz;
27751+ while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
27752+ de = vdir_cache->vd_last.p.de;
27753+ AuDbg("%.*s, off%lld, i%lu, dt%d\n",
392086de 27754+ de->de_str.len, de->de_str.name, ctx->pos,
1facf9fc 27755+ (unsigned long)de->de_ino, de->de_type);
392086de
AM
27756+ if (unlikely(!dir_emit(ctx, de->de_str.name,
27757+ de->de_str.len, de->de_ino,
27758+ de->de_type))) {
1facf9fc 27759+ /* todo: ignore the error caused by udba? */
27760+ /* return err; */
27761+ return 0;
27762+ }
27763+
27764+ l = calc_size(de->de_str.len);
27765+ vdir_cache->vd_last.p.deblk += l;
392086de 27766+ ctx->pos += l;
1facf9fc 27767+ }
27768+ if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
27769+ vdir_cache->vd_last.ul++;
27770+ vdir_cache->vd_last.p.deblk
27771+ = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
392086de 27772+ ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
1facf9fc 27773+ continue;
27774+ }
27775+ break;
27776+ }
27777+
27778+ /* smp_mb(); */
27779+ return 0;
27780+}
7f207e10
AM
27781diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
27782--- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 27783+++ linux/fs/aufs/vfsub.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 27784@@ -0,0 +1,782 @@
1facf9fc 27785+/*
523b37e3 27786+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 27787+ *
27788+ * This program, aufs is free software; you can redistribute it and/or modify
27789+ * it under the terms of the GNU General Public License as published by
27790+ * the Free Software Foundation; either version 2 of the License, or
27791+ * (at your option) any later version.
dece6358
AM
27792+ *
27793+ * This program is distributed in the hope that it will be useful,
27794+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
27795+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27796+ * GNU General Public License for more details.
27797+ *
27798+ * You should have received a copy of the GNU General Public License
523b37e3 27799+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 27800+ */
27801+
27802+/*
27803+ * sub-routines for VFS
27804+ */
27805+
1308ab2a 27806+#include <linux/ima.h>
dece6358
AM
27807+#include <linux/namei.h>
27808+#include <linux/security.h>
27809+#include <linux/splice.h>
1facf9fc 27810+#include "aufs.h"
27811+
27812+int vfsub_update_h_iattr(struct path *h_path, int *did)
27813+{
27814+ int err;
27815+ struct kstat st;
27816+ struct super_block *h_sb;
27817+
27818+ /* for remote fs, leave work for its getattr or d_revalidate */
27819+ /* for bad i_attr fs, handle them in aufs_getattr() */
27820+ /* still some fs may acquire i_mutex. we need to skip them */
27821+ err = 0;
27822+ if (!did)
27823+ did = &err;
27824+ h_sb = h_path->dentry->d_sb;
27825+ *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
27826+ if (*did)
c06a8ce3 27827+ err = vfs_getattr(h_path, &st);
1facf9fc 27828+
27829+ return err;
27830+}
27831+
27832+/* ---------------------------------------------------------------------- */
27833+
4a4d8108 27834+struct file *vfsub_dentry_open(struct path *path, int flags)
1308ab2a 27835+{
27836+ struct file *file;
27837+
b4510431 27838+ file = dentry_open(path, flags /* | __FMODE_NONOTIFY */,
7f207e10 27839+ current_cred());
2cbb1c4b
JR
27840+ if (!IS_ERR_OR_NULL(file)
27841+ && (file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
27842+ i_readcount_inc(path->dentry->d_inode);
4a4d8108 27843+
1308ab2a 27844+ return file;
27845+}
27846+
1facf9fc 27847+struct file *vfsub_filp_open(const char *path, int oflags, int mode)
27848+{
27849+ struct file *file;
27850+
2cbb1c4b 27851+ lockdep_off();
7f207e10 27852+ file = filp_open(path,
2cbb1c4b 27853+ oflags /* | __FMODE_NONOTIFY */,
7f207e10 27854+ mode);
2cbb1c4b 27855+ lockdep_on();
1facf9fc 27856+ if (IS_ERR(file))
27857+ goto out;
27858+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
27859+
4f0767ce 27860+out:
1facf9fc 27861+ return file;
27862+}
27863+
27864+int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
27865+{
27866+ int err;
27867+
1facf9fc 27868+ err = kern_path(name, flags, path);
1facf9fc 27869+ if (!err && path->dentry->d_inode)
27870+ vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
27871+ return err;
27872+}
27873+
27874+struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
27875+ int len)
27876+{
27877+ struct path path = {
27878+ .mnt = NULL
27879+ };
27880+
1308ab2a 27881+ /* VFS checks it too, but by WARN_ON_ONCE() */
1facf9fc 27882+ IMustLock(parent->d_inode);
27883+
27884+ path.dentry = lookup_one_len(name, parent, len);
27885+ if (IS_ERR(path.dentry))
27886+ goto out;
27887+ if (path.dentry->d_inode)
27888+ vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
27889+
4f0767ce 27890+out:
4a4d8108 27891+ AuTraceErrPtr(path.dentry);
1facf9fc 27892+ return path.dentry;
27893+}
27894+
b4510431 27895+void vfsub_call_lkup_one(void *args)
2cbb1c4b 27896+{
b4510431
AM
27897+ struct vfsub_lkup_one_args *a = args;
27898+ *a->errp = vfsub_lkup_one(a->name, a->parent);
2cbb1c4b
JR
27899+}
27900+
1facf9fc 27901+/* ---------------------------------------------------------------------- */
27902+
27903+struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
27904+ struct dentry *d2, struct au_hinode *hdir2)
27905+{
27906+ struct dentry *d;
27907+
2cbb1c4b 27908+ lockdep_off();
1facf9fc 27909+ d = lock_rename(d1, d2);
2cbb1c4b 27910+ lockdep_on();
4a4d8108 27911+ au_hn_suspend(hdir1);
1facf9fc 27912+ if (hdir1 != hdir2)
4a4d8108 27913+ au_hn_suspend(hdir2);
1facf9fc 27914+
27915+ return d;
27916+}
27917+
27918+void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
27919+ struct dentry *d2, struct au_hinode *hdir2)
27920+{
4a4d8108 27921+ au_hn_resume(hdir1);
1facf9fc 27922+ if (hdir1 != hdir2)
4a4d8108 27923+ au_hn_resume(hdir2);
2cbb1c4b 27924+ lockdep_off();
1facf9fc 27925+ unlock_rename(d1, d2);
2cbb1c4b 27926+ lockdep_on();
1facf9fc 27927+}
27928+
27929+/* ---------------------------------------------------------------------- */
27930+
b4510431 27931+int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
1facf9fc 27932+{
27933+ int err;
27934+ struct dentry *d;
27935+
27936+ IMustLock(dir);
27937+
27938+ d = path->dentry;
27939+ path->dentry = d->d_parent;
b752ccd1 27940+ err = security_path_mknod(path, d, mode, 0);
1facf9fc 27941+ path->dentry = d;
27942+ if (unlikely(err))
27943+ goto out;
27944+
b4510431 27945+ err = vfs_create(dir, path->dentry, mode, want_excl);
1facf9fc 27946+ if (!err) {
27947+ struct path tmp = *path;
27948+ int did;
27949+
27950+ vfsub_update_h_iattr(&tmp, &did);
27951+ if (did) {
27952+ tmp.dentry = path->dentry->d_parent;
27953+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
27954+ }
27955+ /*ignore*/
27956+ }
27957+
4f0767ce 27958+out:
1facf9fc 27959+ return err;
27960+}
27961+
27962+int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
27963+{
27964+ int err;
27965+ struct dentry *d;
27966+
27967+ IMustLock(dir);
27968+
27969+ d = path->dentry;
27970+ path->dentry = d->d_parent;
b752ccd1 27971+ err = security_path_symlink(path, d, symname);
1facf9fc 27972+ path->dentry = d;
27973+ if (unlikely(err))
27974+ goto out;
27975+
27976+ err = vfs_symlink(dir, path->dentry, symname);
27977+ if (!err) {
27978+ struct path tmp = *path;
27979+ int did;
27980+
27981+ vfsub_update_h_iattr(&tmp, &did);
27982+ if (did) {
27983+ tmp.dentry = path->dentry->d_parent;
27984+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
27985+ }
27986+ /*ignore*/
27987+ }
27988+
4f0767ce 27989+out:
1facf9fc 27990+ return err;
27991+}
27992+
27993+int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
27994+{
27995+ int err;
27996+ struct dentry *d;
27997+
27998+ IMustLock(dir);
27999+
28000+ d = path->dentry;
28001+ path->dentry = d->d_parent;
027c5e7a 28002+ err = security_path_mknod(path, d, mode, new_encode_dev(dev));
1facf9fc 28003+ path->dentry = d;
28004+ if (unlikely(err))
28005+ goto out;
28006+
28007+ err = vfs_mknod(dir, path->dentry, mode, dev);
28008+ if (!err) {
28009+ struct path tmp = *path;
28010+ int did;
28011+
28012+ vfsub_update_h_iattr(&tmp, &did);
28013+ if (did) {
28014+ tmp.dentry = path->dentry->d_parent;
28015+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
28016+ }
28017+ /*ignore*/
28018+ }
28019+
4f0767ce 28020+out:
1facf9fc 28021+ return err;
28022+}
28023+
28024+static int au_test_nlink(struct inode *inode)
28025+{
28026+ const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
28027+
28028+ if (!au_test_fs_no_limit_nlink(inode->i_sb)
28029+ || inode->i_nlink < link_max)
28030+ return 0;
28031+ return -EMLINK;
28032+}
28033+
523b37e3
AM
28034+int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
28035+ struct inode **delegated_inode)
1facf9fc 28036+{
28037+ int err;
28038+ struct dentry *d;
28039+
28040+ IMustLock(dir);
28041+
28042+ err = au_test_nlink(src_dentry->d_inode);
28043+ if (unlikely(err))
28044+ return err;
28045+
b4510431 28046+ /* we don't call may_linkat() */
1facf9fc 28047+ d = path->dentry;
28048+ path->dentry = d->d_parent;
b752ccd1 28049+ err = security_path_link(src_dentry, path, d);
1facf9fc 28050+ path->dentry = d;
28051+ if (unlikely(err))
28052+ goto out;
28053+
2cbb1c4b 28054+ lockdep_off();
523b37e3 28055+ err = vfs_link(src_dentry, dir, path->dentry, delegated_inode);
2cbb1c4b 28056+ lockdep_on();
1facf9fc 28057+ if (!err) {
28058+ struct path tmp = *path;
28059+ int did;
28060+
28061+ /* fuse has different memory inode for the same inumber */
28062+ vfsub_update_h_iattr(&tmp, &did);
28063+ if (did) {
28064+ tmp.dentry = path->dentry->d_parent;
28065+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
28066+ tmp.dentry = src_dentry;
28067+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
28068+ }
28069+ /*ignore*/
28070+ }
28071+
4f0767ce 28072+out:
1facf9fc 28073+ return err;
28074+}
28075+
28076+int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
523b37e3
AM
28077+ struct inode *dir, struct path *path,
28078+ struct inode **delegated_inode)
1facf9fc 28079+{
28080+ int err;
28081+ struct path tmp = {
28082+ .mnt = path->mnt
28083+ };
28084+ struct dentry *d;
28085+
28086+ IMustLock(dir);
28087+ IMustLock(src_dir);
28088+
28089+ d = path->dentry;
28090+ path->dentry = d->d_parent;
28091+ tmp.dentry = src_dentry->d_parent;
38d290e6 28092+ err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
1facf9fc 28093+ path->dentry = d;
28094+ if (unlikely(err))
28095+ goto out;
28096+
2cbb1c4b 28097+ lockdep_off();
523b37e3 28098+ err = vfs_rename(src_dir, src_dentry, dir, path->dentry,
38d290e6 28099+ delegated_inode, /*flags*/0);
2cbb1c4b 28100+ lockdep_on();
1facf9fc 28101+ if (!err) {
28102+ int did;
28103+
28104+ tmp.dentry = d->d_parent;
28105+ vfsub_update_h_iattr(&tmp, &did);
28106+ if (did) {
28107+ tmp.dentry = src_dentry;
28108+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
28109+ tmp.dentry = src_dentry->d_parent;
28110+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
28111+ }
28112+ /*ignore*/
28113+ }
28114+
4f0767ce 28115+out:
1facf9fc 28116+ return err;
28117+}
28118+
28119+int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
28120+{
28121+ int err;
28122+ struct dentry *d;
28123+
28124+ IMustLock(dir);
28125+
28126+ d = path->dentry;
28127+ path->dentry = d->d_parent;
b752ccd1 28128+ err = security_path_mkdir(path, d, mode);
1facf9fc 28129+ path->dentry = d;
28130+ if (unlikely(err))
28131+ goto out;
28132+
28133+ err = vfs_mkdir(dir, path->dentry, mode);
28134+ if (!err) {
28135+ struct path tmp = *path;
28136+ int did;
28137+
28138+ vfsub_update_h_iattr(&tmp, &did);
28139+ if (did) {
28140+ tmp.dentry = path->dentry->d_parent;
28141+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
28142+ }
28143+ /*ignore*/
28144+ }
28145+
4f0767ce 28146+out:
1facf9fc 28147+ return err;
28148+}
28149+
28150+int vfsub_rmdir(struct inode *dir, struct path *path)
28151+{
28152+ int err;
28153+ struct dentry *d;
28154+
28155+ IMustLock(dir);
28156+
28157+ d = path->dentry;
28158+ path->dentry = d->d_parent;
b752ccd1 28159+ err = security_path_rmdir(path, d);
1facf9fc 28160+ path->dentry = d;
28161+ if (unlikely(err))
28162+ goto out;
28163+
2cbb1c4b 28164+ lockdep_off();
1facf9fc 28165+ err = vfs_rmdir(dir, path->dentry);
2cbb1c4b 28166+ lockdep_on();
1facf9fc 28167+ if (!err) {
28168+ struct path tmp = {
28169+ .dentry = path->dentry->d_parent,
28170+ .mnt = path->mnt
28171+ };
28172+
28173+ vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
28174+ }
28175+
4f0767ce 28176+out:
1facf9fc 28177+ return err;
28178+}
28179+
28180+/* ---------------------------------------------------------------------- */
28181+
9dbd164d 28182+/* todo: support mmap_sem? */
1facf9fc 28183+ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
28184+ loff_t *ppos)
28185+{
28186+ ssize_t err;
28187+
2cbb1c4b 28188+ lockdep_off();
1facf9fc 28189+ err = vfs_read(file, ubuf, count, ppos);
2cbb1c4b 28190+ lockdep_on();
1facf9fc 28191+ if (err >= 0)
28192+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
28193+ return err;
28194+}
28195+
28196+/* todo: kernel_read()? */
28197+ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
28198+ loff_t *ppos)
28199+{
28200+ ssize_t err;
28201+ mm_segment_t oldfs;
b752ccd1
AM
28202+ union {
28203+ void *k;
28204+ char __user *u;
28205+ } buf;
1facf9fc 28206+
b752ccd1 28207+ buf.k = kbuf;
1facf9fc 28208+ oldfs = get_fs();
28209+ set_fs(KERNEL_DS);
b752ccd1 28210+ err = vfsub_read_u(file, buf.u, count, ppos);
1facf9fc 28211+ set_fs(oldfs);
28212+ return err;
28213+}
28214+
28215+ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
28216+ loff_t *ppos)
28217+{
28218+ ssize_t err;
28219+
2cbb1c4b 28220+ lockdep_off();
1facf9fc 28221+ err = vfs_write(file, ubuf, count, ppos);
2cbb1c4b 28222+ lockdep_on();
1facf9fc 28223+ if (err >= 0)
28224+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
28225+ return err;
28226+}
28227+
28228+ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
28229+{
28230+ ssize_t err;
28231+ mm_segment_t oldfs;
b752ccd1
AM
28232+ union {
28233+ void *k;
28234+ const char __user *u;
28235+ } buf;
1facf9fc 28236+
b752ccd1 28237+ buf.k = kbuf;
1facf9fc 28238+ oldfs = get_fs();
28239+ set_fs(KERNEL_DS);
b752ccd1 28240+ err = vfsub_write_u(file, buf.u, count, ppos);
1facf9fc 28241+ set_fs(oldfs);
28242+ return err;
28243+}
28244+
4a4d8108
AM
28245+int vfsub_flush(struct file *file, fl_owner_t id)
28246+{
28247+ int err;
28248+
28249+ err = 0;
523b37e3 28250+ if (file->f_op->flush) {
2cbb1c4b
JR
28251+ if (!au_test_nfs(file->f_dentry->d_sb))
28252+ err = file->f_op->flush(file, id);
28253+ else {
28254+ lockdep_off();
28255+ err = file->f_op->flush(file, id);
28256+ lockdep_on();
28257+ }
4a4d8108
AM
28258+ if (!err)
28259+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
28260+ /*ignore*/
28261+ }
28262+ return err;
28263+}
28264+
392086de 28265+int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
1facf9fc 28266+{
28267+ int err;
28268+
523b37e3 28269+ AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
392086de 28270+
2cbb1c4b 28271+ lockdep_off();
392086de 28272+ err = iterate_dir(file, ctx);
2cbb1c4b 28273+ lockdep_on();
1facf9fc 28274+ if (err >= 0)
28275+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
28276+ return err;
28277+}
28278+
28279+long vfsub_splice_to(struct file *in, loff_t *ppos,
28280+ struct pipe_inode_info *pipe, size_t len,
28281+ unsigned int flags)
28282+{
28283+ long err;
28284+
2cbb1c4b 28285+ lockdep_off();
0fc653ad 28286+ err = do_splice_to(in, ppos, pipe, len, flags);
2cbb1c4b 28287+ lockdep_on();
4a4d8108 28288+ file_accessed(in);
1facf9fc 28289+ if (err >= 0)
28290+ vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
28291+ return err;
28292+}
28293+
28294+long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
28295+ loff_t *ppos, size_t len, unsigned int flags)
28296+{
28297+ long err;
28298+
2cbb1c4b 28299+ lockdep_off();
0fc653ad 28300+ err = do_splice_from(pipe, out, ppos, len, flags);
2cbb1c4b 28301+ lockdep_on();
1facf9fc 28302+ if (err >= 0)
28303+ vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
28304+ return err;
28305+}
28306+
53392da6
AM
28307+int vfsub_fsync(struct file *file, struct path *path, int datasync)
28308+{
28309+ int err;
28310+
28311+ /* file can be NULL */
28312+ lockdep_off();
28313+ err = vfs_fsync(file, datasync);
28314+ lockdep_on();
28315+ if (!err) {
28316+ if (!path) {
28317+ AuDebugOn(!file);
28318+ path = &file->f_path;
28319+ }
28320+ vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
28321+ }
28322+ return err;
28323+}
28324+
1facf9fc 28325+/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
28326+int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
28327+ struct file *h_file)
28328+{
28329+ int err;
28330+ struct inode *h_inode;
c06a8ce3 28331+ struct super_block *h_sb;
1facf9fc 28332+
1facf9fc 28333+ if (!h_file) {
c06a8ce3
AM
28334+ err = vfsub_truncate(h_path, length);
28335+ goto out;
1facf9fc 28336+ }
28337+
c06a8ce3
AM
28338+ h_inode = h_path->dentry->d_inode;
28339+ h_sb = h_inode->i_sb;
28340+ lockdep_off();
28341+ sb_start_write(h_sb);
28342+ lockdep_on();
1facf9fc 28343+ err = locks_verify_truncate(h_inode, h_file, length);
28344+ if (!err)
953406b4 28345+ err = security_path_truncate(h_path);
2cbb1c4b
JR
28346+ if (!err) {
28347+ lockdep_off();
1facf9fc 28348+ err = do_truncate(h_path->dentry, length, attr, h_file);
2cbb1c4b
JR
28349+ lockdep_on();
28350+ }
c06a8ce3
AM
28351+ lockdep_off();
28352+ sb_end_write(h_sb);
28353+ lockdep_on();
1facf9fc 28354+
4f0767ce 28355+out:
1facf9fc 28356+ return err;
28357+}
28358+
28359+/* ---------------------------------------------------------------------- */
28360+
28361+struct au_vfsub_mkdir_args {
28362+ int *errp;
28363+ struct inode *dir;
28364+ struct path *path;
28365+ int mode;
28366+};
28367+
28368+static void au_call_vfsub_mkdir(void *args)
28369+{
28370+ struct au_vfsub_mkdir_args *a = args;
28371+ *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
28372+}
28373+
28374+int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
28375+{
28376+ int err, do_sio, wkq_err;
28377+
28378+ do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
28379+ if (!do_sio)
28380+ err = vfsub_mkdir(dir, path, mode);
28381+ else {
28382+ struct au_vfsub_mkdir_args args = {
28383+ .errp = &err,
28384+ .dir = dir,
28385+ .path = path,
28386+ .mode = mode
28387+ };
28388+ wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
28389+ if (unlikely(wkq_err))
28390+ err = wkq_err;
28391+ }
28392+
28393+ return err;
28394+}
28395+
28396+struct au_vfsub_rmdir_args {
28397+ int *errp;
28398+ struct inode *dir;
28399+ struct path *path;
28400+};
28401+
28402+static void au_call_vfsub_rmdir(void *args)
28403+{
28404+ struct au_vfsub_rmdir_args *a = args;
28405+ *a->errp = vfsub_rmdir(a->dir, a->path);
28406+}
28407+
28408+int vfsub_sio_rmdir(struct inode *dir, struct path *path)
28409+{
28410+ int err, do_sio, wkq_err;
28411+
28412+ do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
28413+ if (!do_sio)
28414+ err = vfsub_rmdir(dir, path);
28415+ else {
28416+ struct au_vfsub_rmdir_args args = {
28417+ .errp = &err,
28418+ .dir = dir,
28419+ .path = path
28420+ };
28421+ wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
28422+ if (unlikely(wkq_err))
28423+ err = wkq_err;
28424+ }
28425+
28426+ return err;
28427+}
28428+
28429+/* ---------------------------------------------------------------------- */
28430+
28431+struct notify_change_args {
28432+ int *errp;
28433+ struct path *path;
28434+ struct iattr *ia;
523b37e3 28435+ struct inode **delegated_inode;
1facf9fc 28436+};
28437+
28438+static void call_notify_change(void *args)
28439+{
28440+ struct notify_change_args *a = args;
28441+ struct inode *h_inode;
28442+
28443+ h_inode = a->path->dentry->d_inode;
28444+ IMustLock(h_inode);
28445+
28446+ *a->errp = -EPERM;
28447+ if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
523b37e3
AM
28448+ *a->errp = notify_change(a->path->dentry, a->ia,
28449+ a->delegated_inode);
1facf9fc 28450+ if (!*a->errp)
28451+ vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
28452+ }
28453+ AuTraceErr(*a->errp);
28454+}
28455+
523b37e3
AM
28456+int vfsub_notify_change(struct path *path, struct iattr *ia,
28457+ struct inode **delegated_inode)
1facf9fc 28458+{
28459+ int err;
28460+ struct notify_change_args args = {
523b37e3
AM
28461+ .errp = &err,
28462+ .path = path,
28463+ .ia = ia,
28464+ .delegated_inode = delegated_inode
1facf9fc 28465+ };
28466+
28467+ call_notify_change(&args);
28468+
28469+ return err;
28470+}
28471+
523b37e3
AM
28472+int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
28473+ struct inode **delegated_inode)
1facf9fc 28474+{
28475+ int err, wkq_err;
28476+ struct notify_change_args args = {
523b37e3
AM
28477+ .errp = &err,
28478+ .path = path,
28479+ .ia = ia,
28480+ .delegated_inode = delegated_inode
1facf9fc 28481+ };
28482+
28483+ wkq_err = au_wkq_wait(call_notify_change, &args);
28484+ if (unlikely(wkq_err))
28485+ err = wkq_err;
28486+
28487+ return err;
28488+}
28489+
28490+/* ---------------------------------------------------------------------- */
28491+
28492+struct unlink_args {
28493+ int *errp;
28494+ struct inode *dir;
28495+ struct path *path;
523b37e3 28496+ struct inode **delegated_inode;
1facf9fc 28497+};
28498+
28499+static void call_unlink(void *args)
28500+{
28501+ struct unlink_args *a = args;
28502+ struct dentry *d = a->path->dentry;
28503+ struct inode *h_inode;
28504+ const int stop_sillyrename = (au_test_nfs(d->d_sb)
392086de 28505+ && d_count(d) == 1);
1facf9fc 28506+
28507+ IMustLock(a->dir);
28508+
28509+ a->path->dentry = d->d_parent;
28510+ *a->errp = security_path_unlink(a->path, d);
28511+ a->path->dentry = d;
28512+ if (unlikely(*a->errp))
28513+ return;
28514+
28515+ if (!stop_sillyrename)
28516+ dget(d);
28517+ h_inode = d->d_inode;
28518+ if (h_inode)
027c5e7a 28519+ ihold(h_inode);
1facf9fc 28520+
2cbb1c4b 28521+ lockdep_off();
523b37e3 28522+ *a->errp = vfs_unlink(a->dir, d, a->delegated_inode);
2cbb1c4b 28523+ lockdep_on();
1facf9fc 28524+ if (!*a->errp) {
28525+ struct path tmp = {
28526+ .dentry = d->d_parent,
28527+ .mnt = a->path->mnt
28528+ };
28529+ vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
28530+ }
28531+
28532+ if (!stop_sillyrename)
28533+ dput(d);
28534+ if (h_inode)
28535+ iput(h_inode);
28536+
28537+ AuTraceErr(*a->errp);
28538+}
28539+
28540+/*
28541+ * @dir: must be locked.
28542+ * @dentry: target dentry.
28543+ */
523b37e3
AM
28544+int vfsub_unlink(struct inode *dir, struct path *path,
28545+ struct inode **delegated_inode, int force)
1facf9fc 28546+{
28547+ int err;
28548+ struct unlink_args args = {
523b37e3
AM
28549+ .errp = &err,
28550+ .dir = dir,
28551+ .path = path,
28552+ .delegated_inode = delegated_inode
1facf9fc 28553+ };
28554+
28555+ if (!force)
28556+ call_unlink(&args);
28557+ else {
28558+ int wkq_err;
28559+
28560+ wkq_err = au_wkq_wait(call_unlink, &args);
28561+ if (unlikely(wkq_err))
28562+ err = wkq_err;
28563+ }
28564+
28565+ return err;
28566+}
7f207e10
AM
28567diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
28568--- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 28569+++ linux/fs/aufs/vfsub.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 28570@@ -0,0 +1,282 @@
1facf9fc 28571+/*
523b37e3 28572+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 28573+ *
28574+ * This program, aufs is free software; you can redistribute it and/or modify
28575+ * it under the terms of the GNU General Public License as published by
28576+ * the Free Software Foundation; either version 2 of the License, or
28577+ * (at your option) any later version.
dece6358
AM
28578+ *
28579+ * This program is distributed in the hope that it will be useful,
28580+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28581+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28582+ * GNU General Public License for more details.
28583+ *
28584+ * You should have received a copy of the GNU General Public License
523b37e3 28585+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 28586+ */
28587+
28588+/*
28589+ * sub-routines for VFS
28590+ */
28591+
28592+#ifndef __AUFS_VFSUB_H__
28593+#define __AUFS_VFSUB_H__
28594+
28595+#ifdef __KERNEL__
28596+
28597+#include <linux/fs.h>
0c5527e5 28598+#include <linux/lglock.h>
b4510431 28599+#include <linux/mount.h>
7f207e10 28600+#include "debug.h"
1facf9fc 28601+
7f207e10 28602+/* copied from linux/fs/internal.h */
2cbb1c4b 28603+/* todo: BAD approach!! */
c06a8ce3 28604+extern void __mnt_drop_write(struct vfsmount *);
2cbb1c4b 28605+extern spinlock_t inode_sb_list_lock;
7f207e10
AM
28606+
28607+/* ---------------------------------------------------------------------- */
1facf9fc 28608+
28609+/* lock subclass for lower inode */
28610+/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
28611+/* reduce? gave up. */
28612+enum {
523b37e3 28613+ AuLsc_I_Begin = I_MUTEX_NONDIR2, /* 4 */
1facf9fc 28614+ AuLsc_I_PARENT, /* lower inode, parent first */
28615+ AuLsc_I_PARENT2, /* copyup dirs */
dece6358 28616+ AuLsc_I_PARENT3, /* copyup wh */
1facf9fc 28617+ AuLsc_I_CHILD,
28618+ AuLsc_I_CHILD2,
28619+ AuLsc_I_End
28620+};
28621+
28622+/* to debug easier, do not make them inlined functions */
28623+#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
28624+#define IMustLock(i) MtxMustLock(&(i)->i_mutex)
28625+
28626+/* ---------------------------------------------------------------------- */
28627+
7f207e10
AM
28628+static inline void vfsub_drop_nlink(struct inode *inode)
28629+{
28630+ AuDebugOn(!inode->i_nlink);
28631+ drop_nlink(inode);
28632+}
28633+
027c5e7a
AM
28634+static inline void vfsub_dead_dir(struct inode *inode)
28635+{
28636+ AuDebugOn(!S_ISDIR(inode->i_mode));
28637+ inode->i_flags |= S_DEAD;
28638+ clear_nlink(inode);
28639+}
28640+
392086de
AM
28641+static inline int vfsub_native_ro(struct inode *inode)
28642+{
28643+ return (inode->i_sb->s_flags & MS_RDONLY)
28644+ || IS_RDONLY(inode)
28645+ /* || IS_APPEND(inode) */
28646+ || IS_IMMUTABLE(inode);
28647+}
28648+
7f207e10
AM
28649+/* ---------------------------------------------------------------------- */
28650+
28651+int vfsub_update_h_iattr(struct path *h_path, int *did);
28652+struct file *vfsub_dentry_open(struct path *path, int flags);
28653+struct file *vfsub_filp_open(const char *path, int oflags, int mode);
1facf9fc 28654+int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
b4510431 28655+
1facf9fc 28656+struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
28657+ int len);
b4510431
AM
28658+
28659+struct vfsub_lkup_one_args {
28660+ struct dentry **errp;
28661+ struct qstr *name;
28662+ struct dentry *parent;
28663+};
28664+
28665+static inline struct dentry *vfsub_lkup_one(struct qstr *name,
28666+ struct dentry *parent)
28667+{
28668+ return vfsub_lookup_one_len(name->name, parent, name->len);
28669+}
28670+
28671+void vfsub_call_lkup_one(void *args);
28672+
28673+/* ---------------------------------------------------------------------- */
28674+
28675+static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
28676+{
28677+ int err;
28678+ lockdep_off();
28679+ err = mnt_want_write(mnt);
28680+ lockdep_on();
28681+ return err;
28682+}
28683+
28684+static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
28685+{
28686+ lockdep_off();
28687+ mnt_drop_write(mnt);
28688+ lockdep_on();
28689+}
1facf9fc 28690+
c06a8ce3
AM
28691+static inline void vfsub_mnt_drop_write_file(struct file *file)
28692+{
28693+ lockdep_off();
28694+ mnt_drop_write_file(file);
28695+ lockdep_on();
28696+}
28697+
1facf9fc 28698+/* ---------------------------------------------------------------------- */
28699+
28700+struct au_hinode;
28701+struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
28702+ struct dentry *d2, struct au_hinode *hdir2);
28703+void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
28704+ struct dentry *d2, struct au_hinode *hdir2);
28705+
537831f9
AM
28706+int vfsub_create(struct inode *dir, struct path *path, int mode,
28707+ bool want_excl);
1facf9fc 28708+int vfsub_symlink(struct inode *dir, struct path *path,
28709+ const char *symname);
28710+int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
28711+int vfsub_link(struct dentry *src_dentry, struct inode *dir,
523b37e3 28712+ struct path *path, struct inode **delegated_inode);
1facf9fc 28713+int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
523b37e3
AM
28714+ struct inode *hdir, struct path *path,
28715+ struct inode **delegated_inode);
1facf9fc 28716+int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
28717+int vfsub_rmdir(struct inode *dir, struct path *path);
28718+
28719+/* ---------------------------------------------------------------------- */
28720+
28721+ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
28722+ loff_t *ppos);
28723+ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
28724+ loff_t *ppos);
28725+ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
28726+ loff_t *ppos);
28727+ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
28728+ loff_t *ppos);
4a4d8108 28729+int vfsub_flush(struct file *file, fl_owner_t id);
392086de
AM
28730+int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
28731+
28732+/* just for type-check */
28733+static inline filldir_t au_diractor(int (*func)(struct dir_context *,
28734+ const char *, int, loff_t, u64,
28735+ unsigned))
28736+{
28737+ return (filldir_t)func;
28738+}
28739+
1facf9fc 28740+
c06a8ce3
AM
28741+static inline loff_t vfsub_f_size_read(struct file *file)
28742+{
28743+ return i_size_read(file_inode(file));
28744+}
28745+
4a4d8108
AM
28746+static inline unsigned int vfsub_file_flags(struct file *file)
28747+{
28748+ unsigned int flags;
28749+
28750+ spin_lock(&file->f_lock);
28751+ flags = file->f_flags;
28752+ spin_unlock(&file->f_lock);
28753+
28754+ return flags;
28755+}
1308ab2a 28756+
1facf9fc 28757+static inline void vfsub_file_accessed(struct file *h_file)
28758+{
28759+ file_accessed(h_file);
28760+ vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
28761+}
28762+
28763+static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
28764+ struct dentry *h_dentry)
28765+{
28766+ struct path h_path = {
28767+ .dentry = h_dentry,
28768+ .mnt = h_mnt
28769+ };
92d182d2 28770+ touch_atime(&h_path);
1facf9fc 28771+ vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
28772+}
28773+
0c3ec466
AM
28774+static inline int vfsub_update_time(struct inode *h_inode, struct timespec *ts,
28775+ int flags)
28776+{
28777+ return update_time(h_inode, ts, flags);
28778+ /* no vfsub_update_h_iattr() since we don't have struct path */
28779+}
28780+
4a4d8108
AM
28781+long vfsub_splice_to(struct file *in, loff_t *ppos,
28782+ struct pipe_inode_info *pipe, size_t len,
28783+ unsigned int flags);
28784+long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
28785+ loff_t *ppos, size_t len, unsigned int flags);
c06a8ce3
AM
28786+
28787+static inline long vfsub_truncate(struct path *path, loff_t length)
28788+{
28789+ long err;
28790+ lockdep_off();
28791+ err = vfs_truncate(path, length);
28792+ lockdep_on();
28793+ return err;
28794+}
28795+
4a4d8108
AM
28796+int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
28797+ struct file *h_file);
53392da6 28798+int vfsub_fsync(struct file *file, struct path *path, int datasync);
4a4d8108 28799+
1facf9fc 28800+/* ---------------------------------------------------------------------- */
28801+
28802+static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
28803+{
28804+ loff_t err;
28805+
2cbb1c4b 28806+ lockdep_off();
1facf9fc 28807+ err = vfs_llseek(file, offset, origin);
2cbb1c4b 28808+ lockdep_on();
1facf9fc 28809+ return err;
28810+}
28811+
28812+/* ---------------------------------------------------------------------- */
28813+
28814+/* dirty workaround for strict type of fmode_t */
28815+union vfsub_fmu {
28816+ fmode_t fm;
28817+ unsigned int ui;
28818+};
28819+
28820+static inline unsigned int vfsub_fmode_to_uint(fmode_t fm)
28821+{
28822+ union vfsub_fmu u = {
28823+ .fm = fm
28824+ };
28825+
28826+ BUILD_BUG_ON(sizeof(u.fm) != sizeof(u.ui));
28827+
28828+ return u.ui;
28829+}
28830+
28831+static inline fmode_t vfsub_uint_to_fmode(unsigned int ui)
28832+{
28833+ union vfsub_fmu u = {
28834+ .ui = ui
28835+ };
28836+
28837+ return u.fm;
28838+}
28839+
4a4d8108
AM
28840+/* ---------------------------------------------------------------------- */
28841+
28842+int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
28843+int vfsub_sio_rmdir(struct inode *dir, struct path *path);
523b37e3
AM
28844+int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
28845+ struct inode **delegated_inode);
28846+int vfsub_notify_change(struct path *path, struct iattr *ia,
28847+ struct inode **delegated_inode);
28848+int vfsub_unlink(struct inode *dir, struct path *path,
28849+ struct inode **delegated_inode, int force);
4a4d8108 28850+
1facf9fc 28851+#endif /* __KERNEL__ */
28852+#endif /* __AUFS_VFSUB_H__ */
7f207e10
AM
28853diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
28854--- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 28855+++ linux/fs/aufs/wbr_policy.c 2014-06-24 00:14:06.789386337 +0200
392086de 28856@@ -0,0 +1,756 @@
1facf9fc 28857+/*
523b37e3 28858+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 28859+ *
28860+ * This program, aufs is free software; you can redistribute it and/or modify
28861+ * it under the terms of the GNU General Public License as published by
28862+ * the Free Software Foundation; either version 2 of the License, or
28863+ * (at your option) any later version.
dece6358
AM
28864+ *
28865+ * This program is distributed in the hope that it will be useful,
28866+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28867+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28868+ * GNU General Public License for more details.
28869+ *
28870+ * You should have received a copy of the GNU General Public License
523b37e3 28871+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 28872+ */
28873+
28874+/*
28875+ * policies for selecting one among multiple writable branches
28876+ */
28877+
28878+#include <linux/statfs.h>
28879+#include "aufs.h"
28880+
28881+/* subset of cpup_attr() */
28882+static noinline_for_stack
28883+int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
28884+{
28885+ int err, sbits;
28886+ struct iattr ia;
28887+ struct inode *h_isrc;
28888+
28889+ h_isrc = h_src->d_inode;
28890+ ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
28891+ ia.ia_mode = h_isrc->i_mode;
28892+ ia.ia_uid = h_isrc->i_uid;
28893+ ia.ia_gid = h_isrc->i_gid;
28894+ sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
86dc4139 28895+ au_cpup_attr_flags(h_path->dentry->d_inode, h_isrc->i_flags);
523b37e3
AM
28896+ /* no delegation since it is just created */
28897+ err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
1facf9fc 28898+
28899+ /* is this nfs only? */
28900+ if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
28901+ ia.ia_valid = ATTR_FORCE | ATTR_MODE;
28902+ ia.ia_mode = h_isrc->i_mode;
523b37e3 28903+ err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
1facf9fc 28904+ }
28905+
28906+ return err;
28907+}
28908+
28909+#define AuCpdown_PARENT_OPQ 1
28910+#define AuCpdown_WHED (1 << 1)
28911+#define AuCpdown_MADE_DIR (1 << 2)
28912+#define AuCpdown_DIROPQ (1 << 3)
28913+#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
7f207e10
AM
28914+#define au_fset_cpdown(flags, name) \
28915+ do { (flags) |= AuCpdown_##name; } while (0)
28916+#define au_fclr_cpdown(flags, name) \
28917+ do { (flags) &= ~AuCpdown_##name; } while (0)
1facf9fc 28918+
1facf9fc 28919+static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
c2b27bf2 28920+ unsigned int *flags)
1facf9fc 28921+{
28922+ int err;
28923+ struct dentry *opq_dentry;
28924+
28925+ opq_dentry = au_diropq_create(dentry, bdst);
28926+ err = PTR_ERR(opq_dentry);
28927+ if (IS_ERR(opq_dentry))
28928+ goto out;
28929+ dput(opq_dentry);
c2b27bf2 28930+ au_fset_cpdown(*flags, DIROPQ);
1facf9fc 28931+
4f0767ce 28932+out:
1facf9fc 28933+ return err;
28934+}
28935+
28936+static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
28937+ struct inode *dir, aufs_bindex_t bdst)
28938+{
28939+ int err;
28940+ struct path h_path;
28941+ struct au_branch *br;
28942+
28943+ br = au_sbr(dentry->d_sb, bdst);
28944+ h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
28945+ err = PTR_ERR(h_path.dentry);
28946+ if (IS_ERR(h_path.dentry))
28947+ goto out;
28948+
28949+ err = 0;
28950+ if (h_path.dentry->d_inode) {
86dc4139 28951+ h_path.mnt = au_br_mnt(br);
1facf9fc 28952+ err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
28953+ dentry);
28954+ }
28955+ dput(h_path.dentry);
28956+
4f0767ce 28957+out:
1facf9fc 28958+ return err;
28959+}
28960+
28961+static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 28962+ struct au_pin *pin,
1facf9fc 28963+ struct dentry *h_parent, void *arg)
28964+{
28965+ int err, rerr;
4a4d8108 28966+ aufs_bindex_t bopq, bstart;
1facf9fc 28967+ struct path h_path;
28968+ struct dentry *parent;
28969+ struct inode *h_dir, *h_inode, *inode, *dir;
c2b27bf2 28970+ unsigned int *flags = arg;
1facf9fc 28971+
28972+ bstart = au_dbstart(dentry);
28973+ /* dentry is di-locked */
28974+ parent = dget_parent(dentry);
28975+ dir = parent->d_inode;
28976+ h_dir = h_parent->d_inode;
28977+ AuDebugOn(h_dir != au_h_iptr(dir, bdst));
28978+ IMustLock(h_dir);
28979+
86dc4139 28980+ err = au_lkup_neg(dentry, bdst, /*wh*/0);
1facf9fc 28981+ if (unlikely(err < 0))
28982+ goto out;
28983+ h_path.dentry = au_h_dptr(dentry, bdst);
28984+ h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
28985+ err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path,
28986+ S_IRWXU | S_IRUGO | S_IXUGO);
28987+ if (unlikely(err))
28988+ goto out_put;
c2b27bf2 28989+ au_fset_cpdown(*flags, MADE_DIR);
1facf9fc 28990+
1facf9fc 28991+ bopq = au_dbdiropq(dentry);
c2b27bf2
AM
28992+ au_fclr_cpdown(*flags, WHED);
28993+ au_fclr_cpdown(*flags, DIROPQ);
1facf9fc 28994+ if (au_dbwh(dentry) == bdst)
c2b27bf2
AM
28995+ au_fset_cpdown(*flags, WHED);
28996+ if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
28997+ au_fset_cpdown(*flags, PARENT_OPQ);
1facf9fc 28998+ h_inode = h_path.dentry->d_inode;
28999+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
c2b27bf2
AM
29000+ if (au_ftest_cpdown(*flags, WHED)) {
29001+ err = au_cpdown_dir_opq(dentry, bdst, flags);
1facf9fc 29002+ if (unlikely(err)) {
29003+ mutex_unlock(&h_inode->i_mutex);
29004+ goto out_dir;
29005+ }
29006+ }
29007+
29008+ err = au_cpdown_attr(&h_path, au_h_dptr(dentry, bstart));
29009+ mutex_unlock(&h_inode->i_mutex);
29010+ if (unlikely(err))
29011+ goto out_opq;
29012+
c2b27bf2 29013+ if (au_ftest_cpdown(*flags, WHED)) {
1facf9fc 29014+ err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
29015+ if (unlikely(err))
29016+ goto out_opq;
29017+ }
29018+
29019+ inode = dentry->d_inode;
29020+ if (au_ibend(inode) < bdst)
29021+ au_set_ibend(inode, bdst);
29022+ au_set_h_iptr(inode, bdst, au_igrab(h_inode),
29023+ au_hi_flags(inode, /*isdir*/1));
29024+ goto out; /* success */
29025+
29026+ /* revert */
4f0767ce 29027+out_opq:
c2b27bf2 29028+ if (au_ftest_cpdown(*flags, DIROPQ)) {
1facf9fc 29029+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
29030+ rerr = au_diropq_remove(dentry, bdst);
29031+ mutex_unlock(&h_inode->i_mutex);
29032+ if (unlikely(rerr)) {
523b37e3
AM
29033+ AuIOErr("failed removing diropq for %pd b%d (%d)\n",
29034+ dentry, bdst, rerr);
1facf9fc 29035+ err = -EIO;
29036+ goto out;
29037+ }
29038+ }
4f0767ce 29039+out_dir:
c2b27bf2 29040+ if (au_ftest_cpdown(*flags, MADE_DIR)) {
1facf9fc 29041+ rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
29042+ if (unlikely(rerr)) {
523b37e3
AM
29043+ AuIOErr("failed removing %pd b%d (%d)\n",
29044+ dentry, bdst, rerr);
1facf9fc 29045+ err = -EIO;
29046+ }
29047+ }
4f0767ce 29048+out_put:
1facf9fc 29049+ au_set_h_dptr(dentry, bdst, NULL);
29050+ if (au_dbend(dentry) == bdst)
29051+ au_update_dbend(dentry);
4f0767ce 29052+out:
1facf9fc 29053+ dput(parent);
29054+ return err;
29055+}
29056+
29057+int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
29058+{
29059+ int err;
c2b27bf2 29060+ unsigned int flags;
1facf9fc 29061+
c2b27bf2
AM
29062+ flags = 0;
29063+ err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
1facf9fc 29064+
29065+ return err;
29066+}
29067+
29068+/* ---------------------------------------------------------------------- */
29069+
29070+/* policies for create */
29071+
c2b27bf2 29072+int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
4a4d8108
AM
29073+{
29074+ int err, i, j, ndentry;
29075+ aufs_bindex_t bopq;
29076+ struct au_dcsub_pages dpages;
29077+ struct au_dpage *dpage;
29078+ struct dentry **dentries, *parent, *d;
29079+
29080+ err = au_dpages_init(&dpages, GFP_NOFS);
29081+ if (unlikely(err))
29082+ goto out;
29083+ parent = dget_parent(dentry);
027c5e7a 29084+ err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
4a4d8108
AM
29085+ if (unlikely(err))
29086+ goto out_free;
29087+
29088+ err = bindex;
29089+ for (i = 0; i < dpages.ndpage; i++) {
29090+ dpage = dpages.dpages + i;
29091+ dentries = dpage->dentries;
29092+ ndentry = dpage->ndentry;
29093+ for (j = 0; j < ndentry; j++) {
29094+ d = dentries[j];
29095+ di_read_lock_parent2(d, !AuLock_IR);
29096+ bopq = au_dbdiropq(d);
29097+ di_read_unlock(d, !AuLock_IR);
29098+ if (bopq >= 0 && bopq < err)
29099+ err = bopq;
29100+ }
29101+ }
29102+
29103+out_free:
29104+ dput(parent);
29105+ au_dpages_free(&dpages);
29106+out:
29107+ return err;
29108+}
29109+
1facf9fc 29110+static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
29111+{
29112+ for (; bindex >= 0; bindex--)
29113+ if (!au_br_rdonly(au_sbr(sb, bindex)))
29114+ return bindex;
29115+ return -EROFS;
29116+}
29117+
29118+/* top down parent */
392086de
AM
29119+static int au_wbr_create_tdp(struct dentry *dentry,
29120+ unsigned int flags __maybe_unused)
1facf9fc 29121+{
29122+ int err;
29123+ aufs_bindex_t bstart, bindex;
29124+ struct super_block *sb;
29125+ struct dentry *parent, *h_parent;
29126+
29127+ sb = dentry->d_sb;
29128+ bstart = au_dbstart(dentry);
29129+ err = bstart;
29130+ if (!au_br_rdonly(au_sbr(sb, bstart)))
29131+ goto out;
29132+
29133+ err = -EROFS;
29134+ parent = dget_parent(dentry);
29135+ for (bindex = au_dbstart(parent); bindex < bstart; bindex++) {
29136+ h_parent = au_h_dptr(parent, bindex);
29137+ if (!h_parent || !h_parent->d_inode)
29138+ continue;
29139+
29140+ if (!au_br_rdonly(au_sbr(sb, bindex))) {
29141+ err = bindex;
29142+ break;
29143+ }
29144+ }
29145+ dput(parent);
29146+
29147+ /* bottom up here */
4a4d8108 29148+ if (unlikely(err < 0)) {
1facf9fc 29149+ err = au_wbr_bu(sb, bstart - 1);
4a4d8108
AM
29150+ if (err >= 0)
29151+ err = au_wbr_nonopq(dentry, err);
29152+ }
1facf9fc 29153+
4f0767ce 29154+out:
1facf9fc 29155+ AuDbg("b%d\n", err);
29156+ return err;
29157+}
29158+
29159+/* ---------------------------------------------------------------------- */
29160+
29161+/* an exception for the policy other than tdp */
29162+static int au_wbr_create_exp(struct dentry *dentry)
29163+{
29164+ int err;
29165+ aufs_bindex_t bwh, bdiropq;
29166+ struct dentry *parent;
29167+
29168+ err = -1;
29169+ bwh = au_dbwh(dentry);
29170+ parent = dget_parent(dentry);
29171+ bdiropq = au_dbdiropq(parent);
29172+ if (bwh >= 0) {
29173+ if (bdiropq >= 0)
29174+ err = min(bdiropq, bwh);
29175+ else
29176+ err = bwh;
29177+ AuDbg("%d\n", err);
29178+ } else if (bdiropq >= 0) {
29179+ err = bdiropq;
29180+ AuDbg("%d\n", err);
29181+ }
29182+ dput(parent);
29183+
4a4d8108
AM
29184+ if (err >= 0)
29185+ err = au_wbr_nonopq(dentry, err);
29186+
1facf9fc 29187+ if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
29188+ err = -1;
29189+
29190+ AuDbg("%d\n", err);
29191+ return err;
29192+}
29193+
29194+/* ---------------------------------------------------------------------- */
29195+
29196+/* round robin */
29197+static int au_wbr_create_init_rr(struct super_block *sb)
29198+{
29199+ int err;
29200+
29201+ err = au_wbr_bu(sb, au_sbend(sb));
29202+ atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
dece6358 29203+ /* smp_mb(); */
1facf9fc 29204+
29205+ AuDbg("b%d\n", err);
29206+ return err;
29207+}
29208+
392086de 29209+static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
1facf9fc 29210+{
29211+ int err, nbr;
29212+ unsigned int u;
29213+ aufs_bindex_t bindex, bend;
29214+ struct super_block *sb;
29215+ atomic_t *next;
29216+
29217+ err = au_wbr_create_exp(dentry);
29218+ if (err >= 0)
29219+ goto out;
29220+
29221+ sb = dentry->d_sb;
29222+ next = &au_sbi(sb)->si_wbr_rr_next;
29223+ bend = au_sbend(sb);
29224+ nbr = bend + 1;
29225+ for (bindex = 0; bindex <= bend; bindex++) {
392086de 29226+ if (!au_ftest_wbr(flags, DIR)) {
1facf9fc 29227+ err = atomic_dec_return(next) + 1;
29228+ /* modulo for 0 is meaningless */
29229+ if (unlikely(!err))
29230+ err = atomic_dec_return(next) + 1;
29231+ } else
29232+ err = atomic_read(next);
29233+ AuDbg("%d\n", err);
29234+ u = err;
29235+ err = u % nbr;
29236+ AuDbg("%d\n", err);
29237+ if (!au_br_rdonly(au_sbr(sb, err)))
29238+ break;
29239+ err = -EROFS;
29240+ }
29241+
4a4d8108
AM
29242+ if (err >= 0)
29243+ err = au_wbr_nonopq(dentry, err);
29244+
4f0767ce 29245+out:
1facf9fc 29246+ AuDbg("%d\n", err);
29247+ return err;
29248+}
29249+
29250+/* ---------------------------------------------------------------------- */
29251+
29252+/* most free space */
392086de 29253+static void au_mfs(struct dentry *dentry, struct dentry *parent)
1facf9fc 29254+{
29255+ struct super_block *sb;
29256+ struct au_branch *br;
29257+ struct au_wbr_mfs *mfs;
392086de 29258+ struct dentry *h_parent;
1facf9fc 29259+ aufs_bindex_t bindex, bend;
29260+ int err;
29261+ unsigned long long b, bavail;
7f207e10 29262+ struct path h_path;
1facf9fc 29263+ /* reduce the stack usage */
29264+ struct kstatfs *st;
29265+
29266+ st = kmalloc(sizeof(*st), GFP_NOFS);
29267+ if (unlikely(!st)) {
29268+ AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
29269+ return;
29270+ }
29271+
29272+ bavail = 0;
29273+ sb = dentry->d_sb;
29274+ mfs = &au_sbi(sb)->si_wbr_mfs;
dece6358 29275+ MtxMustLock(&mfs->mfs_lock);
1facf9fc 29276+ mfs->mfs_bindex = -EROFS;
29277+ mfs->mfsrr_bytes = 0;
392086de
AM
29278+ if (!parent) {
29279+ bindex = 0;
29280+ bend = au_sbend(sb);
29281+ } else {
29282+ bindex = au_dbstart(parent);
29283+ bend = au_dbtaildir(parent);
29284+ }
29285+
29286+ for (; bindex <= bend; bindex++) {
29287+ if (parent) {
29288+ h_parent = au_h_dptr(parent, bindex);
29289+ if (!h_parent || !h_parent->d_inode)
29290+ continue;
29291+ }
1facf9fc 29292+ br = au_sbr(sb, bindex);
29293+ if (au_br_rdonly(br))
29294+ continue;
29295+
29296+ /* sb->s_root for NFS is unreliable */
86dc4139 29297+ h_path.mnt = au_br_mnt(br);
7f207e10
AM
29298+ h_path.dentry = h_path.mnt->mnt_root;
29299+ err = vfs_statfs(&h_path, st);
1facf9fc 29300+ if (unlikely(err)) {
29301+ AuWarn1("failed statfs, b%d, %d\n", bindex, err);
29302+ continue;
29303+ }
29304+
29305+ /* when the available size is equal, select the lower one */
29306+ BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
29307+ || sizeof(b) < sizeof(st->f_bsize));
29308+ b = st->f_bavail * st->f_bsize;
29309+ br->br_wbr->wbr_bytes = b;
29310+ if (b >= bavail) {
29311+ bavail = b;
29312+ mfs->mfs_bindex = bindex;
29313+ mfs->mfs_jiffy = jiffies;
29314+ }
29315+ }
29316+
29317+ mfs->mfsrr_bytes = bavail;
29318+ AuDbg("b%d\n", mfs->mfs_bindex);
29319+ kfree(st);
29320+}
29321+
392086de 29322+static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
1facf9fc 29323+{
29324+ int err;
392086de 29325+ struct dentry *parent;
1facf9fc 29326+ struct super_block *sb;
29327+ struct au_wbr_mfs *mfs;
29328+
29329+ err = au_wbr_create_exp(dentry);
29330+ if (err >= 0)
29331+ goto out;
29332+
29333+ sb = dentry->d_sb;
392086de
AM
29334+ parent = NULL;
29335+ if (au_ftest_wbr(flags, PARENT))
29336+ parent = dget_parent(dentry);
1facf9fc 29337+ mfs = &au_sbi(sb)->si_wbr_mfs;
29338+ mutex_lock(&mfs->mfs_lock);
29339+ if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
29340+ || mfs->mfs_bindex < 0
29341+ || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
392086de 29342+ au_mfs(dentry, parent);
1facf9fc 29343+ mutex_unlock(&mfs->mfs_lock);
29344+ err = mfs->mfs_bindex;
392086de 29345+ dput(parent);
1facf9fc 29346+
4a4d8108
AM
29347+ if (err >= 0)
29348+ err = au_wbr_nonopq(dentry, err);
29349+
4f0767ce 29350+out:
1facf9fc 29351+ AuDbg("b%d\n", err);
29352+ return err;
29353+}
29354+
29355+static int au_wbr_create_init_mfs(struct super_block *sb)
29356+{
29357+ struct au_wbr_mfs *mfs;
29358+
29359+ mfs = &au_sbi(sb)->si_wbr_mfs;
29360+ mutex_init(&mfs->mfs_lock);
29361+ mfs->mfs_jiffy = 0;
29362+ mfs->mfs_bindex = -EROFS;
29363+
29364+ return 0;
29365+}
29366+
29367+static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
29368+{
29369+ mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
29370+ return 0;
29371+}
29372+
29373+/* ---------------------------------------------------------------------- */
29374+
29375+/* most free space and then round robin */
392086de 29376+static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
1facf9fc 29377+{
29378+ int err;
29379+ struct au_wbr_mfs *mfs;
29380+
392086de 29381+ err = au_wbr_create_mfs(dentry, flags);
1facf9fc 29382+ if (err >= 0) {
29383+ mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
dece6358 29384+ mutex_lock(&mfs->mfs_lock);
1facf9fc 29385+ if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
392086de 29386+ err = au_wbr_create_rr(dentry, flags);
dece6358 29387+ mutex_unlock(&mfs->mfs_lock);
1facf9fc 29388+ }
29389+
29390+ AuDbg("b%d\n", err);
29391+ return err;
29392+}
29393+
29394+static int au_wbr_create_init_mfsrr(struct super_block *sb)
29395+{
29396+ int err;
29397+
29398+ au_wbr_create_init_mfs(sb); /* ignore */
29399+ err = au_wbr_create_init_rr(sb);
29400+
29401+ return err;
29402+}
29403+
29404+/* ---------------------------------------------------------------------- */
29405+
29406+/* top down parent and most free space */
392086de 29407+static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
1facf9fc 29408+{
29409+ int err, e2;
29410+ unsigned long long b;
29411+ aufs_bindex_t bindex, bstart, bend;
29412+ struct super_block *sb;
29413+ struct dentry *parent, *h_parent;
29414+ struct au_branch *br;
29415+
392086de 29416+ err = au_wbr_create_tdp(dentry, flags);
1facf9fc 29417+ if (unlikely(err < 0))
29418+ goto out;
29419+ parent = dget_parent(dentry);
29420+ bstart = au_dbstart(parent);
29421+ bend = au_dbtaildir(parent);
29422+ if (bstart == bend)
29423+ goto out_parent; /* success */
29424+
392086de 29425+ e2 = au_wbr_create_mfs(dentry, flags);
1facf9fc 29426+ if (e2 < 0)
29427+ goto out_parent; /* success */
29428+
29429+ /* when the available size is equal, select upper one */
29430+ sb = dentry->d_sb;
29431+ br = au_sbr(sb, err);
29432+ b = br->br_wbr->wbr_bytes;
29433+ AuDbg("b%d, %llu\n", err, b);
29434+
29435+ for (bindex = bstart; bindex <= bend; bindex++) {
29436+ h_parent = au_h_dptr(parent, bindex);
29437+ if (!h_parent || !h_parent->d_inode)
29438+ continue;
29439+
29440+ br = au_sbr(sb, bindex);
29441+ if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
29442+ b = br->br_wbr->wbr_bytes;
29443+ err = bindex;
29444+ AuDbg("b%d, %llu\n", err, b);
29445+ }
29446+ }
29447+
4a4d8108
AM
29448+ if (err >= 0)
29449+ err = au_wbr_nonopq(dentry, err);
29450+
4f0767ce 29451+out_parent:
1facf9fc 29452+ dput(parent);
4f0767ce 29453+out:
1facf9fc 29454+ AuDbg("b%d\n", err);
29455+ return err;
29456+}
29457+
29458+/* ---------------------------------------------------------------------- */
29459+
392086de
AM
29460+/*
29461+ * - top down parent
29462+ * - most free space with parent
29463+ * - most free space round-robin regardless parent
29464+ */
29465+static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
29466+{
29467+ int err;
29468+ unsigned long long watermark;
29469+ struct super_block *sb;
29470+ struct au_branch *br;
29471+ struct au_wbr_mfs *mfs;
29472+
29473+ err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
29474+ if (unlikely(err < 0))
29475+ goto out;
29476+
29477+ sb = dentry->d_sb;
29478+ br = au_sbr(sb, err);
29479+ mfs = &au_sbi(sb)->si_wbr_mfs;
29480+ mutex_lock(&mfs->mfs_lock);
29481+ watermark = mfs->mfsrr_watermark;
29482+ mutex_unlock(&mfs->mfs_lock);
29483+ if (br->br_wbr->wbr_bytes < watermark)
29484+ /* regardless the parent dir */
29485+ err = au_wbr_create_mfsrr(dentry, flags);
29486+
29487+out:
29488+ AuDbg("b%d\n", err);
29489+ return err;
29490+}
29491+
29492+/* ---------------------------------------------------------------------- */
29493+
1facf9fc 29494+/* policies for copyup */
29495+
29496+/* top down parent */
29497+static int au_wbr_copyup_tdp(struct dentry *dentry)
29498+{
392086de 29499+ return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
1facf9fc 29500+}
29501+
29502+/* bottom up parent */
29503+static int au_wbr_copyup_bup(struct dentry *dentry)
29504+{
29505+ int err;
29506+ aufs_bindex_t bindex, bstart;
29507+ struct dentry *parent, *h_parent;
29508+ struct super_block *sb;
29509+
29510+ err = -EROFS;
29511+ sb = dentry->d_sb;
29512+ parent = dget_parent(dentry);
29513+ bstart = au_dbstart(parent);
29514+ for (bindex = au_dbstart(dentry); bindex >= bstart; bindex--) {
29515+ h_parent = au_h_dptr(parent, bindex);
29516+ if (!h_parent || !h_parent->d_inode)
29517+ continue;
29518+
29519+ if (!au_br_rdonly(au_sbr(sb, bindex))) {
29520+ err = bindex;
29521+ break;
29522+ }
29523+ }
29524+ dput(parent);
29525+
29526+ /* bottom up here */
29527+ if (unlikely(err < 0))
29528+ err = au_wbr_bu(sb, bstart - 1);
29529+
29530+ AuDbg("b%d\n", err);
29531+ return err;
29532+}
29533+
29534+/* bottom up */
29535+static int au_wbr_copyup_bu(struct dentry *dentry)
29536+{
29537+ int err;
4a4d8108 29538+ aufs_bindex_t bstart;
1facf9fc 29539+
4a4d8108
AM
29540+ bstart = au_dbstart(dentry);
29541+ err = au_wbr_bu(dentry->d_sb, bstart);
29542+ AuDbg("b%d\n", err);
29543+ if (err > bstart)
29544+ err = au_wbr_nonopq(dentry, err);
1facf9fc 29545+
29546+ AuDbg("b%d\n", err);
29547+ return err;
29548+}
29549+
29550+/* ---------------------------------------------------------------------- */
29551+
29552+struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
29553+ [AuWbrCopyup_TDP] = {
29554+ .copyup = au_wbr_copyup_tdp
29555+ },
29556+ [AuWbrCopyup_BUP] = {
29557+ .copyup = au_wbr_copyup_bup
29558+ },
29559+ [AuWbrCopyup_BU] = {
29560+ .copyup = au_wbr_copyup_bu
29561+ }
29562+};
29563+
29564+struct au_wbr_create_operations au_wbr_create_ops[] = {
29565+ [AuWbrCreate_TDP] = {
29566+ .create = au_wbr_create_tdp
29567+ },
29568+ [AuWbrCreate_RR] = {
29569+ .create = au_wbr_create_rr,
29570+ .init = au_wbr_create_init_rr
29571+ },
29572+ [AuWbrCreate_MFS] = {
29573+ .create = au_wbr_create_mfs,
29574+ .init = au_wbr_create_init_mfs,
29575+ .fin = au_wbr_create_fin_mfs
29576+ },
29577+ [AuWbrCreate_MFSV] = {
29578+ .create = au_wbr_create_mfs,
29579+ .init = au_wbr_create_init_mfs,
29580+ .fin = au_wbr_create_fin_mfs
29581+ },
29582+ [AuWbrCreate_MFSRR] = {
29583+ .create = au_wbr_create_mfsrr,
29584+ .init = au_wbr_create_init_mfsrr,
29585+ .fin = au_wbr_create_fin_mfs
29586+ },
29587+ [AuWbrCreate_MFSRRV] = {
29588+ .create = au_wbr_create_mfsrr,
29589+ .init = au_wbr_create_init_mfsrr,
29590+ .fin = au_wbr_create_fin_mfs
29591+ },
29592+ [AuWbrCreate_PMFS] = {
29593+ .create = au_wbr_create_pmfs,
29594+ .init = au_wbr_create_init_mfs,
29595+ .fin = au_wbr_create_fin_mfs
29596+ },
29597+ [AuWbrCreate_PMFSV] = {
29598+ .create = au_wbr_create_pmfs,
29599+ .init = au_wbr_create_init_mfs,
29600+ .fin = au_wbr_create_fin_mfs
392086de
AM
29601+ },
29602+ [AuWbrCreate_PMFSRR] = {
29603+ .create = au_wbr_create_pmfsrr,
29604+ .init = au_wbr_create_init_mfsrr,
29605+ .fin = au_wbr_create_fin_mfs
29606+ },
29607+ [AuWbrCreate_PMFSRRV] = {
29608+ .create = au_wbr_create_pmfsrr,
29609+ .init = au_wbr_create_init_mfsrr,
29610+ .fin = au_wbr_create_fin_mfs
1facf9fc 29611+ }
29612+};
7f207e10
AM
29613diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
29614--- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
38d290e6 29615+++ linux/fs/aufs/whout.c 2014-06-24 00:14:06.789386337 +0200
523b37e3 29616@@ -0,0 +1,1052 @@
1facf9fc 29617+/*
523b37e3 29618+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 29619+ *
29620+ * This program, aufs is free software; you can redistribute it and/or modify
29621+ * it under the terms of the GNU General Public License as published by
29622+ * the Free Software Foundation; either version 2 of the License, or
29623+ * (at your option) any later version.
dece6358
AM
29624+ *
29625+ * This program is distributed in the hope that it will be useful,
29626+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
29627+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29628+ * GNU General Public License for more details.
29629+ *
29630+ * You should have received a copy of the GNU General Public License
523b37e3 29631+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 29632+ */
29633+
29634+/*
29635+ * whiteout for logical deletion and opaque directory
29636+ */
29637+
1facf9fc 29638+#include "aufs.h"
29639+
29640+#define WH_MASK S_IRUGO
29641+
29642+/*
29643+ * If a directory contains this file, then it is opaque. We start with the
29644+ * .wh. flag so that it is blocked by lookup.
29645+ */
0c3ec466
AM
29646+static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
29647+ sizeof(AUFS_WH_DIROPQ) - 1);
1facf9fc 29648+
29649+/*
29650+ * generate whiteout name, which is NOT terminated by NULL.
29651+ * @name: original d_name.name
29652+ * @len: original d_name.len
29653+ * @wh: whiteout qstr
29654+ * returns zero when succeeds, otherwise error.
29655+ * succeeded value as wh->name should be freed by kfree().
29656+ */
29657+int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
29658+{
29659+ char *p;
29660+
29661+ if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
29662+ return -ENAMETOOLONG;
29663+
29664+ wh->len = name->len + AUFS_WH_PFX_LEN;
29665+ p = kmalloc(wh->len, GFP_NOFS);
29666+ wh->name = p;
29667+ if (p) {
29668+ memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
29669+ memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
29670+ /* smp_mb(); */
29671+ return 0;
29672+ }
29673+ return -ENOMEM;
29674+}
29675+
29676+/* ---------------------------------------------------------------------- */
29677+
29678+/*
29679+ * test if the @wh_name exists under @h_parent.
29680+ * @try_sio specifies the necessary of super-io.
29681+ */
29682+int au_wh_test(struct dentry *h_parent, struct qstr *wh_name,
29683+ struct au_branch *br, int try_sio)
29684+{
29685+ int err;
29686+ struct dentry *wh_dentry;
1facf9fc 29687+
1facf9fc 29688+ if (!try_sio)
b4510431 29689+ wh_dentry = vfsub_lkup_one(wh_name, h_parent);
1facf9fc 29690+ else
29691+ wh_dentry = au_sio_lkup_one(wh_name, h_parent, br);
29692+ err = PTR_ERR(wh_dentry);
29693+ if (IS_ERR(wh_dentry))
29694+ goto out;
29695+
29696+ err = 0;
29697+ if (!wh_dentry->d_inode)
29698+ goto out_wh; /* success */
29699+
29700+ err = 1;
29701+ if (S_ISREG(wh_dentry->d_inode->i_mode))
29702+ goto out_wh; /* success */
29703+
29704+ err = -EIO;
523b37e3
AM
29705+ AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
29706+ wh_dentry, wh_dentry->d_inode->i_mode);
1facf9fc 29707+
4f0767ce 29708+out_wh:
1facf9fc 29709+ dput(wh_dentry);
4f0767ce 29710+out:
1facf9fc 29711+ return err;
29712+}
29713+
29714+/*
29715+ * test if the @h_dentry sets opaque or not.
29716+ */
29717+int au_diropq_test(struct dentry *h_dentry, struct au_branch *br)
29718+{
29719+ int err;
29720+ struct inode *h_dir;
29721+
29722+ h_dir = h_dentry->d_inode;
29723+ err = au_wh_test(h_dentry, &diropq_name, br,
29724+ au_test_h_perm_sio(h_dir, MAY_EXEC));
29725+ return err;
29726+}
29727+
29728+/*
29729+ * returns a negative dentry whose name is unique and temporary.
29730+ */
29731+struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
29732+ struct qstr *prefix)
29733+{
1facf9fc 29734+ struct dentry *dentry;
29735+ int i;
027c5e7a 29736+ char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
4a4d8108 29737+ *name, *p;
027c5e7a 29738+ /* strict atomic_t is unnecessary here */
1facf9fc 29739+ static unsigned short cnt;
29740+ struct qstr qs;
29741+
4a4d8108
AM
29742+ BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
29743+
1facf9fc 29744+ name = defname;
027c5e7a
AM
29745+ qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
29746+ if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
1facf9fc 29747+ dentry = ERR_PTR(-ENAMETOOLONG);
4a4d8108 29748+ if (unlikely(qs.len > NAME_MAX))
1facf9fc 29749+ goto out;
29750+ dentry = ERR_PTR(-ENOMEM);
29751+ name = kmalloc(qs.len + 1, GFP_NOFS);
29752+ if (unlikely(!name))
29753+ goto out;
29754+ }
29755+
29756+ /* doubly whiteout-ed */
29757+ memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
29758+ p = name + AUFS_WH_PFX_LEN * 2;
29759+ memcpy(p, prefix->name, prefix->len);
29760+ p += prefix->len;
29761+ *p++ = '.';
4a4d8108 29762+ AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
1facf9fc 29763+
29764+ qs.name = name;
29765+ for (i = 0; i < 3; i++) {
b752ccd1 29766+ sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
1facf9fc 29767+ dentry = au_sio_lkup_one(&qs, h_parent, br);
29768+ if (IS_ERR(dentry) || !dentry->d_inode)
29769+ goto out_name;
29770+ dput(dentry);
29771+ }
0c3ec466 29772+ /* pr_warn("could not get random name\n"); */
1facf9fc 29773+ dentry = ERR_PTR(-EEXIST);
29774+ AuDbg("%.*s\n", AuLNPair(&qs));
29775+ BUG();
29776+
4f0767ce 29777+out_name:
1facf9fc 29778+ if (name != defname)
29779+ kfree(name);
4f0767ce 29780+out:
4a4d8108 29781+ AuTraceErrPtr(dentry);
1facf9fc 29782+ return dentry;
1facf9fc 29783+}
29784+
29785+/*
29786+ * rename the @h_dentry on @br to the whiteouted temporary name.
29787+ */
29788+int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
29789+{
29790+ int err;
29791+ struct path h_path = {
86dc4139 29792+ .mnt = au_br_mnt(br)
1facf9fc 29793+ };
523b37e3 29794+ struct inode *h_dir, *delegated;
1facf9fc 29795+ struct dentry *h_parent;
29796+
29797+ h_parent = h_dentry->d_parent; /* dir inode is locked */
29798+ h_dir = h_parent->d_inode;
29799+ IMustLock(h_dir);
29800+
29801+ h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
29802+ err = PTR_ERR(h_path.dentry);
29803+ if (IS_ERR(h_path.dentry))
29804+ goto out;
29805+
29806+ /* under the same dir, no need to lock_rename() */
523b37e3
AM
29807+ delegated = NULL;
29808+ err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated);
1facf9fc 29809+ AuTraceErr(err);
523b37e3
AM
29810+ if (unlikely(err == -EWOULDBLOCK)) {
29811+ pr_warn("cannot retry for NFSv4 delegation"
29812+ " for an internal rename\n");
29813+ iput(delegated);
29814+ }
1facf9fc 29815+ dput(h_path.dentry);
29816+
4f0767ce 29817+out:
4a4d8108 29818+ AuTraceErr(err);
1facf9fc 29819+ return err;
29820+}
29821+
29822+/* ---------------------------------------------------------------------- */
29823+/*
29824+ * functions for removing a whiteout
29825+ */
29826+
29827+static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
29828+{
523b37e3
AM
29829+ int err, force;
29830+ struct inode *delegated;
1facf9fc 29831+
29832+ /*
29833+ * forces superio when the dir has a sticky bit.
29834+ * this may be a violation of unix fs semantics.
29835+ */
29836+ force = (h_dir->i_mode & S_ISVTX)
0c3ec466 29837+ && !uid_eq(current_fsuid(), h_path->dentry->d_inode->i_uid);
523b37e3
AM
29838+ delegated = NULL;
29839+ err = vfsub_unlink(h_dir, h_path, &delegated, force);
29840+ if (unlikely(err == -EWOULDBLOCK)) {
29841+ pr_warn("cannot retry for NFSv4 delegation"
29842+ " for an internal unlink\n");
29843+ iput(delegated);
29844+ }
29845+ return err;
1facf9fc 29846+}
29847+
29848+int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
29849+ struct dentry *dentry)
29850+{
29851+ int err;
29852+
29853+ err = do_unlink_wh(h_dir, h_path);
29854+ if (!err && dentry)
29855+ au_set_dbwh(dentry, -1);
29856+
29857+ return err;
29858+}
29859+
29860+static int unlink_wh_name(struct dentry *h_parent, struct qstr *wh,
29861+ struct au_branch *br)
29862+{
29863+ int err;
29864+ struct path h_path = {
86dc4139 29865+ .mnt = au_br_mnt(br)
1facf9fc 29866+ };
29867+
29868+ err = 0;
b4510431 29869+ h_path.dentry = vfsub_lkup_one(wh, h_parent);
1facf9fc 29870+ if (IS_ERR(h_path.dentry))
29871+ err = PTR_ERR(h_path.dentry);
29872+ else {
29873+ if (h_path.dentry->d_inode
29874+ && S_ISREG(h_path.dentry->d_inode->i_mode))
29875+ err = do_unlink_wh(h_parent->d_inode, &h_path);
29876+ dput(h_path.dentry);
29877+ }
29878+
29879+ return err;
29880+}
29881+
29882+/* ---------------------------------------------------------------------- */
29883+/*
29884+ * initialize/clean whiteout for a branch
29885+ */
29886+
29887+static void au_wh_clean(struct inode *h_dir, struct path *whpath,
29888+ const int isdir)
29889+{
29890+ int err;
523b37e3 29891+ struct inode *delegated;
1facf9fc 29892+
29893+ if (!whpath->dentry->d_inode)
29894+ return;
29895+
86dc4139
AM
29896+ if (isdir)
29897+ err = vfsub_rmdir(h_dir, whpath);
523b37e3
AM
29898+ else {
29899+ delegated = NULL;
29900+ err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
29901+ if (unlikely(err == -EWOULDBLOCK)) {
29902+ pr_warn("cannot retry for NFSv4 delegation"
29903+ " for an internal unlink\n");
29904+ iput(delegated);
29905+ }
29906+ }
1facf9fc 29907+ if (unlikely(err))
523b37e3
AM
29908+ pr_warn("failed removing %pd (%d), ignored.\n",
29909+ whpath->dentry, err);
1facf9fc 29910+}
29911+
29912+static int test_linkable(struct dentry *h_root)
29913+{
29914+ struct inode *h_dir = h_root->d_inode;
29915+
29916+ if (h_dir->i_op->link)
29917+ return 0;
29918+
523b37e3
AM
29919+ pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
29920+ h_root, au_sbtype(h_root->d_sb));
1facf9fc 29921+ return -ENOSYS;
29922+}
29923+
29924+/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
29925+static int au_whdir(struct inode *h_dir, struct path *path)
29926+{
29927+ int err;
29928+
29929+ err = -EEXIST;
29930+ if (!path->dentry->d_inode) {
29931+ int mode = S_IRWXU;
29932+
29933+ if (au_test_nfs(path->dentry->d_sb))
29934+ mode |= S_IXUGO;
86dc4139 29935+ err = vfsub_mkdir(h_dir, path, mode);
1facf9fc 29936+ } else if (S_ISDIR(path->dentry->d_inode->i_mode))
29937+ err = 0;
29938+ else
523b37e3 29939+ pr_err("unknown %pd exists\n", path->dentry);
1facf9fc 29940+
29941+ return err;
29942+}
29943+
29944+struct au_wh_base {
29945+ const struct qstr *name;
29946+ struct dentry *dentry;
29947+};
29948+
29949+static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
29950+ struct path *h_path)
29951+{
29952+ h_path->dentry = base[AuBrWh_BASE].dentry;
29953+ au_wh_clean(h_dir, h_path, /*isdir*/0);
29954+ h_path->dentry = base[AuBrWh_PLINK].dentry;
29955+ au_wh_clean(h_dir, h_path, /*isdir*/1);
29956+ h_path->dentry = base[AuBrWh_ORPH].dentry;
29957+ au_wh_clean(h_dir, h_path, /*isdir*/1);
29958+}
29959+
29960+/*
29961+ * returns tri-state,
29962+ * minus: error, caller should print the mesage
29963+ * zero: succuess
29964+ * plus: error, caller should NOT print the mesage
29965+ */
29966+static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
29967+ int do_plink, struct au_wh_base base[],
29968+ struct path *h_path)
29969+{
29970+ int err;
29971+ struct inode *h_dir;
29972+
29973+ h_dir = h_root->d_inode;
29974+ h_path->dentry = base[AuBrWh_BASE].dentry;
29975+ au_wh_clean(h_dir, h_path, /*isdir*/0);
29976+ h_path->dentry = base[AuBrWh_PLINK].dentry;
29977+ if (do_plink) {
29978+ err = test_linkable(h_root);
29979+ if (unlikely(err)) {
29980+ err = 1;
29981+ goto out;
29982+ }
29983+
29984+ err = au_whdir(h_dir, h_path);
29985+ if (unlikely(err))
29986+ goto out;
29987+ wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
29988+ } else
29989+ au_wh_clean(h_dir, h_path, /*isdir*/1);
29990+ h_path->dentry = base[AuBrWh_ORPH].dentry;
29991+ err = au_whdir(h_dir, h_path);
29992+ if (unlikely(err))
29993+ goto out;
29994+ wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
29995+
4f0767ce 29996+out:
1facf9fc 29997+ return err;
29998+}
29999+
30000+/*
30001+ * for the moment, aufs supports the branch filesystem which does not support
30002+ * link(2). testing on FAT which does not support i_op->setattr() fully either,
30003+ * copyup failed. finally, such filesystem will not be used as the writable
30004+ * branch.
30005+ *
30006+ * returns tri-state, see above.
30007+ */
30008+static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
30009+ int do_plink, struct au_wh_base base[],
30010+ struct path *h_path)
30011+{
30012+ int err;
30013+ struct inode *h_dir;
30014+
1308ab2a 30015+ WbrWhMustWriteLock(wbr);
30016+
1facf9fc 30017+ err = test_linkable(h_root);
30018+ if (unlikely(err)) {
30019+ err = 1;
30020+ goto out;
30021+ }
30022+
30023+ /*
30024+ * todo: should this create be done in /sbin/mount.aufs helper?
30025+ */
30026+ err = -EEXIST;
30027+ h_dir = h_root->d_inode;
30028+ if (!base[AuBrWh_BASE].dentry->d_inode) {
86dc4139
AM
30029+ h_path->dentry = base[AuBrWh_BASE].dentry;
30030+ err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
1facf9fc 30031+ } else if (S_ISREG(base[AuBrWh_BASE].dentry->d_inode->i_mode))
30032+ err = 0;
30033+ else
523b37e3 30034+ pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
1facf9fc 30035+ if (unlikely(err))
30036+ goto out;
30037+
30038+ h_path->dentry = base[AuBrWh_PLINK].dentry;
30039+ if (do_plink) {
30040+ err = au_whdir(h_dir, h_path);
30041+ if (unlikely(err))
30042+ goto out;
30043+ wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
30044+ } else
30045+ au_wh_clean(h_dir, h_path, /*isdir*/1);
30046+ wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
30047+
30048+ h_path->dentry = base[AuBrWh_ORPH].dentry;
30049+ err = au_whdir(h_dir, h_path);
30050+ if (unlikely(err))
30051+ goto out;
30052+ wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
30053+
4f0767ce 30054+out:
1facf9fc 30055+ return err;
30056+}
30057+
30058+/*
30059+ * initialize the whiteout base file/dir for @br.
30060+ */
86dc4139 30061+int au_wh_init(struct au_branch *br, struct super_block *sb)
1facf9fc 30062+{
30063+ int err, i;
30064+ const unsigned char do_plink
30065+ = !!au_opt_test(au_mntflags(sb), PLINK);
1facf9fc 30066+ struct inode *h_dir;
86dc4139
AM
30067+ struct path path = br->br_path;
30068+ struct dentry *h_root = path.dentry;
1facf9fc 30069+ struct au_wbr *wbr = br->br_wbr;
30070+ static const struct qstr base_name[] = {
0c3ec466
AM
30071+ [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
30072+ sizeof(AUFS_BASE_NAME) - 1),
30073+ [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
30074+ sizeof(AUFS_PLINKDIR_NAME) - 1),
30075+ [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
30076+ sizeof(AUFS_ORPHDIR_NAME) - 1)
1facf9fc 30077+ };
30078+ struct au_wh_base base[] = {
30079+ [AuBrWh_BASE] = {
30080+ .name = base_name + AuBrWh_BASE,
30081+ .dentry = NULL
30082+ },
30083+ [AuBrWh_PLINK] = {
30084+ .name = base_name + AuBrWh_PLINK,
30085+ .dentry = NULL
30086+ },
30087+ [AuBrWh_ORPH] = {
30088+ .name = base_name + AuBrWh_ORPH,
30089+ .dentry = NULL
30090+ }
30091+ };
30092+
1308ab2a 30093+ if (wbr)
30094+ WbrWhMustWriteLock(wbr);
1facf9fc 30095+
1facf9fc 30096+ for (i = 0; i < AuBrWh_Last; i++) {
30097+ /* doubly whiteouted */
30098+ struct dentry *d;
30099+
30100+ d = au_wh_lkup(h_root, (void *)base[i].name, br);
30101+ err = PTR_ERR(d);
30102+ if (IS_ERR(d))
30103+ goto out;
30104+
30105+ base[i].dentry = d;
30106+ AuDebugOn(wbr
30107+ && wbr->wbr_wh[i]
30108+ && wbr->wbr_wh[i] != base[i].dentry);
30109+ }
30110+
30111+ if (wbr)
30112+ for (i = 0; i < AuBrWh_Last; i++) {
30113+ dput(wbr->wbr_wh[i]);
30114+ wbr->wbr_wh[i] = NULL;
30115+ }
30116+
30117+ err = 0;
1e00d052 30118+ if (!au_br_writable(br->br_perm)) {
4a4d8108 30119+ h_dir = h_root->d_inode;
1facf9fc 30120+ au_wh_init_ro(h_dir, base, &path);
1e00d052 30121+ } else if (!au_br_wh_linkable(br->br_perm)) {
1facf9fc 30122+ err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
30123+ if (err > 0)
30124+ goto out;
30125+ else if (err)
30126+ goto out_err;
1e00d052 30127+ } else {
1facf9fc 30128+ err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
30129+ if (err > 0)
30130+ goto out;
30131+ else if (err)
30132+ goto out_err;
1facf9fc 30133+ }
30134+ goto out; /* success */
30135+
4f0767ce 30136+out_err:
523b37e3
AM
30137+ pr_err("an error(%d) on the writable branch %pd(%s)\n",
30138+ err, h_root, au_sbtype(h_root->d_sb));
4f0767ce 30139+out:
1facf9fc 30140+ for (i = 0; i < AuBrWh_Last; i++)
30141+ dput(base[i].dentry);
30142+ return err;
30143+}
30144+
30145+/* ---------------------------------------------------------------------- */
30146+/*
30147+ * whiteouts are all hard-linked usually.
30148+ * when its link count reaches a ceiling, we create a new whiteout base
30149+ * asynchronously.
30150+ */
30151+
30152+struct reinit_br_wh {
30153+ struct super_block *sb;
30154+ struct au_branch *br;
30155+};
30156+
30157+static void reinit_br_wh(void *arg)
30158+{
30159+ int err;
30160+ aufs_bindex_t bindex;
30161+ struct path h_path;
30162+ struct reinit_br_wh *a = arg;
30163+ struct au_wbr *wbr;
523b37e3 30164+ struct inode *dir, *delegated;
1facf9fc 30165+ struct dentry *h_root;
30166+ struct au_hinode *hdir;
30167+
30168+ err = 0;
30169+ wbr = a->br->br_wbr;
30170+ /* big aufs lock */
30171+ si_noflush_write_lock(a->sb);
30172+ if (!au_br_writable(a->br->br_perm))
30173+ goto out;
30174+ bindex = au_br_index(a->sb, a->br->br_id);
30175+ if (unlikely(bindex < 0))
30176+ goto out;
30177+
1308ab2a 30178+ di_read_lock_parent(a->sb->s_root, AuLock_IR);
1facf9fc 30179+ dir = a->sb->s_root->d_inode;
1facf9fc 30180+ hdir = au_hi(dir, bindex);
30181+ h_root = au_h_dptr(a->sb->s_root, bindex);
86dc4139 30182+ AuDebugOn(h_root != au_br_dentry(a->br));
1facf9fc 30183+
4a4d8108 30184+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
1facf9fc 30185+ wbr_wh_write_lock(wbr);
30186+ err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
30187+ h_root, a->br);
30188+ if (!err) {
86dc4139
AM
30189+ h_path.dentry = wbr->wbr_whbase;
30190+ h_path.mnt = au_br_mnt(a->br);
523b37e3
AM
30191+ delegated = NULL;
30192+ err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
30193+ /*force*/0);
30194+ if (unlikely(err == -EWOULDBLOCK)) {
30195+ pr_warn("cannot retry for NFSv4 delegation"
30196+ " for an internal unlink\n");
30197+ iput(delegated);
30198+ }
1facf9fc 30199+ } else {
523b37e3 30200+ pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
1facf9fc 30201+ err = 0;
30202+ }
30203+ dput(wbr->wbr_whbase);
30204+ wbr->wbr_whbase = NULL;
30205+ if (!err)
86dc4139 30206+ err = au_wh_init(a->br, a->sb);
1facf9fc 30207+ wbr_wh_write_unlock(wbr);
4a4d8108 30208+ au_hn_imtx_unlock(hdir);
1308ab2a 30209+ di_read_unlock(a->sb->s_root, AuLock_IR);
1facf9fc 30210+
4f0767ce 30211+out:
1facf9fc 30212+ if (wbr)
30213+ atomic_dec(&wbr->wbr_wh_running);
30214+ atomic_dec(&a->br->br_count);
1facf9fc 30215+ si_write_unlock(a->sb);
027c5e7a 30216+ au_nwt_done(&au_sbi(a->sb)->si_nowait);
1facf9fc 30217+ kfree(arg);
30218+ if (unlikely(err))
30219+ AuIOErr("err %d\n", err);
30220+}
30221+
30222+static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
30223+{
30224+ int do_dec, wkq_err;
30225+ struct reinit_br_wh *arg;
30226+
30227+ do_dec = 1;
30228+ if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
30229+ goto out;
30230+
30231+ /* ignore ENOMEM */
30232+ arg = kmalloc(sizeof(*arg), GFP_NOFS);
30233+ if (arg) {
30234+ /*
30235+ * dec(wh_running), kfree(arg) and dec(br_count)
30236+ * in reinit function
30237+ */
30238+ arg->sb = sb;
30239+ arg->br = br;
30240+ atomic_inc(&br->br_count);
53392da6 30241+ wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
1facf9fc 30242+ if (unlikely(wkq_err)) {
30243+ atomic_dec(&br->br_wbr->wbr_wh_running);
30244+ atomic_dec(&br->br_count);
30245+ kfree(arg);
30246+ }
30247+ do_dec = 0;
30248+ }
30249+
4f0767ce 30250+out:
1facf9fc 30251+ if (do_dec)
30252+ atomic_dec(&br->br_wbr->wbr_wh_running);
30253+}
30254+
30255+/* ---------------------------------------------------------------------- */
30256+
30257+/*
30258+ * create the whiteout @wh.
30259+ */
30260+static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
30261+ struct dentry *wh)
30262+{
30263+ int err;
30264+ struct path h_path = {
30265+ .dentry = wh
30266+ };
30267+ struct au_branch *br;
30268+ struct au_wbr *wbr;
30269+ struct dentry *h_parent;
523b37e3 30270+ struct inode *h_dir, *delegated;
1facf9fc 30271+
30272+ h_parent = wh->d_parent; /* dir inode is locked */
30273+ h_dir = h_parent->d_inode;
30274+ IMustLock(h_dir);
30275+
30276+ br = au_sbr(sb, bindex);
86dc4139 30277+ h_path.mnt = au_br_mnt(br);
1facf9fc 30278+ wbr = br->br_wbr;
30279+ wbr_wh_read_lock(wbr);
30280+ if (wbr->wbr_whbase) {
523b37e3
AM
30281+ delegated = NULL;
30282+ err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
30283+ if (unlikely(err == -EWOULDBLOCK)) {
30284+ pr_warn("cannot retry for NFSv4 delegation"
30285+ " for an internal link\n");
30286+ iput(delegated);
30287+ }
1facf9fc 30288+ if (!err || err != -EMLINK)
30289+ goto out;
30290+
30291+ /* link count full. re-initialize br_whbase. */
30292+ kick_reinit_br_wh(sb, br);
30293+ }
30294+
30295+ /* return this error in this context */
b4510431 30296+ err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
1facf9fc 30297+
4f0767ce 30298+out:
1facf9fc 30299+ wbr_wh_read_unlock(wbr);
30300+ return err;
30301+}
30302+
30303+/* ---------------------------------------------------------------------- */
30304+
30305+/*
30306+ * create or remove the diropq.
30307+ */
30308+static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
30309+ unsigned int flags)
30310+{
30311+ struct dentry *opq_dentry, *h_dentry;
30312+ struct super_block *sb;
30313+ struct au_branch *br;
30314+ int err;
30315+
30316+ sb = dentry->d_sb;
30317+ br = au_sbr(sb, bindex);
30318+ h_dentry = au_h_dptr(dentry, bindex);
b4510431 30319+ opq_dentry = vfsub_lkup_one(&diropq_name, h_dentry);
1facf9fc 30320+ if (IS_ERR(opq_dentry))
30321+ goto out;
30322+
30323+ if (au_ftest_diropq(flags, CREATE)) {
30324+ err = link_or_create_wh(sb, bindex, opq_dentry);
30325+ if (!err) {
30326+ au_set_dbdiropq(dentry, bindex);
30327+ goto out; /* success */
30328+ }
30329+ } else {
30330+ struct path tmp = {
30331+ .dentry = opq_dentry,
86dc4139 30332+ .mnt = au_br_mnt(br)
1facf9fc 30333+ };
30334+ err = do_unlink_wh(au_h_iptr(dentry->d_inode, bindex), &tmp);
30335+ if (!err)
30336+ au_set_dbdiropq(dentry, -1);
30337+ }
30338+ dput(opq_dentry);
30339+ opq_dentry = ERR_PTR(err);
30340+
4f0767ce 30341+out:
1facf9fc 30342+ return opq_dentry;
30343+}
30344+
30345+struct do_diropq_args {
30346+ struct dentry **errp;
30347+ struct dentry *dentry;
30348+ aufs_bindex_t bindex;
30349+ unsigned int flags;
30350+};
30351+
30352+static void call_do_diropq(void *args)
30353+{
30354+ struct do_diropq_args *a = args;
30355+ *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
30356+}
30357+
30358+struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
30359+ unsigned int flags)
30360+{
30361+ struct dentry *diropq, *h_dentry;
30362+
30363+ h_dentry = au_h_dptr(dentry, bindex);
30364+ if (!au_test_h_perm_sio(h_dentry->d_inode, MAY_EXEC | MAY_WRITE))
30365+ diropq = do_diropq(dentry, bindex, flags);
30366+ else {
30367+ int wkq_err;
30368+ struct do_diropq_args args = {
30369+ .errp = &diropq,
30370+ .dentry = dentry,
30371+ .bindex = bindex,
30372+ .flags = flags
30373+ };
30374+
30375+ wkq_err = au_wkq_wait(call_do_diropq, &args);
30376+ if (unlikely(wkq_err))
30377+ diropq = ERR_PTR(wkq_err);
30378+ }
30379+
30380+ return diropq;
30381+}
30382+
30383+/* ---------------------------------------------------------------------- */
30384+
30385+/*
30386+ * lookup whiteout dentry.
30387+ * @h_parent: lower parent dentry which must exist and be locked
30388+ * @base_name: name of dentry which will be whiteouted
30389+ * returns dentry for whiteout.
30390+ */
30391+struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
30392+ struct au_branch *br)
30393+{
30394+ int err;
30395+ struct qstr wh_name;
30396+ struct dentry *wh_dentry;
30397+
30398+ err = au_wh_name_alloc(&wh_name, base_name);
30399+ wh_dentry = ERR_PTR(err);
30400+ if (!err) {
b4510431 30401+ wh_dentry = vfsub_lkup_one(&wh_name, h_parent);
1facf9fc 30402+ kfree(wh_name.name);
30403+ }
30404+ return wh_dentry;
30405+}
30406+
30407+/*
30408+ * link/create a whiteout for @dentry on @bindex.
30409+ */
30410+struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
30411+ struct dentry *h_parent)
30412+{
30413+ struct dentry *wh_dentry;
30414+ struct super_block *sb;
30415+ int err;
30416+
30417+ sb = dentry->d_sb;
30418+ wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
30419+ if (!IS_ERR(wh_dentry) && !wh_dentry->d_inode) {
30420+ err = link_or_create_wh(sb, bindex, wh_dentry);
30421+ if (!err)
30422+ au_set_dbwh(dentry, bindex);
30423+ else {
30424+ dput(wh_dentry);
30425+ wh_dentry = ERR_PTR(err);
30426+ }
30427+ }
30428+
30429+ return wh_dentry;
30430+}
30431+
30432+/* ---------------------------------------------------------------------- */
30433+
30434+/* Delete all whiteouts in this directory on branch bindex. */
30435+static int del_wh_children(struct dentry *h_dentry, struct au_nhash *whlist,
30436+ aufs_bindex_t bindex, struct au_branch *br)
30437+{
30438+ int err;
30439+ unsigned long ul, n;
30440+ struct qstr wh_name;
30441+ char *p;
30442+ struct hlist_head *head;
c06a8ce3 30443+ struct au_vdir_wh *pos;
1facf9fc 30444+ struct au_vdir_destr *str;
30445+
30446+ err = -ENOMEM;
537831f9 30447+ p = (void *)__get_free_page(GFP_NOFS);
1facf9fc 30448+ wh_name.name = p;
30449+ if (unlikely(!wh_name.name))
30450+ goto out;
30451+
30452+ err = 0;
30453+ memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
30454+ p += AUFS_WH_PFX_LEN;
30455+ n = whlist->nh_num;
30456+ head = whlist->nh_head;
30457+ for (ul = 0; !err && ul < n; ul++, head++) {
c06a8ce3
AM
30458+ hlist_for_each_entry(pos, head, wh_hash) {
30459+ if (pos->wh_bindex != bindex)
1facf9fc 30460+ continue;
30461+
c06a8ce3 30462+ str = &pos->wh_str;
1facf9fc 30463+ if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
30464+ memcpy(p, str->name, str->len);
30465+ wh_name.len = AUFS_WH_PFX_LEN + str->len;
30466+ err = unlink_wh_name(h_dentry, &wh_name, br);
30467+ if (!err)
30468+ continue;
30469+ break;
30470+ }
30471+ AuIOErr("whiteout name too long %.*s\n",
30472+ str->len, str->name);
30473+ err = -EIO;
30474+ break;
30475+ }
30476+ }
537831f9 30477+ free_page((unsigned long)wh_name.name);
1facf9fc 30478+
4f0767ce 30479+out:
1facf9fc 30480+ return err;
30481+}
30482+
30483+struct del_wh_children_args {
30484+ int *errp;
30485+ struct dentry *h_dentry;
1308ab2a 30486+ struct au_nhash *whlist;
1facf9fc 30487+ aufs_bindex_t bindex;
30488+ struct au_branch *br;
30489+};
30490+
30491+static void call_del_wh_children(void *args)
30492+{
30493+ struct del_wh_children_args *a = args;
1308ab2a 30494+ *a->errp = del_wh_children(a->h_dentry, a->whlist, a->bindex, a->br);
1facf9fc 30495+}
30496+
30497+/* ---------------------------------------------------------------------- */
30498+
30499+struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
30500+{
30501+ struct au_whtmp_rmdir *whtmp;
dece6358 30502+ int err;
1308ab2a 30503+ unsigned int rdhash;
dece6358
AM
30504+
30505+ SiMustAnyLock(sb);
1facf9fc 30506+
30507+ whtmp = kmalloc(sizeof(*whtmp), gfp);
dece6358
AM
30508+ if (unlikely(!whtmp)) {
30509+ whtmp = ERR_PTR(-ENOMEM);
1facf9fc 30510+ goto out;
dece6358 30511+ }
1facf9fc 30512+
30513+ whtmp->dir = NULL;
027c5e7a 30514+ whtmp->br = NULL;
1facf9fc 30515+ whtmp->wh_dentry = NULL;
1308ab2a 30516+ /* no estimation for dir size */
30517+ rdhash = au_sbi(sb)->si_rdhash;
30518+ if (!rdhash)
30519+ rdhash = AUFS_RDHASH_DEF;
30520+ err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
30521+ if (unlikely(err)) {
30522+ kfree(whtmp);
30523+ whtmp = ERR_PTR(err);
30524+ }
dece6358 30525+
4f0767ce 30526+out:
dece6358 30527+ return whtmp;
1facf9fc 30528+}
30529+
30530+void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
30531+{
027c5e7a
AM
30532+ if (whtmp->br)
30533+ atomic_dec(&whtmp->br->br_count);
1facf9fc 30534+ dput(whtmp->wh_dentry);
30535+ iput(whtmp->dir);
dece6358 30536+ au_nhash_wh_free(&whtmp->whlist);
1facf9fc 30537+ kfree(whtmp);
30538+}
30539+
30540+/*
30541+ * rmdir the whiteouted temporary named dir @h_dentry.
30542+ * @whlist: whiteouted children.
30543+ */
30544+int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
30545+ struct dentry *wh_dentry, struct au_nhash *whlist)
30546+{
30547+ int err;
30548+ struct path h_tmp;
30549+ struct inode *wh_inode, *h_dir;
30550+ struct au_branch *br;
30551+
30552+ h_dir = wh_dentry->d_parent->d_inode; /* dir inode is locked */
30553+ IMustLock(h_dir);
30554+
30555+ br = au_sbr(dir->i_sb, bindex);
30556+ wh_inode = wh_dentry->d_inode;
30557+ mutex_lock_nested(&wh_inode->i_mutex, AuLsc_I_CHILD);
30558+
30559+ /*
30560+ * someone else might change some whiteouts while we were sleeping.
30561+ * it means this whlist may have an obsoleted entry.
30562+ */
30563+ if (!au_test_h_perm_sio(wh_inode, MAY_EXEC | MAY_WRITE))
30564+ err = del_wh_children(wh_dentry, whlist, bindex, br);
30565+ else {
30566+ int wkq_err;
30567+ struct del_wh_children_args args = {
30568+ .errp = &err,
30569+ .h_dentry = wh_dentry,
1308ab2a 30570+ .whlist = whlist,
1facf9fc 30571+ .bindex = bindex,
30572+ .br = br
30573+ };
30574+
30575+ wkq_err = au_wkq_wait(call_del_wh_children, &args);
30576+ if (unlikely(wkq_err))
30577+ err = wkq_err;
30578+ }
30579+ mutex_unlock(&wh_inode->i_mutex);
30580+
30581+ if (!err) {
30582+ h_tmp.dentry = wh_dentry;
86dc4139 30583+ h_tmp.mnt = au_br_mnt(br);
1facf9fc 30584+ err = vfsub_rmdir(h_dir, &h_tmp);
1facf9fc 30585+ }
30586+
30587+ if (!err) {
30588+ if (au_ibstart(dir) == bindex) {
7f207e10 30589+ /* todo: dir->i_mutex is necessary */
1facf9fc 30590+ au_cpup_attr_timesizes(dir);
7f207e10 30591+ vfsub_drop_nlink(dir);
1facf9fc 30592+ }
30593+ return 0; /* success */
30594+ }
30595+
523b37e3 30596+ pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
1facf9fc 30597+ return err;
30598+}
30599+
30600+static void call_rmdir_whtmp(void *args)
30601+{
30602+ int err;
e49829fe 30603+ aufs_bindex_t bindex;
1facf9fc 30604+ struct au_whtmp_rmdir *a = args;
30605+ struct super_block *sb;
30606+ struct dentry *h_parent;
30607+ struct inode *h_dir;
1facf9fc 30608+ struct au_hinode *hdir;
30609+
30610+ /* rmdir by nfsd may cause deadlock with this i_mutex */
30611+ /* mutex_lock(&a->dir->i_mutex); */
e49829fe 30612+ err = -EROFS;
1facf9fc 30613+ sb = a->dir->i_sb;
e49829fe
JR
30614+ si_read_lock(sb, !AuLock_FLUSH);
30615+ if (!au_br_writable(a->br->br_perm))
30616+ goto out;
30617+ bindex = au_br_index(sb, a->br->br_id);
30618+ if (unlikely(bindex < 0))
1facf9fc 30619+ goto out;
30620+
30621+ err = -EIO;
1facf9fc 30622+ ii_write_lock_parent(a->dir);
30623+ h_parent = dget_parent(a->wh_dentry);
30624+ h_dir = h_parent->d_inode;
e49829fe 30625+ hdir = au_hi(a->dir, bindex);
86dc4139
AM
30626+ err = vfsub_mnt_want_write(au_br_mnt(a->br));
30627+ if (unlikely(err))
30628+ goto out_mnt;
4a4d8108 30629+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
e49829fe
JR
30630+ err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
30631+ a->br);
86dc4139
AM
30632+ if (!err)
30633+ err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
4a4d8108 30634+ au_hn_imtx_unlock(hdir);
86dc4139
AM
30635+ vfsub_mnt_drop_write(au_br_mnt(a->br));
30636+
30637+out_mnt:
1facf9fc 30638+ dput(h_parent);
30639+ ii_write_unlock(a->dir);
4f0767ce 30640+out:
1facf9fc 30641+ /* mutex_unlock(&a->dir->i_mutex); */
1facf9fc 30642+ au_whtmp_rmdir_free(a);
027c5e7a
AM
30643+ si_read_unlock(sb);
30644+ au_nwt_done(&au_sbi(sb)->si_nowait);
1facf9fc 30645+ if (unlikely(err))
30646+ AuIOErr("err %d\n", err);
30647+}
30648+
30649+void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
30650+ struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
30651+{
30652+ int wkq_err;
e49829fe 30653+ struct super_block *sb;
1facf9fc 30654+
30655+ IMustLock(dir);
30656+
30657+ /* all post-process will be done in do_rmdir_whtmp(). */
e49829fe 30658+ sb = dir->i_sb;
1facf9fc 30659+ args->dir = au_igrab(dir);
e49829fe
JR
30660+ args->br = au_sbr(sb, bindex);
30661+ atomic_inc(&args->br->br_count);
1facf9fc 30662+ args->wh_dentry = dget(wh_dentry);
53392da6 30663+ wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
1facf9fc 30664+ if (unlikely(wkq_err)) {
523b37e3 30665+ pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
1facf9fc 30666+ au_whtmp_rmdir_free(args);
30667+ }
30668+}
7f207e10
AM
30669diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
30670--- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 30671+++ linux/fs/aufs/whout.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 30672@@ -0,0 +1,86 @@
1facf9fc 30673+/*
523b37e3 30674+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 30675+ *
30676+ * This program, aufs is free software; you can redistribute it and/or modify
30677+ * it under the terms of the GNU General Public License as published by
30678+ * the Free Software Foundation; either version 2 of the License, or
30679+ * (at your option) any later version.
dece6358
AM
30680+ *
30681+ * This program is distributed in the hope that it will be useful,
30682+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
30683+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30684+ * GNU General Public License for more details.
30685+ *
30686+ * You should have received a copy of the GNU General Public License
523b37e3 30687+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 30688+ */
30689+
30690+/*
30691+ * whiteout for logical deletion and opaque directory
30692+ */
30693+
30694+#ifndef __AUFS_WHOUT_H__
30695+#define __AUFS_WHOUT_H__
30696+
30697+#ifdef __KERNEL__
30698+
1facf9fc 30699+#include "dir.h"
30700+
30701+/* whout.c */
30702+int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
30703+struct au_branch;
30704+int au_wh_test(struct dentry *h_parent, struct qstr *wh_name,
30705+ struct au_branch *br, int try_sio);
30706+int au_diropq_test(struct dentry *h_dentry, struct au_branch *br);
30707+struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
30708+ struct qstr *prefix);
30709+int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
30710+int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
30711+ struct dentry *dentry);
86dc4139 30712+int au_wh_init(struct au_branch *br, struct super_block *sb);
1facf9fc 30713+
30714+/* diropq flags */
30715+#define AuDiropq_CREATE 1
30716+#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
7f207e10
AM
30717+#define au_fset_diropq(flags, name) \
30718+ do { (flags) |= AuDiropq_##name; } while (0)
30719+#define au_fclr_diropq(flags, name) \
30720+ do { (flags) &= ~AuDiropq_##name; } while (0)
1facf9fc 30721+
30722+struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
30723+ unsigned int flags);
30724+struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
30725+ struct au_branch *br);
30726+struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
30727+ struct dentry *h_parent);
30728+
30729+/* real rmdir for the whiteout-ed dir */
30730+struct au_whtmp_rmdir {
30731+ struct inode *dir;
e49829fe 30732+ struct au_branch *br;
1facf9fc 30733+ struct dentry *wh_dentry;
dece6358 30734+ struct au_nhash whlist;
1facf9fc 30735+};
30736+
30737+struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
30738+void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
30739+int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
30740+ struct dentry *wh_dentry, struct au_nhash *whlist);
30741+void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
30742+ struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
30743+
30744+/* ---------------------------------------------------------------------- */
30745+
30746+static inline struct dentry *au_diropq_create(struct dentry *dentry,
30747+ aufs_bindex_t bindex)
30748+{
30749+ return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
30750+}
30751+
30752+static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
30753+{
30754+ return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
30755+}
30756+
30757+#endif /* __KERNEL__ */
30758+#endif /* __AUFS_WHOUT_H__ */
7f207e10
AM
30759diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
30760--- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
30761+++ linux/fs/aufs/wkq.c 2014-06-24 00:14:06.789386337 +0200
30762@@ -0,0 +1,213 @@
1facf9fc 30763+/*
523b37e3 30764+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 30765+ *
30766+ * This program, aufs is free software; you can redistribute it and/or modify
30767+ * it under the terms of the GNU General Public License as published by
30768+ * the Free Software Foundation; either version 2 of the License, or
30769+ * (at your option) any later version.
dece6358
AM
30770+ *
30771+ * This program is distributed in the hope that it will be useful,
30772+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
30773+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30774+ * GNU General Public License for more details.
30775+ *
30776+ * You should have received a copy of the GNU General Public License
523b37e3 30777+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 30778+ */
30779+
30780+/*
30781+ * workqueue for asynchronous/super-io operations
30782+ * todo: try new dredential scheme
30783+ */
30784+
dece6358 30785+#include <linux/module.h>
1facf9fc 30786+#include "aufs.h"
30787+
9dbd164d 30788+/* internal workqueue named AUFS_WKQ_NAME */
b752ccd1 30789+
9dbd164d 30790+static struct workqueue_struct *au_wkq;
1facf9fc 30791+
30792+struct au_wkinfo {
30793+ struct work_struct wk;
7f207e10 30794+ struct kobject *kobj;
1facf9fc 30795+
30796+ unsigned int flags; /* see wkq.h */
30797+
30798+ au_wkq_func_t func;
30799+ void *args;
30800+
1facf9fc 30801+ struct completion *comp;
30802+};
30803+
30804+/* ---------------------------------------------------------------------- */
30805+
1facf9fc 30806+static void wkq_func(struct work_struct *wk)
30807+{
30808+ struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
30809+
2dfbb274 30810+ AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
7f207e10
AM
30811+ AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
30812+
1facf9fc 30813+ wkinfo->func(wkinfo->args);
1facf9fc 30814+ if (au_ftest_wkq(wkinfo->flags, WAIT))
30815+ complete(wkinfo->comp);
30816+ else {
7f207e10 30817+ kobject_put(wkinfo->kobj);
9dbd164d 30818+ module_put(THIS_MODULE); /* todo: ?? */
1facf9fc 30819+ kfree(wkinfo);
30820+ }
30821+}
30822+
30823+/*
30824+ * Since struct completion is large, try allocating it dynamically.
30825+ */
c2b27bf2 30826+#if 1 /* defined(CONFIG_4KSTACKS) || defined(AuTest4KSTACKS) */
1facf9fc 30827+#define AuWkqCompDeclare(name) struct completion *comp = NULL
30828+
30829+static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
30830+{
30831+ *comp = kmalloc(sizeof(**comp), GFP_NOFS);
30832+ if (*comp) {
30833+ init_completion(*comp);
30834+ wkinfo->comp = *comp;
30835+ return 0;
30836+ }
30837+ return -ENOMEM;
30838+}
30839+
30840+static void au_wkq_comp_free(struct completion *comp)
30841+{
30842+ kfree(comp);
30843+}
30844+
30845+#else
30846+
30847+/* no braces */
30848+#define AuWkqCompDeclare(name) \
30849+ DECLARE_COMPLETION_ONSTACK(_ ## name); \
30850+ struct completion *comp = &_ ## name
30851+
30852+static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
30853+{
30854+ wkinfo->comp = *comp;
30855+ return 0;
30856+}
30857+
30858+static void au_wkq_comp_free(struct completion *comp __maybe_unused)
30859+{
30860+ /* empty */
30861+}
30862+#endif /* 4KSTACKS */
30863+
53392da6 30864+static void au_wkq_run(struct au_wkinfo *wkinfo)
1facf9fc 30865+{
53392da6
AM
30866+ if (au_ftest_wkq(wkinfo->flags, NEST)) {
30867+ if (au_wkq_test()) {
38d290e6
JR
30868+ AuWarn1("wkq from wkq, unless silly-rename on NFS,"
30869+ " due to a dead dir by UDBA?\n");
53392da6
AM
30870+ AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
30871+ }
30872+ } else
30873+ au_dbg_verify_kthread();
30874+
30875+ if (au_ftest_wkq(wkinfo->flags, WAIT)) {
a1f66529 30876+ INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
9dbd164d 30877+ queue_work(au_wkq, &wkinfo->wk);
4a4d8108
AM
30878+ } else {
30879+ INIT_WORK(&wkinfo->wk, wkq_func);
30880+ schedule_work(&wkinfo->wk);
30881+ }
1facf9fc 30882+}
30883+
7f207e10
AM
30884+/*
30885+ * Be careful. It is easy to make deadlock happen.
30886+ * processA: lock, wkq and wait
30887+ * processB: wkq and wait, lock in wkq
30888+ * --> deadlock
30889+ */
b752ccd1 30890+int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
1facf9fc 30891+{
30892+ int err;
30893+ AuWkqCompDeclare(comp);
30894+ struct au_wkinfo wkinfo = {
b752ccd1 30895+ .flags = flags,
1facf9fc 30896+ .func = func,
30897+ .args = args
30898+ };
30899+
30900+ err = au_wkq_comp_alloc(&wkinfo, &comp);
30901+ if (!err) {
53392da6 30902+ au_wkq_run(&wkinfo);
1facf9fc 30903+ /* no timeout, no interrupt */
30904+ wait_for_completion(wkinfo.comp);
30905+ au_wkq_comp_free(comp);
4a4d8108 30906+ destroy_work_on_stack(&wkinfo.wk);
1facf9fc 30907+ }
30908+
30909+ return err;
30910+
30911+}
30912+
027c5e7a
AM
30913+/*
30914+ * Note: dget/dput() in func for aufs dentries are not supported. It will be a
30915+ * problem in a concurrent umounting.
30916+ */
53392da6
AM
30917+int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
30918+ unsigned int flags)
1facf9fc 30919+{
30920+ int err;
30921+ struct au_wkinfo *wkinfo;
30922+
30923+ atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
30924+
30925+ /*
30926+ * wkq_func() must free this wkinfo.
30927+ * it highly depends upon the implementation of workqueue.
30928+ */
30929+ err = 0;
30930+ wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
30931+ if (wkinfo) {
7f207e10 30932+ wkinfo->kobj = &au_sbi(sb)->si_kobj;
53392da6 30933+ wkinfo->flags = flags & ~AuWkq_WAIT;
1facf9fc 30934+ wkinfo->func = func;
30935+ wkinfo->args = args;
30936+ wkinfo->comp = NULL;
7f207e10 30937+ kobject_get(wkinfo->kobj);
9dbd164d 30938+ __module_get(THIS_MODULE); /* todo: ?? */
1facf9fc 30939+
53392da6 30940+ au_wkq_run(wkinfo);
1facf9fc 30941+ } else {
30942+ err = -ENOMEM;
e49829fe 30943+ au_nwt_done(&au_sbi(sb)->si_nowait);
1facf9fc 30944+ }
30945+
30946+ return err;
30947+}
30948+
30949+/* ---------------------------------------------------------------------- */
30950+
30951+void au_nwt_init(struct au_nowait_tasks *nwt)
30952+{
30953+ atomic_set(&nwt->nw_len, 0);
4a4d8108 30954+ /* smp_mb(); */ /* atomic_set */
1facf9fc 30955+ init_waitqueue_head(&nwt->nw_wq);
30956+}
30957+
30958+void au_wkq_fin(void)
30959+{
9dbd164d 30960+ destroy_workqueue(au_wkq);
1facf9fc 30961+}
30962+
30963+int __init au_wkq_init(void)
30964+{
9dbd164d 30965+ int err;
b752ccd1
AM
30966+
30967+ err = 0;
86dc4139 30968+ au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
9dbd164d
AM
30969+ if (IS_ERR(au_wkq))
30970+ err = PTR_ERR(au_wkq);
30971+ else if (!au_wkq)
30972+ err = -ENOMEM;
b752ccd1
AM
30973+
30974+ return err;
1facf9fc 30975+}
7f207e10
AM
30976diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
30977--- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 30978+++ linux/fs/aufs/wkq.h 2014-06-24 00:14:06.789386337 +0200
523b37e3 30979@@ -0,0 +1,91 @@
1facf9fc 30980+/*
523b37e3 30981+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 30982+ *
30983+ * This program, aufs is free software; you can redistribute it and/or modify
30984+ * it under the terms of the GNU General Public License as published by
30985+ * the Free Software Foundation; either version 2 of the License, or
30986+ * (at your option) any later version.
dece6358
AM
30987+ *
30988+ * This program is distributed in the hope that it will be useful,
30989+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
30990+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30991+ * GNU General Public License for more details.
30992+ *
30993+ * You should have received a copy of the GNU General Public License
523b37e3 30994+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 30995+ */
30996+
30997+/*
30998+ * workqueue for asynchronous/super-io operations
30999+ * todo: try new credentials management scheme
31000+ */
31001+
31002+#ifndef __AUFS_WKQ_H__
31003+#define __AUFS_WKQ_H__
31004+
31005+#ifdef __KERNEL__
31006+
dece6358
AM
31007+struct super_block;
31008+
1facf9fc 31009+/* ---------------------------------------------------------------------- */
31010+
31011+/*
31012+ * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
31013+ */
31014+struct au_nowait_tasks {
31015+ atomic_t nw_len;
31016+ wait_queue_head_t nw_wq;
31017+};
31018+
31019+/* ---------------------------------------------------------------------- */
31020+
31021+typedef void (*au_wkq_func_t)(void *args);
31022+
31023+/* wkq flags */
31024+#define AuWkq_WAIT 1
9dbd164d 31025+#define AuWkq_NEST (1 << 1)
1facf9fc 31026+#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
7f207e10
AM
31027+#define au_fset_wkq(flags, name) \
31028+ do { (flags) |= AuWkq_##name; } while (0)
31029+#define au_fclr_wkq(flags, name) \
31030+ do { (flags) &= ~AuWkq_##name; } while (0)
1facf9fc 31031+
9dbd164d
AM
31032+#ifndef CONFIG_AUFS_HNOTIFY
31033+#undef AuWkq_NEST
31034+#define AuWkq_NEST 0
31035+#endif
31036+
1facf9fc 31037+/* wkq.c */
b752ccd1 31038+int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
53392da6
AM
31039+int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
31040+ unsigned int flags);
1facf9fc 31041+void au_nwt_init(struct au_nowait_tasks *nwt);
31042+int __init au_wkq_init(void);
31043+void au_wkq_fin(void);
31044+
31045+/* ---------------------------------------------------------------------- */
31046+
53392da6
AM
31047+static inline int au_wkq_test(void)
31048+{
31049+ return current->flags & PF_WQ_WORKER;
31050+}
31051+
b752ccd1 31052+static inline int au_wkq_wait(au_wkq_func_t func, void *args)
1facf9fc 31053+{
b752ccd1 31054+ return au_wkq_do_wait(AuWkq_WAIT, func, args);
1facf9fc 31055+}
31056+
31057+static inline void au_nwt_done(struct au_nowait_tasks *nwt)
31058+{
e49829fe 31059+ if (atomic_dec_and_test(&nwt->nw_len))
1facf9fc 31060+ wake_up_all(&nwt->nw_wq);
31061+}
31062+
31063+static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
31064+{
31065+ wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
31066+ return 0;
31067+}
31068+
31069+#endif /* __KERNEL__ */
31070+#endif /* __AUFS_WKQ_H__ */
7f207e10
AM
31071diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
31072--- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
38d290e6
JR
31073+++ linux/fs/aufs/xino.c 2014-06-24 00:14:06.792719853 +0200
31074@@ -0,0 +1,1312 @@
1facf9fc 31075+/*
523b37e3 31076+ * Copyright (C) 2005-2014 Junjiro R. Okajima
1facf9fc 31077+ *
31078+ * This program, aufs is free software; you can redistribute it and/or modify
31079+ * it under the terms of the GNU General Public License as published by
31080+ * the Free Software Foundation; either version 2 of the License, or
31081+ * (at your option) any later version.
dece6358
AM
31082+ *
31083+ * This program is distributed in the hope that it will be useful,
31084+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
31085+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31086+ * GNU General Public License for more details.
31087+ *
31088+ * You should have received a copy of the GNU General Public License
523b37e3 31089+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 31090+ */
31091+
31092+/*
31093+ * external inode number translation table and bitmap
31094+ */
31095+
31096+#include <linux/seq_file.h>
392086de 31097+#include <linux/statfs.h>
1facf9fc 31098+#include "aufs.h"
31099+
9dbd164d 31100+/* todo: unnecessary to support mmap_sem since kernel-space? */
b752ccd1 31101+ssize_t xino_fread(au_readf_t func, struct file *file, void *kbuf, size_t size,
1facf9fc 31102+ loff_t *pos)
31103+{
31104+ ssize_t err;
31105+ mm_segment_t oldfs;
b752ccd1
AM
31106+ union {
31107+ void *k;
31108+ char __user *u;
31109+ } buf;
1facf9fc 31110+
b752ccd1 31111+ buf.k = kbuf;
1facf9fc 31112+ oldfs = get_fs();
31113+ set_fs(KERNEL_DS);
31114+ do {
31115+ /* todo: signal_pending? */
b752ccd1 31116+ err = func(file, buf.u, size, pos);
1facf9fc 31117+ } while (err == -EAGAIN || err == -EINTR);
31118+ set_fs(oldfs);
31119+
31120+#if 0 /* reserved for future use */
31121+ if (err > 0)
31122+ fsnotify_access(file->f_dentry);
31123+#endif
31124+
31125+ return err;
31126+}
31127+
31128+/* ---------------------------------------------------------------------- */
31129+
b752ccd1 31130+static ssize_t do_xino_fwrite(au_writef_t func, struct file *file, void *kbuf,
1facf9fc 31131+ size_t size, loff_t *pos)
31132+{
31133+ ssize_t err;
31134+ mm_segment_t oldfs;
b752ccd1
AM
31135+ union {
31136+ void *k;
31137+ const char __user *u;
31138+ } buf;
1facf9fc 31139+
b752ccd1 31140+ buf.k = kbuf;
1facf9fc 31141+ oldfs = get_fs();
31142+ set_fs(KERNEL_DS);
1facf9fc 31143+ do {
31144+ /* todo: signal_pending? */
b752ccd1 31145+ err = func(file, buf.u, size, pos);
1facf9fc 31146+ } while (err == -EAGAIN || err == -EINTR);
1facf9fc 31147+ set_fs(oldfs);
31148+
31149+#if 0 /* reserved for future use */
31150+ if (err > 0)
31151+ fsnotify_modify(file->f_dentry);
31152+#endif
31153+
31154+ return err;
31155+}
31156+
31157+struct do_xino_fwrite_args {
31158+ ssize_t *errp;
31159+ au_writef_t func;
31160+ struct file *file;
31161+ void *buf;
31162+ size_t size;
31163+ loff_t *pos;
31164+};
31165+
31166+static void call_do_xino_fwrite(void *args)
31167+{
31168+ struct do_xino_fwrite_args *a = args;
31169+ *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos);
31170+}
31171+
31172+ssize_t xino_fwrite(au_writef_t func, struct file *file, void *buf, size_t size,
31173+ loff_t *pos)
31174+{
31175+ ssize_t err;
31176+
31177+ /* todo: signal block and no wkq? */
b752ccd1
AM
31178+ if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
31179+ lockdep_off();
31180+ err = do_xino_fwrite(func, file, buf, size, pos);
31181+ lockdep_on();
31182+ } else {
31183+ /*
31184+ * it breaks RLIMIT_FSIZE and normal user's limit,
31185+ * users should care about quota and real 'filesystem full.'
31186+ */
1facf9fc 31187+ int wkq_err;
31188+ struct do_xino_fwrite_args args = {
31189+ .errp = &err,
31190+ .func = func,
31191+ .file = file,
31192+ .buf = buf,
31193+ .size = size,
31194+ .pos = pos
31195+ };
31196+
31197+ wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
31198+ if (unlikely(wkq_err))
31199+ err = wkq_err;
b752ccd1 31200+ }
1facf9fc 31201+
31202+ return err;
31203+}
31204+
31205+/* ---------------------------------------------------------------------- */
31206+
31207+/*
31208+ * create a new xinofile at the same place/path as @base_file.
31209+ */
31210+struct file *au_xino_create2(struct file *base_file, struct file *copy_src)
31211+{
31212+ struct file *file;
4a4d8108 31213+ struct dentry *base, *parent;
523b37e3 31214+ struct inode *dir, *delegated;
1facf9fc 31215+ struct qstr *name;
1308ab2a 31216+ struct path path;
4a4d8108 31217+ int err;
1facf9fc 31218+
31219+ base = base_file->f_dentry;
31220+ parent = base->d_parent; /* dir inode is locked */
31221+ dir = parent->d_inode;
31222+ IMustLock(dir);
31223+
31224+ file = ERR_PTR(-EINVAL);
31225+ name = &base->d_name;
4a4d8108
AM
31226+ path.dentry = vfsub_lookup_one_len(name->name, parent, name->len);
31227+ if (IS_ERR(path.dentry)) {
31228+ file = (void *)path.dentry;
523b37e3
AM
31229+ pr_err("%pd lookup err %ld\n",
31230+ base, PTR_ERR(path.dentry));
1facf9fc 31231+ goto out;
31232+ }
31233+
31234+ /* no need to mnt_want_write() since we call dentry_open() later */
4a4d8108 31235+ err = vfs_create(dir, path.dentry, S_IRUGO | S_IWUGO, NULL);
1facf9fc 31236+ if (unlikely(err)) {
31237+ file = ERR_PTR(err);
523b37e3 31238+ pr_err("%pd create err %d\n", base, err);
1facf9fc 31239+ goto out_dput;
31240+ }
31241+
c06a8ce3 31242+ path.mnt = base_file->f_path.mnt;
4a4d8108 31243+ file = vfsub_dentry_open(&path,
7f207e10 31244+ O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
2cbb1c4b 31245+ /* | __FMODE_NONOTIFY */);
1facf9fc 31246+ if (IS_ERR(file)) {
523b37e3 31247+ pr_err("%pd open err %ld\n", base, PTR_ERR(file));
1facf9fc 31248+ goto out_dput;
31249+ }
31250+
523b37e3
AM
31251+ delegated = NULL;
31252+ err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
31253+ if (unlikely(err == -EWOULDBLOCK)) {
31254+ pr_warn("cannot retry for NFSv4 delegation"
31255+ " for an internal unlink\n");
31256+ iput(delegated);
31257+ }
1facf9fc 31258+ if (unlikely(err)) {
523b37e3 31259+ pr_err("%pd unlink err %d\n", base, err);
1facf9fc 31260+ goto out_fput;
31261+ }
31262+
31263+ if (copy_src) {
31264+ /* no one can touch copy_src xino */
c06a8ce3 31265+ err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
1facf9fc 31266+ if (unlikely(err)) {
523b37e3 31267+ pr_err("%pd copy err %d\n", base, err);
1facf9fc 31268+ goto out_fput;
31269+ }
31270+ }
31271+ goto out_dput; /* success */
31272+
4f0767ce 31273+out_fput:
1facf9fc 31274+ fput(file);
31275+ file = ERR_PTR(err);
4f0767ce 31276+out_dput:
4a4d8108 31277+ dput(path.dentry);
4f0767ce 31278+out:
1facf9fc 31279+ return file;
31280+}
31281+
31282+struct au_xino_lock_dir {
31283+ struct au_hinode *hdir;
31284+ struct dentry *parent;
31285+ struct mutex *mtx;
31286+};
31287+
31288+static void au_xino_lock_dir(struct super_block *sb, struct file *xino,
31289+ struct au_xino_lock_dir *ldir)
31290+{
31291+ aufs_bindex_t brid, bindex;
31292+
31293+ ldir->hdir = NULL;
31294+ bindex = -1;
31295+ brid = au_xino_brid(sb);
31296+ if (brid >= 0)
31297+ bindex = au_br_index(sb, brid);
31298+ if (bindex >= 0) {
31299+ ldir->hdir = au_hi(sb->s_root->d_inode, bindex);
4a4d8108 31300+ au_hn_imtx_lock_nested(ldir->hdir, AuLsc_I_PARENT);
1facf9fc 31301+ } else {
31302+ ldir->parent = dget_parent(xino->f_dentry);
31303+ ldir->mtx = &ldir->parent->d_inode->i_mutex;
31304+ mutex_lock_nested(ldir->mtx, AuLsc_I_PARENT);
31305+ }
31306+}
31307+
31308+static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
31309+{
31310+ if (ldir->hdir)
4a4d8108 31311+ au_hn_imtx_unlock(ldir->hdir);
1facf9fc 31312+ else {
31313+ mutex_unlock(ldir->mtx);
31314+ dput(ldir->parent);
31315+ }
31316+}
31317+
31318+/* ---------------------------------------------------------------------- */
31319+
31320+/* trucate xino files asynchronously */
31321+
31322+int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex)
31323+{
31324+ int err;
392086de
AM
31325+ unsigned long jiffy;
31326+ blkcnt_t blocks;
1facf9fc 31327+ aufs_bindex_t bi, bend;
392086de 31328+ struct kstatfs *st;
1facf9fc 31329+ struct au_branch *br;
31330+ struct file *new_xino, *file;
31331+ struct super_block *h_sb;
31332+ struct au_xino_lock_dir ldir;
31333+
392086de
AM
31334+ err = -ENOMEM;
31335+ st = kzalloc(sizeof(*st), GFP_NOFS);
31336+ if (unlikely(!st))
31337+ goto out;
31338+
1facf9fc 31339+ err = -EINVAL;
31340+ bend = au_sbend(sb);
31341+ if (unlikely(bindex < 0 || bend < bindex))
392086de 31342+ goto out_st;
1facf9fc 31343+ br = au_sbr(sb, bindex);
31344+ file = br->br_xino.xi_file;
31345+ if (!file)
392086de
AM
31346+ goto out_st;
31347+
31348+ err = vfs_statfs(&file->f_path, st);
31349+ if (unlikely(err))
31350+ AuErr1("statfs err %d, ignored\n", err);
31351+ jiffy = jiffies;
31352+ blocks = file_inode(file)->i_blocks;
31353+ pr_info("begin truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
31354+ bindex, (u64)blocks, st->f_bfree, st->f_blocks);
1facf9fc 31355+
31356+ au_xino_lock_dir(sb, file, &ldir);
31357+ /* mnt_want_write() is unnecessary here */
31358+ new_xino = au_xino_create2(file, file);
31359+ au_xino_unlock_dir(&ldir);
31360+ err = PTR_ERR(new_xino);
392086de
AM
31361+ if (IS_ERR(new_xino)) {
31362+ pr_err("err %d, ignored\n", err);
31363+ goto out_st;
31364+ }
1facf9fc 31365+ err = 0;
31366+ fput(file);
31367+ br->br_xino.xi_file = new_xino;
31368+
86dc4139 31369+ h_sb = au_br_sb(br);
1facf9fc 31370+ for (bi = 0; bi <= bend; bi++) {
31371+ if (unlikely(bi == bindex))
31372+ continue;
31373+ br = au_sbr(sb, bi);
86dc4139 31374+ if (au_br_sb(br) != h_sb)
1facf9fc 31375+ continue;
31376+
31377+ fput(br->br_xino.xi_file);
31378+ br->br_xino.xi_file = new_xino;
31379+ get_file(new_xino);
31380+ }
31381+
392086de
AM
31382+ err = vfs_statfs(&new_xino->f_path, st);
31383+ if (!err) {
31384+ pr_info("end truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
31385+ bindex, (u64)file_inode(new_xino)->i_blocks,
31386+ st->f_bfree, st->f_blocks);
31387+ if (file_inode(new_xino)->i_blocks < blocks)
31388+ au_sbi(sb)->si_xino_jiffy = jiffy;
31389+ } else
31390+ AuErr1("statfs err %d, ignored\n", err);
31391+
31392+out_st:
31393+ kfree(st);
4f0767ce 31394+out:
1facf9fc 31395+ return err;
31396+}
31397+
31398+struct xino_do_trunc_args {
31399+ struct super_block *sb;
31400+ struct au_branch *br;
31401+};
31402+
31403+static void xino_do_trunc(void *_args)
31404+{
31405+ struct xino_do_trunc_args *args = _args;
31406+ struct super_block *sb;
31407+ struct au_branch *br;
31408+ struct inode *dir;
31409+ int err;
31410+ aufs_bindex_t bindex;
31411+
31412+ err = 0;
31413+ sb = args->sb;
31414+ dir = sb->s_root->d_inode;
31415+ br = args->br;
31416+
31417+ si_noflush_write_lock(sb);
31418+ ii_read_lock_parent(dir);
31419+ bindex = au_br_index(sb, br->br_id);
31420+ err = au_xino_trunc(sb, bindex);
1facf9fc 31421+ ii_read_unlock(dir);
31422+ if (unlikely(err))
392086de 31423+ pr_warn("err b%d, (%d)\n", bindex, err);
1facf9fc 31424+ atomic_dec(&br->br_xino_running);
31425+ atomic_dec(&br->br_count);
1facf9fc 31426+ si_write_unlock(sb);
027c5e7a 31427+ au_nwt_done(&au_sbi(sb)->si_nowait);
1facf9fc 31428+ kfree(args);
31429+}
31430+
392086de
AM
31431+static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
31432+{
31433+ int err;
31434+ struct kstatfs st;
31435+ struct au_sbinfo *sbinfo;
31436+
31437+ /* todo: si_xino_expire and the ratio should be customizable */
31438+ sbinfo = au_sbi(sb);
31439+ if (time_before(jiffies,
31440+ sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
31441+ return 0;
31442+
31443+ /* truncation border */
31444+ err = vfs_statfs(&br->br_xino.xi_file->f_path, &st);
31445+ if (unlikely(err)) {
31446+ AuErr1("statfs err %d, ignored\n", err);
31447+ return 0;
31448+ }
31449+ if (div64_u64(st.f_bfree * 100, st.f_blocks) >= AUFS_XINO_DEF_TRUNC)
31450+ return 0;
31451+
31452+ return 1;
31453+}
31454+
1facf9fc 31455+static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
31456+{
31457+ struct xino_do_trunc_args *args;
31458+ int wkq_err;
31459+
392086de 31460+ if (!xino_trunc_test(sb, br))
1facf9fc 31461+ return;
31462+
31463+ if (atomic_inc_return(&br->br_xino_running) > 1)
31464+ goto out;
31465+
31466+ /* lock and kfree() will be called in trunc_xino() */
31467+ args = kmalloc(sizeof(*args), GFP_NOFS);
31468+ if (unlikely(!args)) {
31469+ AuErr1("no memory\n");
31470+ goto out_args;
31471+ }
31472+
e49829fe 31473+ atomic_inc(&br->br_count);
1facf9fc 31474+ args->sb = sb;
31475+ args->br = br;
53392da6 31476+ wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
1facf9fc 31477+ if (!wkq_err)
31478+ return; /* success */
31479+
4a4d8108 31480+ pr_err("wkq %d\n", wkq_err);
e49829fe 31481+ atomic_dec(&br->br_count);
1facf9fc 31482+
4f0767ce 31483+out_args:
1facf9fc 31484+ kfree(args);
4f0767ce 31485+out:
e49829fe 31486+ atomic_dec(&br->br_xino_running);
1facf9fc 31487+}
31488+
31489+/* ---------------------------------------------------------------------- */
31490+
31491+static int au_xino_do_write(au_writef_t write, struct file *file,
31492+ ino_t h_ino, ino_t ino)
31493+{
31494+ loff_t pos;
31495+ ssize_t sz;
31496+
31497+ pos = h_ino;
31498+ if (unlikely(au_loff_max / sizeof(ino) - 1 < pos)) {
31499+ AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
31500+ return -EFBIG;
31501+ }
31502+ pos *= sizeof(ino);
31503+ sz = xino_fwrite(write, file, &ino, sizeof(ino), &pos);
31504+ if (sz == sizeof(ino))
31505+ return 0; /* success */
31506+
31507+ AuIOErr("write failed (%zd)\n", sz);
31508+ return -EIO;
31509+}
31510+
31511+/*
31512+ * write @ino to the xinofile for the specified branch{@sb, @bindex}
31513+ * at the position of @h_ino.
31514+ * even if @ino is zero, it is written to the xinofile and means no entry.
31515+ * if the size of the xino file on a specific filesystem exceeds the watermark,
31516+ * try truncating it.
31517+ */
31518+int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
31519+ ino_t ino)
31520+{
31521+ int err;
31522+ unsigned int mnt_flags;
31523+ struct au_branch *br;
31524+
31525+ BUILD_BUG_ON(sizeof(long long) != sizeof(au_loff_max)
31526+ || ((loff_t)-1) > 0);
dece6358 31527+ SiMustAnyLock(sb);
1facf9fc 31528+
31529+ mnt_flags = au_mntflags(sb);
31530+ if (!au_opt_test(mnt_flags, XINO))
31531+ return 0;
31532+
31533+ br = au_sbr(sb, bindex);
31534+ err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
31535+ h_ino, ino);
31536+ if (!err) {
31537+ if (au_opt_test(mnt_flags, TRUNC_XINO)
86dc4139 31538+ && au_test_fs_trunc_xino(au_br_sb(br)))
1facf9fc 31539+ xino_try_trunc(sb, br);
31540+ return 0; /* success */
31541+ }
31542+
31543+ AuIOErr("write failed (%d)\n", err);
31544+ return -EIO;
31545+}
31546+
31547+/* ---------------------------------------------------------------------- */
31548+
31549+/* aufs inode number bitmap */
31550+
31551+static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
31552+static ino_t xib_calc_ino(unsigned long pindex, int bit)
31553+{
31554+ ino_t ino;
31555+
31556+ AuDebugOn(bit < 0 || page_bits <= bit);
31557+ ino = AUFS_FIRST_INO + pindex * page_bits + bit;
31558+ return ino;
31559+}
31560+
31561+static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
31562+{
31563+ AuDebugOn(ino < AUFS_FIRST_INO);
31564+ ino -= AUFS_FIRST_INO;
31565+ *pindex = ino / page_bits;
31566+ *bit = ino % page_bits;
31567+}
31568+
31569+static int xib_pindex(struct super_block *sb, unsigned long pindex)
31570+{
31571+ int err;
31572+ loff_t pos;
31573+ ssize_t sz;
31574+ struct au_sbinfo *sbinfo;
31575+ struct file *xib;
31576+ unsigned long *p;
31577+
31578+ sbinfo = au_sbi(sb);
31579+ MtxMustLock(&sbinfo->si_xib_mtx);
31580+ AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
31581+ || !au_opt_test(sbinfo->si_mntflags, XINO));
31582+
31583+ if (pindex == sbinfo->si_xib_last_pindex)
31584+ return 0;
31585+
31586+ xib = sbinfo->si_xib;
31587+ p = sbinfo->si_xib_buf;
31588+ pos = sbinfo->si_xib_last_pindex;
31589+ pos *= PAGE_SIZE;
31590+ sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
31591+ if (unlikely(sz != PAGE_SIZE))
31592+ goto out;
31593+
31594+ pos = pindex;
31595+ pos *= PAGE_SIZE;
c06a8ce3 31596+ if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
1facf9fc 31597+ sz = xino_fread(sbinfo->si_xread, xib, p, PAGE_SIZE, &pos);
31598+ else {
31599+ memset(p, 0, PAGE_SIZE);
31600+ sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
31601+ }
31602+ if (sz == PAGE_SIZE) {
31603+ sbinfo->si_xib_last_pindex = pindex;
31604+ return 0; /* success */
31605+ }
31606+
4f0767ce 31607+out:
b752ccd1
AM
31608+ AuIOErr1("write failed (%zd)\n", sz);
31609+ err = sz;
31610+ if (sz >= 0)
31611+ err = -EIO;
31612+ return err;
31613+}
31614+
31615+/* ---------------------------------------------------------------------- */
31616+
31617+static void au_xib_clear_bit(struct inode *inode)
31618+{
31619+ int err, bit;
31620+ unsigned long pindex;
31621+ struct super_block *sb;
31622+ struct au_sbinfo *sbinfo;
31623+
31624+ AuDebugOn(inode->i_nlink);
31625+
31626+ sb = inode->i_sb;
31627+ xib_calc_bit(inode->i_ino, &pindex, &bit);
31628+ AuDebugOn(page_bits <= bit);
31629+ sbinfo = au_sbi(sb);
31630+ mutex_lock(&sbinfo->si_xib_mtx);
31631+ err = xib_pindex(sb, pindex);
31632+ if (!err) {
31633+ clear_bit(bit, sbinfo->si_xib_buf);
31634+ sbinfo->si_xib_next_bit = bit;
31635+ }
31636+ mutex_unlock(&sbinfo->si_xib_mtx);
31637+}
31638+
31639+/* for s_op->delete_inode() */
31640+void au_xino_delete_inode(struct inode *inode, const int unlinked)
31641+{
31642+ int err;
31643+ unsigned int mnt_flags;
31644+ aufs_bindex_t bindex, bend, bi;
31645+ unsigned char try_trunc;
31646+ struct au_iinfo *iinfo;
31647+ struct super_block *sb;
31648+ struct au_hinode *hi;
31649+ struct inode *h_inode;
31650+ struct au_branch *br;
31651+ au_writef_t xwrite;
31652+
31653+ sb = inode->i_sb;
31654+ mnt_flags = au_mntflags(sb);
31655+ if (!au_opt_test(mnt_flags, XINO)
31656+ || inode->i_ino == AUFS_ROOT_INO)
31657+ return;
31658+
31659+ if (unlinked) {
31660+ au_xigen_inc(inode);
31661+ au_xib_clear_bit(inode);
31662+ }
31663+
31664+ iinfo = au_ii(inode);
31665+ if (!iinfo)
31666+ return;
1facf9fc 31667+
b752ccd1
AM
31668+ bindex = iinfo->ii_bstart;
31669+ if (bindex < 0)
31670+ return;
1facf9fc 31671+
b752ccd1
AM
31672+ xwrite = au_sbi(sb)->si_xwrite;
31673+ try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
31674+ hi = iinfo->ii_hinode + bindex;
31675+ bend = iinfo->ii_bend;
31676+ for (; bindex <= bend; bindex++, hi++) {
31677+ h_inode = hi->hi_inode;
31678+ if (!h_inode
31679+ || (!unlinked && h_inode->i_nlink))
31680+ continue;
1facf9fc 31681+
b752ccd1
AM
31682+ /* inode may not be revalidated */
31683+ bi = au_br_index(sb, hi->hi_id);
31684+ if (bi < 0)
31685+ continue;
1facf9fc 31686+
b752ccd1
AM
31687+ br = au_sbr(sb, bi);
31688+ err = au_xino_do_write(xwrite, br->br_xino.xi_file,
31689+ h_inode->i_ino, /*ino*/0);
31690+ if (!err && try_trunc
86dc4139 31691+ && au_test_fs_trunc_xino(au_br_sb(br)))
b752ccd1 31692+ xino_try_trunc(sb, br);
1facf9fc 31693+ }
1facf9fc 31694+}
31695+
31696+/* get an unused inode number from bitmap */
31697+ino_t au_xino_new_ino(struct super_block *sb)
31698+{
31699+ ino_t ino;
31700+ unsigned long *p, pindex, ul, pend;
31701+ struct au_sbinfo *sbinfo;
31702+ struct file *file;
31703+ int free_bit, err;
31704+
31705+ if (!au_opt_test(au_mntflags(sb), XINO))
31706+ return iunique(sb, AUFS_FIRST_INO);
31707+
31708+ sbinfo = au_sbi(sb);
31709+ mutex_lock(&sbinfo->si_xib_mtx);
31710+ p = sbinfo->si_xib_buf;
31711+ free_bit = sbinfo->si_xib_next_bit;
31712+ if (free_bit < page_bits && !test_bit(free_bit, p))
31713+ goto out; /* success */
31714+ free_bit = find_first_zero_bit(p, page_bits);
31715+ if (free_bit < page_bits)
31716+ goto out; /* success */
31717+
31718+ pindex = sbinfo->si_xib_last_pindex;
31719+ for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
31720+ err = xib_pindex(sb, ul);
31721+ if (unlikely(err))
31722+ goto out_err;
31723+ free_bit = find_first_zero_bit(p, page_bits);
31724+ if (free_bit < page_bits)
31725+ goto out; /* success */
31726+ }
31727+
31728+ file = sbinfo->si_xib;
c06a8ce3 31729+ pend = vfsub_f_size_read(file) / PAGE_SIZE;
1facf9fc 31730+ for (ul = pindex + 1; ul <= pend; ul++) {
31731+ err = xib_pindex(sb, ul);
31732+ if (unlikely(err))
31733+ goto out_err;
31734+ free_bit = find_first_zero_bit(p, page_bits);
31735+ if (free_bit < page_bits)
31736+ goto out; /* success */
31737+ }
31738+ BUG();
31739+
4f0767ce 31740+out:
1facf9fc 31741+ set_bit(free_bit, p);
7f207e10 31742+ sbinfo->si_xib_next_bit = free_bit + 1;
1facf9fc 31743+ pindex = sbinfo->si_xib_last_pindex;
31744+ mutex_unlock(&sbinfo->si_xib_mtx);
31745+ ino = xib_calc_ino(pindex, free_bit);
31746+ AuDbg("i%lu\n", (unsigned long)ino);
31747+ return ino;
4f0767ce 31748+out_err:
1facf9fc 31749+ mutex_unlock(&sbinfo->si_xib_mtx);
31750+ AuDbg("i0\n");
31751+ return 0;
31752+}
31753+
31754+/*
31755+ * read @ino from xinofile for the specified branch{@sb, @bindex}
31756+ * at the position of @h_ino.
31757+ * if @ino does not exist and @do_new is true, get new one.
31758+ */
31759+int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
31760+ ino_t *ino)
31761+{
31762+ int err;
31763+ ssize_t sz;
31764+ loff_t pos;
31765+ struct file *file;
31766+ struct au_sbinfo *sbinfo;
31767+
31768+ *ino = 0;
31769+ if (!au_opt_test(au_mntflags(sb), XINO))
31770+ return 0; /* no xino */
31771+
31772+ err = 0;
31773+ sbinfo = au_sbi(sb);
31774+ pos = h_ino;
31775+ if (unlikely(au_loff_max / sizeof(*ino) - 1 < pos)) {
31776+ AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
31777+ return -EFBIG;
31778+ }
31779+ pos *= sizeof(*ino);
31780+
31781+ file = au_sbr(sb, bindex)->br_xino.xi_file;
c06a8ce3 31782+ if (vfsub_f_size_read(file) < pos + sizeof(*ino))
1facf9fc 31783+ return 0; /* no ino */
31784+
31785+ sz = xino_fread(sbinfo->si_xread, file, ino, sizeof(*ino), &pos);
31786+ if (sz == sizeof(*ino))
31787+ return 0; /* success */
31788+
31789+ err = sz;
31790+ if (unlikely(sz >= 0)) {
31791+ err = -EIO;
31792+ AuIOErr("xino read error (%zd)\n", sz);
31793+ }
31794+
31795+ return err;
31796+}
31797+
31798+/* ---------------------------------------------------------------------- */
31799+
31800+/* create and set a new xino file */
31801+
31802+struct file *au_xino_create(struct super_block *sb, char *fname, int silent)
31803+{
31804+ struct file *file;
31805+ struct dentry *h_parent, *d;
31806+ struct inode *h_dir;
31807+ int err;
31808+
31809+ /*
31810+ * at mount-time, and the xino file is the default path,
4a4d8108 31811+ * hnotify is disabled so we have no notify events to ignore.
1facf9fc 31812+ * when a user specified the xino, we cannot get au_hdir to be ignored.
31813+ */
7f207e10 31814+ file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
2cbb1c4b 31815+ /* | __FMODE_NONOTIFY */,
1facf9fc 31816+ S_IRUGO | S_IWUGO);
31817+ if (IS_ERR(file)) {
31818+ if (!silent)
4a4d8108 31819+ pr_err("open %s(%ld)\n", fname, PTR_ERR(file));
1facf9fc 31820+ return file;
31821+ }
31822+
31823+ /* keep file count */
31824+ h_parent = dget_parent(file->f_dentry);
31825+ h_dir = h_parent->d_inode;
31826+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
31827+ /* mnt_want_write() is unnecessary here */
523b37e3
AM
31828+ /* no delegation since it is just created */
31829+ err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL, /*force*/0);
1facf9fc 31830+ mutex_unlock(&h_dir->i_mutex);
31831+ dput(h_parent);
31832+ if (unlikely(err)) {
31833+ if (!silent)
4a4d8108 31834+ pr_err("unlink %s(%d)\n", fname, err);
1facf9fc 31835+ goto out;
31836+ }
31837+
31838+ err = -EINVAL;
31839+ d = file->f_dentry;
31840+ if (unlikely(sb == d->d_sb)) {
31841+ if (!silent)
4a4d8108 31842+ pr_err("%s must be outside\n", fname);
1facf9fc 31843+ goto out;
31844+ }
31845+ if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
31846+ if (!silent)
4a4d8108
AM
31847+ pr_err("xino doesn't support %s(%s)\n",
31848+ fname, au_sbtype(d->d_sb));
1facf9fc 31849+ goto out;
31850+ }
31851+ return file; /* success */
31852+
4f0767ce 31853+out:
1facf9fc 31854+ fput(file);
31855+ file = ERR_PTR(err);
31856+ return file;
31857+}
31858+
31859+/*
31860+ * find another branch who is on the same filesystem of the specified
31861+ * branch{@btgt}. search until @bend.
31862+ */
31863+static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
31864+ aufs_bindex_t bend)
31865+{
31866+ aufs_bindex_t bindex;
31867+ struct super_block *tgt_sb = au_sbr_sb(sb, btgt);
31868+
31869+ for (bindex = 0; bindex < btgt; bindex++)
31870+ if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
31871+ return bindex;
31872+ for (bindex++; bindex <= bend; bindex++)
31873+ if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
31874+ return bindex;
31875+ return -1;
31876+}
31877+
31878+/* ---------------------------------------------------------------------- */
31879+
31880+/*
31881+ * initialize the xinofile for the specified branch @br
31882+ * at the place/path where @base_file indicates.
31883+ * test whether another branch is on the same filesystem or not,
31884+ * if @do_test is true.
31885+ */
31886+int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
31887+ struct file *base_file, int do_test)
31888+{
31889+ int err;
31890+ ino_t ino;
31891+ aufs_bindex_t bend, bindex;
31892+ struct au_branch *shared_br, *b;
31893+ struct file *file;
31894+ struct super_block *tgt_sb;
31895+
31896+ shared_br = NULL;
31897+ bend = au_sbend(sb);
31898+ if (do_test) {
86dc4139 31899+ tgt_sb = au_br_sb(br);
1facf9fc 31900+ for (bindex = 0; bindex <= bend; bindex++) {
31901+ b = au_sbr(sb, bindex);
86dc4139 31902+ if (tgt_sb == au_br_sb(b)) {
1facf9fc 31903+ shared_br = b;
31904+ break;
31905+ }
31906+ }
31907+ }
31908+
31909+ if (!shared_br || !shared_br->br_xino.xi_file) {
31910+ struct au_xino_lock_dir ldir;
31911+
31912+ au_xino_lock_dir(sb, base_file, &ldir);
31913+ /* mnt_want_write() is unnecessary here */
31914+ file = au_xino_create2(base_file, NULL);
31915+ au_xino_unlock_dir(&ldir);
31916+ err = PTR_ERR(file);
31917+ if (IS_ERR(file))
31918+ goto out;
31919+ br->br_xino.xi_file = file;
31920+ } else {
31921+ br->br_xino.xi_file = shared_br->br_xino.xi_file;
31922+ get_file(br->br_xino.xi_file);
31923+ }
31924+
31925+ ino = AUFS_ROOT_INO;
31926+ err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
31927+ h_ino, ino);
b752ccd1
AM
31928+ if (unlikely(err)) {
31929+ fput(br->br_xino.xi_file);
31930+ br->br_xino.xi_file = NULL;
31931+ }
1facf9fc 31932+
4f0767ce 31933+out:
1facf9fc 31934+ return err;
31935+}
31936+
31937+/* ---------------------------------------------------------------------- */
31938+
31939+/* trucate a xino bitmap file */
31940+
31941+/* todo: slow */
31942+static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
31943+{
31944+ int err, bit;
31945+ ssize_t sz;
31946+ unsigned long pindex;
31947+ loff_t pos, pend;
31948+ struct au_sbinfo *sbinfo;
31949+ au_readf_t func;
31950+ ino_t *ino;
31951+ unsigned long *p;
31952+
31953+ err = 0;
31954+ sbinfo = au_sbi(sb);
dece6358 31955+ MtxMustLock(&sbinfo->si_xib_mtx);
1facf9fc 31956+ p = sbinfo->si_xib_buf;
31957+ func = sbinfo->si_xread;
c06a8ce3 31958+ pend = vfsub_f_size_read(file);
1facf9fc 31959+ pos = 0;
31960+ while (pos < pend) {
31961+ sz = xino_fread(func, file, page, PAGE_SIZE, &pos);
31962+ err = sz;
31963+ if (unlikely(sz <= 0))
31964+ goto out;
31965+
31966+ err = 0;
31967+ for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
31968+ if (unlikely(*ino < AUFS_FIRST_INO))
31969+ continue;
31970+
31971+ xib_calc_bit(*ino, &pindex, &bit);
31972+ AuDebugOn(page_bits <= bit);
31973+ err = xib_pindex(sb, pindex);
31974+ if (!err)
31975+ set_bit(bit, p);
31976+ else
31977+ goto out;
31978+ }
31979+ }
31980+
4f0767ce 31981+out:
1facf9fc 31982+ return err;
31983+}
31984+
31985+static int xib_restore(struct super_block *sb)
31986+{
31987+ int err;
31988+ aufs_bindex_t bindex, bend;
31989+ void *page;
31990+
31991+ err = -ENOMEM;
31992+ page = (void *)__get_free_page(GFP_NOFS);
31993+ if (unlikely(!page))
31994+ goto out;
31995+
31996+ err = 0;
31997+ bend = au_sbend(sb);
31998+ for (bindex = 0; !err && bindex <= bend; bindex++)
31999+ if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0)
32000+ err = do_xib_restore
32001+ (sb, au_sbr(sb, bindex)->br_xino.xi_file, page);
32002+ else
32003+ AuDbg("b%d\n", bindex);
32004+ free_page((unsigned long)page);
32005+
4f0767ce 32006+out:
1facf9fc 32007+ return err;
32008+}
32009+
32010+int au_xib_trunc(struct super_block *sb)
32011+{
32012+ int err;
32013+ ssize_t sz;
32014+ loff_t pos;
32015+ struct au_xino_lock_dir ldir;
32016+ struct au_sbinfo *sbinfo;
32017+ unsigned long *p;
32018+ struct file *file;
32019+
dece6358
AM
32020+ SiMustWriteLock(sb);
32021+
1facf9fc 32022+ err = 0;
32023+ sbinfo = au_sbi(sb);
32024+ if (!au_opt_test(sbinfo->si_mntflags, XINO))
32025+ goto out;
32026+
32027+ file = sbinfo->si_xib;
c06a8ce3 32028+ if (vfsub_f_size_read(file) <= PAGE_SIZE)
1facf9fc 32029+ goto out;
32030+
32031+ au_xino_lock_dir(sb, file, &ldir);
32032+ /* mnt_want_write() is unnecessary here */
32033+ file = au_xino_create2(sbinfo->si_xib, NULL);
32034+ au_xino_unlock_dir(&ldir);
32035+ err = PTR_ERR(file);
32036+ if (IS_ERR(file))
32037+ goto out;
32038+ fput(sbinfo->si_xib);
32039+ sbinfo->si_xib = file;
32040+
32041+ p = sbinfo->si_xib_buf;
32042+ memset(p, 0, PAGE_SIZE);
32043+ pos = 0;
32044+ sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xib, p, PAGE_SIZE, &pos);
32045+ if (unlikely(sz != PAGE_SIZE)) {
32046+ err = sz;
32047+ AuIOErr("err %d\n", err);
32048+ if (sz >= 0)
32049+ err = -EIO;
32050+ goto out;
32051+ }
32052+
32053+ mutex_lock(&sbinfo->si_xib_mtx);
32054+ /* mnt_want_write() is unnecessary here */
32055+ err = xib_restore(sb);
32056+ mutex_unlock(&sbinfo->si_xib_mtx);
32057+
32058+out:
32059+ return err;
32060+}
32061+
32062+/* ---------------------------------------------------------------------- */
32063+
32064+/*
32065+ * xino mount option handlers
32066+ */
32067+static au_readf_t find_readf(struct file *h_file)
32068+{
32069+ const struct file_operations *fop = h_file->f_op;
32070+
523b37e3
AM
32071+ if (fop->read)
32072+ return fop->read;
32073+ if (fop->aio_read)
32074+ return do_sync_read;
1facf9fc 32075+ return ERR_PTR(-ENOSYS);
32076+}
32077+
32078+static au_writef_t find_writef(struct file *h_file)
32079+{
32080+ const struct file_operations *fop = h_file->f_op;
32081+
523b37e3
AM
32082+ if (fop->write)
32083+ return fop->write;
32084+ if (fop->aio_write)
32085+ return do_sync_write;
1facf9fc 32086+ return ERR_PTR(-ENOSYS);
32087+}
32088+
32089+/* xino bitmap */
32090+static void xino_clear_xib(struct super_block *sb)
32091+{
32092+ struct au_sbinfo *sbinfo;
32093+
dece6358
AM
32094+ SiMustWriteLock(sb);
32095+
1facf9fc 32096+ sbinfo = au_sbi(sb);
32097+ sbinfo->si_xread = NULL;
32098+ sbinfo->si_xwrite = NULL;
32099+ if (sbinfo->si_xib)
32100+ fput(sbinfo->si_xib);
32101+ sbinfo->si_xib = NULL;
32102+ free_page((unsigned long)sbinfo->si_xib_buf);
32103+ sbinfo->si_xib_buf = NULL;
32104+}
32105+
32106+static int au_xino_set_xib(struct super_block *sb, struct file *base)
32107+{
32108+ int err;
32109+ loff_t pos;
32110+ struct au_sbinfo *sbinfo;
32111+ struct file *file;
32112+
dece6358
AM
32113+ SiMustWriteLock(sb);
32114+
1facf9fc 32115+ sbinfo = au_sbi(sb);
32116+ file = au_xino_create2(base, sbinfo->si_xib);
32117+ err = PTR_ERR(file);
32118+ if (IS_ERR(file))
32119+ goto out;
32120+ if (sbinfo->si_xib)
32121+ fput(sbinfo->si_xib);
32122+ sbinfo->si_xib = file;
32123+ sbinfo->si_xread = find_readf(file);
32124+ sbinfo->si_xwrite = find_writef(file);
32125+
32126+ err = -ENOMEM;
32127+ if (!sbinfo->si_xib_buf)
32128+ sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
32129+ if (unlikely(!sbinfo->si_xib_buf))
32130+ goto out_unset;
32131+
32132+ sbinfo->si_xib_last_pindex = 0;
32133+ sbinfo->si_xib_next_bit = 0;
c06a8ce3 32134+ if (vfsub_f_size_read(file) < PAGE_SIZE) {
1facf9fc 32135+ pos = 0;
32136+ err = xino_fwrite(sbinfo->si_xwrite, file, sbinfo->si_xib_buf,
32137+ PAGE_SIZE, &pos);
32138+ if (unlikely(err != PAGE_SIZE))
32139+ goto out_free;
32140+ }
32141+ err = 0;
32142+ goto out; /* success */
32143+
4f0767ce 32144+out_free:
1facf9fc 32145+ free_page((unsigned long)sbinfo->si_xib_buf);
b752ccd1
AM
32146+ sbinfo->si_xib_buf = NULL;
32147+ if (err >= 0)
32148+ err = -EIO;
4f0767ce 32149+out_unset:
b752ccd1
AM
32150+ fput(sbinfo->si_xib);
32151+ sbinfo->si_xib = NULL;
32152+ sbinfo->si_xread = NULL;
32153+ sbinfo->si_xwrite = NULL;
4f0767ce 32154+out:
b752ccd1 32155+ return err;
1facf9fc 32156+}
32157+
b752ccd1
AM
32158+/* xino for each branch */
32159+static void xino_clear_br(struct super_block *sb)
32160+{
32161+ aufs_bindex_t bindex, bend;
32162+ struct au_branch *br;
1facf9fc 32163+
b752ccd1
AM
32164+ bend = au_sbend(sb);
32165+ for (bindex = 0; bindex <= bend; bindex++) {
32166+ br = au_sbr(sb, bindex);
32167+ if (!br || !br->br_xino.xi_file)
32168+ continue;
32169+
32170+ fput(br->br_xino.xi_file);
32171+ br->br_xino.xi_file = NULL;
32172+ }
32173+}
32174+
32175+static int au_xino_set_br(struct super_block *sb, struct file *base)
1facf9fc 32176+{
32177+ int err;
b752ccd1
AM
32178+ ino_t ino;
32179+ aufs_bindex_t bindex, bend, bshared;
32180+ struct {
32181+ struct file *old, *new;
32182+ } *fpair, *p;
32183+ struct au_branch *br;
32184+ struct inode *inode;
32185+ au_writef_t writef;
1facf9fc 32186+
b752ccd1
AM
32187+ SiMustWriteLock(sb);
32188+
32189+ err = -ENOMEM;
32190+ bend = au_sbend(sb);
32191+ fpair = kcalloc(bend + 1, sizeof(*fpair), GFP_NOFS);
32192+ if (unlikely(!fpair))
1facf9fc 32193+ goto out;
32194+
b752ccd1
AM
32195+ inode = sb->s_root->d_inode;
32196+ ino = AUFS_ROOT_INO;
32197+ writef = au_sbi(sb)->si_xwrite;
32198+ for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
32199+ br = au_sbr(sb, bindex);
32200+ bshared = is_sb_shared(sb, bindex, bindex - 1);
32201+ if (bshared >= 0) {
32202+ /* shared xino */
32203+ *p = fpair[bshared];
32204+ get_file(p->new);
32205+ }
32206+
32207+ if (!p->new) {
32208+ /* new xino */
32209+ p->old = br->br_xino.xi_file;
32210+ p->new = au_xino_create2(base, br->br_xino.xi_file);
32211+ err = PTR_ERR(p->new);
32212+ if (IS_ERR(p->new)) {
32213+ p->new = NULL;
32214+ goto out_pair;
32215+ }
32216+ }
32217+
32218+ err = au_xino_do_write(writef, p->new,
32219+ au_h_iptr(inode, bindex)->i_ino, ino);
32220+ if (unlikely(err))
32221+ goto out_pair;
32222+ }
32223+
32224+ for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
32225+ br = au_sbr(sb, bindex);
32226+ if (br->br_xino.xi_file)
32227+ fput(br->br_xino.xi_file);
32228+ get_file(p->new);
32229+ br->br_xino.xi_file = p->new;
32230+ }
1facf9fc 32231+
4f0767ce 32232+out_pair:
b752ccd1
AM
32233+ for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++)
32234+ if (p->new)
32235+ fput(p->new);
32236+ else
32237+ break;
32238+ kfree(fpair);
4f0767ce 32239+out:
1facf9fc 32240+ return err;
32241+}
b752ccd1
AM
32242+
32243+void au_xino_clr(struct super_block *sb)
32244+{
32245+ struct au_sbinfo *sbinfo;
32246+
32247+ au_xigen_clr(sb);
32248+ xino_clear_xib(sb);
32249+ xino_clear_br(sb);
32250+ sbinfo = au_sbi(sb);
32251+ /* lvalue, do not call au_mntflags() */
32252+ au_opt_clr(sbinfo->si_mntflags, XINO);
32253+}
32254+
32255+int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount)
32256+{
32257+ int err, skip;
32258+ struct dentry *parent, *cur_parent;
32259+ struct qstr *dname, *cur_name;
32260+ struct file *cur_xino;
32261+ struct inode *dir;
32262+ struct au_sbinfo *sbinfo;
32263+
32264+ SiMustWriteLock(sb);
32265+
32266+ err = 0;
32267+ sbinfo = au_sbi(sb);
32268+ parent = dget_parent(xino->file->f_dentry);
32269+ if (remount) {
32270+ skip = 0;
32271+ dname = &xino->file->f_dentry->d_name;
32272+ cur_xino = sbinfo->si_xib;
32273+ if (cur_xino) {
32274+ cur_parent = dget_parent(cur_xino->f_dentry);
32275+ cur_name = &cur_xino->f_dentry->d_name;
32276+ skip = (cur_parent == parent
38d290e6 32277+ && au_qstreq(dname, cur_name));
b752ccd1
AM
32278+ dput(cur_parent);
32279+ }
32280+ if (skip)
32281+ goto out;
32282+ }
32283+
32284+ au_opt_set(sbinfo->si_mntflags, XINO);
32285+ dir = parent->d_inode;
32286+ mutex_lock_nested(&dir->i_mutex, AuLsc_I_PARENT);
32287+ /* mnt_want_write() is unnecessary here */
32288+ err = au_xino_set_xib(sb, xino->file);
32289+ if (!err)
32290+ err = au_xigen_set(sb, xino->file);
32291+ if (!err)
32292+ err = au_xino_set_br(sb, xino->file);
32293+ mutex_unlock(&dir->i_mutex);
32294+ if (!err)
32295+ goto out; /* success */
32296+
32297+ /* reset all */
32298+ AuIOErr("failed creating xino(%d).\n", err);
32299+
4f0767ce 32300+out:
b752ccd1
AM
32301+ dput(parent);
32302+ return err;
32303+}
32304+
32305+/* ---------------------------------------------------------------------- */
32306+
32307+/*
32308+ * create a xinofile at the default place/path.
32309+ */
32310+struct file *au_xino_def(struct super_block *sb)
32311+{
32312+ struct file *file;
32313+ char *page, *p;
32314+ struct au_branch *br;
32315+ struct super_block *h_sb;
32316+ struct path path;
32317+ aufs_bindex_t bend, bindex, bwr;
32318+
32319+ br = NULL;
32320+ bend = au_sbend(sb);
32321+ bwr = -1;
32322+ for (bindex = 0; bindex <= bend; bindex++) {
32323+ br = au_sbr(sb, bindex);
32324+ if (au_br_writable(br->br_perm)
86dc4139 32325+ && !au_test_fs_bad_xino(au_br_sb(br))) {
b752ccd1
AM
32326+ bwr = bindex;
32327+ break;
32328+ }
32329+ }
32330+
7f207e10
AM
32331+ if (bwr >= 0) {
32332+ file = ERR_PTR(-ENOMEM);
537831f9 32333+ page = (void *)__get_free_page(GFP_NOFS);
7f207e10
AM
32334+ if (unlikely(!page))
32335+ goto out;
86dc4139 32336+ path.mnt = au_br_mnt(br);
7f207e10
AM
32337+ path.dentry = au_h_dptr(sb->s_root, bwr);
32338+ p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
32339+ file = (void *)p;
32340+ if (!IS_ERR(p)) {
32341+ strcat(p, "/" AUFS_XINO_FNAME);
32342+ AuDbg("%s\n", p);
32343+ file = au_xino_create(sb, p, /*silent*/0);
32344+ if (!IS_ERR(file))
32345+ au_xino_brid_set(sb, br->br_id);
32346+ }
537831f9 32347+ free_page((unsigned long)page);
7f207e10
AM
32348+ } else {
32349+ file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0);
32350+ if (IS_ERR(file))
32351+ goto out;
32352+ h_sb = file->f_dentry->d_sb;
32353+ if (unlikely(au_test_fs_bad_xino(h_sb))) {
32354+ pr_err("xino doesn't support %s(%s)\n",
32355+ AUFS_XINO_DEFPATH, au_sbtype(h_sb));
32356+ fput(file);
32357+ file = ERR_PTR(-EINVAL);
32358+ }
32359+ if (!IS_ERR(file))
32360+ au_xino_brid_set(sb, -1);
32361+ }
0c5527e5 32362+
7f207e10
AM
32363+out:
32364+ return file;
32365+}
32366+
32367+/* ---------------------------------------------------------------------- */
32368+
32369+int au_xino_path(struct seq_file *seq, struct file *file)
32370+{
32371+ int err;
32372+
32373+ err = au_seq_path(seq, &file->f_path);
32374+ if (unlikely(err < 0))
32375+ goto out;
32376+
32377+ err = 0;
32378+#define Deleted "\\040(deleted)"
32379+ seq->count -= sizeof(Deleted) - 1;
32380+ AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
32381+ sizeof(Deleted) - 1));
32382+#undef Deleted
32383+
32384+out:
32385+ return err;
32386+}
537831f9
AM
32387diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
32388--- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
38d290e6 32389+++ linux/include/uapi/linux/aufs_type.h 2014-06-24 00:14:06.792719853 +0200
523b37e3 32390@@ -0,0 +1,281 @@
7f207e10 32391+/*
523b37e3 32392+ * Copyright (C) 2005-2014 Junjiro R. Okajima
7f207e10
AM
32393+ *
32394+ * This program, aufs is free software; you can redistribute it and/or modify
32395+ * it under the terms of the GNU General Public License as published by
32396+ * the Free Software Foundation; either version 2 of the License, or
32397+ * (at your option) any later version.
32398+ *
32399+ * This program is distributed in the hope that it will be useful,
32400+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
32401+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32402+ * GNU General Public License for more details.
32403+ *
32404+ * You should have received a copy of the GNU General Public License
523b37e3 32405+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
7f207e10
AM
32406+ */
32407+
32408+#ifndef __AUFS_TYPE_H__
32409+#define __AUFS_TYPE_H__
32410+
f6c5ef8b
AM
32411+#define AUFS_NAME "aufs"
32412+
9dbd164d 32413+#ifdef __KERNEL__
f6c5ef8b
AM
32414+/*
32415+ * define it before including all other headers.
32416+ * sched.h may use pr_* macros before defining "current", so define the
32417+ * no-current version first, and re-define later.
32418+ */
32419+#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
32420+#include <linux/sched.h>
32421+#undef pr_fmt
a2a7ad62
AM
32422+#define pr_fmt(fmt) \
32423+ AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
32424+ (int)sizeof(current->comm), current->comm, current->pid
9dbd164d
AM
32425+#else
32426+#include <stdint.h>
32427+#include <sys/types.h>
f6c5ef8b 32428+#endif /* __KERNEL__ */
7f207e10 32429+
f6c5ef8b
AM
32430+#include <linux/limits.h>
32431+
38d290e6 32432+#define AUFS_VERSION "3.15-20140623"
7f207e10
AM
32433+
32434+/* todo? move this to linux-2.6.19/include/magic.h */
32435+#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
32436+
32437+/* ---------------------------------------------------------------------- */
32438+
32439+#ifdef CONFIG_AUFS_BRANCH_MAX_127
9dbd164d 32440+typedef int8_t aufs_bindex_t;
7f207e10
AM
32441+#define AUFS_BRANCH_MAX 127
32442+#else
9dbd164d 32443+typedef int16_t aufs_bindex_t;
7f207e10
AM
32444+#ifdef CONFIG_AUFS_BRANCH_MAX_511
32445+#define AUFS_BRANCH_MAX 511
32446+#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
32447+#define AUFS_BRANCH_MAX 1023
32448+#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
32449+#define AUFS_BRANCH_MAX 32767
32450+#endif
32451+#endif
32452+
32453+#ifdef __KERNEL__
32454+#ifndef AUFS_BRANCH_MAX
32455+#error unknown CONFIG_AUFS_BRANCH_MAX value
32456+#endif
32457+#endif /* __KERNEL__ */
32458+
32459+/* ---------------------------------------------------------------------- */
32460+
7f207e10
AM
32461+#define AUFS_FSTYPE AUFS_NAME
32462+
32463+#define AUFS_ROOT_INO 2
32464+#define AUFS_FIRST_INO 11
32465+
32466+#define AUFS_WH_PFX ".wh."
32467+#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
32468+#define AUFS_WH_TMP_LEN 4
86dc4139 32469+/* a limit for rmdir/rename a dir and copyup */
7f207e10
AM
32470+#define AUFS_MAX_NAMELEN (NAME_MAX \
32471+ - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
32472+ - 1 /* dot */\
32473+ - AUFS_WH_TMP_LEN) /* hex */
32474+#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
32475+#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
392086de
AM
32476+#define AUFS_XINO_DEF_SEC 30 /* seconds */
32477+#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
7f207e10
AM
32478+#define AUFS_DIRWH_DEF 3
32479+#define AUFS_RDCACHE_DEF 10 /* seconds */
027c5e7a 32480+#define AUFS_RDCACHE_MAX 3600 /* seconds */
7f207e10
AM
32481+#define AUFS_RDBLK_DEF 512 /* bytes */
32482+#define AUFS_RDHASH_DEF 32
32483+#define AUFS_WKQ_NAME AUFS_NAME "d"
027c5e7a
AM
32484+#define AUFS_MFS_DEF_SEC 30 /* seconds */
32485+#define AUFS_MFS_MAX_SEC 3600 /* seconds */
86dc4139 32486+#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
7f207e10
AM
32487+
32488+/* pseudo-link maintenace under /proc */
32489+#define AUFS_PLINK_MAINT_NAME "plink_maint"
32490+#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
32491+#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
32492+
32493+#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
32494+#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
32495+
32496+#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
32497+#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
32498+#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
32499+
32500+/* doubly whiteouted */
32501+#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
32502+#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
32503+#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
32504+
1e00d052 32505+/* branch permissions and attributes */
7f207e10
AM
32506+#define AUFS_BRPERM_RW "rw"
32507+#define AUFS_BRPERM_RO "ro"
32508+#define AUFS_BRPERM_RR "rr"
1e00d052
AM
32509+#define AUFS_BRRATTR_WH "wh"
32510+#define AUFS_BRWATTR_NLWH "nolwh"
86dc4139 32511+#define AUFS_BRATTR_UNPIN "unpin"
7f207e10
AM
32512+
32513+/* ---------------------------------------------------------------------- */
32514+
32515+/* ioctl */
32516+enum {
32517+ /* readdir in userspace */
32518+ AuCtl_RDU,
32519+ AuCtl_RDU_INO,
32520+
32521+ /* pathconf wrapper */
027c5e7a
AM
32522+ AuCtl_WBR_FD,
32523+
32524+ /* busy inode */
c2b27bf2
AM
32525+ AuCtl_IBUSY,
32526+
32527+ /* move-down */
32528+ AuCtl_MVDOWN
7f207e10
AM
32529+};
32530+
32531+/* borrowed from linux/include/linux/kernel.h */
32532+#ifndef ALIGN
32533+#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
32534+#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
32535+#endif
32536+
32537+/* borrowed from linux/include/linux/compiler-gcc3.h */
32538+#ifndef __aligned
32539+#define __aligned(x) __attribute__((aligned(x)))
53392da6
AM
32540+#endif
32541+
32542+#ifdef __KERNEL__
32543+#ifndef __packed
7f207e10
AM
32544+#define __packed __attribute__((packed))
32545+#endif
53392da6 32546+#endif
7f207e10
AM
32547+
32548+struct au_rdu_cookie {
9dbd164d
AM
32549+ uint64_t h_pos;
32550+ int16_t bindex;
32551+ uint8_t flags;
32552+ uint8_t pad;
32553+ uint32_t generation;
7f207e10
AM
32554+} __aligned(8);
32555+
32556+struct au_rdu_ent {
9dbd164d
AM
32557+ uint64_t ino;
32558+ int16_t bindex;
32559+ uint8_t type;
32560+ uint8_t nlen;
32561+ uint8_t wh;
7f207e10
AM
32562+ char name[0];
32563+} __aligned(8);
32564+
32565+static inline int au_rdu_len(int nlen)
32566+{
32567+ /* include the terminating NULL */
32568+ return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
9dbd164d 32569+ sizeof(uint64_t));
7f207e10
AM
32570+}
32571+
32572+union au_rdu_ent_ul {
32573+ struct au_rdu_ent __user *e;
9dbd164d 32574+ uint64_t ul;
7f207e10
AM
32575+};
32576+
32577+enum {
32578+ AufsCtlRduV_SZ,
32579+ AufsCtlRduV_End
32580+};
32581+
32582+struct aufs_rdu {
32583+ /* input */
32584+ union {
9dbd164d
AM
32585+ uint64_t sz; /* AuCtl_RDU */
32586+ uint64_t nent; /* AuCtl_RDU_INO */
7f207e10
AM
32587+ };
32588+ union au_rdu_ent_ul ent;
9dbd164d 32589+ uint16_t verify[AufsCtlRduV_End];
7f207e10
AM
32590+
32591+ /* input/output */
9dbd164d 32592+ uint32_t blk;
7f207e10
AM
32593+
32594+ /* output */
32595+ union au_rdu_ent_ul tail;
32596+ /* number of entries which were added in a single call */
9dbd164d
AM
32597+ uint64_t rent;
32598+ uint8_t full;
32599+ uint8_t shwh;
7f207e10
AM
32600+
32601+ struct au_rdu_cookie cookie;
32602+} __aligned(8);
32603+
1e00d052
AM
32604+/* ---------------------------------------------------------------------- */
32605+
32606+struct aufs_wbr_fd {
9dbd164d
AM
32607+ uint32_t oflags;
32608+ int16_t brid;
1e00d052
AM
32609+} __aligned(8);
32610+
32611+/* ---------------------------------------------------------------------- */
32612+
027c5e7a 32613+struct aufs_ibusy {
9dbd164d
AM
32614+ uint64_t ino, h_ino;
32615+ int16_t bindex;
027c5e7a
AM
32616+} __aligned(8);
32617+
1e00d052
AM
32618+/* ---------------------------------------------------------------------- */
32619+
392086de
AM
32620+/* error code for move-down */
32621+/* the actual message strings are implemented in aufs-util.git */
32622+enum {
32623+ EAU_MVDOWN_OPAQUE = 1,
32624+ EAU_MVDOWN_WHITEOUT,
32625+ EAU_MVDOWN_UPPER,
32626+ EAU_MVDOWN_BOTTOM,
32627+ EAU_MVDOWN_NOUPPER,
32628+ EAU_MVDOWN_NOLOWERBR,
32629+ EAU_Last
32630+};
32631+
c2b27bf2 32632+/* flags for move-down */
392086de
AM
32633+#define AUFS_MVDOWN_DMSG 1
32634+#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
32635+#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
32636+#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
32637+#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
32638+#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
32639+#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
32640+#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
32641+#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
c2b27bf2
AM
32642+/* will be added more */
32643+
392086de
AM
32644+enum {
32645+ AUFS_MVDOWN_UPPER,
32646+ AUFS_MVDOWN_LOWER,
32647+ AUFS_MVDOWN_NARRAY
32648+};
32649+
c2b27bf2 32650+struct aufs_mvdown {
392086de
AM
32651+ uint32_t flags;
32652+ struct {
32653+ int16_t bindex;
32654+ int16_t brid;
32655+ } a[AUFS_MVDOWN_NARRAY];
32656+ int8_t au_errno;
c2b27bf2
AM
32657+ /* will be added more */
32658+} __aligned(8);
32659+
32660+/* ---------------------------------------------------------------------- */
32661+
7f207e10
AM
32662+#define AuCtlType 'A'
32663+#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
32664+#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
1e00d052
AM
32665+#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
32666+ struct aufs_wbr_fd)
027c5e7a 32667+#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
392086de
AM
32668+#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
32669+ struct aufs_mvdown)
7f207e10
AM
32670+
32671+#endif /* __AUFS_TYPE_H__ */
38d290e6 32672aufs3.15 loopback patch
93a1a2a2
JR
32673
32674diff --git a/drivers/block/loop.c b/drivers/block/loop.c
38d290e6 32675index 138104b..227f7fa 100644
93a1a2a2
JR
32676--- a/drivers/block/loop.c
32677+++ b/drivers/block/loop.c
32678@@ -514,7 +514,7 @@ out:
32679 }
32680
32681 struct switch_request {
32682- struct file *file;
32683+ struct file *file, *virt_file;
32684 struct completion wait;
32685 };
32686
32687@@ -576,7 +576,8 @@ static int loop_thread(void *data)
32688 * First it needs to flush existing IO, it does this by sending a magic
32689 * BIO down the pipe. The completion of this BIO does the actual switch.
32690 */
32691-static int loop_switch(struct loop_device *lo, struct file *file)
32692+static int loop_switch(struct loop_device *lo, struct file *file,
32693+ struct file *virt_file)
32694 {
32695 struct switch_request w;
32696 struct bio *bio = bio_alloc(GFP_KERNEL, 0);
32697@@ -584,6 +585,7 @@ static int loop_switch(struct loop_device *lo, struct file *file)
32698 return -ENOMEM;
32699 init_completion(&w.wait);
32700 w.file = file;
32701+ w.virt_file = virt_file;
32702 bio->bi_private = &w;
32703 bio->bi_bdev = NULL;
32704 loop_make_request(lo->lo_queue, bio);
32705@@ -600,7 +602,7 @@ static int loop_flush(struct loop_device *lo)
32706 if (!lo->lo_thread)
32707 return 0;
32708
32709- return loop_switch(lo, NULL);
32710+ return loop_switch(lo, NULL, NULL);
32711 }
32712
32713 /*
32714@@ -619,6 +621,7 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
32715 mapping = file->f_mapping;
32716 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
32717 lo->lo_backing_file = file;
32718+ lo->lo_backing_virt_file = p->virt_file;
32719 lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ?
32720 mapping->host->i_bdev->bd_block_size : PAGE_SIZE;
32721 lo->old_gfp_mask = mapping_gfp_mask(mapping);
32722@@ -627,6 +630,13 @@ out:
32723 complete(&p->wait);
32724 }
32725
32726+static struct file *loop_real_file(struct file *file)
32727+{
32728+ struct file *f = NULL;
32729+ if (file->f_dentry->d_sb->s_op->real_loop)
32730+ f = file->f_dentry->d_sb->s_op->real_loop(file);
32731+ return f;
32732+}
32733
32734 /*
32735 * loop_change_fd switched the backing store of a loopback device to
32736@@ -640,6 +650,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
32737 unsigned int arg)
32738 {
32739 struct file *file, *old_file;
32740+ struct file *f, *virt_file = NULL, *old_virt_file;
32741 struct inode *inode;
32742 int error;
32743
32744@@ -656,9 +667,16 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
32745 file = fget(arg);
32746 if (!file)
32747 goto out;
32748+ f = loop_real_file(file);
32749+ if (f) {
32750+ virt_file = file;
32751+ file = f;
32752+ get_file(file);
32753+ }
32754
32755 inode = file->f_mapping->host;
32756 old_file = lo->lo_backing_file;
32757+ old_virt_file = lo->lo_backing_virt_file;
32758
32759 error = -EINVAL;
32760
32761@@ -670,17 +688,21 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
32762 goto out_putf;
32763
32764 /* and ... switch */
32765- error = loop_switch(lo, file);
32766+ error = loop_switch(lo, file, virt_file);
32767 if (error)
32768 goto out_putf;
32769
32770 fput(old_file);
32771+ if (old_virt_file)
32772+ fput(old_virt_file);
32773 if (lo->lo_flags & LO_FLAGS_PARTSCAN)
32774 ioctl_by_bdev(bdev, BLKRRPART, 0);
32775 return 0;
32776
32777 out_putf:
32778 fput(file);
32779+ if (virt_file)
32780+ fput(virt_file);
32781 out:
32782 return error;
32783 }
32784@@ -841,7 +863,7 @@ static void loop_config_discard(struct loop_device *lo)
32785 static int loop_set_fd(struct loop_device *lo, fmode_t mode,
32786 struct block_device *bdev, unsigned int arg)
32787 {
32788- struct file *file, *f;
32789+ struct file *file, *f, *virt_file = NULL;
32790 struct inode *inode;
32791 struct address_space *mapping;
32792 unsigned lo_blocksize;
32793@@ -856,6 +878,12 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
32794 file = fget(arg);
32795 if (!file)
32796 goto out;
32797+ f = loop_real_file(file);
32798+ if (f) {
32799+ virt_file = file;
32800+ file = f;
32801+ get_file(file);
32802+ }
32803
32804 error = -EBUSY;
32805 if (lo->lo_state != Lo_unbound)
32806@@ -904,6 +932,7 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
32807 lo->lo_device = bdev;
32808 lo->lo_flags = lo_flags;
32809 lo->lo_backing_file = file;
32810+ lo->lo_backing_virt_file = virt_file;
32811 lo->transfer = transfer_none;
32812 lo->ioctl = NULL;
32813 lo->lo_sizelimit = 0;
32814@@ -948,6 +977,7 @@ out_clr:
32815 lo->lo_thread = NULL;
32816 lo->lo_device = NULL;
32817 lo->lo_backing_file = NULL;
32818+ lo->lo_backing_virt_file = NULL;
32819 lo->lo_flags = 0;
32820 set_capacity(lo->lo_disk, 0);
32821 invalidate_bdev(bdev);
32822@@ -957,6 +987,8 @@ out_clr:
32823 lo->lo_state = Lo_unbound;
32824 out_putf:
32825 fput(file);
32826+ if (virt_file)
32827+ fput(virt_file);
32828 out:
32829 /* This is safe: open() is still holding a reference. */
32830 module_put(THIS_MODULE);
32831@@ -1003,6 +1035,7 @@ loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
32832 static int loop_clr_fd(struct loop_device *lo)
32833 {
32834 struct file *filp = lo->lo_backing_file;
32835+ struct file *virt_filp = lo->lo_backing_virt_file;
32836 gfp_t gfp = lo->old_gfp_mask;
32837 struct block_device *bdev = lo->lo_device;
32838
32839@@ -1036,6 +1069,7 @@ static int loop_clr_fd(struct loop_device *lo)
32840
32841 spin_lock_irq(&lo->lo_lock);
32842 lo->lo_backing_file = NULL;
32843+ lo->lo_backing_virt_file = NULL;
32844 spin_unlock_irq(&lo->lo_lock);
32845
32846 loop_release_xfer(lo);
32847@@ -1078,6 +1112,8 @@ static int loop_clr_fd(struct loop_device *lo)
32848 * bd_mutex which is usually taken before lo_ctl_mutex.
32849 */
32850 fput(filp);
32851+ if (virt_filp)
32852+ fput(virt_filp);
32853 return 0;
32854 }
32855
32856diff --git a/drivers/block/loop.h b/drivers/block/loop.h
32857index 90df5d6..cb91822 100644
32858--- a/drivers/block/loop.h
32859+++ b/drivers/block/loop.h
32860@@ -44,7 +44,7 @@ struct loop_device {
32861 int (*ioctl)(struct loop_device *, int cmd,
32862 unsigned long arg);
32863
32864- struct file * lo_backing_file;
32865+ struct file * lo_backing_file, *lo_backing_virt_file;
32866 struct block_device *lo_device;
32867 unsigned lo_blocksize;
32868 void *key_data;
32869diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
38d290e6 32870index 510919e..f9317c9 100644
93a1a2a2
JR
32871--- a/fs/aufs/f_op.c
32872+++ b/fs/aufs/f_op.c
38d290e6 32873@@ -345,7 +345,7 @@ static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
93a1a2a2
JR
32874 err = -EINVAL;
32875 h_file = au_hf_top(file);
32876 get_file(h_file);
32877- if (au_test_loopback_kthread()) {
32878+ if (0 && au_test_loopback_kthread()) {
32879 au_warn_loopback(h_file->f_dentry->d_sb);
32880 if (file->f_mapping != h_file->f_mapping) {
32881 file->f_mapping = h_file->f_mapping;
32882diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
32883index 3b03b52..4ab749d 100644
32884--- a/fs/aufs/loop.c
32885+++ b/fs/aufs/loop.c
32886@@ -130,3 +130,19 @@ void au_loopback_fin(void)
32887 symbol_put(loop_backing_file);
32888 kfree(au_warn_loopback_array);
32889 }
32890+
32891+/* ---------------------------------------------------------------------- */
32892+
32893+/* support the loopback block device insude aufs */
32894+
32895+struct file *aufs_real_loop(struct file *file)
32896+{
32897+ struct file *f;
32898+
32899+ BUG_ON(!au_test_aufs(file->f_dentry->d_sb));
32900+ fi_read_lock(file);
32901+ f = au_hf_top(file);
32902+ fi_read_unlock(file);
32903+ AuDebugOn(!f);
32904+ return f;
32905+}
32906diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
32907index da8b756..28cb7ea 100644
32908--- a/fs/aufs/loop.h
32909+++ b/fs/aufs/loop.h
32910@@ -25,7 +25,11 @@ void au_warn_loopback(struct super_block *h_sb);
32911
32912 int au_loopback_init(void);
32913 void au_loopback_fin(void);
32914+
32915+struct file *aufs_real_loop(struct file *file);
32916 #else
32917+AuStub(struct file *, loop_backing_file, return NULL)
32918+
32919 AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
32920 struct dentry *h_adding)
32921 AuStubInt0(au_test_loopback_kthread, void)
32922@@ -33,6 +37,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
32923
32924 AuStubInt0(au_loopback_init, void)
32925 AuStubVoid(au_loopback_fin, void)
32926+
32927+AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
32928 #endif /* BLK_DEV_LOOP */
32929
32930 #endif /* __KERNEL__ */
32931diff --git a/fs/aufs/super.c b/fs/aufs/super.c
32932index b609e5a..e3909ed 100644
32933--- a/fs/aufs/super.c
32934+++ b/fs/aufs/super.c
32935@@ -807,7 +807,10 @@ static const struct super_operations aufs_sop = {
32936 .statfs = aufs_statfs,
32937 .put_super = aufs_put_super,
32938 .sync_fs = aufs_sync_fs,
32939- .remount_fs = aufs_remount_fs
32940+ .remount_fs = aufs_remount_fs,
32941+#ifdef CONFIG_AUFS_BDEV_LOOP
32942+ .real_loop = aufs_real_loop
32943+#endif
32944 };
32945
32946 /* ---------------------------------------------------------------------- */
32947diff --git a/include/linux/fs.h b/include/linux/fs.h
38d290e6 32948index 3ff3e89..13017e6 100644
93a1a2a2
JR
32949--- a/include/linux/fs.h
32950+++ b/include/linux/fs.h
38d290e6 32951@@ -1554,6 +1554,10 @@ struct super_operations {
93a1a2a2
JR
32952 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
32953 long (*nr_cached_objects)(struct super_block *, int);
32954 long (*free_cached_objects)(struct super_block *, long, int);
32955+#if defined(CONFIG_BLK_DEV_LOOP) || defined(CONFIG_BLK_DEV_LOOP_MODULE)
32956+ /* and aufs */
32957+ struct file *(*real_loop)(struct file *);
32958+#endif
32959 };
32960
32961 /*
This page took 6.219806 seconds and 4 git commands to generate.