]> git.pld-linux.org Git - packages/kernel.git/blame - kernel-aufs4.patch
- 4.2.2
[packages/kernel.git] / kernel-aufs4.patch
CommitLineData
5527c038 1aufs4.x-rcN kbuild patch
7f207e10
AM
2
3diff --git a/fs/Kconfig b/fs/Kconfig
5527c038 4index 011f433..b1083f6 100644
7f207e10
AM
5--- a/fs/Kconfig
6+++ b/fs/Kconfig
5527c038
JR
7@@ -218,6 +218,7 @@ source "fs/pstore/Kconfig"
8 source "fs/sysv/Kconfig"
7e9cd9fe 9 source "fs/ufs/Kconfig"
7f207e10
AM
10 source "fs/exofs/Kconfig"
11+source "fs/aufs/Kconfig"
12
13 endif # MISC_FILESYSTEMS
14
15diff --git a/fs/Makefile b/fs/Makefile
c2c0f25c 16index cb20e4b..dd81418 100644
7f207e10
AM
17--- a/fs/Makefile
18+++ b/fs/Makefile
c2c0f25c 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
c2c0f25c 25index 1ff9942..31efc0a 100644
c06a8ce3
AM
26--- a/include/uapi/linux/Kbuild
27+++ b/include/uapi/linux/Kbuild
5527c038 28@@ -59,6 +59,7 @@ header-y += atmsvc.h
03673fb0
JR
29 header-y += atm_tcp.h
30 header-y += atm_zatm.h
c06a8ce3
AM
31 header-y += audit.h
32+header-y += aufs_type.h
c06a8ce3 33 header-y += auto_fs4.h
03673fb0 34 header-y += auto_fs.h
c06a8ce3 35 header-y += auxvec.h
5527c038 36aufs4.x-rcN base patch
7f207e10 37
c1595e42 38diff --git a/MAINTAINERS b/MAINTAINERS
c2c0f25c 39index 8133cef..04beb19 100644
c1595e42
JR
40--- a/MAINTAINERS
41+++ b/MAINTAINERS
c2c0f25c 42@@ -1939,6 +1939,19 @@ F: include/linux/audit.h
c1595e42
JR
43 F: include/uapi/linux/audit.h
44 F: kernel/audit*
45
46+AUFS (advanced multi layered unification filesystem) FILESYSTEM
47+M: "J. R. Okajima" <hooanon05g@gmail.com>
48+L: linux-unionfs@vger.kernel.org
49+L: aufs-users@lists.sourceforge.net (members only)
50+W: http://aufs.sourceforge.net
5527c038 51+T: git://github.com/sfjro/aufs4-linux.git
c1595e42
JR
52+S: Supported
53+F: Documentation/filesystems/aufs/
54+F: Documentation/ABI/testing/debugfs-aufs
55+F: Documentation/ABI/testing/sysfs-aufs
56+F: fs/aufs/
57+F: include/uapi/linux/aufs_type.h
58+
59 AUXILIARY DISPLAY DRIVERS
60 M: Miguel Ojeda Sandonis <miguel.ojeda.sandonis@gmail.com>
61 W: http://miguelojeda.es/auxdisplay.htm
392086de 62diff --git a/drivers/block/loop.c b/drivers/block/loop.c
c2c0f25c 63index f7a4c9d..23103ad 100644
392086de
AM
64--- a/drivers/block/loop.c
65+++ b/drivers/block/loop.c
c2c0f25c 66@@ -560,6 +560,24 @@ static inline int is_loop_device(struct file *file)
392086de
AM
67 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
68 }
69
70+/*
71+ * for AUFS
72+ * no get/put for file.
73+ */
74+struct file *loop_backing_file(struct super_block *sb)
75+{
76+ struct file *ret;
77+ struct loop_device *l;
78+
79+ ret = NULL;
80+ if (MAJOR(sb->s_dev) == LOOP_MAJOR) {
81+ l = sb->s_bdev->bd_disk->private_data;
82+ ret = l->lo_backing_file;
83+ }
84+ return ret;
85+}
86+EXPORT_SYMBOL(loop_backing_file);
87+
88 /* loop sysfs attributes */
89
90 static ssize_t loop_attr_show(struct device *dev, char *page,
c1595e42 91diff --git a/fs/dcache.c b/fs/dcache.c
c2c0f25c 92index 7a3f3e5..0b40298 100644
c1595e42
JR
93--- a/fs/dcache.c
94+++ b/fs/dcache.c
5527c038 95@@ -1164,7 +1164,7 @@ enum d_walk_ret {
c1595e42
JR
96 *
97 * The @enter() and @finish() callbacks are called with d_lock held.
98 */
99-static void d_walk(struct dentry *parent, void *data,
100+void d_walk(struct dentry *parent, void *data,
101 enum d_walk_ret (*enter)(void *, struct dentry *),
102 void (*finish)(void *))
103 {
5527c038
JR
104diff --git a/fs/read_write.c b/fs/read_write.c
105index 819ef3f..fd0414e 100644
106--- a/fs/read_write.c
107+++ b/fs/read_write.c
108@@ -494,6 +494,28 @@ ssize_t __vfs_write(struct file *file, const char __user *p, size_t count,
109 }
110 EXPORT_SYMBOL(__vfs_write);
111
112+vfs_readf_t vfs_readf(struct file *file)
113+{
114+ const struct file_operations *fop = file->f_op;
115+
116+ if (fop->read)
117+ return fop->read;
118+ if (fop->read_iter)
119+ return new_sync_read;
120+ return ERR_PTR(-ENOSYS);
121+}
122+
123+vfs_writef_t vfs_writef(struct file *file)
124+{
125+ const struct file_operations *fop = file->f_op;
126+
127+ if (fop->write)
128+ return fop->write;
129+ if (fop->write_iter)
130+ return new_sync_write;
131+ return ERR_PTR(-ENOSYS);
132+}
133+
134 ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
135 {
136 mm_segment_t old_fs;
7f207e10 137diff --git a/fs/splice.c b/fs/splice.c
c2c0f25c 138index 5fc1e50..5f8385a 100644
7f207e10
AM
139--- a/fs/splice.c
140+++ b/fs/splice.c
c2c0f25c 141@@ -1102,8 +1102,8 @@ EXPORT_SYMBOL(generic_splice_sendpage);
7f207e10
AM
142 /*
143 * Attempt to initiate a splice from pipe to file.
144 */
145-static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
146- loff_t *ppos, size_t len, unsigned int flags)
147+long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
148+ loff_t *ppos, size_t len, unsigned int flags)
149 {
150 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
151 loff_t *, size_t, unsigned int);
c2c0f25c 152@@ -1119,9 +1119,9 @@ static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
7f207e10
AM
153 /*
154 * Attempt to initiate a splice from a file to a pipe.
155 */
156-static long do_splice_to(struct file *in, loff_t *ppos,
157- struct pipe_inode_info *pipe, size_t len,
158- unsigned int flags)
159+long do_splice_to(struct file *in, loff_t *ppos,
160+ struct pipe_inode_info *pipe, size_t len,
161+ unsigned int flags)
162 {
163 ssize_t (*splice_read)(struct file *, loff_t *,
164 struct pipe_inode_info *, size_t, unsigned int);
b912730e
AM
165diff --git a/include/linux/file.h b/include/linux/file.h
166index f87d308..9a290b3 100644
167--- a/include/linux/file.h
168+++ b/include/linux/file.h
169@@ -19,6 +19,7 @@ struct dentry;
170 struct path;
171 extern struct file *alloc_file(struct path *, fmode_t mode,
172 const struct file_operations *fop);
173+extern struct file *get_empty_filp(void);
174
175 static inline void fput_light(struct file *file, int fput_needed)
176 {
5527c038 177diff --git a/include/linux/fs.h b/include/linux/fs.h
c2c0f25c 178index a0653e5..86080ea 100644
5527c038
JR
179--- a/include/linux/fs.h
180+++ b/include/linux/fs.h
c2c0f25c 181@@ -1661,6 +1661,12 @@ ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
5527c038
JR
182 struct iovec *fast_pointer,
183 struct iovec **ret_pointer);
184
185+typedef ssize_t (*vfs_readf_t)(struct file *, char __user *, size_t, loff_t *);
186+typedef ssize_t (*vfs_writef_t)(struct file *, const char __user *, size_t,
187+ loff_t *);
188+vfs_readf_t vfs_readf(struct file *file);
189+vfs_writef_t vfs_writef(struct file *file);
190+
191 extern ssize_t __vfs_read(struct file *, char __user *, size_t, loff_t *);
192 extern ssize_t __vfs_write(struct file *, const char __user *, size_t, loff_t *);
193 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
1e00d052 194diff --git a/include/linux/splice.h b/include/linux/splice.h
076b876e 195index da2751d..2e0fca6 100644
1e00d052
AM
196--- a/include/linux/splice.h
197+++ b/include/linux/splice.h
076b876e 198@@ -83,4 +83,10 @@ extern void splice_shrink_spd(struct splice_pipe_desc *);
4b3da204
AM
199 extern void spd_release_page(struct splice_pipe_desc *, unsigned int);
200
201 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
1e00d052
AM
202+
203+extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
204+ loff_t *ppos, size_t len, unsigned int flags);
205+extern long do_splice_to(struct file *in, loff_t *ppos,
206+ struct pipe_inode_info *pipe, size_t len,
207+ unsigned int flags);
208 #endif
5527c038 209aufs4.x-rcN mmap patch
fb47a38f
JR
210
211diff --git a/fs/buffer.c b/fs/buffer.c
c2c0f25c 212index 1cf7a53..076aff86 100644
fb47a38f
JR
213--- a/fs/buffer.c
214+++ b/fs/buffer.c
c2c0f25c 215@@ -2473,7 +2473,7 @@ int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
fb47a38f
JR
216 * Update file times before taking page lock. We may end up failing the
217 * fault so this update may be superfluous but who really cares...
218 */
219- file_update_time(vma->vm_file);
220+ vma_file_update_time(vma);
221
222 ret = __block_page_mkwrite(vma, vmf, get_block);
223 sb_end_pagefault(sb);
c1595e42 224diff --git a/fs/proc/base.c b/fs/proc/base.c
c2c0f25c 225index 87782e8..b287e64 100644
c1595e42
JR
226--- a/fs/proc/base.c
227+++ b/fs/proc/base.c
c2c0f25c 228@@ -1934,7 +1934,7 @@ static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
c1595e42
JR
229 down_read(&mm->mmap_sem);
230 vma = find_exact_vma(mm, vm_start, vm_end);
231 if (vma && vma->vm_file) {
232- *path = vma->vm_file->f_path;
233+ *path = vma_pr_or_file(vma)->f_path;
234 path_get(path);
235 rc = 0;
236 }
fb47a38f 237diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
c2c0f25c 238index f8595e8..cb8eda0 100644
fb47a38f
JR
239--- a/fs/proc/nommu.c
240+++ b/fs/proc/nommu.c
076b876e 241@@ -45,7 +45,10 @@ static int nommu_region_show(struct seq_file *m, struct vm_region *region)
fb47a38f
JR
242 file = region->vm_file;
243
244 if (file) {
245- struct inode *inode = file_inode(region->vm_file);
246+ struct inode *inode;
076b876e 247+
fb47a38f
JR
248+ file = vmr_pr_or_file(region);
249+ inode = file_inode(file);
250 dev = inode->i_sb->s_dev;
251 ino = inode->i_ino;
252 }
253diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
c2c0f25c 254index ca1e091..8940e47 100644
fb47a38f
JR
255--- a/fs/proc/task_mmu.c
256+++ b/fs/proc/task_mmu.c
7e9cd9fe 257@@ -279,7 +279,10 @@ show_map_vma(struct seq_file *m, struct vm_area_struct *vma, int is_pid)
fb47a38f
JR
258 const char *name = NULL;
259
260 if (file) {
261- struct inode *inode = file_inode(vma->vm_file);
262+ struct inode *inode;
076b876e 263+
fb47a38f
JR
264+ file = vma_pr_or_file(vma);
265+ inode = file_inode(file);
266 dev = inode->i_sb->s_dev;
267 ino = inode->i_ino;
268 pgoff = ((loff_t)vma->vm_pgoff) << PAGE_SHIFT;
7e9cd9fe 269@@ -1479,7 +1482,7 @@ static int show_numa_map(struct seq_file *m, void *v, int is_pid)
076b876e
AM
270 struct proc_maps_private *proc_priv = &numa_priv->proc_maps;
271 struct vm_area_struct *vma = v;
272 struct numa_maps *md = &numa_priv->md;
273- struct file *file = vma->vm_file;
274+ struct file *file = vma_pr_or_file(vma);
076b876e 275 struct mm_struct *mm = vma->vm_mm;
7e9cd9fe
AM
276 struct mm_walk walk = {
277 .hugetlb_entry = gather_hugetlb_stats,
fb47a38f 278diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
c2c0f25c 279index e0d64c9..7aa92db 100644
fb47a38f
JR
280--- a/fs/proc/task_nommu.c
281+++ b/fs/proc/task_nommu.c
c1595e42 282@@ -160,7 +160,10 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma,
fb47a38f
JR
283 file = vma->vm_file;
284
285 if (file) {
286- struct inode *inode = file_inode(vma->vm_file);
287+ struct inode *inode;
076b876e 288+
b912730e 289+ file = vma_pr_or_file(vma);
fb47a38f
JR
290+ inode = file_inode(file);
291 dev = inode->i_sb->s_dev;
292 ino = inode->i_ino;
293 pgoff = (loff_t)vma->vm_pgoff << PAGE_SHIFT;
294diff --git a/include/linux/mm.h b/include/linux/mm.h
c2c0f25c 295index 2e872f9..2494ed2 100644
fb47a38f
JR
296--- a/include/linux/mm.h
297+++ b/include/linux/mm.h
c2c0f25c 298@@ -1173,6 +1173,28 @@ static inline int fixup_user_fault(struct task_struct *tsk,
fb47a38f
JR
299 }
300 #endif
301
076b876e
AM
302+extern void vma_do_file_update_time(struct vm_area_struct *, const char[], int);
303+extern struct file *vma_do_pr_or_file(struct vm_area_struct *, const char[],
304+ int);
305+extern void vma_do_get_file(struct vm_area_struct *, const char[], int);
306+extern void vma_do_fput(struct vm_area_struct *, const char[], int);
fb47a38f 307+
fb47a38f
JR
308+#define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
309+ __LINE__)
310+#define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
311+ __LINE__)
312+#define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
313+#define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
b912730e
AM
314+
315+#ifndef CONFIG_MMU
076b876e
AM
316+extern struct file *vmr_do_pr_or_file(struct vm_region *, const char[], int);
317+extern void vmr_do_fput(struct vm_region *, const char[], int);
318+
319+#define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
320+ __LINE__)
321+#define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
b912730e 322+#endif /* !CONFIG_MMU */
fb47a38f
JR
323+
324 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
325 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
326 void *buf, int len, int write);
327diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
c2c0f25c 328index 0038ac7..409940d 100644
fb47a38f
JR
329--- a/include/linux/mm_types.h
330+++ b/include/linux/mm_types.h
c2c0f25c 331@@ -259,6 +259,7 @@ struct vm_region {
fb47a38f
JR
332 unsigned long vm_top; /* region allocated to here */
333 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
334 struct file *vm_file; /* the backing file or NULL */
335+ struct file *vm_prfile; /* the virtual backing file or NULL */
336
337 int vm_usage; /* region usage count (access under nommu_region_sem) */
338 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
c2c0f25c 339@@ -323,6 +324,7 @@ struct vm_area_struct {
fb47a38f
JR
340 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
341 units, *not* PAGE_CACHE_SIZE */
342 struct file * vm_file; /* File we map to (can be NULL). */
343+ struct file *vm_prfile; /* shadow of vm_file */
344 void * vm_private_data; /* was vm_pte (shared mem) */
345
346 #ifndef CONFIG_MMU
347diff --git a/kernel/fork.c b/kernel/fork.c
c2c0f25c 348index 1bfefc6..a86bd7f 100644
fb47a38f
JR
349--- a/kernel/fork.c
350+++ b/kernel/fork.c
5527c038 351@@ -456,7 +456,7 @@ static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
fb47a38f
JR
352 struct inode *inode = file_inode(file);
353 struct address_space *mapping = file->f_mapping;
354
355- get_file(file);
356+ vma_get_file(tmp);
357 if (tmp->vm_flags & VM_DENYWRITE)
358 atomic_dec(&inode->i_writecount);
2000de60 359 i_mmap_lock_write(mapping);
076b876e 360diff --git a/mm/Makefile b/mm/Makefile
5527c038 361index 98c4eae..3f0c9b9 100644
076b876e
AM
362--- a/mm/Makefile
363+++ b/mm/Makefile
7e9cd9fe 364@@ -21,7 +21,7 @@ obj-y := filemap.o mempool.o oom_kill.o \
076b876e 365 mm_init.o mmu_context.o percpu.o slab_common.o \
c1595e42 366 compaction.o vmacache.o \
076b876e 367 interval_tree.o list_lru.o workingset.o \
7e9cd9fe
AM
368- debug.o $(mmu-y)
369+ prfile.o debug.o $(mmu-y)
076b876e
AM
370
371 obj-y += init-mm.o
372
fb47a38f 373diff --git a/mm/filemap.c b/mm/filemap.c
c2c0f25c 374index 1283fc8..128f18f 100644
fb47a38f
JR
375--- a/mm/filemap.c
376+++ b/mm/filemap.c
c2c0f25c 377@@ -2089,7 +2089,7 @@ int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
fb47a38f
JR
378 int ret = VM_FAULT_LOCKED;
379
380 sb_start_pagefault(inode->i_sb);
381- file_update_time(vma->vm_file);
382+ vma_file_update_time(vma);
383 lock_page(page);
384 if (page->mapping != inode->i_mapping) {
385 unlock_page(page);
fb47a38f 386diff --git a/mm/memory.c b/mm/memory.c
c2c0f25c 387index a84fbb7..13973d2 100644
fb47a38f
JR
388--- a/mm/memory.c
389+++ b/mm/memory.c
5527c038 390@@ -2034,7 +2034,7 @@ static inline int wp_page_reuse(struct mm_struct *mm,
fb47a38f 391 }
7e9cd9fe 392
b912730e
AM
393 if (!page_mkwrite)
394- file_update_time(vma->vm_file);
395+ vma_file_update_time(vma);
396 }
397
398 return VM_FAULT_WRITE;
fb47a38f 399diff --git a/mm/mmap.c b/mm/mmap.c
c2c0f25c 400index aa632ad..3ff75d3 100644
fb47a38f
JR
401--- a/mm/mmap.c
402+++ b/mm/mmap.c
7e9cd9fe 403@@ -274,7 +274,7 @@ static struct vm_area_struct *remove_vma(struct vm_area_struct *vma)
fb47a38f
JR
404 if (vma->vm_ops && vma->vm_ops->close)
405 vma->vm_ops->close(vma);
406 if (vma->vm_file)
407- fput(vma->vm_file);
408+ vma_fput(vma);
409 mpol_put(vma_policy(vma));
410 kmem_cache_free(vm_area_cachep, vma);
411 return next;
7e9cd9fe 412@@ -886,7 +886,7 @@ again: remove_next = 1 + (end > next->vm_end);
fb47a38f
JR
413 if (remove_next) {
414 if (file) {
415 uprobe_munmap(next, next->vm_start, next->vm_end);
416- fput(file);
417+ vma_fput(vma);
418 }
419 if (next->anon_vma)
420 anon_vma_merge(vma, next);
7e9cd9fe 421@@ -1671,8 +1671,8 @@ out:
35939ee7
JR
422 return addr;
423
fb47a38f 424 unmap_and_free_vma:
fb47a38f
JR
425+ vma_fput(vma);
426 vma->vm_file = NULL;
427- fput(file);
428
429 /* Undo any partial mapping done by a device driver. */
430 unmap_region(mm, vma, prev, vma->vm_start, vma->vm_end);
7e9cd9fe 431@@ -2473,7 +2473,7 @@ static int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
fb47a38f
JR
432 goto out_free_mpol;
433
434 if (new->vm_file)
435- get_file(new->vm_file);
436+ vma_get_file(new);
437
438 if (new->vm_ops && new->vm_ops->open)
439 new->vm_ops->open(new);
7e9cd9fe 440@@ -2492,7 +2492,7 @@ static int __split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
fb47a38f
JR
441 if (new->vm_ops && new->vm_ops->close)
442 new->vm_ops->close(new);
443 if (new->vm_file)
444- fput(new->vm_file);
445+ vma_fput(new);
446 unlink_anon_vmas(new);
447 out_free_mpol:
448 mpol_put(vma_policy(new));
7e9cd9fe
AM
449@@ -2635,7 +2635,6 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
450 struct vm_area_struct *vma;
451 unsigned long populate = 0;
452 unsigned long ret = -EINVAL;
453- struct file *file;
454
455 pr_warn_once("%s (%d) uses deprecated remap_file_pages() syscall. "
456 "See Documentation/vm/remap_file_pages.txt.\n",
457@@ -2679,10 +2678,10 @@ SYSCALL_DEFINE5(remap_file_pages, unsigned long, start, unsigned long, size,
458 munlock_vma_pages_range(vma, start, start + size);
459 }
460
461- file = get_file(vma->vm_file);
462+ vma_get_file(vma);
463 ret = do_mmap_pgoff(vma->vm_file, start, size,
464 prot, flags, pgoff, &populate);
465- fput(file);
466+ vma_fput(vma);
467 out:
468 up_write(&mm->mmap_sem);
469 if (populate)
5527c038 470@@ -2949,7 +2948,7 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
fb47a38f
JR
471 if (anon_vma_clone(new_vma, vma))
472 goto out_free_mempol;
473 if (new_vma->vm_file)
474- get_file(new_vma->vm_file);
475+ vma_get_file(new_vma);
476 if (new_vma->vm_ops && new_vma->vm_ops->open)
477 new_vma->vm_ops->open(new_vma);
478 vma_link(mm, new_vma, prev, rb_link, rb_parent);
fb47a38f 479diff --git a/mm/nommu.c b/mm/nommu.c
c2c0f25c 480index 58ea364..f937b7e 100644
fb47a38f
JR
481--- a/mm/nommu.c
482+++ b/mm/nommu.c
c2c0f25c 483@@ -671,7 +671,7 @@ static void __put_nommu_region(struct vm_region *region)
fb47a38f
JR
484 up_write(&nommu_region_sem);
485
486 if (region->vm_file)
487- fput(region->vm_file);
488+ vmr_fput(region);
489
490 /* IO memory and memory shared directly out of the pagecache
491 * from ramfs/tmpfs mustn't be released here */
c2c0f25c 492@@ -829,7 +829,7 @@ static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
fb47a38f
JR
493 if (vma->vm_ops && vma->vm_ops->close)
494 vma->vm_ops->close(vma);
495 if (vma->vm_file)
496- fput(vma->vm_file);
497+ vma_fput(vma);
498 put_nommu_region(vma->vm_region);
499 kmem_cache_free(vm_area_cachep, vma);
500 }
c2c0f25c 501@@ -1354,7 +1354,7 @@ unsigned long do_mmap_pgoff(struct file *file,
fb47a38f
JR
502 goto error_just_free;
503 }
504 }
505- fput(region->vm_file);
506+ vmr_fput(region);
507 kmem_cache_free(vm_region_jar, region);
508 region = pregion;
509 result = start;
c2c0f25c 510@@ -1429,10 +1429,10 @@ error_just_free:
fb47a38f
JR
511 up_write(&nommu_region_sem);
512 error:
513 if (region->vm_file)
514- fput(region->vm_file);
515+ vmr_fput(region);
516 kmem_cache_free(vm_region_jar, region);
517 if (vma->vm_file)
518- fput(vma->vm_file);
519+ vma_fput(vma);
520 kmem_cache_free(vm_area_cachep, vma);
fb47a38f 521 return ret;
c2c0f25c 522
076b876e
AM
523diff --git a/mm/prfile.c b/mm/prfile.c
524new file mode 100644
c2c0f25c 525index 0000000..b323b8a
076b876e
AM
526--- /dev/null
527+++ b/mm/prfile.c
528@@ -0,0 +1,86 @@
529+/*
530+ * Mainly for aufs which mmap(2) diffrent file and wants to print different path
531+ * in /proc/PID/maps.
532+ * Call these functions via macros defined in linux/mm.h.
533+ *
534+ * See Documentation/filesystems/aufs/design/06mmap.txt
535+ *
536+ * Copyright (c) 2014 Junjro R. Okajima
537+ * Copyright (c) 2014 Ian Campbell
538+ */
539+
540+#include <linux/mm.h>
541+#include <linux/file.h>
542+#include <linux/fs.h>
543+
544+/* #define PRFILE_TRACE */
545+static inline void prfile_trace(struct file *f, struct file *pr,
546+ const char func[], int line, const char func2[])
547+{
548+#ifdef PRFILE_TRACE
549+ if (pr)
c2c0f25c 550+ pr_info("%s:%d: %s, %s\n", func, line, func2,
7e9cd9fe 551+ f ? (char *)f->f_path.dentry->d_name.name : "(null)");
076b876e
AM
552+#endif
553+}
554+
076b876e
AM
555+void vma_do_file_update_time(struct vm_area_struct *vma, const char func[],
556+ int line)
557+{
558+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
559+
560+ prfile_trace(f, pr, func, line, __func__);
561+ file_update_time(f);
562+ if (f && pr)
563+ file_update_time(pr);
564+}
565+
566+struct file *vma_do_pr_or_file(struct vm_area_struct *vma, const char func[],
567+ int line)
568+{
569+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
570+
571+ prfile_trace(f, pr, func, line, __func__);
572+ return (f && pr) ? pr : f;
573+}
574+
575+void vma_do_get_file(struct vm_area_struct *vma, const char func[], int line)
576+{
577+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
578+
579+ prfile_trace(f, pr, func, line, __func__);
580+ get_file(f);
581+ if (f && pr)
582+ get_file(pr);
583+}
584+
585+void vma_do_fput(struct vm_area_struct *vma, const char func[], int line)
586+{
587+ struct file *f = vma->vm_file, *pr = vma->vm_prfile;
588+
589+ prfile_trace(f, pr, func, line, __func__);
590+ fput(f);
591+ if (f && pr)
592+ fput(pr);
593+}
b912730e
AM
594+
595+#ifndef CONFIG_MMU
076b876e
AM
596+struct file *vmr_do_pr_or_file(struct vm_region *region, const char func[],
597+ int line)
598+{
599+ struct file *f = region->vm_file, *pr = region->vm_prfile;
600+
601+ prfile_trace(f, pr, func, line, __func__);
602+ return (f && pr) ? pr : f;
603+}
604+
605+void vmr_do_fput(struct vm_region *region, const char func[], int line)
606+{
607+ struct file *f = region->vm_file, *pr = region->vm_prfile;
608+
609+ prfile_trace(f, pr, func, line, __func__);
610+ fput(f);
611+ if (f && pr)
612+ fput(pr);
613+}
b912730e 614+#endif /* !CONFIG_MMU */
5527c038 615aufs4.x-rcN standalone patch
7f207e10 616
c1595e42 617diff --git a/fs/dcache.c b/fs/dcache.c
c2c0f25c 618index 0b40298..4844e61 100644
c1595e42
JR
619--- a/fs/dcache.c
620+++ b/fs/dcache.c
5527c038 621@@ -1269,6 +1269,7 @@ rename_retry:
c1595e42
JR
622 seq = 1;
623 goto again;
624 }
625+EXPORT_SYMBOL(d_walk);
626
627 /*
628 * Search for at least 1 mount point in the dentry's subdirs.
b912730e 629diff --git a/fs/file_table.c b/fs/file_table.c
c2c0f25c 630index 7f9d407..8c9ec1d 100644
b912730e
AM
631--- a/fs/file_table.c
632+++ b/fs/file_table.c
c2c0f25c 633@@ -146,6 +146,7 @@ over:
b912730e
AM
634 }
635 return ERR_PTR(-ENFILE);
636 }
637+EXPORT_SYMBOL(get_empty_filp);
638
639 /**
640 * alloc_file - allocate and initialize a 'struct file'
c2c0f25c 641@@ -307,6 +308,7 @@ void put_filp(struct file *file)
b912730e
AM
642 file_free(file);
643 }
644 }
645+EXPORT_SYMBOL(put_filp);
646
647 void __init files_init(unsigned long mempages)
648 {
1e00d052 649diff --git a/fs/inode.c b/fs/inode.c
c2c0f25c 650index d30640f..1ecc715 100644
1e00d052
AM
651--- a/fs/inode.c
652+++ b/fs/inode.c
7e9cd9fe 653@@ -58,6 +58,7 @@ static struct hlist_head *inode_hashtable __read_mostly;
4b3da204 654 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
2cbb1c4b
JR
655
656 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
2cbb1c4b 657+EXPORT_SYMBOL(inode_sb_list_lock);
7f207e10
AM
658
659 /*
4b3da204 660 * Empty aops. Can be used for the cases where the user does not
7f207e10 661diff --git a/fs/namespace.c b/fs/namespace.c
c2c0f25c 662index c7cb8a5..08723de 100644
7f207e10
AM
663--- a/fs/namespace.c
664+++ b/fs/namespace.c
7e9cd9fe 665@@ -463,6 +463,7 @@ void __mnt_drop_write(struct vfsmount *mnt)
c06a8ce3
AM
666 mnt_dec_writers(real_mount(mnt));
667 preempt_enable();
668 }
669+EXPORT_SYMBOL_GPL(__mnt_drop_write);
670
671 /**
672 * mnt_drop_write - give up write access to a mount
c2c0f25c 673@@ -1779,6 +1780,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
7f207e10
AM
674 }
675 return 0;
676 }
677+EXPORT_SYMBOL(iterate_mounts);
678
7eafdf33 679 static void cleanup_group_ids(struct mount *mnt, struct mount *end)
7f207e10
AM
680 {
681diff --git a/fs/notify/group.c b/fs/notify/group.c
c1595e42 682index d16b62c..06ca6bc 100644
7f207e10
AM
683--- a/fs/notify/group.c
684+++ b/fs/notify/group.c
685@@ -22,6 +22,7 @@
686 #include <linux/srcu.h>
687 #include <linux/rculist.h>
688 #include <linux/wait.h>
689+#include <linux/module.h>
690
691 #include <linux/fsnotify_backend.h>
692 #include "fsnotify.h"
fb47a38f 693@@ -72,6 +73,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
1716fcea
AM
694 {
695 atomic_inc(&group->refcnt);
696 }
697+EXPORT_SYMBOL(fsnotify_get_group);
698
699 /*
700 * Drop a reference to a group. Free it if it's through.
fb47a38f 701@@ -81,6 +83,7 @@ void fsnotify_put_group(struct fsnotify_group *group)
7f207e10 702 if (atomic_dec_and_test(&group->refcnt))
1716fcea 703 fsnotify_final_destroy_group(group);
7f207e10
AM
704 }
705+EXPORT_SYMBOL(fsnotify_put_group);
706
707 /*
708 * Create a new fsnotify_group and hold a reference for the group returned.
fb47a38f 709@@ -109,6 +112,7 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
7f207e10
AM
710
711 return group;
712 }
713+EXPORT_SYMBOL(fsnotify_alloc_group);
1716fcea
AM
714
715 int fsnotify_fasync(int fd, struct file *file, int on)
716 {
7f207e10 717diff --git a/fs/notify/mark.c b/fs/notify/mark.c
2000de60 718index 92e48c7..d2c4b68 100644
7f207e10
AM
719--- a/fs/notify/mark.c
720+++ b/fs/notify/mark.c
392086de 721@@ -109,6 +109,7 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
7f207e10 722 mark->free_mark(mark);
1716fcea 723 }
7f207e10
AM
724 }
725+EXPORT_SYMBOL(fsnotify_put_mark);
726
2000de60
JR
727 /* Calculate mask of events for a list of marks */
728 u32 fsnotify_recalc_mask(struct hlist_head *head)
729@@ -202,6 +203,7 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark,
1716fcea
AM
730 fsnotify_destroy_mark_locked(mark, group);
731 mutex_unlock(&group->mark_mutex);
7f207e10
AM
732 }
733+EXPORT_SYMBOL(fsnotify_destroy_mark);
734
2000de60
JR
735 /*
736 * Destroy all marks in the given list. The marks must be already detached from
737@@ -376,6 +378,7 @@ err:
7f207e10
AM
738
739 return ret;
740 }
741+EXPORT_SYMBOL(fsnotify_add_mark);
742
1716fcea
AM
743 int fsnotify_add_mark(struct fsnotify_mark *mark, struct fsnotify_group *group,
744 struct inode *inode, struct vfsmount *mnt, int allow_dups)
2000de60 745@@ -455,6 +458,7 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
7f207e10
AM
746 atomic_set(&mark->refcnt, 1);
747 mark->free_mark = free_mark;
748 }
749+EXPORT_SYMBOL(fsnotify_init_mark);
750
751 static int fsnotify_mark_destroy(void *ignored)
752 {
753diff --git a/fs/open.c b/fs/open.c
c2c0f25c 754index e33dab2..b84b828 100644
7f207e10
AM
755--- a/fs/open.c
756+++ b/fs/open.c
c2c0f25c 757@@ -64,6 +64,7 @@ int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
7f207e10
AM
758 mutex_unlock(&dentry->d_inode->i_mutex);
759 return ret;
760 }
761+EXPORT_SYMBOL(do_truncate);
762
1716fcea 763 long vfs_truncate(struct path *path, loff_t length)
7f207e10 764 {
c2c0f25c 765@@ -678,6 +679,7 @@ int open_check_o_direct(struct file *f)
b912730e
AM
766 }
767 return 0;
768 }
769+EXPORT_SYMBOL(open_check_o_direct);
770
771 static int do_dentry_open(struct file *f,
c2c0f25c 772 struct inode *inode,
5527c038
JR
773diff --git a/fs/read_write.c b/fs/read_write.c
774index fd0414e..8ace6ec 100644
775--- a/fs/read_write.c
776+++ b/fs/read_write.c
777@@ -504,6 +504,7 @@ vfs_readf_t vfs_readf(struct file *file)
778 return new_sync_read;
779 return ERR_PTR(-ENOSYS);
780 }
781+EXPORT_SYMBOL(vfs_readf);
782
783 vfs_writef_t vfs_writef(struct file *file)
784 {
785@@ -515,6 +516,7 @@ vfs_writef_t vfs_writef(struct file *file)
786 return new_sync_write;
787 return ERR_PTR(-ENOSYS);
788 }
789+EXPORT_SYMBOL(vfs_writef);
790
791 ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
792 {
7f207e10 793diff --git a/fs/splice.c b/fs/splice.c
c2c0f25c 794index 5f8385a..f76067e 100644
7f207e10
AM
795--- a/fs/splice.c
796+++ b/fs/splice.c
c2c0f25c 797@@ -1115,6 +1115,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
392086de
AM
798
799 return splice_write(pipe, out, ppos, len, flags);
7f207e10
AM
800 }
801+EXPORT_SYMBOL(do_splice_from);
802
803 /*
804 * Attempt to initiate a splice from a file to a pipe.
c2c0f25c 805@@ -1141,6 +1142,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
7f207e10
AM
806
807 return splice_read(in, ppos, pipe, len, flags);
808 }
809+EXPORT_SYMBOL(do_splice_to);
810
811 /**
812 * splice_direct_to_actor - splices data directly between two non-pipes
c1595e42 813diff --git a/fs/xattr.c b/fs/xattr.c
c2c0f25c 814index 072fee1..a7677af 100644
c1595e42
JR
815--- a/fs/xattr.c
816+++ b/fs/xattr.c
817@@ -207,6 +207,7 @@ vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
818 *xattr_value = value;
819 return error;
820 }
821+EXPORT_SYMBOL(vfs_getxattr_alloc);
822
823 /* Compare an extended attribute value with the given value */
824 int vfs_xattr_cmp(struct dentry *dentry, const char *xattr_name,
7f207e10 825diff --git a/security/commoncap.c b/security/commoncap.c
c2c0f25c 826index d103f5a4..393654e 100644
7f207e10
AM
827--- a/security/commoncap.c
828+++ b/security/commoncap.c
c2c0f25c 829@@ -970,12 +970,14 @@ int cap_mmap_addr(unsigned long addr)
94337f0d 830 }
7f207e10
AM
831 return ret;
832 }
0c3ec466
AM
833+EXPORT_SYMBOL(cap_mmap_addr);
834
835 int cap_mmap_file(struct file *file, unsigned long reqprot,
836 unsigned long prot, unsigned long flags)
837 {
838 return 0;
839 }
840+EXPORT_SYMBOL(cap_mmap_file);
c2c0f25c
AM
841
842 #ifdef CONFIG_SECURITY
843
7f207e10 844diff --git a/security/device_cgroup.c b/security/device_cgroup.c
c1595e42 845index 188c1d2..426d9af 100644
7f207e10
AM
846--- a/security/device_cgroup.c
847+++ b/security/device_cgroup.c
f6c5ef8b
AM
848@@ -7,6 +7,7 @@
849 #include <linux/device_cgroup.h>
850 #include <linux/cgroup.h>
851 #include <linux/ctype.h>
852+#include <linux/export.h>
853 #include <linux/list.h>
854 #include <linux/uaccess.h>
855 #include <linux/seq_file.h>
076b876e 856@@ -849,6 +850,7 @@ int __devcgroup_inode_permission(struct inode *inode, int mask)
537831f9
AM
857 return __devcgroup_check_permission(type, imajor(inode), iminor(inode),
858 access);
7f207e10 859 }
2cbb1c4b 860+EXPORT_SYMBOL(__devcgroup_inode_permission);
7f207e10
AM
861
862 int devcgroup_inode_mknod(int mode, dev_t dev)
863 {
864diff --git a/security/security.c b/security/security.c
c2c0f25c 865index 595fffa..346bad6 100644
7f207e10
AM
866--- a/security/security.c
867+++ b/security/security.c
c2c0f25c 868@@ -438,6 +438,7 @@ int security_path_rmdir(struct path *dir, struct dentry *dentry)
7f207e10 869 return 0;
c2c0f25c 870 return call_int_hook(path_rmdir, 0, dir, dentry);
7f207e10
AM
871 }
872+EXPORT_SYMBOL(security_path_rmdir);
873
874 int security_path_unlink(struct path *dir, struct dentry *dentry)
875 {
c2c0f25c 876@@ -454,6 +455,7 @@ int security_path_symlink(struct path *dir, struct dentry *dentry,
7f207e10 877 return 0;
c2c0f25c 878 return call_int_hook(path_symlink, 0, dir, dentry, old_name);
7f207e10
AM
879 }
880+EXPORT_SYMBOL(security_path_symlink);
881
882 int security_path_link(struct dentry *old_dentry, struct path *new_dir,
883 struct dentry *new_dentry)
c2c0f25c 884@@ -462,6 +464,7 @@ int security_path_link(struct dentry *old_dentry, struct path *new_dir,
7f207e10 885 return 0;
c2c0f25c 886 return call_int_hook(path_link, 0, old_dentry, new_dir, new_dentry);
7f207e10
AM
887 }
888+EXPORT_SYMBOL(security_path_link);
889
890 int security_path_rename(struct path *old_dir, struct dentry *old_dentry,
38d290e6 891 struct path *new_dir, struct dentry *new_dentry,
c2c0f25c 892@@ -489,6 +492,7 @@ int security_path_truncate(struct path *path)
7f207e10 893 return 0;
c2c0f25c 894 return call_int_hook(path_truncate, 0, path);
7f207e10
AM
895 }
896+EXPORT_SYMBOL(security_path_truncate);
897
7eafdf33
AM
898 int security_path_chmod(struct path *path, umode_t mode)
899 {
c2c0f25c 900@@ -496,6 +500,7 @@ int security_path_chmod(struct path *path, umode_t mode)
7f207e10 901 return 0;
c2c0f25c 902 return call_int_hook(path_chmod, 0, path, mode);
7f207e10
AM
903 }
904+EXPORT_SYMBOL(security_path_chmod);
905
537831f9 906 int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
7f207e10 907 {
c2c0f25c 908@@ -503,6 +508,7 @@ int security_path_chown(struct path *path, kuid_t uid, kgid_t gid)
7f207e10 909 return 0;
c2c0f25c 910 return call_int_hook(path_chown, 0, path, uid, gid);
7f207e10
AM
911 }
912+EXPORT_SYMBOL(security_path_chown);
913
914 int security_path_chroot(struct path *path)
915 {
c2c0f25c 916@@ -588,6 +594,7 @@ int security_inode_readlink(struct dentry *dentry)
7f207e10 917 return 0;
c2c0f25c 918 return call_int_hook(inode_readlink, 0, dentry);
7f207e10
AM
919 }
920+EXPORT_SYMBOL(security_inode_readlink);
921
c2c0f25c
AM
922 int security_inode_follow_link(struct dentry *dentry, struct inode *inode,
923 bool rcu)
924@@ -603,6 +610,7 @@ int security_inode_permission(struct inode *inode, int mask)
7f207e10 925 return 0;
c2c0f25c 926 return call_int_hook(inode_permission, 0, inode, mask);
7f207e10
AM
927 }
928+EXPORT_SYMBOL(security_inode_permission);
929
1e00d052 930 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
7f207e10 931 {
c2c0f25c 932@@ -741,6 +749,7 @@ int security_file_permission(struct file *file, int mask)
7f207e10
AM
933
934 return fsnotify_perm(file, mask);
935 }
936+EXPORT_SYMBOL(security_file_permission);
937
938 int security_file_alloc(struct file *file)
939 {
c2c0f25c 940@@ -800,6 +809,7 @@ int security_mmap_file(struct file *file, unsigned long prot,
7f207e10
AM
941 return ret;
942 return ima_file_mmap(file, prot);
943 }
0c3ec466 944+EXPORT_SYMBOL(security_mmap_file);
7f207e10 945
0c3ec466
AM
946 int security_mmap_addr(unsigned long addr)
947 {
7f207e10
AM
948diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
949--- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 950+++ linux/Documentation/ABI/testing/debugfs-aufs 2015-09-24 10:47:58.244719488 +0200
86dc4139 951@@ -0,0 +1,50 @@
7f207e10
AM
952+What: /debug/aufs/si_<id>/
953+Date: March 2009
f6b6e03d 954+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
955+Description:
956+ Under /debug/aufs, a directory named si_<id> is created
957+ per aufs mount, where <id> is a unique id generated
958+ internally.
1facf9fc 959+
86dc4139
AM
960+What: /debug/aufs/si_<id>/plink
961+Date: Apr 2013
f6b6e03d 962+Contact: J. R. Okajima <hooanon05g@gmail.com>
86dc4139
AM
963+Description:
964+ It has three lines and shows the information about the
965+ pseudo-link. The first line is a single number
966+ representing a number of buckets. The second line is a
967+ number of pseudo-links per buckets (separated by a
968+ blank). The last line is a single number representing a
969+ total number of psedo-links.
970+ When the aufs mount option 'noplink' is specified, it
971+ will show "1\n0\n0\n".
972+
7f207e10
AM
973+What: /debug/aufs/si_<id>/xib
974+Date: March 2009
f6b6e03d 975+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
976+Description:
977+ It shows the consumed blocks by xib (External Inode Number
978+ Bitmap), its block size and file size.
979+ When the aufs mount option 'noxino' is specified, it
980+ will be empty. About XINO files, see the aufs manual.
981+
982+What: /debug/aufs/si_<id>/xino0, xino1 ... xinoN
983+Date: March 2009
f6b6e03d 984+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
985+Description:
986+ It shows the consumed blocks by xino (External Inode Number
987+ Translation Table), its link count, block size and file
988+ size.
989+ When the aufs mount option 'noxino' is specified, it
990+ will be empty. About XINO files, see the aufs manual.
991+
992+What: /debug/aufs/si_<id>/xigen
993+Date: March 2009
f6b6e03d 994+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
995+Description:
996+ It shows the consumed blocks by xigen (External Inode
997+ Generation Table), its block size and file size.
998+ If CONFIG_AUFS_EXPORT is disabled, this entry will not
999+ be created.
1000+ When the aufs mount option 'noxino' is specified, it
1001+ will be empty. About XINO files, see the aufs manual.
1002diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
1003--- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1004+++ linux/Documentation/ABI/testing/sysfs-aufs 2015-09-24 10:47:58.244719488 +0200
392086de 1005@@ -0,0 +1,31 @@
7f207e10
AM
1006+What: /sys/fs/aufs/si_<id>/
1007+Date: March 2009
f6b6e03d 1008+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
1009+Description:
1010+ Under /sys/fs/aufs, a directory named si_<id> is created
1011+ per aufs mount, where <id> is a unique id generated
1012+ internally.
1013+
1014+What: /sys/fs/aufs/si_<id>/br0, br1 ... brN
1015+Date: March 2009
f6b6e03d 1016+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
1017+Description:
1018+ It shows the abolute path of a member directory (which
1019+ is called branch) in aufs, and its permission.
1020+
392086de
AM
1021+What: /sys/fs/aufs/si_<id>/brid0, brid1 ... bridN
1022+Date: July 2013
f6b6e03d 1023+Contact: J. R. Okajima <hooanon05g@gmail.com>
392086de
AM
1024+Description:
1025+ It shows the id of a member directory (which is called
1026+ branch) in aufs.
1027+
7f207e10
AM
1028+What: /sys/fs/aufs/si_<id>/xi_path
1029+Date: March 2009
f6b6e03d 1030+Contact: J. R. Okajima <hooanon05g@gmail.com>
7f207e10
AM
1031+Description:
1032+ It shows the abolute path of XINO (External Inode Number
1033+ Bitmap, Translation Table and Generation Table) file
1034+ even if it is the default path.
1035+ When the aufs mount option 'noxino' is specified, it
1036+ will be empty. About XINO files, see the aufs manual.
53392da6
AM
1037diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
1038--- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1039+++ linux/Documentation/filesystems/aufs/design/01intro.txt 2015-09-24 10:47:58.244719488 +0200
7e9cd9fe 1040@@ -0,0 +1,170 @@
53392da6 1041+
2000de60 1042+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
1043+#
1044+# This program is free software; you can redistribute it and/or modify
1045+# it under the terms of the GNU General Public License as published by
1046+# the Free Software Foundation; either version 2 of the License, or
1047+# (at your option) any later version.
1048+#
1049+# This program is distributed in the hope that it will be useful,
1050+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1051+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1052+# GNU General Public License for more details.
1053+#
1054+# You should have received a copy of the GNU General Public License
523b37e3 1055+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1056+
1057+Introduction
1058+----------------------------------------
1059+
1060+aufs [ei ju: ef es] | [a u f s]
1061+1. abbrev. for "advanced multi-layered unification filesystem".
1062+2. abbrev. for "another unionfs".
1063+3. abbrev. for "auf das" in German which means "on the" in English.
1064+ Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
1065+ But "Filesystem aufs Filesystem" is hard to understand.
1066+
1067+AUFS is a filesystem with features:
1068+- multi layered stackable unification filesystem, the member directory
1069+ is called as a branch.
1070+- branch permission and attribute, 'readonly', 'real-readonly',
7e9cd9fe 1071+ 'readwrite', 'whiteout-able', 'link-able whiteout', etc. and their
53392da6
AM
1072+ combination.
1073+- internal "file copy-on-write".
1074+- logical deletion, whiteout.
1075+- dynamic branch manipulation, adding, deleting and changing permission.
1076+- allow bypassing aufs, user's direct branch access.
1077+- external inode number translation table and bitmap which maintains the
1078+ persistent aufs inode number.
1079+- seekable directory, including NFS readdir.
1080+- file mapping, mmap and sharing pages.
1081+- pseudo-link, hardlink over branches.
1082+- loopback mounted filesystem as a branch.
1083+- several policies to select one among multiple writable branches.
1084+- revert a single systemcall when an error occurs in aufs.
1085+- and more...
1086+
1087+
1088+Multi Layered Stackable Unification Filesystem
1089+----------------------------------------------------------------------
1090+Most people already knows what it is.
1091+It is a filesystem which unifies several directories and provides a
1092+merged single directory. When users access a file, the access will be
1093+passed/re-directed/converted (sorry, I am not sure which English word is
1094+correct) to the real file on the member filesystem. The member
1095+filesystem is called 'lower filesystem' or 'branch' and has a mode
1096+'readonly' and 'readwrite.' And the deletion for a file on the lower
1097+readonly branch is handled by creating 'whiteout' on the upper writable
1098+branch.
1099+
1100+On LKML, there have been discussions about UnionMount (Jan Blunck,
1101+Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
1102+different approaches to implement the merged-view.
1103+The former tries putting it into VFS, and the latter implements as a
1104+separate filesystem.
1105+(If I misunderstand about these implementations, please let me know and
1106+I shall correct it. Because it is a long time ago when I read their
1107+source files last time).
1108+
1109+UnionMount's approach will be able to small, but may be hard to share
1110+branches between several UnionMount since the whiteout in it is
1111+implemented in the inode on branch filesystem and always
1112+shared. According to Bharata's post, readdir does not seems to be
1113+finished yet.
1114+There are several missing features known in this implementations such as
1115+- for users, the inode number may change silently. eg. copy-up.
1116+- link(2) may break by copy-up.
1117+- read(2) may get an obsoleted filedata (fstat(2) too).
1118+- fcntl(F_SETLK) may be broken by copy-up.
1119+- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
1120+ open(O_RDWR).
1121+
7e9cd9fe
AM
1122+In linux-3.18, "overlay" filesystem (formerly known as "overlayfs") was
1123+merged into mainline. This is another implementation of UnionMount as a
1124+separated filesystem. All the limitations and known problems which
1125+UnionMount are equally inherited to "overlay" filesystem.
1126+
1127+Unionfs has a longer history. When I started implementing a stackable
1128+filesystem (Aug 2005), it already existed. It has virtual super_block,
1129+inode, dentry and file objects and they have an array pointing lower
1130+same kind objects. After contributing many patches for Unionfs, I
1131+re-started my project AUFS (Jun 2006).
53392da6
AM
1132+
1133+In AUFS, the structure of filesystem resembles to Unionfs, but I
1134+implemented my own ideas, approaches and enhancements and it became
1135+totally different one.
1136+
1137+Comparing DM snapshot and fs based implementation
1138+- the number of bytes to be copied between devices is much smaller.
1139+- the type of filesystem must be one and only.
1140+- the fs must be writable, no readonly fs, even for the lower original
1141+ device. so the compression fs will not be usable. but if we use
1142+ loopback mount, we may address this issue.
1143+ for instance,
1144+ mount /cdrom/squashfs.img /sq
1145+ losetup /sq/ext2.img
1146+ losetup /somewhere/cow
1147+ dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
1148+- it will be difficult (or needs more operations) to extract the
1149+ difference between the original device and COW.
1150+- DM snapshot-merge may help a lot when users try merging. in the
1151+ fs-layer union, users will use rsync(1).
1152+
7e9cd9fe
AM
1153+You may want to read my old paper "Filesystems in LiveCD"
1154+(http://aufs.sourceforge.net/aufs2/report/sq/sq.pdf).
53392da6 1155+
7e9cd9fe
AM
1156+
1157+Several characters/aspects/persona of aufs
53392da6
AM
1158+----------------------------------------------------------------------
1159+
7e9cd9fe 1160+Aufs has several characters, aspects or persona.
53392da6
AM
1161+1. a filesystem, callee of VFS helper
1162+2. sub-VFS, caller of VFS helper for branches
1163+3. a virtual filesystem which maintains persistent inode number
1164+4. reader/writer of files on branches such like an application
1165+
1166+1. Callee of VFS Helper
1167+As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
1168+unlink(2) from an application reaches sys_unlink() kernel function and
1169+then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
1170+calls filesystem specific unlink operation. Actually aufs implements the
1171+unlink operation but it behaves like a redirector.
1172+
1173+2. Caller of VFS Helper for Branches
1174+aufs_unlink() passes the unlink request to the branch filesystem as if
1175+it were called from VFS. So the called unlink operation of the branch
1176+filesystem acts as usual. As a caller of VFS helper, aufs should handle
1177+every necessary pre/post operation for the branch filesystem.
1178+- acquire the lock for the parent dir on a branch
1179+- lookup in a branch
1180+- revalidate dentry on a branch
1181+- mnt_want_write() for a branch
1182+- vfs_unlink() for a branch
1183+- mnt_drop_write() for a branch
1184+- release the lock on a branch
1185+
1186+3. Persistent Inode Number
1187+One of the most important issue for a filesystem is to maintain inode
1188+numbers. This is particularly important to support exporting a
1189+filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
1190+backend block device for its own. But some storage is necessary to
7e9cd9fe
AM
1191+keep and maintain the inode numbers. It may be a large space and may not
1192+suit to keep in memory. Aufs rents some space from its first writable
1193+branch filesystem (by default) and creates file(s) on it. These files
1194+are created by aufs internally and removed soon (currently) keeping
1195+opened.
53392da6
AM
1196+Note: Because these files are removed, they are totally gone after
1197+ unmounting aufs. It means the inode numbers are not persistent
1198+ across unmount or reboot. I have a plan to make them really
1199+ persistent which will be important for aufs on NFS server.
1200+
1201+4. Read/Write Files Internally (copy-on-write)
1202+Because a branch can be readonly, when you write a file on it, aufs will
1203+"copy-up" it to the upper writable branch internally. And then write the
1204+originally requested thing to the file. Generally kernel doesn't
1205+open/read/write file actively. In aufs, even a single write may cause a
1206+internal "file copy". This behaviour is very similar to cp(1) command.
1207+
1208+Some people may think it is better to pass such work to user space
1209+helper, instead of doing in kernel space. Actually I am still thinking
1210+about it. But currently I have implemented it in kernel space.
1211diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
1212--- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1213+++ linux/Documentation/filesystems/aufs/design/02struct.txt 2015-09-24 10:47:58.244719488 +0200
7e9cd9fe 1214@@ -0,0 +1,258 @@
53392da6 1215+
2000de60 1216+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
1217+#
1218+# This program is free software; you can redistribute it and/or modify
1219+# it under the terms of the GNU General Public License as published by
1220+# the Free Software Foundation; either version 2 of the License, or
1221+# (at your option) any later version.
1222+#
1223+# This program is distributed in the hope that it will be useful,
1224+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1225+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1226+# GNU General Public License for more details.
1227+#
1228+# You should have received a copy of the GNU General Public License
523b37e3 1229+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1230+
1231+Basic Aufs Internal Structure
1232+
1233+Superblock/Inode/Dentry/File Objects
1234+----------------------------------------------------------------------
1235+As like an ordinary filesystem, aufs has its own
1236+superblock/inode/dentry/file objects. All these objects have a
1237+dynamically allocated array and store the same kind of pointers to the
1238+lower filesystem, branch.
1239+For example, when you build a union with one readwrite branch and one
1240+readonly, mounted /au, /rw and /ro respectively.
1241+- /au = /rw + /ro
1242+- /ro/fileA exists but /rw/fileA
1243+
1244+Aufs lookup operation finds /ro/fileA and gets dentry for that. These
1245+pointers are stored in a aufs dentry. The array in aufs dentry will be,
7e9cd9fe 1246+- [0] = NULL (because /rw/fileA doesn't exist)
53392da6
AM
1247+- [1] = /ro/fileA
1248+
1249+This style of an array is essentially same to the aufs
1250+superblock/inode/dentry/file objects.
1251+
1252+Because aufs supports manipulating branches, ie. add/delete/change
7e9cd9fe
AM
1253+branches dynamically, these objects has its own generation. When
1254+branches are changed, the generation in aufs superblock is
1255+incremented. And a generation in other object are compared when it is
1256+accessed. When a generation in other objects are obsoleted, aufs
1257+refreshes the internal array.
53392da6
AM
1258+
1259+
1260+Superblock
1261+----------------------------------------------------------------------
1262+Additionally aufs superblock has some data for policies to select one
1263+among multiple writable branches, XIB files, pseudo-links and kobject.
1264+See below in detail.
7e9cd9fe
AM
1265+About the policies which supports copy-down a directory, see
1266+wbr_policy.txt too.
53392da6
AM
1267+
1268+
1269+Branch and XINO(External Inode Number Translation Table)
1270+----------------------------------------------------------------------
1271+Every branch has its own xino (external inode number translation table)
1272+file. The xino file is created and unlinked by aufs internally. When two
1273+members of a union exist on the same filesystem, they share the single
1274+xino file.
1275+The struct of a xino file is simple, just a sequence of aufs inode
1276+numbers which is indexed by the lower inode number.
1277+In the above sample, assume the inode number of /ro/fileA is i111 and
1278+aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
1279+4(8) bytes at 111 * 4(8) bytes offset in the xino file.
1280+
1281+When the inode numbers are not contiguous, the xino file will be sparse
1282+which has a hole in it and doesn't consume as much disk space as it
1283+might appear. If your branch filesystem consumes disk space for such
1284+holes, then you should specify 'xino=' option at mounting aufs.
1285+
7e9cd9fe
AM
1286+Aufs has a mount option to free the disk blocks for such holes in XINO
1287+files on tmpfs or ramdisk. But it is not so effective actually. If you
1288+meet a problem of disk shortage due to XINO files, then you should try
1289+"tmpfs-ino.patch" (and "vfs-ino.patch" too) in aufs4-standalone.git.
1290+The patch localizes the assignment inumbers per tmpfs-mount and avoid
1291+the holes in XINO files.
1292+
53392da6 1293+Also a writable branch has three kinds of "whiteout bases". All these
7e9cd9fe 1294+are existed when the branch is joined to aufs, and their names are
53392da6
AM
1295+whiteout-ed doubly, so that users will never see their names in aufs
1296+hierarchy.
7e9cd9fe 1297+1. a regular file which will be hardlinked to all whiteouts.
53392da6 1298+2. a directory to store a pseudo-link.
7e9cd9fe 1299+3. a directory to store an "orphan"-ed file temporary.
53392da6
AM
1300+
1301+1. Whiteout Base
1302+ When you remove a file on a readonly branch, aufs handles it as a
1303+ logical deletion and creates a whiteout on the upper writable branch
1304+ as a hardlink of this file in order not to consume inode on the
1305+ writable branch.
1306+2. Pseudo-link Dir
1307+ See below, Pseudo-link.
1308+3. Step-Parent Dir
1309+ When "fileC" exists on the lower readonly branch only and it is
1310+ opened and removed with its parent dir, and then user writes
1311+ something into it, then aufs copies-up fileC to this
1312+ directory. Because there is no other dir to store fileC. After
1313+ creating a file under this dir, the file is unlinked.
1314+
1315+Because aufs supports manipulating branches, ie. add/delete/change
7e9cd9fe
AM
1316+dynamically, a branch has its own id. When the branch order changes,
1317+aufs finds the new index by searching the branch id.
53392da6
AM
1318+
1319+
1320+Pseudo-link
1321+----------------------------------------------------------------------
1322+Assume "fileA" exists on the lower readonly branch only and it is
1323+hardlinked to "fileB" on the branch. When you write something to fileA,
1324+aufs copies-up it to the upper writable branch. Additionally aufs
1325+creates a hardlink under the Pseudo-link Directory of the writable
1326+branch. The inode of a pseudo-link is kept in aufs super_block as a
1327+simple list. If fileB is read after unlinking fileA, aufs returns
1328+filedata from the pseudo-link instead of the lower readonly
1329+branch. Because the pseudo-link is based upon the inode, to keep the
7e9cd9fe 1330+inode number by xino (see above) is essentially necessary.
53392da6
AM
1331+
1332+All the hardlinks under the Pseudo-link Directory of the writable branch
1333+should be restored in a proper location later. Aufs provides a utility
1334+to do this. The userspace helpers executed at remounting and unmounting
1335+aufs by default.
1336+During this utility is running, it puts aufs into the pseudo-link
1337+maintenance mode. In this mode, only the process which began the
1338+maintenance mode (and its child processes) is allowed to operate in
1339+aufs. Some other processes which are not related to the pseudo-link will
1340+be allowed to run too, but the rest have to return an error or wait
1341+until the maintenance mode ends. If a process already acquires an inode
1342+mutex (in VFS), it has to return an error.
1343+
1344+
1345+XIB(external inode number bitmap)
1346+----------------------------------------------------------------------
1347+Addition to the xino file per a branch, aufs has an external inode number
7e9cd9fe
AM
1348+bitmap in a superblock object. It is also an internal file such like a
1349+xino file.
53392da6
AM
1350+It is a simple bitmap to mark whether the aufs inode number is in-use or
1351+not.
1352+To reduce the file I/O, aufs prepares a single memory page to cache xib.
1353+
7e9cd9fe 1354+As well as XINO files, aufs has a feature to truncate/refresh XIB to
53392da6
AM
1355+reduce the number of consumed disk blocks for these files.
1356+
1357+
1358+Virtual or Vertical Dir, and Readdir in Userspace
1359+----------------------------------------------------------------------
1360+In order to support multiple layers (branches), aufs readdir operation
1361+constructs a virtual dir block on memory. For readdir, aufs calls
1362+vfs_readdir() internally for each dir on branches, merges their entries
1363+with eliminating the whiteout-ed ones, and sets it to file (dir)
1364+object. So the file object has its entry list until it is closed. The
1365+entry list will be updated when the file position is zero and becomes
7e9cd9fe 1366+obsoleted. This decision is made in aufs automatically.
53392da6
AM
1367+
1368+The dynamically allocated memory block for the name of entries has a
1369+unit of 512 bytes (by default) and stores the names contiguously (no
1370+padding). Another block for each entry is handled by kmem_cache too.
1371+During building dir blocks, aufs creates hash list and judging whether
1372+the entry is whiteouted by its upper branch or already listed.
1373+The merged result is cached in the corresponding inode object and
1374+maintained by a customizable life-time option.
1375+
1376+Some people may call it can be a security hole or invite DoS attack
1377+since the opened and once readdir-ed dir (file object) holds its entry
1378+list and becomes a pressure for system memory. But I'd say it is similar
1379+to files under /proc or /sys. The virtual files in them also holds a
1380+memory page (generally) while they are opened. When an idea to reduce
1381+memory for them is introduced, it will be applied to aufs too.
1382+For those who really hate this situation, I've developed readdir(3)
1383+library which operates this merging in userspace. You just need to set
1384+LD_PRELOAD environment variable, and aufs will not consume no memory in
1385+kernel space for readdir(3).
1386+
1387+
1388+Workqueue
1389+----------------------------------------------------------------------
1390+Aufs sometimes requires privilege access to a branch. For instance,
1391+in copy-up/down operation. When a user process is going to make changes
1392+to a file which exists in the lower readonly branch only, and the mode
1393+of one of ancestor directories may not be writable by a user
1394+process. Here aufs copy-up the file with its ancestors and they may
1395+require privilege to set its owner/group/mode/etc.
1396+This is a typical case of a application character of aufs (see
1397+Introduction).
1398+
1399+Aufs uses workqueue synchronously for this case. It creates its own
1400+workqueue. The workqueue is a kernel thread and has privilege. Aufs
1401+passes the request to call mkdir or write (for example), and wait for
1402+its completion. This approach solves a problem of a signal handler
1403+simply.
1404+If aufs didn't adopt the workqueue and changed the privilege of the
7e9cd9fe
AM
1405+process, then the process may receive the unexpected SIGXFSZ or other
1406+signals.
53392da6
AM
1407+
1408+Also aufs uses the system global workqueue ("events" kernel thread) too
1409+for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
1410+whiteout base and etc. This is unrelated to a privilege.
1411+Most of aufs operation tries acquiring a rw_semaphore for aufs
1412+superblock at the beginning, at the same time waits for the completion
1413+of all queued asynchronous tasks.
1414+
1415+
1416+Whiteout
1417+----------------------------------------------------------------------
1418+The whiteout in aufs is very similar to Unionfs's. That is represented
1419+by its filename. UnionMount takes an approach of a file mode, but I am
1420+afraid several utilities (find(1) or something) will have to support it.
1421+
1422+Basically the whiteout represents "logical deletion" which stops aufs to
1423+lookup further, but also it represents "dir is opaque" which also stop
7e9cd9fe 1424+further lookup.
53392da6
AM
1425+
1426+In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
1427+In order to make several functions in a single systemcall to be
1428+revertible, aufs adopts an approach to rename a directory to a temporary
1429+unique whiteouted name.
1430+For example, in rename(2) dir where the target dir already existed, aufs
1431+renames the target dir to a temporary unique whiteouted name before the
7e9cd9fe 1432+actual rename on a branch, and then handles other actions (make it opaque,
53392da6
AM
1433+update the attributes, etc). If an error happens in these actions, aufs
1434+simply renames the whiteouted name back and returns an error. If all are
1435+succeeded, aufs registers a function to remove the whiteouted unique
1436+temporary name completely and asynchronously to the system global
1437+workqueue.
1438+
1439+
1440+Copy-up
1441+----------------------------------------------------------------------
1442+It is a well-known feature or concept.
1443+When user modifies a file on a readonly branch, aufs operate "copy-up"
1444+internally and makes change to the new file on the upper writable branch.
1445+When the trigger systemcall does not update the timestamps of the parent
1446+dir, aufs reverts it after copy-up.
c2b27bf2
AM
1447+
1448+
1449+Move-down (aufs3.9 and later)
1450+----------------------------------------------------------------------
1451+"Copy-up" is one of the essential feature in aufs. It copies a file from
1452+the lower readonly branch to the upper writable branch when a user
1453+changes something about the file.
1454+"Move-down" is an opposite action of copy-up. Basically this action is
1455+ran manually instead of automatically and internally.
076b876e
AM
1456+For desgin and implementation, aufs has to consider these issues.
1457+- whiteout for the file may exist on the lower branch.
1458+- ancestor directories may not exist on the lower branch.
1459+- diropq for the ancestor directories may exist on the upper branch.
1460+- free space on the lower branch will reduce.
1461+- another access to the file may happen during moving-down, including
7e9cd9fe 1462+ UDBA (see "Revalidate Dentry and UDBA").
076b876e
AM
1463+- the file should not be hard-linked nor pseudo-linked. they should be
1464+ handled by auplink utility later.
c2b27bf2
AM
1465+
1466+Sometimes users want to move-down a file from the upper writable branch
1467+to the lower readonly or writable branch. For instance,
1468+- the free space of the upper writable branch is going to run out.
1469+- create a new intermediate branch between the upper and lower branch.
1470+- etc.
1471+
1472+For this purpose, use "aumvdown" command in aufs-util.git.
b912730e
AM
1473diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt linux/Documentation/filesystems/aufs/design/03atomic_open.txt
1474--- /usr/share/empty/Documentation/filesystems/aufs/design/03atomic_open.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1475+++ linux/Documentation/filesystems/aufs/design/03atomic_open.txt 2015-09-24 10:47:58.244719488 +0200
b912730e
AM
1476@@ -0,0 +1,85 @@
1477+
1478+# Copyright (C) 2015 Junjiro R. Okajima
1479+#
1480+# This program is free software; you can redistribute it and/or modify
1481+# it under the terms of the GNU General Public License as published by
1482+# the Free Software Foundation; either version 2 of the License, or
1483+# (at your option) any later version.
1484+#
1485+# This program is distributed in the hope that it will be useful,
1486+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1487+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1488+# GNU General Public License for more details.
1489+#
1490+# You should have received a copy of the GNU General Public License
1491+# along with this program. If not, see <http://www.gnu.org/licenses/>.
1492+
1493+Support for a branch who has its ->atomic_open()
1494+----------------------------------------------------------------------
1495+The filesystems who implement its ->atomic_open() are not majority. For
1496+example NFSv4 does, and aufs should call NFSv4 ->atomic_open,
1497+particularly for open(O_CREAT|O_EXCL, 0400) case. Other than
1498+->atomic_open(), NFSv4 returns an error for this open(2). While I am not
1499+sure whether all filesystems who have ->atomic_open() behave like this,
1500+but NFSv4 surely returns the error.
1501+
1502+In order to support ->atomic_open() for aufs, there are a few
1503+approaches.
1504+
1505+A. Introduce aufs_atomic_open()
1506+ - calls one of VFS:do_last(), lookup_open() or atomic_open() for
1507+ branch fs.
1508+B. Introduce aufs_atomic_open() calling create, open and chmod. this is
1509+ an aufs user Pip Cet's approach
1510+ - calls aufs_create(), VFS finish_open() and notify_change().
1511+ - pass fake-mode to finish_open(), and then correct the mode by
1512+ notify_change().
1513+C. Extend aufs_open() to call branch fs's ->atomic_open()
1514+ - no aufs_atomic_open().
1515+ - aufs_lookup() registers the TID to an aufs internal object.
1516+ - aufs_create() does nothing when the matching TID is registered, but
1517+ registers the mode.
1518+ - aufs_open() calls branch fs's ->atomic_open() when the matching
1519+ TID is registered.
1520+D. Extend aufs_open() to re-try branch fs's ->open() with superuser's
1521+ credential
1522+ - no aufs_atomic_open().
1523+ - aufs_create() registers the TID to an internal object. this info
1524+ represents "this process created this file just now."
1525+ - when aufs gets EACCES from branch fs's ->open(), then confirm the
1526+ registered TID and re-try open() with superuser's credential.
1527+
1528+Pros and cons for each approach.
1529+
1530+A.
1531+ - straightforward but highly depends upon VFS internal.
1532+ - the atomic behavaiour is kept.
1533+ - some of parameters such as nameidata are hard to reproduce for
1534+ branch fs.
1535+ - large overhead.
1536+B.
1537+ - easy to implement.
1538+ - the atomic behavaiour is lost.
1539+C.
1540+ - the atomic behavaiour is kept.
1541+ - dirty and tricky.
1542+ - VFS checks whether the file is created correctly after calling
1543+ ->create(), which means this approach doesn't work.
1544+D.
1545+ - easy to implement.
1546+ - the atomic behavaiour is lost.
1547+ - to open a file with superuser's credential and give it to a user
1548+ process is a bad idea, since the file object keeps the credential
1549+ in it. It may affect LSM or something. This approach doesn't work
1550+ either.
1551+
1552+The approach A is ideal, but it hard to implement. So here is a
1553+variation of A, which is to be implemented.
1554+
1555+A-1. Introduce aufs_atomic_open()
1556+ - calls branch fs ->atomic_open() if exists. otherwise calls
1557+ vfs_create() and finish_open().
1558+ - the demerit is that the several checks after branch fs
1559+ ->atomic_open() are lost. in the ordinary case, the checks are
1560+ done by VFS:do_last(), lookup_open() and atomic_open(). some can
1561+ be implemented in aufs, but not all I am afraid.
53392da6
AM
1562diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
1563--- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1564+++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2015-09-24 10:47:58.244719488 +0200
7e9cd9fe 1565@@ -0,0 +1,113 @@
53392da6 1566+
2000de60 1567+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
1568+#
1569+# This program is free software; you can redistribute it and/or modify
1570+# it under the terms of the GNU General Public License as published by
1571+# the Free Software Foundation; either version 2 of the License, or
1572+# (at your option) any later version.
1573+#
1574+# This program is distributed in the hope that it will be useful,
1575+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1576+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1577+# GNU General Public License for more details.
1578+#
1579+# You should have received a copy of the GNU General Public License
523b37e3 1580+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1581+
1582+Lookup in a Branch
1583+----------------------------------------------------------------------
1584+Since aufs has a character of sub-VFS (see Introduction), it operates
7e9cd9fe
AM
1585+lookup for branches as VFS does. It may be a heavy work. But almost all
1586+lookup operation in aufs is the simplest case, ie. lookup only an entry
1587+directly connected to its parent. Digging down the directory hierarchy
1588+is unnecessary. VFS has a function lookup_one_len() for that use, and
1589+aufs calls it.
1590+
1591+When a branch is a remote filesystem, aufs basically relies upon its
53392da6
AM
1592+->d_revalidate(), also aufs forces the hardest revalidate tests for
1593+them.
1594+For d_revalidate, aufs implements three levels of revalidate tests. See
1595+"Revalidate Dentry and UDBA" in detail.
1596+
1597+
076b876e
AM
1598+Test Only the Highest One for the Directory Permission (dirperm1 option)
1599+----------------------------------------------------------------------
1600+Let's try case study.
1601+- aufs has two branches, upper readwrite and lower readonly.
1602+ /au = /rw + /ro
1603+- "dirA" exists under /ro, but /rw. and its mode is 0700.
1604+- user invoked "chmod a+rx /au/dirA"
1605+- the internal copy-up is activated and "/rw/dirA" is created and its
7e9cd9fe 1606+ permission bits are set to world readable.
076b876e
AM
1607+- then "/au/dirA" becomes world readable?
1608+
1609+In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1610+or it may be a natively readonly filesystem. If aufs respects the lower
1611+branch, it should not respond readdir request from other users. But user
1612+allowed it by chmod. Should really aufs rejects showing the entries
1613+under /ro/dirA?
1614+
7e9cd9fe
AM
1615+To be honest, I don't have a good solution for this case. So aufs
1616+implements 'dirperm1' and 'nodirperm1' mount options, and leave it to
1617+users.
076b876e
AM
1618+When dirperm1 is specified, aufs checks only the highest one for the
1619+directory permission, and shows the entries. Otherwise, as usual, checks
1620+every dir existing on all branches and rejects the request.
1621+
1622+As a side effect, dirperm1 option improves the performance of aufs
1623+because the number of permission check is reduced when the number of
1624+branch is many.
1625+
1626+
53392da6
AM
1627+Revalidate Dentry and UDBA (User's Direct Branch Access)
1628+----------------------------------------------------------------------
1629+Generally VFS helpers re-validate a dentry as a part of lookup.
1630+0. digging down the directory hierarchy.
1631+1. lock the parent dir by its i_mutex.
1632+2. lookup the final (child) entry.
1633+3. revalidate it.
1634+4. call the actual operation (create, unlink, etc.)
1635+5. unlock the parent dir
1636+
1637+If the filesystem implements its ->d_revalidate() (step 3), then it is
1638+called. Actually aufs implements it and checks the dentry on a branch is
1639+still valid.
1640+But it is not enough. Because aufs has to release the lock for the
1641+parent dir on a branch at the end of ->lookup() (step 2) and
1642+->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
1643+held by VFS.
1644+If the file on a branch is changed directly, eg. bypassing aufs, after
1645+aufs released the lock, then the subsequent operation may cause
1646+something unpleasant result.
1647+
1648+This situation is a result of VFS architecture, ->lookup() and
1649+->d_revalidate() is separated. But I never say it is wrong. It is a good
1650+design from VFS's point of view. It is just not suitable for sub-VFS
1651+character in aufs.
1652+
1653+Aufs supports such case by three level of revalidation which is
1654+selectable by user.
1655+1. Simple Revalidate
1656+ Addition to the native flow in VFS's, confirm the child-parent
1657+ relationship on the branch just after locking the parent dir on the
1658+ branch in the "actual operation" (step 4). When this validation
1659+ fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
1660+ checks the validation of the dentry on branches.
1661+2. Monitor Changes Internally by Inotify/Fsnotify
1662+ Addition to above, in the "actual operation" (step 4) aufs re-lookup
1663+ the dentry on the branch, and returns EBUSY if it finds different
1664+ dentry.
1665+ Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
1666+ during it is in cache. When the event is notified, aufs registers a
1667+ function to kernel 'events' thread by schedule_work(). And the
1668+ function sets some special status to the cached aufs dentry and inode
1669+ private data. If they are not cached, then aufs has nothing to
1670+ do. When the same file is accessed through aufs (step 0-3) later,
1671+ aufs will detect the status and refresh all necessary data.
1672+ In this mode, aufs has to ignore the event which is fired by aufs
1673+ itself.
1674+3. No Extra Validation
1675+ This is the simplest test and doesn't add any additional revalidation
7e9cd9fe 1676+ test, and skip the revalidation in step 4. It is useful and improves
53392da6
AM
1677+ aufs performance when system surely hide the aufs branches from user,
1678+ by over-mounting something (or another method).
1679diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
1680--- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1681+++ linux/Documentation/filesystems/aufs/design/04branch.txt 2015-09-24 10:47:58.244719488 +0200
7e9cd9fe 1682@@ -0,0 +1,74 @@
53392da6 1683+
2000de60 1684+# Copyright (C) 2005-2015 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+Branch Manipulation
1700+
1701+Since aufs supports dynamic branch manipulation, ie. add/remove a branch
1702+and changing its permission/attribute, there are a lot of works to do.
1703+
1704+
1705+Add a Branch
1706+----------------------------------------------------------------------
1707+o Confirm the adding dir exists outside of aufs, including loopback
7e9cd9fe 1708+ mount, and its various attributes.
53392da6
AM
1709+o Initialize the xino file and whiteout bases if necessary.
1710+ See struct.txt.
1711+
1712+o Check the owner/group/mode of the directory
1713+ When the owner/group/mode of the adding directory differs from the
1714+ existing branch, aufs issues a warning because it may impose a
1715+ security risk.
1716+ For example, when a upper writable branch has a world writable empty
1717+ top directory, a malicious user can create any files on the writable
1718+ branch directly, like copy-up and modify manually. If something like
1719+ /etc/{passwd,shadow} exists on the lower readonly branch but the upper
1720+ writable branch, and the writable branch is world-writable, then a
1721+ malicious guy may create /etc/passwd on the writable branch directly
1722+ and the infected file will be valid in aufs.
7e9cd9fe 1723+ I am afraid it can be a security issue, but aufs can do nothing except
53392da6
AM
1724+ producing a warning.
1725+
1726+
1727+Delete a Branch
1728+----------------------------------------------------------------------
1729+o Confirm the deleting branch is not busy
1730+ To be general, there is one merit to adopt "remount" interface to
1731+ manipulate branches. It is to discard caches. At deleting a branch,
1732+ aufs checks the still cached (and connected) dentries and inodes. If
1733+ there are any, then they are all in-use. An inode without its
1734+ corresponding dentry can be alive alone (for example, inotify/fsnotify case).
1735+
1736+ For the cached one, aufs checks whether the same named entry exists on
1737+ other branches.
1738+ If the cached one is a directory, because aufs provides a merged view
1739+ to users, as long as one dir is left on any branch aufs can show the
1740+ dir to users. In this case, the branch can be removed from aufs.
1741+ Otherwise aufs rejects deleting the branch.
1742+
1743+ If any file on the deleting branch is opened by aufs, then aufs
1744+ rejects deleting.
1745+
1746+
1747+Modify the Permission of a Branch
1748+----------------------------------------------------------------------
1749+o Re-initialize or remove the xino file and whiteout bases if necessary.
1750+ See struct.txt.
1751+
1752+o rw --> ro: Confirm the modifying branch is not busy
1753+ Aufs rejects the request if any of these conditions are true.
1754+ - a file on the branch is mmap-ed.
1755+ - a regular file on the branch is opened for write and there is no
1756+ same named entry on the upper branch.
1757diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1758--- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1759+++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2015-09-24 10:47:58.244719488 +0200
523b37e3 1760@@ -0,0 +1,64 @@
53392da6 1761+
2000de60 1762+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
1763+#
1764+# This program is free software; you can redistribute it and/or modify
1765+# it under the terms of the GNU General Public License as published by
1766+# the Free Software Foundation; either version 2 of the License, or
1767+# (at your option) any later version.
1768+#
1769+# This program is distributed in the hope that it will be useful,
1770+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1771+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1772+# GNU General Public License for more details.
1773+#
1774+# You should have received a copy of the GNU General Public License
523b37e3 1775+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1776+
1777+Policies to Select One among Multiple Writable Branches
1778+----------------------------------------------------------------------
1779+When the number of writable branch is more than one, aufs has to decide
1780+the target branch for file creation or copy-up. By default, the highest
1781+writable branch which has the parent (or ancestor) dir of the target
1782+file is chosen (top-down-parent policy).
1783+By user's request, aufs implements some other policies to select the
7e9cd9fe
AM
1784+writable branch, for file creation several policies, round-robin,
1785+most-free-space, and other policies. For copy-up, top-down-parent,
1786+bottom-up-parent, bottom-up and others.
53392da6
AM
1787+
1788+As expected, the round-robin policy selects the branch in circular. When
1789+you have two writable branches and creates 10 new files, 5 files will be
1790+created for each branch. mkdir(2) systemcall is an exception. When you
1791+create 10 new directories, all will be created on the same branch.
1792+And the most-free-space policy selects the one which has most free
1793+space among the writable branches. The amount of free space will be
1794+checked by aufs internally, and users can specify its time interval.
1795+
1796+The policies for copy-up is more simple,
1797+top-down-parent is equivalent to the same named on in create policy,
1798+bottom-up-parent selects the writable branch where the parent dir
1799+exists and the nearest upper one from the copyup-source,
1800+bottom-up selects the nearest upper writable branch from the
1801+copyup-source, regardless the existence of the parent dir.
1802+
1803+There are some rules or exceptions to apply these policies.
1804+- If there is a readonly branch above the policy-selected branch and
1805+ the parent dir is marked as opaque (a variation of whiteout), or the
1806+ target (creating) file is whiteout-ed on the upper readonly branch,
1807+ then the result of the policy is ignored and the target file will be
1808+ created on the nearest upper writable branch than the readonly branch.
1809+- If there is a writable branch above the policy-selected branch and
1810+ the parent dir is marked as opaque or the target file is whiteouted
1811+ on the branch, then the result of the policy is ignored and the target
1812+ file will be created on the highest one among the upper writable
1813+ branches who has diropq or whiteout. In case of whiteout, aufs removes
1814+ it as usual.
1815+- link(2) and rename(2) systemcalls are exceptions in every policy.
1816+ They try selecting the branch where the source exists as possible
1817+ since copyup a large file will take long time. If it can't be,
1818+ ie. the branch where the source exists is readonly, then they will
1819+ follow the copyup policy.
1820+- There is an exception for rename(2) when the target exists.
1821+ If the rename target exists, aufs compares the index of the branches
1822+ where the source and the target exists and selects the higher
1823+ one. If the selected branch is readonly, then aufs follows the
1824+ copyup policy.
076b876e
AM
1825diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt linux/Documentation/filesystems/aufs/design/06fhsm.txt
1826--- /usr/share/empty/Documentation/filesystems/aufs/design/06fhsm.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1827+++ linux/Documentation/filesystems/aufs/design/06fhsm.txt 2015-09-24 10:47:58.244719488 +0200
076b876e
AM
1828@@ -0,0 +1,120 @@
1829+
2000de60 1830+# Copyright (C) 2011-2015 Junjiro R. Okajima
076b876e
AM
1831+#
1832+# This program is free software; you can redistribute it and/or modify
1833+# it under the terms of the GNU General Public License as published by
1834+# the Free Software Foundation; either version 2 of the License, or
1835+# (at your option) any later version.
1836+#
1837+# This program is distributed in the hope that it will be useful,
1838+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1839+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1840+# GNU General Public License for more details.
1841+#
1842+# You should have received a copy of the GNU General Public License
1843+# along with this program; if not, write to the Free Software
1844+# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1845+
1846+
1847+File-based Hierarchical Storage Management (FHSM)
1848+----------------------------------------------------------------------
1849+Hierarchical Storage Management (or HSM) is a well-known feature in the
1850+storage world. Aufs provides this feature as file-based with multiple
7e9cd9fe 1851+writable branches, based upon the principle of "Colder, the Lower".
076b876e 1852+Here the word "colder" means that the less used files, and "lower" means
7e9cd9fe 1853+that the position in the order of the stacked branches vertically.
076b876e
AM
1854+These multiple writable branches are prioritized, ie. the topmost one
1855+should be the fastest drive and be used heavily.
1856+
1857+o Characters in aufs FHSM story
1858+- aufs itself and a new branch attribute.
1859+- a new ioctl interface to move-down and to establish a connection with
1860+ the daemon ("move-down" is a converse of "copy-up").
1861+- userspace tool and daemon.
1862+
1863+The userspace daemon establishes a connection with aufs and waits for
1864+the notification. The notified information is very similar to struct
1865+statfs containing the number of consumed blocks and inodes.
1866+When the consumed blocks/inodes of a branch exceeds the user-specified
1867+upper watermark, the daemon activates its move-down process until the
1868+consumed blocks/inodes reaches the user-specified lower watermark.
1869+
1870+The actual move-down is done by aufs based upon the request from
1871+user-space since we need to maintain the inode number and the internal
1872+pointer arrays in aufs.
1873+
1874+Currently aufs FHSM handles the regular files only. Additionally they
1875+must not be hard-linked nor pseudo-linked.
1876+
1877+
1878+o Cowork of aufs and the user-space daemon
1879+ During the userspace daemon established the connection, aufs sends a
1880+ small notification to it whenever aufs writes something into the
1881+ writable branch. But it may cost high since aufs issues statfs(2)
1882+ internally. So user can specify a new option to cache the
1883+ info. Actually the notification is controlled by these factors.
1884+ + the specified cache time.
1885+ + classified as "force" by aufs internally.
1886+ Until the specified time expires, aufs doesn't send the info
1887+ except the forced cases. When aufs decide forcing, the info is always
1888+ notified to userspace.
1889+ For example, the number of free inodes is generally large enough and
1890+ the shortage of it happens rarely. So aufs doesn't force the
1891+ notification when creating a new file, directory and others. This is
1892+ the typical case which aufs doesn't force.
1893+ When aufs writes the actual filedata and the files consumes any of new
1894+ blocks, the aufs forces notifying.
1895+
1896+
1897+o Interfaces in aufs
1898+- New branch attribute.
1899+ + fhsm
1900+ Specifies that the branch is managed by FHSM feature. In other word,
1901+ participant in the FHSM.
1902+ When nofhsm is set to the branch, it will not be the source/target
1903+ branch of the move-down operation. This attribute is set
1904+ independently from coo and moo attributes, and if you want full
1905+ FHSM, you should specify them as well.
1906+- New mount option.
1907+ + fhsm_sec
1908+ Specifies a second to suppress many less important info to be
1909+ notified.
1910+- New ioctl.
1911+ + AUFS_CTL_FHSM_FD
1912+ create a new file descriptor which userspace can read the notification
1913+ (a subset of struct statfs) from aufs.
1914+- Module parameter 'brs'
1915+ It has to be set to 1. Otherwise the new mount option 'fhsm' will not
1916+ be set.
1917+- mount helpers /sbin/mount.aufs and /sbin/umount.aufs
1918+ When there are two or more branches with fhsm attributes,
1919+ /sbin/mount.aufs invokes the user-space daemon and /sbin/umount.aufs
1920+ terminates it. As a result of remounting and branch-manipulation, the
1921+ number of branches with fhsm attribute can be one. In this case,
1922+ /sbin/mount.aufs will terminate the user-space daemon.
1923+
1924+
1925+Finally the operation is done as these steps in kernel-space.
1926+- make sure that,
1927+ + no one else is using the file.
1928+ + the file is not hard-linked.
1929+ + the file is not pseudo-linked.
1930+ + the file is a regular file.
1931+ + the parent dir is not opaqued.
1932+- find the target writable branch.
1933+- make sure the file is not whiteout-ed by the upper (than the target)
1934+ branch.
1935+- make the parent dir on the target branch.
1936+- mutex lock the inode on the branch.
1937+- unlink the whiteout on the target branch (if exists).
1938+- lookup and create the whiteout-ed temporary name on the target branch.
1939+- copy the file as the whiteout-ed temporary name on the target branch.
1940+- rename the whiteout-ed temporary name to the original name.
1941+- unlink the file on the source branch.
1942+- maintain the internal pointer array and the external inode number
1943+ table (XINO).
1944+- maintain the timestamps and other attributes of the parent dir and the
1945+ file.
1946+
1947+And of course, in every step, an error may happen. So the operation
1948+should restore the original file state after an error happens.
53392da6
AM
1949diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
1950--- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 1951+++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2015-09-24 10:47:58.244719488 +0200
b912730e 1952@@ -0,0 +1,72 @@
53392da6 1953+
2000de60 1954+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
1955+#
1956+# This program is free software; you can redistribute it and/or modify
1957+# it under the terms of the GNU General Public License as published by
1958+# the Free Software Foundation; either version 2 of the License, or
1959+# (at your option) any later version.
1960+#
1961+# This program is distributed in the hope that it will be useful,
1962+# but WITHOUT ANY WARRANTY; without even the implied warranty of
1963+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1964+# GNU General Public License for more details.
1965+#
1966+# You should have received a copy of the GNU General Public License
523b37e3 1967+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
1968+
1969+mmap(2) -- File Memory Mapping
1970+----------------------------------------------------------------------
1971+In aufs, the file-mapped pages are handled by a branch fs directly, no
1972+interaction with aufs. It means aufs_mmap() calls the branch fs's
1973+->mmap().
1974+This approach is simple and good, but there is one problem.
7e9cd9fe 1975+Under /proc, several entries show the mmapped files by its path (with
53392da6
AM
1976+device and inode number), and the printed path will be the path on the
1977+branch fs's instead of virtual aufs's.
1978+This is not a problem in most cases, but some utilities lsof(1) (and its
1979+user) may expect the path on aufs.
1980+
1981+To address this issue, aufs adds a new member called vm_prfile in struct
1982+vm_area_struct (and struct vm_region). The original vm_file points to
1983+the file on the branch fs in order to handle everything correctly as
1984+usual. The new vm_prfile points to a virtual file in aufs, and the
1985+show-functions in procfs refers to vm_prfile if it is set.
1986+Also we need to maintain several other places where touching vm_file
1987+such like
1988+- fork()/clone() copies vma and the reference count of vm_file is
1989+ incremented.
1990+- merging vma maintains the ref count too.
1991+
7e9cd9fe 1992+This is not a good approach. It just fakes the printed path. But it
53392da6
AM
1993+leaves all behaviour around f_mapping unchanged. This is surely an
1994+advantage.
1995+Actually aufs had adopted another complicated approach which calls
1996+generic_file_mmap() and handles struct vm_operations_struct. In this
1997+approach, aufs met a hard problem and I could not solve it without
1998+switching the approach.
b912730e
AM
1999+
2000+There may be one more another approach which is
2001+- bind-mount the branch-root onto the aufs-root internally
2002+- grab the new vfsmount (ie. struct mount)
2003+- lazy-umount the branch-root internally
2004+- in open(2) the aufs-file, open the branch-file with the hidden
2005+ vfsmount (instead of the original branch's vfsmount)
2006+- ideally this "bind-mount and lazy-umount" should be done atomically,
2007+ but it may be possible from userspace by the mount helper.
2008+
2009+Adding the internal hidden vfsmount and using it in opening a file, the
2010+file path under /proc will be printed correctly. This approach looks
2011+smarter, but is not possible I am afraid.
2012+- aufs-root may be bind-mount later. when it happens, another hidden
2013+ vfsmount will be required.
2014+- it is hard to get the chance to bind-mount and lazy-umount
2015+ + in kernel-space, FS can have vfsmount in open(2) via
2016+ file->f_path, and aufs can know its vfsmount. But several locks are
2017+ already acquired, and if aufs tries to bind-mount and lazy-umount
2018+ here, then it may cause a deadlock.
2019+ + in user-space, bind-mount doesn't invoke the mount helper.
2020+- since /proc shows dev and ino, aufs has to give vma these info. it
2021+ means a new member vm_prinode will be necessary. this is essentially
2022+ equivalent to vm_prfile described above.
2023+
2024+I have to give up this "looks-smater" approach.
c1595e42
JR
2025diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt linux/Documentation/filesystems/aufs/design/06xattr.txt
2026--- /usr/share/empty/Documentation/filesystems/aufs/design/06xattr.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2027+++ linux/Documentation/filesystems/aufs/design/06xattr.txt 2015-09-24 10:47:58.244719488 +0200
c1595e42
JR
2028@@ -0,0 +1,96 @@
2029+
2000de60 2030+# Copyright (C) 2014-2015 Junjiro R. Okajima
c1595e42
JR
2031+#
2032+# This program is free software; you can redistribute it and/or modify
2033+# it under the terms of the GNU General Public License as published by
2034+# the Free Software Foundation; either version 2 of the License, or
2035+# (at your option) any later version.
2036+#
2037+# This program is distributed in the hope that it will be useful,
2038+# but WITHOUT ANY WARRANTY; without even the implied warranty of
2039+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2040+# GNU General Public License for more details.
2041+#
2042+# You should have received a copy of the GNU General Public License
2043+# along with this program; if not, write to the Free Software
2044+# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2045+
2046+
2047+Listing XATTR/EA and getting the value
2048+----------------------------------------------------------------------
2049+For the inode standard attributes (owner, group, timestamps, etc.), aufs
2050+shows the values from the topmost existing file. This behaviour is good
7e9cd9fe 2051+for the non-dir entries since the bahaviour exactly matches the shown
c1595e42
JR
2052+information. But for the directories, aufs considers all the same named
2053+entries on the lower branches. Which means, if one of the lower entry
2054+rejects readdir call, then aufs returns an error even if the topmost
2055+entry allows it. This behaviour is necessary to respect the branch fs's
2056+security, but can make users confused since the user-visible standard
2057+attributes don't match the behaviour.
2058+To address this issue, aufs has a mount option called dirperm1 which
2059+checks the permission for the topmost entry only, and ignores the lower
2060+entry's permission.
2061+
2062+A similar issue can happen around XATTR.
2063+getxattr(2) and listxattr(2) families behave as if dirperm1 option is
7e9cd9fe
AM
2064+always set. Otherwise these very unpleasant situation would happen.
2065+- listxattr(2) may return the duplicated entries.
c1595e42
JR
2066+- users may not be able to remove or reset the XATTR forever,
2067+
2068+
2069+XATTR/EA support in the internal (copy,move)-(up,down)
2070+----------------------------------------------------------------------
7e9cd9fe 2071+Generally the extended attributes of inode are categorized as these.
c1595e42
JR
2072+- "security" for LSM and capability.
2073+- "system" for posix ACL, 'acl' mount option is required for the branch
2074+ fs generally.
2075+- "trusted" for userspace, CAP_SYS_ADMIN is required.
2076+- "user" for userspace, 'user_xattr' mount option is required for the
2077+ branch fs generally.
2078+
2079+Moreover there are some other categories. Aufs handles these rather
2080+unpopular categories as the ordinary ones, ie. there is no special
2081+condition nor exception.
2082+
2083+In copy-up, the support for XATTR on the dst branch may differ from the
2084+src branch. In this case, the copy-up operation will get an error and
7e9cd9fe
AM
2085+the original user operation which triggered the copy-up will fail. It
2086+can happen that even all copy-up will fail.
c1595e42
JR
2087+When both of src and dst branches support XATTR and if an error occurs
2088+during copying XATTR, then the copy-up should fail obviously. That is a
2089+good reason and aufs should return an error to userspace. But when only
7e9cd9fe 2090+the src branch support that XATTR, aufs should not return an error.
c1595e42
JR
2091+For example, the src branch supports ACL but the dst branch doesn't
2092+because the dst branch may natively un-support it or temporary
2093+un-support it due to "noacl" mount option. Of course, the dst branch fs
2094+may NOT return an error even if the XATTR is not supported. It is
2095+totally up to the branch fs.
2096+
2097+Anyway when the aufs internal copy-up gets an error from the dst branch
2098+fs, then aufs tries removing the just copied entry and returns the error
2099+to the userspace. The worst case of this situation will be all copy-up
2100+will fail.
2101+
2102+For the copy-up operation, there two basic approaches.
2103+- copy the specified XATTR only (by category above), and return the
7e9cd9fe 2104+ error unconditionally if it happens.
c1595e42
JR
2105+- copy all XATTR, and ignore the error on the specified category only.
2106+
2107+In order to support XATTR and to implement the correct behaviour, aufs
7e9cd9fe
AM
2108+chooses the latter approach and introduces some new branch attributes,
2109+"icexsec", "icexsys", "icextr", "icexusr", and "icexoth".
c1595e42 2110+They correspond to the XATTR namespaces (see above). Additionally, to be
7e9cd9fe
AM
2111+convenient, "icex" is also provided which means all "icex*" attributes
2112+are set (here the word "icex" stands for "ignore copy-error on XATTR").
c1595e42
JR
2113+
2114+The meaning of these attributes is to ignore the error from setting
2115+XATTR on that branch.
2116+Note that aufs tries copying all XATTR unconditionally, and ignores the
2117+error from the dst branch according to the specified attributes.
2118+
2119+Some XATTR may have its default value. The default value may come from
2120+the parent dir or the environment. If the default value is set at the
2121+file creating-time, it will be overwritten by copy-up.
2122+Some contradiction may happen I am afraid.
2123+Do we need another attribute to stop copying XATTR? I am unsure. For
2124+now, aufs implements the branch attributes to ignore the error.
53392da6
AM
2125diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
2126--- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2127+++ linux/Documentation/filesystems/aufs/design/07export.txt 2015-09-24 10:47:58.248052907 +0200
523b37e3 2128@@ -0,0 +1,58 @@
53392da6 2129+
2000de60 2130+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
2131+#
2132+# This program is free software; you can redistribute it and/or modify
2133+# it under the terms of the GNU General Public License as published by
2134+# the Free Software Foundation; either version 2 of the License, or
2135+# (at your option) any later version.
2136+#
2137+# This program is distributed in the hope that it will be useful,
2138+# but WITHOUT ANY WARRANTY; without even the implied warranty of
2139+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2140+# GNU General Public License for more details.
2141+#
2142+# You should have received a copy of the GNU General Public License
523b37e3 2143+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
2144+
2145+Export Aufs via NFS
2146+----------------------------------------------------------------------
2147+Here is an approach.
2148+- like xino/xib, add a new file 'xigen' which stores aufs inode
2149+ generation.
2150+- iget_locked(): initialize aufs inode generation for a new inode, and
2151+ store it in xigen file.
2152+- destroy_inode(): increment aufs inode generation and store it in xigen
2153+ file. it is necessary even if it is not unlinked, because any data of
2154+ inode may be changed by UDBA.
2155+- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
2156+ build file handle by
2157+ + branch id (4 bytes)
2158+ + superblock generation (4 bytes)
2159+ + inode number (4 or 8 bytes)
2160+ + parent dir inode number (4 or 8 bytes)
2161+ + inode generation (4 bytes))
2162+ + return value of exportfs_encode_fh() for the parent on a branch (4
2163+ bytes)
2164+ + file handle for a branch (by exportfs_encode_fh())
2165+- fh_to_dentry():
2166+ + find the index of a branch from its id in handle, and check it is
2167+ still exist in aufs.
2168+ + 1st level: get the inode number from handle and search it in cache.
7e9cd9fe
AM
2169+ + 2nd level: if not found in cache, get the parent inode number from
2170+ the handle and search it in cache. and then open the found parent
2171+ dir, find the matching inode number by vfs_readdir() and get its
2172+ name, and call lookup_one_len() for the target dentry.
53392da6
AM
2173+ + 3rd level: if the parent dir is not cached, call
2174+ exportfs_decode_fh() for a branch and get the parent on a branch,
2175+ build a pathname of it, convert it a pathname in aufs, call
2176+ path_lookup(). now aufs gets a parent dir dentry, then handle it as
2177+ the 2nd level.
2178+ + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
2179+ for every branch, but not itself. to get this, (currently) aufs
2180+ searches in current->nsproxy->mnt_ns list. it may not be a good
2181+ idea, but I didn't get other approach.
2182+ + test the generation of the gotten inode.
2183+- every inode operation: they may get EBUSY due to UDBA. in this case,
2184+ convert it into ESTALE for NFSD.
2185+- readdir(): call lockdep_on/off() because filldir in NFSD calls
2186+ lookup_one_len(), vfs_getattr(), encode_fh() and others.
2187diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
2188--- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2189+++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2015-09-24 10:47:58.248052907 +0200
523b37e3 2190@@ -0,0 +1,52 @@
53392da6 2191+
2000de60 2192+# Copyright (C) 2005-2015 Junjiro R. Okajima
53392da6
AM
2193+#
2194+# This program is free software; you can redistribute it and/or modify
2195+# it under the terms of the GNU General Public License as published by
2196+# the Free Software Foundation; either version 2 of the License, or
2197+# (at your option) any later version.
2198+#
2199+# This program is distributed in the hope that it will be useful,
2200+# but WITHOUT ANY WARRANTY; without even the implied warranty of
2201+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2202+# GNU General Public License for more details.
2203+#
2204+# You should have received a copy of the GNU General Public License
523b37e3 2205+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
2206+
2207+Show Whiteout Mode (shwh)
2208+----------------------------------------------------------------------
2209+Generally aufs hides the name of whiteouts. But in some cases, to show
2210+them is very useful for users. For instance, creating a new middle layer
2211+(branch) by merging existing layers.
2212+
2213+(borrowing aufs1 HOW-TO from a user, Michael Towers)
2214+When you have three branches,
2215+- Bottom: 'system', squashfs (underlying base system), read-only
2216+- Middle: 'mods', squashfs, read-only
2217+- Top: 'overlay', ram (tmpfs), read-write
2218+
2219+The top layer is loaded at boot time and saved at shutdown, to preserve
2220+the changes made to the system during the session.
2221+When larger changes have been made, or smaller changes have accumulated,
2222+the size of the saved top layer data grows. At this point, it would be
2223+nice to be able to merge the two overlay branches ('mods' and 'overlay')
2224+and rewrite the 'mods' squashfs, clearing the top layer and thus
2225+restoring save and load speed.
2226+
2227+This merging is simplified by the use of another aufs mount, of just the
2228+two overlay branches using the 'shwh' option.
2229+# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
2230+ aufs /livesys/merge_union
2231+
2232+A merged view of these two branches is then available at
2233+/livesys/merge_union, and the new feature is that the whiteouts are
2234+visible!
2235+Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
2236+writing to all branches. Also the default mode for all branches is 'ro'.
2237+It is now possible to save the combined contents of the two overlay
2238+branches to a new squashfs, e.g.:
2239+# mksquashfs /livesys/merge_union /path/to/newmods.squash
2240+
2241+This new squashfs archive can be stored on the boot device and the
2242+initramfs will use it to replace the old one at the next boot.
2243diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
2244--- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2245+++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2015-09-24 10:47:58.248052907 +0200
7e9cd9fe 2246@@ -0,0 +1,47 @@
53392da6 2247+
2000de60 2248+# Copyright (C) 2010-2015 Junjiro R. Okajima
53392da6
AM
2249+#
2250+# This program is free software; you can redistribute it and/or modify
2251+# it under the terms of the GNU General Public License as published by
2252+# the Free Software Foundation; either version 2 of the License, or
2253+# (at your option) any later version.
2254+#
2255+# This program is distributed in the hope that it will be useful,
2256+# but WITHOUT ANY WARRANTY; without even the implied warranty of
2257+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2258+# GNU General Public License for more details.
2259+#
2260+# You should have received a copy of the GNU General Public License
523b37e3 2261+# along with this program. If not, see <http://www.gnu.org/licenses/>.
53392da6
AM
2262+
2263+Dynamically customizable FS operations
2264+----------------------------------------------------------------------
2265+Generally FS operations (struct inode_operations, struct
2266+address_space_operations, struct file_operations, etc.) are defined as
2267+"static const", but it never means that FS have only one set of
2268+operation. Some FS have multiple sets of them. For instance, ext2 has
2269+three sets, one for XIP, for NOBH, and for normal.
2270+Since aufs overrides and redirects these operations, sometimes aufs has
7e9cd9fe 2271+to change its behaviour according to the branch FS type. More importantly
53392da6
AM
2272+VFS acts differently if a function (member in the struct) is set or
2273+not. It means aufs should have several sets of operations and select one
2274+among them according to the branch FS definition.
2275+
7e9cd9fe 2276+In order to solve this problem and not to affect the behaviour of VFS,
53392da6 2277+aufs defines these operations dynamically. For instance, aufs defines
7e9cd9fe
AM
2278+dummy direct_IO function for struct address_space_operations, but it may
2279+not be set to the address_space_operations actually. When the branch FS
2280+doesn't have it, aufs doesn't set it to its address_space_operations
2281+while the function definition itself is still alive. So the behaviour
2282+itself will not change, and it will return an error when direct_IO is
2283+not set.
53392da6
AM
2284+
2285+The lifetime of these dynamically generated operation object is
2286+maintained by aufs branch object. When the branch is removed from aufs,
2287+the reference counter of the object is decremented. When it reaches
2288+zero, the dynamically generated operation object will be freed.
2289+
7e9cd9fe
AM
2290+This approach is designed to support AIO (io_submit), Direct I/O and
2291+XIP (DAX) mainly.
2292+Currently this approach is applied to address_space_operations for
2293+regular files only.
53392da6
AM
2294diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
2295--- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2296+++ linux/Documentation/filesystems/aufs/README 2015-09-24 10:47:58.244719488 +0200
5527c038 2297@@ -0,0 +1,383 @@
53392da6 2298+
5527c038 2299+Aufs4 -- advanced multi layered unification filesystem version 4.x
53392da6
AM
2300+http://aufs.sf.net
2301+Junjiro R. Okajima
2302+
2303+
2304+0. Introduction
2305+----------------------------------------
2306+In the early days, aufs was entirely re-designed and re-implemented
7e9cd9fe 2307+Unionfs Version 1.x series. Adding many original ideas, approaches,
53392da6
AM
2308+improvements and implementations, it becomes totally different from
2309+Unionfs while keeping the basic features.
2310+Recently, Unionfs Version 2.x series begin taking some of the same
2311+approaches to aufs1's.
2312+Unionfs is being developed by Professor Erez Zadok at Stony Brook
2313+University and his team.
2314+
5527c038 2315+Aufs4 supports linux-4.0 and later, and for linux-3.x series try aufs3.
53392da6
AM
2316+If you want older kernel version support, try aufs2-2.6.git or
2317+aufs2-standalone.git repository, aufs1 from CVS on SourceForge.
2318+
2319+Note: it becomes clear that "Aufs was rejected. Let's give it up."
38d290e6
JR
2320+ According to Christoph Hellwig, linux rejects all union-type
2321+ filesystems but UnionMount.
53392da6
AM
2322+<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
2323+
38d290e6
JR
2324+PS. Al Viro seems have a plan to merge aufs as well as overlayfs and
2325+ UnionMount, and he pointed out an issue around a directory mutex
2326+ lock and aufs addressed it. But it is still unsure whether aufs will
2327+ be merged (or any other union solution).
076b876e 2328+<http://marc.info/?l=linux-kernel&m=136312705029295&w=1>
38d290e6 2329+
53392da6
AM
2330+
2331+1. Features
2332+----------------------------------------
2333+- unite several directories into a single virtual filesystem. The member
2334+ directory is called as a branch.
2335+- you can specify the permission flags to the branch, which are 'readonly',
2336+ 'readwrite' and 'whiteout-able.'
2337+- by upper writable branch, internal copyup and whiteout, files/dirs on
2338+ readonly branch are modifiable logically.
2339+- dynamic branch manipulation, add, del.
2340+- etc...
2341+
7e9cd9fe
AM
2342+Also there are many enhancements in aufs, such as:
2343+- test only the highest one for the directory permission (dirperm1)
2344+- copyup on open (coo=)
2345+- 'move' policy for copy-up between two writable branches, after
2346+ checking free space.
2347+- xattr, acl
53392da6
AM
2348+- readdir(3) in userspace.
2349+- keep inode number by external inode number table
2350+- keep the timestamps of file/dir in internal copyup operation
2351+- seekable directory, supporting NFS readdir.
2352+- whiteout is hardlinked in order to reduce the consumption of inodes
2353+ on branch
2354+- do not copyup, nor create a whiteout when it is unnecessary
2355+- revert a single systemcall when an error occurs in aufs
2356+- remount interface instead of ioctl
2357+- maintain /etc/mtab by an external command, /sbin/mount.aufs.
2358+- loopback mounted filesystem as a branch
2359+- kernel thread for removing the dir who has a plenty of whiteouts
2360+- support copyup sparse file (a file which has a 'hole' in it)
2361+- default permission flags for branches
2362+- selectable permission flags for ro branch, whether whiteout can
2363+ exist or not
2364+- export via NFS.
2365+- support <sysfs>/fs/aufs and <debugfs>/aufs.
2366+- support multiple writable branches, some policies to select one
2367+ among multiple writable branches.
2368+- a new semantics for link(2) and rename(2) to support multiple
2369+ writable branches.
2370+- no glibc changes are required.
2371+- pseudo hardlink (hardlink over branches)
2372+- allow a direct access manually to a file on branch, e.g. bypassing aufs.
2373+ including NFS or remote filesystem branch.
2374+- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
2375+- and more...
2376+
5527c038 2377+Currently these features are dropped temporary from aufs4.
53392da6 2378+See design/08plan.txt in detail.
53392da6
AM
2379+- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
2380+ (robr)
2381+- statistics of aufs thread (/sys/fs/aufs/stat)
53392da6
AM
2382+
2383+Features or just an idea in the future (see also design/*.txt),
2384+- reorder the branch index without del/re-add.
2385+- permanent xino files for NFSD
2386+- an option for refreshing the opened files after add/del branches
53392da6
AM
2387+- light version, without branch manipulation. (unnecessary?)
2388+- copyup in userspace
2389+- inotify in userspace
2390+- readv/writev
53392da6
AM
2391+
2392+
2393+2. Download
2394+----------------------------------------
5527c038
JR
2395+There are three GIT trees for aufs4, aufs4-linux.git,
2396+aufs4-standalone.git, and aufs-util.git. Note that there is no "4" in
1e00d052 2397+"aufs-util.git."
5527c038
JR
2398+While the aufs-util is always necessary, you need either of aufs4-linux
2399+or aufs4-standalone.
1e00d052 2400+
5527c038 2401+The aufs4-linux tree includes the whole linux mainline GIT tree,
1e00d052
AM
2402+git://git.kernel.org/.../torvalds/linux.git.
2403+And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
5527c038 2404+build aufs4 as an external kernel module.
2000de60 2405+Several extra patches are not included in this tree. Only
5527c038 2406+aufs4-standalone tree contains them. They are describe in the later
2000de60 2407+section "Configuration and Compilation."
1e00d052 2408+
5527c038 2409+On the other hand, the aufs4-standalone tree has only aufs source files
53392da6 2410+and necessary patches, and you can select CONFIG_AUFS_FS=m.
2000de60 2411+But you need to apply all aufs patches manually.
53392da6 2412+
5527c038
JR
2413+You will find GIT branches whose name is in form of "aufs4.x" where "x"
2414+represents the linux kernel version, "linux-4.x". For instance,
2415+"aufs4.0" is for linux-4.0. For latest "linux-4.x-rcN", use
2416+"aufs4.x-rcN" branch.
1e00d052 2417+
5527c038 2418+o aufs4-linux tree
1e00d052 2419+$ git clone --reference /your/linux/git/tree \
5527c038 2420+ git://github.com/sfjro/aufs4-linux.git aufs4-linux.git
1e00d052 2421+- if you don't have linux GIT tree, then remove "--reference ..."
5527c038
JR
2422+$ cd aufs4-linux.git
2423+$ git checkout origin/aufs4.0
53392da6 2424+
2000de60
JR
2425+Or You may want to directly git-pull aufs into your linux GIT tree, and
2426+leave the patch-work to GIT.
2427+$ cd /your/linux/git/tree
5527c038
JR
2428+$ git remote add aufs4 git://github.com/sfjro/aufs4-linux.git
2429+$ git fetch aufs4
2430+$ git checkout -b my4.0 v4.0
2431+$ (add your local change...)
2432+$ git pull aufs4 aufs4.0
2433+- now you have v4.0 + your_changes + aufs4.0 in you my4.0 branch.
2000de60 2434+- you may need to solve some conflicts between your_changes and
5527c038
JR
2435+ aufs4.0. in this case, git-rerere is recommended so that you can
2436+ solve the similar conflicts automatically when you upgrade to 4.1 or
2000de60
JR
2437+ later in the future.
2438+
5527c038
JR
2439+o aufs4-standalone tree
2440+$ git clone git://github.com/sfjro/aufs4-standalone.git aufs4-standalone.git
2441+$ cd aufs4-standalone.git
2442+$ git checkout origin/aufs4.0
53392da6
AM
2443+
2444+o aufs-util tree
5527c038
JR
2445+$ git clone git://git.code.sf.net/p/aufs/aufs-util aufs-util.git
2446+- note that the public aufs-util.git is on SourceForge instead of
2447+ GitHUB.
53392da6 2448+$ cd aufs-util.git
5527c038 2449+$ git checkout origin/aufs4.0
53392da6 2450+
5527c038
JR
2451+Note: The 4.x-rcN branch is to be used with `rc' kernel versions ONLY.
2452+The minor version number, 'x' in '4.x', of aufs may not always
9dbd164d
AM
2453+follow the minor version number of the kernel.
2454+Because changes in the kernel that cause the use of a new
2455+minor version number do not always require changes to aufs-util.
2456+
2457+Since aufs-util has its own minor version number, you may not be
2458+able to find a GIT branch in aufs-util for your kernel's
2459+exact minor version number.
2460+In this case, you should git-checkout the branch for the
53392da6 2461+nearest lower number.
9dbd164d
AM
2462+
2463+For (an unreleased) example:
5527c038
JR
2464+If you are using "linux-4.10" and the "aufs4.10" branch
2465+does not exist in aufs-util repository, then "aufs4.9", "aufs4.8"
9dbd164d
AM
2466+or something numerically smaller is the branch for your kernel.
2467+
53392da6
AM
2468+Also you can view all branches by
2469+ $ git branch -a
2470+
2471+
2472+3. Configuration and Compilation
2473+----------------------------------------
2474+Make sure you have git-checkout'ed the correct branch.
2475+
5527c038 2476+For aufs4-linux tree,
c06a8ce3 2477+- enable CONFIG_AUFS_FS.
1e00d052
AM
2478+- set other aufs configurations if necessary.
2479+
5527c038 2480+For aufs4-standalone tree,
53392da6
AM
2481+There are several ways to build.
2482+
2483+1.
5527c038
JR
2484+- apply ./aufs4-kbuild.patch to your kernel source files.
2485+- apply ./aufs4-base.patch too.
2486+- apply ./aufs4-mmap.patch too.
2487+- apply ./aufs4-standalone.patch too, if you have a plan to set
2488+ CONFIG_AUFS_FS=m. otherwise you don't need ./aufs4-standalone.patch.
537831f9
AM
2489+- copy ./{Documentation,fs,include/uapi/linux/aufs_type.h} files to your
2490+ kernel source tree. Never copy $PWD/include/uapi/linux/Kbuild.
c06a8ce3 2491+- enable CONFIG_AUFS_FS, you can select either
53392da6
AM
2492+ =m or =y.
2493+- and build your kernel as usual.
2494+- install the built kernel.
c06a8ce3
AM
2495+ Note: Since linux-3.9, every filesystem module requires an alias
2496+ "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2497+ modules.aliases file if you set CONFIG_AUFS_FS=m.
7eafdf33
AM
2498+- install the header files too by "make headers_install" to the
2499+ directory where you specify. By default, it is $PWD/usr.
b4510431 2500+ "make help" shows a brief note for headers_install.
53392da6
AM
2501+- and reboot your system.
2502+
2503+2.
2504+- module only (CONFIG_AUFS_FS=m).
5527c038
JR
2505+- apply ./aufs4-base.patch to your kernel source files.
2506+- apply ./aufs4-mmap.patch too.
2507+- apply ./aufs4-standalone.patch too.
53392da6
AM
2508+- build your kernel, don't forget "make headers_install", and reboot.
2509+- edit ./config.mk and set other aufs configurations if necessary.
b4510431 2510+ Note: You should read $PWD/fs/aufs/Kconfig carefully which describes
53392da6
AM
2511+ every aufs configurations.
2512+- build the module by simple "make".
c06a8ce3
AM
2513+ Note: Since linux-3.9, every filesystem module requires an alias
2514+ "fs-<fsname>". You should make sure that "fs-aufs" is listed in your
2515+ modules.aliases file.
53392da6
AM
2516+- you can specify ${KDIR} make variable which points to your kernel
2517+ source tree.
2518+- install the files
2519+ + run "make install" to install the aufs module, or copy the built
b4510431
AM
2520+ $PWD/aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
2521+ + run "make install_headers" (instead of headers_install) to install
2522+ the modified aufs header file (you can specify DESTDIR which is
2523+ available in aufs standalone version's Makefile only), or copy
2524+ $PWD/usr/include/linux/aufs_type.h to /usr/include/linux or wherever
2525+ you like manually. By default, the target directory is $PWD/usr.
5527c038 2526+- no need to apply aufs4-kbuild.patch, nor copying source files to your
53392da6
AM
2527+ kernel source tree.
2528+
b4510431 2529+Note: The header file aufs_type.h is necessary to build aufs-util
53392da6
AM
2530+ as well as "make headers_install" in the kernel source tree.
2531+ headers_install is subject to be forgotten, but it is essentially
2532+ necessary, not only for building aufs-util.
2533+ You may not meet problems without headers_install in some older
2534+ version though.
2535+
2536+And then,
2537+- read README in aufs-util, build and install it
9dbd164d
AM
2538+- note that your distribution may contain an obsoleted version of
2539+ aufs_type.h in /usr/include/linux or something. When you build aufs
2540+ utilities, make sure that your compiler refers the correct aufs header
2541+ file which is built by "make headers_install."
53392da6
AM
2542+- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
2543+ then run "make install_ulib" too. And refer to the aufs manual in
2544+ detail.
2545+
5527c038 2546+There several other patches in aufs4-standalone.git. They are all
38d290e6 2547+optional. When you meet some problems, they will help you.
5527c038 2548+- aufs4-loopback.patch
38d290e6
JR
2549+ Supports a nested loopback mount in a branch-fs. This patch is
2550+ unnecessary until aufs produces a message like "you may want to try
2551+ another patch for loopback file".
2552+- vfs-ino.patch
2553+ Modifies a system global kernel internal function get_next_ino() in
2554+ order to stop assigning 0 for an inode-number. Not directly related to
2555+ aufs, but recommended generally.
2556+- tmpfs-idr.patch
2557+ Keeps the tmpfs inode number as the lowest value. Effective to reduce
2558+ the size of aufs XINO files for tmpfs branch. Also it prevents the
2559+ duplication of inode number, which is important for backup tools and
2560+ other utilities. When you find aufs XINO files for tmpfs branch
2561+ growing too much, try this patch.
2562+
53392da6
AM
2563+
2564+4. Usage
2565+----------------------------------------
2566+At first, make sure aufs-util are installed, and please read the aufs
2567+manual, aufs.5 in aufs-util.git tree.
2568+$ man -l aufs.5
2569+
2570+And then,
2571+$ mkdir /tmp/rw /tmp/aufs
2572+# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
2573+
2574+Here is another example. The result is equivalent.
2575+# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
2576+ Or
2577+# mount -t aufs -o br:/tmp/rw none /tmp/aufs
2578+# mount -o remount,append:${HOME} /tmp/aufs
2579+
2580+Then, you can see whole tree of your home dir through /tmp/aufs. If
2581+you modify a file under /tmp/aufs, the one on your home directory is
2582+not affected, instead the same named file will be newly created under
2583+/tmp/rw. And all of your modification to a file will be applied to
2584+the one under /tmp/rw. This is called the file based Copy on Write
2585+(COW) method.
2586+Aufs mount options are described in aufs.5.
2587+If you run chroot or something and make your aufs as a root directory,
2588+then you need to customize the shutdown script. See the aufs manual in
2589+detail.
2590+
2591+Additionally, there are some sample usages of aufs which are a
2592+diskless system with network booting, and LiveCD over NFS.
2593+See sample dir in CVS tree on SourceForge.
2594+
2595+
2596+5. Contact
2597+----------------------------------------
2598+When you have any problems or strange behaviour in aufs, please let me
2599+know with:
2600+- /proc/mounts (instead of the output of mount(8))
2601+- /sys/module/aufs/*
2602+- /sys/fs/aufs/* (if you have them)
2603+- /debug/aufs/* (if you have them)
2604+- linux kernel version
2605+ if your kernel is not plain, for example modified by distributor,
2606+ the url where i can download its source is necessary too.
2607+- aufs version which was printed at loading the module or booting the
2608+ system, instead of the date you downloaded.
2609+- configuration (define/undefine CONFIG_AUFS_xxx)
2610+- kernel configuration or /proc/config.gz (if you have it)
2611+- behaviour which you think to be incorrect
2612+- actual operation, reproducible one is better
2613+- mailto: aufs-users at lists.sourceforge.net
2614+
2615+Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
2616+and Feature Requests) on SourceForge. Please join and write to
2617+aufs-users ML.
2618+
2619+
2620+6. Acknowledgements
2621+----------------------------------------
2622+Thanks to everyone who have tried and are using aufs, whoever
2623+have reported a bug or any feedback.
2624+
2625+Especially donators:
2626+Tomas Matejicek(slax.org) made a donation (much more than once).
2627+ Since Apr 2010, Tomas M (the author of Slax and Linux Live
2628+ scripts) is making "doubling" donations.
2629+ Unfortunately I cannot list all of the donators, but I really
b4510431 2630+ appreciate.
53392da6
AM
2631+ It ends Aug 2010, but the ordinary donation URL is still available.
2632+ <http://sourceforge.net/donate/index.php?group_id=167503>
2633+Dai Itasaka made a donation (2007/8).
2634+Chuck Smith made a donation (2008/4, 10 and 12).
2635+Henk Schoneveld made a donation (2008/9).
2636+Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
2637+Francois Dupoux made a donation (2008/11).
2638+Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
2639+ aufs2 GIT tree (2009/2).
2640+William Grant made a donation (2009/3).
2641+Patrick Lane made a donation (2009/4).
2642+The Mail Archive (mail-archive.com) made donations (2009/5).
2643+Nippy Networks (Ed Wildgoose) made a donation (2009/7).
2644+New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
2645+Pavel Pronskiy made a donation (2011/2).
2646+Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
2647+ Networks (Ed Wildgoose) made a donation for hardware (2011/3).
537831f9
AM
2648+Max Lekomcev (DOM-TV project) made a donation (2011/7, 12, 2012/3, 6 and
2649+11).
1e00d052 2650+Sam Liddicott made a donation (2011/9).
86dc4139
AM
2651+Era Scarecrow made a donation (2013/4).
2652+Bor Ratajc made a donation (2013/4).
2653+Alessandro Gorreta made a donation (2013/4).
2654+POIRETTE Marc made a donation (2013/4).
2655+Alessandro Gorreta made a donation (2013/4).
2656+lauri kasvandik made a donation (2013/5).
392086de 2657+"pemasu from Finland" made a donation (2013/7).
523b37e3
AM
2658+The Parted Magic Project made a donation (2013/9 and 11).
2659+Pavel Barta made a donation (2013/10).
38d290e6 2660+Nikolay Pertsev made a donation (2014/5).
c2c0f25c 2661+James B made a donation (2014/7 and 2015/7).
076b876e 2662+Stefano Di Biase made a donation (2014/8).
2000de60 2663+Daniel Epellei made a donation (2015/1).
53392da6
AM
2664+
2665+Thank you very much.
2666+Donations are always, including future donations, very important and
2667+helpful for me to keep on developing aufs.
2668+
2669+
2670+7.
2671+----------------------------------------
2672+If you are an experienced user, no explanation is needed. Aufs is
2673+just a linux filesystem.
2674+
2675+
2676+Enjoy!
2677+
2678+# Local variables: ;
2679+# mode: text;
2680+# End: ;
7f207e10
AM
2681diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
2682--- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2683+++ linux/fs/aufs/aufs.h 2015-09-24 10:47:58.248052907 +0200
523b37e3 2684@@ -0,0 +1,59 @@
7f207e10 2685+/*
2000de60 2686+ * Copyright (C) 2005-2015 Junjiro R. Okajima
7f207e10
AM
2687+ *
2688+ * This program, aufs is free software; you can redistribute it and/or modify
2689+ * it under the terms of the GNU General Public License as published by
2690+ * the Free Software Foundation; either version 2 of the License, or
2691+ * (at your option) any later version.
2692+ *
2693+ * This program is distributed in the hope that it will be useful,
2694+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
2695+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2696+ * GNU General Public License for more details.
2697+ *
2698+ * You should have received a copy of the GNU General Public License
523b37e3 2699+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
7f207e10
AM
2700+ */
2701+
2702+/*
2703+ * all header files
2704+ */
2705+
2706+#ifndef __AUFS_H__
2707+#define __AUFS_H__
2708+
2709+#ifdef __KERNEL__
2710+
2711+#define AuStub(type, name, body, ...) \
2712+ static inline type name(__VA_ARGS__) { body; }
2713+
2714+#define AuStubVoid(name, ...) \
2715+ AuStub(void, name, , __VA_ARGS__)
2716+#define AuStubInt0(name, ...) \
2717+ AuStub(int, name, return 0, __VA_ARGS__)
2718+
2719+#include "debug.h"
2720+
2721+#include "branch.h"
2722+#include "cpup.h"
2723+#include "dcsub.h"
2724+#include "dbgaufs.h"
2725+#include "dentry.h"
2726+#include "dir.h"
2727+#include "dynop.h"
2728+#include "file.h"
2729+#include "fstype.h"
2730+#include "inode.h"
2731+#include "loop.h"
2732+#include "module.h"
7f207e10
AM
2733+#include "opts.h"
2734+#include "rwsem.h"
2735+#include "spl.h"
2736+#include "super.h"
2737+#include "sysaufs.h"
2738+#include "vfsub.h"
2739+#include "whout.h"
2740+#include "wkq.h"
2741+
2742+#endif /* __KERNEL__ */
2743+#endif /* __AUFS_H__ */
2744diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
2745--- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 2746+++ linux/fs/aufs/branch.c 2015-09-24 10:47:58.248052907 +0200
5527c038 2747@@ -0,0 +1,1414 @@
7f207e10 2748+/*
2000de60 2749+ * Copyright (C) 2005-2015 Junjiro R. Okajima
7f207e10
AM
2750+ *
2751+ * This program, aufs is free software; you can redistribute it and/or modify
2752+ * it under the terms of the GNU General Public License as published by
2753+ * the Free Software Foundation; either version 2 of the License, or
2754+ * (at your option) any later version.
2755+ *
2756+ * This program is distributed in the hope that it will be useful,
2757+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
2758+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2759+ * GNU General Public License for more details.
2760+ *
2761+ * You should have received a copy of the GNU General Public License
523b37e3 2762+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
7f207e10
AM
2763+ */
2764+
2765+/*
2766+ * branch management
2767+ */
2768+
027c5e7a 2769+#include <linux/compat.h>
7f207e10
AM
2770+#include <linux/statfs.h>
2771+#include "aufs.h"
2772+
2773+/*
2774+ * free a single branch
1facf9fc 2775+ */
2776+static void au_br_do_free(struct au_branch *br)
2777+{
2778+ int i;
2779+ struct au_wbr *wbr;
4a4d8108 2780+ struct au_dykey **key;
1facf9fc 2781+
027c5e7a
AM
2782+ au_hnotify_fin_br(br);
2783+
1facf9fc 2784+ if (br->br_xino.xi_file)
2785+ fput(br->br_xino.xi_file);
2786+ mutex_destroy(&br->br_xino.xi_nondir_mtx);
2787+
2788+ AuDebugOn(atomic_read(&br->br_count));
2789+
2790+ wbr = br->br_wbr;
2791+ if (wbr) {
2792+ for (i = 0; i < AuBrWh_Last; i++)
2793+ dput(wbr->wbr_wh[i]);
2794+ AuDebugOn(atomic_read(&wbr->wbr_wh_running));
dece6358 2795+ AuRwDestroy(&wbr->wbr_wh_rwsem);
1facf9fc 2796+ }
2797+
076b876e
AM
2798+ if (br->br_fhsm) {
2799+ au_br_fhsm_fin(br->br_fhsm);
2800+ kfree(br->br_fhsm);
2801+ }
2802+
4a4d8108
AM
2803+ key = br->br_dykey;
2804+ for (i = 0; i < AuBrDynOp; i++, key++)
2805+ if (*key)
2806+ au_dy_put(*key);
2807+ else
2808+ break;
2809+
537831f9
AM
2810+ /* recursive lock, s_umount of branch's */
2811+ lockdep_off();
86dc4139 2812+ path_put(&br->br_path);
537831f9 2813+ lockdep_on();
1facf9fc 2814+ kfree(wbr);
2815+ kfree(br);
2816+}
2817+
2818+/*
2819+ * frees all branches
2820+ */
2821+void au_br_free(struct au_sbinfo *sbinfo)
2822+{
2823+ aufs_bindex_t bmax;
2824+ struct au_branch **br;
2825+
dece6358
AM
2826+ AuRwMustWriteLock(&sbinfo->si_rwsem);
2827+
1facf9fc 2828+ bmax = sbinfo->si_bend + 1;
2829+ br = sbinfo->si_branch;
2830+ while (bmax--)
2831+ au_br_do_free(*br++);
2832+}
2833+
2834+/*
2835+ * find the index of a branch which is specified by @br_id.
2836+ */
2837+int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
2838+{
2839+ aufs_bindex_t bindex, bend;
2840+
2841+ bend = au_sbend(sb);
2842+ for (bindex = 0; bindex <= bend; bindex++)
2843+ if (au_sbr_id(sb, bindex) == br_id)
2844+ return bindex;
2845+ return -1;
2846+}
2847+
2848+/* ---------------------------------------------------------------------- */
2849+
2850+/*
2851+ * add a branch
2852+ */
2853+
b752ccd1
AM
2854+static int test_overlap(struct super_block *sb, struct dentry *h_adding,
2855+ struct dentry *h_root)
1facf9fc 2856+{
b752ccd1
AM
2857+ if (unlikely(h_adding == h_root
2858+ || au_test_loopback_overlap(sb, h_adding)))
1facf9fc 2859+ return 1;
b752ccd1
AM
2860+ if (h_adding->d_sb != h_root->d_sb)
2861+ return 0;
2862+ return au_test_subdir(h_adding, h_root)
2863+ || au_test_subdir(h_root, h_adding);
1facf9fc 2864+}
2865+
2866+/*
2867+ * returns a newly allocated branch. @new_nbranch is a number of branches
2868+ * after adding a branch.
2869+ */
2870+static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
2871+ int perm)
2872+{
2873+ struct au_branch *add_branch;
2874+ struct dentry *root;
5527c038 2875+ struct inode *inode;
4a4d8108 2876+ int err;
1facf9fc 2877+
4a4d8108 2878+ err = -ENOMEM;
1facf9fc 2879+ root = sb->s_root;
2880+ add_branch = kmalloc(sizeof(*add_branch), GFP_NOFS);
2881+ if (unlikely(!add_branch))
2882+ goto out;
2883+
027c5e7a
AM
2884+ err = au_hnotify_init_br(add_branch, perm);
2885+ if (unlikely(err))
2886+ goto out_br;
2887+
1facf9fc 2888+ add_branch->br_wbr = NULL;
2889+ if (au_br_writable(perm)) {
2890+ /* may be freed separately at changing the branch permission */
2891+ add_branch->br_wbr = kmalloc(sizeof(*add_branch->br_wbr),
2892+ GFP_NOFS);
2893+ if (unlikely(!add_branch->br_wbr))
027c5e7a 2894+ goto out_hnotify;
1facf9fc 2895+ }
2896+
076b876e
AM
2897+ add_branch->br_fhsm = NULL;
2898+ if (au_br_fhsm(perm)) {
2899+ err = au_fhsm_br_alloc(add_branch);
2900+ if (unlikely(err))
2901+ goto out_wbr;
2902+ }
2903+
4a4d8108
AM
2904+ err = au_sbr_realloc(au_sbi(sb), new_nbranch);
2905+ if (!err)
2906+ err = au_di_realloc(au_di(root), new_nbranch);
5527c038
JR
2907+ if (!err) {
2908+ inode = d_inode(root);
2909+ err = au_ii_realloc(au_ii(inode), new_nbranch);
2910+ }
4a4d8108
AM
2911+ if (!err)
2912+ return add_branch; /* success */
1facf9fc 2913+
076b876e 2914+out_wbr:
1facf9fc 2915+ kfree(add_branch->br_wbr);
027c5e7a
AM
2916+out_hnotify:
2917+ au_hnotify_fin_br(add_branch);
4f0767ce 2918+out_br:
1facf9fc 2919+ kfree(add_branch);
4f0767ce 2920+out:
4a4d8108 2921+ return ERR_PTR(err);
1facf9fc 2922+}
2923+
2924+/*
2925+ * test if the branch permission is legal or not.
2926+ */
2927+static int test_br(struct inode *inode, int brperm, char *path)
2928+{
2929+ int err;
2930+
4a4d8108
AM
2931+ err = (au_br_writable(brperm) && IS_RDONLY(inode));
2932+ if (!err)
2933+ goto out;
1facf9fc 2934+
4a4d8108
AM
2935+ err = -EINVAL;
2936+ pr_err("write permission for readonly mount or inode, %s\n", path);
2937+
4f0767ce 2938+out:
1facf9fc 2939+ return err;
2940+}
2941+
2942+/*
2943+ * returns:
2944+ * 0: success, the caller will add it
2945+ * plus: success, it is already unified, the caller should ignore it
2946+ * minus: error
2947+ */
2948+static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
2949+{
2950+ int err;
2951+ aufs_bindex_t bend, bindex;
5527c038 2952+ struct dentry *root, *h_dentry;
1facf9fc 2953+ struct inode *inode, *h_inode;
2954+
2955+ root = sb->s_root;
2956+ bend = au_sbend(sb);
2957+ if (unlikely(bend >= 0
2958+ && au_find_dbindex(root, add->path.dentry) >= 0)) {
2959+ err = 1;
2960+ if (!remount) {
2961+ err = -EINVAL;
4a4d8108 2962+ pr_err("%s duplicated\n", add->pathname);
1facf9fc 2963+ }
2964+ goto out;
2965+ }
2966+
2967+ err = -ENOSPC; /* -E2BIG; */
2968+ if (unlikely(AUFS_BRANCH_MAX <= add->bindex
2969+ || AUFS_BRANCH_MAX - 1 <= bend)) {
4a4d8108 2970+ pr_err("number of branches exceeded %s\n", add->pathname);
1facf9fc 2971+ goto out;
2972+ }
2973+
2974+ err = -EDOM;
2975+ if (unlikely(add->bindex < 0 || bend + 1 < add->bindex)) {
4a4d8108 2976+ pr_err("bad index %d\n", add->bindex);
1facf9fc 2977+ goto out;
2978+ }
2979+
5527c038 2980+ inode = d_inode(add->path.dentry);
1facf9fc 2981+ err = -ENOENT;
2982+ if (unlikely(!inode->i_nlink)) {
4a4d8108 2983+ pr_err("no existence %s\n", add->pathname);
1facf9fc 2984+ goto out;
2985+ }
2986+
2987+ err = -EINVAL;
2988+ if (unlikely(inode->i_sb == sb)) {
4a4d8108 2989+ pr_err("%s must be outside\n", add->pathname);
1facf9fc 2990+ goto out;
2991+ }
2992+
2993+ if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
4a4d8108
AM
2994+ pr_err("unsupported filesystem, %s (%s)\n",
2995+ add->pathname, au_sbtype(inode->i_sb));
1facf9fc 2996+ goto out;
2997+ }
2998+
c1595e42
JR
2999+ if (unlikely(inode->i_sb->s_stack_depth)) {
3000+ pr_err("already stacked, %s (%s)\n",
3001+ add->pathname, au_sbtype(inode->i_sb));
3002+ goto out;
3003+ }
3004+
5527c038 3005+ err = test_br(d_inode(add->path.dentry), add->perm, add->pathname);
1facf9fc 3006+ if (unlikely(err))
3007+ goto out;
3008+
3009+ if (bend < 0)
3010+ return 0; /* success */
3011+
3012+ err = -EINVAL;
3013+ for (bindex = 0; bindex <= bend; bindex++)
3014+ if (unlikely(test_overlap(sb, add->path.dentry,
3015+ au_h_dptr(root, bindex)))) {
4a4d8108 3016+ pr_err("%s is overlapped\n", add->pathname);
1facf9fc 3017+ goto out;
3018+ }
3019+
3020+ err = 0;
3021+ if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
5527c038
JR
3022+ h_dentry = au_h_dptr(root, 0);
3023+ h_inode = d_inode(h_dentry);
1facf9fc 3024+ if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
0c3ec466
AM
3025+ || !uid_eq(h_inode->i_uid, inode->i_uid)
3026+ || !gid_eq(h_inode->i_gid, inode->i_gid))
3027+ pr_warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
3028+ add->pathname,
3029+ i_uid_read(inode), i_gid_read(inode),
3030+ (inode->i_mode & S_IALLUGO),
3031+ i_uid_read(h_inode), i_gid_read(h_inode),
3032+ (h_inode->i_mode & S_IALLUGO));
1facf9fc 3033+ }
3034+
4f0767ce 3035+out:
1facf9fc 3036+ return err;
3037+}
3038+
3039+/*
3040+ * initialize or clean the whiteouts for an adding branch
3041+ */
3042+static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
86dc4139 3043+ int new_perm)
1facf9fc 3044+{
3045+ int err, old_perm;
3046+ aufs_bindex_t bindex;
3047+ struct mutex *h_mtx;
3048+ struct au_wbr *wbr;
3049+ struct au_hinode *hdir;
5527c038 3050+ struct dentry *h_dentry;
1facf9fc 3051+
86dc4139
AM
3052+ err = vfsub_mnt_want_write(au_br_mnt(br));
3053+ if (unlikely(err))
3054+ goto out;
3055+
1facf9fc 3056+ wbr = br->br_wbr;
3057+ old_perm = br->br_perm;
3058+ br->br_perm = new_perm;
3059+ hdir = NULL;
3060+ h_mtx = NULL;
3061+ bindex = au_br_index(sb, br->br_id);
3062+ if (0 <= bindex) {
5527c038 3063+ hdir = au_hi(d_inode(sb->s_root), bindex);
4a4d8108 3064+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
1facf9fc 3065+ } else {
5527c038
JR
3066+ h_dentry = au_br_dentry(br);
3067+ h_mtx = &d_inode(h_dentry)->i_mutex;
1facf9fc 3068+ mutex_lock_nested(h_mtx, AuLsc_I_PARENT);
3069+ }
3070+ if (!wbr)
86dc4139 3071+ err = au_wh_init(br, sb);
1facf9fc 3072+ else {
3073+ wbr_wh_write_lock(wbr);
86dc4139 3074+ err = au_wh_init(br, sb);
1facf9fc 3075+ wbr_wh_write_unlock(wbr);
3076+ }
3077+ if (hdir)
4a4d8108 3078+ au_hn_imtx_unlock(hdir);
1facf9fc 3079+ else
3080+ mutex_unlock(h_mtx);
86dc4139 3081+ vfsub_mnt_drop_write(au_br_mnt(br));
1facf9fc 3082+ br->br_perm = old_perm;
3083+
3084+ if (!err && wbr && !au_br_writable(new_perm)) {
3085+ kfree(wbr);
3086+ br->br_wbr = NULL;
3087+ }
3088+
86dc4139 3089+out:
1facf9fc 3090+ return err;
3091+}
3092+
3093+static int au_wbr_init(struct au_branch *br, struct super_block *sb,
86dc4139 3094+ int perm)
1facf9fc 3095+{
3096+ int err;
4a4d8108 3097+ struct kstatfs kst;
1facf9fc 3098+ struct au_wbr *wbr;
3099+
3100+ wbr = br->br_wbr;
dece6358 3101+ au_rw_init(&wbr->wbr_wh_rwsem);
1facf9fc 3102+ memset(wbr->wbr_wh, 0, sizeof(wbr->wbr_wh));
3103+ atomic_set(&wbr->wbr_wh_running, 0);
3104+ wbr->wbr_bytes = 0;
3105+
4a4d8108
AM
3106+ /*
3107+ * a limit for rmdir/rename a dir
523b37e3 3108+ * cf. AUFS_MAX_NAMELEN in include/uapi/linux/aufs_type.h
4a4d8108 3109+ */
86dc4139 3110+ err = vfs_statfs(&br->br_path, &kst);
4a4d8108
AM
3111+ if (unlikely(err))
3112+ goto out;
3113+ err = -EINVAL;
3114+ if (kst.f_namelen >= NAME_MAX)
86dc4139 3115+ err = au_br_init_wh(sb, br, perm);
4a4d8108 3116+ else
523b37e3
AM
3117+ pr_err("%pd(%s), unsupported namelen %ld\n",
3118+ au_br_dentry(br),
86dc4139 3119+ au_sbtype(au_br_dentry(br)->d_sb), kst.f_namelen);
1facf9fc 3120+
4f0767ce 3121+out:
1facf9fc 3122+ return err;
3123+}
3124+
c1595e42 3125+/* initialize a new branch */
1facf9fc 3126+static int au_br_init(struct au_branch *br, struct super_block *sb,
3127+ struct au_opt_add *add)
3128+{
3129+ int err;
5527c038 3130+ struct inode *h_inode;
1facf9fc 3131+
3132+ err = 0;
3133+ memset(&br->br_xino, 0, sizeof(br->br_xino));
3134+ mutex_init(&br->br_xino.xi_nondir_mtx);
3135+ br->br_perm = add->perm;
86dc4139 3136+ br->br_path = add->path; /* set first, path_get() later */
4a4d8108
AM
3137+ spin_lock_init(&br->br_dykey_lock);
3138+ memset(br->br_dykey, 0, sizeof(br->br_dykey));
1facf9fc 3139+ atomic_set(&br->br_count, 0);
1facf9fc 3140+ atomic_set(&br->br_xino_running, 0);
3141+ br->br_id = au_new_br_id(sb);
7f207e10 3142+ AuDebugOn(br->br_id < 0);
1facf9fc 3143+
3144+ if (au_br_writable(add->perm)) {
86dc4139 3145+ err = au_wbr_init(br, sb, add->perm);
1facf9fc 3146+ if (unlikely(err))
b752ccd1 3147+ goto out_err;
1facf9fc 3148+ }
3149+
3150+ if (au_opt_test(au_mntflags(sb), XINO)) {
5527c038
JR
3151+ h_inode = d_inode(add->path.dentry);
3152+ err = au_xino_br(sb, br, h_inode->i_ino,
1facf9fc 3153+ au_sbr(sb, 0)->br_xino.xi_file, /*do_test*/1);
3154+ if (unlikely(err)) {
3155+ AuDebugOn(br->br_xino.xi_file);
b752ccd1 3156+ goto out_err;
1facf9fc 3157+ }
3158+ }
3159+
3160+ sysaufs_br_init(br);
86dc4139 3161+ path_get(&br->br_path);
b752ccd1 3162+ goto out; /* success */
1facf9fc 3163+
4f0767ce 3164+out_err:
86dc4139 3165+ memset(&br->br_path, 0, sizeof(br->br_path));
4f0767ce 3166+out:
1facf9fc 3167+ return err;
3168+}
3169+
3170+static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
3171+ struct au_branch *br, aufs_bindex_t bend,
3172+ aufs_bindex_t amount)
3173+{
3174+ struct au_branch **brp;
3175+
dece6358
AM
3176+ AuRwMustWriteLock(&sbinfo->si_rwsem);
3177+
1facf9fc 3178+ brp = sbinfo->si_branch + bindex;
3179+ memmove(brp + 1, brp, sizeof(*brp) * amount);
3180+ *brp = br;
3181+ sbinfo->si_bend++;
3182+ if (unlikely(bend < 0))
3183+ sbinfo->si_bend = 0;
3184+}
3185+
3186+static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
3187+ aufs_bindex_t bend, aufs_bindex_t amount)
3188+{
3189+ struct au_hdentry *hdp;
3190+
1308ab2a 3191+ AuRwMustWriteLock(&dinfo->di_rwsem);
3192+
1facf9fc 3193+ hdp = dinfo->di_hdentry + bindex;
3194+ memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
3195+ au_h_dentry_init(hdp);
3196+ dinfo->di_bend++;
3197+ if (unlikely(bend < 0))
3198+ dinfo->di_bstart = 0;
3199+}
3200+
3201+static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
3202+ aufs_bindex_t bend, aufs_bindex_t amount)
3203+{
3204+ struct au_hinode *hip;
3205+
1308ab2a 3206+ AuRwMustWriteLock(&iinfo->ii_rwsem);
3207+
1facf9fc 3208+ hip = iinfo->ii_hinode + bindex;
3209+ memmove(hip + 1, hip, sizeof(*hip) * amount);
3210+ hip->hi_inode = NULL;
4a4d8108 3211+ au_hn_init(hip);
1facf9fc 3212+ iinfo->ii_bend++;
3213+ if (unlikely(bend < 0))
3214+ iinfo->ii_bstart = 0;
3215+}
3216+
86dc4139
AM
3217+static void au_br_do_add(struct super_block *sb, struct au_branch *br,
3218+ aufs_bindex_t bindex)
1facf9fc 3219+{
86dc4139 3220+ struct dentry *root, *h_dentry;
5527c038 3221+ struct inode *root_inode, *h_inode;
1facf9fc 3222+ aufs_bindex_t bend, amount;
3223+
3224+ root = sb->s_root;
5527c038 3225+ root_inode = d_inode(root);
1facf9fc 3226+ bend = au_sbend(sb);
3227+ amount = bend + 1 - bindex;
86dc4139 3228+ h_dentry = au_br_dentry(br);
53392da6 3229+ au_sbilist_lock();
1facf9fc 3230+ au_br_do_add_brp(au_sbi(sb), bindex, br, bend, amount);
3231+ au_br_do_add_hdp(au_di(root), bindex, bend, amount);
3232+ au_br_do_add_hip(au_ii(root_inode), bindex, bend, amount);
3233+ au_set_h_dptr(root, bindex, dget(h_dentry));
5527c038
JR
3234+ h_inode = d_inode(h_dentry);
3235+ au_set_h_iptr(root_inode, bindex, au_igrab(h_inode), /*flags*/0);
53392da6 3236+ au_sbilist_unlock();
1facf9fc 3237+}
3238+
3239+int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
3240+{
3241+ int err;
1facf9fc 3242+ aufs_bindex_t bend, add_bindex;
3243+ struct dentry *root, *h_dentry;
3244+ struct inode *root_inode;
3245+ struct au_branch *add_branch;
3246+
3247+ root = sb->s_root;
5527c038 3248+ root_inode = d_inode(root);
1facf9fc 3249+ IMustLock(root_inode);
3250+ err = test_add(sb, add, remount);
3251+ if (unlikely(err < 0))
3252+ goto out;
3253+ if (err) {
3254+ err = 0;
3255+ goto out; /* success */
3256+ }
3257+
3258+ bend = au_sbend(sb);
3259+ add_branch = au_br_alloc(sb, bend + 2, add->perm);
3260+ err = PTR_ERR(add_branch);
3261+ if (IS_ERR(add_branch))
3262+ goto out;
3263+
3264+ err = au_br_init(add_branch, sb, add);
3265+ if (unlikely(err)) {
3266+ au_br_do_free(add_branch);
3267+ goto out;
3268+ }
3269+
3270+ add_bindex = add->bindex;
1facf9fc 3271+ if (!remount)
86dc4139 3272+ au_br_do_add(sb, add_branch, add_bindex);
1facf9fc 3273+ else {
3274+ sysaufs_brs_del(sb, add_bindex);
86dc4139 3275+ au_br_do_add(sb, add_branch, add_bindex);
1facf9fc 3276+ sysaufs_brs_add(sb, add_bindex);
3277+ }
3278+
86dc4139 3279+ h_dentry = add->path.dentry;
1308ab2a 3280+ if (!add_bindex) {
1facf9fc 3281+ au_cpup_attr_all(root_inode, /*force*/1);
1308ab2a 3282+ sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
3283+ } else
5527c038 3284+ au_add_nlink(root_inode, d_inode(h_dentry));
1facf9fc 3285+
3286+ /*
4a4d8108 3287+ * this test/set prevents aufs from handling unnecesary notify events
027c5e7a 3288+ * of xino files, in case of re-adding a writable branch which was
1facf9fc 3289+ * once detached from aufs.
3290+ */
3291+ if (au_xino_brid(sb) < 0
3292+ && au_br_writable(add_branch->br_perm)
3293+ && !au_test_fs_bad_xino(h_dentry->d_sb)
3294+ && add_branch->br_xino.xi_file
2000de60 3295+ && add_branch->br_xino.xi_file->f_path.dentry->d_parent == h_dentry)
1facf9fc 3296+ au_xino_brid_set(sb, add_branch->br_id);
3297+
4f0767ce 3298+out:
1facf9fc 3299+ return err;
3300+}
3301+
3302+/* ---------------------------------------------------------------------- */
3303+
076b876e
AM
3304+static unsigned long long au_farray_cb(void *a,
3305+ unsigned long long max __maybe_unused,
3306+ void *arg)
3307+{
3308+ unsigned long long n;
3309+ struct file **p, *f;
3310+ struct au_sphlhead *files;
3311+ struct au_finfo *finfo;
3312+ struct super_block *sb = arg;
3313+
3314+ n = 0;
3315+ p = a;
3316+ files = &au_sbi(sb)->si_files;
3317+ spin_lock(&files->spin);
3318+ hlist_for_each_entry(finfo, &files->head, fi_hlist) {
3319+ f = finfo->fi_file;
3320+ if (file_count(f)
3321+ && !special_file(file_inode(f)->i_mode)) {
3322+ get_file(f);
3323+ *p++ = f;
3324+ n++;
3325+ AuDebugOn(n > max);
3326+ }
3327+ }
3328+ spin_unlock(&files->spin);
3329+
3330+ return n;
3331+}
3332+
3333+static struct file **au_farray_alloc(struct super_block *sb,
3334+ unsigned long long *max)
3335+{
3336+ *max = atomic_long_read(&au_sbi(sb)->si_nfiles);
3337+ return au_array_alloc(max, au_farray_cb, sb);
3338+}
3339+
3340+static void au_farray_free(struct file **a, unsigned long long max)
3341+{
3342+ unsigned long long ull;
3343+
3344+ for (ull = 0; ull < max; ull++)
3345+ if (a[ull])
3346+ fput(a[ull]);
3347+ au_array_free(a);
3348+}
3349+
3350+/* ---------------------------------------------------------------------- */
3351+
1facf9fc 3352+/*
3353+ * delete a branch
3354+ */
3355+
3356+/* to show the line number, do not make it inlined function */
4a4d8108 3357+#define AuVerbose(do_info, fmt, ...) do { \
1facf9fc 3358+ if (do_info) \
4a4d8108 3359+ pr_info(fmt, ##__VA_ARGS__); \
1facf9fc 3360+} while (0)
3361+
027c5e7a
AM
3362+static int au_test_ibusy(struct inode *inode, aufs_bindex_t bstart,
3363+ aufs_bindex_t bend)
3364+{
3365+ return (inode && !S_ISDIR(inode->i_mode)) || bstart == bend;
3366+}
3367+
3368+static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t bstart,
3369+ aufs_bindex_t bend)
3370+{
5527c038 3371+ return au_test_ibusy(d_inode(dentry), bstart, bend);
027c5e7a
AM
3372+}
3373+
1facf9fc 3374+/*
3375+ * test if the branch is deletable or not.
3376+ */
3377+static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
b752ccd1 3378+ unsigned int sigen, const unsigned int verbose)
1facf9fc 3379+{
3380+ int err, i, j, ndentry;
3381+ aufs_bindex_t bstart, bend;
1facf9fc 3382+ struct au_dcsub_pages dpages;
3383+ struct au_dpage *dpage;
3384+ struct dentry *d;
1facf9fc 3385+
3386+ err = au_dpages_init(&dpages, GFP_NOFS);
3387+ if (unlikely(err))
3388+ goto out;
3389+ err = au_dcsub_pages(&dpages, root, NULL, NULL);
3390+ if (unlikely(err))
3391+ goto out_dpages;
3392+
1facf9fc 3393+ for (i = 0; !err && i < dpages.ndpage; i++) {
3394+ dpage = dpages.dpages + i;
3395+ ndentry = dpage->ndentry;
3396+ for (j = 0; !err && j < ndentry; j++) {
3397+ d = dpage->dentries[j];
c1595e42 3398+ AuDebugOn(au_dcount(d) <= 0);
027c5e7a 3399+ if (!au_digen_test(d, sigen)) {
1facf9fc 3400+ di_read_lock_child(d, AuLock_IR);
027c5e7a
AM
3401+ if (unlikely(au_dbrange_test(d))) {
3402+ di_read_unlock(d, AuLock_IR);
3403+ continue;
3404+ }
3405+ } else {
1facf9fc 3406+ di_write_lock_child(d);
027c5e7a
AM
3407+ if (unlikely(au_dbrange_test(d))) {
3408+ di_write_unlock(d);
3409+ continue;
3410+ }
1facf9fc 3411+ err = au_reval_dpath(d, sigen);
3412+ if (!err)
3413+ di_downgrade_lock(d, AuLock_IR);
3414+ else {
3415+ di_write_unlock(d);
3416+ break;
3417+ }
3418+ }
3419+
027c5e7a 3420+ /* AuDbgDentry(d); */
1facf9fc 3421+ bstart = au_dbstart(d);
3422+ bend = au_dbend(d);
3423+ if (bstart <= bindex
3424+ && bindex <= bend
3425+ && au_h_dptr(d, bindex)
027c5e7a 3426+ && au_test_dbusy(d, bstart, bend)) {
1facf9fc 3427+ err = -EBUSY;
523b37e3 3428+ AuVerbose(verbose, "busy %pd\n", d);
027c5e7a 3429+ AuDbgDentry(d);
1facf9fc 3430+ }
3431+ di_read_unlock(d, AuLock_IR);
3432+ }
3433+ }
3434+
4f0767ce 3435+out_dpages:
1facf9fc 3436+ au_dpages_free(&dpages);
4f0767ce 3437+out:
1facf9fc 3438+ return err;
3439+}
3440+
3441+static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
b752ccd1 3442+ unsigned int sigen, const unsigned int verbose)
1facf9fc 3443+{
3444+ int err;
7f207e10
AM
3445+ unsigned long long max, ull;
3446+ struct inode *i, **array;
1facf9fc 3447+ aufs_bindex_t bstart, bend;
1facf9fc 3448+
7f207e10
AM
3449+ array = au_iarray_alloc(sb, &max);
3450+ err = PTR_ERR(array);
3451+ if (IS_ERR(array))
3452+ goto out;
3453+
1facf9fc 3454+ err = 0;
7f207e10
AM
3455+ AuDbg("b%d\n", bindex);
3456+ for (ull = 0; !err && ull < max; ull++) {
3457+ i = array[ull];
076b876e
AM
3458+ if (unlikely(!i))
3459+ break;
7f207e10 3460+ if (i->i_ino == AUFS_ROOT_INO)
1facf9fc 3461+ continue;
3462+
7f207e10 3463+ /* AuDbgInode(i); */
537831f9 3464+ if (au_iigen(i, NULL) == sigen)
1facf9fc 3465+ ii_read_lock_child(i);
3466+ else {
3467+ ii_write_lock_child(i);
027c5e7a
AM
3468+ err = au_refresh_hinode_self(i);
3469+ au_iigen_dec(i);
1facf9fc 3470+ if (!err)
3471+ ii_downgrade_lock(i);
3472+ else {
3473+ ii_write_unlock(i);
3474+ break;
3475+ }
3476+ }
3477+
3478+ bstart = au_ibstart(i);
3479+ bend = au_ibend(i);
3480+ if (bstart <= bindex
3481+ && bindex <= bend
3482+ && au_h_iptr(i, bindex)
027c5e7a 3483+ && au_test_ibusy(i, bstart, bend)) {
1facf9fc 3484+ err = -EBUSY;
3485+ AuVerbose(verbose, "busy i%lu\n", i->i_ino);
7f207e10 3486+ AuDbgInode(i);
1facf9fc 3487+ }
3488+ ii_read_unlock(i);
3489+ }
7f207e10 3490+ au_iarray_free(array, max);
1facf9fc 3491+
7f207e10 3492+out:
1facf9fc 3493+ return err;
3494+}
3495+
b752ccd1
AM
3496+static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
3497+ const unsigned int verbose)
1facf9fc 3498+{
3499+ int err;
3500+ unsigned int sigen;
3501+
3502+ sigen = au_sigen(root->d_sb);
3503+ DiMustNoWaiters(root);
5527c038 3504+ IiMustNoWaiters(d_inode(root));
1facf9fc 3505+ di_write_unlock(root);
b752ccd1 3506+ err = test_dentry_busy(root, bindex, sigen, verbose);
1facf9fc 3507+ if (!err)
b752ccd1 3508+ err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
1facf9fc 3509+ di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
3510+
3511+ return err;
3512+}
3513+
076b876e
AM
3514+static int test_dir_busy(struct file *file, aufs_bindex_t br_id,
3515+ struct file **to_free, int *idx)
3516+{
3517+ int err;
c1595e42 3518+ unsigned char matched, root;
076b876e
AM
3519+ aufs_bindex_t bindex, bend;
3520+ struct au_fidir *fidir;
3521+ struct au_hfile *hfile;
3522+
3523+ err = 0;
2000de60 3524+ root = IS_ROOT(file->f_path.dentry);
c1595e42
JR
3525+ if (root) {
3526+ get_file(file);
3527+ to_free[*idx] = file;
3528+ (*idx)++;
3529+ goto out;
3530+ }
3531+
076b876e 3532+ matched = 0;
076b876e
AM
3533+ fidir = au_fi(file)->fi_hdir;
3534+ AuDebugOn(!fidir);
3535+ bend = au_fbend_dir(file);
3536+ for (bindex = au_fbstart(file); bindex <= bend; bindex++) {
3537+ hfile = fidir->fd_hfile + bindex;
3538+ if (!hfile->hf_file)
3539+ continue;
3540+
c1595e42 3541+ if (hfile->hf_br->br_id == br_id) {
076b876e 3542+ matched = 1;
076b876e 3543+ break;
c1595e42 3544+ }
076b876e 3545+ }
c1595e42 3546+ if (matched)
076b876e
AM
3547+ err = -EBUSY;
3548+
3549+out:
3550+ return err;
3551+}
3552+
3553+static int test_file_busy(struct super_block *sb, aufs_bindex_t br_id,
3554+ struct file **to_free, int opened)
3555+{
3556+ int err, idx;
3557+ unsigned long long ull, max;
3558+ aufs_bindex_t bstart;
3559+ struct file *file, **array;
076b876e
AM
3560+ struct dentry *root;
3561+ struct au_hfile *hfile;
3562+
3563+ array = au_farray_alloc(sb, &max);
3564+ err = PTR_ERR(array);
3565+ if (IS_ERR(array))
3566+ goto out;
3567+
3568+ err = 0;
3569+ idx = 0;
3570+ root = sb->s_root;
3571+ di_write_unlock(root);
3572+ for (ull = 0; ull < max; ull++) {
3573+ file = array[ull];
3574+ if (unlikely(!file))
3575+ break;
3576+
3577+ /* AuDbg("%pD\n", file); */
3578+ fi_read_lock(file);
3579+ bstart = au_fbstart(file);
2000de60 3580+ if (!d_is_dir(file->f_path.dentry)) {
076b876e
AM
3581+ hfile = &au_fi(file)->fi_htop;
3582+ if (hfile->hf_br->br_id == br_id)
3583+ err = -EBUSY;
3584+ } else
3585+ err = test_dir_busy(file, br_id, to_free, &idx);
3586+ fi_read_unlock(file);
3587+ if (unlikely(err))
3588+ break;
3589+ }
3590+ di_write_lock_child(root);
3591+ au_farray_free(array, max);
3592+ AuDebugOn(idx > opened);
3593+
3594+out:
3595+ return err;
3596+}
3597+
3598+static void br_del_file(struct file **to_free, unsigned long long opened,
3599+ aufs_bindex_t br_id)
3600+{
3601+ unsigned long long ull;
3602+ aufs_bindex_t bindex, bstart, bend, bfound;
3603+ struct file *file;
3604+ struct au_fidir *fidir;
3605+ struct au_hfile *hfile;
3606+
3607+ for (ull = 0; ull < opened; ull++) {
3608+ file = to_free[ull];
3609+ if (unlikely(!file))
3610+ break;
3611+
3612+ /* AuDbg("%pD\n", file); */
2000de60 3613+ AuDebugOn(!d_is_dir(file->f_path.dentry));
076b876e
AM
3614+ bfound = -1;
3615+ fidir = au_fi(file)->fi_hdir;
3616+ AuDebugOn(!fidir);
3617+ fi_write_lock(file);
3618+ bstart = au_fbstart(file);
3619+ bend = au_fbend_dir(file);
3620+ for (bindex = bstart; bindex <= bend; bindex++) {
3621+ hfile = fidir->fd_hfile + bindex;
3622+ if (!hfile->hf_file)
3623+ continue;
3624+
3625+ if (hfile->hf_br->br_id == br_id) {
3626+ bfound = bindex;
3627+ break;
3628+ }
3629+ }
3630+ AuDebugOn(bfound < 0);
3631+ au_set_h_fptr(file, bfound, NULL);
3632+ if (bfound == bstart) {
3633+ for (bstart++; bstart <= bend; bstart++)
3634+ if (au_hf_dir(file, bstart)) {
3635+ au_set_fbstart(file, bstart);
3636+ break;
3637+ }
3638+ }
3639+ fi_write_unlock(file);
3640+ }
3641+}
3642+
1facf9fc 3643+static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
3644+ const aufs_bindex_t bindex,
3645+ const aufs_bindex_t bend)
3646+{
3647+ struct au_branch **brp, **p;
3648+
dece6358
AM
3649+ AuRwMustWriteLock(&sbinfo->si_rwsem);
3650+
1facf9fc 3651+ brp = sbinfo->si_branch + bindex;
3652+ if (bindex < bend)
3653+ memmove(brp, brp + 1, sizeof(*brp) * (bend - bindex));
3654+ sbinfo->si_branch[0 + bend] = NULL;
3655+ sbinfo->si_bend--;
3656+
53392da6 3657+ p = krealloc(sbinfo->si_branch, sizeof(*p) * bend, AuGFP_SBILIST);
1facf9fc 3658+ if (p)
3659+ sbinfo->si_branch = p;
4a4d8108 3660+ /* harmless error */
1facf9fc 3661+}
3662+
3663+static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
3664+ const aufs_bindex_t bend)
3665+{
3666+ struct au_hdentry *hdp, *p;
3667+
1308ab2a 3668+ AuRwMustWriteLock(&dinfo->di_rwsem);
3669+
4a4d8108 3670+ hdp = dinfo->di_hdentry;
1facf9fc 3671+ if (bindex < bend)
4a4d8108
AM
3672+ memmove(hdp + bindex, hdp + bindex + 1,
3673+ sizeof(*hdp) * (bend - bindex));
3674+ hdp[0 + bend].hd_dentry = NULL;
1facf9fc 3675+ dinfo->di_bend--;
3676+
53392da6 3677+ p = krealloc(hdp, sizeof(*p) * bend, AuGFP_SBILIST);
1facf9fc 3678+ if (p)
3679+ dinfo->di_hdentry = p;
4a4d8108 3680+ /* harmless error */
1facf9fc 3681+}
3682+
3683+static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
3684+ const aufs_bindex_t bend)
3685+{
3686+ struct au_hinode *hip, *p;
3687+
1308ab2a 3688+ AuRwMustWriteLock(&iinfo->ii_rwsem);
3689+
1facf9fc 3690+ hip = iinfo->ii_hinode + bindex;
3691+ if (bindex < bend)
3692+ memmove(hip, hip + 1, sizeof(*hip) * (bend - bindex));
3693+ iinfo->ii_hinode[0 + bend].hi_inode = NULL;
4a4d8108 3694+ au_hn_init(iinfo->ii_hinode + bend);
1facf9fc 3695+ iinfo->ii_bend--;
3696+
53392da6 3697+ p = krealloc(iinfo->ii_hinode, sizeof(*p) * bend, AuGFP_SBILIST);
1facf9fc 3698+ if (p)
3699+ iinfo->ii_hinode = p;
4a4d8108 3700+ /* harmless error */
1facf9fc 3701+}
3702+
3703+static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
3704+ struct au_branch *br)
3705+{
3706+ aufs_bindex_t bend;
3707+ struct au_sbinfo *sbinfo;
53392da6
AM
3708+ struct dentry *root, *h_root;
3709+ struct inode *inode, *h_inode;
3710+ struct au_hinode *hinode;
1facf9fc 3711+
dece6358
AM
3712+ SiMustWriteLock(sb);
3713+
1facf9fc 3714+ root = sb->s_root;
5527c038 3715+ inode = d_inode(root);
1facf9fc 3716+ sbinfo = au_sbi(sb);
3717+ bend = sbinfo->si_bend;
3718+
53392da6
AM
3719+ h_root = au_h_dptr(root, bindex);
3720+ hinode = au_hi(inode, bindex);
3721+ h_inode = au_igrab(hinode->hi_inode);
3722+ au_hiput(hinode);
1facf9fc 3723+
53392da6 3724+ au_sbilist_lock();
1facf9fc 3725+ au_br_do_del_brp(sbinfo, bindex, bend);
3726+ au_br_do_del_hdp(au_di(root), bindex, bend);
3727+ au_br_do_del_hip(au_ii(inode), bindex, bend);
53392da6
AM
3728+ au_sbilist_unlock();
3729+
3730+ dput(h_root);
3731+ iput(h_inode);
3732+ au_br_do_free(br);
1facf9fc 3733+}
3734+
076b876e
AM
3735+static unsigned long long empty_cb(void *array, unsigned long long max,
3736+ void *arg)
3737+{
3738+ return max;
3739+}
3740+
1facf9fc 3741+int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
3742+{
3743+ int err, rerr, i;
076b876e 3744+ unsigned long long opened;
1facf9fc 3745+ unsigned int mnt_flags;
3746+ aufs_bindex_t bindex, bend, br_id;
3747+ unsigned char do_wh, verbose;
3748+ struct au_branch *br;
3749+ struct au_wbr *wbr;
076b876e
AM
3750+ struct dentry *root;
3751+ struct file **to_free;
1facf9fc 3752+
3753+ err = 0;
076b876e
AM
3754+ opened = 0;
3755+ to_free = NULL;
3756+ root = sb->s_root;
3757+ bindex = au_find_dbindex(root, del->h_path.dentry);
1facf9fc 3758+ if (bindex < 0) {
3759+ if (remount)
3760+ goto out; /* success */
3761+ err = -ENOENT;
4a4d8108 3762+ pr_err("%s no such branch\n", del->pathname);
1facf9fc 3763+ goto out;
3764+ }
3765+ AuDbg("bindex b%d\n", bindex);
3766+
3767+ err = -EBUSY;
3768+ mnt_flags = au_mntflags(sb);
3769+ verbose = !!au_opt_test(mnt_flags, VERBOSE);
3770+ bend = au_sbend(sb);
3771+ if (unlikely(!bend)) {
3772+ AuVerbose(verbose, "no more branches left\n");
3773+ goto out;
3774+ }
3775+ br = au_sbr(sb, bindex);
86dc4139 3776+ AuDebugOn(!path_equal(&br->br_path, &del->h_path));
076b876e
AM
3777+
3778+ br_id = br->br_id;
3779+ opened = atomic_read(&br->br_count);
3780+ if (unlikely(opened)) {
3781+ to_free = au_array_alloc(&opened, empty_cb, NULL);
3782+ err = PTR_ERR(to_free);
3783+ if (IS_ERR(to_free))
3784+ goto out;
3785+
3786+ err = test_file_busy(sb, br_id, to_free, opened);
3787+ if (unlikely(err)) {
3788+ AuVerbose(verbose, "%llu file(s) opened\n", opened);
3789+ goto out;
3790+ }
1facf9fc 3791+ }
3792+
3793+ wbr = br->br_wbr;
3794+ do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
3795+ if (do_wh) {
1308ab2a 3796+ /* instead of WbrWhMustWriteLock(wbr) */
3797+ SiMustWriteLock(sb);
1facf9fc 3798+ for (i = 0; i < AuBrWh_Last; i++) {
3799+ dput(wbr->wbr_wh[i]);
3800+ wbr->wbr_wh[i] = NULL;
3801+ }
3802+ }
3803+
076b876e 3804+ err = test_children_busy(root, bindex, verbose);
1facf9fc 3805+ if (unlikely(err)) {
3806+ if (do_wh)
3807+ goto out_wh;
3808+ goto out;
3809+ }
3810+
3811+ err = 0;
076b876e
AM
3812+ if (to_free) {
3813+ /*
3814+ * now we confirmed the branch is deletable.
3815+ * let's free the remaining opened dirs on the branch.
3816+ */
3817+ di_write_unlock(root);
3818+ br_del_file(to_free, opened, br_id);
3819+ di_write_lock_child(root);
3820+ }
3821+
1facf9fc 3822+ if (!remount)
3823+ au_br_do_del(sb, bindex, br);
3824+ else {
3825+ sysaufs_brs_del(sb, bindex);
3826+ au_br_do_del(sb, bindex, br);
3827+ sysaufs_brs_add(sb, bindex);
3828+ }
3829+
1308ab2a 3830+ if (!bindex) {
5527c038 3831+ au_cpup_attr_all(d_inode(root), /*force*/1);
1308ab2a 3832+ sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
3833+ } else
5527c038 3834+ au_sub_nlink(d_inode(root), d_inode(del->h_path.dentry));
1facf9fc 3835+ if (au_opt_test(mnt_flags, PLINK))
3836+ au_plink_half_refresh(sb, br_id);
3837+
b752ccd1 3838+ if (au_xino_brid(sb) == br_id)
1facf9fc 3839+ au_xino_brid_set(sb, -1);
3840+ goto out; /* success */
3841+
4f0767ce 3842+out_wh:
1facf9fc 3843+ /* revert */
86dc4139 3844+ rerr = au_br_init_wh(sb, br, br->br_perm);
1facf9fc 3845+ if (rerr)
0c3ec466
AM
3846+ pr_warn("failed re-creating base whiteout, %s. (%d)\n",
3847+ del->pathname, rerr);
4f0767ce 3848+out:
076b876e
AM
3849+ if (to_free)
3850+ au_farray_free(to_free, opened);
1facf9fc 3851+ return err;
3852+}
3853+
3854+/* ---------------------------------------------------------------------- */
3855+
027c5e7a
AM
3856+static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
3857+{
3858+ int err;
3859+ aufs_bindex_t bstart, bend;
3860+ struct aufs_ibusy ibusy;
3861+ struct inode *inode, *h_inode;
3862+
3863+ err = -EPERM;
3864+ if (unlikely(!capable(CAP_SYS_ADMIN)))
3865+ goto out;
3866+
3867+ err = copy_from_user(&ibusy, arg, sizeof(ibusy));
3868+ if (!err)
3869+ err = !access_ok(VERIFY_WRITE, &arg->h_ino, sizeof(arg->h_ino));
3870+ if (unlikely(err)) {
3871+ err = -EFAULT;
3872+ AuTraceErr(err);
3873+ goto out;
3874+ }
3875+
3876+ err = -EINVAL;
3877+ si_read_lock(sb, AuLock_FLUSH);
3878+ if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbend(sb)))
3879+ goto out_unlock;
3880+
3881+ err = 0;
3882+ ibusy.h_ino = 0; /* invalid */
3883+ inode = ilookup(sb, ibusy.ino);
3884+ if (!inode
3885+ || inode->i_ino == AUFS_ROOT_INO
3886+ || is_bad_inode(inode))
3887+ goto out_unlock;
3888+
3889+ ii_read_lock_child(inode);
3890+ bstart = au_ibstart(inode);
3891+ bend = au_ibend(inode);
3892+ if (bstart <= ibusy.bindex && ibusy.bindex <= bend) {
3893+ h_inode = au_h_iptr(inode, ibusy.bindex);
3894+ if (h_inode && au_test_ibusy(inode, bstart, bend))
3895+ ibusy.h_ino = h_inode->i_ino;
3896+ }
3897+ ii_read_unlock(inode);
3898+ iput(inode);
3899+
3900+out_unlock:
3901+ si_read_unlock(sb);
3902+ if (!err) {
3903+ err = __put_user(ibusy.h_ino, &arg->h_ino);
3904+ if (unlikely(err)) {
3905+ err = -EFAULT;
3906+ AuTraceErr(err);
3907+ }
3908+ }
3909+out:
3910+ return err;
3911+}
3912+
3913+long au_ibusy_ioctl(struct file *file, unsigned long arg)
3914+{
2000de60 3915+ return au_ibusy(file->f_path.dentry->d_sb, (void __user *)arg);
027c5e7a
AM
3916+}
3917+
3918+#ifdef CONFIG_COMPAT
3919+long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
3920+{
2000de60 3921+ return au_ibusy(file->f_path.dentry->d_sb, compat_ptr(arg));
027c5e7a
AM
3922+}
3923+#endif
3924+
3925+/* ---------------------------------------------------------------------- */
3926+
1facf9fc 3927+/*
3928+ * change a branch permission
3929+ */
3930+
dece6358
AM
3931+static void au_warn_ima(void)
3932+{
3933+#ifdef CONFIG_IMA
1308ab2a 3934+ /* since it doesn't support mark_files_ro() */
027c5e7a 3935+ AuWarn1("RW -> RO makes IMA to produce wrong message\n");
dece6358
AM
3936+#endif
3937+}
3938+
1facf9fc 3939+static int do_need_sigen_inc(int a, int b)
3940+{
3941+ return au_br_whable(a) && !au_br_whable(b);
3942+}
3943+
3944+static int need_sigen_inc(int old, int new)
3945+{
3946+ return do_need_sigen_inc(old, new)
3947+ || do_need_sigen_inc(new, old);
3948+}
3949+
3950+static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
3951+{
7f207e10 3952+ int err, do_warn;
027c5e7a 3953+ unsigned int mnt_flags;
7f207e10 3954+ unsigned long long ull, max;
e49829fe 3955+ aufs_bindex_t br_id;
38d290e6 3956+ unsigned char verbose, writer;
7f207e10 3957+ struct file *file, *hf, **array;
e49829fe 3958+ struct au_hfile *hfile;
1facf9fc 3959+
027c5e7a
AM
3960+ mnt_flags = au_mntflags(sb);
3961+ verbose = !!au_opt_test(mnt_flags, VERBOSE);
3962+
7f207e10
AM
3963+ array = au_farray_alloc(sb, &max);
3964+ err = PTR_ERR(array);
3965+ if (IS_ERR(array))
1facf9fc 3966+ goto out;
3967+
7f207e10 3968+ do_warn = 0;
e49829fe 3969+ br_id = au_sbr_id(sb, bindex);
7f207e10
AM
3970+ for (ull = 0; ull < max; ull++) {
3971+ file = array[ull];
076b876e
AM
3972+ if (unlikely(!file))
3973+ break;
1facf9fc 3974+
523b37e3 3975+ /* AuDbg("%pD\n", file); */
1facf9fc 3976+ fi_read_lock(file);
3977+ if (unlikely(au_test_mmapped(file))) {
3978+ err = -EBUSY;
523b37e3 3979+ AuVerbose(verbose, "mmapped %pD\n", file);
7f207e10 3980+ AuDbgFile(file);
1facf9fc 3981+ FiMustNoWaiters(file);
3982+ fi_read_unlock(file);
7f207e10 3983+ goto out_array;
1facf9fc 3984+ }
3985+
e49829fe
JR
3986+ hfile = &au_fi(file)->fi_htop;
3987+ hf = hfile->hf_file;
7e9cd9fe 3988+ if (!d_is_reg(file->f_path.dentry)
1facf9fc 3989+ || !(file->f_mode & FMODE_WRITE)
e49829fe 3990+ || hfile->hf_br->br_id != br_id
7f207e10
AM
3991+ || !(hf->f_mode & FMODE_WRITE))
3992+ array[ull] = NULL;
3993+ else {
3994+ do_warn = 1;
3995+ get_file(file);
1facf9fc 3996+ }
3997+
1facf9fc 3998+ FiMustNoWaiters(file);
3999+ fi_read_unlock(file);
7f207e10
AM
4000+ fput(file);
4001+ }
1facf9fc 4002+
4003+ err = 0;
7f207e10 4004+ if (do_warn)
dece6358 4005+ au_warn_ima();
7f207e10
AM
4006+
4007+ for (ull = 0; ull < max; ull++) {
4008+ file = array[ull];
4009+ if (!file)
4010+ continue;
4011+
1facf9fc 4012+ /* todo: already flushed? */
523b37e3
AM
4013+ /*
4014+ * fs/super.c:mark_files_ro() is gone, but aufs keeps its
4015+ * approach which resets f_mode and calls mnt_drop_write() and
4016+ * file_release_write() for each file, because the branch
4017+ * attribute in aufs world is totally different from the native
4018+ * fs rw/ro mode.
4019+ */
7f207e10
AM
4020+ /* fi_read_lock(file); */
4021+ hfile = &au_fi(file)->fi_htop;
4022+ hf = hfile->hf_file;
4023+ /* fi_read_unlock(file); */
027c5e7a 4024+ spin_lock(&hf->f_lock);
38d290e6
JR
4025+ writer = !!(hf->f_mode & FMODE_WRITER);
4026+ hf->f_mode &= ~(FMODE_WRITE | FMODE_WRITER);
027c5e7a 4027+ spin_unlock(&hf->f_lock);
38d290e6
JR
4028+ if (writer) {
4029+ put_write_access(file_inode(hf));
c06a8ce3 4030+ __mnt_drop_write(hf->f_path.mnt);
1facf9fc 4031+ }
4032+ }
4033+
7f207e10
AM
4034+out_array:
4035+ au_farray_free(array, max);
4f0767ce 4036+out:
7f207e10 4037+ AuTraceErr(err);
1facf9fc 4038+ return err;
4039+}
4040+
4041+int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
7f207e10 4042+ int *do_refresh)
1facf9fc 4043+{
4044+ int err, rerr;
4045+ aufs_bindex_t bindex;
4046+ struct dentry *root;
4047+ struct au_branch *br;
076b876e 4048+ struct au_br_fhsm *bf;
1facf9fc 4049+
4050+ root = sb->s_root;
1facf9fc 4051+ bindex = au_find_dbindex(root, mod->h_root);
4052+ if (bindex < 0) {
4053+ if (remount)
4054+ return 0; /* success */
4055+ err = -ENOENT;
4a4d8108 4056+ pr_err("%s no such branch\n", mod->path);
1facf9fc 4057+ goto out;
4058+ }
4059+ AuDbg("bindex b%d\n", bindex);
4060+
5527c038 4061+ err = test_br(d_inode(mod->h_root), mod->perm, mod->path);
1facf9fc 4062+ if (unlikely(err))
4063+ goto out;
4064+
4065+ br = au_sbr(sb, bindex);
86dc4139 4066+ AuDebugOn(mod->h_root != au_br_dentry(br));
1facf9fc 4067+ if (br->br_perm == mod->perm)
4068+ return 0; /* success */
4069+
076b876e
AM
4070+ /* pre-allocate for non-fhsm --> fhsm */
4071+ bf = NULL;
4072+ if (!au_br_fhsm(br->br_perm) && au_br_fhsm(mod->perm)) {
4073+ err = au_fhsm_br_alloc(br);
4074+ if (unlikely(err))
4075+ goto out;
4076+ bf = br->br_fhsm;
4077+ br->br_fhsm = NULL;
4078+ }
4079+
1facf9fc 4080+ if (au_br_writable(br->br_perm)) {
4081+ /* remove whiteout base */
86dc4139 4082+ err = au_br_init_wh(sb, br, mod->perm);
1facf9fc 4083+ if (unlikely(err))
076b876e 4084+ goto out_bf;
1facf9fc 4085+
4086+ if (!au_br_writable(mod->perm)) {
4087+ /* rw --> ro, file might be mmapped */
4088+ DiMustNoWaiters(root);
5527c038 4089+ IiMustNoWaiters(d_inode(root));
1facf9fc 4090+ di_write_unlock(root);
4091+ err = au_br_mod_files_ro(sb, bindex);
4092+ /* aufs_write_lock() calls ..._child() */
4093+ di_write_lock_child(root);
4094+
4095+ if (unlikely(err)) {
4096+ rerr = -ENOMEM;
4097+ br->br_wbr = kmalloc(sizeof(*br->br_wbr),
4098+ GFP_NOFS);
86dc4139
AM
4099+ if (br->br_wbr)
4100+ rerr = au_wbr_init(br, sb, br->br_perm);
1facf9fc 4101+ if (unlikely(rerr)) {
4102+ AuIOErr("nested error %d (%d)\n",
4103+ rerr, err);
4104+ br->br_perm = mod->perm;
4105+ }
4106+ }
4107+ }
4108+ } else if (au_br_writable(mod->perm)) {
4109+ /* ro --> rw */
4110+ err = -ENOMEM;
4111+ br->br_wbr = kmalloc(sizeof(*br->br_wbr), GFP_NOFS);
4112+ if (br->br_wbr) {
86dc4139 4113+ err = au_wbr_init(br, sb, mod->perm);
1facf9fc 4114+ if (unlikely(err)) {
4115+ kfree(br->br_wbr);
4116+ br->br_wbr = NULL;
4117+ }
4118+ }
4119+ }
076b876e
AM
4120+ if (unlikely(err))
4121+ goto out_bf;
4122+
4123+ if (au_br_fhsm(br->br_perm)) {
4124+ if (!au_br_fhsm(mod->perm)) {
4125+ /* fhsm --> non-fhsm */
4126+ au_br_fhsm_fin(br->br_fhsm);
4127+ kfree(br->br_fhsm);
4128+ br->br_fhsm = NULL;
4129+ }
4130+ } else if (au_br_fhsm(mod->perm))
4131+ /* non-fhsm --> fhsm */
4132+ br->br_fhsm = bf;
4133+
076b876e
AM
4134+ *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
4135+ br->br_perm = mod->perm;
4136+ goto out; /* success */
1facf9fc 4137+
076b876e
AM
4138+out_bf:
4139+ kfree(bf);
4140+out:
4141+ AuTraceErr(err);
4142+ return err;
4143+}
4144+
4145+/* ---------------------------------------------------------------------- */
4146+
4147+int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs)
4148+{
4149+ int err;
4150+ struct kstatfs kstfs;
4151+
4152+ err = vfs_statfs(&br->br_path, &kstfs);
1facf9fc 4153+ if (!err) {
076b876e
AM
4154+ stfs->f_blocks = kstfs.f_blocks;
4155+ stfs->f_bavail = kstfs.f_bavail;
4156+ stfs->f_files = kstfs.f_files;
4157+ stfs->f_ffree = kstfs.f_ffree;
1facf9fc 4158+ }
4159+
1facf9fc 4160+ return err;
4161+}
7f207e10
AM
4162diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
4163--- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 4164+++ linux/fs/aufs/branch.h 2015-09-24 10:47:58.248052907 +0200
b912730e 4165@@ -0,0 +1,279 @@
1facf9fc 4166+/*
2000de60 4167+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 4168+ *
4169+ * This program, aufs is free software; you can redistribute it and/or modify
4170+ * it under the terms of the GNU General Public License as published by
4171+ * the Free Software Foundation; either version 2 of the License, or
4172+ * (at your option) any later version.
dece6358
AM
4173+ *
4174+ * This program is distributed in the hope that it will be useful,
4175+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
4176+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4177+ * GNU General Public License for more details.
4178+ *
4179+ * You should have received a copy of the GNU General Public License
523b37e3 4180+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 4181+ */
4182+
4183+/*
4184+ * branch filesystems and xino for them
4185+ */
4186+
4187+#ifndef __AUFS_BRANCH_H__
4188+#define __AUFS_BRANCH_H__
4189+
4190+#ifdef __KERNEL__
4191+
1facf9fc 4192+#include <linux/mount.h>
4a4d8108 4193+#include "dynop.h"
1facf9fc 4194+#include "rwsem.h"
4195+#include "super.h"
4196+
4197+/* ---------------------------------------------------------------------- */
4198+
4199+/* a xino file */
4200+struct au_xino_file {
4201+ struct file *xi_file;
4202+ struct mutex xi_nondir_mtx;
4203+
4204+ /* todo: make xino files an array to support huge inode number */
4205+
4206+#ifdef CONFIG_DEBUG_FS
4207+ struct dentry *xi_dbgaufs;
4208+#endif
4209+};
4210+
076b876e
AM
4211+/* File-based Hierarchical Storage Management */
4212+struct au_br_fhsm {
4213+#ifdef CONFIG_AUFS_FHSM
4214+ struct mutex bf_lock;
4215+ unsigned long bf_jiffy;
4216+ struct aufs_stfs bf_stfs;
4217+ int bf_readable;
4218+#endif
4219+};
4220+
1facf9fc 4221+/* members for writable branch only */
4222+enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
4223+struct au_wbr {
dece6358 4224+ struct au_rwsem wbr_wh_rwsem;
1facf9fc 4225+ struct dentry *wbr_wh[AuBrWh_Last];
4a4d8108 4226+ atomic_t wbr_wh_running;
1facf9fc 4227+#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
4228+#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
4229+#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
4230+
4231+ /* mfs mode */
4232+ unsigned long long wbr_bytes;
4233+};
4234+
4a4d8108
AM
4235+/* ext2 has 3 types of operations at least, ext3 has 4 */
4236+#define AuBrDynOp (AuDyLast * 4)
4237+
1716fcea
AM
4238+#ifdef CONFIG_AUFS_HFSNOTIFY
4239+/* support for asynchronous destruction */
4240+struct au_br_hfsnotify {
4241+ struct fsnotify_group *hfsn_group;
4242+};
4243+#endif
4244+
392086de
AM
4245+/* sysfs entries */
4246+struct au_brsysfs {
4247+ char name[16];
4248+ struct attribute attr;
4249+};
4250+
4251+enum {
4252+ AuBrSysfs_BR,
4253+ AuBrSysfs_BRID,
4254+ AuBrSysfs_Last
4255+};
4256+
1facf9fc 4257+/* protected by superblock rwsem */
4258+struct au_branch {
4259+ struct au_xino_file br_xino;
4260+
4261+ aufs_bindex_t br_id;
4262+
4263+ int br_perm;
86dc4139 4264+ struct path br_path;
4a4d8108
AM
4265+ spinlock_t br_dykey_lock;
4266+ struct au_dykey *br_dykey[AuBrDynOp];
1facf9fc 4267+ atomic_t br_count;
4268+
4269+ struct au_wbr *br_wbr;
076b876e 4270+ struct au_br_fhsm *br_fhsm;
1facf9fc 4271+
4272+ /* xino truncation */
1facf9fc 4273+ atomic_t br_xino_running;
4274+
027c5e7a 4275+#ifdef CONFIG_AUFS_HFSNOTIFY
1716fcea 4276+ struct au_br_hfsnotify *br_hfsn;
027c5e7a
AM
4277+#endif
4278+
1facf9fc 4279+#ifdef CONFIG_SYSFS
392086de
AM
4280+ /* entries under sysfs per mount-point */
4281+ struct au_brsysfs br_sysfs[AuBrSysfs_Last];
1facf9fc 4282+#endif
4283+};
4284+
4285+/* ---------------------------------------------------------------------- */
4286+
86dc4139
AM
4287+static inline struct vfsmount *au_br_mnt(struct au_branch *br)
4288+{
4289+ return br->br_path.mnt;
4290+}
4291+
4292+static inline struct dentry *au_br_dentry(struct au_branch *br)
4293+{
4294+ return br->br_path.dentry;
4295+}
4296+
4297+static inline struct super_block *au_br_sb(struct au_branch *br)
4298+{
4299+ return au_br_mnt(br)->mnt_sb;
4300+}
4301+
1facf9fc 4302+static inline int au_br_rdonly(struct au_branch *br)
4303+{
86dc4139 4304+ return ((au_br_sb(br)->s_flags & MS_RDONLY)
1facf9fc 4305+ || !au_br_writable(br->br_perm))
4306+ ? -EROFS : 0;
4307+}
4308+
4a4d8108 4309+static inline int au_br_hnotifyable(int brperm __maybe_unused)
1facf9fc 4310+{
4a4d8108 4311+#ifdef CONFIG_AUFS_HNOTIFY
1e00d052 4312+ return !(brperm & AuBrPerm_RR);
1facf9fc 4313+#else
4314+ return 0;
4315+#endif
4316+}
4317+
b912730e
AM
4318+static inline int au_br_test_oflag(int oflag, struct au_branch *br)
4319+{
4320+ int err, exec_flag;
4321+
4322+ err = 0;
4323+ exec_flag = oflag & __FMODE_EXEC;
4324+ if (unlikely(exec_flag && (au_br_mnt(br)->mnt_flags & MNT_NOEXEC)))
4325+ err = -EACCES;
4326+
4327+ return err;
4328+}
4329+
1facf9fc 4330+/* ---------------------------------------------------------------------- */
4331+
4332+/* branch.c */
4333+struct au_sbinfo;
4334+void au_br_free(struct au_sbinfo *sinfo);
4335+int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
4336+struct au_opt_add;
4337+int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
4338+struct au_opt_del;
4339+int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
027c5e7a
AM
4340+long au_ibusy_ioctl(struct file *file, unsigned long arg);
4341+#ifdef CONFIG_COMPAT
4342+long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
4343+#endif
1facf9fc 4344+struct au_opt_mod;
4345+int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
7f207e10 4346+ int *do_refresh);
076b876e
AM
4347+struct aufs_stfs;
4348+int au_br_stfs(struct au_branch *br, struct aufs_stfs *stfs);
1facf9fc 4349+
4350+/* xino.c */
4351+static const loff_t au_loff_max = LLONG_MAX;
4352+
4353+int au_xib_trunc(struct super_block *sb);
5527c038 4354+ssize_t xino_fread(vfs_readf_t func, struct file *file, void *buf, size_t size,
1facf9fc 4355+ loff_t *pos);
5527c038
JR
4356+ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
4357+ size_t size, loff_t *pos);
1facf9fc 4358+struct file *au_xino_create2(struct file *base_file, struct file *copy_src);
4359+struct file *au_xino_create(struct super_block *sb, char *fname, int silent);
4360+ino_t au_xino_new_ino(struct super_block *sb);
b752ccd1 4361+void au_xino_delete_inode(struct inode *inode, const int unlinked);
1facf9fc 4362+int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4363+ ino_t ino);
4364+int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
4365+ ino_t *ino);
4366+int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t hino,
4367+ struct file *base_file, int do_test);
4368+int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex);
4369+
4370+struct au_opt_xino;
4371+int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount);
4372+void au_xino_clr(struct super_block *sb);
4373+struct file *au_xino_def(struct super_block *sb);
4374+int au_xino_path(struct seq_file *seq, struct file *file);
4375+
4376+/* ---------------------------------------------------------------------- */
4377+
4378+/* Superblock to branch */
4379+static inline
4380+aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
4381+{
4382+ return au_sbr(sb, bindex)->br_id;
4383+}
4384+
4385+static inline
4386+struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
4387+{
86dc4139 4388+ return au_br_mnt(au_sbr(sb, bindex));
1facf9fc 4389+}
4390+
4391+static inline
4392+struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
4393+{
86dc4139 4394+ return au_br_sb(au_sbr(sb, bindex));
1facf9fc 4395+}
4396+
4397+static inline void au_sbr_put(struct super_block *sb, aufs_bindex_t bindex)
4398+{
e49829fe 4399+ atomic_dec(&au_sbr(sb, bindex)->br_count);
1facf9fc 4400+}
4401+
4402+static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
4403+{
4404+ return au_sbr(sb, bindex)->br_perm;
4405+}
4406+
4407+static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
4408+{
4409+ return au_br_whable(au_sbr_perm(sb, bindex));
4410+}
4411+
4412+/* ---------------------------------------------------------------------- */
4413+
4414+/*
4415+ * wbr_wh_read_lock, wbr_wh_write_lock
4416+ * wbr_wh_read_unlock, wbr_wh_write_unlock, wbr_wh_downgrade_lock
4417+ */
4418+AuSimpleRwsemFuncs(wbr_wh, struct au_wbr *wbr, &wbr->wbr_wh_rwsem);
4419+
dece6358
AM
4420+#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&wbr->wbr_wh_rwsem)
4421+#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&wbr->wbr_wh_rwsem)
4422+#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&wbr->wbr_wh_rwsem)
4423+
076b876e
AM
4424+/* ---------------------------------------------------------------------- */
4425+
4426+#ifdef CONFIG_AUFS_FHSM
4427+static inline void au_br_fhsm_init(struct au_br_fhsm *brfhsm)
4428+{
4429+ mutex_init(&brfhsm->bf_lock);
4430+ brfhsm->bf_jiffy = 0;
4431+ brfhsm->bf_readable = 0;
4432+}
4433+
4434+static inline void au_br_fhsm_fin(struct au_br_fhsm *brfhsm)
4435+{
4436+ mutex_destroy(&brfhsm->bf_lock);
4437+}
4438+#else
4439+AuStubVoid(au_br_fhsm_init, struct au_br_fhsm *brfhsm)
4440+AuStubVoid(au_br_fhsm_fin, struct au_br_fhsm *brfhsm)
4441+#endif
4442+
1facf9fc 4443+#endif /* __KERNEL__ */
4444+#endif /* __AUFS_BRANCH_H__ */
7f207e10
AM
4445diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
4446--- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 4447+++ linux/fs/aufs/conf.mk 2015-09-24 10:47:58.248052907 +0200
c1595e42 4448@@ -0,0 +1,38 @@
4a4d8108
AM
4449+
4450+AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
4451+
4452+define AuConf
4453+ifdef ${1}
4454+AuConfStr += ${1}=${${1}}
4455+endif
4456+endef
4457+
b752ccd1 4458+AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
e49829fe 4459+ SBILIST \
7f207e10 4460+ HNOTIFY HFSNOTIFY \
4a4d8108 4461+ EXPORT INO_T_64 \
c1595e42 4462+ XATTR \
076b876e 4463+ FHSM \
4a4d8108 4464+ RDU \
4a4d8108
AM
4465+ SHWH \
4466+ BR_RAMFS \
4467+ BR_FUSE POLL \
4468+ BR_HFSPLUS \
4469+ BDEV_LOOP \
b752ccd1
AM
4470+ DEBUG MAGIC_SYSRQ
4471+$(foreach i, ${AuConfAll}, \
4a4d8108
AM
4472+ $(eval $(call AuConf,CONFIG_AUFS_${i})))
4473+
4474+AuConfName = ${obj}/conf.str
4475+${AuConfName}.tmp: FORCE
4476+ @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
4477+${AuConfName}: ${AuConfName}.tmp
4478+ @diff -q $< $@ > /dev/null 2>&1 || { \
4479+ echo ' GEN ' $@; \
4480+ cp -p $< $@; \
4481+ }
4482+FORCE:
4483+clean-files += ${AuConfName} ${AuConfName}.tmp
4484+${obj}/sysfs.o: ${AuConfName}
b752ccd1
AM
4485+
4486+-include ${srctree}/${src}/conf_priv.mk
7f207e10
AM
4487diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
4488--- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 4489+++ linux/fs/aufs/cpup.c 2015-09-24 10:47:58.248052907 +0200
5527c038 4490@@ -0,0 +1,1319 @@
1facf9fc 4491+/*
2000de60 4492+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 4493+ *
4494+ * This program, aufs is free software; you can redistribute it and/or modify
4495+ * it under the terms of the GNU General Public License as published by
4496+ * the Free Software Foundation; either version 2 of the License, or
4497+ * (at your option) any later version.
dece6358
AM
4498+ *
4499+ * This program is distributed in the hope that it will be useful,
4500+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
4501+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4502+ * GNU General Public License for more details.
4503+ *
4504+ * You should have received a copy of the GNU General Public License
523b37e3 4505+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 4506+ */
4507+
4508+/*
4509+ * copy-up functions, see wbr_policy.c for copy-down
4510+ */
4511+
4512+#include <linux/fs_stack.h>
dece6358 4513+#include <linux/mm.h>
1facf9fc 4514+#include "aufs.h"
4515+
86dc4139 4516+void au_cpup_attr_flags(struct inode *dst, unsigned int iflags)
1facf9fc 4517+{
4518+ const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
367653fa 4519+ | S_NOATIME | S_NOCMTIME | S_AUTOMOUNT;
1facf9fc 4520+
86dc4139
AM
4521+ BUILD_BUG_ON(sizeof(iflags) != sizeof(dst->i_flags));
4522+
4523+ dst->i_flags |= iflags & ~mask;
1facf9fc 4524+ if (au_test_fs_notime(dst->i_sb))
4525+ dst->i_flags |= S_NOATIME | S_NOCMTIME;
4526+}
4527+
4528+void au_cpup_attr_timesizes(struct inode *inode)
4529+{
4530+ struct inode *h_inode;
4531+
4532+ h_inode = au_h_iptr(inode, au_ibstart(inode));
4533+ fsstack_copy_attr_times(inode, h_inode);
4a4d8108 4534+ fsstack_copy_inode_size(inode, h_inode);
1facf9fc 4535+}
4536+
4537+void au_cpup_attr_nlink(struct inode *inode, int force)
4538+{
4539+ struct inode *h_inode;
4540+ struct super_block *sb;
4541+ aufs_bindex_t bindex, bend;
4542+
4543+ sb = inode->i_sb;
4544+ bindex = au_ibstart(inode);
4545+ h_inode = au_h_iptr(inode, bindex);
4546+ if (!force
4547+ && !S_ISDIR(h_inode->i_mode)
4548+ && au_opt_test(au_mntflags(sb), PLINK)
4549+ && au_plink_test(inode))
4550+ return;
4551+
7eafdf33
AM
4552+ /*
4553+ * 0 can happen in revalidating.
38d290e6
JR
4554+ * h_inode->i_mutex may not be held here, but it is harmless since once
4555+ * i_nlink reaches 0, it will never become positive except O_TMPFILE
4556+ * case.
4557+ * todo: O_TMPFILE+linkat(AT_SYMLINK_FOLLOW) bypassing aufs may cause
4558+ * the incorrect link count.
7eafdf33 4559+ */
92d182d2 4560+ set_nlink(inode, h_inode->i_nlink);
1facf9fc 4561+
4562+ /*
4563+ * fewer nlink makes find(1) noisy, but larger nlink doesn't.
4564+ * it may includes whplink directory.
4565+ */
4566+ if (S_ISDIR(h_inode->i_mode)) {
4567+ bend = au_ibend(inode);
4568+ for (bindex++; bindex <= bend; bindex++) {
4569+ h_inode = au_h_iptr(inode, bindex);
4570+ if (h_inode)
4571+ au_add_nlink(inode, h_inode);
4572+ }
4573+ }
4574+}
4575+
4576+void au_cpup_attr_changeable(struct inode *inode)
4577+{
4578+ struct inode *h_inode;
4579+
4580+ h_inode = au_h_iptr(inode, au_ibstart(inode));
4581+ inode->i_mode = h_inode->i_mode;
4582+ inode->i_uid = h_inode->i_uid;
4583+ inode->i_gid = h_inode->i_gid;
4584+ au_cpup_attr_timesizes(inode);
86dc4139 4585+ au_cpup_attr_flags(inode, h_inode->i_flags);
1facf9fc 4586+}
4587+
4588+void au_cpup_igen(struct inode *inode, struct inode *h_inode)
4589+{
4590+ struct au_iinfo *iinfo = au_ii(inode);
4591+
1308ab2a 4592+ IiMustWriteLock(inode);
4593+
1facf9fc 4594+ iinfo->ii_higen = h_inode->i_generation;
4595+ iinfo->ii_hsb1 = h_inode->i_sb;
4596+}
4597+
4598+void au_cpup_attr_all(struct inode *inode, int force)
4599+{
4600+ struct inode *h_inode;
4601+
4602+ h_inode = au_h_iptr(inode, au_ibstart(inode));
4603+ au_cpup_attr_changeable(inode);
4604+ if (inode->i_nlink > 0)
4605+ au_cpup_attr_nlink(inode, force);
4606+ inode->i_rdev = h_inode->i_rdev;
4607+ inode->i_blkbits = h_inode->i_blkbits;
4608+ au_cpup_igen(inode, h_inode);
4609+}
4610+
4611+/* ---------------------------------------------------------------------- */
4612+
4613+/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
4614+
4615+/* keep the timestamps of the parent dir when cpup */
4616+void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4617+ struct path *h_path)
4618+{
4619+ struct inode *h_inode;
4620+
4621+ dt->dt_dentry = dentry;
4622+ dt->dt_h_path = *h_path;
5527c038 4623+ h_inode = d_inode(h_path->dentry);
1facf9fc 4624+ dt->dt_atime = h_inode->i_atime;
4625+ dt->dt_mtime = h_inode->i_mtime;
4626+ /* smp_mb(); */
4627+}
4628+
4629+void au_dtime_revert(struct au_dtime *dt)
4630+{
4631+ struct iattr attr;
4632+ int err;
4633+
4634+ attr.ia_atime = dt->dt_atime;
4635+ attr.ia_mtime = dt->dt_mtime;
4636+ attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
4637+ | ATTR_ATIME | ATTR_ATIME_SET;
4638+
523b37e3
AM
4639+ /* no delegation since this is a directory */
4640+ err = vfsub_notify_change(&dt->dt_h_path, &attr, /*delegated*/NULL);
1facf9fc 4641+ if (unlikely(err))
0c3ec466 4642+ pr_warn("restoring timestamps failed(%d). ignored\n", err);
1facf9fc 4643+}
4644+
4645+/* ---------------------------------------------------------------------- */
4646+
86dc4139
AM
4647+/* internal use only */
4648+struct au_cpup_reg_attr {
4649+ int valid;
4650+ struct kstat st;
4651+ unsigned int iflags; /* inode->i_flags */
4652+};
4653+
1facf9fc 4654+static noinline_for_stack
86dc4139
AM
4655+int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct dentry *h_src,
4656+ struct au_cpup_reg_attr *h_src_attr)
1facf9fc 4657+{
c1595e42 4658+ int err, sbits, icex;
7e9cd9fe
AM
4659+ unsigned int mnt_flags;
4660+ unsigned char verbose;
1facf9fc 4661+ struct iattr ia;
4662+ struct path h_path;
1308ab2a 4663+ struct inode *h_isrc, *h_idst;
86dc4139 4664+ struct kstat *h_st;
c1595e42 4665+ struct au_branch *br;
1facf9fc 4666+
4667+ h_path.dentry = au_h_dptr(dst, bindex);
5527c038 4668+ h_idst = d_inode(h_path.dentry);
c1595e42
JR
4669+ br = au_sbr(dst->d_sb, bindex);
4670+ h_path.mnt = au_br_mnt(br);
5527c038 4671+ h_isrc = d_inode(h_src);
1308ab2a 4672+ ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
1facf9fc 4673+ | ATTR_ATIME | ATTR_MTIME
4674+ | ATTR_ATIME_SET | ATTR_MTIME_SET;
86dc4139
AM
4675+ if (h_src_attr && h_src_attr->valid) {
4676+ h_st = &h_src_attr->st;
4677+ ia.ia_uid = h_st->uid;
4678+ ia.ia_gid = h_st->gid;
4679+ ia.ia_atime = h_st->atime;
4680+ ia.ia_mtime = h_st->mtime;
4681+ if (h_idst->i_mode != h_st->mode
4682+ && !S_ISLNK(h_idst->i_mode)) {
4683+ ia.ia_valid |= ATTR_MODE;
4684+ ia.ia_mode = h_st->mode;
4685+ }
4686+ sbits = !!(h_st->mode & (S_ISUID | S_ISGID));
4687+ au_cpup_attr_flags(h_idst, h_src_attr->iflags);
4688+ } else {
4689+ ia.ia_uid = h_isrc->i_uid;
4690+ ia.ia_gid = h_isrc->i_gid;
4691+ ia.ia_atime = h_isrc->i_atime;
4692+ ia.ia_mtime = h_isrc->i_mtime;
4693+ if (h_idst->i_mode != h_isrc->i_mode
4694+ && !S_ISLNK(h_idst->i_mode)) {
4695+ ia.ia_valid |= ATTR_MODE;
4696+ ia.ia_mode = h_isrc->i_mode;
4697+ }
4698+ sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
4699+ au_cpup_attr_flags(h_idst, h_isrc->i_flags);
1308ab2a 4700+ }
523b37e3
AM
4701+ /* no delegation since it is just created */
4702+ err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
1facf9fc 4703+
4704+ /* is this nfs only? */
4705+ if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
4706+ ia.ia_valid = ATTR_FORCE | ATTR_MODE;
4707+ ia.ia_mode = h_isrc->i_mode;
523b37e3 4708+ err = vfsub_notify_change(&h_path, &ia, /*delegated*/NULL);
1facf9fc 4709+ }
4710+
c1595e42 4711+ icex = br->br_perm & AuBrAttr_ICEX;
7e9cd9fe
AM
4712+ if (!err) {
4713+ mnt_flags = au_mntflags(dst->d_sb);
4714+ verbose = !!au_opt_test(mnt_flags, VERBOSE);
4715+ err = au_cpup_xattr(h_path.dentry, h_src, icex, verbose);
4716+ }
c1595e42 4717+
1facf9fc 4718+ return err;
4719+}
4720+
4721+/* ---------------------------------------------------------------------- */
4722+
4723+static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
4724+ char *buf, unsigned long blksize)
4725+{
4726+ int err;
4727+ size_t sz, rbytes, wbytes;
4728+ unsigned char all_zero;
4729+ char *p, *zp;
4730+ struct mutex *h_mtx;
4731+ /* reduce stack usage */
4732+ struct iattr *ia;
4733+
4734+ zp = page_address(ZERO_PAGE(0));
4735+ if (unlikely(!zp))
4736+ return -ENOMEM; /* possible? */
4737+
4738+ err = 0;
4739+ all_zero = 0;
4740+ while (len) {
4741+ AuDbg("len %lld\n", len);
4742+ sz = blksize;
4743+ if (len < blksize)
4744+ sz = len;
4745+
4746+ rbytes = 0;
4747+ /* todo: signal_pending? */
4748+ while (!rbytes || err == -EAGAIN || err == -EINTR) {
4749+ rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
4750+ err = rbytes;
4751+ }
4752+ if (unlikely(err < 0))
4753+ break;
4754+
4755+ all_zero = 0;
4756+ if (len >= rbytes && rbytes == blksize)
4757+ all_zero = !memcmp(buf, zp, rbytes);
4758+ if (!all_zero) {
4759+ wbytes = rbytes;
4760+ p = buf;
4761+ while (wbytes) {
4762+ size_t b;
4763+
4764+ b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
4765+ err = b;
4766+ /* todo: signal_pending? */
4767+ if (unlikely(err == -EAGAIN || err == -EINTR))
4768+ continue;
4769+ if (unlikely(err < 0))
4770+ break;
4771+ wbytes -= b;
4772+ p += b;
4773+ }
392086de
AM
4774+ if (unlikely(err < 0))
4775+ break;
1facf9fc 4776+ } else {
4777+ loff_t res;
4778+
4779+ AuLabel(hole);
4780+ res = vfsub_llseek(dst, rbytes, SEEK_CUR);
4781+ err = res;
4782+ if (unlikely(res < 0))
4783+ break;
4784+ }
4785+ len -= rbytes;
4786+ err = 0;
4787+ }
4788+
4789+ /* the last block may be a hole */
4790+ if (!err && all_zero) {
4791+ AuLabel(last hole);
4792+
4793+ err = 1;
2000de60 4794+ if (au_test_nfs(dst->f_path.dentry->d_sb)) {
1facf9fc 4795+ /* nfs requires this step to make last hole */
4796+ /* is this only nfs? */
4797+ do {
4798+ /* todo: signal_pending? */
4799+ err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
4800+ } while (err == -EAGAIN || err == -EINTR);
4801+ if (err == 1)
4802+ dst->f_pos--;
4803+ }
4804+
4805+ if (err == 1) {
4806+ ia = (void *)buf;
4807+ ia->ia_size = dst->f_pos;
4808+ ia->ia_valid = ATTR_SIZE | ATTR_FILE;
4809+ ia->ia_file = dst;
c06a8ce3 4810+ h_mtx = &file_inode(dst)->i_mutex;
1facf9fc 4811+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
523b37e3
AM
4812+ /* no delegation since it is just created */
4813+ err = vfsub_notify_change(&dst->f_path, ia,
4814+ /*delegated*/NULL);
1facf9fc 4815+ mutex_unlock(h_mtx);
4816+ }
4817+ }
4818+
4819+ return err;
4820+}
4821+
4822+int au_copy_file(struct file *dst, struct file *src, loff_t len)
4823+{
4824+ int err;
4825+ unsigned long blksize;
4826+ unsigned char do_kfree;
4827+ char *buf;
4828+
4829+ err = -ENOMEM;
2000de60 4830+ blksize = dst->f_path.dentry->d_sb->s_blocksize;
1facf9fc 4831+ if (!blksize || PAGE_SIZE < blksize)
4832+ blksize = PAGE_SIZE;
4833+ AuDbg("blksize %lu\n", blksize);
4834+ do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
4835+ if (do_kfree)
4836+ buf = kmalloc(blksize, GFP_NOFS);
4837+ else
4838+ buf = (void *)__get_free_page(GFP_NOFS);
4839+ if (unlikely(!buf))
4840+ goto out;
4841+
4842+ if (len > (1 << 22))
4843+ AuDbg("copying a large file %lld\n", (long long)len);
4844+
4845+ src->f_pos = 0;
4846+ dst->f_pos = 0;
4847+ err = au_do_copy_file(dst, src, len, buf, blksize);
4848+ if (do_kfree)
4849+ kfree(buf);
4850+ else
4851+ free_page((unsigned long)buf);
4852+
4f0767ce 4853+out:
1facf9fc 4854+ return err;
4855+}
4856+
4857+/*
4858+ * to support a sparse file which is opened with O_APPEND,
4859+ * we need to close the file.
4860+ */
c2b27bf2 4861+static int au_cp_regular(struct au_cp_generic *cpg)
1facf9fc 4862+{
4863+ int err, i;
4864+ enum { SRC, DST };
4865+ struct {
4866+ aufs_bindex_t bindex;
4867+ unsigned int flags;
4868+ struct dentry *dentry;
392086de 4869+ int force_wr;
1facf9fc 4870+ struct file *file;
523b37e3 4871+ void *label;
1facf9fc 4872+ } *f, file[] = {
4873+ {
c2b27bf2 4874+ .bindex = cpg->bsrc,
1facf9fc 4875+ .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
523b37e3 4876+ .label = &&out
1facf9fc 4877+ },
4878+ {
c2b27bf2 4879+ .bindex = cpg->bdst,
1facf9fc 4880+ .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
392086de 4881+ .force_wr = !!au_ftest_cpup(cpg->flags, RWDST),
523b37e3 4882+ .label = &&out_src
1facf9fc 4883+ }
4884+ };
4885+ struct super_block *sb;
4886+
4887+ /* bsrc branch can be ro/rw. */
c2b27bf2 4888+ sb = cpg->dentry->d_sb;
1facf9fc 4889+ f = file;
4890+ for (i = 0; i < 2; i++, f++) {
c2b27bf2
AM
4891+ f->dentry = au_h_dptr(cpg->dentry, f->bindex);
4892+ f->file = au_h_open(cpg->dentry, f->bindex, f->flags,
392086de 4893+ /*file*/NULL, f->force_wr);
1facf9fc 4894+ err = PTR_ERR(f->file);
4895+ if (IS_ERR(f->file))
4896+ goto *f->label;
1facf9fc 4897+ }
4898+
4899+ /* try stopping to update while we copyup */
5527c038 4900+ IMustLock(d_inode(file[SRC].dentry));
c2b27bf2 4901+ err = au_copy_file(file[DST].file, file[SRC].file, cpg->len);
1facf9fc 4902+
1facf9fc 4903+ fput(file[DST].file);
4904+ au_sbr_put(sb, file[DST].bindex);
523b37e3 4905+
4f0767ce 4906+out_src:
1facf9fc 4907+ fput(file[SRC].file);
4908+ au_sbr_put(sb, file[SRC].bindex);
4f0767ce 4909+out:
1facf9fc 4910+ return err;
4911+}
4912+
c2b27bf2 4913+static int au_do_cpup_regular(struct au_cp_generic *cpg,
86dc4139 4914+ struct au_cpup_reg_attr *h_src_attr)
1facf9fc 4915+{
4916+ int err, rerr;
4917+ loff_t l;
86dc4139 4918+ struct path h_path;
38d290e6 4919+ struct inode *h_src_inode, *h_dst_inode;
1facf9fc 4920+
4921+ err = 0;
5527c038 4922+ h_src_inode = au_h_iptr(d_inode(cpg->dentry), cpg->bsrc);
86dc4139 4923+ l = i_size_read(h_src_inode);
c2b27bf2
AM
4924+ if (cpg->len == -1 || l < cpg->len)
4925+ cpg->len = l;
4926+ if (cpg->len) {
86dc4139
AM
4927+ /* try stopping to update while we are referencing */
4928+ mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
c2b27bf2 4929+ au_pin_hdir_unlock(cpg->pin);
1facf9fc 4930+
c2b27bf2
AM
4931+ h_path.dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
4932+ h_path.mnt = au_sbr_mnt(cpg->dentry->d_sb, cpg->bsrc);
86dc4139 4933+ h_src_attr->iflags = h_src_inode->i_flags;
5527c038
JR
4934+ if (!au_test_nfs(h_src_inode->i_sb))
4935+ err = vfs_getattr(&h_path, &h_src_attr->st);
4936+ else {
4937+ mutex_unlock(&h_src_inode->i_mutex);
4938+ err = vfs_getattr(&h_path, &h_src_attr->st);
4939+ mutex_lock_nested(&h_src_inode->i_mutex, AuLsc_I_CHILD);
4940+ }
86dc4139
AM
4941+ if (unlikely(err)) {
4942+ mutex_unlock(&h_src_inode->i_mutex);
4943+ goto out;
4944+ }
4945+ h_src_attr->valid = 1;
c2b27bf2 4946+ err = au_cp_regular(cpg);
86dc4139 4947+ mutex_unlock(&h_src_inode->i_mutex);
c2b27bf2 4948+ rerr = au_pin_hdir_relock(cpg->pin);
86dc4139
AM
4949+ if (!err && rerr)
4950+ err = rerr;
1facf9fc 4951+ }
38d290e6
JR
4952+ if (!err && (h_src_inode->i_state & I_LINKABLE)) {
4953+ h_path.dentry = au_h_dptr(cpg->dentry, cpg->bdst);
5527c038 4954+ h_dst_inode = d_inode(h_path.dentry);
38d290e6
JR
4955+ spin_lock(&h_dst_inode->i_lock);
4956+ h_dst_inode->i_state |= I_LINKABLE;
4957+ spin_unlock(&h_dst_inode->i_lock);
4958+ }
1facf9fc 4959+
4f0767ce 4960+out:
1facf9fc 4961+ return err;
4962+}
4963+
4964+static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
4965+ struct inode *h_dir)
4966+{
4967+ int err, symlen;
4968+ mm_segment_t old_fs;
b752ccd1
AM
4969+ union {
4970+ char *k;
4971+ char __user *u;
4972+ } sym;
5527c038
JR
4973+ struct inode *h_inode = d_inode(h_src);
4974+ const struct inode_operations *h_iop = h_inode->i_op;
1facf9fc 4975+
4976+ err = -ENOSYS;
5527c038 4977+ if (unlikely(!h_iop->readlink))
1facf9fc 4978+ goto out;
4979+
4980+ err = -ENOMEM;
537831f9 4981+ sym.k = (void *)__get_free_page(GFP_NOFS);
b752ccd1 4982+ if (unlikely(!sym.k))
1facf9fc 4983+ goto out;
4984+
9dbd164d 4985+ /* unnecessary to support mmap_sem since symlink is not mmap-able */
1facf9fc 4986+ old_fs = get_fs();
4987+ set_fs(KERNEL_DS);
5527c038 4988+ symlen = h_iop->readlink(h_src, sym.u, PATH_MAX);
1facf9fc 4989+ err = symlen;
4990+ set_fs(old_fs);
4991+
4992+ if (symlen > 0) {
b752ccd1
AM
4993+ sym.k[symlen] = 0;
4994+ err = vfsub_symlink(h_dir, h_path, sym.k);
1facf9fc 4995+ }
537831f9 4996+ free_page((unsigned long)sym.k);
1facf9fc 4997+
4f0767ce 4998+out:
1facf9fc 4999+ return err;
5000+}
5001+
1facf9fc 5002+static noinline_for_stack
c2b27bf2 5003+int cpup_entry(struct au_cp_generic *cpg, struct dentry *dst_parent,
86dc4139 5004+ struct au_cpup_reg_attr *h_src_attr)
1facf9fc 5005+{
5006+ int err;
5007+ umode_t mode;
5008+ unsigned int mnt_flags;
076b876e 5009+ unsigned char isdir, isreg, force;
c2b27bf2 5010+ const unsigned char do_dt = !!au_ftest_cpup(cpg->flags, DTIME);
1facf9fc 5011+ struct au_dtime dt;
5012+ struct path h_path;
5013+ struct dentry *h_src, *h_dst, *h_parent;
5527c038 5014+ struct inode *h_inode, *h_dir, *dir, *inode;
1facf9fc 5015+ struct super_block *sb;
5016+
5017+ /* bsrc branch can be ro/rw. */
c2b27bf2 5018+ h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5527c038
JR
5019+ h_inode = d_inode(h_src);
5020+ AuDebugOn(h_inode != au_h_iptr(d_inode(cpg->dentry), cpg->bsrc));
1facf9fc 5021+
5022+ /* try stopping to be referenced while we are creating */
c2b27bf2
AM
5023+ h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
5024+ if (au_ftest_cpup(cpg->flags, RENAME))
86dc4139
AM
5025+ AuDebugOn(strncmp(h_dst->d_name.name, AUFS_WH_PFX,
5026+ AUFS_WH_PFX_LEN));
1facf9fc 5027+ h_parent = h_dst->d_parent; /* dir inode is locked */
5527c038 5028+ h_dir = d_inode(h_parent);
1facf9fc 5029+ IMustLock(h_dir);
5030+ AuDebugOn(h_parent != h_dst->d_parent);
5031+
c2b27bf2
AM
5032+ sb = cpg->dentry->d_sb;
5033+ h_path.mnt = au_sbr_mnt(sb, cpg->bdst);
1facf9fc 5034+ if (do_dt) {
5035+ h_path.dentry = h_parent;
5036+ au_dtime_store(&dt, dst_parent, &h_path);
5037+ }
5038+ h_path.dentry = h_dst;
5039+
076b876e 5040+ isreg = 0;
1facf9fc 5041+ isdir = 0;
5042+ mode = h_inode->i_mode;
5043+ switch (mode & S_IFMT) {
5044+ case S_IFREG:
076b876e 5045+ isreg = 1;
b4510431
AM
5046+ err = vfsub_create(h_dir, &h_path, mode | S_IWUSR,
5047+ /*want_excl*/true);
1facf9fc 5048+ if (!err)
c2b27bf2 5049+ err = au_do_cpup_regular(cpg, h_src_attr);
1facf9fc 5050+ break;
5051+ case S_IFDIR:
5052+ isdir = 1;
5053+ err = vfsub_mkdir(h_dir, &h_path, mode);
5054+ if (!err) {
5055+ /*
5056+ * strange behaviour from the users view,
5057+ * particularry setattr case
5058+ */
5527c038
JR
5059+ dir = d_inode(dst_parent);
5060+ if (au_ibstart(dir) == cpg->bdst)
5061+ au_cpup_attr_nlink(dir, /*force*/1);
5062+ inode = d_inode(cpg->dentry);
5063+ au_cpup_attr_nlink(inode, /*force*/1);
1facf9fc 5064+ }
5065+ break;
5066+ case S_IFLNK:
5067+ err = au_do_cpup_symlink(&h_path, h_src, h_dir);
5068+ break;
5069+ case S_IFCHR:
5070+ case S_IFBLK:
5071+ AuDebugOn(!capable(CAP_MKNOD));
5072+ /*FALLTHROUGH*/
5073+ case S_IFIFO:
5074+ case S_IFSOCK:
5075+ err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
5076+ break;
5077+ default:
5078+ AuIOErr("Unknown inode type 0%o\n", mode);
5079+ err = -EIO;
5080+ }
5081+
5082+ mnt_flags = au_mntflags(sb);
5083+ if (!au_opt_test(mnt_flags, UDBA_NONE)
5084+ && !isdir
5085+ && au_opt_test(mnt_flags, XINO)
38d290e6
JR
5086+ && (h_inode->i_nlink == 1
5087+ || (h_inode->i_state & I_LINKABLE))
1facf9fc 5088+ /* todo: unnecessary? */
5527c038 5089+ /* && d_inode(cpg->dentry)->i_nlink == 1 */
c2b27bf2
AM
5090+ && cpg->bdst < cpg->bsrc
5091+ && !au_ftest_cpup(cpg->flags, KEEPLINO))
5092+ au_xino_write(sb, cpg->bsrc, h_inode->i_ino, /*ino*/0);
1facf9fc 5093+ /* ignore this error */
5094+
076b876e
AM
5095+ if (!err) {
5096+ force = 0;
5097+ if (isreg) {
5098+ force = !!cpg->len;
5099+ if (cpg->len == -1)
5100+ force = !!i_size_read(h_inode);
5101+ }
5102+ au_fhsm_wrote(sb, cpg->bdst, force);
5103+ }
5104+
1facf9fc 5105+ if (do_dt)
5106+ au_dtime_revert(&dt);
5107+ return err;
5108+}
5109+
392086de 5110+static int au_do_ren_after_cpup(struct au_cp_generic *cpg, struct path *h_path)
86dc4139
AM
5111+{
5112+ int err;
392086de 5113+ struct dentry *dentry, *h_dentry, *h_parent, *parent;
86dc4139 5114+ struct inode *h_dir;
392086de 5115+ aufs_bindex_t bdst;
86dc4139 5116+
392086de
AM
5117+ dentry = cpg->dentry;
5118+ bdst = cpg->bdst;
5119+ h_dentry = au_h_dptr(dentry, bdst);
5120+ if (!au_ftest_cpup(cpg->flags, OVERWRITE)) {
5121+ dget(h_dentry);
5122+ au_set_h_dptr(dentry, bdst, NULL);
5123+ err = au_lkup_neg(dentry, bdst, /*wh*/0);
5124+ if (!err)
5125+ h_path->dentry = dget(au_h_dptr(dentry, bdst));
86dc4139 5126+ au_set_h_dptr(dentry, bdst, h_dentry);
392086de
AM
5127+ } else {
5128+ err = 0;
5129+ parent = dget_parent(dentry);
5130+ h_parent = au_h_dptr(parent, bdst);
5131+ dput(parent);
5132+ h_path->dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
5133+ if (IS_ERR(h_path->dentry))
5134+ err = PTR_ERR(h_path->dentry);
86dc4139 5135+ }
392086de
AM
5136+ if (unlikely(err))
5137+ goto out;
86dc4139 5138+
86dc4139 5139+ h_parent = h_dentry->d_parent; /* dir inode is locked */
5527c038 5140+ h_dir = d_inode(h_parent);
86dc4139 5141+ IMustLock(h_dir);
523b37e3
AM
5142+ AuDbg("%pd %pd\n", h_dentry, h_path->dentry);
5143+ /* no delegation since it is just created */
5144+ err = vfsub_rename(h_dir, h_dentry, h_dir, h_path, /*delegated*/NULL);
86dc4139
AM
5145+ dput(h_path->dentry);
5146+
5147+out:
5148+ return err;
5149+}
5150+
1facf9fc 5151+/*
5152+ * copyup the @dentry from @bsrc to @bdst.
5153+ * the caller must set the both of lower dentries.
5154+ * @len is for truncating when it is -1 copyup the entire file.
5155+ * in link/rename cases, @dst_parent may be different from the real one.
c2b27bf2 5156+ * basic->bsrc can be larger than basic->bdst.
1facf9fc 5157+ */
c2b27bf2 5158+static int au_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
1facf9fc 5159+{
5160+ int err, rerr;
5161+ aufs_bindex_t old_ibstart;
5162+ unsigned char isdir, plink;
1facf9fc 5163+ struct dentry *h_src, *h_dst, *h_parent;
5527c038 5164+ struct inode *dst_inode, *h_dir, *inode, *delegated, *src_inode;
1facf9fc 5165+ struct super_block *sb;
86dc4139 5166+ struct au_branch *br;
c2b27bf2
AM
5167+ /* to reuduce stack size */
5168+ struct {
5169+ struct au_dtime dt;
5170+ struct path h_path;
5171+ struct au_cpup_reg_attr h_src_attr;
5172+ } *a;
1facf9fc 5173+
c2b27bf2
AM
5174+ err = -ENOMEM;
5175+ a = kmalloc(sizeof(*a), GFP_NOFS);
5176+ if (unlikely(!a))
5177+ goto out;
5178+ a->h_src_attr.valid = 0;
1facf9fc 5179+
c2b27bf2
AM
5180+ sb = cpg->dentry->d_sb;
5181+ br = au_sbr(sb, cpg->bdst);
5182+ a->h_path.mnt = au_br_mnt(br);
5183+ h_dst = au_h_dptr(cpg->dentry, cpg->bdst);
1facf9fc 5184+ h_parent = h_dst->d_parent; /* dir inode is locked */
5527c038 5185+ h_dir = d_inode(h_parent);
1facf9fc 5186+ IMustLock(h_dir);
5187+
c2b27bf2 5188+ h_src = au_h_dptr(cpg->dentry, cpg->bsrc);
5527c038 5189+ inode = d_inode(cpg->dentry);
1facf9fc 5190+
5191+ if (!dst_parent)
c2b27bf2 5192+ dst_parent = dget_parent(cpg->dentry);
1facf9fc 5193+ else
5194+ dget(dst_parent);
5195+
5196+ plink = !!au_opt_test(au_mntflags(sb), PLINK);
c2b27bf2 5197+ dst_inode = au_h_iptr(inode, cpg->bdst);
1facf9fc 5198+ if (dst_inode) {
5199+ if (unlikely(!plink)) {
5200+ err = -EIO;
027c5e7a
AM
5201+ AuIOErr("hi%lu(i%lu) exists on b%d "
5202+ "but plink is disabled\n",
c2b27bf2
AM
5203+ dst_inode->i_ino, inode->i_ino, cpg->bdst);
5204+ goto out_parent;
1facf9fc 5205+ }
5206+
5207+ if (dst_inode->i_nlink) {
c2b27bf2 5208+ const int do_dt = au_ftest_cpup(cpg->flags, DTIME);
1facf9fc 5209+
c2b27bf2 5210+ h_src = au_plink_lkup(inode, cpg->bdst);
1facf9fc 5211+ err = PTR_ERR(h_src);
5212+ if (IS_ERR(h_src))
c2b27bf2 5213+ goto out_parent;
5527c038 5214+ if (unlikely(d_is_negative(h_src))) {
1facf9fc 5215+ err = -EIO;
5216+ AuIOErr("i%lu exists on a upper branch "
027c5e7a
AM
5217+ "but not pseudo-linked\n",
5218+ inode->i_ino);
1facf9fc 5219+ dput(h_src);
c2b27bf2 5220+ goto out_parent;
1facf9fc 5221+ }
5222+
5223+ if (do_dt) {
c2b27bf2
AM
5224+ a->h_path.dentry = h_parent;
5225+ au_dtime_store(&a->dt, dst_parent, &a->h_path);
1facf9fc 5226+ }
86dc4139 5227+
c2b27bf2 5228+ a->h_path.dentry = h_dst;
523b37e3
AM
5229+ delegated = NULL;
5230+ err = vfsub_link(h_src, h_dir, &a->h_path, &delegated);
c2b27bf2 5231+ if (!err && au_ftest_cpup(cpg->flags, RENAME))
392086de 5232+ err = au_do_ren_after_cpup(cpg, &a->h_path);
1facf9fc 5233+ if (do_dt)
c2b27bf2 5234+ au_dtime_revert(&a->dt);
523b37e3
AM
5235+ if (unlikely(err == -EWOULDBLOCK)) {
5236+ pr_warn("cannot retry for NFSv4 delegation"
5237+ " for an internal link\n");
5238+ iput(delegated);
5239+ }
1facf9fc 5240+ dput(h_src);
c2b27bf2 5241+ goto out_parent;
1facf9fc 5242+ } else
5243+ /* todo: cpup_wh_file? */
5244+ /* udba work */
4a4d8108 5245+ au_update_ibrange(inode, /*do_put_zero*/1);
1facf9fc 5246+ }
5247+
86dc4139 5248+ isdir = S_ISDIR(inode->i_mode);
1facf9fc 5249+ old_ibstart = au_ibstart(inode);
c2b27bf2 5250+ err = cpup_entry(cpg, dst_parent, &a->h_src_attr);
1facf9fc 5251+ if (unlikely(err))
86dc4139 5252+ goto out_rev;
5527c038 5253+ dst_inode = d_inode(h_dst);
1facf9fc 5254+ mutex_lock_nested(&dst_inode->i_mutex, AuLsc_I_CHILD2);
86dc4139 5255+ /* todo: necessary? */
c2b27bf2 5256+ /* au_pin_hdir_unlock(cpg->pin); */
1facf9fc 5257+
c2b27bf2 5258+ err = cpup_iattr(cpg->dentry, cpg->bdst, h_src, &a->h_src_attr);
86dc4139
AM
5259+ if (unlikely(err)) {
5260+ /* todo: necessary? */
c2b27bf2 5261+ /* au_pin_hdir_relock(cpg->pin); */ /* ignore an error */
86dc4139
AM
5262+ mutex_unlock(&dst_inode->i_mutex);
5263+ goto out_rev;
5264+ }
5265+
c2b27bf2 5266+ if (cpg->bdst < old_ibstart) {
86dc4139 5267+ if (S_ISREG(inode->i_mode)) {
c2b27bf2 5268+ err = au_dy_iaop(inode, cpg->bdst, dst_inode);
86dc4139 5269+ if (unlikely(err)) {
c2b27bf2
AM
5270+ /* ignore an error */
5271+ /* au_pin_hdir_relock(cpg->pin); */
86dc4139
AM
5272+ mutex_unlock(&dst_inode->i_mutex);
5273+ goto out_rev;
4a4d8108 5274+ }
4a4d8108 5275+ }
c2b27bf2
AM
5276+ au_set_ibstart(inode, cpg->bdst);
5277+ } else
5278+ au_set_ibend(inode, cpg->bdst);
5279+ au_set_h_iptr(inode, cpg->bdst, au_igrab(dst_inode),
86dc4139
AM
5280+ au_hi_flags(inode, isdir));
5281+
5282+ /* todo: necessary? */
c2b27bf2 5283+ /* err = au_pin_hdir_relock(cpg->pin); */
86dc4139
AM
5284+ mutex_unlock(&dst_inode->i_mutex);
5285+ if (unlikely(err))
5286+ goto out_rev;
5287+
5527c038 5288+ src_inode = d_inode(h_src);
86dc4139 5289+ if (!isdir
5527c038
JR
5290+ && (src_inode->i_nlink > 1
5291+ || src_inode->i_state & I_LINKABLE)
86dc4139 5292+ && plink)
c2b27bf2 5293+ au_plink_append(inode, cpg->bdst, h_dst);
86dc4139 5294+
c2b27bf2
AM
5295+ if (au_ftest_cpup(cpg->flags, RENAME)) {
5296+ a->h_path.dentry = h_dst;
392086de 5297+ err = au_do_ren_after_cpup(cpg, &a->h_path);
86dc4139
AM
5298+ }
5299+ if (!err)
c2b27bf2 5300+ goto out_parent; /* success */
1facf9fc 5301+
5302+ /* revert */
4a4d8108 5303+out_rev:
c2b27bf2
AM
5304+ a->h_path.dentry = h_parent;
5305+ au_dtime_store(&a->dt, dst_parent, &a->h_path);
5306+ a->h_path.dentry = h_dst;
86dc4139 5307+ rerr = 0;
5527c038 5308+ if (d_is_positive(h_dst)) {
523b37e3
AM
5309+ if (!isdir) {
5310+ /* no delegation since it is just created */
5311+ rerr = vfsub_unlink(h_dir, &a->h_path,
5312+ /*delegated*/NULL, /*force*/0);
5313+ } else
c2b27bf2 5314+ rerr = vfsub_rmdir(h_dir, &a->h_path);
86dc4139 5315+ }
c2b27bf2 5316+ au_dtime_revert(&a->dt);
1facf9fc 5317+ if (rerr) {
5318+ AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
5319+ err = -EIO;
5320+ }
c2b27bf2 5321+out_parent:
1facf9fc 5322+ dput(dst_parent);
c2b27bf2
AM
5323+ kfree(a);
5324+out:
1facf9fc 5325+ return err;
5326+}
5327+
7e9cd9fe 5328+#if 0 /* reserved */
1facf9fc 5329+struct au_cpup_single_args {
5330+ int *errp;
c2b27bf2 5331+ struct au_cp_generic *cpg;
1facf9fc 5332+ struct dentry *dst_parent;
5333+};
5334+
5335+static void au_call_cpup_single(void *args)
5336+{
5337+ struct au_cpup_single_args *a = args;
86dc4139 5338+
c2b27bf2
AM
5339+ au_pin_hdir_acquire_nest(a->cpg->pin);
5340+ *a->errp = au_cpup_single(a->cpg, a->dst_parent);
5341+ au_pin_hdir_release(a->cpg->pin);
1facf9fc 5342+}
c2b27bf2 5343+#endif
1facf9fc 5344+
53392da6
AM
5345+/*
5346+ * prevent SIGXFSZ in copy-up.
5347+ * testing CAP_MKNOD is for generic fs,
5348+ * but CAP_FSETID is for xfs only, currently.
5349+ */
86dc4139 5350+static int au_cpup_sio_test(struct au_pin *pin, umode_t mode)
53392da6
AM
5351+{
5352+ int do_sio;
86dc4139
AM
5353+ struct super_block *sb;
5354+ struct inode *h_dir;
53392da6
AM
5355+
5356+ do_sio = 0;
86dc4139 5357+ sb = au_pinned_parent(pin)->d_sb;
53392da6
AM
5358+ if (!au_wkq_test()
5359+ && (!au_sbi(sb)->si_plink_maint_pid
5360+ || au_plink_maint(sb, AuLock_NOPLM))) {
5361+ switch (mode & S_IFMT) {
5362+ case S_IFREG:
5363+ /* no condition about RLIMIT_FSIZE and the file size */
5364+ do_sio = 1;
5365+ break;
5366+ case S_IFCHR:
5367+ case S_IFBLK:
5368+ do_sio = !capable(CAP_MKNOD);
5369+ break;
5370+ }
5371+ if (!do_sio)
5372+ do_sio = ((mode & (S_ISUID | S_ISGID))
5373+ && !capable(CAP_FSETID));
86dc4139
AM
5374+ /* this workaround may be removed in the future */
5375+ if (!do_sio) {
5376+ h_dir = au_pinned_h_dir(pin);
5377+ do_sio = h_dir->i_mode & S_ISVTX;
5378+ }
53392da6
AM
5379+ }
5380+
5381+ return do_sio;
5382+}
5383+
7e9cd9fe 5384+#if 0 /* reserved */
c2b27bf2 5385+int au_sio_cpup_single(struct au_cp_generic *cpg, struct dentry *dst_parent)
1facf9fc 5386+{
5387+ int err, wkq_err;
1facf9fc 5388+ struct dentry *h_dentry;
5389+
c2b27bf2 5390+ h_dentry = au_h_dptr(cpg->dentry, cpg->bsrc);
5527c038 5391+ if (!au_cpup_sio_test(pin, d_inode(h_dentry)->i_mode))
c2b27bf2 5392+ err = au_cpup_single(cpg, dst_parent);
1facf9fc 5393+ else {
5394+ struct au_cpup_single_args args = {
5395+ .errp = &err,
c2b27bf2
AM
5396+ .cpg = cpg,
5397+ .dst_parent = dst_parent
1facf9fc 5398+ };
5399+ wkq_err = au_wkq_wait(au_call_cpup_single, &args);
5400+ if (unlikely(wkq_err))
5401+ err = wkq_err;
5402+ }
5403+
5404+ return err;
5405+}
c2b27bf2 5406+#endif
1facf9fc 5407+
5408+/*
5409+ * copyup the @dentry from the first active lower branch to @bdst,
5410+ * using au_cpup_single().
5411+ */
c2b27bf2 5412+static int au_cpup_simple(struct au_cp_generic *cpg)
1facf9fc 5413+{
5414+ int err;
c2b27bf2
AM
5415+ unsigned int flags_orig;
5416+ struct dentry *dentry;
5417+
5418+ AuDebugOn(cpg->bsrc < 0);
1facf9fc 5419+
c2b27bf2 5420+ dentry = cpg->dentry;
86dc4139 5421+ DiMustWriteLock(dentry);
1facf9fc 5422+
c2b27bf2 5423+ err = au_lkup_neg(dentry, cpg->bdst, /*wh*/1);
1facf9fc 5424+ if (!err) {
c2b27bf2
AM
5425+ flags_orig = cpg->flags;
5426+ au_fset_cpup(cpg->flags, RENAME);
5427+ err = au_cpup_single(cpg, NULL);
5428+ cpg->flags = flags_orig;
1facf9fc 5429+ if (!err)
5430+ return 0; /* success */
5431+
5432+ /* revert */
c2b27bf2
AM
5433+ au_set_h_dptr(dentry, cpg->bdst, NULL);
5434+ au_set_dbstart(dentry, cpg->bsrc);
1facf9fc 5435+ }
5436+
5437+ return err;
5438+}
5439+
5440+struct au_cpup_simple_args {
5441+ int *errp;
c2b27bf2 5442+ struct au_cp_generic *cpg;
1facf9fc 5443+};
5444+
5445+static void au_call_cpup_simple(void *args)
5446+{
5447+ struct au_cpup_simple_args *a = args;
86dc4139 5448+
c2b27bf2
AM
5449+ au_pin_hdir_acquire_nest(a->cpg->pin);
5450+ *a->errp = au_cpup_simple(a->cpg);
5451+ au_pin_hdir_release(a->cpg->pin);
1facf9fc 5452+}
5453+
c2b27bf2 5454+static int au_do_sio_cpup_simple(struct au_cp_generic *cpg)
1facf9fc 5455+{
5456+ int err, wkq_err;
c2b27bf2
AM
5457+ struct dentry *dentry, *parent;
5458+ struct file *h_file;
1facf9fc 5459+ struct inode *h_dir;
5460+
c2b27bf2
AM
5461+ dentry = cpg->dentry;
5462+ h_file = NULL;
5463+ if (au_ftest_cpup(cpg->flags, HOPEN)) {
5464+ AuDebugOn(cpg->bsrc < 0);
392086de 5465+ h_file = au_h_open_pre(dentry, cpg->bsrc, /*force_wr*/0);
c2b27bf2
AM
5466+ err = PTR_ERR(h_file);
5467+ if (IS_ERR(h_file))
5468+ goto out;
5469+ }
5470+
1facf9fc 5471+ parent = dget_parent(dentry);
5527c038 5472+ h_dir = au_h_iptr(d_inode(parent), cpg->bdst);
53392da6 5473+ if (!au_test_h_perm_sio(h_dir, MAY_EXEC | MAY_WRITE)
5527c038 5474+ && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
c2b27bf2 5475+ err = au_cpup_simple(cpg);
1facf9fc 5476+ else {
5477+ struct au_cpup_simple_args args = {
5478+ .errp = &err,
c2b27bf2 5479+ .cpg = cpg
1facf9fc 5480+ };
5481+ wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
5482+ if (unlikely(wkq_err))
5483+ err = wkq_err;
5484+ }
5485+
5486+ dput(parent);
c2b27bf2
AM
5487+ if (h_file)
5488+ au_h_open_post(dentry, cpg->bsrc, h_file);
5489+
5490+out:
1facf9fc 5491+ return err;
5492+}
5493+
c2b27bf2 5494+int au_sio_cpup_simple(struct au_cp_generic *cpg)
367653fa 5495+{
c2b27bf2
AM
5496+ aufs_bindex_t bsrc, bend;
5497+ struct dentry *dentry, *h_dentry;
367653fa 5498+
c2b27bf2
AM
5499+ if (cpg->bsrc < 0) {
5500+ dentry = cpg->dentry;
5501+ bend = au_dbend(dentry);
5502+ for (bsrc = cpg->bdst + 1; bsrc <= bend; bsrc++) {
5503+ h_dentry = au_h_dptr(dentry, bsrc);
5504+ if (h_dentry) {
5527c038 5505+ AuDebugOn(d_is_negative(h_dentry));
c2b27bf2
AM
5506+ break;
5507+ }
5508+ }
5509+ AuDebugOn(bsrc > bend);
5510+ cpg->bsrc = bsrc;
367653fa 5511+ }
c2b27bf2
AM
5512+ AuDebugOn(cpg->bsrc <= cpg->bdst);
5513+ return au_do_sio_cpup_simple(cpg);
5514+}
367653fa 5515+
c2b27bf2
AM
5516+int au_sio_cpdown_simple(struct au_cp_generic *cpg)
5517+{
5518+ AuDebugOn(cpg->bdst <= cpg->bsrc);
5519+ return au_do_sio_cpup_simple(cpg);
367653fa
AM
5520+}
5521+
1facf9fc 5522+/* ---------------------------------------------------------------------- */
5523+
5524+/*
5525+ * copyup the deleted file for writing.
5526+ */
c2b27bf2
AM
5527+static int au_do_cpup_wh(struct au_cp_generic *cpg, struct dentry *wh_dentry,
5528+ struct file *file)
1facf9fc 5529+{
5530+ int err;
c2b27bf2
AM
5531+ unsigned int flags_orig;
5532+ aufs_bindex_t bsrc_orig;
1facf9fc 5533+ struct dentry *h_d_dst, *h_d_start;
c2b27bf2 5534+ struct au_dinfo *dinfo;
4a4d8108 5535+ struct au_hdentry *hdp;
1facf9fc 5536+
c2b27bf2 5537+ dinfo = au_di(cpg->dentry);
1308ab2a 5538+ AuRwMustWriteLock(&dinfo->di_rwsem);
5539+
c2b27bf2
AM
5540+ bsrc_orig = cpg->bsrc;
5541+ cpg->bsrc = dinfo->di_bstart;
4a4d8108 5542+ hdp = dinfo->di_hdentry;
c2b27bf2
AM
5543+ h_d_dst = hdp[0 + cpg->bdst].hd_dentry;
5544+ dinfo->di_bstart = cpg->bdst;
5545+ hdp[0 + cpg->bdst].hd_dentry = wh_dentry;
86dc4139 5546+ h_d_start = NULL;
027c5e7a 5547+ if (file) {
c2b27bf2 5548+ h_d_start = hdp[0 + cpg->bsrc].hd_dentry;
2000de60 5549+ hdp[0 + cpg->bsrc].hd_dentry = au_hf_top(file)->f_path.dentry;
027c5e7a 5550+ }
c2b27bf2
AM
5551+ flags_orig = cpg->flags;
5552+ cpg->flags = !AuCpup_DTIME;
5553+ err = au_cpup_single(cpg, /*h_parent*/NULL);
5554+ cpg->flags = flags_orig;
027c5e7a
AM
5555+ if (file) {
5556+ if (!err)
5557+ err = au_reopen_nondir(file);
c2b27bf2 5558+ hdp[0 + cpg->bsrc].hd_dentry = h_d_start;
1facf9fc 5559+ }
c2b27bf2
AM
5560+ hdp[0 + cpg->bdst].hd_dentry = h_d_dst;
5561+ dinfo->di_bstart = cpg->bsrc;
5562+ cpg->bsrc = bsrc_orig;
1facf9fc 5563+
5564+ return err;
5565+}
5566+
c2b27bf2 5567+static int au_cpup_wh(struct au_cp_generic *cpg, struct file *file)
1facf9fc 5568+{
5569+ int err;
c2b27bf2 5570+ aufs_bindex_t bdst;
1facf9fc 5571+ struct au_dtime dt;
c2b27bf2 5572+ struct dentry *dentry, *parent, *h_parent, *wh_dentry;
1facf9fc 5573+ struct au_branch *br;
5574+ struct path h_path;
5575+
c2b27bf2
AM
5576+ dentry = cpg->dentry;
5577+ bdst = cpg->bdst;
1facf9fc 5578+ br = au_sbr(dentry->d_sb, bdst);
5579+ parent = dget_parent(dentry);
5580+ h_parent = au_h_dptr(parent, bdst);
5581+ wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
5582+ err = PTR_ERR(wh_dentry);
5583+ if (IS_ERR(wh_dentry))
5584+ goto out;
5585+
5586+ h_path.dentry = h_parent;
86dc4139 5587+ h_path.mnt = au_br_mnt(br);
1facf9fc 5588+ au_dtime_store(&dt, parent, &h_path);
c2b27bf2 5589+ err = au_do_cpup_wh(cpg, wh_dentry, file);
1facf9fc 5590+ if (unlikely(err))
5591+ goto out_wh;
5592+
5593+ dget(wh_dentry);
5594+ h_path.dentry = wh_dentry;
2000de60 5595+ if (!d_is_dir(wh_dentry)) {
523b37e3 5596+ /* no delegation since it is just created */
5527c038 5597+ err = vfsub_unlink(d_inode(h_parent), &h_path,
523b37e3
AM
5598+ /*delegated*/NULL, /*force*/0);
5599+ } else
5527c038 5600+ err = vfsub_rmdir(d_inode(h_parent), &h_path);
1facf9fc 5601+ if (unlikely(err)) {
523b37e3
AM
5602+ AuIOErr("failed remove copied-up tmp file %pd(%d)\n",
5603+ wh_dentry, err);
1facf9fc 5604+ err = -EIO;
5605+ }
5606+ au_dtime_revert(&dt);
5527c038 5607+ au_set_hi_wh(d_inode(dentry), bdst, wh_dentry);
1facf9fc 5608+
4f0767ce 5609+out_wh:
1facf9fc 5610+ dput(wh_dentry);
4f0767ce 5611+out:
1facf9fc 5612+ dput(parent);
5613+ return err;
5614+}
5615+
5616+struct au_cpup_wh_args {
5617+ int *errp;
c2b27bf2 5618+ struct au_cp_generic *cpg;
1facf9fc 5619+ struct file *file;
5620+};
5621+
5622+static void au_call_cpup_wh(void *args)
5623+{
5624+ struct au_cpup_wh_args *a = args;
86dc4139 5625+
c2b27bf2
AM
5626+ au_pin_hdir_acquire_nest(a->cpg->pin);
5627+ *a->errp = au_cpup_wh(a->cpg, a->file);
5628+ au_pin_hdir_release(a->cpg->pin);
1facf9fc 5629+}
5630+
c2b27bf2 5631+int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file)
1facf9fc 5632+{
5633+ int err, wkq_err;
c2b27bf2 5634+ aufs_bindex_t bdst;
c1595e42 5635+ struct dentry *dentry, *parent, *h_orph, *h_parent;
86dc4139 5636+ struct inode *dir, *h_dir, *h_tmpdir;
1facf9fc 5637+ struct au_wbr *wbr;
c2b27bf2 5638+ struct au_pin wh_pin, *pin_orig;
1facf9fc 5639+
c2b27bf2
AM
5640+ dentry = cpg->dentry;
5641+ bdst = cpg->bdst;
1facf9fc 5642+ parent = dget_parent(dentry);
5527c038 5643+ dir = d_inode(parent);
1facf9fc 5644+ h_orph = NULL;
5645+ h_parent = NULL;
5646+ h_dir = au_igrab(au_h_iptr(dir, bdst));
5647+ h_tmpdir = h_dir;
c2b27bf2 5648+ pin_orig = NULL;
1facf9fc 5649+ if (!h_dir->i_nlink) {
5650+ wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
5651+ h_orph = wbr->wbr_orph;
5652+
5653+ h_parent = dget(au_h_dptr(parent, bdst));
1facf9fc 5654+ au_set_h_dptr(parent, bdst, dget(h_orph));
5527c038 5655+ h_tmpdir = d_inode(h_orph);
1facf9fc 5656+ au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
5657+
dece6358 5658+ mutex_lock_nested(&h_tmpdir->i_mutex, AuLsc_I_PARENT3);
4a4d8108 5659+ /* todo: au_h_open_pre()? */
86dc4139 5660+
c2b27bf2 5661+ pin_orig = cpg->pin;
86dc4139 5662+ au_pin_init(&wh_pin, dentry, bdst, AuLsc_DI_PARENT,
c2b27bf2
AM
5663+ AuLsc_I_PARENT3, cpg->pin->udba, AuPin_DI_LOCKED);
5664+ cpg->pin = &wh_pin;
1facf9fc 5665+ }
5666+
53392da6 5667+ if (!au_test_h_perm_sio(h_tmpdir, MAY_EXEC | MAY_WRITE)
5527c038 5668+ && !au_cpup_sio_test(cpg->pin, d_inode(dentry)->i_mode))
c2b27bf2 5669+ err = au_cpup_wh(cpg, file);
1facf9fc 5670+ else {
5671+ struct au_cpup_wh_args args = {
5672+ .errp = &err,
c2b27bf2
AM
5673+ .cpg = cpg,
5674+ .file = file
1facf9fc 5675+ };
5676+ wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
5677+ if (unlikely(wkq_err))
5678+ err = wkq_err;
5679+ }
5680+
5681+ if (h_orph) {
5682+ mutex_unlock(&h_tmpdir->i_mutex);
4a4d8108 5683+ /* todo: au_h_open_post()? */
1facf9fc 5684+ au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
1facf9fc 5685+ au_set_h_dptr(parent, bdst, h_parent);
c2b27bf2
AM
5686+ AuDebugOn(!pin_orig);
5687+ cpg->pin = pin_orig;
1facf9fc 5688+ }
5689+ iput(h_dir);
5690+ dput(parent);
5691+
5692+ return err;
5693+}
5694+
5695+/* ---------------------------------------------------------------------- */
5696+
5697+/*
5698+ * generic routine for both of copy-up and copy-down.
5699+ */
5700+/* cf. revalidate function in file.c */
5701+int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5702+ int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 5703+ struct au_pin *pin,
1facf9fc 5704+ struct dentry *h_parent, void *arg),
5705+ void *arg)
5706+{
5707+ int err;
5708+ struct au_pin pin;
5527c038 5709+ struct dentry *d, *parent, *h_parent, *real_parent, *h_dentry;
1facf9fc 5710+
5711+ err = 0;
5712+ parent = dget_parent(dentry);
5713+ if (IS_ROOT(parent))
5714+ goto out;
5715+
5716+ au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
5717+ au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
5718+
5719+ /* do not use au_dpage */
5720+ real_parent = parent;
5721+ while (1) {
5722+ dput(parent);
5723+ parent = dget_parent(dentry);
5724+ h_parent = au_h_dptr(parent, bdst);
5725+ if (h_parent)
5726+ goto out; /* success */
5727+
5728+ /* find top dir which is necessary to cpup */
5729+ do {
5730+ d = parent;
5731+ dput(parent);
5732+ parent = dget_parent(d);
5733+ di_read_lock_parent3(parent, !AuLock_IR);
5734+ h_parent = au_h_dptr(parent, bdst);
5735+ di_read_unlock(parent, !AuLock_IR);
5736+ } while (!h_parent);
5737+
5738+ if (d != real_parent)
5739+ di_write_lock_child3(d);
5740+
5741+ /* somebody else might create while we were sleeping */
5527c038
JR
5742+ h_dentry = au_h_dptr(d, bdst);
5743+ if (!h_dentry || d_is_negative(h_dentry)) {
5744+ if (h_dentry)
1facf9fc 5745+ au_update_dbstart(d);
5746+
5747+ au_pin_set_dentry(&pin, d);
5748+ err = au_do_pin(&pin);
5749+ if (!err) {
86dc4139 5750+ err = cp(d, bdst, &pin, h_parent, arg);
1facf9fc 5751+ au_unpin(&pin);
5752+ }
5753+ }
5754+
5755+ if (d != real_parent)
5756+ di_write_unlock(d);
5757+ if (unlikely(err))
5758+ break;
5759+ }
5760+
4f0767ce 5761+out:
1facf9fc 5762+ dput(parent);
5763+ return err;
5764+}
5765+
5766+static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 5767+ struct au_pin *pin,
2000de60 5768+ struct dentry *h_parent __maybe_unused,
1facf9fc 5769+ void *arg __maybe_unused)
5770+{
c2b27bf2
AM
5771+ struct au_cp_generic cpg = {
5772+ .dentry = dentry,
5773+ .bdst = bdst,
5774+ .bsrc = -1,
5775+ .len = 0,
5776+ .pin = pin,
5777+ .flags = AuCpup_DTIME
5778+ };
5779+ return au_sio_cpup_simple(&cpg);
1facf9fc 5780+}
5781+
5782+int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5783+{
5784+ return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
5785+}
5786+
5787+int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
5788+{
5789+ int err;
5790+ struct dentry *parent;
5791+ struct inode *dir;
5792+
5793+ parent = dget_parent(dentry);
5527c038 5794+ dir = d_inode(parent);
1facf9fc 5795+ err = 0;
5796+ if (au_h_iptr(dir, bdst))
5797+ goto out;
5798+
5799+ di_read_unlock(parent, AuLock_IR);
5800+ di_write_lock_parent(parent);
5801+ /* someone else might change our inode while we were sleeping */
5802+ if (!au_h_iptr(dir, bdst))
5803+ err = au_cpup_dirs(dentry, bdst);
5804+ di_downgrade_lock(parent, AuLock_IR);
5805+
4f0767ce 5806+out:
1facf9fc 5807+ dput(parent);
5808+ return err;
5809+}
7f207e10
AM
5810diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
5811--- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 5812+++ linux/fs/aufs/cpup.h 2015-09-24 10:47:58.248052907 +0200
523b37e3 5813@@ -0,0 +1,94 @@
1facf9fc 5814+/*
2000de60 5815+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 5816+ *
5817+ * This program, aufs is free software; you can redistribute it and/or modify
5818+ * it under the terms of the GNU General Public License as published by
5819+ * the Free Software Foundation; either version 2 of the License, or
5820+ * (at your option) any later version.
dece6358
AM
5821+ *
5822+ * This program is distributed in the hope that it will be useful,
5823+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5824+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5825+ * GNU General Public License for more details.
5826+ *
5827+ * You should have received a copy of the GNU General Public License
523b37e3 5828+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5829+ */
5830+
5831+/*
5832+ * copy-up/down functions
5833+ */
5834+
5835+#ifndef __AUFS_CPUP_H__
5836+#define __AUFS_CPUP_H__
5837+
5838+#ifdef __KERNEL__
5839+
dece6358 5840+#include <linux/path.h>
1facf9fc 5841+
dece6358
AM
5842+struct inode;
5843+struct file;
86dc4139 5844+struct au_pin;
dece6358 5845+
86dc4139 5846+void au_cpup_attr_flags(struct inode *dst, unsigned int iflags);
1facf9fc 5847+void au_cpup_attr_timesizes(struct inode *inode);
5848+void au_cpup_attr_nlink(struct inode *inode, int force);
5849+void au_cpup_attr_changeable(struct inode *inode);
5850+void au_cpup_igen(struct inode *inode, struct inode *h_inode);
5851+void au_cpup_attr_all(struct inode *inode, int force);
5852+
5853+/* ---------------------------------------------------------------------- */
5854+
c2b27bf2
AM
5855+struct au_cp_generic {
5856+ struct dentry *dentry;
5857+ aufs_bindex_t bdst, bsrc;
5858+ loff_t len;
5859+ struct au_pin *pin;
5860+ unsigned int flags;
5861+};
5862+
1facf9fc 5863+/* cpup flags */
392086de
AM
5864+#define AuCpup_DTIME 1 /* do dtime_store/revert */
5865+#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
5866+ for link(2) */
5867+#define AuCpup_RENAME (1 << 2) /* rename after cpup */
5868+#define AuCpup_HOPEN (1 << 3) /* call h_open_pre/post() in
5869+ cpup */
5870+#define AuCpup_OVERWRITE (1 << 4) /* allow overwriting the
5871+ existing entry */
5872+#define AuCpup_RWDST (1 << 5) /* force write target even if
5873+ the branch is marked as RO */
c2b27bf2 5874+
1facf9fc 5875+#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
7f207e10
AM
5876+#define au_fset_cpup(flags, name) \
5877+ do { (flags) |= AuCpup_##name; } while (0)
5878+#define au_fclr_cpup(flags, name) \
5879+ do { (flags) &= ~AuCpup_##name; } while (0)
1facf9fc 5880+
5881+int au_copy_file(struct file *dst, struct file *src, loff_t len);
c2b27bf2
AM
5882+int au_sio_cpup_simple(struct au_cp_generic *cpg);
5883+int au_sio_cpdown_simple(struct au_cp_generic *cpg);
5884+int au_sio_cpup_wh(struct au_cp_generic *cpg, struct file *file);
1facf9fc 5885+
5886+int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
5887+ int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 5888+ struct au_pin *pin,
1facf9fc 5889+ struct dentry *h_parent, void *arg),
5890+ void *arg);
5891+int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5892+int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
5893+
5894+/* ---------------------------------------------------------------------- */
5895+
5896+/* keep timestamps when copyup */
5897+struct au_dtime {
5898+ struct dentry *dt_dentry;
5899+ struct path dt_h_path;
5900+ struct timespec dt_atime, dt_mtime;
5901+};
5902+void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
5903+ struct path *h_path);
5904+void au_dtime_revert(struct au_dtime *dt);
5905+
5906+#endif /* __KERNEL__ */
5907+#endif /* __AUFS_CPUP_H__ */
7f207e10
AM
5908diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
5909--- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 5910+++ linux/fs/aufs/dbgaufs.c 2015-09-24 10:47:58.248052907 +0200
523b37e3 5911@@ -0,0 +1,432 @@
1facf9fc 5912+/*
2000de60 5913+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 5914+ *
5915+ * This program, aufs is free software; you can redistribute it and/or modify
5916+ * it under the terms of the GNU General Public License as published by
5917+ * the Free Software Foundation; either version 2 of the License, or
5918+ * (at your option) any later version.
dece6358
AM
5919+ *
5920+ * This program is distributed in the hope that it will be useful,
5921+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
5922+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5923+ * GNU General Public License for more details.
5924+ *
5925+ * You should have received a copy of the GNU General Public License
523b37e3 5926+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 5927+ */
5928+
5929+/*
5930+ * debugfs interface
5931+ */
5932+
5933+#include <linux/debugfs.h>
5934+#include "aufs.h"
5935+
5936+#ifndef CONFIG_SYSFS
5937+#error DEBUG_FS depends upon SYSFS
5938+#endif
5939+
5940+static struct dentry *dbgaufs;
5941+static const mode_t dbgaufs_mode = S_IRUSR | S_IRGRP | S_IROTH;
5942+
5943+/* 20 is max digits length of ulong 64 */
5944+struct dbgaufs_arg {
5945+ int n;
5946+ char a[20 * 4];
5947+};
5948+
5949+/*
5950+ * common function for all XINO files
5951+ */
5952+static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
5953+ struct file *file)
5954+{
5955+ kfree(file->private_data);
5956+ return 0;
5957+}
5958+
5959+static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt)
5960+{
5961+ int err;
5962+ struct kstat st;
5963+ struct dbgaufs_arg *p;
5964+
5965+ err = -ENOMEM;
5966+ p = kmalloc(sizeof(*p), GFP_NOFS);
5967+ if (unlikely(!p))
5968+ goto out;
5969+
5970+ err = 0;
5971+ p->n = 0;
5972+ file->private_data = p;
5973+ if (!xf)
5974+ goto out;
5975+
c06a8ce3 5976+ err = vfs_getattr(&xf->f_path, &st);
1facf9fc 5977+ if (!err) {
5978+ if (do_fcnt)
5979+ p->n = snprintf
5980+ (p->a, sizeof(p->a), "%ld, %llux%lu %lld\n",
5981+ (long)file_count(xf), st.blocks, st.blksize,
5982+ (long long)st.size);
5983+ else
5984+ p->n = snprintf(p->a, sizeof(p->a), "%llux%lu %lld\n",
5985+ st.blocks, st.blksize,
5986+ (long long)st.size);
5987+ AuDebugOn(p->n >= sizeof(p->a));
5988+ } else {
5989+ p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
5990+ err = 0;
5991+ }
5992+
4f0767ce 5993+out:
1facf9fc 5994+ return err;
5995+
5996+}
5997+
5998+static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
5999+ size_t count, loff_t *ppos)
6000+{
6001+ struct dbgaufs_arg *p;
6002+
6003+ p = file->private_data;
6004+ return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6005+}
6006+
6007+/* ---------------------------------------------------------------------- */
6008+
86dc4139
AM
6009+struct dbgaufs_plink_arg {
6010+ int n;
6011+ char a[];
6012+};
6013+
6014+static int dbgaufs_plink_release(struct inode *inode __maybe_unused,
6015+ struct file *file)
6016+{
6017+ free_page((unsigned long)file->private_data);
6018+ return 0;
6019+}
6020+
6021+static int dbgaufs_plink_open(struct inode *inode, struct file *file)
6022+{
6023+ int err, i, limit;
6024+ unsigned long n, sum;
6025+ struct dbgaufs_plink_arg *p;
6026+ struct au_sbinfo *sbinfo;
6027+ struct super_block *sb;
6028+ struct au_sphlhead *sphl;
6029+
6030+ err = -ENOMEM;
6031+ p = (void *)get_zeroed_page(GFP_NOFS);
6032+ if (unlikely(!p))
6033+ goto out;
6034+
6035+ err = -EFBIG;
6036+ sbinfo = inode->i_private;
6037+ sb = sbinfo->si_sb;
6038+ si_noflush_read_lock(sb);
6039+ if (au_opt_test(au_mntflags(sb), PLINK)) {
6040+ limit = PAGE_SIZE - sizeof(p->n);
6041+
6042+ /* the number of buckets */
6043+ n = snprintf(p->a + p->n, limit, "%d\n", AuPlink_NHASH);
6044+ p->n += n;
6045+ limit -= n;
6046+
6047+ sum = 0;
6048+ for (i = 0, sphl = sbinfo->si_plink;
6049+ i < AuPlink_NHASH;
6050+ i++, sphl++) {
6051+ n = au_sphl_count(sphl);
6052+ sum += n;
6053+
6054+ n = snprintf(p->a + p->n, limit, "%lu ", n);
6055+ p->n += n;
6056+ limit -= n;
6057+ if (unlikely(limit <= 0))
6058+ goto out_free;
6059+ }
6060+ p->a[p->n - 1] = '\n';
6061+
6062+ /* the sum of plinks */
6063+ n = snprintf(p->a + p->n, limit, "%lu\n", sum);
6064+ p->n += n;
6065+ limit -= n;
6066+ if (unlikely(limit <= 0))
6067+ goto out_free;
6068+ } else {
6069+#define str "1\n0\n0\n"
6070+ p->n = sizeof(str) - 1;
6071+ strcpy(p->a, str);
6072+#undef str
6073+ }
6074+ si_read_unlock(sb);
6075+
6076+ err = 0;
6077+ file->private_data = p;
6078+ goto out; /* success */
6079+
6080+out_free:
6081+ free_page((unsigned long)p);
6082+out:
6083+ return err;
6084+}
6085+
6086+static ssize_t dbgaufs_plink_read(struct file *file, char __user *buf,
6087+ size_t count, loff_t *ppos)
6088+{
6089+ struct dbgaufs_plink_arg *p;
6090+
6091+ p = file->private_data;
6092+ return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
6093+}
6094+
6095+static const struct file_operations dbgaufs_plink_fop = {
6096+ .owner = THIS_MODULE,
6097+ .open = dbgaufs_plink_open,
6098+ .release = dbgaufs_plink_release,
6099+ .read = dbgaufs_plink_read
6100+};
6101+
6102+/* ---------------------------------------------------------------------- */
6103+
1facf9fc 6104+static int dbgaufs_xib_open(struct inode *inode, struct file *file)
6105+{
6106+ int err;
6107+ struct au_sbinfo *sbinfo;
6108+ struct super_block *sb;
6109+
6110+ sbinfo = inode->i_private;
6111+ sb = sbinfo->si_sb;
6112+ si_noflush_read_lock(sb);
6113+ err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0);
6114+ si_read_unlock(sb);
6115+ return err;
6116+}
6117+
6118+static const struct file_operations dbgaufs_xib_fop = {
4a4d8108 6119+ .owner = THIS_MODULE,
1facf9fc 6120+ .open = dbgaufs_xib_open,
6121+ .release = dbgaufs_xi_release,
6122+ .read = dbgaufs_xi_read
6123+};
6124+
6125+/* ---------------------------------------------------------------------- */
6126+
6127+#define DbgaufsXi_PREFIX "xi"
6128+
6129+static int dbgaufs_xino_open(struct inode *inode, struct file *file)
6130+{
6131+ int err;
6132+ long l;
6133+ struct au_sbinfo *sbinfo;
6134+ struct super_block *sb;
6135+ struct file *xf;
6136+ struct qstr *name;
6137+
6138+ err = -ENOENT;
6139+ xf = NULL;
2000de60 6140+ name = &file->f_path.dentry->d_name;
1facf9fc 6141+ if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
6142+ || memcmp(name->name, DbgaufsXi_PREFIX,
6143+ sizeof(DbgaufsXi_PREFIX) - 1)))
6144+ goto out;
9dbd164d 6145+ err = kstrtol(name->name + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
1facf9fc 6146+ if (unlikely(err))
6147+ goto out;
6148+
6149+ sbinfo = inode->i_private;
6150+ sb = sbinfo->si_sb;
6151+ si_noflush_read_lock(sb);
6152+ if (l <= au_sbend(sb)) {
6153+ xf = au_sbr(sb, (aufs_bindex_t)l)->br_xino.xi_file;
6154+ err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1);
6155+ } else
6156+ err = -ENOENT;
6157+ si_read_unlock(sb);
6158+
4f0767ce 6159+out:
1facf9fc 6160+ return err;
6161+}
6162+
6163+static const struct file_operations dbgaufs_xino_fop = {
4a4d8108 6164+ .owner = THIS_MODULE,
1facf9fc 6165+ .open = dbgaufs_xino_open,
6166+ .release = dbgaufs_xi_release,
6167+ .read = dbgaufs_xi_read
6168+};
6169+
6170+void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
6171+{
6172+ aufs_bindex_t bend;
6173+ struct au_branch *br;
6174+ struct au_xino_file *xi;
6175+
6176+ if (!au_sbi(sb)->si_dbgaufs)
6177+ return;
6178+
6179+ bend = au_sbend(sb);
6180+ for (; bindex <= bend; bindex++) {
6181+ br = au_sbr(sb, bindex);
6182+ xi = &br->br_xino;
c06a8ce3
AM
6183+ debugfs_remove(xi->xi_dbgaufs);
6184+ xi->xi_dbgaufs = NULL;
1facf9fc 6185+ }
6186+}
6187+
6188+void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
6189+{
6190+ struct au_sbinfo *sbinfo;
6191+ struct dentry *parent;
6192+ struct au_branch *br;
6193+ struct au_xino_file *xi;
6194+ aufs_bindex_t bend;
6195+ char name[sizeof(DbgaufsXi_PREFIX) + 5]; /* "xi" bindex NULL */
6196+
6197+ sbinfo = au_sbi(sb);
6198+ parent = sbinfo->si_dbgaufs;
6199+ if (!parent)
6200+ return;
6201+
6202+ bend = au_sbend(sb);
6203+ for (; bindex <= bend; bindex++) {
6204+ snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
6205+ br = au_sbr(sb, bindex);
6206+ xi = &br->br_xino;
6207+ AuDebugOn(xi->xi_dbgaufs);
6208+ xi->xi_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
6209+ sbinfo, &dbgaufs_xino_fop);
6210+ /* ignore an error */
6211+ if (unlikely(!xi->xi_dbgaufs))
6212+ AuWarn1("failed %s under debugfs\n", name);
6213+ }
6214+}
6215+
6216+/* ---------------------------------------------------------------------- */
6217+
6218+#ifdef CONFIG_AUFS_EXPORT
6219+static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
6220+{
6221+ int err;
6222+ struct au_sbinfo *sbinfo;
6223+ struct super_block *sb;
6224+
6225+ sbinfo = inode->i_private;
6226+ sb = sbinfo->si_sb;
6227+ si_noflush_read_lock(sb);
6228+ err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0);
6229+ si_read_unlock(sb);
6230+ return err;
6231+}
6232+
6233+static const struct file_operations dbgaufs_xigen_fop = {
4a4d8108 6234+ .owner = THIS_MODULE,
1facf9fc 6235+ .open = dbgaufs_xigen_open,
6236+ .release = dbgaufs_xi_release,
6237+ .read = dbgaufs_xi_read
6238+};
6239+
6240+static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6241+{
6242+ int err;
6243+
dece6358 6244+ /*
c1595e42 6245+ * This function is a dynamic '__init' function actually,
dece6358
AM
6246+ * so the tiny check for si_rwsem is unnecessary.
6247+ */
6248+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6249+
1facf9fc 6250+ err = -EIO;
6251+ sbinfo->si_dbgaufs_xigen = debugfs_create_file
6252+ ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6253+ &dbgaufs_xigen_fop);
6254+ if (sbinfo->si_dbgaufs_xigen)
6255+ err = 0;
6256+
6257+ return err;
6258+}
6259+#else
6260+static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
6261+{
6262+ return 0;
6263+}
6264+#endif /* CONFIG_AUFS_EXPORT */
6265+
6266+/* ---------------------------------------------------------------------- */
6267+
6268+void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
6269+{
dece6358 6270+ /*
7e9cd9fe 6271+ * This function is a dynamic '__fin' function actually,
dece6358
AM
6272+ * so the tiny check for si_rwsem is unnecessary.
6273+ */
6274+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6275+
1facf9fc 6276+ debugfs_remove_recursive(sbinfo->si_dbgaufs);
6277+ sbinfo->si_dbgaufs = NULL;
6278+ kobject_put(&sbinfo->si_kobj);
6279+}
6280+
6281+int dbgaufs_si_init(struct au_sbinfo *sbinfo)
6282+{
6283+ int err;
6284+ char name[SysaufsSiNameLen];
6285+
dece6358 6286+ /*
c1595e42 6287+ * This function is a dynamic '__init' function actually,
dece6358
AM
6288+ * so the tiny check for si_rwsem is unnecessary.
6289+ */
6290+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
6291+
1facf9fc 6292+ err = -ENOENT;
6293+ if (!dbgaufs) {
6294+ AuErr1("/debug/aufs is uninitialized\n");
6295+ goto out;
6296+ }
6297+
6298+ err = -EIO;
6299+ sysaufs_name(sbinfo, name);
6300+ sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
6301+ if (unlikely(!sbinfo->si_dbgaufs))
6302+ goto out;
6303+ kobject_get(&sbinfo->si_kobj);
6304+
6305+ sbinfo->si_dbgaufs_xib = debugfs_create_file
6306+ ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6307+ &dbgaufs_xib_fop);
6308+ if (unlikely(!sbinfo->si_dbgaufs_xib))
6309+ goto out_dir;
6310+
86dc4139
AM
6311+ sbinfo->si_dbgaufs_plink = debugfs_create_file
6312+ ("plink", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
6313+ &dbgaufs_plink_fop);
6314+ if (unlikely(!sbinfo->si_dbgaufs_plink))
6315+ goto out_dir;
6316+
1facf9fc 6317+ err = dbgaufs_xigen_init(sbinfo);
6318+ if (!err)
6319+ goto out; /* success */
6320+
4f0767ce 6321+out_dir:
1facf9fc 6322+ dbgaufs_si_fin(sbinfo);
4f0767ce 6323+out:
1facf9fc 6324+ return err;
6325+}
6326+
6327+/* ---------------------------------------------------------------------- */
6328+
6329+void dbgaufs_fin(void)
6330+{
6331+ debugfs_remove(dbgaufs);
6332+}
6333+
6334+int __init dbgaufs_init(void)
6335+{
6336+ int err;
6337+
6338+ err = -EIO;
6339+ dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
6340+ if (dbgaufs)
6341+ err = 0;
6342+ return err;
6343+}
7f207e10
AM
6344diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
6345--- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 6346+++ linux/fs/aufs/dbgaufs.h 2015-09-24 10:47:58.248052907 +0200
523b37e3 6347@@ -0,0 +1,48 @@
1facf9fc 6348+/*
2000de60 6349+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 6350+ *
6351+ * This program, aufs is free software; you can redistribute it and/or modify
6352+ * it under the terms of the GNU General Public License as published by
6353+ * the Free Software Foundation; either version 2 of the License, or
6354+ * (at your option) any later version.
dece6358
AM
6355+ *
6356+ * This program is distributed in the hope that it will be useful,
6357+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6358+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6359+ * GNU General Public License for more details.
6360+ *
6361+ * You should have received a copy of the GNU General Public License
523b37e3 6362+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6363+ */
6364+
6365+/*
6366+ * debugfs interface
6367+ */
6368+
6369+#ifndef __DBGAUFS_H__
6370+#define __DBGAUFS_H__
6371+
6372+#ifdef __KERNEL__
6373+
dece6358 6374+struct super_block;
1facf9fc 6375+struct au_sbinfo;
dece6358 6376+
1facf9fc 6377+#ifdef CONFIG_DEBUG_FS
6378+/* dbgaufs.c */
6379+void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
6380+void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
6381+void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
6382+int dbgaufs_si_init(struct au_sbinfo *sbinfo);
6383+void dbgaufs_fin(void);
6384+int __init dbgaufs_init(void);
1facf9fc 6385+#else
4a4d8108
AM
6386+AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
6387+AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
6388+AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
6389+AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
6390+AuStubVoid(dbgaufs_fin, void)
6391+AuStubInt0(__init dbgaufs_init, void)
1facf9fc 6392+#endif /* CONFIG_DEBUG_FS */
6393+
6394+#endif /* __KERNEL__ */
6395+#endif /* __DBGAUFS_H__ */
7f207e10
AM
6396diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
6397--- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 6398+++ linux/fs/aufs/dcsub.c 2015-09-24 10:47:58.248052907 +0200
c1595e42 6399@@ -0,0 +1,224 @@
1facf9fc 6400+/*
2000de60 6401+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 6402+ *
6403+ * This program, aufs is free software; you can redistribute it and/or modify
6404+ * it under the terms of the GNU General Public License as published by
6405+ * the Free Software Foundation; either version 2 of the License, or
6406+ * (at your option) any later version.
dece6358
AM
6407+ *
6408+ * This program is distributed in the hope that it will be useful,
6409+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6410+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6411+ * GNU General Public License for more details.
6412+ *
6413+ * You should have received a copy of the GNU General Public License
523b37e3 6414+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6415+ */
6416+
6417+/*
6418+ * sub-routines for dentry cache
6419+ */
6420+
6421+#include "aufs.h"
6422+
6423+static void au_dpage_free(struct au_dpage *dpage)
6424+{
6425+ int i;
6426+ struct dentry **p;
6427+
6428+ p = dpage->dentries;
6429+ for (i = 0; i < dpage->ndentry; i++)
6430+ dput(*p++);
6431+ free_page((unsigned long)dpage->dentries);
6432+}
6433+
6434+int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
6435+{
6436+ int err;
6437+ void *p;
6438+
6439+ err = -ENOMEM;
6440+ dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
6441+ if (unlikely(!dpages->dpages))
6442+ goto out;
6443+
6444+ p = (void *)__get_free_page(gfp);
6445+ if (unlikely(!p))
6446+ goto out_dpages;
6447+
6448+ dpages->dpages[0].ndentry = 0;
6449+ dpages->dpages[0].dentries = p;
6450+ dpages->ndpage = 1;
6451+ return 0; /* success */
6452+
4f0767ce 6453+out_dpages:
1facf9fc 6454+ kfree(dpages->dpages);
4f0767ce 6455+out:
1facf9fc 6456+ return err;
6457+}
6458+
6459+void au_dpages_free(struct au_dcsub_pages *dpages)
6460+{
6461+ int i;
6462+ struct au_dpage *p;
6463+
6464+ p = dpages->dpages;
6465+ for (i = 0; i < dpages->ndpage; i++)
6466+ au_dpage_free(p++);
6467+ kfree(dpages->dpages);
6468+}
6469+
6470+static int au_dpages_append(struct au_dcsub_pages *dpages,
6471+ struct dentry *dentry, gfp_t gfp)
6472+{
6473+ int err, sz;
6474+ struct au_dpage *dpage;
6475+ void *p;
6476+
6477+ dpage = dpages->dpages + dpages->ndpage - 1;
6478+ sz = PAGE_SIZE / sizeof(dentry);
6479+ if (unlikely(dpage->ndentry >= sz)) {
6480+ AuLabel(new dpage);
6481+ err = -ENOMEM;
6482+ sz = dpages->ndpage * sizeof(*dpages->dpages);
6483+ p = au_kzrealloc(dpages->dpages, sz,
6484+ sz + sizeof(*dpages->dpages), gfp);
6485+ if (unlikely(!p))
6486+ goto out;
6487+
6488+ dpages->dpages = p;
6489+ dpage = dpages->dpages + dpages->ndpage;
6490+ p = (void *)__get_free_page(gfp);
6491+ if (unlikely(!p))
6492+ goto out;
6493+
6494+ dpage->ndentry = 0;
6495+ dpage->dentries = p;
6496+ dpages->ndpage++;
6497+ }
6498+
c1595e42 6499+ AuDebugOn(au_dcount(dentry) <= 0);
027c5e7a 6500+ dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
1facf9fc 6501+ return 0; /* success */
6502+
4f0767ce 6503+out:
1facf9fc 6504+ return err;
6505+}
6506+
c1595e42
JR
6507+/* todo: BAD approach */
6508+/* copied from linux/fs/dcache.c */
6509+enum d_walk_ret {
6510+ D_WALK_CONTINUE,
6511+ D_WALK_QUIT,
6512+ D_WALK_NORETRY,
6513+ D_WALK_SKIP,
6514+};
6515+
6516+extern void d_walk(struct dentry *parent, void *data,
6517+ enum d_walk_ret (*enter)(void *, struct dentry *),
6518+ void (*finish)(void *));
6519+
6520+struct ac_dpages_arg {
1facf9fc 6521+ int err;
c1595e42
JR
6522+ struct au_dcsub_pages *dpages;
6523+ struct super_block *sb;
6524+ au_dpages_test test;
6525+ void *arg;
6526+};
1facf9fc 6527+
c1595e42
JR
6528+static enum d_walk_ret au_call_dpages_append(void *_arg, struct dentry *dentry)
6529+{
6530+ enum d_walk_ret ret;
6531+ struct ac_dpages_arg *arg = _arg;
1facf9fc 6532+
c1595e42
JR
6533+ ret = D_WALK_CONTINUE;
6534+ if (dentry->d_sb == arg->sb
6535+ && !IS_ROOT(dentry)
6536+ && au_dcount(dentry) > 0
6537+ && au_di(dentry)
6538+ && (!arg->test || arg->test(dentry, arg->arg))) {
6539+ arg->err = au_dpages_append(arg->dpages, dentry, GFP_ATOMIC);
6540+ if (unlikely(arg->err))
6541+ ret = D_WALK_QUIT;
1facf9fc 6542+ }
6543+
c1595e42
JR
6544+ return ret;
6545+}
027c5e7a 6546+
c1595e42
JR
6547+int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6548+ au_dpages_test test, void *arg)
6549+{
6550+ struct ac_dpages_arg args = {
6551+ .err = 0,
6552+ .dpages = dpages,
6553+ .sb = root->d_sb,
6554+ .test = test,
6555+ .arg = arg
6556+ };
027c5e7a 6557+
c1595e42
JR
6558+ d_walk(root, &args, au_call_dpages_append, NULL);
6559+
6560+ return args.err;
1facf9fc 6561+}
6562+
6563+int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6564+ int do_include, au_dpages_test test, void *arg)
6565+{
6566+ int err;
6567+
6568+ err = 0;
027c5e7a
AM
6569+ write_seqlock(&rename_lock);
6570+ spin_lock(&dentry->d_lock);
6571+ if (do_include
c1595e42 6572+ && au_dcount(dentry) > 0
027c5e7a 6573+ && (!test || test(dentry, arg)))
1facf9fc 6574+ err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
027c5e7a
AM
6575+ spin_unlock(&dentry->d_lock);
6576+ if (unlikely(err))
6577+ goto out;
6578+
6579+ /*
523b37e3 6580+ * RCU for vfsmount is unnecessary since this is a traverse in a single
027c5e7a
AM
6581+ * mount
6582+ */
1facf9fc 6583+ while (!IS_ROOT(dentry)) {
027c5e7a
AM
6584+ dentry = dentry->d_parent; /* rename_lock is locked */
6585+ spin_lock(&dentry->d_lock);
c1595e42 6586+ if (au_dcount(dentry) > 0
027c5e7a 6587+ && (!test || test(dentry, arg)))
1facf9fc 6588+ err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
027c5e7a
AM
6589+ spin_unlock(&dentry->d_lock);
6590+ if (unlikely(err))
6591+ break;
1facf9fc 6592+ }
6593+
4f0767ce 6594+out:
027c5e7a 6595+ write_sequnlock(&rename_lock);
1facf9fc 6596+ return err;
6597+}
6598+
027c5e7a
AM
6599+static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
6600+{
6601+ return au_di(dentry) && dentry->d_sb == arg;
6602+}
6603+
6604+int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6605+ struct dentry *dentry, int do_include)
6606+{
6607+ return au_dcsub_pages_rev(dpages, dentry, do_include,
6608+ au_dcsub_dpages_aufs, dentry->d_sb);
6609+}
6610+
4a4d8108 6611+int au_test_subdir(struct dentry *d1, struct dentry *d2)
1facf9fc 6612+{
4a4d8108
AM
6613+ struct path path[2] = {
6614+ {
6615+ .dentry = d1
6616+ },
6617+ {
6618+ .dentry = d2
6619+ }
6620+ };
1facf9fc 6621+
4a4d8108 6622+ return path_is_under(path + 0, path + 1);
1facf9fc 6623+}
7f207e10
AM
6624diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
6625--- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 6626+++ linux/fs/aufs/dcsub.h 2015-09-24 10:47:58.251386326 +0200
5527c038 6627@@ -0,0 +1,136 @@
1facf9fc 6628+/*
2000de60 6629+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 6630+ *
6631+ * This program, aufs is free software; you can redistribute it and/or modify
6632+ * it under the terms of the GNU General Public License as published by
6633+ * the Free Software Foundation; either version 2 of the License, or
6634+ * (at your option) any later version.
dece6358
AM
6635+ *
6636+ * This program is distributed in the hope that it will be useful,
6637+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6638+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6639+ * GNU General Public License for more details.
6640+ *
6641+ * You should have received a copy of the GNU General Public License
523b37e3 6642+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6643+ */
6644+
6645+/*
6646+ * sub-routines for dentry cache
6647+ */
6648+
6649+#ifndef __AUFS_DCSUB_H__
6650+#define __AUFS_DCSUB_H__
6651+
6652+#ifdef __KERNEL__
6653+
7f207e10 6654+#include <linux/dcache.h>
027c5e7a 6655+#include <linux/fs.h>
dece6358 6656+
1facf9fc 6657+struct au_dpage {
6658+ int ndentry;
6659+ struct dentry **dentries;
6660+};
6661+
6662+struct au_dcsub_pages {
6663+ int ndpage;
6664+ struct au_dpage *dpages;
6665+};
6666+
6667+/* ---------------------------------------------------------------------- */
6668+
7f207e10 6669+/* dcsub.c */
1facf9fc 6670+int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
6671+void au_dpages_free(struct au_dcsub_pages *dpages);
6672+typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
6673+int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
6674+ au_dpages_test test, void *arg);
6675+int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
6676+ int do_include, au_dpages_test test, void *arg);
027c5e7a
AM
6677+int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
6678+ struct dentry *dentry, int do_include);
4a4d8108 6679+int au_test_subdir(struct dentry *d1, struct dentry *d2);
1facf9fc 6680+
7f207e10
AM
6681+/* ---------------------------------------------------------------------- */
6682+
523b37e3
AM
6683+/*
6684+ * todo: in linux-3.13, several similar (but faster) helpers are added to
6685+ * include/linux/dcache.h. Try them (in the future).
6686+ */
6687+
027c5e7a
AM
6688+static inline int au_d_hashed_positive(struct dentry *d)
6689+{
6690+ int err;
5527c038 6691+ struct inode *inode = d_inode(d);
076b876e 6692+
027c5e7a 6693+ err = 0;
5527c038
JR
6694+ if (unlikely(d_unhashed(d)
6695+ || d_is_negative(d)
6696+ || !inode->i_nlink))
027c5e7a
AM
6697+ err = -ENOENT;
6698+ return err;
6699+}
6700+
38d290e6
JR
6701+static inline int au_d_linkable(struct dentry *d)
6702+{
6703+ int err;
5527c038 6704+ struct inode *inode = d_inode(d);
076b876e 6705+
38d290e6
JR
6706+ err = au_d_hashed_positive(d);
6707+ if (err
5527c038 6708+ && d_is_positive(d)
38d290e6
JR
6709+ && (inode->i_state & I_LINKABLE))
6710+ err = 0;
6711+ return err;
6712+}
6713+
027c5e7a
AM
6714+static inline int au_d_alive(struct dentry *d)
6715+{
6716+ int err;
6717+ struct inode *inode;
076b876e 6718+
027c5e7a
AM
6719+ err = 0;
6720+ if (!IS_ROOT(d))
6721+ err = au_d_hashed_positive(d);
6722+ else {
5527c038
JR
6723+ inode = d_inode(d);
6724+ if (unlikely(d_unlinked(d)
6725+ || d_is_negative(d)
6726+ || !inode->i_nlink))
027c5e7a
AM
6727+ err = -ENOENT;
6728+ }
6729+ return err;
6730+}
6731+
6732+static inline int au_alive_dir(struct dentry *d)
7f207e10 6733+{
027c5e7a 6734+ int err;
076b876e 6735+
027c5e7a 6736+ err = au_d_alive(d);
5527c038 6737+ if (unlikely(err || IS_DEADDIR(d_inode(d))))
027c5e7a
AM
6738+ err = -ENOENT;
6739+ return err;
7f207e10
AM
6740+}
6741+
38d290e6
JR
6742+static inline int au_qstreq(struct qstr *a, struct qstr *b)
6743+{
6744+ return a->len == b->len
6745+ && !memcmp(a->name, b->name, a->len);
6746+}
6747+
7e9cd9fe
AM
6748+/*
6749+ * by the commit
6750+ * 360f547 2015-01-25 dcache: let the dentry count go down to zero without
6751+ * taking d_lock
6752+ * the type of d_lockref.count became int, but the inlined function d_count()
6753+ * still returns unsigned int.
6754+ * I don't know why. Maybe it is for every d_count() users?
6755+ * Anyway au_dcount() lives on.
6756+ */
c1595e42
JR
6757+static inline int au_dcount(struct dentry *d)
6758+{
6759+ return (int)d_count(d);
6760+}
6761+
1facf9fc 6762+#endif /* __KERNEL__ */
6763+#endif /* __AUFS_DCSUB_H__ */
7f207e10
AM
6764diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
6765--- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 6766+++ linux/fs/aufs/debug.c 2015-09-24 10:47:58.251386326 +0200
5527c038 6767@@ -0,0 +1,440 @@
1facf9fc 6768+/*
2000de60 6769+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 6770+ *
6771+ * This program, aufs is free software; you can redistribute it and/or modify
6772+ * it under the terms of the GNU General Public License as published by
6773+ * the Free Software Foundation; either version 2 of the License, or
6774+ * (at your option) any later version.
dece6358
AM
6775+ *
6776+ * This program is distributed in the hope that it will be useful,
6777+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
6778+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
6779+ * GNU General Public License for more details.
6780+ *
6781+ * You should have received a copy of the GNU General Public License
523b37e3 6782+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 6783+ */
6784+
6785+/*
6786+ * debug print functions
6787+ */
6788+
6789+#include "aufs.h"
6790+
392086de
AM
6791+/* Returns 0, or -errno. arg is in kp->arg. */
6792+static int param_atomic_t_set(const char *val, const struct kernel_param *kp)
6793+{
6794+ int err, n;
6795+
6796+ err = kstrtoint(val, 0, &n);
6797+ if (!err) {
6798+ if (n > 0)
6799+ au_debug_on();
6800+ else
6801+ au_debug_off();
6802+ }
6803+ return err;
6804+}
6805+
6806+/* Returns length written or -errno. Buffer is 4k (ie. be short!) */
6807+static int param_atomic_t_get(char *buffer, const struct kernel_param *kp)
6808+{
6809+ atomic_t *a;
6810+
6811+ a = kp->arg;
6812+ return sprintf(buffer, "%d", atomic_read(a));
6813+}
6814+
6815+static struct kernel_param_ops param_ops_atomic_t = {
6816+ .set = param_atomic_t_set,
6817+ .get = param_atomic_t_get
6818+ /* void (*free)(void *arg) */
6819+};
6820+
6821+atomic_t aufs_debug = ATOMIC_INIT(0);
1facf9fc 6822+MODULE_PARM_DESC(debug, "debug print");
392086de 6823+module_param_named(debug, aufs_debug, atomic_t, S_IRUGO | S_IWUSR | S_IWGRP);
1facf9fc 6824+
c1595e42 6825+DEFINE_MUTEX(au_dbg_mtx); /* just to serialize the dbg msgs */
1facf9fc 6826+char *au_plevel = KERN_DEBUG;
e49829fe
JR
6827+#define dpri(fmt, ...) do { \
6828+ if ((au_plevel \
6829+ && strcmp(au_plevel, KERN_DEBUG)) \
6830+ || au_debug_test()) \
6831+ printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
1facf9fc 6832+} while (0)
6833+
6834+/* ---------------------------------------------------------------------- */
6835+
6836+void au_dpri_whlist(struct au_nhash *whlist)
6837+{
6838+ unsigned long ul, n;
6839+ struct hlist_head *head;
c06a8ce3 6840+ struct au_vdir_wh *pos;
1facf9fc 6841+
6842+ n = whlist->nh_num;
6843+ head = whlist->nh_head;
6844+ for (ul = 0; ul < n; ul++) {
c06a8ce3 6845+ hlist_for_each_entry(pos, head, wh_hash)
1facf9fc 6846+ dpri("b%d, %.*s, %d\n",
c06a8ce3
AM
6847+ pos->wh_bindex,
6848+ pos->wh_str.len, pos->wh_str.name,
6849+ pos->wh_str.len);
1facf9fc 6850+ head++;
6851+ }
6852+}
6853+
6854+void au_dpri_vdir(struct au_vdir *vdir)
6855+{
6856+ unsigned long ul;
6857+ union au_vdir_deblk_p p;
6858+ unsigned char *o;
6859+
6860+ if (!vdir || IS_ERR(vdir)) {
6861+ dpri("err %ld\n", PTR_ERR(vdir));
6862+ return;
6863+ }
6864+
6865+ dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %lu\n",
6866+ vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
6867+ vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
6868+ for (ul = 0; ul < vdir->vd_nblk; ul++) {
6869+ p.deblk = vdir->vd_deblk[ul];
6870+ o = p.deblk;
6871+ dpri("[%lu]: %p\n", ul, o);
6872+ }
6873+}
6874+
53392da6 6875+static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
1facf9fc 6876+ struct dentry *wh)
6877+{
6878+ char *n = NULL;
6879+ int l = 0;
6880+
6881+ if (!inode || IS_ERR(inode)) {
6882+ dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
6883+ return -1;
6884+ }
6885+
c2b27bf2 6886+ /* the type of i_blocks depends upon CONFIG_LBDAF */
1facf9fc 6887+ BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
6888+ && sizeof(inode->i_blocks) != sizeof(u64));
6889+ if (wh) {
6890+ n = (void *)wh->d_name.name;
6891+ l = wh->d_name.len;
6892+ }
6893+
53392da6
AM
6894+ dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
6895+ " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
6896+ bindex, inode,
1facf9fc 6897+ inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
6898+ atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
6899+ i_size_read(inode), (unsigned long long)inode->i_blocks,
53392da6 6900+ hn, (long long)timespec_to_ns(&inode->i_ctime) & 0x0ffff,
1facf9fc 6901+ inode->i_mapping ? inode->i_mapping->nrpages : 0,
b752ccd1
AM
6902+ inode->i_state, inode->i_flags, inode->i_version,
6903+ inode->i_generation,
1facf9fc 6904+ l ? ", wh " : "", l, n);
6905+ return 0;
6906+}
6907+
6908+void au_dpri_inode(struct inode *inode)
6909+{
6910+ struct au_iinfo *iinfo;
6911+ aufs_bindex_t bindex;
53392da6 6912+ int err, hn;
1facf9fc 6913+
53392da6 6914+ err = do_pri_inode(-1, inode, -1, NULL);
1facf9fc 6915+ if (err || !au_test_aufs(inode->i_sb))
6916+ return;
6917+
6918+ iinfo = au_ii(inode);
6919+ if (!iinfo)
6920+ return;
6921+ dpri("i-1: bstart %d, bend %d, gen %d\n",
537831f9 6922+ iinfo->ii_bstart, iinfo->ii_bend, au_iigen(inode, NULL));
1facf9fc 6923+ if (iinfo->ii_bstart < 0)
6924+ return;
53392da6
AM
6925+ hn = 0;
6926+ for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; bindex++) {
6927+ hn = !!au_hn(iinfo->ii_hinode + bindex);
6928+ do_pri_inode(bindex, iinfo->ii_hinode[0 + bindex].hi_inode, hn,
1facf9fc 6929+ iinfo->ii_hinode[0 + bindex].hi_whdentry);
53392da6 6930+ }
1facf9fc 6931+}
6932+
2cbb1c4b
JR
6933+void au_dpri_dalias(struct inode *inode)
6934+{
6935+ struct dentry *d;
6936+
6937+ spin_lock(&inode->i_lock);
c1595e42 6938+ hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias)
2cbb1c4b
JR
6939+ au_dpri_dentry(d);
6940+ spin_unlock(&inode->i_lock);
6941+}
6942+
1facf9fc 6943+static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
6944+{
6945+ struct dentry *wh = NULL;
53392da6 6946+ int hn;
076b876e 6947+ struct au_iinfo *iinfo;
1facf9fc 6948+
6949+ if (!dentry || IS_ERR(dentry)) {
6950+ dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
6951+ return -1;
6952+ }
6953+ /* do not call dget_parent() here */
027c5e7a 6954+ /* note: access d_xxx without d_lock */
523b37e3
AM
6955+ dpri("d%d: %p, %pd2?, %s, cnt %d, flags 0x%x, %shashed\n",
6956+ bindex, dentry, dentry,
1facf9fc 6957+ dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
c1595e42 6958+ au_dcount(dentry), dentry->d_flags,
523b37e3 6959+ d_unhashed(dentry) ? "un" : "");
53392da6 6960+ hn = -1;
5527c038
JR
6961+ if (bindex >= 0
6962+ && d_is_positive(dentry)
6963+ && au_test_aufs(dentry->d_sb)) {
6964+ iinfo = au_ii(d_inode(dentry));
53392da6
AM
6965+ if (iinfo) {
6966+ hn = !!au_hn(iinfo->ii_hinode + bindex);
1facf9fc 6967+ wh = iinfo->ii_hinode[0 + bindex].hi_whdentry;
53392da6 6968+ }
1facf9fc 6969+ }
5527c038 6970+ do_pri_inode(bindex, d_inode(dentry), hn, wh);
1facf9fc 6971+ return 0;
6972+}
6973+
6974+void au_dpri_dentry(struct dentry *dentry)
6975+{
6976+ struct au_dinfo *dinfo;
6977+ aufs_bindex_t bindex;
6978+ int err;
4a4d8108 6979+ struct au_hdentry *hdp;
1facf9fc 6980+
6981+ err = do_pri_dentry(-1, dentry);
6982+ if (err || !au_test_aufs(dentry->d_sb))
6983+ return;
6984+
6985+ dinfo = au_di(dentry);
6986+ if (!dinfo)
6987+ return;
38d290e6 6988+ dpri("d-1: bstart %d, bend %d, bwh %d, bdiropq %d, gen %d, tmp %d\n",
1facf9fc 6989+ dinfo->di_bstart, dinfo->di_bend,
38d290e6
JR
6990+ dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry),
6991+ dinfo->di_tmpfile);
1facf9fc 6992+ if (dinfo->di_bstart < 0)
6993+ return;
4a4d8108 6994+ hdp = dinfo->di_hdentry;
1facf9fc 6995+ for (bindex = dinfo->di_bstart; bindex <= dinfo->di_bend; bindex++)
4a4d8108 6996+ do_pri_dentry(bindex, hdp[0 + bindex].hd_dentry);
1facf9fc 6997+}
6998+
6999+static int do_pri_file(aufs_bindex_t bindex, struct file *file)
7000+{
7001+ char a[32];
7002+
7003+ if (!file || IS_ERR(file)) {
7004+ dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
7005+ return -1;
7006+ }
7007+ a[0] = 0;
7008+ if (bindex < 0
b912730e 7009+ && !IS_ERR_OR_NULL(file->f_path.dentry)
2000de60 7010+ && au_test_aufs(file->f_path.dentry->d_sb)
1facf9fc 7011+ && au_fi(file))
e49829fe 7012+ snprintf(a, sizeof(a), ", gen %d, mmapped %d",
2cbb1c4b 7013+ au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
b752ccd1 7014+ dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
1facf9fc 7015+ bindex, file->f_mode, file->f_flags, (long)file_count(file),
b752ccd1 7016+ file->f_version, file->f_pos, a);
b912730e 7017+ if (!IS_ERR_OR_NULL(file->f_path.dentry))
2000de60 7018+ do_pri_dentry(bindex, file->f_path.dentry);
1facf9fc 7019+ return 0;
7020+}
7021+
7022+void au_dpri_file(struct file *file)
7023+{
7024+ struct au_finfo *finfo;
4a4d8108
AM
7025+ struct au_fidir *fidir;
7026+ struct au_hfile *hfile;
1facf9fc 7027+ aufs_bindex_t bindex;
7028+ int err;
7029+
7030+ err = do_pri_file(-1, file);
2000de60 7031+ if (err
b912730e 7032+ || IS_ERR_OR_NULL(file->f_path.dentry)
2000de60 7033+ || !au_test_aufs(file->f_path.dentry->d_sb))
1facf9fc 7034+ return;
7035+
7036+ finfo = au_fi(file);
7037+ if (!finfo)
7038+ return;
4a4d8108 7039+ if (finfo->fi_btop < 0)
1facf9fc 7040+ return;
4a4d8108
AM
7041+ fidir = finfo->fi_hdir;
7042+ if (!fidir)
7043+ do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
7044+ else
e49829fe
JR
7045+ for (bindex = finfo->fi_btop;
7046+ bindex >= 0 && bindex <= fidir->fd_bbot;
4a4d8108
AM
7047+ bindex++) {
7048+ hfile = fidir->fd_hfile + bindex;
7049+ do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
7050+ }
1facf9fc 7051+}
7052+
7053+static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
7054+{
7055+ struct vfsmount *mnt;
7056+ struct super_block *sb;
7057+
7058+ if (!br || IS_ERR(br))
7059+ goto out;
86dc4139 7060+ mnt = au_br_mnt(br);
1facf9fc 7061+ if (!mnt || IS_ERR(mnt))
7062+ goto out;
7063+ sb = mnt->mnt_sb;
7064+ if (!sb || IS_ERR(sb))
7065+ goto out;
7066+
1e00d052 7067+ dpri("s%d: {perm 0x%x, id %d, cnt %d, wbr %p}, "
b752ccd1 7068+ "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
1facf9fc 7069+ "xino %d\n",
1e00d052
AM
7070+ bindex, br->br_perm, br->br_id, atomic_read(&br->br_count),
7071+ br->br_wbr, au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
b752ccd1 7072+ sb->s_flags, sb->s_count,
1facf9fc 7073+ atomic_read(&sb->s_active), !!br->br_xino.xi_file);
7074+ return 0;
7075+
4f0767ce 7076+out:
1facf9fc 7077+ dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
7078+ return -1;
7079+}
7080+
7081+void au_dpri_sb(struct super_block *sb)
7082+{
7083+ struct au_sbinfo *sbinfo;
7084+ aufs_bindex_t bindex;
7085+ int err;
7086+ /* to reuduce stack size */
7087+ struct {
7088+ struct vfsmount mnt;
7089+ struct au_branch fake;
7090+ } *a;
7091+
7092+ /* this function can be called from magic sysrq */
7093+ a = kzalloc(sizeof(*a), GFP_ATOMIC);
7094+ if (unlikely(!a)) {
7095+ dpri("no memory\n");
7096+ return;
7097+ }
7098+
7099+ a->mnt.mnt_sb = sb;
7100+ a->fake.br_perm = 0;
86dc4139 7101+ a->fake.br_path.mnt = &a->mnt;
1facf9fc 7102+ a->fake.br_xino.xi_file = NULL;
7103+ atomic_set(&a->fake.br_count, 0);
7104+ smp_mb(); /* atomic_set */
7105+ err = do_pri_br(-1, &a->fake);
7106+ kfree(a);
7107+ dpri("dev 0x%x\n", sb->s_dev);
7108+ if (err || !au_test_aufs(sb))
7109+ return;
7110+
7111+ sbinfo = au_sbi(sb);
7112+ if (!sbinfo)
7113+ return;
7114+ dpri("nw %d, gen %u, kobj %d\n",
7115+ atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
7116+ atomic_read(&sbinfo->si_kobj.kref.refcount));
7117+ for (bindex = 0; bindex <= sbinfo->si_bend; bindex++)
7118+ do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
7119+}
7120+
7121+/* ---------------------------------------------------------------------- */
7122+
027c5e7a
AM
7123+void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
7124+{
5527c038 7125+ struct inode *h_inode, *inode = d_inode(dentry);
027c5e7a
AM
7126+ struct dentry *h_dentry;
7127+ aufs_bindex_t bindex, bend, bi;
7128+
7129+ if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
7130+ return;
7131+
7132+ bend = au_dbend(dentry);
7133+ bi = au_ibend(inode);
7134+ if (bi < bend)
7135+ bend = bi;
7136+ bindex = au_dbstart(dentry);
7137+ bi = au_ibstart(inode);
7138+ if (bi > bindex)
7139+ bindex = bi;
7140+
7141+ for (; bindex <= bend; bindex++) {
7142+ h_dentry = au_h_dptr(dentry, bindex);
7143+ if (!h_dentry)
7144+ continue;
7145+ h_inode = au_h_iptr(inode, bindex);
5527c038 7146+ if (unlikely(h_inode != d_inode(h_dentry))) {
392086de 7147+ au_debug_on();
027c5e7a
AM
7148+ AuDbg("b%d, %s:%d\n", bindex, func, line);
7149+ AuDbgDentry(dentry);
7150+ AuDbgInode(inode);
392086de 7151+ au_debug_off();
027c5e7a
AM
7152+ BUG();
7153+ }
7154+ }
7155+}
7156+
1facf9fc 7157+void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
7158+{
7159+ int err, i, j;
7160+ struct au_dcsub_pages dpages;
7161+ struct au_dpage *dpage;
7162+ struct dentry **dentries;
7163+
7164+ err = au_dpages_init(&dpages, GFP_NOFS);
7165+ AuDebugOn(err);
027c5e7a 7166+ err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
1facf9fc 7167+ AuDebugOn(err);
7168+ for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
7169+ dpage = dpages.dpages + i;
7170+ dentries = dpage->dentries;
7171+ for (j = dpage->ndentry - 1; !err && j >= 0; j--)
027c5e7a 7172+ AuDebugOn(au_digen_test(dentries[j], sigen));
1facf9fc 7173+ }
7174+ au_dpages_free(&dpages);
7175+}
7176+
1facf9fc 7177+void au_dbg_verify_kthread(void)
7178+{
53392da6 7179+ if (au_wkq_test()) {
1facf9fc 7180+ au_dbg_blocked();
1e00d052
AM
7181+ /*
7182+ * It may be recursive, but udba=notify between two aufs mounts,
7183+ * where a single ro branch is shared, is not a problem.
7184+ */
7185+ /* WARN_ON(1); */
1facf9fc 7186+ }
7187+}
7188+
7189+/* ---------------------------------------------------------------------- */
7190+
1facf9fc 7191+int __init au_debug_init(void)
7192+{
7193+ aufs_bindex_t bindex;
7194+ struct au_vdir_destr destr;
7195+
7196+ bindex = -1;
7197+ AuDebugOn(bindex >= 0);
7198+
7199+ destr.len = -1;
7200+ AuDebugOn(destr.len < NAME_MAX);
7201+
7202+#ifdef CONFIG_4KSTACKS
0c3ec466 7203+ pr_warn("CONFIG_4KSTACKS is defined.\n");
1facf9fc 7204+#endif
7205+
1facf9fc 7206+ return 0;
7207+}
7f207e10
AM
7208diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
7209--- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 7210+++ linux/fs/aufs/debug.h 2015-09-24 10:47:58.251386326 +0200
5527c038 7211@@ -0,0 +1,225 @@
1facf9fc 7212+/*
2000de60 7213+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 7214+ *
7215+ * This program, aufs is free software; you can redistribute it and/or modify
7216+ * it under the terms of the GNU General Public License as published by
7217+ * the Free Software Foundation; either version 2 of the License, or
7218+ * (at your option) any later version.
dece6358
AM
7219+ *
7220+ * This program is distributed in the hope that it will be useful,
7221+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
7222+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7223+ * GNU General Public License for more details.
7224+ *
7225+ * You should have received a copy of the GNU General Public License
523b37e3 7226+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 7227+ */
7228+
7229+/*
7230+ * debug print functions
7231+ */
7232+
7233+#ifndef __AUFS_DEBUG_H__
7234+#define __AUFS_DEBUG_H__
7235+
7236+#ifdef __KERNEL__
7237+
392086de 7238+#include <linux/atomic.h>
4a4d8108
AM
7239+#include <linux/module.h>
7240+#include <linux/kallsyms.h>
1facf9fc 7241+#include <linux/sysrq.h>
4a4d8108 7242+
1facf9fc 7243+#ifdef CONFIG_AUFS_DEBUG
7244+#define AuDebugOn(a) BUG_ON(a)
7245+
7246+/* module parameter */
392086de
AM
7247+extern atomic_t aufs_debug;
7248+static inline void au_debug_on(void)
1facf9fc 7249+{
392086de
AM
7250+ atomic_inc(&aufs_debug);
7251+}
7252+static inline void au_debug_off(void)
7253+{
7254+ atomic_dec_if_positive(&aufs_debug);
1facf9fc 7255+}
7256+
7257+static inline int au_debug_test(void)
7258+{
392086de 7259+ return atomic_read(&aufs_debug) > 0;
1facf9fc 7260+}
7261+#else
7262+#define AuDebugOn(a) do {} while (0)
392086de
AM
7263+AuStubVoid(au_debug_on, void)
7264+AuStubVoid(au_debug_off, void)
4a4d8108 7265+AuStubInt0(au_debug_test, void)
1facf9fc 7266+#endif /* CONFIG_AUFS_DEBUG */
7267+
392086de
AM
7268+#define param_check_atomic_t(name, p) __param_check(name, p, atomic_t)
7269+
1facf9fc 7270+/* ---------------------------------------------------------------------- */
7271+
7272+/* debug print */
7273+
4a4d8108 7274+#define AuDbg(fmt, ...) do { \
1facf9fc 7275+ if (au_debug_test()) \
4a4d8108 7276+ pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
1facf9fc 7277+} while (0)
4a4d8108
AM
7278+#define AuLabel(l) AuDbg(#l "\n")
7279+#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
7280+#define AuWarn1(fmt, ...) do { \
1facf9fc 7281+ static unsigned char _c; \
7282+ if (!_c++) \
0c3ec466 7283+ pr_warn(fmt, ##__VA_ARGS__); \
1facf9fc 7284+} while (0)
7285+
4a4d8108 7286+#define AuErr1(fmt, ...) do { \
1facf9fc 7287+ static unsigned char _c; \
7288+ if (!_c++) \
4a4d8108 7289+ pr_err(fmt, ##__VA_ARGS__); \
1facf9fc 7290+} while (0)
7291+
4a4d8108 7292+#define AuIOErr1(fmt, ...) do { \
1facf9fc 7293+ static unsigned char _c; \
7294+ if (!_c++) \
4a4d8108 7295+ AuIOErr(fmt, ##__VA_ARGS__); \
1facf9fc 7296+} while (0)
7297+
7298+#define AuUnsupportMsg "This operation is not supported." \
7299+ " Please report this application to aufs-users ML."
4a4d8108
AM
7300+#define AuUnsupport(fmt, ...) do { \
7301+ pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
1facf9fc 7302+ dump_stack(); \
7303+} while (0)
7304+
7305+#define AuTraceErr(e) do { \
7306+ if (unlikely((e) < 0)) \
7307+ AuDbg("err %d\n", (int)(e)); \
7308+} while (0)
7309+
7310+#define AuTraceErrPtr(p) do { \
7311+ if (IS_ERR(p)) \
7312+ AuDbg("err %ld\n", PTR_ERR(p)); \
7313+} while (0)
7314+
7315+/* dirty macros for debug print, use with "%.*s" and caution */
7316+#define AuLNPair(qstr) (qstr)->len, (qstr)->name
1facf9fc 7317+
7318+/* ---------------------------------------------------------------------- */
7319+
dece6358 7320+struct dentry;
1facf9fc 7321+#ifdef CONFIG_AUFS_DEBUG
c1595e42 7322+extern struct mutex au_dbg_mtx;
1facf9fc 7323+extern char *au_plevel;
7324+struct au_nhash;
7325+void au_dpri_whlist(struct au_nhash *whlist);
7326+struct au_vdir;
7327+void au_dpri_vdir(struct au_vdir *vdir);
dece6358 7328+struct inode;
1facf9fc 7329+void au_dpri_inode(struct inode *inode);
2cbb1c4b 7330+void au_dpri_dalias(struct inode *inode);
1facf9fc 7331+void au_dpri_dentry(struct dentry *dentry);
dece6358 7332+struct file;
1facf9fc 7333+void au_dpri_file(struct file *filp);
dece6358 7334+struct super_block;
1facf9fc 7335+void au_dpri_sb(struct super_block *sb);
7336+
027c5e7a
AM
7337+#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
7338+void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
1facf9fc 7339+void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
1facf9fc 7340+void au_dbg_verify_kthread(void);
7341+
7342+int __init au_debug_init(void);
7e9cd9fe 7343+
1facf9fc 7344+#define AuDbgWhlist(w) do { \
c1595e42 7345+ mutex_lock(&au_dbg_mtx); \
1facf9fc 7346+ AuDbg(#w "\n"); \
7347+ au_dpri_whlist(w); \
c1595e42 7348+ mutex_unlock(&au_dbg_mtx); \
1facf9fc 7349+} while (0)
7350+
7351+#define AuDbgVdir(v) do { \
c1595e42 7352+ mutex_lock(&au_dbg_mtx); \
1facf9fc 7353+ AuDbg(#v "\n"); \
7354+ au_dpri_vdir(v); \
c1595e42 7355+ mutex_unlock(&au_dbg_mtx); \
1facf9fc 7356+} while (0)
7357+
7358+#define AuDbgInode(i) do { \
c1595e42 7359+ mutex_lock(&au_dbg_mtx); \
1facf9fc 7360+ AuDbg(#i "\n"); \
7361+ au_dpri_inode(i); \
c1595e42 7362+ mutex_unlock(&au_dbg_mtx); \
1facf9fc 7363+} while (0)
7364+
2cbb1c4b 7365+#define AuDbgDAlias(i) do { \
c1595e42 7366+ mutex_lock(&au_dbg_mtx); \
2cbb1c4b
JR
7367+ AuDbg(#i "\n"); \
7368+ au_dpri_dalias(i); \
c1595e42 7369+ mutex_unlock(&au_dbg_mtx); \
2cbb1c4b
JR
7370+} while (0)
7371+
1facf9fc 7372+#define AuDbgDentry(d) do { \
c1595e42 7373+ mutex_lock(&au_dbg_mtx); \
1facf9fc 7374+ AuDbg(#d "\n"); \
7375+ au_dpri_dentry(d); \
c1595e42 7376+ mutex_unlock(&au_dbg_mtx); \
1facf9fc 7377+} while (0)
7378+
7379+#define AuDbgFile(f) do { \
c1595e42 7380+ mutex_lock(&au_dbg_mtx); \
1facf9fc 7381+ AuDbg(#f "\n"); \
7382+ au_dpri_file(f); \
c1595e42 7383+ mutex_unlock(&au_dbg_mtx); \
1facf9fc 7384+} while (0)
7385+
7386+#define AuDbgSb(sb) do { \
c1595e42 7387+ mutex_lock(&au_dbg_mtx); \
1facf9fc 7388+ AuDbg(#sb "\n"); \
7389+ au_dpri_sb(sb); \
c1595e42 7390+ mutex_unlock(&au_dbg_mtx); \
1facf9fc 7391+} while (0)
7392+
4a4d8108
AM
7393+#define AuDbgSym(addr) do { \
7394+ char sym[KSYM_SYMBOL_LEN]; \
7395+ sprint_symbol(sym, (unsigned long)addr); \
7396+ AuDbg("%s\n", sym); \
7397+} while (0)
1facf9fc 7398+#else
027c5e7a 7399+AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
4a4d8108
AM
7400+AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
7401+AuStubVoid(au_dbg_verify_kthread, void)
7402+AuStubInt0(__init au_debug_init, void)
1facf9fc 7403+
1facf9fc 7404+#define AuDbgWhlist(w) do {} while (0)
7405+#define AuDbgVdir(v) do {} while (0)
7406+#define AuDbgInode(i) do {} while (0)
2cbb1c4b 7407+#define AuDbgDAlias(i) do {} while (0)
1facf9fc 7408+#define AuDbgDentry(d) do {} while (0)
7409+#define AuDbgFile(f) do {} while (0)
7410+#define AuDbgSb(sb) do {} while (0)
4a4d8108 7411+#define AuDbgSym(addr) do {} while (0)
1facf9fc 7412+#endif /* CONFIG_AUFS_DEBUG */
7413+
7414+/* ---------------------------------------------------------------------- */
7415+
7416+#ifdef CONFIG_AUFS_MAGIC_SYSRQ
7417+int __init au_sysrq_init(void);
7418+void au_sysrq_fin(void);
7419+
7420+#ifdef CONFIG_HW_CONSOLE
7421+#define au_dbg_blocked() do { \
7422+ WARN_ON(1); \
0c5527e5 7423+ handle_sysrq('w'); \
1facf9fc 7424+} while (0)
7425+#else
4a4d8108 7426+AuStubVoid(au_dbg_blocked, void)
1facf9fc 7427+#endif
7428+
7429+#else
4a4d8108
AM
7430+AuStubInt0(__init au_sysrq_init, void)
7431+AuStubVoid(au_sysrq_fin, void)
7432+AuStubVoid(au_dbg_blocked, void)
1facf9fc 7433+#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
7434+
7435+#endif /* __KERNEL__ */
7436+#endif /* __AUFS_DEBUG_H__ */
7f207e10
AM
7437diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
7438--- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 7439+++ linux/fs/aufs/dentry.c 2015-09-24 10:47:58.251386326 +0200
5527c038 7440@@ -0,0 +1,1105 @@
1facf9fc 7441+/*
2000de60 7442+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 7443+ *
7444+ * This program, aufs is free software; you can redistribute it and/or modify
7445+ * it under the terms of the GNU General Public License as published by
7446+ * the Free Software Foundation; either version 2 of the License, or
7447+ * (at your option) any later version.
dece6358
AM
7448+ *
7449+ * This program is distributed in the hope that it will be useful,
7450+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
7451+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7452+ * GNU General Public License for more details.
7453+ *
7454+ * You should have received a copy of the GNU General Public License
523b37e3 7455+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 7456+ */
7457+
7458+/*
7459+ * lookup and dentry operations
7460+ */
7461+
dece6358 7462+#include <linux/namei.h>
1facf9fc 7463+#include "aufs.h"
7464+
1facf9fc 7465+#define AuLkup_ALLOW_NEG 1
076b876e 7466+#define AuLkup_IGNORE_PERM (1 << 1)
1facf9fc 7467+#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
7f207e10
AM
7468+#define au_fset_lkup(flags, name) \
7469+ do { (flags) |= AuLkup_##name; } while (0)
7470+#define au_fclr_lkup(flags, name) \
7471+ do { (flags) &= ~AuLkup_##name; } while (0)
1facf9fc 7472+
7473+struct au_do_lookup_args {
7474+ unsigned int flags;
7475+ mode_t type;
1facf9fc 7476+};
7477+
7478+/*
7479+ * returns positive/negative dentry, NULL or an error.
7480+ * NULL means whiteout-ed or not-found.
7481+ */
7482+static struct dentry*
7483+au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
7484+ aufs_bindex_t bindex, struct qstr *wh_name,
7485+ struct au_do_lookup_args *args)
7486+{
7487+ struct dentry *h_dentry;
2000de60 7488+ struct inode *h_inode;
1facf9fc 7489+ struct au_branch *br;
7490+ int wh_found, opq;
7491+ unsigned char wh_able;
7492+ const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
076b876e
AM
7493+ const unsigned char ignore_perm = !!au_ftest_lkup(args->flags,
7494+ IGNORE_PERM);
1facf9fc 7495+
1facf9fc 7496+ wh_found = 0;
7497+ br = au_sbr(dentry->d_sb, bindex);
7498+ wh_able = !!au_br_whable(br->br_perm);
7499+ if (wh_able)
076b876e 7500+ wh_found = au_wh_test(h_parent, wh_name, /*try_sio*/0);
1facf9fc 7501+ h_dentry = ERR_PTR(wh_found);
7502+ if (!wh_found)
7503+ goto real_lookup;
7504+ if (unlikely(wh_found < 0))
7505+ goto out;
7506+
7507+ /* We found a whiteout */
7508+ /* au_set_dbend(dentry, bindex); */
7509+ au_set_dbwh(dentry, bindex);
7510+ if (!allow_neg)
7511+ return NULL; /* success */
7512+
4f0767ce 7513+real_lookup:
076b876e
AM
7514+ if (!ignore_perm)
7515+ h_dentry = vfsub_lkup_one(&dentry->d_name, h_parent);
7516+ else
7517+ h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
2000de60
JR
7518+ if (IS_ERR(h_dentry)) {
7519+ if (PTR_ERR(h_dentry) == -ENAMETOOLONG
7520+ && !allow_neg)
7521+ h_dentry = NULL;
1facf9fc 7522+ goto out;
2000de60 7523+ }
1facf9fc 7524+
5527c038
JR
7525+ h_inode = d_inode(h_dentry);
7526+ if (d_is_negative(h_dentry)) {
1facf9fc 7527+ if (!allow_neg)
7528+ goto out_neg;
7529+ } else if (wh_found
7530+ || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
7531+ goto out_neg;
7532+
7533+ if (au_dbend(dentry) <= bindex)
7534+ au_set_dbend(dentry, bindex);
7535+ if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
7536+ au_set_dbstart(dentry, bindex);
7537+ au_set_h_dptr(dentry, bindex, h_dentry);
7538+
2000de60
JR
7539+ if (!d_is_dir(h_dentry)
7540+ || !wh_able
5527c038 7541+ || (d_really_is_positive(dentry) && !d_is_dir(dentry)))
1facf9fc 7542+ goto out; /* success */
7543+
7544+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
076b876e 7545+ opq = au_diropq_test(h_dentry);
1facf9fc 7546+ mutex_unlock(&h_inode->i_mutex);
7547+ if (opq > 0)
7548+ au_set_dbdiropq(dentry, bindex);
7549+ else if (unlikely(opq < 0)) {
7550+ au_set_h_dptr(dentry, bindex, NULL);
7551+ h_dentry = ERR_PTR(opq);
7552+ }
7553+ goto out;
7554+
4f0767ce 7555+out_neg:
1facf9fc 7556+ dput(h_dentry);
7557+ h_dentry = NULL;
4f0767ce 7558+out:
1facf9fc 7559+ return h_dentry;
7560+}
7561+
dece6358
AM
7562+static int au_test_shwh(struct super_block *sb, const struct qstr *name)
7563+{
7564+ if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
7565+ && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
7566+ return -EPERM;
7567+ return 0;
7568+}
7569+
1facf9fc 7570+/*
7571+ * returns the number of lower positive dentries,
7572+ * otherwise an error.
7573+ * can be called at unlinking with @type is zero.
7574+ */
537831f9 7575+int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type)
1facf9fc 7576+{
7577+ int npositive, err;
7578+ aufs_bindex_t bindex, btail, bdiropq;
076b876e 7579+ unsigned char isdir, dirperm1;
1facf9fc 7580+ struct qstr whname;
7581+ struct au_do_lookup_args args = {
b4510431 7582+ .flags = 0,
537831f9 7583+ .type = type
1facf9fc 7584+ };
7585+ const struct qstr *name = &dentry->d_name;
7586+ struct dentry *parent;
076b876e 7587+ struct super_block *sb;
1facf9fc 7588+
076b876e
AM
7589+ sb = dentry->d_sb;
7590+ err = au_test_shwh(sb, name);
dece6358 7591+ if (unlikely(err))
1facf9fc 7592+ goto out;
7593+
7594+ err = au_wh_name_alloc(&whname, name);
7595+ if (unlikely(err))
7596+ goto out;
7597+
2000de60 7598+ isdir = !!d_is_dir(dentry);
1facf9fc 7599+ if (!type)
7600+ au_fset_lkup(args.flags, ALLOW_NEG);
076b876e 7601+ dirperm1 = !!au_opt_test(au_mntflags(sb), DIRPERM1);
1facf9fc 7602+
7603+ npositive = 0;
4a4d8108 7604+ parent = dget_parent(dentry);
1facf9fc 7605+ btail = au_dbtaildir(parent);
7606+ for (bindex = bstart; bindex <= btail; bindex++) {
7607+ struct dentry *h_parent, *h_dentry;
7608+ struct inode *h_inode, *h_dir;
7609+
7610+ h_dentry = au_h_dptr(dentry, bindex);
7611+ if (h_dentry) {
5527c038 7612+ if (d_is_positive(h_dentry))
1facf9fc 7613+ npositive++;
7614+ if (type != S_IFDIR)
7615+ break;
7616+ continue;
7617+ }
7618+ h_parent = au_h_dptr(parent, bindex);
2000de60 7619+ if (!h_parent || !d_is_dir(h_parent))
1facf9fc 7620+ continue;
7621+
5527c038 7622+ h_dir = d_inode(h_parent);
1facf9fc 7623+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
7624+ h_dentry = au_do_lookup(h_parent, dentry, bindex, &whname,
7625+ &args);
7626+ mutex_unlock(&h_dir->i_mutex);
7627+ err = PTR_ERR(h_dentry);
7628+ if (IS_ERR(h_dentry))
4a4d8108 7629+ goto out_parent;
2000de60
JR
7630+ if (h_dentry)
7631+ au_fclr_lkup(args.flags, ALLOW_NEG);
076b876e
AM
7632+ if (dirperm1)
7633+ au_fset_lkup(args.flags, IGNORE_PERM);
1facf9fc 7634+
7635+ if (au_dbwh(dentry) >= 0)
7636+ break;
7637+ if (!h_dentry)
7638+ continue;
5527c038 7639+ if (d_is_negative(h_dentry))
1facf9fc 7640+ continue;
5527c038 7641+ h_inode = d_inode(h_dentry);
1facf9fc 7642+ npositive++;
7643+ if (!args.type)
7644+ args.type = h_inode->i_mode & S_IFMT;
7645+ if (args.type != S_IFDIR)
7646+ break;
7647+ else if (isdir) {
7648+ /* the type of lower may be different */
7649+ bdiropq = au_dbdiropq(dentry);
7650+ if (bdiropq >= 0 && bdiropq <= bindex)
7651+ break;
7652+ }
7653+ }
7654+
7655+ if (npositive) {
7656+ AuLabel(positive);
7657+ au_update_dbstart(dentry);
7658+ }
7659+ err = npositive;
076b876e 7660+ if (unlikely(!au_opt_test(au_mntflags(sb), UDBA_NONE)
027c5e7a 7661+ && au_dbstart(dentry) < 0)) {
1facf9fc 7662+ err = -EIO;
523b37e3
AM
7663+ AuIOErr("both of real entry and whiteout found, %pd, err %d\n",
7664+ dentry, err);
027c5e7a 7665+ }
1facf9fc 7666+
4f0767ce 7667+out_parent:
4a4d8108 7668+ dput(parent);
1facf9fc 7669+ kfree(whname.name);
4f0767ce 7670+out:
1facf9fc 7671+ return err;
7672+}
7673+
076b876e 7674+struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent)
1facf9fc 7675+{
7676+ struct dentry *dentry;
7677+ int wkq_err;
7678+
5527c038 7679+ if (!au_test_h_perm_sio(d_inode(parent), MAY_EXEC))
b4510431 7680+ dentry = vfsub_lkup_one(name, parent);
1facf9fc 7681+ else {
b4510431
AM
7682+ struct vfsub_lkup_one_args args = {
7683+ .errp = &dentry,
7684+ .name = name,
7685+ .parent = parent
1facf9fc 7686+ };
7687+
b4510431 7688+ wkq_err = au_wkq_wait(vfsub_call_lkup_one, &args);
1facf9fc 7689+ if (unlikely(wkq_err))
7690+ dentry = ERR_PTR(wkq_err);
7691+ }
7692+
7693+ return dentry;
7694+}
7695+
7696+/*
7697+ * lookup @dentry on @bindex which should be negative.
7698+ */
86dc4139 7699+int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh)
1facf9fc 7700+{
7701+ int err;
7702+ struct dentry *parent, *h_parent, *h_dentry;
86dc4139 7703+ struct au_branch *br;
1facf9fc 7704+
1facf9fc 7705+ parent = dget_parent(dentry);
7706+ h_parent = au_h_dptr(parent, bindex);
86dc4139
AM
7707+ br = au_sbr(dentry->d_sb, bindex);
7708+ if (wh)
7709+ h_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
7710+ else
076b876e 7711+ h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent);
1facf9fc 7712+ err = PTR_ERR(h_dentry);
7713+ if (IS_ERR(h_dentry))
7714+ goto out;
5527c038 7715+ if (unlikely(d_is_positive(h_dentry))) {
1facf9fc 7716+ err = -EIO;
523b37e3 7717+ AuIOErr("%pd should be negative on b%d.\n", h_dentry, bindex);
1facf9fc 7718+ dput(h_dentry);
7719+ goto out;
7720+ }
7721+
4a4d8108 7722+ err = 0;
1facf9fc 7723+ if (bindex < au_dbstart(dentry))
7724+ au_set_dbstart(dentry, bindex);
7725+ if (au_dbend(dentry) < bindex)
7726+ au_set_dbend(dentry, bindex);
7727+ au_set_h_dptr(dentry, bindex, h_dentry);
1facf9fc 7728+
4f0767ce 7729+out:
1facf9fc 7730+ dput(parent);
7731+ return err;
7732+}
7733+
7734+/* ---------------------------------------------------------------------- */
7735+
7736+/* subset of struct inode */
7737+struct au_iattr {
7738+ unsigned long i_ino;
7739+ /* unsigned int i_nlink; */
0c3ec466
AM
7740+ kuid_t i_uid;
7741+ kgid_t i_gid;
1facf9fc 7742+ u64 i_version;
7743+/*
7744+ loff_t i_size;
7745+ blkcnt_t i_blocks;
7746+*/
7747+ umode_t i_mode;
7748+};
7749+
7750+static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
7751+{
7752+ ia->i_ino = h_inode->i_ino;
7753+ /* ia->i_nlink = h_inode->i_nlink; */
7754+ ia->i_uid = h_inode->i_uid;
7755+ ia->i_gid = h_inode->i_gid;
7756+ ia->i_version = h_inode->i_version;
7757+/*
7758+ ia->i_size = h_inode->i_size;
7759+ ia->i_blocks = h_inode->i_blocks;
7760+*/
7761+ ia->i_mode = (h_inode->i_mode & S_IFMT);
7762+}
7763+
7764+static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
7765+{
7766+ return ia->i_ino != h_inode->i_ino
7767+ /* || ia->i_nlink != h_inode->i_nlink */
0c3ec466 7768+ || !uid_eq(ia->i_uid, h_inode->i_uid)
2dfbb274 7769+ || !gid_eq(ia->i_gid, h_inode->i_gid)
1facf9fc 7770+ || ia->i_version != h_inode->i_version
7771+/*
7772+ || ia->i_size != h_inode->i_size
7773+ || ia->i_blocks != h_inode->i_blocks
7774+*/
7775+ || ia->i_mode != (h_inode->i_mode & S_IFMT);
7776+}
7777+
7778+static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
7779+ struct au_branch *br)
7780+{
7781+ int err;
7782+ struct au_iattr ia;
7783+ struct inode *h_inode;
7784+ struct dentry *h_d;
7785+ struct super_block *h_sb;
7786+
7787+ err = 0;
7788+ memset(&ia, -1, sizeof(ia));
7789+ h_sb = h_dentry->d_sb;
5527c038
JR
7790+ h_inode = NULL;
7791+ if (d_is_positive(h_dentry)) {
7792+ h_inode = d_inode(h_dentry);
1facf9fc 7793+ au_iattr_save(&ia, h_inode);
5527c038 7794+ } else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
1facf9fc 7795+ /* nfs d_revalidate may return 0 for negative dentry */
7796+ /* fuse d_revalidate always return 0 for negative dentry */
7797+ goto out;
7798+
7799+ /* main purpose is namei.c:cached_lookup() and d_revalidate */
b4510431 7800+ h_d = vfsub_lkup_one(&h_dentry->d_name, h_parent);
1facf9fc 7801+ err = PTR_ERR(h_d);
7802+ if (IS_ERR(h_d))
7803+ goto out;
7804+
7805+ err = 0;
7806+ if (unlikely(h_d != h_dentry
5527c038 7807+ || d_inode(h_d) != h_inode
1facf9fc 7808+ || (h_inode && au_iattr_test(&ia, h_inode))))
7809+ err = au_busy_or_stale();
7810+ dput(h_d);
7811+
4f0767ce 7812+out:
1facf9fc 7813+ AuTraceErr(err);
7814+ return err;
7815+}
7816+
7817+int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
7818+ struct dentry *h_parent, struct au_branch *br)
7819+{
7820+ int err;
7821+
7822+ err = 0;
027c5e7a
AM
7823+ if (udba == AuOpt_UDBA_REVAL
7824+ && !au_test_fs_remote(h_dentry->d_sb)) {
1facf9fc 7825+ IMustLock(h_dir);
5527c038 7826+ err = (d_inode(h_dentry->d_parent) != h_dir);
027c5e7a 7827+ } else if (udba != AuOpt_UDBA_NONE)
1facf9fc 7828+ err = au_h_verify_dentry(h_dentry, h_parent, br);
7829+
7830+ return err;
7831+}
7832+
7833+/* ---------------------------------------------------------------------- */
7834+
027c5e7a 7835+static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
1facf9fc 7836+{
027c5e7a 7837+ int err;
1facf9fc 7838+ aufs_bindex_t new_bindex, bindex, bend, bwh, bdiropq;
027c5e7a
AM
7839+ struct au_hdentry tmp, *p, *q;
7840+ struct au_dinfo *dinfo;
7841+ struct super_block *sb;
1facf9fc 7842+
027c5e7a 7843+ DiMustWriteLock(dentry);
1308ab2a 7844+
027c5e7a
AM
7845+ sb = dentry->d_sb;
7846+ dinfo = au_di(dentry);
1facf9fc 7847+ bend = dinfo->di_bend;
7848+ bwh = dinfo->di_bwh;
7849+ bdiropq = dinfo->di_bdiropq;
027c5e7a 7850+ p = dinfo->di_hdentry + dinfo->di_bstart;
1facf9fc 7851+ for (bindex = dinfo->di_bstart; bindex <= bend; bindex++, p++) {
027c5e7a 7852+ if (!p->hd_dentry)
1facf9fc 7853+ continue;
7854+
027c5e7a
AM
7855+ new_bindex = au_br_index(sb, p->hd_id);
7856+ if (new_bindex == bindex)
1facf9fc 7857+ continue;
1facf9fc 7858+
1facf9fc 7859+ if (dinfo->di_bwh == bindex)
7860+ bwh = new_bindex;
7861+ if (dinfo->di_bdiropq == bindex)
7862+ bdiropq = new_bindex;
7863+ if (new_bindex < 0) {
7864+ au_hdput(p);
7865+ p->hd_dentry = NULL;
7866+ continue;
7867+ }
7868+
7869+ /* swap two lower dentries, and loop again */
7870+ q = dinfo->di_hdentry + new_bindex;
7871+ tmp = *q;
7872+ *q = *p;
7873+ *p = tmp;
7874+ if (tmp.hd_dentry) {
7875+ bindex--;
7876+ p--;
7877+ }
7878+ }
7879+
1facf9fc 7880+ dinfo->di_bwh = -1;
7881+ if (bwh >= 0 && bwh <= au_sbend(sb) && au_sbr_whable(sb, bwh))
7882+ dinfo->di_bwh = bwh;
7883+
7884+ dinfo->di_bdiropq = -1;
7885+ if (bdiropq >= 0
7886+ && bdiropq <= au_sbend(sb)
7887+ && au_sbr_whable(sb, bdiropq))
7888+ dinfo->di_bdiropq = bdiropq;
7889+
027c5e7a
AM
7890+ err = -EIO;
7891+ dinfo->di_bstart = -1;
7892+ dinfo->di_bend = -1;
1facf9fc 7893+ bend = au_dbend(parent);
7894+ p = dinfo->di_hdentry;
7895+ for (bindex = 0; bindex <= bend; bindex++, p++)
7896+ if (p->hd_dentry) {
7897+ dinfo->di_bstart = bindex;
7898+ break;
7899+ }
7900+
027c5e7a
AM
7901+ if (dinfo->di_bstart >= 0) {
7902+ p = dinfo->di_hdentry + bend;
7903+ for (bindex = bend; bindex >= 0; bindex--, p--)
7904+ if (p->hd_dentry) {
7905+ dinfo->di_bend = bindex;
7906+ err = 0;
7907+ break;
7908+ }
7909+ }
7910+
7911+ return err;
1facf9fc 7912+}
7913+
027c5e7a 7914+static void au_do_hide(struct dentry *dentry)
1facf9fc 7915+{
027c5e7a 7916+ struct inode *inode;
1facf9fc 7917+
5527c038
JR
7918+ if (d_really_is_positive(dentry)) {
7919+ inode = d_inode(dentry);
7920+ if (!d_is_dir(dentry)) {
027c5e7a
AM
7921+ if (inode->i_nlink && !d_unhashed(dentry))
7922+ drop_nlink(inode);
7923+ } else {
7924+ clear_nlink(inode);
7925+ /* stop next lookup */
7926+ inode->i_flags |= S_DEAD;
7927+ }
7928+ smp_mb(); /* necessary? */
7929+ }
7930+ d_drop(dentry);
7931+}
1308ab2a 7932+
027c5e7a
AM
7933+static int au_hide_children(struct dentry *parent)
7934+{
7935+ int err, i, j, ndentry;
7936+ struct au_dcsub_pages dpages;
7937+ struct au_dpage *dpage;
7938+ struct dentry *dentry;
1facf9fc 7939+
027c5e7a 7940+ err = au_dpages_init(&dpages, GFP_NOFS);
1facf9fc 7941+ if (unlikely(err))
7942+ goto out;
027c5e7a
AM
7943+ err = au_dcsub_pages(&dpages, parent, NULL, NULL);
7944+ if (unlikely(err))
7945+ goto out_dpages;
1facf9fc 7946+
027c5e7a
AM
7947+ /* in reverse order */
7948+ for (i = dpages.ndpage - 1; i >= 0; i--) {
7949+ dpage = dpages.dpages + i;
7950+ ndentry = dpage->ndentry;
7951+ for (j = ndentry - 1; j >= 0; j--) {
7952+ dentry = dpage->dentries[j];
7953+ if (dentry != parent)
7954+ au_do_hide(dentry);
7955+ }
7956+ }
1facf9fc 7957+
027c5e7a
AM
7958+out_dpages:
7959+ au_dpages_free(&dpages);
4f0767ce 7960+out:
027c5e7a 7961+ return err;
1facf9fc 7962+}
7963+
027c5e7a 7964+static void au_hide(struct dentry *dentry)
1facf9fc 7965+{
027c5e7a 7966+ int err;
1facf9fc 7967+
027c5e7a 7968+ AuDbgDentry(dentry);
2000de60 7969+ if (d_is_dir(dentry)) {
027c5e7a
AM
7970+ /* shrink_dcache_parent(dentry); */
7971+ err = au_hide_children(dentry);
7972+ if (unlikely(err))
523b37e3
AM
7973+ AuIOErr("%pd, failed hiding children, ignored %d\n",
7974+ dentry, err);
027c5e7a
AM
7975+ }
7976+ au_do_hide(dentry);
7977+}
1facf9fc 7978+
027c5e7a
AM
7979+/*
7980+ * By adding a dirty branch, a cached dentry may be affected in various ways.
7981+ *
7982+ * a dirty branch is added
7983+ * - on the top of layers
7984+ * - in the middle of layers
7985+ * - to the bottom of layers
7986+ *
7987+ * on the added branch there exists
7988+ * - a whiteout
7989+ * - a diropq
7990+ * - a same named entry
7991+ * + exist
7992+ * * negative --> positive
7993+ * * positive --> positive
7994+ * - type is unchanged
7995+ * - type is changed
7996+ * + doesn't exist
7997+ * * negative --> negative
7998+ * * positive --> negative (rejected by au_br_del() for non-dir case)
7999+ * - none
8000+ */
8001+static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
8002+ struct au_dinfo *tmp)
8003+{
8004+ int err;
8005+ aufs_bindex_t bindex, bend;
8006+ struct {
8007+ struct dentry *dentry;
8008+ struct inode *inode;
8009+ mode_t mode;
8010+ } orig_h, tmp_h;
8011+ struct au_hdentry *hd;
8012+ struct inode *inode, *h_inode;
8013+ struct dentry *h_dentry;
8014+
8015+ err = 0;
8016+ AuDebugOn(dinfo->di_bstart < 0);
027c5e7a 8017+ orig_h.mode = 0;
5527c038
JR
8018+ orig_h.dentry = dinfo->di_hdentry[dinfo->di_bstart].hd_dentry;
8019+ orig_h.inode = NULL;
8020+ if (d_is_positive(orig_h.dentry)) {
8021+ orig_h.inode = d_inode(orig_h.dentry);
027c5e7a 8022+ orig_h.mode = orig_h.inode->i_mode & S_IFMT;
5527c038 8023+ }
027c5e7a
AM
8024+ memset(&tmp_h, 0, sizeof(tmp_h));
8025+ if (tmp->di_bstart >= 0) {
8026+ tmp_h.dentry = tmp->di_hdentry[tmp->di_bstart].hd_dentry;
5527c038
JR
8027+ tmp_h.inode = NULL;
8028+ if (d_is_positive(tmp_h.dentry)) {
8029+ tmp_h.inode = d_inode(tmp_h.dentry);
027c5e7a 8030+ tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
5527c038 8031+ }
027c5e7a
AM
8032+ }
8033+
5527c038
JR
8034+ inode = NULL;
8035+ if (d_really_is_positive(dentry))
8036+ inode = d_inode(dentry);
027c5e7a
AM
8037+ if (!orig_h.inode) {
8038+ AuDbg("nagative originally\n");
8039+ if (inode) {
8040+ au_hide(dentry);
8041+ goto out;
8042+ }
8043+ AuDebugOn(inode);
8044+ AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
8045+ AuDebugOn(dinfo->di_bdiropq != -1);
8046+
8047+ if (!tmp_h.inode) {
8048+ AuDbg("negative --> negative\n");
8049+ /* should have only one negative lower */
8050+ if (tmp->di_bstart >= 0
8051+ && tmp->di_bstart < dinfo->di_bstart) {
8052+ AuDebugOn(tmp->di_bstart != tmp->di_bend);
8053+ AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
8054+ au_set_h_dptr(dentry, dinfo->di_bstart, NULL);
8055+ au_di_cp(dinfo, tmp);
8056+ hd = tmp->di_hdentry + tmp->di_bstart;
8057+ au_set_h_dptr(dentry, tmp->di_bstart,
8058+ dget(hd->hd_dentry));
8059+ }
8060+ au_dbg_verify_dinode(dentry);
8061+ } else {
8062+ AuDbg("negative --> positive\n");
8063+ /*
8064+ * similar to the behaviour of creating with bypassing
8065+ * aufs.
8066+ * unhash it in order to force an error in the
8067+ * succeeding create operation.
8068+ * we should not set S_DEAD here.
8069+ */
8070+ d_drop(dentry);
8071+ /* au_di_swap(tmp, dinfo); */
8072+ au_dbg_verify_dinode(dentry);
8073+ }
8074+ } else {
8075+ AuDbg("positive originally\n");
8076+ /* inode may be NULL */
8077+ AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
8078+ if (!tmp_h.inode) {
8079+ AuDbg("positive --> negative\n");
8080+ /* or bypassing aufs */
8081+ au_hide(dentry);
8082+ if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_bstart)
8083+ dinfo->di_bwh = tmp->di_bwh;
8084+ if (inode)
8085+ err = au_refresh_hinode_self(inode);
8086+ au_dbg_verify_dinode(dentry);
8087+ } else if (orig_h.mode == tmp_h.mode) {
8088+ AuDbg("positive --> positive, same type\n");
8089+ if (!S_ISDIR(orig_h.mode)
8090+ && dinfo->di_bstart > tmp->di_bstart) {
8091+ /*
8092+ * similar to the behaviour of removing and
8093+ * creating.
8094+ */
8095+ au_hide(dentry);
8096+ if (inode)
8097+ err = au_refresh_hinode_self(inode);
8098+ au_dbg_verify_dinode(dentry);
8099+ } else {
8100+ /* fill empty slots */
8101+ if (dinfo->di_bstart > tmp->di_bstart)
8102+ dinfo->di_bstart = tmp->di_bstart;
8103+ if (dinfo->di_bend < tmp->di_bend)
8104+ dinfo->di_bend = tmp->di_bend;
8105+ dinfo->di_bwh = tmp->di_bwh;
8106+ dinfo->di_bdiropq = tmp->di_bdiropq;
8107+ hd = tmp->di_hdentry;
8108+ bend = dinfo->di_bend;
8109+ for (bindex = tmp->di_bstart; bindex <= bend;
8110+ bindex++) {
8111+ if (au_h_dptr(dentry, bindex))
8112+ continue;
8113+ h_dentry = hd[bindex].hd_dentry;
8114+ if (!h_dentry)
8115+ continue;
5527c038
JR
8116+ AuDebugOn(d_is_negative(h_dentry));
8117+ h_inode = d_inode(h_dentry);
027c5e7a
AM
8118+ AuDebugOn(orig_h.mode
8119+ != (h_inode->i_mode
8120+ & S_IFMT));
8121+ au_set_h_dptr(dentry, bindex,
8122+ dget(h_dentry));
8123+ }
8124+ err = au_refresh_hinode(inode, dentry);
8125+ au_dbg_verify_dinode(dentry);
8126+ }
8127+ } else {
8128+ AuDbg("positive --> positive, different type\n");
8129+ /* similar to the behaviour of removing and creating */
8130+ au_hide(dentry);
8131+ if (inode)
8132+ err = au_refresh_hinode_self(inode);
8133+ au_dbg_verify_dinode(dentry);
8134+ }
8135+ }
8136+
8137+out:
8138+ return err;
8139+}
8140+
8141+int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
8142+{
8143+ int err, ebrange;
8144+ unsigned int sigen;
8145+ struct au_dinfo *dinfo, *tmp;
8146+ struct super_block *sb;
8147+ struct inode *inode;
8148+
8149+ DiMustWriteLock(dentry);
8150+ AuDebugOn(IS_ROOT(dentry));
5527c038 8151+ AuDebugOn(d_really_is_negative(parent));
027c5e7a
AM
8152+
8153+ sb = dentry->d_sb;
027c5e7a
AM
8154+ sigen = au_sigen(sb);
8155+ err = au_digen_test(parent, sigen);
8156+ if (unlikely(err))
8157+ goto out;
8158+
8159+ dinfo = au_di(dentry);
8160+ err = au_di_realloc(dinfo, au_sbend(sb) + 1);
8161+ if (unlikely(err))
8162+ goto out;
8163+ ebrange = au_dbrange_test(dentry);
8164+ if (!ebrange)
8165+ ebrange = au_do_refresh_hdentry(dentry, parent);
8166+
38d290e6 8167+ if (d_unhashed(dentry) || ebrange /* || dinfo->di_tmpfile */) {
027c5e7a 8168+ AuDebugOn(au_dbstart(dentry) < 0 && au_dbend(dentry) >= 0);
5527c038
JR
8169+ if (d_really_is_positive(dentry)) {
8170+ inode = d_inode(dentry);
027c5e7a 8171+ err = au_refresh_hinode_self(inode);
5527c038 8172+ }
027c5e7a
AM
8173+ au_dbg_verify_dinode(dentry);
8174+ if (!err)
8175+ goto out_dgen; /* success */
8176+ goto out;
8177+ }
8178+
8179+ /* temporary dinfo */
8180+ AuDbgDentry(dentry);
8181+ err = -ENOMEM;
8182+ tmp = au_di_alloc(sb, AuLsc_DI_TMP);
8183+ if (unlikely(!tmp))
8184+ goto out;
8185+ au_di_swap(tmp, dinfo);
8186+ /* returns the number of positive dentries */
8187+ /*
8188+ * if current working dir is removed, it returns an error.
8189+ * but the dentry is legal.
8190+ */
537831f9 8191+ err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
027c5e7a
AM
8192+ AuDbgDentry(dentry);
8193+ au_di_swap(tmp, dinfo);
8194+ if (err == -ENOENT)
8195+ err = 0;
8196+ if (err >= 0) {
8197+ /* compare/refresh by dinfo */
8198+ AuDbgDentry(dentry);
8199+ err = au_refresh_by_dinfo(dentry, dinfo, tmp);
8200+ au_dbg_verify_dinode(dentry);
8201+ AuTraceErr(err);
8202+ }
8203+ au_rw_write_unlock(&tmp->di_rwsem);
8204+ au_di_free(tmp);
8205+ if (unlikely(err))
8206+ goto out;
8207+
8208+out_dgen:
8209+ au_update_digen(dentry);
8210+out:
8211+ if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
523b37e3 8212+ AuIOErr("failed refreshing %pd, %d\n", dentry, err);
027c5e7a
AM
8213+ AuDbgDentry(dentry);
8214+ }
8215+ AuTraceErr(err);
8216+ return err;
8217+}
8218+
b4510431
AM
8219+static int au_do_h_d_reval(struct dentry *h_dentry, unsigned int flags,
8220+ struct dentry *dentry, aufs_bindex_t bindex)
027c5e7a
AM
8221+{
8222+ int err, valid;
027c5e7a
AM
8223+
8224+ err = 0;
8225+ if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
8226+ goto out;
027c5e7a
AM
8227+
8228+ AuDbg("b%d\n", bindex);
b4510431
AM
8229+ /*
8230+ * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
8231+ * due to whiteout and branch permission.
8232+ */
8233+ flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
8234+ | LOOKUP_FOLLOW | LOOKUP_EXCL);
8235+ /* it may return tri-state */
8236+ valid = h_dentry->d_op->d_revalidate(h_dentry, flags);
1facf9fc 8237+
8238+ if (unlikely(valid < 0))
8239+ err = valid;
8240+ else if (!valid)
8241+ err = -EINVAL;
8242+
4f0767ce 8243+out:
1facf9fc 8244+ AuTraceErr(err);
8245+ return err;
8246+}
8247+
8248+/* todo: remove this */
8249+static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
b4510431 8250+ unsigned int flags, int do_udba)
1facf9fc 8251+{
8252+ int err;
8253+ umode_t mode, h_mode;
8254+ aufs_bindex_t bindex, btail, bstart, ibs, ibe;
38d290e6 8255+ unsigned char plus, unhashed, is_root, h_plus, h_nfs, tmpfile;
4a4d8108 8256+ struct inode *h_inode, *h_cached_inode;
1facf9fc 8257+ struct dentry *h_dentry;
8258+ struct qstr *name, *h_name;
8259+
8260+ err = 0;
8261+ plus = 0;
8262+ mode = 0;
1facf9fc 8263+ ibs = -1;
8264+ ibe = -1;
8265+ unhashed = !!d_unhashed(dentry);
8266+ is_root = !!IS_ROOT(dentry);
8267+ name = &dentry->d_name;
38d290e6 8268+ tmpfile = au_di(dentry)->di_tmpfile;
1facf9fc 8269+
8270+ /*
7f207e10
AM
8271+ * Theoretically, REVAL test should be unnecessary in case of
8272+ * {FS,I}NOTIFY.
8273+ * But {fs,i}notify doesn't fire some necessary events,
1facf9fc 8274+ * IN_ATTRIB for atime/nlink/pageio
1facf9fc 8275+ * Let's do REVAL test too.
8276+ */
8277+ if (do_udba && inode) {
8278+ mode = (inode->i_mode & S_IFMT);
8279+ plus = (inode->i_nlink > 0);
1facf9fc 8280+ ibs = au_ibstart(inode);
8281+ ibe = au_ibend(inode);
8282+ }
8283+
8284+ bstart = au_dbstart(dentry);
8285+ btail = bstart;
8286+ if (inode && S_ISDIR(inode->i_mode))
8287+ btail = au_dbtaildir(dentry);
8288+ for (bindex = bstart; bindex <= btail; bindex++) {
8289+ h_dentry = au_h_dptr(dentry, bindex);
8290+ if (!h_dentry)
8291+ continue;
8292+
523b37e3
AM
8293+ AuDbg("b%d, %pd\n", bindex, h_dentry);
8294+ h_nfs = !!au_test_nfs(h_dentry->d_sb);
027c5e7a 8295+ spin_lock(&h_dentry->d_lock);
1facf9fc 8296+ h_name = &h_dentry->d_name;
8297+ if (unlikely(do_udba
8298+ && !is_root
523b37e3
AM
8299+ && ((!h_nfs
8300+ && (unhashed != !!d_unhashed(h_dentry)
38d290e6
JR
8301+ || (!tmpfile
8302+ && !au_qstreq(name, h_name))
8303+ ))
523b37e3
AM
8304+ || (h_nfs
8305+ && !(flags & LOOKUP_OPEN)
8306+ && (h_dentry->d_flags
8307+ & DCACHE_NFSFS_RENAMED)))
1facf9fc 8308+ )) {
38d290e6
JR
8309+ int h_unhashed;
8310+
8311+ h_unhashed = d_unhashed(h_dentry);
027c5e7a 8312+ spin_unlock(&h_dentry->d_lock);
38d290e6
JR
8313+ AuDbg("unhash 0x%x 0x%x, %pd %pd\n",
8314+ unhashed, h_unhashed, dentry, h_dentry);
1facf9fc 8315+ goto err;
8316+ }
027c5e7a 8317+ spin_unlock(&h_dentry->d_lock);
1facf9fc 8318+
b4510431 8319+ err = au_do_h_d_reval(h_dentry, flags, dentry, bindex);
1facf9fc 8320+ if (unlikely(err))
8321+ /* do not goto err, to keep the errno */
8322+ break;
8323+
8324+ /* todo: plink too? */
8325+ if (!do_udba)
8326+ continue;
8327+
8328+ /* UDBA tests */
5527c038 8329+ if (unlikely(!!inode != d_is_positive(h_dentry)))
1facf9fc 8330+ goto err;
8331+
5527c038
JR
8332+ h_inode = NULL;
8333+ if (d_is_positive(h_dentry))
8334+ h_inode = d_inode(h_dentry);
1facf9fc 8335+ h_plus = plus;
8336+ h_mode = mode;
8337+ h_cached_inode = h_inode;
8338+ if (h_inode) {
8339+ h_mode = (h_inode->i_mode & S_IFMT);
8340+ h_plus = (h_inode->i_nlink > 0);
8341+ }
8342+ if (inode && ibs <= bindex && bindex <= ibe)
8343+ h_cached_inode = au_h_iptr(inode, bindex);
8344+
523b37e3 8345+ if (!h_nfs) {
38d290e6 8346+ if (unlikely(plus != h_plus && !tmpfile))
523b37e3
AM
8347+ goto err;
8348+ } else {
8349+ if (unlikely(!(h_dentry->d_flags & DCACHE_NFSFS_RENAMED)
8350+ && !is_root
8351+ && !IS_ROOT(h_dentry)
8352+ && unhashed != d_unhashed(h_dentry)))
8353+ goto err;
8354+ }
8355+ if (unlikely(mode != h_mode
1facf9fc 8356+ || h_cached_inode != h_inode))
8357+ goto err;
8358+ continue;
8359+
f6b6e03d 8360+err:
1facf9fc 8361+ err = -EINVAL;
8362+ break;
8363+ }
8364+
523b37e3 8365+ AuTraceErr(err);
1facf9fc 8366+ return err;
8367+}
8368+
027c5e7a 8369+/* todo: consolidate with do_refresh() and au_reval_for_attr() */
1facf9fc 8370+static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
8371+{
8372+ int err;
8373+ struct dentry *parent;
1facf9fc 8374+
027c5e7a 8375+ if (!au_digen_test(dentry, sigen))
1facf9fc 8376+ return 0;
8377+
8378+ parent = dget_parent(dentry);
8379+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a 8380+ AuDebugOn(au_digen_test(parent, sigen));
1facf9fc 8381+ au_dbg_verify_gen(parent, sigen);
027c5e7a 8382+ err = au_refresh_dentry(dentry, parent);
1facf9fc 8383+ di_read_unlock(parent, AuLock_IR);
8384+ dput(parent);
027c5e7a 8385+ AuTraceErr(err);
1facf9fc 8386+ return err;
8387+}
8388+
8389+int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
8390+{
8391+ int err;
8392+ struct dentry *d, *parent;
1facf9fc 8393+
027c5e7a 8394+ if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
1facf9fc 8395+ return simple_reval_dpath(dentry, sigen);
8396+
8397+ /* slow loop, keep it simple and stupid */
8398+ /* cf: au_cpup_dirs() */
8399+ err = 0;
8400+ parent = NULL;
027c5e7a 8401+ while (au_digen_test(dentry, sigen)) {
1facf9fc 8402+ d = dentry;
8403+ while (1) {
8404+ dput(parent);
8405+ parent = dget_parent(d);
027c5e7a 8406+ if (!au_digen_test(parent, sigen))
1facf9fc 8407+ break;
8408+ d = parent;
8409+ }
8410+
1facf9fc 8411+ if (d != dentry)
027c5e7a 8412+ di_write_lock_child2(d);
1facf9fc 8413+
8414+ /* someone might update our dentry while we were sleeping */
027c5e7a
AM
8415+ if (au_digen_test(d, sigen)) {
8416+ /*
8417+ * todo: consolidate with simple_reval_dpath(),
8418+ * do_refresh() and au_reval_for_attr().
8419+ */
1facf9fc 8420+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a 8421+ err = au_refresh_dentry(d, parent);
1facf9fc 8422+ di_read_unlock(parent, AuLock_IR);
8423+ }
8424+
8425+ if (d != dentry)
8426+ di_write_unlock(d);
8427+ dput(parent);
8428+ if (unlikely(err))
8429+ break;
8430+ }
8431+
8432+ return err;
8433+}
8434+
8435+/*
8436+ * if valid returns 1, otherwise 0.
8437+ */
b4510431 8438+static int aufs_d_revalidate(struct dentry *dentry, unsigned int flags)
1facf9fc 8439+{
8440+ int valid, err;
8441+ unsigned int sigen;
8442+ unsigned char do_udba;
8443+ struct super_block *sb;
8444+ struct inode *inode;
8445+
027c5e7a 8446+ /* todo: support rcu-walk? */
b4510431 8447+ if (flags & LOOKUP_RCU)
027c5e7a
AM
8448+ return -ECHILD;
8449+
8450+ valid = 0;
8451+ if (unlikely(!au_di(dentry)))
8452+ goto out;
8453+
e49829fe 8454+ valid = 1;
1facf9fc 8455+ sb = dentry->d_sb;
e49829fe
JR
8456+ /*
8457+ * todo: very ugly
8458+ * i_mutex of parent dir may be held,
8459+ * but we should not return 'invalid' due to busy.
8460+ */
8461+ err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
8462+ if (unlikely(err)) {
8463+ valid = err;
027c5e7a 8464+ AuTraceErr(err);
e49829fe
JR
8465+ goto out;
8466+ }
5527c038
JR
8467+ inode = NULL;
8468+ if (d_really_is_positive(dentry))
8469+ inode = d_inode(dentry);
c1595e42
JR
8470+ if (unlikely(inode && is_bad_inode(inode))) {
8471+ err = -EINVAL;
8472+ AuTraceErr(err);
8473+ goto out_dgrade;
8474+ }
027c5e7a
AM
8475+ if (unlikely(au_dbrange_test(dentry))) {
8476+ err = -EINVAL;
8477+ AuTraceErr(err);
8478+ goto out_dgrade;
1facf9fc 8479+ }
027c5e7a
AM
8480+
8481+ sigen = au_sigen(sb);
8482+ if (au_digen_test(dentry, sigen)) {
1facf9fc 8483+ AuDebugOn(IS_ROOT(dentry));
027c5e7a
AM
8484+ err = au_reval_dpath(dentry, sigen);
8485+ if (unlikely(err)) {
8486+ AuTraceErr(err);
1facf9fc 8487+ goto out_dgrade;
027c5e7a 8488+ }
1facf9fc 8489+ }
8490+ di_downgrade_lock(dentry, AuLock_IR);
8491+
1facf9fc 8492+ err = -EINVAL;
c1595e42 8493+ if (!(flags & (LOOKUP_OPEN | LOOKUP_EMPTY))
523b37e3 8494+ && inode
38d290e6 8495+ && !(inode->i_state && I_LINKABLE)
523b37e3 8496+ && (IS_DEADDIR(inode) || !inode->i_nlink))
027c5e7a
AM
8497+ goto out_inval;
8498+
1facf9fc 8499+ do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
8500+ if (do_udba && inode) {
8501+ aufs_bindex_t bstart = au_ibstart(inode);
027c5e7a 8502+ struct inode *h_inode;
1facf9fc 8503+
027c5e7a
AM
8504+ if (bstart >= 0) {
8505+ h_inode = au_h_iptr(inode, bstart);
8506+ if (h_inode && au_test_higen(inode, h_inode))
8507+ goto out_inval;
8508+ }
1facf9fc 8509+ }
8510+
b4510431 8511+ err = h_d_revalidate(dentry, inode, flags, do_udba);
027c5e7a 8512+ if (unlikely(!err && do_udba && au_dbstart(dentry) < 0)) {
1facf9fc 8513+ err = -EIO;
523b37e3
AM
8514+ AuDbg("both of real entry and whiteout found, %p, err %d\n",
8515+ dentry, err);
027c5e7a 8516+ }
e49829fe 8517+ goto out_inval;
1facf9fc 8518+
4f0767ce 8519+out_dgrade:
1facf9fc 8520+ di_downgrade_lock(dentry, AuLock_IR);
e49829fe 8521+out_inval:
1facf9fc 8522+ aufs_read_unlock(dentry, AuLock_IR);
8523+ AuTraceErr(err);
8524+ valid = !err;
e49829fe 8525+out:
027c5e7a 8526+ if (!valid) {
523b37e3 8527+ AuDbg("%pd invalid, %d\n", dentry, valid);
027c5e7a
AM
8528+ d_drop(dentry);
8529+ }
1facf9fc 8530+ return valid;
8531+}
8532+
8533+static void aufs_d_release(struct dentry *dentry)
8534+{
027c5e7a 8535+ if (au_di(dentry)) {
4a4d8108
AM
8536+ au_di_fin(dentry);
8537+ au_hn_di_reinit(dentry);
1facf9fc 8538+ }
1facf9fc 8539+}
8540+
4a4d8108 8541+const struct dentry_operations aufs_dop = {
c06a8ce3
AM
8542+ .d_revalidate = aufs_d_revalidate,
8543+ .d_weak_revalidate = aufs_d_revalidate,
8544+ .d_release = aufs_d_release
1facf9fc 8545+};
7f207e10
AM
8546diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
8547--- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 8548+++ linux/fs/aufs/dentry.h 2015-09-24 10:47:58.251386326 +0200
076b876e 8549@@ -0,0 +1,233 @@
1facf9fc 8550+/*
2000de60 8551+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 8552+ *
8553+ * This program, aufs is free software; you can redistribute it and/or modify
8554+ * it under the terms of the GNU General Public License as published by
8555+ * the Free Software Foundation; either version 2 of the License, or
8556+ * (at your option) any later version.
dece6358
AM
8557+ *
8558+ * This program is distributed in the hope that it will be useful,
8559+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
8560+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8561+ * GNU General Public License for more details.
8562+ *
8563+ * You should have received a copy of the GNU General Public License
523b37e3 8564+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 8565+ */
8566+
8567+/*
8568+ * lookup and dentry operations
8569+ */
8570+
8571+#ifndef __AUFS_DENTRY_H__
8572+#define __AUFS_DENTRY_H__
8573+
8574+#ifdef __KERNEL__
8575+
dece6358 8576+#include <linux/dcache.h>
1facf9fc 8577+#include "rwsem.h"
8578+
1facf9fc 8579+struct au_hdentry {
8580+ struct dentry *hd_dentry;
027c5e7a 8581+ aufs_bindex_t hd_id;
1facf9fc 8582+};
8583+
8584+struct au_dinfo {
8585+ atomic_t di_generation;
8586+
dece6358 8587+ struct au_rwsem di_rwsem;
1facf9fc 8588+ aufs_bindex_t di_bstart, di_bend, di_bwh, di_bdiropq;
38d290e6 8589+ unsigned char di_tmpfile; /* to allow the different name */
1facf9fc 8590+ struct au_hdentry *di_hdentry;
4a4d8108 8591+} ____cacheline_aligned_in_smp;
1facf9fc 8592+
8593+/* ---------------------------------------------------------------------- */
8594+
8595+/* dentry.c */
4a4d8108 8596+extern const struct dentry_operations aufs_dop;
1facf9fc 8597+struct au_branch;
076b876e 8598+struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent);
1facf9fc 8599+int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
8600+ struct dentry *h_parent, struct au_branch *br);
8601+
537831f9 8602+int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type);
86dc4139 8603+int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex, int wh);
027c5e7a 8604+int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
1facf9fc 8605+int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
8606+
8607+/* dinfo.c */
4a4d8108 8608+void au_di_init_once(void *_di);
027c5e7a
AM
8609+struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
8610+void au_di_free(struct au_dinfo *dinfo);
8611+void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
8612+void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
4a4d8108
AM
8613+int au_di_init(struct dentry *dentry);
8614+void au_di_fin(struct dentry *dentry);
1facf9fc 8615+int au_di_realloc(struct au_dinfo *dinfo, int nbr);
8616+
8617+void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
8618+void di_read_unlock(struct dentry *d, int flags);
8619+void di_downgrade_lock(struct dentry *d, int flags);
8620+void di_write_lock(struct dentry *d, unsigned int lsc);
8621+void di_write_unlock(struct dentry *d);
8622+void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
8623+void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
8624+void di_write_unlock2(struct dentry *d1, struct dentry *d2);
8625+
8626+struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
2cbb1c4b 8627+struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
1facf9fc 8628+aufs_bindex_t au_dbtail(struct dentry *dentry);
8629+aufs_bindex_t au_dbtaildir(struct dentry *dentry);
8630+
8631+void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
8632+ struct dentry *h_dentry);
027c5e7a
AM
8633+int au_digen_test(struct dentry *dentry, unsigned int sigen);
8634+int au_dbrange_test(struct dentry *dentry);
1facf9fc 8635+void au_update_digen(struct dentry *dentry);
8636+void au_update_dbrange(struct dentry *dentry, int do_put_zero);
8637+void au_update_dbstart(struct dentry *dentry);
8638+void au_update_dbend(struct dentry *dentry);
8639+int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
8640+
8641+/* ---------------------------------------------------------------------- */
8642+
8643+static inline struct au_dinfo *au_di(struct dentry *dentry)
8644+{
8645+ return dentry->d_fsdata;
8646+}
8647+
8648+/* ---------------------------------------------------------------------- */
8649+
8650+/* lock subclass for dinfo */
8651+enum {
8652+ AuLsc_DI_CHILD, /* child first */
4a4d8108 8653+ AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
1facf9fc 8654+ AuLsc_DI_CHILD3, /* copyup dirs */
8655+ AuLsc_DI_PARENT,
8656+ AuLsc_DI_PARENT2,
027c5e7a
AM
8657+ AuLsc_DI_PARENT3,
8658+ AuLsc_DI_TMP /* temp for replacing dinfo */
1facf9fc 8659+};
8660+
8661+/*
8662+ * di_read_lock_child, di_write_lock_child,
8663+ * di_read_lock_child2, di_write_lock_child2,
8664+ * di_read_lock_child3, di_write_lock_child3,
8665+ * di_read_lock_parent, di_write_lock_parent,
8666+ * di_read_lock_parent2, di_write_lock_parent2,
8667+ * di_read_lock_parent3, di_write_lock_parent3,
8668+ */
8669+#define AuReadLockFunc(name, lsc) \
8670+static inline void di_read_lock_##name(struct dentry *d, int flags) \
8671+{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
8672+
8673+#define AuWriteLockFunc(name, lsc) \
8674+static inline void di_write_lock_##name(struct dentry *d) \
8675+{ di_write_lock(d, AuLsc_DI_##lsc); }
8676+
8677+#define AuRWLockFuncs(name, lsc) \
8678+ AuReadLockFunc(name, lsc) \
8679+ AuWriteLockFunc(name, lsc)
8680+
8681+AuRWLockFuncs(child, CHILD);
8682+AuRWLockFuncs(child2, CHILD2);
8683+AuRWLockFuncs(child3, CHILD3);
8684+AuRWLockFuncs(parent, PARENT);
8685+AuRWLockFuncs(parent2, PARENT2);
8686+AuRWLockFuncs(parent3, PARENT3);
8687+
8688+#undef AuReadLockFunc
8689+#undef AuWriteLockFunc
8690+#undef AuRWLockFuncs
8691+
8692+#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
dece6358
AM
8693+#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
8694+#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
1facf9fc 8695+
8696+/* ---------------------------------------------------------------------- */
8697+
8698+/* todo: memory barrier? */
8699+static inline unsigned int au_digen(struct dentry *d)
8700+{
8701+ return atomic_read(&au_di(d)->di_generation);
8702+}
8703+
8704+static inline void au_h_dentry_init(struct au_hdentry *hdentry)
8705+{
8706+ hdentry->hd_dentry = NULL;
8707+}
8708+
8709+static inline void au_hdput(struct au_hdentry *hd)
8710+{
4a4d8108
AM
8711+ if (hd)
8712+ dput(hd->hd_dentry);
1facf9fc 8713+}
8714+
8715+static inline aufs_bindex_t au_dbstart(struct dentry *dentry)
8716+{
1308ab2a 8717+ DiMustAnyLock(dentry);
1facf9fc 8718+ return au_di(dentry)->di_bstart;
8719+}
8720+
8721+static inline aufs_bindex_t au_dbend(struct dentry *dentry)
8722+{
1308ab2a 8723+ DiMustAnyLock(dentry);
1facf9fc 8724+ return au_di(dentry)->di_bend;
8725+}
8726+
8727+static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
8728+{
1308ab2a 8729+ DiMustAnyLock(dentry);
1facf9fc 8730+ return au_di(dentry)->di_bwh;
8731+}
8732+
8733+static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
8734+{
1308ab2a 8735+ DiMustAnyLock(dentry);
1facf9fc 8736+ return au_di(dentry)->di_bdiropq;
8737+}
8738+
8739+/* todo: hard/soft set? */
8740+static inline void au_set_dbstart(struct dentry *dentry, aufs_bindex_t bindex)
8741+{
1308ab2a 8742+ DiMustWriteLock(dentry);
1facf9fc 8743+ au_di(dentry)->di_bstart = bindex;
8744+}
8745+
8746+static inline void au_set_dbend(struct dentry *dentry, aufs_bindex_t bindex)
8747+{
1308ab2a 8748+ DiMustWriteLock(dentry);
1facf9fc 8749+ au_di(dentry)->di_bend = bindex;
8750+}
8751+
8752+static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
8753+{
1308ab2a 8754+ DiMustWriteLock(dentry);
1facf9fc 8755+ /* dbwh can be outside of bstart - bend range */
8756+ au_di(dentry)->di_bwh = bindex;
8757+}
8758+
8759+static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
8760+{
1308ab2a 8761+ DiMustWriteLock(dentry);
1facf9fc 8762+ au_di(dentry)->di_bdiropq = bindex;
8763+}
8764+
8765+/* ---------------------------------------------------------------------- */
8766+
4a4d8108 8767+#ifdef CONFIG_AUFS_HNOTIFY
1facf9fc 8768+static inline void au_digen_dec(struct dentry *d)
8769+{
e49829fe 8770+ atomic_dec(&au_di(d)->di_generation);
1facf9fc 8771+}
8772+
4a4d8108 8773+static inline void au_hn_di_reinit(struct dentry *dentry)
1facf9fc 8774+{
8775+ dentry->d_fsdata = NULL;
8776+}
8777+#else
4a4d8108
AM
8778+AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
8779+#endif /* CONFIG_AUFS_HNOTIFY */
1facf9fc 8780+
8781+#endif /* __KERNEL__ */
8782+#endif /* __AUFS_DENTRY_H__ */
7f207e10
AM
8783diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
8784--- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 8785+++ linux/fs/aufs/dinfo.c 2015-09-24 10:47:58.251386326 +0200
5527c038 8786@@ -0,0 +1,550 @@
1facf9fc 8787+/*
2000de60 8788+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 8789+ *
8790+ * This program, aufs is free software; you can redistribute it and/or modify
8791+ * it under the terms of the GNU General Public License as published by
8792+ * the Free Software Foundation; either version 2 of the License, or
8793+ * (at your option) any later version.
dece6358
AM
8794+ *
8795+ * This program is distributed in the hope that it will be useful,
8796+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
8797+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8798+ * GNU General Public License for more details.
8799+ *
8800+ * You should have received a copy of the GNU General Public License
523b37e3 8801+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 8802+ */
8803+
8804+/*
8805+ * dentry private data
8806+ */
8807+
8808+#include "aufs.h"
8809+
e49829fe 8810+void au_di_init_once(void *_dinfo)
4a4d8108 8811+{
e49829fe
JR
8812+ struct au_dinfo *dinfo = _dinfo;
8813+ static struct lock_class_key aufs_di;
4a4d8108 8814+
e49829fe
JR
8815+ au_rw_init(&dinfo->di_rwsem);
8816+ au_rw_class(&dinfo->di_rwsem, &aufs_di);
4a4d8108
AM
8817+}
8818+
027c5e7a 8819+struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
1facf9fc 8820+{
8821+ struct au_dinfo *dinfo;
027c5e7a 8822+ int nbr, i;
1facf9fc 8823+
8824+ dinfo = au_cache_alloc_dinfo();
8825+ if (unlikely(!dinfo))
8826+ goto out;
8827+
1facf9fc 8828+ nbr = au_sbend(sb) + 1;
8829+ if (nbr <= 0)
8830+ nbr = 1;
8831+ dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
027c5e7a
AM
8832+ if (dinfo->di_hdentry) {
8833+ au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
8834+ dinfo->di_bstart = -1;
8835+ dinfo->di_bend = -1;
8836+ dinfo->di_bwh = -1;
8837+ dinfo->di_bdiropq = -1;
38d290e6 8838+ dinfo->di_tmpfile = 0;
027c5e7a
AM
8839+ for (i = 0; i < nbr; i++)
8840+ dinfo->di_hdentry[i].hd_id = -1;
8841+ goto out;
8842+ }
1facf9fc 8843+
1facf9fc 8844+ au_cache_free_dinfo(dinfo);
027c5e7a
AM
8845+ dinfo = NULL;
8846+
4f0767ce 8847+out:
027c5e7a 8848+ return dinfo;
1facf9fc 8849+}
8850+
027c5e7a 8851+void au_di_free(struct au_dinfo *dinfo)
4a4d8108 8852+{
4a4d8108
AM
8853+ struct au_hdentry *p;
8854+ aufs_bindex_t bend, bindex;
8855+
8856+ /* dentry may not be revalidated */
027c5e7a 8857+ bindex = dinfo->di_bstart;
4a4d8108 8858+ if (bindex >= 0) {
027c5e7a
AM
8859+ bend = dinfo->di_bend;
8860+ p = dinfo->di_hdentry + bindex;
4a4d8108
AM
8861+ while (bindex++ <= bend)
8862+ au_hdput(p++);
8863+ }
027c5e7a
AM
8864+ kfree(dinfo->di_hdentry);
8865+ au_cache_free_dinfo(dinfo);
8866+}
8867+
8868+void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
8869+{
8870+ struct au_hdentry *p;
8871+ aufs_bindex_t bi;
8872+
8873+ AuRwMustWriteLock(&a->di_rwsem);
8874+ AuRwMustWriteLock(&b->di_rwsem);
8875+
8876+#define DiSwap(v, name) \
8877+ do { \
8878+ v = a->di_##name; \
8879+ a->di_##name = b->di_##name; \
8880+ b->di_##name = v; \
8881+ } while (0)
8882+
8883+ DiSwap(p, hdentry);
8884+ DiSwap(bi, bstart);
8885+ DiSwap(bi, bend);
8886+ DiSwap(bi, bwh);
8887+ DiSwap(bi, bdiropq);
8888+ /* smp_mb(); */
8889+
8890+#undef DiSwap
8891+}
8892+
8893+void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
8894+{
8895+ AuRwMustWriteLock(&dst->di_rwsem);
8896+ AuRwMustWriteLock(&src->di_rwsem);
8897+
8898+ dst->di_bstart = src->di_bstart;
8899+ dst->di_bend = src->di_bend;
8900+ dst->di_bwh = src->di_bwh;
8901+ dst->di_bdiropq = src->di_bdiropq;
8902+ /* smp_mb(); */
8903+}
8904+
8905+int au_di_init(struct dentry *dentry)
8906+{
8907+ int err;
8908+ struct super_block *sb;
8909+ struct au_dinfo *dinfo;
8910+
8911+ err = 0;
8912+ sb = dentry->d_sb;
8913+ dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
8914+ if (dinfo) {
8915+ atomic_set(&dinfo->di_generation, au_sigen(sb));
8916+ /* smp_mb(); */ /* atomic_set */
8917+ dentry->d_fsdata = dinfo;
8918+ } else
8919+ err = -ENOMEM;
8920+
8921+ return err;
8922+}
8923+
8924+void au_di_fin(struct dentry *dentry)
8925+{
8926+ struct au_dinfo *dinfo;
8927+
8928+ dinfo = au_di(dentry);
8929+ AuRwDestroy(&dinfo->di_rwsem);
8930+ au_di_free(dinfo);
4a4d8108
AM
8931+}
8932+
1facf9fc 8933+int au_di_realloc(struct au_dinfo *dinfo, int nbr)
8934+{
8935+ int err, sz;
8936+ struct au_hdentry *hdp;
8937+
1308ab2a 8938+ AuRwMustWriteLock(&dinfo->di_rwsem);
8939+
1facf9fc 8940+ err = -ENOMEM;
8941+ sz = sizeof(*hdp) * (dinfo->di_bend + 1);
8942+ if (!sz)
8943+ sz = sizeof(*hdp);
8944+ hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS);
8945+ if (hdp) {
8946+ dinfo->di_hdentry = hdp;
8947+ err = 0;
8948+ }
8949+
8950+ return err;
8951+}
8952+
8953+/* ---------------------------------------------------------------------- */
8954+
8955+static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
8956+{
8957+ switch (lsc) {
8958+ case AuLsc_DI_CHILD:
8959+ ii_write_lock_child(inode);
8960+ break;
8961+ case AuLsc_DI_CHILD2:
8962+ ii_write_lock_child2(inode);
8963+ break;
8964+ case AuLsc_DI_CHILD3:
8965+ ii_write_lock_child3(inode);
8966+ break;
8967+ case AuLsc_DI_PARENT:
8968+ ii_write_lock_parent(inode);
8969+ break;
8970+ case AuLsc_DI_PARENT2:
8971+ ii_write_lock_parent2(inode);
8972+ break;
8973+ case AuLsc_DI_PARENT3:
8974+ ii_write_lock_parent3(inode);
8975+ break;
8976+ default:
8977+ BUG();
8978+ }
8979+}
8980+
8981+static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
8982+{
8983+ switch (lsc) {
8984+ case AuLsc_DI_CHILD:
8985+ ii_read_lock_child(inode);
8986+ break;
8987+ case AuLsc_DI_CHILD2:
8988+ ii_read_lock_child2(inode);
8989+ break;
8990+ case AuLsc_DI_CHILD3:
8991+ ii_read_lock_child3(inode);
8992+ break;
8993+ case AuLsc_DI_PARENT:
8994+ ii_read_lock_parent(inode);
8995+ break;
8996+ case AuLsc_DI_PARENT2:
8997+ ii_read_lock_parent2(inode);
8998+ break;
8999+ case AuLsc_DI_PARENT3:
9000+ ii_read_lock_parent3(inode);
9001+ break;
9002+ default:
9003+ BUG();
9004+ }
9005+}
9006+
9007+void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
9008+{
5527c038
JR
9009+ struct inode *inode;
9010+
dece6358 9011+ au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
5527c038
JR
9012+ if (d_really_is_positive(d)) {
9013+ inode = d_inode(d);
1facf9fc 9014+ if (au_ftest_lock(flags, IW))
5527c038 9015+ do_ii_write_lock(inode, lsc);
1facf9fc 9016+ else if (au_ftest_lock(flags, IR))
5527c038 9017+ do_ii_read_lock(inode, lsc);
1facf9fc 9018+ }
9019+}
9020+
9021+void di_read_unlock(struct dentry *d, int flags)
9022+{
5527c038
JR
9023+ struct inode *inode;
9024+
9025+ if (d_really_is_positive(d)) {
9026+ inode = d_inode(d);
027c5e7a
AM
9027+ if (au_ftest_lock(flags, IW)) {
9028+ au_dbg_verify_dinode(d);
5527c038 9029+ ii_write_unlock(inode);
027c5e7a
AM
9030+ } else if (au_ftest_lock(flags, IR)) {
9031+ au_dbg_verify_dinode(d);
5527c038 9032+ ii_read_unlock(inode);
027c5e7a 9033+ }
1facf9fc 9034+ }
dece6358 9035+ au_rw_read_unlock(&au_di(d)->di_rwsem);
1facf9fc 9036+}
9037+
9038+void di_downgrade_lock(struct dentry *d, int flags)
9039+{
5527c038
JR
9040+ if (d_really_is_positive(d) && au_ftest_lock(flags, IR))
9041+ ii_downgrade_lock(d_inode(d));
dece6358 9042+ au_rw_dgrade_lock(&au_di(d)->di_rwsem);
1facf9fc 9043+}
9044+
9045+void di_write_lock(struct dentry *d, unsigned int lsc)
9046+{
dece6358 9047+ au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
5527c038
JR
9048+ if (d_really_is_positive(d))
9049+ do_ii_write_lock(d_inode(d), lsc);
1facf9fc 9050+}
9051+
9052+void di_write_unlock(struct dentry *d)
9053+{
027c5e7a 9054+ au_dbg_verify_dinode(d);
5527c038
JR
9055+ if (d_really_is_positive(d))
9056+ ii_write_unlock(d_inode(d));
dece6358 9057+ au_rw_write_unlock(&au_di(d)->di_rwsem);
1facf9fc 9058+}
9059+
9060+void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
9061+{
9062+ AuDebugOn(d1 == d2
5527c038 9063+ || d_inode(d1) == d_inode(d2)
1facf9fc 9064+ || d1->d_sb != d2->d_sb);
9065+
9066+ if (isdir && au_test_subdir(d1, d2)) {
9067+ di_write_lock_child(d1);
9068+ di_write_lock_child2(d2);
9069+ } else {
9070+ /* there should be no races */
9071+ di_write_lock_child(d2);
9072+ di_write_lock_child2(d1);
9073+ }
9074+}
9075+
9076+void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
9077+{
9078+ AuDebugOn(d1 == d2
5527c038 9079+ || d_inode(d1) == d_inode(d2)
1facf9fc 9080+ || d1->d_sb != d2->d_sb);
9081+
9082+ if (isdir && au_test_subdir(d1, d2)) {
9083+ di_write_lock_parent(d1);
9084+ di_write_lock_parent2(d2);
9085+ } else {
9086+ /* there should be no races */
9087+ di_write_lock_parent(d2);
9088+ di_write_lock_parent2(d1);
9089+ }
9090+}
9091+
9092+void di_write_unlock2(struct dentry *d1, struct dentry *d2)
9093+{
9094+ di_write_unlock(d1);
5527c038 9095+ if (d_inode(d1) == d_inode(d2))
dece6358 9096+ au_rw_write_unlock(&au_di(d2)->di_rwsem);
1facf9fc 9097+ else
9098+ di_write_unlock(d2);
9099+}
9100+
9101+/* ---------------------------------------------------------------------- */
9102+
9103+struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
9104+{
9105+ struct dentry *d;
9106+
1308ab2a 9107+ DiMustAnyLock(dentry);
9108+
1facf9fc 9109+ if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
9110+ return NULL;
9111+ AuDebugOn(bindex < 0);
9112+ d = au_di(dentry)->di_hdentry[0 + bindex].hd_dentry;
c1595e42 9113+ AuDebugOn(d && au_dcount(d) <= 0);
1facf9fc 9114+ return d;
9115+}
9116+
2cbb1c4b
JR
9117+/*
9118+ * extended version of au_h_dptr().
38d290e6
JR
9119+ * returns a hashed and positive (or linkable) h_dentry in bindex, NULL, or
9120+ * error.
2cbb1c4b
JR
9121+ */
9122+struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
9123+{
9124+ struct dentry *h_dentry;
9125+ struct inode *inode, *h_inode;
9126+
5527c038 9127+ AuDebugOn(d_really_is_negative(dentry));
2cbb1c4b
JR
9128+
9129+ h_dentry = NULL;
9130+ if (au_dbstart(dentry) <= bindex
9131+ && bindex <= au_dbend(dentry))
9132+ h_dentry = au_h_dptr(dentry, bindex);
38d290e6 9133+ if (h_dentry && !au_d_linkable(h_dentry)) {
2cbb1c4b
JR
9134+ dget(h_dentry);
9135+ goto out; /* success */
9136+ }
9137+
5527c038 9138+ inode = d_inode(dentry);
2cbb1c4b
JR
9139+ AuDebugOn(bindex < au_ibstart(inode));
9140+ AuDebugOn(au_ibend(inode) < bindex);
9141+ h_inode = au_h_iptr(inode, bindex);
9142+ h_dentry = d_find_alias(h_inode);
9143+ if (h_dentry) {
9144+ if (!IS_ERR(h_dentry)) {
38d290e6 9145+ if (!au_d_linkable(h_dentry))
2cbb1c4b
JR
9146+ goto out; /* success */
9147+ dput(h_dentry);
9148+ } else
9149+ goto out;
9150+ }
9151+
9152+ if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
9153+ h_dentry = au_plink_lkup(inode, bindex);
9154+ AuDebugOn(!h_dentry);
9155+ if (!IS_ERR(h_dentry)) {
9156+ if (!au_d_hashed_positive(h_dentry))
9157+ goto out; /* success */
9158+ dput(h_dentry);
9159+ h_dentry = NULL;
9160+ }
9161+ }
9162+
9163+out:
9164+ AuDbgDentry(h_dentry);
9165+ return h_dentry;
9166+}
9167+
1facf9fc 9168+aufs_bindex_t au_dbtail(struct dentry *dentry)
9169+{
9170+ aufs_bindex_t bend, bwh;
9171+
9172+ bend = au_dbend(dentry);
9173+ if (0 <= bend) {
9174+ bwh = au_dbwh(dentry);
9175+ if (!bwh)
9176+ return bwh;
9177+ if (0 < bwh && bwh < bend)
9178+ return bwh - 1;
9179+ }
9180+ return bend;
9181+}
9182+
9183+aufs_bindex_t au_dbtaildir(struct dentry *dentry)
9184+{
9185+ aufs_bindex_t bend, bopq;
9186+
9187+ bend = au_dbtail(dentry);
9188+ if (0 <= bend) {
9189+ bopq = au_dbdiropq(dentry);
9190+ if (0 <= bopq && bopq < bend)
9191+ bend = bopq;
9192+ }
9193+ return bend;
9194+}
9195+
9196+/* ---------------------------------------------------------------------- */
9197+
9198+void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
9199+ struct dentry *h_dentry)
9200+{
9201+ struct au_hdentry *hd = au_di(dentry)->di_hdentry + bindex;
027c5e7a 9202+ struct au_branch *br;
1facf9fc 9203+
1308ab2a 9204+ DiMustWriteLock(dentry);
9205+
4a4d8108 9206+ au_hdput(hd);
1facf9fc 9207+ hd->hd_dentry = h_dentry;
027c5e7a
AM
9208+ if (h_dentry) {
9209+ br = au_sbr(dentry->d_sb, bindex);
9210+ hd->hd_id = br->br_id;
9211+ }
9212+}
9213+
9214+int au_dbrange_test(struct dentry *dentry)
9215+{
9216+ int err;
9217+ aufs_bindex_t bstart, bend;
9218+
9219+ err = 0;
9220+ bstart = au_dbstart(dentry);
9221+ bend = au_dbend(dentry);
9222+ if (bstart >= 0)
9223+ AuDebugOn(bend < 0 && bstart > bend);
9224+ else {
9225+ err = -EIO;
9226+ AuDebugOn(bend >= 0);
9227+ }
9228+
9229+ return err;
9230+}
9231+
9232+int au_digen_test(struct dentry *dentry, unsigned int sigen)
9233+{
9234+ int err;
9235+
9236+ err = 0;
9237+ if (unlikely(au_digen(dentry) != sigen
5527c038 9238+ || au_iigen_test(d_inode(dentry), sigen)))
027c5e7a
AM
9239+ err = -EIO;
9240+
9241+ return err;
1facf9fc 9242+}
9243+
9244+void au_update_digen(struct dentry *dentry)
9245+{
9246+ atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
9247+ /* smp_mb(); */ /* atomic_set */
9248+}
9249+
9250+void au_update_dbrange(struct dentry *dentry, int do_put_zero)
9251+{
9252+ struct au_dinfo *dinfo;
9253+ struct dentry *h_d;
4a4d8108 9254+ struct au_hdentry *hdp;
1facf9fc 9255+
1308ab2a 9256+ DiMustWriteLock(dentry);
9257+
1facf9fc 9258+ dinfo = au_di(dentry);
9259+ if (!dinfo || dinfo->di_bstart < 0)
9260+ return;
9261+
4a4d8108 9262+ hdp = dinfo->di_hdentry;
1facf9fc 9263+ if (do_put_zero) {
9264+ aufs_bindex_t bindex, bend;
9265+
9266+ bend = dinfo->di_bend;
9267+ for (bindex = dinfo->di_bstart; bindex <= bend; bindex++) {
4a4d8108 9268+ h_d = hdp[0 + bindex].hd_dentry;
5527c038 9269+ if (h_d && d_is_negative(h_d))
1facf9fc 9270+ au_set_h_dptr(dentry, bindex, NULL);
9271+ }
9272+ }
9273+
9274+ dinfo->di_bstart = -1;
9275+ while (++dinfo->di_bstart <= dinfo->di_bend)
4a4d8108 9276+ if (hdp[0 + dinfo->di_bstart].hd_dentry)
1facf9fc 9277+ break;
9278+ if (dinfo->di_bstart > dinfo->di_bend) {
9279+ dinfo->di_bstart = -1;
9280+ dinfo->di_bend = -1;
9281+ return;
9282+ }
9283+
9284+ dinfo->di_bend++;
9285+ while (0 <= --dinfo->di_bend)
4a4d8108 9286+ if (hdp[0 + dinfo->di_bend].hd_dentry)
1facf9fc 9287+ break;
9288+ AuDebugOn(dinfo->di_bstart > dinfo->di_bend || dinfo->di_bend < 0);
9289+}
9290+
9291+void au_update_dbstart(struct dentry *dentry)
9292+{
9293+ aufs_bindex_t bindex, bend;
9294+ struct dentry *h_dentry;
9295+
9296+ bend = au_dbend(dentry);
9297+ for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
9298+ h_dentry = au_h_dptr(dentry, bindex);
9299+ if (!h_dentry)
9300+ continue;
5527c038 9301+ if (d_is_positive(h_dentry)) {
1facf9fc 9302+ au_set_dbstart(dentry, bindex);
9303+ return;
9304+ }
9305+ au_set_h_dptr(dentry, bindex, NULL);
9306+ }
9307+}
9308+
9309+void au_update_dbend(struct dentry *dentry)
9310+{
9311+ aufs_bindex_t bindex, bstart;
9312+ struct dentry *h_dentry;
9313+
9314+ bstart = au_dbstart(dentry);
7f207e10 9315+ for (bindex = au_dbend(dentry); bindex >= bstart; bindex--) {
1facf9fc 9316+ h_dentry = au_h_dptr(dentry, bindex);
9317+ if (!h_dentry)
9318+ continue;
5527c038 9319+ if (d_is_positive(h_dentry)) {
1facf9fc 9320+ au_set_dbend(dentry, bindex);
9321+ return;
9322+ }
9323+ au_set_h_dptr(dentry, bindex, NULL);
9324+ }
9325+}
9326+
9327+int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
9328+{
9329+ aufs_bindex_t bindex, bend;
9330+
9331+ bend = au_dbend(dentry);
9332+ for (bindex = au_dbstart(dentry); bindex <= bend; bindex++)
9333+ if (au_h_dptr(dentry, bindex) == h_dentry)
9334+ return bindex;
9335+ return -1;
9336+}
7f207e10
AM
9337diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
9338--- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 9339+++ linux/fs/aufs/dir.c 2015-09-24 10:47:58.251386326 +0200
5527c038 9340@@ -0,0 +1,753 @@
1facf9fc 9341+/*
2000de60 9342+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 9343+ *
9344+ * This program, aufs is free software; you can redistribute it and/or modify
9345+ * it under the terms of the GNU General Public License as published by
9346+ * the Free Software Foundation; either version 2 of the License, or
9347+ * (at your option) any later version.
dece6358
AM
9348+ *
9349+ * This program is distributed in the hope that it will be useful,
9350+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
9351+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9352+ * GNU General Public License for more details.
9353+ *
9354+ * You should have received a copy of the GNU General Public License
523b37e3 9355+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 9356+ */
9357+
9358+/*
9359+ * directory operations
9360+ */
9361+
9362+#include <linux/fs_stack.h>
9363+#include "aufs.h"
9364+
9365+void au_add_nlink(struct inode *dir, struct inode *h_dir)
9366+{
9dbd164d
AM
9367+ unsigned int nlink;
9368+
1facf9fc 9369+ AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9370+
9dbd164d
AM
9371+ nlink = dir->i_nlink;
9372+ nlink += h_dir->i_nlink - 2;
1facf9fc 9373+ if (h_dir->i_nlink < 2)
9dbd164d 9374+ nlink += 2;
f6b6e03d 9375+ smp_mb(); /* for i_nlink */
7eafdf33 9376+ /* 0 can happen in revaliding */
92d182d2 9377+ set_nlink(dir, nlink);
1facf9fc 9378+}
9379+
9380+void au_sub_nlink(struct inode *dir, struct inode *h_dir)
9381+{
9dbd164d
AM
9382+ unsigned int nlink;
9383+
1facf9fc 9384+ AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
9385+
9dbd164d
AM
9386+ nlink = dir->i_nlink;
9387+ nlink -= h_dir->i_nlink - 2;
1facf9fc 9388+ if (h_dir->i_nlink < 2)
9dbd164d 9389+ nlink -= 2;
f6b6e03d 9390+ smp_mb(); /* for i_nlink */
92d182d2 9391+ /* nlink == 0 means the branch-fs is broken */
9dbd164d 9392+ set_nlink(dir, nlink);
1facf9fc 9393+}
9394+
1308ab2a 9395+loff_t au_dir_size(struct file *file, struct dentry *dentry)
9396+{
9397+ loff_t sz;
9398+ aufs_bindex_t bindex, bend;
9399+ struct file *h_file;
9400+ struct dentry *h_dentry;
9401+
9402+ sz = 0;
9403+ if (file) {
2000de60 9404+ AuDebugOn(!d_is_dir(file->f_path.dentry));
1308ab2a 9405+
4a4d8108 9406+ bend = au_fbend_dir(file);
1308ab2a 9407+ for (bindex = au_fbstart(file);
9408+ bindex <= bend && sz < KMALLOC_MAX_SIZE;
9409+ bindex++) {
4a4d8108 9410+ h_file = au_hf_dir(file, bindex);
c06a8ce3
AM
9411+ if (h_file && file_inode(h_file))
9412+ sz += vfsub_f_size_read(h_file);
1308ab2a 9413+ }
9414+ } else {
9415+ AuDebugOn(!dentry);
2000de60 9416+ AuDebugOn(!d_is_dir(dentry));
1308ab2a 9417+
9418+ bend = au_dbtaildir(dentry);
9419+ for (bindex = au_dbstart(dentry);
9420+ bindex <= bend && sz < KMALLOC_MAX_SIZE;
9421+ bindex++) {
9422+ h_dentry = au_h_dptr(dentry, bindex);
5527c038
JR
9423+ if (h_dentry && d_is_positive(h_dentry))
9424+ sz += i_size_read(d_inode(h_dentry));
1308ab2a 9425+ }
9426+ }
9427+ if (sz < KMALLOC_MAX_SIZE)
9428+ sz = roundup_pow_of_two(sz);
9429+ if (sz > KMALLOC_MAX_SIZE)
9430+ sz = KMALLOC_MAX_SIZE;
9431+ else if (sz < NAME_MAX) {
9432+ BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
9433+ sz = AUFS_RDBLK_DEF;
9434+ }
9435+ return sz;
9436+}
9437+
b912730e
AM
9438+struct au_dir_ts_arg {
9439+ struct dentry *dentry;
9440+ aufs_bindex_t brid;
9441+};
9442+
9443+static void au_do_dir_ts(void *arg)
9444+{
9445+ struct au_dir_ts_arg *a = arg;
9446+ struct au_dtime dt;
9447+ struct path h_path;
9448+ struct inode *dir, *h_dir;
9449+ struct super_block *sb;
9450+ struct au_branch *br;
9451+ struct au_hinode *hdir;
9452+ int err;
9453+ aufs_bindex_t bstart, bindex;
9454+
9455+ sb = a->dentry->d_sb;
5527c038 9456+ if (d_really_is_negative(a->dentry))
b912730e 9457+ goto out;
b912730e
AM
9458+ aufs_read_lock(a->dentry, AuLock_DW | AuLock_DIR); /* noflush */
9459+
5527c038
JR
9460+ /* no dir->i_mutex lock */
9461+ dir = d_inode(a->dentry);
b912730e
AM
9462+ bstart = au_ibstart(dir);
9463+ bindex = au_br_index(sb, a->brid);
9464+ if (bindex < bstart)
9465+ goto out_unlock;
9466+
9467+ br = au_sbr(sb, bindex);
9468+ h_path.dentry = au_h_dptr(a->dentry, bindex);
9469+ if (!h_path.dentry)
9470+ goto out_unlock;
9471+ h_path.mnt = au_br_mnt(br);
9472+ au_dtime_store(&dt, a->dentry, &h_path);
9473+
9474+ br = au_sbr(sb, bstart);
9475+ if (!au_br_writable(br->br_perm))
9476+ goto out_unlock;
9477+ h_path.dentry = au_h_dptr(a->dentry, bstart);
9478+ h_path.mnt = au_br_mnt(br);
9479+ err = vfsub_mnt_want_write(h_path.mnt);
9480+ if (err)
9481+ goto out_unlock;
9482+ hdir = au_hi(dir, bstart);
9483+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
9484+ h_dir = au_h_iptr(dir, bstart);
9485+ if (h_dir->i_nlink
9486+ && timespec_compare(&h_dir->i_mtime, &dt.dt_mtime) < 0) {
9487+ dt.dt_h_path = h_path;
9488+ au_dtime_revert(&dt);
9489+ }
9490+ au_hn_imtx_unlock(hdir);
9491+ vfsub_mnt_drop_write(h_path.mnt);
9492+ au_cpup_attr_timesizes(dir);
9493+
9494+out_unlock:
9495+ aufs_read_unlock(a->dentry, AuLock_DW);
9496+out:
9497+ dput(a->dentry);
9498+ au_nwt_done(&au_sbi(sb)->si_nowait);
9499+ kfree(arg);
9500+}
9501+
9502+void au_dir_ts(struct inode *dir, aufs_bindex_t bindex)
9503+{
9504+ int perm, wkq_err;
9505+ aufs_bindex_t bstart;
9506+ struct au_dir_ts_arg *arg;
9507+ struct dentry *dentry;
9508+ struct super_block *sb;
9509+
9510+ IMustLock(dir);
9511+
9512+ dentry = d_find_any_alias(dir);
9513+ AuDebugOn(!dentry);
9514+ sb = dentry->d_sb;
9515+ bstart = au_ibstart(dir);
9516+ if (bstart == bindex) {
9517+ au_cpup_attr_timesizes(dir);
9518+ goto out;
9519+ }
9520+
9521+ perm = au_sbr_perm(sb, bstart);
9522+ if (!au_br_writable(perm))
9523+ goto out;
9524+
9525+ arg = kmalloc(sizeof(*arg), GFP_NOFS);
9526+ if (!arg)
9527+ goto out;
9528+
9529+ arg->dentry = dget(dentry); /* will be dput-ted by au_do_dir_ts() */
9530+ arg->brid = au_sbr_id(sb, bindex);
9531+ wkq_err = au_wkq_nowait(au_do_dir_ts, arg, sb, /*flags*/0);
9532+ if (unlikely(wkq_err)) {
9533+ pr_err("wkq %d\n", wkq_err);
9534+ dput(dentry);
9535+ kfree(arg);
9536+ }
9537+
9538+out:
9539+ dput(dentry);
9540+}
9541+
1facf9fc 9542+/* ---------------------------------------------------------------------- */
9543+
9544+static int reopen_dir(struct file *file)
9545+{
9546+ int err;
9547+ unsigned int flags;
9548+ aufs_bindex_t bindex, btail, bstart;
9549+ struct dentry *dentry, *h_dentry;
9550+ struct file *h_file;
9551+
9552+ /* open all lower dirs */
2000de60 9553+ dentry = file->f_path.dentry;
1facf9fc 9554+ bstart = au_dbstart(dentry);
9555+ for (bindex = au_fbstart(file); bindex < bstart; bindex++)
9556+ au_set_h_fptr(file, bindex, NULL);
9557+ au_set_fbstart(file, bstart);
9558+
9559+ btail = au_dbtaildir(dentry);
4a4d8108 9560+ for (bindex = au_fbend_dir(file); btail < bindex; bindex--)
1facf9fc 9561+ au_set_h_fptr(file, bindex, NULL);
4a4d8108 9562+ au_set_fbend_dir(file, btail);
1facf9fc 9563+
4a4d8108 9564+ flags = vfsub_file_flags(file);
1facf9fc 9565+ for (bindex = bstart; bindex <= btail; bindex++) {
9566+ h_dentry = au_h_dptr(dentry, bindex);
9567+ if (!h_dentry)
9568+ continue;
4a4d8108 9569+ h_file = au_hf_dir(file, bindex);
1facf9fc 9570+ if (h_file)
9571+ continue;
9572+
392086de 9573+ h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
1facf9fc 9574+ err = PTR_ERR(h_file);
9575+ if (IS_ERR(h_file))
9576+ goto out; /* close all? */
9577+ au_set_h_fptr(file, bindex, h_file);
9578+ }
9579+ au_update_figen(file);
9580+ /* todo: necessary? */
9581+ /* file->f_ra = h_file->f_ra; */
9582+ err = 0;
9583+
4f0767ce 9584+out:
1facf9fc 9585+ return err;
9586+}
9587+
b912730e 9588+static int do_open_dir(struct file *file, int flags, struct file *h_file)
1facf9fc 9589+{
9590+ int err;
9591+ aufs_bindex_t bindex, btail;
9592+ struct dentry *dentry, *h_dentry;
1facf9fc 9593+
1308ab2a 9594+ FiMustWriteLock(file);
b912730e 9595+ AuDebugOn(h_file);
1308ab2a 9596+
523b37e3 9597+ err = 0;
2000de60 9598+ dentry = file->f_path.dentry;
5527c038 9599+ file->f_version = d_inode(dentry)->i_version;
1facf9fc 9600+ bindex = au_dbstart(dentry);
9601+ au_set_fbstart(file, bindex);
9602+ btail = au_dbtaildir(dentry);
4a4d8108 9603+ au_set_fbend_dir(file, btail);
1facf9fc 9604+ for (; !err && bindex <= btail; bindex++) {
9605+ h_dentry = au_h_dptr(dentry, bindex);
9606+ if (!h_dentry)
9607+ continue;
9608+
392086de 9609+ h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
1facf9fc 9610+ if (IS_ERR(h_file)) {
9611+ err = PTR_ERR(h_file);
9612+ break;
9613+ }
9614+ au_set_h_fptr(file, bindex, h_file);
9615+ }
9616+ au_update_figen(file);
9617+ /* todo: necessary? */
9618+ /* file->f_ra = h_file->f_ra; */
9619+ if (!err)
9620+ return 0; /* success */
9621+
9622+ /* close all */
9623+ for (bindex = au_fbstart(file); bindex <= btail; bindex++)
9624+ au_set_h_fptr(file, bindex, NULL);
9625+ au_set_fbstart(file, -1);
4a4d8108
AM
9626+ au_set_fbend_dir(file, -1);
9627+
1facf9fc 9628+ return err;
9629+}
9630+
9631+static int aufs_open_dir(struct inode *inode __maybe_unused,
9632+ struct file *file)
9633+{
4a4d8108
AM
9634+ int err;
9635+ struct super_block *sb;
9636+ struct au_fidir *fidir;
9637+
9638+ err = -ENOMEM;
2000de60 9639+ sb = file->f_path.dentry->d_sb;
4a4d8108 9640+ si_read_lock(sb, AuLock_FLUSH);
e49829fe 9641+ fidir = au_fidir_alloc(sb);
4a4d8108 9642+ if (fidir) {
b912730e
AM
9643+ struct au_do_open_args args = {
9644+ .open = do_open_dir,
9645+ .fidir = fidir
9646+ };
9647+ err = au_do_open(file, &args);
4a4d8108
AM
9648+ if (unlikely(err))
9649+ kfree(fidir);
9650+ }
9651+ si_read_unlock(sb);
9652+ return err;
1facf9fc 9653+}
9654+
9655+static int aufs_release_dir(struct inode *inode __maybe_unused,
9656+ struct file *file)
9657+{
9658+ struct au_vdir *vdir_cache;
4a4d8108
AM
9659+ struct au_finfo *finfo;
9660+ struct au_fidir *fidir;
9661+ aufs_bindex_t bindex, bend;
1facf9fc 9662+
4a4d8108
AM
9663+ finfo = au_fi(file);
9664+ fidir = finfo->fi_hdir;
9665+ if (fidir) {
076b876e 9666+ au_sphl_del(&finfo->fi_hlist,
2000de60 9667+ &au_sbi(file->f_path.dentry->d_sb)->si_files);
4a4d8108
AM
9668+ vdir_cache = fidir->fd_vdir_cache; /* lock-free */
9669+ if (vdir_cache)
9670+ au_vdir_free(vdir_cache);
9671+
9672+ bindex = finfo->fi_btop;
9673+ if (bindex >= 0) {
9674+ /*
9675+ * calls fput() instead of filp_close(),
9676+ * since no dnotify or lock for the lower file.
9677+ */
9678+ bend = fidir->fd_bbot;
9679+ for (; bindex <= bend; bindex++)
9680+ au_set_h_fptr(file, bindex, NULL);
9681+ }
9682+ kfree(fidir);
9683+ finfo->fi_hdir = NULL;
1facf9fc 9684+ }
1facf9fc 9685+ au_finfo_fin(file);
1facf9fc 9686+ return 0;
9687+}
9688+
9689+/* ---------------------------------------------------------------------- */
9690+
4a4d8108
AM
9691+static int au_do_flush_dir(struct file *file, fl_owner_t id)
9692+{
9693+ int err;
9694+ aufs_bindex_t bindex, bend;
9695+ struct file *h_file;
9696+
9697+ err = 0;
9698+ bend = au_fbend_dir(file);
9699+ for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
9700+ h_file = au_hf_dir(file, bindex);
9701+ if (h_file)
9702+ err = vfsub_flush(h_file, id);
9703+ }
9704+ return err;
9705+}
9706+
9707+static int aufs_flush_dir(struct file *file, fl_owner_t id)
9708+{
9709+ return au_do_flush(file, id, au_do_flush_dir);
9710+}
9711+
9712+/* ---------------------------------------------------------------------- */
9713+
1facf9fc 9714+static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
9715+{
9716+ int err;
9717+ aufs_bindex_t bend, bindex;
9718+ struct inode *inode;
9719+ struct super_block *sb;
9720+
9721+ err = 0;
9722+ sb = dentry->d_sb;
5527c038 9723+ inode = d_inode(dentry);
1facf9fc 9724+ IMustLock(inode);
9725+ bend = au_dbend(dentry);
9726+ for (bindex = au_dbstart(dentry); !err && bindex <= bend; bindex++) {
9727+ struct path h_path;
1facf9fc 9728+
9729+ if (au_test_ro(sb, bindex, inode))
9730+ continue;
9731+ h_path.dentry = au_h_dptr(dentry, bindex);
9732+ if (!h_path.dentry)
9733+ continue;
1facf9fc 9734+
1facf9fc 9735+ h_path.mnt = au_sbr_mnt(sb, bindex);
53392da6 9736+ err = vfsub_fsync(NULL, &h_path, datasync);
1facf9fc 9737+ }
9738+
9739+ return err;
9740+}
9741+
9742+static int au_do_fsync_dir(struct file *file, int datasync)
9743+{
9744+ int err;
9745+ aufs_bindex_t bend, bindex;
9746+ struct file *h_file;
9747+ struct super_block *sb;
9748+ struct inode *inode;
1facf9fc 9749+
9750+ err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9751+ if (unlikely(err))
9752+ goto out;
9753+
c06a8ce3 9754+ inode = file_inode(file);
b912730e 9755+ sb = inode->i_sb;
4a4d8108 9756+ bend = au_fbend_dir(file);
1facf9fc 9757+ for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
4a4d8108 9758+ h_file = au_hf_dir(file, bindex);
1facf9fc 9759+ if (!h_file || au_test_ro(sb, bindex, inode))
9760+ continue;
9761+
53392da6 9762+ err = vfsub_fsync(h_file, &h_file->f_path, datasync);
1facf9fc 9763+ }
9764+
4f0767ce 9765+out:
1facf9fc 9766+ return err;
9767+}
9768+
9769+/*
9770+ * @file may be NULL
9771+ */
1e00d052
AM
9772+static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
9773+ int datasync)
1facf9fc 9774+{
9775+ int err;
b752ccd1 9776+ struct dentry *dentry;
5527c038 9777+ struct inode *inode;
1facf9fc 9778+ struct super_block *sb;
1e00d052 9779+ struct mutex *mtx;
1facf9fc 9780+
9781+ err = 0;
2000de60 9782+ dentry = file->f_path.dentry;
5527c038
JR
9783+ inode = d_inode(dentry);
9784+ mtx = &inode->i_mutex;
1e00d052 9785+ mutex_lock(mtx);
1facf9fc 9786+ sb = dentry->d_sb;
9787+ si_noflush_read_lock(sb);
9788+ if (file)
9789+ err = au_do_fsync_dir(file, datasync);
9790+ else {
9791+ di_write_lock_child(dentry);
9792+ err = au_do_fsync_dir_no_file(dentry, datasync);
9793+ }
5527c038 9794+ au_cpup_attr_timesizes(inode);
1facf9fc 9795+ di_write_unlock(dentry);
9796+ if (file)
9797+ fi_write_unlock(file);
9798+
9799+ si_read_unlock(sb);
1e00d052 9800+ mutex_unlock(mtx);
1facf9fc 9801+ return err;
9802+}
9803+
9804+/* ---------------------------------------------------------------------- */
9805+
392086de 9806+static int aufs_iterate(struct file *file, struct dir_context *ctx)
1facf9fc 9807+{
9808+ int err;
9809+ struct dentry *dentry;
9dbd164d 9810+ struct inode *inode, *h_inode;
1facf9fc 9811+ struct super_block *sb;
9812+
523b37e3 9813+ AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
392086de 9814+
2000de60 9815+ dentry = file->f_path.dentry;
5527c038 9816+ inode = d_inode(dentry);
1facf9fc 9817+ IMustLock(inode);
9818+
9819+ sb = dentry->d_sb;
9820+ si_read_lock(sb, AuLock_FLUSH);
9821+ err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
9822+ if (unlikely(err))
9823+ goto out;
027c5e7a
AM
9824+ err = au_alive_dir(dentry);
9825+ if (!err)
9826+ err = au_vdir_init(file);
1facf9fc 9827+ di_downgrade_lock(dentry, AuLock_IR);
9828+ if (unlikely(err))
9829+ goto out_unlock;
9830+
9dbd164d 9831+ h_inode = au_h_iptr(inode, au_ibstart(inode));
b752ccd1 9832+ if (!au_test_nfsd()) {
392086de 9833+ err = au_vdir_fill_de(file, ctx);
9dbd164d 9834+ fsstack_copy_attr_atime(inode, h_inode);
1facf9fc 9835+ } else {
9836+ /*
9837+ * nfsd filldir may call lookup_one_len(), vfs_getattr(),
9838+ * encode_fh() and others.
9839+ */
9dbd164d 9840+ atomic_inc(&h_inode->i_count);
1facf9fc 9841+ di_read_unlock(dentry, AuLock_IR);
9842+ si_read_unlock(sb);
392086de 9843+ err = au_vdir_fill_de(file, ctx);
1facf9fc 9844+ fsstack_copy_attr_atime(inode, h_inode);
9845+ fi_write_unlock(file);
9dbd164d 9846+ iput(h_inode);
1facf9fc 9847+
9848+ AuTraceErr(err);
9849+ return err;
9850+ }
9851+
4f0767ce 9852+out_unlock:
1facf9fc 9853+ di_read_unlock(dentry, AuLock_IR);
9854+ fi_write_unlock(file);
4f0767ce 9855+out:
1facf9fc 9856+ si_read_unlock(sb);
9857+ return err;
9858+}
9859+
9860+/* ---------------------------------------------------------------------- */
9861+
9862+#define AuTestEmpty_WHONLY 1
dece6358
AM
9863+#define AuTestEmpty_CALLED (1 << 1)
9864+#define AuTestEmpty_SHWH (1 << 2)
1facf9fc 9865+#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
7f207e10
AM
9866+#define au_fset_testempty(flags, name) \
9867+ do { (flags) |= AuTestEmpty_##name; } while (0)
9868+#define au_fclr_testempty(flags, name) \
9869+ do { (flags) &= ~AuTestEmpty_##name; } while (0)
1facf9fc 9870+
dece6358
AM
9871+#ifndef CONFIG_AUFS_SHWH
9872+#undef AuTestEmpty_SHWH
9873+#define AuTestEmpty_SHWH 0
9874+#endif
9875+
1facf9fc 9876+struct test_empty_arg {
392086de 9877+ struct dir_context ctx;
1308ab2a 9878+ struct au_nhash *whlist;
1facf9fc 9879+ unsigned int flags;
9880+ int err;
9881+ aufs_bindex_t bindex;
9882+};
9883+
392086de
AM
9884+static int test_empty_cb(struct dir_context *ctx, const char *__name,
9885+ int namelen, loff_t offset __maybe_unused, u64 ino,
dece6358 9886+ unsigned int d_type)
1facf9fc 9887+{
392086de
AM
9888+ struct test_empty_arg *arg = container_of(ctx, struct test_empty_arg,
9889+ ctx);
1facf9fc 9890+ char *name = (void *)__name;
9891+
9892+ arg->err = 0;
9893+ au_fset_testempty(arg->flags, CALLED);
9894+ /* smp_mb(); */
9895+ if (name[0] == '.'
9896+ && (namelen == 1 || (name[1] == '.' && namelen == 2)))
9897+ goto out; /* success */
9898+
9899+ if (namelen <= AUFS_WH_PFX_LEN
9900+ || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
9901+ if (au_ftest_testempty(arg->flags, WHONLY)
1308ab2a 9902+ && !au_nhash_test_known_wh(arg->whlist, name, namelen))
1facf9fc 9903+ arg->err = -ENOTEMPTY;
9904+ goto out;
9905+ }
9906+
9907+ name += AUFS_WH_PFX_LEN;
9908+ namelen -= AUFS_WH_PFX_LEN;
1308ab2a 9909+ if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
1facf9fc 9910+ arg->err = au_nhash_append_wh
1308ab2a 9911+ (arg->whlist, name, namelen, ino, d_type, arg->bindex,
dece6358 9912+ au_ftest_testempty(arg->flags, SHWH));
1facf9fc 9913+
4f0767ce 9914+out:
1facf9fc 9915+ /* smp_mb(); */
9916+ AuTraceErr(arg->err);
9917+ return arg->err;
9918+}
9919+
9920+static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9921+{
9922+ int err;
9923+ struct file *h_file;
9924+
9925+ h_file = au_h_open(dentry, arg->bindex,
9926+ O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
392086de 9927+ /*file*/NULL, /*force_wr*/0);
1facf9fc 9928+ err = PTR_ERR(h_file);
9929+ if (IS_ERR(h_file))
9930+ goto out;
9931+
9932+ err = 0;
9933+ if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
c06a8ce3 9934+ && !file_inode(h_file)->i_nlink)
1facf9fc 9935+ goto out_put;
9936+
9937+ do {
9938+ arg->err = 0;
9939+ au_fclr_testempty(arg->flags, CALLED);
9940+ /* smp_mb(); */
392086de 9941+ err = vfsub_iterate_dir(h_file, &arg->ctx);
1facf9fc 9942+ if (err >= 0)
9943+ err = arg->err;
9944+ } while (!err && au_ftest_testempty(arg->flags, CALLED));
9945+
4f0767ce 9946+out_put:
1facf9fc 9947+ fput(h_file);
9948+ au_sbr_put(dentry->d_sb, arg->bindex);
4f0767ce 9949+out:
1facf9fc 9950+ return err;
9951+}
9952+
9953+struct do_test_empty_args {
9954+ int *errp;
9955+ struct dentry *dentry;
9956+ struct test_empty_arg *arg;
9957+};
9958+
9959+static void call_do_test_empty(void *args)
9960+{
9961+ struct do_test_empty_args *a = args;
9962+ *a->errp = do_test_empty(a->dentry, a->arg);
9963+}
9964+
9965+static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
9966+{
9967+ int err, wkq_err;
9968+ struct dentry *h_dentry;
9969+ struct inode *h_inode;
9970+
9971+ h_dentry = au_h_dptr(dentry, arg->bindex);
5527c038 9972+ h_inode = d_inode(h_dentry);
53392da6 9973+ /* todo: i_mode changes anytime? */
1facf9fc 9974+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
9975+ err = au_test_h_perm_sio(h_inode, MAY_EXEC | MAY_READ);
9976+ mutex_unlock(&h_inode->i_mutex);
9977+ if (!err)
9978+ err = do_test_empty(dentry, arg);
9979+ else {
9980+ struct do_test_empty_args args = {
9981+ .errp = &err,
9982+ .dentry = dentry,
9983+ .arg = arg
9984+ };
9985+ unsigned int flags = arg->flags;
9986+
9987+ wkq_err = au_wkq_wait(call_do_test_empty, &args);
9988+ if (unlikely(wkq_err))
9989+ err = wkq_err;
9990+ arg->flags = flags;
9991+ }
9992+
9993+ return err;
9994+}
9995+
9996+int au_test_empty_lower(struct dentry *dentry)
9997+{
9998+ int err;
1308ab2a 9999+ unsigned int rdhash;
1facf9fc 10000+ aufs_bindex_t bindex, bstart, btail;
1308ab2a 10001+ struct au_nhash whlist;
392086de
AM
10002+ struct test_empty_arg arg = {
10003+ .ctx = {
2000de60 10004+ .actor = test_empty_cb
392086de
AM
10005+ }
10006+ };
076b876e 10007+ int (*test_empty)(struct dentry *dentry, struct test_empty_arg *arg);
1facf9fc 10008+
dece6358
AM
10009+ SiMustAnyLock(dentry->d_sb);
10010+
1308ab2a 10011+ rdhash = au_sbi(dentry->d_sb)->si_rdhash;
10012+ if (!rdhash)
10013+ rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
10014+ err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
dece6358 10015+ if (unlikely(err))
1facf9fc 10016+ goto out;
10017+
1facf9fc 10018+ arg.flags = 0;
1308ab2a 10019+ arg.whlist = &whlist;
10020+ bstart = au_dbstart(dentry);
dece6358
AM
10021+ if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10022+ au_fset_testempty(arg.flags, SHWH);
076b876e
AM
10023+ test_empty = do_test_empty;
10024+ if (au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1))
10025+ test_empty = sio_test_empty;
1facf9fc 10026+ arg.bindex = bstart;
076b876e 10027+ err = test_empty(dentry, &arg);
1facf9fc 10028+ if (unlikely(err))
10029+ goto out_whlist;
10030+
10031+ au_fset_testempty(arg.flags, WHONLY);
10032+ btail = au_dbtaildir(dentry);
10033+ for (bindex = bstart + 1; !err && bindex <= btail; bindex++) {
10034+ struct dentry *h_dentry;
10035+
10036+ h_dentry = au_h_dptr(dentry, bindex);
5527c038 10037+ if (h_dentry && d_is_positive(h_dentry)) {
1facf9fc 10038+ arg.bindex = bindex;
076b876e 10039+ err = test_empty(dentry, &arg);
1facf9fc 10040+ }
10041+ }
10042+
4f0767ce 10043+out_whlist:
1308ab2a 10044+ au_nhash_wh_free(&whlist);
4f0767ce 10045+out:
1facf9fc 10046+ return err;
10047+}
10048+
10049+int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
10050+{
10051+ int err;
392086de
AM
10052+ struct test_empty_arg arg = {
10053+ .ctx = {
2000de60 10054+ .actor = test_empty_cb
392086de
AM
10055+ }
10056+ };
1facf9fc 10057+ aufs_bindex_t bindex, btail;
10058+
10059+ err = 0;
1308ab2a 10060+ arg.whlist = whlist;
1facf9fc 10061+ arg.flags = AuTestEmpty_WHONLY;
dece6358
AM
10062+ if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
10063+ au_fset_testempty(arg.flags, SHWH);
1facf9fc 10064+ btail = au_dbtaildir(dentry);
10065+ for (bindex = au_dbstart(dentry); !err && bindex <= btail; bindex++) {
10066+ struct dentry *h_dentry;
10067+
10068+ h_dentry = au_h_dptr(dentry, bindex);
5527c038 10069+ if (h_dentry && d_is_positive(h_dentry)) {
1facf9fc 10070+ arg.bindex = bindex;
10071+ err = sio_test_empty(dentry, &arg);
10072+ }
10073+ }
10074+
10075+ return err;
10076+}
10077+
10078+/* ---------------------------------------------------------------------- */
10079+
10080+const struct file_operations aufs_dir_fop = {
4a4d8108 10081+ .owner = THIS_MODULE,
027c5e7a 10082+ .llseek = default_llseek,
1facf9fc 10083+ .read = generic_read_dir,
392086de 10084+ .iterate = aufs_iterate,
1facf9fc 10085+ .unlocked_ioctl = aufs_ioctl_dir,
b752ccd1
AM
10086+#ifdef CONFIG_COMPAT
10087+ .compat_ioctl = aufs_compat_ioctl_dir,
10088+#endif
1facf9fc 10089+ .open = aufs_open_dir,
10090+ .release = aufs_release_dir,
4a4d8108 10091+ .flush = aufs_flush_dir,
1facf9fc 10092+ .fsync = aufs_fsync_dir
10093+};
7f207e10
AM
10094diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
10095--- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 10096+++ linux/fs/aufs/dir.h 2015-09-24 10:47:58.251386326 +0200
b912730e 10097@@ -0,0 +1,131 @@
1facf9fc 10098+/*
2000de60 10099+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 10100+ *
10101+ * This program, aufs is free software; you can redistribute it and/or modify
10102+ * it under the terms of the GNU General Public License as published by
10103+ * the Free Software Foundation; either version 2 of the License, or
10104+ * (at your option) any later version.
dece6358
AM
10105+ *
10106+ * This program is distributed in the hope that it will be useful,
10107+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10108+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10109+ * GNU General Public License for more details.
10110+ *
10111+ * You should have received a copy of the GNU General Public License
523b37e3 10112+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 10113+ */
10114+
10115+/*
10116+ * directory operations
10117+ */
10118+
10119+#ifndef __AUFS_DIR_H__
10120+#define __AUFS_DIR_H__
10121+
10122+#ifdef __KERNEL__
10123+
10124+#include <linux/fs.h>
1facf9fc 10125+
10126+/* ---------------------------------------------------------------------- */
10127+
10128+/* need to be faster and smaller */
10129+
10130+struct au_nhash {
dece6358
AM
10131+ unsigned int nh_num;
10132+ struct hlist_head *nh_head;
1facf9fc 10133+};
10134+
10135+struct au_vdir_destr {
10136+ unsigned char len;
10137+ unsigned char name[0];
10138+} __packed;
10139+
10140+struct au_vdir_dehstr {
10141+ struct hlist_node hash;
10142+ struct au_vdir_destr *str;
4a4d8108 10143+} ____cacheline_aligned_in_smp;
1facf9fc 10144+
10145+struct au_vdir_de {
10146+ ino_t de_ino;
10147+ unsigned char de_type;
10148+ /* caution: packed */
10149+ struct au_vdir_destr de_str;
10150+} __packed;
10151+
10152+struct au_vdir_wh {
10153+ struct hlist_node wh_hash;
dece6358
AM
10154+#ifdef CONFIG_AUFS_SHWH
10155+ ino_t wh_ino;
1facf9fc 10156+ aufs_bindex_t wh_bindex;
dece6358
AM
10157+ unsigned char wh_type;
10158+#else
10159+ aufs_bindex_t wh_bindex;
10160+#endif
10161+ /* caution: packed */
1facf9fc 10162+ struct au_vdir_destr wh_str;
10163+} __packed;
10164+
10165+union au_vdir_deblk_p {
10166+ unsigned char *deblk;
10167+ struct au_vdir_de *de;
10168+};
10169+
10170+struct au_vdir {
10171+ unsigned char **vd_deblk;
10172+ unsigned long vd_nblk;
1facf9fc 10173+ struct {
10174+ unsigned long ul;
10175+ union au_vdir_deblk_p p;
10176+ } vd_last;
10177+
10178+ unsigned long vd_version;
dece6358 10179+ unsigned int vd_deblk_sz;
1facf9fc 10180+ unsigned long vd_jiffy;
4a4d8108 10181+} ____cacheline_aligned_in_smp;
1facf9fc 10182+
10183+/* ---------------------------------------------------------------------- */
10184+
10185+/* dir.c */
10186+extern const struct file_operations aufs_dir_fop;
10187+void au_add_nlink(struct inode *dir, struct inode *h_dir);
10188+void au_sub_nlink(struct inode *dir, struct inode *h_dir);
1308ab2a 10189+loff_t au_dir_size(struct file *file, struct dentry *dentry);
b912730e 10190+void au_dir_ts(struct inode *dir, aufs_bindex_t bsrc);
1facf9fc 10191+int au_test_empty_lower(struct dentry *dentry);
10192+int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
10193+
10194+/* vdir.c */
1308ab2a 10195+unsigned int au_rdhash_est(loff_t sz);
dece6358
AM
10196+int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
10197+void au_nhash_wh_free(struct au_nhash *whlist);
1facf9fc 10198+int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
10199+ int limit);
dece6358
AM
10200+int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
10201+int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
10202+ unsigned int d_type, aufs_bindex_t bindex,
10203+ unsigned char shwh);
1facf9fc 10204+void au_vdir_free(struct au_vdir *vdir);
10205+int au_vdir_init(struct file *file);
392086de 10206+int au_vdir_fill_de(struct file *file, struct dir_context *ctx);
1facf9fc 10207+
10208+/* ioctl.c */
10209+long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
10210+
1308ab2a 10211+#ifdef CONFIG_AUFS_RDU
10212+/* rdu.c */
10213+long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
b752ccd1
AM
10214+#ifdef CONFIG_COMPAT
10215+long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
10216+ unsigned long arg);
10217+#endif
1308ab2a 10218+#else
c1595e42
JR
10219+AuStub(long, au_rdu_ioctl, return -EINVAL, struct file *file,
10220+ unsigned int cmd, unsigned long arg)
b752ccd1 10221+#ifdef CONFIG_COMPAT
c1595e42
JR
10222+AuStub(long, au_rdu_compat_ioctl, return -EINVAL, struct file *file,
10223+ unsigned int cmd, unsigned long arg)
b752ccd1 10224+#endif
1308ab2a 10225+#endif
10226+
1facf9fc 10227+#endif /* __KERNEL__ */
10228+#endif /* __AUFS_DIR_H__ */
7f207e10
AM
10229diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
10230--- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 10231+++ linux/fs/aufs/dynop.c 2015-09-24 10:47:58.251386326 +0200
7e9cd9fe 10232@@ -0,0 +1,369 @@
1facf9fc 10233+/*
2000de60 10234+ * Copyright (C) 2010-2015 Junjiro R. Okajima
1facf9fc 10235+ *
10236+ * This program, aufs is free software; you can redistribute it and/or modify
10237+ * it under the terms of the GNU General Public License as published by
10238+ * the Free Software Foundation; either version 2 of the License, or
10239+ * (at your option) any later version.
dece6358
AM
10240+ *
10241+ * This program is distributed in the hope that it will be useful,
10242+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10243+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10244+ * GNU General Public License for more details.
10245+ *
10246+ * You should have received a copy of the GNU General Public License
523b37e3 10247+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 10248+ */
10249+
10250+/*
4a4d8108 10251+ * dynamically customizable operations for regular files
1facf9fc 10252+ */
10253+
1facf9fc 10254+#include "aufs.h"
10255+
4a4d8108 10256+#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
1facf9fc 10257+
4a4d8108
AM
10258+/*
10259+ * How large will these lists be?
10260+ * Usually just a few elements, 20-30 at most for each, I guess.
10261+ */
10262+static struct au_splhead dynop[AuDyLast];
10263+
10264+static struct au_dykey *dy_gfind_get(struct au_splhead *spl, const void *h_op)
1facf9fc 10265+{
4a4d8108
AM
10266+ struct au_dykey *key, *tmp;
10267+ struct list_head *head;
1facf9fc 10268+
4a4d8108
AM
10269+ key = NULL;
10270+ head = &spl->head;
10271+ rcu_read_lock();
10272+ list_for_each_entry_rcu(tmp, head, dk_list)
10273+ if (tmp->dk_op.dy_hop == h_op) {
10274+ key = tmp;
10275+ kref_get(&key->dk_kref);
10276+ break;
10277+ }
10278+ rcu_read_unlock();
10279+
10280+ return key;
1facf9fc 10281+}
10282+
4a4d8108 10283+static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
1facf9fc 10284+{
4a4d8108
AM
10285+ struct au_dykey **k, *found;
10286+ const void *h_op = key->dk_op.dy_hop;
10287+ int i;
1facf9fc 10288+
4a4d8108
AM
10289+ found = NULL;
10290+ k = br->br_dykey;
10291+ for (i = 0; i < AuBrDynOp; i++)
10292+ if (k[i]) {
10293+ if (k[i]->dk_op.dy_hop == h_op) {
10294+ found = k[i];
10295+ break;
10296+ }
10297+ } else
10298+ break;
10299+ if (!found) {
10300+ spin_lock(&br->br_dykey_lock);
10301+ for (; i < AuBrDynOp; i++)
10302+ if (k[i]) {
10303+ if (k[i]->dk_op.dy_hop == h_op) {
10304+ found = k[i];
10305+ break;
10306+ }
10307+ } else {
10308+ k[i] = key;
10309+ break;
10310+ }
10311+ spin_unlock(&br->br_dykey_lock);
10312+ BUG_ON(i == AuBrDynOp); /* expand the array */
10313+ }
10314+
10315+ return found;
1facf9fc 10316+}
10317+
4a4d8108
AM
10318+/* kref_get() if @key is already added */
10319+static struct au_dykey *dy_gadd(struct au_splhead *spl, struct au_dykey *key)
10320+{
10321+ struct au_dykey *tmp, *found;
10322+ struct list_head *head;
10323+ const void *h_op = key->dk_op.dy_hop;
1facf9fc 10324+
4a4d8108
AM
10325+ found = NULL;
10326+ head = &spl->head;
10327+ spin_lock(&spl->spin);
10328+ list_for_each_entry(tmp, head, dk_list)
10329+ if (tmp->dk_op.dy_hop == h_op) {
10330+ kref_get(&tmp->dk_kref);
10331+ found = tmp;
10332+ break;
10333+ }
10334+ if (!found)
10335+ list_add_rcu(&key->dk_list, head);
10336+ spin_unlock(&spl->spin);
1facf9fc 10337+
4a4d8108
AM
10338+ if (!found)
10339+ DyPrSym(key);
10340+ return found;
10341+}
10342+
10343+static void dy_free_rcu(struct rcu_head *rcu)
1facf9fc 10344+{
4a4d8108
AM
10345+ struct au_dykey *key;
10346+
10347+ key = container_of(rcu, struct au_dykey, dk_rcu);
10348+ DyPrSym(key);
10349+ kfree(key);
1facf9fc 10350+}
10351+
4a4d8108
AM
10352+static void dy_free(struct kref *kref)
10353+{
10354+ struct au_dykey *key;
10355+ struct au_splhead *spl;
1facf9fc 10356+
4a4d8108
AM
10357+ key = container_of(kref, struct au_dykey, dk_kref);
10358+ spl = dynop + key->dk_op.dy_type;
10359+ au_spl_del_rcu(&key->dk_list, spl);
10360+ call_rcu(&key->dk_rcu, dy_free_rcu);
10361+}
10362+
10363+void au_dy_put(struct au_dykey *key)
1facf9fc 10364+{
4a4d8108
AM
10365+ kref_put(&key->dk_kref, dy_free);
10366+}
1facf9fc 10367+
4a4d8108
AM
10368+/* ---------------------------------------------------------------------- */
10369+
10370+#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
10371+
10372+#ifdef CONFIG_AUFS_DEBUG
10373+#define DyDbgDeclare(cnt) unsigned int cnt = 0
4f0767ce 10374+#define DyDbgInc(cnt) do { cnt++; } while (0)
4a4d8108
AM
10375+#else
10376+#define DyDbgDeclare(cnt) do {} while (0)
10377+#define DyDbgInc(cnt) do {} while (0)
10378+#endif
10379+
10380+#define DySet(func, dst, src, h_op, h_sb) do { \
10381+ DyDbgInc(cnt); \
10382+ if (h_op->func) { \
10383+ if (src.func) \
10384+ dst.func = src.func; \
10385+ else \
10386+ AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
10387+ } \
10388+} while (0)
10389+
10390+#define DySetForce(func, dst, src) do { \
10391+ AuDebugOn(!src.func); \
10392+ DyDbgInc(cnt); \
10393+ dst.func = src.func; \
10394+} while (0)
10395+
10396+#define DySetAop(func) \
10397+ DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
10398+#define DySetAopForce(func) \
10399+ DySetForce(func, dyaop->da_op, aufs_aop)
10400+
10401+static void dy_aop(struct au_dykey *key, const void *h_op,
10402+ struct super_block *h_sb __maybe_unused)
10403+{
10404+ struct au_dyaop *dyaop = (void *)key;
10405+ const struct address_space_operations *h_aop = h_op;
10406+ DyDbgDeclare(cnt);
10407+
10408+ AuDbg("%s\n", au_sbtype(h_sb));
10409+
10410+ DySetAop(writepage);
10411+ DySetAopForce(readpage); /* force */
4a4d8108
AM
10412+ DySetAop(writepages);
10413+ DySetAop(set_page_dirty);
10414+ DySetAop(readpages);
10415+ DySetAop(write_begin);
10416+ DySetAop(write_end);
10417+ DySetAop(bmap);
10418+ DySetAop(invalidatepage);
10419+ DySetAop(releasepage);
027c5e7a 10420+ DySetAop(freepage);
7e9cd9fe 10421+ /* this one will be changed according to an aufs mount option */
4a4d8108 10422+ DySetAop(direct_IO);
4a4d8108
AM
10423+ DySetAop(migratepage);
10424+ DySetAop(launder_page);
10425+ DySetAop(is_partially_uptodate);
392086de 10426+ DySetAop(is_dirty_writeback);
4a4d8108 10427+ DySetAop(error_remove_page);
b4510431
AM
10428+ DySetAop(swap_activate);
10429+ DySetAop(swap_deactivate);
4a4d8108
AM
10430+
10431+ DyDbgSize(cnt, *h_aop);
4a4d8108
AM
10432+}
10433+
4a4d8108
AM
10434+/* ---------------------------------------------------------------------- */
10435+
10436+static void dy_bug(struct kref *kref)
10437+{
10438+ BUG();
10439+}
10440+
10441+static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
10442+{
10443+ struct au_dykey *key, *old;
10444+ struct au_splhead *spl;
b752ccd1 10445+ struct op {
4a4d8108 10446+ unsigned int sz;
b752ccd1
AM
10447+ void (*set)(struct au_dykey *key, const void *h_op,
10448+ struct super_block *h_sb __maybe_unused);
10449+ };
10450+ static const struct op a[] = {
4a4d8108
AM
10451+ [AuDy_AOP] = {
10452+ .sz = sizeof(struct au_dyaop),
b752ccd1 10453+ .set = dy_aop
4a4d8108 10454+ }
b752ccd1
AM
10455+ };
10456+ const struct op *p;
4a4d8108
AM
10457+
10458+ spl = dynop + op->dy_type;
10459+ key = dy_gfind_get(spl, op->dy_hop);
10460+ if (key)
10461+ goto out_add; /* success */
10462+
10463+ p = a + op->dy_type;
10464+ key = kzalloc(p->sz, GFP_NOFS);
10465+ if (unlikely(!key)) {
10466+ key = ERR_PTR(-ENOMEM);
10467+ goto out;
10468+ }
10469+
10470+ key->dk_op.dy_hop = op->dy_hop;
10471+ kref_init(&key->dk_kref);
86dc4139 10472+ p->set(key, op->dy_hop, au_br_sb(br));
4a4d8108
AM
10473+ old = dy_gadd(spl, key);
10474+ if (old) {
10475+ kfree(key);
10476+ key = old;
10477+ }
10478+
10479+out_add:
10480+ old = dy_bradd(br, key);
10481+ if (old)
10482+ /* its ref-count should never be zero here */
10483+ kref_put(&key->dk_kref, dy_bug);
10484+out:
10485+ return key;
10486+}
10487+
10488+/* ---------------------------------------------------------------------- */
10489+/*
10490+ * Aufs prohibits O_DIRECT by defaut even if the branch supports it.
c1595e42 10491+ * This behaviour is necessary to return an error from open(O_DIRECT) instead
4a4d8108
AM
10492+ * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
10493+ * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
10494+ * See the aufs manual in detail.
4a4d8108
AM
10495+ */
10496+static void dy_adx(struct au_dyaop *dyaop, int do_dx)
10497+{
7e9cd9fe 10498+ if (!do_dx)
4a4d8108 10499+ dyaop->da_op.direct_IO = NULL;
7e9cd9fe 10500+ else
4a4d8108 10501+ dyaop->da_op.direct_IO = aufs_aop.direct_IO;
4a4d8108
AM
10502+}
10503+
10504+static struct au_dyaop *dy_aget(struct au_branch *br,
10505+ const struct address_space_operations *h_aop,
10506+ int do_dx)
10507+{
10508+ struct au_dyaop *dyaop;
10509+ struct au_dynop op;
10510+
10511+ op.dy_type = AuDy_AOP;
10512+ op.dy_haop = h_aop;
10513+ dyaop = (void *)dy_get(&op, br);
10514+ if (IS_ERR(dyaop))
10515+ goto out;
10516+ dy_adx(dyaop, do_dx);
10517+
10518+out:
10519+ return dyaop;
10520+}
10521+
10522+int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
10523+ struct inode *h_inode)
10524+{
10525+ int err, do_dx;
10526+ struct super_block *sb;
10527+ struct au_branch *br;
10528+ struct au_dyaop *dyaop;
10529+
10530+ AuDebugOn(!S_ISREG(h_inode->i_mode));
10531+ IiMustWriteLock(inode);
10532+
10533+ sb = inode->i_sb;
10534+ br = au_sbr(sb, bindex);
10535+ do_dx = !!au_opt_test(au_mntflags(sb), DIO);
10536+ dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
10537+ err = PTR_ERR(dyaop);
10538+ if (IS_ERR(dyaop))
10539+ /* unnecessary to call dy_fput() */
10540+ goto out;
10541+
10542+ err = 0;
10543+ inode->i_mapping->a_ops = &dyaop->da_op;
10544+
10545+out:
10546+ return err;
10547+}
10548+
b752ccd1
AM
10549+/*
10550+ * Is it safe to replace a_ops during the inode/file is in operation?
10551+ * Yes, I hope so.
10552+ */
10553+int au_dy_irefresh(struct inode *inode)
10554+{
10555+ int err;
10556+ aufs_bindex_t bstart;
10557+ struct inode *h_inode;
10558+
10559+ err = 0;
10560+ if (S_ISREG(inode->i_mode)) {
10561+ bstart = au_ibstart(inode);
10562+ h_inode = au_h_iptr(inode, bstart);
10563+ err = au_dy_iaop(inode, bstart, h_inode);
10564+ }
10565+ return err;
10566+}
10567+
4a4d8108
AM
10568+void au_dy_arefresh(int do_dx)
10569+{
10570+ struct au_splhead *spl;
10571+ struct list_head *head;
10572+ struct au_dykey *key;
10573+
10574+ spl = dynop + AuDy_AOP;
10575+ head = &spl->head;
10576+ spin_lock(&spl->spin);
10577+ list_for_each_entry(key, head, dk_list)
10578+ dy_adx((void *)key, do_dx);
10579+ spin_unlock(&spl->spin);
10580+}
10581+
4a4d8108
AM
10582+/* ---------------------------------------------------------------------- */
10583+
10584+void __init au_dy_init(void)
10585+{
10586+ int i;
10587+
10588+ /* make sure that 'struct au_dykey *' can be any type */
10589+ BUILD_BUG_ON(offsetof(struct au_dyaop, da_key));
4a4d8108
AM
10590+
10591+ for (i = 0; i < AuDyLast; i++)
10592+ au_spl_init(dynop + i);
10593+}
10594+
10595+void au_dy_fin(void)
10596+{
10597+ int i;
10598+
10599+ for (i = 0; i < AuDyLast; i++)
10600+ WARN_ON(!list_empty(&dynop[i].head));
10601+}
7f207e10
AM
10602diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
10603--- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 10604+++ linux/fs/aufs/dynop.h 2015-09-24 10:47:58.251386326 +0200
7e9cd9fe 10605@@ -0,0 +1,74 @@
4a4d8108 10606+/*
2000de60 10607+ * Copyright (C) 2010-2015 Junjiro R. Okajima
4a4d8108
AM
10608+ *
10609+ * This program, aufs is free software; you can redistribute it and/or modify
10610+ * it under the terms of the GNU General Public License as published by
10611+ * the Free Software Foundation; either version 2 of the License, or
10612+ * (at your option) any later version.
10613+ *
10614+ * This program is distributed in the hope that it will be useful,
10615+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10616+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10617+ * GNU General Public License for more details.
10618+ *
10619+ * You should have received a copy of the GNU General Public License
523b37e3 10620+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
10621+ */
10622+
10623+/*
10624+ * dynamically customizable operations (for regular files only)
10625+ */
10626+
10627+#ifndef __AUFS_DYNOP_H__
10628+#define __AUFS_DYNOP_H__
10629+
10630+#ifdef __KERNEL__
10631+
7e9cd9fe
AM
10632+#include <linux/fs.h>
10633+#include <linux/kref.h>
4a4d8108 10634+
2cbb1c4b 10635+enum {AuDy_AOP, AuDyLast};
4a4d8108
AM
10636+
10637+struct au_dynop {
10638+ int dy_type;
10639+ union {
10640+ const void *dy_hop;
10641+ const struct address_space_operations *dy_haop;
4a4d8108
AM
10642+ };
10643+};
10644+
10645+struct au_dykey {
10646+ union {
10647+ struct list_head dk_list;
10648+ struct rcu_head dk_rcu;
10649+ };
10650+ struct au_dynop dk_op;
10651+
10652+ /*
10653+ * during I am in the branch local array, kref is gotten. when the
10654+ * branch is removed, kref is put.
10655+ */
10656+ struct kref dk_kref;
10657+};
10658+
10659+/* stop unioning since their sizes are very different from each other */
10660+struct au_dyaop {
10661+ struct au_dykey da_key;
10662+ struct address_space_operations da_op; /* not const */
4a4d8108
AM
10663+};
10664+
4a4d8108
AM
10665+/* ---------------------------------------------------------------------- */
10666+
10667+/* dynop.c */
10668+struct au_branch;
10669+void au_dy_put(struct au_dykey *key);
10670+int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
10671+ struct inode *h_inode);
b752ccd1 10672+int au_dy_irefresh(struct inode *inode);
4a4d8108 10673+void au_dy_arefresh(int do_dio);
4a4d8108
AM
10674+
10675+void __init au_dy_init(void);
10676+void au_dy_fin(void);
10677+
4a4d8108
AM
10678+#endif /* __KERNEL__ */
10679+#endif /* __AUFS_DYNOP_H__ */
7f207e10
AM
10680diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
10681--- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 10682+++ linux/fs/aufs/export.c 2015-09-24 10:47:58.251386326 +0200
5527c038 10683@@ -0,0 +1,832 @@
4a4d8108 10684+/*
2000de60 10685+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
10686+ *
10687+ * This program, aufs is free software; you can redistribute it and/or modify
10688+ * it under the terms of the GNU General Public License as published by
10689+ * the Free Software Foundation; either version 2 of the License, or
10690+ * (at your option) any later version.
10691+ *
10692+ * This program is distributed in the hope that it will be useful,
10693+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
10694+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10695+ * GNU General Public License for more details.
10696+ *
10697+ * You should have received a copy of the GNU General Public License
523b37e3 10698+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
10699+ */
10700+
10701+/*
10702+ * export via nfs
10703+ */
10704+
10705+#include <linux/exportfs.h>
7eafdf33 10706+#include <linux/fs_struct.h>
4a4d8108
AM
10707+#include <linux/namei.h>
10708+#include <linux/nsproxy.h>
10709+#include <linux/random.h>
10710+#include <linux/writeback.h>
7eafdf33 10711+#include "../fs/mount.h"
4a4d8108
AM
10712+#include "aufs.h"
10713+
10714+union conv {
10715+#ifdef CONFIG_AUFS_INO_T_64
10716+ __u32 a[2];
10717+#else
10718+ __u32 a[1];
10719+#endif
10720+ ino_t ino;
10721+};
10722+
10723+static ino_t decode_ino(__u32 *a)
10724+{
10725+ union conv u;
10726+
10727+ BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
10728+ u.a[0] = a[0];
10729+#ifdef CONFIG_AUFS_INO_T_64
10730+ u.a[1] = a[1];
10731+#endif
10732+ return u.ino;
10733+}
10734+
10735+static void encode_ino(__u32 *a, ino_t ino)
10736+{
10737+ union conv u;
10738+
10739+ u.ino = ino;
10740+ a[0] = u.a[0];
10741+#ifdef CONFIG_AUFS_INO_T_64
10742+ a[1] = u.a[1];
10743+#endif
10744+}
10745+
10746+/* NFS file handle */
10747+enum {
10748+ Fh_br_id,
10749+ Fh_sigen,
10750+#ifdef CONFIG_AUFS_INO_T_64
10751+ /* support 64bit inode number */
10752+ Fh_ino1,
10753+ Fh_ino2,
10754+ Fh_dir_ino1,
10755+ Fh_dir_ino2,
10756+#else
10757+ Fh_ino1,
10758+ Fh_dir_ino1,
10759+#endif
10760+ Fh_igen,
10761+ Fh_h_type,
10762+ Fh_tail,
10763+
10764+ Fh_ino = Fh_ino1,
10765+ Fh_dir_ino = Fh_dir_ino1
10766+};
10767+
10768+static int au_test_anon(struct dentry *dentry)
10769+{
027c5e7a 10770+ /* note: read d_flags without d_lock */
4a4d8108
AM
10771+ return !!(dentry->d_flags & DCACHE_DISCONNECTED);
10772+}
10773+
a2a7ad62
AM
10774+int au_test_nfsd(void)
10775+{
10776+ int ret;
10777+ struct task_struct *tsk = current;
10778+ char comm[sizeof(tsk->comm)];
10779+
10780+ ret = 0;
10781+ if (tsk->flags & PF_KTHREAD) {
10782+ get_task_comm(comm, tsk);
10783+ ret = !strcmp(comm, "nfsd");
10784+ }
10785+
10786+ return ret;
10787+}
10788+
4a4d8108
AM
10789+/* ---------------------------------------------------------------------- */
10790+/* inode generation external table */
10791+
b752ccd1 10792+void au_xigen_inc(struct inode *inode)
4a4d8108 10793+{
4a4d8108
AM
10794+ loff_t pos;
10795+ ssize_t sz;
10796+ __u32 igen;
10797+ struct super_block *sb;
10798+ struct au_sbinfo *sbinfo;
10799+
4a4d8108 10800+ sb = inode->i_sb;
b752ccd1 10801+ AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
1facf9fc 10802+
b752ccd1 10803+ sbinfo = au_sbi(sb);
1facf9fc 10804+ pos = inode->i_ino;
10805+ pos *= sizeof(igen);
10806+ igen = inode->i_generation + 1;
1facf9fc 10807+ sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xigen, &igen,
10808+ sizeof(igen), &pos);
10809+ if (sz == sizeof(igen))
b752ccd1 10810+ return; /* success */
1facf9fc 10811+
b752ccd1 10812+ if (unlikely(sz >= 0))
1facf9fc 10813+ AuIOErr("xigen error (%zd)\n", sz);
1facf9fc 10814+}
10815+
10816+int au_xigen_new(struct inode *inode)
10817+{
10818+ int err;
10819+ loff_t pos;
10820+ ssize_t sz;
10821+ struct super_block *sb;
10822+ struct au_sbinfo *sbinfo;
10823+ struct file *file;
10824+
10825+ err = 0;
10826+ /* todo: dirty, at mount time */
10827+ if (inode->i_ino == AUFS_ROOT_INO)
10828+ goto out;
10829+ sb = inode->i_sb;
dece6358 10830+ SiMustAnyLock(sb);
1facf9fc 10831+ if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
10832+ goto out;
10833+
10834+ err = -EFBIG;
10835+ pos = inode->i_ino;
10836+ if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
10837+ AuIOErr1("too large i%lld\n", pos);
10838+ goto out;
10839+ }
10840+ pos *= sizeof(inode->i_generation);
10841+
10842+ err = 0;
10843+ sbinfo = au_sbi(sb);
10844+ file = sbinfo->si_xigen;
10845+ BUG_ON(!file);
10846+
c06a8ce3 10847+ if (vfsub_f_size_read(file)
1facf9fc 10848+ < pos + sizeof(inode->i_generation)) {
10849+ inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
10850+ sz = xino_fwrite(sbinfo->si_xwrite, file, &inode->i_generation,
10851+ sizeof(inode->i_generation), &pos);
10852+ } else
10853+ sz = xino_fread(sbinfo->si_xread, file, &inode->i_generation,
10854+ sizeof(inode->i_generation), &pos);
10855+ if (sz == sizeof(inode->i_generation))
10856+ goto out; /* success */
10857+
10858+ err = sz;
10859+ if (unlikely(sz >= 0)) {
10860+ err = -EIO;
10861+ AuIOErr("xigen error (%zd)\n", sz);
10862+ }
10863+
4f0767ce 10864+out:
1facf9fc 10865+ return err;
10866+}
10867+
10868+int au_xigen_set(struct super_block *sb, struct file *base)
10869+{
10870+ int err;
10871+ struct au_sbinfo *sbinfo;
10872+ struct file *file;
10873+
dece6358
AM
10874+ SiMustWriteLock(sb);
10875+
1facf9fc 10876+ sbinfo = au_sbi(sb);
10877+ file = au_xino_create2(base, sbinfo->si_xigen);
10878+ err = PTR_ERR(file);
10879+ if (IS_ERR(file))
10880+ goto out;
10881+ err = 0;
10882+ if (sbinfo->si_xigen)
10883+ fput(sbinfo->si_xigen);
10884+ sbinfo->si_xigen = file;
10885+
4f0767ce 10886+out:
1facf9fc 10887+ return err;
10888+}
10889+
10890+void au_xigen_clr(struct super_block *sb)
10891+{
10892+ struct au_sbinfo *sbinfo;
10893+
dece6358
AM
10894+ SiMustWriteLock(sb);
10895+
1facf9fc 10896+ sbinfo = au_sbi(sb);
10897+ if (sbinfo->si_xigen) {
10898+ fput(sbinfo->si_xigen);
10899+ sbinfo->si_xigen = NULL;
10900+ }
10901+}
10902+
10903+/* ---------------------------------------------------------------------- */
10904+
10905+static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
10906+ ino_t dir_ino)
10907+{
10908+ struct dentry *dentry, *d;
10909+ struct inode *inode;
10910+ unsigned int sigen;
10911+
10912+ dentry = NULL;
10913+ inode = ilookup(sb, ino);
10914+ if (!inode)
10915+ goto out;
10916+
10917+ dentry = ERR_PTR(-ESTALE);
10918+ sigen = au_sigen(sb);
10919+ if (unlikely(is_bad_inode(inode)
10920+ || IS_DEADDIR(inode)
537831f9 10921+ || sigen != au_iigen(inode, NULL)))
1facf9fc 10922+ goto out_iput;
10923+
10924+ dentry = NULL;
10925+ if (!dir_ino || S_ISDIR(inode->i_mode))
10926+ dentry = d_find_alias(inode);
10927+ else {
027c5e7a 10928+ spin_lock(&inode->i_lock);
c1595e42 10929+ hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
027c5e7a 10930+ spin_lock(&d->d_lock);
1facf9fc 10931+ if (!au_test_anon(d)
5527c038 10932+ && d_inode(d->d_parent)->i_ino == dir_ino) {
027c5e7a
AM
10933+ dentry = dget_dlock(d);
10934+ spin_unlock(&d->d_lock);
1facf9fc 10935+ break;
10936+ }
027c5e7a
AM
10937+ spin_unlock(&d->d_lock);
10938+ }
10939+ spin_unlock(&inode->i_lock);
1facf9fc 10940+ }
027c5e7a 10941+ if (unlikely(dentry && au_digen_test(dentry, sigen))) {
2cbb1c4b 10942+ /* need to refresh */
1facf9fc 10943+ dput(dentry);
2cbb1c4b 10944+ dentry = NULL;
1facf9fc 10945+ }
10946+
4f0767ce 10947+out_iput:
1facf9fc 10948+ iput(inode);
4f0767ce 10949+out:
2cbb1c4b 10950+ AuTraceErrPtr(dentry);
1facf9fc 10951+ return dentry;
10952+}
10953+
10954+/* ---------------------------------------------------------------------- */
10955+
10956+/* todo: dirty? */
10957+/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
4a4d8108
AM
10958+
10959+struct au_compare_mnt_args {
10960+ /* input */
10961+ struct super_block *sb;
10962+
10963+ /* output */
10964+ struct vfsmount *mnt;
10965+};
10966+
10967+static int au_compare_mnt(struct vfsmount *mnt, void *arg)
10968+{
10969+ struct au_compare_mnt_args *a = arg;
10970+
10971+ if (mnt->mnt_sb != a->sb)
10972+ return 0;
10973+ a->mnt = mntget(mnt);
10974+ return 1;
10975+}
10976+
1facf9fc 10977+static struct vfsmount *au_mnt_get(struct super_block *sb)
10978+{
4a4d8108 10979+ int err;
7eafdf33 10980+ struct path root;
4a4d8108
AM
10981+ struct au_compare_mnt_args args = {
10982+ .sb = sb
10983+ };
1facf9fc 10984+
7eafdf33 10985+ get_fs_root(current->fs, &root);
523b37e3 10986+ rcu_read_lock();
7eafdf33 10987+ err = iterate_mounts(au_compare_mnt, &args, root.mnt);
523b37e3 10988+ rcu_read_unlock();
7eafdf33 10989+ path_put(&root);
4a4d8108
AM
10990+ AuDebugOn(!err);
10991+ AuDebugOn(!args.mnt);
10992+ return args.mnt;
1facf9fc 10993+}
10994+
10995+struct au_nfsd_si_lock {
4a4d8108 10996+ unsigned int sigen;
027c5e7a 10997+ aufs_bindex_t bindex, br_id;
1facf9fc 10998+ unsigned char force_lock;
10999+};
11000+
027c5e7a
AM
11001+static int si_nfsd_read_lock(struct super_block *sb,
11002+ struct au_nfsd_si_lock *nsi_lock)
1facf9fc 11003+{
027c5e7a 11004+ int err;
1facf9fc 11005+ aufs_bindex_t bindex;
11006+
11007+ si_read_lock(sb, AuLock_FLUSH);
11008+
11009+ /* branch id may be wrapped around */
027c5e7a 11010+ err = 0;
1facf9fc 11011+ bindex = au_br_index(sb, nsi_lock->br_id);
11012+ if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
11013+ goto out; /* success */
11014+
027c5e7a
AM
11015+ err = -ESTALE;
11016+ bindex = -1;
1facf9fc 11017+ if (!nsi_lock->force_lock)
11018+ si_read_unlock(sb);
1facf9fc 11019+
4f0767ce 11020+out:
027c5e7a
AM
11021+ nsi_lock->bindex = bindex;
11022+ return err;
1facf9fc 11023+}
11024+
11025+struct find_name_by_ino {
392086de 11026+ struct dir_context ctx;
1facf9fc 11027+ int called, found;
11028+ ino_t ino;
11029+ char *name;
11030+ int namelen;
11031+};
11032+
11033+static int
392086de
AM
11034+find_name_by_ino(struct dir_context *ctx, const char *name, int namelen,
11035+ loff_t offset, u64 ino, unsigned int d_type)
1facf9fc 11036+{
392086de
AM
11037+ struct find_name_by_ino *a = container_of(ctx, struct find_name_by_ino,
11038+ ctx);
1facf9fc 11039+
11040+ a->called++;
11041+ if (a->ino != ino)
11042+ return 0;
11043+
11044+ memcpy(a->name, name, namelen);
11045+ a->namelen = namelen;
11046+ a->found = 1;
11047+ return 1;
11048+}
11049+
11050+static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
11051+ struct au_nfsd_si_lock *nsi_lock)
11052+{
11053+ struct dentry *dentry, *parent;
11054+ struct file *file;
11055+ struct inode *dir;
392086de
AM
11056+ struct find_name_by_ino arg = {
11057+ .ctx = {
2000de60 11058+ .actor = find_name_by_ino
392086de
AM
11059+ }
11060+ };
1facf9fc 11061+ int err;
11062+
11063+ parent = path->dentry;
11064+ if (nsi_lock)
11065+ si_read_unlock(parent->d_sb);
4a4d8108 11066+ file = vfsub_dentry_open(path, au_dir_roflags);
1facf9fc 11067+ dentry = (void *)file;
11068+ if (IS_ERR(file))
11069+ goto out;
11070+
11071+ dentry = ERR_PTR(-ENOMEM);
537831f9 11072+ arg.name = (void *)__get_free_page(GFP_NOFS);
1facf9fc 11073+ if (unlikely(!arg.name))
11074+ goto out_file;
11075+ arg.ino = ino;
11076+ arg.found = 0;
11077+ do {
11078+ arg.called = 0;
11079+ /* smp_mb(); */
392086de 11080+ err = vfsub_iterate_dir(file, &arg.ctx);
1facf9fc 11081+ } while (!err && !arg.found && arg.called);
11082+ dentry = ERR_PTR(err);
11083+ if (unlikely(err))
11084+ goto out_name;
1716fcea
AM
11085+ /* instead of ENOENT */
11086+ dentry = ERR_PTR(-ESTALE);
1facf9fc 11087+ if (!arg.found)
11088+ goto out_name;
11089+
b4510431 11090+ /* do not call vfsub_lkup_one() */
5527c038 11091+ dir = d_inode(parent);
1facf9fc 11092+ mutex_lock(&dir->i_mutex);
11093+ dentry = vfsub_lookup_one_len(arg.name, parent, arg.namelen);
11094+ mutex_unlock(&dir->i_mutex);
11095+ AuTraceErrPtr(dentry);
11096+ if (IS_ERR(dentry))
11097+ goto out_name;
11098+ AuDebugOn(au_test_anon(dentry));
5527c038 11099+ if (unlikely(d_really_is_negative(dentry))) {
1facf9fc 11100+ dput(dentry);
11101+ dentry = ERR_PTR(-ENOENT);
11102+ }
11103+
4f0767ce 11104+out_name:
537831f9 11105+ free_page((unsigned long)arg.name);
4f0767ce 11106+out_file:
1facf9fc 11107+ fput(file);
4f0767ce 11108+out:
1facf9fc 11109+ if (unlikely(nsi_lock
11110+ && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
11111+ if (!IS_ERR(dentry)) {
11112+ dput(dentry);
11113+ dentry = ERR_PTR(-ESTALE);
11114+ }
11115+ AuTraceErrPtr(dentry);
11116+ return dentry;
11117+}
11118+
11119+static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
11120+ ino_t dir_ino,
11121+ struct au_nfsd_si_lock *nsi_lock)
11122+{
11123+ struct dentry *dentry;
11124+ struct path path;
11125+
11126+ if (dir_ino != AUFS_ROOT_INO) {
11127+ path.dentry = decode_by_ino(sb, dir_ino, 0);
11128+ dentry = path.dentry;
11129+ if (!path.dentry || IS_ERR(path.dentry))
11130+ goto out;
11131+ AuDebugOn(au_test_anon(path.dentry));
11132+ } else
11133+ path.dentry = dget(sb->s_root);
11134+
11135+ path.mnt = au_mnt_get(sb);
11136+ dentry = au_lkup_by_ino(&path, ino, nsi_lock);
11137+ path_put(&path);
11138+
4f0767ce 11139+out:
1facf9fc 11140+ AuTraceErrPtr(dentry);
11141+ return dentry;
11142+}
11143+
11144+/* ---------------------------------------------------------------------- */
11145+
11146+static int h_acceptable(void *expv, struct dentry *dentry)
11147+{
11148+ return 1;
11149+}
11150+
11151+static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
11152+ char *buf, int len, struct super_block *sb)
11153+{
11154+ char *p;
11155+ int n;
11156+ struct path path;
11157+
11158+ p = d_path(h_rootpath, buf, len);
11159+ if (IS_ERR(p))
11160+ goto out;
11161+ n = strlen(p);
11162+
11163+ path.mnt = h_rootpath->mnt;
11164+ path.dentry = h_parent;
11165+ p = d_path(&path, buf, len);
11166+ if (IS_ERR(p))
11167+ goto out;
11168+ if (n != 1)
11169+ p += n;
11170+
11171+ path.mnt = au_mnt_get(sb);
11172+ path.dentry = sb->s_root;
11173+ p = d_path(&path, buf, len - strlen(p));
11174+ mntput(path.mnt);
11175+ if (IS_ERR(p))
11176+ goto out;
11177+ if (n != 1)
11178+ p[strlen(p)] = '/';
11179+
4f0767ce 11180+out:
1facf9fc 11181+ AuTraceErrPtr(p);
11182+ return p;
11183+}
11184+
11185+static
027c5e7a
AM
11186+struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
11187+ int fh_len, struct au_nfsd_si_lock *nsi_lock)
1facf9fc 11188+{
11189+ struct dentry *dentry, *h_parent, *root;
11190+ struct super_block *h_sb;
11191+ char *pathname, *p;
11192+ struct vfsmount *h_mnt;
11193+ struct au_branch *br;
11194+ int err;
11195+ struct path path;
11196+
027c5e7a 11197+ br = au_sbr(sb, nsi_lock->bindex);
86dc4139 11198+ h_mnt = au_br_mnt(br);
1facf9fc 11199+ h_sb = h_mnt->mnt_sb;
11200+ /* todo: call lower fh_to_dentry()? fh_to_parent()? */
11201+ h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
11202+ fh_len - Fh_tail, fh[Fh_h_type],
11203+ h_acceptable, /*context*/NULL);
11204+ dentry = h_parent;
11205+ if (unlikely(!h_parent || IS_ERR(h_parent))) {
11206+ AuWarn1("%s decode_fh failed, %ld\n",
11207+ au_sbtype(h_sb), PTR_ERR(h_parent));
11208+ goto out;
11209+ }
11210+ dentry = NULL;
11211+ if (unlikely(au_test_anon(h_parent))) {
11212+ AuWarn1("%s decode_fh returned a disconnected dentry\n",
11213+ au_sbtype(h_sb));
11214+ goto out_h_parent;
11215+ }
11216+
11217+ dentry = ERR_PTR(-ENOMEM);
11218+ pathname = (void *)__get_free_page(GFP_NOFS);
11219+ if (unlikely(!pathname))
11220+ goto out_h_parent;
11221+
11222+ root = sb->s_root;
11223+ path.mnt = h_mnt;
11224+ di_read_lock_parent(root, !AuLock_IR);
027c5e7a 11225+ path.dentry = au_h_dptr(root, nsi_lock->bindex);
1facf9fc 11226+ di_read_unlock(root, !AuLock_IR);
11227+ p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
11228+ dentry = (void *)p;
11229+ if (IS_ERR(p))
11230+ goto out_pathname;
11231+
11232+ si_read_unlock(sb);
11233+ err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
11234+ dentry = ERR_PTR(err);
11235+ if (unlikely(err))
11236+ goto out_relock;
11237+
11238+ dentry = ERR_PTR(-ENOENT);
11239+ AuDebugOn(au_test_anon(path.dentry));
5527c038 11240+ if (unlikely(d_really_is_negative(path.dentry)))
1facf9fc 11241+ goto out_path;
11242+
5527c038 11243+ if (ino != d_inode(path.dentry)->i_ino)
1facf9fc 11244+ dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
11245+ else
11246+ dentry = dget(path.dentry);
11247+
4f0767ce 11248+out_path:
1facf9fc 11249+ path_put(&path);
4f0767ce 11250+out_relock:
1facf9fc 11251+ if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
11252+ if (!IS_ERR(dentry)) {
11253+ dput(dentry);
11254+ dentry = ERR_PTR(-ESTALE);
11255+ }
4f0767ce 11256+out_pathname:
1facf9fc 11257+ free_page((unsigned long)pathname);
4f0767ce 11258+out_h_parent:
1facf9fc 11259+ dput(h_parent);
4f0767ce 11260+out:
1facf9fc 11261+ AuTraceErrPtr(dentry);
11262+ return dentry;
11263+}
11264+
11265+/* ---------------------------------------------------------------------- */
11266+
11267+static struct dentry *
11268+aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
11269+ int fh_type)
11270+{
11271+ struct dentry *dentry;
11272+ __u32 *fh = fid->raw;
027c5e7a 11273+ struct au_branch *br;
1facf9fc 11274+ ino_t ino, dir_ino;
1facf9fc 11275+ struct au_nfsd_si_lock nsi_lock = {
1facf9fc 11276+ .force_lock = 0
11277+ };
11278+
1facf9fc 11279+ dentry = ERR_PTR(-ESTALE);
4a4d8108
AM
11280+ /* it should never happen, but the file handle is unreliable */
11281+ if (unlikely(fh_len < Fh_tail))
11282+ goto out;
11283+ nsi_lock.sigen = fh[Fh_sigen];
11284+ nsi_lock.br_id = fh[Fh_br_id];
11285+
1facf9fc 11286+ /* branch id may be wrapped around */
027c5e7a
AM
11287+ br = NULL;
11288+ if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
1facf9fc 11289+ goto out;
11290+ nsi_lock.force_lock = 1;
11291+
11292+ /* is this inode still cached? */
11293+ ino = decode_ino(fh + Fh_ino);
4a4d8108
AM
11294+ /* it should never happen */
11295+ if (unlikely(ino == AUFS_ROOT_INO))
11296+ goto out;
11297+
1facf9fc 11298+ dir_ino = decode_ino(fh + Fh_dir_ino);
11299+ dentry = decode_by_ino(sb, ino, dir_ino);
11300+ if (IS_ERR(dentry))
11301+ goto out_unlock;
11302+ if (dentry)
11303+ goto accept;
11304+
11305+ /* is the parent dir cached? */
027c5e7a
AM
11306+ br = au_sbr(sb, nsi_lock.bindex);
11307+ atomic_inc(&br->br_count);
1facf9fc 11308+ dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
11309+ if (IS_ERR(dentry))
11310+ goto out_unlock;
11311+ if (dentry)
11312+ goto accept;
11313+
11314+ /* lookup path */
027c5e7a 11315+ dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
1facf9fc 11316+ if (IS_ERR(dentry))
11317+ goto out_unlock;
11318+ if (unlikely(!dentry))
11319+ /* todo?: make it ESTALE */
11320+ goto out_unlock;
11321+
4f0767ce 11322+accept:
027c5e7a 11323+ if (!au_digen_test(dentry, au_sigen(sb))
5527c038 11324+ && d_inode(dentry)->i_generation == fh[Fh_igen])
1facf9fc 11325+ goto out_unlock; /* success */
11326+
11327+ dput(dentry);
11328+ dentry = ERR_PTR(-ESTALE);
4f0767ce 11329+out_unlock:
027c5e7a
AM
11330+ if (br)
11331+ atomic_dec(&br->br_count);
1facf9fc 11332+ si_read_unlock(sb);
4f0767ce 11333+out:
1facf9fc 11334+ AuTraceErrPtr(dentry);
11335+ return dentry;
11336+}
11337+
11338+#if 0 /* reserved for future use */
11339+/* support subtreecheck option */
11340+static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
11341+ int fh_len, int fh_type)
11342+{
11343+ struct dentry *parent;
11344+ __u32 *fh = fid->raw;
11345+ ino_t dir_ino;
11346+
11347+ dir_ino = decode_ino(fh + Fh_dir_ino);
11348+ parent = decode_by_ino(sb, dir_ino, 0);
11349+ if (IS_ERR(parent))
11350+ goto out;
11351+ if (!parent)
11352+ parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
11353+ dir_ino, fh, fh_len);
11354+
4f0767ce 11355+out:
1facf9fc 11356+ AuTraceErrPtr(parent);
11357+ return parent;
11358+}
11359+#endif
11360+
11361+/* ---------------------------------------------------------------------- */
11362+
0c3ec466
AM
11363+static int aufs_encode_fh(struct inode *inode, __u32 *fh, int *max_len,
11364+ struct inode *dir)
1facf9fc 11365+{
11366+ int err;
0c3ec466 11367+ aufs_bindex_t bindex;
1facf9fc 11368+ struct super_block *sb, *h_sb;
0c3ec466
AM
11369+ struct dentry *dentry, *parent, *h_parent;
11370+ struct inode *h_dir;
1facf9fc 11371+ struct au_branch *br;
11372+
1facf9fc 11373+ err = -ENOSPC;
11374+ if (unlikely(*max_len <= Fh_tail)) {
11375+ AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
11376+ goto out;
11377+ }
11378+
11379+ err = FILEID_ROOT;
0c3ec466
AM
11380+ if (inode->i_ino == AUFS_ROOT_INO) {
11381+ AuDebugOn(inode->i_ino != AUFS_ROOT_INO);
1facf9fc 11382+ goto out;
11383+ }
11384+
1facf9fc 11385+ h_parent = NULL;
0c3ec466
AM
11386+ sb = inode->i_sb;
11387+ err = si_read_lock(sb, AuLock_FLUSH);
027c5e7a
AM
11388+ if (unlikely(err))
11389+ goto out;
11390+
1facf9fc 11391+#ifdef CONFIG_AUFS_DEBUG
11392+ if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
11393+ AuWarn1("NFS-exporting requires xino\n");
11394+#endif
027c5e7a 11395+ err = -EIO;
0c3ec466
AM
11396+ parent = NULL;
11397+ ii_read_lock_child(inode);
11398+ bindex = au_ibstart(inode);
11399+ if (!dir) {
c1595e42 11400+ dentry = d_find_any_alias(inode);
0c3ec466
AM
11401+ if (unlikely(!dentry))
11402+ goto out_unlock;
11403+ AuDebugOn(au_test_anon(dentry));
11404+ parent = dget_parent(dentry);
11405+ dput(dentry);
11406+ if (unlikely(!parent))
11407+ goto out_unlock;
5527c038
JR
11408+ if (d_really_is_positive(parent))
11409+ dir = d_inode(parent);
1facf9fc 11410+ }
0c3ec466
AM
11411+
11412+ ii_read_lock_parent(dir);
11413+ h_dir = au_h_iptr(dir, bindex);
11414+ ii_read_unlock(dir);
11415+ if (unlikely(!h_dir))
11416+ goto out_parent;
c1595e42 11417+ h_parent = d_find_any_alias(h_dir);
1facf9fc 11418+ if (unlikely(!h_parent))
0c3ec466 11419+ goto out_hparent;
1facf9fc 11420+
11421+ err = -EPERM;
11422+ br = au_sbr(sb, bindex);
86dc4139 11423+ h_sb = au_br_sb(br);
1facf9fc 11424+ if (unlikely(!h_sb->s_export_op)) {
11425+ AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
0c3ec466 11426+ goto out_hparent;
1facf9fc 11427+ }
11428+
11429+ fh[Fh_br_id] = br->br_id;
11430+ fh[Fh_sigen] = au_sigen(sb);
11431+ encode_ino(fh + Fh_ino, inode->i_ino);
0c3ec466 11432+ encode_ino(fh + Fh_dir_ino, dir->i_ino);
1facf9fc 11433+ fh[Fh_igen] = inode->i_generation;
11434+
11435+ *max_len -= Fh_tail;
11436+ fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
11437+ max_len,
11438+ /*connectable or subtreecheck*/0);
11439+ err = fh[Fh_h_type];
11440+ *max_len += Fh_tail;
11441+ /* todo: macros? */
1716fcea 11442+ if (err != FILEID_INVALID)
1facf9fc 11443+ err = 99;
11444+ else
11445+ AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
11446+
0c3ec466 11447+out_hparent:
1facf9fc 11448+ dput(h_parent);
0c3ec466 11449+out_parent:
1facf9fc 11450+ dput(parent);
0c3ec466
AM
11451+out_unlock:
11452+ ii_read_unlock(inode);
11453+ si_read_unlock(sb);
4f0767ce 11454+out:
1facf9fc 11455+ if (unlikely(err < 0))
1716fcea 11456+ err = FILEID_INVALID;
1facf9fc 11457+ return err;
11458+}
11459+
11460+/* ---------------------------------------------------------------------- */
11461+
4a4d8108
AM
11462+static int aufs_commit_metadata(struct inode *inode)
11463+{
11464+ int err;
11465+ aufs_bindex_t bindex;
11466+ struct super_block *sb;
11467+ struct inode *h_inode;
11468+ int (*f)(struct inode *inode);
11469+
11470+ sb = inode->i_sb;
e49829fe 11471+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108
AM
11472+ ii_write_lock_child(inode);
11473+ bindex = au_ibstart(inode);
11474+ AuDebugOn(bindex < 0);
11475+ h_inode = au_h_iptr(inode, bindex);
11476+
11477+ f = h_inode->i_sb->s_export_op->commit_metadata;
11478+ if (f)
11479+ err = f(h_inode);
11480+ else {
11481+ struct writeback_control wbc = {
11482+ .sync_mode = WB_SYNC_ALL,
11483+ .nr_to_write = 0 /* metadata only */
11484+ };
11485+
11486+ err = sync_inode(h_inode, &wbc);
11487+ }
11488+
11489+ au_cpup_attr_timesizes(inode);
11490+ ii_write_unlock(inode);
11491+ si_read_unlock(sb);
11492+ return err;
11493+}
11494+
11495+/* ---------------------------------------------------------------------- */
11496+
1facf9fc 11497+static struct export_operations aufs_export_op = {
4a4d8108 11498+ .fh_to_dentry = aufs_fh_to_dentry,
1facf9fc 11499+ /* .fh_to_parent = aufs_fh_to_parent, */
4a4d8108
AM
11500+ .encode_fh = aufs_encode_fh,
11501+ .commit_metadata = aufs_commit_metadata
1facf9fc 11502+};
11503+
11504+void au_export_init(struct super_block *sb)
11505+{
11506+ struct au_sbinfo *sbinfo;
11507+ __u32 u;
11508+
11509+ sb->s_export_op = &aufs_export_op;
11510+ sbinfo = au_sbi(sb);
11511+ sbinfo->si_xigen = NULL;
11512+ get_random_bytes(&u, sizeof(u));
11513+ BUILD_BUG_ON(sizeof(u) != sizeof(int));
11514+ atomic_set(&sbinfo->si_xigen_next, u);
11515+}
076b876e
AM
11516diff -urN /usr/share/empty/fs/aufs/fhsm.c linux/fs/aufs/fhsm.c
11517--- /usr/share/empty/fs/aufs/fhsm.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 11518+++ linux/fs/aufs/fhsm.c 2015-09-24 10:47:58.251386326 +0200
c1595e42 11519@@ -0,0 +1,426 @@
076b876e 11520+/*
2000de60 11521+ * Copyright (C) 2011-2015 Junjiro R. Okajima
076b876e
AM
11522+ *
11523+ * This program, aufs is free software; you can redistribute it and/or modify
11524+ * it under the terms of the GNU General Public License as published by
11525+ * the Free Software Foundation; either version 2 of the License, or
11526+ * (at your option) any later version.
11527+ *
11528+ * This program is distributed in the hope that it will be useful,
11529+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
11530+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11531+ * GNU General Public License for more details.
11532+ *
11533+ * You should have received a copy of the GNU General Public License
11534+ * along with this program; if not, write to the Free Software
11535+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
11536+ */
11537+
11538+/*
11539+ * File-based Hierarchy Storage Management
11540+ */
11541+
11542+#include <linux/anon_inodes.h>
11543+#include <linux/poll.h>
11544+#include <linux/seq_file.h>
11545+#include <linux/statfs.h>
11546+#include "aufs.h"
11547+
c1595e42
JR
11548+static aufs_bindex_t au_fhsm_bottom(struct super_block *sb)
11549+{
11550+ struct au_sbinfo *sbinfo;
11551+ struct au_fhsm *fhsm;
11552+
11553+ SiMustAnyLock(sb);
11554+
11555+ sbinfo = au_sbi(sb);
11556+ fhsm = &sbinfo->si_fhsm;
11557+ AuDebugOn(!fhsm);
11558+ return fhsm->fhsm_bottom;
11559+}
11560+
11561+void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex)
11562+{
11563+ struct au_sbinfo *sbinfo;
11564+ struct au_fhsm *fhsm;
11565+
11566+ SiMustWriteLock(sb);
11567+
11568+ sbinfo = au_sbi(sb);
11569+ fhsm = &sbinfo->si_fhsm;
11570+ AuDebugOn(!fhsm);
11571+ fhsm->fhsm_bottom = bindex;
11572+}
11573+
11574+/* ---------------------------------------------------------------------- */
11575+
076b876e
AM
11576+static int au_fhsm_test_jiffy(struct au_sbinfo *sbinfo, struct au_branch *br)
11577+{
11578+ struct au_br_fhsm *bf;
11579+
11580+ bf = br->br_fhsm;
11581+ MtxMustLock(&bf->bf_lock);
11582+
11583+ return !bf->bf_readable
11584+ || time_after(jiffies,
11585+ bf->bf_jiffy + sbinfo->si_fhsm.fhsm_expire);
11586+}
11587+
11588+/* ---------------------------------------------------------------------- */
11589+
11590+static void au_fhsm_notify(struct super_block *sb, int val)
11591+{
11592+ struct au_sbinfo *sbinfo;
11593+ struct au_fhsm *fhsm;
11594+
11595+ SiMustAnyLock(sb);
11596+
11597+ sbinfo = au_sbi(sb);
11598+ fhsm = &sbinfo->si_fhsm;
11599+ if (au_fhsm_pid(fhsm)
11600+ && atomic_read(&fhsm->fhsm_readable) != -1) {
11601+ atomic_set(&fhsm->fhsm_readable, val);
11602+ if (val)
11603+ wake_up(&fhsm->fhsm_wqh);
11604+ }
11605+}
11606+
11607+static int au_fhsm_stfs(struct super_block *sb, aufs_bindex_t bindex,
11608+ struct aufs_stfs *rstfs, int do_lock, int do_notify)
11609+{
11610+ int err;
11611+ struct au_branch *br;
11612+ struct au_br_fhsm *bf;
11613+
11614+ br = au_sbr(sb, bindex);
11615+ AuDebugOn(au_br_rdonly(br));
11616+ bf = br->br_fhsm;
11617+ AuDebugOn(!bf);
11618+
11619+ if (do_lock)
11620+ mutex_lock(&bf->bf_lock);
11621+ else
11622+ MtxMustLock(&bf->bf_lock);
11623+
11624+ /* sb->s_root for NFS is unreliable */
11625+ err = au_br_stfs(br, &bf->bf_stfs);
11626+ if (unlikely(err)) {
11627+ AuErr1("FHSM failed (%d), b%d, ignored.\n", bindex, err);
11628+ goto out;
11629+ }
11630+
11631+ bf->bf_jiffy = jiffies;
11632+ bf->bf_readable = 1;
11633+ if (do_notify)
11634+ au_fhsm_notify(sb, /*val*/1);
11635+ if (rstfs)
11636+ *rstfs = bf->bf_stfs;
11637+
11638+out:
11639+ if (do_lock)
11640+ mutex_unlock(&bf->bf_lock);
11641+ au_fhsm_notify(sb, /*val*/1);
11642+
11643+ return err;
11644+}
11645+
11646+void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force)
11647+{
11648+ int err;
076b876e
AM
11649+ struct au_sbinfo *sbinfo;
11650+ struct au_fhsm *fhsm;
11651+ struct au_branch *br;
11652+ struct au_br_fhsm *bf;
11653+
11654+ AuDbg("b%d, force %d\n", bindex, force);
11655+ SiMustAnyLock(sb);
11656+
11657+ sbinfo = au_sbi(sb);
11658+ fhsm = &sbinfo->si_fhsm;
c1595e42
JR
11659+ if (!au_ftest_si(sbinfo, FHSM)
11660+ || fhsm->fhsm_bottom == bindex)
076b876e
AM
11661+ return;
11662+
11663+ br = au_sbr(sb, bindex);
11664+ bf = br->br_fhsm;
11665+ AuDebugOn(!bf);
11666+ mutex_lock(&bf->bf_lock);
11667+ if (force
11668+ || au_fhsm_pid(fhsm)
11669+ || au_fhsm_test_jiffy(sbinfo, br))
11670+ err = au_fhsm_stfs(sb, bindex, /*rstfs*/NULL, /*do_lock*/0,
11671+ /*do_notify*/1);
11672+ mutex_unlock(&bf->bf_lock);
11673+}
11674+
11675+void au_fhsm_wrote_all(struct super_block *sb, int force)
11676+{
11677+ aufs_bindex_t bindex, bend;
11678+ struct au_branch *br;
11679+
11680+ /* exclude the bottom */
c1595e42 11681+ bend = au_fhsm_bottom(sb);
076b876e
AM
11682+ for (bindex = 0; bindex < bend; bindex++) {
11683+ br = au_sbr(sb, bindex);
11684+ if (au_br_fhsm(br->br_perm))
11685+ au_fhsm_wrote(sb, bindex, force);
11686+ }
11687+}
11688+
11689+/* ---------------------------------------------------------------------- */
11690+
11691+static unsigned int au_fhsm_poll(struct file *file,
11692+ struct poll_table_struct *wait)
11693+{
11694+ unsigned int mask;
11695+ struct au_sbinfo *sbinfo;
11696+ struct au_fhsm *fhsm;
11697+
11698+ mask = 0;
11699+ sbinfo = file->private_data;
11700+ fhsm = &sbinfo->si_fhsm;
11701+ poll_wait(file, &fhsm->fhsm_wqh, wait);
11702+ if (atomic_read(&fhsm->fhsm_readable))
11703+ mask = POLLIN /* | POLLRDNORM */;
11704+
11705+ AuTraceErr((int)mask);
11706+ return mask;
11707+}
11708+
11709+static int au_fhsm_do_read_one(struct aufs_stbr __user *stbr,
11710+ struct aufs_stfs *stfs, __s16 brid)
11711+{
11712+ int err;
11713+
11714+ err = copy_to_user(&stbr->stfs, stfs, sizeof(*stfs));
11715+ if (!err)
11716+ err = __put_user(brid, &stbr->brid);
11717+ if (unlikely(err))
11718+ err = -EFAULT;
11719+
11720+ return err;
11721+}
11722+
11723+static ssize_t au_fhsm_do_read(struct super_block *sb,
11724+ struct aufs_stbr __user *stbr, size_t count)
11725+{
11726+ ssize_t err;
11727+ int nstbr;
11728+ aufs_bindex_t bindex, bend;
11729+ struct au_branch *br;
11730+ struct au_br_fhsm *bf;
11731+
11732+ /* except the bottom branch */
11733+ err = 0;
11734+ nstbr = 0;
c1595e42 11735+ bend = au_fhsm_bottom(sb);
076b876e
AM
11736+ for (bindex = 0; !err && bindex < bend; bindex++) {
11737+ br = au_sbr(sb, bindex);
11738+ if (!au_br_fhsm(br->br_perm))
11739+ continue;
11740+
11741+ bf = br->br_fhsm;
11742+ mutex_lock(&bf->bf_lock);
11743+ if (bf->bf_readable) {
11744+ err = -EFAULT;
11745+ if (count >= sizeof(*stbr))
11746+ err = au_fhsm_do_read_one(stbr++, &bf->bf_stfs,
11747+ br->br_id);
11748+ if (!err) {
11749+ bf->bf_readable = 0;
11750+ count -= sizeof(*stbr);
11751+ nstbr++;
11752+ }
11753+ }
11754+ mutex_unlock(&bf->bf_lock);
11755+ }
11756+ if (!err)
11757+ err = sizeof(*stbr) * nstbr;
11758+
11759+ return err;
11760+}
11761+
11762+static ssize_t au_fhsm_read(struct file *file, char __user *buf, size_t count,
11763+ loff_t *pos)
11764+{
11765+ ssize_t err;
11766+ int readable;
11767+ aufs_bindex_t nfhsm, bindex, bend;
11768+ struct au_sbinfo *sbinfo;
11769+ struct au_fhsm *fhsm;
11770+ struct au_branch *br;
11771+ struct super_block *sb;
11772+
11773+ err = 0;
11774+ sbinfo = file->private_data;
11775+ fhsm = &sbinfo->si_fhsm;
11776+need_data:
11777+ spin_lock_irq(&fhsm->fhsm_wqh.lock);
11778+ if (!atomic_read(&fhsm->fhsm_readable)) {
11779+ if (vfsub_file_flags(file) & O_NONBLOCK)
11780+ err = -EAGAIN;
11781+ else
11782+ err = wait_event_interruptible_locked_irq
11783+ (fhsm->fhsm_wqh,
11784+ atomic_read(&fhsm->fhsm_readable));
11785+ }
11786+ spin_unlock_irq(&fhsm->fhsm_wqh.lock);
11787+ if (unlikely(err))
11788+ goto out;
11789+
11790+ /* sb may already be dead */
11791+ au_rw_read_lock(&sbinfo->si_rwsem);
11792+ readable = atomic_read(&fhsm->fhsm_readable);
11793+ if (readable > 0) {
11794+ sb = sbinfo->si_sb;
11795+ AuDebugOn(!sb);
11796+ /* exclude the bottom branch */
11797+ nfhsm = 0;
c1595e42 11798+ bend = au_fhsm_bottom(sb);
076b876e
AM
11799+ for (bindex = 0; bindex < bend; bindex++) {
11800+ br = au_sbr(sb, bindex);
11801+ if (au_br_fhsm(br->br_perm))
11802+ nfhsm++;
11803+ }
11804+ err = -EMSGSIZE;
11805+ if (nfhsm * sizeof(struct aufs_stbr) <= count) {
11806+ atomic_set(&fhsm->fhsm_readable, 0);
11807+ err = au_fhsm_do_read(sbinfo->si_sb, (void __user *)buf,
11808+ count);
11809+ }
11810+ }
11811+ au_rw_read_unlock(&sbinfo->si_rwsem);
11812+ if (!readable)
11813+ goto need_data;
11814+
11815+out:
11816+ return err;
11817+}
11818+
11819+static int au_fhsm_release(struct inode *inode, struct file *file)
11820+{
11821+ struct au_sbinfo *sbinfo;
11822+ struct au_fhsm *fhsm;
11823+
11824+ /* sb may already be dead */
11825+ sbinfo = file->private_data;
11826+ fhsm = &sbinfo->si_fhsm;
11827+ spin_lock(&fhsm->fhsm_spin);
11828+ fhsm->fhsm_pid = 0;
11829+ spin_unlock(&fhsm->fhsm_spin);
11830+ kobject_put(&sbinfo->si_kobj);
11831+
11832+ return 0;
11833+}
11834+
11835+static const struct file_operations au_fhsm_fops = {
11836+ .owner = THIS_MODULE,
11837+ .llseek = noop_llseek,
11838+ .read = au_fhsm_read,
11839+ .poll = au_fhsm_poll,
11840+ .release = au_fhsm_release
11841+};
11842+
11843+int au_fhsm_fd(struct super_block *sb, int oflags)
11844+{
11845+ int err, fd;
11846+ struct au_sbinfo *sbinfo;
11847+ struct au_fhsm *fhsm;
11848+
11849+ err = -EPERM;
11850+ if (unlikely(!capable(CAP_SYS_ADMIN)))
11851+ goto out;
11852+
11853+ err = -EINVAL;
11854+ if (unlikely(oflags & ~(O_CLOEXEC | O_NONBLOCK)))
11855+ goto out;
11856+
11857+ err = 0;
11858+ sbinfo = au_sbi(sb);
11859+ fhsm = &sbinfo->si_fhsm;
11860+ spin_lock(&fhsm->fhsm_spin);
11861+ if (!fhsm->fhsm_pid)
11862+ fhsm->fhsm_pid = current->pid;
11863+ else
11864+ err = -EBUSY;
11865+ spin_unlock(&fhsm->fhsm_spin);
11866+ if (unlikely(err))
11867+ goto out;
11868+
11869+ oflags |= O_RDONLY;
11870+ /* oflags |= FMODE_NONOTIFY; */
11871+ fd = anon_inode_getfd("[aufs_fhsm]", &au_fhsm_fops, sbinfo, oflags);
11872+ err = fd;
11873+ if (unlikely(fd < 0))
11874+ goto out_pid;
11875+
11876+ /* succeed reglardless 'fhsm' status */
11877+ kobject_get(&sbinfo->si_kobj);
11878+ si_noflush_read_lock(sb);
11879+ if (au_ftest_si(sbinfo, FHSM))
11880+ au_fhsm_wrote_all(sb, /*force*/0);
11881+ si_read_unlock(sb);
11882+ goto out; /* success */
11883+
11884+out_pid:
11885+ spin_lock(&fhsm->fhsm_spin);
11886+ fhsm->fhsm_pid = 0;
11887+ spin_unlock(&fhsm->fhsm_spin);
11888+out:
11889+ AuTraceErr(err);
11890+ return err;
11891+}
11892+
11893+/* ---------------------------------------------------------------------- */
11894+
11895+int au_fhsm_br_alloc(struct au_branch *br)
11896+{
11897+ int err;
11898+
11899+ err = 0;
11900+ br->br_fhsm = kmalloc(sizeof(*br->br_fhsm), GFP_NOFS);
11901+ if (br->br_fhsm)
11902+ au_br_fhsm_init(br->br_fhsm);
11903+ else
11904+ err = -ENOMEM;
11905+
11906+ return err;
11907+}
11908+
11909+/* ---------------------------------------------------------------------- */
11910+
11911+void au_fhsm_fin(struct super_block *sb)
11912+{
11913+ au_fhsm_notify(sb, /*val*/-1);
11914+}
11915+
11916+void au_fhsm_init(struct au_sbinfo *sbinfo)
11917+{
11918+ struct au_fhsm *fhsm;
11919+
11920+ fhsm = &sbinfo->si_fhsm;
11921+ spin_lock_init(&fhsm->fhsm_spin);
11922+ init_waitqueue_head(&fhsm->fhsm_wqh);
11923+ atomic_set(&fhsm->fhsm_readable, 0);
11924+ fhsm->fhsm_expire
11925+ = msecs_to_jiffies(AUFS_FHSM_CACHE_DEF_SEC * MSEC_PER_SEC);
c1595e42 11926+ fhsm->fhsm_bottom = -1;
076b876e
AM
11927+}
11928+
11929+void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec)
11930+{
11931+ sbinfo->si_fhsm.fhsm_expire
11932+ = msecs_to_jiffies(sec * MSEC_PER_SEC);
11933+}
11934+
11935+void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo)
11936+{
11937+ unsigned int u;
11938+
11939+ if (!au_ftest_si(sbinfo, FHSM))
11940+ return;
11941+
11942+ u = jiffies_to_msecs(sbinfo->si_fhsm.fhsm_expire) / MSEC_PER_SEC;
11943+ if (u != AUFS_FHSM_CACHE_DEF_SEC)
11944+ seq_printf(seq, ",fhsm_sec=%u", u);
11945+}
7f207e10
AM
11946diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
11947--- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 11948+++ linux/fs/aufs/file.c 2015-09-24 10:47:58.251386326 +0200
5527c038 11949@@ -0,0 +1,841 @@
1facf9fc 11950+/*
2000de60 11951+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 11952+ *
11953+ * This program, aufs is free software; you can redistribute it and/or modify
11954+ * it under the terms of the GNU General Public License as published by
11955+ * the Free Software Foundation; either version 2 of the License, or
11956+ * (at your option) any later version.
dece6358
AM
11957+ *
11958+ * This program is distributed in the hope that it will be useful,
11959+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
11960+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11961+ * GNU General Public License for more details.
11962+ *
11963+ * You should have received a copy of the GNU General Public License
523b37e3 11964+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 11965+ */
11966+
11967+/*
4a4d8108 11968+ * handling file/dir, and address_space operation
1facf9fc 11969+ */
11970+
7eafdf33
AM
11971+#ifdef CONFIG_AUFS_DEBUG
11972+#include <linux/migrate.h>
11973+#endif
4a4d8108 11974+#include <linux/pagemap.h>
1facf9fc 11975+#include "aufs.h"
11976+
4a4d8108
AM
11977+/* drop flags for writing */
11978+unsigned int au_file_roflags(unsigned int flags)
11979+{
11980+ flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
11981+ flags |= O_RDONLY | O_NOATIME;
11982+ return flags;
11983+}
11984+
11985+/* common functions to regular file and dir */
11986+struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
392086de 11987+ struct file *file, int force_wr)
1facf9fc 11988+{
1308ab2a 11989+ struct file *h_file;
4a4d8108
AM
11990+ struct dentry *h_dentry;
11991+ struct inode *h_inode;
11992+ struct super_block *sb;
11993+ struct au_branch *br;
11994+ struct path h_path;
b912730e 11995+ int err;
1facf9fc 11996+
4a4d8108
AM
11997+ /* a race condition can happen between open and unlink/rmdir */
11998+ h_file = ERR_PTR(-ENOENT);
11999+ h_dentry = au_h_dptr(dentry, bindex);
5527c038 12000+ if (au_test_nfsd() && (!h_dentry || d_is_negative(h_dentry)))
4a4d8108 12001+ goto out;
5527c038 12002+ h_inode = d_inode(h_dentry);
027c5e7a
AM
12003+ spin_lock(&h_dentry->d_lock);
12004+ err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
5527c038 12005+ /* || !d_inode(dentry)->i_nlink */
027c5e7a
AM
12006+ ;
12007+ spin_unlock(&h_dentry->d_lock);
12008+ if (unlikely(err))
4a4d8108 12009+ goto out;
1facf9fc 12010+
4a4d8108
AM
12011+ sb = dentry->d_sb;
12012+ br = au_sbr(sb, bindex);
b912730e
AM
12013+ err = au_br_test_oflag(flags, br);
12014+ h_file = ERR_PTR(err);
12015+ if (unlikely(err))
027c5e7a 12016+ goto out;
1facf9fc 12017+
4a4d8108 12018+ /* drop flags for writing */
5527c038 12019+ if (au_test_ro(sb, bindex, d_inode(dentry))) {
392086de
AM
12020+ if (force_wr && !(flags & O_WRONLY))
12021+ force_wr = 0;
4a4d8108 12022+ flags = au_file_roflags(flags);
392086de
AM
12023+ if (force_wr) {
12024+ h_file = ERR_PTR(-EROFS);
12025+ flags = au_file_roflags(flags);
12026+ if (unlikely(vfsub_native_ro(h_inode)
12027+ || IS_APPEND(h_inode)))
12028+ goto out;
12029+ flags &= ~O_ACCMODE;
12030+ flags |= O_WRONLY;
12031+ }
12032+ }
4a4d8108
AM
12033+ flags &= ~O_CREAT;
12034+ atomic_inc(&br->br_count);
12035+ h_path.dentry = h_dentry;
86dc4139 12036+ h_path.mnt = au_br_mnt(br);
38d290e6 12037+ h_file = vfsub_dentry_open(&h_path, flags);
4a4d8108
AM
12038+ if (IS_ERR(h_file))
12039+ goto out_br;
dece6358 12040+
b912730e 12041+ if (flags & __FMODE_EXEC) {
4a4d8108
AM
12042+ err = deny_write_access(h_file);
12043+ if (unlikely(err)) {
12044+ fput(h_file);
12045+ h_file = ERR_PTR(err);
12046+ goto out_br;
12047+ }
12048+ }
953406b4 12049+ fsnotify_open(h_file);
4a4d8108 12050+ goto out; /* success */
1facf9fc 12051+
4f0767ce 12052+out_br:
4a4d8108 12053+ atomic_dec(&br->br_count);
4f0767ce 12054+out:
4a4d8108
AM
12055+ return h_file;
12056+}
1308ab2a 12057+
076b876e
AM
12058+static int au_cmoo(struct dentry *dentry)
12059+{
12060+ int err, cmoo;
12061+ unsigned int udba;
12062+ struct path h_path;
12063+ struct au_pin pin;
12064+ struct au_cp_generic cpg = {
12065+ .dentry = dentry,
12066+ .bdst = -1,
12067+ .bsrc = -1,
12068+ .len = -1,
12069+ .pin = &pin,
12070+ .flags = AuCpup_DTIME | AuCpup_HOPEN
12071+ };
7e9cd9fe 12072+ struct inode *delegated;
076b876e
AM
12073+ struct super_block *sb;
12074+ struct au_sbinfo *sbinfo;
12075+ struct au_fhsm *fhsm;
12076+ pid_t pid;
12077+ struct au_branch *br;
12078+ struct dentry *parent;
12079+ struct au_hinode *hdir;
12080+
12081+ DiMustWriteLock(dentry);
5527c038 12082+ IiMustWriteLock(d_inode(dentry));
076b876e
AM
12083+
12084+ err = 0;
12085+ if (IS_ROOT(dentry))
12086+ goto out;
12087+ cpg.bsrc = au_dbstart(dentry);
12088+ if (!cpg.bsrc)
12089+ goto out;
12090+
12091+ sb = dentry->d_sb;
12092+ sbinfo = au_sbi(sb);
12093+ fhsm = &sbinfo->si_fhsm;
12094+ pid = au_fhsm_pid(fhsm);
12095+ if (pid
12096+ && (current->pid == pid
12097+ || current->real_parent->pid == pid))
12098+ goto out;
12099+
12100+ br = au_sbr(sb, cpg.bsrc);
12101+ cmoo = au_br_cmoo(br->br_perm);
12102+ if (!cmoo)
12103+ goto out;
7e9cd9fe 12104+ if (!d_is_reg(dentry))
076b876e
AM
12105+ cmoo &= AuBrAttr_COO_ALL;
12106+ if (!cmoo)
12107+ goto out;
12108+
12109+ parent = dget_parent(dentry);
12110+ di_write_lock_parent(parent);
12111+ err = au_wbr_do_copyup_bu(dentry, cpg.bsrc - 1);
12112+ cpg.bdst = err;
12113+ if (unlikely(err < 0)) {
12114+ err = 0; /* there is no upper writable branch */
12115+ goto out_dgrade;
12116+ }
12117+ AuDbg("bsrc %d, bdst %d\n", cpg.bsrc, cpg.bdst);
12118+
12119+ /* do not respect the coo attrib for the target branch */
12120+ err = au_cpup_dirs(dentry, cpg.bdst);
12121+ if (unlikely(err))
12122+ goto out_dgrade;
12123+
12124+ di_downgrade_lock(parent, AuLock_IR);
12125+ udba = au_opt_udba(sb);
12126+ err = au_pin(&pin, dentry, cpg.bdst, udba,
12127+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12128+ if (unlikely(err))
12129+ goto out_parent;
12130+
12131+ err = au_sio_cpup_simple(&cpg);
12132+ au_unpin(&pin);
12133+ if (unlikely(err))
12134+ goto out_parent;
12135+ if (!(cmoo & AuBrWAttr_MOO))
12136+ goto out_parent; /* success */
12137+
12138+ err = au_pin(&pin, dentry, cpg.bsrc, udba,
12139+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12140+ if (unlikely(err))
12141+ goto out_parent;
12142+
12143+ h_path.mnt = au_br_mnt(br);
12144+ h_path.dentry = au_h_dptr(dentry, cpg.bsrc);
5527c038 12145+ hdir = au_hi(d_inode(parent), cpg.bsrc);
076b876e
AM
12146+ delegated = NULL;
12147+ err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated, /*force*/1);
12148+ au_unpin(&pin);
12149+ /* todo: keep h_dentry or not? */
12150+ if (unlikely(err == -EWOULDBLOCK)) {
12151+ pr_warn("cannot retry for NFSv4 delegation"
12152+ " for an internal unlink\n");
12153+ iput(delegated);
12154+ }
12155+ if (unlikely(err)) {
12156+ pr_err("unlink %pd after coo failed (%d), ignored\n",
12157+ dentry, err);
12158+ err = 0;
12159+ }
12160+ goto out_parent; /* success */
12161+
12162+out_dgrade:
12163+ di_downgrade_lock(parent, AuLock_IR);
12164+out_parent:
12165+ di_read_unlock(parent, AuLock_IR);
12166+ dput(parent);
12167+out:
12168+ AuTraceErr(err);
12169+ return err;
12170+}
12171+
b912730e 12172+int au_do_open(struct file *file, struct au_do_open_args *args)
1facf9fc 12173+{
b912730e 12174+ int err, no_lock = args->no_lock;
1facf9fc 12175+ struct dentry *dentry;
076b876e 12176+ struct au_finfo *finfo;
1308ab2a 12177+
b912730e
AM
12178+ if (!no_lock)
12179+ err = au_finfo_init(file, args->fidir);
12180+ else {
12181+ lockdep_off();
12182+ err = au_finfo_init(file, args->fidir);
12183+ lockdep_on();
12184+ }
4a4d8108
AM
12185+ if (unlikely(err))
12186+ goto out;
1facf9fc 12187+
2000de60 12188+ dentry = file->f_path.dentry;
b912730e
AM
12189+ AuDebugOn(IS_ERR_OR_NULL(dentry));
12190+ if (!no_lock) {
12191+ di_write_lock_child(dentry);
12192+ err = au_cmoo(dentry);
12193+ di_downgrade_lock(dentry, AuLock_IR);
12194+ if (!err)
12195+ err = args->open(file, vfsub_file_flags(file), NULL);
12196+ di_read_unlock(dentry, AuLock_IR);
12197+ } else {
12198+ err = au_cmoo(dentry);
12199+ if (!err)
12200+ err = args->open(file, vfsub_file_flags(file),
12201+ args->h_file);
12202+ if (!err && au_fbstart(file) != au_dbstart(dentry))
12203+ /*
12204+ * cmoo happens after h_file was opened.
12205+ * need to refresh file later.
12206+ */
12207+ atomic_dec(&au_fi(file)->fi_generation);
12208+ }
1facf9fc 12209+
076b876e
AM
12210+ finfo = au_fi(file);
12211+ if (!err) {
12212+ finfo->fi_file = file;
12213+ au_sphl_add(&finfo->fi_hlist,
2000de60 12214+ &au_sbi(file->f_path.dentry->d_sb)->si_files);
076b876e 12215+ }
b912730e
AM
12216+ if (!no_lock)
12217+ fi_write_unlock(file);
12218+ else {
12219+ lockdep_off();
12220+ fi_write_unlock(file);
12221+ lockdep_on();
12222+ }
4a4d8108 12223+ if (unlikely(err)) {
076b876e 12224+ finfo->fi_hdir = NULL;
4a4d8108 12225+ au_finfo_fin(file);
1308ab2a 12226+ }
4a4d8108 12227+
4f0767ce 12228+out:
1308ab2a 12229+ return err;
12230+}
dece6358 12231+
4a4d8108 12232+int au_reopen_nondir(struct file *file)
1308ab2a 12233+{
4a4d8108
AM
12234+ int err;
12235+ aufs_bindex_t bstart;
12236+ struct dentry *dentry;
12237+ struct file *h_file, *h_file_tmp;
1308ab2a 12238+
2000de60 12239+ dentry = file->f_path.dentry;
4a4d8108
AM
12240+ bstart = au_dbstart(dentry);
12241+ h_file_tmp = NULL;
12242+ if (au_fbstart(file) == bstart) {
12243+ h_file = au_hf_top(file);
12244+ if (file->f_mode == h_file->f_mode)
12245+ return 0; /* success */
12246+ h_file_tmp = h_file;
12247+ get_file(h_file_tmp);
12248+ au_set_h_fptr(file, bstart, NULL);
12249+ }
12250+ AuDebugOn(au_fi(file)->fi_hdir);
86dc4139
AM
12251+ /*
12252+ * it can happen
12253+ * file exists on both of rw and ro
12254+ * open --> dbstart and fbstart are both 0
12255+ * prepend a branch as rw, "rw" become ro
12256+ * remove rw/file
12257+ * delete the top branch, "rw" becomes rw again
12258+ * --> dbstart is 1, fbstart is still 0
12259+ * write --> fbstart is 0 but dbstart is 1
12260+ */
12261+ /* AuDebugOn(au_fbstart(file) < bstart); */
1308ab2a 12262+
4a4d8108 12263+ h_file = au_h_open(dentry, bstart, vfsub_file_flags(file) & ~O_TRUNC,
392086de 12264+ file, /*force_wr*/0);
4a4d8108 12265+ err = PTR_ERR(h_file);
86dc4139
AM
12266+ if (IS_ERR(h_file)) {
12267+ if (h_file_tmp) {
12268+ atomic_inc(&au_sbr(dentry->d_sb, bstart)->br_count);
12269+ au_set_h_fptr(file, bstart, h_file_tmp);
12270+ h_file_tmp = NULL;
12271+ }
4a4d8108 12272+ goto out; /* todo: close all? */
86dc4139 12273+ }
4a4d8108
AM
12274+
12275+ err = 0;
12276+ au_set_fbstart(file, bstart);
12277+ au_set_h_fptr(file, bstart, h_file);
12278+ au_update_figen(file);
12279+ /* todo: necessary? */
12280+ /* file->f_ra = h_file->f_ra; */
12281+
4f0767ce 12282+out:
4a4d8108
AM
12283+ if (h_file_tmp)
12284+ fput(h_file_tmp);
12285+ return err;
1facf9fc 12286+}
12287+
1308ab2a 12288+/* ---------------------------------------------------------------------- */
12289+
4a4d8108
AM
12290+static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
12291+ struct dentry *hi_wh)
1facf9fc 12292+{
4a4d8108
AM
12293+ int err;
12294+ aufs_bindex_t bstart;
12295+ struct au_dinfo *dinfo;
12296+ struct dentry *h_dentry;
12297+ struct au_hdentry *hdp;
1facf9fc 12298+
2000de60 12299+ dinfo = au_di(file->f_path.dentry);
4a4d8108 12300+ AuRwMustWriteLock(&dinfo->di_rwsem);
dece6358 12301+
4a4d8108
AM
12302+ bstart = dinfo->di_bstart;
12303+ dinfo->di_bstart = btgt;
12304+ hdp = dinfo->di_hdentry;
12305+ h_dentry = hdp[0 + btgt].hd_dentry;
12306+ hdp[0 + btgt].hd_dentry = hi_wh;
12307+ err = au_reopen_nondir(file);
12308+ hdp[0 + btgt].hd_dentry = h_dentry;
12309+ dinfo->di_bstart = bstart;
1facf9fc 12310+
1facf9fc 12311+ return err;
12312+}
12313+
4a4d8108 12314+static int au_ready_to_write_wh(struct file *file, loff_t len,
86dc4139 12315+ aufs_bindex_t bcpup, struct au_pin *pin)
1facf9fc 12316+{
4a4d8108 12317+ int err;
027c5e7a 12318+ struct inode *inode, *h_inode;
c2b27bf2
AM
12319+ struct dentry *h_dentry, *hi_wh;
12320+ struct au_cp_generic cpg = {
2000de60 12321+ .dentry = file->f_path.dentry,
c2b27bf2
AM
12322+ .bdst = bcpup,
12323+ .bsrc = -1,
12324+ .len = len,
12325+ .pin = pin
12326+ };
1facf9fc 12327+
c2b27bf2 12328+ au_update_dbstart(cpg.dentry);
5527c038 12329+ inode = d_inode(cpg.dentry);
027c5e7a 12330+ h_inode = NULL;
c2b27bf2
AM
12331+ if (au_dbstart(cpg.dentry) <= bcpup
12332+ && au_dbend(cpg.dentry) >= bcpup) {
12333+ h_dentry = au_h_dptr(cpg.dentry, bcpup);
5527c038
JR
12334+ if (h_dentry && d_is_positive(h_dentry))
12335+ h_inode = d_inode(h_dentry);
027c5e7a 12336+ }
4a4d8108 12337+ hi_wh = au_hi_wh(inode, bcpup);
027c5e7a 12338+ if (!hi_wh && !h_inode)
c2b27bf2 12339+ err = au_sio_cpup_wh(&cpg, file);
4a4d8108
AM
12340+ else
12341+ /* already copied-up after unlink */
12342+ err = au_reopen_wh(file, bcpup, hi_wh);
1facf9fc 12343+
4a4d8108 12344+ if (!err
38d290e6
JR
12345+ && (inode->i_nlink > 1
12346+ || (inode->i_state & I_LINKABLE))
c2b27bf2
AM
12347+ && au_opt_test(au_mntflags(cpg.dentry->d_sb), PLINK))
12348+ au_plink_append(inode, bcpup, au_h_dptr(cpg.dentry, bcpup));
1308ab2a 12349+
dece6358 12350+ return err;
1facf9fc 12351+}
12352+
4a4d8108
AM
12353+/*
12354+ * prepare the @file for writing.
12355+ */
12356+int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
1facf9fc 12357+{
4a4d8108 12358+ int err;
c2b27bf2 12359+ aufs_bindex_t dbstart;
c1595e42 12360+ struct dentry *parent;
86dc4139 12361+ struct inode *inode;
1facf9fc 12362+ struct super_block *sb;
4a4d8108 12363+ struct file *h_file;
c2b27bf2 12364+ struct au_cp_generic cpg = {
2000de60 12365+ .dentry = file->f_path.dentry,
c2b27bf2
AM
12366+ .bdst = -1,
12367+ .bsrc = -1,
12368+ .len = len,
12369+ .pin = pin,
12370+ .flags = AuCpup_DTIME
12371+ };
1facf9fc 12372+
c2b27bf2 12373+ sb = cpg.dentry->d_sb;
5527c038 12374+ inode = d_inode(cpg.dentry);
c2b27bf2
AM
12375+ cpg.bsrc = au_fbstart(file);
12376+ err = au_test_ro(sb, cpg.bsrc, inode);
4a4d8108 12377+ if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
c2b27bf2
AM
12378+ err = au_pin(pin, cpg.dentry, cpg.bsrc, AuOpt_UDBA_NONE,
12379+ /*flags*/0);
1facf9fc 12380+ goto out;
4a4d8108 12381+ }
1facf9fc 12382+
027c5e7a 12383+ /* need to cpup or reopen */
c2b27bf2 12384+ parent = dget_parent(cpg.dentry);
4a4d8108 12385+ di_write_lock_parent(parent);
c2b27bf2
AM
12386+ err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
12387+ cpg.bdst = err;
4a4d8108
AM
12388+ if (unlikely(err < 0))
12389+ goto out_dgrade;
12390+ err = 0;
12391+
c2b27bf2
AM
12392+ if (!d_unhashed(cpg.dentry) && !au_h_dptr(parent, cpg.bdst)) {
12393+ err = au_cpup_dirs(cpg.dentry, cpg.bdst);
1facf9fc 12394+ if (unlikely(err))
4a4d8108
AM
12395+ goto out_dgrade;
12396+ }
12397+
c2b27bf2 12398+ err = au_pin(pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
4a4d8108
AM
12399+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
12400+ if (unlikely(err))
12401+ goto out_dgrade;
12402+
c2b27bf2 12403+ dbstart = au_dbstart(cpg.dentry);
c1595e42 12404+ if (dbstart <= cpg.bdst)
c2b27bf2 12405+ cpg.bsrc = cpg.bdst;
027c5e7a 12406+
c2b27bf2
AM
12407+ if (dbstart <= cpg.bdst /* just reopen */
12408+ || !d_unhashed(cpg.dentry) /* copyup and reopen */
027c5e7a 12409+ ) {
392086de 12410+ h_file = au_h_open_pre(cpg.dentry, cpg.bsrc, /*force_wr*/0);
86dc4139 12411+ if (IS_ERR(h_file))
027c5e7a 12412+ err = PTR_ERR(h_file);
86dc4139 12413+ else {
027c5e7a 12414+ di_downgrade_lock(parent, AuLock_IR);
c2b27bf2
AM
12415+ if (dbstart > cpg.bdst)
12416+ err = au_sio_cpup_simple(&cpg);
027c5e7a
AM
12417+ if (!err)
12418+ err = au_reopen_nondir(file);
c2b27bf2 12419+ au_h_open_post(cpg.dentry, cpg.bsrc, h_file);
027c5e7a 12420+ }
027c5e7a
AM
12421+ } else { /* copyup as wh and reopen */
12422+ /*
12423+ * since writable hfsplus branch is not supported,
12424+ * h_open_pre/post() are unnecessary.
12425+ */
c2b27bf2 12426+ err = au_ready_to_write_wh(file, len, cpg.bdst, pin);
4a4d8108 12427+ di_downgrade_lock(parent, AuLock_IR);
4a4d8108 12428+ }
4a4d8108
AM
12429+
12430+ if (!err) {
12431+ au_pin_set_parent_lflag(pin, /*lflag*/0);
12432+ goto out_dput; /* success */
12433+ }
12434+ au_unpin(pin);
12435+ goto out_unlock;
1facf9fc 12436+
4f0767ce 12437+out_dgrade:
4a4d8108 12438+ di_downgrade_lock(parent, AuLock_IR);
4f0767ce 12439+out_unlock:
4a4d8108 12440+ di_read_unlock(parent, AuLock_IR);
4f0767ce 12441+out_dput:
4a4d8108 12442+ dput(parent);
4f0767ce 12443+out:
1facf9fc 12444+ return err;
12445+}
12446+
4a4d8108
AM
12447+/* ---------------------------------------------------------------------- */
12448+
12449+int au_do_flush(struct file *file, fl_owner_t id,
12450+ int (*flush)(struct file *file, fl_owner_t id))
1facf9fc 12451+{
4a4d8108 12452+ int err;
1facf9fc 12453+ struct super_block *sb;
4a4d8108 12454+ struct inode *inode;
1facf9fc 12455+
c06a8ce3
AM
12456+ inode = file_inode(file);
12457+ sb = inode->i_sb;
4a4d8108
AM
12458+ si_noflush_read_lock(sb);
12459+ fi_read_lock(file);
b752ccd1 12460+ ii_read_lock_child(inode);
1facf9fc 12461+
4a4d8108
AM
12462+ err = flush(file, id);
12463+ au_cpup_attr_timesizes(inode);
1facf9fc 12464+
b752ccd1 12465+ ii_read_unlock(inode);
4a4d8108 12466+ fi_read_unlock(file);
1308ab2a 12467+ si_read_unlock(sb);
dece6358 12468+ return err;
1facf9fc 12469+}
12470+
4a4d8108
AM
12471+/* ---------------------------------------------------------------------- */
12472+
12473+static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
1facf9fc 12474+{
4a4d8108 12475+ int err;
4a4d8108
AM
12476+ struct au_pin pin;
12477+ struct au_finfo *finfo;
c2b27bf2 12478+ struct dentry *parent, *hi_wh;
4a4d8108 12479+ struct inode *inode;
1facf9fc 12480+ struct super_block *sb;
c2b27bf2 12481+ struct au_cp_generic cpg = {
2000de60 12482+ .dentry = file->f_path.dentry,
c2b27bf2
AM
12483+ .bdst = -1,
12484+ .bsrc = -1,
12485+ .len = -1,
12486+ .pin = &pin,
12487+ .flags = AuCpup_DTIME
12488+ };
1facf9fc 12489+
4a4d8108
AM
12490+ FiMustWriteLock(file);
12491+
12492+ err = 0;
12493+ finfo = au_fi(file);
c2b27bf2 12494+ sb = cpg.dentry->d_sb;
5527c038 12495+ inode = d_inode(cpg.dentry);
c2b27bf2
AM
12496+ cpg.bdst = au_ibstart(inode);
12497+ if (cpg.bdst == finfo->fi_btop || IS_ROOT(cpg.dentry))
1308ab2a 12498+ goto out;
dece6358 12499+
c2b27bf2
AM
12500+ parent = dget_parent(cpg.dentry);
12501+ if (au_test_ro(sb, cpg.bdst, inode)) {
4a4d8108 12502+ di_read_lock_parent(parent, !AuLock_IR);
c2b27bf2
AM
12503+ err = AuWbrCopyup(au_sbi(sb), cpg.dentry);
12504+ cpg.bdst = err;
4a4d8108
AM
12505+ di_read_unlock(parent, !AuLock_IR);
12506+ if (unlikely(err < 0))
12507+ goto out_parent;
12508+ err = 0;
1facf9fc 12509+ }
1facf9fc 12510+
4a4d8108 12511+ di_read_lock_parent(parent, AuLock_IR);
c2b27bf2 12512+ hi_wh = au_hi_wh(inode, cpg.bdst);
7f207e10
AM
12513+ if (!S_ISDIR(inode->i_mode)
12514+ && au_opt_test(au_mntflags(sb), PLINK)
4a4d8108 12515+ && au_plink_test(inode)
c2b27bf2
AM
12516+ && !d_unhashed(cpg.dentry)
12517+ && cpg.bdst < au_dbstart(cpg.dentry)) {
12518+ err = au_test_and_cpup_dirs(cpg.dentry, cpg.bdst);
4a4d8108
AM
12519+ if (unlikely(err))
12520+ goto out_unlock;
12521+
12522+ /* always superio. */
c2b27bf2 12523+ err = au_pin(&pin, cpg.dentry, cpg.bdst, AuOpt_UDBA_NONE,
4a4d8108 12524+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
367653fa 12525+ if (!err) {
c2b27bf2 12526+ err = au_sio_cpup_simple(&cpg);
367653fa
AM
12527+ au_unpin(&pin);
12528+ }
4a4d8108
AM
12529+ } else if (hi_wh) {
12530+ /* already copied-up after unlink */
c2b27bf2 12531+ err = au_reopen_wh(file, cpg.bdst, hi_wh);
4a4d8108
AM
12532+ *need_reopen = 0;
12533+ }
1facf9fc 12534+
4f0767ce 12535+out_unlock:
4a4d8108 12536+ di_read_unlock(parent, AuLock_IR);
4f0767ce 12537+out_parent:
4a4d8108 12538+ dput(parent);
4f0767ce 12539+out:
1308ab2a 12540+ return err;
dece6358 12541+}
1facf9fc 12542+
4a4d8108 12543+static void au_do_refresh_dir(struct file *file)
dece6358 12544+{
4a4d8108
AM
12545+ aufs_bindex_t bindex, bend, new_bindex, brid;
12546+ struct au_hfile *p, tmp, *q;
12547+ struct au_finfo *finfo;
1308ab2a 12548+ struct super_block *sb;
4a4d8108 12549+ struct au_fidir *fidir;
1facf9fc 12550+
4a4d8108 12551+ FiMustWriteLock(file);
1facf9fc 12552+
2000de60 12553+ sb = file->f_path.dentry->d_sb;
4a4d8108
AM
12554+ finfo = au_fi(file);
12555+ fidir = finfo->fi_hdir;
12556+ AuDebugOn(!fidir);
12557+ p = fidir->fd_hfile + finfo->fi_btop;
12558+ brid = p->hf_br->br_id;
12559+ bend = fidir->fd_bbot;
12560+ for (bindex = finfo->fi_btop; bindex <= bend; bindex++, p++) {
12561+ if (!p->hf_file)
12562+ continue;
1308ab2a 12563+
4a4d8108
AM
12564+ new_bindex = au_br_index(sb, p->hf_br->br_id);
12565+ if (new_bindex == bindex)
12566+ continue;
12567+ if (new_bindex < 0) {
12568+ au_set_h_fptr(file, bindex, NULL);
12569+ continue;
12570+ }
1308ab2a 12571+
4a4d8108
AM
12572+ /* swap two lower inode, and loop again */
12573+ q = fidir->fd_hfile + new_bindex;
12574+ tmp = *q;
12575+ *q = *p;
12576+ *p = tmp;
12577+ if (tmp.hf_file) {
12578+ bindex--;
12579+ p--;
12580+ }
12581+ }
1308ab2a 12582+
4a4d8108 12583+ p = fidir->fd_hfile;
2000de60 12584+ if (!au_test_mmapped(file) && !d_unlinked(file->f_path.dentry)) {
4a4d8108
AM
12585+ bend = au_sbend(sb);
12586+ for (finfo->fi_btop = 0; finfo->fi_btop <= bend;
12587+ finfo->fi_btop++, p++)
12588+ if (p->hf_file) {
c06a8ce3 12589+ if (file_inode(p->hf_file))
4a4d8108 12590+ break;
c1595e42 12591+ au_hfput(p, file);
4a4d8108
AM
12592+ }
12593+ } else {
12594+ bend = au_br_index(sb, brid);
12595+ for (finfo->fi_btop = 0; finfo->fi_btop < bend;
12596+ finfo->fi_btop++, p++)
12597+ if (p->hf_file)
12598+ au_hfput(p, file);
12599+ bend = au_sbend(sb);
12600+ }
1308ab2a 12601+
4a4d8108
AM
12602+ p = fidir->fd_hfile + bend;
12603+ for (fidir->fd_bbot = bend; fidir->fd_bbot >= finfo->fi_btop;
12604+ fidir->fd_bbot--, p--)
12605+ if (p->hf_file) {
c06a8ce3 12606+ if (file_inode(p->hf_file))
4a4d8108 12607+ break;
c1595e42 12608+ au_hfput(p, file);
4a4d8108
AM
12609+ }
12610+ AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
1308ab2a 12611+}
12612+
4a4d8108
AM
12613+/*
12614+ * after branch manipulating, refresh the file.
12615+ */
12616+static int refresh_file(struct file *file, int (*reopen)(struct file *file))
1facf9fc 12617+{
4a4d8108
AM
12618+ int err, need_reopen;
12619+ aufs_bindex_t bend, bindex;
12620+ struct dentry *dentry;
1308ab2a 12621+ struct au_finfo *finfo;
4a4d8108 12622+ struct au_hfile *hfile;
1facf9fc 12623+
2000de60 12624+ dentry = file->f_path.dentry;
1308ab2a 12625+ finfo = au_fi(file);
4a4d8108
AM
12626+ if (!finfo->fi_hdir) {
12627+ hfile = &finfo->fi_htop;
12628+ AuDebugOn(!hfile->hf_file);
12629+ bindex = au_br_index(dentry->d_sb, hfile->hf_br->br_id);
12630+ AuDebugOn(bindex < 0);
12631+ if (bindex != finfo->fi_btop)
12632+ au_set_fbstart(file, bindex);
12633+ } else {
12634+ err = au_fidir_realloc(finfo, au_sbend(dentry->d_sb) + 1);
12635+ if (unlikely(err))
12636+ goto out;
12637+ au_do_refresh_dir(file);
12638+ }
1facf9fc 12639+
4a4d8108
AM
12640+ err = 0;
12641+ need_reopen = 1;
12642+ if (!au_test_mmapped(file))
12643+ err = au_file_refresh_by_inode(file, &need_reopen);
027c5e7a 12644+ if (!err && need_reopen && !d_unlinked(dentry))
4a4d8108
AM
12645+ err = reopen(file);
12646+ if (!err) {
12647+ au_update_figen(file);
12648+ goto out; /* success */
12649+ }
12650+
12651+ /* error, close all lower files */
12652+ if (finfo->fi_hdir) {
12653+ bend = au_fbend_dir(file);
12654+ for (bindex = au_fbstart(file); bindex <= bend; bindex++)
12655+ au_set_h_fptr(file, bindex, NULL);
12656+ }
1facf9fc 12657+
4f0767ce 12658+out:
1facf9fc 12659+ return err;
12660+}
12661+
4a4d8108
AM
12662+/* common function to regular file and dir */
12663+int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
12664+ int wlock)
dece6358 12665+{
1308ab2a 12666+ int err;
4a4d8108
AM
12667+ unsigned int sigen, figen;
12668+ aufs_bindex_t bstart;
12669+ unsigned char pseudo_link;
12670+ struct dentry *dentry;
12671+ struct inode *inode;
1facf9fc 12672+
4a4d8108 12673+ err = 0;
2000de60 12674+ dentry = file->f_path.dentry;
5527c038 12675+ inode = d_inode(dentry);
4a4d8108
AM
12676+ sigen = au_sigen(dentry->d_sb);
12677+ fi_write_lock(file);
12678+ figen = au_figen(file);
12679+ di_write_lock_child(dentry);
12680+ bstart = au_dbstart(dentry);
12681+ pseudo_link = (bstart != au_ibstart(inode));
12682+ if (sigen == figen && !pseudo_link && au_fbstart(file) == bstart) {
12683+ if (!wlock) {
12684+ di_downgrade_lock(dentry, AuLock_IR);
12685+ fi_downgrade_lock(file);
12686+ }
12687+ goto out; /* success */
12688+ }
dece6358 12689+
4a4d8108 12690+ AuDbg("sigen %d, figen %d\n", sigen, figen);
027c5e7a 12691+ if (au_digen_test(dentry, sigen)) {
4a4d8108 12692+ err = au_reval_dpath(dentry, sigen);
027c5e7a 12693+ AuDebugOn(!err && au_digen_test(dentry, sigen));
4a4d8108 12694+ }
dece6358 12695+
027c5e7a
AM
12696+ if (!err)
12697+ err = refresh_file(file, reopen);
4a4d8108
AM
12698+ if (!err) {
12699+ if (!wlock) {
12700+ di_downgrade_lock(dentry, AuLock_IR);
12701+ fi_downgrade_lock(file);
12702+ }
12703+ } else {
12704+ di_write_unlock(dentry);
12705+ fi_write_unlock(file);
12706+ }
1facf9fc 12707+
4f0767ce 12708+out:
1308ab2a 12709+ return err;
12710+}
1facf9fc 12711+
4a4d8108
AM
12712+/* ---------------------------------------------------------------------- */
12713+
12714+/* cf. aufs_nopage() */
12715+/* for madvise(2) */
12716+static int aufs_readpage(struct file *file __maybe_unused, struct page *page)
1308ab2a 12717+{
4a4d8108
AM
12718+ unlock_page(page);
12719+ return 0;
12720+}
1facf9fc 12721+
4a4d8108 12722+/* it will never be called, but necessary to support O_DIRECT */
5527c038
JR
12723+static ssize_t aufs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
12724+ loff_t offset)
4a4d8108 12725+{ BUG(); return 0; }
1facf9fc 12726+
4a4d8108
AM
12727+/* they will never be called. */
12728+#ifdef CONFIG_AUFS_DEBUG
12729+static int aufs_write_begin(struct file *file, struct address_space *mapping,
12730+ loff_t pos, unsigned len, unsigned flags,
12731+ struct page **pagep, void **fsdata)
12732+{ AuUnsupport(); return 0; }
12733+static int aufs_write_end(struct file *file, struct address_space *mapping,
12734+ loff_t pos, unsigned len, unsigned copied,
12735+ struct page *page, void *fsdata)
12736+{ AuUnsupport(); return 0; }
12737+static int aufs_writepage(struct page *page, struct writeback_control *wbc)
12738+{ AuUnsupport(); return 0; }
1308ab2a 12739+
4a4d8108
AM
12740+static int aufs_set_page_dirty(struct page *page)
12741+{ AuUnsupport(); return 0; }
392086de
AM
12742+static void aufs_invalidatepage(struct page *page, unsigned int offset,
12743+ unsigned int length)
4a4d8108
AM
12744+{ AuUnsupport(); }
12745+static int aufs_releasepage(struct page *page, gfp_t gfp)
12746+{ AuUnsupport(); return 0; }
12747+static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
7eafdf33 12748+ struct page *page, enum migrate_mode mode)
4a4d8108
AM
12749+{ AuUnsupport(); return 0; }
12750+static int aufs_launder_page(struct page *page)
12751+{ AuUnsupport(); return 0; }
12752+static int aufs_is_partially_uptodate(struct page *page,
38d290e6
JR
12753+ unsigned long from,
12754+ unsigned long count)
4a4d8108 12755+{ AuUnsupport(); return 0; }
392086de
AM
12756+static void aufs_is_dirty_writeback(struct page *page, bool *dirty,
12757+ bool *writeback)
12758+{ AuUnsupport(); }
4a4d8108
AM
12759+static int aufs_error_remove_page(struct address_space *mapping,
12760+ struct page *page)
12761+{ AuUnsupport(); return 0; }
b4510431
AM
12762+static int aufs_swap_activate(struct swap_info_struct *sis, struct file *file,
12763+ sector_t *span)
12764+{ AuUnsupport(); return 0; }
12765+static void aufs_swap_deactivate(struct file *file)
12766+{ AuUnsupport(); }
4a4d8108
AM
12767+#endif /* CONFIG_AUFS_DEBUG */
12768+
12769+const struct address_space_operations aufs_aop = {
12770+ .readpage = aufs_readpage,
12771+ .direct_IO = aufs_direct_IO,
4a4d8108
AM
12772+#ifdef CONFIG_AUFS_DEBUG
12773+ .writepage = aufs_writepage,
4a4d8108
AM
12774+ /* no writepages, because of writepage */
12775+ .set_page_dirty = aufs_set_page_dirty,
12776+ /* no readpages, because of readpage */
12777+ .write_begin = aufs_write_begin,
12778+ .write_end = aufs_write_end,
12779+ /* no bmap, no block device */
12780+ .invalidatepage = aufs_invalidatepage,
12781+ .releasepage = aufs_releasepage,
12782+ .migratepage = aufs_migratepage,
12783+ .launder_page = aufs_launder_page,
12784+ .is_partially_uptodate = aufs_is_partially_uptodate,
392086de 12785+ .is_dirty_writeback = aufs_is_dirty_writeback,
b4510431
AM
12786+ .error_remove_page = aufs_error_remove_page,
12787+ .swap_activate = aufs_swap_activate,
12788+ .swap_deactivate = aufs_swap_deactivate
4a4d8108 12789+#endif /* CONFIG_AUFS_DEBUG */
dece6358 12790+};
7f207e10
AM
12791diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
12792--- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 12793+++ linux/fs/aufs/file.h 2015-09-24 10:47:58.251386326 +0200
b912730e 12794@@ -0,0 +1,291 @@
4a4d8108 12795+/*
2000de60 12796+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
12797+ *
12798+ * This program, aufs is free software; you can redistribute it and/or modify
12799+ * it under the terms of the GNU General Public License as published by
12800+ * the Free Software Foundation; either version 2 of the License, or
12801+ * (at your option) any later version.
12802+ *
12803+ * This program is distributed in the hope that it will be useful,
12804+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12805+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12806+ * GNU General Public License for more details.
12807+ *
12808+ * You should have received a copy of the GNU General Public License
523b37e3 12809+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 12810+ */
1facf9fc 12811+
4a4d8108
AM
12812+/*
12813+ * file operations
12814+ */
1facf9fc 12815+
4a4d8108
AM
12816+#ifndef __AUFS_FILE_H__
12817+#define __AUFS_FILE_H__
1facf9fc 12818+
4a4d8108 12819+#ifdef __KERNEL__
1facf9fc 12820+
2cbb1c4b 12821+#include <linux/file.h>
4a4d8108
AM
12822+#include <linux/fs.h>
12823+#include <linux/poll.h>
4a4d8108 12824+#include "rwsem.h"
1facf9fc 12825+
4a4d8108
AM
12826+struct au_branch;
12827+struct au_hfile {
12828+ struct file *hf_file;
12829+ struct au_branch *hf_br;
12830+};
1facf9fc 12831+
4a4d8108
AM
12832+struct au_vdir;
12833+struct au_fidir {
12834+ aufs_bindex_t fd_bbot;
12835+ aufs_bindex_t fd_nent;
12836+ struct au_vdir *fd_vdir_cache;
12837+ struct au_hfile fd_hfile[];
12838+};
1facf9fc 12839+
4a4d8108 12840+static inline int au_fidir_sz(int nent)
dece6358 12841+{
4f0767ce
JR
12842+ AuDebugOn(nent < 0);
12843+ return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
4a4d8108 12844+}
1facf9fc 12845+
4a4d8108
AM
12846+struct au_finfo {
12847+ atomic_t fi_generation;
dece6358 12848+
4a4d8108
AM
12849+ struct au_rwsem fi_rwsem;
12850+ aufs_bindex_t fi_btop;
12851+
12852+ /* do not union them */
12853+ struct { /* for non-dir */
12854+ struct au_hfile fi_htop;
2cbb1c4b 12855+ atomic_t fi_mmapped;
4a4d8108
AM
12856+ };
12857+ struct au_fidir *fi_hdir; /* for dir only */
523b37e3
AM
12858+
12859+ struct hlist_node fi_hlist;
12860+ struct file *fi_file; /* very ugly */
4a4d8108 12861+} ____cacheline_aligned_in_smp;
1facf9fc 12862+
4a4d8108 12863+/* ---------------------------------------------------------------------- */
1facf9fc 12864+
4a4d8108
AM
12865+/* file.c */
12866+extern const struct address_space_operations aufs_aop;
12867+unsigned int au_file_roflags(unsigned int flags);
12868+struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
392086de 12869+ struct file *file, int force_wr);
b912730e
AM
12870+struct au_do_open_args {
12871+ int no_lock;
12872+ int (*open)(struct file *file, int flags,
12873+ struct file *h_file);
12874+ struct au_fidir *fidir;
12875+ struct file *h_file;
12876+};
12877+int au_do_open(struct file *file, struct au_do_open_args *args);
4a4d8108
AM
12878+int au_reopen_nondir(struct file *file);
12879+struct au_pin;
12880+int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
12881+int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
12882+ int wlock);
12883+int au_do_flush(struct file *file, fl_owner_t id,
12884+ int (*flush)(struct file *file, fl_owner_t id));
1facf9fc 12885+
4a4d8108
AM
12886+/* poll.c */
12887+#ifdef CONFIG_AUFS_POLL
12888+unsigned int aufs_poll(struct file *file, poll_table *wait);
12889+#endif
1facf9fc 12890+
4a4d8108
AM
12891+#ifdef CONFIG_AUFS_BR_HFSPLUS
12892+/* hfsplus.c */
392086de
AM
12893+struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
12894+ int force_wr);
4a4d8108
AM
12895+void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
12896+ struct file *h_file);
12897+#else
c1595e42
JR
12898+AuStub(struct file *, au_h_open_pre, return NULL, struct dentry *dentry,
12899+ aufs_bindex_t bindex, int force_wr)
4a4d8108
AM
12900+AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
12901+ struct file *h_file);
12902+#endif
1facf9fc 12903+
4a4d8108
AM
12904+/* f_op.c */
12905+extern const struct file_operations aufs_file_fop;
b912730e 12906+int au_do_open_nondir(struct file *file, int flags, struct file *h_file);
4a4d8108 12907+int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
b912730e 12908+struct file *au_read_pre(struct file *file, int keep_fi);
4a4d8108 12909+
4a4d8108
AM
12910+/* finfo.c */
12911+void au_hfput(struct au_hfile *hf, struct file *file);
12912+void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
12913+ struct file *h_file);
1facf9fc 12914+
4a4d8108 12915+void au_update_figen(struct file *file);
4a4d8108
AM
12916+struct au_fidir *au_fidir_alloc(struct super_block *sb);
12917+int au_fidir_realloc(struct au_finfo *finfo, int nbr);
1facf9fc 12918+
4a4d8108
AM
12919+void au_fi_init_once(void *_fi);
12920+void au_finfo_fin(struct file *file);
12921+int au_finfo_init(struct file *file, struct au_fidir *fidir);
1facf9fc 12922+
4a4d8108
AM
12923+/* ioctl.c */
12924+long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
b752ccd1
AM
12925+#ifdef CONFIG_COMPAT
12926+long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
12927+ unsigned long arg);
c2b27bf2
AM
12928+long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
12929+ unsigned long arg);
b752ccd1 12930+#endif
1facf9fc 12931+
4a4d8108 12932+/* ---------------------------------------------------------------------- */
1facf9fc 12933+
4a4d8108
AM
12934+static inline struct au_finfo *au_fi(struct file *file)
12935+{
38d290e6 12936+ return file->private_data;
4a4d8108 12937+}
1facf9fc 12938+
4a4d8108 12939+/* ---------------------------------------------------------------------- */
1facf9fc 12940+
4a4d8108
AM
12941+/*
12942+ * fi_read_lock, fi_write_lock,
12943+ * fi_read_unlock, fi_write_unlock, fi_downgrade_lock
12944+ */
12945+AuSimpleRwsemFuncs(fi, struct file *f, &au_fi(f)->fi_rwsem);
1308ab2a 12946+
4a4d8108
AM
12947+#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
12948+#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
12949+#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
1facf9fc 12950+
1308ab2a 12951+/* ---------------------------------------------------------------------- */
12952+
4a4d8108
AM
12953+/* todo: hard/soft set? */
12954+static inline aufs_bindex_t au_fbstart(struct file *file)
dece6358 12955+{
4a4d8108
AM
12956+ FiMustAnyLock(file);
12957+ return au_fi(file)->fi_btop;
12958+}
dece6358 12959+
4a4d8108
AM
12960+static inline aufs_bindex_t au_fbend_dir(struct file *file)
12961+{
12962+ FiMustAnyLock(file);
12963+ AuDebugOn(!au_fi(file)->fi_hdir);
12964+ return au_fi(file)->fi_hdir->fd_bbot;
12965+}
1facf9fc 12966+
4a4d8108
AM
12967+static inline struct au_vdir *au_fvdir_cache(struct file *file)
12968+{
12969+ FiMustAnyLock(file);
12970+ AuDebugOn(!au_fi(file)->fi_hdir);
12971+ return au_fi(file)->fi_hdir->fd_vdir_cache;
12972+}
1facf9fc 12973+
4a4d8108
AM
12974+static inline void au_set_fbstart(struct file *file, aufs_bindex_t bindex)
12975+{
12976+ FiMustWriteLock(file);
12977+ au_fi(file)->fi_btop = bindex;
12978+}
1facf9fc 12979+
4a4d8108
AM
12980+static inline void au_set_fbend_dir(struct file *file, aufs_bindex_t bindex)
12981+{
12982+ FiMustWriteLock(file);
12983+ AuDebugOn(!au_fi(file)->fi_hdir);
12984+ au_fi(file)->fi_hdir->fd_bbot = bindex;
12985+}
1308ab2a 12986+
4a4d8108
AM
12987+static inline void au_set_fvdir_cache(struct file *file,
12988+ struct au_vdir *vdir_cache)
12989+{
12990+ FiMustWriteLock(file);
12991+ AuDebugOn(!au_fi(file)->fi_hdir);
12992+ au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
12993+}
dece6358 12994+
4a4d8108
AM
12995+static inline struct file *au_hf_top(struct file *file)
12996+{
12997+ FiMustAnyLock(file);
12998+ AuDebugOn(au_fi(file)->fi_hdir);
12999+ return au_fi(file)->fi_htop.hf_file;
13000+}
1facf9fc 13001+
4a4d8108
AM
13002+static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
13003+{
13004+ FiMustAnyLock(file);
13005+ AuDebugOn(!au_fi(file)->fi_hdir);
13006+ return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
dece6358
AM
13007+}
13008+
4a4d8108
AM
13009+/* todo: memory barrier? */
13010+static inline unsigned int au_figen(struct file *f)
dece6358 13011+{
4a4d8108
AM
13012+ return atomic_read(&au_fi(f)->fi_generation);
13013+}
dece6358 13014+
2cbb1c4b
JR
13015+static inline void au_set_mmapped(struct file *f)
13016+{
13017+ if (atomic_inc_return(&au_fi(f)->fi_mmapped))
13018+ return;
0c3ec466 13019+ pr_warn("fi_mmapped wrapped around\n");
2cbb1c4b
JR
13020+ while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
13021+ ;
13022+}
13023+
13024+static inline void au_unset_mmapped(struct file *f)
13025+{
13026+ atomic_dec(&au_fi(f)->fi_mmapped);
13027+}
13028+
4a4d8108
AM
13029+static inline int au_test_mmapped(struct file *f)
13030+{
2cbb1c4b
JR
13031+ return atomic_read(&au_fi(f)->fi_mmapped);
13032+}
13033+
13034+/* customize vma->vm_file */
13035+
13036+static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
13037+ struct file *file)
13038+{
53392da6
AM
13039+ struct file *f;
13040+
13041+ f = vma->vm_file;
2cbb1c4b
JR
13042+ get_file(file);
13043+ vma->vm_file = file;
53392da6 13044+ fput(f);
2cbb1c4b
JR
13045+}
13046+
13047+#ifdef CONFIG_MMU
13048+#define AuDbgVmRegion(file, vma) do {} while (0)
13049+
13050+static inline void au_vm_file_reset(struct vm_area_struct *vma,
13051+ struct file *file)
13052+{
13053+ au_do_vm_file_reset(vma, file);
13054+}
13055+#else
13056+#define AuDbgVmRegion(file, vma) \
13057+ AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
13058+
13059+static inline void au_vm_file_reset(struct vm_area_struct *vma,
13060+ struct file *file)
13061+{
53392da6
AM
13062+ struct file *f;
13063+
2cbb1c4b 13064+ au_do_vm_file_reset(vma, file);
53392da6 13065+ f = vma->vm_region->vm_file;
2cbb1c4b
JR
13066+ get_file(file);
13067+ vma->vm_region->vm_file = file;
53392da6 13068+ fput(f);
2cbb1c4b
JR
13069+}
13070+#endif /* CONFIG_MMU */
13071+
13072+/* handle vma->vm_prfile */
fb47a38f 13073+static inline void au_vm_prfile_set(struct vm_area_struct *vma,
2cbb1c4b
JR
13074+ struct file *file)
13075+{
2cbb1c4b
JR
13076+ get_file(file);
13077+ vma->vm_prfile = file;
13078+#ifndef CONFIG_MMU
13079+ get_file(file);
13080+ vma->vm_region->vm_prfile = file;
13081+#endif
fb47a38f 13082+}
1308ab2a 13083+
4a4d8108
AM
13084+#endif /* __KERNEL__ */
13085+#endif /* __AUFS_FILE_H__ */
7f207e10
AM
13086diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
13087--- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 13088+++ linux/fs/aufs/finfo.c 2015-09-24 10:47:58.251386326 +0200
b912730e 13089@@ -0,0 +1,157 @@
4a4d8108 13090+/*
2000de60 13091+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
13092+ *
13093+ * This program, aufs is free software; you can redistribute it and/or modify
13094+ * it under the terms of the GNU General Public License as published by
13095+ * the Free Software Foundation; either version 2 of the License, or
13096+ * (at your option) any later version.
13097+ *
13098+ * This program is distributed in the hope that it will be useful,
13099+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
13100+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13101+ * GNU General Public License for more details.
13102+ *
13103+ * You should have received a copy of the GNU General Public License
523b37e3 13104+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 13105+ */
1308ab2a 13106+
4a4d8108
AM
13107+/*
13108+ * file private data
13109+ */
1facf9fc 13110+
4a4d8108 13111+#include "aufs.h"
1facf9fc 13112+
4a4d8108
AM
13113+void au_hfput(struct au_hfile *hf, struct file *file)
13114+{
13115+ /* todo: direct access f_flags */
2cbb1c4b 13116+ if (vfsub_file_flags(file) & __FMODE_EXEC)
4a4d8108
AM
13117+ allow_write_access(hf->hf_file);
13118+ fput(hf->hf_file);
13119+ hf->hf_file = NULL;
e49829fe 13120+ atomic_dec(&hf->hf_br->br_count);
4a4d8108
AM
13121+ hf->hf_br = NULL;
13122+}
1facf9fc 13123+
4a4d8108
AM
13124+void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
13125+{
13126+ struct au_finfo *finfo = au_fi(file);
13127+ struct au_hfile *hf;
13128+ struct au_fidir *fidir;
13129+
13130+ fidir = finfo->fi_hdir;
13131+ if (!fidir) {
13132+ AuDebugOn(finfo->fi_btop != bindex);
13133+ hf = &finfo->fi_htop;
13134+ } else
13135+ hf = fidir->fd_hfile + bindex;
13136+
13137+ if (hf && hf->hf_file)
13138+ au_hfput(hf, file);
13139+ if (val) {
13140+ FiMustWriteLock(file);
b912730e 13141+ AuDebugOn(IS_ERR_OR_NULL(file->f_path.dentry));
4a4d8108 13142+ hf->hf_file = val;
2000de60 13143+ hf->hf_br = au_sbr(file->f_path.dentry->d_sb, bindex);
1308ab2a 13144+ }
4a4d8108 13145+}
1facf9fc 13146+
4a4d8108
AM
13147+void au_update_figen(struct file *file)
13148+{
2000de60 13149+ atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_path.dentry));
4a4d8108 13150+ /* smp_mb(); */ /* atomic_set */
1facf9fc 13151+}
13152+
4a4d8108
AM
13153+/* ---------------------------------------------------------------------- */
13154+
4a4d8108
AM
13155+struct au_fidir *au_fidir_alloc(struct super_block *sb)
13156+{
13157+ struct au_fidir *fidir;
13158+ int nbr;
13159+
13160+ nbr = au_sbend(sb) + 1;
13161+ if (nbr < 2)
13162+ nbr = 2; /* initial allocate for 2 branches */
13163+ fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
13164+ if (fidir) {
13165+ fidir->fd_bbot = -1;
13166+ fidir->fd_nent = nbr;
13167+ fidir->fd_vdir_cache = NULL;
13168+ }
13169+
13170+ return fidir;
13171+}
13172+
13173+int au_fidir_realloc(struct au_finfo *finfo, int nbr)
13174+{
13175+ int err;
13176+ struct au_fidir *fidir, *p;
13177+
13178+ AuRwMustWriteLock(&finfo->fi_rwsem);
13179+ fidir = finfo->fi_hdir;
13180+ AuDebugOn(!fidir);
13181+
13182+ err = -ENOMEM;
13183+ p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
13184+ GFP_NOFS);
13185+ if (p) {
13186+ p->fd_nent = nbr;
13187+ finfo->fi_hdir = p;
13188+ err = 0;
13189+ }
1facf9fc 13190+
dece6358 13191+ return err;
1facf9fc 13192+}
1308ab2a 13193+
13194+/* ---------------------------------------------------------------------- */
13195+
4a4d8108 13196+void au_finfo_fin(struct file *file)
1308ab2a 13197+{
4a4d8108
AM
13198+ struct au_finfo *finfo;
13199+
2000de60 13200+ au_nfiles_dec(file->f_path.dentry->d_sb);
7f207e10 13201+
4a4d8108
AM
13202+ finfo = au_fi(file);
13203+ AuDebugOn(finfo->fi_hdir);
13204+ AuRwDestroy(&finfo->fi_rwsem);
13205+ au_cache_free_finfo(finfo);
1308ab2a 13206+}
1308ab2a 13207+
e49829fe 13208+void au_fi_init_once(void *_finfo)
4a4d8108 13209+{
e49829fe 13210+ struct au_finfo *finfo = _finfo;
2cbb1c4b 13211+ static struct lock_class_key aufs_fi;
1308ab2a 13212+
e49829fe
JR
13213+ au_rw_init(&finfo->fi_rwsem);
13214+ au_rw_class(&finfo->fi_rwsem, &aufs_fi);
4a4d8108 13215+}
1308ab2a 13216+
4a4d8108
AM
13217+int au_finfo_init(struct file *file, struct au_fidir *fidir)
13218+{
1716fcea 13219+ int err;
4a4d8108
AM
13220+ struct au_finfo *finfo;
13221+ struct dentry *dentry;
13222+
13223+ err = -ENOMEM;
2000de60 13224+ dentry = file->f_path.dentry;
4a4d8108
AM
13225+ finfo = au_cache_alloc_finfo();
13226+ if (unlikely(!finfo))
13227+ goto out;
13228+
13229+ err = 0;
7f207e10 13230+ au_nfiles_inc(dentry->d_sb);
1716fcea
AM
13231+ /* verbose coding for lock class name */
13232+ if (!fidir)
13233+ au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcNonDir_FIINFO);
13234+ else
13235+ au_rw_class(&finfo->fi_rwsem, au_lc_key + AuLcDir_FIINFO);
4a4d8108
AM
13236+ au_rw_write_lock(&finfo->fi_rwsem);
13237+ finfo->fi_btop = -1;
13238+ finfo->fi_hdir = fidir;
13239+ atomic_set(&finfo->fi_generation, au_digen(dentry));
13240+ /* smp_mb(); */ /* atomic_set */
13241+
13242+ file->private_data = finfo;
13243+
13244+out:
13245+ return err;
13246+}
7f207e10
AM
13247diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
13248--- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 13249+++ linux/fs/aufs/f_op.c 2015-09-24 10:47:58.251386326 +0200
5527c038 13250@@ -0,0 +1,738 @@
dece6358 13251+/*
2000de60 13252+ * Copyright (C) 2005-2015 Junjiro R. Okajima
dece6358
AM
13253+ *
13254+ * This program, aufs is free software; you can redistribute it and/or modify
13255+ * it under the terms of the GNU General Public License as published by
13256+ * the Free Software Foundation; either version 2 of the License, or
13257+ * (at your option) any later version.
13258+ *
13259+ * This program is distributed in the hope that it will be useful,
13260+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
13261+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13262+ * GNU General Public License for more details.
13263+ *
13264+ * You should have received a copy of the GNU General Public License
523b37e3 13265+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
dece6358 13266+ */
1facf9fc 13267+
13268+/*
4a4d8108 13269+ * file and vm operations
1facf9fc 13270+ */
dece6358 13271+
86dc4139 13272+#include <linux/aio.h>
4a4d8108
AM
13273+#include <linux/fs_stack.h>
13274+#include <linux/mman.h>
4a4d8108 13275+#include <linux/security.h>
dece6358
AM
13276+#include "aufs.h"
13277+
b912730e 13278+int au_do_open_nondir(struct file *file, int flags, struct file *h_file)
1facf9fc 13279+{
4a4d8108
AM
13280+ int err;
13281+ aufs_bindex_t bindex;
4a4d8108
AM
13282+ struct dentry *dentry;
13283+ struct au_finfo *finfo;
38d290e6 13284+ struct inode *h_inode;
4a4d8108
AM
13285+
13286+ FiMustWriteLock(file);
13287+
523b37e3 13288+ err = 0;
2000de60 13289+ dentry = file->f_path.dentry;
b912730e 13290+ AuDebugOn(IS_ERR_OR_NULL(dentry));
4a4d8108
AM
13291+ finfo = au_fi(file);
13292+ memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
2cbb1c4b 13293+ atomic_set(&finfo->fi_mmapped, 0);
4a4d8108 13294+ bindex = au_dbstart(dentry);
b912730e
AM
13295+ if (!h_file)
13296+ h_file = au_h_open(dentry, bindex, flags, file, /*force_wr*/0);
13297+ else
13298+ get_file(h_file);
4a4d8108
AM
13299+ if (IS_ERR(h_file))
13300+ err = PTR_ERR(h_file);
13301+ else {
38d290e6
JR
13302+ if ((flags & __O_TMPFILE)
13303+ && !(flags & O_EXCL)) {
13304+ h_inode = file_inode(h_file);
13305+ spin_lock(&h_inode->i_lock);
13306+ h_inode->i_state |= I_LINKABLE;
13307+ spin_unlock(&h_inode->i_lock);
13308+ }
4a4d8108
AM
13309+ au_set_fbstart(file, bindex);
13310+ au_set_h_fptr(file, bindex, h_file);
13311+ au_update_figen(file);
13312+ /* todo: necessary? */
13313+ /* file->f_ra = h_file->f_ra; */
13314+ }
027c5e7a 13315+
4a4d8108 13316+ return err;
1facf9fc 13317+}
13318+
4a4d8108
AM
13319+static int aufs_open_nondir(struct inode *inode __maybe_unused,
13320+ struct file *file)
1facf9fc 13321+{
4a4d8108 13322+ int err;
1308ab2a 13323+ struct super_block *sb;
b912730e
AM
13324+ struct au_do_open_args args = {
13325+ .open = au_do_open_nondir
13326+ };
1facf9fc 13327+
523b37e3
AM
13328+ AuDbg("%pD, f_flags 0x%x, f_mode 0x%x\n",
13329+ file, vfsub_file_flags(file), file->f_mode);
1facf9fc 13330+
2000de60 13331+ sb = file->f_path.dentry->d_sb;
4a4d8108 13332+ si_read_lock(sb, AuLock_FLUSH);
b912730e 13333+ err = au_do_open(file, &args);
4a4d8108
AM
13334+ si_read_unlock(sb);
13335+ return err;
13336+}
1facf9fc 13337+
4a4d8108
AM
13338+int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
13339+{
13340+ struct au_finfo *finfo;
13341+ aufs_bindex_t bindex;
1facf9fc 13342+
4a4d8108 13343+ finfo = au_fi(file);
2000de60
JR
13344+ au_sphl_del(&finfo->fi_hlist,
13345+ &au_sbi(file->f_path.dentry->d_sb)->si_files);
4a4d8108 13346+ bindex = finfo->fi_btop;
b4510431 13347+ if (bindex >= 0)
4a4d8108 13348+ au_set_h_fptr(file, bindex, NULL);
7f207e10 13349+
4a4d8108
AM
13350+ au_finfo_fin(file);
13351+ return 0;
1facf9fc 13352+}
13353+
4a4d8108
AM
13354+/* ---------------------------------------------------------------------- */
13355+
13356+static int au_do_flush_nondir(struct file *file, fl_owner_t id)
dece6358 13357+{
1308ab2a 13358+ int err;
4a4d8108
AM
13359+ struct file *h_file;
13360+
13361+ err = 0;
13362+ h_file = au_hf_top(file);
13363+ if (h_file)
13364+ err = vfsub_flush(h_file, id);
13365+ return err;
13366+}
13367+
13368+static int aufs_flush_nondir(struct file *file, fl_owner_t id)
13369+{
13370+ return au_do_flush(file, id, au_do_flush_nondir);
13371+}
13372+
13373+/* ---------------------------------------------------------------------- */
9dbd164d
AM
13374+/*
13375+ * read and write functions acquire [fdi]_rwsem once, but release before
13376+ * mmap_sem. This is because to stop a race condition between mmap(2).
13377+ * Releasing these aufs-rwsem should be safe, no branch-mamagement (by keeping
13378+ * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
13379+ * read functions after [fdi]_rwsem are released, but it should be harmless.
13380+ */
4a4d8108 13381+
b912730e
AM
13382+/* Callers should call au_read_post() or fput() in the end */
13383+struct file *au_read_pre(struct file *file, int keep_fi)
4a4d8108 13384+{
4a4d8108 13385+ struct file *h_file;
b912730e 13386+ int err;
1facf9fc 13387+
4a4d8108 13388+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
b912730e
AM
13389+ if (!err) {
13390+ di_read_unlock(file->f_path.dentry, AuLock_IR);
13391+ h_file = au_hf_top(file);
13392+ get_file(h_file);
13393+ if (!keep_fi)
13394+ fi_read_unlock(file);
13395+ } else
13396+ h_file = ERR_PTR(err);
13397+
13398+ return h_file;
13399+}
13400+
13401+static void au_read_post(struct inode *inode, struct file *h_file)
13402+{
13403+ /* update without lock, I don't think it a problem */
13404+ fsstack_copy_attr_atime(inode, file_inode(h_file));
13405+ fput(h_file);
13406+}
13407+
13408+struct au_write_pre {
13409+ blkcnt_t blks;
13410+ aufs_bindex_t bstart;
13411+};
13412+
13413+/*
13414+ * return with iinfo is write-locked
13415+ * callers should call au_write_post() or iinfo_write_unlock() + fput() in the
13416+ * end
13417+ */
13418+static struct file *au_write_pre(struct file *file, int do_ready,
13419+ struct au_write_pre *wpre)
13420+{
13421+ struct file *h_file;
13422+ struct dentry *dentry;
13423+ int err;
13424+ struct au_pin pin;
13425+
13426+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
13427+ h_file = ERR_PTR(err);
dece6358
AM
13428+ if (unlikely(err))
13429+ goto out;
1facf9fc 13430+
b912730e
AM
13431+ dentry = file->f_path.dentry;
13432+ if (do_ready) {
13433+ err = au_ready_to_write(file, -1, &pin);
13434+ if (unlikely(err)) {
13435+ h_file = ERR_PTR(err);
13436+ di_write_unlock(dentry);
13437+ goto out_fi;
13438+ }
13439+ }
13440+
13441+ di_downgrade_lock(dentry, /*flags*/0);
13442+ if (wpre)
13443+ wpre->bstart = au_fbstart(file);
4a4d8108 13444+ h_file = au_hf_top(file);
9dbd164d 13445+ get_file(h_file);
b912730e
AM
13446+ if (wpre)
13447+ wpre->blks = file_inode(h_file)->i_blocks;
13448+ if (do_ready)
13449+ au_unpin(&pin);
13450+ di_read_unlock(dentry, /*flags*/0);
13451+
13452+out_fi:
13453+ fi_write_unlock(file);
13454+out:
13455+ return h_file;
13456+}
13457+
13458+static void au_write_post(struct inode *inode, struct file *h_file,
13459+ struct au_write_pre *wpre, ssize_t written)
13460+{
13461+ struct inode *h_inode;
13462+
13463+ au_cpup_attr_timesizes(inode);
13464+ AuDebugOn(au_ibstart(inode) != wpre->bstart);
13465+ h_inode = file_inode(h_file);
13466+ inode->i_mode = h_inode->i_mode;
13467+ ii_write_unlock(inode);
13468+ fput(h_file);
13469+
13470+ /* AuDbg("blks %llu, %llu\n", (u64)blks, (u64)h_inode->i_blocks); */
13471+ if (written > 0)
13472+ au_fhsm_wrote(inode->i_sb, wpre->bstart,
13473+ /*force*/h_inode->i_blocks > wpre->blks);
13474+}
13475+
13476+static ssize_t aufs_read(struct file *file, char __user *buf, size_t count,
13477+ loff_t *ppos)
13478+{
13479+ ssize_t err;
13480+ struct inode *inode;
13481+ struct file *h_file;
13482+ struct super_block *sb;
13483+
13484+ inode = file_inode(file);
13485+ sb = inode->i_sb;
13486+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
13487+
13488+ h_file = au_read_pre(file, /*keep_fi*/0);
13489+ err = PTR_ERR(h_file);
13490+ if (IS_ERR(h_file))
13491+ goto out;
9dbd164d
AM
13492+
13493+ /* filedata may be obsoleted by concurrent copyup, but no problem */
4a4d8108
AM
13494+ err = vfsub_read_u(h_file, buf, count, ppos);
13495+ /* todo: necessary? */
13496+ /* file->f_ra = h_file->f_ra; */
b912730e 13497+ au_read_post(inode, h_file);
1308ab2a 13498+
4f0767ce 13499+out:
dece6358
AM
13500+ si_read_unlock(sb);
13501+ return err;
13502+}
1facf9fc 13503+
e49829fe
JR
13504+/*
13505+ * todo: very ugly
13506+ * it locks both of i_mutex and si_rwsem for read in safe.
13507+ * if the plink maintenance mode continues forever (that is the problem),
13508+ * may loop forever.
13509+ */
13510+static void au_mtx_and_read_lock(struct inode *inode)
13511+{
13512+ int err;
13513+ struct super_block *sb = inode->i_sb;
13514+
13515+ while (1) {
13516+ mutex_lock(&inode->i_mutex);
13517+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
13518+ if (!err)
13519+ break;
13520+ mutex_unlock(&inode->i_mutex);
13521+ si_read_lock(sb, AuLock_NOPLMW);
13522+ si_read_unlock(sb);
13523+ }
13524+}
13525+
4a4d8108
AM
13526+static ssize_t aufs_write(struct file *file, const char __user *ubuf,
13527+ size_t count, loff_t *ppos)
dece6358 13528+{
4a4d8108 13529+ ssize_t err;
b912730e
AM
13530+ struct au_write_pre wpre;
13531+ struct inode *inode;
4a4d8108
AM
13532+ struct file *h_file;
13533+ char __user *buf = (char __user *)ubuf;
1facf9fc 13534+
b912730e 13535+ inode = file_inode(file);
e49829fe 13536+ au_mtx_and_read_lock(inode);
1facf9fc 13537+
b912730e
AM
13538+ h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13539+ err = PTR_ERR(h_file);
13540+ if (IS_ERR(h_file))
9dbd164d 13541+ goto out;
9dbd164d 13542+
4a4d8108 13543+ err = vfsub_write_u(h_file, buf, count, ppos);
b912730e 13544+ au_write_post(inode, h_file, &wpre, err);
1facf9fc 13545+
4f0767ce 13546+out:
b912730e 13547+ si_read_unlock(inode->i_sb);
4a4d8108 13548+ mutex_unlock(&inode->i_mutex);
dece6358
AM
13549+ return err;
13550+}
1facf9fc 13551+
076b876e
AM
13552+static ssize_t au_do_iter(struct file *h_file, int rw, struct kiocb *kio,
13553+ struct iov_iter *iov_iter)
dece6358 13554+{
4a4d8108
AM
13555+ ssize_t err;
13556+ struct file *file;
076b876e 13557+ ssize_t (*iter)(struct kiocb *, struct iov_iter *);
1facf9fc 13558+
4a4d8108
AM
13559+ err = security_file_permission(h_file, rw);
13560+ if (unlikely(err))
13561+ goto out;
1facf9fc 13562+
4a4d8108 13563+ err = -ENOSYS;
076b876e 13564+ iter = NULL;
5527c038 13565+ if (rw == MAY_READ)
076b876e 13566+ iter = h_file->f_op->read_iter;
5527c038 13567+ else if (rw == MAY_WRITE)
076b876e 13568+ iter = h_file->f_op->write_iter;
076b876e
AM
13569+
13570+ file = kio->ki_filp;
13571+ kio->ki_filp = h_file;
13572+ if (iter) {
2cbb1c4b 13573+ lockdep_off();
076b876e
AM
13574+ err = iter(kio, iov_iter);
13575+ lockdep_on();
4a4d8108
AM
13576+ } else
13577+ /* currently there is no such fs */
13578+ WARN_ON_ONCE(1);
076b876e 13579+ kio->ki_filp = file;
1facf9fc 13580+
4f0767ce 13581+out:
dece6358
AM
13582+ return err;
13583+}
1facf9fc 13584+
076b876e 13585+static ssize_t aufs_read_iter(struct kiocb *kio, struct iov_iter *iov_iter)
1facf9fc 13586+{
4a4d8108
AM
13587+ ssize_t err;
13588+ struct file *file, *h_file;
b912730e 13589+ struct inode *inode;
dece6358 13590+ struct super_block *sb;
1facf9fc 13591+
4a4d8108 13592+ file = kio->ki_filp;
b912730e
AM
13593+ inode = file_inode(file);
13594+ sb = inode->i_sb;
e49829fe 13595+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
4a4d8108 13596+
b912730e
AM
13597+ h_file = au_read_pre(file, /*keep_fi*/0);
13598+ err = PTR_ERR(h_file);
13599+ if (IS_ERR(h_file))
13600+ goto out;
9dbd164d 13601+
076b876e 13602+ err = au_do_iter(h_file, MAY_READ, kio, iov_iter);
4a4d8108
AM
13603+ /* todo: necessary? */
13604+ /* file->f_ra = h_file->f_ra; */
b912730e 13605+ au_read_post(inode, h_file);
1facf9fc 13606+
4f0767ce 13607+out:
4a4d8108 13608+ si_read_unlock(sb);
1308ab2a 13609+ return err;
13610+}
1facf9fc 13611+
076b876e 13612+static ssize_t aufs_write_iter(struct kiocb *kio, struct iov_iter *iov_iter)
1308ab2a 13613+{
4a4d8108 13614+ ssize_t err;
b912730e
AM
13615+ struct au_write_pre wpre;
13616+ struct inode *inode;
4a4d8108 13617+ struct file *file, *h_file;
1308ab2a 13618+
4a4d8108 13619+ file = kio->ki_filp;
b912730e 13620+ inode = file_inode(file);
e49829fe
JR
13621+ au_mtx_and_read_lock(inode);
13622+
b912730e
AM
13623+ h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13624+ err = PTR_ERR(h_file);
13625+ if (IS_ERR(h_file))
9dbd164d 13626+ goto out;
9dbd164d 13627+
076b876e 13628+ err = au_do_iter(h_file, MAY_WRITE, kio, iov_iter);
b912730e 13629+ au_write_post(inode, h_file, &wpre, err);
1facf9fc 13630+
4f0767ce 13631+out:
b912730e 13632+ si_read_unlock(inode->i_sb);
4a4d8108 13633+ mutex_unlock(&inode->i_mutex);
dece6358 13634+ return err;
1facf9fc 13635+}
13636+
4a4d8108
AM
13637+static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
13638+ struct pipe_inode_info *pipe, size_t len,
13639+ unsigned int flags)
1facf9fc 13640+{
4a4d8108
AM
13641+ ssize_t err;
13642+ struct file *h_file;
b912730e 13643+ struct inode *inode;
dece6358 13644+ struct super_block *sb;
1facf9fc 13645+
b912730e
AM
13646+ inode = file_inode(file);
13647+ sb = inode->i_sb;
e49829fe 13648+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
b912730e
AM
13649+
13650+ h_file = au_read_pre(file, /*keep_fi*/1);
13651+ err = PTR_ERR(h_file);
13652+ if (IS_ERR(h_file))
dece6358 13653+ goto out;
1facf9fc 13654+
4a4d8108 13655+ if (au_test_loopback_kthread()) {
2000de60 13656+ au_warn_loopback(h_file->f_path.dentry->d_sb);
87a755f4
AM
13657+ if (file->f_mapping != h_file->f_mapping) {
13658+ file->f_mapping = h_file->f_mapping;
13659+ smp_mb(); /* unnecessary? */
13660+ }
1308ab2a 13661+ }
9dbd164d
AM
13662+ fi_read_unlock(file);
13663+
4a4d8108
AM
13664+ err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
13665+ /* todo: necessasry? */
13666+ /* file->f_ra = h_file->f_ra; */
b912730e 13667+ au_read_post(inode, h_file);
1facf9fc 13668+
4f0767ce 13669+out:
4a4d8108 13670+ si_read_unlock(sb);
dece6358 13671+ return err;
1facf9fc 13672+}
13673+
4a4d8108
AM
13674+static ssize_t
13675+aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
13676+ size_t len, unsigned int flags)
1facf9fc 13677+{
4a4d8108 13678+ ssize_t err;
b912730e
AM
13679+ struct au_write_pre wpre;
13680+ struct inode *inode;
076b876e 13681+ struct file *h_file;
1facf9fc 13682+
b912730e 13683+ inode = file_inode(file);
e49829fe 13684+ au_mtx_and_read_lock(inode);
9dbd164d 13685+
b912730e
AM
13686+ h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13687+ err = PTR_ERR(h_file);
13688+ if (IS_ERR(h_file))
9dbd164d 13689+ goto out;
9dbd164d 13690+
4a4d8108 13691+ err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
b912730e 13692+ au_write_post(inode, h_file, &wpre, err);
1facf9fc 13693+
4f0767ce 13694+out:
b912730e 13695+ si_read_unlock(inode->i_sb);
4a4d8108
AM
13696+ mutex_unlock(&inode->i_mutex);
13697+ return err;
13698+}
1facf9fc 13699+
38d290e6
JR
13700+static long aufs_fallocate(struct file *file, int mode, loff_t offset,
13701+ loff_t len)
13702+{
13703+ long err;
b912730e 13704+ struct au_write_pre wpre;
38d290e6
JR
13705+ struct inode *inode;
13706+ struct file *h_file;
13707+
b912730e 13708+ inode = file_inode(file);
38d290e6
JR
13709+ au_mtx_and_read_lock(inode);
13710+
b912730e
AM
13711+ h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13712+ err = PTR_ERR(h_file);
13713+ if (IS_ERR(h_file))
38d290e6 13714+ goto out;
38d290e6
JR
13715+
13716+ lockdep_off();
03673fb0 13717+ err = vfs_fallocate(h_file, mode, offset, len);
38d290e6 13718+ lockdep_on();
b912730e 13719+ au_write_post(inode, h_file, &wpre, /*written*/1);
38d290e6
JR
13720+
13721+out:
b912730e 13722+ si_read_unlock(inode->i_sb);
38d290e6
JR
13723+ mutex_unlock(&inode->i_mutex);
13724+ return err;
13725+}
13726+
4a4d8108
AM
13727+/* ---------------------------------------------------------------------- */
13728+
9dbd164d
AM
13729+/*
13730+ * The locking order around current->mmap_sem.
13731+ * - in most and regular cases
13732+ * file I/O syscall -- aufs_read() or something
13733+ * -- si_rwsem for read -- mmap_sem
13734+ * (Note that [fdi]i_rwsem are released before mmap_sem).
13735+ * - in mmap case
13736+ * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
13737+ * This AB-BA order is definitly bad, but is not a problem since "si_rwsem for
13738+ * read" allows muliple processes to acquire it and [fdi]i_rwsem are not held in
13739+ * file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
13740+ * It means that when aufs acquires si_rwsem for write, the process should never
13741+ * acquire mmap_sem.
13742+ *
392086de 13743+ * Actually aufs_iterate() holds [fdi]i_rwsem before mmap_sem, but this is not a
9dbd164d
AM
13744+ * problem either since any directory is not able to be mmap-ed.
13745+ * The similar scenario is applied to aufs_readlink() too.
13746+ */
13747+
38d290e6 13748+#if 0 /* stop calling security_file_mmap() */
2dfbb274
AM
13749+/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
13750+#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
13751+
13752+static unsigned long au_arch_prot_conv(unsigned long flags)
13753+{
13754+ /* currently ppc64 only */
13755+#ifdef CONFIG_PPC64
13756+ /* cf. linux/arch/powerpc/include/asm/mman.h */
13757+ AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
13758+ return AuConv_VM_PROT(flags, SAO);
13759+#else
13760+ AuDebugOn(arch_calc_vm_prot_bits(-1));
13761+ return 0;
13762+#endif
13763+}
13764+
13765+static unsigned long au_prot_conv(unsigned long flags)
13766+{
13767+ return AuConv_VM_PROT(flags, READ)
13768+ | AuConv_VM_PROT(flags, WRITE)
13769+ | AuConv_VM_PROT(flags, EXEC)
13770+ | au_arch_prot_conv(flags);
13771+}
13772+
13773+/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
13774+#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
13775+
13776+static unsigned long au_flag_conv(unsigned long flags)
13777+{
13778+ return AuConv_VM_MAP(flags, GROWSDOWN)
13779+ | AuConv_VM_MAP(flags, DENYWRITE)
2dfbb274
AM
13780+ | AuConv_VM_MAP(flags, LOCKED);
13781+}
38d290e6 13782+#endif
2dfbb274 13783+
9dbd164d 13784+static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
dece6358 13785+{
4a4d8108 13786+ int err;
4a4d8108 13787+ const unsigned char wlock
9dbd164d 13788+ = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
4a4d8108 13789+ struct super_block *sb;
9dbd164d 13790+ struct file *h_file;
b912730e 13791+ struct inode *inode;
9dbd164d
AM
13792+
13793+ AuDbgVmRegion(file, vma);
1308ab2a 13794+
b912730e
AM
13795+ inode = file_inode(file);
13796+ sb = inode->i_sb;
9dbd164d 13797+ lockdep_off();
e49829fe 13798+ si_read_lock(sb, AuLock_NOPLMW);
4a4d8108 13799+
b912730e 13800+ h_file = au_write_pre(file, wlock, /*wpre*/NULL);
9dbd164d 13801+ lockdep_on();
b912730e
AM
13802+ err = PTR_ERR(h_file);
13803+ if (IS_ERR(h_file))
13804+ goto out;
1308ab2a 13805+
b912730e
AM
13806+ err = 0;
13807+ au_set_mmapped(file);
9dbd164d 13808+ au_vm_file_reset(vma, h_file);
38d290e6
JR
13809+ /*
13810+ * we cannot call security_mmap_file() here since it may acquire
13811+ * mmap_sem or i_mutex.
13812+ *
13813+ * err = security_mmap_file(h_file, au_prot_conv(vma->vm_flags),
13814+ * au_flag_conv(vma->vm_flags));
13815+ */
9dbd164d
AM
13816+ if (!err)
13817+ err = h_file->f_op->mmap(h_file, vma);
b912730e
AM
13818+ if (!err) {
13819+ au_vm_prfile_set(vma, file);
13820+ fsstack_copy_attr_atime(inode, file_inode(h_file));
13821+ goto out_fput; /* success */
13822+ }
2cbb1c4b
JR
13823+ au_unset_mmapped(file);
13824+ au_vm_file_reset(vma, file);
b912730e 13825+
2cbb1c4b 13826+out_fput:
9dbd164d 13827+ lockdep_off();
b912730e
AM
13828+ ii_write_unlock(inode);
13829+ lockdep_on();
13830+ fput(h_file);
4f0767ce 13831+out:
b912730e 13832+ lockdep_off();
9dbd164d
AM
13833+ si_read_unlock(sb);
13834+ lockdep_on();
13835+ AuTraceErr(err);
4a4d8108
AM
13836+ return err;
13837+}
13838+
13839+/* ---------------------------------------------------------------------- */
13840+
1e00d052
AM
13841+static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
13842+ int datasync)
4a4d8108
AM
13843+{
13844+ int err;
b912730e 13845+ struct au_write_pre wpre;
4a4d8108
AM
13846+ struct inode *inode;
13847+ struct file *h_file;
4a4d8108
AM
13848+
13849+ err = 0; /* -EBADF; */ /* posix? */
13850+ if (unlikely(!(file->f_mode & FMODE_WRITE)))
b912730e 13851+ goto out;
4a4d8108 13852+
b912730e
AM
13853+ inode = file_inode(file);
13854+ au_mtx_and_read_lock(inode);
13855+
13856+ h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13857+ err = PTR_ERR(h_file);
13858+ if (IS_ERR(h_file))
4a4d8108 13859+ goto out_unlock;
4a4d8108 13860+
53392da6 13861+ err = vfsub_fsync(h_file, &h_file->f_path, datasync);
b912730e 13862+ au_write_post(inode, h_file, &wpre, /*written*/0);
4a4d8108 13863+
4f0767ce 13864+out_unlock:
b912730e 13865+ si_read_unlock(inode->i_sb);
1e00d052 13866+ mutex_unlock(&inode->i_mutex);
b912730e 13867+out:
4a4d8108 13868+ return err;
dece6358
AM
13869+}
13870+
4a4d8108
AM
13871+/* no one supports this operation, currently */
13872+#if 0
13873+static int aufs_aio_fsync_nondir(struct kiocb *kio, int datasync)
dece6358 13874+{
4a4d8108 13875+ int err;
b912730e 13876+ struct au_write_pre wpre;
4a4d8108
AM
13877+ struct inode *inode;
13878+ struct file *file, *h_file;
1308ab2a 13879+
4a4d8108
AM
13880+ err = 0; /* -EBADF; */ /* posix? */
13881+ if (unlikely(!(file->f_mode & FMODE_WRITE)))
13882+ goto out;
1308ab2a 13883+
b912730e
AM
13884+ file = kio->ki_filp;
13885+ inode = file_inode(file);
13886+ au_mtx_and_read_lock(inode);
13887+
13888+ h_file = au_write_pre(file, /*do_ready*/1, &wpre);
13889+ err = PTR_ERR(h_file);
13890+ if (IS_ERR(h_file))
4a4d8108 13891+ goto out_unlock;
1308ab2a 13892+
4a4d8108
AM
13893+ err = -ENOSYS;
13894+ h_file = au_hf_top(file);
523b37e3 13895+ if (h_file->f_op->aio_fsync) {
4a4d8108 13896+ struct mutex *h_mtx;
1308ab2a 13897+
c06a8ce3 13898+ h_mtx = &file_inode(h_file)->i_mutex;
4a4d8108
AM
13899+ if (!is_sync_kiocb(kio)) {
13900+ get_file(h_file);
13901+ fput(file);
13902+ }
13903+ kio->ki_filp = h_file;
13904+ err = h_file->f_op->aio_fsync(kio, datasync);
13905+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
13906+ if (!err)
13907+ vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL);
13908+ /*ignore*/
4a4d8108
AM
13909+ mutex_unlock(h_mtx);
13910+ }
b912730e 13911+ au_write_post(inode, h_file, &wpre, /*written*/0);
1308ab2a 13912+
4f0767ce 13913+out_unlock:
e49829fe 13914+ si_read_unlock(inode->sb);
4a4d8108 13915+ mutex_unlock(&inode->i_mutex);
b912730e 13916+out:
4a4d8108 13917+ return err;
dece6358 13918+}
4a4d8108 13919+#endif
dece6358 13920+
4a4d8108 13921+static int aufs_fasync(int fd, struct file *file, int flag)
dece6358 13922+{
4a4d8108
AM
13923+ int err;
13924+ struct file *h_file;
4a4d8108 13925+ struct super_block *sb;
1308ab2a 13926+
b912730e 13927+ sb = file->f_path.dentry->d_sb;
e49829fe 13928+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
b912730e
AM
13929+
13930+ h_file = au_read_pre(file, /*keep_fi*/0);
13931+ err = PTR_ERR(h_file);
13932+ if (IS_ERR(h_file))
4a4d8108
AM
13933+ goto out;
13934+
523b37e3 13935+ if (h_file->f_op->fasync)
4a4d8108 13936+ err = h_file->f_op->fasync(fd, h_file, flag);
b912730e 13937+ fput(h_file); /* instead of au_read_post() */
1308ab2a 13938+
4f0767ce 13939+out:
4a4d8108 13940+ si_read_unlock(sb);
1308ab2a 13941+ return err;
dece6358 13942+}
4a4d8108
AM
13943+
13944+/* ---------------------------------------------------------------------- */
13945+
13946+/* no one supports this operation, currently */
13947+#if 0
13948+static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
2000de60 13949+ size_t len, loff_t *pos, int more)
4a4d8108
AM
13950+{
13951+}
13952+#endif
13953+
13954+/* ---------------------------------------------------------------------- */
13955+
13956+const struct file_operations aufs_file_fop = {
13957+ .owner = THIS_MODULE,
2cbb1c4b 13958+
027c5e7a 13959+ .llseek = default_llseek,
4a4d8108
AM
13960+
13961+ .read = aufs_read,
13962+ .write = aufs_write,
076b876e
AM
13963+ .read_iter = aufs_read_iter,
13964+ .write_iter = aufs_write_iter,
13965+
4a4d8108
AM
13966+#ifdef CONFIG_AUFS_POLL
13967+ .poll = aufs_poll,
13968+#endif
13969+ .unlocked_ioctl = aufs_ioctl_nondir,
b752ccd1 13970+#ifdef CONFIG_COMPAT
c2b27bf2 13971+ .compat_ioctl = aufs_compat_ioctl_nondir,
b752ccd1 13972+#endif
4a4d8108
AM
13973+ .mmap = aufs_mmap,
13974+ .open = aufs_open_nondir,
13975+ .flush = aufs_flush_nondir,
13976+ .release = aufs_release_nondir,
13977+ .fsync = aufs_fsync_nondir,
13978+ /* .aio_fsync = aufs_aio_fsync_nondir, */
13979+ .fasync = aufs_fasync,
13980+ /* .sendpage = aufs_sendpage, */
13981+ .splice_write = aufs_splice_write,
13982+ .splice_read = aufs_splice_read,
13983+#if 0
13984+ .aio_splice_write = aufs_aio_splice_write,
38d290e6 13985+ .aio_splice_read = aufs_aio_splice_read,
4a4d8108 13986+#endif
38d290e6 13987+ .fallocate = aufs_fallocate
4a4d8108 13988+};
7f207e10
AM
13989diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
13990--- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 13991+++ linux/fs/aufs/fstype.h 2015-09-24 10:47:58.254719746 +0200
b912730e 13992@@ -0,0 +1,400 @@
4a4d8108 13993+/*
2000de60 13994+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
13995+ *
13996+ * This program, aufs is free software; you can redistribute it and/or modify
13997+ * it under the terms of the GNU General Public License as published by
13998+ * the Free Software Foundation; either version 2 of the License, or
13999+ * (at your option) any later version.
14000+ *
14001+ * This program is distributed in the hope that it will be useful,
14002+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14003+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14004+ * GNU General Public License for more details.
14005+ *
14006+ * You should have received a copy of the GNU General Public License
523b37e3 14007+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
14008+ */
14009+
14010+/*
14011+ * judging filesystem type
14012+ */
14013+
14014+#ifndef __AUFS_FSTYPE_H__
14015+#define __AUFS_FSTYPE_H__
14016+
14017+#ifdef __KERNEL__
14018+
14019+#include <linux/fs.h>
14020+#include <linux/magic.h>
14021+#include <linux/romfs_fs.h>
b912730e 14022+#include <linux/nfs_fs.h>
4a4d8108
AM
14023+
14024+static inline int au_test_aufs(struct super_block *sb)
14025+{
14026+ return sb->s_magic == AUFS_SUPER_MAGIC;
14027+}
14028+
14029+static inline const char *au_sbtype(struct super_block *sb)
14030+{
14031+ return sb->s_type->name;
14032+}
1308ab2a 14033+
14034+static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
14035+{
2000de60
JR
14036+#if defined(CONFIG_ISO9660_FS) || defined(CONFIG_ISO9660_FS_MODULE)
14037+ return sb->s_magic == ISOFS_SUPER_MAGIC;
dece6358
AM
14038+#else
14039+ return 0;
14040+#endif
14041+}
14042+
1308ab2a 14043+static inline int au_test_romfs(struct super_block *sb __maybe_unused)
dece6358 14044+{
2000de60
JR
14045+#if defined(CONFIG_ROMFS_FS) || defined(CONFIG_ROMFS_FS_MODULE)
14046+ return sb->s_magic == ROMFS_MAGIC;
dece6358
AM
14047+#else
14048+ return 0;
14049+#endif
14050+}
14051+
1308ab2a 14052+static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
dece6358 14053+{
1308ab2a 14054+#if defined(CONFIG_CRAMFS) || defined(CONFIG_CRAMFS_MODULE)
14055+ return sb->s_magic == CRAMFS_MAGIC;
14056+#endif
14057+ return 0;
14058+}
14059+
14060+static inline int au_test_nfs(struct super_block *sb __maybe_unused)
14061+{
14062+#if defined(CONFIG_NFS_FS) || defined(CONFIG_NFS_FS_MODULE)
14063+ return sb->s_magic == NFS_SUPER_MAGIC;
dece6358
AM
14064+#else
14065+ return 0;
14066+#endif
14067+}
14068+
1308ab2a 14069+static inline int au_test_fuse(struct super_block *sb __maybe_unused)
dece6358 14070+{
1308ab2a 14071+#if defined(CONFIG_FUSE_FS) || defined(CONFIG_FUSE_FS_MODULE)
14072+ return sb->s_magic == FUSE_SUPER_MAGIC;
dece6358
AM
14073+#else
14074+ return 0;
14075+#endif
14076+}
14077+
1308ab2a 14078+static inline int au_test_xfs(struct super_block *sb __maybe_unused)
dece6358 14079+{
1308ab2a 14080+#if defined(CONFIG_XFS_FS) || defined(CONFIG_XFS_FS_MODULE)
14081+ return sb->s_magic == XFS_SB_MAGIC;
dece6358
AM
14082+#else
14083+ return 0;
14084+#endif
14085+}
14086+
1308ab2a 14087+static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
dece6358 14088+{
1308ab2a 14089+#ifdef CONFIG_TMPFS
14090+ return sb->s_magic == TMPFS_MAGIC;
14091+#else
14092+ return 0;
dece6358 14093+#endif
dece6358
AM
14094+}
14095+
1308ab2a 14096+static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
1facf9fc 14097+{
1308ab2a 14098+#if defined(CONFIG_ECRYPT_FS) || defined(CONFIG_ECRYPT_FS_MODULE)
14099+ return !strcmp(au_sbtype(sb), "ecryptfs");
14100+#else
14101+ return 0;
14102+#endif
1facf9fc 14103+}
14104+
1308ab2a 14105+static inline int au_test_ramfs(struct super_block *sb)
14106+{
14107+ return sb->s_magic == RAMFS_MAGIC;
14108+}
14109+
14110+static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
14111+{
14112+#if defined(CONFIG_UBIFS_FS) || defined(CONFIG_UBIFS_FS_MODULE)
14113+ return sb->s_magic == UBIFS_SUPER_MAGIC;
14114+#else
14115+ return 0;
14116+#endif
14117+}
14118+
14119+static inline int au_test_procfs(struct super_block *sb __maybe_unused)
14120+{
14121+#ifdef CONFIG_PROC_FS
14122+ return sb->s_magic == PROC_SUPER_MAGIC;
14123+#else
14124+ return 0;
14125+#endif
14126+}
14127+
14128+static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
14129+{
14130+#ifdef CONFIG_SYSFS
14131+ return sb->s_magic == SYSFS_MAGIC;
14132+#else
14133+ return 0;
14134+#endif
14135+}
14136+
14137+static inline int au_test_configfs(struct super_block *sb __maybe_unused)
14138+{
14139+#if defined(CONFIG_CONFIGFS_FS) || defined(CONFIG_CONFIGFS_FS_MODULE)
14140+ return sb->s_magic == CONFIGFS_MAGIC;
14141+#else
14142+ return 0;
14143+#endif
14144+}
14145+
14146+static inline int au_test_minix(struct super_block *sb __maybe_unused)
14147+{
14148+#if defined(CONFIG_MINIX_FS) || defined(CONFIG_MINIX_FS_MODULE)
14149+ return sb->s_magic == MINIX3_SUPER_MAGIC
14150+ || sb->s_magic == MINIX2_SUPER_MAGIC
14151+ || sb->s_magic == MINIX2_SUPER_MAGIC2
14152+ || sb->s_magic == MINIX_SUPER_MAGIC
14153+ || sb->s_magic == MINIX_SUPER_MAGIC2;
14154+#else
14155+ return 0;
14156+#endif
14157+}
14158+
1308ab2a 14159+static inline int au_test_fat(struct super_block *sb __maybe_unused)
14160+{
14161+#if defined(CONFIG_FAT_FS) || defined(CONFIG_FAT_FS_MODULE)
14162+ return sb->s_magic == MSDOS_SUPER_MAGIC;
14163+#else
14164+ return 0;
14165+#endif
14166+}
14167+
14168+static inline int au_test_msdos(struct super_block *sb)
14169+{
14170+ return au_test_fat(sb);
14171+}
14172+
14173+static inline int au_test_vfat(struct super_block *sb)
14174+{
14175+ return au_test_fat(sb);
14176+}
14177+
14178+static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
14179+{
14180+#ifdef CONFIG_SECURITYFS
14181+ return sb->s_magic == SECURITYFS_MAGIC;
14182+#else
14183+ return 0;
14184+#endif
14185+}
14186+
14187+static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
14188+{
14189+#if defined(CONFIG_SQUASHFS) || defined(CONFIG_SQUASHFS_MODULE)
14190+ return sb->s_magic == SQUASHFS_MAGIC;
14191+#else
14192+ return 0;
14193+#endif
14194+}
14195+
14196+static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
14197+{
14198+#if defined(CONFIG_BTRFS_FS) || defined(CONFIG_BTRFS_FS_MODULE)
14199+ return sb->s_magic == BTRFS_SUPER_MAGIC;
14200+#else
14201+ return 0;
14202+#endif
14203+}
14204+
14205+static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
14206+{
14207+#if defined(CONFIG_XENFS) || defined(CONFIG_XENFS_MODULE)
14208+ return sb->s_magic == XENFS_SUPER_MAGIC;
14209+#else
14210+ return 0;
14211+#endif
14212+}
14213+
14214+static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
14215+{
14216+#ifdef CONFIG_DEBUG_FS
14217+ return sb->s_magic == DEBUGFS_MAGIC;
14218+#else
14219+ return 0;
14220+#endif
14221+}
14222+
14223+static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
14224+{
14225+#if defined(CONFIG_NILFS) || defined(CONFIG_NILFS_MODULE)
14226+ return sb->s_magic == NILFS_SUPER_MAGIC;
14227+#else
14228+ return 0;
14229+#endif
14230+}
14231+
4a4d8108
AM
14232+static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
14233+{
14234+#if defined(CONFIG_HFSPLUS_FS) || defined(CONFIG_HFSPLUS_FS_MODULE)
14235+ return sb->s_magic == HFSPLUS_SUPER_MAGIC;
14236+#else
14237+ return 0;
14238+#endif
14239+}
14240+
1308ab2a 14241+/* ---------------------------------------------------------------------- */
14242+/*
14243+ * they can't be an aufs branch.
14244+ */
14245+static inline int au_test_fs_unsuppoted(struct super_block *sb)
14246+{
14247+ return
14248+#ifndef CONFIG_AUFS_BR_RAMFS
14249+ au_test_ramfs(sb) ||
14250+#endif
14251+ au_test_procfs(sb)
14252+ || au_test_sysfs(sb)
14253+ || au_test_configfs(sb)
14254+ || au_test_debugfs(sb)
14255+ || au_test_securityfs(sb)
14256+ || au_test_xenfs(sb)
14257+ || au_test_ecryptfs(sb)
14258+ /* || !strcmp(au_sbtype(sb), "unionfs") */
14259+ || au_test_aufs(sb); /* will be supported in next version */
14260+}
14261+
1308ab2a 14262+static inline int au_test_fs_remote(struct super_block *sb)
14263+{
14264+ return !au_test_tmpfs(sb)
14265+#ifdef CONFIG_AUFS_BR_RAMFS
14266+ && !au_test_ramfs(sb)
14267+#endif
14268+ && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
14269+}
14270+
14271+/* ---------------------------------------------------------------------- */
14272+
14273+/*
14274+ * Note: these functions (below) are created after reading ->getattr() in all
14275+ * filesystems under linux/fs. it means we have to do so in every update...
14276+ */
14277+
14278+/*
14279+ * some filesystems require getattr to refresh the inode attributes before
14280+ * referencing.
14281+ * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
14282+ * and leave the work for d_revalidate()
14283+ */
14284+static inline int au_test_fs_refresh_iattr(struct super_block *sb)
14285+{
14286+ return au_test_nfs(sb)
14287+ || au_test_fuse(sb)
1308ab2a 14288+ /* || au_test_btrfs(sb) */ /* untested */
1308ab2a 14289+ ;
14290+}
14291+
14292+/*
14293+ * filesystems which don't maintain i_size or i_blocks.
14294+ */
14295+static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
14296+{
14297+ return au_test_xfs(sb)
4a4d8108
AM
14298+ || au_test_btrfs(sb)
14299+ || au_test_ubifs(sb)
14300+ || au_test_hfsplus(sb) /* maintained, but incorrect */
1308ab2a 14301+ /* || au_test_minix(sb) */ /* untested */
14302+ ;
14303+}
14304+
14305+/*
14306+ * filesystems which don't store the correct value in some of their inode
14307+ * attributes.
14308+ */
14309+static inline int au_test_fs_bad_iattr(struct super_block *sb)
14310+{
14311+ return au_test_fs_bad_iattr_size(sb)
1308ab2a 14312+ || au_test_fat(sb)
14313+ || au_test_msdos(sb)
14314+ || au_test_vfat(sb);
1facf9fc 14315+}
14316+
14317+/* they don't check i_nlink in link(2) */
14318+static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
14319+{
14320+ return au_test_tmpfs(sb)
14321+#ifdef CONFIG_AUFS_BR_RAMFS
14322+ || au_test_ramfs(sb)
14323+#endif
4a4d8108 14324+ || au_test_ubifs(sb)
4a4d8108 14325+ || au_test_hfsplus(sb);
1facf9fc 14326+}
14327+
14328+/*
14329+ * filesystems which sets S_NOATIME and S_NOCMTIME.
14330+ */
14331+static inline int au_test_fs_notime(struct super_block *sb)
14332+{
14333+ return au_test_nfs(sb)
14334+ || au_test_fuse(sb)
dece6358 14335+ || au_test_ubifs(sb)
1facf9fc 14336+ ;
14337+}
14338+
1facf9fc 14339+/* temporary support for i#1 in cramfs */
14340+static inline int au_test_fs_unique_ino(struct inode *inode)
14341+{
14342+ if (au_test_cramfs(inode->i_sb))
14343+ return inode->i_ino != 1;
14344+ return 1;
14345+}
14346+
14347+/* ---------------------------------------------------------------------- */
14348+
14349+/*
14350+ * the filesystem where the xino files placed must support i/o after unlink and
14351+ * maintain i_size and i_blocks.
14352+ */
14353+static inline int au_test_fs_bad_xino(struct super_block *sb)
14354+{
14355+ return au_test_fs_remote(sb)
14356+ || au_test_fs_bad_iattr_size(sb)
1facf9fc 14357+ /* don't want unnecessary work for xino */
14358+ || au_test_aufs(sb)
1308ab2a 14359+ || au_test_ecryptfs(sb)
14360+ || au_test_nilfs(sb);
1facf9fc 14361+}
14362+
14363+static inline int au_test_fs_trunc_xino(struct super_block *sb)
14364+{
14365+ return au_test_tmpfs(sb)
14366+ || au_test_ramfs(sb);
14367+}
14368+
14369+/*
14370+ * test if the @sb is real-readonly.
14371+ */
14372+static inline int au_test_fs_rr(struct super_block *sb)
14373+{
14374+ return au_test_squashfs(sb)
14375+ || au_test_iso9660(sb)
14376+ || au_test_cramfs(sb)
14377+ || au_test_romfs(sb);
14378+}
14379+
b912730e
AM
14380+/*
14381+ * test if the @inode is nfs with 'noacl' option
14382+ * NFS always sets MS_POSIXACL regardless its mount option 'noacl.'
14383+ */
14384+static inline int au_test_nfs_noacl(struct inode *inode)
14385+{
14386+ return au_test_nfs(inode->i_sb)
14387+ /* && IS_POSIXACL(inode) */
14388+ && !nfs_server_capable(inode, NFS_CAP_ACLS);
14389+}
14390+
1facf9fc 14391+#endif /* __KERNEL__ */
14392+#endif /* __AUFS_FSTYPE_H__ */
7f207e10
AM
14393diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
14394--- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 14395+++ linux/fs/aufs/hfsnotify.c 2015-09-24 10:47:58.254719746 +0200
c1595e42 14396@@ -0,0 +1,288 @@
1facf9fc 14397+/*
2000de60 14398+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 14399+ *
14400+ * This program, aufs is free software; you can redistribute it and/or modify
14401+ * it under the terms of the GNU General Public License as published by
14402+ * the Free Software Foundation; either version 2 of the License, or
14403+ * (at your option) any later version.
dece6358
AM
14404+ *
14405+ * This program is distributed in the hope that it will be useful,
14406+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14407+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14408+ * GNU General Public License for more details.
14409+ *
14410+ * You should have received a copy of the GNU General Public License
523b37e3 14411+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 14412+ */
14413+
14414+/*
4a4d8108 14415+ * fsnotify for the lower directories
1facf9fc 14416+ */
14417+
14418+#include "aufs.h"
14419+
4a4d8108
AM
14420+/* FS_IN_IGNORED is unnecessary */
14421+static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
14422+ | FS_CREATE | FS_EVENT_ON_CHILD);
7f207e10 14423+static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
7eafdf33 14424+static __cacheline_aligned_in_smp atomic64_t au_hfsn_ifree = ATOMIC64_INIT(0);
1facf9fc 14425+
0c5527e5 14426+static void au_hfsn_free_mark(struct fsnotify_mark *mark)
1facf9fc 14427+{
0c5527e5
AM
14428+ struct au_hnotify *hn = container_of(mark, struct au_hnotify,
14429+ hn_mark);
4a4d8108 14430+ AuDbg("here\n");
7eafdf33 14431+ au_cache_free_hnotify(hn);
076b876e 14432+ smp_mb__before_atomic();
1716fcea
AM
14433+ if (atomic64_dec_and_test(&au_hfsn_ifree))
14434+ wake_up(&au_hfsn_wq);
4a4d8108 14435+}
1facf9fc 14436+
027c5e7a 14437+static int au_hfsn_alloc(struct au_hinode *hinode)
4a4d8108 14438+{
1716fcea 14439+ int err;
027c5e7a
AM
14440+ struct au_hnotify *hn;
14441+ struct super_block *sb;
14442+ struct au_branch *br;
0c5527e5 14443+ struct fsnotify_mark *mark;
027c5e7a 14444+ aufs_bindex_t bindex;
1facf9fc 14445+
027c5e7a
AM
14446+ hn = hinode->hi_notify;
14447+ sb = hn->hn_aufs_inode->i_sb;
14448+ bindex = au_br_index(sb, hinode->hi_id);
14449+ br = au_sbr(sb, bindex);
1716fcea
AM
14450+ AuDebugOn(!br->br_hfsn);
14451+
0c5527e5
AM
14452+ mark = &hn->hn_mark;
14453+ fsnotify_init_mark(mark, au_hfsn_free_mark);
14454+ mark->mask = AuHfsnMask;
7f207e10
AM
14455+ /*
14456+ * by udba rename or rmdir, aufs assign a new inode to the known
14457+ * h_inode, so specify 1 to allow dups.
14458+ */
c1595e42 14459+ lockdep_off();
1716fcea 14460+ err = fsnotify_add_mark(mark, br->br_hfsn->hfsn_group, hinode->hi_inode,
027c5e7a 14461+ /*mnt*/NULL, /*allow_dups*/1);
1716fcea
AM
14462+ /* even if err */
14463+ fsnotify_put_mark(mark);
c1595e42 14464+ lockdep_on();
1716fcea
AM
14465+
14466+ return err;
1facf9fc 14467+}
14468+
7eafdf33 14469+static int au_hfsn_free(struct au_hinode *hinode, struct au_hnotify *hn)
1facf9fc 14470+{
0c5527e5 14471+ struct fsnotify_mark *mark;
7eafdf33 14472+ unsigned long long ull;
1716fcea 14473+ struct fsnotify_group *group;
7eafdf33
AM
14474+
14475+ ull = atomic64_inc_return(&au_hfsn_ifree);
14476+ BUG_ON(!ull);
953406b4 14477+
0c5527e5 14478+ mark = &hn->hn_mark;
1716fcea
AM
14479+ spin_lock(&mark->lock);
14480+ group = mark->group;
14481+ fsnotify_get_group(group);
14482+ spin_unlock(&mark->lock);
c1595e42 14483+ lockdep_off();
1716fcea
AM
14484+ fsnotify_destroy_mark(mark, group);
14485+ fsnotify_put_group(group);
c1595e42 14486+ lockdep_on();
7f207e10 14487+
7eafdf33
AM
14488+ /* free hn by myself */
14489+ return 0;
1facf9fc 14490+}
14491+
14492+/* ---------------------------------------------------------------------- */
14493+
4a4d8108 14494+static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
1facf9fc 14495+{
0c5527e5 14496+ struct fsnotify_mark *mark;
1facf9fc 14497+
0c5527e5
AM
14498+ mark = &hinode->hi_notify->hn_mark;
14499+ spin_lock(&mark->lock);
1facf9fc 14500+ if (do_set) {
0c5527e5
AM
14501+ AuDebugOn(mark->mask & AuHfsnMask);
14502+ mark->mask |= AuHfsnMask;
1facf9fc 14503+ } else {
0c5527e5
AM
14504+ AuDebugOn(!(mark->mask & AuHfsnMask));
14505+ mark->mask &= ~AuHfsnMask;
1facf9fc 14506+ }
0c5527e5 14507+ spin_unlock(&mark->lock);
4a4d8108 14508+ /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
1facf9fc 14509+}
14510+
4a4d8108 14511+/* ---------------------------------------------------------------------- */
1facf9fc 14512+
4a4d8108
AM
14513+/* #define AuDbgHnotify */
14514+#ifdef AuDbgHnotify
14515+static char *au_hfsn_name(u32 mask)
14516+{
14517+#ifdef CONFIG_AUFS_DEBUG
c06a8ce3
AM
14518+#define test_ret(flag) \
14519+ do { \
14520+ if (mask & flag) \
14521+ return #flag; \
14522+ } while (0)
4a4d8108
AM
14523+ test_ret(FS_ACCESS);
14524+ test_ret(FS_MODIFY);
14525+ test_ret(FS_ATTRIB);
14526+ test_ret(FS_CLOSE_WRITE);
14527+ test_ret(FS_CLOSE_NOWRITE);
14528+ test_ret(FS_OPEN);
14529+ test_ret(FS_MOVED_FROM);
14530+ test_ret(FS_MOVED_TO);
14531+ test_ret(FS_CREATE);
14532+ test_ret(FS_DELETE);
14533+ test_ret(FS_DELETE_SELF);
14534+ test_ret(FS_MOVE_SELF);
14535+ test_ret(FS_UNMOUNT);
14536+ test_ret(FS_Q_OVERFLOW);
14537+ test_ret(FS_IN_IGNORED);
b912730e 14538+ test_ret(FS_ISDIR);
4a4d8108
AM
14539+ test_ret(FS_IN_ONESHOT);
14540+ test_ret(FS_EVENT_ON_CHILD);
14541+ return "";
14542+#undef test_ret
14543+#else
14544+ return "??";
14545+#endif
1facf9fc 14546+}
4a4d8108 14547+#endif
1facf9fc 14548+
14549+/* ---------------------------------------------------------------------- */
14550+
1716fcea
AM
14551+static void au_hfsn_free_group(struct fsnotify_group *group)
14552+{
14553+ struct au_br_hfsnotify *hfsn = group->private;
14554+
14555+ AuDbg("here\n");
14556+ kfree(hfsn);
14557+}
14558+
4a4d8108 14559+static int au_hfsn_handle_event(struct fsnotify_group *group,
fb47a38f 14560+ struct inode *inode,
0c5527e5
AM
14561+ struct fsnotify_mark *inode_mark,
14562+ struct fsnotify_mark *vfsmount_mark,
fb47a38f
JR
14563+ u32 mask, void *data, int data_type,
14564+ const unsigned char *file_name, u32 cookie)
1facf9fc 14565+{
14566+ int err;
4a4d8108
AM
14567+ struct au_hnotify *hnotify;
14568+ struct inode *h_dir, *h_inode;
fb47a38f 14569+ struct qstr h_child_qstr = QSTR_INIT(file_name, strlen(file_name));
4a4d8108 14570+
fb47a38f 14571+ AuDebugOn(data_type != FSNOTIFY_EVENT_INODE);
1facf9fc 14572+
14573+ err = 0;
0c5527e5 14574+ /* if FS_UNMOUNT happens, there must be another bug */
4a4d8108 14575+ AuDebugOn(mask & FS_UNMOUNT);
0c5527e5 14576+ if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
1facf9fc 14577+ goto out;
1facf9fc 14578+
fb47a38f
JR
14579+ h_dir = inode;
14580+ h_inode = NULL;
4a4d8108 14581+#ifdef AuDbgHnotify
392086de 14582+ au_debug_on();
4a4d8108
AM
14583+ if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
14584+ || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
14585+ AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
14586+ h_dir->i_ino, mask, au_hfsn_name(mask),
14587+ AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
14588+ /* WARN_ON(1); */
1facf9fc 14589+ }
392086de 14590+ au_debug_off();
1facf9fc 14591+#endif
4a4d8108 14592+
0c5527e5
AM
14593+ AuDebugOn(!inode_mark);
14594+ hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
14595+ err = au_hnotify(h_dir, hnotify, mask, &h_child_qstr, h_inode);
1facf9fc 14596+
4a4d8108
AM
14597+out:
14598+ return err;
14599+}
1facf9fc 14600+
4a4d8108 14601+static struct fsnotify_ops au_hfsn_ops = {
1716fcea
AM
14602+ .handle_event = au_hfsn_handle_event,
14603+ .free_group_priv = au_hfsn_free_group
4a4d8108
AM
14604+};
14605+
14606+/* ---------------------------------------------------------------------- */
14607+
027c5e7a
AM
14608+static void au_hfsn_fin_br(struct au_branch *br)
14609+{
1716fcea 14610+ struct au_br_hfsnotify *hfsn;
027c5e7a 14611+
1716fcea 14612+ hfsn = br->br_hfsn;
c1595e42
JR
14613+ if (hfsn) {
14614+ lockdep_off();
1716fcea 14615+ fsnotify_put_group(hfsn->hfsn_group);
c1595e42
JR
14616+ lockdep_on();
14617+ }
027c5e7a
AM
14618+}
14619+
1716fcea 14620+static int au_hfsn_init_br(struct au_branch *br, int perm)
4a4d8108
AM
14621+{
14622+ int err;
1716fcea
AM
14623+ struct fsnotify_group *group;
14624+ struct au_br_hfsnotify *hfsn;
1facf9fc 14625+
4a4d8108 14626+ err = 0;
1716fcea
AM
14627+ br->br_hfsn = NULL;
14628+ if (!au_br_hnotifyable(perm))
027c5e7a 14629+ goto out;
027c5e7a 14630+
1716fcea
AM
14631+ err = -ENOMEM;
14632+ hfsn = kmalloc(sizeof(*hfsn), GFP_NOFS);
14633+ if (unlikely(!hfsn))
027c5e7a
AM
14634+ goto out;
14635+
1716fcea
AM
14636+ err = 0;
14637+ group = fsnotify_alloc_group(&au_hfsn_ops);
14638+ if (IS_ERR(group)) {
14639+ err = PTR_ERR(group);
0c5527e5 14640+ pr_err("fsnotify_alloc_group() failed, %d\n", err);
1716fcea 14641+ goto out_hfsn;
4a4d8108 14642+ }
1facf9fc 14643+
1716fcea
AM
14644+ group->private = hfsn;
14645+ hfsn->hfsn_group = group;
14646+ br->br_hfsn = hfsn;
14647+ goto out; /* success */
14648+
14649+out_hfsn:
14650+ kfree(hfsn);
027c5e7a 14651+out:
1716fcea
AM
14652+ return err;
14653+}
14654+
14655+static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
14656+{
14657+ int err;
14658+
14659+ err = 0;
14660+ if (!br->br_hfsn)
14661+ err = au_hfsn_init_br(br, perm);
14662+
1facf9fc 14663+ return err;
14664+}
14665+
7eafdf33
AM
14666+/* ---------------------------------------------------------------------- */
14667+
14668+static void au_hfsn_fin(void)
14669+{
14670+ AuDbg("au_hfsn_ifree %lld\n", (long long)atomic64_read(&au_hfsn_ifree));
14671+ wait_event(au_hfsn_wq, !atomic64_read(&au_hfsn_ifree));
14672+}
14673+
4a4d8108
AM
14674+const struct au_hnotify_op au_hnotify_op = {
14675+ .ctl = au_hfsn_ctl,
14676+ .alloc = au_hfsn_alloc,
14677+ .free = au_hfsn_free,
1facf9fc 14678+
7eafdf33
AM
14679+ .fin = au_hfsn_fin,
14680+
027c5e7a
AM
14681+ .reset_br = au_hfsn_reset_br,
14682+ .fin_br = au_hfsn_fin_br,
14683+ .init_br = au_hfsn_init_br
4a4d8108 14684+};
7f207e10
AM
14685diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
14686--- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 14687+++ linux/fs/aufs/hfsplus.c 2015-09-24 10:47:58.254719746 +0200
523b37e3 14688@@ -0,0 +1,56 @@
4a4d8108 14689+/*
2000de60 14690+ * Copyright (C) 2010-2015 Junjiro R. Okajima
4a4d8108
AM
14691+ *
14692+ * This program, aufs is free software; you can redistribute it and/or modify
14693+ * it under the terms of the GNU General Public License as published by
14694+ * the Free Software Foundation; either version 2 of the License, or
14695+ * (at your option) any later version.
14696+ *
14697+ * This program is distributed in the hope that it will be useful,
14698+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14699+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14700+ * GNU General Public License for more details.
14701+ *
14702+ * You should have received a copy of the GNU General Public License
523b37e3 14703+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 14704+ */
1facf9fc 14705+
4a4d8108
AM
14706+/*
14707+ * special support for filesystems which aqucires an inode mutex
14708+ * at final closing a file, eg, hfsplus.
14709+ *
14710+ * This trick is very simple and stupid, just to open the file before really
14711+ * neceeary open to tell hfsplus that this is not the final closing.
14712+ * The caller should call au_h_open_pre() after acquiring the inode mutex,
14713+ * and au_h_open_post() after releasing it.
14714+ */
1facf9fc 14715+
4a4d8108 14716+#include "aufs.h"
1facf9fc 14717+
392086de
AM
14718+struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex,
14719+ int force_wr)
4a4d8108
AM
14720+{
14721+ struct file *h_file;
14722+ struct dentry *h_dentry;
1facf9fc 14723+
4a4d8108
AM
14724+ h_dentry = au_h_dptr(dentry, bindex);
14725+ AuDebugOn(!h_dentry);
5527c038 14726+ AuDebugOn(d_is_negative(h_dentry));
4a4d8108
AM
14727+
14728+ h_file = NULL;
14729+ if (au_test_hfsplus(h_dentry->d_sb)
7e9cd9fe 14730+ && d_is_reg(h_dentry))
4a4d8108
AM
14731+ h_file = au_h_open(dentry, bindex,
14732+ O_RDONLY | O_NOATIME | O_LARGEFILE,
392086de 14733+ /*file*/NULL, force_wr);
4a4d8108 14734+ return h_file;
1facf9fc 14735+}
14736+
4a4d8108
AM
14737+void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
14738+ struct file *h_file)
14739+{
14740+ if (h_file) {
14741+ fput(h_file);
14742+ au_sbr_put(dentry->d_sb, bindex);
14743+ }
14744+}
7f207e10
AM
14745diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
14746--- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 14747+++ linux/fs/aufs/hnotify.c 2015-09-24 10:47:58.254719746 +0200
5527c038 14748@@ -0,0 +1,710 @@
e49829fe 14749+/*
2000de60 14750+ * Copyright (C) 2005-2015 Junjiro R. Okajima
e49829fe
JR
14751+ *
14752+ * This program, aufs is free software; you can redistribute it and/or modify
14753+ * it under the terms of the GNU General Public License as published by
14754+ * the Free Software Foundation; either version 2 of the License, or
14755+ * (at your option) any later version.
14756+ *
14757+ * This program is distributed in the hope that it will be useful,
14758+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14759+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14760+ * GNU General Public License for more details.
14761+ *
14762+ * You should have received a copy of the GNU General Public License
523b37e3 14763+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
e49829fe
JR
14764+ */
14765+
14766+/*
7f207e10 14767+ * abstraction to notify the direct changes on lower directories
e49829fe
JR
14768+ */
14769+
14770+#include "aufs.h"
14771+
027c5e7a 14772+int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
e49829fe
JR
14773+{
14774+ int err;
7f207e10 14775+ struct au_hnotify *hn;
1facf9fc 14776+
4a4d8108
AM
14777+ err = -ENOMEM;
14778+ hn = au_cache_alloc_hnotify();
14779+ if (hn) {
14780+ hn->hn_aufs_inode = inode;
027c5e7a
AM
14781+ hinode->hi_notify = hn;
14782+ err = au_hnotify_op.alloc(hinode);
14783+ AuTraceErr(err);
14784+ if (unlikely(err)) {
14785+ hinode->hi_notify = NULL;
4a4d8108
AM
14786+ au_cache_free_hnotify(hn);
14787+ /*
14788+ * The upper dir was removed by udba, but the same named
14789+ * dir left. In this case, aufs assignes a new inode
14790+ * number and set the monitor again.
14791+ * For the lower dir, the old monitnor is still left.
14792+ */
14793+ if (err == -EEXIST)
14794+ err = 0;
14795+ }
1308ab2a 14796+ }
1308ab2a 14797+
027c5e7a 14798+ AuTraceErr(err);
1308ab2a 14799+ return err;
dece6358 14800+}
1facf9fc 14801+
4a4d8108 14802+void au_hn_free(struct au_hinode *hinode)
dece6358 14803+{
4a4d8108 14804+ struct au_hnotify *hn;
1facf9fc 14805+
4a4d8108
AM
14806+ hn = hinode->hi_notify;
14807+ if (hn) {
4a4d8108 14808+ hinode->hi_notify = NULL;
7eafdf33
AM
14809+ if (au_hnotify_op.free(hinode, hn))
14810+ au_cache_free_hnotify(hn);
4a4d8108
AM
14811+ }
14812+}
dece6358 14813+
4a4d8108 14814+/* ---------------------------------------------------------------------- */
dece6358 14815+
4a4d8108
AM
14816+void au_hn_ctl(struct au_hinode *hinode, int do_set)
14817+{
14818+ if (hinode->hi_notify)
14819+ au_hnotify_op.ctl(hinode, do_set);
14820+}
14821+
14822+void au_hn_reset(struct inode *inode, unsigned int flags)
14823+{
14824+ aufs_bindex_t bindex, bend;
14825+ struct inode *hi;
14826+ struct dentry *iwhdentry;
1facf9fc 14827+
1308ab2a 14828+ bend = au_ibend(inode);
4a4d8108
AM
14829+ for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
14830+ hi = au_h_iptr(inode, bindex);
14831+ if (!hi)
14832+ continue;
1308ab2a 14833+
4a4d8108
AM
14834+ /* mutex_lock_nested(&hi->i_mutex, AuLsc_I_CHILD); */
14835+ iwhdentry = au_hi_wh(inode, bindex);
14836+ if (iwhdentry)
14837+ dget(iwhdentry);
14838+ au_igrab(hi);
14839+ au_set_h_iptr(inode, bindex, NULL, 0);
14840+ au_set_h_iptr(inode, bindex, au_igrab(hi),
14841+ flags & ~AuHi_XINO);
14842+ iput(hi);
14843+ dput(iwhdentry);
14844+ /* mutex_unlock(&hi->i_mutex); */
1facf9fc 14845+ }
1facf9fc 14846+}
14847+
1308ab2a 14848+/* ---------------------------------------------------------------------- */
1facf9fc 14849+
4a4d8108 14850+static int hn_xino(struct inode *inode, struct inode *h_inode)
1facf9fc 14851+{
4a4d8108
AM
14852+ int err;
14853+ aufs_bindex_t bindex, bend, bfound, bstart;
14854+ struct inode *h_i;
1facf9fc 14855+
4a4d8108
AM
14856+ err = 0;
14857+ if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
0c3ec466 14858+ pr_warn("branch root dir was changed\n");
4a4d8108
AM
14859+ goto out;
14860+ }
1facf9fc 14861+
4a4d8108
AM
14862+ bfound = -1;
14863+ bend = au_ibend(inode);
14864+ bstart = au_ibstart(inode);
14865+#if 0 /* reserved for future use */
14866+ if (bindex == bend) {
14867+ /* keep this ino in rename case */
14868+ goto out;
14869+ }
14870+#endif
14871+ for (bindex = bstart; bindex <= bend; bindex++)
14872+ if (au_h_iptr(inode, bindex) == h_inode) {
14873+ bfound = bindex;
14874+ break;
14875+ }
14876+ if (bfound < 0)
1308ab2a 14877+ goto out;
1facf9fc 14878+
4a4d8108
AM
14879+ for (bindex = bstart; bindex <= bend; bindex++) {
14880+ h_i = au_h_iptr(inode, bindex);
14881+ if (!h_i)
14882+ continue;
1facf9fc 14883+
4a4d8108
AM
14884+ err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
14885+ /* ignore this error */
14886+ /* bad action? */
1facf9fc 14887+ }
1facf9fc 14888+
4a4d8108 14889+ /* children inode number will be broken */
1facf9fc 14890+
4f0767ce 14891+out:
4a4d8108
AM
14892+ AuTraceErr(err);
14893+ return err;
1facf9fc 14894+}
14895+
4a4d8108 14896+static int hn_gen_tree(struct dentry *dentry)
1facf9fc 14897+{
4a4d8108
AM
14898+ int err, i, j, ndentry;
14899+ struct au_dcsub_pages dpages;
14900+ struct au_dpage *dpage;
14901+ struct dentry **dentries;
1facf9fc 14902+
4a4d8108
AM
14903+ err = au_dpages_init(&dpages, GFP_NOFS);
14904+ if (unlikely(err))
14905+ goto out;
14906+ err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
14907+ if (unlikely(err))
14908+ goto out_dpages;
1facf9fc 14909+
4a4d8108
AM
14910+ for (i = 0; i < dpages.ndpage; i++) {
14911+ dpage = dpages.dpages + i;
14912+ dentries = dpage->dentries;
14913+ ndentry = dpage->ndentry;
14914+ for (j = 0; j < ndentry; j++) {
14915+ struct dentry *d;
14916+
14917+ d = dentries[j];
14918+ if (IS_ROOT(d))
14919+ continue;
14920+
4a4d8108 14921+ au_digen_dec(d);
5527c038 14922+ if (d_really_is_positive(d))
4a4d8108
AM
14923+ /* todo: reset children xino?
14924+ cached children only? */
5527c038 14925+ au_iigen_dec(d_inode(d));
1308ab2a 14926+ }
dece6358 14927+ }
1facf9fc 14928+
4f0767ce 14929+out_dpages:
4a4d8108 14930+ au_dpages_free(&dpages);
dece6358 14931+
027c5e7a 14932+#if 0
4a4d8108
AM
14933+ /* discard children */
14934+ dentry_unhash(dentry);
14935+ dput(dentry);
027c5e7a 14936+#endif
4f0767ce 14937+out:
dece6358
AM
14938+ return err;
14939+}
14940+
1308ab2a 14941+/*
4a4d8108 14942+ * return 0 if processed.
1308ab2a 14943+ */
4a4d8108
AM
14944+static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
14945+ const unsigned int isdir)
dece6358 14946+{
1308ab2a 14947+ int err;
4a4d8108
AM
14948+ struct dentry *d;
14949+ struct qstr *dname;
1facf9fc 14950+
4a4d8108
AM
14951+ err = 1;
14952+ if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
0c3ec466 14953+ pr_warn("branch root dir was changed\n");
4a4d8108
AM
14954+ err = 0;
14955+ goto out;
14956+ }
dece6358 14957+
4a4d8108
AM
14958+ if (!isdir) {
14959+ AuDebugOn(!name);
14960+ au_iigen_dec(inode);
027c5e7a 14961+ spin_lock(&inode->i_lock);
c1595e42 14962+ hlist_for_each_entry(d, &inode->i_dentry, d_u.d_alias) {
027c5e7a 14963+ spin_lock(&d->d_lock);
4a4d8108
AM
14964+ dname = &d->d_name;
14965+ if (dname->len != nlen
027c5e7a
AM
14966+ && memcmp(dname->name, name, nlen)) {
14967+ spin_unlock(&d->d_lock);
4a4d8108 14968+ continue;
027c5e7a 14969+ }
4a4d8108 14970+ err = 0;
4a4d8108
AM
14971+ au_digen_dec(d);
14972+ spin_unlock(&d->d_lock);
14973+ break;
1facf9fc 14974+ }
027c5e7a 14975+ spin_unlock(&inode->i_lock);
1308ab2a 14976+ } else {
027c5e7a 14977+ au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
c1595e42 14978+ d = d_find_any_alias(inode);
4a4d8108
AM
14979+ if (!d) {
14980+ au_iigen_dec(inode);
14981+ goto out;
14982+ }
1facf9fc 14983+
027c5e7a 14984+ spin_lock(&d->d_lock);
4a4d8108 14985+ dname = &d->d_name;
027c5e7a
AM
14986+ if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
14987+ spin_unlock(&d->d_lock);
4a4d8108 14988+ err = hn_gen_tree(d);
027c5e7a
AM
14989+ spin_lock(&d->d_lock);
14990+ }
14991+ spin_unlock(&d->d_lock);
4a4d8108
AM
14992+ dput(d);
14993+ }
1facf9fc 14994+
4f0767ce 14995+out:
4a4d8108 14996+ AuTraceErr(err);
1308ab2a 14997+ return err;
14998+}
dece6358 14999+
4a4d8108 15000+static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
1facf9fc 15001+{
4a4d8108 15002+ int err;
1facf9fc 15003+
5527c038 15004+ if (IS_ROOT(dentry)) {
0c3ec466 15005+ pr_warn("branch root dir was changed\n");
4a4d8108
AM
15006+ return 0;
15007+ }
1308ab2a 15008+
4a4d8108
AM
15009+ err = 0;
15010+ if (!isdir) {
4a4d8108 15011+ au_digen_dec(dentry);
5527c038
JR
15012+ if (d_really_is_positive(dentry))
15013+ au_iigen_dec(d_inode(dentry));
4a4d8108 15014+ } else {
027c5e7a 15015+ au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
5527c038 15016+ if (d_really_is_positive(dentry))
4a4d8108
AM
15017+ err = hn_gen_tree(dentry);
15018+ }
15019+
15020+ AuTraceErr(err);
15021+ return err;
1facf9fc 15022+}
15023+
4a4d8108 15024+/* ---------------------------------------------------------------------- */
1facf9fc 15025+
4a4d8108
AM
15026+/* hnotify job flags */
15027+#define AuHnJob_XINO0 1
15028+#define AuHnJob_GEN (1 << 1)
15029+#define AuHnJob_DIRENT (1 << 2)
15030+#define AuHnJob_ISDIR (1 << 3)
15031+#define AuHnJob_TRYXINO0 (1 << 4)
15032+#define AuHnJob_MNTPNT (1 << 5)
15033+#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
7f207e10
AM
15034+#define au_fset_hnjob(flags, name) \
15035+ do { (flags) |= AuHnJob_##name; } while (0)
15036+#define au_fclr_hnjob(flags, name) \
15037+ do { (flags) &= ~AuHnJob_##name; } while (0)
1facf9fc 15038+
4a4d8108
AM
15039+enum {
15040+ AuHn_CHILD,
15041+ AuHn_PARENT,
15042+ AuHnLast
15043+};
1facf9fc 15044+
4a4d8108
AM
15045+struct au_hnotify_args {
15046+ struct inode *h_dir, *dir, *h_child_inode;
15047+ u32 mask;
15048+ unsigned int flags[AuHnLast];
15049+ unsigned int h_child_nlen;
15050+ char h_child_name[];
15051+};
1facf9fc 15052+
4a4d8108
AM
15053+struct hn_job_args {
15054+ unsigned int flags;
15055+ struct inode *inode, *h_inode, *dir, *h_dir;
15056+ struct dentry *dentry;
15057+ char *h_name;
15058+ int h_nlen;
15059+};
1308ab2a 15060+
4a4d8108
AM
15061+static int hn_job(struct hn_job_args *a)
15062+{
15063+ const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
076b876e 15064+ int e;
1308ab2a 15065+
4a4d8108
AM
15066+ /* reset xino */
15067+ if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
15068+ hn_xino(a->inode, a->h_inode); /* ignore this error */
1308ab2a 15069+
4a4d8108
AM
15070+ if (au_ftest_hnjob(a->flags, TRYXINO0)
15071+ && a->inode
15072+ && a->h_inode) {
15073+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
38d290e6
JR
15074+ if (!a->h_inode->i_nlink
15075+ && !(a->h_inode->i_state & I_LINKABLE))
4a4d8108
AM
15076+ hn_xino(a->inode, a->h_inode); /* ignore this error */
15077+ mutex_unlock(&a->h_inode->i_mutex);
1308ab2a 15078+ }
1facf9fc 15079+
4a4d8108
AM
15080+ /* make the generation obsolete */
15081+ if (au_ftest_hnjob(a->flags, GEN)) {
076b876e 15082+ e = -1;
4a4d8108 15083+ if (a->inode)
076b876e 15084+ e = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
4a4d8108 15085+ isdir);
076b876e 15086+ if (e && a->dentry)
4a4d8108
AM
15087+ hn_gen_by_name(a->dentry, isdir);
15088+ /* ignore this error */
1facf9fc 15089+ }
1facf9fc 15090+
4a4d8108
AM
15091+ /* make dir entries obsolete */
15092+ if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
15093+ struct au_vdir *vdir;
1facf9fc 15094+
4a4d8108
AM
15095+ vdir = au_ivdir(a->inode);
15096+ if (vdir)
15097+ vdir->vd_jiffy = 0;
15098+ /* IMustLock(a->inode); */
15099+ /* a->inode->i_version++; */
15100+ }
1facf9fc 15101+
4a4d8108
AM
15102+ /* can do nothing but warn */
15103+ if (au_ftest_hnjob(a->flags, MNTPNT)
15104+ && a->dentry
15105+ && d_mountpoint(a->dentry))
523b37e3 15106+ pr_warn("mount-point %pd is removed or renamed\n", a->dentry);
1facf9fc 15107+
4a4d8108 15108+ return 0;
1308ab2a 15109+}
1facf9fc 15110+
1308ab2a 15111+/* ---------------------------------------------------------------------- */
1facf9fc 15112+
4a4d8108
AM
15113+static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
15114+ struct inode *dir)
1308ab2a 15115+{
4a4d8108
AM
15116+ struct dentry *dentry, *d, *parent;
15117+ struct qstr *dname;
1308ab2a 15118+
c1595e42 15119+ parent = d_find_any_alias(dir);
4a4d8108
AM
15120+ if (!parent)
15121+ return NULL;
1308ab2a 15122+
4a4d8108 15123+ dentry = NULL;
027c5e7a 15124+ spin_lock(&parent->d_lock);
c1595e42 15125+ list_for_each_entry(d, &parent->d_subdirs, d_child) {
523b37e3 15126+ /* AuDbg("%pd\n", d); */
027c5e7a 15127+ spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
4a4d8108
AM
15128+ dname = &d->d_name;
15129+ if (dname->len != nlen || memcmp(dname->name, name, nlen))
027c5e7a
AM
15130+ goto cont_unlock;
15131+ if (au_di(d))
15132+ au_digen_dec(d);
15133+ else
15134+ goto cont_unlock;
c1595e42 15135+ if (au_dcount(d) > 0) {
027c5e7a 15136+ dentry = dget_dlock(d);
4a4d8108 15137+ spin_unlock(&d->d_lock);
027c5e7a 15138+ break;
dece6358 15139+ }
1facf9fc 15140+
f6b6e03d 15141+cont_unlock:
027c5e7a 15142+ spin_unlock(&d->d_lock);
1308ab2a 15143+ }
027c5e7a 15144+ spin_unlock(&parent->d_lock);
4a4d8108 15145+ dput(parent);
1facf9fc 15146+
4a4d8108
AM
15147+ if (dentry)
15148+ di_write_lock_child(dentry);
1308ab2a 15149+
4a4d8108
AM
15150+ return dentry;
15151+}
dece6358 15152+
4a4d8108
AM
15153+static struct inode *lookup_wlock_by_ino(struct super_block *sb,
15154+ aufs_bindex_t bindex, ino_t h_ino)
15155+{
15156+ struct inode *inode;
15157+ ino_t ino;
15158+ int err;
15159+
15160+ inode = NULL;
15161+ err = au_xino_read(sb, bindex, h_ino, &ino);
15162+ if (!err && ino)
15163+ inode = ilookup(sb, ino);
15164+ if (!inode)
15165+ goto out;
15166+
15167+ if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
0c3ec466 15168+ pr_warn("wrong root branch\n");
4a4d8108
AM
15169+ iput(inode);
15170+ inode = NULL;
15171+ goto out;
1308ab2a 15172+ }
15173+
4a4d8108 15174+ ii_write_lock_child(inode);
1308ab2a 15175+
4f0767ce 15176+out:
4a4d8108 15177+ return inode;
dece6358
AM
15178+}
15179+
4a4d8108 15180+static void au_hn_bh(void *_args)
1facf9fc 15181+{
4a4d8108
AM
15182+ struct au_hnotify_args *a = _args;
15183+ struct super_block *sb;
15184+ aufs_bindex_t bindex, bend, bfound;
15185+ unsigned char xino, try_iput;
1facf9fc 15186+ int err;
1308ab2a 15187+ struct inode *inode;
4a4d8108
AM
15188+ ino_t h_ino;
15189+ struct hn_job_args args;
15190+ struct dentry *dentry;
15191+ struct au_sbinfo *sbinfo;
1facf9fc 15192+
4a4d8108
AM
15193+ AuDebugOn(!_args);
15194+ AuDebugOn(!a->h_dir);
15195+ AuDebugOn(!a->dir);
15196+ AuDebugOn(!a->mask);
15197+ AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
15198+ a->mask, a->dir->i_ino, a->h_dir->i_ino,
15199+ a->h_child_inode ? a->h_child_inode->i_ino : 0);
1facf9fc 15200+
4a4d8108
AM
15201+ inode = NULL;
15202+ dentry = NULL;
15203+ /*
15204+ * do not lock a->dir->i_mutex here
15205+ * because of d_revalidate() may cause a deadlock.
15206+ */
15207+ sb = a->dir->i_sb;
15208+ AuDebugOn(!sb);
15209+ sbinfo = au_sbi(sb);
15210+ AuDebugOn(!sbinfo);
7f207e10 15211+ si_write_lock(sb, AuLock_NOPLMW);
1facf9fc 15212+
4a4d8108
AM
15213+ ii_read_lock_parent(a->dir);
15214+ bfound = -1;
15215+ bend = au_ibend(a->dir);
15216+ for (bindex = au_ibstart(a->dir); bindex <= bend; bindex++)
15217+ if (au_h_iptr(a->dir, bindex) == a->h_dir) {
15218+ bfound = bindex;
15219+ break;
15220+ }
15221+ ii_read_unlock(a->dir);
15222+ if (unlikely(bfound < 0))
15223+ goto out;
1facf9fc 15224+
4a4d8108
AM
15225+ xino = !!au_opt_test(au_mntflags(sb), XINO);
15226+ h_ino = 0;
15227+ if (a->h_child_inode)
15228+ h_ino = a->h_child_inode->i_ino;
1facf9fc 15229+
4a4d8108
AM
15230+ if (a->h_child_nlen
15231+ && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
15232+ || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
15233+ dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
15234+ a->dir);
15235+ try_iput = 0;
5527c038
JR
15236+ if (dentry && d_really_is_positive(dentry))
15237+ inode = d_inode(dentry);
4a4d8108
AM
15238+ if (xino && !inode && h_ino
15239+ && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
15240+ || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
15241+ || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
15242+ inode = lookup_wlock_by_ino(sb, bfound, h_ino);
15243+ try_iput = 1;
15244+ }
1facf9fc 15245+
4a4d8108
AM
15246+ args.flags = a->flags[AuHn_CHILD];
15247+ args.dentry = dentry;
15248+ args.inode = inode;
15249+ args.h_inode = a->h_child_inode;
15250+ args.dir = a->dir;
15251+ args.h_dir = a->h_dir;
15252+ args.h_name = a->h_child_name;
15253+ args.h_nlen = a->h_child_nlen;
15254+ err = hn_job(&args);
15255+ if (dentry) {
027c5e7a 15256+ if (au_di(dentry))
4a4d8108
AM
15257+ di_write_unlock(dentry);
15258+ dput(dentry);
15259+ }
15260+ if (inode && try_iput) {
15261+ ii_write_unlock(inode);
15262+ iput(inode);
15263+ }
1facf9fc 15264+
4a4d8108
AM
15265+ ii_write_lock_parent(a->dir);
15266+ args.flags = a->flags[AuHn_PARENT];
15267+ args.dentry = NULL;
15268+ args.inode = a->dir;
15269+ args.h_inode = a->h_dir;
15270+ args.dir = NULL;
15271+ args.h_dir = NULL;
15272+ args.h_name = NULL;
15273+ args.h_nlen = 0;
15274+ err = hn_job(&args);
15275+ ii_write_unlock(a->dir);
1facf9fc 15276+
4f0767ce 15277+out:
4a4d8108
AM
15278+ iput(a->h_child_inode);
15279+ iput(a->h_dir);
15280+ iput(a->dir);
027c5e7a
AM
15281+ si_write_unlock(sb);
15282+ au_nwt_done(&sbinfo->si_nowait);
1308ab2a 15283+ kfree(a);
dece6358 15284+}
1facf9fc 15285+
4a4d8108
AM
15286+/* ---------------------------------------------------------------------- */
15287+
15288+int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
15289+ struct qstr *h_child_qstr, struct inode *h_child_inode)
dece6358 15290+{
4a4d8108 15291+ int err, len;
53392da6 15292+ unsigned int flags[AuHnLast], f;
4a4d8108
AM
15293+ unsigned char isdir, isroot, wh;
15294+ struct inode *dir;
15295+ struct au_hnotify_args *args;
15296+ char *p, *h_child_name;
dece6358 15297+
1308ab2a 15298+ err = 0;
4a4d8108
AM
15299+ AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
15300+ dir = igrab(hnotify->hn_aufs_inode);
15301+ if (!dir)
15302+ goto out;
1facf9fc 15303+
4a4d8108
AM
15304+ isroot = (dir->i_ino == AUFS_ROOT_INO);
15305+ wh = 0;
15306+ h_child_name = (void *)h_child_qstr->name;
15307+ len = h_child_qstr->len;
15308+ if (h_child_name) {
15309+ if (len > AUFS_WH_PFX_LEN
15310+ && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
15311+ h_child_name += AUFS_WH_PFX_LEN;
15312+ len -= AUFS_WH_PFX_LEN;
15313+ wh = 1;
15314+ }
1facf9fc 15315+ }
dece6358 15316+
4a4d8108
AM
15317+ isdir = 0;
15318+ if (h_child_inode)
15319+ isdir = !!S_ISDIR(h_child_inode->i_mode);
15320+ flags[AuHn_PARENT] = AuHnJob_ISDIR;
15321+ flags[AuHn_CHILD] = 0;
15322+ if (isdir)
15323+ flags[AuHn_CHILD] = AuHnJob_ISDIR;
15324+ au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
15325+ au_fset_hnjob(flags[AuHn_CHILD], GEN);
15326+ switch (mask & FS_EVENTS_POSS_ON_CHILD) {
15327+ case FS_MOVED_FROM:
15328+ case FS_MOVED_TO:
15329+ au_fset_hnjob(flags[AuHn_CHILD], XINO0);
15330+ au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
15331+ /*FALLTHROUGH*/
15332+ case FS_CREATE:
fb47a38f 15333+ AuDebugOn(!h_child_name);
4a4d8108 15334+ break;
1facf9fc 15335+
4a4d8108
AM
15336+ case FS_DELETE:
15337+ /*
15338+ * aufs never be able to get this child inode.
15339+ * revalidation should be in d_revalidate()
15340+ * by checking i_nlink, i_generation or d_unhashed().
15341+ */
15342+ AuDebugOn(!h_child_name);
15343+ au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
15344+ au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
15345+ break;
dece6358 15346+
4a4d8108
AM
15347+ default:
15348+ AuDebugOn(1);
15349+ }
1308ab2a 15350+
4a4d8108
AM
15351+ if (wh)
15352+ h_child_inode = NULL;
1308ab2a 15353+
4a4d8108
AM
15354+ err = -ENOMEM;
15355+ /* iput() and kfree() will be called in au_hnotify() */
4a4d8108 15356+ args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
4a4d8108
AM
15357+ if (unlikely(!args)) {
15358+ AuErr1("no memory\n");
15359+ iput(dir);
15360+ goto out;
15361+ }
15362+ args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
15363+ args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
15364+ args->mask = mask;
15365+ args->dir = dir;
15366+ args->h_dir = igrab(h_dir);
15367+ if (h_child_inode)
15368+ h_child_inode = igrab(h_child_inode); /* can be NULL */
15369+ args->h_child_inode = h_child_inode;
15370+ args->h_child_nlen = len;
15371+ if (len) {
15372+ p = (void *)args;
15373+ p += sizeof(*args);
15374+ memcpy(p, h_child_name, len);
15375+ p[len] = 0;
1308ab2a 15376+ }
1308ab2a 15377+
38d290e6 15378+ /* NFS fires the event for silly-renamed one from kworker */
53392da6 15379+ f = 0;
38d290e6
JR
15380+ if (!dir->i_nlink
15381+ || (au_test_nfs(h_dir->i_sb) && (mask & FS_DELETE)))
53392da6
AM
15382+ f = AuWkq_NEST;
15383+ err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
4a4d8108
AM
15384+ if (unlikely(err)) {
15385+ pr_err("wkq %d\n", err);
15386+ iput(args->h_child_inode);
15387+ iput(args->h_dir);
15388+ iput(args->dir);
15389+ kfree(args);
1facf9fc 15390+ }
1facf9fc 15391+
4a4d8108 15392+out:
1facf9fc 15393+ return err;
15394+}
15395+
027c5e7a
AM
15396+/* ---------------------------------------------------------------------- */
15397+
15398+int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
15399+{
15400+ int err;
15401+
15402+ AuDebugOn(!(udba & AuOptMask_UDBA));
15403+
15404+ err = 0;
15405+ if (au_hnotify_op.reset_br)
15406+ err = au_hnotify_op.reset_br(udba, br, perm);
15407+
15408+ return err;
15409+}
15410+
15411+int au_hnotify_init_br(struct au_branch *br, int perm)
15412+{
15413+ int err;
15414+
15415+ err = 0;
15416+ if (au_hnotify_op.init_br)
15417+ err = au_hnotify_op.init_br(br, perm);
15418+
15419+ return err;
15420+}
15421+
15422+void au_hnotify_fin_br(struct au_branch *br)
15423+{
15424+ if (au_hnotify_op.fin_br)
15425+ au_hnotify_op.fin_br(br);
15426+}
15427+
4a4d8108
AM
15428+static void au_hn_destroy_cache(void)
15429+{
15430+ kmem_cache_destroy(au_cachep[AuCache_HNOTIFY]);
15431+ au_cachep[AuCache_HNOTIFY] = NULL;
15432+}
1308ab2a 15433+
4a4d8108 15434+int __init au_hnotify_init(void)
1facf9fc 15435+{
1308ab2a 15436+ int err;
1308ab2a 15437+
4a4d8108
AM
15438+ err = -ENOMEM;
15439+ au_cachep[AuCache_HNOTIFY] = AuCache(au_hnotify);
15440+ if (au_cachep[AuCache_HNOTIFY]) {
027c5e7a
AM
15441+ err = 0;
15442+ if (au_hnotify_op.init)
15443+ err = au_hnotify_op.init();
4a4d8108
AM
15444+ if (unlikely(err))
15445+ au_hn_destroy_cache();
1308ab2a 15446+ }
1308ab2a 15447+ AuTraceErr(err);
4a4d8108 15448+ return err;
1308ab2a 15449+}
15450+
4a4d8108 15451+void au_hnotify_fin(void)
1308ab2a 15452+{
027c5e7a
AM
15453+ if (au_hnotify_op.fin)
15454+ au_hnotify_op.fin();
4a4d8108
AM
15455+ /* cf. au_cache_fin() */
15456+ if (au_cachep[AuCache_HNOTIFY])
15457+ au_hn_destroy_cache();
dece6358 15458+}
7f207e10
AM
15459diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
15460--- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 15461+++ linux/fs/aufs/iinfo.c 2015-09-24 10:47:58.254719746 +0200
38d290e6 15462@@ -0,0 +1,277 @@
dece6358 15463+/*
2000de60 15464+ * Copyright (C) 2005-2015 Junjiro R. Okajima
dece6358
AM
15465+ *
15466+ * This program, aufs is free software; you can redistribute it and/or modify
15467+ * it under the terms of the GNU General Public License as published by
15468+ * the Free Software Foundation; either version 2 of the License, or
15469+ * (at your option) any later version.
15470+ *
15471+ * This program is distributed in the hope that it will be useful,
15472+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
15473+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15474+ * GNU General Public License for more details.
15475+ *
15476+ * You should have received a copy of the GNU General Public License
523b37e3 15477+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
dece6358 15478+ */
1facf9fc 15479+
dece6358 15480+/*
4a4d8108 15481+ * inode private data
dece6358 15482+ */
1facf9fc 15483+
1308ab2a 15484+#include "aufs.h"
1facf9fc 15485+
4a4d8108 15486+struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
1308ab2a 15487+{
4a4d8108 15488+ struct inode *h_inode;
1facf9fc 15489+
4a4d8108 15490+ IiMustAnyLock(inode);
1facf9fc 15491+
4a4d8108
AM
15492+ h_inode = au_ii(inode)->ii_hinode[0 + bindex].hi_inode;
15493+ AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
15494+ return h_inode;
15495+}
1facf9fc 15496+
4a4d8108
AM
15497+/* todo: hard/soft set? */
15498+void au_hiput(struct au_hinode *hinode)
15499+{
15500+ au_hn_free(hinode);
15501+ dput(hinode->hi_whdentry);
15502+ iput(hinode->hi_inode);
15503+}
1facf9fc 15504+
4a4d8108
AM
15505+unsigned int au_hi_flags(struct inode *inode, int isdir)
15506+{
15507+ unsigned int flags;
15508+ const unsigned int mnt_flags = au_mntflags(inode->i_sb);
1facf9fc 15509+
4a4d8108
AM
15510+ flags = 0;
15511+ if (au_opt_test(mnt_flags, XINO))
15512+ au_fset_hi(flags, XINO);
15513+ if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
15514+ au_fset_hi(flags, HNOTIFY);
15515+ return flags;
1facf9fc 15516+}
15517+
4a4d8108
AM
15518+void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
15519+ struct inode *h_inode, unsigned int flags)
1308ab2a 15520+{
4a4d8108
AM
15521+ struct au_hinode *hinode;
15522+ struct inode *hi;
15523+ struct au_iinfo *iinfo = au_ii(inode);
1facf9fc 15524+
4a4d8108 15525+ IiMustWriteLock(inode);
dece6358 15526+
4a4d8108
AM
15527+ hinode = iinfo->ii_hinode + bindex;
15528+ hi = hinode->hi_inode;
15529+ AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
15530+
15531+ if (hi)
15532+ au_hiput(hinode);
15533+ hinode->hi_inode = h_inode;
15534+ if (h_inode) {
15535+ int err;
15536+ struct super_block *sb = inode->i_sb;
15537+ struct au_branch *br;
15538+
027c5e7a
AM
15539+ AuDebugOn(inode->i_mode
15540+ && (h_inode->i_mode & S_IFMT)
15541+ != (inode->i_mode & S_IFMT));
4a4d8108
AM
15542+ if (bindex == iinfo->ii_bstart)
15543+ au_cpup_igen(inode, h_inode);
15544+ br = au_sbr(sb, bindex);
15545+ hinode->hi_id = br->br_id;
15546+ if (au_ftest_hi(flags, XINO)) {
15547+ err = au_xino_write(sb, bindex, h_inode->i_ino,
15548+ inode->i_ino);
15549+ if (unlikely(err))
15550+ AuIOErr1("failed au_xino_write() %d\n", err);
15551+ }
15552+
15553+ if (au_ftest_hi(flags, HNOTIFY)
15554+ && au_br_hnotifyable(br->br_perm)) {
027c5e7a 15555+ err = au_hn_alloc(hinode, inode);
4a4d8108
AM
15556+ if (unlikely(err))
15557+ AuIOErr1("au_hn_alloc() %d\n", err);
1308ab2a 15558+ }
15559+ }
4a4d8108 15560+}
dece6358 15561+
4a4d8108
AM
15562+void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
15563+ struct dentry *h_wh)
15564+{
15565+ struct au_hinode *hinode;
dece6358 15566+
4a4d8108
AM
15567+ IiMustWriteLock(inode);
15568+
15569+ hinode = au_ii(inode)->ii_hinode + bindex;
15570+ AuDebugOn(hinode->hi_whdentry);
15571+ hinode->hi_whdentry = h_wh;
1facf9fc 15572+}
15573+
537831f9 15574+void au_update_iigen(struct inode *inode, int half)
1308ab2a 15575+{
537831f9
AM
15576+ struct au_iinfo *iinfo;
15577+ struct au_iigen *iigen;
15578+ unsigned int sigen;
15579+
15580+ sigen = au_sigen(inode->i_sb);
15581+ iinfo = au_ii(inode);
15582+ iigen = &iinfo->ii_generation;
15583+ spin_lock(&iinfo->ii_genspin);
15584+ iigen->ig_generation = sigen;
15585+ if (half)
15586+ au_ig_fset(iigen->ig_flags, HALF_REFRESHED);
15587+ else
15588+ au_ig_fclr(iigen->ig_flags, HALF_REFRESHED);
15589+ spin_unlock(&iinfo->ii_genspin);
4a4d8108 15590+}
1facf9fc 15591+
4a4d8108
AM
15592+/* it may be called at remount time, too */
15593+void au_update_ibrange(struct inode *inode, int do_put_zero)
15594+{
15595+ struct au_iinfo *iinfo;
027c5e7a 15596+ aufs_bindex_t bindex, bend;
1facf9fc 15597+
4a4d8108 15598+ iinfo = au_ii(inode);
027c5e7a 15599+ if (!iinfo)
4a4d8108 15600+ return;
1facf9fc 15601+
4a4d8108 15602+ IiMustWriteLock(inode);
1facf9fc 15603+
027c5e7a 15604+ if (do_put_zero && iinfo->ii_bstart >= 0) {
4a4d8108
AM
15605+ for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
15606+ bindex++) {
15607+ struct inode *h_i;
1facf9fc 15608+
4a4d8108 15609+ h_i = iinfo->ii_hinode[0 + bindex].hi_inode;
38d290e6
JR
15610+ if (h_i
15611+ && !h_i->i_nlink
15612+ && !(h_i->i_state & I_LINKABLE))
027c5e7a
AM
15613+ au_set_h_iptr(inode, bindex, NULL, 0);
15614+ }
4a4d8108
AM
15615+ }
15616+
027c5e7a
AM
15617+ iinfo->ii_bstart = -1;
15618+ iinfo->ii_bend = -1;
15619+ bend = au_sbend(inode->i_sb);
15620+ for (bindex = 0; bindex <= bend; bindex++)
15621+ if (iinfo->ii_hinode[0 + bindex].hi_inode) {
15622+ iinfo->ii_bstart = bindex;
4a4d8108 15623+ break;
027c5e7a
AM
15624+ }
15625+ if (iinfo->ii_bstart >= 0)
15626+ for (bindex = bend; bindex >= iinfo->ii_bstart; bindex--)
15627+ if (iinfo->ii_hinode[0 + bindex].hi_inode) {
15628+ iinfo->ii_bend = bindex;
15629+ break;
15630+ }
15631+ AuDebugOn(iinfo->ii_bstart > iinfo->ii_bend);
1308ab2a 15632+}
1facf9fc 15633+
dece6358 15634+/* ---------------------------------------------------------------------- */
1facf9fc 15635+
4a4d8108 15636+void au_icntnr_init_once(void *_c)
dece6358 15637+{
4a4d8108
AM
15638+ struct au_icntnr *c = _c;
15639+ struct au_iinfo *iinfo = &c->iinfo;
e49829fe 15640+ static struct lock_class_key aufs_ii;
1facf9fc 15641+
537831f9 15642+ spin_lock_init(&iinfo->ii_genspin);
4a4d8108 15643+ au_rw_init(&iinfo->ii_rwsem);
e49829fe 15644+ au_rw_class(&iinfo->ii_rwsem, &aufs_ii);
4a4d8108
AM
15645+ inode_init_once(&c->vfs_inode);
15646+}
1facf9fc 15647+
4a4d8108
AM
15648+int au_iinfo_init(struct inode *inode)
15649+{
15650+ struct au_iinfo *iinfo;
15651+ struct super_block *sb;
15652+ int nbr, i;
1facf9fc 15653+
4a4d8108
AM
15654+ sb = inode->i_sb;
15655+ iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
15656+ nbr = au_sbend(sb) + 1;
15657+ if (unlikely(nbr <= 0))
15658+ nbr = 1;
15659+ iinfo->ii_hinode = kcalloc(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
15660+ if (iinfo->ii_hinode) {
7f207e10 15661+ au_ninodes_inc(sb);
4a4d8108
AM
15662+ for (i = 0; i < nbr; i++)
15663+ iinfo->ii_hinode[i].hi_id = -1;
1facf9fc 15664+
537831f9 15665+ iinfo->ii_generation.ig_generation = au_sigen(sb);
4a4d8108
AM
15666+ iinfo->ii_bstart = -1;
15667+ iinfo->ii_bend = -1;
15668+ iinfo->ii_vdir = NULL;
15669+ return 0;
1308ab2a 15670+ }
4a4d8108
AM
15671+ return -ENOMEM;
15672+}
1facf9fc 15673+
4a4d8108
AM
15674+int au_ii_realloc(struct au_iinfo *iinfo, int nbr)
15675+{
15676+ int err, sz;
15677+ struct au_hinode *hip;
1facf9fc 15678+
4a4d8108
AM
15679+ AuRwMustWriteLock(&iinfo->ii_rwsem);
15680+
15681+ err = -ENOMEM;
15682+ sz = sizeof(*hip) * (iinfo->ii_bend + 1);
15683+ if (!sz)
15684+ sz = sizeof(*hip);
15685+ hip = au_kzrealloc(iinfo->ii_hinode, sz, sizeof(*hip) * nbr, GFP_NOFS);
15686+ if (hip) {
15687+ iinfo->ii_hinode = hip;
15688+ err = 0;
1308ab2a 15689+ }
4a4d8108 15690+
1308ab2a 15691+ return err;
1facf9fc 15692+}
15693+
4a4d8108 15694+void au_iinfo_fin(struct inode *inode)
1facf9fc 15695+{
4a4d8108
AM
15696+ struct au_iinfo *iinfo;
15697+ struct au_hinode *hi;
15698+ struct super_block *sb;
b752ccd1
AM
15699+ aufs_bindex_t bindex, bend;
15700+ const unsigned char unlinked = !inode->i_nlink;
1308ab2a 15701+
4a4d8108
AM
15702+ iinfo = au_ii(inode);
15703+ /* bad_inode case */
15704+ if (!iinfo)
15705+ return;
1308ab2a 15706+
b752ccd1 15707+ sb = inode->i_sb;
7f207e10 15708+ au_ninodes_dec(sb);
b752ccd1
AM
15709+ if (si_pid_test(sb))
15710+ au_xino_delete_inode(inode, unlinked);
15711+ else {
15712+ /*
15713+ * it is safe to hide the dependency between sbinfo and
15714+ * sb->s_umount.
15715+ */
15716+ lockdep_off();
15717+ si_noflush_read_lock(sb);
15718+ au_xino_delete_inode(inode, unlinked);
15719+ si_read_unlock(sb);
15720+ lockdep_on();
15721+ }
15722+
4a4d8108
AM
15723+ if (iinfo->ii_vdir)
15724+ au_vdir_free(iinfo->ii_vdir);
1308ab2a 15725+
b752ccd1
AM
15726+ bindex = iinfo->ii_bstart;
15727+ if (bindex >= 0) {
15728+ hi = iinfo->ii_hinode + bindex;
4a4d8108 15729+ bend = iinfo->ii_bend;
b752ccd1
AM
15730+ while (bindex++ <= bend) {
15731+ if (hi->hi_inode)
4a4d8108 15732+ au_hiput(hi);
4a4d8108
AM
15733+ hi++;
15734+ }
15735+ }
4a4d8108 15736+ kfree(iinfo->ii_hinode);
027c5e7a 15737+ iinfo->ii_hinode = NULL;
4a4d8108 15738+ AuRwDestroy(&iinfo->ii_rwsem);
dece6358 15739+}
7f207e10
AM
15740diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
15741--- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 15742+++ linux/fs/aufs/inode.c 2015-09-24 10:47:58.254719746 +0200
5527c038 15743@@ -0,0 +1,500 @@
4a4d8108 15744+/*
2000de60 15745+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
15746+ *
15747+ * This program, aufs is free software; you can redistribute it and/or modify
15748+ * it under the terms of the GNU General Public License as published by
15749+ * the Free Software Foundation; either version 2 of the License, or
15750+ * (at your option) any later version.
15751+ *
15752+ * This program is distributed in the hope that it will be useful,
15753+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
15754+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15755+ * GNU General Public License for more details.
15756+ *
15757+ * You should have received a copy of the GNU General Public License
523b37e3 15758+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 15759+ */
1facf9fc 15760+
4a4d8108
AM
15761+/*
15762+ * inode functions
15763+ */
1facf9fc 15764+
4a4d8108 15765+#include "aufs.h"
1308ab2a 15766+
4a4d8108
AM
15767+struct inode *au_igrab(struct inode *inode)
15768+{
15769+ if (inode) {
15770+ AuDebugOn(!atomic_read(&inode->i_count));
027c5e7a 15771+ ihold(inode);
1facf9fc 15772+ }
4a4d8108
AM
15773+ return inode;
15774+}
1facf9fc 15775+
4a4d8108
AM
15776+static void au_refresh_hinode_attr(struct inode *inode, int do_version)
15777+{
15778+ au_cpup_attr_all(inode, /*force*/0);
537831f9 15779+ au_update_iigen(inode, /*half*/1);
4a4d8108
AM
15780+ if (do_version)
15781+ inode->i_version++;
dece6358 15782+}
1facf9fc 15783+
027c5e7a 15784+static int au_ii_refresh(struct inode *inode, int *update)
dece6358 15785+{
4a4d8108 15786+ int err, e;
027c5e7a 15787+ umode_t type;
4a4d8108 15788+ aufs_bindex_t bindex, new_bindex;
1308ab2a 15789+ struct super_block *sb;
4a4d8108 15790+ struct au_iinfo *iinfo;
027c5e7a 15791+ struct au_hinode *p, *q, tmp;
1facf9fc 15792+
4a4d8108 15793+ IiMustWriteLock(inode);
1facf9fc 15794+
027c5e7a 15795+ *update = 0;
4a4d8108 15796+ sb = inode->i_sb;
027c5e7a 15797+ type = inode->i_mode & S_IFMT;
4a4d8108
AM
15798+ iinfo = au_ii(inode);
15799+ err = au_ii_realloc(iinfo, au_sbend(sb) + 1);
15800+ if (unlikely(err))
1308ab2a 15801+ goto out;
1facf9fc 15802+
027c5e7a 15803+ AuDebugOn(iinfo->ii_bstart < 0);
4a4d8108 15804+ p = iinfo->ii_hinode + iinfo->ii_bstart;
4a4d8108
AM
15805+ for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
15806+ bindex++, p++) {
15807+ if (!p->hi_inode)
15808+ continue;
1facf9fc 15809+
027c5e7a 15810+ AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
4a4d8108
AM
15811+ new_bindex = au_br_index(sb, p->hi_id);
15812+ if (new_bindex == bindex)
15813+ continue;
1facf9fc 15814+
4a4d8108 15815+ if (new_bindex < 0) {
027c5e7a 15816+ *update = 1;
4a4d8108
AM
15817+ au_hiput(p);
15818+ p->hi_inode = NULL;
15819+ continue;
1308ab2a 15820+ }
4a4d8108
AM
15821+
15822+ if (new_bindex < iinfo->ii_bstart)
15823+ iinfo->ii_bstart = new_bindex;
15824+ if (iinfo->ii_bend < new_bindex)
15825+ iinfo->ii_bend = new_bindex;
15826+ /* swap two lower inode, and loop again */
15827+ q = iinfo->ii_hinode + new_bindex;
15828+ tmp = *q;
15829+ *q = *p;
15830+ *p = tmp;
15831+ if (tmp.hi_inode) {
15832+ bindex--;
15833+ p--;
1308ab2a 15834+ }
15835+ }
4a4d8108
AM
15836+ au_update_ibrange(inode, /*do_put_zero*/0);
15837+ e = au_dy_irefresh(inode);
15838+ if (unlikely(e && !err))
15839+ err = e;
1facf9fc 15840+
4f0767ce 15841+out:
027c5e7a
AM
15842+ AuTraceErr(err);
15843+ return err;
15844+}
15845+
15846+int au_refresh_hinode_self(struct inode *inode)
15847+{
15848+ int err, update;
15849+
15850+ err = au_ii_refresh(inode, &update);
15851+ if (!err)
15852+ au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
15853+
15854+ AuTraceErr(err);
4a4d8108
AM
15855+ return err;
15856+}
1facf9fc 15857+
4a4d8108
AM
15858+int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
15859+{
027c5e7a 15860+ int err, e, update;
4a4d8108 15861+ unsigned int flags;
027c5e7a 15862+ umode_t mode;
4a4d8108 15863+ aufs_bindex_t bindex, bend;
027c5e7a 15864+ unsigned char isdir;
4a4d8108
AM
15865+ struct au_hinode *p;
15866+ struct au_iinfo *iinfo;
1facf9fc 15867+
027c5e7a 15868+ err = au_ii_refresh(inode, &update);
4a4d8108
AM
15869+ if (unlikely(err))
15870+ goto out;
15871+
15872+ update = 0;
15873+ iinfo = au_ii(inode);
15874+ p = iinfo->ii_hinode + iinfo->ii_bstart;
027c5e7a
AM
15875+ mode = (inode->i_mode & S_IFMT);
15876+ isdir = S_ISDIR(mode);
4a4d8108
AM
15877+ flags = au_hi_flags(inode, isdir);
15878+ bend = au_dbend(dentry);
15879+ for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
5527c038 15880+ struct inode *h_i, *h_inode;
4a4d8108
AM
15881+ struct dentry *h_d;
15882+
15883+ h_d = au_h_dptr(dentry, bindex);
5527c038 15884+ if (!h_d || d_is_negative(h_d))
4a4d8108
AM
15885+ continue;
15886+
5527c038
JR
15887+ h_inode = d_inode(h_d);
15888+ AuDebugOn(mode != (h_inode->i_mode & S_IFMT));
4a4d8108
AM
15889+ if (iinfo->ii_bstart <= bindex && bindex <= iinfo->ii_bend) {
15890+ h_i = au_h_iptr(inode, bindex);
15891+ if (h_i) {
5527c038 15892+ if (h_i == h_inode)
4a4d8108
AM
15893+ continue;
15894+ err = -EIO;
15895+ break;
15896+ }
15897+ }
15898+ if (bindex < iinfo->ii_bstart)
15899+ iinfo->ii_bstart = bindex;
15900+ if (iinfo->ii_bend < bindex)
15901+ iinfo->ii_bend = bindex;
5527c038 15902+ au_set_h_iptr(inode, bindex, au_igrab(h_inode), flags);
4a4d8108 15903+ update = 1;
1308ab2a 15904+ }
4a4d8108
AM
15905+ au_update_ibrange(inode, /*do_put_zero*/0);
15906+ e = au_dy_irefresh(inode);
15907+ if (unlikely(e && !err))
15908+ err = e;
027c5e7a
AM
15909+ if (!err)
15910+ au_refresh_hinode_attr(inode, update && isdir);
4a4d8108 15911+
4f0767ce 15912+out:
4a4d8108 15913+ AuTraceErr(err);
1308ab2a 15914+ return err;
dece6358
AM
15915+}
15916+
4a4d8108 15917+static int set_inode(struct inode *inode, struct dentry *dentry)
dece6358 15918+{
4a4d8108
AM
15919+ int err;
15920+ unsigned int flags;
15921+ umode_t mode;
15922+ aufs_bindex_t bindex, bstart, btail;
15923+ unsigned char isdir;
15924+ struct dentry *h_dentry;
15925+ struct inode *h_inode;
15926+ struct au_iinfo *iinfo;
dece6358 15927+
4a4d8108 15928+ IiMustWriteLock(inode);
dece6358 15929+
4a4d8108
AM
15930+ err = 0;
15931+ isdir = 0;
15932+ bstart = au_dbstart(dentry);
5527c038
JR
15933+ h_dentry = au_h_dptr(dentry, bstart);
15934+ h_inode = d_inode(h_dentry);
4a4d8108
AM
15935+ mode = h_inode->i_mode;
15936+ switch (mode & S_IFMT) {
15937+ case S_IFREG:
15938+ btail = au_dbtail(dentry);
15939+ inode->i_op = &aufs_iop;
15940+ inode->i_fop = &aufs_file_fop;
15941+ err = au_dy_iaop(inode, bstart, h_inode);
15942+ if (unlikely(err))
15943+ goto out;
15944+ break;
15945+ case S_IFDIR:
15946+ isdir = 1;
15947+ btail = au_dbtaildir(dentry);
15948+ inode->i_op = &aufs_dir_iop;
15949+ inode->i_fop = &aufs_dir_fop;
15950+ break;
15951+ case S_IFLNK:
15952+ btail = au_dbtail(dentry);
15953+ inode->i_op = &aufs_symlink_iop;
15954+ break;
15955+ case S_IFBLK:
15956+ case S_IFCHR:
15957+ case S_IFIFO:
15958+ case S_IFSOCK:
15959+ btail = au_dbtail(dentry);
15960+ inode->i_op = &aufs_iop;
38d290e6 15961+ init_special_inode(inode, mode, h_inode->i_rdev);
4a4d8108
AM
15962+ break;
15963+ default:
15964+ AuIOErr("Unknown file type 0%o\n", mode);
15965+ err = -EIO;
1308ab2a 15966+ goto out;
4a4d8108 15967+ }
dece6358 15968+
4a4d8108
AM
15969+ /* do not set hnotify for whiteouted dirs (SHWH mode) */
15970+ flags = au_hi_flags(inode, isdir);
15971+ if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
15972+ && au_ftest_hi(flags, HNOTIFY)
15973+ && dentry->d_name.len > AUFS_WH_PFX_LEN
15974+ && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
15975+ au_fclr_hi(flags, HNOTIFY);
15976+ iinfo = au_ii(inode);
15977+ iinfo->ii_bstart = bstart;
15978+ iinfo->ii_bend = btail;
15979+ for (bindex = bstart; bindex <= btail; bindex++) {
15980+ h_dentry = au_h_dptr(dentry, bindex);
15981+ if (h_dentry)
15982+ au_set_h_iptr(inode, bindex,
5527c038 15983+ au_igrab(d_inode(h_dentry)), flags);
4a4d8108
AM
15984+ }
15985+ au_cpup_attr_all(inode, /*force*/1);
c1595e42
JR
15986+ /*
15987+ * to force calling aufs_get_acl() every time,
15988+ * do not call cache_no_acl() for aufs inode.
15989+ */
dece6358 15990+
4f0767ce 15991+out:
4a4d8108
AM
15992+ return err;
15993+}
dece6358 15994+
027c5e7a
AM
15995+/*
15996+ * successful returns with iinfo write_locked
15997+ * minus: errno
15998+ * zero: success, matched
15999+ * plus: no error, but unmatched
16000+ */
16001+static int reval_inode(struct inode *inode, struct dentry *dentry)
4a4d8108
AM
16002+{
16003+ int err;
537831f9
AM
16004+ unsigned int gen;
16005+ struct au_iigen iigen;
4a4d8108
AM
16006+ aufs_bindex_t bindex, bend;
16007+ struct inode *h_inode, *h_dinode;
5527c038 16008+ struct dentry *h_dentry;
dece6358 16009+
4a4d8108
AM
16010+ /*
16011+ * before this function, if aufs got any iinfo lock, it must be only
16012+ * one, the parent dir.
16013+ * it can happen by UDBA and the obsoleted inode number.
16014+ */
16015+ err = -EIO;
16016+ if (unlikely(inode->i_ino == parent_ino(dentry)))
16017+ goto out;
16018+
027c5e7a 16019+ err = 1;
4a4d8108 16020+ ii_write_lock_new_child(inode);
5527c038
JR
16021+ h_dentry = au_h_dptr(dentry, au_dbstart(dentry));
16022+ h_dinode = d_inode(h_dentry);
4a4d8108
AM
16023+ bend = au_ibend(inode);
16024+ for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
16025+ h_inode = au_h_iptr(inode, bindex);
537831f9
AM
16026+ if (!h_inode || h_inode != h_dinode)
16027+ continue;
16028+
16029+ err = 0;
16030+ gen = au_iigen(inode, &iigen);
16031+ if (gen == au_digen(dentry)
16032+ && !au_ig_ftest(iigen.ig_flags, HALF_REFRESHED))
4a4d8108 16033+ break;
537831f9
AM
16034+
16035+ /* fully refresh inode using dentry */
16036+ err = au_refresh_hinode(inode, dentry);
16037+ if (!err)
16038+ au_update_iigen(inode, /*half*/0);
16039+ break;
1facf9fc 16040+ }
dece6358 16041+
4a4d8108
AM
16042+ if (unlikely(err))
16043+ ii_write_unlock(inode);
4f0767ce 16044+out:
1facf9fc 16045+ return err;
16046+}
1facf9fc 16047+
4a4d8108
AM
16048+int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
16049+ unsigned int d_type, ino_t *ino)
1facf9fc 16050+{
4a4d8108
AM
16051+ int err;
16052+ struct mutex *mtx;
1facf9fc 16053+
b752ccd1 16054+ /* prevent hardlinked inode number from race condition */
4a4d8108 16055+ mtx = NULL;
b752ccd1 16056+ if (d_type != DT_DIR) {
4a4d8108
AM
16057+ mtx = &au_sbr(sb, bindex)->br_xino.xi_nondir_mtx;
16058+ mutex_lock(mtx);
16059+ }
16060+ err = au_xino_read(sb, bindex, h_ino, ino);
16061+ if (unlikely(err))
16062+ goto out;
1308ab2a 16063+
4a4d8108
AM
16064+ if (!*ino) {
16065+ err = -EIO;
16066+ *ino = au_xino_new_ino(sb);
16067+ if (unlikely(!*ino))
1facf9fc 16068+ goto out;
4a4d8108
AM
16069+ err = au_xino_write(sb, bindex, h_ino, *ino);
16070+ if (unlikely(err))
1308ab2a 16071+ goto out;
1308ab2a 16072+ }
1facf9fc 16073+
4f0767ce 16074+out:
b752ccd1 16075+ if (mtx)
4a4d8108 16076+ mutex_unlock(mtx);
1facf9fc 16077+ return err;
16078+}
16079+
4a4d8108
AM
16080+/* successful returns with iinfo write_locked */
16081+/* todo: return with unlocked? */
16082+struct inode *au_new_inode(struct dentry *dentry, int must_new)
1facf9fc 16083+{
5527c038 16084+ struct inode *inode, *h_inode;
4a4d8108
AM
16085+ struct dentry *h_dentry;
16086+ struct super_block *sb;
b752ccd1 16087+ struct mutex *mtx;
4a4d8108 16088+ ino_t h_ino, ino;
1716fcea 16089+ int err;
4a4d8108 16090+ aufs_bindex_t bstart;
1facf9fc 16091+
4a4d8108
AM
16092+ sb = dentry->d_sb;
16093+ bstart = au_dbstart(dentry);
16094+ h_dentry = au_h_dptr(dentry, bstart);
5527c038
JR
16095+ h_inode = d_inode(h_dentry);
16096+ h_ino = h_inode->i_ino;
b752ccd1
AM
16097+
16098+ /*
16099+ * stop 'race'-ing between hardlinks under different
16100+ * parents.
16101+ */
16102+ mtx = NULL;
2000de60 16103+ if (!d_is_dir(h_dentry))
b752ccd1
AM
16104+ mtx = &au_sbr(sb, bstart)->br_xino.xi_nondir_mtx;
16105+
4f0767ce 16106+new_ino:
b752ccd1
AM
16107+ if (mtx)
16108+ mutex_lock(mtx);
4a4d8108
AM
16109+ err = au_xino_read(sb, bstart, h_ino, &ino);
16110+ inode = ERR_PTR(err);
16111+ if (unlikely(err))
16112+ goto out;
b752ccd1 16113+
4a4d8108
AM
16114+ if (!ino) {
16115+ ino = au_xino_new_ino(sb);
16116+ if (unlikely(!ino)) {
16117+ inode = ERR_PTR(-EIO);
dece6358
AM
16118+ goto out;
16119+ }
16120+ }
1facf9fc 16121+
4a4d8108
AM
16122+ AuDbg("i%lu\n", (unsigned long)ino);
16123+ inode = au_iget_locked(sb, ino);
16124+ err = PTR_ERR(inode);
16125+ if (IS_ERR(inode))
1facf9fc 16126+ goto out;
1facf9fc 16127+
4a4d8108
AM
16128+ AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
16129+ if (inode->i_state & I_NEW) {
1716fcea 16130+ /* verbose coding for lock class name */
2000de60 16131+ if (unlikely(d_is_symlink(h_dentry)))
1716fcea
AM
16132+ au_rw_class(&au_ii(inode)->ii_rwsem,
16133+ au_lc_key + AuLcSymlink_IIINFO);
2000de60 16134+ else if (unlikely(d_is_dir(h_dentry)))
1716fcea
AM
16135+ au_rw_class(&au_ii(inode)->ii_rwsem,
16136+ au_lc_key + AuLcDir_IIINFO);
16137+ else /* likely */
16138+ au_rw_class(&au_ii(inode)->ii_rwsem,
16139+ au_lc_key + AuLcNonDir_IIINFO);
2dfbb274 16140+
4a4d8108
AM
16141+ ii_write_lock_new_child(inode);
16142+ err = set_inode(inode, dentry);
16143+ if (!err) {
16144+ unlock_new_inode(inode);
16145+ goto out; /* success */
16146+ }
1308ab2a 16147+
027c5e7a
AM
16148+ /*
16149+ * iget_failed() calls iput(), but we need to call
16150+ * ii_write_unlock() after iget_failed(). so dirty hack for
16151+ * i_count.
16152+ */
16153+ atomic_inc(&inode->i_count);
4a4d8108 16154+ iget_failed(inode);
027c5e7a
AM
16155+ ii_write_unlock(inode);
16156+ au_xino_write(sb, bstart, h_ino, /*ino*/0);
16157+ /* ignore this error */
16158+ goto out_iput;
16159+ } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
b752ccd1
AM
16160+ /*
16161+ * horrible race condition between lookup, readdir and copyup
16162+ * (or something).
16163+ */
16164+ if (mtx)
16165+ mutex_unlock(mtx);
027c5e7a
AM
16166+ err = reval_inode(inode, dentry);
16167+ if (unlikely(err < 0)) {
16168+ mtx = NULL;
16169+ goto out_iput;
16170+ }
16171+
b752ccd1
AM
16172+ if (!err) {
16173+ mtx = NULL;
4a4d8108 16174+ goto out; /* success */
b752ccd1
AM
16175+ } else if (mtx)
16176+ mutex_lock(mtx);
4a4d8108
AM
16177+ }
16178+
5527c038 16179+ if (unlikely(au_test_fs_unique_ino(h_inode)))
4a4d8108 16180+ AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
523b37e3
AM
16181+ " b%d, %s, %pd, hi%lu, i%lu.\n",
16182+ bstart, au_sbtype(h_dentry->d_sb), dentry,
4a4d8108
AM
16183+ (unsigned long)h_ino, (unsigned long)ino);
16184+ ino = 0;
16185+ err = au_xino_write(sb, bstart, h_ino, /*ino*/0);
16186+ if (!err) {
16187+ iput(inode);
b752ccd1
AM
16188+ if (mtx)
16189+ mutex_unlock(mtx);
4a4d8108
AM
16190+ goto new_ino;
16191+ }
1308ab2a 16192+
4f0767ce 16193+out_iput:
4a4d8108 16194+ iput(inode);
4a4d8108 16195+ inode = ERR_PTR(err);
4f0767ce 16196+out:
b752ccd1
AM
16197+ if (mtx)
16198+ mutex_unlock(mtx);
4a4d8108 16199+ return inode;
1facf9fc 16200+}
16201+
4a4d8108 16202+/* ---------------------------------------------------------------------- */
1facf9fc 16203+
4a4d8108
AM
16204+int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
16205+ struct inode *inode)
16206+{
16207+ int err;
076b876e 16208+ struct inode *hi;
1facf9fc 16209+
4a4d8108 16210+ err = au_br_rdonly(au_sbr(sb, bindex));
1facf9fc 16211+
4a4d8108
AM
16212+ /* pseudo-link after flushed may happen out of bounds */
16213+ if (!err
16214+ && inode
16215+ && au_ibstart(inode) <= bindex
16216+ && bindex <= au_ibend(inode)) {
16217+ /*
16218+ * permission check is unnecessary since vfsub routine
16219+ * will be called later
16220+ */
076b876e 16221+ hi = au_h_iptr(inode, bindex);
4a4d8108
AM
16222+ if (hi)
16223+ err = IS_IMMUTABLE(hi) ? -EROFS : 0;
1facf9fc 16224+ }
16225+
4a4d8108
AM
16226+ return err;
16227+}
dece6358 16228+
4a4d8108
AM
16229+int au_test_h_perm(struct inode *h_inode, int mask)
16230+{
2dfbb274 16231+ if (uid_eq(current_fsuid(), GLOBAL_ROOT_UID))
4a4d8108
AM
16232+ return 0;
16233+ return inode_permission(h_inode, mask);
16234+}
1facf9fc 16235+
4a4d8108
AM
16236+int au_test_h_perm_sio(struct inode *h_inode, int mask)
16237+{
16238+ if (au_test_nfs(h_inode->i_sb)
16239+ && (mask & MAY_WRITE)
16240+ && S_ISDIR(h_inode->i_mode))
16241+ mask |= MAY_READ; /* force permission check */
16242+ return au_test_h_perm(h_inode, mask);
1facf9fc 16243+}
7f207e10
AM
16244diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
16245--- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 16246+++ linux/fs/aufs/inode.h 2015-09-24 10:47:58.254719746 +0200
b912730e 16247@@ -0,0 +1,673 @@
4a4d8108 16248+/*
2000de60 16249+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
16250+ *
16251+ * This program, aufs is free software; you can redistribute it and/or modify
16252+ * it under the terms of the GNU General Public License as published by
16253+ * the Free Software Foundation; either version 2 of the License, or
16254+ * (at your option) any later version.
16255+ *
16256+ * This program is distributed in the hope that it will be useful,
16257+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
16258+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16259+ * GNU General Public License for more details.
16260+ *
16261+ * You should have received a copy of the GNU General Public License
523b37e3 16262+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 16263+ */
1facf9fc 16264+
1308ab2a 16265+/*
4a4d8108 16266+ * inode operations
1308ab2a 16267+ */
dece6358 16268+
4a4d8108
AM
16269+#ifndef __AUFS_INODE_H__
16270+#define __AUFS_INODE_H__
dece6358 16271+
4a4d8108 16272+#ifdef __KERNEL__
1308ab2a 16273+
4a4d8108 16274+#include <linux/fsnotify.h>
4a4d8108 16275+#include "rwsem.h"
1308ab2a 16276+
4a4d8108 16277+struct vfsmount;
1facf9fc 16278+
4a4d8108
AM
16279+struct au_hnotify {
16280+#ifdef CONFIG_AUFS_HNOTIFY
16281+#ifdef CONFIG_AUFS_HFSNOTIFY
7f207e10 16282+ /* never use fsnotify_add_vfsmount_mark() */
0c5527e5 16283+ struct fsnotify_mark hn_mark;
4a4d8108 16284+#endif
7f207e10 16285+ struct inode *hn_aufs_inode; /* no get/put */
4a4d8108
AM
16286+#endif
16287+} ____cacheline_aligned_in_smp;
1facf9fc 16288+
4a4d8108
AM
16289+struct au_hinode {
16290+ struct inode *hi_inode;
16291+ aufs_bindex_t hi_id;
16292+#ifdef CONFIG_AUFS_HNOTIFY
16293+ struct au_hnotify *hi_notify;
16294+#endif
dece6358 16295+
4a4d8108
AM
16296+ /* reference to the copied-up whiteout with get/put */
16297+ struct dentry *hi_whdentry;
16298+};
dece6358 16299+
537831f9
AM
16300+/* ig_flags */
16301+#define AuIG_HALF_REFRESHED 1
16302+#define au_ig_ftest(flags, name) ((flags) & AuIG_##name)
16303+#define au_ig_fset(flags, name) \
16304+ do { (flags) |= AuIG_##name; } while (0)
16305+#define au_ig_fclr(flags, name) \
16306+ do { (flags) &= ~AuIG_##name; } while (0)
16307+
16308+struct au_iigen {
16309+ __u32 ig_generation, ig_flags;
16310+};
16311+
4a4d8108
AM
16312+struct au_vdir;
16313+struct au_iinfo {
537831f9 16314+ spinlock_t ii_genspin;
7a9e40b8 16315+ struct au_iigen ii_generation;
4a4d8108 16316+ struct super_block *ii_hsb1; /* no get/put */
1facf9fc 16317+
4a4d8108
AM
16318+ struct au_rwsem ii_rwsem;
16319+ aufs_bindex_t ii_bstart, ii_bend;
16320+ __u32 ii_higen;
16321+ struct au_hinode *ii_hinode;
16322+ struct au_vdir *ii_vdir;
16323+};
1facf9fc 16324+
4a4d8108
AM
16325+struct au_icntnr {
16326+ struct au_iinfo iinfo;
16327+ struct inode vfs_inode;
16328+} ____cacheline_aligned_in_smp;
1308ab2a 16329+
4a4d8108
AM
16330+/* au_pin flags */
16331+#define AuPin_DI_LOCKED 1
16332+#define AuPin_MNT_WRITE (1 << 1)
16333+#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
7f207e10
AM
16334+#define au_fset_pin(flags, name) \
16335+ do { (flags) |= AuPin_##name; } while (0)
16336+#define au_fclr_pin(flags, name) \
16337+ do { (flags) &= ~AuPin_##name; } while (0)
4a4d8108
AM
16338+
16339+struct au_pin {
16340+ /* input */
16341+ struct dentry *dentry;
16342+ unsigned int udba;
16343+ unsigned char lsc_di, lsc_hi, flags;
16344+ aufs_bindex_t bindex;
16345+
16346+ /* output */
16347+ struct dentry *parent;
16348+ struct au_hinode *hdir;
16349+ struct vfsmount *h_mnt;
86dc4139
AM
16350+
16351+ /* temporary unlock/relock for copyup */
16352+ struct dentry *h_dentry, *h_parent;
16353+ struct au_branch *br;
16354+ struct task_struct *task;
4a4d8108 16355+};
1facf9fc 16356+
86dc4139 16357+void au_pin_hdir_unlock(struct au_pin *p);
c1595e42 16358+int au_pin_hdir_lock(struct au_pin *p);
86dc4139
AM
16359+int au_pin_hdir_relock(struct au_pin *p);
16360+void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task);
16361+void au_pin_hdir_acquire_nest(struct au_pin *p);
16362+void au_pin_hdir_release(struct au_pin *p);
16363+
1308ab2a 16364+/* ---------------------------------------------------------------------- */
16365+
4a4d8108 16366+static inline struct au_iinfo *au_ii(struct inode *inode)
1facf9fc 16367+{
4a4d8108 16368+ struct au_iinfo *iinfo;
1facf9fc 16369+
4a4d8108
AM
16370+ iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
16371+ if (iinfo->ii_hinode)
16372+ return iinfo;
16373+ return NULL; /* debugging bad_inode case */
16374+}
1facf9fc 16375+
4a4d8108 16376+/* ---------------------------------------------------------------------- */
1facf9fc 16377+
4a4d8108
AM
16378+/* inode.c */
16379+struct inode *au_igrab(struct inode *inode);
027c5e7a 16380+int au_refresh_hinode_self(struct inode *inode);
4a4d8108
AM
16381+int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
16382+int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
16383+ unsigned int d_type, ino_t *ino);
16384+struct inode *au_new_inode(struct dentry *dentry, int must_new);
16385+int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
16386+ struct inode *inode);
16387+int au_test_h_perm(struct inode *h_inode, int mask);
16388+int au_test_h_perm_sio(struct inode *h_inode, int mask);
1facf9fc 16389+
4a4d8108
AM
16390+static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
16391+ ino_t h_ino, unsigned int d_type, ino_t *ino)
16392+{
16393+#ifdef CONFIG_AUFS_SHWH
16394+ return au_ino(sb, bindex, h_ino, d_type, ino);
16395+#else
16396+ return 0;
16397+#endif
16398+}
1facf9fc 16399+
4a4d8108
AM
16400+/* i_op.c */
16401+extern struct inode_operations aufs_iop, aufs_symlink_iop, aufs_dir_iop;
1308ab2a 16402+
4a4d8108
AM
16403+/* au_wr_dir flags */
16404+#define AuWrDir_ADD_ENTRY 1
7e9cd9fe
AM
16405+#define AuWrDir_ISDIR (1 << 1)
16406+#define AuWrDir_TMPFILE (1 << 2)
4a4d8108 16407+#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
7f207e10
AM
16408+#define au_fset_wrdir(flags, name) \
16409+ do { (flags) |= AuWrDir_##name; } while (0)
16410+#define au_fclr_wrdir(flags, name) \
16411+ do { (flags) &= ~AuWrDir_##name; } while (0)
1facf9fc 16412+
4a4d8108
AM
16413+struct au_wr_dir_args {
16414+ aufs_bindex_t force_btgt;
16415+ unsigned char flags;
16416+};
16417+int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
16418+ struct au_wr_dir_args *args);
dece6358 16419+
4a4d8108
AM
16420+struct dentry *au_pinned_h_parent(struct au_pin *pin);
16421+void au_pin_init(struct au_pin *pin, struct dentry *dentry,
16422+ aufs_bindex_t bindex, int lsc_di, int lsc_hi,
16423+ unsigned int udba, unsigned char flags);
16424+int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
16425+ unsigned int udba, unsigned char flags) __must_check;
16426+int au_do_pin(struct au_pin *pin) __must_check;
16427+void au_unpin(struct au_pin *pin);
c1595e42
JR
16428+int au_reval_for_attr(struct dentry *dentry, unsigned int sigen);
16429+
16430+#define AuIcpup_DID_CPUP 1
16431+#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
16432+#define au_fset_icpup(flags, name) \
16433+ do { (flags) |= AuIcpup_##name; } while (0)
16434+#define au_fclr_icpup(flags, name) \
16435+ do { (flags) &= ~AuIcpup_##name; } while (0)
16436+
16437+struct au_icpup_args {
16438+ unsigned char flags;
16439+ unsigned char pin_flags;
16440+ aufs_bindex_t btgt;
16441+ unsigned int udba;
16442+ struct au_pin pin;
16443+ struct path h_path;
16444+ struct inode *h_inode;
16445+};
16446+
16447+int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
16448+ struct au_icpup_args *a);
16449+
16450+int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path);
1facf9fc 16451+
4a4d8108
AM
16452+/* i_op_add.c */
16453+int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
16454+ struct dentry *h_parent, int isdir);
7eafdf33
AM
16455+int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
16456+ dev_t dev);
4a4d8108 16457+int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname);
7eafdf33 16458+int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
b4510431 16459+ bool want_excl);
b912730e
AM
16460+struct vfsub_aopen_args;
16461+int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
16462+ struct vfsub_aopen_args *args);
38d290e6 16463+int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode);
4a4d8108
AM
16464+int aufs_link(struct dentry *src_dentry, struct inode *dir,
16465+ struct dentry *dentry);
7eafdf33 16466+int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
1facf9fc 16467+
4a4d8108
AM
16468+/* i_op_del.c */
16469+int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
16470+int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
16471+ struct dentry *h_parent, int isdir);
16472+int aufs_unlink(struct inode *dir, struct dentry *dentry);
16473+int aufs_rmdir(struct inode *dir, struct dentry *dentry);
1308ab2a 16474+
4a4d8108
AM
16475+/* i_op_ren.c */
16476+int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
16477+int aufs_rename(struct inode *src_dir, struct dentry *src_dentry,
16478+ struct inode *dir, struct dentry *dentry);
1facf9fc 16479+
4a4d8108
AM
16480+/* iinfo.c */
16481+struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
16482+void au_hiput(struct au_hinode *hinode);
16483+void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
16484+ struct dentry *h_wh);
16485+unsigned int au_hi_flags(struct inode *inode, int isdir);
1308ab2a 16486+
4a4d8108
AM
16487+/* hinode flags */
16488+#define AuHi_XINO 1
16489+#define AuHi_HNOTIFY (1 << 1)
16490+#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
7f207e10
AM
16491+#define au_fset_hi(flags, name) \
16492+ do { (flags) |= AuHi_##name; } while (0)
16493+#define au_fclr_hi(flags, name) \
16494+ do { (flags) &= ~AuHi_##name; } while (0)
1facf9fc 16495+
4a4d8108
AM
16496+#ifndef CONFIG_AUFS_HNOTIFY
16497+#undef AuHi_HNOTIFY
16498+#define AuHi_HNOTIFY 0
16499+#endif
1facf9fc 16500+
4a4d8108
AM
16501+void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
16502+ struct inode *h_inode, unsigned int flags);
1facf9fc 16503+
537831f9 16504+void au_update_iigen(struct inode *inode, int half);
4a4d8108 16505+void au_update_ibrange(struct inode *inode, int do_put_zero);
1facf9fc 16506+
4a4d8108
AM
16507+void au_icntnr_init_once(void *_c);
16508+int au_iinfo_init(struct inode *inode);
16509+void au_iinfo_fin(struct inode *inode);
16510+int au_ii_realloc(struct au_iinfo *iinfo, int nbr);
1308ab2a 16511+
e49829fe 16512+#ifdef CONFIG_PROC_FS
4a4d8108 16513+/* plink.c */
e49829fe 16514+int au_plink_maint(struct super_block *sb, int flags);
7e9cd9fe 16515+struct au_sbinfo;
e49829fe
JR
16516+void au_plink_maint_leave(struct au_sbinfo *sbinfo);
16517+int au_plink_maint_enter(struct super_block *sb);
4a4d8108
AM
16518+#ifdef CONFIG_AUFS_DEBUG
16519+void au_plink_list(struct super_block *sb);
16520+#else
16521+AuStubVoid(au_plink_list, struct super_block *sb)
16522+#endif
16523+int au_plink_test(struct inode *inode);
16524+struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
16525+void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
16526+ struct dentry *h_dentry);
e49829fe
JR
16527+void au_plink_put(struct super_block *sb, int verbose);
16528+void au_plink_clean(struct super_block *sb, int verbose);
4a4d8108 16529+void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
e49829fe
JR
16530+#else
16531+AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
16532+AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
16533+AuStubInt0(au_plink_maint_enter, struct super_block *sb);
16534+AuStubVoid(au_plink_list, struct super_block *sb);
16535+AuStubInt0(au_plink_test, struct inode *inode);
16536+AuStub(struct dentry *, au_plink_lkup, return NULL,
16537+ struct inode *inode, aufs_bindex_t bindex);
16538+AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
16539+ struct dentry *h_dentry);
16540+AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
16541+AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
16542+AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
16543+#endif /* CONFIG_PROC_FS */
1facf9fc 16544+
c1595e42
JR
16545+#ifdef CONFIG_AUFS_XATTR
16546+/* xattr.c */
7e9cd9fe
AM
16547+int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
16548+ unsigned int verbose);
c1595e42
JR
16549+ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size);
16550+ssize_t aufs_getxattr(struct dentry *dentry, const char *name, void *value,
16551+ size_t size);
16552+int aufs_setxattr(struct dentry *dentry, const char *name, const void *value,
16553+ size_t size, int flags);
16554+int aufs_removexattr(struct dentry *dentry, const char *name);
16555+
16556+/* void au_xattr_init(struct super_block *sb); */
16557+#else
16558+AuStubInt0(au_cpup_xattr, struct dentry *h_dst, struct dentry *h_src,
7e9cd9fe 16559+ int ignore_flags, unsigned int verbose);
c1595e42
JR
16560+/* AuStubVoid(au_xattr_init, struct super_block *sb); */
16561+#endif
16562+
16563+#ifdef CONFIG_FS_POSIX_ACL
16564+struct posix_acl *aufs_get_acl(struct inode *inode, int type);
16565+int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
16566+#endif
16567+
16568+#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
16569+enum {
16570+ AU_XATTR_SET,
16571+ AU_XATTR_REMOVE,
16572+ AU_ACL_SET
16573+};
16574+
16575+struct au_srxattr {
16576+ int type;
16577+ union {
16578+ struct {
16579+ const char *name;
16580+ const void *value;
16581+ size_t size;
16582+ int flags;
16583+ } set;
16584+ struct {
16585+ const char *name;
16586+ } remove;
16587+ struct {
16588+ struct posix_acl *acl;
16589+ int type;
16590+ } acl_set;
16591+ } u;
16592+};
16593+ssize_t au_srxattr(struct dentry *dentry, struct au_srxattr *arg);
16594+#endif
16595+
4a4d8108 16596+/* ---------------------------------------------------------------------- */
1308ab2a 16597+
4a4d8108
AM
16598+/* lock subclass for iinfo */
16599+enum {
16600+ AuLsc_II_CHILD, /* child first */
16601+ AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
16602+ AuLsc_II_CHILD3, /* copyup dirs */
16603+ AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
16604+ AuLsc_II_PARENT2,
16605+ AuLsc_II_PARENT3, /* copyup dirs */
16606+ AuLsc_II_NEW_CHILD
16607+};
1308ab2a 16608+
1facf9fc 16609+/*
4a4d8108
AM
16610+ * ii_read_lock_child, ii_write_lock_child,
16611+ * ii_read_lock_child2, ii_write_lock_child2,
16612+ * ii_read_lock_child3, ii_write_lock_child3,
16613+ * ii_read_lock_parent, ii_write_lock_parent,
16614+ * ii_read_lock_parent2, ii_write_lock_parent2,
16615+ * ii_read_lock_parent3, ii_write_lock_parent3,
16616+ * ii_read_lock_new_child, ii_write_lock_new_child,
1facf9fc 16617+ */
4a4d8108
AM
16618+#define AuReadLockFunc(name, lsc) \
16619+static inline void ii_read_lock_##name(struct inode *i) \
16620+{ \
16621+ au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
16622+}
16623+
16624+#define AuWriteLockFunc(name, lsc) \
16625+static inline void ii_write_lock_##name(struct inode *i) \
16626+{ \
16627+ au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
16628+}
16629+
16630+#define AuRWLockFuncs(name, lsc) \
16631+ AuReadLockFunc(name, lsc) \
16632+ AuWriteLockFunc(name, lsc)
16633+
16634+AuRWLockFuncs(child, CHILD);
16635+AuRWLockFuncs(child2, CHILD2);
16636+AuRWLockFuncs(child3, CHILD3);
16637+AuRWLockFuncs(parent, PARENT);
16638+AuRWLockFuncs(parent2, PARENT2);
16639+AuRWLockFuncs(parent3, PARENT3);
16640+AuRWLockFuncs(new_child, NEW_CHILD);
16641+
16642+#undef AuReadLockFunc
16643+#undef AuWriteLockFunc
16644+#undef AuRWLockFuncs
1facf9fc 16645+
16646+/*
4a4d8108 16647+ * ii_read_unlock, ii_write_unlock, ii_downgrade_lock
1facf9fc 16648+ */
4a4d8108 16649+AuSimpleUnlockRwsemFuncs(ii, struct inode *i, &au_ii(i)->ii_rwsem);
1facf9fc 16650+
4a4d8108
AM
16651+#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
16652+#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
16653+#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
1facf9fc 16654+
4a4d8108 16655+/* ---------------------------------------------------------------------- */
1308ab2a 16656+
027c5e7a
AM
16657+static inline void au_icntnr_init(struct au_icntnr *c)
16658+{
16659+#ifdef CONFIG_AUFS_DEBUG
16660+ c->vfs_inode.i_mode = 0;
16661+#endif
16662+}
16663+
537831f9 16664+static inline unsigned int au_iigen(struct inode *inode, struct au_iigen *iigen)
4a4d8108 16665+{
537831f9
AM
16666+ unsigned int gen;
16667+ struct au_iinfo *iinfo;
16668+
16669+ iinfo = au_ii(inode);
16670+ spin_lock(&iinfo->ii_genspin);
16671+ if (iigen)
16672+ *iigen = iinfo->ii_generation;
16673+ gen = iinfo->ii_generation.ig_generation;
16674+ spin_unlock(&iinfo->ii_genspin);
16675+
16676+ return gen;
4a4d8108 16677+}
1308ab2a 16678+
4a4d8108
AM
16679+/* tiny test for inode number */
16680+/* tmpfs generation is too rough */
16681+static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
16682+{
16683+ struct au_iinfo *iinfo;
1308ab2a 16684+
4a4d8108
AM
16685+ iinfo = au_ii(inode);
16686+ AuRwMustAnyLock(&iinfo->ii_rwsem);
16687+ return !(iinfo->ii_hsb1 == h_inode->i_sb
16688+ && iinfo->ii_higen == h_inode->i_generation);
16689+}
1308ab2a 16690+
4a4d8108
AM
16691+static inline void au_iigen_dec(struct inode *inode)
16692+{
537831f9
AM
16693+ struct au_iinfo *iinfo;
16694+
16695+ iinfo = au_ii(inode);
16696+ spin_lock(&iinfo->ii_genspin);
16697+ iinfo->ii_generation.ig_generation--;
16698+ spin_unlock(&iinfo->ii_genspin);
027c5e7a
AM
16699+}
16700+
16701+static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
16702+{
16703+ int err;
16704+
16705+ err = 0;
537831f9 16706+ if (unlikely(inode && au_iigen(inode, NULL) != sigen))
027c5e7a
AM
16707+ err = -EIO;
16708+
16709+ return err;
4a4d8108 16710+}
1308ab2a 16711+
4a4d8108 16712+/* ---------------------------------------------------------------------- */
1308ab2a 16713+
4a4d8108
AM
16714+static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
16715+ aufs_bindex_t bindex)
16716+{
16717+ IiMustAnyLock(inode);
16718+ return au_ii(inode)->ii_hinode[0 + bindex].hi_id;
16719+}
1308ab2a 16720+
4a4d8108
AM
16721+static inline aufs_bindex_t au_ibstart(struct inode *inode)
16722+{
16723+ IiMustAnyLock(inode);
16724+ return au_ii(inode)->ii_bstart;
16725+}
1308ab2a 16726+
4a4d8108
AM
16727+static inline aufs_bindex_t au_ibend(struct inode *inode)
16728+{
16729+ IiMustAnyLock(inode);
16730+ return au_ii(inode)->ii_bend;
16731+}
1308ab2a 16732+
4a4d8108
AM
16733+static inline struct au_vdir *au_ivdir(struct inode *inode)
16734+{
16735+ IiMustAnyLock(inode);
16736+ return au_ii(inode)->ii_vdir;
16737+}
1308ab2a 16738+
4a4d8108
AM
16739+static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
16740+{
16741+ IiMustAnyLock(inode);
16742+ return au_ii(inode)->ii_hinode[0 + bindex].hi_whdentry;
16743+}
1308ab2a 16744+
4a4d8108 16745+static inline void au_set_ibstart(struct inode *inode, aufs_bindex_t bindex)
1308ab2a 16746+{
4a4d8108
AM
16747+ IiMustWriteLock(inode);
16748+ au_ii(inode)->ii_bstart = bindex;
16749+}
1308ab2a 16750+
4a4d8108
AM
16751+static inline void au_set_ibend(struct inode *inode, aufs_bindex_t bindex)
16752+{
16753+ IiMustWriteLock(inode);
16754+ au_ii(inode)->ii_bend = bindex;
1308ab2a 16755+}
16756+
4a4d8108
AM
16757+static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
16758+{
16759+ IiMustWriteLock(inode);
16760+ au_ii(inode)->ii_vdir = vdir;
16761+}
1facf9fc 16762+
4a4d8108 16763+static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
1308ab2a 16764+{
4a4d8108
AM
16765+ IiMustAnyLock(inode);
16766+ return au_ii(inode)->ii_hinode + bindex;
16767+}
dece6358 16768+
4a4d8108 16769+/* ---------------------------------------------------------------------- */
1facf9fc 16770+
4a4d8108
AM
16771+static inline struct dentry *au_pinned_parent(struct au_pin *pin)
16772+{
16773+ if (pin)
16774+ return pin->parent;
16775+ return NULL;
1facf9fc 16776+}
16777+
4a4d8108 16778+static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
1facf9fc 16779+{
4a4d8108
AM
16780+ if (pin && pin->hdir)
16781+ return pin->hdir->hi_inode;
16782+ return NULL;
1308ab2a 16783+}
1facf9fc 16784+
4a4d8108
AM
16785+static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
16786+{
16787+ if (pin)
16788+ return pin->hdir;
16789+ return NULL;
16790+}
1facf9fc 16791+
4a4d8108 16792+static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
1308ab2a 16793+{
4a4d8108
AM
16794+ if (pin)
16795+ pin->dentry = dentry;
16796+}
1308ab2a 16797+
4a4d8108
AM
16798+static inline void au_pin_set_parent_lflag(struct au_pin *pin,
16799+ unsigned char lflag)
16800+{
16801+ if (pin) {
7f207e10 16802+ if (lflag)
4a4d8108 16803+ au_fset_pin(pin->flags, DI_LOCKED);
7f207e10 16804+ else
4a4d8108 16805+ au_fclr_pin(pin->flags, DI_LOCKED);
1308ab2a 16806+ }
4a4d8108
AM
16807+}
16808+
7e9cd9fe 16809+#if 0 /* reserved */
4a4d8108
AM
16810+static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
16811+{
16812+ if (pin) {
16813+ dput(pin->parent);
16814+ pin->parent = dget(parent);
1facf9fc 16815+ }
4a4d8108 16816+}
7e9cd9fe 16817+#endif
1facf9fc 16818+
4a4d8108
AM
16819+/* ---------------------------------------------------------------------- */
16820+
027c5e7a 16821+struct au_branch;
4a4d8108
AM
16822+#ifdef CONFIG_AUFS_HNOTIFY
16823+struct au_hnotify_op {
16824+ void (*ctl)(struct au_hinode *hinode, int do_set);
027c5e7a 16825+ int (*alloc)(struct au_hinode *hinode);
7eafdf33
AM
16826+
16827+ /*
16828+ * if it returns true, the the caller should free hinode->hi_notify,
16829+ * otherwise ->free() frees it.
16830+ */
16831+ int (*free)(struct au_hinode *hinode,
16832+ struct au_hnotify *hn) __must_check;
4a4d8108
AM
16833+
16834+ void (*fin)(void);
16835+ int (*init)(void);
027c5e7a
AM
16836+
16837+ int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
16838+ void (*fin_br)(struct au_branch *br);
16839+ int (*init_br)(struct au_branch *br, int perm);
4a4d8108
AM
16840+};
16841+
16842+/* hnotify.c */
027c5e7a 16843+int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
4a4d8108
AM
16844+void au_hn_free(struct au_hinode *hinode);
16845+void au_hn_ctl(struct au_hinode *hinode, int do_set);
16846+void au_hn_reset(struct inode *inode, unsigned int flags);
16847+int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
16848+ struct qstr *h_child_qstr, struct inode *h_child_inode);
027c5e7a
AM
16849+int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
16850+int au_hnotify_init_br(struct au_branch *br, int perm);
16851+void au_hnotify_fin_br(struct au_branch *br);
4a4d8108
AM
16852+int __init au_hnotify_init(void);
16853+void au_hnotify_fin(void);
16854+
7f207e10 16855+/* hfsnotify.c */
4a4d8108
AM
16856+extern const struct au_hnotify_op au_hnotify_op;
16857+
16858+static inline
16859+void au_hn_init(struct au_hinode *hinode)
16860+{
16861+ hinode->hi_notify = NULL;
1308ab2a 16862+}
16863+
53392da6
AM
16864+static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
16865+{
16866+ return hinode->hi_notify;
16867+}
16868+
4a4d8108 16869+#else
c1595e42
JR
16870+AuStub(int, au_hn_alloc, return -EOPNOTSUPP,
16871+ struct au_hinode *hinode __maybe_unused,
16872+ struct inode *inode __maybe_unused)
16873+AuStub(struct au_hnotify *, au_hn, return NULL, struct au_hinode *hinode)
4a4d8108
AM
16874+AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
16875+AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
16876+ int do_set __maybe_unused)
16877+AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
16878+ unsigned int flags __maybe_unused)
027c5e7a
AM
16879+AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
16880+ struct au_branch *br __maybe_unused,
16881+ int perm __maybe_unused)
16882+AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
16883+ int perm __maybe_unused)
16884+AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
4a4d8108
AM
16885+AuStubInt0(__init au_hnotify_init, void)
16886+AuStubVoid(au_hnotify_fin, void)
16887+AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
16888+#endif /* CONFIG_AUFS_HNOTIFY */
16889+
16890+static inline void au_hn_suspend(struct au_hinode *hdir)
16891+{
16892+ au_hn_ctl(hdir, /*do_set*/0);
1308ab2a 16893+}
16894+
4a4d8108 16895+static inline void au_hn_resume(struct au_hinode *hdir)
1308ab2a 16896+{
4a4d8108
AM
16897+ au_hn_ctl(hdir, /*do_set*/1);
16898+}
1308ab2a 16899+
4a4d8108
AM
16900+static inline void au_hn_imtx_lock(struct au_hinode *hdir)
16901+{
16902+ mutex_lock(&hdir->hi_inode->i_mutex);
16903+ au_hn_suspend(hdir);
16904+}
dece6358 16905+
4a4d8108
AM
16906+static inline void au_hn_imtx_lock_nested(struct au_hinode *hdir,
16907+ unsigned int sc __maybe_unused)
16908+{
16909+ mutex_lock_nested(&hdir->hi_inode->i_mutex, sc);
16910+ au_hn_suspend(hdir);
1facf9fc 16911+}
1facf9fc 16912+
4a4d8108
AM
16913+static inline void au_hn_imtx_unlock(struct au_hinode *hdir)
16914+{
16915+ au_hn_resume(hdir);
16916+ mutex_unlock(&hdir->hi_inode->i_mutex);
16917+}
16918+
16919+#endif /* __KERNEL__ */
16920+#endif /* __AUFS_INODE_H__ */
7f207e10
AM
16921diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
16922--- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 16923+++ linux/fs/aufs/ioctl.c 2015-09-24 10:47:58.254719746 +0200
c1595e42 16924@@ -0,0 +1,219 @@
4a4d8108 16925+/*
2000de60 16926+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
16927+ *
16928+ * This program, aufs is free software; you can redistribute it and/or modify
16929+ * it under the terms of the GNU General Public License as published by
16930+ * the Free Software Foundation; either version 2 of the License, or
16931+ * (at your option) any later version.
16932+ *
16933+ * This program is distributed in the hope that it will be useful,
16934+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
16935+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16936+ * GNU General Public License for more details.
16937+ *
16938+ * You should have received a copy of the GNU General Public License
523b37e3 16939+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
16940+ */
16941+
16942+/*
16943+ * ioctl
16944+ * plink-management and readdir in userspace.
16945+ * assist the pathconf(3) wrapper library.
c2b27bf2 16946+ * move-down
076b876e 16947+ * File-based Hierarchical Storage Management.
4a4d8108
AM
16948+ */
16949+
c2b27bf2
AM
16950+#include <linux/compat.h>
16951+#include <linux/file.h>
4a4d8108
AM
16952+#include "aufs.h"
16953+
1e00d052 16954+static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
4a4d8108
AM
16955+{
16956+ int err, fd;
16957+ aufs_bindex_t wbi, bindex, bend;
16958+ struct file *h_file;
16959+ struct super_block *sb;
16960+ struct dentry *root;
1e00d052
AM
16961+ struct au_branch *br;
16962+ struct aufs_wbr_fd wbrfd = {
16963+ .oflags = au_dir_roflags,
16964+ .brid = -1
16965+ };
16966+ const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
16967+ | O_NOATIME | O_CLOEXEC;
4a4d8108 16968+
1e00d052
AM
16969+ AuDebugOn(wbrfd.oflags & ~valid);
16970+
16971+ if (arg) {
16972+ err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
16973+ if (unlikely(err)) {
16974+ err = -EFAULT;
16975+ goto out;
16976+ }
16977+
16978+ err = -EINVAL;
16979+ AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
16980+ wbrfd.oflags |= au_dir_roflags;
16981+ AuDbg("0%o\n", wbrfd.oflags);
16982+ if (unlikely(wbrfd.oflags & ~valid))
16983+ goto out;
16984+ }
16985+
2000de60 16986+ fd = get_unused_fd_flags(0);
1e00d052
AM
16987+ err = fd;
16988+ if (unlikely(fd < 0))
4a4d8108 16989+ goto out;
4a4d8108 16990+
1e00d052 16991+ h_file = ERR_PTR(-EINVAL);
4a4d8108 16992+ wbi = 0;
1e00d052 16993+ br = NULL;
4a4d8108
AM
16994+ sb = path->dentry->d_sb;
16995+ root = sb->s_root;
16996+ aufs_read_lock(root, AuLock_IR);
1e00d052
AM
16997+ bend = au_sbend(sb);
16998+ if (wbrfd.brid >= 0) {
16999+ wbi = au_br_index(sb, wbrfd.brid);
17000+ if (unlikely(wbi < 0 || wbi > bend))
17001+ goto out_unlock;
17002+ }
17003+
17004+ h_file = ERR_PTR(-ENOENT);
17005+ br = au_sbr(sb, wbi);
17006+ if (!au_br_writable(br->br_perm)) {
17007+ if (arg)
17008+ goto out_unlock;
17009+
17010+ bindex = wbi + 1;
17011+ wbi = -1;
17012+ for (; bindex <= bend; bindex++) {
17013+ br = au_sbr(sb, bindex);
17014+ if (au_br_writable(br->br_perm)) {
4a4d8108 17015+ wbi = bindex;
1e00d052 17016+ br = au_sbr(sb, wbi);
4a4d8108
AM
17017+ break;
17018+ }
17019+ }
4a4d8108
AM
17020+ }
17021+ AuDbg("wbi %d\n", wbi);
1e00d052 17022+ if (wbi >= 0)
392086de
AM
17023+ h_file = au_h_open(root, wbi, wbrfd.oflags, NULL,
17024+ /*force_wr*/0);
1e00d052
AM
17025+
17026+out_unlock:
4a4d8108
AM
17027+ aufs_read_unlock(root, AuLock_IR);
17028+ err = PTR_ERR(h_file);
17029+ if (IS_ERR(h_file))
17030+ goto out_fd;
17031+
1e00d052 17032+ atomic_dec(&br->br_count); /* cf. au_h_open() */
4a4d8108
AM
17033+ fd_install(fd, h_file);
17034+ err = fd;
17035+ goto out; /* success */
17036+
4f0767ce 17037+out_fd:
4a4d8108 17038+ put_unused_fd(fd);
4f0767ce 17039+out:
1e00d052 17040+ AuTraceErr(err);
4a4d8108
AM
17041+ return err;
17042+}
17043+
17044+/* ---------------------------------------------------------------------- */
17045+
17046+long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
17047+{
17048+ long err;
c1595e42 17049+ struct dentry *dentry;
4a4d8108
AM
17050+
17051+ switch (cmd) {
4a4d8108
AM
17052+ case AUFS_CTL_RDU:
17053+ case AUFS_CTL_RDU_INO:
17054+ err = au_rdu_ioctl(file, cmd, arg);
17055+ break;
17056+
17057+ case AUFS_CTL_WBR_FD:
1e00d052 17058+ err = au_wbr_fd(&file->f_path, (void __user *)arg);
4a4d8108
AM
17059+ break;
17060+
027c5e7a
AM
17061+ case AUFS_CTL_IBUSY:
17062+ err = au_ibusy_ioctl(file, arg);
17063+ break;
17064+
076b876e
AM
17065+ case AUFS_CTL_BRINFO:
17066+ err = au_brinfo_ioctl(file, arg);
17067+ break;
17068+
17069+ case AUFS_CTL_FHSM_FD:
2000de60 17070+ dentry = file->f_path.dentry;
c1595e42
JR
17071+ if (IS_ROOT(dentry))
17072+ err = au_fhsm_fd(dentry->d_sb, arg);
17073+ else
17074+ err = -ENOTTY;
076b876e
AM
17075+ break;
17076+
4a4d8108
AM
17077+ default:
17078+ /* do not call the lower */
17079+ AuDbg("0x%x\n", cmd);
17080+ err = -ENOTTY;
17081+ }
17082+
17083+ AuTraceErr(err);
17084+ return err;
17085+}
17086+
17087+long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
17088+{
17089+ long err;
17090+
17091+ switch (cmd) {
c2b27bf2 17092+ case AUFS_CTL_MVDOWN:
2000de60 17093+ err = au_mvdown(file->f_path.dentry, (void __user *)arg);
c2b27bf2
AM
17094+ break;
17095+
4a4d8108 17096+ case AUFS_CTL_WBR_FD:
1e00d052 17097+ err = au_wbr_fd(&file->f_path, (void __user *)arg);
4a4d8108
AM
17098+ break;
17099+
17100+ default:
17101+ /* do not call the lower */
17102+ AuDbg("0x%x\n", cmd);
17103+ err = -ENOTTY;
17104+ }
17105+
17106+ AuTraceErr(err);
17107+ return err;
17108+}
b752ccd1
AM
17109+
17110+#ifdef CONFIG_COMPAT
17111+long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
17112+ unsigned long arg)
17113+{
17114+ long err;
17115+
17116+ switch (cmd) {
17117+ case AUFS_CTL_RDU:
17118+ case AUFS_CTL_RDU_INO:
17119+ err = au_rdu_compat_ioctl(file, cmd, arg);
17120+ break;
17121+
027c5e7a
AM
17122+ case AUFS_CTL_IBUSY:
17123+ err = au_ibusy_compat_ioctl(file, arg);
17124+ break;
17125+
076b876e
AM
17126+ case AUFS_CTL_BRINFO:
17127+ err = au_brinfo_compat_ioctl(file, arg);
17128+ break;
17129+
b752ccd1
AM
17130+ default:
17131+ err = aufs_ioctl_dir(file, cmd, arg);
17132+ }
17133+
17134+ AuTraceErr(err);
17135+ return err;
17136+}
17137+
b752ccd1
AM
17138+long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
17139+ unsigned long arg)
17140+{
17141+ return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
17142+}
17143+#endif
7f207e10
AM
17144diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
17145--- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 17146+++ linux/fs/aufs/i_op_add.c 2015-09-24 10:47:58.254719746 +0200
5527c038 17147@@ -0,0 +1,932 @@
4a4d8108 17148+/*
2000de60 17149+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
17150+ *
17151+ * This program, aufs is free software; you can redistribute it and/or modify
17152+ * it under the terms of the GNU General Public License as published by
17153+ * the Free Software Foundation; either version 2 of the License, or
17154+ * (at your option) any later version.
17155+ *
17156+ * This program is distributed in the hope that it will be useful,
17157+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
17158+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17159+ * GNU General Public License for more details.
17160+ *
17161+ * You should have received a copy of the GNU General Public License
523b37e3 17162+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108
AM
17163+ */
17164+
17165+/*
17166+ * inode operations (add entry)
17167+ */
17168+
17169+#include "aufs.h"
17170+
17171+/*
17172+ * final procedure of adding a new entry, except link(2).
17173+ * remove whiteout, instantiate, copyup the parent dir's times and size
17174+ * and update version.
17175+ * if it failed, re-create the removed whiteout.
17176+ */
17177+static int epilog(struct inode *dir, aufs_bindex_t bindex,
17178+ struct dentry *wh_dentry, struct dentry *dentry)
17179+{
17180+ int err, rerr;
17181+ aufs_bindex_t bwh;
17182+ struct path h_path;
076b876e 17183+ struct super_block *sb;
4a4d8108
AM
17184+ struct inode *inode, *h_dir;
17185+ struct dentry *wh;
17186+
17187+ bwh = -1;
076b876e 17188+ sb = dir->i_sb;
4a4d8108 17189+ if (wh_dentry) {
5527c038 17190+ h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
4a4d8108
AM
17191+ IMustLock(h_dir);
17192+ AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
17193+ bwh = au_dbwh(dentry);
17194+ h_path.dentry = wh_dentry;
076b876e 17195+ h_path.mnt = au_sbr_mnt(sb, bindex);
4a4d8108
AM
17196+ err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
17197+ dentry);
17198+ if (unlikely(err))
17199+ goto out;
17200+ }
17201+
17202+ inode = au_new_inode(dentry, /*must_new*/1);
17203+ if (!IS_ERR(inode)) {
17204+ d_instantiate(dentry, inode);
5527c038 17205+ dir = d_inode(dentry->d_parent); /* dir inode is locked */
4a4d8108 17206+ IMustLock(dir);
b912730e 17207+ au_dir_ts(dir, bindex);
4a4d8108 17208+ dir->i_version++;
076b876e 17209+ au_fhsm_wrote(sb, bindex, /*force*/0);
4a4d8108
AM
17210+ return 0; /* success */
17211+ }
17212+
17213+ err = PTR_ERR(inode);
17214+ if (!wh_dentry)
17215+ goto out;
17216+
17217+ /* revert */
17218+ /* dir inode is locked */
17219+ wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
17220+ rerr = PTR_ERR(wh);
17221+ if (IS_ERR(wh)) {
523b37e3
AM
17222+ AuIOErr("%pd reverting whiteout failed(%d, %d)\n",
17223+ dentry, err, rerr);
4a4d8108
AM
17224+ err = -EIO;
17225+ } else
17226+ dput(wh);
17227+
4f0767ce 17228+out:
4a4d8108
AM
17229+ return err;
17230+}
17231+
027c5e7a
AM
17232+static int au_d_may_add(struct dentry *dentry)
17233+{
17234+ int err;
17235+
17236+ err = 0;
17237+ if (unlikely(d_unhashed(dentry)))
17238+ err = -ENOENT;
5527c038 17239+ if (unlikely(d_really_is_positive(dentry)))
027c5e7a
AM
17240+ err = -EEXIST;
17241+ return err;
17242+}
17243+
4a4d8108
AM
17244+/*
17245+ * simple tests for the adding inode operations.
17246+ * following the checks in vfs, plus the parent-child relationship.
17247+ */
17248+int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
17249+ struct dentry *h_parent, int isdir)
17250+{
17251+ int err;
17252+ umode_t h_mode;
17253+ struct dentry *h_dentry;
17254+ struct inode *h_inode;
17255+
17256+ err = -ENAMETOOLONG;
17257+ if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
17258+ goto out;
17259+
17260+ h_dentry = au_h_dptr(dentry, bindex);
5527c038 17261+ if (d_really_is_negative(dentry)) {
4a4d8108 17262+ err = -EEXIST;
5527c038 17263+ if (unlikely(d_is_positive(h_dentry)))
4a4d8108
AM
17264+ goto out;
17265+ } else {
17266+ /* rename(2) case */
17267+ err = -EIO;
5527c038
JR
17268+ if (unlikely(d_is_negative(h_dentry)))
17269+ goto out;
17270+ h_inode = d_inode(h_dentry);
17271+ if (unlikely(!h_inode->i_nlink))
4a4d8108
AM
17272+ goto out;
17273+
17274+ h_mode = h_inode->i_mode;
17275+ if (!isdir) {
17276+ err = -EISDIR;
17277+ if (unlikely(S_ISDIR(h_mode)))
17278+ goto out;
17279+ } else if (unlikely(!S_ISDIR(h_mode))) {
17280+ err = -ENOTDIR;
17281+ goto out;
17282+ }
17283+ }
17284+
17285+ err = 0;
17286+ /* expected parent dir is locked */
17287+ if (unlikely(h_parent != h_dentry->d_parent))
17288+ err = -EIO;
17289+
4f0767ce 17290+out:
4a4d8108
AM
17291+ AuTraceErr(err);
17292+ return err;
17293+}
17294+
17295+/*
17296+ * initial procedure of adding a new entry.
17297+ * prepare writable branch and the parent dir, lock it,
17298+ * and lookup whiteout for the new entry.
17299+ */
17300+static struct dentry*
17301+lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
17302+ struct dentry *src_dentry, struct au_pin *pin,
17303+ struct au_wr_dir_args *wr_dir_args)
17304+{
17305+ struct dentry *wh_dentry, *h_parent;
17306+ struct super_block *sb;
17307+ struct au_branch *br;
17308+ int err;
17309+ unsigned int udba;
17310+ aufs_bindex_t bcpup;
17311+
523b37e3 17312+ AuDbg("%pd\n", dentry);
4a4d8108
AM
17313+
17314+ err = au_wr_dir(dentry, src_dentry, wr_dir_args);
17315+ bcpup = err;
17316+ wh_dentry = ERR_PTR(err);
17317+ if (unlikely(err < 0))
17318+ goto out;
17319+
17320+ sb = dentry->d_sb;
17321+ udba = au_opt_udba(sb);
17322+ err = au_pin(pin, dentry, bcpup, udba,
17323+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17324+ wh_dentry = ERR_PTR(err);
17325+ if (unlikely(err))
17326+ goto out;
17327+
17328+ h_parent = au_pinned_h_parent(pin);
17329+ if (udba != AuOpt_UDBA_NONE
17330+ && au_dbstart(dentry) == bcpup)
17331+ err = au_may_add(dentry, bcpup, h_parent,
17332+ au_ftest_wrdir(wr_dir_args->flags, ISDIR));
17333+ else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
17334+ err = -ENAMETOOLONG;
17335+ wh_dentry = ERR_PTR(err);
17336+ if (unlikely(err))
17337+ goto out_unpin;
17338+
17339+ br = au_sbr(sb, bcpup);
17340+ if (dt) {
17341+ struct path tmp = {
17342+ .dentry = h_parent,
86dc4139 17343+ .mnt = au_br_mnt(br)
4a4d8108
AM
17344+ };
17345+ au_dtime_store(dt, au_pinned_parent(pin), &tmp);
17346+ }
17347+
17348+ wh_dentry = NULL;
17349+ if (bcpup != au_dbwh(dentry))
17350+ goto out; /* success */
17351+
2000de60
JR
17352+ /*
17353+ * ENAMETOOLONG here means that if we allowed create such name, then it
17354+ * would not be able to removed in the future. So we don't allow such
17355+ * name here and we don't handle ENAMETOOLONG differently here.
17356+ */
4a4d8108
AM
17357+ wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
17358+
4f0767ce 17359+out_unpin:
4a4d8108
AM
17360+ if (IS_ERR(wh_dentry))
17361+ au_unpin(pin);
4f0767ce 17362+out:
4a4d8108
AM
17363+ return wh_dentry;
17364+}
17365+
17366+/* ---------------------------------------------------------------------- */
17367+
17368+enum { Mknod, Symlink, Creat };
17369+struct simple_arg {
17370+ int type;
17371+ union {
17372+ struct {
b912730e
AM
17373+ umode_t mode;
17374+ bool want_excl;
17375+ bool try_aopen;
17376+ struct vfsub_aopen_args *aopen;
4a4d8108
AM
17377+ } c;
17378+ struct {
17379+ const char *symname;
17380+ } s;
17381+ struct {
7eafdf33 17382+ umode_t mode;
4a4d8108
AM
17383+ dev_t dev;
17384+ } m;
17385+ } u;
17386+};
17387+
17388+static int add_simple(struct inode *dir, struct dentry *dentry,
17389+ struct simple_arg *arg)
17390+{
076b876e 17391+ int err, rerr;
4a4d8108
AM
17392+ aufs_bindex_t bstart;
17393+ unsigned char created;
b912730e
AM
17394+ const unsigned char try_aopen
17395+ = (arg->type == Creat && arg->u.c.try_aopen);
4a4d8108
AM
17396+ struct dentry *wh_dentry, *parent;
17397+ struct inode *h_dir;
b912730e
AM
17398+ struct super_block *sb;
17399+ struct au_branch *br;
c2b27bf2
AM
17400+ /* to reuduce stack size */
17401+ struct {
17402+ struct au_dtime dt;
17403+ struct au_pin pin;
17404+ struct path h_path;
17405+ struct au_wr_dir_args wr_dir_args;
17406+ } *a;
4a4d8108 17407+
523b37e3 17408+ AuDbg("%pd\n", dentry);
4a4d8108
AM
17409+ IMustLock(dir);
17410+
c2b27bf2
AM
17411+ err = -ENOMEM;
17412+ a = kmalloc(sizeof(*a), GFP_NOFS);
17413+ if (unlikely(!a))
17414+ goto out;
17415+ a->wr_dir_args.force_btgt = -1;
17416+ a->wr_dir_args.flags = AuWrDir_ADD_ENTRY;
17417+
4a4d8108 17418+ parent = dentry->d_parent; /* dir inode is locked */
b912730e
AM
17419+ if (!try_aopen) {
17420+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
17421+ if (unlikely(err))
17422+ goto out_free;
17423+ }
027c5e7a
AM
17424+ err = au_d_may_add(dentry);
17425+ if (unlikely(err))
17426+ goto out_unlock;
b912730e
AM
17427+ if (!try_aopen)
17428+ di_write_lock_parent(parent);
c2b27bf2
AM
17429+ wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
17430+ &a->pin, &a->wr_dir_args);
4a4d8108
AM
17431+ err = PTR_ERR(wh_dentry);
17432+ if (IS_ERR(wh_dentry))
027c5e7a 17433+ goto out_parent;
4a4d8108
AM
17434+
17435+ bstart = au_dbstart(dentry);
b912730e
AM
17436+ sb = dentry->d_sb;
17437+ br = au_sbr(sb, bstart);
c2b27bf2 17438+ a->h_path.dentry = au_h_dptr(dentry, bstart);
b912730e 17439+ a->h_path.mnt = au_br_mnt(br);
c2b27bf2 17440+ h_dir = au_pinned_h_dir(&a->pin);
4a4d8108
AM
17441+ switch (arg->type) {
17442+ case Creat:
b912730e
AM
17443+ err = 0;
17444+ if (!try_aopen || !h_dir->i_op->atomic_open)
17445+ err = vfsub_create(h_dir, &a->h_path, arg->u.c.mode,
17446+ arg->u.c.want_excl);
17447+ else
17448+ err = vfsub_atomic_open(h_dir, a->h_path.dentry,
17449+ arg->u.c.aopen, br);
4a4d8108
AM
17450+ break;
17451+ case Symlink:
c2b27bf2 17452+ err = vfsub_symlink(h_dir, &a->h_path, arg->u.s.symname);
4a4d8108
AM
17453+ break;
17454+ case Mknod:
c2b27bf2
AM
17455+ err = vfsub_mknod(h_dir, &a->h_path, arg->u.m.mode,
17456+ arg->u.m.dev);
4a4d8108
AM
17457+ break;
17458+ default:
17459+ BUG();
17460+ }
17461+ created = !err;
17462+ if (!err)
17463+ err = epilog(dir, bstart, wh_dentry, dentry);
17464+
17465+ /* revert */
5527c038 17466+ if (unlikely(created && err && d_is_positive(a->h_path.dentry))) {
523b37e3
AM
17467+ /* no delegation since it is just created */
17468+ rerr = vfsub_unlink(h_dir, &a->h_path, /*delegated*/NULL,
17469+ /*force*/0);
4a4d8108 17470+ if (rerr) {
523b37e3
AM
17471+ AuIOErr("%pd revert failure(%d, %d)\n",
17472+ dentry, err, rerr);
4a4d8108
AM
17473+ err = -EIO;
17474+ }
c2b27bf2 17475+ au_dtime_revert(&a->dt);
4a4d8108
AM
17476+ }
17477+
b912730e
AM
17478+ if (!err && try_aopen && !h_dir->i_op->atomic_open)
17479+ *arg->u.c.aopen->opened |= FILE_CREATED;
17480+
c2b27bf2 17481+ au_unpin(&a->pin);
4a4d8108
AM
17482+ dput(wh_dentry);
17483+
027c5e7a 17484+out_parent:
b912730e
AM
17485+ if (!try_aopen)
17486+ di_write_unlock(parent);
027c5e7a 17487+out_unlock:
4a4d8108
AM
17488+ if (unlikely(err)) {
17489+ au_update_dbstart(dentry);
17490+ d_drop(dentry);
17491+ }
b912730e
AM
17492+ if (!try_aopen)
17493+ aufs_read_unlock(dentry, AuLock_DW);
c2b27bf2
AM
17494+out_free:
17495+ kfree(a);
027c5e7a 17496+out:
4a4d8108
AM
17497+ return err;
17498+}
17499+
7eafdf33
AM
17500+int aufs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
17501+ dev_t dev)
4a4d8108
AM
17502+{
17503+ struct simple_arg arg = {
17504+ .type = Mknod,
17505+ .u.m = {
17506+ .mode = mode,
17507+ .dev = dev
17508+ }
17509+ };
17510+ return add_simple(dir, dentry, &arg);
17511+}
17512+
17513+int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
17514+{
17515+ struct simple_arg arg = {
17516+ .type = Symlink,
17517+ .u.s.symname = symname
17518+ };
17519+ return add_simple(dir, dentry, &arg);
17520+}
17521+
7eafdf33 17522+int aufs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
b4510431 17523+ bool want_excl)
4a4d8108
AM
17524+{
17525+ struct simple_arg arg = {
17526+ .type = Creat,
17527+ .u.c = {
b4510431
AM
17528+ .mode = mode,
17529+ .want_excl = want_excl
4a4d8108
AM
17530+ }
17531+ };
17532+ return add_simple(dir, dentry, &arg);
17533+}
17534+
b912730e
AM
17535+int au_aopen_or_create(struct inode *dir, struct dentry *dentry,
17536+ struct vfsub_aopen_args *aopen_args)
17537+{
17538+ struct simple_arg arg = {
17539+ .type = Creat,
17540+ .u.c = {
17541+ .mode = aopen_args->create_mode,
17542+ .want_excl = aopen_args->open_flag & O_EXCL,
17543+ .try_aopen = true,
17544+ .aopen = aopen_args
17545+ }
17546+ };
17547+ return add_simple(dir, dentry, &arg);
17548+}
17549+
38d290e6
JR
17550+int aufs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
17551+{
17552+ int err;
17553+ aufs_bindex_t bindex;
17554+ struct super_block *sb;
17555+ struct dentry *parent, *h_parent, *h_dentry;
17556+ struct inode *h_dir, *inode;
17557+ struct vfsmount *h_mnt;
17558+ struct au_wr_dir_args wr_dir_args = {
17559+ .force_btgt = -1,
17560+ .flags = AuWrDir_TMPFILE
17561+ };
17562+
17563+ /* copy-up may happen */
17564+ mutex_lock(&dir->i_mutex);
17565+
17566+ sb = dir->i_sb;
17567+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
17568+ if (unlikely(err))
17569+ goto out;
17570+
17571+ err = au_di_init(dentry);
17572+ if (unlikely(err))
17573+ goto out_si;
17574+
17575+ err = -EBUSY;
17576+ parent = d_find_any_alias(dir);
17577+ AuDebugOn(!parent);
17578+ di_write_lock_parent(parent);
5527c038 17579+ if (unlikely(d_inode(parent) != dir))
38d290e6
JR
17580+ goto out_parent;
17581+
17582+ err = au_digen_test(parent, au_sigen(sb));
17583+ if (unlikely(err))
17584+ goto out_parent;
17585+
17586+ bindex = au_dbstart(parent);
17587+ au_set_dbstart(dentry, bindex);
17588+ au_set_dbend(dentry, bindex);
17589+ err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
17590+ bindex = err;
17591+ if (unlikely(err < 0))
17592+ goto out_parent;
17593+
17594+ err = -EOPNOTSUPP;
17595+ h_dir = au_h_iptr(dir, bindex);
17596+ if (unlikely(!h_dir->i_op->tmpfile))
17597+ goto out_parent;
17598+
17599+ h_mnt = au_sbr_mnt(sb, bindex);
17600+ err = vfsub_mnt_want_write(h_mnt);
17601+ if (unlikely(err))
17602+ goto out_parent;
17603+
17604+ h_parent = au_h_dptr(parent, bindex);
5527c038 17605+ err = inode_permission(d_inode(h_parent), MAY_WRITE | MAY_EXEC);
38d290e6
JR
17606+ if (unlikely(err))
17607+ goto out_mnt;
17608+
17609+ err = -ENOMEM;
17610+ h_dentry = d_alloc(h_parent, &dentry->d_name);
17611+ if (unlikely(!h_dentry))
17612+ goto out_mnt;
17613+
17614+ err = h_dir->i_op->tmpfile(h_dir, h_dentry, mode);
17615+ if (unlikely(err))
17616+ goto out_dentry;
17617+
17618+ au_set_dbstart(dentry, bindex);
17619+ au_set_dbend(dentry, bindex);
17620+ au_set_h_dptr(dentry, bindex, dget(h_dentry));
17621+ inode = au_new_inode(dentry, /*must_new*/1);
17622+ if (IS_ERR(inode)) {
17623+ err = PTR_ERR(inode);
17624+ au_set_h_dptr(dentry, bindex, NULL);
17625+ au_set_dbstart(dentry, -1);
17626+ au_set_dbend(dentry, -1);
17627+ } else {
17628+ if (!inode->i_nlink)
17629+ set_nlink(inode, 1);
17630+ d_tmpfile(dentry, inode);
17631+ au_di(dentry)->di_tmpfile = 1;
17632+
17633+ /* update without i_mutex */
17634+ if (au_ibstart(dir) == au_dbstart(dentry))
17635+ au_cpup_attr_timesizes(dir);
17636+ }
17637+
17638+out_dentry:
17639+ dput(h_dentry);
17640+out_mnt:
17641+ vfsub_mnt_drop_write(h_mnt);
17642+out_parent:
17643+ di_write_unlock(parent);
17644+ dput(parent);
17645+ di_write_unlock(dentry);
17646+ if (!err)
17647+#if 0
17648+ /* verbose coding for lock class name */
17649+ au_rw_class(&au_di(dentry)->di_rwsem,
17650+ au_lc_key + AuLcNonDir_DIINFO);
17651+#else
17652+ ;
17653+#endif
17654+ else {
17655+ au_di_fin(dentry);
17656+ dentry->d_fsdata = NULL;
17657+ }
17658+out_si:
17659+ si_read_unlock(sb);
17660+out:
17661+ mutex_unlock(&dir->i_mutex);
17662+ return err;
17663+}
17664+
4a4d8108
AM
17665+/* ---------------------------------------------------------------------- */
17666+
17667+struct au_link_args {
17668+ aufs_bindex_t bdst, bsrc;
17669+ struct au_pin pin;
17670+ struct path h_path;
17671+ struct dentry *src_parent, *parent;
17672+};
17673+
17674+static int au_cpup_before_link(struct dentry *src_dentry,
17675+ struct au_link_args *a)
17676+{
17677+ int err;
17678+ struct dentry *h_src_dentry;
c2b27bf2
AM
17679+ struct au_cp_generic cpg = {
17680+ .dentry = src_dentry,
17681+ .bdst = a->bdst,
17682+ .bsrc = a->bsrc,
17683+ .len = -1,
17684+ .pin = &a->pin,
17685+ .flags = AuCpup_DTIME | AuCpup_HOPEN /* | AuCpup_KEEPLINO */
17686+ };
4a4d8108
AM
17687+
17688+ di_read_lock_parent(a->src_parent, AuLock_IR);
17689+ err = au_test_and_cpup_dirs(src_dentry, a->bdst);
17690+ if (unlikely(err))
17691+ goto out;
17692+
17693+ h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
4a4d8108
AM
17694+ err = au_pin(&a->pin, src_dentry, a->bdst,
17695+ au_opt_udba(src_dentry->d_sb),
17696+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17697+ if (unlikely(err))
17698+ goto out;
367653fa 17699+
c2b27bf2 17700+ err = au_sio_cpup_simple(&cpg);
4a4d8108
AM
17701+ au_unpin(&a->pin);
17702+
4f0767ce 17703+out:
4a4d8108
AM
17704+ di_read_unlock(a->src_parent, AuLock_IR);
17705+ return err;
17706+}
17707+
86dc4139
AM
17708+static int au_cpup_or_link(struct dentry *src_dentry, struct dentry *dentry,
17709+ struct au_link_args *a)
4a4d8108
AM
17710+{
17711+ int err;
17712+ unsigned char plink;
86dc4139 17713+ aufs_bindex_t bend;
4a4d8108 17714+ struct dentry *h_src_dentry;
523b37e3 17715+ struct inode *h_inode, *inode, *delegated;
4a4d8108
AM
17716+ struct super_block *sb;
17717+ struct file *h_file;
17718+
17719+ plink = 0;
17720+ h_inode = NULL;
17721+ sb = src_dentry->d_sb;
5527c038 17722+ inode = d_inode(src_dentry);
4a4d8108
AM
17723+ if (au_ibstart(inode) <= a->bdst)
17724+ h_inode = au_h_iptr(inode, a->bdst);
17725+ if (!h_inode || !h_inode->i_nlink) {
17726+ /* copyup src_dentry as the name of dentry. */
86dc4139
AM
17727+ bend = au_dbend(dentry);
17728+ if (bend < a->bsrc)
17729+ au_set_dbend(dentry, a->bsrc);
17730+ au_set_h_dptr(dentry, a->bsrc,
17731+ dget(au_h_dptr(src_dentry, a->bsrc)));
17732+ dget(a->h_path.dentry);
17733+ au_set_h_dptr(dentry, a->bdst, NULL);
c1595e42
JR
17734+ AuDbg("temporary d_inode...\n");
17735+ spin_lock(&dentry->d_lock);
5527c038 17736+ dentry->d_inode = d_inode(src_dentry); /* tmp */
c1595e42 17737+ spin_unlock(&dentry->d_lock);
392086de 17738+ h_file = au_h_open_pre(dentry, a->bsrc, /*force_wr*/0);
86dc4139 17739+ if (IS_ERR(h_file))
4a4d8108 17740+ err = PTR_ERR(h_file);
86dc4139 17741+ else {
c2b27bf2
AM
17742+ struct au_cp_generic cpg = {
17743+ .dentry = dentry,
17744+ .bdst = a->bdst,
17745+ .bsrc = -1,
17746+ .len = -1,
17747+ .pin = &a->pin,
17748+ .flags = AuCpup_KEEPLINO
17749+ };
17750+ err = au_sio_cpup_simple(&cpg);
86dc4139
AM
17751+ au_h_open_post(dentry, a->bsrc, h_file);
17752+ if (!err) {
17753+ dput(a->h_path.dentry);
17754+ a->h_path.dentry = au_h_dptr(dentry, a->bdst);
17755+ } else
17756+ au_set_h_dptr(dentry, a->bdst,
17757+ a->h_path.dentry);
17758+ }
c1595e42 17759+ spin_lock(&dentry->d_lock);
86dc4139 17760+ dentry->d_inode = NULL; /* restore */
c1595e42
JR
17761+ spin_unlock(&dentry->d_lock);
17762+ AuDbg("temporary d_inode...done\n");
86dc4139
AM
17763+ au_set_h_dptr(dentry, a->bsrc, NULL);
17764+ au_set_dbend(dentry, bend);
4a4d8108
AM
17765+ } else {
17766+ /* the inode of src_dentry already exists on a.bdst branch */
17767+ h_src_dentry = d_find_alias(h_inode);
17768+ if (!h_src_dentry && au_plink_test(inode)) {
17769+ plink = 1;
17770+ h_src_dentry = au_plink_lkup(inode, a->bdst);
17771+ err = PTR_ERR(h_src_dentry);
17772+ if (IS_ERR(h_src_dentry))
17773+ goto out;
17774+
5527c038 17775+ if (unlikely(d_is_negative(h_src_dentry))) {
4a4d8108
AM
17776+ dput(h_src_dentry);
17777+ h_src_dentry = NULL;
17778+ }
17779+
17780+ }
17781+ if (h_src_dentry) {
523b37e3 17782+ delegated = NULL;
4a4d8108 17783+ err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
523b37e3
AM
17784+ &a->h_path, &delegated);
17785+ if (unlikely(err == -EWOULDBLOCK)) {
17786+ pr_warn("cannot retry for NFSv4 delegation"
17787+ " for an internal link\n");
17788+ iput(delegated);
17789+ }
4a4d8108
AM
17790+ dput(h_src_dentry);
17791+ } else {
17792+ AuIOErr("no dentry found for hi%lu on b%d\n",
17793+ h_inode->i_ino, a->bdst);
17794+ err = -EIO;
17795+ }
17796+ }
17797+
17798+ if (!err && !plink)
17799+ au_plink_append(inode, a->bdst, a->h_path.dentry);
17800+
17801+out:
2cbb1c4b 17802+ AuTraceErr(err);
4a4d8108
AM
17803+ return err;
17804+}
17805+
17806+int aufs_link(struct dentry *src_dentry, struct inode *dir,
17807+ struct dentry *dentry)
17808+{
17809+ int err, rerr;
17810+ struct au_dtime dt;
17811+ struct au_link_args *a;
17812+ struct dentry *wh_dentry, *h_src_dentry;
523b37e3 17813+ struct inode *inode, *delegated;
4a4d8108
AM
17814+ struct super_block *sb;
17815+ struct au_wr_dir_args wr_dir_args = {
17816+ /* .force_btgt = -1, */
17817+ .flags = AuWrDir_ADD_ENTRY
17818+ };
17819+
17820+ IMustLock(dir);
5527c038 17821+ inode = d_inode(src_dentry);
4a4d8108
AM
17822+ IMustLock(inode);
17823+
4a4d8108
AM
17824+ err = -ENOMEM;
17825+ a = kzalloc(sizeof(*a), GFP_NOFS);
17826+ if (unlikely(!a))
17827+ goto out;
17828+
17829+ a->parent = dentry->d_parent; /* dir inode is locked */
027c5e7a
AM
17830+ err = aufs_read_and_write_lock2(dentry, src_dentry,
17831+ AuLock_NOPLM | AuLock_GEN);
e49829fe
JR
17832+ if (unlikely(err))
17833+ goto out_kfree;
38d290e6 17834+ err = au_d_linkable(src_dentry);
027c5e7a
AM
17835+ if (unlikely(err))
17836+ goto out_unlock;
17837+ err = au_d_may_add(dentry);
17838+ if (unlikely(err))
17839+ goto out_unlock;
e49829fe 17840+
4a4d8108 17841+ a->src_parent = dget_parent(src_dentry);
2cbb1c4b 17842+ wr_dir_args.force_btgt = au_ibstart(inode);
4a4d8108
AM
17843+
17844+ di_write_lock_parent(a->parent);
17845+ wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
17846+ wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
17847+ &wr_dir_args);
17848+ err = PTR_ERR(wh_dentry);
17849+ if (IS_ERR(wh_dentry))
027c5e7a 17850+ goto out_parent;
4a4d8108
AM
17851+
17852+ err = 0;
17853+ sb = dentry->d_sb;
17854+ a->bdst = au_dbstart(dentry);
17855+ a->h_path.dentry = au_h_dptr(dentry, a->bdst);
17856+ a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
2cbb1c4b
JR
17857+ a->bsrc = au_ibstart(inode);
17858+ h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
38d290e6
JR
17859+ if (!h_src_dentry && au_di(src_dentry)->di_tmpfile)
17860+ h_src_dentry = dget(au_hi_wh(inode, a->bsrc));
2cbb1c4b
JR
17861+ if (!h_src_dentry) {
17862+ a->bsrc = au_dbstart(src_dentry);
17863+ h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
17864+ AuDebugOn(!h_src_dentry);
38d290e6
JR
17865+ } else if (IS_ERR(h_src_dentry)) {
17866+ err = PTR_ERR(h_src_dentry);
2cbb1c4b 17867+ goto out_parent;
38d290e6 17868+ }
2cbb1c4b 17869+
4a4d8108
AM
17870+ if (au_opt_test(au_mntflags(sb), PLINK)) {
17871+ if (a->bdst < a->bsrc
17872+ /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
86dc4139 17873+ err = au_cpup_or_link(src_dentry, dentry, a);
523b37e3
AM
17874+ else {
17875+ delegated = NULL;
4a4d8108 17876+ err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
523b37e3
AM
17877+ &a->h_path, &delegated);
17878+ if (unlikely(err == -EWOULDBLOCK)) {
17879+ pr_warn("cannot retry for NFSv4 delegation"
17880+ " for an internal link\n");
17881+ iput(delegated);
17882+ }
17883+ }
2cbb1c4b 17884+ dput(h_src_dentry);
4a4d8108
AM
17885+ } else {
17886+ /*
17887+ * copyup src_dentry to the branch we process,
17888+ * and then link(2) to it.
17889+ */
2cbb1c4b 17890+ dput(h_src_dentry);
4a4d8108
AM
17891+ if (a->bdst < a->bsrc
17892+ /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
17893+ au_unpin(&a->pin);
17894+ di_write_unlock(a->parent);
17895+ err = au_cpup_before_link(src_dentry, a);
17896+ di_write_lock_parent(a->parent);
17897+ if (!err)
17898+ err = au_pin(&a->pin, dentry, a->bdst,
17899+ au_opt_udba(sb),
17900+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17901+ if (unlikely(err))
17902+ goto out_wh;
17903+ }
17904+ if (!err) {
17905+ h_src_dentry = au_h_dptr(src_dentry, a->bdst);
17906+ err = -ENOENT;
5527c038 17907+ if (h_src_dentry && d_is_positive(h_src_dentry)) {
523b37e3 17908+ delegated = NULL;
4a4d8108
AM
17909+ err = vfsub_link(h_src_dentry,
17910+ au_pinned_h_dir(&a->pin),
523b37e3
AM
17911+ &a->h_path, &delegated);
17912+ if (unlikely(err == -EWOULDBLOCK)) {
17913+ pr_warn("cannot retry"
17914+ " for NFSv4 delegation"
17915+ " for an internal link\n");
17916+ iput(delegated);
17917+ }
17918+ }
4a4d8108
AM
17919+ }
17920+ }
17921+ if (unlikely(err))
17922+ goto out_unpin;
17923+
17924+ if (wh_dentry) {
17925+ a->h_path.dentry = wh_dentry;
17926+ err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
17927+ dentry);
17928+ if (unlikely(err))
17929+ goto out_revert;
17930+ }
17931+
b912730e 17932+ au_dir_ts(dir, a->bdst);
4a4d8108 17933+ dir->i_version++;
4a4d8108
AM
17934+ inc_nlink(inode);
17935+ inode->i_ctime = dir->i_ctime;
027c5e7a
AM
17936+ d_instantiate(dentry, au_igrab(inode));
17937+ if (d_unhashed(a->h_path.dentry))
4a4d8108
AM
17938+ /* some filesystem calls d_drop() */
17939+ d_drop(dentry);
076b876e
AM
17940+ /* some filesystems consume an inode even hardlink */
17941+ au_fhsm_wrote(sb, a->bdst, /*force*/0);
4a4d8108
AM
17942+ goto out_unpin; /* success */
17943+
4f0767ce 17944+out_revert:
523b37e3
AM
17945+ /* no delegation since it is just created */
17946+ rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path,
17947+ /*delegated*/NULL, /*force*/0);
027c5e7a 17948+ if (unlikely(rerr)) {
523b37e3 17949+ AuIOErr("%pd reverting failed(%d, %d)\n", dentry, err, rerr);
027c5e7a
AM
17950+ err = -EIO;
17951+ }
4a4d8108 17952+ au_dtime_revert(&dt);
4f0767ce 17953+out_unpin:
4a4d8108 17954+ au_unpin(&a->pin);
4f0767ce 17955+out_wh:
4a4d8108 17956+ dput(wh_dentry);
027c5e7a
AM
17957+out_parent:
17958+ di_write_unlock(a->parent);
17959+ dput(a->src_parent);
4f0767ce 17960+out_unlock:
4a4d8108
AM
17961+ if (unlikely(err)) {
17962+ au_update_dbstart(dentry);
17963+ d_drop(dentry);
17964+ }
4a4d8108 17965+ aufs_read_and_write_unlock2(dentry, src_dentry);
e49829fe 17966+out_kfree:
4a4d8108 17967+ kfree(a);
4f0767ce 17968+out:
86dc4139 17969+ AuTraceErr(err);
4a4d8108
AM
17970+ return err;
17971+}
17972+
7eafdf33 17973+int aufs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4a4d8108
AM
17974+{
17975+ int err, rerr;
17976+ aufs_bindex_t bindex;
17977+ unsigned char diropq;
17978+ struct path h_path;
17979+ struct dentry *wh_dentry, *parent, *opq_dentry;
17980+ struct mutex *h_mtx;
17981+ struct super_block *sb;
17982+ struct {
17983+ struct au_pin pin;
17984+ struct au_dtime dt;
17985+ } *a; /* reduce the stack usage */
17986+ struct au_wr_dir_args wr_dir_args = {
17987+ .force_btgt = -1,
17988+ .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
17989+ };
17990+
17991+ IMustLock(dir);
17992+
17993+ err = -ENOMEM;
17994+ a = kmalloc(sizeof(*a), GFP_NOFS);
17995+ if (unlikely(!a))
17996+ goto out;
17997+
027c5e7a
AM
17998+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
17999+ if (unlikely(err))
18000+ goto out_free;
18001+ err = au_d_may_add(dentry);
18002+ if (unlikely(err))
18003+ goto out_unlock;
18004+
4a4d8108
AM
18005+ parent = dentry->d_parent; /* dir inode is locked */
18006+ di_write_lock_parent(parent);
18007+ wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
18008+ &a->pin, &wr_dir_args);
18009+ err = PTR_ERR(wh_dentry);
18010+ if (IS_ERR(wh_dentry))
027c5e7a 18011+ goto out_parent;
4a4d8108
AM
18012+
18013+ sb = dentry->d_sb;
18014+ bindex = au_dbstart(dentry);
18015+ h_path.dentry = au_h_dptr(dentry, bindex);
18016+ h_path.mnt = au_sbr_mnt(sb, bindex);
18017+ err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
18018+ if (unlikely(err))
027c5e7a 18019+ goto out_unpin;
4a4d8108
AM
18020+
18021+ /* make the dir opaque */
18022+ diropq = 0;
5527c038 18023+ h_mtx = &d_inode(h_path.dentry)->i_mutex;
4a4d8108
AM
18024+ if (wh_dentry
18025+ || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
18026+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18027+ opq_dentry = au_diropq_create(dentry, bindex);
18028+ mutex_unlock(h_mtx);
18029+ err = PTR_ERR(opq_dentry);
18030+ if (IS_ERR(opq_dentry))
18031+ goto out_dir;
18032+ dput(opq_dentry);
18033+ diropq = 1;
18034+ }
18035+
18036+ err = epilog(dir, bindex, wh_dentry, dentry);
18037+ if (!err) {
18038+ inc_nlink(dir);
027c5e7a 18039+ goto out_unpin; /* success */
4a4d8108
AM
18040+ }
18041+
18042+ /* revert */
18043+ if (diropq) {
18044+ AuLabel(revert opq);
18045+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18046+ rerr = au_diropq_remove(dentry, bindex);
18047+ mutex_unlock(h_mtx);
18048+ if (rerr) {
523b37e3
AM
18049+ AuIOErr("%pd reverting diropq failed(%d, %d)\n",
18050+ dentry, err, rerr);
4a4d8108
AM
18051+ err = -EIO;
18052+ }
18053+ }
18054+
4f0767ce 18055+out_dir:
4a4d8108
AM
18056+ AuLabel(revert dir);
18057+ rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
18058+ if (rerr) {
523b37e3
AM
18059+ AuIOErr("%pd reverting dir failed(%d, %d)\n",
18060+ dentry, err, rerr);
4a4d8108
AM
18061+ err = -EIO;
18062+ }
4a4d8108 18063+ au_dtime_revert(&a->dt);
027c5e7a 18064+out_unpin:
4a4d8108
AM
18065+ au_unpin(&a->pin);
18066+ dput(wh_dentry);
027c5e7a
AM
18067+out_parent:
18068+ di_write_unlock(parent);
18069+out_unlock:
4a4d8108
AM
18070+ if (unlikely(err)) {
18071+ au_update_dbstart(dentry);
18072+ d_drop(dentry);
18073+ }
4a4d8108 18074+ aufs_read_unlock(dentry, AuLock_DW);
027c5e7a 18075+out_free:
4a4d8108 18076+ kfree(a);
4f0767ce 18077+out:
4a4d8108
AM
18078+ return err;
18079+}
7f207e10
AM
18080diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
18081--- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c
AM
18082+++ linux/fs/aufs/i_op.c 2015-09-24 10:47:58.254719746 +0200
18083@@ -0,0 +1,1483 @@
4a4d8108 18084+/*
2000de60 18085+ * Copyright (C) 2005-2015 Junjiro R. Okajima
4a4d8108
AM
18086+ *
18087+ * This program, aufs is free software; you can redistribute it and/or modify
18088+ * it under the terms of the GNU General Public License as published by
18089+ * the Free Software Foundation; either version 2 of the License, or
18090+ * (at your option) any later version.
18091+ *
18092+ * This program is distributed in the hope that it will be useful,
18093+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
18094+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18095+ * GNU General Public License for more details.
18096+ *
18097+ * You should have received a copy of the GNU General Public License
523b37e3 18098+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
4a4d8108 18099+ */
1facf9fc 18100+
1308ab2a 18101+/*
4a4d8108 18102+ * inode operations (except add/del/rename)
1308ab2a 18103+ */
4a4d8108
AM
18104+
18105+#include <linux/device_cgroup.h>
18106+#include <linux/fs_stack.h>
92d182d2 18107+#include <linux/mm.h>
4a4d8108
AM
18108+#include <linux/namei.h>
18109+#include <linux/security.h>
4a4d8108
AM
18110+#include "aufs.h"
18111+
1e00d052 18112+static int h_permission(struct inode *h_inode, int mask,
4a4d8108 18113+ struct vfsmount *h_mnt, int brperm)
1facf9fc 18114+{
1308ab2a 18115+ int err;
4a4d8108 18116+ const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
1facf9fc 18117+
4a4d8108
AM
18118+ err = -EACCES;
18119+ if ((write_mask && IS_IMMUTABLE(h_inode))
18120+ || ((mask & MAY_EXEC)
18121+ && S_ISREG(h_inode->i_mode)
18122+ && ((h_mnt->mnt_flags & MNT_NOEXEC)
18123+ || !(h_inode->i_mode & S_IXUGO))))
18124+ goto out;
18125+
18126+ /*
18127+ * - skip the lower fs test in the case of write to ro branch.
18128+ * - nfs dir permission write check is optimized, but a policy for
18129+ * link/rename requires a real check.
b912730e
AM
18130+ * - nfs always sets MS_POSIXACL regardless its mount option 'noacl.'
18131+ * in this case, generic_permission() returns -EOPNOTSUPP.
4a4d8108
AM
18132+ */
18133+ if ((write_mask && !au_br_writable(brperm))
18134+ || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
18135+ && write_mask && !(mask & MAY_READ))
18136+ || !h_inode->i_op->permission) {
18137+ /* AuLabel(generic_permission); */
b912730e 18138+ /* AuDbg("get_acl %pf\n", h_inode->i_op->get_acl); */
1e00d052 18139+ err = generic_permission(h_inode, mask);
b912730e
AM
18140+ if (err == -EOPNOTSUPP && au_test_nfs_noacl(h_inode))
18141+ err = h_inode->i_op->permission(h_inode, mask);
18142+ AuTraceErr(err);
1308ab2a 18143+ } else {
4a4d8108 18144+ /* AuLabel(h_inode->permission); */
1e00d052 18145+ err = h_inode->i_op->permission(h_inode, mask);
4a4d8108
AM
18146+ AuTraceErr(err);
18147+ }
1facf9fc 18148+
4a4d8108
AM
18149+ if (!err)
18150+ err = devcgroup_inode_permission(h_inode, mask);
7f207e10 18151+ if (!err)
4a4d8108 18152+ err = security_inode_permission(h_inode, mask);
4a4d8108
AM
18153+
18154+#if 0
18155+ if (!err) {
18156+ /* todo: do we need to call ima_path_check()? */
18157+ struct path h_path = {
18158+ .dentry =
18159+ .mnt = h_mnt
18160+ };
18161+ err = ima_path_check(&h_path,
18162+ mask & (MAY_READ | MAY_WRITE | MAY_EXEC),
18163+ IMA_COUNT_LEAVE);
1308ab2a 18164+ }
4a4d8108 18165+#endif
dece6358 18166+
4f0767ce 18167+out:
1308ab2a 18168+ return err;
18169+}
dece6358 18170+
1e00d052 18171+static int aufs_permission(struct inode *inode, int mask)
1308ab2a 18172+{
18173+ int err;
4a4d8108
AM
18174+ aufs_bindex_t bindex, bend;
18175+ const unsigned char isdir = !!S_ISDIR(inode->i_mode),
18176+ write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
18177+ struct inode *h_inode;
18178+ struct super_block *sb;
18179+ struct au_branch *br;
1facf9fc 18180+
027c5e7a 18181+ /* todo: support rcu-walk? */
1e00d052 18182+ if (mask & MAY_NOT_BLOCK)
027c5e7a
AM
18183+ return -ECHILD;
18184+
4a4d8108
AM
18185+ sb = inode->i_sb;
18186+ si_read_lock(sb, AuLock_FLUSH);
18187+ ii_read_lock_child(inode);
027c5e7a
AM
18188+#if 0
18189+ err = au_iigen_test(inode, au_sigen(sb));
18190+ if (unlikely(err))
18191+ goto out;
18192+#endif
dece6358 18193+
076b876e
AM
18194+ if (!isdir
18195+ || write_mask
18196+ || au_opt_test(au_mntflags(sb), DIRPERM1)) {
4a4d8108
AM
18197+ err = au_busy_or_stale();
18198+ h_inode = au_h_iptr(inode, au_ibstart(inode));
18199+ if (unlikely(!h_inode
18200+ || (h_inode->i_mode & S_IFMT)
18201+ != (inode->i_mode & S_IFMT)))
18202+ goto out;
1facf9fc 18203+
4a4d8108
AM
18204+ err = 0;
18205+ bindex = au_ibstart(inode);
18206+ br = au_sbr(sb, bindex);
86dc4139 18207+ err = h_permission(h_inode, mask, au_br_mnt(br), br->br_perm);
4a4d8108
AM
18208+ if (write_mask
18209+ && !err
18210+ && !special_file(h_inode->i_mode)) {
18211+ /* test whether the upper writable branch exists */
18212+ err = -EROFS;
18213+ for (; bindex >= 0; bindex--)
18214+ if (!au_br_rdonly(au_sbr(sb, bindex))) {
18215+ err = 0;
18216+ break;
18217+ }
18218+ }
18219+ goto out;
18220+ }
dece6358 18221+
4a4d8108 18222+ /* non-write to dir */
1308ab2a 18223+ err = 0;
4a4d8108
AM
18224+ bend = au_ibend(inode);
18225+ for (bindex = au_ibstart(inode); !err && bindex <= bend; bindex++) {
18226+ h_inode = au_h_iptr(inode, bindex);
18227+ if (h_inode) {
18228+ err = au_busy_or_stale();
18229+ if (unlikely(!S_ISDIR(h_inode->i_mode)))
18230+ break;
18231+
18232+ br = au_sbr(sb, bindex);
86dc4139 18233+ err = h_permission(h_inode, mask, au_br_mnt(br),
4a4d8108
AM
18234+ br->br_perm);
18235+ }
18236+ }
1308ab2a 18237+
4f0767ce 18238+out:
4a4d8108
AM
18239+ ii_read_unlock(inode);
18240+ si_read_unlock(sb);
1308ab2a 18241+ return err;
18242+}
18243+
4a4d8108 18244+/* ---------------------------------------------------------------------- */
1facf9fc 18245+
4a4d8108 18246+static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
b4510431 18247+ unsigned int flags)
4a4d8108
AM
18248+{
18249+ struct dentry *ret, *parent;
b752ccd1 18250+ struct inode *inode;
4a4d8108 18251+ struct super_block *sb;
1716fcea 18252+ int err, npositive;
dece6358 18253+
4a4d8108 18254+ IMustLock(dir);
1308ab2a 18255+
537831f9
AM
18256+ /* todo: support rcu-walk? */
18257+ ret = ERR_PTR(-ECHILD);
18258+ if (flags & LOOKUP_RCU)
18259+ goto out;
18260+
18261+ ret = ERR_PTR(-ENAMETOOLONG);
18262+ if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
18263+ goto out;
18264+
4a4d8108 18265+ sb = dir->i_sb;
7f207e10
AM
18266+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
18267+ ret = ERR_PTR(err);
18268+ if (unlikely(err))
18269+ goto out;
18270+
4a4d8108
AM
18271+ err = au_di_init(dentry);
18272+ ret = ERR_PTR(err);
18273+ if (unlikely(err))
7f207e10 18274+ goto out_si;
1308ab2a 18275+
9dbd164d 18276+ inode = NULL;
027c5e7a 18277+ npositive = 0; /* suppress a warning */
4a4d8108
AM
18278+ parent = dentry->d_parent; /* dir inode is locked */
18279+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a
AM
18280+ err = au_alive_dir(parent);
18281+ if (!err)
18282+ err = au_digen_test(parent, au_sigen(sb));
18283+ if (!err) {
18284+ npositive = au_lkup_dentry(dentry, au_dbstart(parent),
537831f9 18285+ /*type*/0);
027c5e7a
AM
18286+ err = npositive;
18287+ }
4a4d8108 18288+ di_read_unlock(parent, AuLock_IR);
4a4d8108
AM
18289+ ret = ERR_PTR(err);
18290+ if (unlikely(err < 0))
18291+ goto out_unlock;
1308ab2a 18292+
4a4d8108 18293+ if (npositive) {
b752ccd1 18294+ inode = au_new_inode(dentry, /*must_new*/0);
c1595e42
JR
18295+ if (IS_ERR(inode)) {
18296+ ret = (void *)inode;
18297+ inode = NULL;
18298+ goto out_unlock;
18299+ }
9dbd164d 18300+ }
4a4d8108 18301+
c1595e42
JR
18302+ if (inode)
18303+ atomic_inc(&inode->i_count);
4a4d8108 18304+ ret = d_splice_alias(inode, dentry);
537831f9
AM
18305+#if 0
18306+ if (unlikely(d_need_lookup(dentry))) {
18307+ spin_lock(&dentry->d_lock);
18308+ dentry->d_flags &= ~DCACHE_NEED_LOOKUP;
18309+ spin_unlock(&dentry->d_lock);
18310+ } else
18311+#endif
c1595e42 18312+ if (inode) {
2000de60 18313+ if (!IS_ERR(ret)) {
c1595e42 18314+ iput(inode);
2000de60
JR
18315+ if (ret && ret != dentry)
18316+ ii_write_unlock(inode);
18317+ } else {
c1595e42
JR
18318+ ii_write_unlock(inode);
18319+ iput(inode);
18320+ inode = NULL;
18321+ }
7f207e10 18322+ }
1facf9fc 18323+
4f0767ce 18324+out_unlock:
4a4d8108 18325+ di_write_unlock(dentry);
2dfbb274 18326+ if (inode) {
1716fcea
AM
18327+ /* verbose coding for lock class name */
18328+ if (unlikely(S_ISLNK(inode->i_mode)))
18329+ au_rw_class(&au_di(dentry)->di_rwsem,
18330+ au_lc_key + AuLcSymlink_DIINFO);
18331+ else if (unlikely(S_ISDIR(inode->i_mode)))
18332+ au_rw_class(&au_di(dentry)->di_rwsem,
18333+ au_lc_key + AuLcDir_DIINFO);
18334+ else /* likely */
18335+ au_rw_class(&au_di(dentry)->di_rwsem,
18336+ au_lc_key + AuLcNonDir_DIINFO);
9dbd164d 18337+ }
7f207e10 18338+out_si:
4a4d8108 18339+ si_read_unlock(sb);
7f207e10 18340+out:
4a4d8108
AM
18341+ return ret;
18342+}
1facf9fc 18343+
4a4d8108 18344+/* ---------------------------------------------------------------------- */
1facf9fc 18345+
b912730e
AM
18346+struct aopen_node {
18347+ struct hlist_node hlist;
18348+ struct file *file, *h_file;
18349+};
18350+
18351+static int au_do_aopen(struct inode *inode, struct file *file)
18352+{
18353+ struct au_sphlhead *aopen;
18354+ struct aopen_node *node;
18355+ struct au_do_open_args args = {
18356+ .no_lock = 1,
18357+ .open = au_do_open_nondir
18358+ };
18359+
18360+ aopen = &au_sbi(inode->i_sb)->si_aopen;
18361+ spin_lock(&aopen->spin);
18362+ hlist_for_each_entry(node, &aopen->head, hlist)
18363+ if (node->file == file) {
18364+ args.h_file = node->h_file;
18365+ break;
18366+ }
18367+ spin_unlock(&aopen->spin);
18368+ /* AuDebugOn(!args.h_file); */
18369+
18370+ return au_do_open(file, &args);
18371+}
18372+
18373+static int aufs_atomic_open(struct inode *dir, struct dentry *dentry,
18374+ struct file *file, unsigned int open_flag,
18375+ umode_t create_mode, int *opened)
18376+{
18377+ int err, h_opened = *opened;
18378+ struct dentry *parent;
18379+ struct dentry *d;
18380+ struct au_sphlhead *aopen;
18381+ struct vfsub_aopen_args args = {
18382+ .open_flag = open_flag,
18383+ .create_mode = create_mode,
18384+ .opened = &h_opened
18385+ };
18386+ struct aopen_node aopen_node = {
18387+ .file = file
18388+ };
18389+
18390+ IMustLock(dir);
18391+ AuDbg("open_flag 0x%x\n", open_flag);
18392+ AuDbgDentry(dentry);
18393+
18394+ err = 0;
18395+ if (!au_di(dentry)) {
18396+ d = aufs_lookup(dir, dentry, /*flags*/0);
18397+ if (IS_ERR(d)) {
18398+ err = PTR_ERR(d);
18399+ goto out;
18400+ } else if (d) {
18401+ /*
18402+ * obsoleted dentry found.
18403+ * another error will be returned later.
18404+ */
18405+ d_drop(d);
18406+ dput(d);
18407+ AuDbgDentry(d);
18408+ }
18409+ AuDbgDentry(dentry);
18410+ }
18411+
18412+ if (d_is_positive(dentry)
18413+ || d_unhashed(dentry)
18414+ || d_unlinked(dentry)
18415+ || !(open_flag & O_CREAT))
18416+ goto out_no_open;
18417+
18418+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
18419+ if (unlikely(err))
18420+ goto out;
18421+
18422+ parent = dentry->d_parent; /* dir is locked */
18423+ di_write_lock_parent(parent);
18424+ err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0);
18425+ if (unlikely(err))
18426+ goto out_unlock;
18427+
18428+ AuDbgDentry(dentry);
18429+ if (d_is_positive(dentry))
18430+ goto out_unlock;
18431+
18432+ args.file = get_empty_filp();
18433+ err = PTR_ERR(args.file);
18434+ if (IS_ERR(args.file))
18435+ goto out_unlock;
18436+
18437+ args.file->f_flags = file->f_flags;
18438+ err = au_aopen_or_create(dir, dentry, &args);
18439+ AuTraceErr(err);
18440+ AuDbgFile(args.file);
18441+ if (unlikely(err < 0)) {
18442+ if (h_opened & FILE_OPENED)
18443+ fput(args.file);
18444+ else
18445+ put_filp(args.file);
18446+ goto out_unlock;
18447+ }
18448+
18449+ /* some filesystems don't set FILE_CREATED while succeeded? */
18450+ *opened |= FILE_CREATED;
18451+ if (h_opened & FILE_OPENED)
18452+ aopen_node.h_file = args.file;
18453+ else {
18454+ put_filp(args.file);
18455+ args.file = NULL;
18456+ }
18457+ aopen = &au_sbi(dir->i_sb)->si_aopen;
18458+ au_sphl_add(&aopen_node.hlist, aopen);
18459+ err = finish_open(file, dentry, au_do_aopen, opened);
18460+ au_sphl_del(&aopen_node.hlist, aopen);
18461+ AuTraceErr(err);
18462+ AuDbgFile(file);
18463+ if (aopen_node.h_file)
18464+ fput(aopen_node.h_file);
18465+
18466+out_unlock:
18467+ di_write_unlock(parent);
18468+ aufs_read_unlock(dentry, AuLock_DW);
18469+ AuDbgDentry(dentry);
18470+ if (unlikely(err))
18471+ goto out;
18472+out_no_open:
18473+ if (!err && !(*opened & FILE_CREATED)) {
18474+ AuLabel(out_no_open);
18475+ dget(dentry);
18476+ err = finish_no_open(file, dentry);
18477+ }
18478+out:
18479+ AuDbg("%pd%s%s\n", dentry,
18480+ (*opened & FILE_CREATED) ? " created" : "",
18481+ (*opened & FILE_OPENED) ? " opened" : "");
18482+ AuTraceErr(err);
18483+ return err;
18484+}
18485+
18486+
18487+/* ---------------------------------------------------------------------- */
18488+
4a4d8108
AM
18489+static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
18490+ const unsigned char add_entry, aufs_bindex_t bcpup,
18491+ aufs_bindex_t bstart)
18492+{
18493+ int err;
18494+ struct dentry *h_parent;
18495+ struct inode *h_dir;
1facf9fc 18496+
027c5e7a 18497+ if (add_entry)
5527c038 18498+ IMustLock(d_inode(parent));
027c5e7a 18499+ else
4a4d8108
AM
18500+ di_write_lock_parent(parent);
18501+
18502+ err = 0;
18503+ if (!au_h_dptr(parent, bcpup)) {
c2b27bf2
AM
18504+ if (bstart > bcpup)
18505+ err = au_cpup_dirs(dentry, bcpup);
18506+ else if (bstart < bcpup)
4a4d8108
AM
18507+ err = au_cpdown_dirs(dentry, bcpup);
18508+ else
c2b27bf2 18509+ BUG();
4a4d8108 18510+ }
38d290e6 18511+ if (!err && add_entry && !au_ftest_wrdir(add_entry, TMPFILE)) {
4a4d8108 18512+ h_parent = au_h_dptr(parent, bcpup);
5527c038 18513+ h_dir = d_inode(h_parent);
4a4d8108 18514+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
7e9cd9fe 18515+ err = au_lkup_neg(dentry, bcpup, /*wh*/0);
4a4d8108
AM
18516+ /* todo: no unlock here */
18517+ mutex_unlock(&h_dir->i_mutex);
027c5e7a
AM
18518+
18519+ AuDbg("bcpup %d\n", bcpup);
18520+ if (!err) {
5527c038 18521+ if (d_really_is_negative(dentry))
027c5e7a 18522+ au_set_h_dptr(dentry, bstart, NULL);
4a4d8108
AM
18523+ au_update_dbrange(dentry, /*do_put_zero*/0);
18524+ }
1308ab2a 18525+ }
1facf9fc 18526+
4a4d8108
AM
18527+ if (!add_entry)
18528+ di_write_unlock(parent);
18529+ if (!err)
18530+ err = bcpup; /* success */
1308ab2a 18531+
027c5e7a 18532+ AuTraceErr(err);
4a4d8108
AM
18533+ return err;
18534+}
1facf9fc 18535+
4a4d8108
AM
18536+/*
18537+ * decide the branch and the parent dir where we will create a new entry.
18538+ * returns new bindex or an error.
18539+ * copyup the parent dir if needed.
18540+ */
18541+int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
18542+ struct au_wr_dir_args *args)
18543+{
18544+ int err;
392086de 18545+ unsigned int flags;
4a4d8108 18546+ aufs_bindex_t bcpup, bstart, src_bstart;
86dc4139
AM
18547+ const unsigned char add_entry
18548+ = au_ftest_wrdir(args->flags, ADD_ENTRY)
38d290e6 18549+ | au_ftest_wrdir(args->flags, TMPFILE);
4a4d8108
AM
18550+ struct super_block *sb;
18551+ struct dentry *parent;
18552+ struct au_sbinfo *sbinfo;
1facf9fc 18553+
4a4d8108
AM
18554+ sb = dentry->d_sb;
18555+ sbinfo = au_sbi(sb);
18556+ parent = dget_parent(dentry);
18557+ bstart = au_dbstart(dentry);
18558+ bcpup = bstart;
18559+ if (args->force_btgt < 0) {
18560+ if (src_dentry) {
18561+ src_bstart = au_dbstart(src_dentry);
18562+ if (src_bstart < bstart)
18563+ bcpup = src_bstart;
18564+ } else if (add_entry) {
392086de
AM
18565+ flags = 0;
18566+ if (au_ftest_wrdir(args->flags, ISDIR))
18567+ au_fset_wbr(flags, DIR);
18568+ err = AuWbrCreate(sbinfo, dentry, flags);
4a4d8108
AM
18569+ bcpup = err;
18570+ }
1facf9fc 18571+
5527c038 18572+ if (bcpup < 0 || au_test_ro(sb, bcpup, d_inode(dentry))) {
4a4d8108
AM
18573+ if (add_entry)
18574+ err = AuWbrCopyup(sbinfo, dentry);
18575+ else {
18576+ if (!IS_ROOT(dentry)) {
18577+ di_read_lock_parent(parent, !AuLock_IR);
18578+ err = AuWbrCopyup(sbinfo, dentry);
18579+ di_read_unlock(parent, !AuLock_IR);
18580+ } else
18581+ err = AuWbrCopyup(sbinfo, dentry);
18582+ }
18583+ bcpup = err;
18584+ if (unlikely(err < 0))
18585+ goto out;
18586+ }
18587+ } else {
18588+ bcpup = args->force_btgt;
5527c038 18589+ AuDebugOn(au_test_ro(sb, bcpup, d_inode(dentry)));
1308ab2a 18590+ }
027c5e7a 18591+
4a4d8108
AM
18592+ AuDbg("bstart %d, bcpup %d\n", bstart, bcpup);
18593+ err = bcpup;
18594+ if (bcpup == bstart)
18595+ goto out; /* success */
4a4d8108
AM
18596+
18597+ /* copyup the new parent into the branch we process */
18598+ err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, bstart);
027c5e7a 18599+ if (err >= 0) {
5527c038 18600+ if (d_really_is_negative(dentry)) {
027c5e7a
AM
18601+ au_set_h_dptr(dentry, bstart, NULL);
18602+ au_set_dbstart(dentry, bcpup);
18603+ au_set_dbend(dentry, bcpup);
18604+ }
38d290e6
JR
18605+ AuDebugOn(add_entry
18606+ && !au_ftest_wrdir(args->flags, TMPFILE)
18607+ && !au_h_dptr(dentry, bcpup));
027c5e7a 18608+ }
86dc4139
AM
18609+
18610+out:
18611+ dput(parent);
18612+ return err;
18613+}
18614+
18615+/* ---------------------------------------------------------------------- */
18616+
18617+void au_pin_hdir_unlock(struct au_pin *p)
18618+{
18619+ if (p->hdir)
18620+ au_hn_imtx_unlock(p->hdir);
18621+}
18622+
c1595e42 18623+int au_pin_hdir_lock(struct au_pin *p)
86dc4139
AM
18624+{
18625+ int err;
18626+
18627+ err = 0;
18628+ if (!p->hdir)
18629+ goto out;
18630+
18631+ /* even if an error happens later, keep this lock */
18632+ au_hn_imtx_lock_nested(p->hdir, p->lsc_hi);
18633+
18634+ err = -EBUSY;
5527c038 18635+ if (unlikely(p->hdir->hi_inode != d_inode(p->h_parent)))
86dc4139
AM
18636+ goto out;
18637+
18638+ err = 0;
18639+ if (p->h_dentry)
18640+ err = au_h_verify(p->h_dentry, p->udba, p->hdir->hi_inode,
18641+ p->h_parent, p->br);
18642+
18643+out:
18644+ return err;
18645+}
18646+
18647+int au_pin_hdir_relock(struct au_pin *p)
18648+{
18649+ int err, i;
18650+ struct inode *h_i;
18651+ struct dentry *h_d[] = {
18652+ p->h_dentry,
18653+ p->h_parent
18654+ };
18655+
18656+ err = au_pin_hdir_lock(p);
18657+ if (unlikely(err))
18658+ goto out;
18659+
18660+ for (i = 0; !err && i < sizeof(h_d)/sizeof(*h_d); i++) {
18661+ if (!h_d[i])
18662+ continue;
5527c038
JR
18663+ if (d_is_positive(h_d[i])) {
18664+ h_i = d_inode(h_d[i]);
86dc4139 18665+ err = !h_i->i_nlink;
5527c038 18666+ }
86dc4139
AM
18667+ }
18668+
18669+out:
18670+ return err;
18671+}
18672+
18673+void au_pin_hdir_set_owner(struct au_pin *p, struct task_struct *task)
18674+{
18675+#if defined(CONFIG_DEBUG_MUTEXES) || defined(CONFIG_SMP)
18676+ p->hdir->hi_inode->i_mutex.owner = task;
18677+#endif
18678+}
18679+
18680+void au_pin_hdir_acquire_nest(struct au_pin *p)
18681+{
18682+ if (p->hdir) {
18683+ mutex_acquire_nest(&p->hdir->hi_inode->i_mutex.dep_map,
18684+ p->lsc_hi, 0, NULL, _RET_IP_);
18685+ au_pin_hdir_set_owner(p, current);
18686+ }
dece6358 18687+}
1facf9fc 18688+
86dc4139
AM
18689+void au_pin_hdir_release(struct au_pin *p)
18690+{
18691+ if (p->hdir) {
18692+ au_pin_hdir_set_owner(p, p->task);
18693+ mutex_release(&p->hdir->hi_inode->i_mutex.dep_map, 1, _RET_IP_);
18694+ }
18695+}
1308ab2a 18696+
4a4d8108 18697+struct dentry *au_pinned_h_parent(struct au_pin *pin)
1308ab2a 18698+{
4a4d8108
AM
18699+ if (pin && pin->parent)
18700+ return au_h_dptr(pin->parent, pin->bindex);
18701+ return NULL;
dece6358 18702+}
1facf9fc 18703+
4a4d8108 18704+void au_unpin(struct au_pin *p)
dece6358 18705+{
86dc4139
AM
18706+ if (p->hdir)
18707+ au_pin_hdir_unlock(p);
e49829fe 18708+ if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
b4510431 18709+ vfsub_mnt_drop_write(p->h_mnt);
4a4d8108
AM
18710+ if (!p->hdir)
18711+ return;
1facf9fc 18712+
4a4d8108
AM
18713+ if (!au_ftest_pin(p->flags, DI_LOCKED))
18714+ di_read_unlock(p->parent, AuLock_IR);
18715+ iput(p->hdir->hi_inode);
18716+ dput(p->parent);
18717+ p->parent = NULL;
18718+ p->hdir = NULL;
18719+ p->h_mnt = NULL;
86dc4139 18720+ /* do not clear p->task */
4a4d8108 18721+}
1308ab2a 18722+
4a4d8108
AM
18723+int au_do_pin(struct au_pin *p)
18724+{
18725+ int err;
18726+ struct super_block *sb;
4a4d8108
AM
18727+ struct inode *h_dir;
18728+
18729+ err = 0;
18730+ sb = p->dentry->d_sb;
86dc4139 18731+ p->br = au_sbr(sb, p->bindex);
4a4d8108
AM
18732+ if (IS_ROOT(p->dentry)) {
18733+ if (au_ftest_pin(p->flags, MNT_WRITE)) {
86dc4139 18734+ p->h_mnt = au_br_mnt(p->br);
b4510431 18735+ err = vfsub_mnt_want_write(p->h_mnt);
4a4d8108
AM
18736+ if (unlikely(err)) {
18737+ au_fclr_pin(p->flags, MNT_WRITE);
18738+ goto out_err;
18739+ }
18740+ }
dece6358 18741+ goto out;
1facf9fc 18742+ }
18743+
86dc4139 18744+ p->h_dentry = NULL;
4a4d8108 18745+ if (p->bindex <= au_dbend(p->dentry))
86dc4139 18746+ p->h_dentry = au_h_dptr(p->dentry, p->bindex);
dece6358 18747+
4a4d8108
AM
18748+ p->parent = dget_parent(p->dentry);
18749+ if (!au_ftest_pin(p->flags, DI_LOCKED))
18750+ di_read_lock(p->parent, AuLock_IR, p->lsc_di);
dece6358 18751+
4a4d8108 18752+ h_dir = NULL;
86dc4139 18753+ p->h_parent = au_h_dptr(p->parent, p->bindex);
5527c038 18754+ p->hdir = au_hi(d_inode(p->parent), p->bindex);
4a4d8108
AM
18755+ if (p->hdir)
18756+ h_dir = p->hdir->hi_inode;
dece6358 18757+
b752ccd1
AM
18758+ /*
18759+ * udba case, or
18760+ * if DI_LOCKED is not set, then p->parent may be different
18761+ * and h_parent can be NULL.
18762+ */
86dc4139 18763+ if (unlikely(!p->hdir || !h_dir || !p->h_parent)) {
e49829fe 18764+ err = -EBUSY;
4a4d8108
AM
18765+ if (!au_ftest_pin(p->flags, DI_LOCKED))
18766+ di_read_unlock(p->parent, AuLock_IR);
18767+ dput(p->parent);
18768+ p->parent = NULL;
18769+ goto out_err;
18770+ }
1308ab2a 18771+
4a4d8108 18772+ if (au_ftest_pin(p->flags, MNT_WRITE)) {
86dc4139 18773+ p->h_mnt = au_br_mnt(p->br);
b4510431 18774+ err = vfsub_mnt_want_write(p->h_mnt);
dece6358 18775+ if (unlikely(err)) {
4a4d8108 18776+ au_fclr_pin(p->flags, MNT_WRITE);
86dc4139
AM
18777+ if (!au_ftest_pin(p->flags, DI_LOCKED))
18778+ di_read_unlock(p->parent, AuLock_IR);
18779+ dput(p->parent);
18780+ p->parent = NULL;
18781+ goto out_err;
dece6358
AM
18782+ }
18783+ }
4a4d8108 18784+
86dc4139
AM
18785+ au_igrab(h_dir);
18786+ err = au_pin_hdir_lock(p);
18787+ if (!err)
18788+ goto out; /* success */
18789+
076b876e
AM
18790+ au_unpin(p);
18791+
4f0767ce 18792+out_err:
4a4d8108
AM
18793+ pr_err("err %d\n", err);
18794+ err = au_busy_or_stale();
4f0767ce 18795+out:
1facf9fc 18796+ return err;
18797+}
18798+
4a4d8108
AM
18799+void au_pin_init(struct au_pin *p, struct dentry *dentry,
18800+ aufs_bindex_t bindex, int lsc_di, int lsc_hi,
18801+ unsigned int udba, unsigned char flags)
18802+{
18803+ p->dentry = dentry;
18804+ p->udba = udba;
18805+ p->lsc_di = lsc_di;
18806+ p->lsc_hi = lsc_hi;
18807+ p->flags = flags;
18808+ p->bindex = bindex;
18809+
18810+ p->parent = NULL;
18811+ p->hdir = NULL;
18812+ p->h_mnt = NULL;
86dc4139
AM
18813+
18814+ p->h_dentry = NULL;
18815+ p->h_parent = NULL;
18816+ p->br = NULL;
18817+ p->task = current;
4a4d8108
AM
18818+}
18819+
18820+int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
18821+ unsigned int udba, unsigned char flags)
18822+{
18823+ au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
18824+ udba, flags);
18825+ return au_do_pin(pin);
18826+}
18827+
dece6358
AM
18828+/* ---------------------------------------------------------------------- */
18829+
1308ab2a 18830+/*
4a4d8108
AM
18831+ * ->setattr() and ->getattr() are called in various cases.
18832+ * chmod, stat: dentry is revalidated.
18833+ * fchmod, fstat: file and dentry are not revalidated, additionally they may be
18834+ * unhashed.
18835+ * for ->setattr(), ia->ia_file is passed from ftruncate only.
1308ab2a 18836+ */
027c5e7a 18837+/* todo: consolidate with do_refresh() and simple_reval_dpath() */
c1595e42 18838+int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
1facf9fc 18839+{
4a4d8108 18840+ int err;
4a4d8108 18841+ struct dentry *parent;
1facf9fc 18842+
1308ab2a 18843+ err = 0;
027c5e7a 18844+ if (au_digen_test(dentry, sigen)) {
4a4d8108
AM
18845+ parent = dget_parent(dentry);
18846+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a 18847+ err = au_refresh_dentry(dentry, parent);
4a4d8108
AM
18848+ di_read_unlock(parent, AuLock_IR);
18849+ dput(parent);
dece6358 18850+ }
1facf9fc 18851+
4a4d8108 18852+ AuTraceErr(err);
1308ab2a 18853+ return err;
18854+}
dece6358 18855+
c1595e42
JR
18856+int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
18857+ struct au_icpup_args *a)
1308ab2a 18858+{
18859+ int err;
4a4d8108 18860+ loff_t sz;
e49829fe 18861+ aufs_bindex_t bstart, ibstart;
4a4d8108
AM
18862+ struct dentry *hi_wh, *parent;
18863+ struct inode *inode;
4a4d8108
AM
18864+ struct au_wr_dir_args wr_dir_args = {
18865+ .force_btgt = -1,
18866+ .flags = 0
18867+ };
18868+
2000de60 18869+ if (d_is_dir(dentry))
4a4d8108
AM
18870+ au_fset_wrdir(wr_dir_args.flags, ISDIR);
18871+ /* plink or hi_wh() case */
2000de60 18872+ bstart = au_dbstart(dentry);
5527c038 18873+ inode = d_inode(dentry);
e49829fe 18874+ ibstart = au_ibstart(inode);
027c5e7a 18875+ if (bstart != ibstart && !au_test_ro(inode->i_sb, ibstart, inode))
e49829fe 18876+ wr_dir_args.force_btgt = ibstart;
4a4d8108
AM
18877+ err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
18878+ if (unlikely(err < 0))
18879+ goto out;
18880+ a->btgt = err;
18881+ if (err != bstart)
18882+ au_fset_icpup(a->flags, DID_CPUP);
18883+
18884+ err = 0;
18885+ a->pin_flags = AuPin_MNT_WRITE;
18886+ parent = NULL;
18887+ if (!IS_ROOT(dentry)) {
18888+ au_fset_pin(a->pin_flags, DI_LOCKED);
18889+ parent = dget_parent(dentry);
18890+ di_write_lock_parent(parent);
18891+ }
18892+
18893+ err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
18894+ if (unlikely(err))
18895+ goto out_parent;
18896+
18897+ a->h_path.dentry = au_h_dptr(dentry, bstart);
4a4d8108 18898+ sz = -1;
5527c038 18899+ a->h_inode = d_inode(a->h_path.dentry);
c1595e42
JR
18900+ if (ia && (ia->ia_valid & ATTR_SIZE)) {
18901+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
18902+ if (ia->ia_size < i_size_read(a->h_inode))
18903+ sz = ia->ia_size;
18904+ mutex_unlock(&a->h_inode->i_mutex);
18905+ }
4a4d8108 18906+
4a4d8108 18907+ hi_wh = NULL;
027c5e7a 18908+ if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
4a4d8108
AM
18909+ hi_wh = au_hi_wh(inode, a->btgt);
18910+ if (!hi_wh) {
c2b27bf2
AM
18911+ struct au_cp_generic cpg = {
18912+ .dentry = dentry,
18913+ .bdst = a->btgt,
18914+ .bsrc = -1,
18915+ .len = sz,
18916+ .pin = &a->pin
18917+ };
18918+ err = au_sio_cpup_wh(&cpg, /*file*/NULL);
4a4d8108
AM
18919+ if (unlikely(err))
18920+ goto out_unlock;
18921+ hi_wh = au_hi_wh(inode, a->btgt);
18922+ /* todo: revalidate hi_wh? */
18923+ }
18924+ }
18925+
18926+ if (parent) {
18927+ au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
18928+ di_downgrade_lock(parent, AuLock_IR);
18929+ dput(parent);
18930+ parent = NULL;
18931+ }
18932+ if (!au_ftest_icpup(a->flags, DID_CPUP))
18933+ goto out; /* success */
18934+
18935+ if (!d_unhashed(dentry)) {
c2b27bf2
AM
18936+ struct au_cp_generic cpg = {
18937+ .dentry = dentry,
18938+ .bdst = a->btgt,
18939+ .bsrc = bstart,
18940+ .len = sz,
18941+ .pin = &a->pin,
18942+ .flags = AuCpup_DTIME | AuCpup_HOPEN
18943+ };
18944+ err = au_sio_cpup_simple(&cpg);
4a4d8108
AM
18945+ if (!err)
18946+ a->h_path.dentry = au_h_dptr(dentry, a->btgt);
18947+ } else if (!hi_wh)
18948+ a->h_path.dentry = au_h_dptr(dentry, a->btgt);
18949+ else
18950+ a->h_path.dentry = hi_wh; /* do not dget here */
1308ab2a 18951+
4f0767ce 18952+out_unlock:
5527c038 18953+ a->h_inode = d_inode(a->h_path.dentry);
86dc4139 18954+ if (!err)
dece6358 18955+ goto out; /* success */
4a4d8108 18956+ au_unpin(&a->pin);
4f0767ce 18957+out_parent:
4a4d8108
AM
18958+ if (parent) {
18959+ di_write_unlock(parent);
18960+ dput(parent);
18961+ }
4f0767ce 18962+out:
86dc4139
AM
18963+ if (!err)
18964+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
1facf9fc 18965+ return err;
18966+}
18967+
4a4d8108 18968+static int aufs_setattr(struct dentry *dentry, struct iattr *ia)
1facf9fc 18969+{
4a4d8108 18970+ int err;
523b37e3 18971+ struct inode *inode, *delegated;
4a4d8108
AM
18972+ struct super_block *sb;
18973+ struct file *file;
18974+ struct au_icpup_args *a;
1facf9fc 18975+
5527c038 18976+ inode = d_inode(dentry);
4a4d8108 18977+ IMustLock(inode);
dece6358 18978+
4a4d8108
AM
18979+ err = -ENOMEM;
18980+ a = kzalloc(sizeof(*a), GFP_NOFS);
18981+ if (unlikely(!a))
18982+ goto out;
1facf9fc 18983+
4a4d8108
AM
18984+ if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
18985+ ia->ia_valid &= ~ATTR_MODE;
dece6358 18986+
4a4d8108
AM
18987+ file = NULL;
18988+ sb = dentry->d_sb;
e49829fe
JR
18989+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
18990+ if (unlikely(err))
18991+ goto out_kfree;
18992+
4a4d8108
AM
18993+ if (ia->ia_valid & ATTR_FILE) {
18994+ /* currently ftruncate(2) only */
7e9cd9fe 18995+ AuDebugOn(!d_is_reg(dentry));
4a4d8108
AM
18996+ file = ia->ia_file;
18997+ err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
18998+ if (unlikely(err))
18999+ goto out_si;
19000+ ia->ia_file = au_hf_top(file);
19001+ a->udba = AuOpt_UDBA_NONE;
19002+ } else {
19003+ /* fchmod() doesn't pass ia_file */
19004+ a->udba = au_opt_udba(sb);
027c5e7a
AM
19005+ di_write_lock_child(dentry);
19006+ /* no d_unlinked(), to set UDBA_NONE for root */
4a4d8108
AM
19007+ if (d_unhashed(dentry))
19008+ a->udba = AuOpt_UDBA_NONE;
4a4d8108
AM
19009+ if (a->udba != AuOpt_UDBA_NONE) {
19010+ AuDebugOn(IS_ROOT(dentry));
19011+ err = au_reval_for_attr(dentry, au_sigen(sb));
19012+ if (unlikely(err))
19013+ goto out_dentry;
19014+ }
dece6358 19015+ }
dece6358 19016+
4a4d8108
AM
19017+ err = au_pin_and_icpup(dentry, ia, a);
19018+ if (unlikely(err < 0))
19019+ goto out_dentry;
19020+ if (au_ftest_icpup(a->flags, DID_CPUP)) {
19021+ ia->ia_file = NULL;
19022+ ia->ia_valid &= ~ATTR_FILE;
1308ab2a 19023+ }
dece6358 19024+
4a4d8108
AM
19025+ a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
19026+ if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
19027+ == (ATTR_MODE | ATTR_CTIME)) {
7eafdf33 19028+ err = security_path_chmod(&a->h_path, ia->ia_mode);
4a4d8108
AM
19029+ if (unlikely(err))
19030+ goto out_unlock;
19031+ } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
19032+ && (ia->ia_valid & ATTR_CTIME)) {
86dc4139 19033+ err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
4a4d8108
AM
19034+ if (unlikely(err))
19035+ goto out_unlock;
19036+ }
dece6358 19037+
4a4d8108
AM
19038+ if (ia->ia_valid & ATTR_SIZE) {
19039+ struct file *f;
1308ab2a 19040+
953406b4 19041+ if (ia->ia_size < i_size_read(inode))
4a4d8108 19042+ /* unmap only */
953406b4 19043+ truncate_setsize(inode, ia->ia_size);
1308ab2a 19044+
4a4d8108
AM
19045+ f = NULL;
19046+ if (ia->ia_valid & ATTR_FILE)
19047+ f = ia->ia_file;
19048+ mutex_unlock(&a->h_inode->i_mutex);
19049+ err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
19050+ mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
523b37e3
AM
19051+ } else {
19052+ delegated = NULL;
19053+ while (1) {
19054+ err = vfsub_notify_change(&a->h_path, ia, &delegated);
19055+ if (delegated) {
19056+ err = break_deleg_wait(&delegated);
19057+ if (!err)
19058+ continue;
19059+ }
19060+ break;
19061+ }
19062+ }
4a4d8108
AM
19063+ if (!err)
19064+ au_cpup_attr_changeable(inode);
1308ab2a 19065+
4f0767ce 19066+out_unlock:
4a4d8108
AM
19067+ mutex_unlock(&a->h_inode->i_mutex);
19068+ au_unpin(&a->pin);
027c5e7a
AM
19069+ if (unlikely(err))
19070+ au_update_dbstart(dentry);
4f0767ce 19071+out_dentry:
4a4d8108
AM
19072+ di_write_unlock(dentry);
19073+ if (file) {
19074+ fi_write_unlock(file);
19075+ ia->ia_file = file;
19076+ ia->ia_valid |= ATTR_FILE;
19077+ }
4f0767ce 19078+out_si:
4a4d8108 19079+ si_read_unlock(sb);
e49829fe 19080+out_kfree:
4a4d8108 19081+ kfree(a);
4f0767ce 19082+out:
4a4d8108
AM
19083+ AuTraceErr(err);
19084+ return err;
1facf9fc 19085+}
19086+
c1595e42
JR
19087+#if IS_ENABLED(CONFIG_AUFS_XATTR) || IS_ENABLED(CONFIG_FS_POSIX_ACL)
19088+static int au_h_path_to_set_attr(struct dentry *dentry,
19089+ struct au_icpup_args *a, struct path *h_path)
19090+{
19091+ int err;
19092+ struct super_block *sb;
19093+
19094+ sb = dentry->d_sb;
19095+ a->udba = au_opt_udba(sb);
19096+ /* no d_unlinked(), to set UDBA_NONE for root */
19097+ if (d_unhashed(dentry))
19098+ a->udba = AuOpt_UDBA_NONE;
19099+ if (a->udba != AuOpt_UDBA_NONE) {
19100+ AuDebugOn(IS_ROOT(dentry));
19101+ err = au_reval_for_attr(dentry, au_sigen(sb));
19102+ if (unlikely(err))
19103+ goto out;
19104+ }
19105+ err = au_pin_and_icpup(dentry, /*ia*/NULL, a);
19106+ if (unlikely(err < 0))
19107+ goto out;
19108+
19109+ h_path->dentry = a->h_path.dentry;
19110+ h_path->mnt = au_sbr_mnt(sb, a->btgt);
19111+
19112+out:
19113+ return err;
19114+}
19115+
19116+ssize_t au_srxattr(struct dentry *dentry, struct au_srxattr *arg)
19117+{
19118+ int err;
19119+ struct path h_path;
19120+ struct super_block *sb;
19121+ struct au_icpup_args *a;
19122+ struct inode *inode, *h_inode;
19123+
5527c038 19124+ inode = d_inode(dentry);
c1595e42
JR
19125+ IMustLock(inode);
19126+
19127+ err = -ENOMEM;
19128+ a = kzalloc(sizeof(*a), GFP_NOFS);
19129+ if (unlikely(!a))
19130+ goto out;
19131+
19132+ sb = dentry->d_sb;
19133+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19134+ if (unlikely(err))
19135+ goto out_kfree;
19136+
19137+ h_path.dentry = NULL; /* silence gcc */
19138+ di_write_lock_child(dentry);
19139+ err = au_h_path_to_set_attr(dentry, a, &h_path);
19140+ if (unlikely(err))
19141+ goto out_di;
19142+
19143+ mutex_unlock(&a->h_inode->i_mutex);
19144+ switch (arg->type) {
19145+ case AU_XATTR_SET:
19146+ err = vfsub_setxattr(h_path.dentry,
19147+ arg->u.set.name, arg->u.set.value,
19148+ arg->u.set.size, arg->u.set.flags);
19149+ break;
19150+ case AU_XATTR_REMOVE:
19151+ err = vfsub_removexattr(h_path.dentry, arg->u.remove.name);
19152+ break;
19153+ case AU_ACL_SET:
19154+ err = -EOPNOTSUPP;
5527c038 19155+ h_inode = d_inode(h_path.dentry);
c1595e42
JR
19156+ if (h_inode->i_op->set_acl)
19157+ err = h_inode->i_op->set_acl(h_inode,
19158+ arg->u.acl_set.acl,
19159+ arg->u.acl_set.type);
19160+ break;
19161+ }
19162+ if (!err)
19163+ au_cpup_attr_timesizes(inode);
19164+
19165+ au_unpin(&a->pin);
19166+ if (unlikely(err))
19167+ au_update_dbstart(dentry);
19168+
19169+out_di:
19170+ di_write_unlock(dentry);
19171+ si_read_unlock(sb);
19172+out_kfree:
19173+ kfree(a);
19174+out:
19175+ AuTraceErr(err);
19176+ return err;
19177+}
19178+#endif
19179+
4a4d8108
AM
19180+static void au_refresh_iattr(struct inode *inode, struct kstat *st,
19181+ unsigned int nlink)
1facf9fc 19182+{
9dbd164d
AM
19183+ unsigned int n;
19184+
4a4d8108 19185+ inode->i_mode = st->mode;
86dc4139
AM
19186+ /* don't i_[ug]id_write() here */
19187+ inode->i_uid = st->uid;
19188+ inode->i_gid = st->gid;
4a4d8108
AM
19189+ inode->i_atime = st->atime;
19190+ inode->i_mtime = st->mtime;
19191+ inode->i_ctime = st->ctime;
1facf9fc 19192+
4a4d8108
AM
19193+ au_cpup_attr_nlink(inode, /*force*/0);
19194+ if (S_ISDIR(inode->i_mode)) {
9dbd164d
AM
19195+ n = inode->i_nlink;
19196+ n -= nlink;
19197+ n += st->nlink;
f6b6e03d 19198+ smp_mb(); /* for i_nlink */
7eafdf33 19199+ /* 0 can happen */
92d182d2 19200+ set_nlink(inode, n);
4a4d8108 19201+ }
1facf9fc 19202+
4a4d8108
AM
19203+ spin_lock(&inode->i_lock);
19204+ inode->i_blocks = st->blocks;
19205+ i_size_write(inode, st->size);
19206+ spin_unlock(&inode->i_lock);
1facf9fc 19207+}
19208+
c1595e42
JR
19209+/*
19210+ * common routine for aufs_getattr() and aufs_getxattr().
19211+ * returns zero or negative (an error).
19212+ * @dentry will be read-locked in success.
19213+ */
19214+int au_h_path_getattr(struct dentry *dentry, int force, struct path *h_path)
1facf9fc 19215+{
4a4d8108 19216+ int err;
076b876e 19217+ unsigned int mnt_flags, sigen;
c1595e42 19218+ unsigned char udba_none;
4a4d8108 19219+ aufs_bindex_t bindex;
4a4d8108
AM
19220+ struct super_block *sb, *h_sb;
19221+ struct inode *inode;
1facf9fc 19222+
c1595e42
JR
19223+ h_path->mnt = NULL;
19224+ h_path->dentry = NULL;
19225+
19226+ err = 0;
4a4d8108 19227+ sb = dentry->d_sb;
4a4d8108
AM
19228+ mnt_flags = au_mntflags(sb);
19229+ udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
1facf9fc 19230+
4a4d8108 19231+ /* support fstat(2) */
027c5e7a 19232+ if (!d_unlinked(dentry) && !udba_none) {
076b876e 19233+ sigen = au_sigen(sb);
027c5e7a
AM
19234+ err = au_digen_test(dentry, sigen);
19235+ if (!err) {
4a4d8108 19236+ di_read_lock_child(dentry, AuLock_IR);
027c5e7a 19237+ err = au_dbrange_test(dentry);
c1595e42
JR
19238+ if (unlikely(err)) {
19239+ di_read_unlock(dentry, AuLock_IR);
19240+ goto out;
19241+ }
027c5e7a 19242+ } else {
4a4d8108
AM
19243+ AuDebugOn(IS_ROOT(dentry));
19244+ di_write_lock_child(dentry);
027c5e7a
AM
19245+ err = au_dbrange_test(dentry);
19246+ if (!err)
19247+ err = au_reval_for_attr(dentry, sigen);
c1595e42
JR
19248+ if (!err)
19249+ di_downgrade_lock(dentry, AuLock_IR);
19250+ else {
19251+ di_write_unlock(dentry);
19252+ goto out;
19253+ }
4a4d8108
AM
19254+ }
19255+ } else
19256+ di_read_lock_child(dentry, AuLock_IR);
1facf9fc 19257+
5527c038 19258+ inode = d_inode(dentry);
4a4d8108 19259+ bindex = au_ibstart(inode);
c1595e42
JR
19260+ h_path->mnt = au_sbr_mnt(sb, bindex);
19261+ h_sb = h_path->mnt->mnt_sb;
19262+ if (!force
19263+ && !au_test_fs_bad_iattr(h_sb)
19264+ && udba_none)
19265+ goto out; /* success */
1facf9fc 19266+
4a4d8108 19267+ if (au_dbstart(dentry) == bindex)
c1595e42 19268+ h_path->dentry = au_h_dptr(dentry, bindex);
4a4d8108 19269+ else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
c1595e42
JR
19270+ h_path->dentry = au_plink_lkup(inode, bindex);
19271+ if (IS_ERR(h_path->dentry))
19272+ /* pretending success */
19273+ h_path->dentry = NULL;
19274+ else
19275+ dput(h_path->dentry);
4a4d8108 19276+ }
c1595e42
JR
19277+
19278+out:
19279+ return err;
19280+}
19281+
19282+static int aufs_getattr(struct vfsmount *mnt __maybe_unused,
19283+ struct dentry *dentry, struct kstat *st)
19284+{
19285+ int err;
19286+ unsigned char positive;
19287+ struct path h_path;
19288+ struct inode *inode;
19289+ struct super_block *sb;
19290+
5527c038 19291+ inode = d_inode(dentry);
c1595e42
JR
19292+ sb = dentry->d_sb;
19293+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
19294+ if (unlikely(err))
19295+ goto out;
19296+ err = au_h_path_getattr(dentry, /*force*/0, &h_path);
19297+ if (unlikely(err))
19298+ goto out_si;
c06a8ce3 19299+ if (unlikely(!h_path.dentry))
c1595e42 19300+ /* illegally overlapped or something */
4a4d8108
AM
19301+ goto out_fill; /* pretending success */
19302+
5527c038 19303+ positive = d_is_positive(h_path.dentry);
4a4d8108 19304+ if (positive)
c06a8ce3 19305+ err = vfs_getattr(&h_path, st);
4a4d8108
AM
19306+ if (!err) {
19307+ if (positive)
c06a8ce3 19308+ au_refresh_iattr(inode, st,
5527c038 19309+ d_inode(h_path.dentry)->i_nlink);
4a4d8108 19310+ goto out_fill; /* success */
1facf9fc 19311+ }
7f207e10 19312+ AuTraceErr(err);
c1595e42 19313+ goto out_di;
4a4d8108 19314+
4f0767ce 19315+out_fill:
4a4d8108 19316+ generic_fillattr(inode, st);
c1595e42 19317+out_di:
4a4d8108 19318+ di_read_unlock(dentry, AuLock_IR);
c1595e42 19319+out_si:
4a4d8108 19320+ si_read_unlock(sb);
7f207e10
AM
19321+out:
19322+ AuTraceErr(err);
4a4d8108 19323+ return err;
1facf9fc 19324+}
19325+
19326+/* ---------------------------------------------------------------------- */
19327+
c2c0f25c
AM
19328+/*
19329+ * Assumption:
19330+ * - the number of symlinks is not so many.
19331+ *
19332+ * Structure:
19333+ * - sbinfo (instead of iinfo) contains an hlist of struct au_symlink.
19334+ * If iinfo contained the hlist, then it would be rather large waste of memory
19335+ * I am afraid.
19336+ * - struct au_symlink contains the necessary info for h_inode follow_link() and
19337+ * put_link().
19338+ */
1facf9fc 19339+
c2c0f25c
AM
19340+struct au_symlink {
19341+ union {
19342+ struct hlist_node hlist;
19343+ struct rcu_head rcu;
19344+ };
1facf9fc 19345+
c2c0f25c
AM
19346+ struct inode *h_inode;
19347+ void *h_cookie;
19348+};
1facf9fc 19349+
c2c0f25c
AM
19350+static void au_symlink_add(struct super_block *sb, struct au_symlink *slink,
19351+ struct inode *h_inode, void *cookie)
19352+{
19353+ struct au_sbinfo *sbinfo;
1facf9fc 19354+
c2c0f25c
AM
19355+ ihold(h_inode);
19356+ slink->h_inode = h_inode;
19357+ slink->h_cookie = cookie;
19358+ sbinfo = au_sbi(sb);
19359+ au_sphl_add(&slink->hlist, &sbinfo->si_symlink);
4a4d8108 19360+}
1facf9fc 19361+
c2c0f25c 19362+static void au_symlink_del(struct super_block *sb, struct au_symlink *slink)
4a4d8108 19363+{
c2c0f25c 19364+ struct au_sbinfo *sbinfo;
1facf9fc 19365+
c2c0f25c
AM
19366+ /* do not iput() within rcu */
19367+ iput(slink->h_inode);
19368+ slink->h_inode = NULL;
19369+ sbinfo = au_sbi(sb);
19370+ au_sphl_del_rcu(&slink->hlist, &sbinfo->si_symlink);
19371+ kfree_rcu(slink, rcu);
4a4d8108 19372+}
1facf9fc 19373+
c2c0f25c 19374+static const char *aufs_follow_link(struct dentry *dentry, void **cookie)
4a4d8108 19375+{
c2c0f25c
AM
19376+ const char *ret;
19377+ struct inode *inode, *h_inode;
19378+ struct dentry *h_dentry;
19379+ struct au_symlink *slink;
4a4d8108 19380+ int err;
c2c0f25c 19381+ aufs_bindex_t bindex;
1facf9fc 19382+
c2c0f25c 19383+ ret = NULL; /* supress a warning */
027c5e7a
AM
19384+ err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
19385+ if (unlikely(err))
c2c0f25c 19386+ goto out;
027c5e7a
AM
19387+
19388+ err = au_d_hashed_positive(dentry);
c2c0f25c
AM
19389+ if (unlikely(err))
19390+ goto out_unlock;
19391+
19392+ err = -EINVAL;
19393+ inode = d_inode(dentry);
19394+ bindex = au_ibstart(inode);
19395+ h_inode = au_h_iptr(inode, bindex);
19396+ if (unlikely(!h_inode->i_op->follow_link))
19397+ goto out_unlock;
19398+
19399+ err = -ENOMEM;
19400+ slink = kmalloc(sizeof(*slink), GFP_NOFS);
19401+ if (unlikely(!slink))
19402+ goto out_unlock;
19403+
19404+ err = -EBUSY;
19405+ h_dentry = NULL;
19406+ if (au_dbstart(dentry) <= bindex) {
19407+ h_dentry = au_h_dptr(dentry, bindex);
19408+ if (h_dentry)
19409+ dget(h_dentry);
027c5e7a 19410+ }
c2c0f25c
AM
19411+ if (!h_dentry) {
19412+ h_dentry = d_find_any_alias(h_inode);
19413+ if (IS_ERR(h_dentry)) {
19414+ err = PTR_ERR(h_dentry);
19415+ goto out_free;
19416+ }
19417+ }
19418+ if (unlikely(!h_dentry))
19419+ goto out_free;
1facf9fc 19420+
c2c0f25c
AM
19421+ err = 0;
19422+ AuDbg("%pf\n", h_inode->i_op->follow_link);
19423+ AuDbgDentry(h_dentry);
19424+ ret = h_inode->i_op->follow_link(h_dentry, cookie);
19425+ dput(h_dentry);
19426+
19427+ if (!IS_ERR_OR_NULL(ret)) {
19428+ au_symlink_add(inode->i_sb, slink, h_inode, *cookie);
19429+ *cookie = slink;
19430+ AuDbg("slink %p\n", slink);
19431+ goto out_unlock; /* success */
1308ab2a 19432+ }
1facf9fc 19433+
c2c0f25c
AM
19434+out_free:
19435+ slink->h_inode = NULL;
19436+ kfree_rcu(slink, rcu);
19437+out_unlock:
19438+ aufs_read_unlock(dentry, AuLock_IR);
4f0767ce 19439+out:
c2c0f25c
AM
19440+ if (unlikely(err))
19441+ ret = ERR_PTR(err);
19442+ AuTraceErrPtr(ret);
19443+ return ret;
4a4d8108 19444+}
1facf9fc 19445+
c2c0f25c 19446+static void aufs_put_link(struct inode *inode, void *cookie)
4a4d8108 19447+{
c2c0f25c
AM
19448+ struct au_symlink *slink;
19449+ struct inode *h_inode;
537831f9 19450+
c2c0f25c
AM
19451+ slink = cookie;
19452+ AuDbg("slink %p\n", slink);
19453+ h_inode = slink->h_inode;
19454+ AuDbg("%pf\n", h_inode->i_op->put_link);
19455+ AuDbgInode(h_inode);
19456+ if (h_inode->i_op->put_link)
19457+ h_inode->i_op->put_link(h_inode, slink->h_cookie);
19458+ au_symlink_del(inode->i_sb, slink);
4a4d8108 19459+}
1facf9fc 19460+
4a4d8108 19461+/* ---------------------------------------------------------------------- */
1facf9fc 19462+
0c3ec466 19463+static int aufs_update_time(struct inode *inode, struct timespec *ts, int flags)
4a4d8108 19464+{
0c3ec466
AM
19465+ int err;
19466+ struct super_block *sb;
19467+ struct inode *h_inode;
19468+
19469+ sb = inode->i_sb;
19470+ /* mmap_sem might be acquired already, cf. aufs_mmap() */
19471+ lockdep_off();
19472+ si_read_lock(sb, AuLock_FLUSH);
19473+ ii_write_lock_child(inode);
19474+ lockdep_on();
19475+ h_inode = au_h_iptr(inode, au_ibstart(inode));
19476+ err = vfsub_update_time(h_inode, ts, flags);
19477+ lockdep_off();
38d290e6
JR
19478+ if (!err)
19479+ au_cpup_attr_timesizes(inode);
0c3ec466
AM
19480+ ii_write_unlock(inode);
19481+ si_read_unlock(sb);
19482+ lockdep_on();
38d290e6
JR
19483+
19484+ if (!err && (flags & S_VERSION))
19485+ inode_inc_iversion(inode);
19486+
0c3ec466 19487+ return err;
4a4d8108 19488+}
1facf9fc 19489+
4a4d8108 19490+/* ---------------------------------------------------------------------- */
1308ab2a 19491+
4a4d8108
AM
19492+struct inode_operations aufs_symlink_iop = {
19493+ .permission = aufs_permission,
c1595e42
JR
19494+#ifdef CONFIG_FS_POSIX_ACL
19495+ .get_acl = aufs_get_acl,
19496+ .set_acl = aufs_set_acl, /* unsupport for symlink? */
19497+#endif
19498+
4a4d8108
AM
19499+ .setattr = aufs_setattr,
19500+ .getattr = aufs_getattr,
0c3ec466 19501+
c1595e42
JR
19502+#ifdef CONFIG_AUFS_XATTR
19503+ .setxattr = aufs_setxattr,
19504+ .getxattr = aufs_getxattr,
19505+ .listxattr = aufs_listxattr,
19506+ .removexattr = aufs_removexattr,
19507+#endif
19508+
c2c0f25c 19509+ .readlink = generic_readlink,
4a4d8108 19510+ .follow_link = aufs_follow_link,
0c3ec466
AM
19511+ .put_link = aufs_put_link,
19512+
19513+ /* .update_time = aufs_update_time */
4a4d8108
AM
19514+};
19515+
19516+struct inode_operations aufs_dir_iop = {
19517+ .create = aufs_create,
19518+ .lookup = aufs_lookup,
19519+ .link = aufs_link,
19520+ .unlink = aufs_unlink,
19521+ .symlink = aufs_symlink,
19522+ .mkdir = aufs_mkdir,
19523+ .rmdir = aufs_rmdir,
19524+ .mknod = aufs_mknod,
19525+ .rename = aufs_rename,
19526+
19527+ .permission = aufs_permission,
c1595e42
JR
19528+#ifdef CONFIG_FS_POSIX_ACL
19529+ .get_acl = aufs_get_acl,
19530+ .set_acl = aufs_set_acl,
19531+#endif
19532+
4a4d8108 19533+ .setattr = aufs_setattr,
0c3ec466
AM
19534+ .getattr = aufs_getattr,
19535+
c1595e42
JR
19536+#ifdef CONFIG_AUFS_XATTR
19537+ .setxattr = aufs_setxattr,
19538+ .getxattr = aufs_getxattr,
19539+ .listxattr = aufs_listxattr,
19540+ .removexattr = aufs_removexattr,
19541+#endif
19542+
38d290e6 19543+ .update_time = aufs_update_time,
b912730e 19544+ .atomic_open = aufs_atomic_open,
38d290e6 19545+ .tmpfile = aufs_tmpfile
4a4d8108
AM
19546+};
19547+
19548+struct inode_operations aufs_iop = {
19549+ .permission = aufs_permission,
c1595e42
JR
19550+#ifdef CONFIG_FS_POSIX_ACL
19551+ .get_acl = aufs_get_acl,
19552+ .set_acl = aufs_set_acl,
19553+#endif
19554+
4a4d8108
AM
19555+ .setattr = aufs_setattr,
19556+ .getattr = aufs_getattr,
0c3ec466 19557+
c1595e42
JR
19558+#ifdef CONFIG_AUFS_XATTR
19559+ .setxattr = aufs_setxattr,
19560+ .getxattr = aufs_getxattr,
19561+ .listxattr = aufs_listxattr,
19562+ .removexattr = aufs_removexattr,
19563+#endif
19564+
0c3ec466 19565+ .update_time = aufs_update_time
4a4d8108 19566+};
7f207e10
AM
19567diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
19568--- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 19569+++ linux/fs/aufs/i_op_del.c 2015-09-24 10:47:58.254719746 +0200
5527c038 19570@@ -0,0 +1,510 @@
1facf9fc 19571+/*
2000de60 19572+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 19573+ *
19574+ * This program, aufs is free software; you can redistribute it and/or modify
19575+ * it under the terms of the GNU General Public License as published by
19576+ * the Free Software Foundation; either version 2 of the License, or
19577+ * (at your option) any later version.
dece6358
AM
19578+ *
19579+ * This program is distributed in the hope that it will be useful,
19580+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
19581+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19582+ * GNU General Public License for more details.
19583+ *
19584+ * You should have received a copy of the GNU General Public License
523b37e3 19585+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 19586+ */
19587+
19588+/*
4a4d8108 19589+ * inode operations (del entry)
1308ab2a 19590+ */
dece6358 19591+
1308ab2a 19592+#include "aufs.h"
dece6358 19593+
4a4d8108
AM
19594+/*
19595+ * decide if a new whiteout for @dentry is necessary or not.
19596+ * when it is necessary, prepare the parent dir for the upper branch whose
19597+ * branch index is @bcpup for creation. the actual creation of the whiteout will
19598+ * be done by caller.
19599+ * return value:
19600+ * 0: wh is unnecessary
19601+ * plus: wh is necessary
19602+ * minus: error
19603+ */
19604+int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
1308ab2a 19605+{
4a4d8108
AM
19606+ int need_wh, err;
19607+ aufs_bindex_t bstart;
19608+ struct super_block *sb;
dece6358 19609+
4a4d8108
AM
19610+ sb = dentry->d_sb;
19611+ bstart = au_dbstart(dentry);
19612+ if (*bcpup < 0) {
19613+ *bcpup = bstart;
5527c038 19614+ if (au_test_ro(sb, bstart, d_inode(dentry))) {
4a4d8108
AM
19615+ err = AuWbrCopyup(au_sbi(sb), dentry);
19616+ *bcpup = err;
19617+ if (unlikely(err < 0))
19618+ goto out;
19619+ }
19620+ } else
19621+ AuDebugOn(bstart < *bcpup
5527c038 19622+ || au_test_ro(sb, *bcpup, d_inode(dentry)));
4a4d8108 19623+ AuDbg("bcpup %d, bstart %d\n", *bcpup, bstart);
1308ab2a 19624+
4a4d8108
AM
19625+ if (*bcpup != bstart) {
19626+ err = au_cpup_dirs(dentry, *bcpup);
19627+ if (unlikely(err))
19628+ goto out;
19629+ need_wh = 1;
19630+ } else {
027c5e7a 19631+ struct au_dinfo *dinfo, *tmp;
4a4d8108 19632+
027c5e7a
AM
19633+ need_wh = -ENOMEM;
19634+ dinfo = au_di(dentry);
19635+ tmp = au_di_alloc(sb, AuLsc_DI_TMP);
19636+ if (tmp) {
19637+ au_di_cp(tmp, dinfo);
19638+ au_di_swap(tmp, dinfo);
19639+ /* returns the number of positive dentries */
537831f9 19640+ need_wh = au_lkup_dentry(dentry, bstart + 1, /*type*/0);
027c5e7a
AM
19641+ au_di_swap(tmp, dinfo);
19642+ au_rw_write_unlock(&tmp->di_rwsem);
19643+ au_di_free(tmp);
4a4d8108
AM
19644+ }
19645+ }
19646+ AuDbg("need_wh %d\n", need_wh);
19647+ err = need_wh;
19648+
4f0767ce 19649+out:
4a4d8108 19650+ return err;
1facf9fc 19651+}
19652+
4a4d8108
AM
19653+/*
19654+ * simple tests for the del-entry operations.
19655+ * following the checks in vfs, plus the parent-child relationship.
19656+ */
19657+int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
19658+ struct dentry *h_parent, int isdir)
1facf9fc 19659+{
4a4d8108
AM
19660+ int err;
19661+ umode_t h_mode;
19662+ struct dentry *h_dentry, *h_latest;
1308ab2a 19663+ struct inode *h_inode;
1facf9fc 19664+
4a4d8108 19665+ h_dentry = au_h_dptr(dentry, bindex);
5527c038 19666+ if (d_really_is_positive(dentry)) {
4a4d8108 19667+ err = -ENOENT;
5527c038
JR
19668+ if (unlikely(d_is_negative(h_dentry)))
19669+ goto out;
19670+ h_inode = d_inode(h_dentry);
19671+ if (unlikely(!h_inode->i_nlink))
4a4d8108 19672+ goto out;
1facf9fc 19673+
4a4d8108
AM
19674+ h_mode = h_inode->i_mode;
19675+ if (!isdir) {
19676+ err = -EISDIR;
19677+ if (unlikely(S_ISDIR(h_mode)))
19678+ goto out;
19679+ } else if (unlikely(!S_ISDIR(h_mode))) {
19680+ err = -ENOTDIR;
19681+ goto out;
19682+ }
19683+ } else {
19684+ /* rename(2) case */
19685+ err = -EIO;
5527c038 19686+ if (unlikely(d_is_positive(h_dentry)))
4a4d8108
AM
19687+ goto out;
19688+ }
1facf9fc 19689+
4a4d8108
AM
19690+ err = -ENOENT;
19691+ /* expected parent dir is locked */
19692+ if (unlikely(h_parent != h_dentry->d_parent))
19693+ goto out;
19694+ err = 0;
19695+
19696+ /*
19697+ * rmdir a dir may break the consistency on some filesystem.
19698+ * let's try heavy test.
19699+ */
19700+ err = -EACCES;
076b876e 19701+ if (unlikely(!au_opt_test(au_mntflags(dentry->d_sb), DIRPERM1)
5527c038 19702+ && au_test_h_perm(d_inode(h_parent),
076b876e 19703+ MAY_EXEC | MAY_WRITE)))
4a4d8108
AM
19704+ goto out;
19705+
076b876e 19706+ h_latest = au_sio_lkup_one(&dentry->d_name, h_parent);
4a4d8108
AM
19707+ err = -EIO;
19708+ if (IS_ERR(h_latest))
19709+ goto out;
19710+ if (h_latest == h_dentry)
19711+ err = 0;
19712+ dput(h_latest);
19713+
4f0767ce 19714+out:
4a4d8108 19715+ return err;
1308ab2a 19716+}
1facf9fc 19717+
4a4d8108
AM
19718+/*
19719+ * decide the branch where we operate for @dentry. the branch index will be set
19720+ * @rbcpup. after diciding it, 'pin' it and store the timestamps of the parent
19721+ * dir for reverting.
19722+ * when a new whiteout is necessary, create it.
19723+ */
19724+static struct dentry*
19725+lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
19726+ struct au_dtime *dt, struct au_pin *pin)
1308ab2a 19727+{
4a4d8108
AM
19728+ struct dentry *wh_dentry;
19729+ struct super_block *sb;
19730+ struct path h_path;
19731+ int err, need_wh;
19732+ unsigned int udba;
19733+ aufs_bindex_t bcpup;
dece6358 19734+
4a4d8108
AM
19735+ need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
19736+ wh_dentry = ERR_PTR(need_wh);
19737+ if (unlikely(need_wh < 0))
19738+ goto out;
19739+
19740+ sb = dentry->d_sb;
19741+ udba = au_opt_udba(sb);
19742+ bcpup = *rbcpup;
19743+ err = au_pin(pin, dentry, bcpup, udba,
19744+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
19745+ wh_dentry = ERR_PTR(err);
19746+ if (unlikely(err))
19747+ goto out;
19748+
19749+ h_path.dentry = au_pinned_h_parent(pin);
19750+ if (udba != AuOpt_UDBA_NONE
19751+ && au_dbstart(dentry) == bcpup) {
19752+ err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
19753+ wh_dentry = ERR_PTR(err);
19754+ if (unlikely(err))
19755+ goto out_unpin;
19756+ }
19757+
19758+ h_path.mnt = au_sbr_mnt(sb, bcpup);
19759+ au_dtime_store(dt, au_pinned_parent(pin), &h_path);
19760+ wh_dentry = NULL;
19761+ if (!need_wh)
19762+ goto out; /* success, no need to create whiteout */
19763+
19764+ wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
19765+ if (IS_ERR(wh_dentry))
19766+ goto out_unpin;
19767+
19768+ /* returns with the parent is locked and wh_dentry is dget-ed */
19769+ goto out; /* success */
19770+
4f0767ce 19771+out_unpin:
4a4d8108 19772+ au_unpin(pin);
4f0767ce 19773+out:
4a4d8108 19774+ return wh_dentry;
1facf9fc 19775+}
19776+
4a4d8108
AM
19777+/*
19778+ * when removing a dir, rename it to a unique temporary whiteout-ed name first
19779+ * in order to be revertible and save time for removing many child whiteouts
19780+ * under the dir.
19781+ * returns 1 when there are too many child whiteout and caller should remove
19782+ * them asynchronously. returns 0 when the number of children is enough small to
19783+ * remove now or the branch fs is a remote fs.
19784+ * otherwise return an error.
19785+ */
19786+static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
19787+ struct au_nhash *whlist, struct inode *dir)
1facf9fc 19788+{
4a4d8108
AM
19789+ int rmdir_later, err, dirwh;
19790+ struct dentry *h_dentry;
19791+ struct super_block *sb;
5527c038 19792+ struct inode *inode;
4a4d8108
AM
19793+
19794+ sb = dentry->d_sb;
19795+ SiMustAnyLock(sb);
19796+ h_dentry = au_h_dptr(dentry, bindex);
19797+ err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
19798+ if (unlikely(err))
19799+ goto out;
19800+
19801+ /* stop monitoring */
5527c038
JR
19802+ inode = d_inode(dentry);
19803+ au_hn_free(au_hi(inode, bindex));
4a4d8108
AM
19804+
19805+ if (!au_test_fs_remote(h_dentry->d_sb)) {
19806+ dirwh = au_sbi(sb)->si_dirwh;
19807+ rmdir_later = (dirwh <= 1);
19808+ if (!rmdir_later)
19809+ rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
19810+ dirwh);
19811+ if (rmdir_later)
19812+ return rmdir_later;
19813+ }
1facf9fc 19814+
4a4d8108
AM
19815+ err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
19816+ if (unlikely(err)) {
523b37e3
AM
19817+ AuIOErr("rmdir %pd, b%d failed, %d. ignored\n",
19818+ h_dentry, bindex, err);
4a4d8108
AM
19819+ err = 0;
19820+ }
dece6358 19821+
4f0767ce 19822+out:
4a4d8108
AM
19823+ AuTraceErr(err);
19824+ return err;
19825+}
1308ab2a 19826+
4a4d8108
AM
19827+/*
19828+ * final procedure for deleting a entry.
19829+ * maintain dentry and iattr.
19830+ */
19831+static void epilog(struct inode *dir, struct dentry *dentry,
19832+ aufs_bindex_t bindex)
19833+{
19834+ struct inode *inode;
1308ab2a 19835+
5527c038 19836+ inode = d_inode(dentry);
4a4d8108
AM
19837+ d_drop(dentry);
19838+ inode->i_ctime = dir->i_ctime;
1308ab2a 19839+
b912730e 19840+ au_dir_ts(dir, bindex);
4a4d8108 19841+ dir->i_version++;
1facf9fc 19842+}
19843+
4a4d8108
AM
19844+/*
19845+ * when an error happened, remove the created whiteout and revert everything.
19846+ */
7f207e10
AM
19847+static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
19848+ aufs_bindex_t bwh, struct dentry *wh_dentry,
19849+ struct dentry *dentry, struct au_dtime *dt)
1facf9fc 19850+{
4a4d8108
AM
19851+ int rerr;
19852+ struct path h_path = {
19853+ .dentry = wh_dentry,
7f207e10 19854+ .mnt = au_sbr_mnt(dir->i_sb, bindex)
4a4d8108 19855+ };
dece6358 19856+
7f207e10 19857+ rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
4a4d8108
AM
19858+ if (!rerr) {
19859+ au_set_dbwh(dentry, bwh);
19860+ au_dtime_revert(dt);
19861+ return 0;
19862+ }
dece6358 19863+
523b37e3 19864+ AuIOErr("%pd reverting whiteout failed(%d, %d)\n", dentry, err, rerr);
4a4d8108 19865+ return -EIO;
1facf9fc 19866+}
19867+
4a4d8108 19868+/* ---------------------------------------------------------------------- */
1facf9fc 19869+
4a4d8108 19870+int aufs_unlink(struct inode *dir, struct dentry *dentry)
1308ab2a 19871+{
4a4d8108
AM
19872+ int err;
19873+ aufs_bindex_t bwh, bindex, bstart;
523b37e3 19874+ struct inode *inode, *h_dir, *delegated;
4a4d8108 19875+ struct dentry *parent, *wh_dentry;
c2b27bf2
AM
19876+ /* to reuduce stack size */
19877+ struct {
19878+ struct au_dtime dt;
19879+ struct au_pin pin;
19880+ struct path h_path;
19881+ } *a;
1facf9fc 19882+
4a4d8108 19883+ IMustLock(dir);
027c5e7a 19884+
c2b27bf2
AM
19885+ err = -ENOMEM;
19886+ a = kmalloc(sizeof(*a), GFP_NOFS);
19887+ if (unlikely(!a))
19888+ goto out;
19889+
027c5e7a
AM
19890+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
19891+ if (unlikely(err))
c2b27bf2 19892+ goto out_free;
027c5e7a
AM
19893+ err = au_d_hashed_positive(dentry);
19894+ if (unlikely(err))
19895+ goto out_unlock;
5527c038 19896+ inode = d_inode(dentry);
4a4d8108 19897+ IMustLock(inode);
027c5e7a 19898+ err = -EISDIR;
2000de60 19899+ if (unlikely(d_is_dir(dentry)))
027c5e7a 19900+ goto out_unlock; /* possible? */
1facf9fc 19901+
4a4d8108
AM
19902+ bstart = au_dbstart(dentry);
19903+ bwh = au_dbwh(dentry);
19904+ bindex = -1;
027c5e7a
AM
19905+ parent = dentry->d_parent; /* dir inode is locked */
19906+ di_write_lock_parent(parent);
c2b27bf2
AM
19907+ wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &a->dt,
19908+ &a->pin);
4a4d8108
AM
19909+ err = PTR_ERR(wh_dentry);
19910+ if (IS_ERR(wh_dentry))
027c5e7a 19911+ goto out_parent;
1facf9fc 19912+
c2b27bf2
AM
19913+ a->h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
19914+ a->h_path.dentry = au_h_dptr(dentry, bstart);
19915+ dget(a->h_path.dentry);
4a4d8108 19916+ if (bindex == bstart) {
c2b27bf2 19917+ h_dir = au_pinned_h_dir(&a->pin);
523b37e3
AM
19918+ delegated = NULL;
19919+ err = vfsub_unlink(h_dir, &a->h_path, &delegated, /*force*/0);
19920+ if (unlikely(err == -EWOULDBLOCK)) {
19921+ pr_warn("cannot retry for NFSv4 delegation"
19922+ " for an internal unlink\n");
19923+ iput(delegated);
19924+ }
4a4d8108
AM
19925+ } else {
19926+ /* dir inode is locked */
5527c038 19927+ h_dir = d_inode(wh_dentry->d_parent);
4a4d8108
AM
19928+ IMustLock(h_dir);
19929+ err = 0;
19930+ }
dece6358 19931+
4a4d8108 19932+ if (!err) {
7f207e10 19933+ vfsub_drop_nlink(inode);
4a4d8108
AM
19934+ epilog(dir, dentry, bindex);
19935+
19936+ /* update target timestamps */
19937+ if (bindex == bstart) {
c2b27bf2
AM
19938+ vfsub_update_h_iattr(&a->h_path, /*did*/NULL);
19939+ /*ignore*/
5527c038 19940+ inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
4a4d8108
AM
19941+ } else
19942+ /* todo: this timestamp may be reverted later */
19943+ inode->i_ctime = h_dir->i_ctime;
027c5e7a 19944+ goto out_unpin; /* success */
1facf9fc 19945+ }
19946+
4a4d8108
AM
19947+ /* revert */
19948+ if (wh_dentry) {
19949+ int rerr;
19950+
c2b27bf2
AM
19951+ rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
19952+ &a->dt);
4a4d8108
AM
19953+ if (rerr)
19954+ err = rerr;
dece6358 19955+ }
1facf9fc 19956+
027c5e7a 19957+out_unpin:
c2b27bf2 19958+ au_unpin(&a->pin);
4a4d8108 19959+ dput(wh_dentry);
c2b27bf2 19960+ dput(a->h_path.dentry);
027c5e7a 19961+out_parent:
4a4d8108 19962+ di_write_unlock(parent);
027c5e7a 19963+out_unlock:
4a4d8108 19964+ aufs_read_unlock(dentry, AuLock_DW);
c2b27bf2
AM
19965+out_free:
19966+ kfree(a);
027c5e7a 19967+out:
4a4d8108 19968+ return err;
dece6358
AM
19969+}
19970+
4a4d8108 19971+int aufs_rmdir(struct inode *dir, struct dentry *dentry)
1308ab2a 19972+{
4a4d8108
AM
19973+ int err, rmdir_later;
19974+ aufs_bindex_t bwh, bindex, bstart;
4a4d8108
AM
19975+ struct inode *inode;
19976+ struct dentry *parent, *wh_dentry, *h_dentry;
19977+ struct au_whtmp_rmdir *args;
c2b27bf2
AM
19978+ /* to reuduce stack size */
19979+ struct {
19980+ struct au_dtime dt;
19981+ struct au_pin pin;
19982+ } *a;
1facf9fc 19983+
4a4d8108 19984+ IMustLock(dir);
027c5e7a 19985+
c2b27bf2
AM
19986+ err = -ENOMEM;
19987+ a = kmalloc(sizeof(*a), GFP_NOFS);
19988+ if (unlikely(!a))
19989+ goto out;
19990+
027c5e7a
AM
19991+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
19992+ if (unlikely(err))
c2b27bf2 19993+ goto out_free;
53392da6
AM
19994+ err = au_alive_dir(dentry);
19995+ if (unlikely(err))
027c5e7a 19996+ goto out_unlock;
5527c038 19997+ inode = d_inode(dentry);
4a4d8108 19998+ IMustLock(inode);
027c5e7a 19999+ err = -ENOTDIR;
2000de60 20000+ if (unlikely(!d_is_dir(dentry)))
027c5e7a 20001+ goto out_unlock; /* possible? */
dece6358 20002+
4a4d8108
AM
20003+ err = -ENOMEM;
20004+ args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
20005+ if (unlikely(!args))
20006+ goto out_unlock;
dece6358 20007+
4a4d8108
AM
20008+ parent = dentry->d_parent; /* dir inode is locked */
20009+ di_write_lock_parent(parent);
20010+ err = au_test_empty(dentry, &args->whlist);
20011+ if (unlikely(err))
027c5e7a 20012+ goto out_parent;
1facf9fc 20013+
4a4d8108
AM
20014+ bstart = au_dbstart(dentry);
20015+ bwh = au_dbwh(dentry);
20016+ bindex = -1;
c2b27bf2
AM
20017+ wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &a->dt,
20018+ &a->pin);
4a4d8108
AM
20019+ err = PTR_ERR(wh_dentry);
20020+ if (IS_ERR(wh_dentry))
027c5e7a 20021+ goto out_parent;
1facf9fc 20022+
4a4d8108
AM
20023+ h_dentry = au_h_dptr(dentry, bstart);
20024+ dget(h_dentry);
20025+ rmdir_later = 0;
20026+ if (bindex == bstart) {
20027+ err = renwh_and_rmdir(dentry, bstart, &args->whlist, dir);
20028+ if (err > 0) {
20029+ rmdir_later = err;
20030+ err = 0;
20031+ }
20032+ } else {
20033+ /* stop monitoring */
20034+ au_hn_free(au_hi(inode, bstart));
20035+
20036+ /* dir inode is locked */
5527c038 20037+ IMustLock(d_inode(wh_dentry->d_parent));
1facf9fc 20038+ err = 0;
20039+ }
20040+
4a4d8108 20041+ if (!err) {
027c5e7a 20042+ vfsub_dead_dir(inode);
4a4d8108
AM
20043+ au_set_dbdiropq(dentry, -1);
20044+ epilog(dir, dentry, bindex);
1308ab2a 20045+
4a4d8108
AM
20046+ if (rmdir_later) {
20047+ au_whtmp_kick_rmdir(dir, bstart, h_dentry, args);
20048+ args = NULL;
20049+ }
1308ab2a 20050+
4a4d8108 20051+ goto out_unpin; /* success */
1facf9fc 20052+ }
20053+
4a4d8108
AM
20054+ /* revert */
20055+ AuLabel(revert);
20056+ if (wh_dentry) {
20057+ int rerr;
1308ab2a 20058+
c2b27bf2
AM
20059+ rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry,
20060+ &a->dt);
4a4d8108
AM
20061+ if (rerr)
20062+ err = rerr;
1facf9fc 20063+ }
20064+
4f0767ce 20065+out_unpin:
c2b27bf2 20066+ au_unpin(&a->pin);
4a4d8108
AM
20067+ dput(wh_dentry);
20068+ dput(h_dentry);
027c5e7a 20069+out_parent:
4a4d8108
AM
20070+ di_write_unlock(parent);
20071+ if (args)
20072+ au_whtmp_rmdir_free(args);
4f0767ce 20073+out_unlock:
4a4d8108 20074+ aufs_read_unlock(dentry, AuLock_DW);
c2b27bf2
AM
20075+out_free:
20076+ kfree(a);
4f0767ce 20077+out:
4a4d8108
AM
20078+ AuTraceErr(err);
20079+ return err;
dece6358 20080+}
7f207e10
AM
20081diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
20082--- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 20083+++ linux/fs/aufs/i_op_ren.c 2015-09-24 10:47:58.254719746 +0200
5527c038 20084@@ -0,0 +1,1017 @@
1facf9fc 20085+/*
2000de60 20086+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 20087+ *
20088+ * This program, aufs is free software; you can redistribute it and/or modify
20089+ * it under the terms of the GNU General Public License as published by
20090+ * the Free Software Foundation; either version 2 of the License, or
20091+ * (at your option) any later version.
dece6358
AM
20092+ *
20093+ * This program is distributed in the hope that it will be useful,
20094+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
20095+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20096+ * GNU General Public License for more details.
20097+ *
20098+ * You should have received a copy of the GNU General Public License
523b37e3 20099+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 20100+ */
20101+
20102+/*
4a4d8108
AM
20103+ * inode operation (rename entry)
20104+ * todo: this is crazy monster
1facf9fc 20105+ */
20106+
20107+#include "aufs.h"
20108+
4a4d8108
AM
20109+enum { AuSRC, AuDST, AuSrcDst };
20110+enum { AuPARENT, AuCHILD, AuParentChild };
1facf9fc 20111+
4a4d8108
AM
20112+#define AuRen_ISDIR 1
20113+#define AuRen_ISSAMEDIR (1 << 1)
20114+#define AuRen_WHSRC (1 << 2)
20115+#define AuRen_WHDST (1 << 3)
20116+#define AuRen_MNT_WRITE (1 << 4)
20117+#define AuRen_DT_DSTDIR (1 << 5)
20118+#define AuRen_DIROPQ (1 << 6)
4a4d8108 20119+#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
7f207e10
AM
20120+#define au_fset_ren(flags, name) \
20121+ do { (flags) |= AuRen_##name; } while (0)
20122+#define au_fclr_ren(flags, name) \
20123+ do { (flags) &= ~AuRen_##name; } while (0)
1facf9fc 20124+
4a4d8108
AM
20125+struct au_ren_args {
20126+ struct {
20127+ struct dentry *dentry, *h_dentry, *parent, *h_parent,
20128+ *wh_dentry;
20129+ struct inode *dir, *inode;
20130+ struct au_hinode *hdir;
20131+ struct au_dtime dt[AuParentChild];
20132+ aufs_bindex_t bstart;
20133+ } sd[AuSrcDst];
1facf9fc 20134+
4a4d8108
AM
20135+#define src_dentry sd[AuSRC].dentry
20136+#define src_dir sd[AuSRC].dir
20137+#define src_inode sd[AuSRC].inode
20138+#define src_h_dentry sd[AuSRC].h_dentry
20139+#define src_parent sd[AuSRC].parent
20140+#define src_h_parent sd[AuSRC].h_parent
20141+#define src_wh_dentry sd[AuSRC].wh_dentry
20142+#define src_hdir sd[AuSRC].hdir
20143+#define src_h_dir sd[AuSRC].hdir->hi_inode
20144+#define src_dt sd[AuSRC].dt
20145+#define src_bstart sd[AuSRC].bstart
1facf9fc 20146+
4a4d8108
AM
20147+#define dst_dentry sd[AuDST].dentry
20148+#define dst_dir sd[AuDST].dir
20149+#define dst_inode sd[AuDST].inode
20150+#define dst_h_dentry sd[AuDST].h_dentry
20151+#define dst_parent sd[AuDST].parent
20152+#define dst_h_parent sd[AuDST].h_parent
20153+#define dst_wh_dentry sd[AuDST].wh_dentry
20154+#define dst_hdir sd[AuDST].hdir
20155+#define dst_h_dir sd[AuDST].hdir->hi_inode
20156+#define dst_dt sd[AuDST].dt
20157+#define dst_bstart sd[AuDST].bstart
20158+
20159+ struct dentry *h_trap;
20160+ struct au_branch *br;
20161+ struct au_hinode *src_hinode;
20162+ struct path h_path;
20163+ struct au_nhash whlist;
027c5e7a 20164+ aufs_bindex_t btgt, src_bwh, src_bdiropq;
1facf9fc 20165+
1308ab2a 20166+ unsigned int flags;
1facf9fc 20167+
4a4d8108
AM
20168+ struct au_whtmp_rmdir *thargs;
20169+ struct dentry *h_dst;
20170+};
1308ab2a 20171+
4a4d8108 20172+/* ---------------------------------------------------------------------- */
1308ab2a 20173+
4a4d8108
AM
20174+/*
20175+ * functions for reverting.
20176+ * when an error happened in a single rename systemcall, we should revert
20177+ * everything as if nothing happend.
20178+ * we don't need to revert the copied-up/down the parent dir since they are
20179+ * harmless.
20180+ */
1facf9fc 20181+
4a4d8108
AM
20182+#define RevertFailure(fmt, ...) do { \
20183+ AuIOErr("revert failure: " fmt " (%d, %d)\n", \
20184+ ##__VA_ARGS__, err, rerr); \
20185+ err = -EIO; \
20186+} while (0)
1facf9fc 20187+
4a4d8108 20188+static void au_ren_rev_diropq(int err, struct au_ren_args *a)
1facf9fc 20189+{
4a4d8108 20190+ int rerr;
1facf9fc 20191+
4a4d8108
AM
20192+ au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
20193+ rerr = au_diropq_remove(a->src_dentry, a->btgt);
20194+ au_hn_imtx_unlock(a->src_hinode);
027c5e7a 20195+ au_set_dbdiropq(a->src_dentry, a->src_bdiropq);
4a4d8108 20196+ if (rerr)
523b37e3 20197+ RevertFailure("remove diropq %pd", a->src_dentry);
4a4d8108 20198+}
1facf9fc 20199+
4a4d8108
AM
20200+static void au_ren_rev_rename(int err, struct au_ren_args *a)
20201+{
20202+ int rerr;
523b37e3 20203+ struct inode *delegated;
1facf9fc 20204+
b4510431
AM
20205+ a->h_path.dentry = vfsub_lkup_one(&a->src_dentry->d_name,
20206+ a->src_h_parent);
4a4d8108
AM
20207+ rerr = PTR_ERR(a->h_path.dentry);
20208+ if (IS_ERR(a->h_path.dentry)) {
523b37e3 20209+ RevertFailure("lkup one %pd", a->src_dentry);
4a4d8108 20210+ return;
1facf9fc 20211+ }
20212+
523b37e3 20213+ delegated = NULL;
4a4d8108
AM
20214+ rerr = vfsub_rename(a->dst_h_dir,
20215+ au_h_dptr(a->src_dentry, a->btgt),
523b37e3
AM
20216+ a->src_h_dir, &a->h_path, &delegated);
20217+ if (unlikely(rerr == -EWOULDBLOCK)) {
20218+ pr_warn("cannot retry for NFSv4 delegation"
20219+ " for an internal rename\n");
20220+ iput(delegated);
20221+ }
4a4d8108
AM
20222+ d_drop(a->h_path.dentry);
20223+ dput(a->h_path.dentry);
20224+ /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
20225+ if (rerr)
523b37e3 20226+ RevertFailure("rename %pd", a->src_dentry);
1facf9fc 20227+}
20228+
4a4d8108 20229+static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
1facf9fc 20230+{
4a4d8108 20231+ int rerr;
523b37e3 20232+ struct inode *delegated;
dece6358 20233+
b4510431
AM
20234+ a->h_path.dentry = vfsub_lkup_one(&a->dst_dentry->d_name,
20235+ a->dst_h_parent);
4a4d8108
AM
20236+ rerr = PTR_ERR(a->h_path.dentry);
20237+ if (IS_ERR(a->h_path.dentry)) {
523b37e3 20238+ RevertFailure("lkup one %pd", a->dst_dentry);
4a4d8108
AM
20239+ return;
20240+ }
5527c038 20241+ if (d_is_positive(a->h_path.dentry)) {
4a4d8108
AM
20242+ d_drop(a->h_path.dentry);
20243+ dput(a->h_path.dentry);
20244+ return;
dece6358
AM
20245+ }
20246+
523b37e3
AM
20247+ delegated = NULL;
20248+ rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path,
20249+ &delegated);
20250+ if (unlikely(rerr == -EWOULDBLOCK)) {
20251+ pr_warn("cannot retry for NFSv4 delegation"
20252+ " for an internal rename\n");
20253+ iput(delegated);
20254+ }
4a4d8108
AM
20255+ d_drop(a->h_path.dentry);
20256+ dput(a->h_path.dentry);
20257+ if (!rerr)
20258+ au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
20259+ else
523b37e3 20260+ RevertFailure("rename %pd", a->h_dst);
4a4d8108 20261+}
1308ab2a 20262+
4a4d8108
AM
20263+static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
20264+{
20265+ int rerr;
1308ab2a 20266+
4a4d8108
AM
20267+ a->h_path.dentry = a->src_wh_dentry;
20268+ rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
027c5e7a 20269+ au_set_dbwh(a->src_dentry, a->src_bwh);
4a4d8108 20270+ if (rerr)
523b37e3 20271+ RevertFailure("unlink %pd", a->src_wh_dentry);
4a4d8108 20272+}
4a4d8108 20273+#undef RevertFailure
1facf9fc 20274+
1308ab2a 20275+/* ---------------------------------------------------------------------- */
20276+
4a4d8108
AM
20277+/*
20278+ * when we have to copyup the renaming entry, do it with the rename-target name
20279+ * in order to minimize the cost (the later actual rename is unnecessary).
20280+ * otherwise rename it on the target branch.
20281+ */
20282+static int au_ren_or_cpup(struct au_ren_args *a)
1facf9fc 20283+{
dece6358 20284+ int err;
4a4d8108 20285+ struct dentry *d;
523b37e3 20286+ struct inode *delegated;
1facf9fc 20287+
4a4d8108
AM
20288+ d = a->src_dentry;
20289+ if (au_dbstart(d) == a->btgt) {
20290+ a->h_path.dentry = a->dst_h_dentry;
20291+ if (au_ftest_ren(a->flags, DIROPQ)
20292+ && au_dbdiropq(d) == a->btgt)
20293+ au_fclr_ren(a->flags, DIROPQ);
20294+ AuDebugOn(au_dbstart(d) != a->btgt);
523b37e3 20295+ delegated = NULL;
4a4d8108 20296+ err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
523b37e3
AM
20297+ a->dst_h_dir, &a->h_path, &delegated);
20298+ if (unlikely(err == -EWOULDBLOCK)) {
20299+ pr_warn("cannot retry for NFSv4 delegation"
20300+ " for an internal rename\n");
20301+ iput(delegated);
20302+ }
c2b27bf2 20303+ } else
86dc4139 20304+ BUG();
1308ab2a 20305+
027c5e7a
AM
20306+ if (!err && a->h_dst)
20307+ /* it will be set to dinfo later */
20308+ dget(a->h_dst);
1facf9fc 20309+
dece6358
AM
20310+ return err;
20311+}
1facf9fc 20312+
4a4d8108
AM
20313+/* cf. aufs_rmdir() */
20314+static int au_ren_del_whtmp(struct au_ren_args *a)
dece6358 20315+{
4a4d8108
AM
20316+ int err;
20317+ struct inode *dir;
1facf9fc 20318+
4a4d8108
AM
20319+ dir = a->dst_dir;
20320+ SiMustAnyLock(dir->i_sb);
20321+ if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
20322+ au_sbi(dir->i_sb)->si_dirwh)
20323+ || au_test_fs_remote(a->h_dst->d_sb)) {
20324+ err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
20325+ if (unlikely(err))
523b37e3
AM
20326+ pr_warn("failed removing whtmp dir %pd (%d), "
20327+ "ignored.\n", a->h_dst, err);
4a4d8108
AM
20328+ } else {
20329+ au_nhash_wh_free(&a->thargs->whlist);
20330+ a->thargs->whlist = a->whlist;
20331+ a->whlist.nh_num = 0;
20332+ au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
20333+ dput(a->h_dst);
20334+ a->thargs = NULL;
20335+ }
20336+
20337+ return 0;
1308ab2a 20338+}
1facf9fc 20339+
4a4d8108
AM
20340+/* make it 'opaque' dir. */
20341+static int au_ren_diropq(struct au_ren_args *a)
20342+{
20343+ int err;
20344+ struct dentry *diropq;
1facf9fc 20345+
4a4d8108 20346+ err = 0;
027c5e7a 20347+ a->src_bdiropq = au_dbdiropq(a->src_dentry);
4a4d8108
AM
20348+ a->src_hinode = au_hi(a->src_inode, a->btgt);
20349+ au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
20350+ diropq = au_diropq_create(a->src_dentry, a->btgt);
20351+ au_hn_imtx_unlock(a->src_hinode);
20352+ if (IS_ERR(diropq))
20353+ err = PTR_ERR(diropq);
076b876e
AM
20354+ else
20355+ dput(diropq);
1facf9fc 20356+
4a4d8108
AM
20357+ return err;
20358+}
1facf9fc 20359+
4a4d8108
AM
20360+static int do_rename(struct au_ren_args *a)
20361+{
20362+ int err;
20363+ struct dentry *d, *h_d;
1facf9fc 20364+
4a4d8108
AM
20365+ /* prepare workqueue args for asynchronous rmdir */
20366+ h_d = a->dst_h_dentry;
5527c038 20367+ if (au_ftest_ren(a->flags, ISDIR) && d_is_positive(h_d)) {
4a4d8108
AM
20368+ err = -ENOMEM;
20369+ a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb, GFP_NOFS);
20370+ if (unlikely(!a->thargs))
20371+ goto out;
20372+ a->h_dst = dget(h_d);
20373+ }
1facf9fc 20374+
4a4d8108
AM
20375+ /* create whiteout for src_dentry */
20376+ if (au_ftest_ren(a->flags, WHSRC)) {
027c5e7a
AM
20377+ a->src_bwh = au_dbwh(a->src_dentry);
20378+ AuDebugOn(a->src_bwh >= 0);
4a4d8108
AM
20379+ a->src_wh_dentry
20380+ = au_wh_create(a->src_dentry, a->btgt, a->src_h_parent);
20381+ err = PTR_ERR(a->src_wh_dentry);
20382+ if (IS_ERR(a->src_wh_dentry))
20383+ goto out_thargs;
20384+ }
1facf9fc 20385+
4a4d8108
AM
20386+ /* lookup whiteout for dentry */
20387+ if (au_ftest_ren(a->flags, WHDST)) {
20388+ h_d = au_wh_lkup(a->dst_h_parent, &a->dst_dentry->d_name,
20389+ a->br);
20390+ err = PTR_ERR(h_d);
20391+ if (IS_ERR(h_d))
20392+ goto out_whsrc;
5527c038 20393+ if (d_is_negative(h_d))
4a4d8108
AM
20394+ dput(h_d);
20395+ else
20396+ a->dst_wh_dentry = h_d;
20397+ }
1facf9fc 20398+
4a4d8108
AM
20399+ /* rename dentry to tmpwh */
20400+ if (a->thargs) {
20401+ err = au_whtmp_ren(a->dst_h_dentry, a->br);
20402+ if (unlikely(err))
20403+ goto out_whdst;
dece6358 20404+
4a4d8108
AM
20405+ d = a->dst_dentry;
20406+ au_set_h_dptr(d, a->btgt, NULL);
86dc4139 20407+ err = au_lkup_neg(d, a->btgt, /*wh*/0);
4a4d8108
AM
20408+ if (unlikely(err))
20409+ goto out_whtmp;
20410+ a->dst_h_dentry = au_h_dptr(d, a->btgt);
20411+ }
1facf9fc 20412+
5527c038 20413+ BUG_ON(d_is_positive(a->dst_h_dentry) && a->src_bstart != a->btgt);
1facf9fc 20414+
4a4d8108
AM
20415+ /* rename by vfs_rename or cpup */
20416+ d = a->dst_dentry;
20417+ if (au_ftest_ren(a->flags, ISDIR)
20418+ && (a->dst_wh_dentry
20419+ || au_dbdiropq(d) == a->btgt
20420+ /* hide the lower to keep xino */
20421+ || a->btgt < au_dbend(d)
20422+ || au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ)))
20423+ au_fset_ren(a->flags, DIROPQ);
20424+ err = au_ren_or_cpup(a);
20425+ if (unlikely(err))
20426+ /* leave the copied-up one */
20427+ goto out_whtmp;
1308ab2a 20428+
4a4d8108
AM
20429+ /* make dir opaque */
20430+ if (au_ftest_ren(a->flags, DIROPQ)) {
20431+ err = au_ren_diropq(a);
20432+ if (unlikely(err))
20433+ goto out_rename;
20434+ }
1308ab2a 20435+
4a4d8108
AM
20436+ /* update target timestamps */
20437+ AuDebugOn(au_dbstart(a->src_dentry) != a->btgt);
20438+ a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
20439+ vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
5527c038 20440+ a->src_inode->i_ctime = d_inode(a->h_path.dentry)->i_ctime;
1facf9fc 20441+
4a4d8108
AM
20442+ /* remove whiteout for dentry */
20443+ if (a->dst_wh_dentry) {
20444+ a->h_path.dentry = a->dst_wh_dentry;
20445+ err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
20446+ a->dst_dentry);
20447+ if (unlikely(err))
20448+ goto out_diropq;
20449+ }
1facf9fc 20450+
4a4d8108
AM
20451+ /* remove whtmp */
20452+ if (a->thargs)
20453+ au_ren_del_whtmp(a); /* ignore this error */
1308ab2a 20454+
076b876e 20455+ au_fhsm_wrote(a->src_dentry->d_sb, a->btgt, /*force*/0);
4a4d8108
AM
20456+ err = 0;
20457+ goto out_success;
20458+
4f0767ce 20459+out_diropq:
4a4d8108
AM
20460+ if (au_ftest_ren(a->flags, DIROPQ))
20461+ au_ren_rev_diropq(err, a);
4f0767ce 20462+out_rename:
7e9cd9fe 20463+ au_ren_rev_rename(err, a);
027c5e7a 20464+ dput(a->h_dst);
4f0767ce 20465+out_whtmp:
4a4d8108
AM
20466+ if (a->thargs)
20467+ au_ren_rev_whtmp(err, a);
4f0767ce 20468+out_whdst:
4a4d8108
AM
20469+ dput(a->dst_wh_dentry);
20470+ a->dst_wh_dentry = NULL;
4f0767ce 20471+out_whsrc:
4a4d8108
AM
20472+ if (a->src_wh_dentry)
20473+ au_ren_rev_whsrc(err, a);
4f0767ce 20474+out_success:
4a4d8108
AM
20475+ dput(a->src_wh_dentry);
20476+ dput(a->dst_wh_dentry);
4f0767ce 20477+out_thargs:
4a4d8108
AM
20478+ if (a->thargs) {
20479+ dput(a->h_dst);
20480+ au_whtmp_rmdir_free(a->thargs);
20481+ a->thargs = NULL;
20482+ }
4f0767ce 20483+out:
4a4d8108 20484+ return err;
dece6358 20485+}
1facf9fc 20486+
1308ab2a 20487+/* ---------------------------------------------------------------------- */
1facf9fc 20488+
4a4d8108
AM
20489+/*
20490+ * test if @dentry dir can be rename destination or not.
20491+ * success means, it is a logically empty dir.
20492+ */
20493+static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
1308ab2a 20494+{
4a4d8108 20495+ return au_test_empty(dentry, whlist);
1308ab2a 20496+}
1facf9fc 20497+
4a4d8108
AM
20498+/*
20499+ * test if @dentry dir can be rename source or not.
20500+ * if it can, return 0 and @children is filled.
20501+ * success means,
20502+ * - it is a logically empty dir.
20503+ * - or, it exists on writable branch and has no children including whiteouts
20504+ * on the lower branch.
20505+ */
20506+static int may_rename_srcdir(struct dentry *dentry, aufs_bindex_t btgt)
20507+{
20508+ int err;
20509+ unsigned int rdhash;
20510+ aufs_bindex_t bstart;
1facf9fc 20511+
4a4d8108
AM
20512+ bstart = au_dbstart(dentry);
20513+ if (bstart != btgt) {
20514+ struct au_nhash whlist;
dece6358 20515+
4a4d8108
AM
20516+ SiMustAnyLock(dentry->d_sb);
20517+ rdhash = au_sbi(dentry->d_sb)->si_rdhash;
20518+ if (!rdhash)
20519+ rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
20520+ dentry));
20521+ err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
20522+ if (unlikely(err))
20523+ goto out;
20524+ err = au_test_empty(dentry, &whlist);
20525+ au_nhash_wh_free(&whlist);
20526+ goto out;
20527+ }
dece6358 20528+
4a4d8108
AM
20529+ if (bstart == au_dbtaildir(dentry))
20530+ return 0; /* success */
dece6358 20531+
4a4d8108 20532+ err = au_test_empty_lower(dentry);
1facf9fc 20533+
4f0767ce 20534+out:
4a4d8108
AM
20535+ if (err == -ENOTEMPTY) {
20536+ AuWarn1("renaming dir who has child(ren) on multiple branches,"
20537+ " is not supported\n");
20538+ err = -EXDEV;
20539+ }
20540+ return err;
20541+}
1308ab2a 20542+
4a4d8108
AM
20543+/* side effect: sets whlist and h_dentry */
20544+static int au_ren_may_dir(struct au_ren_args *a)
1308ab2a 20545+{
4a4d8108
AM
20546+ int err;
20547+ unsigned int rdhash;
20548+ struct dentry *d;
1facf9fc 20549+
4a4d8108
AM
20550+ d = a->dst_dentry;
20551+ SiMustAnyLock(d->d_sb);
1facf9fc 20552+
4a4d8108
AM
20553+ err = 0;
20554+ if (au_ftest_ren(a->flags, ISDIR) && a->dst_inode) {
20555+ rdhash = au_sbi(d->d_sb)->si_rdhash;
20556+ if (!rdhash)
20557+ rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
20558+ err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
20559+ if (unlikely(err))
20560+ goto out;
1308ab2a 20561+
4a4d8108
AM
20562+ au_set_dbstart(d, a->dst_bstart);
20563+ err = may_rename_dstdir(d, &a->whlist);
20564+ au_set_dbstart(d, a->btgt);
20565+ }
20566+ a->dst_h_dentry = au_h_dptr(d, au_dbstart(d));
20567+ if (unlikely(err))
20568+ goto out;
20569+
20570+ d = a->src_dentry;
20571+ a->src_h_dentry = au_h_dptr(d, au_dbstart(d));
20572+ if (au_ftest_ren(a->flags, ISDIR)) {
20573+ err = may_rename_srcdir(d, a->btgt);
20574+ if (unlikely(err)) {
20575+ au_nhash_wh_free(&a->whlist);
20576+ a->whlist.nh_num = 0;
20577+ }
20578+ }
4f0767ce 20579+out:
4a4d8108 20580+ return err;
1facf9fc 20581+}
20582+
4a4d8108 20583+/* ---------------------------------------------------------------------- */
1facf9fc 20584+
4a4d8108
AM
20585+/*
20586+ * simple tests for rename.
20587+ * following the checks in vfs, plus the parent-child relationship.
20588+ */
20589+static int au_may_ren(struct au_ren_args *a)
20590+{
20591+ int err, isdir;
20592+ struct inode *h_inode;
1facf9fc 20593+
4a4d8108
AM
20594+ if (a->src_bstart == a->btgt) {
20595+ err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
20596+ au_ftest_ren(a->flags, ISDIR));
20597+ if (unlikely(err))
20598+ goto out;
20599+ err = -EINVAL;
20600+ if (unlikely(a->src_h_dentry == a->h_trap))
20601+ goto out;
20602+ }
1facf9fc 20603+
4a4d8108
AM
20604+ err = 0;
20605+ if (a->dst_bstart != a->btgt)
20606+ goto out;
1facf9fc 20607+
027c5e7a
AM
20608+ err = -ENOTEMPTY;
20609+ if (unlikely(a->dst_h_dentry == a->h_trap))
20610+ goto out;
20611+
4a4d8108 20612+ err = -EIO;
4a4d8108 20613+ isdir = !!au_ftest_ren(a->flags, ISDIR);
5527c038
JR
20614+ if (d_really_is_negative(a->dst_dentry)) {
20615+ if (d_is_negative(a->dst_h_dentry))
20616+ err = au_may_add(a->dst_dentry, a->btgt,
20617+ a->dst_h_parent, isdir);
4a4d8108 20618+ } else {
5527c038 20619+ if (unlikely(d_is_negative(a->dst_h_dentry)))
4a4d8108 20620+ goto out;
5527c038
JR
20621+ h_inode = d_inode(a->dst_h_dentry);
20622+ if (h_inode->i_nlink)
20623+ err = au_may_del(a->dst_dentry, a->btgt,
20624+ a->dst_h_parent, isdir);
4a4d8108 20625+ }
1facf9fc 20626+
4f0767ce 20627+out:
4a4d8108
AM
20628+ if (unlikely(err == -ENOENT || err == -EEXIST))
20629+ err = -EIO;
20630+ AuTraceErr(err);
20631+ return err;
20632+}
1facf9fc 20633+
1308ab2a 20634+/* ---------------------------------------------------------------------- */
1facf9fc 20635+
4a4d8108
AM
20636+/*
20637+ * locking order
20638+ * (VFS)
20639+ * - src_dir and dir by lock_rename()
20640+ * - inode if exitsts
20641+ * (aufs)
20642+ * - lock all
20643+ * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
20644+ * + si_read_lock
20645+ * + di_write_lock2_child()
20646+ * + di_write_lock_child()
20647+ * + ii_write_lock_child()
20648+ * + di_write_lock_child2()
20649+ * + ii_write_lock_child2()
20650+ * + src_parent and parent
20651+ * + di_write_lock_parent()
20652+ * + ii_write_lock_parent()
20653+ * + di_write_lock_parent2()
20654+ * + ii_write_lock_parent2()
20655+ * + lower src_dir and dir by vfsub_lock_rename()
20656+ * + verify the every relationships between child and parent. if any
20657+ * of them failed, unlock all and return -EBUSY.
20658+ */
20659+static void au_ren_unlock(struct au_ren_args *a)
1308ab2a 20660+{
4a4d8108
AM
20661+ vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
20662+ a->dst_h_parent, a->dst_hdir);
86dc4139
AM
20663+ if (au_ftest_ren(a->flags, MNT_WRITE))
20664+ vfsub_mnt_drop_write(au_br_mnt(a->br));
1308ab2a 20665+}
20666+
4a4d8108 20667+static int au_ren_lock(struct au_ren_args *a)
1308ab2a 20668+{
4a4d8108
AM
20669+ int err;
20670+ unsigned int udba;
1308ab2a 20671+
4a4d8108
AM
20672+ err = 0;
20673+ a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
20674+ a->src_hdir = au_hi(a->src_dir, a->btgt);
20675+ a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
20676+ a->dst_hdir = au_hi(a->dst_dir, a->btgt);
86dc4139
AM
20677+
20678+ err = vfsub_mnt_want_write(au_br_mnt(a->br));
20679+ if (unlikely(err))
20680+ goto out;
20681+ au_fset_ren(a->flags, MNT_WRITE);
4a4d8108
AM
20682+ a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
20683+ a->dst_h_parent, a->dst_hdir);
20684+ udba = au_opt_udba(a->src_dentry->d_sb);
5527c038
JR
20685+ if (unlikely(a->src_hdir->hi_inode != d_inode(a->src_h_parent)
20686+ || a->dst_hdir->hi_inode != d_inode(a->dst_h_parent)))
4a4d8108
AM
20687+ err = au_busy_or_stale();
20688+ if (!err && au_dbstart(a->src_dentry) == a->btgt)
20689+ err = au_h_verify(a->src_h_dentry, udba,
5527c038 20690+ d_inode(a->src_h_parent), a->src_h_parent,
4a4d8108
AM
20691+ a->br);
20692+ if (!err && au_dbstart(a->dst_dentry) == a->btgt)
20693+ err = au_h_verify(a->dst_h_dentry, udba,
5527c038 20694+ d_inode(a->dst_h_parent), a->dst_h_parent,
4a4d8108 20695+ a->br);
86dc4139 20696+ if (!err)
4a4d8108 20697+ goto out; /* success */
4a4d8108
AM
20698+
20699+ err = au_busy_or_stale();
4a4d8108 20700+ au_ren_unlock(a);
86dc4139 20701+
4f0767ce 20702+out:
4a4d8108 20703+ return err;
1facf9fc 20704+}
20705+
20706+/* ---------------------------------------------------------------------- */
20707+
4a4d8108 20708+static void au_ren_refresh_dir(struct au_ren_args *a)
1facf9fc 20709+{
4a4d8108 20710+ struct inode *dir;
dece6358 20711+
4a4d8108
AM
20712+ dir = a->dst_dir;
20713+ dir->i_version++;
20714+ if (au_ftest_ren(a->flags, ISDIR)) {
20715+ /* is this updating defined in POSIX? */
20716+ au_cpup_attr_timesizes(a->src_inode);
20717+ au_cpup_attr_nlink(dir, /*force*/1);
4a4d8108 20718+ }
027c5e7a 20719+
b912730e 20720+ au_dir_ts(dir, a->btgt);
dece6358 20721+
4a4d8108
AM
20722+ if (au_ftest_ren(a->flags, ISSAMEDIR))
20723+ return;
dece6358 20724+
4a4d8108
AM
20725+ dir = a->src_dir;
20726+ dir->i_version++;
20727+ if (au_ftest_ren(a->flags, ISDIR))
20728+ au_cpup_attr_nlink(dir, /*force*/1);
b912730e 20729+ au_dir_ts(dir, a->btgt);
1facf9fc 20730+}
20731+
4a4d8108 20732+static void au_ren_refresh(struct au_ren_args *a)
1facf9fc 20733+{
4a4d8108
AM
20734+ aufs_bindex_t bend, bindex;
20735+ struct dentry *d, *h_d;
20736+ struct inode *i, *h_i;
20737+ struct super_block *sb;
dece6358 20738+
027c5e7a
AM
20739+ d = a->dst_dentry;
20740+ d_drop(d);
20741+ if (a->h_dst)
20742+ /* already dget-ed by au_ren_or_cpup() */
20743+ au_set_h_dptr(d, a->btgt, a->h_dst);
20744+
20745+ i = a->dst_inode;
20746+ if (i) {
20747+ if (!au_ftest_ren(a->flags, ISDIR))
20748+ vfsub_drop_nlink(i);
20749+ else {
20750+ vfsub_dead_dir(i);
20751+ au_cpup_attr_timesizes(i);
20752+ }
20753+ au_update_dbrange(d, /*do_put_zero*/1);
20754+ } else {
20755+ bend = a->btgt;
20756+ for (bindex = au_dbstart(d); bindex < bend; bindex++)
20757+ au_set_h_dptr(d, bindex, NULL);
20758+ bend = au_dbend(d);
20759+ for (bindex = a->btgt + 1; bindex <= bend; bindex++)
20760+ au_set_h_dptr(d, bindex, NULL);
20761+ au_update_dbrange(d, /*do_put_zero*/0);
20762+ }
20763+
4a4d8108
AM
20764+ d = a->src_dentry;
20765+ au_set_dbwh(d, -1);
20766+ bend = au_dbend(d);
20767+ for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
20768+ h_d = au_h_dptr(d, bindex);
20769+ if (h_d)
20770+ au_set_h_dptr(d, bindex, NULL);
20771+ }
20772+ au_set_dbend(d, a->btgt);
20773+
20774+ sb = d->d_sb;
20775+ i = a->src_inode;
20776+ if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
20777+ return; /* success */
20778+
20779+ bend = au_ibend(i);
20780+ for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
20781+ h_i = au_h_iptr(i, bindex);
20782+ if (h_i) {
20783+ au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
20784+ /* ignore this error */
20785+ au_set_h_iptr(i, bindex, NULL, 0);
20786+ }
20787+ }
20788+ au_set_ibend(i, a->btgt);
1308ab2a 20789+}
dece6358 20790+
4a4d8108
AM
20791+/* ---------------------------------------------------------------------- */
20792+
20793+/* mainly for link(2) and rename(2) */
20794+int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
1308ab2a 20795+{
4a4d8108
AM
20796+ aufs_bindex_t bdiropq, bwh;
20797+ struct dentry *parent;
20798+ struct au_branch *br;
20799+
20800+ parent = dentry->d_parent;
5527c038 20801+ IMustLock(d_inode(parent)); /* dir is locked */
4a4d8108
AM
20802+
20803+ bdiropq = au_dbdiropq(parent);
20804+ bwh = au_dbwh(dentry);
20805+ br = au_sbr(dentry->d_sb, btgt);
20806+ if (au_br_rdonly(br)
20807+ || (0 <= bdiropq && bdiropq < btgt)
20808+ || (0 <= bwh && bwh < btgt))
20809+ btgt = -1;
20810+
20811+ AuDbg("btgt %d\n", btgt);
20812+ return btgt;
1facf9fc 20813+}
20814+
4a4d8108
AM
20815+/* sets src_bstart, dst_bstart and btgt */
20816+static int au_ren_wbr(struct au_ren_args *a)
1facf9fc 20817+{
4a4d8108
AM
20818+ int err;
20819+ struct au_wr_dir_args wr_dir_args = {
20820+ /* .force_btgt = -1, */
20821+ .flags = AuWrDir_ADD_ENTRY
20822+ };
dece6358 20823+
4a4d8108
AM
20824+ a->src_bstart = au_dbstart(a->src_dentry);
20825+ a->dst_bstart = au_dbstart(a->dst_dentry);
20826+ if (au_ftest_ren(a->flags, ISDIR))
20827+ au_fset_wrdir(wr_dir_args.flags, ISDIR);
20828+ wr_dir_args.force_btgt = a->src_bstart;
20829+ if (a->dst_inode && a->dst_bstart < a->src_bstart)
20830+ wr_dir_args.force_btgt = a->dst_bstart;
20831+ wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
20832+ err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
20833+ a->btgt = err;
dece6358 20834+
4a4d8108 20835+ return err;
1facf9fc 20836+}
20837+
4a4d8108 20838+static void au_ren_dt(struct au_ren_args *a)
1facf9fc 20839+{
4a4d8108
AM
20840+ a->h_path.dentry = a->src_h_parent;
20841+ au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
20842+ if (!au_ftest_ren(a->flags, ISSAMEDIR)) {
20843+ a->h_path.dentry = a->dst_h_parent;
20844+ au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
20845+ }
1facf9fc 20846+
4a4d8108
AM
20847+ au_fclr_ren(a->flags, DT_DSTDIR);
20848+ if (!au_ftest_ren(a->flags, ISDIR))
20849+ return;
dece6358 20850+
4a4d8108
AM
20851+ a->h_path.dentry = a->src_h_dentry;
20852+ au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
5527c038 20853+ if (d_is_positive(a->dst_h_dentry)) {
4a4d8108
AM
20854+ au_fset_ren(a->flags, DT_DSTDIR);
20855+ a->h_path.dentry = a->dst_h_dentry;
20856+ au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
20857+ }
1308ab2a 20858+}
dece6358 20859+
4a4d8108 20860+static void au_ren_rev_dt(int err, struct au_ren_args *a)
1308ab2a 20861+{
4a4d8108
AM
20862+ struct dentry *h_d;
20863+ struct mutex *h_mtx;
20864+
20865+ au_dtime_revert(a->src_dt + AuPARENT);
20866+ if (!au_ftest_ren(a->flags, ISSAMEDIR))
20867+ au_dtime_revert(a->dst_dt + AuPARENT);
20868+
20869+ if (au_ftest_ren(a->flags, ISDIR) && err != -EIO) {
20870+ h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
5527c038 20871+ h_mtx = &d_inode(h_d)->i_mutex;
4a4d8108
AM
20872+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
20873+ au_dtime_revert(a->src_dt + AuCHILD);
20874+ mutex_unlock(h_mtx);
20875+
20876+ if (au_ftest_ren(a->flags, DT_DSTDIR)) {
20877+ h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
5527c038 20878+ h_mtx = &d_inode(h_d)->i_mutex;
4a4d8108
AM
20879+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
20880+ au_dtime_revert(a->dst_dt + AuCHILD);
20881+ mutex_unlock(h_mtx);
1facf9fc 20882+ }
20883+ }
20884+}
20885+
4a4d8108
AM
20886+/* ---------------------------------------------------------------------- */
20887+
20888+int aufs_rename(struct inode *_src_dir, struct dentry *_src_dentry,
20889+ struct inode *_dst_dir, struct dentry *_dst_dentry)
1facf9fc 20890+{
e49829fe 20891+ int err, flags;
4a4d8108
AM
20892+ /* reduce stack space */
20893+ struct au_ren_args *a;
20894+
523b37e3 20895+ AuDbg("%pd, %pd\n", _src_dentry, _dst_dentry);
4a4d8108
AM
20896+ IMustLock(_src_dir);
20897+ IMustLock(_dst_dir);
20898+
20899+ err = -ENOMEM;
20900+ BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
20901+ a = kzalloc(sizeof(*a), GFP_NOFS);
20902+ if (unlikely(!a))
20903+ goto out;
20904+
20905+ a->src_dir = _src_dir;
20906+ a->src_dentry = _src_dentry;
5527c038
JR
20907+ a->src_inode = NULL;
20908+ if (d_really_is_positive(a->src_dentry))
20909+ a->src_inode = d_inode(a->src_dentry);
4a4d8108
AM
20910+ a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
20911+ a->dst_dir = _dst_dir;
20912+ a->dst_dentry = _dst_dentry;
5527c038
JR
20913+ a->dst_inode = NULL;
20914+ if (d_really_is_positive(a->dst_dentry))
20915+ a->dst_inode = d_inode(a->dst_dentry);
4a4d8108
AM
20916+ a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
20917+ if (a->dst_inode) {
20918+ IMustLock(a->dst_inode);
20919+ au_igrab(a->dst_inode);
1facf9fc 20920+ }
1facf9fc 20921+
4a4d8108 20922+ err = -ENOTDIR;
027c5e7a 20923+ flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
2000de60 20924+ if (d_is_dir(a->src_dentry)) {
4a4d8108 20925+ au_fset_ren(a->flags, ISDIR);
5527c038 20926+ if (unlikely(d_really_is_positive(a->dst_dentry)
2000de60 20927+ && !d_is_dir(a->dst_dentry)))
4a4d8108 20928+ goto out_free;
e49829fe
JR
20929+ err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
20930+ AuLock_DIR | flags);
4a4d8108 20931+ } else
e49829fe
JR
20932+ err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
20933+ flags);
20934+ if (unlikely(err))
20935+ goto out_free;
1facf9fc 20936+
027c5e7a
AM
20937+ err = au_d_hashed_positive(a->src_dentry);
20938+ if (unlikely(err))
20939+ goto out_unlock;
20940+ err = -ENOENT;
20941+ if (a->dst_inode) {
20942+ /*
20943+ * If it is a dir, VFS unhash dst_dentry before this
20944+ * function. It means we cannot rely upon d_unhashed().
20945+ */
20946+ if (unlikely(!a->dst_inode->i_nlink))
20947+ goto out_unlock;
20948+ if (!S_ISDIR(a->dst_inode->i_mode)) {
20949+ err = au_d_hashed_positive(a->dst_dentry);
20950+ if (unlikely(err))
20951+ goto out_unlock;
20952+ } else if (unlikely(IS_DEADDIR(a->dst_inode)))
20953+ goto out_unlock;
20954+ } else if (unlikely(d_unhashed(a->dst_dentry)))
20955+ goto out_unlock;
20956+
7eafdf33
AM
20957+ /*
20958+ * is it possible?
20959+ * yes, it happend (in linux-3.3-rcN) but I don't know why.
20960+ * there may exist a problem somewhere else.
20961+ */
20962+ err = -EINVAL;
5527c038 20963+ if (unlikely(d_inode(a->dst_parent) == d_inode(a->src_dentry)))
7eafdf33
AM
20964+ goto out_unlock;
20965+
4a4d8108
AM
20966+ au_fset_ren(a->flags, ISSAMEDIR); /* temporary */
20967+ di_write_lock_parent(a->dst_parent);
1facf9fc 20968+
4a4d8108
AM
20969+ /* which branch we process */
20970+ err = au_ren_wbr(a);
20971+ if (unlikely(err < 0))
027c5e7a 20972+ goto out_parent;
4a4d8108 20973+ a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
86dc4139 20974+ a->h_path.mnt = au_br_mnt(a->br);
1facf9fc 20975+
4a4d8108
AM
20976+ /* are they available to be renamed */
20977+ err = au_ren_may_dir(a);
20978+ if (unlikely(err))
20979+ goto out_children;
1facf9fc 20980+
4a4d8108
AM
20981+ /* prepare the writable parent dir on the same branch */
20982+ if (a->dst_bstart == a->btgt) {
20983+ au_fset_ren(a->flags, WHDST);
20984+ } else {
20985+ err = au_cpup_dirs(a->dst_dentry, a->btgt);
20986+ if (unlikely(err))
20987+ goto out_children;
20988+ }
1facf9fc 20989+
4a4d8108
AM
20990+ if (a->src_dir != a->dst_dir) {
20991+ /*
20992+ * this temporary unlock is safe,
20993+ * because both dir->i_mutex are locked.
20994+ */
20995+ di_write_unlock(a->dst_parent);
20996+ di_write_lock_parent(a->src_parent);
20997+ err = au_wr_dir_need_wh(a->src_dentry,
20998+ au_ftest_ren(a->flags, ISDIR),
20999+ &a->btgt);
21000+ di_write_unlock(a->src_parent);
21001+ di_write_lock2_parent(a->src_parent, a->dst_parent, /*isdir*/1);
21002+ au_fclr_ren(a->flags, ISSAMEDIR);
21003+ } else
21004+ err = au_wr_dir_need_wh(a->src_dentry,
21005+ au_ftest_ren(a->flags, ISDIR),
21006+ &a->btgt);
21007+ if (unlikely(err < 0))
21008+ goto out_children;
21009+ if (err)
21010+ au_fset_ren(a->flags, WHSRC);
1facf9fc 21011+
86dc4139
AM
21012+ /* cpup src */
21013+ if (a->src_bstart != a->btgt) {
86dc4139
AM
21014+ struct au_pin pin;
21015+
21016+ err = au_pin(&pin, a->src_dentry, a->btgt,
21017+ au_opt_udba(a->src_dentry->d_sb),
21018+ AuPin_DI_LOCKED | AuPin_MNT_WRITE);
367653fa 21019+ if (!err) {
c2b27bf2
AM
21020+ struct au_cp_generic cpg = {
21021+ .dentry = a->src_dentry,
21022+ .bdst = a->btgt,
21023+ .bsrc = a->src_bstart,
21024+ .len = -1,
21025+ .pin = &pin,
21026+ .flags = AuCpup_DTIME | AuCpup_HOPEN
21027+ };
367653fa 21028+ AuDebugOn(au_dbstart(a->src_dentry) != a->src_bstart);
c2b27bf2 21029+ err = au_sio_cpup_simple(&cpg);
367653fa 21030+ au_unpin(&pin);
86dc4139 21031+ }
86dc4139
AM
21032+ if (unlikely(err))
21033+ goto out_children;
21034+ a->src_bstart = a->btgt;
21035+ a->src_h_dentry = au_h_dptr(a->src_dentry, a->btgt);
21036+ au_fset_ren(a->flags, WHSRC);
21037+ }
21038+
4a4d8108
AM
21039+ /* lock them all */
21040+ err = au_ren_lock(a);
21041+ if (unlikely(err))
86dc4139 21042+ /* leave the copied-up one */
4a4d8108 21043+ goto out_children;
1facf9fc 21044+
4a4d8108
AM
21045+ if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
21046+ err = au_may_ren(a);
21047+ else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
21048+ err = -ENAMETOOLONG;
21049+ if (unlikely(err))
21050+ goto out_hdir;
1facf9fc 21051+
4a4d8108
AM
21052+ /* store timestamps to be revertible */
21053+ au_ren_dt(a);
1facf9fc 21054+
4a4d8108
AM
21055+ /* here we go */
21056+ err = do_rename(a);
21057+ if (unlikely(err))
21058+ goto out_dt;
21059+
21060+ /* update dir attributes */
21061+ au_ren_refresh_dir(a);
21062+
21063+ /* dput/iput all lower dentries */
21064+ au_ren_refresh(a);
21065+
21066+ goto out_hdir; /* success */
21067+
4f0767ce 21068+out_dt:
4a4d8108 21069+ au_ren_rev_dt(err, a);
4f0767ce 21070+out_hdir:
4a4d8108 21071+ au_ren_unlock(a);
4f0767ce 21072+out_children:
4a4d8108 21073+ au_nhash_wh_free(&a->whlist);
027c5e7a
AM
21074+ if (err && a->dst_inode && a->dst_bstart != a->btgt) {
21075+ AuDbg("bstart %d, btgt %d\n", a->dst_bstart, a->btgt);
21076+ au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
21077+ au_set_dbstart(a->dst_dentry, a->dst_bstart);
4a4d8108 21078+ }
027c5e7a 21079+out_parent:
4a4d8108
AM
21080+ if (!err)
21081+ d_move(a->src_dentry, a->dst_dentry);
027c5e7a
AM
21082+ else {
21083+ au_update_dbstart(a->dst_dentry);
21084+ if (!a->dst_inode)
21085+ d_drop(a->dst_dentry);
21086+ }
4a4d8108
AM
21087+ if (au_ftest_ren(a->flags, ISSAMEDIR))
21088+ di_write_unlock(a->dst_parent);
21089+ else
21090+ di_write_unlock2(a->src_parent, a->dst_parent);
027c5e7a 21091+out_unlock:
4a4d8108 21092+ aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
4f0767ce 21093+out_free:
4a4d8108
AM
21094+ iput(a->dst_inode);
21095+ if (a->thargs)
21096+ au_whtmp_rmdir_free(a->thargs);
21097+ kfree(a);
4f0767ce 21098+out:
4a4d8108
AM
21099+ AuTraceErr(err);
21100+ return err;
1308ab2a 21101+}
7f207e10
AM
21102diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
21103--- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21104+++ linux/fs/aufs/Kconfig 2015-09-24 10:47:58.248052907 +0200
c1595e42 21105@@ -0,0 +1,185 @@
4a4d8108
AM
21106+config AUFS_FS
21107+ tristate "Aufs (Advanced multi layered unification filesystem) support"
4a4d8108
AM
21108+ help
21109+ Aufs is a stackable unification filesystem such as Unionfs,
21110+ which unifies several directories and provides a merged single
21111+ directory.
21112+ In the early days, aufs was entirely re-designed and
21113+ re-implemented Unionfs Version 1.x series. Introducing many
21114+ original ideas, approaches and improvements, it becomes totally
21115+ different from Unionfs while keeping the basic features.
1facf9fc 21116+
4a4d8108
AM
21117+if AUFS_FS
21118+choice
21119+ prompt "Maximum number of branches"
21120+ default AUFS_BRANCH_MAX_127
21121+ help
21122+ Specifies the maximum number of branches (or member directories)
21123+ in a single aufs. The larger value consumes more system
21124+ resources and has a minor impact to performance.
21125+config AUFS_BRANCH_MAX_127
21126+ bool "127"
21127+ help
21128+ Specifies the maximum number of branches (or member directories)
21129+ in a single aufs. The larger value consumes more system
21130+ resources and has a minor impact to performance.
21131+config AUFS_BRANCH_MAX_511
21132+ bool "511"
21133+ help
21134+ Specifies the maximum number of branches (or member directories)
21135+ in a single aufs. The larger value consumes more system
21136+ resources and has a minor impact to performance.
21137+config AUFS_BRANCH_MAX_1023
21138+ bool "1023"
21139+ help
21140+ Specifies the maximum number of branches (or member directories)
21141+ in a single aufs. The larger value consumes more system
21142+ resources and has a minor impact to performance.
21143+config AUFS_BRANCH_MAX_32767
21144+ bool "32767"
21145+ help
21146+ Specifies the maximum number of branches (or member directories)
21147+ in a single aufs. The larger value consumes more system
21148+ resources and has a minor impact to performance.
21149+endchoice
1facf9fc 21150+
e49829fe
JR
21151+config AUFS_SBILIST
21152+ bool
21153+ depends on AUFS_MAGIC_SYSRQ || PROC_FS
21154+ default y
21155+ help
21156+ Automatic configuration for internal use.
21157+ When aufs supports Magic SysRq or /proc, enabled automatically.
21158+
4a4d8108
AM
21159+config AUFS_HNOTIFY
21160+ bool "Detect direct branch access (bypassing aufs)"
21161+ help
21162+ If you want to modify files on branches directly, eg. bypassing aufs,
21163+ and want aufs to detect the changes of them fully, then enable this
21164+ option and use 'udba=notify' mount option.
7f207e10 21165+ Currently there is only one available configuration, "fsnotify".
4a4d8108
AM
21166+ It will have a negative impact to the performance.
21167+ See detail in aufs.5.
dece6358 21168+
4a4d8108
AM
21169+choice
21170+ prompt "method" if AUFS_HNOTIFY
21171+ default AUFS_HFSNOTIFY
21172+config AUFS_HFSNOTIFY
21173+ bool "fsnotify"
21174+ select FSNOTIFY
4a4d8108 21175+endchoice
1facf9fc 21176+
4a4d8108
AM
21177+config AUFS_EXPORT
21178+ bool "NFS-exportable aufs"
2cbb1c4b 21179+ depends on EXPORTFS
4a4d8108
AM
21180+ help
21181+ If you want to export your mounted aufs via NFS, then enable this
21182+ option. There are several requirements for this configuration.
21183+ See detail in aufs.5.
1facf9fc 21184+
4a4d8108
AM
21185+config AUFS_INO_T_64
21186+ bool
21187+ depends on AUFS_EXPORT
21188+ depends on 64BIT && !(ALPHA || S390)
21189+ default y
21190+ help
21191+ Automatic configuration for internal use.
21192+ /* typedef unsigned long/int __kernel_ino_t */
21193+ /* alpha and s390x are int */
1facf9fc 21194+
c1595e42
JR
21195+config AUFS_XATTR
21196+ bool "support for XATTR/EA (including Security Labels)"
21197+ help
21198+ If your branch fs supports XATTR/EA and you want to make them
21199+ available in aufs too, then enable this opsion and specify the
21200+ branch attributes for EA.
21201+ See detail in aufs.5.
21202+
076b876e
AM
21203+config AUFS_FHSM
21204+ bool "File-based Hierarchical Storage Management"
21205+ help
21206+ Hierarchical Storage Management (or HSM) is a well-known feature
21207+ in the storage world. Aufs provides this feature as file-based.
21208+ with multiple branches.
21209+ These multiple branches are prioritized, ie. the topmost one
21210+ should be the fastest drive and be used heavily.
21211+
4a4d8108
AM
21212+config AUFS_RDU
21213+ bool "Readdir in userspace"
21214+ help
21215+ Aufs has two methods to provide a merged view for a directory,
21216+ by a user-space library and by kernel-space natively. The latter
21217+ is always enabled but sometimes large and slow.
21218+ If you enable this option, install the library in aufs2-util
21219+ package, and set some environment variables for your readdir(3),
21220+ then the work will be handled in user-space which generally
21221+ shows better performance in most cases.
21222+ See detail in aufs.5.
1facf9fc 21223+
4a4d8108
AM
21224+config AUFS_SHWH
21225+ bool "Show whiteouts"
21226+ help
21227+ If you want to make the whiteouts in aufs visible, then enable
21228+ this option and specify 'shwh' mount option. Although it may
21229+ sounds like philosophy or something, but in technically it
21230+ simply shows the name of whiteout with keeping its behaviour.
1facf9fc 21231+
4a4d8108
AM
21232+config AUFS_BR_RAMFS
21233+ bool "Ramfs (initramfs/rootfs) as an aufs branch"
21234+ help
21235+ If you want to use ramfs as an aufs branch fs, then enable this
21236+ option. Generally tmpfs is recommended.
21237+ Aufs prohibited them to be a branch fs by default, because
21238+ initramfs becomes unusable after switch_root or something
21239+ generally. If you sets initramfs as an aufs branch and boot your
21240+ system by switch_root, you will meet a problem easily since the
21241+ files in initramfs may be inaccessible.
21242+ Unless you are going to use ramfs as an aufs branch fs without
21243+ switch_root or something, leave it N.
1facf9fc 21244+
4a4d8108
AM
21245+config AUFS_BR_FUSE
21246+ bool "Fuse fs as an aufs branch"
21247+ depends on FUSE_FS
21248+ select AUFS_POLL
21249+ help
21250+ If you want to use fuse-based userspace filesystem as an aufs
21251+ branch fs, then enable this option.
21252+ It implements the internal poll(2) operation which is
21253+ implemented by fuse only (curretnly).
1facf9fc 21254+
4a4d8108
AM
21255+config AUFS_POLL
21256+ bool
21257+ help
21258+ Automatic configuration for internal use.
1facf9fc 21259+
4a4d8108
AM
21260+config AUFS_BR_HFSPLUS
21261+ bool "Hfsplus as an aufs branch"
21262+ depends on HFSPLUS_FS
21263+ default y
21264+ help
21265+ If you want to use hfsplus fs as an aufs branch fs, then enable
21266+ this option. This option introduces a small overhead at
21267+ copying-up a file on hfsplus.
1facf9fc 21268+
4a4d8108
AM
21269+config AUFS_BDEV_LOOP
21270+ bool
21271+ depends on BLK_DEV_LOOP
21272+ default y
21273+ help
21274+ Automatic configuration for internal use.
21275+ Convert =[ym] into =y.
1308ab2a 21276+
4a4d8108
AM
21277+config AUFS_DEBUG
21278+ bool "Debug aufs"
21279+ help
21280+ Enable this to compile aufs internal debug code.
21281+ It will have a negative impact to the performance.
21282+
21283+config AUFS_MAGIC_SYSRQ
21284+ bool
21285+ depends on AUFS_DEBUG && MAGIC_SYSRQ
21286+ default y
21287+ help
21288+ Automatic configuration for internal use.
21289+ When aufs supports Magic SysRq, enabled automatically.
21290+endif
7f207e10
AM
21291diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
21292--- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21293+++ linux/fs/aufs/loop.c 2015-09-24 10:47:58.254719746 +0200
523b37e3 21294@@ -0,0 +1,145 @@
1facf9fc 21295+/*
2000de60 21296+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 21297+ *
21298+ * This program, aufs is free software; you can redistribute it and/or modify
21299+ * it under the terms of the GNU General Public License as published by
21300+ * the Free Software Foundation; either version 2 of the License, or
21301+ * (at your option) any later version.
dece6358
AM
21302+ *
21303+ * This program is distributed in the hope that it will be useful,
21304+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
21305+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21306+ * GNU General Public License for more details.
21307+ *
21308+ * You should have received a copy of the GNU General Public License
523b37e3 21309+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 21310+ */
21311+
21312+/*
21313+ * support for loopback block device as a branch
21314+ */
21315+
1facf9fc 21316+#include "aufs.h"
21317+
392086de
AM
21318+/* added into drivers/block/loop.c */
21319+static struct file *(*backing_file_func)(struct super_block *sb);
21320+
1facf9fc 21321+/*
21322+ * test if two lower dentries have overlapping branches.
21323+ */
b752ccd1 21324+int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
1facf9fc 21325+{
b752ccd1 21326+ struct super_block *h_sb;
392086de
AM
21327+ struct file *backing_file;
21328+
21329+ if (unlikely(!backing_file_func)) {
21330+ /* don't load "loop" module here */
21331+ backing_file_func = symbol_get(loop_backing_file);
21332+ if (unlikely(!backing_file_func))
21333+ /* "loop" module is not loaded */
21334+ return 0;
21335+ }
1facf9fc 21336+
b752ccd1 21337+ h_sb = h_adding->d_sb;
392086de
AM
21338+ backing_file = backing_file_func(h_sb);
21339+ if (!backing_file)
1facf9fc 21340+ return 0;
21341+
2000de60 21342+ h_adding = backing_file->f_path.dentry;
b752ccd1
AM
21343+ /*
21344+ * h_adding can be local NFS.
21345+ * in this case aufs cannot detect the loop.
21346+ */
21347+ if (unlikely(h_adding->d_sb == sb))
1facf9fc 21348+ return 1;
b752ccd1 21349+ return !!au_test_subdir(h_adding, sb->s_root);
1facf9fc 21350+}
21351+
21352+/* true if a kernel thread named 'loop[0-9].*' accesses a file */
21353+int au_test_loopback_kthread(void)
21354+{
b752ccd1
AM
21355+ int ret;
21356+ struct task_struct *tsk = current;
a2a7ad62 21357+ char c, comm[sizeof(tsk->comm)];
b752ccd1
AM
21358+
21359+ ret = 0;
21360+ if (tsk->flags & PF_KTHREAD) {
a2a7ad62
AM
21361+ get_task_comm(comm, tsk);
21362+ c = comm[4];
b752ccd1 21363+ ret = ('0' <= c && c <= '9'
a2a7ad62 21364+ && !strncmp(comm, "loop", 4));
b752ccd1 21365+ }
1facf9fc 21366+
b752ccd1 21367+ return ret;
1facf9fc 21368+}
87a755f4
AM
21369+
21370+/* ---------------------------------------------------------------------- */
21371+
21372+#define au_warn_loopback_step 16
21373+static int au_warn_loopback_nelem = au_warn_loopback_step;
21374+static unsigned long *au_warn_loopback_array;
21375+
21376+void au_warn_loopback(struct super_block *h_sb)
21377+{
21378+ int i, new_nelem;
21379+ unsigned long *a, magic;
21380+ static DEFINE_SPINLOCK(spin);
21381+
21382+ magic = h_sb->s_magic;
21383+ spin_lock(&spin);
21384+ a = au_warn_loopback_array;
21385+ for (i = 0; i < au_warn_loopback_nelem && *a; i++)
21386+ if (a[i] == magic) {
21387+ spin_unlock(&spin);
21388+ return;
21389+ }
21390+
21391+ /* h_sb is new to us, print it */
21392+ if (i < au_warn_loopback_nelem) {
21393+ a[i] = magic;
21394+ goto pr;
21395+ }
21396+
21397+ /* expand the array */
21398+ new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
21399+ a = au_kzrealloc(au_warn_loopback_array,
21400+ au_warn_loopback_nelem * sizeof(unsigned long),
21401+ new_nelem * sizeof(unsigned long), GFP_ATOMIC);
21402+ if (a) {
21403+ au_warn_loopback_nelem = new_nelem;
21404+ au_warn_loopback_array = a;
21405+ a[i] = magic;
21406+ goto pr;
21407+ }
21408+
21409+ spin_unlock(&spin);
21410+ AuWarn1("realloc failed, ignored\n");
21411+ return;
21412+
21413+pr:
21414+ spin_unlock(&spin);
0c3ec466
AM
21415+ pr_warn("you may want to try another patch for loopback file "
21416+ "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
87a755f4
AM
21417+}
21418+
21419+int au_loopback_init(void)
21420+{
21421+ int err;
21422+ struct super_block *sb __maybe_unused;
21423+
21424+ AuDebugOn(sizeof(sb->s_magic) != sizeof(unsigned long));
21425+
21426+ err = 0;
21427+ au_warn_loopback_array = kcalloc(au_warn_loopback_step,
21428+ sizeof(unsigned long), GFP_NOFS);
21429+ if (unlikely(!au_warn_loopback_array))
21430+ err = -ENOMEM;
21431+
21432+ return err;
21433+}
21434+
21435+void au_loopback_fin(void)
21436+{
392086de 21437+ symbol_put(loop_backing_file);
87a755f4
AM
21438+ kfree(au_warn_loopback_array);
21439+}
7f207e10
AM
21440diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
21441--- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21442+++ linux/fs/aufs/loop.h 2015-09-24 10:47:58.254719746 +0200
523b37e3 21443@@ -0,0 +1,52 @@
1facf9fc 21444+/*
2000de60 21445+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 21446+ *
21447+ * This program, aufs is free software; you can redistribute it and/or modify
21448+ * it under the terms of the GNU General Public License as published by
21449+ * the Free Software Foundation; either version 2 of the License, or
21450+ * (at your option) any later version.
dece6358
AM
21451+ *
21452+ * This program is distributed in the hope that it will be useful,
21453+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
21454+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21455+ * GNU General Public License for more details.
21456+ *
21457+ * You should have received a copy of the GNU General Public License
523b37e3 21458+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 21459+ */
21460+
21461+/*
21462+ * support for loopback mount as a branch
21463+ */
21464+
21465+#ifndef __AUFS_LOOP_H__
21466+#define __AUFS_LOOP_H__
21467+
21468+#ifdef __KERNEL__
21469+
dece6358
AM
21470+struct dentry;
21471+struct super_block;
1facf9fc 21472+
21473+#ifdef CONFIG_AUFS_BDEV_LOOP
392086de
AM
21474+/* drivers/block/loop.c */
21475+struct file *loop_backing_file(struct super_block *sb);
21476+
1facf9fc 21477+/* loop.c */
b752ccd1 21478+int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
1facf9fc 21479+int au_test_loopback_kthread(void);
87a755f4
AM
21480+void au_warn_loopback(struct super_block *h_sb);
21481+
21482+int au_loopback_init(void);
21483+void au_loopback_fin(void);
1facf9fc 21484+#else
4a4d8108 21485+AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
b752ccd1 21486+ struct dentry *h_adding)
4a4d8108 21487+AuStubInt0(au_test_loopback_kthread, void)
87a755f4
AM
21488+AuStubVoid(au_warn_loopback, struct super_block *h_sb)
21489+
21490+AuStubInt0(au_loopback_init, void)
21491+AuStubVoid(au_loopback_fin, void)
1facf9fc 21492+#endif /* BLK_DEV_LOOP */
21493+
21494+#endif /* __KERNEL__ */
21495+#endif /* __AUFS_LOOP_H__ */
7f207e10
AM
21496diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
21497--- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21498+++ linux/fs/aufs/magic.mk 2015-09-24 10:47:58.254719746 +0200
7e9cd9fe 21499@@ -0,0 +1,30 @@
1facf9fc 21500+
21501+# defined in ${srctree}/fs/fuse/inode.c
21502+# tristate
21503+ifdef CONFIG_FUSE_FS
21504+ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
21505+endif
21506+
1facf9fc 21507+# defined in ${srctree}/fs/xfs/xfs_sb.h
21508+# tristate
21509+ifdef CONFIG_XFS_FS
21510+ccflags-y += -DXFS_SB_MAGIC=0x58465342
21511+endif
21512+
21513+# defined in ${srctree}/fs/configfs/mount.c
21514+# tristate
21515+ifdef CONFIG_CONFIGFS_FS
21516+ccflags-y += -DCONFIGFS_MAGIC=0x62656570
21517+endif
21518+
1facf9fc 21519+# defined in ${srctree}/fs/ubifs/ubifs.h
21520+# tristate
21521+ifdef CONFIG_UBIFS_FS
21522+ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
21523+endif
4a4d8108
AM
21524+
21525+# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
21526+# tristate
21527+ifdef CONFIG_HFSPLUS_FS
21528+ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
21529+endif
7f207e10
AM
21530diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
21531--- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21532+++ linux/fs/aufs/Makefile 2015-09-24 10:47:58.248052907 +0200
c1595e42 21533@@ -0,0 +1,44 @@
4a4d8108
AM
21534+
21535+include ${src}/magic.mk
21536+ifeq (${CONFIG_AUFS_FS},m)
21537+include ${src}/conf.mk
21538+endif
21539+-include ${src}/priv_def.mk
21540+
21541+# cf. include/linux/kernel.h
21542+# enable pr_debug
21543+ccflags-y += -DDEBUG
f6c5ef8b
AM
21544+# sparse requires the full pathname
21545+ifdef M
523b37e3 21546+ccflags-y += -include ${M}/../../include/uapi/linux/aufs_type.h
f6c5ef8b 21547+else
523b37e3 21548+ccflags-y += -include ${srctree}/include/uapi/linux/aufs_type.h
f6c5ef8b 21549+endif
4a4d8108
AM
21550+
21551+obj-$(CONFIG_AUFS_FS) += aufs.o
21552+aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o \
21553+ wkq.o vfsub.o dcsub.o \
e49829fe 21554+ cpup.o whout.o wbr_policy.o \
4a4d8108
AM
21555+ dinfo.o dentry.o \
21556+ dynop.o \
21557+ finfo.o file.o f_op.o \
21558+ dir.o vdir.o \
21559+ iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
c2b27bf2 21560+ mvdown.o ioctl.o
4a4d8108
AM
21561+
21562+# all are boolean
e49829fe 21563+aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
4a4d8108
AM
21564+aufs-$(CONFIG_SYSFS) += sysfs.o
21565+aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
21566+aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
21567+aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
21568+aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
4a4d8108 21569+aufs-$(CONFIG_AUFS_EXPORT) += export.o
c1595e42
JR
21570+aufs-$(CONFIG_AUFS_XATTR) += xattr.o
21571+aufs-$(CONFIG_FS_POSIX_ACL) += posix_acl.o
076b876e 21572+aufs-$(CONFIG_AUFS_FHSM) += fhsm.o
4a4d8108
AM
21573+aufs-$(CONFIG_AUFS_POLL) += poll.o
21574+aufs-$(CONFIG_AUFS_RDU) += rdu.o
4a4d8108
AM
21575+aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
21576+aufs-$(CONFIG_AUFS_DEBUG) += debug.o
21577+aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
7f207e10
AM
21578diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
21579--- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21580+++ linux/fs/aufs/module.c 2015-09-24 10:47:58.254719746 +0200
076b876e 21581@@ -0,0 +1,210 @@
1facf9fc 21582+/*
2000de60 21583+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 21584+ *
21585+ * This program, aufs is free software; you can redistribute it and/or modify
21586+ * it under the terms of the GNU General Public License as published by
21587+ * the Free Software Foundation; either version 2 of the License, or
21588+ * (at your option) any later version.
dece6358
AM
21589+ *
21590+ * This program is distributed in the hope that it will be useful,
21591+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
21592+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21593+ * GNU General Public License for more details.
21594+ *
21595+ * You should have received a copy of the GNU General Public License
523b37e3 21596+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 21597+ */
21598+
21599+/*
21600+ * module global variables and operations
21601+ */
21602+
21603+#include <linux/module.h>
21604+#include <linux/seq_file.h>
21605+#include "aufs.h"
21606+
21607+void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp)
21608+{
21609+ if (new_sz <= nused)
21610+ return p;
21611+
21612+ p = krealloc(p, new_sz, gfp);
21613+ if (p)
21614+ memset(p + nused, 0, new_sz - nused);
21615+ return p;
21616+}
21617+
21618+/* ---------------------------------------------------------------------- */
21619+
21620+/*
21621+ * aufs caches
21622+ */
21623+struct kmem_cache *au_cachep[AuCache_Last];
21624+static int __init au_cache_init(void)
21625+{
4a4d8108 21626+ au_cachep[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
1facf9fc 21627+ if (au_cachep[AuCache_DINFO])
027c5e7a 21628+ /* SLAB_DESTROY_BY_RCU */
4a4d8108
AM
21629+ au_cachep[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
21630+ au_icntnr_init_once);
1facf9fc 21631+ if (au_cachep[AuCache_ICNTNR])
4a4d8108
AM
21632+ au_cachep[AuCache_FINFO] = AuCacheCtor(au_finfo,
21633+ au_fi_init_once);
1facf9fc 21634+ if (au_cachep[AuCache_FINFO])
21635+ au_cachep[AuCache_VDIR] = AuCache(au_vdir);
21636+ if (au_cachep[AuCache_VDIR])
21637+ au_cachep[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
21638+ if (au_cachep[AuCache_DEHSTR])
21639+ return 0;
21640+
21641+ return -ENOMEM;
21642+}
21643+
21644+static void au_cache_fin(void)
21645+{
21646+ int i;
4a4d8108 21647+
537831f9
AM
21648+ /*
21649+ * Make sure all delayed rcu free inodes are flushed before we
21650+ * destroy cache.
21651+ */
21652+ rcu_barrier();
21653+
7eafdf33
AM
21654+ /* excluding AuCache_HNOTIFY */
21655+ BUILD_BUG_ON(AuCache_HNOTIFY + 1 != AuCache_Last);
21656+ for (i = 0; i < AuCache_HNOTIFY; i++)
1facf9fc 21657+ if (au_cachep[i]) {
21658+ kmem_cache_destroy(au_cachep[i]);
21659+ au_cachep[i] = NULL;
21660+ }
21661+}
21662+
21663+/* ---------------------------------------------------------------------- */
21664+
21665+int au_dir_roflags;
21666+
e49829fe 21667+#ifdef CONFIG_AUFS_SBILIST
1e00d052
AM
21668+/*
21669+ * iterate_supers_type() doesn't protect us from
21670+ * remounting (branch management)
21671+ */
e49829fe
JR
21672+struct au_splhead au_sbilist;
21673+#endif
21674+
9dbd164d
AM
21675+struct lock_class_key au_lc_key[AuLcKey_Last];
21676+
1facf9fc 21677+/*
21678+ * functions for module interface.
21679+ */
21680+MODULE_LICENSE("GPL");
21681+/* MODULE_LICENSE("GPL v2"); */
dece6358 21682+MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
1facf9fc 21683+MODULE_DESCRIPTION(AUFS_NAME
21684+ " -- Advanced multi layered unification filesystem");
21685+MODULE_VERSION(AUFS_VERSION);
c06a8ce3 21686+MODULE_ALIAS_FS(AUFS_NAME);
1facf9fc 21687+
1facf9fc 21688+/* this module parameter has no meaning when SYSFS is disabled */
21689+int sysaufs_brs = 1;
21690+MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
21691+module_param_named(brs, sysaufs_brs, int, S_IRUGO);
21692+
076b876e
AM
21693+/* this module parameter has no meaning when USER_NS is disabled */
21694+static bool au_userns;
21695+MODULE_PARM_DESC(allow_userns, "allow unprivileged to mount under userns");
21696+module_param_named(allow_userns, au_userns, bool, S_IRUGO);
21697+
1facf9fc 21698+/* ---------------------------------------------------------------------- */
21699+
21700+static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
21701+
21702+int au_seq_path(struct seq_file *seq, struct path *path)
21703+{
21704+ return seq_path(seq, path, au_esc_chars);
21705+}
21706+
21707+/* ---------------------------------------------------------------------- */
21708+
21709+static int __init aufs_init(void)
21710+{
21711+ int err, i;
21712+ char *p;
21713+
21714+ p = au_esc_chars;
21715+ for (i = 1; i <= ' '; i++)
21716+ *p++ = i;
21717+ *p++ = '\\';
21718+ *p++ = '\x7f';
21719+ *p = 0;
21720+
21721+ au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
21722+
e49829fe 21723+ au_sbilist_init();
1facf9fc 21724+ sysaufs_brs_init();
21725+ au_debug_init();
4a4d8108 21726+ au_dy_init();
1facf9fc 21727+ err = sysaufs_init();
21728+ if (unlikely(err))
21729+ goto out;
e49829fe 21730+ err = au_procfs_init();
4f0767ce 21731+ if (unlikely(err))
953406b4 21732+ goto out_sysaufs;
e49829fe
JR
21733+ err = au_wkq_init();
21734+ if (unlikely(err))
21735+ goto out_procfs;
87a755f4 21736+ err = au_loopback_init();
1facf9fc 21737+ if (unlikely(err))
21738+ goto out_wkq;
87a755f4
AM
21739+ err = au_hnotify_init();
21740+ if (unlikely(err))
21741+ goto out_loopback;
1facf9fc 21742+ err = au_sysrq_init();
21743+ if (unlikely(err))
21744+ goto out_hin;
21745+ err = au_cache_init();
21746+ if (unlikely(err))
21747+ goto out_sysrq;
076b876e
AM
21748+
21749+ aufs_fs_type.fs_flags |= au_userns ? FS_USERNS_MOUNT : 0;
1facf9fc 21750+ err = register_filesystem(&aufs_fs_type);
21751+ if (unlikely(err))
21752+ goto out_cache;
076b876e 21753+
4a4d8108
AM
21754+ /* since we define pr_fmt, call printk directly */
21755+ printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
1facf9fc 21756+ goto out; /* success */
21757+
4f0767ce 21758+out_cache:
1facf9fc 21759+ au_cache_fin();
4f0767ce 21760+out_sysrq:
1facf9fc 21761+ au_sysrq_fin();
4f0767ce 21762+out_hin:
4a4d8108 21763+ au_hnotify_fin();
87a755f4
AM
21764+out_loopback:
21765+ au_loopback_fin();
4f0767ce 21766+out_wkq:
1facf9fc 21767+ au_wkq_fin();
e49829fe
JR
21768+out_procfs:
21769+ au_procfs_fin();
4f0767ce 21770+out_sysaufs:
1facf9fc 21771+ sysaufs_fin();
4a4d8108 21772+ au_dy_fin();
4f0767ce 21773+out:
1facf9fc 21774+ return err;
21775+}
21776+
21777+static void __exit aufs_exit(void)
21778+{
21779+ unregister_filesystem(&aufs_fs_type);
21780+ au_cache_fin();
21781+ au_sysrq_fin();
4a4d8108 21782+ au_hnotify_fin();
87a755f4 21783+ au_loopback_fin();
1facf9fc 21784+ au_wkq_fin();
e49829fe 21785+ au_procfs_fin();
1facf9fc 21786+ sysaufs_fin();
4a4d8108 21787+ au_dy_fin();
1facf9fc 21788+}
21789+
21790+module_init(aufs_init);
21791+module_exit(aufs_exit);
7f207e10
AM
21792diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
21793--- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21794+++ linux/fs/aufs/module.h 2015-09-24 10:47:58.254719746 +0200
523b37e3 21795@@ -0,0 +1,104 @@
1facf9fc 21796+/*
2000de60 21797+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 21798+ *
21799+ * This program, aufs is free software; you can redistribute it and/or modify
21800+ * it under the terms of the GNU General Public License as published by
21801+ * the Free Software Foundation; either version 2 of the License, or
21802+ * (at your option) any later version.
dece6358
AM
21803+ *
21804+ * This program is distributed in the hope that it will be useful,
21805+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
21806+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21807+ * GNU General Public License for more details.
21808+ *
21809+ * You should have received a copy of the GNU General Public License
523b37e3 21810+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 21811+ */
21812+
21813+/*
21814+ * module initialization and module-global
21815+ */
21816+
21817+#ifndef __AUFS_MODULE_H__
21818+#define __AUFS_MODULE_H__
21819+
21820+#ifdef __KERNEL__
21821+
21822+#include <linux/slab.h>
21823+
dece6358
AM
21824+struct path;
21825+struct seq_file;
21826+
1facf9fc 21827+/* module parameters */
1facf9fc 21828+extern int sysaufs_brs;
21829+
21830+/* ---------------------------------------------------------------------- */
21831+
21832+extern int au_dir_roflags;
21833+
9dbd164d
AM
21834+enum {
21835+ AuLcNonDir_FIINFO,
21836+ AuLcNonDir_DIINFO,
21837+ AuLcNonDir_IIINFO,
21838+
21839+ AuLcDir_FIINFO,
21840+ AuLcDir_DIINFO,
21841+ AuLcDir_IIINFO,
21842+
21843+ AuLcSymlink_DIINFO,
21844+ AuLcSymlink_IIINFO,
21845+
21846+ AuLcKey_Last
21847+};
21848+extern struct lock_class_key au_lc_key[AuLcKey_Last];
21849+
1facf9fc 21850+void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp);
21851+int au_seq_path(struct seq_file *seq, struct path *path);
21852+
e49829fe
JR
21853+#ifdef CONFIG_PROC_FS
21854+/* procfs.c */
21855+int __init au_procfs_init(void);
21856+void au_procfs_fin(void);
21857+#else
21858+AuStubInt0(au_procfs_init, void);
21859+AuStubVoid(au_procfs_fin, void);
21860+#endif
21861+
4f0767ce
JR
21862+/* ---------------------------------------------------------------------- */
21863+
21864+/* kmem cache */
1facf9fc 21865+enum {
21866+ AuCache_DINFO,
21867+ AuCache_ICNTNR,
21868+ AuCache_FINFO,
21869+ AuCache_VDIR,
21870+ AuCache_DEHSTR,
7eafdf33 21871+ AuCache_HNOTIFY, /* must be last */
1facf9fc 21872+ AuCache_Last
21873+};
21874+
4a4d8108
AM
21875+#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
21876+#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
21877+#define AuCacheCtor(type, ctor) \
21878+ kmem_cache_create(#type, sizeof(struct type), \
21879+ __alignof__(struct type), AuCacheFlags, ctor)
1facf9fc 21880+
21881+extern struct kmem_cache *au_cachep[];
21882+
21883+#define AuCacheFuncs(name, index) \
4a4d8108 21884+static inline struct au_##name *au_cache_alloc_##name(void) \
1facf9fc 21885+{ return kmem_cache_alloc(au_cachep[AuCache_##index], GFP_NOFS); } \
4a4d8108 21886+static inline void au_cache_free_##name(struct au_##name *p) \
1facf9fc 21887+{ kmem_cache_free(au_cachep[AuCache_##index], p); }
21888+
21889+AuCacheFuncs(dinfo, DINFO);
21890+AuCacheFuncs(icntnr, ICNTNR);
21891+AuCacheFuncs(finfo, FINFO);
21892+AuCacheFuncs(vdir, VDIR);
4a4d8108
AM
21893+AuCacheFuncs(vdir_dehstr, DEHSTR);
21894+#ifdef CONFIG_AUFS_HNOTIFY
21895+AuCacheFuncs(hnotify, HNOTIFY);
21896+#endif
1facf9fc 21897+
4a4d8108
AM
21898+#endif /* __KERNEL__ */
21899+#endif /* __AUFS_MODULE_H__ */
c2b27bf2
AM
21900diff -urN /usr/share/empty/fs/aufs/mvdown.c linux/fs/aufs/mvdown.c
21901--- /usr/share/empty/fs/aufs/mvdown.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 21902+++ linux/fs/aufs/mvdown.c 2015-09-24 10:47:58.254719746 +0200
c1595e42 21903@@ -0,0 +1,694 @@
c2b27bf2 21904+/*
2000de60 21905+ * Copyright (C) 2011-2015 Junjiro R. Okajima
c2b27bf2
AM
21906+ *
21907+ * This program, aufs is free software; you can redistribute it and/or modify
21908+ * it under the terms of the GNU General Public License as published by
21909+ * the Free Software Foundation; either version 2 of the License, or
21910+ * (at your option) any later version.
21911+ *
21912+ * This program is distributed in the hope that it will be useful,
21913+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
21914+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21915+ * GNU General Public License for more details.
21916+ *
21917+ * You should have received a copy of the GNU General Public License
523b37e3
AM
21918+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
21919+ */
21920+
21921+/*
21922+ * move-down, opposite of copy-up
c2b27bf2
AM
21923+ */
21924+
21925+#include "aufs.h"
21926+
c2b27bf2
AM
21927+struct au_mvd_args {
21928+ struct {
c2b27bf2
AM
21929+ struct super_block *h_sb;
21930+ struct dentry *h_parent;
21931+ struct au_hinode *hdir;
392086de 21932+ struct inode *h_dir, *h_inode;
c1595e42 21933+ struct au_pin pin;
c2b27bf2
AM
21934+ } info[AUFS_MVDOWN_NARRAY];
21935+
21936+ struct aufs_mvdown mvdown;
21937+ struct dentry *dentry, *parent;
21938+ struct inode *inode, *dir;
21939+ struct super_block *sb;
21940+ aufs_bindex_t bopq, bwh, bfound;
21941+ unsigned char rename_lock;
c2b27bf2
AM
21942+};
21943+
392086de 21944+#define mvd_errno mvdown.au_errno
076b876e
AM
21945+#define mvd_bsrc mvdown.stbr[AUFS_MVDOWN_UPPER].bindex
21946+#define mvd_src_brid mvdown.stbr[AUFS_MVDOWN_UPPER].brid
21947+#define mvd_bdst mvdown.stbr[AUFS_MVDOWN_LOWER].bindex
21948+#define mvd_dst_brid mvdown.stbr[AUFS_MVDOWN_LOWER].brid
c2b27bf2 21949+
392086de
AM
21950+#define mvd_h_src_sb info[AUFS_MVDOWN_UPPER].h_sb
21951+#define mvd_h_src_parent info[AUFS_MVDOWN_UPPER].h_parent
21952+#define mvd_hdir_src info[AUFS_MVDOWN_UPPER].hdir
21953+#define mvd_h_src_dir info[AUFS_MVDOWN_UPPER].h_dir
21954+#define mvd_h_src_inode info[AUFS_MVDOWN_UPPER].h_inode
c1595e42 21955+#define mvd_pin_src info[AUFS_MVDOWN_UPPER].pin
392086de
AM
21956+
21957+#define mvd_h_dst_sb info[AUFS_MVDOWN_LOWER].h_sb
21958+#define mvd_h_dst_parent info[AUFS_MVDOWN_LOWER].h_parent
21959+#define mvd_hdir_dst info[AUFS_MVDOWN_LOWER].hdir
21960+#define mvd_h_dst_dir info[AUFS_MVDOWN_LOWER].h_dir
21961+#define mvd_h_dst_inode info[AUFS_MVDOWN_LOWER].h_inode
c1595e42 21962+#define mvd_pin_dst info[AUFS_MVDOWN_LOWER].pin
c2b27bf2
AM
21963+
21964+#define AU_MVD_PR(flag, ...) do { \
21965+ if (flag) \
21966+ pr_err(__VA_ARGS__); \
21967+ } while (0)
21968+
076b876e
AM
21969+static int find_lower_writable(struct au_mvd_args *a)
21970+{
21971+ struct super_block *sb;
21972+ aufs_bindex_t bindex, bend;
21973+ struct au_branch *br;
21974+
21975+ sb = a->sb;
21976+ bindex = a->mvd_bsrc;
21977+ bend = au_sbend(sb);
21978+ if (a->mvdown.flags & AUFS_MVDOWN_FHSM_LOWER)
21979+ for (bindex++; bindex <= bend; bindex++) {
21980+ br = au_sbr(sb, bindex);
21981+ if (au_br_fhsm(br->br_perm)
21982+ && (!(au_br_sb(br)->s_flags & MS_RDONLY)))
21983+ return bindex;
21984+ }
21985+ else if (!(a->mvdown.flags & AUFS_MVDOWN_ROLOWER))
21986+ for (bindex++; bindex <= bend; bindex++) {
21987+ br = au_sbr(sb, bindex);
21988+ if (!au_br_rdonly(br))
21989+ return bindex;
21990+ }
21991+ else
21992+ for (bindex++; bindex <= bend; bindex++) {
21993+ br = au_sbr(sb, bindex);
21994+ if (!(au_br_sb(br)->s_flags & MS_RDONLY)) {
21995+ if (au_br_rdonly(br))
21996+ a->mvdown.flags
21997+ |= AUFS_MVDOWN_ROLOWER_R;
21998+ return bindex;
21999+ }
22000+ }
22001+
22002+ return -1;
22003+}
22004+
c2b27bf2 22005+/* make the parent dir on bdst */
392086de 22006+static int au_do_mkdir(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22007+{
22008+ int err;
22009+
22010+ err = 0;
22011+ a->mvd_hdir_src = au_hi(a->dir, a->mvd_bsrc);
22012+ a->mvd_hdir_dst = au_hi(a->dir, a->mvd_bdst);
22013+ a->mvd_h_src_parent = au_h_dptr(a->parent, a->mvd_bsrc);
22014+ a->mvd_h_dst_parent = NULL;
22015+ if (au_dbend(a->parent) >= a->mvd_bdst)
22016+ a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
22017+ if (!a->mvd_h_dst_parent) {
22018+ err = au_cpdown_dirs(a->dentry, a->mvd_bdst);
22019+ if (unlikely(err)) {
392086de 22020+ AU_MVD_PR(dmsg, "cpdown_dirs failed\n");
c2b27bf2
AM
22021+ goto out;
22022+ }
22023+ a->mvd_h_dst_parent = au_h_dptr(a->parent, a->mvd_bdst);
22024+ }
22025+
22026+out:
22027+ AuTraceErr(err);
22028+ return err;
22029+}
22030+
22031+/* lock them all */
392086de 22032+static int au_do_lock(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22033+{
22034+ int err;
22035+ struct dentry *h_trap;
22036+
22037+ a->mvd_h_src_sb = au_sbr_sb(a->sb, a->mvd_bsrc);
22038+ a->mvd_h_dst_sb = au_sbr_sb(a->sb, a->mvd_bdst);
c1595e42
JR
22039+ err = au_pin(&a->mvd_pin_dst, a->dentry, a->mvd_bdst,
22040+ au_opt_udba(a->sb),
22041+ AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22042+ AuTraceErr(err);
22043+ if (unlikely(err)) {
22044+ AU_MVD_PR(dmsg, "pin_dst failed\n");
22045+ goto out;
22046+ }
22047+
c2b27bf2
AM
22048+ if (a->mvd_h_src_sb != a->mvd_h_dst_sb) {
22049+ a->rename_lock = 0;
c1595e42
JR
22050+ au_pin_init(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
22051+ AuLsc_DI_PARENT, AuLsc_I_PARENT3,
22052+ au_opt_udba(a->sb),
22053+ AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22054+ err = au_do_pin(&a->mvd_pin_src);
22055+ AuTraceErr(err);
5527c038 22056+ a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
c1595e42
JR
22057+ if (unlikely(err)) {
22058+ AU_MVD_PR(dmsg, "pin_src failed\n");
22059+ goto out_dst;
22060+ }
22061+ goto out; /* success */
c2b27bf2
AM
22062+ }
22063+
c2b27bf2 22064+ a->rename_lock = 1;
c1595e42
JR
22065+ au_pin_hdir_unlock(&a->mvd_pin_dst);
22066+ err = au_pin(&a->mvd_pin_src, a->dentry, a->mvd_bsrc,
22067+ au_opt_udba(a->sb),
22068+ AuPin_MNT_WRITE | AuPin_DI_LOCKED);
22069+ AuTraceErr(err);
5527c038 22070+ a->mvd_h_src_dir = d_inode(a->mvd_h_src_parent);
c1595e42
JR
22071+ if (unlikely(err)) {
22072+ AU_MVD_PR(dmsg, "pin_src failed\n");
22073+ au_pin_hdir_lock(&a->mvd_pin_dst);
22074+ goto out_dst;
22075+ }
22076+ au_pin_hdir_unlock(&a->mvd_pin_src);
c2b27bf2
AM
22077+ h_trap = vfsub_lock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22078+ a->mvd_h_dst_parent, a->mvd_hdir_dst);
22079+ if (h_trap) {
22080+ err = (h_trap != a->mvd_h_src_parent);
22081+ if (err)
22082+ err = (h_trap != a->mvd_h_dst_parent);
22083+ }
22084+ BUG_ON(err); /* it should never happen */
c1595e42
JR
22085+ if (unlikely(a->mvd_h_src_dir != au_pinned_h_dir(&a->mvd_pin_src))) {
22086+ err = -EBUSY;
22087+ AuTraceErr(err);
22088+ vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22089+ a->mvd_h_dst_parent, a->mvd_hdir_dst);
22090+ au_pin_hdir_lock(&a->mvd_pin_src);
22091+ au_unpin(&a->mvd_pin_src);
22092+ au_pin_hdir_lock(&a->mvd_pin_dst);
22093+ goto out_dst;
22094+ }
22095+ goto out; /* success */
c2b27bf2 22096+
c1595e42
JR
22097+out_dst:
22098+ au_unpin(&a->mvd_pin_dst);
c2b27bf2
AM
22099+out:
22100+ AuTraceErr(err);
22101+ return err;
22102+}
22103+
392086de 22104+static void au_do_unlock(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2 22105+{
c1595e42
JR
22106+ if (!a->rename_lock)
22107+ au_unpin(&a->mvd_pin_src);
22108+ else {
c2b27bf2
AM
22109+ vfsub_unlock_rename(a->mvd_h_src_parent, a->mvd_hdir_src,
22110+ a->mvd_h_dst_parent, a->mvd_hdir_dst);
c1595e42
JR
22111+ au_pin_hdir_lock(&a->mvd_pin_src);
22112+ au_unpin(&a->mvd_pin_src);
22113+ au_pin_hdir_lock(&a->mvd_pin_dst);
22114+ }
22115+ au_unpin(&a->mvd_pin_dst);
c2b27bf2
AM
22116+}
22117+
22118+/* copy-down the file */
392086de 22119+static int au_do_cpdown(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22120+{
22121+ int err;
22122+ struct au_cp_generic cpg = {
22123+ .dentry = a->dentry,
22124+ .bdst = a->mvd_bdst,
22125+ .bsrc = a->mvd_bsrc,
22126+ .len = -1,
c1595e42 22127+ .pin = &a->mvd_pin_dst,
c2b27bf2
AM
22128+ .flags = AuCpup_DTIME | AuCpup_HOPEN
22129+ };
22130+
22131+ AuDbg("b%d, b%d\n", cpg.bsrc, cpg.bdst);
392086de
AM
22132+ if (a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
22133+ au_fset_cpup(cpg.flags, OVERWRITE);
22134+ if (a->mvdown.flags & AUFS_MVDOWN_ROLOWER)
22135+ au_fset_cpup(cpg.flags, RWDST);
c2b27bf2
AM
22136+ err = au_sio_cpdown_simple(&cpg);
22137+ if (unlikely(err))
392086de 22138+ AU_MVD_PR(dmsg, "cpdown failed\n");
c2b27bf2
AM
22139+
22140+ AuTraceErr(err);
22141+ return err;
22142+}
22143+
22144+/*
22145+ * unlink the whiteout on bdst if exist which may be created by UDBA while we
22146+ * were sleeping
22147+ */
392086de 22148+static int au_do_unlink_wh(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22149+{
22150+ int err;
22151+ struct path h_path;
22152+ struct au_branch *br;
523b37e3 22153+ struct inode *delegated;
c2b27bf2
AM
22154+
22155+ br = au_sbr(a->sb, a->mvd_bdst);
22156+ h_path.dentry = au_wh_lkup(a->mvd_h_dst_parent, &a->dentry->d_name, br);
22157+ err = PTR_ERR(h_path.dentry);
22158+ if (IS_ERR(h_path.dentry)) {
392086de 22159+ AU_MVD_PR(dmsg, "wh_lkup failed\n");
c2b27bf2
AM
22160+ goto out;
22161+ }
22162+
22163+ err = 0;
5527c038 22164+ if (d_is_positive(h_path.dentry)) {
c2b27bf2 22165+ h_path.mnt = au_br_mnt(br);
523b37e3 22166+ delegated = NULL;
5527c038 22167+ err = vfsub_unlink(d_inode(a->mvd_h_dst_parent), &h_path,
523b37e3
AM
22168+ &delegated, /*force*/0);
22169+ if (unlikely(err == -EWOULDBLOCK)) {
22170+ pr_warn("cannot retry for NFSv4 delegation"
22171+ " for an internal unlink\n");
22172+ iput(delegated);
22173+ }
c2b27bf2 22174+ if (unlikely(err))
392086de 22175+ AU_MVD_PR(dmsg, "wh_unlink failed\n");
c2b27bf2
AM
22176+ }
22177+ dput(h_path.dentry);
22178+
22179+out:
22180+ AuTraceErr(err);
22181+ return err;
22182+}
22183+
22184+/*
22185+ * unlink the topmost h_dentry
c2b27bf2 22186+ */
392086de 22187+static int au_do_unlink(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22188+{
22189+ int err;
22190+ struct path h_path;
523b37e3 22191+ struct inode *delegated;
c2b27bf2
AM
22192+
22193+ h_path.mnt = au_sbr_mnt(a->sb, a->mvd_bsrc);
22194+ h_path.dentry = au_h_dptr(a->dentry, a->mvd_bsrc);
523b37e3
AM
22195+ delegated = NULL;
22196+ err = vfsub_unlink(a->mvd_h_src_dir, &h_path, &delegated, /*force*/0);
22197+ if (unlikely(err == -EWOULDBLOCK)) {
22198+ pr_warn("cannot retry for NFSv4 delegation"
22199+ " for an internal unlink\n");
22200+ iput(delegated);
22201+ }
c2b27bf2 22202+ if (unlikely(err))
392086de 22203+ AU_MVD_PR(dmsg, "unlink failed\n");
c2b27bf2
AM
22204+
22205+ AuTraceErr(err);
22206+ return err;
22207+}
22208+
076b876e
AM
22209+/* Since mvdown succeeded, we ignore an error of this function */
22210+static void au_do_stfs(const unsigned char dmsg, struct au_mvd_args *a)
22211+{
22212+ int err;
22213+ struct au_branch *br;
22214+
22215+ a->mvdown.flags |= AUFS_MVDOWN_STFS_FAILED;
22216+ br = au_sbr(a->sb, a->mvd_bsrc);
22217+ err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_UPPER].stfs);
22218+ if (!err) {
22219+ br = au_sbr(a->sb, a->mvd_bdst);
22220+ a->mvdown.stbr[AUFS_MVDOWN_LOWER].brid = br->br_id;
22221+ err = au_br_stfs(br, &a->mvdown.stbr[AUFS_MVDOWN_LOWER].stfs);
22222+ }
22223+ if (!err)
22224+ a->mvdown.flags &= ~AUFS_MVDOWN_STFS_FAILED;
22225+ else
22226+ AU_MVD_PR(dmsg, "statfs failed (%d), ignored\n", err);
22227+}
22228+
c2b27bf2
AM
22229+/*
22230+ * copy-down the file and unlink the bsrc file.
22231+ * - unlink the bdst whout if exist
22232+ * - copy-down the file (with whtmp name and rename)
22233+ * - unlink the bsrc file
22234+ */
392086de 22235+static int au_do_mvdown(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22236+{
22237+ int err;
22238+
392086de 22239+ err = au_do_mkdir(dmsg, a);
c2b27bf2 22240+ if (!err)
392086de 22241+ err = au_do_lock(dmsg, a);
c2b27bf2
AM
22242+ if (unlikely(err))
22243+ goto out;
22244+
22245+ /*
22246+ * do not revert the activities we made on bdst since they should be
22247+ * harmless in aufs.
22248+ */
22249+
392086de 22250+ err = au_do_cpdown(dmsg, a);
c2b27bf2 22251+ if (!err)
392086de
AM
22252+ err = au_do_unlink_wh(dmsg, a);
22253+ if (!err && !(a->mvdown.flags & AUFS_MVDOWN_KUPPER))
22254+ err = au_do_unlink(dmsg, a);
c2b27bf2
AM
22255+ if (unlikely(err))
22256+ goto out_unlock;
22257+
c1595e42
JR
22258+ AuDbg("%pd2, 0x%x, %d --> %d\n",
22259+ a->dentry, a->mvdown.flags, a->mvd_bsrc, a->mvd_bdst);
076b876e
AM
22260+ if (find_lower_writable(a) < 0)
22261+ a->mvdown.flags |= AUFS_MVDOWN_BOTTOM;
22262+
22263+ if (a->mvdown.flags & AUFS_MVDOWN_STFS)
22264+ au_do_stfs(dmsg, a);
22265+
c2b27bf2 22266+ /* maintain internal array */
392086de
AM
22267+ if (!(a->mvdown.flags & AUFS_MVDOWN_KUPPER)) {
22268+ au_set_h_dptr(a->dentry, a->mvd_bsrc, NULL);
22269+ au_set_dbstart(a->dentry, a->mvd_bdst);
22270+ au_set_h_iptr(a->inode, a->mvd_bsrc, NULL, /*flags*/0);
22271+ au_set_ibstart(a->inode, a->mvd_bdst);
22272+ }
c2b27bf2
AM
22273+ if (au_dbend(a->dentry) < a->mvd_bdst)
22274+ au_set_dbend(a->dentry, a->mvd_bdst);
c2b27bf2
AM
22275+ if (au_ibend(a->inode) < a->mvd_bdst)
22276+ au_set_ibend(a->inode, a->mvd_bdst);
22277+
22278+out_unlock:
392086de 22279+ au_do_unlock(dmsg, a);
c2b27bf2
AM
22280+out:
22281+ AuTraceErr(err);
22282+ return err;
22283+}
22284+
22285+/* ---------------------------------------------------------------------- */
22286+
c2b27bf2 22287+/* make sure the file is idle */
392086de 22288+static int au_mvd_args_busy(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22289+{
22290+ int err, plinked;
c2b27bf2
AM
22291+
22292+ err = 0;
c2b27bf2
AM
22293+ plinked = !!au_opt_test(au_mntflags(a->sb), PLINK);
22294+ if (au_dbstart(a->dentry) == a->mvd_bsrc
c1595e42 22295+ && au_dcount(a->dentry) == 1
c2b27bf2 22296+ && atomic_read(&a->inode->i_count) == 1
392086de 22297+ /* && a->mvd_h_src_inode->i_nlink == 1 */
c2b27bf2
AM
22298+ && (!plinked || !au_plink_test(a->inode))
22299+ && a->inode->i_nlink == 1)
22300+ goto out;
22301+
22302+ err = -EBUSY;
392086de 22303+ AU_MVD_PR(dmsg,
c1595e42
JR
22304+ "b%d, d{b%d, c%d?}, i{c%d?, l%u}, hi{l%u}, p{%d, %d}\n",
22305+ a->mvd_bsrc, au_dbstart(a->dentry), au_dcount(a->dentry),
c2b27bf2 22306+ atomic_read(&a->inode->i_count), a->inode->i_nlink,
392086de 22307+ a->mvd_h_src_inode->i_nlink,
c2b27bf2
AM
22308+ plinked, plinked ? au_plink_test(a->inode) : 0);
22309+
22310+out:
22311+ AuTraceErr(err);
22312+ return err;
22313+}
22314+
22315+/* make sure the parent dir is fine */
392086de 22316+static int au_mvd_args_parent(const unsigned char dmsg,
c2b27bf2
AM
22317+ struct au_mvd_args *a)
22318+{
22319+ int err;
22320+ aufs_bindex_t bindex;
22321+
22322+ err = 0;
22323+ if (unlikely(au_alive_dir(a->parent))) {
22324+ err = -ENOENT;
392086de 22325+ AU_MVD_PR(dmsg, "parent dir is dead\n");
c2b27bf2
AM
22326+ goto out;
22327+ }
22328+
22329+ a->bopq = au_dbdiropq(a->parent);
22330+ bindex = au_wbr_nonopq(a->dentry, a->mvd_bdst);
22331+ AuDbg("b%d\n", bindex);
22332+ if (unlikely((bindex >= 0 && bindex < a->mvd_bdst)
22333+ || (a->bopq != -1 && a->bopq < a->mvd_bdst))) {
22334+ err = -EINVAL;
392086de
AM
22335+ a->mvd_errno = EAU_MVDOWN_OPAQUE;
22336+ AU_MVD_PR(dmsg, "ancestor is opaque b%d, b%d\n",
c2b27bf2
AM
22337+ a->bopq, a->mvd_bdst);
22338+ }
22339+
22340+out:
22341+ AuTraceErr(err);
22342+ return err;
22343+}
22344+
392086de 22345+static int au_mvd_args_intermediate(const unsigned char dmsg,
c2b27bf2
AM
22346+ struct au_mvd_args *a)
22347+{
22348+ int err;
22349+ struct au_dinfo *dinfo, *tmp;
22350+
22351+ /* lookup the next lower positive entry */
22352+ err = -ENOMEM;
22353+ tmp = au_di_alloc(a->sb, AuLsc_DI_TMP);
22354+ if (unlikely(!tmp))
22355+ goto out;
22356+
22357+ a->bfound = -1;
22358+ a->bwh = -1;
22359+ dinfo = au_di(a->dentry);
22360+ au_di_cp(tmp, dinfo);
22361+ au_di_swap(tmp, dinfo);
22362+
22363+ /* returns the number of positive dentries */
22364+ err = au_lkup_dentry(a->dentry, a->mvd_bsrc + 1, /*type*/0);
22365+ if (!err)
22366+ a->bwh = au_dbwh(a->dentry);
22367+ else if (err > 0)
22368+ a->bfound = au_dbstart(a->dentry);
22369+
22370+ au_di_swap(tmp, dinfo);
22371+ au_rw_write_unlock(&tmp->di_rwsem);
22372+ au_di_free(tmp);
22373+ if (unlikely(err < 0))
392086de 22374+ AU_MVD_PR(dmsg, "failed look-up lower\n");
c2b27bf2
AM
22375+
22376+ /*
22377+ * here, we have these cases.
22378+ * bfound == -1
22379+ * no positive dentry under bsrc. there are more sub-cases.
22380+ * bwh < 0
22381+ * there no whiteout, we can safely move-down.
22382+ * bwh <= bsrc
22383+ * impossible
22384+ * bsrc < bwh && bwh < bdst
22385+ * there is a whiteout on RO branch. cannot proceed.
22386+ * bwh == bdst
22387+ * there is a whiteout on the RW target branch. it should
22388+ * be removed.
22389+ * bdst < bwh
22390+ * there is a whiteout somewhere unrelated branch.
22391+ * -1 < bfound && bfound <= bsrc
22392+ * impossible.
22393+ * bfound < bdst
22394+ * found, but it is on RO branch between bsrc and bdst. cannot
22395+ * proceed.
22396+ * bfound == bdst
22397+ * found, replace it if AUFS_MVDOWN_FORCE is set. otherwise return
22398+ * error.
22399+ * bdst < bfound
22400+ * found, after we create the file on bdst, it will be hidden.
22401+ */
22402+
22403+ AuDebugOn(a->bfound == -1
22404+ && a->bwh != -1
22405+ && a->bwh <= a->mvd_bsrc);
22406+ AuDebugOn(-1 < a->bfound
22407+ && a->bfound <= a->mvd_bsrc);
22408+
22409+ err = -EINVAL;
22410+ if (a->bfound == -1
22411+ && a->mvd_bsrc < a->bwh
22412+ && a->bwh != -1
22413+ && a->bwh < a->mvd_bdst) {
392086de
AM
22414+ a->mvd_errno = EAU_MVDOWN_WHITEOUT;
22415+ AU_MVD_PR(dmsg, "bsrc %d, bdst %d, bfound %d, bwh %d\n",
c2b27bf2
AM
22416+ a->mvd_bsrc, a->mvd_bdst, a->bfound, a->bwh);
22417+ goto out;
22418+ } else if (a->bfound != -1 && a->bfound < a->mvd_bdst) {
392086de
AM
22419+ a->mvd_errno = EAU_MVDOWN_UPPER;
22420+ AU_MVD_PR(dmsg, "bdst %d, bfound %d\n",
c2b27bf2
AM
22421+ a->mvd_bdst, a->bfound);
22422+ goto out;
22423+ }
22424+
22425+ err = 0; /* success */
22426+
22427+out:
22428+ AuTraceErr(err);
22429+ return err;
22430+}
22431+
392086de 22432+static int au_mvd_args_exist(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22433+{
22434+ int err;
22435+
392086de
AM
22436+ err = 0;
22437+ if (!(a->mvdown.flags & AUFS_MVDOWN_OWLOWER)
22438+ && a->bfound == a->mvd_bdst)
22439+ err = -EEXIST;
c2b27bf2
AM
22440+ AuTraceErr(err);
22441+ return err;
22442+}
22443+
392086de 22444+static int au_mvd_args(const unsigned char dmsg, struct au_mvd_args *a)
c2b27bf2
AM
22445+{
22446+ int err;
22447+ struct au_branch *br;
22448+
22449+ err = -EISDIR;
22450+ if (unlikely(S_ISDIR(a->inode->i_mode)))
22451+ goto out;
22452+
22453+ err = -EINVAL;
392086de
AM
22454+ if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_UPPER))
22455+ a->mvd_bsrc = au_ibstart(a->inode);
22456+ else {
22457+ a->mvd_bsrc = au_br_index(a->sb, a->mvd_src_brid);
22458+ if (unlikely(a->mvd_bsrc < 0
22459+ || (a->mvd_bsrc < au_dbstart(a->dentry)
22460+ || au_dbend(a->dentry) < a->mvd_bsrc
22461+ || !au_h_dptr(a->dentry, a->mvd_bsrc))
22462+ || (a->mvd_bsrc < au_ibstart(a->inode)
22463+ || au_ibend(a->inode) < a->mvd_bsrc
22464+ || !au_h_iptr(a->inode, a->mvd_bsrc)))) {
22465+ a->mvd_errno = EAU_MVDOWN_NOUPPER;
22466+ AU_MVD_PR(dmsg, "no upper\n");
22467+ goto out;
22468+ }
22469+ }
c2b27bf2 22470+ if (unlikely(a->mvd_bsrc == au_sbend(a->sb))) {
392086de
AM
22471+ a->mvd_errno = EAU_MVDOWN_BOTTOM;
22472+ AU_MVD_PR(dmsg, "on the bottom\n");
c2b27bf2
AM
22473+ goto out;
22474+ }
392086de 22475+ a->mvd_h_src_inode = au_h_iptr(a->inode, a->mvd_bsrc);
c2b27bf2
AM
22476+ br = au_sbr(a->sb, a->mvd_bsrc);
22477+ err = au_br_rdonly(br);
392086de
AM
22478+ if (!(a->mvdown.flags & AUFS_MVDOWN_ROUPPER)) {
22479+ if (unlikely(err))
22480+ goto out;
22481+ } else if (!(vfsub_native_ro(a->mvd_h_src_inode)
22482+ || IS_APPEND(a->mvd_h_src_inode))) {
22483+ if (err)
22484+ a->mvdown.flags |= AUFS_MVDOWN_ROUPPER_R;
22485+ /* go on */
22486+ } else
c2b27bf2
AM
22487+ goto out;
22488+
22489+ err = -EINVAL;
392086de
AM
22490+ if (!(a->mvdown.flags & AUFS_MVDOWN_BRID_LOWER)) {
22491+ a->mvd_bdst = find_lower_writable(a);
22492+ if (unlikely(a->mvd_bdst < 0)) {
22493+ a->mvd_errno = EAU_MVDOWN_BOTTOM;
22494+ AU_MVD_PR(dmsg, "no writable lower branch\n");
22495+ goto out;
22496+ }
22497+ } else {
22498+ a->mvd_bdst = au_br_index(a->sb, a->mvd_dst_brid);
22499+ if (unlikely(a->mvd_bdst < 0
22500+ || au_sbend(a->sb) < a->mvd_bdst)) {
22501+ a->mvd_errno = EAU_MVDOWN_NOLOWERBR;
22502+ AU_MVD_PR(dmsg, "no lower brid\n");
22503+ goto out;
22504+ }
c2b27bf2
AM
22505+ }
22506+
392086de 22507+ err = au_mvd_args_busy(dmsg, a);
c2b27bf2 22508+ if (!err)
392086de 22509+ err = au_mvd_args_parent(dmsg, a);
c2b27bf2 22510+ if (!err)
392086de 22511+ err = au_mvd_args_intermediate(dmsg, a);
c2b27bf2 22512+ if (!err)
392086de 22513+ err = au_mvd_args_exist(dmsg, a);
c2b27bf2
AM
22514+ if (!err)
22515+ AuDbg("b%d, b%d\n", a->mvd_bsrc, a->mvd_bdst);
22516+
22517+out:
22518+ AuTraceErr(err);
22519+ return err;
22520+}
22521+
22522+int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *uarg)
22523+{
392086de
AM
22524+ int err, e;
22525+ unsigned char dmsg;
22526+ struct au_mvd_args *args;
c2b27bf2
AM
22527+
22528+ err = -EPERM;
22529+ if (unlikely(!capable(CAP_SYS_ADMIN)))
22530+ goto out;
22531+
392086de
AM
22532+ err = -ENOMEM;
22533+ args = kmalloc(sizeof(*args), GFP_NOFS);
22534+ if (unlikely(!args))
22535+ goto out;
22536+
22537+ err = copy_from_user(&args->mvdown, uarg, sizeof(args->mvdown));
22538+ if (!err)
22539+ err = !access_ok(VERIFY_WRITE, uarg, sizeof(*uarg));
c2b27bf2
AM
22540+ if (unlikely(err)) {
22541+ err = -EFAULT;
392086de
AM
22542+ AuTraceErr(err);
22543+ goto out_free;
c2b27bf2 22544+ }
392086de
AM
22545+ AuDbg("flags 0x%x\n", args->mvdown.flags);
22546+ args->mvdown.flags &= ~(AUFS_MVDOWN_ROLOWER_R | AUFS_MVDOWN_ROUPPER_R);
22547+ args->mvdown.au_errno = 0;
22548+ args->dentry = dentry;
5527c038 22549+ args->inode = d_inode(dentry);
392086de 22550+ args->sb = dentry->d_sb;
c2b27bf2 22551+
392086de
AM
22552+ err = -ENOENT;
22553+ dmsg = !!(args->mvdown.flags & AUFS_MVDOWN_DMSG);
22554+ args->parent = dget_parent(dentry);
5527c038 22555+ args->dir = d_inode(args->parent);
392086de
AM
22556+ mutex_lock_nested(&args->dir->i_mutex, I_MUTEX_PARENT);
22557+ dput(args->parent);
22558+ if (unlikely(args->parent != dentry->d_parent)) {
22559+ AU_MVD_PR(dmsg, "parent dir is moved\n");
c2b27bf2
AM
22560+ goto out_dir;
22561+ }
22562+
392086de 22563+ mutex_lock_nested(&args->inode->i_mutex, I_MUTEX_CHILD);
c2b27bf2
AM
22564+ err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH);
22565+ if (unlikely(err))
22566+ goto out_inode;
22567+
392086de
AM
22568+ di_write_lock_parent(args->parent);
22569+ err = au_mvd_args(dmsg, args);
c2b27bf2
AM
22570+ if (unlikely(err))
22571+ goto out_parent;
22572+
392086de 22573+ err = au_do_mvdown(dmsg, args);
c2b27bf2
AM
22574+ if (unlikely(err))
22575+ goto out_parent;
c2b27bf2 22576+
392086de
AM
22577+ au_cpup_attr_timesizes(args->dir);
22578+ au_cpup_attr_timesizes(args->inode);
22579+ au_cpup_igen(args->inode, au_h_iptr(args->inode, args->mvd_bdst));
c2b27bf2
AM
22580+ /* au_digen_dec(dentry); */
22581+
22582+out_parent:
392086de 22583+ di_write_unlock(args->parent);
c2b27bf2
AM
22584+ aufs_read_unlock(dentry, AuLock_DW);
22585+out_inode:
392086de 22586+ mutex_unlock(&args->inode->i_mutex);
c2b27bf2 22587+out_dir:
392086de
AM
22588+ mutex_unlock(&args->dir->i_mutex);
22589+out_free:
22590+ e = copy_to_user(uarg, &args->mvdown, sizeof(args->mvdown));
22591+ if (unlikely(e))
22592+ err = -EFAULT;
22593+ kfree(args);
c2b27bf2
AM
22594+out:
22595+ AuTraceErr(err);
22596+ return err;
22597+}
22598diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
22599--- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 22600+++ linux/fs/aufs/opts.c 2015-09-24 10:47:58.254719746 +0200
5527c038 22601@@ -0,0 +1,1835 @@
1facf9fc 22602+/*
2000de60 22603+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 22604+ *
22605+ * This program, aufs is free software; you can redistribute it and/or modify
22606+ * it under the terms of the GNU General Public License as published by
22607+ * the Free Software Foundation; either version 2 of the License, or
22608+ * (at your option) any later version.
dece6358
AM
22609+ *
22610+ * This program is distributed in the hope that it will be useful,
22611+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
22612+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22613+ * GNU General Public License for more details.
22614+ *
22615+ * You should have received a copy of the GNU General Public License
523b37e3 22616+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 22617+ */
22618+
22619+/*
22620+ * mount options/flags
22621+ */
22622+
dece6358 22623+#include <linux/namei.h>
1facf9fc 22624+#include <linux/types.h> /* a distribution requires */
22625+#include <linux/parser.h>
22626+#include "aufs.h"
22627+
22628+/* ---------------------------------------------------------------------- */
22629+
22630+enum {
22631+ Opt_br,
7e9cd9fe
AM
22632+ Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend,
22633+ Opt_idel, Opt_imod,
22634+ Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash,
dece6358 22635+ Opt_rdblk_def, Opt_rdhash_def,
7e9cd9fe 22636+ Opt_xino, Opt_noxino,
1facf9fc 22637+ Opt_trunc_xino, Opt_trunc_xino_v, Opt_notrunc_xino,
22638+ Opt_trunc_xino_path, Opt_itrunc_xino,
22639+ Opt_trunc_xib, Opt_notrunc_xib,
dece6358 22640+ Opt_shwh, Opt_noshwh,
1facf9fc 22641+ Opt_plink, Opt_noplink, Opt_list_plink,
22642+ Opt_udba,
4a4d8108 22643+ Opt_dio, Opt_nodio,
1facf9fc 22644+ Opt_diropq_a, Opt_diropq_w,
22645+ Opt_warn_perm, Opt_nowarn_perm,
22646+ Opt_wbr_copyup, Opt_wbr_create,
076b876e 22647+ Opt_fhsm_sec,
1facf9fc 22648+ Opt_verbose, Opt_noverbose,
22649+ Opt_sum, Opt_nosum, Opt_wsum,
076b876e 22650+ Opt_dirperm1, Opt_nodirperm1,
c1595e42 22651+ Opt_acl, Opt_noacl,
1facf9fc 22652+ Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
22653+};
22654+
22655+static match_table_t options = {
22656+ {Opt_br, "br=%s"},
22657+ {Opt_br, "br:%s"},
22658+
22659+ {Opt_add, "add=%d:%s"},
22660+ {Opt_add, "add:%d:%s"},
22661+ {Opt_add, "ins=%d:%s"},
22662+ {Opt_add, "ins:%d:%s"},
22663+ {Opt_append, "append=%s"},
22664+ {Opt_append, "append:%s"},
22665+ {Opt_prepend, "prepend=%s"},
22666+ {Opt_prepend, "prepend:%s"},
22667+
22668+ {Opt_del, "del=%s"},
22669+ {Opt_del, "del:%s"},
22670+ /* {Opt_idel, "idel:%d"}, */
22671+ {Opt_mod, "mod=%s"},
22672+ {Opt_mod, "mod:%s"},
22673+ /* {Opt_imod, "imod:%d:%s"}, */
22674+
22675+ {Opt_dirwh, "dirwh=%d"},
22676+
22677+ {Opt_xino, "xino=%s"},
22678+ {Opt_noxino, "noxino"},
22679+ {Opt_trunc_xino, "trunc_xino"},
22680+ {Opt_trunc_xino_v, "trunc_xino_v=%d:%d"},
22681+ {Opt_notrunc_xino, "notrunc_xino"},
22682+ {Opt_trunc_xino_path, "trunc_xino=%s"},
22683+ {Opt_itrunc_xino, "itrunc_xino=%d"},
22684+ /* {Opt_zxino, "zxino=%s"}, */
22685+ {Opt_trunc_xib, "trunc_xib"},
22686+ {Opt_notrunc_xib, "notrunc_xib"},
22687+
e49829fe 22688+#ifdef CONFIG_PROC_FS
1facf9fc 22689+ {Opt_plink, "plink"},
e49829fe
JR
22690+#else
22691+ {Opt_ignore_silent, "plink"},
22692+#endif
22693+
1facf9fc 22694+ {Opt_noplink, "noplink"},
e49829fe 22695+
1facf9fc 22696+#ifdef CONFIG_AUFS_DEBUG
22697+ {Opt_list_plink, "list_plink"},
22698+#endif
22699+
22700+ {Opt_udba, "udba=%s"},
22701+
4a4d8108
AM
22702+ {Opt_dio, "dio"},
22703+ {Opt_nodio, "nodio"},
22704+
076b876e
AM
22705+#ifdef CONFIG_AUFS_FHSM
22706+ {Opt_fhsm_sec, "fhsm_sec=%d"},
22707+#else
22708+ {Opt_ignore_silent, "fhsm_sec=%d"},
22709+#endif
22710+
1facf9fc 22711+ {Opt_diropq_a, "diropq=always"},
22712+ {Opt_diropq_a, "diropq=a"},
22713+ {Opt_diropq_w, "diropq=whiteouted"},
22714+ {Opt_diropq_w, "diropq=w"},
22715+
22716+ {Opt_warn_perm, "warn_perm"},
22717+ {Opt_nowarn_perm, "nowarn_perm"},
22718+
22719+ /* keep them temporary */
1facf9fc 22720+ {Opt_ignore_silent, "nodlgt"},
1facf9fc 22721+ {Opt_ignore_silent, "clean_plink"},
22722+
dece6358
AM
22723+#ifdef CONFIG_AUFS_SHWH
22724+ {Opt_shwh, "shwh"},
22725+#endif
22726+ {Opt_noshwh, "noshwh"},
22727+
076b876e
AM
22728+ {Opt_dirperm1, "dirperm1"},
22729+ {Opt_nodirperm1, "nodirperm1"},
22730+
1facf9fc 22731+ {Opt_verbose, "verbose"},
22732+ {Opt_verbose, "v"},
22733+ {Opt_noverbose, "noverbose"},
22734+ {Opt_noverbose, "quiet"},
22735+ {Opt_noverbose, "q"},
22736+ {Opt_noverbose, "silent"},
22737+
22738+ {Opt_sum, "sum"},
22739+ {Opt_nosum, "nosum"},
22740+ {Opt_wsum, "wsum"},
22741+
22742+ {Opt_rdcache, "rdcache=%d"},
22743+ {Opt_rdblk, "rdblk=%d"},
dece6358 22744+ {Opt_rdblk_def, "rdblk=def"},
1facf9fc 22745+ {Opt_rdhash, "rdhash=%d"},
dece6358 22746+ {Opt_rdhash_def, "rdhash=def"},
1facf9fc 22747+
22748+ {Opt_wbr_create, "create=%s"},
22749+ {Opt_wbr_create, "create_policy=%s"},
22750+ {Opt_wbr_copyup, "cpup=%s"},
22751+ {Opt_wbr_copyup, "copyup=%s"},
22752+ {Opt_wbr_copyup, "copyup_policy=%s"},
22753+
c1595e42
JR
22754+ /* generic VFS flag */
22755+#ifdef CONFIG_FS_POSIX_ACL
22756+ {Opt_acl, "acl"},
22757+ {Opt_noacl, "noacl"},
22758+#else
22759+ {Opt_ignore_silent, "acl"},
22760+ {Opt_ignore_silent, "noacl"},
22761+#endif
22762+
1facf9fc 22763+ /* internal use for the scripts */
22764+ {Opt_ignore_silent, "si=%s"},
22765+
22766+ {Opt_br, "dirs=%s"},
22767+ {Opt_ignore, "debug=%d"},
22768+ {Opt_ignore, "delete=whiteout"},
22769+ {Opt_ignore, "delete=all"},
22770+ {Opt_ignore, "imap=%s"},
22771+
1308ab2a 22772+ /* temporary workaround, due to old mount(8)? */
22773+ {Opt_ignore_silent, "relatime"},
22774+
1facf9fc 22775+ {Opt_err, NULL}
22776+};
22777+
22778+/* ---------------------------------------------------------------------- */
22779+
076b876e 22780+static const char *au_parser_pattern(int val, match_table_t tbl)
1facf9fc 22781+{
076b876e
AM
22782+ struct match_token *p;
22783+
22784+ p = tbl;
22785+ while (p->pattern) {
22786+ if (p->token == val)
22787+ return p->pattern;
22788+ p++;
1facf9fc 22789+ }
22790+ BUG();
22791+ return "??";
22792+}
22793+
076b876e
AM
22794+static const char *au_optstr(int *val, match_table_t tbl)
22795+{
22796+ struct match_token *p;
22797+ int v;
22798+
22799+ v = *val;
2000de60
JR
22800+ if (!v)
22801+ goto out;
076b876e 22802+ p = tbl;
2000de60
JR
22803+ while (p->pattern) {
22804+ if (p->token
22805+ && (v & p->token) == p->token) {
076b876e
AM
22806+ *val &= ~p->token;
22807+ return p->pattern;
22808+ }
22809+ p++;
22810+ }
2000de60
JR
22811+
22812+out:
076b876e
AM
22813+ return NULL;
22814+}
22815+
1facf9fc 22816+/* ---------------------------------------------------------------------- */
22817+
1e00d052 22818+static match_table_t brperm = {
1facf9fc 22819+ {AuBrPerm_RO, AUFS_BRPERM_RO},
22820+ {AuBrPerm_RR, AUFS_BRPERM_RR},
22821+ {AuBrPerm_RW, AUFS_BRPERM_RW},
1e00d052
AM
22822+ {0, NULL}
22823+};
1facf9fc 22824+
86dc4139 22825+static match_table_t brattr = {
076b876e
AM
22826+ /* general */
22827+ {AuBrAttr_COO_REG, AUFS_BRATTR_COO_REG},
22828+ {AuBrAttr_COO_ALL, AUFS_BRATTR_COO_ALL},
c1595e42 22829+ /* 'unpin' attrib is meaningless since linux-3.18-rc1 */
86dc4139 22830+ {AuBrAttr_UNPIN, AUFS_BRATTR_UNPIN},
2000de60 22831+#ifdef CONFIG_AUFS_FHSM
076b876e 22832+ {AuBrAttr_FHSM, AUFS_BRATTR_FHSM},
2000de60
JR
22833+#endif
22834+#ifdef CONFIG_AUFS_XATTR
c1595e42
JR
22835+ {AuBrAttr_ICEX, AUFS_BRATTR_ICEX},
22836+ {AuBrAttr_ICEX_SEC, AUFS_BRATTR_ICEX_SEC},
22837+ {AuBrAttr_ICEX_SYS, AUFS_BRATTR_ICEX_SYS},
22838+ {AuBrAttr_ICEX_TR, AUFS_BRATTR_ICEX_TR},
22839+ {AuBrAttr_ICEX_USR, AUFS_BRATTR_ICEX_USR},
22840+ {AuBrAttr_ICEX_OTH, AUFS_BRATTR_ICEX_OTH},
2000de60 22841+#endif
076b876e
AM
22842+
22843+ /* ro/rr branch */
1e00d052 22844+ {AuBrRAttr_WH, AUFS_BRRATTR_WH},
076b876e
AM
22845+
22846+ /* rw branch */
22847+ {AuBrWAttr_MOO, AUFS_BRWATTR_MOO},
1e00d052 22848+ {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
076b876e 22849+
1e00d052 22850+ {0, NULL}
1facf9fc 22851+};
22852+
1e00d052
AM
22853+static int br_attr_val(char *str, match_table_t table, substring_t args[])
22854+{
22855+ int attr, v;
22856+ char *p;
22857+
22858+ attr = 0;
22859+ do {
22860+ p = strchr(str, '+');
22861+ if (p)
22862+ *p = 0;
22863+ v = match_token(str, table, args);
076b876e
AM
22864+ if (v) {
22865+ if (v & AuBrAttr_CMOO_Mask)
22866+ attr &= ~AuBrAttr_CMOO_Mask;
1e00d052 22867+ attr |= v;
076b876e 22868+ } else {
1e00d052
AM
22869+ if (p)
22870+ *p = '+';
0c3ec466 22871+ pr_warn("ignored branch attribute %s\n", str);
1e00d052
AM
22872+ break;
22873+ }
22874+ if (p)
22875+ str = p + 1;
22876+ } while (p);
22877+
22878+ return attr;
22879+}
22880+
076b876e
AM
22881+static int au_do_optstr_br_attr(au_br_perm_str_t *str, int perm)
22882+{
22883+ int sz;
22884+ const char *p;
22885+ char *q;
22886+
076b876e
AM
22887+ q = str->a;
22888+ *q = 0;
22889+ p = au_optstr(&perm, brattr);
22890+ if (p) {
22891+ sz = strlen(p);
22892+ memcpy(q, p, sz + 1);
22893+ q += sz;
22894+ } else
22895+ goto out;
22896+
22897+ do {
22898+ p = au_optstr(&perm, brattr);
22899+ if (p) {
22900+ *q++ = '+';
22901+ sz = strlen(p);
22902+ memcpy(q, p, sz + 1);
22903+ q += sz;
22904+ }
22905+ } while (p);
22906+
22907+out:
c1595e42 22908+ return q - str->a;
076b876e
AM
22909+}
22910+
4a4d8108 22911+static int noinline_for_stack br_perm_val(char *perm)
1facf9fc 22912+{
076b876e
AM
22913+ int val, bad, sz;
22914+ char *p;
1facf9fc 22915+ substring_t args[MAX_OPT_ARGS];
076b876e 22916+ au_br_perm_str_t attr;
1facf9fc 22917+
1e00d052
AM
22918+ p = strchr(perm, '+');
22919+ if (p)
22920+ *p = 0;
22921+ val = match_token(perm, brperm, args);
22922+ if (!val) {
22923+ if (p)
22924+ *p = '+';
0c3ec466 22925+ pr_warn("ignored branch permission %s\n", perm);
1e00d052
AM
22926+ val = AuBrPerm_RO;
22927+ goto out;
22928+ }
22929+ if (!p)
22930+ goto out;
22931+
076b876e
AM
22932+ val |= br_attr_val(p + 1, brattr, args);
22933+
22934+ bad = 0;
86dc4139 22935+ switch (val & AuBrPerm_Mask) {
1e00d052
AM
22936+ case AuBrPerm_RO:
22937+ case AuBrPerm_RR:
076b876e
AM
22938+ bad = val & AuBrWAttr_Mask;
22939+ val &= ~AuBrWAttr_Mask;
1e00d052
AM
22940+ break;
22941+ case AuBrPerm_RW:
076b876e
AM
22942+ bad = val & AuBrRAttr_Mask;
22943+ val &= ~AuBrRAttr_Mask;
1e00d052
AM
22944+ break;
22945+ }
c1595e42
JR
22946+
22947+ /*
22948+ * 'unpin' attrib becomes meaningless since linux-3.18-rc1, but aufs
22949+ * does not treat it as an error, just warning.
22950+ * this is a tiny guard for the user operation.
22951+ */
22952+ if (val & AuBrAttr_UNPIN) {
22953+ bad |= AuBrAttr_UNPIN;
22954+ val &= ~AuBrAttr_UNPIN;
22955+ }
22956+
076b876e
AM
22957+ if (unlikely(bad)) {
22958+ sz = au_do_optstr_br_attr(&attr, bad);
22959+ AuDebugOn(!sz);
22960+ pr_warn("ignored branch attribute %s\n", attr.a);
22961+ }
1e00d052
AM
22962+
22963+out:
1facf9fc 22964+ return val;
22965+}
22966+
076b876e 22967+void au_optstr_br_perm(au_br_perm_str_t *str, int perm)
1facf9fc 22968+{
076b876e
AM
22969+ au_br_perm_str_t attr;
22970+ const char *p;
22971+ char *q;
1e00d052
AM
22972+ int sz;
22973+
076b876e
AM
22974+ q = str->a;
22975+ p = au_optstr(&perm, brperm);
22976+ AuDebugOn(!p || !*p);
22977+ sz = strlen(p);
22978+ memcpy(q, p, sz + 1);
22979+ q += sz;
1e00d052 22980+
076b876e
AM
22981+ sz = au_do_optstr_br_attr(&attr, perm);
22982+ if (sz) {
22983+ *q++ = '+';
22984+ memcpy(q, attr.a, sz + 1);
1e00d052
AM
22985+ }
22986+
076b876e 22987+ AuDebugOn(strlen(str->a) >= sizeof(str->a));
1facf9fc 22988+}
22989+
22990+/* ---------------------------------------------------------------------- */
22991+
22992+static match_table_t udbalevel = {
22993+ {AuOpt_UDBA_REVAL, "reval"},
22994+ {AuOpt_UDBA_NONE, "none"},
4a4d8108
AM
22995+#ifdef CONFIG_AUFS_HNOTIFY
22996+ {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
22997+#ifdef CONFIG_AUFS_HFSNOTIFY
22998+ {AuOpt_UDBA_HNOTIFY, "fsnotify"},
4a4d8108 22999+#endif
1facf9fc 23000+#endif
23001+ {-1, NULL}
23002+};
23003+
4a4d8108 23004+static int noinline_for_stack udba_val(char *str)
1facf9fc 23005+{
23006+ substring_t args[MAX_OPT_ARGS];
23007+
7f207e10 23008+ return match_token(str, udbalevel, args);
1facf9fc 23009+}
23010+
23011+const char *au_optstr_udba(int udba)
23012+{
076b876e 23013+ return au_parser_pattern(udba, udbalevel);
1facf9fc 23014+}
23015+
23016+/* ---------------------------------------------------------------------- */
23017+
23018+static match_table_t au_wbr_create_policy = {
23019+ {AuWbrCreate_TDP, "tdp"},
23020+ {AuWbrCreate_TDP, "top-down-parent"},
23021+ {AuWbrCreate_RR, "rr"},
23022+ {AuWbrCreate_RR, "round-robin"},
23023+ {AuWbrCreate_MFS, "mfs"},
23024+ {AuWbrCreate_MFS, "most-free-space"},
23025+ {AuWbrCreate_MFSV, "mfs:%d"},
23026+ {AuWbrCreate_MFSV, "most-free-space:%d"},
23027+
23028+ {AuWbrCreate_MFSRR, "mfsrr:%d"},
23029+ {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
23030+ {AuWbrCreate_PMFS, "pmfs"},
23031+ {AuWbrCreate_PMFSV, "pmfs:%d"},
392086de
AM
23032+ {AuWbrCreate_PMFSRR, "pmfsrr:%d"},
23033+ {AuWbrCreate_PMFSRRV, "pmfsrr:%d:%d"},
1facf9fc 23034+
23035+ {-1, NULL}
23036+};
23037+
dece6358
AM
23038+/*
23039+ * cf. linux/lib/parser.c and cmdline.c
23040+ * gave up calling memparse() since it uses simple_strtoull() instead of
9dbd164d 23041+ * kstrto...().
dece6358 23042+ */
4a4d8108
AM
23043+static int noinline_for_stack
23044+au_match_ull(substring_t *s, unsigned long long *result)
1facf9fc 23045+{
23046+ int err;
23047+ unsigned int len;
23048+ char a[32];
23049+
23050+ err = -ERANGE;
23051+ len = s->to - s->from;
23052+ if (len + 1 <= sizeof(a)) {
23053+ memcpy(a, s->from, len);
23054+ a[len] = '\0';
9dbd164d 23055+ err = kstrtoull(a, 0, result);
1facf9fc 23056+ }
23057+ return err;
23058+}
23059+
23060+static int au_wbr_mfs_wmark(substring_t *arg, char *str,
23061+ struct au_opt_wbr_create *create)
23062+{
23063+ int err;
23064+ unsigned long long ull;
23065+
23066+ err = 0;
23067+ if (!au_match_ull(arg, &ull))
23068+ create->mfsrr_watermark = ull;
23069+ else {
4a4d8108 23070+ pr_err("bad integer in %s\n", str);
1facf9fc 23071+ err = -EINVAL;
23072+ }
23073+
23074+ return err;
23075+}
23076+
23077+static int au_wbr_mfs_sec(substring_t *arg, char *str,
23078+ struct au_opt_wbr_create *create)
23079+{
23080+ int n, err;
23081+
23082+ err = 0;
027c5e7a 23083+ if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
1facf9fc 23084+ create->mfs_second = n;
23085+ else {
4a4d8108 23086+ pr_err("bad integer in %s\n", str);
1facf9fc 23087+ err = -EINVAL;
23088+ }
23089+
23090+ return err;
23091+}
23092+
4a4d8108
AM
23093+static int noinline_for_stack
23094+au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
1facf9fc 23095+{
23096+ int err, e;
23097+ substring_t args[MAX_OPT_ARGS];
23098+
23099+ err = match_token(str, au_wbr_create_policy, args);
23100+ create->wbr_create = err;
23101+ switch (err) {
23102+ case AuWbrCreate_MFSRRV:
392086de 23103+ case AuWbrCreate_PMFSRRV:
1facf9fc 23104+ e = au_wbr_mfs_wmark(&args[0], str, create);
23105+ if (!e)
23106+ e = au_wbr_mfs_sec(&args[1], str, create);
23107+ if (unlikely(e))
23108+ err = e;
23109+ break;
23110+ case AuWbrCreate_MFSRR:
392086de 23111+ case AuWbrCreate_PMFSRR:
1facf9fc 23112+ e = au_wbr_mfs_wmark(&args[0], str, create);
23113+ if (unlikely(e)) {
23114+ err = e;
23115+ break;
23116+ }
23117+ /*FALLTHROUGH*/
23118+ case AuWbrCreate_MFS:
23119+ case AuWbrCreate_PMFS:
027c5e7a 23120+ create->mfs_second = AUFS_MFS_DEF_SEC;
1facf9fc 23121+ break;
23122+ case AuWbrCreate_MFSV:
23123+ case AuWbrCreate_PMFSV:
23124+ e = au_wbr_mfs_sec(&args[0], str, create);
23125+ if (unlikely(e))
23126+ err = e;
23127+ break;
23128+ }
23129+
23130+ return err;
23131+}
23132+
23133+const char *au_optstr_wbr_create(int wbr_create)
23134+{
076b876e 23135+ return au_parser_pattern(wbr_create, au_wbr_create_policy);
1facf9fc 23136+}
23137+
23138+static match_table_t au_wbr_copyup_policy = {
23139+ {AuWbrCopyup_TDP, "tdp"},
23140+ {AuWbrCopyup_TDP, "top-down-parent"},
23141+ {AuWbrCopyup_BUP, "bup"},
23142+ {AuWbrCopyup_BUP, "bottom-up-parent"},
23143+ {AuWbrCopyup_BU, "bu"},
23144+ {AuWbrCopyup_BU, "bottom-up"},
23145+ {-1, NULL}
23146+};
23147+
4a4d8108 23148+static int noinline_for_stack au_wbr_copyup_val(char *str)
1facf9fc 23149+{
23150+ substring_t args[MAX_OPT_ARGS];
23151+
23152+ return match_token(str, au_wbr_copyup_policy, args);
23153+}
23154+
23155+const char *au_optstr_wbr_copyup(int wbr_copyup)
23156+{
076b876e 23157+ return au_parser_pattern(wbr_copyup, au_wbr_copyup_policy);
1facf9fc 23158+}
23159+
23160+/* ---------------------------------------------------------------------- */
23161+
23162+static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
23163+
23164+static void dump_opts(struct au_opts *opts)
23165+{
23166+#ifdef CONFIG_AUFS_DEBUG
23167+ /* reduce stack space */
23168+ union {
23169+ struct au_opt_add *add;
23170+ struct au_opt_del *del;
23171+ struct au_opt_mod *mod;
23172+ struct au_opt_xino *xino;
23173+ struct au_opt_xino_itrunc *xino_itrunc;
23174+ struct au_opt_wbr_create *create;
23175+ } u;
23176+ struct au_opt *opt;
23177+
23178+ opt = opts->opt;
23179+ while (opt->type != Opt_tail) {
23180+ switch (opt->type) {
23181+ case Opt_add:
23182+ u.add = &opt->add;
23183+ AuDbg("add {b%d, %s, 0x%x, %p}\n",
23184+ u.add->bindex, u.add->pathname, u.add->perm,
23185+ u.add->path.dentry);
23186+ break;
23187+ case Opt_del:
23188+ case Opt_idel:
23189+ u.del = &opt->del;
23190+ AuDbg("del {%s, %p}\n",
23191+ u.del->pathname, u.del->h_path.dentry);
23192+ break;
23193+ case Opt_mod:
23194+ case Opt_imod:
23195+ u.mod = &opt->mod;
23196+ AuDbg("mod {%s, 0x%x, %p}\n",
23197+ u.mod->path, u.mod->perm, u.mod->h_root);
23198+ break;
23199+ case Opt_append:
23200+ u.add = &opt->add;
23201+ AuDbg("append {b%d, %s, 0x%x, %p}\n",
23202+ u.add->bindex, u.add->pathname, u.add->perm,
23203+ u.add->path.dentry);
23204+ break;
23205+ case Opt_prepend:
23206+ u.add = &opt->add;
23207+ AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
23208+ u.add->bindex, u.add->pathname, u.add->perm,
23209+ u.add->path.dentry);
23210+ break;
23211+ case Opt_dirwh:
23212+ AuDbg("dirwh %d\n", opt->dirwh);
23213+ break;
23214+ case Opt_rdcache:
23215+ AuDbg("rdcache %d\n", opt->rdcache);
23216+ break;
23217+ case Opt_rdblk:
23218+ AuDbg("rdblk %u\n", opt->rdblk);
23219+ break;
dece6358
AM
23220+ case Opt_rdblk_def:
23221+ AuDbg("rdblk_def\n");
23222+ break;
1facf9fc 23223+ case Opt_rdhash:
23224+ AuDbg("rdhash %u\n", opt->rdhash);
23225+ break;
dece6358
AM
23226+ case Opt_rdhash_def:
23227+ AuDbg("rdhash_def\n");
23228+ break;
1facf9fc 23229+ case Opt_xino:
23230+ u.xino = &opt->xino;
523b37e3 23231+ AuDbg("xino {%s %pD}\n", u.xino->path, u.xino->file);
1facf9fc 23232+ break;
23233+ case Opt_trunc_xino:
23234+ AuLabel(trunc_xino);
23235+ break;
23236+ case Opt_notrunc_xino:
23237+ AuLabel(notrunc_xino);
23238+ break;
23239+ case Opt_trunc_xino_path:
23240+ case Opt_itrunc_xino:
23241+ u.xino_itrunc = &opt->xino_itrunc;
23242+ AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
23243+ break;
1facf9fc 23244+ case Opt_noxino:
23245+ AuLabel(noxino);
23246+ break;
23247+ case Opt_trunc_xib:
23248+ AuLabel(trunc_xib);
23249+ break;
23250+ case Opt_notrunc_xib:
23251+ AuLabel(notrunc_xib);
23252+ break;
dece6358
AM
23253+ case Opt_shwh:
23254+ AuLabel(shwh);
23255+ break;
23256+ case Opt_noshwh:
23257+ AuLabel(noshwh);
23258+ break;
076b876e
AM
23259+ case Opt_dirperm1:
23260+ AuLabel(dirperm1);
23261+ break;
23262+ case Opt_nodirperm1:
23263+ AuLabel(nodirperm1);
23264+ break;
1facf9fc 23265+ case Opt_plink:
23266+ AuLabel(plink);
23267+ break;
23268+ case Opt_noplink:
23269+ AuLabel(noplink);
23270+ break;
23271+ case Opt_list_plink:
23272+ AuLabel(list_plink);
23273+ break;
23274+ case Opt_udba:
23275+ AuDbg("udba %d, %s\n",
23276+ opt->udba, au_optstr_udba(opt->udba));
23277+ break;
4a4d8108
AM
23278+ case Opt_dio:
23279+ AuLabel(dio);
23280+ break;
23281+ case Opt_nodio:
23282+ AuLabel(nodio);
23283+ break;
1facf9fc 23284+ case Opt_diropq_a:
23285+ AuLabel(diropq_a);
23286+ break;
23287+ case Opt_diropq_w:
23288+ AuLabel(diropq_w);
23289+ break;
23290+ case Opt_warn_perm:
23291+ AuLabel(warn_perm);
23292+ break;
23293+ case Opt_nowarn_perm:
23294+ AuLabel(nowarn_perm);
23295+ break;
1facf9fc 23296+ case Opt_verbose:
23297+ AuLabel(verbose);
23298+ break;
23299+ case Opt_noverbose:
23300+ AuLabel(noverbose);
23301+ break;
23302+ case Opt_sum:
23303+ AuLabel(sum);
23304+ break;
23305+ case Opt_nosum:
23306+ AuLabel(nosum);
23307+ break;
23308+ case Opt_wsum:
23309+ AuLabel(wsum);
23310+ break;
23311+ case Opt_wbr_create:
23312+ u.create = &opt->wbr_create;
23313+ AuDbg("create %d, %s\n", u.create->wbr_create,
23314+ au_optstr_wbr_create(u.create->wbr_create));
23315+ switch (u.create->wbr_create) {
23316+ case AuWbrCreate_MFSV:
23317+ case AuWbrCreate_PMFSV:
23318+ AuDbg("%d sec\n", u.create->mfs_second);
23319+ break;
23320+ case AuWbrCreate_MFSRR:
23321+ AuDbg("%llu watermark\n",
23322+ u.create->mfsrr_watermark);
23323+ break;
23324+ case AuWbrCreate_MFSRRV:
392086de 23325+ case AuWbrCreate_PMFSRRV:
1facf9fc 23326+ AuDbg("%llu watermark, %d sec\n",
23327+ u.create->mfsrr_watermark,
23328+ u.create->mfs_second);
23329+ break;
23330+ }
23331+ break;
23332+ case Opt_wbr_copyup:
23333+ AuDbg("copyup %d, %s\n", opt->wbr_copyup,
23334+ au_optstr_wbr_copyup(opt->wbr_copyup));
23335+ break;
076b876e
AM
23336+ case Opt_fhsm_sec:
23337+ AuDbg("fhsm_sec %u\n", opt->fhsm_second);
23338+ break;
c1595e42
JR
23339+ case Opt_acl:
23340+ AuLabel(acl);
23341+ break;
23342+ case Opt_noacl:
23343+ AuLabel(noacl);
23344+ break;
1facf9fc 23345+ default:
23346+ BUG();
23347+ }
23348+ opt++;
23349+ }
23350+#endif
23351+}
23352+
23353+void au_opts_free(struct au_opts *opts)
23354+{
23355+ struct au_opt *opt;
23356+
23357+ opt = opts->opt;
23358+ while (opt->type != Opt_tail) {
23359+ switch (opt->type) {
23360+ case Opt_add:
23361+ case Opt_append:
23362+ case Opt_prepend:
23363+ path_put(&opt->add.path);
23364+ break;
23365+ case Opt_del:
23366+ case Opt_idel:
23367+ path_put(&opt->del.h_path);
23368+ break;
23369+ case Opt_mod:
23370+ case Opt_imod:
23371+ dput(opt->mod.h_root);
23372+ break;
23373+ case Opt_xino:
23374+ fput(opt->xino.file);
23375+ break;
23376+ }
23377+ opt++;
23378+ }
23379+}
23380+
23381+static int opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
23382+ aufs_bindex_t bindex)
23383+{
23384+ int err;
23385+ struct au_opt_add *add = &opt->add;
23386+ char *p;
23387+
23388+ add->bindex = bindex;
1e00d052 23389+ add->perm = AuBrPerm_RO;
1facf9fc 23390+ add->pathname = opt_str;
23391+ p = strchr(opt_str, '=');
23392+ if (p) {
23393+ *p++ = 0;
23394+ if (*p)
23395+ add->perm = br_perm_val(p);
23396+ }
23397+
23398+ err = vfsub_kern_path(add->pathname, lkup_dirflags, &add->path);
23399+ if (!err) {
23400+ if (!p) {
23401+ add->perm = AuBrPerm_RO;
23402+ if (au_test_fs_rr(add->path.dentry->d_sb))
23403+ add->perm = AuBrPerm_RR;
23404+ else if (!bindex && !(sb_flags & MS_RDONLY))
23405+ add->perm = AuBrPerm_RW;
23406+ }
23407+ opt->type = Opt_add;
23408+ goto out;
23409+ }
4a4d8108 23410+ pr_err("lookup failed %s (%d)\n", add->pathname, err);
1facf9fc 23411+ err = -EINVAL;
23412+
4f0767ce 23413+out:
1facf9fc 23414+ return err;
23415+}
23416+
23417+static int au_opts_parse_del(struct au_opt_del *del, substring_t args[])
23418+{
23419+ int err;
23420+
23421+ del->pathname = args[0].from;
23422+ AuDbg("del path %s\n", del->pathname);
23423+
23424+ err = vfsub_kern_path(del->pathname, lkup_dirflags, &del->h_path);
23425+ if (unlikely(err))
4a4d8108 23426+ pr_err("lookup failed %s (%d)\n", del->pathname, err);
1facf9fc 23427+
23428+ return err;
23429+}
23430+
23431+#if 0 /* reserved for future use */
23432+static int au_opts_parse_idel(struct super_block *sb, aufs_bindex_t bindex,
23433+ struct au_opt_del *del, substring_t args[])
23434+{
23435+ int err;
23436+ struct dentry *root;
23437+
23438+ err = -EINVAL;
23439+ root = sb->s_root;
23440+ aufs_read_lock(root, AuLock_FLUSH);
23441+ if (bindex < 0 || au_sbend(sb) < bindex) {
4a4d8108 23442+ pr_err("out of bounds, %d\n", bindex);
1facf9fc 23443+ goto out;
23444+ }
23445+
23446+ err = 0;
23447+ del->h_path.dentry = dget(au_h_dptr(root, bindex));
23448+ del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
23449+
4f0767ce 23450+out:
1facf9fc 23451+ aufs_read_unlock(root, !AuLock_IR);
23452+ return err;
23453+}
23454+#endif
23455+
4a4d8108
AM
23456+static int noinline_for_stack
23457+au_opts_parse_mod(struct au_opt_mod *mod, substring_t args[])
1facf9fc 23458+{
23459+ int err;
23460+ struct path path;
23461+ char *p;
23462+
23463+ err = -EINVAL;
23464+ mod->path = args[0].from;
23465+ p = strchr(mod->path, '=');
23466+ if (unlikely(!p)) {
4a4d8108 23467+ pr_err("no permssion %s\n", args[0].from);
1facf9fc 23468+ goto out;
23469+ }
23470+
23471+ *p++ = 0;
23472+ err = vfsub_kern_path(mod->path, lkup_dirflags, &path);
23473+ if (unlikely(err)) {
4a4d8108 23474+ pr_err("lookup failed %s (%d)\n", mod->path, err);
1facf9fc 23475+ goto out;
23476+ }
23477+
23478+ mod->perm = br_perm_val(p);
23479+ AuDbg("mod path %s, perm 0x%x, %s\n", mod->path, mod->perm, p);
23480+ mod->h_root = dget(path.dentry);
23481+ path_put(&path);
23482+
4f0767ce 23483+out:
1facf9fc 23484+ return err;
23485+}
23486+
23487+#if 0 /* reserved for future use */
23488+static int au_opts_parse_imod(struct super_block *sb, aufs_bindex_t bindex,
23489+ struct au_opt_mod *mod, substring_t args[])
23490+{
23491+ int err;
23492+ struct dentry *root;
23493+
23494+ err = -EINVAL;
23495+ root = sb->s_root;
23496+ aufs_read_lock(root, AuLock_FLUSH);
23497+ if (bindex < 0 || au_sbend(sb) < bindex) {
4a4d8108 23498+ pr_err("out of bounds, %d\n", bindex);
1facf9fc 23499+ goto out;
23500+ }
23501+
23502+ err = 0;
23503+ mod->perm = br_perm_val(args[1].from);
23504+ AuDbg("mod path %s, perm 0x%x, %s\n",
23505+ mod->path, mod->perm, args[1].from);
23506+ mod->h_root = dget(au_h_dptr(root, bindex));
23507+
4f0767ce 23508+out:
1facf9fc 23509+ aufs_read_unlock(root, !AuLock_IR);
23510+ return err;
23511+}
23512+#endif
23513+
23514+static int au_opts_parse_xino(struct super_block *sb, struct au_opt_xino *xino,
23515+ substring_t args[])
23516+{
23517+ int err;
23518+ struct file *file;
23519+
23520+ file = au_xino_create(sb, args[0].from, /*silent*/0);
23521+ err = PTR_ERR(file);
23522+ if (IS_ERR(file))
23523+ goto out;
23524+
23525+ err = -EINVAL;
2000de60 23526+ if (unlikely(file->f_path.dentry->d_sb == sb)) {
1facf9fc 23527+ fput(file);
4a4d8108 23528+ pr_err("%s must be outside\n", args[0].from);
1facf9fc 23529+ goto out;
23530+ }
23531+
23532+ err = 0;
23533+ xino->file = file;
23534+ xino->path = args[0].from;
23535+
4f0767ce 23536+out:
1facf9fc 23537+ return err;
23538+}
23539+
4a4d8108
AM
23540+static int noinline_for_stack
23541+au_opts_parse_xino_itrunc_path(struct super_block *sb,
23542+ struct au_opt_xino_itrunc *xino_itrunc,
23543+ substring_t args[])
1facf9fc 23544+{
23545+ int err;
23546+ aufs_bindex_t bend, bindex;
23547+ struct path path;
23548+ struct dentry *root;
23549+
23550+ err = vfsub_kern_path(args[0].from, lkup_dirflags, &path);
23551+ if (unlikely(err)) {
4a4d8108 23552+ pr_err("lookup failed %s (%d)\n", args[0].from, err);
1facf9fc 23553+ goto out;
23554+ }
23555+
23556+ xino_itrunc->bindex = -1;
23557+ root = sb->s_root;
23558+ aufs_read_lock(root, AuLock_FLUSH);
23559+ bend = au_sbend(sb);
23560+ for (bindex = 0; bindex <= bend; bindex++) {
23561+ if (au_h_dptr(root, bindex) == path.dentry) {
23562+ xino_itrunc->bindex = bindex;
23563+ break;
23564+ }
23565+ }
23566+ aufs_read_unlock(root, !AuLock_IR);
23567+ path_put(&path);
23568+
23569+ if (unlikely(xino_itrunc->bindex < 0)) {
4a4d8108 23570+ pr_err("no such branch %s\n", args[0].from);
1facf9fc 23571+ err = -EINVAL;
23572+ }
23573+
4f0767ce 23574+out:
1facf9fc 23575+ return err;
23576+}
23577+
23578+/* called without aufs lock */
23579+int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts)
23580+{
23581+ int err, n, token;
23582+ aufs_bindex_t bindex;
23583+ unsigned char skipped;
23584+ struct dentry *root;
23585+ struct au_opt *opt, *opt_tail;
23586+ char *opt_str;
23587+ /* reduce the stack space */
23588+ union {
23589+ struct au_opt_xino_itrunc *xino_itrunc;
23590+ struct au_opt_wbr_create *create;
23591+ } u;
23592+ struct {
23593+ substring_t args[MAX_OPT_ARGS];
23594+ } *a;
23595+
23596+ err = -ENOMEM;
23597+ a = kmalloc(sizeof(*a), GFP_NOFS);
23598+ if (unlikely(!a))
23599+ goto out;
23600+
23601+ root = sb->s_root;
23602+ err = 0;
23603+ bindex = 0;
23604+ opt = opts->opt;
23605+ opt_tail = opt + opts->max_opt - 1;
23606+ opt->type = Opt_tail;
23607+ while (!err && (opt_str = strsep(&str, ",")) && *opt_str) {
23608+ err = -EINVAL;
23609+ skipped = 0;
23610+ token = match_token(opt_str, options, a->args);
23611+ switch (token) {
23612+ case Opt_br:
23613+ err = 0;
23614+ while (!err && (opt_str = strsep(&a->args[0].from, ":"))
23615+ && *opt_str) {
23616+ err = opt_add(opt, opt_str, opts->sb_flags,
23617+ bindex++);
23618+ if (unlikely(!err && ++opt > opt_tail)) {
23619+ err = -E2BIG;
23620+ break;
23621+ }
23622+ opt->type = Opt_tail;
23623+ skipped = 1;
23624+ }
23625+ break;
23626+ case Opt_add:
23627+ if (unlikely(match_int(&a->args[0], &n))) {
4a4d8108 23628+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23629+ break;
23630+ }
23631+ bindex = n;
23632+ err = opt_add(opt, a->args[1].from, opts->sb_flags,
23633+ bindex);
23634+ if (!err)
23635+ opt->type = token;
23636+ break;
23637+ case Opt_append:
23638+ err = opt_add(opt, a->args[0].from, opts->sb_flags,
23639+ /*dummy bindex*/1);
23640+ if (!err)
23641+ opt->type = token;
23642+ break;
23643+ case Opt_prepend:
23644+ err = opt_add(opt, a->args[0].from, opts->sb_flags,
23645+ /*bindex*/0);
23646+ if (!err)
23647+ opt->type = token;
23648+ break;
23649+ case Opt_del:
23650+ err = au_opts_parse_del(&opt->del, a->args);
23651+ if (!err)
23652+ opt->type = token;
23653+ break;
23654+#if 0 /* reserved for future use */
23655+ case Opt_idel:
23656+ del->pathname = "(indexed)";
23657+ if (unlikely(match_int(&args[0], &n))) {
4a4d8108 23658+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23659+ break;
23660+ }
23661+ err = au_opts_parse_idel(sb, n, &opt->del, a->args);
23662+ if (!err)
23663+ opt->type = token;
23664+ break;
23665+#endif
23666+ case Opt_mod:
23667+ err = au_opts_parse_mod(&opt->mod, a->args);
23668+ if (!err)
23669+ opt->type = token;
23670+ break;
23671+#ifdef IMOD /* reserved for future use */
23672+ case Opt_imod:
23673+ u.mod->path = "(indexed)";
23674+ if (unlikely(match_int(&a->args[0], &n))) {
4a4d8108 23675+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23676+ break;
23677+ }
23678+ err = au_opts_parse_imod(sb, n, &opt->mod, a->args);
23679+ if (!err)
23680+ opt->type = token;
23681+ break;
23682+#endif
23683+ case Opt_xino:
23684+ err = au_opts_parse_xino(sb, &opt->xino, a->args);
23685+ if (!err)
23686+ opt->type = token;
23687+ break;
23688+
23689+ case Opt_trunc_xino_path:
23690+ err = au_opts_parse_xino_itrunc_path
23691+ (sb, &opt->xino_itrunc, a->args);
23692+ if (!err)
23693+ opt->type = token;
23694+ break;
23695+
23696+ case Opt_itrunc_xino:
23697+ u.xino_itrunc = &opt->xino_itrunc;
23698+ if (unlikely(match_int(&a->args[0], &n))) {
4a4d8108 23699+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23700+ break;
23701+ }
23702+ u.xino_itrunc->bindex = n;
23703+ aufs_read_lock(root, AuLock_FLUSH);
23704+ if (n < 0 || au_sbend(sb) < n) {
4a4d8108 23705+ pr_err("out of bounds, %d\n", n);
1facf9fc 23706+ aufs_read_unlock(root, !AuLock_IR);
23707+ break;
23708+ }
23709+ aufs_read_unlock(root, !AuLock_IR);
23710+ err = 0;
23711+ opt->type = token;
23712+ break;
23713+
23714+ case Opt_dirwh:
23715+ if (unlikely(match_int(&a->args[0], &opt->dirwh)))
23716+ break;
23717+ err = 0;
23718+ opt->type = token;
23719+ break;
23720+
23721+ case Opt_rdcache:
027c5e7a
AM
23722+ if (unlikely(match_int(&a->args[0], &n))) {
23723+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23724+ break;
027c5e7a
AM
23725+ }
23726+ if (unlikely(n > AUFS_RDCACHE_MAX)) {
23727+ pr_err("rdcache must be smaller than %d\n",
23728+ AUFS_RDCACHE_MAX);
23729+ break;
23730+ }
23731+ opt->rdcache = n;
1facf9fc 23732+ err = 0;
23733+ opt->type = token;
23734+ break;
23735+ case Opt_rdblk:
23736+ if (unlikely(match_int(&a->args[0], &n)
1308ab2a 23737+ || n < 0
1facf9fc 23738+ || n > KMALLOC_MAX_SIZE)) {
4a4d8108 23739+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23740+ break;
23741+ }
1308ab2a 23742+ if (unlikely(n && n < NAME_MAX)) {
4a4d8108
AM
23743+ pr_err("rdblk must be larger than %d\n",
23744+ NAME_MAX);
1facf9fc 23745+ break;
23746+ }
23747+ opt->rdblk = n;
23748+ err = 0;
23749+ opt->type = token;
23750+ break;
23751+ case Opt_rdhash:
23752+ if (unlikely(match_int(&a->args[0], &n)
1308ab2a 23753+ || n < 0
1facf9fc 23754+ || n * sizeof(struct hlist_head)
23755+ > KMALLOC_MAX_SIZE)) {
4a4d8108 23756+ pr_err("bad integer in %s\n", opt_str);
1facf9fc 23757+ break;
23758+ }
23759+ opt->rdhash = n;
23760+ err = 0;
23761+ opt->type = token;
23762+ break;
23763+
23764+ case Opt_trunc_xino:
23765+ case Opt_notrunc_xino:
23766+ case Opt_noxino:
23767+ case Opt_trunc_xib:
23768+ case Opt_notrunc_xib:
dece6358
AM
23769+ case Opt_shwh:
23770+ case Opt_noshwh:
076b876e
AM
23771+ case Opt_dirperm1:
23772+ case Opt_nodirperm1:
1facf9fc 23773+ case Opt_plink:
23774+ case Opt_noplink:
23775+ case Opt_list_plink:
4a4d8108
AM
23776+ case Opt_dio:
23777+ case Opt_nodio:
1facf9fc 23778+ case Opt_diropq_a:
23779+ case Opt_diropq_w:
23780+ case Opt_warn_perm:
23781+ case Opt_nowarn_perm:
1facf9fc 23782+ case Opt_verbose:
23783+ case Opt_noverbose:
23784+ case Opt_sum:
23785+ case Opt_nosum:
23786+ case Opt_wsum:
dece6358
AM
23787+ case Opt_rdblk_def:
23788+ case Opt_rdhash_def:
c1595e42
JR
23789+ case Opt_acl:
23790+ case Opt_noacl:
1facf9fc 23791+ err = 0;
23792+ opt->type = token;
23793+ break;
23794+
23795+ case Opt_udba:
23796+ opt->udba = udba_val(a->args[0].from);
23797+ if (opt->udba >= 0) {
23798+ err = 0;
23799+ opt->type = token;
23800+ } else
4a4d8108 23801+ pr_err("wrong value, %s\n", opt_str);
1facf9fc 23802+ break;
23803+
23804+ case Opt_wbr_create:
23805+ u.create = &opt->wbr_create;
23806+ u.create->wbr_create
23807+ = au_wbr_create_val(a->args[0].from, u.create);
23808+ if (u.create->wbr_create >= 0) {
23809+ err = 0;
23810+ opt->type = token;
23811+ } else
4a4d8108 23812+ pr_err("wrong value, %s\n", opt_str);
1facf9fc 23813+ break;
23814+ case Opt_wbr_copyup:
23815+ opt->wbr_copyup = au_wbr_copyup_val(a->args[0].from);
23816+ if (opt->wbr_copyup >= 0) {
23817+ err = 0;
23818+ opt->type = token;
23819+ } else
4a4d8108 23820+ pr_err("wrong value, %s\n", opt_str);
1facf9fc 23821+ break;
23822+
076b876e
AM
23823+ case Opt_fhsm_sec:
23824+ if (unlikely(match_int(&a->args[0], &n)
23825+ || n < 0)) {
23826+ pr_err("bad integer in %s\n", opt_str);
23827+ break;
23828+ }
23829+ if (sysaufs_brs) {
23830+ opt->fhsm_second = n;
23831+ opt->type = token;
23832+ } else
23833+ pr_warn("ignored %s\n", opt_str);
23834+ err = 0;
23835+ break;
23836+
1facf9fc 23837+ case Opt_ignore:
0c3ec466 23838+ pr_warn("ignored %s\n", opt_str);
1facf9fc 23839+ /*FALLTHROUGH*/
23840+ case Opt_ignore_silent:
23841+ skipped = 1;
23842+ err = 0;
23843+ break;
23844+ case Opt_err:
4a4d8108 23845+ pr_err("unknown option %s\n", opt_str);
1facf9fc 23846+ break;
23847+ }
23848+
23849+ if (!err && !skipped) {
23850+ if (unlikely(++opt > opt_tail)) {
23851+ err = -E2BIG;
23852+ opt--;
23853+ opt->type = Opt_tail;
23854+ break;
23855+ }
23856+ opt->type = Opt_tail;
23857+ }
23858+ }
23859+
23860+ kfree(a);
23861+ dump_opts(opts);
23862+ if (unlikely(err))
23863+ au_opts_free(opts);
23864+
4f0767ce 23865+out:
1facf9fc 23866+ return err;
23867+}
23868+
23869+static int au_opt_wbr_create(struct super_block *sb,
23870+ struct au_opt_wbr_create *create)
23871+{
23872+ int err;
23873+ struct au_sbinfo *sbinfo;
23874+
dece6358
AM
23875+ SiMustWriteLock(sb);
23876+
1facf9fc 23877+ err = 1; /* handled */
23878+ sbinfo = au_sbi(sb);
23879+ if (sbinfo->si_wbr_create_ops->fin) {
23880+ err = sbinfo->si_wbr_create_ops->fin(sb);
23881+ if (!err)
23882+ err = 1;
23883+ }
23884+
23885+ sbinfo->si_wbr_create = create->wbr_create;
23886+ sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
23887+ switch (create->wbr_create) {
23888+ case AuWbrCreate_MFSRRV:
23889+ case AuWbrCreate_MFSRR:
392086de
AM
23890+ case AuWbrCreate_PMFSRR:
23891+ case AuWbrCreate_PMFSRRV:
1facf9fc 23892+ sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
23893+ /*FALLTHROUGH*/
23894+ case AuWbrCreate_MFS:
23895+ case AuWbrCreate_MFSV:
23896+ case AuWbrCreate_PMFS:
23897+ case AuWbrCreate_PMFSV:
e49829fe
JR
23898+ sbinfo->si_wbr_mfs.mfs_expire
23899+ = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
1facf9fc 23900+ break;
23901+ }
23902+
23903+ if (sbinfo->si_wbr_create_ops->init)
23904+ sbinfo->si_wbr_create_ops->init(sb); /* ignore */
23905+
23906+ return err;
23907+}
23908+
23909+/*
23910+ * returns,
23911+ * plus: processed without an error
23912+ * zero: unprocessed
23913+ */
23914+static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
23915+ struct au_opts *opts)
23916+{
23917+ int err;
23918+ struct au_sbinfo *sbinfo;
23919+
dece6358
AM
23920+ SiMustWriteLock(sb);
23921+
1facf9fc 23922+ err = 1; /* handled */
23923+ sbinfo = au_sbi(sb);
23924+ switch (opt->type) {
23925+ case Opt_udba:
23926+ sbinfo->si_mntflags &= ~AuOptMask_UDBA;
23927+ sbinfo->si_mntflags |= opt->udba;
23928+ opts->given_udba |= opt->udba;
23929+ break;
23930+
23931+ case Opt_plink:
23932+ au_opt_set(sbinfo->si_mntflags, PLINK);
23933+ break;
23934+ case Opt_noplink:
23935+ if (au_opt_test(sbinfo->si_mntflags, PLINK))
e49829fe 23936+ au_plink_put(sb, /*verbose*/1);
1facf9fc 23937+ au_opt_clr(sbinfo->si_mntflags, PLINK);
23938+ break;
23939+ case Opt_list_plink:
23940+ if (au_opt_test(sbinfo->si_mntflags, PLINK))
23941+ au_plink_list(sb);
23942+ break;
23943+
4a4d8108
AM
23944+ case Opt_dio:
23945+ au_opt_set(sbinfo->si_mntflags, DIO);
23946+ au_fset_opts(opts->flags, REFRESH_DYAOP);
23947+ break;
23948+ case Opt_nodio:
23949+ au_opt_clr(sbinfo->si_mntflags, DIO);
23950+ au_fset_opts(opts->flags, REFRESH_DYAOP);
23951+ break;
23952+
076b876e
AM
23953+ case Opt_fhsm_sec:
23954+ au_fhsm_set(sbinfo, opt->fhsm_second);
23955+ break;
23956+
1facf9fc 23957+ case Opt_diropq_a:
23958+ au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
23959+ break;
23960+ case Opt_diropq_w:
23961+ au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
23962+ break;
23963+
23964+ case Opt_warn_perm:
23965+ au_opt_set(sbinfo->si_mntflags, WARN_PERM);
23966+ break;
23967+ case Opt_nowarn_perm:
23968+ au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
23969+ break;
23970+
1facf9fc 23971+ case Opt_verbose:
23972+ au_opt_set(sbinfo->si_mntflags, VERBOSE);
23973+ break;
23974+ case Opt_noverbose:
23975+ au_opt_clr(sbinfo->si_mntflags, VERBOSE);
23976+ break;
23977+
23978+ case Opt_sum:
23979+ au_opt_set(sbinfo->si_mntflags, SUM);
23980+ break;
23981+ case Opt_wsum:
23982+ au_opt_clr(sbinfo->si_mntflags, SUM);
23983+ au_opt_set(sbinfo->si_mntflags, SUM_W);
23984+ case Opt_nosum:
23985+ au_opt_clr(sbinfo->si_mntflags, SUM);
23986+ au_opt_clr(sbinfo->si_mntflags, SUM_W);
23987+ break;
23988+
23989+ case Opt_wbr_create:
23990+ err = au_opt_wbr_create(sb, &opt->wbr_create);
23991+ break;
23992+ case Opt_wbr_copyup:
23993+ sbinfo->si_wbr_copyup = opt->wbr_copyup;
23994+ sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
23995+ break;
23996+
23997+ case Opt_dirwh:
23998+ sbinfo->si_dirwh = opt->dirwh;
23999+ break;
24000+
24001+ case Opt_rdcache:
e49829fe
JR
24002+ sbinfo->si_rdcache
24003+ = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
1facf9fc 24004+ break;
24005+ case Opt_rdblk:
24006+ sbinfo->si_rdblk = opt->rdblk;
24007+ break;
dece6358
AM
24008+ case Opt_rdblk_def:
24009+ sbinfo->si_rdblk = AUFS_RDBLK_DEF;
24010+ break;
1facf9fc 24011+ case Opt_rdhash:
24012+ sbinfo->si_rdhash = opt->rdhash;
24013+ break;
dece6358
AM
24014+ case Opt_rdhash_def:
24015+ sbinfo->si_rdhash = AUFS_RDHASH_DEF;
24016+ break;
24017+
24018+ case Opt_shwh:
24019+ au_opt_set(sbinfo->si_mntflags, SHWH);
24020+ break;
24021+ case Opt_noshwh:
24022+ au_opt_clr(sbinfo->si_mntflags, SHWH);
24023+ break;
1facf9fc 24024+
076b876e
AM
24025+ case Opt_dirperm1:
24026+ au_opt_set(sbinfo->si_mntflags, DIRPERM1);
24027+ break;
24028+ case Opt_nodirperm1:
24029+ au_opt_clr(sbinfo->si_mntflags, DIRPERM1);
24030+ break;
24031+
1facf9fc 24032+ case Opt_trunc_xino:
24033+ au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
24034+ break;
24035+ case Opt_notrunc_xino:
24036+ au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
24037+ break;
24038+
24039+ case Opt_trunc_xino_path:
24040+ case Opt_itrunc_xino:
24041+ err = au_xino_trunc(sb, opt->xino_itrunc.bindex);
24042+ if (!err)
24043+ err = 1;
24044+ break;
24045+
24046+ case Opt_trunc_xib:
24047+ au_fset_opts(opts->flags, TRUNC_XIB);
24048+ break;
24049+ case Opt_notrunc_xib:
24050+ au_fclr_opts(opts->flags, TRUNC_XIB);
24051+ break;
24052+
c1595e42
JR
24053+ case Opt_acl:
24054+ sb->s_flags |= MS_POSIXACL;
24055+ break;
24056+ case Opt_noacl:
24057+ sb->s_flags &= ~MS_POSIXACL;
24058+ break;
24059+
1facf9fc 24060+ default:
24061+ err = 0;
24062+ break;
24063+ }
24064+
24065+ return err;
24066+}
24067+
24068+/*
24069+ * returns tri-state.
24070+ * plus: processed without an error
24071+ * zero: unprocessed
24072+ * minus: error
24073+ */
24074+static int au_opt_br(struct super_block *sb, struct au_opt *opt,
24075+ struct au_opts *opts)
24076+{
24077+ int err, do_refresh;
24078+
24079+ err = 0;
24080+ switch (opt->type) {
24081+ case Opt_append:
24082+ opt->add.bindex = au_sbend(sb) + 1;
24083+ if (opt->add.bindex < 0)
24084+ opt->add.bindex = 0;
24085+ goto add;
24086+ case Opt_prepend:
24087+ opt->add.bindex = 0;
f6b6e03d 24088+ add: /* indented label */
1facf9fc 24089+ case Opt_add:
24090+ err = au_br_add(sb, &opt->add,
24091+ au_ftest_opts(opts->flags, REMOUNT));
24092+ if (!err) {
24093+ err = 1;
027c5e7a 24094+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 24095+ }
24096+ break;
24097+
24098+ case Opt_del:
24099+ case Opt_idel:
24100+ err = au_br_del(sb, &opt->del,
24101+ au_ftest_opts(opts->flags, REMOUNT));
24102+ if (!err) {
24103+ err = 1;
24104+ au_fset_opts(opts->flags, TRUNC_XIB);
027c5e7a 24105+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 24106+ }
24107+ break;
24108+
24109+ case Opt_mod:
24110+ case Opt_imod:
24111+ err = au_br_mod(sb, &opt->mod,
24112+ au_ftest_opts(opts->flags, REMOUNT),
24113+ &do_refresh);
24114+ if (!err) {
24115+ err = 1;
027c5e7a
AM
24116+ if (do_refresh)
24117+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 24118+ }
24119+ break;
24120+ }
24121+
24122+ return err;
24123+}
24124+
24125+static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
24126+ struct au_opt_xino **opt_xino,
24127+ struct au_opts *opts)
24128+{
24129+ int err;
24130+ aufs_bindex_t bend, bindex;
24131+ struct dentry *root, *parent, *h_root;
24132+
24133+ err = 0;
24134+ switch (opt->type) {
24135+ case Opt_xino:
24136+ err = au_xino_set(sb, &opt->xino,
24137+ !!au_ftest_opts(opts->flags, REMOUNT));
24138+ if (unlikely(err))
24139+ break;
24140+
24141+ *opt_xino = &opt->xino;
24142+ au_xino_brid_set(sb, -1);
24143+
24144+ /* safe d_parent access */
2000de60 24145+ parent = opt->xino.file->f_path.dentry->d_parent;
1facf9fc 24146+ root = sb->s_root;
24147+ bend = au_sbend(sb);
24148+ for (bindex = 0; bindex <= bend; bindex++) {
24149+ h_root = au_h_dptr(root, bindex);
24150+ if (h_root == parent) {
24151+ au_xino_brid_set(sb, au_sbr_id(sb, bindex));
24152+ break;
24153+ }
24154+ }
24155+ break;
24156+
24157+ case Opt_noxino:
24158+ au_xino_clr(sb);
24159+ au_xino_brid_set(sb, -1);
24160+ *opt_xino = (void *)-1;
24161+ break;
24162+ }
24163+
24164+ return err;
24165+}
24166+
24167+int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
24168+ unsigned int pending)
24169+{
076b876e 24170+ int err, fhsm;
1facf9fc 24171+ aufs_bindex_t bindex, bend;
24172+ unsigned char do_plink, skip, do_free;
24173+ struct au_branch *br;
24174+ struct au_wbr *wbr;
24175+ struct dentry *root;
24176+ struct inode *dir, *h_dir;
24177+ struct au_sbinfo *sbinfo;
24178+ struct au_hinode *hdir;
24179+
dece6358
AM
24180+ SiMustAnyLock(sb);
24181+
1facf9fc 24182+ sbinfo = au_sbi(sb);
24183+ AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
24184+
dece6358
AM
24185+ if (!(sb_flags & MS_RDONLY)) {
24186+ if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
0c3ec466 24187+ pr_warn("first branch should be rw\n");
dece6358 24188+ if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
0c3ec466 24189+ pr_warn("shwh should be used with ro\n");
dece6358 24190+ }
1facf9fc 24191+
4a4d8108 24192+ if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
1facf9fc 24193+ && !au_opt_test(sbinfo->si_mntflags, XINO))
0c3ec466 24194+ pr_warn("udba=*notify requires xino\n");
1facf9fc 24195+
076b876e
AM
24196+ if (au_opt_test(sbinfo->si_mntflags, DIRPERM1))
24197+ pr_warn("dirperm1 breaks the protection"
24198+ " by the permission bits on the lower branch\n");
24199+
1facf9fc 24200+ err = 0;
076b876e 24201+ fhsm = 0;
1facf9fc 24202+ root = sb->s_root;
5527c038 24203+ dir = d_inode(root);
1facf9fc 24204+ do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
24205+ bend = au_sbend(sb);
24206+ for (bindex = 0; !err && bindex <= bend; bindex++) {
24207+ skip = 0;
24208+ h_dir = au_h_iptr(dir, bindex);
24209+ br = au_sbr(sb, bindex);
1facf9fc 24210+
c1595e42
JR
24211+ if ((br->br_perm & AuBrAttr_ICEX)
24212+ && !h_dir->i_op->listxattr)
24213+ br->br_perm &= ~AuBrAttr_ICEX;
24214+#if 0
24215+ if ((br->br_perm & AuBrAttr_ICEX_SEC)
24216+ && (au_br_sb(br)->s_flags & MS_NOSEC))
24217+ br->br_perm &= ~AuBrAttr_ICEX_SEC;
24218+#endif
24219+
24220+ do_free = 0;
1facf9fc 24221+ wbr = br->br_wbr;
24222+ if (wbr)
24223+ wbr_wh_read_lock(wbr);
24224+
1e00d052 24225+ if (!au_br_writable(br->br_perm)) {
1facf9fc 24226+ do_free = !!wbr;
24227+ skip = (!wbr
24228+ || (!wbr->wbr_whbase
24229+ && !wbr->wbr_plink
24230+ && !wbr->wbr_orph));
1e00d052 24231+ } else if (!au_br_wh_linkable(br->br_perm)) {
1facf9fc 24232+ /* skip = (!br->br_whbase && !br->br_orph); */
24233+ skip = (!wbr || !wbr->wbr_whbase);
24234+ if (skip && wbr) {
24235+ if (do_plink)
24236+ skip = !!wbr->wbr_plink;
24237+ else
24238+ skip = !wbr->wbr_plink;
24239+ }
1e00d052 24240+ } else {
1facf9fc 24241+ /* skip = (br->br_whbase && br->br_ohph); */
24242+ skip = (wbr && wbr->wbr_whbase);
24243+ if (skip) {
24244+ if (do_plink)
24245+ skip = !!wbr->wbr_plink;
24246+ else
24247+ skip = !wbr->wbr_plink;
24248+ }
1facf9fc 24249+ }
24250+ if (wbr)
24251+ wbr_wh_read_unlock(wbr);
24252+
076b876e
AM
24253+ if (au_br_fhsm(br->br_perm)) {
24254+ fhsm++;
24255+ AuDebugOn(!br->br_fhsm);
24256+ }
24257+
1facf9fc 24258+ if (skip)
24259+ continue;
24260+
24261+ hdir = au_hi(dir, bindex);
4a4d8108 24262+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
1facf9fc 24263+ if (wbr)
24264+ wbr_wh_write_lock(wbr);
86dc4139 24265+ err = au_wh_init(br, sb);
1facf9fc 24266+ if (wbr)
24267+ wbr_wh_write_unlock(wbr);
4a4d8108 24268+ au_hn_imtx_unlock(hdir);
1facf9fc 24269+
24270+ if (!err && do_free) {
24271+ kfree(wbr);
24272+ br->br_wbr = NULL;
24273+ }
24274+ }
24275+
c1595e42 24276+ if (fhsm >= 2) {
076b876e 24277+ au_fset_si(sbinfo, FHSM);
c1595e42
JR
24278+ for (bindex = bend; bindex >= 0; bindex--) {
24279+ br = au_sbr(sb, bindex);
24280+ if (au_br_fhsm(br->br_perm)) {
24281+ au_fhsm_set_bottom(sb, bindex);
24282+ break;
24283+ }
24284+ }
24285+ } else {
076b876e 24286+ au_fclr_si(sbinfo, FHSM);
c1595e42
JR
24287+ au_fhsm_set_bottom(sb, -1);
24288+ }
076b876e 24289+
1facf9fc 24290+ return err;
24291+}
24292+
24293+int au_opts_mount(struct super_block *sb, struct au_opts *opts)
24294+{
24295+ int err;
24296+ unsigned int tmp;
027c5e7a 24297+ aufs_bindex_t bindex, bend;
1facf9fc 24298+ struct au_opt *opt;
24299+ struct au_opt_xino *opt_xino, xino;
24300+ struct au_sbinfo *sbinfo;
027c5e7a 24301+ struct au_branch *br;
076b876e 24302+ struct inode *dir;
1facf9fc 24303+
dece6358
AM
24304+ SiMustWriteLock(sb);
24305+
1facf9fc 24306+ err = 0;
24307+ opt_xino = NULL;
24308+ opt = opts->opt;
24309+ while (err >= 0 && opt->type != Opt_tail)
24310+ err = au_opt_simple(sb, opt++, opts);
24311+ if (err > 0)
24312+ err = 0;
24313+ else if (unlikely(err < 0))
24314+ goto out;
24315+
24316+ /* disable xino and udba temporary */
24317+ sbinfo = au_sbi(sb);
24318+ tmp = sbinfo->si_mntflags;
24319+ au_opt_clr(sbinfo->si_mntflags, XINO);
24320+ au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
24321+
24322+ opt = opts->opt;
24323+ while (err >= 0 && opt->type != Opt_tail)
24324+ err = au_opt_br(sb, opt++, opts);
24325+ if (err > 0)
24326+ err = 0;
24327+ else if (unlikely(err < 0))
24328+ goto out;
24329+
24330+ bend = au_sbend(sb);
24331+ if (unlikely(bend < 0)) {
24332+ err = -EINVAL;
4a4d8108 24333+ pr_err("no branches\n");
1facf9fc 24334+ goto out;
24335+ }
24336+
24337+ if (au_opt_test(tmp, XINO))
24338+ au_opt_set(sbinfo->si_mntflags, XINO);
24339+ opt = opts->opt;
24340+ while (!err && opt->type != Opt_tail)
24341+ err = au_opt_xino(sb, opt++, &opt_xino, opts);
24342+ if (unlikely(err))
24343+ goto out;
24344+
24345+ err = au_opts_verify(sb, sb->s_flags, tmp);
24346+ if (unlikely(err))
24347+ goto out;
24348+
24349+ /* restore xino */
24350+ if (au_opt_test(tmp, XINO) && !opt_xino) {
24351+ xino.file = au_xino_def(sb);
24352+ err = PTR_ERR(xino.file);
24353+ if (IS_ERR(xino.file))
24354+ goto out;
24355+
24356+ err = au_xino_set(sb, &xino, /*remount*/0);
24357+ fput(xino.file);
24358+ if (unlikely(err))
24359+ goto out;
24360+ }
24361+
24362+ /* restore udba */
027c5e7a 24363+ tmp &= AuOptMask_UDBA;
1facf9fc 24364+ sbinfo->si_mntflags &= ~AuOptMask_UDBA;
027c5e7a
AM
24365+ sbinfo->si_mntflags |= tmp;
24366+ bend = au_sbend(sb);
24367+ for (bindex = 0; bindex <= bend; bindex++) {
24368+ br = au_sbr(sb, bindex);
24369+ err = au_hnotify_reset_br(tmp, br, br->br_perm);
24370+ if (unlikely(err))
24371+ AuIOErr("hnotify failed on br %d, %d, ignored\n",
24372+ bindex, err);
24373+ /* go on even if err */
24374+ }
4a4d8108 24375+ if (au_opt_test(tmp, UDBA_HNOTIFY)) {
5527c038 24376+ dir = d_inode(sb->s_root);
4a4d8108 24377+ au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
1facf9fc 24378+ }
24379+
4f0767ce 24380+out:
1facf9fc 24381+ return err;
24382+}
24383+
24384+int au_opts_remount(struct super_block *sb, struct au_opts *opts)
24385+{
24386+ int err, rerr;
24387+ struct inode *dir;
24388+ struct au_opt_xino *opt_xino;
24389+ struct au_opt *opt;
24390+ struct au_sbinfo *sbinfo;
24391+
dece6358
AM
24392+ SiMustWriteLock(sb);
24393+
5527c038 24394+ dir = d_inode(sb->s_root);
1facf9fc 24395+ sbinfo = au_sbi(sb);
24396+ err = 0;
24397+ opt_xino = NULL;
24398+ opt = opts->opt;
24399+ while (err >= 0 && opt->type != Opt_tail) {
24400+ err = au_opt_simple(sb, opt, opts);
24401+ if (!err)
24402+ err = au_opt_br(sb, opt, opts);
24403+ if (!err)
24404+ err = au_opt_xino(sb, opt, &opt_xino, opts);
24405+ opt++;
24406+ }
24407+ if (err > 0)
24408+ err = 0;
24409+ AuTraceErr(err);
24410+ /* go on even err */
24411+
24412+ rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
24413+ if (unlikely(rerr && !err))
24414+ err = rerr;
24415+
24416+ if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
24417+ rerr = au_xib_trunc(sb);
24418+ if (unlikely(rerr && !err))
24419+ err = rerr;
24420+ }
24421+
24422+ /* will be handled by the caller */
027c5e7a 24423+ if (!au_ftest_opts(opts->flags, REFRESH)
1facf9fc 24424+ && (opts->given_udba || au_opt_test(sbinfo->si_mntflags, XINO)))
027c5e7a 24425+ au_fset_opts(opts->flags, REFRESH);
1facf9fc 24426+
24427+ AuDbg("status 0x%x\n", opts->flags);
24428+ return err;
24429+}
24430+
24431+/* ---------------------------------------------------------------------- */
24432+
24433+unsigned int au_opt_udba(struct super_block *sb)
24434+{
24435+ return au_mntflags(sb) & AuOptMask_UDBA;
24436+}
7f207e10
AM
24437diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
24438--- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 24439+++ linux/fs/aufs/opts.h 2015-09-24 10:47:58.254719746 +0200
5527c038 24440@@ -0,0 +1,210 @@
1facf9fc 24441+/*
2000de60 24442+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 24443+ *
24444+ * This program, aufs is free software; you can redistribute it and/or modify
24445+ * it under the terms of the GNU General Public License as published by
24446+ * the Free Software Foundation; either version 2 of the License, or
24447+ * (at your option) any later version.
dece6358
AM
24448+ *
24449+ * This program is distributed in the hope that it will be useful,
24450+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
24451+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24452+ * GNU General Public License for more details.
24453+ *
24454+ * You should have received a copy of the GNU General Public License
523b37e3 24455+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 24456+ */
24457+
24458+/*
24459+ * mount options/flags
24460+ */
24461+
24462+#ifndef __AUFS_OPTS_H__
24463+#define __AUFS_OPTS_H__
24464+
24465+#ifdef __KERNEL__
24466+
dece6358 24467+#include <linux/path.h>
1facf9fc 24468+
dece6358
AM
24469+struct file;
24470+struct super_block;
24471+
1facf9fc 24472+/* ---------------------------------------------------------------------- */
24473+
24474+/* mount flags */
24475+#define AuOpt_XINO 1 /* external inode number bitmap
24476+ and translation table */
24477+#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
24478+#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
24479+#define AuOpt_UDBA_REVAL (1 << 3)
4a4d8108 24480+#define AuOpt_UDBA_HNOTIFY (1 << 4)
dece6358
AM
24481+#define AuOpt_SHWH (1 << 5) /* show whiteout */
24482+#define AuOpt_PLINK (1 << 6) /* pseudo-link */
076b876e
AM
24483+#define AuOpt_DIRPERM1 (1 << 7) /* ignore the lower dir's perm
24484+ bits */
dece6358
AM
24485+#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
24486+#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
24487+#define AuOpt_SUM_W (1 << 11) /* unimplemented */
24488+#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
24489+#define AuOpt_VERBOSE (1 << 13) /* busy inode when del-branch */
4a4d8108 24490+#define AuOpt_DIO (1 << 14) /* direct io */
1facf9fc 24491+
4a4d8108
AM
24492+#ifndef CONFIG_AUFS_HNOTIFY
24493+#undef AuOpt_UDBA_HNOTIFY
24494+#define AuOpt_UDBA_HNOTIFY 0
1facf9fc 24495+#endif
dece6358
AM
24496+#ifndef CONFIG_AUFS_SHWH
24497+#undef AuOpt_SHWH
24498+#define AuOpt_SHWH 0
24499+#endif
1facf9fc 24500+
24501+#define AuOpt_Def (AuOpt_XINO \
24502+ | AuOpt_UDBA_REVAL \
24503+ | AuOpt_PLINK \
24504+ /* | AuOpt_DIRPERM1 */ \
24505+ | AuOpt_WARN_PERM)
24506+#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
24507+ | AuOpt_UDBA_REVAL \
4a4d8108 24508+ | AuOpt_UDBA_HNOTIFY)
1facf9fc 24509+
24510+#define au_opt_test(flags, name) (flags & AuOpt_##name)
24511+#define au_opt_set(flags, name) do { \
24512+ BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
24513+ ((flags) |= AuOpt_##name); \
24514+} while (0)
24515+#define au_opt_set_udba(flags, name) do { \
24516+ (flags) &= ~AuOptMask_UDBA; \
24517+ ((flags) |= AuOpt_##name); \
24518+} while (0)
7f207e10
AM
24519+#define au_opt_clr(flags, name) do { \
24520+ ((flags) &= ~AuOpt_##name); \
24521+} while (0)
1facf9fc 24522+
e49829fe
JR
24523+static inline unsigned int au_opts_plink(unsigned int mntflags)
24524+{
24525+#ifdef CONFIG_PROC_FS
24526+ return mntflags;
24527+#else
24528+ return mntflags & ~AuOpt_PLINK;
24529+#endif
24530+}
24531+
1facf9fc 24532+/* ---------------------------------------------------------------------- */
24533+
24534+/* policies to select one among multiple writable branches */
24535+enum {
24536+ AuWbrCreate_TDP, /* top down parent */
24537+ AuWbrCreate_RR, /* round robin */
24538+ AuWbrCreate_MFS, /* most free space */
24539+ AuWbrCreate_MFSV, /* mfs with seconds */
24540+ AuWbrCreate_MFSRR, /* mfs then rr */
24541+ AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
24542+ AuWbrCreate_PMFS, /* parent and mfs */
24543+ AuWbrCreate_PMFSV, /* parent and mfs with seconds */
392086de
AM
24544+ AuWbrCreate_PMFSRR, /* parent, mfs and round-robin */
24545+ AuWbrCreate_PMFSRRV, /* plus seconds */
1facf9fc 24546+
24547+ AuWbrCreate_Def = AuWbrCreate_TDP
24548+};
24549+
24550+enum {
24551+ AuWbrCopyup_TDP, /* top down parent */
24552+ AuWbrCopyup_BUP, /* bottom up parent */
24553+ AuWbrCopyup_BU, /* bottom up */
24554+
24555+ AuWbrCopyup_Def = AuWbrCopyup_TDP
24556+};
24557+
24558+/* ---------------------------------------------------------------------- */
24559+
24560+struct au_opt_add {
24561+ aufs_bindex_t bindex;
24562+ char *pathname;
24563+ int perm;
24564+ struct path path;
24565+};
24566+
24567+struct au_opt_del {
24568+ char *pathname;
24569+ struct path h_path;
24570+};
24571+
24572+struct au_opt_mod {
24573+ char *path;
24574+ int perm;
24575+ struct dentry *h_root;
24576+};
24577+
24578+struct au_opt_xino {
24579+ char *path;
24580+ struct file *file;
24581+};
24582+
24583+struct au_opt_xino_itrunc {
24584+ aufs_bindex_t bindex;
24585+};
24586+
24587+struct au_opt_wbr_create {
24588+ int wbr_create;
24589+ int mfs_second;
24590+ unsigned long long mfsrr_watermark;
24591+};
24592+
24593+struct au_opt {
24594+ int type;
24595+ union {
24596+ struct au_opt_xino xino;
24597+ struct au_opt_xino_itrunc xino_itrunc;
24598+ struct au_opt_add add;
24599+ struct au_opt_del del;
24600+ struct au_opt_mod mod;
24601+ int dirwh;
24602+ int rdcache;
24603+ unsigned int rdblk;
24604+ unsigned int rdhash;
24605+ int udba;
24606+ struct au_opt_wbr_create wbr_create;
24607+ int wbr_copyup;
076b876e 24608+ unsigned int fhsm_second;
1facf9fc 24609+ };
24610+};
24611+
24612+/* opts flags */
24613+#define AuOpts_REMOUNT 1
027c5e7a
AM
24614+#define AuOpts_REFRESH (1 << 1)
24615+#define AuOpts_TRUNC_XIB (1 << 2)
24616+#define AuOpts_REFRESH_DYAOP (1 << 3)
1facf9fc 24617+#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
7f207e10
AM
24618+#define au_fset_opts(flags, name) \
24619+ do { (flags) |= AuOpts_##name; } while (0)
24620+#define au_fclr_opts(flags, name) \
24621+ do { (flags) &= ~AuOpts_##name; } while (0)
1facf9fc 24622+
24623+struct au_opts {
24624+ struct au_opt *opt;
24625+ int max_opt;
24626+
24627+ unsigned int given_udba;
24628+ unsigned int flags;
24629+ unsigned long sb_flags;
24630+};
24631+
24632+/* ---------------------------------------------------------------------- */
24633+
7e9cd9fe 24634+/* opts.c */
076b876e 24635+void au_optstr_br_perm(au_br_perm_str_t *str, int perm);
1facf9fc 24636+const char *au_optstr_udba(int udba);
24637+const char *au_optstr_wbr_copyup(int wbr_copyup);
24638+const char *au_optstr_wbr_create(int wbr_create);
24639+
24640+void au_opts_free(struct au_opts *opts);
24641+int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts);
24642+int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
24643+ unsigned int pending);
24644+int au_opts_mount(struct super_block *sb, struct au_opts *opts);
24645+int au_opts_remount(struct super_block *sb, struct au_opts *opts);
24646+
24647+unsigned int au_opt_udba(struct super_block *sb);
24648+
1facf9fc 24649+#endif /* __KERNEL__ */
24650+#endif /* __AUFS_OPTS_H__ */
7f207e10
AM
24651diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
24652--- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 24653+++ linux/fs/aufs/plink.c 2015-09-24 10:47:58.254719746 +0200
5527c038 24654@@ -0,0 +1,528 @@
1facf9fc 24655+/*
2000de60 24656+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 24657+ *
24658+ * This program, aufs is free software; you can redistribute it and/or modify
24659+ * it under the terms of the GNU General Public License as published by
24660+ * the Free Software Foundation; either version 2 of the License, or
24661+ * (at your option) any later version.
dece6358
AM
24662+ *
24663+ * This program is distributed in the hope that it will be useful,
24664+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
24665+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24666+ * GNU General Public License for more details.
24667+ *
24668+ * You should have received a copy of the GNU General Public License
523b37e3 24669+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 24670+ */
24671+
24672+/*
24673+ * pseudo-link
24674+ */
24675+
24676+#include "aufs.h"
24677+
24678+/*
e49829fe 24679+ * the pseudo-link maintenance mode.
1facf9fc 24680+ * during a user process maintains the pseudo-links,
24681+ * prohibit adding a new plink and branch manipulation.
e49829fe
JR
24682+ *
24683+ * Flags
24684+ * NOPLM:
24685+ * For entry functions which will handle plink, and i_mutex is already held
24686+ * in VFS.
24687+ * They cannot wait and should return an error at once.
24688+ * Callers has to check the error.
24689+ * NOPLMW:
24690+ * For entry functions which will handle plink, but i_mutex is not held
24691+ * in VFS.
24692+ * They can wait the plink maintenance mode to finish.
24693+ *
24694+ * They behave like F_SETLK and F_SETLKW.
24695+ * If the caller never handle plink, then both flags are unnecessary.
1facf9fc 24696+ */
e49829fe
JR
24697+
24698+int au_plink_maint(struct super_block *sb, int flags)
1facf9fc 24699+{
e49829fe
JR
24700+ int err;
24701+ pid_t pid, ppid;
24702+ struct au_sbinfo *sbi;
dece6358
AM
24703+
24704+ SiMustAnyLock(sb);
24705+
e49829fe
JR
24706+ err = 0;
24707+ if (!au_opt_test(au_mntflags(sb), PLINK))
24708+ goto out;
24709+
24710+ sbi = au_sbi(sb);
24711+ pid = sbi->si_plink_maint_pid;
24712+ if (!pid || pid == current->pid)
24713+ goto out;
24714+
24715+ /* todo: it highly depends upon /sbin/mount.aufs */
24716+ rcu_read_lock();
24717+ ppid = task_pid_vnr(rcu_dereference(current->real_parent));
24718+ rcu_read_unlock();
24719+ if (pid == ppid)
24720+ goto out;
24721+
24722+ if (au_ftest_lock(flags, NOPLMW)) {
027c5e7a
AM
24723+ /* if there is no i_mutex lock in VFS, we don't need to wait */
24724+ /* AuDebugOn(!lockdep_depth(current)); */
e49829fe
JR
24725+ while (sbi->si_plink_maint_pid) {
24726+ si_read_unlock(sb);
24727+ /* gave up wake_up_bit() */
24728+ wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
24729+
24730+ if (au_ftest_lock(flags, FLUSH))
24731+ au_nwt_flush(&sbi->si_nowait);
24732+ si_noflush_read_lock(sb);
24733+ }
24734+ } else if (au_ftest_lock(flags, NOPLM)) {
24735+ AuDbg("ppid %d, pid %d\n", ppid, pid);
24736+ err = -EAGAIN;
24737+ }
24738+
24739+out:
24740+ return err;
4a4d8108
AM
24741+}
24742+
e49829fe 24743+void au_plink_maint_leave(struct au_sbinfo *sbinfo)
4a4d8108 24744+{
4a4d8108 24745+ spin_lock(&sbinfo->si_plink_maint_lock);
027c5e7a 24746+ sbinfo->si_plink_maint_pid = 0;
4a4d8108 24747+ spin_unlock(&sbinfo->si_plink_maint_lock);
027c5e7a 24748+ wake_up_all(&sbinfo->si_plink_wq);
4a4d8108
AM
24749+}
24750+
e49829fe 24751+int au_plink_maint_enter(struct super_block *sb)
4a4d8108
AM
24752+{
24753+ int err;
4a4d8108
AM
24754+ struct au_sbinfo *sbinfo;
24755+
24756+ err = 0;
4a4d8108
AM
24757+ sbinfo = au_sbi(sb);
24758+ /* make sure i am the only one in this fs */
e49829fe
JR
24759+ si_write_lock(sb, AuLock_FLUSH);
24760+ if (au_opt_test(au_mntflags(sb), PLINK)) {
24761+ spin_lock(&sbinfo->si_plink_maint_lock);
24762+ if (!sbinfo->si_plink_maint_pid)
24763+ sbinfo->si_plink_maint_pid = current->pid;
24764+ else
24765+ err = -EBUSY;
24766+ spin_unlock(&sbinfo->si_plink_maint_lock);
24767+ }
4a4d8108
AM
24768+ si_write_unlock(sb);
24769+
24770+ return err;
1facf9fc 24771+}
24772+
24773+/* ---------------------------------------------------------------------- */
24774+
1facf9fc 24775+#ifdef CONFIG_AUFS_DEBUG
24776+void au_plink_list(struct super_block *sb)
24777+{
86dc4139 24778+ int i;
1facf9fc 24779+ struct au_sbinfo *sbinfo;
86dc4139 24780+ struct hlist_head *plink_hlist;
1facf9fc 24781+ struct pseudo_link *plink;
24782+
dece6358
AM
24783+ SiMustAnyLock(sb);
24784+
1facf9fc 24785+ sbinfo = au_sbi(sb);
24786+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 24787+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 24788+
86dc4139
AM
24789+ for (i = 0; i < AuPlink_NHASH; i++) {
24790+ plink_hlist = &sbinfo->si_plink[i].head;
24791+ rcu_read_lock();
24792+ hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
24793+ AuDbg("%lu\n", plink->inode->i_ino);
24794+ rcu_read_unlock();
24795+ }
1facf9fc 24796+}
24797+#endif
24798+
24799+/* is the inode pseudo-linked? */
24800+int au_plink_test(struct inode *inode)
24801+{
86dc4139 24802+ int found, i;
1facf9fc 24803+ struct au_sbinfo *sbinfo;
86dc4139 24804+ struct hlist_head *plink_hlist;
1facf9fc 24805+ struct pseudo_link *plink;
24806+
24807+ sbinfo = au_sbi(inode->i_sb);
dece6358 24808+ AuRwMustAnyLock(&sbinfo->si_rwsem);
1facf9fc 24809+ AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
e49829fe 24810+ AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
1facf9fc 24811+
24812+ found = 0;
86dc4139
AM
24813+ i = au_plink_hash(inode->i_ino);
24814+ plink_hlist = &sbinfo->si_plink[i].head;
4a4d8108 24815+ rcu_read_lock();
86dc4139 24816+ hlist_for_each_entry_rcu(plink, plink_hlist, hlist)
1facf9fc 24817+ if (plink->inode == inode) {
24818+ found = 1;
24819+ break;
24820+ }
4a4d8108 24821+ rcu_read_unlock();
1facf9fc 24822+ return found;
24823+}
24824+
24825+/* ---------------------------------------------------------------------- */
24826+
24827+/*
24828+ * generate a name for plink.
24829+ * the file will be stored under AUFS_WH_PLINKDIR.
24830+ */
24831+/* 20 is max digits length of ulong 64 */
24832+#define PLINK_NAME_LEN ((20 + 1) * 2)
24833+
24834+static int plink_name(char *name, int len, struct inode *inode,
24835+ aufs_bindex_t bindex)
24836+{
24837+ int rlen;
24838+ struct inode *h_inode;
24839+
24840+ h_inode = au_h_iptr(inode, bindex);
24841+ rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
24842+ return rlen;
24843+}
24844+
7f207e10
AM
24845+struct au_do_plink_lkup_args {
24846+ struct dentry **errp;
24847+ struct qstr *tgtname;
24848+ struct dentry *h_parent;
24849+ struct au_branch *br;
24850+};
24851+
24852+static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
24853+ struct dentry *h_parent,
24854+ struct au_branch *br)
24855+{
24856+ struct dentry *h_dentry;
24857+ struct mutex *h_mtx;
24858+
5527c038 24859+ h_mtx = &d_inode(h_parent)->i_mutex;
7f207e10 24860+ mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
b4510431 24861+ h_dentry = vfsub_lkup_one(tgtname, h_parent);
7f207e10
AM
24862+ mutex_unlock(h_mtx);
24863+ return h_dentry;
24864+}
24865+
24866+static void au_call_do_plink_lkup(void *args)
24867+{
24868+ struct au_do_plink_lkup_args *a = args;
24869+ *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br);
24870+}
24871+
1facf9fc 24872+/* lookup the plink-ed @inode under the branch at @bindex */
24873+struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
24874+{
24875+ struct dentry *h_dentry, *h_parent;
24876+ struct au_branch *br;
7f207e10 24877+ int wkq_err;
1facf9fc 24878+ char a[PLINK_NAME_LEN];
0c3ec466 24879+ struct qstr tgtname = QSTR_INIT(a, 0);
1facf9fc 24880+
e49829fe
JR
24881+ AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
24882+
1facf9fc 24883+ br = au_sbr(inode->i_sb, bindex);
24884+ h_parent = br->br_wbr->wbr_plink;
1facf9fc 24885+ tgtname.len = plink_name(a, sizeof(a), inode, bindex);
24886+
2dfbb274 24887+ if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
7f207e10
AM
24888+ struct au_do_plink_lkup_args args = {
24889+ .errp = &h_dentry,
24890+ .tgtname = &tgtname,
24891+ .h_parent = h_parent,
24892+ .br = br
24893+ };
24894+
24895+ wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
24896+ if (unlikely(wkq_err))
24897+ h_dentry = ERR_PTR(wkq_err);
24898+ } else
24899+ h_dentry = au_do_plink_lkup(&tgtname, h_parent, br);
24900+
1facf9fc 24901+ return h_dentry;
24902+}
24903+
24904+/* create a pseudo-link */
24905+static int do_whplink(struct qstr *tgt, struct dentry *h_parent,
24906+ struct dentry *h_dentry, struct au_branch *br)
24907+{
24908+ int err;
24909+ struct path h_path = {
86dc4139 24910+ .mnt = au_br_mnt(br)
1facf9fc 24911+ };
523b37e3 24912+ struct inode *h_dir, *delegated;
1facf9fc 24913+
5527c038 24914+ h_dir = d_inode(h_parent);
7f207e10 24915+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_CHILD2);
4f0767ce 24916+again:
b4510431 24917+ h_path.dentry = vfsub_lkup_one(tgt, h_parent);
1facf9fc 24918+ err = PTR_ERR(h_path.dentry);
24919+ if (IS_ERR(h_path.dentry))
24920+ goto out;
24921+
24922+ err = 0;
24923+ /* wh.plink dir is not monitored */
7f207e10 24924+ /* todo: is it really safe? */
5527c038
JR
24925+ if (d_is_positive(h_path.dentry)
24926+ && d_inode(h_path.dentry) != d_inode(h_dentry)) {
523b37e3
AM
24927+ delegated = NULL;
24928+ err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0);
24929+ if (unlikely(err == -EWOULDBLOCK)) {
24930+ pr_warn("cannot retry for NFSv4 delegation"
24931+ " for an internal unlink\n");
24932+ iput(delegated);
24933+ }
1facf9fc 24934+ dput(h_path.dentry);
24935+ h_path.dentry = NULL;
24936+ if (!err)
24937+ goto again;
24938+ }
5527c038 24939+ if (!err && d_is_negative(h_path.dentry)) {
523b37e3
AM
24940+ delegated = NULL;
24941+ err = vfsub_link(h_dentry, h_dir, &h_path, &delegated);
24942+ if (unlikely(err == -EWOULDBLOCK)) {
24943+ pr_warn("cannot retry for NFSv4 delegation"
24944+ " for an internal link\n");
24945+ iput(delegated);
24946+ }
24947+ }
1facf9fc 24948+ dput(h_path.dentry);
24949+
4f0767ce 24950+out:
7f207e10 24951+ mutex_unlock(&h_dir->i_mutex);
1facf9fc 24952+ return err;
24953+}
24954+
24955+struct do_whplink_args {
24956+ int *errp;
24957+ struct qstr *tgt;
24958+ struct dentry *h_parent;
24959+ struct dentry *h_dentry;
24960+ struct au_branch *br;
24961+};
24962+
24963+static void call_do_whplink(void *args)
24964+{
24965+ struct do_whplink_args *a = args;
24966+ *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br);
24967+}
24968+
24969+static int whplink(struct dentry *h_dentry, struct inode *inode,
24970+ aufs_bindex_t bindex, struct au_branch *br)
24971+{
24972+ int err, wkq_err;
24973+ struct au_wbr *wbr;
24974+ struct dentry *h_parent;
1facf9fc 24975+ char a[PLINK_NAME_LEN];
0c3ec466 24976+ struct qstr tgtname = QSTR_INIT(a, 0);
1facf9fc 24977+
24978+ wbr = au_sbr(inode->i_sb, bindex)->br_wbr;
24979+ h_parent = wbr->wbr_plink;
1facf9fc 24980+ tgtname.len = plink_name(a, sizeof(a), inode, bindex);
24981+
24982+ /* always superio. */
2dfbb274 24983+ if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) {
1facf9fc 24984+ struct do_whplink_args args = {
24985+ .errp = &err,
24986+ .tgt = &tgtname,
24987+ .h_parent = h_parent,
24988+ .h_dentry = h_dentry,
24989+ .br = br
24990+ };
24991+ wkq_err = au_wkq_wait(call_do_whplink, &args);
24992+ if (unlikely(wkq_err))
24993+ err = wkq_err;
24994+ } else
24995+ err = do_whplink(&tgtname, h_parent, h_dentry, br);
1facf9fc 24996+
24997+ return err;
24998+}
24999+
25000+/* free a single plink */
25001+static void do_put_plink(struct pseudo_link *plink, int do_del)
25002+{
1facf9fc 25003+ if (do_del)
86dc4139 25004+ hlist_del(&plink->hlist);
4a4d8108
AM
25005+ iput(plink->inode);
25006+ kfree(plink);
25007+}
25008+
25009+static void do_put_plink_rcu(struct rcu_head *rcu)
25010+{
25011+ struct pseudo_link *plink;
25012+
25013+ plink = container_of(rcu, struct pseudo_link, rcu);
25014+ iput(plink->inode);
1facf9fc 25015+ kfree(plink);
25016+}
25017+
25018+/*
25019+ * create a new pseudo-link for @h_dentry on @bindex.
25020+ * the linked inode is held in aufs @inode.
25021+ */
25022+void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
25023+ struct dentry *h_dentry)
25024+{
25025+ struct super_block *sb;
25026+ struct au_sbinfo *sbinfo;
86dc4139 25027+ struct hlist_head *plink_hlist;
4a4d8108 25028+ struct pseudo_link *plink, *tmp;
86dc4139
AM
25029+ struct au_sphlhead *sphl;
25030+ int found, err, cnt, i;
1facf9fc 25031+
25032+ sb = inode->i_sb;
25033+ sbinfo = au_sbi(sb);
25034+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 25035+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 25036+
86dc4139 25037+ found = au_plink_test(inode);
4a4d8108 25038+ if (found)
1facf9fc 25039+ return;
4a4d8108 25040+
86dc4139
AM
25041+ i = au_plink_hash(inode->i_ino);
25042+ sphl = sbinfo->si_plink + i;
25043+ plink_hlist = &sphl->head;
4a4d8108
AM
25044+ tmp = kmalloc(sizeof(*plink), GFP_NOFS);
25045+ if (tmp)
25046+ tmp->inode = au_igrab(inode);
25047+ else {
25048+ err = -ENOMEM;
25049+ goto out;
1facf9fc 25050+ }
25051+
86dc4139
AM
25052+ spin_lock(&sphl->spin);
25053+ hlist_for_each_entry(plink, plink_hlist, hlist) {
4a4d8108
AM
25054+ if (plink->inode == inode) {
25055+ found = 1;
25056+ break;
25057+ }
1facf9fc 25058+ }
4a4d8108 25059+ if (!found)
86dc4139
AM
25060+ hlist_add_head_rcu(&tmp->hlist, plink_hlist);
25061+ spin_unlock(&sphl->spin);
4a4d8108 25062+ if (!found) {
86dc4139
AM
25063+ cnt = au_sphl_count(sphl);
25064+#define msg "unexpectedly unblanced or too many pseudo-links"
25065+ if (cnt > AUFS_PLINK_WARN)
25066+ AuWarn1(msg ", %d\n", cnt);
25067+#undef msg
1facf9fc 25068+ err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex));
4a4d8108
AM
25069+ } else {
25070+ do_put_plink(tmp, 0);
25071+ return;
1facf9fc 25072+ }
25073+
4a4d8108 25074+out:
1facf9fc 25075+ if (unlikely(err)) {
0c3ec466 25076+ pr_warn("err %d, damaged pseudo link.\n", err);
4a4d8108 25077+ if (tmp) {
86dc4139 25078+ au_sphl_del_rcu(&tmp->hlist, sphl);
4a4d8108
AM
25079+ call_rcu(&tmp->rcu, do_put_plink_rcu);
25080+ }
1facf9fc 25081+ }
25082+}
25083+
25084+/* free all plinks */
e49829fe 25085+void au_plink_put(struct super_block *sb, int verbose)
1facf9fc 25086+{
86dc4139 25087+ int i, warned;
1facf9fc 25088+ struct au_sbinfo *sbinfo;
86dc4139
AM
25089+ struct hlist_head *plink_hlist;
25090+ struct hlist_node *tmp;
25091+ struct pseudo_link *plink;
1facf9fc 25092+
dece6358
AM
25093+ SiMustWriteLock(sb);
25094+
1facf9fc 25095+ sbinfo = au_sbi(sb);
25096+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 25097+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 25098+
1facf9fc 25099+ /* no spin_lock since sbinfo is write-locked */
86dc4139
AM
25100+ warned = 0;
25101+ for (i = 0; i < AuPlink_NHASH; i++) {
25102+ plink_hlist = &sbinfo->si_plink[i].head;
25103+ if (!warned && verbose && !hlist_empty(plink_hlist)) {
25104+ pr_warn("pseudo-link is not flushed");
25105+ warned = 1;
25106+ }
25107+ hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist)
25108+ do_put_plink(plink, 0);
25109+ INIT_HLIST_HEAD(plink_hlist);
25110+ }
1facf9fc 25111+}
25112+
e49829fe
JR
25113+void au_plink_clean(struct super_block *sb, int verbose)
25114+{
25115+ struct dentry *root;
25116+
25117+ root = sb->s_root;
25118+ aufs_write_lock(root);
25119+ if (au_opt_test(au_mntflags(sb), PLINK))
25120+ au_plink_put(sb, verbose);
25121+ aufs_write_unlock(root);
25122+}
25123+
86dc4139
AM
25124+static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id)
25125+{
25126+ int do_put;
25127+ aufs_bindex_t bstart, bend, bindex;
25128+
25129+ do_put = 0;
25130+ bstart = au_ibstart(inode);
25131+ bend = au_ibend(inode);
25132+ if (bstart >= 0) {
25133+ for (bindex = bstart; bindex <= bend; bindex++) {
25134+ if (!au_h_iptr(inode, bindex)
25135+ || au_ii_br_id(inode, bindex) != br_id)
25136+ continue;
25137+ au_set_h_iptr(inode, bindex, NULL, 0);
25138+ do_put = 1;
25139+ break;
25140+ }
25141+ if (do_put)
25142+ for (bindex = bstart; bindex <= bend; bindex++)
25143+ if (au_h_iptr(inode, bindex)) {
25144+ do_put = 0;
25145+ break;
25146+ }
25147+ } else
25148+ do_put = 1;
25149+
25150+ return do_put;
25151+}
25152+
1facf9fc 25153+/* free the plinks on a branch specified by @br_id */
25154+void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
25155+{
25156+ struct au_sbinfo *sbinfo;
86dc4139
AM
25157+ struct hlist_head *plink_hlist;
25158+ struct hlist_node *tmp;
25159+ struct pseudo_link *plink;
1facf9fc 25160+ struct inode *inode;
86dc4139 25161+ int i, do_put;
1facf9fc 25162+
dece6358
AM
25163+ SiMustWriteLock(sb);
25164+
1facf9fc 25165+ sbinfo = au_sbi(sb);
25166+ AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
e49829fe 25167+ AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
1facf9fc 25168+
1facf9fc 25169+ /* no spin_lock since sbinfo is write-locked */
86dc4139
AM
25170+ for (i = 0; i < AuPlink_NHASH; i++) {
25171+ plink_hlist = &sbinfo->si_plink[i].head;
25172+ hlist_for_each_entry_safe(plink, tmp, plink_hlist, hlist) {
25173+ inode = au_igrab(plink->inode);
25174+ ii_write_lock_child(inode);
25175+ do_put = au_plink_do_half_refresh(inode, br_id);
dece6358
AM
25176+ if (do_put)
25177+ do_put_plink(plink, 1);
86dc4139
AM
25178+ ii_write_unlock(inode);
25179+ iput(inode);
dece6358 25180+ }
dece6358
AM
25181+ }
25182+}
7f207e10
AM
25183diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
25184--- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 25185+++ linux/fs/aufs/poll.c 2015-09-24 10:47:58.254719746 +0200
b912730e 25186@@ -0,0 +1,52 @@
dece6358 25187+/*
2000de60 25188+ * Copyright (C) 2005-2015 Junjiro R. Okajima
dece6358
AM
25189+ *
25190+ * This program, aufs is free software; you can redistribute it and/or modify
25191+ * it under the terms of the GNU General Public License as published by
25192+ * the Free Software Foundation; either version 2 of the License, or
25193+ * (at your option) any later version.
25194+ *
25195+ * This program is distributed in the hope that it will be useful,
25196+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
25197+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25198+ * GNU General Public License for more details.
25199+ *
25200+ * You should have received a copy of the GNU General Public License
523b37e3 25201+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
dece6358
AM
25202+ */
25203+
1308ab2a 25204+/*
25205+ * poll operation
25206+ * There is only one filesystem which implements ->poll operation, currently.
25207+ */
25208+
25209+#include "aufs.h"
25210+
25211+unsigned int aufs_poll(struct file *file, poll_table *wait)
25212+{
25213+ unsigned int mask;
25214+ int err;
25215+ struct file *h_file;
1308ab2a 25216+ struct super_block *sb;
25217+
25218+ /* We should pretend an error happened. */
25219+ mask = POLLERR /* | POLLIN | POLLOUT */;
b912730e 25220+ sb = file->f_path.dentry->d_sb;
e49829fe 25221+ si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
b912730e
AM
25222+
25223+ h_file = au_read_pre(file, /*keep_fi*/0);
25224+ err = PTR_ERR(h_file);
25225+ if (IS_ERR(h_file))
1308ab2a 25226+ goto out;
25227+
25228+ /* it is not an error if h_file has no operation */
25229+ mask = DEFAULT_POLLMASK;
523b37e3 25230+ if (h_file->f_op->poll)
1308ab2a 25231+ mask = h_file->f_op->poll(h_file, wait);
b912730e 25232+ fput(h_file); /* instead of au_read_post() */
1308ab2a 25233+
4f0767ce 25234+out:
1308ab2a 25235+ si_read_unlock(sb);
25236+ AuTraceErr((int)mask);
25237+ return mask;
25238+}
c1595e42
JR
25239diff -urN /usr/share/empty/fs/aufs/posix_acl.c linux/fs/aufs/posix_acl.c
25240--- /usr/share/empty/fs/aufs/posix_acl.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 25241+++ linux/fs/aufs/posix_acl.c 2015-09-24 10:47:58.254719746 +0200
c1595e42
JR
25242@@ -0,0 +1,99 @@
25243+/*
2000de60 25244+ * Copyright (C) 2014-2015 Junjiro R. Okajima
c1595e42
JR
25245+ *
25246+ * This program, aufs is free software; you can redistribute it and/or modify
25247+ * it under the terms of the GNU General Public License as published by
25248+ * the Free Software Foundation; either version 2 of the License, or
25249+ * (at your option) any later version.
25250+ *
25251+ * This program is distributed in the hope that it will be useful,
25252+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
25253+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25254+ * GNU General Public License for more details.
25255+ *
25256+ * You should have received a copy of the GNU General Public License
25257+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
25258+ */
25259+
25260+/*
25261+ * posix acl operations
25262+ */
25263+
25264+#include <linux/fs.h>
25265+#include <linux/posix_acl.h>
25266+#include "aufs.h"
25267+
25268+struct posix_acl *aufs_get_acl(struct inode *inode, int type)
25269+{
25270+ struct posix_acl *acl;
25271+ int err;
25272+ aufs_bindex_t bindex;
25273+ struct inode *h_inode;
25274+ struct super_block *sb;
25275+
25276+ acl = NULL;
25277+ sb = inode->i_sb;
25278+ si_read_lock(sb, AuLock_FLUSH);
25279+ ii_read_lock_child(inode);
25280+ if (!(sb->s_flags & MS_POSIXACL))
25281+ goto out;
25282+
25283+ bindex = au_ibstart(inode);
25284+ h_inode = au_h_iptr(inode, bindex);
25285+ if (unlikely(!h_inode
25286+ || ((h_inode->i_mode & S_IFMT)
25287+ != (inode->i_mode & S_IFMT)))) {
25288+ err = au_busy_or_stale();
25289+ acl = ERR_PTR(err);
25290+ goto out;
25291+ }
25292+
25293+ /* always topmost only */
25294+ acl = get_acl(h_inode, type);
25295+
25296+out:
25297+ ii_read_unlock(inode);
25298+ si_read_unlock(sb);
25299+
25300+ AuTraceErrPtr(acl);
25301+ return acl;
25302+}
25303+
25304+int aufs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
25305+{
25306+ int err;
25307+ ssize_t ssz;
25308+ struct dentry *dentry;
25309+ struct au_srxattr arg = {
25310+ .type = AU_ACL_SET,
25311+ .u.acl_set = {
25312+ .acl = acl,
25313+ .type = type
25314+ },
25315+ };
25316+
25317+ mutex_lock(&inode->i_mutex);
25318+ if (inode->i_ino == AUFS_ROOT_INO)
25319+ dentry = dget(inode->i_sb->s_root);
25320+ else {
25321+ dentry = d_find_alias(inode);
25322+ if (!dentry)
25323+ dentry = d_find_any_alias(inode);
25324+ if (!dentry) {
25325+ pr_warn("cannot handle this inode, "
25326+ "please report to aufs-users ML\n");
25327+ err = -ENOENT;
25328+ goto out;
25329+ }
25330+ }
25331+
25332+ ssz = au_srxattr(dentry, &arg);
25333+ dput(dentry);
25334+ err = ssz;
25335+ if (ssz >= 0)
25336+ err = 0;
25337+
25338+out:
25339+ mutex_unlock(&inode->i_mutex);
25340+ return err;
25341+}
7f207e10
AM
25342diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
25343--- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 25344+++ linux/fs/aufs/procfs.c 2015-09-24 10:47:58.254719746 +0200
523b37e3 25345@@ -0,0 +1,169 @@
e49829fe 25346+/*
2000de60 25347+ * Copyright (C) 2010-2015 Junjiro R. Okajima
e49829fe
JR
25348+ *
25349+ * This program, aufs is free software; you can redistribute it and/or modify
25350+ * it under the terms of the GNU General Public License as published by
25351+ * the Free Software Foundation; either version 2 of the License, or
25352+ * (at your option) any later version.
25353+ *
25354+ * This program is distributed in the hope that it will be useful,
25355+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
25356+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25357+ * GNU General Public License for more details.
25358+ *
25359+ * You should have received a copy of the GNU General Public License
523b37e3 25360+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
e49829fe
JR
25361+ */
25362+
25363+/*
25364+ * procfs interfaces
25365+ */
25366+
25367+#include <linux/proc_fs.h>
25368+#include "aufs.h"
25369+
25370+static int au_procfs_plm_release(struct inode *inode, struct file *file)
25371+{
25372+ struct au_sbinfo *sbinfo;
25373+
25374+ sbinfo = file->private_data;
25375+ if (sbinfo) {
25376+ au_plink_maint_leave(sbinfo);
25377+ kobject_put(&sbinfo->si_kobj);
25378+ }
25379+
25380+ return 0;
25381+}
25382+
25383+static void au_procfs_plm_write_clean(struct file *file)
25384+{
25385+ struct au_sbinfo *sbinfo;
25386+
25387+ sbinfo = file->private_data;
25388+ if (sbinfo)
25389+ au_plink_clean(sbinfo->si_sb, /*verbose*/0);
25390+}
25391+
25392+static int au_procfs_plm_write_si(struct file *file, unsigned long id)
25393+{
25394+ int err;
25395+ struct super_block *sb;
25396+ struct au_sbinfo *sbinfo;
25397+
25398+ err = -EBUSY;
25399+ if (unlikely(file->private_data))
25400+ goto out;
25401+
25402+ sb = NULL;
53392da6 25403+ /* don't use au_sbilist_lock() here */
e49829fe
JR
25404+ spin_lock(&au_sbilist.spin);
25405+ list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
25406+ if (id == sysaufs_si_id(sbinfo)) {
25407+ kobject_get(&sbinfo->si_kobj);
25408+ sb = sbinfo->si_sb;
25409+ break;
25410+ }
25411+ spin_unlock(&au_sbilist.spin);
25412+
25413+ err = -EINVAL;
25414+ if (unlikely(!sb))
25415+ goto out;
25416+
25417+ err = au_plink_maint_enter(sb);
25418+ if (!err)
25419+ /* keep kobject_get() */
25420+ file->private_data = sbinfo;
25421+ else
25422+ kobject_put(&sbinfo->si_kobj);
25423+out:
25424+ return err;
25425+}
25426+
25427+/*
25428+ * Accept a valid "si=xxxx" only.
25429+ * Once it is accepted successfully, accept "clean" too.
25430+ */
25431+static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
25432+ size_t count, loff_t *ppos)
25433+{
25434+ ssize_t err;
25435+ unsigned long id;
25436+ /* last newline is allowed */
25437+ char buf[3 + sizeof(unsigned long) * 2 + 1];
25438+
25439+ err = -EACCES;
25440+ if (unlikely(!capable(CAP_SYS_ADMIN)))
25441+ goto out;
25442+
25443+ err = -EINVAL;
25444+ if (unlikely(count > sizeof(buf)))
25445+ goto out;
25446+
25447+ err = copy_from_user(buf, ubuf, count);
25448+ if (unlikely(err)) {
25449+ err = -EFAULT;
25450+ goto out;
25451+ }
25452+ buf[count] = 0;
25453+
25454+ err = -EINVAL;
25455+ if (!strcmp("clean", buf)) {
25456+ au_procfs_plm_write_clean(file);
25457+ goto out_success;
25458+ } else if (unlikely(strncmp("si=", buf, 3)))
25459+ goto out;
25460+
9dbd164d 25461+ err = kstrtoul(buf + 3, 16, &id);
e49829fe
JR
25462+ if (unlikely(err))
25463+ goto out;
25464+
25465+ err = au_procfs_plm_write_si(file, id);
25466+ if (unlikely(err))
25467+ goto out;
25468+
25469+out_success:
25470+ err = count; /* success */
25471+out:
25472+ return err;
25473+}
25474+
25475+static const struct file_operations au_procfs_plm_fop = {
25476+ .write = au_procfs_plm_write,
25477+ .release = au_procfs_plm_release,
25478+ .owner = THIS_MODULE
25479+};
25480+
25481+/* ---------------------------------------------------------------------- */
25482+
25483+static struct proc_dir_entry *au_procfs_dir;
25484+
25485+void au_procfs_fin(void)
25486+{
25487+ remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
25488+ remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
25489+}
25490+
25491+int __init au_procfs_init(void)
25492+{
25493+ int err;
25494+ struct proc_dir_entry *entry;
25495+
25496+ err = -ENOMEM;
25497+ au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
25498+ if (unlikely(!au_procfs_dir))
25499+ goto out;
25500+
25501+ entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | S_IWUSR,
25502+ au_procfs_dir, &au_procfs_plm_fop);
25503+ if (unlikely(!entry))
25504+ goto out_dir;
25505+
25506+ err = 0;
25507+ goto out; /* success */
25508+
25509+
25510+out_dir:
25511+ remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
25512+out:
25513+ return err;
25514+}
7f207e10
AM
25515diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
25516--- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 25517+++ linux/fs/aufs/rdu.c 2015-09-24 10:47:58.254719746 +0200
523b37e3 25518@@ -0,0 +1,388 @@
1308ab2a 25519+/*
2000de60 25520+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1308ab2a 25521+ *
25522+ * This program, aufs is free software; you can redistribute it and/or modify
25523+ * it under the terms of the GNU General Public License as published by
25524+ * the Free Software Foundation; either version 2 of the License, or
25525+ * (at your option) any later version.
25526+ *
25527+ * This program is distributed in the hope that it will be useful,
25528+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
25529+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25530+ * GNU General Public License for more details.
25531+ *
25532+ * You should have received a copy of the GNU General Public License
523b37e3 25533+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1308ab2a 25534+ */
25535+
25536+/*
25537+ * readdir in userspace.
25538+ */
25539+
b752ccd1 25540+#include <linux/compat.h>
4a4d8108 25541+#include <linux/fs_stack.h>
1308ab2a 25542+#include <linux/security.h>
1308ab2a 25543+#include "aufs.h"
25544+
25545+/* bits for struct aufs_rdu.flags */
25546+#define AuRdu_CALLED 1
25547+#define AuRdu_CONT (1 << 1)
25548+#define AuRdu_FULL (1 << 2)
25549+#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
7f207e10
AM
25550+#define au_fset_rdu(flags, name) \
25551+ do { (flags) |= AuRdu_##name; } while (0)
25552+#define au_fclr_rdu(flags, name) \
25553+ do { (flags) &= ~AuRdu_##name; } while (0)
1308ab2a 25554+
25555+struct au_rdu_arg {
392086de 25556+ struct dir_context ctx;
1308ab2a 25557+ struct aufs_rdu *rdu;
25558+ union au_rdu_ent_ul ent;
25559+ unsigned long end;
25560+
25561+ struct super_block *sb;
25562+ int err;
25563+};
25564+
392086de 25565+static int au_rdu_fill(struct dir_context *ctx, const char *name, int nlen,
1308ab2a 25566+ loff_t offset, u64 h_ino, unsigned int d_type)
25567+{
25568+ int err, len;
392086de 25569+ struct au_rdu_arg *arg = container_of(ctx, struct au_rdu_arg, ctx);
1308ab2a 25570+ struct aufs_rdu *rdu = arg->rdu;
25571+ struct au_rdu_ent ent;
25572+
25573+ err = 0;
25574+ arg->err = 0;
25575+ au_fset_rdu(rdu->cookie.flags, CALLED);
25576+ len = au_rdu_len(nlen);
25577+ if (arg->ent.ul + len < arg->end) {
25578+ ent.ino = h_ino;
25579+ ent.bindex = rdu->cookie.bindex;
25580+ ent.type = d_type;
25581+ ent.nlen = nlen;
4a4d8108
AM
25582+ if (unlikely(nlen > AUFS_MAX_NAMELEN))
25583+ ent.type = DT_UNKNOWN;
1308ab2a 25584+
9dbd164d 25585+ /* unnecessary to support mmap_sem since this is a dir */
1308ab2a 25586+ err = -EFAULT;
25587+ if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
25588+ goto out;
25589+ if (copy_to_user(arg->ent.e->name, name, nlen))
25590+ goto out;
25591+ /* the terminating NULL */
25592+ if (__put_user(0, arg->ent.e->name + nlen))
25593+ goto out;
25594+ err = 0;
25595+ /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
25596+ arg->ent.ul += len;
25597+ rdu->rent++;
25598+ } else {
25599+ err = -EFAULT;
25600+ au_fset_rdu(rdu->cookie.flags, FULL);
25601+ rdu->full = 1;
25602+ rdu->tail = arg->ent;
25603+ }
25604+
4f0767ce 25605+out:
1308ab2a 25606+ /* AuTraceErr(err); */
25607+ return err;
25608+}
25609+
25610+static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
25611+{
25612+ int err;
25613+ loff_t offset;
25614+ struct au_rdu_cookie *cookie = &arg->rdu->cookie;
25615+
92d182d2 25616+ /* we don't have to care (FMODE_32BITHASH | FMODE_64BITHASH) for ext4 */
1308ab2a 25617+ offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
25618+ err = offset;
25619+ if (unlikely(offset != cookie->h_pos))
25620+ goto out;
25621+
25622+ err = 0;
25623+ do {
25624+ arg->err = 0;
25625+ au_fclr_rdu(cookie->flags, CALLED);
25626+ /* smp_mb(); */
392086de 25627+ err = vfsub_iterate_dir(h_file, &arg->ctx);
1308ab2a 25628+ if (err >= 0)
25629+ err = arg->err;
25630+ } while (!err
25631+ && au_ftest_rdu(cookie->flags, CALLED)
25632+ && !au_ftest_rdu(cookie->flags, FULL));
25633+ cookie->h_pos = h_file->f_pos;
25634+
4f0767ce 25635+out:
1308ab2a 25636+ AuTraceErr(err);
25637+ return err;
25638+}
25639+
25640+static int au_rdu(struct file *file, struct aufs_rdu *rdu)
25641+{
25642+ int err;
25643+ aufs_bindex_t bend;
392086de
AM
25644+ struct au_rdu_arg arg = {
25645+ .ctx = {
2000de60 25646+ .actor = au_rdu_fill
392086de
AM
25647+ }
25648+ };
1308ab2a 25649+ struct dentry *dentry;
25650+ struct inode *inode;
25651+ struct file *h_file;
25652+ struct au_rdu_cookie *cookie = &rdu->cookie;
25653+
25654+ err = !access_ok(VERIFY_WRITE, rdu->ent.e, rdu->sz);
25655+ if (unlikely(err)) {
25656+ err = -EFAULT;
25657+ AuTraceErr(err);
25658+ goto out;
25659+ }
25660+ rdu->rent = 0;
25661+ rdu->tail = rdu->ent;
25662+ rdu->full = 0;
25663+ arg.rdu = rdu;
25664+ arg.ent = rdu->ent;
25665+ arg.end = arg.ent.ul;
25666+ arg.end += rdu->sz;
25667+
25668+ err = -ENOTDIR;
523b37e3 25669+ if (unlikely(!file->f_op->iterate))
1308ab2a 25670+ goto out;
25671+
25672+ err = security_file_permission(file, MAY_READ);
25673+ AuTraceErr(err);
25674+ if (unlikely(err))
25675+ goto out;
25676+
2000de60 25677+ dentry = file->f_path.dentry;
5527c038 25678+ inode = d_inode(dentry);
1308ab2a 25679+#if 1
25680+ mutex_lock(&inode->i_mutex);
25681+#else
25682+ err = mutex_lock_killable(&inode->i_mutex);
25683+ AuTraceErr(err);
25684+ if (unlikely(err))
25685+ goto out;
25686+#endif
1308ab2a 25687+
25688+ arg.sb = inode->i_sb;
e49829fe
JR
25689+ err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
25690+ if (unlikely(err))
25691+ goto out_mtx;
027c5e7a
AM
25692+ err = au_alive_dir(dentry);
25693+ if (unlikely(err))
25694+ goto out_si;
e49829fe 25695+ /* todo: reval? */
1308ab2a 25696+ fi_read_lock(file);
25697+
25698+ err = -EAGAIN;
25699+ if (unlikely(au_ftest_rdu(cookie->flags, CONT)
25700+ && cookie->generation != au_figen(file)))
25701+ goto out_unlock;
25702+
25703+ err = 0;
25704+ if (!rdu->blk) {
25705+ rdu->blk = au_sbi(arg.sb)->si_rdblk;
25706+ if (!rdu->blk)
25707+ rdu->blk = au_dir_size(file, /*dentry*/NULL);
25708+ }
25709+ bend = au_fbstart(file);
25710+ if (cookie->bindex < bend)
25711+ cookie->bindex = bend;
4a4d8108 25712+ bend = au_fbend_dir(file);
1308ab2a 25713+ /* AuDbg("b%d, b%d\n", cookie->bindex, bend); */
25714+ for (; !err && cookie->bindex <= bend;
25715+ cookie->bindex++, cookie->h_pos = 0) {
4a4d8108 25716+ h_file = au_hf_dir(file, cookie->bindex);
1308ab2a 25717+ if (!h_file)
25718+ continue;
25719+
25720+ au_fclr_rdu(cookie->flags, FULL);
25721+ err = au_rdu_do(h_file, &arg);
25722+ AuTraceErr(err);
25723+ if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
25724+ break;
25725+ }
25726+ AuDbg("rent %llu\n", rdu->rent);
25727+
25728+ if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
25729+ rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
25730+ au_fset_rdu(cookie->flags, CONT);
25731+ cookie->generation = au_figen(file);
25732+ }
25733+
25734+ ii_read_lock_child(inode);
25735+ fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibstart(inode)));
25736+ ii_read_unlock(inode);
25737+
4f0767ce 25738+out_unlock:
1308ab2a 25739+ fi_read_unlock(file);
027c5e7a 25740+out_si:
1308ab2a 25741+ si_read_unlock(arg.sb);
4f0767ce 25742+out_mtx:
1308ab2a 25743+ mutex_unlock(&inode->i_mutex);
4f0767ce 25744+out:
1308ab2a 25745+ AuTraceErr(err);
25746+ return err;
25747+}
25748+
25749+static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
25750+{
25751+ int err;
25752+ ino_t ino;
25753+ unsigned long long nent;
25754+ union au_rdu_ent_ul *u;
25755+ struct au_rdu_ent ent;
25756+ struct super_block *sb;
25757+
25758+ err = 0;
25759+ nent = rdu->nent;
25760+ u = &rdu->ent;
2000de60 25761+ sb = file->f_path.dentry->d_sb;
1308ab2a 25762+ si_read_lock(sb, AuLock_FLUSH);
25763+ while (nent-- > 0) {
9dbd164d 25764+ /* unnecessary to support mmap_sem since this is a dir */
1308ab2a 25765+ err = copy_from_user(&ent, u->e, sizeof(ent));
4a4d8108
AM
25766+ if (!err)
25767+ err = !access_ok(VERIFY_WRITE, &u->e->ino, sizeof(ino));
1308ab2a 25768+ if (unlikely(err)) {
25769+ err = -EFAULT;
25770+ AuTraceErr(err);
25771+ break;
25772+ }
25773+
25774+ /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
25775+ if (!ent.wh)
25776+ err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
25777+ else
25778+ err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
25779+ &ino);
25780+ if (unlikely(err)) {
25781+ AuTraceErr(err);
25782+ break;
25783+ }
25784+
25785+ err = __put_user(ino, &u->e->ino);
25786+ if (unlikely(err)) {
25787+ err = -EFAULT;
25788+ AuTraceErr(err);
25789+ break;
25790+ }
25791+ u->ul += au_rdu_len(ent.nlen);
25792+ }
25793+ si_read_unlock(sb);
25794+
25795+ return err;
25796+}
25797+
25798+/* ---------------------------------------------------------------------- */
25799+
25800+static int au_rdu_verify(struct aufs_rdu *rdu)
25801+{
b752ccd1 25802+ AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
1308ab2a 25803+ "%llu, b%d, 0x%x, g%u}\n",
b752ccd1 25804+ rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
1308ab2a 25805+ rdu->blk,
25806+ rdu->rent, rdu->shwh, rdu->full,
25807+ rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
25808+ rdu->cookie.generation);
dece6358 25809+
b752ccd1 25810+ if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
1308ab2a 25811+ return 0;
dece6358 25812+
b752ccd1
AM
25813+ AuDbg("%u:%u\n",
25814+ rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
1308ab2a 25815+ return -EINVAL;
25816+}
25817+
25818+long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
dece6358 25819+{
1308ab2a 25820+ long err, e;
25821+ struct aufs_rdu rdu;
25822+ void __user *p = (void __user *)arg;
dece6358 25823+
1308ab2a 25824+ err = copy_from_user(&rdu, p, sizeof(rdu));
25825+ if (unlikely(err)) {
25826+ err = -EFAULT;
25827+ AuTraceErr(err);
25828+ goto out;
25829+ }
25830+ err = au_rdu_verify(&rdu);
dece6358
AM
25831+ if (unlikely(err))
25832+ goto out;
25833+
1308ab2a 25834+ switch (cmd) {
25835+ case AUFS_CTL_RDU:
25836+ err = au_rdu(file, &rdu);
25837+ if (unlikely(err))
25838+ break;
dece6358 25839+
1308ab2a 25840+ e = copy_to_user(p, &rdu, sizeof(rdu));
25841+ if (unlikely(e)) {
25842+ err = -EFAULT;
25843+ AuTraceErr(err);
25844+ }
25845+ break;
25846+ case AUFS_CTL_RDU_INO:
25847+ err = au_rdu_ino(file, &rdu);
25848+ break;
25849+
25850+ default:
4a4d8108 25851+ /* err = -ENOTTY; */
1308ab2a 25852+ err = -EINVAL;
25853+ }
dece6358 25854+
4f0767ce 25855+out:
1308ab2a 25856+ AuTraceErr(err);
25857+ return err;
1facf9fc 25858+}
b752ccd1
AM
25859+
25860+#ifdef CONFIG_COMPAT
25861+long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
25862+{
25863+ long err, e;
25864+ struct aufs_rdu rdu;
25865+ void __user *p = compat_ptr(arg);
25866+
25867+ /* todo: get_user()? */
25868+ err = copy_from_user(&rdu, p, sizeof(rdu));
25869+ if (unlikely(err)) {
25870+ err = -EFAULT;
25871+ AuTraceErr(err);
25872+ goto out;
25873+ }
25874+ rdu.ent.e = compat_ptr(rdu.ent.ul);
25875+ err = au_rdu_verify(&rdu);
25876+ if (unlikely(err))
25877+ goto out;
25878+
25879+ switch (cmd) {
25880+ case AUFS_CTL_RDU:
25881+ err = au_rdu(file, &rdu);
25882+ if (unlikely(err))
25883+ break;
25884+
25885+ rdu.ent.ul = ptr_to_compat(rdu.ent.e);
25886+ rdu.tail.ul = ptr_to_compat(rdu.tail.e);
25887+ e = copy_to_user(p, &rdu, sizeof(rdu));
25888+ if (unlikely(e)) {
25889+ err = -EFAULT;
25890+ AuTraceErr(err);
25891+ }
25892+ break;
25893+ case AUFS_CTL_RDU_INO:
25894+ err = au_rdu_ino(file, &rdu);
25895+ break;
25896+
25897+ default:
25898+ /* err = -ENOTTY; */
25899+ err = -EINVAL;
25900+ }
25901+
4f0767ce 25902+out:
b752ccd1
AM
25903+ AuTraceErr(err);
25904+ return err;
25905+}
25906+#endif
7f207e10
AM
25907diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
25908--- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 25909+++ linux/fs/aufs/rwsem.h 2015-09-24 10:47:58.254719746 +0200
076b876e 25910@@ -0,0 +1,191 @@
1facf9fc 25911+/*
2000de60 25912+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 25913+ *
25914+ * This program, aufs is free software; you can redistribute it and/or modify
25915+ * it under the terms of the GNU General Public License as published by
25916+ * the Free Software Foundation; either version 2 of the License, or
25917+ * (at your option) any later version.
dece6358
AM
25918+ *
25919+ * This program is distributed in the hope that it will be useful,
25920+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
25921+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25922+ * GNU General Public License for more details.
25923+ *
25924+ * You should have received a copy of the GNU General Public License
523b37e3 25925+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 25926+ */
25927+
25928+/*
25929+ * simple read-write semaphore wrappers
25930+ */
25931+
25932+#ifndef __AUFS_RWSEM_H__
25933+#define __AUFS_RWSEM_H__
25934+
25935+#ifdef __KERNEL__
25936+
4a4d8108 25937+#include "debug.h"
dece6358
AM
25938+
25939+struct au_rwsem {
25940+ struct rw_semaphore rwsem;
25941+#ifdef CONFIG_AUFS_DEBUG
25942+ /* just for debugging, not almighty counter */
25943+ atomic_t rcnt, wcnt;
25944+#endif
25945+};
25946+
25947+#ifdef CONFIG_AUFS_DEBUG
25948+#define AuDbgCntInit(rw) do { \
25949+ atomic_set(&(rw)->rcnt, 0); \
25950+ atomic_set(&(rw)->wcnt, 0); \
25951+ smp_mb(); /* atomic set */ \
25952+} while (0)
25953+
e49829fe 25954+#define AuDbgRcntInc(rw) atomic_inc(&(rw)->rcnt)
dece6358 25955+#define AuDbgRcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->rcnt) < 0)
e49829fe 25956+#define AuDbgWcntInc(rw) atomic_inc(&(rw)->wcnt)
dece6358
AM
25957+#define AuDbgWcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->wcnt) < 0)
25958+#else
25959+#define AuDbgCntInit(rw) do {} while (0)
25960+#define AuDbgRcntInc(rw) do {} while (0)
25961+#define AuDbgRcntDec(rw) do {} while (0)
25962+#define AuDbgWcntInc(rw) do {} while (0)
25963+#define AuDbgWcntDec(rw) do {} while (0)
25964+#endif /* CONFIG_AUFS_DEBUG */
25965+
25966+/* to debug easier, do not make them inlined functions */
25967+#define AuRwMustNoWaiters(rw) AuDebugOn(!list_empty(&(rw)->rwsem.wait_list))
25968+/* rwsem_is_locked() is unusable */
25969+#define AuRwMustReadLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0)
25970+#define AuRwMustWriteLock(rw) AuDebugOn(atomic_read(&(rw)->wcnt) <= 0)
25971+#define AuRwMustAnyLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0 \
25972+ && atomic_read(&(rw)->wcnt) <= 0)
25973+#define AuRwDestroy(rw) AuDebugOn(atomic_read(&(rw)->rcnt) \
25974+ || atomic_read(&(rw)->wcnt))
25975+
e49829fe
JR
25976+#define au_rw_class(rw, key) lockdep_set_class(&(rw)->rwsem, key)
25977+
dece6358
AM
25978+static inline void au_rw_init(struct au_rwsem *rw)
25979+{
25980+ AuDbgCntInit(rw);
25981+ init_rwsem(&rw->rwsem);
25982+}
25983+
25984+static inline void au_rw_init_wlock(struct au_rwsem *rw)
25985+{
25986+ au_rw_init(rw);
25987+ down_write(&rw->rwsem);
25988+ AuDbgWcntInc(rw);
25989+}
25990+
25991+static inline void au_rw_init_wlock_nested(struct au_rwsem *rw,
25992+ unsigned int lsc)
25993+{
25994+ au_rw_init(rw);
25995+ down_write_nested(&rw->rwsem, lsc);
25996+ AuDbgWcntInc(rw);
25997+}
25998+
25999+static inline void au_rw_read_lock(struct au_rwsem *rw)
26000+{
26001+ down_read(&rw->rwsem);
26002+ AuDbgRcntInc(rw);
26003+}
26004+
26005+static inline void au_rw_read_lock_nested(struct au_rwsem *rw, unsigned int lsc)
26006+{
26007+ down_read_nested(&rw->rwsem, lsc);
26008+ AuDbgRcntInc(rw);
26009+}
26010+
26011+static inline void au_rw_read_unlock(struct au_rwsem *rw)
26012+{
26013+ AuRwMustReadLock(rw);
26014+ AuDbgRcntDec(rw);
26015+ up_read(&rw->rwsem);
26016+}
26017+
26018+static inline void au_rw_dgrade_lock(struct au_rwsem *rw)
26019+{
26020+ AuRwMustWriteLock(rw);
26021+ AuDbgRcntInc(rw);
26022+ AuDbgWcntDec(rw);
26023+ downgrade_write(&rw->rwsem);
26024+}
26025+
26026+static inline void au_rw_write_lock(struct au_rwsem *rw)
26027+{
26028+ down_write(&rw->rwsem);
26029+ AuDbgWcntInc(rw);
26030+}
26031+
26032+static inline void au_rw_write_lock_nested(struct au_rwsem *rw,
26033+ unsigned int lsc)
26034+{
26035+ down_write_nested(&rw->rwsem, lsc);
26036+ AuDbgWcntInc(rw);
26037+}
1facf9fc 26038+
dece6358
AM
26039+static inline void au_rw_write_unlock(struct au_rwsem *rw)
26040+{
26041+ AuRwMustWriteLock(rw);
26042+ AuDbgWcntDec(rw);
26043+ up_write(&rw->rwsem);
26044+}
26045+
26046+/* why is not _nested version defined */
26047+static inline int au_rw_read_trylock(struct au_rwsem *rw)
26048+{
076b876e
AM
26049+ int ret;
26050+
26051+ ret = down_read_trylock(&rw->rwsem);
dece6358
AM
26052+ if (ret)
26053+ AuDbgRcntInc(rw);
26054+ return ret;
26055+}
26056+
26057+static inline int au_rw_write_trylock(struct au_rwsem *rw)
26058+{
076b876e
AM
26059+ int ret;
26060+
26061+ ret = down_write_trylock(&rw->rwsem);
dece6358
AM
26062+ if (ret)
26063+ AuDbgWcntInc(rw);
26064+ return ret;
26065+}
26066+
26067+#undef AuDbgCntInit
26068+#undef AuDbgRcntInc
26069+#undef AuDbgRcntDec
26070+#undef AuDbgWcntInc
26071+#undef AuDbgWcntDec
1facf9fc 26072+
26073+#define AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
26074+static inline void prefix##_read_lock(param) \
dece6358 26075+{ au_rw_read_lock(rwsem); } \
1facf9fc 26076+static inline void prefix##_write_lock(param) \
dece6358 26077+{ au_rw_write_lock(rwsem); } \
1facf9fc 26078+static inline int prefix##_read_trylock(param) \
dece6358 26079+{ return au_rw_read_trylock(rwsem); } \
1facf9fc 26080+static inline int prefix##_write_trylock(param) \
dece6358 26081+{ return au_rw_write_trylock(rwsem); }
1facf9fc 26082+/* why is not _nested version defined */
26083+/* static inline void prefix##_read_trylock_nested(param, lsc)
dece6358 26084+{ au_rw_read_trylock_nested(rwsem, lsc)); }
1facf9fc 26085+static inline void prefix##_write_trylock_nestd(param, lsc)
dece6358 26086+{ au_rw_write_trylock_nested(rwsem, lsc); } */
1facf9fc 26087+
26088+#define AuSimpleUnlockRwsemFuncs(prefix, param, rwsem) \
26089+static inline void prefix##_read_unlock(param) \
dece6358 26090+{ au_rw_read_unlock(rwsem); } \
1facf9fc 26091+static inline void prefix##_write_unlock(param) \
dece6358 26092+{ au_rw_write_unlock(rwsem); } \
1facf9fc 26093+static inline void prefix##_downgrade_lock(param) \
dece6358 26094+{ au_rw_dgrade_lock(rwsem); }
1facf9fc 26095+
26096+#define AuSimpleRwsemFuncs(prefix, param, rwsem) \
26097+ AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
26098+ AuSimpleUnlockRwsemFuncs(prefix, param, rwsem)
26099+
26100+#endif /* __KERNEL__ */
26101+#endif /* __AUFS_RWSEM_H__ */
7f207e10
AM
26102diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
26103--- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c
AM
26104+++ linux/fs/aufs/sbinfo.c 2015-09-24 10:47:58.254719746 +0200
26105@@ -0,0 +1,360 @@
1facf9fc 26106+/*
2000de60 26107+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 26108+ *
26109+ * This program, aufs is free software; you can redistribute it and/or modify
26110+ * it under the terms of the GNU General Public License as published by
26111+ * the Free Software Foundation; either version 2 of the License, or
26112+ * (at your option) any later version.
dece6358
AM
26113+ *
26114+ * This program is distributed in the hope that it will be useful,
26115+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26116+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26117+ * GNU General Public License for more details.
26118+ *
26119+ * You should have received a copy of the GNU General Public License
523b37e3 26120+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26121+ */
26122+
26123+/*
26124+ * superblock private data
26125+ */
26126+
26127+#include "aufs.h"
26128+
26129+/*
26130+ * they are necessary regardless sysfs is disabled.
26131+ */
26132+void au_si_free(struct kobject *kobj)
26133+{
86dc4139 26134+ int i;
1facf9fc 26135+ struct au_sbinfo *sbinfo;
b752ccd1 26136+ char *locked __maybe_unused; /* debug only */
1facf9fc 26137+
26138+ sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
86dc4139
AM
26139+ for (i = 0; i < AuPlink_NHASH; i++)
26140+ AuDebugOn(!hlist_empty(&sbinfo->si_plink[i].head));
e49829fe 26141+ AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
1facf9fc 26142+
c2c0f25c
AM
26143+ AuDebugOn(!hlist_empty(&sbinfo->si_symlink.head));
26144+
e49829fe 26145+ au_rw_write_lock(&sbinfo->si_rwsem);
1facf9fc 26146+ au_br_free(sbinfo);
e49829fe 26147+ au_rw_write_unlock(&sbinfo->si_rwsem);
b752ccd1
AM
26148+
26149+ AuDebugOn(radix_tree_gang_lookup
26150+ (&sbinfo->au_si_pid.tree, (void **)&locked,
26151+ /*first_index*/PID_MAX_DEFAULT - 1,
26152+ /*max_items*/sizeof(locked)/sizeof(*locked)));
26153+
1facf9fc 26154+ kfree(sbinfo->si_branch);
b752ccd1 26155+ kfree(sbinfo->au_si_pid.bitmap);
1facf9fc 26156+ mutex_destroy(&sbinfo->si_xib_mtx);
dece6358 26157+ AuRwDestroy(&sbinfo->si_rwsem);
1facf9fc 26158+
26159+ kfree(sbinfo);
26160+}
26161+
26162+int au_si_alloc(struct super_block *sb)
26163+{
86dc4139 26164+ int err, i;
1facf9fc 26165+ struct au_sbinfo *sbinfo;
e49829fe 26166+ static struct lock_class_key aufs_si;
1facf9fc 26167+
26168+ err = -ENOMEM;
4a4d8108 26169+ sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
1facf9fc 26170+ if (unlikely(!sbinfo))
26171+ goto out;
26172+
b752ccd1
AM
26173+ BUILD_BUG_ON(sizeof(unsigned long) !=
26174+ sizeof(*sbinfo->au_si_pid.bitmap));
26175+ sbinfo->au_si_pid.bitmap = kcalloc(BITS_TO_LONGS(PID_MAX_DEFAULT),
26176+ sizeof(*sbinfo->au_si_pid.bitmap),
26177+ GFP_NOFS);
26178+ if (unlikely(!sbinfo->au_si_pid.bitmap))
26179+ goto out_sbinfo;
26180+
1facf9fc 26181+ /* will be reallocated separately */
26182+ sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
26183+ if (unlikely(!sbinfo->si_branch))
b752ccd1 26184+ goto out_pidmap;
1facf9fc 26185+
1facf9fc 26186+ err = sysaufs_si_init(sbinfo);
26187+ if (unlikely(err))
26188+ goto out_br;
26189+
26190+ au_nwt_init(&sbinfo->si_nowait);
dece6358 26191+ au_rw_init_wlock(&sbinfo->si_rwsem);
e49829fe 26192+ au_rw_class(&sbinfo->si_rwsem, &aufs_si);
b752ccd1
AM
26193+ spin_lock_init(&sbinfo->au_si_pid.tree_lock);
26194+ INIT_RADIX_TREE(&sbinfo->au_si_pid.tree, GFP_ATOMIC | __GFP_NOFAIL);
26195+
7f207e10 26196+ atomic_long_set(&sbinfo->si_ninodes, 0);
7f207e10
AM
26197+ atomic_long_set(&sbinfo->si_nfiles, 0);
26198+
1facf9fc 26199+ sbinfo->si_bend = -1;
392086de 26200+ sbinfo->si_last_br_id = AUFS_BRANCH_MAX / 2;
1facf9fc 26201+
26202+ sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
26203+ sbinfo->si_wbr_create = AuWbrCreate_Def;
4a4d8108
AM
26204+ sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
26205+ sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
1facf9fc 26206+
076b876e
AM
26207+ au_fhsm_init(sbinfo);
26208+
e49829fe 26209+ sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
1facf9fc 26210+
c2c0f25c
AM
26211+ au_sphl_init(&sbinfo->si_symlink);
26212+
392086de
AM
26213+ sbinfo->si_xino_jiffy = jiffies;
26214+ sbinfo->si_xino_expire
26215+ = msecs_to_jiffies(AUFS_XINO_DEF_SEC * MSEC_PER_SEC);
1facf9fc 26216+ mutex_init(&sbinfo->si_xib_mtx);
1facf9fc 26217+ sbinfo->si_xino_brid = -1;
26218+ /* leave si_xib_last_pindex and si_xib_next_bit */
26219+
b912730e
AM
26220+ au_sphl_init(&sbinfo->si_aopen);
26221+
e49829fe 26222+ sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
1facf9fc 26223+ sbinfo->si_rdblk = AUFS_RDBLK_DEF;
26224+ sbinfo->si_rdhash = AUFS_RDHASH_DEF;
26225+ sbinfo->si_dirwh = AUFS_DIRWH_DEF;
26226+
86dc4139
AM
26227+ for (i = 0; i < AuPlink_NHASH; i++)
26228+ au_sphl_init(sbinfo->si_plink + i);
1facf9fc 26229+ init_waitqueue_head(&sbinfo->si_plink_wq);
4a4d8108 26230+ spin_lock_init(&sbinfo->si_plink_maint_lock);
1facf9fc 26231+
523b37e3
AM
26232+ au_sphl_init(&sbinfo->si_files);
26233+
1facf9fc 26234+ /* leave other members for sysaufs and si_mnt. */
26235+ sbinfo->si_sb = sb;
26236+ sb->s_fs_info = sbinfo;
b752ccd1 26237+ si_pid_set(sb);
1facf9fc 26238+ return 0; /* success */
26239+
4f0767ce 26240+out_br:
1facf9fc 26241+ kfree(sbinfo->si_branch);
4f0767ce 26242+out_pidmap:
b752ccd1 26243+ kfree(sbinfo->au_si_pid.bitmap);
4f0767ce 26244+out_sbinfo:
1facf9fc 26245+ kfree(sbinfo);
4f0767ce 26246+out:
1facf9fc 26247+ return err;
26248+}
26249+
26250+int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr)
26251+{
26252+ int err, sz;
26253+ struct au_branch **brp;
26254+
dece6358
AM
26255+ AuRwMustWriteLock(&sbinfo->si_rwsem);
26256+
1facf9fc 26257+ err = -ENOMEM;
26258+ sz = sizeof(*brp) * (sbinfo->si_bend + 1);
26259+ if (unlikely(!sz))
26260+ sz = sizeof(*brp);
26261+ brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS);
26262+ if (brp) {
26263+ sbinfo->si_branch = brp;
26264+ err = 0;
26265+ }
26266+
26267+ return err;
26268+}
26269+
26270+/* ---------------------------------------------------------------------- */
26271+
26272+unsigned int au_sigen_inc(struct super_block *sb)
26273+{
26274+ unsigned int gen;
5527c038 26275+ struct inode *inode;
1facf9fc 26276+
dece6358
AM
26277+ SiMustWriteLock(sb);
26278+
1facf9fc 26279+ gen = ++au_sbi(sb)->si_generation;
26280+ au_update_digen(sb->s_root);
5527c038
JR
26281+ inode = d_inode(sb->s_root);
26282+ au_update_iigen(inode, /*half*/0);
26283+ inode->i_version++;
1facf9fc 26284+ return gen;
26285+}
26286+
26287+aufs_bindex_t au_new_br_id(struct super_block *sb)
26288+{
26289+ aufs_bindex_t br_id;
26290+ int i;
26291+ struct au_sbinfo *sbinfo;
26292+
dece6358
AM
26293+ SiMustWriteLock(sb);
26294+
1facf9fc 26295+ sbinfo = au_sbi(sb);
26296+ for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
26297+ br_id = ++sbinfo->si_last_br_id;
7f207e10 26298+ AuDebugOn(br_id < 0);
1facf9fc 26299+ if (br_id && au_br_index(sb, br_id) < 0)
26300+ return br_id;
26301+ }
26302+
26303+ return -1;
26304+}
26305+
26306+/* ---------------------------------------------------------------------- */
26307+
e49829fe
JR
26308+/* it is ok that new 'nwt' tasks are appended while we are sleeping */
26309+int si_read_lock(struct super_block *sb, int flags)
26310+{
26311+ int err;
26312+
26313+ err = 0;
26314+ if (au_ftest_lock(flags, FLUSH))
26315+ au_nwt_flush(&au_sbi(sb)->si_nowait);
26316+
26317+ si_noflush_read_lock(sb);
26318+ err = au_plink_maint(sb, flags);
26319+ if (unlikely(err))
26320+ si_read_unlock(sb);
26321+
26322+ return err;
26323+}
26324+
26325+int si_write_lock(struct super_block *sb, int flags)
26326+{
26327+ int err;
26328+
26329+ if (au_ftest_lock(flags, FLUSH))
26330+ au_nwt_flush(&au_sbi(sb)->si_nowait);
26331+
26332+ si_noflush_write_lock(sb);
26333+ err = au_plink_maint(sb, flags);
26334+ if (unlikely(err))
26335+ si_write_unlock(sb);
26336+
26337+ return err;
26338+}
26339+
1facf9fc 26340+/* dentry and super_block lock. call at entry point */
e49829fe 26341+int aufs_read_lock(struct dentry *dentry, int flags)
1facf9fc 26342+{
e49829fe 26343+ int err;
027c5e7a 26344+ struct super_block *sb;
e49829fe 26345+
027c5e7a
AM
26346+ sb = dentry->d_sb;
26347+ err = si_read_lock(sb, flags);
26348+ if (unlikely(err))
26349+ goto out;
26350+
26351+ if (au_ftest_lock(flags, DW))
26352+ di_write_lock_child(dentry);
26353+ else
26354+ di_read_lock_child(dentry, flags);
26355+
26356+ if (au_ftest_lock(flags, GEN)) {
26357+ err = au_digen_test(dentry, au_sigen(sb));
26358+ AuDebugOn(!err && au_dbrange_test(dentry));
26359+ if (unlikely(err))
26360+ aufs_read_unlock(dentry, flags);
e49829fe
JR
26361+ }
26362+
027c5e7a 26363+out:
e49829fe 26364+ return err;
1facf9fc 26365+}
26366+
26367+void aufs_read_unlock(struct dentry *dentry, int flags)
26368+{
26369+ if (au_ftest_lock(flags, DW))
26370+ di_write_unlock(dentry);
26371+ else
26372+ di_read_unlock(dentry, flags);
26373+ si_read_unlock(dentry->d_sb);
26374+}
26375+
26376+void aufs_write_lock(struct dentry *dentry)
26377+{
e49829fe 26378+ si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
1facf9fc 26379+ di_write_lock_child(dentry);
26380+}
26381+
26382+void aufs_write_unlock(struct dentry *dentry)
26383+{
26384+ di_write_unlock(dentry);
26385+ si_write_unlock(dentry->d_sb);
26386+}
26387+
e49829fe 26388+int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
1facf9fc 26389+{
e49829fe 26390+ int err;
027c5e7a
AM
26391+ unsigned int sigen;
26392+ struct super_block *sb;
e49829fe 26393+
027c5e7a
AM
26394+ sb = d1->d_sb;
26395+ err = si_read_lock(sb, flags);
26396+ if (unlikely(err))
26397+ goto out;
26398+
26399+ di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIR));
26400+
26401+ if (au_ftest_lock(flags, GEN)) {
26402+ sigen = au_sigen(sb);
26403+ err = au_digen_test(d1, sigen);
26404+ AuDebugOn(!err && au_dbrange_test(d1));
26405+ if (!err) {
26406+ err = au_digen_test(d2, sigen);
26407+ AuDebugOn(!err && au_dbrange_test(d2));
26408+ }
26409+ if (unlikely(err))
26410+ aufs_read_and_write_unlock2(d1, d2);
26411+ }
26412+
26413+out:
e49829fe 26414+ return err;
1facf9fc 26415+}
26416+
26417+void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
26418+{
26419+ di_write_unlock2(d1, d2);
26420+ si_read_unlock(d1->d_sb);
26421+}
b752ccd1
AM
26422+
26423+/* ---------------------------------------------------------------------- */
26424+
26425+int si_pid_test_slow(struct super_block *sb)
26426+{
26427+ void *p;
26428+
26429+ rcu_read_lock();
26430+ p = radix_tree_lookup(&au_sbi(sb)->au_si_pid.tree, current->pid);
26431+ rcu_read_unlock();
26432+
027c5e7a 26433+ return (long)!!p;
b752ccd1
AM
26434+}
26435+
26436+void si_pid_set_slow(struct super_block *sb)
26437+{
26438+ int err;
26439+ struct au_sbinfo *sbinfo;
26440+
26441+ AuDebugOn(si_pid_test_slow(sb));
26442+
26443+ sbinfo = au_sbi(sb);
26444+ err = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
26445+ AuDebugOn(err);
26446+ spin_lock(&sbinfo->au_si_pid.tree_lock);
26447+ err = radix_tree_insert(&sbinfo->au_si_pid.tree, current->pid,
027c5e7a 26448+ /*any valid ptr*/sb);
b752ccd1
AM
26449+ spin_unlock(&sbinfo->au_si_pid.tree_lock);
26450+ AuDebugOn(err);
26451+ radix_tree_preload_end();
26452+}
26453+
26454+void si_pid_clr_slow(struct super_block *sb)
26455+{
26456+ void *p;
26457+ struct au_sbinfo *sbinfo;
26458+
26459+ AuDebugOn(!si_pid_test_slow(sb));
26460+
26461+ sbinfo = au_sbi(sb);
26462+ spin_lock(&sbinfo->au_si_pid.tree_lock);
26463+ p = radix_tree_delete(&sbinfo->au_si_pid.tree, current->pid);
26464+ spin_unlock(&sbinfo->au_si_pid.tree_lock);
b752ccd1 26465+}
7f207e10
AM
26466diff -urN /usr/share/empty/fs/aufs/spl.h linux/fs/aufs/spl.h
26467--- /usr/share/empty/fs/aufs/spl.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 26468+++ linux/fs/aufs/spl.h 2015-09-24 10:47:58.254719746 +0200
523b37e3 26469@@ -0,0 +1,111 @@
1facf9fc 26470+/*
2000de60 26471+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 26472+ *
26473+ * This program, aufs is free software; you can redistribute it and/or modify
26474+ * it under the terms of the GNU General Public License as published by
26475+ * the Free Software Foundation; either version 2 of the License, or
26476+ * (at your option) any later version.
dece6358
AM
26477+ *
26478+ * This program is distributed in the hope that it will be useful,
26479+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26480+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26481+ * GNU General Public License for more details.
26482+ *
26483+ * You should have received a copy of the GNU General Public License
523b37e3 26484+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26485+ */
26486+
26487+/*
26488+ * simple list protected by a spinlock
26489+ */
26490+
26491+#ifndef __AUFS_SPL_H__
26492+#define __AUFS_SPL_H__
26493+
26494+#ifdef __KERNEL__
26495+
1facf9fc 26496+struct au_splhead {
26497+ spinlock_t spin;
26498+ struct list_head head;
26499+};
26500+
26501+static inline void au_spl_init(struct au_splhead *spl)
26502+{
26503+ spin_lock_init(&spl->spin);
26504+ INIT_LIST_HEAD(&spl->head);
26505+}
26506+
26507+static inline void au_spl_add(struct list_head *list, struct au_splhead *spl)
26508+{
26509+ spin_lock(&spl->spin);
26510+ list_add(list, &spl->head);
26511+ spin_unlock(&spl->spin);
26512+}
26513+
26514+static inline void au_spl_del(struct list_head *list, struct au_splhead *spl)
26515+{
26516+ spin_lock(&spl->spin);
26517+ list_del(list);
26518+ spin_unlock(&spl->spin);
26519+}
26520+
4a4d8108
AM
26521+static inline void au_spl_del_rcu(struct list_head *list,
26522+ struct au_splhead *spl)
26523+{
26524+ spin_lock(&spl->spin);
26525+ list_del_rcu(list);
26526+ spin_unlock(&spl->spin);
26527+}
26528+
86dc4139
AM
26529+/* ---------------------------------------------------------------------- */
26530+
26531+struct au_sphlhead {
26532+ spinlock_t spin;
26533+ struct hlist_head head;
26534+};
26535+
26536+static inline void au_sphl_init(struct au_sphlhead *sphl)
26537+{
26538+ spin_lock_init(&sphl->spin);
26539+ INIT_HLIST_HEAD(&sphl->head);
26540+}
26541+
26542+static inline void au_sphl_add(struct hlist_node *hlist,
26543+ struct au_sphlhead *sphl)
26544+{
26545+ spin_lock(&sphl->spin);
26546+ hlist_add_head(hlist, &sphl->head);
26547+ spin_unlock(&sphl->spin);
26548+}
26549+
26550+static inline void au_sphl_del(struct hlist_node *hlist,
26551+ struct au_sphlhead *sphl)
26552+{
26553+ spin_lock(&sphl->spin);
26554+ hlist_del(hlist);
26555+ spin_unlock(&sphl->spin);
26556+}
26557+
26558+static inline void au_sphl_del_rcu(struct hlist_node *hlist,
26559+ struct au_sphlhead *sphl)
26560+{
26561+ spin_lock(&sphl->spin);
26562+ hlist_del_rcu(hlist);
26563+ spin_unlock(&sphl->spin);
26564+}
26565+
26566+static inline unsigned long au_sphl_count(struct au_sphlhead *sphl)
26567+{
26568+ unsigned long cnt;
26569+ struct hlist_node *pos;
26570+
26571+ cnt = 0;
26572+ spin_lock(&sphl->spin);
26573+ hlist_for_each(pos, &sphl->head)
26574+ cnt++;
26575+ spin_unlock(&sphl->spin);
26576+ return cnt;
26577+}
26578+
1facf9fc 26579+#endif /* __KERNEL__ */
26580+#endif /* __AUFS_SPL_H__ */
7f207e10
AM
26581diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
26582--- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 26583+++ linux/fs/aufs/super.c 2015-09-24 10:47:58.254719746 +0200
5527c038 26584@@ -0,0 +1,1004 @@
1facf9fc 26585+/*
2000de60 26586+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 26587+ *
26588+ * This program, aufs is free software; you can redistribute it and/or modify
26589+ * it under the terms of the GNU General Public License as published by
26590+ * the Free Software Foundation; either version 2 of the License, or
26591+ * (at your option) any later version.
dece6358
AM
26592+ *
26593+ * This program is distributed in the hope that it will be useful,
26594+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
26595+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26596+ * GNU General Public License for more details.
26597+ *
26598+ * You should have received a copy of the GNU General Public License
523b37e3 26599+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 26600+ */
26601+
26602+/*
26603+ * mount and super_block operations
26604+ */
26605+
f6c5ef8b 26606+#include <linux/mm.h>
1facf9fc 26607+#include <linux/seq_file.h>
26608+#include <linux/statfs.h>
7f207e10 26609+#include <linux/vmalloc.h>
1facf9fc 26610+#include "aufs.h"
26611+
26612+/*
26613+ * super_operations
26614+ */
26615+static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
26616+{
26617+ struct au_icntnr *c;
26618+
26619+ c = au_cache_alloc_icntnr();
26620+ if (c) {
027c5e7a 26621+ au_icntnr_init(c);
1facf9fc 26622+ c->vfs_inode.i_version = 1; /* sigen(sb); */
26623+ c->iinfo.ii_hinode = NULL;
26624+ return &c->vfs_inode;
26625+ }
26626+ return NULL;
26627+}
26628+
027c5e7a
AM
26629+static void aufs_destroy_inode_cb(struct rcu_head *head)
26630+{
26631+ struct inode *inode = container_of(head, struct inode, i_rcu);
26632+
b4510431 26633+ INIT_HLIST_HEAD(&inode->i_dentry);
027c5e7a
AM
26634+ au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
26635+}
26636+
1facf9fc 26637+static void aufs_destroy_inode(struct inode *inode)
26638+{
26639+ au_iinfo_fin(inode);
027c5e7a 26640+ call_rcu(&inode->i_rcu, aufs_destroy_inode_cb);
1facf9fc 26641+}
26642+
26643+struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
26644+{
26645+ struct inode *inode;
26646+ int err;
26647+
26648+ inode = iget_locked(sb, ino);
26649+ if (unlikely(!inode)) {
26650+ inode = ERR_PTR(-ENOMEM);
26651+ goto out;
26652+ }
26653+ if (!(inode->i_state & I_NEW))
26654+ goto out;
26655+
26656+ err = au_xigen_new(inode);
26657+ if (!err)
26658+ err = au_iinfo_init(inode);
26659+ if (!err)
26660+ inode->i_version++;
26661+ else {
26662+ iget_failed(inode);
26663+ inode = ERR_PTR(err);
26664+ }
26665+
4f0767ce 26666+out:
1facf9fc 26667+ /* never return NULL */
26668+ AuDebugOn(!inode);
26669+ AuTraceErrPtr(inode);
26670+ return inode;
26671+}
26672+
26673+/* lock free root dinfo */
26674+static int au_show_brs(struct seq_file *seq, struct super_block *sb)
26675+{
26676+ int err;
26677+ aufs_bindex_t bindex, bend;
26678+ struct path path;
4a4d8108 26679+ struct au_hdentry *hdp;
1facf9fc 26680+ struct au_branch *br;
076b876e 26681+ au_br_perm_str_t perm;
1facf9fc 26682+
26683+ err = 0;
26684+ bend = au_sbend(sb);
4a4d8108 26685+ hdp = au_di(sb->s_root)->di_hdentry;
1facf9fc 26686+ for (bindex = 0; !err && bindex <= bend; bindex++) {
26687+ br = au_sbr(sb, bindex);
86dc4139 26688+ path.mnt = au_br_mnt(br);
4a4d8108 26689+ path.dentry = hdp[bindex].hd_dentry;
1facf9fc 26690+ err = au_seq_path(seq, &path);
1e00d052 26691+ if (err > 0) {
076b876e
AM
26692+ au_optstr_br_perm(&perm, br->br_perm);
26693+ err = seq_printf(seq, "=%s", perm.a);
26694+ if (err == -1)
26695+ err = -E2BIG;
1e00d052 26696+ }
1facf9fc 26697+ if (!err && bindex != bend)
26698+ err = seq_putc(seq, ':');
26699+ }
26700+
26701+ return err;
26702+}
26703+
26704+static void au_show_wbr_create(struct seq_file *m, int v,
26705+ struct au_sbinfo *sbinfo)
26706+{
26707+ const char *pat;
26708+
dece6358
AM
26709+ AuRwMustAnyLock(&sbinfo->si_rwsem);
26710+
c2b27bf2 26711+ seq_puts(m, ",create=");
1facf9fc 26712+ pat = au_optstr_wbr_create(v);
26713+ switch (v) {
26714+ case AuWbrCreate_TDP:
26715+ case AuWbrCreate_RR:
26716+ case AuWbrCreate_MFS:
26717+ case AuWbrCreate_PMFS:
c2b27bf2 26718+ seq_puts(m, pat);
1facf9fc 26719+ break;
26720+ case AuWbrCreate_MFSV:
26721+ seq_printf(m, /*pat*/"mfs:%lu",
e49829fe
JR
26722+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26723+ / MSEC_PER_SEC);
1facf9fc 26724+ break;
26725+ case AuWbrCreate_PMFSV:
26726+ seq_printf(m, /*pat*/"pmfs:%lu",
e49829fe
JR
26727+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26728+ / MSEC_PER_SEC);
1facf9fc 26729+ break;
26730+ case AuWbrCreate_MFSRR:
26731+ seq_printf(m, /*pat*/"mfsrr:%llu",
26732+ sbinfo->si_wbr_mfs.mfsrr_watermark);
26733+ break;
26734+ case AuWbrCreate_MFSRRV:
26735+ seq_printf(m, /*pat*/"mfsrr:%llu:%lu",
26736+ sbinfo->si_wbr_mfs.mfsrr_watermark,
e49829fe
JR
26737+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26738+ / MSEC_PER_SEC);
1facf9fc 26739+ break;
392086de
AM
26740+ case AuWbrCreate_PMFSRR:
26741+ seq_printf(m, /*pat*/"pmfsrr:%llu",
26742+ sbinfo->si_wbr_mfs.mfsrr_watermark);
26743+ break;
26744+ case AuWbrCreate_PMFSRRV:
26745+ seq_printf(m, /*pat*/"pmfsrr:%llu:%lu",
26746+ sbinfo->si_wbr_mfs.mfsrr_watermark,
26747+ jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
26748+ / MSEC_PER_SEC);
26749+ break;
1facf9fc 26750+ }
26751+}
26752+
7eafdf33 26753+static int au_show_xino(struct seq_file *seq, struct super_block *sb)
1facf9fc 26754+{
26755+#ifdef CONFIG_SYSFS
26756+ return 0;
26757+#else
26758+ int err;
26759+ const int len = sizeof(AUFS_XINO_FNAME) - 1;
26760+ aufs_bindex_t bindex, brid;
1facf9fc 26761+ struct qstr *name;
26762+ struct file *f;
26763+ struct dentry *d, *h_root;
4a4d8108 26764+ struct au_hdentry *hdp;
1facf9fc 26765+
dece6358
AM
26766+ AuRwMustAnyLock(&sbinfo->si_rwsem);
26767+
1facf9fc 26768+ err = 0;
1facf9fc 26769+ f = au_sbi(sb)->si_xib;
26770+ if (!f)
26771+ goto out;
26772+
26773+ /* stop printing the default xino path on the first writable branch */
26774+ h_root = NULL;
26775+ brid = au_xino_brid(sb);
26776+ if (brid >= 0) {
26777+ bindex = au_br_index(sb, brid);
4a4d8108
AM
26778+ hdp = au_di(sb->s_root)->di_hdentry;
26779+ h_root = hdp[0 + bindex].hd_dentry;
1facf9fc 26780+ }
2000de60 26781+ d = f->f_path.dentry;
1facf9fc 26782+ name = &d->d_name;
26783+ /* safe ->d_parent because the file is unlinked */
26784+ if (d->d_parent == h_root
26785+ && name->len == len
26786+ && !memcmp(name->name, AUFS_XINO_FNAME, len))
26787+ goto out;
26788+
26789+ seq_puts(seq, ",xino=");
26790+ err = au_xino_path(seq, f);
26791+
4f0767ce 26792+out:
1facf9fc 26793+ return err;
26794+#endif
26795+}
26796+
26797+/* seq_file will re-call me in case of too long string */
7eafdf33 26798+static int aufs_show_options(struct seq_file *m, struct dentry *dentry)
1facf9fc 26799+{
027c5e7a 26800+ int err;
1facf9fc 26801+ unsigned int mnt_flags, v;
26802+ struct super_block *sb;
26803+ struct au_sbinfo *sbinfo;
26804+
26805+#define AuBool(name, str) do { \
26806+ v = au_opt_test(mnt_flags, name); \
26807+ if (v != au_opt_test(AuOpt_Def, name)) \
26808+ seq_printf(m, ",%s" #str, v ? "" : "no"); \
26809+} while (0)
26810+
26811+#define AuStr(name, str) do { \
26812+ v = mnt_flags & AuOptMask_##name; \
26813+ if (v != (AuOpt_Def & AuOptMask_##name)) \
26814+ seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
26815+} while (0)
26816+
26817+#define AuUInt(name, str, val) do { \
26818+ if (val != AUFS_##name##_DEF) \
26819+ seq_printf(m, "," #str "=%u", val); \
26820+} while (0)
26821+
7eafdf33 26822+ sb = dentry->d_sb;
c1595e42
JR
26823+ if (sb->s_flags & MS_POSIXACL)
26824+ seq_puts(m, ",acl");
26825+
26826+ /* lock free root dinfo */
1facf9fc 26827+ si_noflush_read_lock(sb);
26828+ sbinfo = au_sbi(sb);
26829+ seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
26830+
26831+ mnt_flags = au_mntflags(sb);
26832+ if (au_opt_test(mnt_flags, XINO)) {
7eafdf33 26833+ err = au_show_xino(m, sb);
1facf9fc 26834+ if (unlikely(err))
26835+ goto out;
26836+ } else
26837+ seq_puts(m, ",noxino");
26838+
26839+ AuBool(TRUNC_XINO, trunc_xino);
26840+ AuStr(UDBA, udba);
dece6358 26841+ AuBool(SHWH, shwh);
1facf9fc 26842+ AuBool(PLINK, plink);
4a4d8108 26843+ AuBool(DIO, dio);
076b876e 26844+ AuBool(DIRPERM1, dirperm1);
1facf9fc 26845+
26846+ v = sbinfo->si_wbr_create;
26847+ if (v != AuWbrCreate_Def)
26848+ au_show_wbr_create(m, v, sbinfo);
26849+
26850+ v = sbinfo->si_wbr_copyup;
26851+ if (v != AuWbrCopyup_Def)
26852+ seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
26853+
26854+ v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
26855+ if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
26856+ seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
26857+
26858+ AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
26859+
027c5e7a
AM
26860+ v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
26861+ AuUInt(RDCACHE, rdcache, v);
1facf9fc 26862+
26863+ AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
26864+ AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
26865+
076b876e
AM
26866+ au_fhsm_show(m, sbinfo);
26867+
1facf9fc 26868+ AuBool(SUM, sum);
26869+ /* AuBool(SUM_W, wsum); */
26870+ AuBool(WARN_PERM, warn_perm);
26871+ AuBool(VERBOSE, verbose);
26872+
4f0767ce 26873+out:
1facf9fc 26874+ /* be sure to print "br:" last */
26875+ if (!sysaufs_brs) {
26876+ seq_puts(m, ",br:");
26877+ au_show_brs(m, sb);
26878+ }
26879+ si_read_unlock(sb);
26880+ return 0;
26881+
1facf9fc 26882+#undef AuBool
26883+#undef AuStr
4a4d8108 26884+#undef AuUInt
1facf9fc 26885+}
26886+
26887+/* ---------------------------------------------------------------------- */
26888+
26889+/* sum mode which returns the summation for statfs(2) */
26890+
26891+static u64 au_add_till_max(u64 a, u64 b)
26892+{
26893+ u64 old;
26894+
26895+ old = a;
26896+ a += b;
92d182d2
AM
26897+ if (old <= a)
26898+ return a;
26899+ return ULLONG_MAX;
26900+}
26901+
26902+static u64 au_mul_till_max(u64 a, long mul)
26903+{
26904+ u64 old;
26905+
26906+ old = a;
26907+ a *= mul;
26908+ if (old <= a)
1facf9fc 26909+ return a;
26910+ return ULLONG_MAX;
26911+}
26912+
26913+static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
26914+{
26915+ int err;
92d182d2 26916+ long bsize, factor;
1facf9fc 26917+ u64 blocks, bfree, bavail, files, ffree;
26918+ aufs_bindex_t bend, bindex, i;
26919+ unsigned char shared;
7f207e10 26920+ struct path h_path;
1facf9fc 26921+ struct super_block *h_sb;
26922+
92d182d2
AM
26923+ err = 0;
26924+ bsize = LONG_MAX;
26925+ files = 0;
26926+ ffree = 0;
1facf9fc 26927+ blocks = 0;
26928+ bfree = 0;
26929+ bavail = 0;
1facf9fc 26930+ bend = au_sbend(sb);
92d182d2 26931+ for (bindex = 0; bindex <= bend; bindex++) {
7f207e10
AM
26932+ h_path.mnt = au_sbr_mnt(sb, bindex);
26933+ h_sb = h_path.mnt->mnt_sb;
1facf9fc 26934+ shared = 0;
92d182d2 26935+ for (i = 0; !shared && i < bindex; i++)
1facf9fc 26936+ shared = (au_sbr_sb(sb, i) == h_sb);
26937+ if (shared)
26938+ continue;
26939+
26940+ /* sb->s_root for NFS is unreliable */
7f207e10
AM
26941+ h_path.dentry = h_path.mnt->mnt_root;
26942+ err = vfs_statfs(&h_path, buf);
1facf9fc 26943+ if (unlikely(err))
26944+ goto out;
26945+
92d182d2
AM
26946+ if (bsize > buf->f_bsize) {
26947+ /*
26948+ * we will reduce bsize, so we have to expand blocks
26949+ * etc. to match them again
26950+ */
26951+ factor = (bsize / buf->f_bsize);
26952+ blocks = au_mul_till_max(blocks, factor);
26953+ bfree = au_mul_till_max(bfree, factor);
26954+ bavail = au_mul_till_max(bavail, factor);
26955+ bsize = buf->f_bsize;
26956+ }
26957+
26958+ factor = (buf->f_bsize / bsize);
26959+ blocks = au_add_till_max(blocks,
26960+ au_mul_till_max(buf->f_blocks, factor));
26961+ bfree = au_add_till_max(bfree,
26962+ au_mul_till_max(buf->f_bfree, factor));
26963+ bavail = au_add_till_max(bavail,
26964+ au_mul_till_max(buf->f_bavail, factor));
1facf9fc 26965+ files = au_add_till_max(files, buf->f_files);
26966+ ffree = au_add_till_max(ffree, buf->f_ffree);
26967+ }
26968+
92d182d2 26969+ buf->f_bsize = bsize;
1facf9fc 26970+ buf->f_blocks = blocks;
26971+ buf->f_bfree = bfree;
26972+ buf->f_bavail = bavail;
26973+ buf->f_files = files;
26974+ buf->f_ffree = ffree;
92d182d2 26975+ buf->f_frsize = 0;
1facf9fc 26976+
4f0767ce 26977+out:
1facf9fc 26978+ return err;
26979+}
26980+
26981+static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
26982+{
26983+ int err;
7f207e10 26984+ struct path h_path;
1facf9fc 26985+ struct super_block *sb;
26986+
26987+ /* lock free root dinfo */
26988+ sb = dentry->d_sb;
26989+ si_noflush_read_lock(sb);
7f207e10 26990+ if (!au_opt_test(au_mntflags(sb), SUM)) {
1facf9fc 26991+ /* sb->s_root for NFS is unreliable */
7f207e10
AM
26992+ h_path.mnt = au_sbr_mnt(sb, 0);
26993+ h_path.dentry = h_path.mnt->mnt_root;
26994+ err = vfs_statfs(&h_path, buf);
26995+ } else
1facf9fc 26996+ err = au_statfs_sum(sb, buf);
26997+ si_read_unlock(sb);
26998+
26999+ if (!err) {
27000+ buf->f_type = AUFS_SUPER_MAGIC;
4a4d8108 27001+ buf->f_namelen = AUFS_MAX_NAMELEN;
1facf9fc 27002+ memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
27003+ }
27004+ /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
27005+
27006+ return err;
27007+}
27008+
27009+/* ---------------------------------------------------------------------- */
27010+
537831f9
AM
27011+static int aufs_sync_fs(struct super_block *sb, int wait)
27012+{
27013+ int err, e;
27014+ aufs_bindex_t bend, bindex;
27015+ struct au_branch *br;
27016+ struct super_block *h_sb;
27017+
27018+ err = 0;
27019+ si_noflush_read_lock(sb);
27020+ bend = au_sbend(sb);
27021+ for (bindex = 0; bindex <= bend; bindex++) {
27022+ br = au_sbr(sb, bindex);
27023+ if (!au_br_writable(br->br_perm))
27024+ continue;
27025+
27026+ h_sb = au_sbr_sb(sb, bindex);
27027+ if (h_sb->s_op->sync_fs) {
27028+ e = h_sb->s_op->sync_fs(h_sb, wait);
27029+ if (unlikely(e && !err))
27030+ err = e;
27031+ /* go on even if an error happens */
27032+ }
27033+ }
27034+ si_read_unlock(sb);
27035+
27036+ return err;
27037+}
27038+
27039+/* ---------------------------------------------------------------------- */
27040+
1facf9fc 27041+/* final actions when unmounting a file system */
27042+static void aufs_put_super(struct super_block *sb)
27043+{
27044+ struct au_sbinfo *sbinfo;
27045+
27046+ sbinfo = au_sbi(sb);
27047+ if (!sbinfo)
27048+ return;
27049+
1facf9fc 27050+ dbgaufs_si_fin(sbinfo);
27051+ kobject_put(&sbinfo->si_kobj);
27052+}
27053+
27054+/* ---------------------------------------------------------------------- */
27055+
7f207e10
AM
27056+void au_array_free(void *array)
27057+{
27058+ if (array) {
27059+ if (!is_vmalloc_addr(array))
27060+ kfree(array);
27061+ else
27062+ vfree(array);
27063+ }
27064+}
27065+
27066+void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb, void *arg)
27067+{
27068+ void *array;
076b876e 27069+ unsigned long long n, sz;
7f207e10
AM
27070+
27071+ array = NULL;
27072+ n = 0;
27073+ if (!*hint)
27074+ goto out;
27075+
27076+ if (*hint > ULLONG_MAX / sizeof(array)) {
27077+ array = ERR_PTR(-EMFILE);
27078+ pr_err("hint %llu\n", *hint);
27079+ goto out;
27080+ }
27081+
076b876e
AM
27082+ sz = sizeof(array) * *hint;
27083+ array = kzalloc(sz, GFP_NOFS);
7f207e10 27084+ if (unlikely(!array))
076b876e 27085+ array = vzalloc(sz);
7f207e10
AM
27086+ if (unlikely(!array)) {
27087+ array = ERR_PTR(-ENOMEM);
27088+ goto out;
27089+ }
27090+
27091+ n = cb(array, *hint, arg);
27092+ AuDebugOn(n > *hint);
27093+
27094+out:
27095+ *hint = n;
27096+ return array;
27097+}
27098+
27099+static unsigned long long au_iarray_cb(void *a,
27100+ unsigned long long max __maybe_unused,
27101+ void *arg)
27102+{
27103+ unsigned long long n;
27104+ struct inode **p, *inode;
27105+ struct list_head *head;
27106+
27107+ n = 0;
27108+ p = a;
27109+ head = arg;
2cbb1c4b 27110+ spin_lock(&inode_sb_list_lock);
7f207e10
AM
27111+ list_for_each_entry(inode, head, i_sb_list) {
27112+ if (!is_bad_inode(inode)
27113+ && au_ii(inode)->ii_bstart >= 0) {
2cbb1c4b
JR
27114+ spin_lock(&inode->i_lock);
27115+ if (atomic_read(&inode->i_count)) {
27116+ au_igrab(inode);
27117+ *p++ = inode;
27118+ n++;
27119+ AuDebugOn(n > max);
27120+ }
27121+ spin_unlock(&inode->i_lock);
7f207e10
AM
27122+ }
27123+ }
2cbb1c4b 27124+ spin_unlock(&inode_sb_list_lock);
7f207e10
AM
27125+
27126+ return n;
27127+}
27128+
27129+struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
27130+{
27131+ *max = atomic_long_read(&au_sbi(sb)->si_ninodes);
27132+ return au_array_alloc(max, au_iarray_cb, &sb->s_inodes);
27133+}
27134+
27135+void au_iarray_free(struct inode **a, unsigned long long max)
27136+{
27137+ unsigned long long ull;
27138+
27139+ for (ull = 0; ull < max; ull++)
27140+ iput(a[ull]);
27141+ au_array_free(a);
27142+}
27143+
27144+/* ---------------------------------------------------------------------- */
27145+
1facf9fc 27146+/*
27147+ * refresh dentry and inode at remount time.
27148+ */
027c5e7a
AM
27149+/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
27150+static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
27151+ struct dentry *parent)
1facf9fc 27152+{
27153+ int err;
1facf9fc 27154+
27155+ di_write_lock_child(dentry);
1facf9fc 27156+ di_read_lock_parent(parent, AuLock_IR);
027c5e7a
AM
27157+ err = au_refresh_dentry(dentry, parent);
27158+ if (!err && dir_flags)
5527c038 27159+ au_hn_reset(d_inode(dentry), dir_flags);
1facf9fc 27160+ di_read_unlock(parent, AuLock_IR);
1facf9fc 27161+ di_write_unlock(dentry);
27162+
27163+ return err;
27164+}
27165+
027c5e7a
AM
27166+static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
27167+ struct au_sbinfo *sbinfo,
27168+ const unsigned int dir_flags)
1facf9fc 27169+{
027c5e7a
AM
27170+ int err;
27171+ struct dentry *parent;
027c5e7a
AM
27172+
27173+ err = 0;
27174+ parent = dget_parent(dentry);
27175+ if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
5527c038
JR
27176+ if (d_really_is_positive(dentry)) {
27177+ if (!d_is_dir(dentry))
027c5e7a
AM
27178+ err = au_do_refresh(dentry, /*dir_flags*/0,
27179+ parent);
27180+ else {
27181+ err = au_do_refresh(dentry, dir_flags, parent);
27182+ if (unlikely(err))
27183+ au_fset_si(sbinfo, FAILED_REFRESH_DIR);
27184+ }
27185+ } else
27186+ err = au_do_refresh(dentry, /*dir_flags*/0, parent);
27187+ AuDbgDentry(dentry);
27188+ }
27189+ dput(parent);
27190+
27191+ AuTraceErr(err);
27192+ return err;
1facf9fc 27193+}
27194+
027c5e7a 27195+static int au_refresh_d(struct super_block *sb)
1facf9fc 27196+{
27197+ int err, i, j, ndentry, e;
027c5e7a 27198+ unsigned int sigen;
1facf9fc 27199+ struct au_dcsub_pages dpages;
27200+ struct au_dpage *dpage;
027c5e7a
AM
27201+ struct dentry **dentries, *d;
27202+ struct au_sbinfo *sbinfo;
27203+ struct dentry *root = sb->s_root;
5527c038 27204+ const unsigned int dir_flags = au_hi_flags(d_inode(root), /*isdir*/1);
1facf9fc 27205+
027c5e7a
AM
27206+ err = au_dpages_init(&dpages, GFP_NOFS);
27207+ if (unlikely(err))
1facf9fc 27208+ goto out;
027c5e7a
AM
27209+ err = au_dcsub_pages(&dpages, root, NULL, NULL);
27210+ if (unlikely(err))
1facf9fc 27211+ goto out_dpages;
1facf9fc 27212+
027c5e7a
AM
27213+ sigen = au_sigen(sb);
27214+ sbinfo = au_sbi(sb);
27215+ for (i = 0; i < dpages.ndpage; i++) {
1facf9fc 27216+ dpage = dpages.dpages + i;
27217+ dentries = dpage->dentries;
27218+ ndentry = dpage->ndentry;
027c5e7a 27219+ for (j = 0; j < ndentry; j++) {
1facf9fc 27220+ d = dentries[j];
027c5e7a
AM
27221+ e = au_do_refresh_d(d, sigen, sbinfo, dir_flags);
27222+ if (unlikely(e && !err))
27223+ err = e;
27224+ /* go on even err */
1facf9fc 27225+ }
27226+ }
27227+
4f0767ce 27228+out_dpages:
1facf9fc 27229+ au_dpages_free(&dpages);
4f0767ce 27230+out:
1facf9fc 27231+ return err;
27232+}
27233+
027c5e7a 27234+static int au_refresh_i(struct super_block *sb)
1facf9fc 27235+{
027c5e7a
AM
27236+ int err, e;
27237+ unsigned int sigen;
27238+ unsigned long long max, ull;
27239+ struct inode *inode, **array;
1facf9fc 27240+
027c5e7a
AM
27241+ array = au_iarray_alloc(sb, &max);
27242+ err = PTR_ERR(array);
27243+ if (IS_ERR(array))
27244+ goto out;
1facf9fc 27245+
27246+ err = 0;
027c5e7a
AM
27247+ sigen = au_sigen(sb);
27248+ for (ull = 0; ull < max; ull++) {
27249+ inode = array[ull];
076b876e
AM
27250+ if (unlikely(!inode))
27251+ break;
537831f9 27252+ if (au_iigen(inode, NULL) != sigen) {
1facf9fc 27253+ ii_write_lock_child(inode);
027c5e7a 27254+ e = au_refresh_hinode_self(inode);
1facf9fc 27255+ ii_write_unlock(inode);
27256+ if (unlikely(e)) {
027c5e7a 27257+ pr_err("error %d, i%lu\n", e, inode->i_ino);
1facf9fc 27258+ if (!err)
27259+ err = e;
27260+ /* go on even if err */
27261+ }
27262+ }
1facf9fc 27263+ }
27264+
027c5e7a 27265+ au_iarray_free(array, max);
1facf9fc 27266+
4f0767ce 27267+out:
1facf9fc 27268+ return err;
27269+}
27270+
027c5e7a 27271+static void au_remount_refresh(struct super_block *sb)
1facf9fc 27272+{
027c5e7a
AM
27273+ int err, e;
27274+ unsigned int udba;
27275+ aufs_bindex_t bindex, bend;
1facf9fc 27276+ struct dentry *root;
27277+ struct inode *inode;
027c5e7a 27278+ struct au_branch *br;
1facf9fc 27279+
27280+ au_sigen_inc(sb);
027c5e7a 27281+ au_fclr_si(au_sbi(sb), FAILED_REFRESH_DIR);
1facf9fc 27282+
27283+ root = sb->s_root;
27284+ DiMustNoWaiters(root);
5527c038 27285+ inode = d_inode(root);
1facf9fc 27286+ IiMustNoWaiters(inode);
1facf9fc 27287+
027c5e7a
AM
27288+ udba = au_opt_udba(sb);
27289+ bend = au_sbend(sb);
27290+ for (bindex = 0; bindex <= bend; bindex++) {
27291+ br = au_sbr(sb, bindex);
27292+ err = au_hnotify_reset_br(udba, br, br->br_perm);
1facf9fc 27293+ if (unlikely(err))
027c5e7a
AM
27294+ AuIOErr("hnotify failed on br %d, %d, ignored\n",
27295+ bindex, err);
27296+ /* go on even if err */
1facf9fc 27297+ }
027c5e7a 27298+ au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
1facf9fc 27299+
027c5e7a
AM
27300+ di_write_unlock(root);
27301+ err = au_refresh_d(sb);
27302+ e = au_refresh_i(sb);
27303+ if (unlikely(e && !err))
27304+ err = e;
1facf9fc 27305+ /* aufs_write_lock() calls ..._child() */
27306+ di_write_lock_child(root);
027c5e7a
AM
27307+
27308+ au_cpup_attr_all(inode, /*force*/1);
27309+
27310+ if (unlikely(err))
27311+ AuIOErr("refresh failed, ignored, %d\n", err);
1facf9fc 27312+}
27313+
27314+/* stop extra interpretation of errno in mount(8), and strange error messages */
27315+static int cvt_err(int err)
27316+{
27317+ AuTraceErr(err);
27318+
27319+ switch (err) {
27320+ case -ENOENT:
27321+ case -ENOTDIR:
27322+ case -EEXIST:
27323+ case -EIO:
27324+ err = -EINVAL;
27325+ }
27326+ return err;
27327+}
27328+
27329+static int aufs_remount_fs(struct super_block *sb, int *flags, char *data)
27330+{
4a4d8108
AM
27331+ int err, do_dx;
27332+ unsigned int mntflags;
1facf9fc 27333+ struct au_opts opts;
27334+ struct dentry *root;
27335+ struct inode *inode;
27336+ struct au_sbinfo *sbinfo;
27337+
27338+ err = 0;
27339+ root = sb->s_root;
27340+ if (!data || !*data) {
e49829fe
JR
27341+ err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
27342+ if (!err) {
27343+ di_write_lock_child(root);
27344+ err = au_opts_verify(sb, *flags, /*pending*/0);
27345+ aufs_write_unlock(root);
27346+ }
1facf9fc 27347+ goto out;
27348+ }
27349+
27350+ err = -ENOMEM;
27351+ memset(&opts, 0, sizeof(opts));
27352+ opts.opt = (void *)__get_free_page(GFP_NOFS);
27353+ if (unlikely(!opts.opt))
27354+ goto out;
27355+ opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
27356+ opts.flags = AuOpts_REMOUNT;
27357+ opts.sb_flags = *flags;
27358+
27359+ /* parse it before aufs lock */
27360+ err = au_opts_parse(sb, data, &opts);
27361+ if (unlikely(err))
27362+ goto out_opts;
27363+
27364+ sbinfo = au_sbi(sb);
5527c038 27365+ inode = d_inode(root);
1facf9fc 27366+ mutex_lock(&inode->i_mutex);
e49829fe
JR
27367+ err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
27368+ if (unlikely(err))
27369+ goto out_mtx;
27370+ di_write_lock_child(root);
1facf9fc 27371+
27372+ /* au_opts_remount() may return an error */
27373+ err = au_opts_remount(sb, &opts);
27374+ au_opts_free(&opts);
27375+
027c5e7a
AM
27376+ if (au_ftest_opts(opts.flags, REFRESH))
27377+ au_remount_refresh(sb);
1facf9fc 27378+
4a4d8108
AM
27379+ if (au_ftest_opts(opts.flags, REFRESH_DYAOP)) {
27380+ mntflags = au_mntflags(sb);
27381+ do_dx = !!au_opt_test(mntflags, DIO);
27382+ au_dy_arefresh(do_dx);
27383+ }
27384+
076b876e 27385+ au_fhsm_wrote_all(sb, /*force*/1); /* ?? */
1facf9fc 27386+ aufs_write_unlock(root);
953406b4 27387+
e49829fe
JR
27388+out_mtx:
27389+ mutex_unlock(&inode->i_mutex);
4f0767ce 27390+out_opts:
1facf9fc 27391+ free_page((unsigned long)opts.opt);
4f0767ce 27392+out:
1facf9fc 27393+ err = cvt_err(err);
27394+ AuTraceErr(err);
27395+ return err;
27396+}
27397+
4a4d8108 27398+static const struct super_operations aufs_sop = {
1facf9fc 27399+ .alloc_inode = aufs_alloc_inode,
27400+ .destroy_inode = aufs_destroy_inode,
b752ccd1 27401+ /* always deleting, no clearing */
1facf9fc 27402+ .drop_inode = generic_delete_inode,
27403+ .show_options = aufs_show_options,
27404+ .statfs = aufs_statfs,
27405+ .put_super = aufs_put_super,
537831f9 27406+ .sync_fs = aufs_sync_fs,
1facf9fc 27407+ .remount_fs = aufs_remount_fs
27408+};
27409+
27410+/* ---------------------------------------------------------------------- */
27411+
27412+static int alloc_root(struct super_block *sb)
27413+{
27414+ int err;
27415+ struct inode *inode;
27416+ struct dentry *root;
27417+
27418+ err = -ENOMEM;
27419+ inode = au_iget_locked(sb, AUFS_ROOT_INO);
27420+ err = PTR_ERR(inode);
27421+ if (IS_ERR(inode))
27422+ goto out;
27423+
27424+ inode->i_op = &aufs_dir_iop;
27425+ inode->i_fop = &aufs_dir_fop;
27426+ inode->i_mode = S_IFDIR;
9dbd164d 27427+ set_nlink(inode, 2);
1facf9fc 27428+ unlock_new_inode(inode);
27429+
92d182d2 27430+ root = d_make_root(inode);
1facf9fc 27431+ if (unlikely(!root))
92d182d2 27432+ goto out;
1facf9fc 27433+ err = PTR_ERR(root);
27434+ if (IS_ERR(root))
92d182d2 27435+ goto out;
1facf9fc 27436+
4a4d8108 27437+ err = au_di_init(root);
1facf9fc 27438+ if (!err) {
27439+ sb->s_root = root;
27440+ return 0; /* success */
27441+ }
27442+ dput(root);
1facf9fc 27443+
4f0767ce 27444+out:
1facf9fc 27445+ return err;
1facf9fc 27446+}
27447+
27448+static int aufs_fill_super(struct super_block *sb, void *raw_data,
27449+ int silent __maybe_unused)
27450+{
27451+ int err;
27452+ struct au_opts opts;
27453+ struct dentry *root;
27454+ struct inode *inode;
27455+ char *arg = raw_data;
27456+
27457+ if (unlikely(!arg || !*arg)) {
27458+ err = -EINVAL;
4a4d8108 27459+ pr_err("no arg\n");
1facf9fc 27460+ goto out;
27461+ }
27462+
27463+ err = -ENOMEM;
27464+ memset(&opts, 0, sizeof(opts));
27465+ opts.opt = (void *)__get_free_page(GFP_NOFS);
27466+ if (unlikely(!opts.opt))
27467+ goto out;
27468+ opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
27469+ opts.sb_flags = sb->s_flags;
27470+
27471+ err = au_si_alloc(sb);
27472+ if (unlikely(err))
27473+ goto out_opts;
27474+
27475+ /* all timestamps always follow the ones on the branch */
27476+ sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
27477+ sb->s_op = &aufs_sop;
027c5e7a 27478+ sb->s_d_op = &aufs_dop;
1facf9fc 27479+ sb->s_magic = AUFS_SUPER_MAGIC;
27480+ sb->s_maxbytes = 0;
c1595e42 27481+ sb->s_stack_depth = 1;
1facf9fc 27482+ au_export_init(sb);
c1595e42 27483+ /* au_xattr_init(sb); */
1facf9fc 27484+
27485+ err = alloc_root(sb);
27486+ if (unlikely(err)) {
27487+ si_write_unlock(sb);
27488+ goto out_info;
27489+ }
27490+ root = sb->s_root;
5527c038 27491+ inode = d_inode(root);
1facf9fc 27492+
27493+ /*
27494+ * actually we can parse options regardless aufs lock here.
27495+ * but at remount time, parsing must be done before aufs lock.
27496+ * so we follow the same rule.
27497+ */
27498+ ii_write_lock_parent(inode);
27499+ aufs_write_unlock(root);
27500+ err = au_opts_parse(sb, arg, &opts);
27501+ if (unlikely(err))
27502+ goto out_root;
27503+
27504+ /* lock vfs_inode first, then aufs. */
27505+ mutex_lock(&inode->i_mutex);
1facf9fc 27506+ aufs_write_lock(root);
27507+ err = au_opts_mount(sb, &opts);
27508+ au_opts_free(&opts);
1facf9fc 27509+ aufs_write_unlock(root);
27510+ mutex_unlock(&inode->i_mutex);
4a4d8108
AM
27511+ if (!err)
27512+ goto out_opts; /* success */
1facf9fc 27513+
4f0767ce 27514+out_root:
1facf9fc 27515+ dput(root);
27516+ sb->s_root = NULL;
4f0767ce 27517+out_info:
2cbb1c4b 27518+ dbgaufs_si_fin(au_sbi(sb));
1facf9fc 27519+ kobject_put(&au_sbi(sb)->si_kobj);
27520+ sb->s_fs_info = NULL;
4f0767ce 27521+out_opts:
1facf9fc 27522+ free_page((unsigned long)opts.opt);
4f0767ce 27523+out:
1facf9fc 27524+ AuTraceErr(err);
27525+ err = cvt_err(err);
27526+ AuTraceErr(err);
27527+ return err;
27528+}
27529+
27530+/* ---------------------------------------------------------------------- */
27531+
027c5e7a
AM
27532+static struct dentry *aufs_mount(struct file_system_type *fs_type, int flags,
27533+ const char *dev_name __maybe_unused,
27534+ void *raw_data)
1facf9fc 27535+{
027c5e7a 27536+ struct dentry *root;
1facf9fc 27537+ struct super_block *sb;
27538+
27539+ /* all timestamps always follow the ones on the branch */
27540+ /* mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; */
027c5e7a
AM
27541+ root = mount_nodev(fs_type, flags, raw_data, aufs_fill_super);
27542+ if (IS_ERR(root))
27543+ goto out;
27544+
27545+ sb = root->d_sb;
27546+ si_write_lock(sb, !AuLock_FLUSH);
27547+ sysaufs_brs_add(sb, 0);
27548+ si_write_unlock(sb);
27549+ au_sbilist_add(sb);
27550+
27551+out:
27552+ return root;
1facf9fc 27553+}
27554+
e49829fe
JR
27555+static void aufs_kill_sb(struct super_block *sb)
27556+{
27557+ struct au_sbinfo *sbinfo;
27558+
27559+ sbinfo = au_sbi(sb);
27560+ if (sbinfo) {
27561+ au_sbilist_del(sb);
27562+ aufs_write_lock(sb->s_root);
076b876e 27563+ au_fhsm_fin(sb);
e49829fe
JR
27564+ if (sbinfo->si_wbr_create_ops->fin)
27565+ sbinfo->si_wbr_create_ops->fin(sb);
27566+ if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
27567+ au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
027c5e7a 27568+ au_remount_refresh(sb);
e49829fe
JR
27569+ }
27570+ if (au_opt_test(sbinfo->si_mntflags, PLINK))
27571+ au_plink_put(sb, /*verbose*/1);
27572+ au_xino_clr(sb);
1e00d052 27573+ sbinfo->si_sb = NULL;
e49829fe 27574+ aufs_write_unlock(sb->s_root);
e49829fe
JR
27575+ au_nwt_flush(&sbinfo->si_nowait);
27576+ }
98d9a5b1 27577+ kill_anon_super(sb);
e49829fe
JR
27578+}
27579+
1facf9fc 27580+struct file_system_type aufs_fs_type = {
27581+ .name = AUFS_FSTYPE,
c06a8ce3
AM
27582+ /* a race between rename and others */
27583+ .fs_flags = FS_RENAME_DOES_D_MOVE,
027c5e7a 27584+ .mount = aufs_mount,
e49829fe 27585+ .kill_sb = aufs_kill_sb,
1facf9fc 27586+ /* no need to __module_get() and module_put(). */
27587+ .owner = THIS_MODULE,
27588+};
7f207e10
AM
27589diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
27590--- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c
AM
27591+++ linux/fs/aufs/super.h 2015-09-24 10:47:58.254719746 +0200
27592@@ -0,0 +1,638 @@
1facf9fc 27593+/*
2000de60 27594+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 27595+ *
27596+ * This program, aufs is free software; you can redistribute it and/or modify
27597+ * it under the terms of the GNU General Public License as published by
27598+ * the Free Software Foundation; either version 2 of the License, or
27599+ * (at your option) any later version.
dece6358
AM
27600+ *
27601+ * This program is distributed in the hope that it will be useful,
27602+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
27603+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27604+ * GNU General Public License for more details.
27605+ *
27606+ * You should have received a copy of the GNU General Public License
523b37e3 27607+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 27608+ */
27609+
27610+/*
27611+ * super_block operations
27612+ */
27613+
27614+#ifndef __AUFS_SUPER_H__
27615+#define __AUFS_SUPER_H__
27616+
27617+#ifdef __KERNEL__
27618+
27619+#include <linux/fs.h>
5527c038 27620+#include <linux/kobject.h>
1facf9fc 27621+#include "rwsem.h"
27622+#include "spl.h"
27623+#include "wkq.h"
27624+
1facf9fc 27625+/* policies to select one among multiple writable branches */
27626+struct au_wbr_copyup_operations {
27627+ int (*copyup)(struct dentry *dentry);
27628+};
27629+
392086de
AM
27630+#define AuWbr_DIR 1 /* target is a dir */
27631+#define AuWbr_PARENT (1 << 1) /* always require a parent */
27632+
27633+#define au_ftest_wbr(flags, name) ((flags) & AuWbr_##name)
27634+#define au_fset_wbr(flags, name) { (flags) |= AuWbr_##name; }
27635+#define au_fclr_wbr(flags, name) { (flags) &= ~AuWbr_##name; }
27636+
1facf9fc 27637+struct au_wbr_create_operations {
392086de 27638+ int (*create)(struct dentry *dentry, unsigned int flags);
1facf9fc 27639+ int (*init)(struct super_block *sb);
27640+ int (*fin)(struct super_block *sb);
27641+};
27642+
27643+struct au_wbr_mfs {
27644+ struct mutex mfs_lock; /* protect this structure */
27645+ unsigned long mfs_jiffy;
27646+ unsigned long mfs_expire;
27647+ aufs_bindex_t mfs_bindex;
27648+
27649+ unsigned long long mfsrr_bytes;
27650+ unsigned long long mfsrr_watermark;
27651+};
27652+
86dc4139
AM
27653+struct pseudo_link {
27654+ union {
27655+ struct hlist_node hlist;
27656+ struct rcu_head rcu;
27657+ };
27658+ struct inode *inode;
27659+};
27660+
27661+#define AuPlink_NHASH 100
27662+static inline int au_plink_hash(ino_t ino)
27663+{
27664+ return ino % AuPlink_NHASH;
27665+}
27666+
076b876e
AM
27667+/* File-based Hierarchical Storage Management */
27668+struct au_fhsm {
27669+#ifdef CONFIG_AUFS_FHSM
27670+ /* allow only one process who can receive the notification */
27671+ spinlock_t fhsm_spin;
27672+ pid_t fhsm_pid;
27673+ wait_queue_head_t fhsm_wqh;
27674+ atomic_t fhsm_readable;
27675+
c1595e42 27676+ /* these are protected by si_rwsem */
076b876e 27677+ unsigned long fhsm_expire;
c1595e42 27678+ aufs_bindex_t fhsm_bottom;
076b876e
AM
27679+#endif
27680+};
27681+
1facf9fc 27682+struct au_branch;
27683+struct au_sbinfo {
27684+ /* nowait tasks in the system-wide workqueue */
27685+ struct au_nowait_tasks si_nowait;
27686+
b752ccd1
AM
27687+ /*
27688+ * tried sb->s_umount, but failed due to the dependecy between i_mutex.
27689+ * rwsem for au_sbinfo is necessary.
27690+ */
dece6358 27691+ struct au_rwsem si_rwsem;
1facf9fc 27692+
b752ccd1
AM
27693+ /* prevent recursive locking in deleting inode */
27694+ struct {
27695+ unsigned long *bitmap;
27696+ spinlock_t tree_lock;
27697+ struct radix_tree_root tree;
27698+ } au_si_pid;
27699+
7f207e10 27700+ /*
523b37e3
AM
27701+ * dirty approach to protect sb->sb_inodes and ->s_files (gone) from
27702+ * remount.
7f207e10
AM
27703+ */
27704+ atomic_long_t si_ninodes, si_nfiles;
27705+
1facf9fc 27706+ /* branch management */
27707+ unsigned int si_generation;
27708+
2000de60 27709+ /* see AuSi_ flags */
1facf9fc 27710+ unsigned char au_si_status;
27711+
27712+ aufs_bindex_t si_bend;
7f207e10
AM
27713+
27714+ /* dirty trick to keep br_id plus */
27715+ unsigned int si_last_br_id :
27716+ sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
1facf9fc 27717+ struct au_branch **si_branch;
27718+
27719+ /* policy to select a writable branch */
27720+ unsigned char si_wbr_copyup;
27721+ unsigned char si_wbr_create;
27722+ struct au_wbr_copyup_operations *si_wbr_copyup_ops;
27723+ struct au_wbr_create_operations *si_wbr_create_ops;
27724+
27725+ /* round robin */
27726+ atomic_t si_wbr_rr_next;
27727+
27728+ /* most free space */
27729+ struct au_wbr_mfs si_wbr_mfs;
27730+
076b876e
AM
27731+ /* File-based Hierarchical Storage Management */
27732+ struct au_fhsm si_fhsm;
27733+
1facf9fc 27734+ /* mount flags */
27735+ /* include/asm-ia64/siginfo.h defines a macro named si_flags */
27736+ unsigned int si_mntflags;
27737+
c2c0f25c
AM
27738+ /* symlink to follow_link() and put_link() */
27739+ struct au_sphlhead si_symlink;
27740+
1facf9fc 27741+ /* external inode number (bitmap and translation table) */
5527c038
JR
27742+ vfs_readf_t si_xread;
27743+ vfs_writef_t si_xwrite;
1facf9fc 27744+ struct file *si_xib;
27745+ struct mutex si_xib_mtx; /* protect xib members */
27746+ unsigned long *si_xib_buf;
27747+ unsigned long si_xib_last_pindex;
27748+ int si_xib_next_bit;
27749+ aufs_bindex_t si_xino_brid;
392086de
AM
27750+ unsigned long si_xino_jiffy;
27751+ unsigned long si_xino_expire;
1facf9fc 27752+ /* reserved for future use */
27753+ /* unsigned long long si_xib_limit; */ /* Max xib file size */
27754+
27755+#ifdef CONFIG_AUFS_EXPORT
27756+ /* i_generation */
27757+ struct file *si_xigen;
27758+ atomic_t si_xigen_next;
27759+#endif
27760+
b912730e
AM
27761+ /* dirty trick to suppoer atomic_open */
27762+ struct au_sphlhead si_aopen;
27763+
1facf9fc 27764+ /* vdir parameters */
e49829fe 27765+ unsigned long si_rdcache; /* max cache time in jiffies */
1facf9fc 27766+ unsigned int si_rdblk; /* deblk size */
27767+ unsigned int si_rdhash; /* hash size */
27768+
27769+ /*
27770+ * If the number of whiteouts are larger than si_dirwh, leave all of
27771+ * them after au_whtmp_ren to reduce the cost of rmdir(2).
27772+ * future fsck.aufs or kernel thread will remove them later.
27773+ * Otherwise, remove all whiteouts and the dir in rmdir(2).
27774+ */
27775+ unsigned int si_dirwh;
27776+
1facf9fc 27777+ /* pseudo_link list */
86dc4139 27778+ struct au_sphlhead si_plink[AuPlink_NHASH];
1facf9fc 27779+ wait_queue_head_t si_plink_wq;
4a4d8108 27780+ spinlock_t si_plink_maint_lock;
e49829fe 27781+ pid_t si_plink_maint_pid;
1facf9fc 27782+
523b37e3
AM
27783+ /* file list */
27784+ struct au_sphlhead si_files;
27785+
1facf9fc 27786+ /*
27787+ * sysfs and lifetime management.
27788+ * this is not a small structure and it may be a waste of memory in case
27789+ * of sysfs is disabled, particulary when many aufs-es are mounted.
27790+ * but using sysfs is majority.
27791+ */
27792+ struct kobject si_kobj;
27793+#ifdef CONFIG_DEBUG_FS
86dc4139
AM
27794+ struct dentry *si_dbgaufs;
27795+ struct dentry *si_dbgaufs_plink;
27796+ struct dentry *si_dbgaufs_xib;
1facf9fc 27797+#ifdef CONFIG_AUFS_EXPORT
27798+ struct dentry *si_dbgaufs_xigen;
27799+#endif
27800+#endif
27801+
e49829fe
JR
27802+#ifdef CONFIG_AUFS_SBILIST
27803+ struct list_head si_list;
27804+#endif
27805+
1facf9fc 27806+ /* dirty, necessary for unmounting, sysfs and sysrq */
27807+ struct super_block *si_sb;
27808+};
27809+
dece6358
AM
27810+/* sbinfo status flags */
27811+/*
27812+ * set true when refresh_dirs() failed at remount time.
27813+ * then try refreshing dirs at access time again.
27814+ * if it is false, refreshing dirs at access time is unnecesary
27815+ */
027c5e7a 27816+#define AuSi_FAILED_REFRESH_DIR 1
076b876e
AM
27817+
27818+#define AuSi_FHSM (1 << 1) /* fhsm is active now */
27819+
27820+#ifndef CONFIG_AUFS_FHSM
27821+#undef AuSi_FHSM
27822+#define AuSi_FHSM 0
27823+#endif
27824+
dece6358
AM
27825+static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
27826+ unsigned int flag)
27827+{
27828+ AuRwMustAnyLock(&sbi->si_rwsem);
27829+ return sbi->au_si_status & flag;
27830+}
27831+#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
27832+#define au_fset_si(sbinfo, name) do { \
27833+ AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
27834+ (sbinfo)->au_si_status |= AuSi_##name; \
27835+} while (0)
27836+#define au_fclr_si(sbinfo, name) do { \
27837+ AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
27838+ (sbinfo)->au_si_status &= ~AuSi_##name; \
27839+} while (0)
27840+
1facf9fc 27841+/* ---------------------------------------------------------------------- */
27842+
27843+/* policy to select one among writable branches */
4a4d8108
AM
27844+#define AuWbrCopyup(sbinfo, ...) \
27845+ ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
27846+#define AuWbrCreate(sbinfo, ...) \
27847+ ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
1facf9fc 27848+
27849+/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
27850+#define AuLock_DW 1 /* write-lock dentry */
27851+#define AuLock_IR (1 << 1) /* read-lock inode */
27852+#define AuLock_IW (1 << 2) /* write-lock inode */
27853+#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
27854+#define AuLock_DIR (1 << 4) /* target is a dir */
e49829fe
JR
27855+#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
27856+#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
027c5e7a 27857+#define AuLock_GEN (1 << 7) /* test digen/iigen */
1facf9fc 27858+#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
7f207e10
AM
27859+#define au_fset_lock(flags, name) \
27860+ do { (flags) |= AuLock_##name; } while (0)
27861+#define au_fclr_lock(flags, name) \
27862+ do { (flags) &= ~AuLock_##name; } while (0)
1facf9fc 27863+
27864+/* ---------------------------------------------------------------------- */
27865+
27866+/* super.c */
27867+extern struct file_system_type aufs_fs_type;
27868+struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
7f207e10
AM
27869+typedef unsigned long long (*au_arraycb_t)(void *array, unsigned long long max,
27870+ void *arg);
27871+void au_array_free(void *array);
27872+void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb, void *arg);
27873+struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
27874+void au_iarray_free(struct inode **a, unsigned long long max);
1facf9fc 27875+
27876+/* sbinfo.c */
27877+void au_si_free(struct kobject *kobj);
27878+int au_si_alloc(struct super_block *sb);
27879+int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr);
27880+
27881+unsigned int au_sigen_inc(struct super_block *sb);
27882+aufs_bindex_t au_new_br_id(struct super_block *sb);
27883+
e49829fe
JR
27884+int si_read_lock(struct super_block *sb, int flags);
27885+int si_write_lock(struct super_block *sb, int flags);
27886+int aufs_read_lock(struct dentry *dentry, int flags);
1facf9fc 27887+void aufs_read_unlock(struct dentry *dentry, int flags);
27888+void aufs_write_lock(struct dentry *dentry);
27889+void aufs_write_unlock(struct dentry *dentry);
e49829fe 27890+int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
1facf9fc 27891+void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
27892+
b752ccd1
AM
27893+int si_pid_test_slow(struct super_block *sb);
27894+void si_pid_set_slow(struct super_block *sb);
27895+void si_pid_clr_slow(struct super_block *sb);
27896+
1facf9fc 27897+/* wbr_policy.c */
27898+extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
27899+extern struct au_wbr_create_operations au_wbr_create_ops[];
27900+int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
c2b27bf2 27901+int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex);
076b876e 27902+int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t bstart);
c2b27bf2
AM
27903+
27904+/* mvdown.c */
27905+int au_mvdown(struct dentry *dentry, struct aufs_mvdown __user *arg);
1facf9fc 27906+
076b876e
AM
27907+#ifdef CONFIG_AUFS_FHSM
27908+/* fhsm.c */
27909+
27910+static inline pid_t au_fhsm_pid(struct au_fhsm *fhsm)
27911+{
27912+ pid_t pid;
27913+
27914+ spin_lock(&fhsm->fhsm_spin);
27915+ pid = fhsm->fhsm_pid;
27916+ spin_unlock(&fhsm->fhsm_spin);
27917+
27918+ return pid;
27919+}
27920+
27921+void au_fhsm_wrote(struct super_block *sb, aufs_bindex_t bindex, int force);
27922+void au_fhsm_wrote_all(struct super_block *sb, int force);
27923+int au_fhsm_fd(struct super_block *sb, int oflags);
27924+int au_fhsm_br_alloc(struct au_branch *br);
c1595e42 27925+void au_fhsm_set_bottom(struct super_block *sb, aufs_bindex_t bindex);
076b876e
AM
27926+void au_fhsm_fin(struct super_block *sb);
27927+void au_fhsm_init(struct au_sbinfo *sbinfo);
27928+void au_fhsm_set(struct au_sbinfo *sbinfo, unsigned int sec);
27929+void au_fhsm_show(struct seq_file *seq, struct au_sbinfo *sbinfo);
27930+#else
27931+AuStubVoid(au_fhsm_wrote, struct super_block *sb, aufs_bindex_t bindex,
27932+ int force)
27933+AuStubVoid(au_fhsm_wrote_all, struct super_block *sb, int force)
27934+AuStub(int, au_fhsm_fd, return -EOPNOTSUPP, struct super_block *sb, int oflags)
c1595e42
JR
27935+AuStub(pid_t, au_fhsm_pid, return 0, struct au_fhsm *fhsm)
27936+AuStubInt0(au_fhsm_br_alloc, struct au_branch *br)
27937+AuStubVoid(au_fhsm_set_bottom, struct super_block *sb, aufs_bindex_t bindex)
076b876e
AM
27938+AuStubVoid(au_fhsm_fin, struct super_block *sb)
27939+AuStubVoid(au_fhsm_init, struct au_sbinfo *sbinfo)
27940+AuStubVoid(au_fhsm_set, struct au_sbinfo *sbinfo, unsigned int sec)
27941+AuStubVoid(au_fhsm_show, struct seq_file *seq, struct au_sbinfo *sbinfo)
27942+#endif
27943+
1facf9fc 27944+/* ---------------------------------------------------------------------- */
27945+
27946+static inline struct au_sbinfo *au_sbi(struct super_block *sb)
27947+{
27948+ return sb->s_fs_info;
27949+}
27950+
27951+/* ---------------------------------------------------------------------- */
27952+
27953+#ifdef CONFIG_AUFS_EXPORT
a2a7ad62 27954+int au_test_nfsd(void);
1facf9fc 27955+void au_export_init(struct super_block *sb);
b752ccd1 27956+void au_xigen_inc(struct inode *inode);
1facf9fc 27957+int au_xigen_new(struct inode *inode);
27958+int au_xigen_set(struct super_block *sb, struct file *base);
27959+void au_xigen_clr(struct super_block *sb);
27960+
27961+static inline int au_busy_or_stale(void)
27962+{
b752ccd1 27963+ if (!au_test_nfsd())
1facf9fc 27964+ return -EBUSY;
27965+ return -ESTALE;
27966+}
27967+#else
b752ccd1 27968+AuStubInt0(au_test_nfsd, void)
a2a7ad62 27969+AuStubVoid(au_export_init, struct super_block *sb)
b752ccd1 27970+AuStubVoid(au_xigen_inc, struct inode *inode)
4a4d8108
AM
27971+AuStubInt0(au_xigen_new, struct inode *inode)
27972+AuStubInt0(au_xigen_set, struct super_block *sb, struct file *base)
27973+AuStubVoid(au_xigen_clr, struct super_block *sb)
c1595e42 27974+AuStub(int, au_busy_or_stale, return -EBUSY, void)
1facf9fc 27975+#endif /* CONFIG_AUFS_EXPORT */
27976+
27977+/* ---------------------------------------------------------------------- */
27978+
e49829fe
JR
27979+#ifdef CONFIG_AUFS_SBILIST
27980+/* module.c */
27981+extern struct au_splhead au_sbilist;
27982+
27983+static inline void au_sbilist_init(void)
27984+{
27985+ au_spl_init(&au_sbilist);
27986+}
27987+
27988+static inline void au_sbilist_add(struct super_block *sb)
27989+{
27990+ au_spl_add(&au_sbi(sb)->si_list, &au_sbilist);
27991+}
27992+
27993+static inline void au_sbilist_del(struct super_block *sb)
27994+{
27995+ au_spl_del(&au_sbi(sb)->si_list, &au_sbilist);
27996+}
53392da6
AM
27997+
27998+#ifdef CONFIG_AUFS_MAGIC_SYSRQ
27999+static inline void au_sbilist_lock(void)
28000+{
28001+ spin_lock(&au_sbilist.spin);
28002+}
28003+
28004+static inline void au_sbilist_unlock(void)
28005+{
28006+ spin_unlock(&au_sbilist.spin);
28007+}
28008+#define AuGFP_SBILIST GFP_ATOMIC
28009+#else
28010+AuStubVoid(au_sbilist_lock, void)
28011+AuStubVoid(au_sbilist_unlock, void)
28012+#define AuGFP_SBILIST GFP_NOFS
28013+#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
e49829fe
JR
28014+#else
28015+AuStubVoid(au_sbilist_init, void)
c1595e42
JR
28016+AuStubVoid(au_sbilist_add, struct super_block *sb)
28017+AuStubVoid(au_sbilist_del, struct super_block *sb)
53392da6
AM
28018+AuStubVoid(au_sbilist_lock, void)
28019+AuStubVoid(au_sbilist_unlock, void)
28020+#define AuGFP_SBILIST GFP_NOFS
e49829fe
JR
28021+#endif
28022+
28023+/* ---------------------------------------------------------------------- */
28024+
1facf9fc 28025+static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
28026+{
dece6358 28027+ /*
c1595e42 28028+ * This function is a dynamic '__init' function actually,
dece6358
AM
28029+ * so the tiny check for si_rwsem is unnecessary.
28030+ */
28031+ /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
1facf9fc 28032+#ifdef CONFIG_DEBUG_FS
28033+ sbinfo->si_dbgaufs = NULL;
86dc4139 28034+ sbinfo->si_dbgaufs_plink = NULL;
1facf9fc 28035+ sbinfo->si_dbgaufs_xib = NULL;
28036+#ifdef CONFIG_AUFS_EXPORT
28037+ sbinfo->si_dbgaufs_xigen = NULL;
28038+#endif
28039+#endif
28040+}
28041+
28042+/* ---------------------------------------------------------------------- */
28043+
b752ccd1
AM
28044+static inline pid_t si_pid_bit(void)
28045+{
28046+ /* the origin of pid is 1, but the bitmap's is 0 */
28047+ return current->pid - 1;
28048+}
28049+
28050+static inline int si_pid_test(struct super_block *sb)
28051+{
076b876e
AM
28052+ pid_t bit;
28053+
28054+ bit = si_pid_bit();
b752ccd1
AM
28055+ if (bit < PID_MAX_DEFAULT)
28056+ return test_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
c1595e42 28057+ return si_pid_test_slow(sb);
b752ccd1
AM
28058+}
28059+
28060+static inline void si_pid_set(struct super_block *sb)
28061+{
076b876e
AM
28062+ pid_t bit;
28063+
28064+ bit = si_pid_bit();
b752ccd1
AM
28065+ if (bit < PID_MAX_DEFAULT) {
28066+ AuDebugOn(test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
28067+ set_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28068+ /* smp_mb(); */
28069+ } else
28070+ si_pid_set_slow(sb);
28071+}
28072+
28073+static inline void si_pid_clr(struct super_block *sb)
28074+{
076b876e
AM
28075+ pid_t bit;
28076+
28077+ bit = si_pid_bit();
b752ccd1
AM
28078+ if (bit < PID_MAX_DEFAULT) {
28079+ AuDebugOn(!test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
28080+ clear_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
28081+ /* smp_mb(); */
28082+ } else
28083+ si_pid_clr_slow(sb);
28084+}
28085+
28086+/* ---------------------------------------------------------------------- */
28087+
1facf9fc 28088+/* lock superblock. mainly for entry point functions */
28089+/*
b752ccd1
AM
28090+ * __si_read_lock, __si_write_lock,
28091+ * __si_read_unlock, __si_write_unlock, __si_downgrade_lock
1facf9fc 28092+ */
b752ccd1 28093+AuSimpleRwsemFuncs(__si, struct super_block *sb, &au_sbi(sb)->si_rwsem);
1facf9fc 28094+
dece6358
AM
28095+#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
28096+#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
28097+#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
28098+
b752ccd1
AM
28099+static inline void si_noflush_read_lock(struct super_block *sb)
28100+{
28101+ __si_read_lock(sb);
28102+ si_pid_set(sb);
28103+}
28104+
28105+static inline int si_noflush_read_trylock(struct super_block *sb)
28106+{
076b876e
AM
28107+ int locked;
28108+
28109+ locked = __si_read_trylock(sb);
b752ccd1
AM
28110+ if (locked)
28111+ si_pid_set(sb);
28112+ return locked;
28113+}
28114+
28115+static inline void si_noflush_write_lock(struct super_block *sb)
28116+{
28117+ __si_write_lock(sb);
28118+ si_pid_set(sb);
28119+}
28120+
28121+static inline int si_noflush_write_trylock(struct super_block *sb)
28122+{
076b876e
AM
28123+ int locked;
28124+
28125+ locked = __si_write_trylock(sb);
b752ccd1
AM
28126+ if (locked)
28127+ si_pid_set(sb);
28128+ return locked;
28129+}
28130+
7e9cd9fe 28131+#if 0 /* reserved */
1facf9fc 28132+static inline int si_read_trylock(struct super_block *sb, int flags)
28133+{
28134+ if (au_ftest_lock(flags, FLUSH))
28135+ au_nwt_flush(&au_sbi(sb)->si_nowait);
28136+ return si_noflush_read_trylock(sb);
28137+}
e49829fe 28138+#endif
1facf9fc 28139+
b752ccd1
AM
28140+static inline void si_read_unlock(struct super_block *sb)
28141+{
28142+ si_pid_clr(sb);
28143+ __si_read_unlock(sb);
28144+}
28145+
7e9cd9fe 28146+#if 0 /* reserved */
1facf9fc 28147+static inline int si_write_trylock(struct super_block *sb, int flags)
28148+{
28149+ if (au_ftest_lock(flags, FLUSH))
28150+ au_nwt_flush(&au_sbi(sb)->si_nowait);
28151+ return si_noflush_write_trylock(sb);
28152+}
b752ccd1
AM
28153+#endif
28154+
28155+static inline void si_write_unlock(struct super_block *sb)
28156+{
28157+ si_pid_clr(sb);
28158+ __si_write_unlock(sb);
28159+}
28160+
7e9cd9fe 28161+#if 0 /* reserved */
b752ccd1
AM
28162+static inline void si_downgrade_lock(struct super_block *sb)
28163+{
28164+ __si_downgrade_lock(sb);
28165+}
28166+#endif
1facf9fc 28167+
28168+/* ---------------------------------------------------------------------- */
28169+
28170+static inline aufs_bindex_t au_sbend(struct super_block *sb)
28171+{
dece6358 28172+ SiMustAnyLock(sb);
1facf9fc 28173+ return au_sbi(sb)->si_bend;
28174+}
28175+
28176+static inline unsigned int au_mntflags(struct super_block *sb)
28177+{
dece6358 28178+ SiMustAnyLock(sb);
1facf9fc 28179+ return au_sbi(sb)->si_mntflags;
28180+}
28181+
28182+static inline unsigned int au_sigen(struct super_block *sb)
28183+{
dece6358 28184+ SiMustAnyLock(sb);
1facf9fc 28185+ return au_sbi(sb)->si_generation;
28186+}
28187+
7f207e10
AM
28188+static inline void au_ninodes_inc(struct super_block *sb)
28189+{
28190+ atomic_long_inc(&au_sbi(sb)->si_ninodes);
28191+}
28192+
28193+static inline void au_ninodes_dec(struct super_block *sb)
28194+{
28195+ AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_ninodes));
28196+ atomic_long_dec(&au_sbi(sb)->si_ninodes);
28197+}
28198+
28199+static inline void au_nfiles_inc(struct super_block *sb)
28200+{
28201+ atomic_long_inc(&au_sbi(sb)->si_nfiles);
28202+}
28203+
28204+static inline void au_nfiles_dec(struct super_block *sb)
28205+{
28206+ AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_nfiles));
28207+ atomic_long_dec(&au_sbi(sb)->si_nfiles);
28208+}
28209+
1facf9fc 28210+static inline struct au_branch *au_sbr(struct super_block *sb,
28211+ aufs_bindex_t bindex)
28212+{
dece6358 28213+ SiMustAnyLock(sb);
1facf9fc 28214+ return au_sbi(sb)->si_branch[0 + bindex];
28215+}
28216+
28217+static inline void au_xino_brid_set(struct super_block *sb, aufs_bindex_t brid)
28218+{
dece6358 28219+ SiMustWriteLock(sb);
1facf9fc 28220+ au_sbi(sb)->si_xino_brid = brid;
28221+}
28222+
28223+static inline aufs_bindex_t au_xino_brid(struct super_block *sb)
28224+{
dece6358 28225+ SiMustAnyLock(sb);
1facf9fc 28226+ return au_sbi(sb)->si_xino_brid;
28227+}
28228+
28229+#endif /* __KERNEL__ */
28230+#endif /* __AUFS_SUPER_H__ */
7f207e10
AM
28231diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
28232--- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 28233+++ linux/fs/aufs/sysaufs.c 2015-09-24 10:47:58.254719746 +0200
523b37e3 28234@@ -0,0 +1,104 @@
1facf9fc 28235+/*
2000de60 28236+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 28237+ *
28238+ * This program, aufs is free software; you can redistribute it and/or modify
28239+ * it under the terms of the GNU General Public License as published by
28240+ * the Free Software Foundation; either version 2 of the License, or
28241+ * (at your option) any later version.
dece6358
AM
28242+ *
28243+ * This program is distributed in the hope that it will be useful,
28244+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28245+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28246+ * GNU General Public License for more details.
28247+ *
28248+ * You should have received a copy of the GNU General Public License
523b37e3 28249+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 28250+ */
28251+
28252+/*
28253+ * sysfs interface and lifetime management
28254+ * they are necessary regardless sysfs is disabled.
28255+ */
28256+
1facf9fc 28257+#include <linux/random.h>
1facf9fc 28258+#include "aufs.h"
28259+
28260+unsigned long sysaufs_si_mask;
e49829fe 28261+struct kset *sysaufs_kset;
1facf9fc 28262+
28263+#define AuSiAttr(_name) { \
28264+ .attr = { .name = __stringify(_name), .mode = 0444 }, \
28265+ .show = sysaufs_si_##_name, \
28266+}
28267+
28268+static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
28269+struct attribute *sysaufs_si_attrs[] = {
28270+ &sysaufs_si_attr_xi_path.attr,
28271+ NULL,
28272+};
28273+
4a4d8108 28274+static const struct sysfs_ops au_sbi_ops = {
1facf9fc 28275+ .show = sysaufs_si_show
28276+};
28277+
28278+static struct kobj_type au_sbi_ktype = {
28279+ .release = au_si_free,
28280+ .sysfs_ops = &au_sbi_ops,
28281+ .default_attrs = sysaufs_si_attrs
28282+};
28283+
28284+/* ---------------------------------------------------------------------- */
28285+
28286+int sysaufs_si_init(struct au_sbinfo *sbinfo)
28287+{
28288+ int err;
28289+
e49829fe 28290+ sbinfo->si_kobj.kset = sysaufs_kset;
1facf9fc 28291+ /* cf. sysaufs_name() */
28292+ err = kobject_init_and_add
e49829fe 28293+ (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
1facf9fc 28294+ SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
28295+
28296+ dbgaufs_si_null(sbinfo);
28297+ if (!err) {
28298+ err = dbgaufs_si_init(sbinfo);
28299+ if (unlikely(err))
28300+ kobject_put(&sbinfo->si_kobj);
28301+ }
28302+ return err;
28303+}
28304+
28305+void sysaufs_fin(void)
28306+{
28307+ dbgaufs_fin();
e49829fe
JR
28308+ sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
28309+ kset_unregister(sysaufs_kset);
1facf9fc 28310+}
28311+
28312+int __init sysaufs_init(void)
28313+{
28314+ int err;
28315+
28316+ do {
28317+ get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
28318+ } while (!sysaufs_si_mask);
28319+
4a4d8108 28320+ err = -EINVAL;
e49829fe
JR
28321+ sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
28322+ if (unlikely(!sysaufs_kset))
4a4d8108 28323+ goto out;
e49829fe
JR
28324+ err = PTR_ERR(sysaufs_kset);
28325+ if (IS_ERR(sysaufs_kset))
1facf9fc 28326+ goto out;
e49829fe 28327+ err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
1facf9fc 28328+ if (unlikely(err)) {
e49829fe 28329+ kset_unregister(sysaufs_kset);
1facf9fc 28330+ goto out;
28331+ }
28332+
28333+ err = dbgaufs_init();
28334+ if (unlikely(err))
28335+ sysaufs_fin();
4f0767ce 28336+out:
1facf9fc 28337+ return err;
28338+}
7f207e10
AM
28339diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
28340--- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 28341+++ linux/fs/aufs/sysaufs.h 2015-09-24 10:47:58.254719746 +0200
c1595e42 28342@@ -0,0 +1,101 @@
1facf9fc 28343+/*
2000de60 28344+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 28345+ *
28346+ * This program, aufs is free software; you can redistribute it and/or modify
28347+ * it under the terms of the GNU General Public License as published by
28348+ * the Free Software Foundation; either version 2 of the License, or
28349+ * (at your option) any later version.
dece6358
AM
28350+ *
28351+ * This program is distributed in the hope that it will be useful,
28352+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28353+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28354+ * GNU General Public License for more details.
28355+ *
28356+ * You should have received a copy of the GNU General Public License
523b37e3 28357+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 28358+ */
28359+
28360+/*
28361+ * sysfs interface and mount lifetime management
28362+ */
28363+
28364+#ifndef __SYSAUFS_H__
28365+#define __SYSAUFS_H__
28366+
28367+#ifdef __KERNEL__
28368+
1facf9fc 28369+#include <linux/sysfs.h>
1facf9fc 28370+#include "module.h"
28371+
dece6358
AM
28372+struct super_block;
28373+struct au_sbinfo;
28374+
1facf9fc 28375+struct sysaufs_si_attr {
28376+ struct attribute attr;
28377+ int (*show)(struct seq_file *seq, struct super_block *sb);
28378+};
28379+
28380+/* ---------------------------------------------------------------------- */
28381+
28382+/* sysaufs.c */
28383+extern unsigned long sysaufs_si_mask;
e49829fe 28384+extern struct kset *sysaufs_kset;
1facf9fc 28385+extern struct attribute *sysaufs_si_attrs[];
28386+int sysaufs_si_init(struct au_sbinfo *sbinfo);
28387+int __init sysaufs_init(void);
28388+void sysaufs_fin(void);
28389+
28390+/* ---------------------------------------------------------------------- */
28391+
28392+/* some people doesn't like to show a pointer in kernel */
28393+static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
28394+{
28395+ return sysaufs_si_mask ^ (unsigned long)sbinfo;
28396+}
28397+
28398+#define SysaufsSiNamePrefix "si_"
28399+#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
28400+static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
28401+{
28402+ snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
28403+ sysaufs_si_id(sbinfo));
28404+}
28405+
28406+struct au_branch;
28407+#ifdef CONFIG_SYSFS
28408+/* sysfs.c */
28409+extern struct attribute_group *sysaufs_attr_group;
28410+
28411+int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
28412+ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
28413+ char *buf);
076b876e
AM
28414+long au_brinfo_ioctl(struct file *file, unsigned long arg);
28415+#ifdef CONFIG_COMPAT
28416+long au_brinfo_compat_ioctl(struct file *file, unsigned long arg);
28417+#endif
1facf9fc 28418+
28419+void sysaufs_br_init(struct au_branch *br);
28420+void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
28421+void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
28422+
28423+#define sysaufs_brs_init() do {} while (0)
28424+
28425+#else
28426+#define sysaufs_attr_group NULL
28427+
4a4d8108 28428+AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
c1595e42
JR
28429+AuStub(ssize_t, sysaufs_si_show, return 0, struct kobject *kobj,
28430+ struct attribute *attr, char *buf)
4a4d8108
AM
28431+AuStubVoid(sysaufs_br_init, struct au_branch *br)
28432+AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
28433+AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
1facf9fc 28434+
28435+static inline void sysaufs_brs_init(void)
28436+{
28437+ sysaufs_brs = 0;
28438+}
28439+
28440+#endif /* CONFIG_SYSFS */
28441+
28442+#endif /* __KERNEL__ */
28443+#endif /* __SYSAUFS_H__ */
7f207e10
AM
28444diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
28445--- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 28446+++ linux/fs/aufs/sysfs.c 2015-09-24 10:47:58.254719746 +0200
076b876e 28447@@ -0,0 +1,372 @@
1facf9fc 28448+/*
2000de60 28449+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 28450+ *
28451+ * This program, aufs is free software; you can redistribute it and/or modify
28452+ * it under the terms of the GNU General Public License as published by
28453+ * the Free Software Foundation; either version 2 of the License, or
28454+ * (at your option) any later version.
dece6358
AM
28455+ *
28456+ * This program is distributed in the hope that it will be useful,
28457+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28458+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28459+ * GNU General Public License for more details.
28460+ *
28461+ * You should have received a copy of the GNU General Public License
523b37e3 28462+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 28463+ */
28464+
28465+/*
28466+ * sysfs interface
28467+ */
28468+
076b876e 28469+#include <linux/compat.h>
1facf9fc 28470+#include <linux/seq_file.h>
1facf9fc 28471+#include "aufs.h"
28472+
4a4d8108
AM
28473+#ifdef CONFIG_AUFS_FS_MODULE
28474+/* this entry violates the "one line per file" policy of sysfs */
28475+static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
28476+ char *buf)
28477+{
28478+ ssize_t err;
28479+ static char *conf =
28480+/* this file is generated at compiling */
28481+#include "conf.str"
28482+ ;
28483+
28484+ err = snprintf(buf, PAGE_SIZE, conf);
28485+ if (unlikely(err >= PAGE_SIZE))
28486+ err = -EFBIG;
28487+ return err;
28488+}
28489+
28490+static struct kobj_attribute au_config_attr = __ATTR_RO(config);
28491+#endif
28492+
1facf9fc 28493+static struct attribute *au_attr[] = {
4a4d8108
AM
28494+#ifdef CONFIG_AUFS_FS_MODULE
28495+ &au_config_attr.attr,
28496+#endif
1facf9fc 28497+ NULL, /* need to NULL terminate the list of attributes */
28498+};
28499+
28500+static struct attribute_group sysaufs_attr_group_body = {
28501+ .attrs = au_attr
28502+};
28503+
28504+struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
28505+
28506+/* ---------------------------------------------------------------------- */
28507+
28508+int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
28509+{
28510+ int err;
28511+
dece6358
AM
28512+ SiMustAnyLock(sb);
28513+
1facf9fc 28514+ err = 0;
28515+ if (au_opt_test(au_mntflags(sb), XINO)) {
28516+ err = au_xino_path(seq, au_sbi(sb)->si_xib);
28517+ seq_putc(seq, '\n');
28518+ }
28519+ return err;
28520+}
28521+
28522+/*
28523+ * the lifetime of branch is independent from the entry under sysfs.
28524+ * sysfs handles the lifetime of the entry, and never call ->show() after it is
28525+ * unlinked.
28526+ */
28527+static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
392086de 28528+ aufs_bindex_t bindex, int idx)
1facf9fc 28529+{
1e00d052 28530+ int err;
1facf9fc 28531+ struct path path;
28532+ struct dentry *root;
28533+ struct au_branch *br;
076b876e 28534+ au_br_perm_str_t perm;
1facf9fc 28535+
28536+ AuDbg("b%d\n", bindex);
28537+
1e00d052 28538+ err = 0;
1facf9fc 28539+ root = sb->s_root;
28540+ di_read_lock_parent(root, !AuLock_IR);
28541+ br = au_sbr(sb, bindex);
392086de
AM
28542+
28543+ switch (idx) {
28544+ case AuBrSysfs_BR:
28545+ path.mnt = au_br_mnt(br);
28546+ path.dentry = au_h_dptr(root, bindex);
28547+ au_seq_path(seq, &path);
076b876e
AM
28548+ au_optstr_br_perm(&perm, br->br_perm);
28549+ err = seq_printf(seq, "=%s\n", perm.a);
392086de
AM
28550+ break;
28551+ case AuBrSysfs_BRID:
28552+ err = seq_printf(seq, "%d\n", br->br_id);
392086de
AM
28553+ break;
28554+ }
076b876e
AM
28555+ di_read_unlock(root, !AuLock_IR);
28556+ if (err == -1)
28557+ err = -E2BIG;
392086de 28558+
1e00d052 28559+ return err;
1facf9fc 28560+}
28561+
28562+/* ---------------------------------------------------------------------- */
28563+
28564+static struct seq_file *au_seq(char *p, ssize_t len)
28565+{
28566+ struct seq_file *seq;
28567+
28568+ seq = kzalloc(sizeof(*seq), GFP_NOFS);
28569+ if (seq) {
28570+ /* mutex_init(&seq.lock); */
28571+ seq->buf = p;
28572+ seq->size = len;
28573+ return seq; /* success */
28574+ }
28575+
28576+ seq = ERR_PTR(-ENOMEM);
28577+ return seq;
28578+}
28579+
392086de
AM
28580+#define SysaufsBr_PREFIX "br"
28581+#define SysaufsBrid_PREFIX "brid"
1facf9fc 28582+
28583+/* todo: file size may exceed PAGE_SIZE */
28584+ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
1308ab2a 28585+ char *buf)
1facf9fc 28586+{
28587+ ssize_t err;
392086de 28588+ int idx;
1facf9fc 28589+ long l;
28590+ aufs_bindex_t bend;
28591+ struct au_sbinfo *sbinfo;
28592+ struct super_block *sb;
28593+ struct seq_file *seq;
28594+ char *name;
28595+ struct attribute **cattr;
28596+
28597+ sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
28598+ sb = sbinfo->si_sb;
1308ab2a 28599+
28600+ /*
28601+ * prevent a race condition between sysfs and aufs.
28602+ * for instance, sysfs_file_read() calls sysfs_get_active_two() which
28603+ * prohibits maintaining the sysfs entries.
28604+ * hew we acquire read lock after sysfs_get_active_two().
28605+ * on the other hand, the remount process may maintain the sysfs/aufs
28606+ * entries after acquiring write lock.
28607+ * it can cause a deadlock.
28608+ * simply we gave up processing read here.
28609+ */
28610+ err = -EBUSY;
28611+ if (unlikely(!si_noflush_read_trylock(sb)))
28612+ goto out;
1facf9fc 28613+
28614+ seq = au_seq(buf, PAGE_SIZE);
28615+ err = PTR_ERR(seq);
28616+ if (IS_ERR(seq))
1308ab2a 28617+ goto out_unlock;
1facf9fc 28618+
28619+ name = (void *)attr->name;
28620+ cattr = sysaufs_si_attrs;
28621+ while (*cattr) {
28622+ if (!strcmp(name, (*cattr)->name)) {
28623+ err = container_of(*cattr, struct sysaufs_si_attr, attr)
28624+ ->show(seq, sb);
28625+ goto out_seq;
28626+ }
28627+ cattr++;
28628+ }
28629+
392086de
AM
28630+ if (!strncmp(name, SysaufsBrid_PREFIX,
28631+ sizeof(SysaufsBrid_PREFIX) - 1)) {
28632+ idx = AuBrSysfs_BRID;
28633+ name += sizeof(SysaufsBrid_PREFIX) - 1;
28634+ } else if (!strncmp(name, SysaufsBr_PREFIX,
28635+ sizeof(SysaufsBr_PREFIX) - 1)) {
28636+ idx = AuBrSysfs_BR;
1facf9fc 28637+ name += sizeof(SysaufsBr_PREFIX) - 1;
392086de
AM
28638+ } else
28639+ BUG();
28640+
28641+ err = kstrtol(name, 10, &l);
28642+ if (!err) {
28643+ bend = au_sbend(sb);
28644+ if (l <= bend)
28645+ err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l, idx);
28646+ else
28647+ err = -ENOENT;
1facf9fc 28648+ }
1facf9fc 28649+
4f0767ce 28650+out_seq:
1facf9fc 28651+ if (!err) {
28652+ err = seq->count;
28653+ /* sysfs limit */
28654+ if (unlikely(err == PAGE_SIZE))
28655+ err = -EFBIG;
28656+ }
28657+ kfree(seq);
4f0767ce 28658+out_unlock:
1facf9fc 28659+ si_read_unlock(sb);
4f0767ce 28660+out:
1facf9fc 28661+ return err;
28662+}
28663+
28664+/* ---------------------------------------------------------------------- */
28665+
076b876e
AM
28666+static int au_brinfo(struct super_block *sb, union aufs_brinfo __user *arg)
28667+{
28668+ int err;
28669+ int16_t brid;
28670+ aufs_bindex_t bindex, bend;
28671+ size_t sz;
28672+ char *buf;
28673+ struct seq_file *seq;
28674+ struct au_branch *br;
28675+
28676+ si_read_lock(sb, AuLock_FLUSH);
28677+ bend = au_sbend(sb);
28678+ err = bend + 1;
28679+ if (!arg)
28680+ goto out;
28681+
28682+ err = -ENOMEM;
28683+ buf = (void *)__get_free_page(GFP_NOFS);
28684+ if (unlikely(!buf))
28685+ goto out;
28686+
28687+ seq = au_seq(buf, PAGE_SIZE);
28688+ err = PTR_ERR(seq);
28689+ if (IS_ERR(seq))
28690+ goto out_buf;
28691+
28692+ sz = sizeof(*arg) - offsetof(union aufs_brinfo, path);
28693+ for (bindex = 0; bindex <= bend; bindex++, arg++) {
28694+ err = !access_ok(VERIFY_WRITE, arg, sizeof(*arg));
28695+ if (unlikely(err))
28696+ break;
28697+
28698+ br = au_sbr(sb, bindex);
28699+ brid = br->br_id;
28700+ BUILD_BUG_ON(sizeof(brid) != sizeof(arg->id));
28701+ err = __put_user(brid, &arg->id);
28702+ if (unlikely(err))
28703+ break;
28704+
28705+ BUILD_BUG_ON(sizeof(br->br_perm) != sizeof(arg->perm));
28706+ err = __put_user(br->br_perm, &arg->perm);
28707+ if (unlikely(err))
28708+ break;
28709+
28710+ au_seq_path(seq, &br->br_path);
28711+ err = seq_putc(seq, '\0');
28712+ if (!err && seq->count <= sz) {
28713+ err = copy_to_user(arg->path, seq->buf, seq->count);
28714+ seq->count = 0;
28715+ if (unlikely(err))
28716+ break;
28717+ } else {
28718+ err = -E2BIG;
28719+ goto out_seq;
28720+ }
28721+ }
28722+ if (unlikely(err))
28723+ err = -EFAULT;
28724+
28725+out_seq:
28726+ kfree(seq);
28727+out_buf:
28728+ free_page((unsigned long)buf);
28729+out:
28730+ si_read_unlock(sb);
28731+ return err;
28732+}
28733+
28734+long au_brinfo_ioctl(struct file *file, unsigned long arg)
28735+{
2000de60 28736+ return au_brinfo(file->f_path.dentry->d_sb, (void __user *)arg);
076b876e
AM
28737+}
28738+
28739+#ifdef CONFIG_COMPAT
28740+long au_brinfo_compat_ioctl(struct file *file, unsigned long arg)
28741+{
2000de60 28742+ return au_brinfo(file->f_path.dentry->d_sb, compat_ptr(arg));
076b876e
AM
28743+}
28744+#endif
28745+
28746+/* ---------------------------------------------------------------------- */
28747+
1facf9fc 28748+void sysaufs_br_init(struct au_branch *br)
28749+{
392086de
AM
28750+ int i;
28751+ struct au_brsysfs *br_sysfs;
28752+ struct attribute *attr;
4a4d8108 28753+
392086de
AM
28754+ br_sysfs = br->br_sysfs;
28755+ for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28756+ attr = &br_sysfs->attr;
28757+ sysfs_attr_init(attr);
28758+ attr->name = br_sysfs->name;
28759+ attr->mode = S_IRUGO;
28760+ br_sysfs++;
28761+ }
1facf9fc 28762+}
28763+
28764+void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
28765+{
28766+ struct au_branch *br;
28767+ struct kobject *kobj;
392086de
AM
28768+ struct au_brsysfs *br_sysfs;
28769+ int i;
1facf9fc 28770+ aufs_bindex_t bend;
28771+
28772+ dbgaufs_brs_del(sb, bindex);
28773+
28774+ if (!sysaufs_brs)
28775+ return;
28776+
28777+ kobj = &au_sbi(sb)->si_kobj;
28778+ bend = au_sbend(sb);
28779+ for (; bindex <= bend; bindex++) {
28780+ br = au_sbr(sb, bindex);
392086de
AM
28781+ br_sysfs = br->br_sysfs;
28782+ for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28783+ sysfs_remove_file(kobj, &br_sysfs->attr);
28784+ br_sysfs++;
28785+ }
1facf9fc 28786+ }
28787+}
28788+
28789+void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
28790+{
392086de 28791+ int err, i;
1facf9fc 28792+ aufs_bindex_t bend;
28793+ struct kobject *kobj;
28794+ struct au_branch *br;
392086de 28795+ struct au_brsysfs *br_sysfs;
1facf9fc 28796+
28797+ dbgaufs_brs_add(sb, bindex);
28798+
28799+ if (!sysaufs_brs)
28800+ return;
28801+
28802+ kobj = &au_sbi(sb)->si_kobj;
28803+ bend = au_sbend(sb);
28804+ for (; bindex <= bend; bindex++) {
28805+ br = au_sbr(sb, bindex);
392086de
AM
28806+ br_sysfs = br->br_sysfs;
28807+ snprintf(br_sysfs[AuBrSysfs_BR].name, sizeof(br_sysfs->name),
28808+ SysaufsBr_PREFIX "%d", bindex);
28809+ snprintf(br_sysfs[AuBrSysfs_BRID].name, sizeof(br_sysfs->name),
28810+ SysaufsBrid_PREFIX "%d", bindex);
28811+ for (i = 0; i < ARRAY_SIZE(br->br_sysfs); i++) {
28812+ err = sysfs_create_file(kobj, &br_sysfs->attr);
28813+ if (unlikely(err))
28814+ pr_warn("failed %s under sysfs(%d)\n",
28815+ br_sysfs->name, err);
28816+ br_sysfs++;
28817+ }
1facf9fc 28818+ }
28819+}
7f207e10
AM
28820diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
28821--- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 28822+++ linux/fs/aufs/sysrq.c 2015-09-24 10:47:58.254719746 +0200
076b876e 28823@@ -0,0 +1,157 @@
1facf9fc 28824+/*
2000de60 28825+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 28826+ *
28827+ * This program, aufs is free software; you can redistribute it and/or modify
28828+ * it under the terms of the GNU General Public License as published by
28829+ * the Free Software Foundation; either version 2 of the License, or
28830+ * (at your option) any later version.
dece6358
AM
28831+ *
28832+ * This program is distributed in the hope that it will be useful,
28833+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28834+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28835+ * GNU General Public License for more details.
28836+ *
28837+ * You should have received a copy of the GNU General Public License
523b37e3 28838+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 28839+ */
28840+
28841+/*
28842+ * magic sysrq hanlder
28843+ */
28844+
1facf9fc 28845+/* #include <linux/sysrq.h> */
027c5e7a 28846+#include <linux/writeback.h>
1facf9fc 28847+#include "aufs.h"
28848+
28849+/* ---------------------------------------------------------------------- */
28850+
28851+static void sysrq_sb(struct super_block *sb)
28852+{
28853+ char *plevel;
28854+ struct au_sbinfo *sbinfo;
28855+ struct file *file;
523b37e3
AM
28856+ struct au_sphlhead *files;
28857+ struct au_finfo *finfo;
1facf9fc 28858+
28859+ plevel = au_plevel;
28860+ au_plevel = KERN_WARNING;
1facf9fc 28861+
4a4d8108 28862+ /* since we define pr_fmt, call printk directly */
c06a8ce3
AM
28863+#define pr(str) printk(KERN_WARNING AUFS_NAME ": " str)
28864+
28865+ sbinfo = au_sbi(sb);
4a4d8108 28866+ printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
c06a8ce3 28867+ pr("superblock\n");
1facf9fc 28868+ au_dpri_sb(sb);
027c5e7a
AM
28869+
28870+#if 0
c06a8ce3 28871+ pr("root dentry\n");
1facf9fc 28872+ au_dpri_dentry(sb->s_root);
c06a8ce3 28873+ pr("root inode\n");
5527c038 28874+ au_dpri_inode(d_inode(sb->s_root));
027c5e7a
AM
28875+#endif
28876+
1facf9fc 28877+#if 0
027c5e7a
AM
28878+ do {
28879+ int err, i, j, ndentry;
28880+ struct au_dcsub_pages dpages;
28881+ struct au_dpage *dpage;
28882+
28883+ err = au_dpages_init(&dpages, GFP_ATOMIC);
28884+ if (unlikely(err))
28885+ break;
28886+ err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
28887+ if (!err)
28888+ for (i = 0; i < dpages.ndpage; i++) {
28889+ dpage = dpages.dpages + i;
28890+ ndentry = dpage->ndentry;
28891+ for (j = 0; j < ndentry; j++)
28892+ au_dpri_dentry(dpage->dentries[j]);
28893+ }
28894+ au_dpages_free(&dpages);
28895+ } while (0);
28896+#endif
28897+
28898+#if 1
28899+ {
28900+ struct inode *i;
076b876e 28901+
c06a8ce3 28902+ pr("isolated inode\n");
2cbb1c4b
JR
28903+ spin_lock(&inode_sb_list_lock);
28904+ list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
28905+ spin_lock(&i->i_lock);
b4510431 28906+ if (1 || hlist_empty(&i->i_dentry))
027c5e7a 28907+ au_dpri_inode(i);
2cbb1c4b
JR
28908+ spin_unlock(&i->i_lock);
28909+ }
28910+ spin_unlock(&inode_sb_list_lock);
027c5e7a 28911+ }
1facf9fc 28912+#endif
c06a8ce3 28913+ pr("files\n");
523b37e3
AM
28914+ files = &au_sbi(sb)->si_files;
28915+ spin_lock(&files->spin);
28916+ hlist_for_each_entry(finfo, &files->head, fi_hlist) {
4a4d8108 28917+ umode_t mode;
076b876e 28918+
523b37e3 28919+ file = finfo->fi_file;
c06a8ce3 28920+ mode = file_inode(file)->i_mode;
38d290e6 28921+ if (!special_file(mode))
1facf9fc 28922+ au_dpri_file(file);
523b37e3
AM
28923+ }
28924+ spin_unlock(&files->spin);
c06a8ce3 28925+ pr("done\n");
1facf9fc 28926+
c06a8ce3 28927+#undef pr
1facf9fc 28928+ au_plevel = plevel;
1facf9fc 28929+}
28930+
28931+/* ---------------------------------------------------------------------- */
28932+
28933+/* module parameter */
28934+static char *aufs_sysrq_key = "a";
28935+module_param_named(sysrq, aufs_sysrq_key, charp, S_IRUGO);
28936+MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
28937+
0c5527e5 28938+static void au_sysrq(int key __maybe_unused)
1facf9fc 28939+{
1facf9fc 28940+ struct au_sbinfo *sbinfo;
28941+
027c5e7a 28942+ lockdep_off();
53392da6 28943+ au_sbilist_lock();
e49829fe 28944+ list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
1facf9fc 28945+ sysrq_sb(sbinfo->si_sb);
53392da6 28946+ au_sbilist_unlock();
027c5e7a 28947+ lockdep_on();
1facf9fc 28948+}
28949+
28950+static struct sysrq_key_op au_sysrq_op = {
28951+ .handler = au_sysrq,
28952+ .help_msg = "Aufs",
28953+ .action_msg = "Aufs",
28954+ .enable_mask = SYSRQ_ENABLE_DUMP
28955+};
28956+
28957+/* ---------------------------------------------------------------------- */
28958+
28959+int __init au_sysrq_init(void)
28960+{
28961+ int err;
28962+ char key;
28963+
28964+ err = -1;
28965+ key = *aufs_sysrq_key;
28966+ if ('a' <= key && key <= 'z')
28967+ err = register_sysrq_key(key, &au_sysrq_op);
28968+ if (unlikely(err))
4a4d8108 28969+ pr_err("err %d, sysrq=%c\n", err, key);
1facf9fc 28970+ return err;
28971+}
28972+
28973+void au_sysrq_fin(void)
28974+{
28975+ int err;
076b876e 28976+
1facf9fc 28977+ err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
28978+ if (unlikely(err))
4a4d8108 28979+ pr_err("err %d (ignored)\n", err);
1facf9fc 28980+}
7f207e10
AM
28981diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
28982--- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 28983+++ linux/fs/aufs/vdir.c 2015-09-24 10:47:58.258053165 +0200
b912730e 28984@@ -0,0 +1,888 @@
1facf9fc 28985+/*
2000de60 28986+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 28987+ *
28988+ * This program, aufs is free software; you can redistribute it and/or modify
28989+ * it under the terms of the GNU General Public License as published by
28990+ * the Free Software Foundation; either version 2 of the License, or
28991+ * (at your option) any later version.
dece6358
AM
28992+ *
28993+ * This program is distributed in the hope that it will be useful,
28994+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
28995+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28996+ * GNU General Public License for more details.
28997+ *
28998+ * You should have received a copy of the GNU General Public License
523b37e3 28999+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 29000+ */
29001+
29002+/*
29003+ * virtual or vertical directory
29004+ */
29005+
29006+#include "aufs.h"
29007+
dece6358 29008+static unsigned int calc_size(int nlen)
1facf9fc 29009+{
dece6358 29010+ return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
1facf9fc 29011+}
29012+
29013+static int set_deblk_end(union au_vdir_deblk_p *p,
29014+ union au_vdir_deblk_p *deblk_end)
29015+{
29016+ if (calc_size(0) <= deblk_end->deblk - p->deblk) {
29017+ p->de->de_str.len = 0;
29018+ /* smp_mb(); */
29019+ return 0;
29020+ }
29021+ return -1; /* error */
29022+}
29023+
29024+/* returns true or false */
29025+static int is_deblk_end(union au_vdir_deblk_p *p,
29026+ union au_vdir_deblk_p *deblk_end)
29027+{
29028+ if (calc_size(0) <= deblk_end->deblk - p->deblk)
29029+ return !p->de->de_str.len;
29030+ return 1;
29031+}
29032+
29033+static unsigned char *last_deblk(struct au_vdir *vdir)
29034+{
29035+ return vdir->vd_deblk[vdir->vd_nblk - 1];
29036+}
29037+
29038+/* ---------------------------------------------------------------------- */
29039+
1308ab2a 29040+/* estimate the apropriate size for name hash table */
29041+unsigned int au_rdhash_est(loff_t sz)
29042+{
29043+ unsigned int n;
29044+
29045+ n = UINT_MAX;
29046+ sz >>= 10;
29047+ if (sz < n)
29048+ n = sz;
29049+ if (sz < AUFS_RDHASH_DEF)
29050+ n = AUFS_RDHASH_DEF;
4a4d8108 29051+ /* pr_info("n %u\n", n); */
1308ab2a 29052+ return n;
29053+}
29054+
1facf9fc 29055+/*
29056+ * the allocated memory has to be freed by
dece6358 29057+ * au_nhash_wh_free() or au_nhash_de_free().
1facf9fc 29058+ */
dece6358 29059+int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
1facf9fc 29060+{
1facf9fc 29061+ struct hlist_head *head;
dece6358 29062+ unsigned int u;
076b876e 29063+ size_t sz;
1facf9fc 29064+
076b876e
AM
29065+ sz = sizeof(*nhash->nh_head) * num_hash;
29066+ head = kmalloc(sz, gfp);
dece6358
AM
29067+ if (head) {
29068+ nhash->nh_num = num_hash;
29069+ nhash->nh_head = head;
29070+ for (u = 0; u < num_hash; u++)
1facf9fc 29071+ INIT_HLIST_HEAD(head++);
dece6358 29072+ return 0; /* success */
1facf9fc 29073+ }
1facf9fc 29074+
dece6358 29075+ return -ENOMEM;
1facf9fc 29076+}
29077+
dece6358
AM
29078+static void nhash_count(struct hlist_head *head)
29079+{
29080+#if 0
29081+ unsigned long n;
29082+ struct hlist_node *pos;
29083+
29084+ n = 0;
29085+ hlist_for_each(pos, head)
29086+ n++;
4a4d8108 29087+ pr_info("%lu\n", n);
dece6358
AM
29088+#endif
29089+}
29090+
29091+static void au_nhash_wh_do_free(struct hlist_head *head)
1facf9fc 29092+{
c06a8ce3
AM
29093+ struct au_vdir_wh *pos;
29094+ struct hlist_node *node;
1facf9fc 29095+
c06a8ce3
AM
29096+ hlist_for_each_entry_safe(pos, node, head, wh_hash)
29097+ kfree(pos);
1facf9fc 29098+}
29099+
dece6358 29100+static void au_nhash_de_do_free(struct hlist_head *head)
1facf9fc 29101+{
c06a8ce3
AM
29102+ struct au_vdir_dehstr *pos;
29103+ struct hlist_node *node;
1facf9fc 29104+
c06a8ce3
AM
29105+ hlist_for_each_entry_safe(pos, node, head, hash)
29106+ au_cache_free_vdir_dehstr(pos);
1facf9fc 29107+}
29108+
dece6358
AM
29109+static void au_nhash_do_free(struct au_nhash *nhash,
29110+ void (*free)(struct hlist_head *head))
1facf9fc 29111+{
1308ab2a 29112+ unsigned int n;
1facf9fc 29113+ struct hlist_head *head;
1facf9fc 29114+
dece6358 29115+ n = nhash->nh_num;
1308ab2a 29116+ if (!n)
29117+ return;
29118+
dece6358 29119+ head = nhash->nh_head;
1308ab2a 29120+ while (n-- > 0) {
dece6358
AM
29121+ nhash_count(head);
29122+ free(head++);
1facf9fc 29123+ }
dece6358 29124+ kfree(nhash->nh_head);
1facf9fc 29125+}
29126+
dece6358 29127+void au_nhash_wh_free(struct au_nhash *whlist)
1facf9fc 29128+{
dece6358
AM
29129+ au_nhash_do_free(whlist, au_nhash_wh_do_free);
29130+}
1facf9fc 29131+
dece6358
AM
29132+static void au_nhash_de_free(struct au_nhash *delist)
29133+{
29134+ au_nhash_do_free(delist, au_nhash_de_do_free);
1facf9fc 29135+}
29136+
29137+/* ---------------------------------------------------------------------- */
29138+
29139+int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
29140+ int limit)
29141+{
29142+ int num;
29143+ unsigned int u, n;
29144+ struct hlist_head *head;
c06a8ce3 29145+ struct au_vdir_wh *pos;
1facf9fc 29146+
29147+ num = 0;
29148+ n = whlist->nh_num;
29149+ head = whlist->nh_head;
1308ab2a 29150+ for (u = 0; u < n; u++, head++)
c06a8ce3
AM
29151+ hlist_for_each_entry(pos, head, wh_hash)
29152+ if (pos->wh_bindex == btgt && ++num > limit)
1facf9fc 29153+ return 1;
1facf9fc 29154+ return 0;
29155+}
29156+
29157+static struct hlist_head *au_name_hash(struct au_nhash *nhash,
dece6358 29158+ unsigned char *name,
1facf9fc 29159+ unsigned int len)
29160+{
dece6358
AM
29161+ unsigned int v;
29162+ /* const unsigned int magic_bit = 12; */
29163+
1308ab2a 29164+ AuDebugOn(!nhash->nh_num || !nhash->nh_head);
29165+
dece6358
AM
29166+ v = 0;
29167+ while (len--)
29168+ v += *name++;
29169+ /* v = hash_long(v, magic_bit); */
29170+ v %= nhash->nh_num;
29171+ return nhash->nh_head + v;
29172+}
29173+
29174+static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
29175+ int nlen)
29176+{
29177+ return str->len == nlen && !memcmp(str->name, name, nlen);
1facf9fc 29178+}
29179+
29180+/* returns found or not */
dece6358 29181+int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
1facf9fc 29182+{
29183+ struct hlist_head *head;
c06a8ce3 29184+ struct au_vdir_wh *pos;
1facf9fc 29185+ struct au_vdir_destr *str;
29186+
dece6358 29187+ head = au_name_hash(whlist, name, nlen);
c06a8ce3
AM
29188+ hlist_for_each_entry(pos, head, wh_hash) {
29189+ str = &pos->wh_str;
1facf9fc 29190+ AuDbg("%.*s\n", str->len, str->name);
dece6358
AM
29191+ if (au_nhash_test_name(str, name, nlen))
29192+ return 1;
29193+ }
29194+ return 0;
29195+}
29196+
29197+/* returns found(true) or not */
29198+static int test_known(struct au_nhash *delist, char *name, int nlen)
29199+{
29200+ struct hlist_head *head;
c06a8ce3 29201+ struct au_vdir_dehstr *pos;
dece6358
AM
29202+ struct au_vdir_destr *str;
29203+
29204+ head = au_name_hash(delist, name, nlen);
c06a8ce3
AM
29205+ hlist_for_each_entry(pos, head, hash) {
29206+ str = pos->str;
dece6358
AM
29207+ AuDbg("%.*s\n", str->len, str->name);
29208+ if (au_nhash_test_name(str, name, nlen))
1facf9fc 29209+ return 1;
29210+ }
29211+ return 0;
29212+}
29213+
dece6358
AM
29214+static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
29215+ unsigned char d_type)
29216+{
29217+#ifdef CONFIG_AUFS_SHWH
29218+ wh->wh_ino = ino;
29219+ wh->wh_type = d_type;
29220+#endif
29221+}
29222+
29223+/* ---------------------------------------------------------------------- */
29224+
29225+int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
29226+ unsigned int d_type, aufs_bindex_t bindex,
29227+ unsigned char shwh)
1facf9fc 29228+{
29229+ int err;
29230+ struct au_vdir_destr *str;
29231+ struct au_vdir_wh *wh;
29232+
dece6358 29233+ AuDbg("%.*s\n", nlen, name);
1308ab2a 29234+ AuDebugOn(!whlist->nh_num || !whlist->nh_head);
29235+
1facf9fc 29236+ err = -ENOMEM;
dece6358 29237+ wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
1facf9fc 29238+ if (unlikely(!wh))
29239+ goto out;
29240+
29241+ err = 0;
29242+ wh->wh_bindex = bindex;
dece6358
AM
29243+ if (shwh)
29244+ au_shwh_init_wh(wh, ino, d_type);
1facf9fc 29245+ str = &wh->wh_str;
dece6358
AM
29246+ str->len = nlen;
29247+ memcpy(str->name, name, nlen);
29248+ hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
1facf9fc 29249+ /* smp_mb(); */
29250+
4f0767ce 29251+out:
1facf9fc 29252+ return err;
29253+}
29254+
1facf9fc 29255+static int append_deblk(struct au_vdir *vdir)
29256+{
29257+ int err;
dece6358 29258+ unsigned long ul;
1facf9fc 29259+ const unsigned int deblk_sz = vdir->vd_deblk_sz;
29260+ union au_vdir_deblk_p p, deblk_end;
29261+ unsigned char **o;
29262+
29263+ err = -ENOMEM;
dece6358
AM
29264+ o = krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
29265+ GFP_NOFS);
1facf9fc 29266+ if (unlikely(!o))
29267+ goto out;
29268+
29269+ vdir->vd_deblk = o;
29270+ p.deblk = kmalloc(deblk_sz, GFP_NOFS);
29271+ if (p.deblk) {
29272+ ul = vdir->vd_nblk++;
29273+ vdir->vd_deblk[ul] = p.deblk;
29274+ vdir->vd_last.ul = ul;
29275+ vdir->vd_last.p.deblk = p.deblk;
29276+ deblk_end.deblk = p.deblk + deblk_sz;
29277+ err = set_deblk_end(&p, &deblk_end);
29278+ }
29279+
4f0767ce 29280+out:
1facf9fc 29281+ return err;
29282+}
29283+
dece6358
AM
29284+static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
29285+ unsigned int d_type, struct au_nhash *delist)
29286+{
29287+ int err;
29288+ unsigned int sz;
29289+ const unsigned int deblk_sz = vdir->vd_deblk_sz;
29290+ union au_vdir_deblk_p p, *room, deblk_end;
29291+ struct au_vdir_dehstr *dehstr;
29292+
29293+ p.deblk = last_deblk(vdir);
29294+ deblk_end.deblk = p.deblk + deblk_sz;
29295+ room = &vdir->vd_last.p;
29296+ AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
29297+ || !is_deblk_end(room, &deblk_end));
29298+
29299+ sz = calc_size(nlen);
29300+ if (unlikely(sz > deblk_end.deblk - room->deblk)) {
29301+ err = append_deblk(vdir);
29302+ if (unlikely(err))
29303+ goto out;
29304+
29305+ p.deblk = last_deblk(vdir);
29306+ deblk_end.deblk = p.deblk + deblk_sz;
29307+ /* smp_mb(); */
29308+ AuDebugOn(room->deblk != p.deblk);
29309+ }
29310+
29311+ err = -ENOMEM;
4a4d8108 29312+ dehstr = au_cache_alloc_vdir_dehstr();
dece6358
AM
29313+ if (unlikely(!dehstr))
29314+ goto out;
29315+
29316+ dehstr->str = &room->de->de_str;
29317+ hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
29318+ room->de->de_ino = ino;
29319+ room->de->de_type = d_type;
29320+ room->de->de_str.len = nlen;
29321+ memcpy(room->de->de_str.name, name, nlen);
29322+
29323+ err = 0;
29324+ room->deblk += sz;
29325+ if (unlikely(set_deblk_end(room, &deblk_end)))
29326+ err = append_deblk(vdir);
29327+ /* smp_mb(); */
29328+
4f0767ce 29329+out:
dece6358
AM
29330+ return err;
29331+}
29332+
29333+/* ---------------------------------------------------------------------- */
29334+
29335+void au_vdir_free(struct au_vdir *vdir)
29336+{
29337+ unsigned char **deblk;
29338+
29339+ deblk = vdir->vd_deblk;
29340+ while (vdir->vd_nblk--)
29341+ kfree(*deblk++);
29342+ kfree(vdir->vd_deblk);
29343+ au_cache_free_vdir(vdir);
29344+}
29345+
1308ab2a 29346+static struct au_vdir *alloc_vdir(struct file *file)
1facf9fc 29347+{
29348+ struct au_vdir *vdir;
1308ab2a 29349+ struct super_block *sb;
1facf9fc 29350+ int err;
29351+
2000de60 29352+ sb = file->f_path.dentry->d_sb;
dece6358
AM
29353+ SiMustAnyLock(sb);
29354+
1facf9fc 29355+ err = -ENOMEM;
29356+ vdir = au_cache_alloc_vdir();
29357+ if (unlikely(!vdir))
29358+ goto out;
29359+
29360+ vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
29361+ if (unlikely(!vdir->vd_deblk))
29362+ goto out_free;
29363+
29364+ vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
1308ab2a 29365+ if (!vdir->vd_deblk_sz) {
29366+ /* estimate the apropriate size for deblk */
29367+ vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
4a4d8108 29368+ /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
1308ab2a 29369+ }
1facf9fc 29370+ vdir->vd_nblk = 0;
29371+ vdir->vd_version = 0;
29372+ vdir->vd_jiffy = 0;
29373+ err = append_deblk(vdir);
29374+ if (!err)
29375+ return vdir; /* success */
29376+
29377+ kfree(vdir->vd_deblk);
29378+
4f0767ce 29379+out_free:
1facf9fc 29380+ au_cache_free_vdir(vdir);
4f0767ce 29381+out:
1facf9fc 29382+ vdir = ERR_PTR(err);
29383+ return vdir;
29384+}
29385+
29386+static int reinit_vdir(struct au_vdir *vdir)
29387+{
29388+ int err;
29389+ union au_vdir_deblk_p p, deblk_end;
29390+
29391+ while (vdir->vd_nblk > 1) {
29392+ kfree(vdir->vd_deblk[vdir->vd_nblk - 1]);
29393+ /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
29394+ vdir->vd_nblk--;
29395+ }
29396+ p.deblk = vdir->vd_deblk[0];
29397+ deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
29398+ err = set_deblk_end(&p, &deblk_end);
29399+ /* keep vd_dblk_sz */
29400+ vdir->vd_last.ul = 0;
29401+ vdir->vd_last.p.deblk = vdir->vd_deblk[0];
29402+ vdir->vd_version = 0;
29403+ vdir->vd_jiffy = 0;
29404+ /* smp_mb(); */
29405+ return err;
29406+}
29407+
29408+/* ---------------------------------------------------------------------- */
29409+
1facf9fc 29410+#define AuFillVdir_CALLED 1
29411+#define AuFillVdir_WHABLE (1 << 1)
dece6358 29412+#define AuFillVdir_SHWH (1 << 2)
1facf9fc 29413+#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
7f207e10
AM
29414+#define au_fset_fillvdir(flags, name) \
29415+ do { (flags) |= AuFillVdir_##name; } while (0)
29416+#define au_fclr_fillvdir(flags, name) \
29417+ do { (flags) &= ~AuFillVdir_##name; } while (0)
1facf9fc 29418+
dece6358
AM
29419+#ifndef CONFIG_AUFS_SHWH
29420+#undef AuFillVdir_SHWH
29421+#define AuFillVdir_SHWH 0
29422+#endif
29423+
1facf9fc 29424+struct fillvdir_arg {
392086de 29425+ struct dir_context ctx;
1facf9fc 29426+ struct file *file;
29427+ struct au_vdir *vdir;
dece6358
AM
29428+ struct au_nhash delist;
29429+ struct au_nhash whlist;
1facf9fc 29430+ aufs_bindex_t bindex;
29431+ unsigned int flags;
29432+ int err;
29433+};
29434+
392086de 29435+static int fillvdir(struct dir_context *ctx, const char *__name, int nlen,
1facf9fc 29436+ loff_t offset __maybe_unused, u64 h_ino,
29437+ unsigned int d_type)
29438+{
392086de 29439+ struct fillvdir_arg *arg = container_of(ctx, struct fillvdir_arg, ctx);
1facf9fc 29440+ char *name = (void *)__name;
29441+ struct super_block *sb;
1facf9fc 29442+ ino_t ino;
dece6358 29443+ const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
1facf9fc 29444+
1facf9fc 29445+ arg->err = 0;
2000de60 29446+ sb = arg->file->f_path.dentry->d_sb;
1facf9fc 29447+ au_fset_fillvdir(arg->flags, CALLED);
29448+ /* smp_mb(); */
dece6358 29449+ if (nlen <= AUFS_WH_PFX_LEN
1facf9fc 29450+ || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
dece6358
AM
29451+ if (test_known(&arg->delist, name, nlen)
29452+ || au_nhash_test_known_wh(&arg->whlist, name, nlen))
29453+ goto out; /* already exists or whiteouted */
1facf9fc 29454+
dece6358 29455+ arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
4a4d8108
AM
29456+ if (!arg->err) {
29457+ if (unlikely(nlen > AUFS_MAX_NAMELEN))
29458+ d_type = DT_UNKNOWN;
dece6358
AM
29459+ arg->err = append_de(arg->vdir, name, nlen, ino,
29460+ d_type, &arg->delist);
4a4d8108 29461+ }
1facf9fc 29462+ } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
29463+ name += AUFS_WH_PFX_LEN;
dece6358
AM
29464+ nlen -= AUFS_WH_PFX_LEN;
29465+ if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
29466+ goto out; /* already whiteouted */
1facf9fc 29467+
dece6358
AM
29468+ if (shwh)
29469+ arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
29470+ &ino);
4a4d8108
AM
29471+ if (!arg->err) {
29472+ if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
29473+ d_type = DT_UNKNOWN;
1facf9fc 29474+ arg->err = au_nhash_append_wh
dece6358
AM
29475+ (&arg->whlist, name, nlen, ino, d_type,
29476+ arg->bindex, shwh);
4a4d8108 29477+ }
1facf9fc 29478+ }
29479+
4f0767ce 29480+out:
1facf9fc 29481+ if (!arg->err)
29482+ arg->vdir->vd_jiffy = jiffies;
29483+ /* smp_mb(); */
29484+ AuTraceErr(arg->err);
29485+ return arg->err;
29486+}
29487+
dece6358
AM
29488+static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
29489+ struct au_nhash *whlist, struct au_nhash *delist)
29490+{
29491+#ifdef CONFIG_AUFS_SHWH
29492+ int err;
29493+ unsigned int nh, u;
29494+ struct hlist_head *head;
c06a8ce3
AM
29495+ struct au_vdir_wh *pos;
29496+ struct hlist_node *n;
dece6358
AM
29497+ char *p, *o;
29498+ struct au_vdir_destr *destr;
29499+
29500+ AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
29501+
29502+ err = -ENOMEM;
537831f9 29503+ o = p = (void *)__get_free_page(GFP_NOFS);
dece6358
AM
29504+ if (unlikely(!p))
29505+ goto out;
29506+
29507+ err = 0;
29508+ nh = whlist->nh_num;
29509+ memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
29510+ p += AUFS_WH_PFX_LEN;
29511+ for (u = 0; u < nh; u++) {
29512+ head = whlist->nh_head + u;
c06a8ce3
AM
29513+ hlist_for_each_entry_safe(pos, n, head, wh_hash) {
29514+ destr = &pos->wh_str;
dece6358
AM
29515+ memcpy(p, destr->name, destr->len);
29516+ err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
c06a8ce3 29517+ pos->wh_ino, pos->wh_type, delist);
dece6358
AM
29518+ if (unlikely(err))
29519+ break;
29520+ }
29521+ }
29522+
537831f9 29523+ free_page((unsigned long)o);
dece6358 29524+
4f0767ce 29525+out:
dece6358
AM
29526+ AuTraceErr(err);
29527+ return err;
29528+#else
29529+ return 0;
29530+#endif
29531+}
29532+
1facf9fc 29533+static int au_do_read_vdir(struct fillvdir_arg *arg)
29534+{
29535+ int err;
dece6358 29536+ unsigned int rdhash;
1facf9fc 29537+ loff_t offset;
dece6358
AM
29538+ aufs_bindex_t bend, bindex, bstart;
29539+ unsigned char shwh;
1facf9fc 29540+ struct file *hf, *file;
29541+ struct super_block *sb;
29542+
1facf9fc 29543+ file = arg->file;
2000de60 29544+ sb = file->f_path.dentry->d_sb;
dece6358
AM
29545+ SiMustAnyLock(sb);
29546+
29547+ rdhash = au_sbi(sb)->si_rdhash;
1308ab2a 29548+ if (!rdhash)
29549+ rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
dece6358
AM
29550+ err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
29551+ if (unlikely(err))
1facf9fc 29552+ goto out;
dece6358
AM
29553+ err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
29554+ if (unlikely(err))
1facf9fc 29555+ goto out_delist;
29556+
29557+ err = 0;
29558+ arg->flags = 0;
dece6358
AM
29559+ shwh = 0;
29560+ if (au_opt_test(au_mntflags(sb), SHWH)) {
29561+ shwh = 1;
29562+ au_fset_fillvdir(arg->flags, SHWH);
29563+ }
29564+ bstart = au_fbstart(file);
4a4d8108 29565+ bend = au_fbend_dir(file);
dece6358 29566+ for (bindex = bstart; !err && bindex <= bend; bindex++) {
4a4d8108 29567+ hf = au_hf_dir(file, bindex);
1facf9fc 29568+ if (!hf)
29569+ continue;
29570+
29571+ offset = vfsub_llseek(hf, 0, SEEK_SET);
29572+ err = offset;
29573+ if (unlikely(offset))
29574+ break;
29575+
29576+ arg->bindex = bindex;
29577+ au_fclr_fillvdir(arg->flags, WHABLE);
dece6358
AM
29578+ if (shwh
29579+ || (bindex != bend
29580+ && au_br_whable(au_sbr_perm(sb, bindex))))
1facf9fc 29581+ au_fset_fillvdir(arg->flags, WHABLE);
29582+ do {
29583+ arg->err = 0;
29584+ au_fclr_fillvdir(arg->flags, CALLED);
29585+ /* smp_mb(); */
392086de 29586+ err = vfsub_iterate_dir(hf, &arg->ctx);
1facf9fc 29587+ if (err >= 0)
29588+ err = arg->err;
29589+ } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
392086de
AM
29590+
29591+ /*
29592+ * dir_relax() may be good for concurrency, but aufs should not
29593+ * use it since it will cause a lockdep problem.
29594+ */
1facf9fc 29595+ }
dece6358
AM
29596+
29597+ if (!err && shwh)
29598+ err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
29599+
29600+ au_nhash_wh_free(&arg->whlist);
1facf9fc 29601+
4f0767ce 29602+out_delist:
dece6358 29603+ au_nhash_de_free(&arg->delist);
4f0767ce 29604+out:
1facf9fc 29605+ return err;
29606+}
29607+
29608+static int read_vdir(struct file *file, int may_read)
29609+{
29610+ int err;
29611+ unsigned long expire;
29612+ unsigned char do_read;
392086de
AM
29613+ struct fillvdir_arg arg = {
29614+ .ctx = {
2000de60 29615+ .actor = fillvdir
392086de
AM
29616+ }
29617+ };
1facf9fc 29618+ struct inode *inode;
29619+ struct au_vdir *vdir, *allocated;
29620+
29621+ err = 0;
c06a8ce3 29622+ inode = file_inode(file);
1facf9fc 29623+ IMustLock(inode);
dece6358
AM
29624+ SiMustAnyLock(inode->i_sb);
29625+
1facf9fc 29626+ allocated = NULL;
29627+ do_read = 0;
29628+ expire = au_sbi(inode->i_sb)->si_rdcache;
29629+ vdir = au_ivdir(inode);
29630+ if (!vdir) {
29631+ do_read = 1;
1308ab2a 29632+ vdir = alloc_vdir(file);
1facf9fc 29633+ err = PTR_ERR(vdir);
29634+ if (IS_ERR(vdir))
29635+ goto out;
29636+ err = 0;
29637+ allocated = vdir;
29638+ } else if (may_read
29639+ && (inode->i_version != vdir->vd_version
29640+ || time_after(jiffies, vdir->vd_jiffy + expire))) {
29641+ do_read = 1;
29642+ err = reinit_vdir(vdir);
29643+ if (unlikely(err))
29644+ goto out;
29645+ }
29646+
29647+ if (!do_read)
29648+ return 0; /* success */
29649+
29650+ arg.file = file;
29651+ arg.vdir = vdir;
29652+ err = au_do_read_vdir(&arg);
29653+ if (!err) {
392086de 29654+ /* file->f_pos = 0; */ /* todo: ctx->pos? */
1facf9fc 29655+ vdir->vd_version = inode->i_version;
29656+ vdir->vd_last.ul = 0;
29657+ vdir->vd_last.p.deblk = vdir->vd_deblk[0];
29658+ if (allocated)
29659+ au_set_ivdir(inode, allocated);
29660+ } else if (allocated)
29661+ au_vdir_free(allocated);
29662+
4f0767ce 29663+out:
1facf9fc 29664+ return err;
29665+}
29666+
29667+static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
29668+{
29669+ int err, rerr;
29670+ unsigned long ul, n;
29671+ const unsigned int deblk_sz = src->vd_deblk_sz;
29672+
29673+ AuDebugOn(tgt->vd_nblk != 1);
29674+
29675+ err = -ENOMEM;
29676+ if (tgt->vd_nblk < src->vd_nblk) {
29677+ unsigned char **p;
29678+
dece6358
AM
29679+ p = krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
29680+ GFP_NOFS);
1facf9fc 29681+ if (unlikely(!p))
29682+ goto out;
29683+ tgt->vd_deblk = p;
29684+ }
29685+
1308ab2a 29686+ if (tgt->vd_deblk_sz != deblk_sz) {
29687+ unsigned char *p;
29688+
29689+ tgt->vd_deblk_sz = deblk_sz;
29690+ p = krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS);
29691+ if (unlikely(!p))
29692+ goto out;
29693+ tgt->vd_deblk[0] = p;
29694+ }
1facf9fc 29695+ memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
1facf9fc 29696+ tgt->vd_version = src->vd_version;
29697+ tgt->vd_jiffy = src->vd_jiffy;
29698+
29699+ n = src->vd_nblk;
29700+ for (ul = 1; ul < n; ul++) {
dece6358
AM
29701+ tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
29702+ GFP_NOFS);
29703+ if (unlikely(!tgt->vd_deblk[ul]))
1facf9fc 29704+ goto out;
1308ab2a 29705+ tgt->vd_nblk++;
1facf9fc 29706+ }
1308ab2a 29707+ tgt->vd_nblk = n;
29708+ tgt->vd_last.ul = tgt->vd_last.ul;
29709+ tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
29710+ tgt->vd_last.p.deblk += src->vd_last.p.deblk
29711+ - src->vd_deblk[src->vd_last.ul];
1facf9fc 29712+ /* smp_mb(); */
29713+ return 0; /* success */
29714+
4f0767ce 29715+out:
1facf9fc 29716+ rerr = reinit_vdir(tgt);
29717+ BUG_ON(rerr);
29718+ return err;
29719+}
29720+
29721+int au_vdir_init(struct file *file)
29722+{
29723+ int err;
29724+ struct inode *inode;
29725+ struct au_vdir *vdir_cache, *allocated;
29726+
392086de 29727+ /* test file->f_pos here instead of ctx->pos */
1facf9fc 29728+ err = read_vdir(file, !file->f_pos);
29729+ if (unlikely(err))
29730+ goto out;
29731+
29732+ allocated = NULL;
29733+ vdir_cache = au_fvdir_cache(file);
29734+ if (!vdir_cache) {
1308ab2a 29735+ vdir_cache = alloc_vdir(file);
1facf9fc 29736+ err = PTR_ERR(vdir_cache);
29737+ if (IS_ERR(vdir_cache))
29738+ goto out;
29739+ allocated = vdir_cache;
29740+ } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
392086de 29741+ /* test file->f_pos here instead of ctx->pos */
1facf9fc 29742+ err = reinit_vdir(vdir_cache);
29743+ if (unlikely(err))
29744+ goto out;
29745+ } else
29746+ return 0; /* success */
29747+
c06a8ce3 29748+ inode = file_inode(file);
1facf9fc 29749+ err = copy_vdir(vdir_cache, au_ivdir(inode));
29750+ if (!err) {
29751+ file->f_version = inode->i_version;
29752+ if (allocated)
29753+ au_set_fvdir_cache(file, allocated);
29754+ } else if (allocated)
29755+ au_vdir_free(allocated);
29756+
4f0767ce 29757+out:
1facf9fc 29758+ return err;
29759+}
29760+
29761+static loff_t calc_offset(struct au_vdir *vdir)
29762+{
29763+ loff_t offset;
29764+ union au_vdir_deblk_p p;
29765+
29766+ p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
29767+ offset = vdir->vd_last.p.deblk - p.deblk;
29768+ offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
29769+ return offset;
29770+}
29771+
29772+/* returns true or false */
392086de 29773+static int seek_vdir(struct file *file, struct dir_context *ctx)
1facf9fc 29774+{
29775+ int valid;
29776+ unsigned int deblk_sz;
29777+ unsigned long ul, n;
29778+ loff_t offset;
29779+ union au_vdir_deblk_p p, deblk_end;
29780+ struct au_vdir *vdir_cache;
29781+
29782+ valid = 1;
29783+ vdir_cache = au_fvdir_cache(file);
29784+ offset = calc_offset(vdir_cache);
29785+ AuDbg("offset %lld\n", offset);
392086de 29786+ if (ctx->pos == offset)
1facf9fc 29787+ goto out;
29788+
29789+ vdir_cache->vd_last.ul = 0;
29790+ vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
392086de 29791+ if (!ctx->pos)
1facf9fc 29792+ goto out;
29793+
29794+ valid = 0;
29795+ deblk_sz = vdir_cache->vd_deblk_sz;
392086de 29796+ ul = div64_u64(ctx->pos, deblk_sz);
1facf9fc 29797+ AuDbg("ul %lu\n", ul);
29798+ if (ul >= vdir_cache->vd_nblk)
29799+ goto out;
29800+
29801+ n = vdir_cache->vd_nblk;
29802+ for (; ul < n; ul++) {
29803+ p.deblk = vdir_cache->vd_deblk[ul];
29804+ deblk_end.deblk = p.deblk + deblk_sz;
29805+ offset = ul;
29806+ offset *= deblk_sz;
392086de 29807+ while (!is_deblk_end(&p, &deblk_end) && offset < ctx->pos) {
1facf9fc 29808+ unsigned int l;
29809+
29810+ l = calc_size(p.de->de_str.len);
29811+ offset += l;
29812+ p.deblk += l;
29813+ }
29814+ if (!is_deblk_end(&p, &deblk_end)) {
29815+ valid = 1;
29816+ vdir_cache->vd_last.ul = ul;
29817+ vdir_cache->vd_last.p = p;
29818+ break;
29819+ }
29820+ }
29821+
4f0767ce 29822+out:
1facf9fc 29823+ /* smp_mb(); */
29824+ AuTraceErr(!valid);
29825+ return valid;
29826+}
29827+
392086de 29828+int au_vdir_fill_de(struct file *file, struct dir_context *ctx)
1facf9fc 29829+{
1facf9fc 29830+ unsigned int l, deblk_sz;
29831+ union au_vdir_deblk_p deblk_end;
29832+ struct au_vdir *vdir_cache;
29833+ struct au_vdir_de *de;
29834+
29835+ vdir_cache = au_fvdir_cache(file);
392086de 29836+ if (!seek_vdir(file, ctx))
1facf9fc 29837+ return 0;
29838+
29839+ deblk_sz = vdir_cache->vd_deblk_sz;
29840+ while (1) {
29841+ deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
29842+ deblk_end.deblk += deblk_sz;
29843+ while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
29844+ de = vdir_cache->vd_last.p.de;
29845+ AuDbg("%.*s, off%lld, i%lu, dt%d\n",
392086de 29846+ de->de_str.len, de->de_str.name, ctx->pos,
1facf9fc 29847+ (unsigned long)de->de_ino, de->de_type);
392086de
AM
29848+ if (unlikely(!dir_emit(ctx, de->de_str.name,
29849+ de->de_str.len, de->de_ino,
29850+ de->de_type))) {
1facf9fc 29851+ /* todo: ignore the error caused by udba? */
29852+ /* return err; */
29853+ return 0;
29854+ }
29855+
29856+ l = calc_size(de->de_str.len);
29857+ vdir_cache->vd_last.p.deblk += l;
392086de 29858+ ctx->pos += l;
1facf9fc 29859+ }
29860+ if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
29861+ vdir_cache->vd_last.ul++;
29862+ vdir_cache->vd_last.p.deblk
29863+ = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
392086de 29864+ ctx->pos = deblk_sz * vdir_cache->vd_last.ul;
1facf9fc 29865+ continue;
29866+ }
29867+ break;
29868+ }
29869+
29870+ /* smp_mb(); */
29871+ return 0;
29872+}
7f207e10
AM
29873diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
29874--- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 29875+++ linux/fs/aufs/vfsub.c 2015-09-24 10:47:58.258053165 +0200
5527c038 29876@@ -0,0 +1,848 @@
1facf9fc 29877+/*
2000de60 29878+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 29879+ *
29880+ * This program, aufs is free software; you can redistribute it and/or modify
29881+ * it under the terms of the GNU General Public License as published by
29882+ * the Free Software Foundation; either version 2 of the License, or
29883+ * (at your option) any later version.
dece6358
AM
29884+ *
29885+ * This program is distributed in the hope that it will be useful,
29886+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
29887+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29888+ * GNU General Public License for more details.
29889+ *
29890+ * You should have received a copy of the GNU General Public License
523b37e3 29891+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 29892+ */
29893+
29894+/*
29895+ * sub-routines for VFS
29896+ */
29897+
dece6358
AM
29898+#include <linux/namei.h>
29899+#include <linux/security.h>
29900+#include <linux/splice.h>
1facf9fc 29901+#include "aufs.h"
29902+
29903+int vfsub_update_h_iattr(struct path *h_path, int *did)
29904+{
29905+ int err;
29906+ struct kstat st;
29907+ struct super_block *h_sb;
29908+
29909+ /* for remote fs, leave work for its getattr or d_revalidate */
29910+ /* for bad i_attr fs, handle them in aufs_getattr() */
29911+ /* still some fs may acquire i_mutex. we need to skip them */
29912+ err = 0;
29913+ if (!did)
29914+ did = &err;
29915+ h_sb = h_path->dentry->d_sb;
29916+ *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
29917+ if (*did)
c06a8ce3 29918+ err = vfs_getattr(h_path, &st);
1facf9fc 29919+
29920+ return err;
29921+}
29922+
29923+/* ---------------------------------------------------------------------- */
29924+
4a4d8108 29925+struct file *vfsub_dentry_open(struct path *path, int flags)
1308ab2a 29926+{
29927+ struct file *file;
29928+
b4510431 29929+ file = dentry_open(path, flags /* | __FMODE_NONOTIFY */,
7f207e10 29930+ current_cred());
2cbb1c4b
JR
29931+ if (!IS_ERR_OR_NULL(file)
29932+ && (file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
5527c038 29933+ i_readcount_inc(d_inode(path->dentry));
4a4d8108 29934+
1308ab2a 29935+ return file;
29936+}
29937+
1facf9fc 29938+struct file *vfsub_filp_open(const char *path, int oflags, int mode)
29939+{
29940+ struct file *file;
29941+
2cbb1c4b 29942+ lockdep_off();
7f207e10 29943+ file = filp_open(path,
2cbb1c4b 29944+ oflags /* | __FMODE_NONOTIFY */,
7f207e10 29945+ mode);
2cbb1c4b 29946+ lockdep_on();
1facf9fc 29947+ if (IS_ERR(file))
29948+ goto out;
29949+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
29950+
4f0767ce 29951+out:
1facf9fc 29952+ return file;
29953+}
29954+
b912730e
AM
29955+/*
29956+ * Ideally this function should call VFS:do_last() in order to keep all its
29957+ * checkings. But it is very hard for aufs to regenerate several VFS internal
29958+ * structure such as nameidata. This is a second (or third) best approach.
29959+ * cf. linux/fs/namei.c:do_last(), lookup_open() and atomic_open().
29960+ */
29961+int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
29962+ struct vfsub_aopen_args *args, struct au_branch *br)
29963+{
29964+ int err;
29965+ struct file *file = args->file;
29966+ /* copied from linux/fs/namei.c:atomic_open() */
29967+ struct dentry *const DENTRY_NOT_SET = (void *)-1UL;
29968+
29969+ IMustLock(dir);
29970+ AuDebugOn(!dir->i_op->atomic_open);
29971+
29972+ err = au_br_test_oflag(args->open_flag, br);
29973+ if (unlikely(err))
29974+ goto out;
29975+
29976+ args->file->f_path.dentry = DENTRY_NOT_SET;
29977+ args->file->f_path.mnt = au_br_mnt(br);
29978+ err = dir->i_op->atomic_open(dir, dentry, file, args->open_flag,
29979+ args->create_mode, args->opened);
29980+ if (err >= 0) {
29981+ /* some filesystems don't set FILE_CREATED while succeeded? */
29982+ if (*args->opened & FILE_CREATED)
29983+ fsnotify_create(dir, dentry);
29984+ } else
29985+ goto out;
29986+
29987+
29988+ if (!err) {
29989+ /* todo: call VFS:may_open() here */
29990+ err = open_check_o_direct(file);
29991+ /* todo: ima_file_check() too? */
29992+ if (!err && (args->open_flag & __FMODE_EXEC))
29993+ err = deny_write_access(file);
29994+ if (unlikely(err))
29995+ /* note that the file is created and still opened */
29996+ goto out;
29997+ }
29998+
29999+ atomic_inc(&br->br_count);
30000+ fsnotify_open(file);
30001+
30002+out:
30003+ return err;
30004+}
30005+
1facf9fc 30006+int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
30007+{
30008+ int err;
30009+
1facf9fc 30010+ err = kern_path(name, flags, path);
5527c038 30011+ if (!err && d_is_positive(path->dentry))
1facf9fc 30012+ vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
30013+ return err;
30014+}
30015+
30016+struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
30017+ int len)
30018+{
30019+ struct path path = {
30020+ .mnt = NULL
30021+ };
30022+
1308ab2a 30023+ /* VFS checks it too, but by WARN_ON_ONCE() */
5527c038 30024+ IMustLock(d_inode(parent));
1facf9fc 30025+
30026+ path.dentry = lookup_one_len(name, parent, len);
30027+ if (IS_ERR(path.dentry))
30028+ goto out;
5527c038 30029+ if (d_is_positive(path.dentry))
1facf9fc 30030+ vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
30031+
4f0767ce 30032+out:
4a4d8108 30033+ AuTraceErrPtr(path.dentry);
1facf9fc 30034+ return path.dentry;
30035+}
30036+
b4510431 30037+void vfsub_call_lkup_one(void *args)
2cbb1c4b 30038+{
b4510431
AM
30039+ struct vfsub_lkup_one_args *a = args;
30040+ *a->errp = vfsub_lkup_one(a->name, a->parent);
2cbb1c4b
JR
30041+}
30042+
1facf9fc 30043+/* ---------------------------------------------------------------------- */
30044+
30045+struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
30046+ struct dentry *d2, struct au_hinode *hdir2)
30047+{
30048+ struct dentry *d;
30049+
2cbb1c4b 30050+ lockdep_off();
1facf9fc 30051+ d = lock_rename(d1, d2);
2cbb1c4b 30052+ lockdep_on();
4a4d8108 30053+ au_hn_suspend(hdir1);
1facf9fc 30054+ if (hdir1 != hdir2)
4a4d8108 30055+ au_hn_suspend(hdir2);
1facf9fc 30056+
30057+ return d;
30058+}
30059+
30060+void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
30061+ struct dentry *d2, struct au_hinode *hdir2)
30062+{
4a4d8108 30063+ au_hn_resume(hdir1);
1facf9fc 30064+ if (hdir1 != hdir2)
4a4d8108 30065+ au_hn_resume(hdir2);
2cbb1c4b 30066+ lockdep_off();
1facf9fc 30067+ unlock_rename(d1, d2);
2cbb1c4b 30068+ lockdep_on();
1facf9fc 30069+}
30070+
30071+/* ---------------------------------------------------------------------- */
30072+
b4510431 30073+int vfsub_create(struct inode *dir, struct path *path, int mode, bool want_excl)
1facf9fc 30074+{
30075+ int err;
30076+ struct dentry *d;
30077+
30078+ IMustLock(dir);
30079+
30080+ d = path->dentry;
30081+ path->dentry = d->d_parent;
b752ccd1 30082+ err = security_path_mknod(path, d, mode, 0);
1facf9fc 30083+ path->dentry = d;
30084+ if (unlikely(err))
30085+ goto out;
30086+
c1595e42 30087+ lockdep_off();
b4510431 30088+ err = vfs_create(dir, path->dentry, mode, want_excl);
c1595e42 30089+ lockdep_on();
1facf9fc 30090+ if (!err) {
30091+ struct path tmp = *path;
30092+ int did;
30093+
30094+ vfsub_update_h_iattr(&tmp, &did);
30095+ if (did) {
30096+ tmp.dentry = path->dentry->d_parent;
30097+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30098+ }
30099+ /*ignore*/
30100+ }
30101+
4f0767ce 30102+out:
1facf9fc 30103+ return err;
30104+}
30105+
30106+int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
30107+{
30108+ int err;
30109+ struct dentry *d;
30110+
30111+ IMustLock(dir);
30112+
30113+ d = path->dentry;
30114+ path->dentry = d->d_parent;
b752ccd1 30115+ err = security_path_symlink(path, d, symname);
1facf9fc 30116+ path->dentry = d;
30117+ if (unlikely(err))
30118+ goto out;
30119+
c1595e42 30120+ lockdep_off();
1facf9fc 30121+ err = vfs_symlink(dir, path->dentry, symname);
c1595e42 30122+ lockdep_on();
1facf9fc 30123+ if (!err) {
30124+ struct path tmp = *path;
30125+ int did;
30126+
30127+ vfsub_update_h_iattr(&tmp, &did);
30128+ if (did) {
30129+ tmp.dentry = path->dentry->d_parent;
30130+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30131+ }
30132+ /*ignore*/
30133+ }
30134+
4f0767ce 30135+out:
1facf9fc 30136+ return err;
30137+}
30138+
30139+int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
30140+{
30141+ int err;
30142+ struct dentry *d;
30143+
30144+ IMustLock(dir);
30145+
30146+ d = path->dentry;
30147+ path->dentry = d->d_parent;
027c5e7a 30148+ err = security_path_mknod(path, d, mode, new_encode_dev(dev));
1facf9fc 30149+ path->dentry = d;
30150+ if (unlikely(err))
30151+ goto out;
30152+
c1595e42 30153+ lockdep_off();
1facf9fc 30154+ err = vfs_mknod(dir, path->dentry, mode, dev);
c1595e42 30155+ lockdep_on();
1facf9fc 30156+ if (!err) {
30157+ struct path tmp = *path;
30158+ int did;
30159+
30160+ vfsub_update_h_iattr(&tmp, &did);
30161+ if (did) {
30162+ tmp.dentry = path->dentry->d_parent;
30163+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30164+ }
30165+ /*ignore*/
30166+ }
30167+
4f0767ce 30168+out:
1facf9fc 30169+ return err;
30170+}
30171+
30172+static int au_test_nlink(struct inode *inode)
30173+{
30174+ const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
30175+
30176+ if (!au_test_fs_no_limit_nlink(inode->i_sb)
30177+ || inode->i_nlink < link_max)
30178+ return 0;
30179+ return -EMLINK;
30180+}
30181+
523b37e3
AM
30182+int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path,
30183+ struct inode **delegated_inode)
1facf9fc 30184+{
30185+ int err;
30186+ struct dentry *d;
30187+
30188+ IMustLock(dir);
30189+
5527c038 30190+ err = au_test_nlink(d_inode(src_dentry));
1facf9fc 30191+ if (unlikely(err))
30192+ return err;
30193+
b4510431 30194+ /* we don't call may_linkat() */
1facf9fc 30195+ d = path->dentry;
30196+ path->dentry = d->d_parent;
b752ccd1 30197+ err = security_path_link(src_dentry, path, d);
1facf9fc 30198+ path->dentry = d;
30199+ if (unlikely(err))
30200+ goto out;
30201+
2cbb1c4b 30202+ lockdep_off();
523b37e3 30203+ err = vfs_link(src_dentry, dir, path->dentry, delegated_inode);
2cbb1c4b 30204+ lockdep_on();
1facf9fc 30205+ if (!err) {
30206+ struct path tmp = *path;
30207+ int did;
30208+
30209+ /* fuse has different memory inode for the same inumber */
30210+ vfsub_update_h_iattr(&tmp, &did);
30211+ if (did) {
30212+ tmp.dentry = path->dentry->d_parent;
30213+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30214+ tmp.dentry = src_dentry;
30215+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30216+ }
30217+ /*ignore*/
30218+ }
30219+
4f0767ce 30220+out:
1facf9fc 30221+ return err;
30222+}
30223+
30224+int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
523b37e3
AM
30225+ struct inode *dir, struct path *path,
30226+ struct inode **delegated_inode)
1facf9fc 30227+{
30228+ int err;
30229+ struct path tmp = {
30230+ .mnt = path->mnt
30231+ };
30232+ struct dentry *d;
30233+
30234+ IMustLock(dir);
30235+ IMustLock(src_dir);
30236+
30237+ d = path->dentry;
30238+ path->dentry = d->d_parent;
30239+ tmp.dentry = src_dentry->d_parent;
38d290e6 30240+ err = security_path_rename(&tmp, src_dentry, path, d, /*flags*/0);
1facf9fc 30241+ path->dentry = d;
30242+ if (unlikely(err))
30243+ goto out;
30244+
2cbb1c4b 30245+ lockdep_off();
523b37e3 30246+ err = vfs_rename(src_dir, src_dentry, dir, path->dentry,
38d290e6 30247+ delegated_inode, /*flags*/0);
2cbb1c4b 30248+ lockdep_on();
1facf9fc 30249+ if (!err) {
30250+ int did;
30251+
30252+ tmp.dentry = d->d_parent;
30253+ vfsub_update_h_iattr(&tmp, &did);
30254+ if (did) {
30255+ tmp.dentry = src_dentry;
30256+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30257+ tmp.dentry = src_dentry->d_parent;
30258+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30259+ }
30260+ /*ignore*/
30261+ }
30262+
4f0767ce 30263+out:
1facf9fc 30264+ return err;
30265+}
30266+
30267+int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
30268+{
30269+ int err;
30270+ struct dentry *d;
30271+
30272+ IMustLock(dir);
30273+
30274+ d = path->dentry;
30275+ path->dentry = d->d_parent;
b752ccd1 30276+ err = security_path_mkdir(path, d, mode);
1facf9fc 30277+ path->dentry = d;
30278+ if (unlikely(err))
30279+ goto out;
30280+
c1595e42 30281+ lockdep_off();
1facf9fc 30282+ err = vfs_mkdir(dir, path->dentry, mode);
c1595e42 30283+ lockdep_on();
1facf9fc 30284+ if (!err) {
30285+ struct path tmp = *path;
30286+ int did;
30287+
30288+ vfsub_update_h_iattr(&tmp, &did);
30289+ if (did) {
30290+ tmp.dentry = path->dentry->d_parent;
30291+ vfsub_update_h_iattr(&tmp, /*did*/NULL);
30292+ }
30293+ /*ignore*/
30294+ }
30295+
4f0767ce 30296+out:
1facf9fc 30297+ return err;
30298+}
30299+
30300+int vfsub_rmdir(struct inode *dir, struct path *path)
30301+{
30302+ int err;
30303+ struct dentry *d;
30304+
30305+ IMustLock(dir);
30306+
30307+ d = path->dentry;
30308+ path->dentry = d->d_parent;
b752ccd1 30309+ err = security_path_rmdir(path, d);
1facf9fc 30310+ path->dentry = d;
30311+ if (unlikely(err))
30312+ goto out;
30313+
2cbb1c4b 30314+ lockdep_off();
1facf9fc 30315+ err = vfs_rmdir(dir, path->dentry);
2cbb1c4b 30316+ lockdep_on();
1facf9fc 30317+ if (!err) {
30318+ struct path tmp = {
30319+ .dentry = path->dentry->d_parent,
30320+ .mnt = path->mnt
30321+ };
30322+
30323+ vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
30324+ }
30325+
4f0767ce 30326+out:
1facf9fc 30327+ return err;
30328+}
30329+
30330+/* ---------------------------------------------------------------------- */
30331+
9dbd164d 30332+/* todo: support mmap_sem? */
1facf9fc 30333+ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
30334+ loff_t *ppos)
30335+{
30336+ ssize_t err;
30337+
2cbb1c4b 30338+ lockdep_off();
1facf9fc 30339+ err = vfs_read(file, ubuf, count, ppos);
2cbb1c4b 30340+ lockdep_on();
1facf9fc 30341+ if (err >= 0)
30342+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30343+ return err;
30344+}
30345+
30346+/* todo: kernel_read()? */
30347+ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
30348+ loff_t *ppos)
30349+{
30350+ ssize_t err;
30351+ mm_segment_t oldfs;
b752ccd1
AM
30352+ union {
30353+ void *k;
30354+ char __user *u;
30355+ } buf;
1facf9fc 30356+
b752ccd1 30357+ buf.k = kbuf;
1facf9fc 30358+ oldfs = get_fs();
30359+ set_fs(KERNEL_DS);
b752ccd1 30360+ err = vfsub_read_u(file, buf.u, count, ppos);
1facf9fc 30361+ set_fs(oldfs);
30362+ return err;
30363+}
30364+
30365+ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
30366+ loff_t *ppos)
30367+{
30368+ ssize_t err;
30369+
2cbb1c4b 30370+ lockdep_off();
1facf9fc 30371+ err = vfs_write(file, ubuf, count, ppos);
2cbb1c4b 30372+ lockdep_on();
1facf9fc 30373+ if (err >= 0)
30374+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30375+ return err;
30376+}
30377+
30378+ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
30379+{
30380+ ssize_t err;
30381+ mm_segment_t oldfs;
b752ccd1
AM
30382+ union {
30383+ void *k;
30384+ const char __user *u;
30385+ } buf;
1facf9fc 30386+
b752ccd1 30387+ buf.k = kbuf;
1facf9fc 30388+ oldfs = get_fs();
30389+ set_fs(KERNEL_DS);
b752ccd1 30390+ err = vfsub_write_u(file, buf.u, count, ppos);
1facf9fc 30391+ set_fs(oldfs);
30392+ return err;
30393+}
30394+
4a4d8108
AM
30395+int vfsub_flush(struct file *file, fl_owner_t id)
30396+{
30397+ int err;
30398+
30399+ err = 0;
523b37e3 30400+ if (file->f_op->flush) {
2000de60 30401+ if (!au_test_nfs(file->f_path.dentry->d_sb))
2cbb1c4b
JR
30402+ err = file->f_op->flush(file, id);
30403+ else {
30404+ lockdep_off();
30405+ err = file->f_op->flush(file, id);
30406+ lockdep_on();
30407+ }
4a4d8108
AM
30408+ if (!err)
30409+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
30410+ /*ignore*/
30411+ }
30412+ return err;
30413+}
30414+
392086de 30415+int vfsub_iterate_dir(struct file *file, struct dir_context *ctx)
1facf9fc 30416+{
30417+ int err;
30418+
523b37e3 30419+ AuDbg("%pD, ctx{%pf, %llu}\n", file, ctx->actor, ctx->pos);
392086de 30420+
2cbb1c4b 30421+ lockdep_off();
392086de 30422+ err = iterate_dir(file, ctx);
2cbb1c4b 30423+ lockdep_on();
1facf9fc 30424+ if (err >= 0)
30425+ vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
30426+ return err;
30427+}
30428+
30429+long vfsub_splice_to(struct file *in, loff_t *ppos,
30430+ struct pipe_inode_info *pipe, size_t len,
30431+ unsigned int flags)
30432+{
30433+ long err;
30434+
2cbb1c4b 30435+ lockdep_off();
0fc653ad 30436+ err = do_splice_to(in, ppos, pipe, len, flags);
2cbb1c4b 30437+ lockdep_on();
4a4d8108 30438+ file_accessed(in);
1facf9fc 30439+ if (err >= 0)
30440+ vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
30441+ return err;
30442+}
30443+
30444+long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
30445+ loff_t *ppos, size_t len, unsigned int flags)
30446+{
30447+ long err;
30448+
2cbb1c4b 30449+ lockdep_off();
0fc653ad 30450+ err = do_splice_from(pipe, out, ppos, len, flags);
2cbb1c4b 30451+ lockdep_on();
1facf9fc 30452+ if (err >= 0)
30453+ vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
30454+ return err;
30455+}
30456+
53392da6
AM
30457+int vfsub_fsync(struct file *file, struct path *path, int datasync)
30458+{
30459+ int err;
30460+
30461+ /* file can be NULL */
30462+ lockdep_off();
30463+ err = vfs_fsync(file, datasync);
30464+ lockdep_on();
30465+ if (!err) {
30466+ if (!path) {
30467+ AuDebugOn(!file);
30468+ path = &file->f_path;
30469+ }
30470+ vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
30471+ }
30472+ return err;
30473+}
30474+
1facf9fc 30475+/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
30476+int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
30477+ struct file *h_file)
30478+{
30479+ int err;
30480+ struct inode *h_inode;
c06a8ce3 30481+ struct super_block *h_sb;
1facf9fc 30482+
1facf9fc 30483+ if (!h_file) {
c06a8ce3
AM
30484+ err = vfsub_truncate(h_path, length);
30485+ goto out;
1facf9fc 30486+ }
30487+
5527c038 30488+ h_inode = d_inode(h_path->dentry);
c06a8ce3
AM
30489+ h_sb = h_inode->i_sb;
30490+ lockdep_off();
30491+ sb_start_write(h_sb);
30492+ lockdep_on();
1facf9fc 30493+ err = locks_verify_truncate(h_inode, h_file, length);
30494+ if (!err)
953406b4 30495+ err = security_path_truncate(h_path);
2cbb1c4b
JR
30496+ if (!err) {
30497+ lockdep_off();
1facf9fc 30498+ err = do_truncate(h_path->dentry, length, attr, h_file);
2cbb1c4b
JR
30499+ lockdep_on();
30500+ }
c06a8ce3
AM
30501+ lockdep_off();
30502+ sb_end_write(h_sb);
30503+ lockdep_on();
1facf9fc 30504+
4f0767ce 30505+out:
1facf9fc 30506+ return err;
30507+}
30508+
30509+/* ---------------------------------------------------------------------- */
30510+
30511+struct au_vfsub_mkdir_args {
30512+ int *errp;
30513+ struct inode *dir;
30514+ struct path *path;
30515+ int mode;
30516+};
30517+
30518+static void au_call_vfsub_mkdir(void *args)
30519+{
30520+ struct au_vfsub_mkdir_args *a = args;
30521+ *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
30522+}
30523+
30524+int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
30525+{
30526+ int err, do_sio, wkq_err;
30527+
30528+ do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
c1595e42
JR
30529+ if (!do_sio) {
30530+ lockdep_off();
1facf9fc 30531+ err = vfsub_mkdir(dir, path, mode);
c1595e42
JR
30532+ lockdep_on();
30533+ } else {
1facf9fc 30534+ struct au_vfsub_mkdir_args args = {
30535+ .errp = &err,
30536+ .dir = dir,
30537+ .path = path,
30538+ .mode = mode
30539+ };
30540+ wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
30541+ if (unlikely(wkq_err))
30542+ err = wkq_err;
30543+ }
30544+
30545+ return err;
30546+}
30547+
30548+struct au_vfsub_rmdir_args {
30549+ int *errp;
30550+ struct inode *dir;
30551+ struct path *path;
30552+};
30553+
30554+static void au_call_vfsub_rmdir(void *args)
30555+{
30556+ struct au_vfsub_rmdir_args *a = args;
30557+ *a->errp = vfsub_rmdir(a->dir, a->path);
30558+}
30559+
30560+int vfsub_sio_rmdir(struct inode *dir, struct path *path)
30561+{
30562+ int err, do_sio, wkq_err;
30563+
30564+ do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
c1595e42
JR
30565+ if (!do_sio) {
30566+ lockdep_off();
1facf9fc 30567+ err = vfsub_rmdir(dir, path);
c1595e42
JR
30568+ lockdep_on();
30569+ } else {
1facf9fc 30570+ struct au_vfsub_rmdir_args args = {
30571+ .errp = &err,
30572+ .dir = dir,
30573+ .path = path
30574+ };
30575+ wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
30576+ if (unlikely(wkq_err))
30577+ err = wkq_err;
30578+ }
30579+
30580+ return err;
30581+}
30582+
30583+/* ---------------------------------------------------------------------- */
30584+
30585+struct notify_change_args {
30586+ int *errp;
30587+ struct path *path;
30588+ struct iattr *ia;
523b37e3 30589+ struct inode **delegated_inode;
1facf9fc 30590+};
30591+
30592+static void call_notify_change(void *args)
30593+{
30594+ struct notify_change_args *a = args;
30595+ struct inode *h_inode;
30596+
5527c038 30597+ h_inode = d_inode(a->path->dentry);
1facf9fc 30598+ IMustLock(h_inode);
30599+
30600+ *a->errp = -EPERM;
30601+ if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
c1595e42 30602+ lockdep_off();
523b37e3
AM
30603+ *a->errp = notify_change(a->path->dentry, a->ia,
30604+ a->delegated_inode);
c1595e42 30605+ lockdep_on();
1facf9fc 30606+ if (!*a->errp)
30607+ vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
30608+ }
30609+ AuTraceErr(*a->errp);
30610+}
30611+
523b37e3
AM
30612+int vfsub_notify_change(struct path *path, struct iattr *ia,
30613+ struct inode **delegated_inode)
1facf9fc 30614+{
30615+ int err;
30616+ struct notify_change_args args = {
523b37e3
AM
30617+ .errp = &err,
30618+ .path = path,
30619+ .ia = ia,
30620+ .delegated_inode = delegated_inode
1facf9fc 30621+ };
30622+
30623+ call_notify_change(&args);
30624+
30625+ return err;
30626+}
30627+
523b37e3
AM
30628+int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
30629+ struct inode **delegated_inode)
1facf9fc 30630+{
30631+ int err, wkq_err;
30632+ struct notify_change_args args = {
523b37e3
AM
30633+ .errp = &err,
30634+ .path = path,
30635+ .ia = ia,
30636+ .delegated_inode = delegated_inode
1facf9fc 30637+ };
30638+
30639+ wkq_err = au_wkq_wait(call_notify_change, &args);
30640+ if (unlikely(wkq_err))
30641+ err = wkq_err;
30642+
30643+ return err;
30644+}
30645+
30646+/* ---------------------------------------------------------------------- */
30647+
30648+struct unlink_args {
30649+ int *errp;
30650+ struct inode *dir;
30651+ struct path *path;
523b37e3 30652+ struct inode **delegated_inode;
1facf9fc 30653+};
30654+
30655+static void call_unlink(void *args)
30656+{
30657+ struct unlink_args *a = args;
30658+ struct dentry *d = a->path->dentry;
30659+ struct inode *h_inode;
30660+ const int stop_sillyrename = (au_test_nfs(d->d_sb)
c1595e42 30661+ && au_dcount(d) == 1);
1facf9fc 30662+
30663+ IMustLock(a->dir);
30664+
30665+ a->path->dentry = d->d_parent;
30666+ *a->errp = security_path_unlink(a->path, d);
30667+ a->path->dentry = d;
30668+ if (unlikely(*a->errp))
30669+ return;
30670+
30671+ if (!stop_sillyrename)
30672+ dget(d);
5527c038
JR
30673+ h_inode = NULL;
30674+ if (d_is_positive(d)) {
30675+ h_inode = d_inode(d);
027c5e7a 30676+ ihold(h_inode);
5527c038 30677+ }
1facf9fc 30678+
2cbb1c4b 30679+ lockdep_off();
523b37e3 30680+ *a->errp = vfs_unlink(a->dir, d, a->delegated_inode);
2cbb1c4b 30681+ lockdep_on();
1facf9fc 30682+ if (!*a->errp) {
30683+ struct path tmp = {
30684+ .dentry = d->d_parent,
30685+ .mnt = a->path->mnt
30686+ };
30687+ vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
30688+ }
30689+
30690+ if (!stop_sillyrename)
30691+ dput(d);
30692+ if (h_inode)
30693+ iput(h_inode);
30694+
30695+ AuTraceErr(*a->errp);
30696+}
30697+
30698+/*
30699+ * @dir: must be locked.
30700+ * @dentry: target dentry.
30701+ */
523b37e3
AM
30702+int vfsub_unlink(struct inode *dir, struct path *path,
30703+ struct inode **delegated_inode, int force)
1facf9fc 30704+{
30705+ int err;
30706+ struct unlink_args args = {
523b37e3
AM
30707+ .errp = &err,
30708+ .dir = dir,
30709+ .path = path,
30710+ .delegated_inode = delegated_inode
1facf9fc 30711+ };
30712+
30713+ if (!force)
30714+ call_unlink(&args);
30715+ else {
30716+ int wkq_err;
30717+
30718+ wkq_err = au_wkq_wait(call_unlink, &args);
30719+ if (unlikely(wkq_err))
30720+ err = wkq_err;
30721+ }
30722+
30723+ return err;
30724+}
7f207e10
AM
30725diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
30726--- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 30727+++ linux/fs/aufs/vfsub.h 2015-09-24 10:47:58.258053165 +0200
b912730e 30728@@ -0,0 +1,286 @@
1facf9fc 30729+/*
2000de60 30730+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 30731+ *
30732+ * This program, aufs is free software; you can redistribute it and/or modify
30733+ * it under the terms of the GNU General Public License as published by
30734+ * the Free Software Foundation; either version 2 of the License, or
30735+ * (at your option) any later version.
dece6358
AM
30736+ *
30737+ * This program is distributed in the hope that it will be useful,
30738+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
30739+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30740+ * GNU General Public License for more details.
30741+ *
30742+ * You should have received a copy of the GNU General Public License
523b37e3 30743+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 30744+ */
30745+
30746+/*
30747+ * sub-routines for VFS
30748+ */
30749+
30750+#ifndef __AUFS_VFSUB_H__
30751+#define __AUFS_VFSUB_H__
30752+
30753+#ifdef __KERNEL__
30754+
30755+#include <linux/fs.h>
b4510431 30756+#include <linux/mount.h>
c1595e42 30757+#include <linux/xattr.h>
7f207e10 30758+#include "debug.h"
1facf9fc 30759+
7f207e10 30760+/* copied from linux/fs/internal.h */
2cbb1c4b 30761+/* todo: BAD approach!! */
c06a8ce3 30762+extern void __mnt_drop_write(struct vfsmount *);
2cbb1c4b 30763+extern spinlock_t inode_sb_list_lock;
b912730e 30764+extern int open_check_o_direct(struct file *f);
7f207e10
AM
30765+
30766+/* ---------------------------------------------------------------------- */
1facf9fc 30767+
30768+/* lock subclass for lower inode */
30769+/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
30770+/* reduce? gave up. */
30771+enum {
c1595e42 30772+ AuLsc_I_Begin = I_MUTEX_PARENT2, /* 5 */
1facf9fc 30773+ AuLsc_I_PARENT, /* lower inode, parent first */
30774+ AuLsc_I_PARENT2, /* copyup dirs */
dece6358 30775+ AuLsc_I_PARENT3, /* copyup wh */
1facf9fc 30776+ AuLsc_I_CHILD,
30777+ AuLsc_I_CHILD2,
30778+ AuLsc_I_End
30779+};
30780+
30781+/* to debug easier, do not make them inlined functions */
30782+#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
30783+#define IMustLock(i) MtxMustLock(&(i)->i_mutex)
30784+
30785+/* ---------------------------------------------------------------------- */
30786+
7f207e10
AM
30787+static inline void vfsub_drop_nlink(struct inode *inode)
30788+{
30789+ AuDebugOn(!inode->i_nlink);
30790+ drop_nlink(inode);
30791+}
30792+
027c5e7a
AM
30793+static inline void vfsub_dead_dir(struct inode *inode)
30794+{
30795+ AuDebugOn(!S_ISDIR(inode->i_mode));
30796+ inode->i_flags |= S_DEAD;
30797+ clear_nlink(inode);
30798+}
30799+
392086de
AM
30800+static inline int vfsub_native_ro(struct inode *inode)
30801+{
30802+ return (inode->i_sb->s_flags & MS_RDONLY)
30803+ || IS_RDONLY(inode)
30804+ /* || IS_APPEND(inode) */
30805+ || IS_IMMUTABLE(inode);
30806+}
30807+
7f207e10
AM
30808+/* ---------------------------------------------------------------------- */
30809+
30810+int vfsub_update_h_iattr(struct path *h_path, int *did);
30811+struct file *vfsub_dentry_open(struct path *path, int flags);
30812+struct file *vfsub_filp_open(const char *path, int oflags, int mode);
b912730e
AM
30813+struct vfsub_aopen_args {
30814+ struct file *file;
30815+ unsigned int open_flag;
30816+ umode_t create_mode;
30817+ int *opened;
30818+};
30819+struct au_branch;
30820+int vfsub_atomic_open(struct inode *dir, struct dentry *dentry,
30821+ struct vfsub_aopen_args *args, struct au_branch *br);
1facf9fc 30822+int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
b4510431 30823+
1facf9fc 30824+struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
30825+ int len);
b4510431
AM
30826+
30827+struct vfsub_lkup_one_args {
30828+ struct dentry **errp;
30829+ struct qstr *name;
30830+ struct dentry *parent;
30831+};
30832+
30833+static inline struct dentry *vfsub_lkup_one(struct qstr *name,
30834+ struct dentry *parent)
30835+{
30836+ return vfsub_lookup_one_len(name->name, parent, name->len);
30837+}
30838+
30839+void vfsub_call_lkup_one(void *args);
30840+
30841+/* ---------------------------------------------------------------------- */
30842+
30843+static inline int vfsub_mnt_want_write(struct vfsmount *mnt)
30844+{
30845+ int err;
076b876e 30846+
b4510431
AM
30847+ lockdep_off();
30848+ err = mnt_want_write(mnt);
30849+ lockdep_on();
30850+ return err;
30851+}
30852+
30853+static inline void vfsub_mnt_drop_write(struct vfsmount *mnt)
30854+{
30855+ lockdep_off();
30856+ mnt_drop_write(mnt);
30857+ lockdep_on();
30858+}
1facf9fc 30859+
7e9cd9fe 30860+#if 0 /* reserved */
c06a8ce3
AM
30861+static inline void vfsub_mnt_drop_write_file(struct file *file)
30862+{
30863+ lockdep_off();
30864+ mnt_drop_write_file(file);
30865+ lockdep_on();
30866+}
7e9cd9fe 30867+#endif
c06a8ce3 30868+
1facf9fc 30869+/* ---------------------------------------------------------------------- */
30870+
30871+struct au_hinode;
30872+struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
30873+ struct dentry *d2, struct au_hinode *hdir2);
30874+void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
30875+ struct dentry *d2, struct au_hinode *hdir2);
30876+
537831f9
AM
30877+int vfsub_create(struct inode *dir, struct path *path, int mode,
30878+ bool want_excl);
1facf9fc 30879+int vfsub_symlink(struct inode *dir, struct path *path,
30880+ const char *symname);
30881+int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
30882+int vfsub_link(struct dentry *src_dentry, struct inode *dir,
523b37e3 30883+ struct path *path, struct inode **delegated_inode);
1facf9fc 30884+int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
523b37e3
AM
30885+ struct inode *hdir, struct path *path,
30886+ struct inode **delegated_inode);
1facf9fc 30887+int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
30888+int vfsub_rmdir(struct inode *dir, struct path *path);
30889+
30890+/* ---------------------------------------------------------------------- */
30891+
30892+ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
30893+ loff_t *ppos);
30894+ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
30895+ loff_t *ppos);
30896+ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
30897+ loff_t *ppos);
30898+ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
30899+ loff_t *ppos);
4a4d8108 30900+int vfsub_flush(struct file *file, fl_owner_t id);
392086de
AM
30901+int vfsub_iterate_dir(struct file *file, struct dir_context *ctx);
30902+
c06a8ce3
AM
30903+static inline loff_t vfsub_f_size_read(struct file *file)
30904+{
30905+ return i_size_read(file_inode(file));
30906+}
30907+
4a4d8108
AM
30908+static inline unsigned int vfsub_file_flags(struct file *file)
30909+{
30910+ unsigned int flags;
30911+
30912+ spin_lock(&file->f_lock);
30913+ flags = file->f_flags;
30914+ spin_unlock(&file->f_lock);
30915+
30916+ return flags;
30917+}
1308ab2a 30918+
7e9cd9fe 30919+#if 0 /* reserved */
1facf9fc 30920+static inline void vfsub_file_accessed(struct file *h_file)
30921+{
30922+ file_accessed(h_file);
30923+ vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
30924+}
7e9cd9fe 30925+#endif
1facf9fc 30926+
30927+static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
30928+ struct dentry *h_dentry)
30929+{
30930+ struct path h_path = {
30931+ .dentry = h_dentry,
30932+ .mnt = h_mnt
30933+ };
92d182d2 30934+ touch_atime(&h_path);
1facf9fc 30935+ vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
30936+}
30937+
0c3ec466
AM
30938+static inline int vfsub_update_time(struct inode *h_inode, struct timespec *ts,
30939+ int flags)
30940+{
7e9cd9fe 30941+ return generic_update_time(h_inode, ts, flags);
0c3ec466
AM
30942+ /* no vfsub_update_h_iattr() since we don't have struct path */
30943+}
30944+
4a4d8108
AM
30945+long vfsub_splice_to(struct file *in, loff_t *ppos,
30946+ struct pipe_inode_info *pipe, size_t len,
30947+ unsigned int flags);
30948+long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
30949+ loff_t *ppos, size_t len, unsigned int flags);
c06a8ce3
AM
30950+
30951+static inline long vfsub_truncate(struct path *path, loff_t length)
30952+{
30953+ long err;
076b876e 30954+
c06a8ce3
AM
30955+ lockdep_off();
30956+ err = vfs_truncate(path, length);
30957+ lockdep_on();
30958+ return err;
30959+}
30960+
4a4d8108
AM
30961+int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
30962+ struct file *h_file);
53392da6 30963+int vfsub_fsync(struct file *file, struct path *path, int datasync);
4a4d8108 30964+
1facf9fc 30965+/* ---------------------------------------------------------------------- */
30966+
30967+static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
30968+{
30969+ loff_t err;
30970+
2cbb1c4b 30971+ lockdep_off();
1facf9fc 30972+ err = vfs_llseek(file, offset, origin);
2cbb1c4b 30973+ lockdep_on();
1facf9fc 30974+ return err;
30975+}
30976+
30977+/* ---------------------------------------------------------------------- */
30978+
4a4d8108
AM
30979+int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
30980+int vfsub_sio_rmdir(struct inode *dir, struct path *path);
523b37e3
AM
30981+int vfsub_sio_notify_change(struct path *path, struct iattr *ia,
30982+ struct inode **delegated_inode);
30983+int vfsub_notify_change(struct path *path, struct iattr *ia,
30984+ struct inode **delegated_inode);
30985+int vfsub_unlink(struct inode *dir, struct path *path,
30986+ struct inode **delegated_inode, int force);
4a4d8108 30987+
c1595e42
JR
30988+/* ---------------------------------------------------------------------- */
30989+
30990+static inline int vfsub_setxattr(struct dentry *dentry, const char *name,
30991+ const void *value, size_t size, int flags)
30992+{
30993+ int err;
30994+
30995+ lockdep_off();
30996+ err = vfs_setxattr(dentry, name, value, size, flags);
30997+ lockdep_on();
30998+
30999+ return err;
31000+}
31001+
31002+static inline int vfsub_removexattr(struct dentry *dentry, const char *name)
31003+{
31004+ int err;
31005+
31006+ lockdep_off();
31007+ err = vfs_removexattr(dentry, name);
31008+ lockdep_on();
31009+
31010+ return err;
31011+}
31012+
1facf9fc 31013+#endif /* __KERNEL__ */
31014+#endif /* __AUFS_VFSUB_H__ */
7f207e10
AM
31015diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
31016--- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 31017+++ linux/fs/aufs/wbr_policy.c 2015-09-24 10:47:58.258053165 +0200
076b876e 31018@@ -0,0 +1,765 @@
1facf9fc 31019+/*
2000de60 31020+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 31021+ *
31022+ * This program, aufs is free software; you can redistribute it and/or modify
31023+ * it under the terms of the GNU General Public License as published by
31024+ * the Free Software Foundation; either version 2 of the License, or
31025+ * (at your option) any later version.
dece6358
AM
31026+ *
31027+ * This program is distributed in the hope that it will be useful,
31028+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
31029+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31030+ * GNU General Public License for more details.
31031+ *
31032+ * You should have received a copy of the GNU General Public License
523b37e3 31033+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 31034+ */
31035+
31036+/*
31037+ * policies for selecting one among multiple writable branches
31038+ */
31039+
31040+#include <linux/statfs.h>
31041+#include "aufs.h"
31042+
31043+/* subset of cpup_attr() */
31044+static noinline_for_stack
31045+int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
31046+{
31047+ int err, sbits;
31048+ struct iattr ia;
31049+ struct inode *h_isrc;
31050+
5527c038 31051+ h_isrc = d_inode(h_src);
1facf9fc 31052+ ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
31053+ ia.ia_mode = h_isrc->i_mode;
31054+ ia.ia_uid = h_isrc->i_uid;
31055+ ia.ia_gid = h_isrc->i_gid;
31056+ sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
5527c038 31057+ au_cpup_attr_flags(d_inode(h_path->dentry), h_isrc->i_flags);
523b37e3
AM
31058+ /* no delegation since it is just created */
31059+ err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
1facf9fc 31060+
31061+ /* is this nfs only? */
31062+ if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
31063+ ia.ia_valid = ATTR_FORCE | ATTR_MODE;
31064+ ia.ia_mode = h_isrc->i_mode;
523b37e3 31065+ err = vfsub_sio_notify_change(h_path, &ia, /*delegated*/NULL);
1facf9fc 31066+ }
31067+
31068+ return err;
31069+}
31070+
31071+#define AuCpdown_PARENT_OPQ 1
31072+#define AuCpdown_WHED (1 << 1)
31073+#define AuCpdown_MADE_DIR (1 << 2)
31074+#define AuCpdown_DIROPQ (1 << 3)
31075+#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
7f207e10
AM
31076+#define au_fset_cpdown(flags, name) \
31077+ do { (flags) |= AuCpdown_##name; } while (0)
31078+#define au_fclr_cpdown(flags, name) \
31079+ do { (flags) &= ~AuCpdown_##name; } while (0)
1facf9fc 31080+
1facf9fc 31081+static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
c2b27bf2 31082+ unsigned int *flags)
1facf9fc 31083+{
31084+ int err;
31085+ struct dentry *opq_dentry;
31086+
31087+ opq_dentry = au_diropq_create(dentry, bdst);
31088+ err = PTR_ERR(opq_dentry);
31089+ if (IS_ERR(opq_dentry))
31090+ goto out;
31091+ dput(opq_dentry);
c2b27bf2 31092+ au_fset_cpdown(*flags, DIROPQ);
1facf9fc 31093+
4f0767ce 31094+out:
1facf9fc 31095+ return err;
31096+}
31097+
31098+static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
31099+ struct inode *dir, aufs_bindex_t bdst)
31100+{
31101+ int err;
31102+ struct path h_path;
31103+ struct au_branch *br;
31104+
31105+ br = au_sbr(dentry->d_sb, bdst);
31106+ h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
31107+ err = PTR_ERR(h_path.dentry);
31108+ if (IS_ERR(h_path.dentry))
31109+ goto out;
31110+
31111+ err = 0;
5527c038 31112+ if (d_is_positive(h_path.dentry)) {
86dc4139 31113+ h_path.mnt = au_br_mnt(br);
1facf9fc 31114+ err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
31115+ dentry);
31116+ }
31117+ dput(h_path.dentry);
31118+
4f0767ce 31119+out:
1facf9fc 31120+ return err;
31121+}
31122+
31123+static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
86dc4139 31124+ struct au_pin *pin,
1facf9fc 31125+ struct dentry *h_parent, void *arg)
31126+{
31127+ int err, rerr;
4a4d8108 31128+ aufs_bindex_t bopq, bstart;
1facf9fc 31129+ struct path h_path;
31130+ struct dentry *parent;
31131+ struct inode *h_dir, *h_inode, *inode, *dir;
c2b27bf2 31132+ unsigned int *flags = arg;
1facf9fc 31133+
31134+ bstart = au_dbstart(dentry);
31135+ /* dentry is di-locked */
31136+ parent = dget_parent(dentry);
5527c038
JR
31137+ dir = d_inode(parent);
31138+ h_dir = d_inode(h_parent);
1facf9fc 31139+ AuDebugOn(h_dir != au_h_iptr(dir, bdst));
31140+ IMustLock(h_dir);
31141+
86dc4139 31142+ err = au_lkup_neg(dentry, bdst, /*wh*/0);
1facf9fc 31143+ if (unlikely(err < 0))
31144+ goto out;
31145+ h_path.dentry = au_h_dptr(dentry, bdst);
31146+ h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
31147+ err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path,
31148+ S_IRWXU | S_IRUGO | S_IXUGO);
31149+ if (unlikely(err))
31150+ goto out_put;
c2b27bf2 31151+ au_fset_cpdown(*flags, MADE_DIR);
1facf9fc 31152+
1facf9fc 31153+ bopq = au_dbdiropq(dentry);
c2b27bf2
AM
31154+ au_fclr_cpdown(*flags, WHED);
31155+ au_fclr_cpdown(*flags, DIROPQ);
1facf9fc 31156+ if (au_dbwh(dentry) == bdst)
c2b27bf2
AM
31157+ au_fset_cpdown(*flags, WHED);
31158+ if (!au_ftest_cpdown(*flags, PARENT_OPQ) && bopq <= bdst)
31159+ au_fset_cpdown(*flags, PARENT_OPQ);
5527c038 31160+ h_inode = d_inode(h_path.dentry);
1facf9fc 31161+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
c2b27bf2
AM
31162+ if (au_ftest_cpdown(*flags, WHED)) {
31163+ err = au_cpdown_dir_opq(dentry, bdst, flags);
1facf9fc 31164+ if (unlikely(err)) {
31165+ mutex_unlock(&h_inode->i_mutex);
31166+ goto out_dir;
31167+ }
31168+ }
31169+
31170+ err = au_cpdown_attr(&h_path, au_h_dptr(dentry, bstart));
31171+ mutex_unlock(&h_inode->i_mutex);
31172+ if (unlikely(err))
31173+ goto out_opq;
31174+
c2b27bf2 31175+ if (au_ftest_cpdown(*flags, WHED)) {
1facf9fc 31176+ err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
31177+ if (unlikely(err))
31178+ goto out_opq;
31179+ }
31180+
5527c038 31181+ inode = d_inode(dentry);
1facf9fc 31182+ if (au_ibend(inode) < bdst)
31183+ au_set_ibend(inode, bdst);
31184+ au_set_h_iptr(inode, bdst, au_igrab(h_inode),
31185+ au_hi_flags(inode, /*isdir*/1));
076b876e 31186+ au_fhsm_wrote(dentry->d_sb, bdst, /*force*/0);
1facf9fc 31187+ goto out; /* success */
31188+
31189+ /* revert */
4f0767ce 31190+out_opq:
c2b27bf2 31191+ if (au_ftest_cpdown(*flags, DIROPQ)) {
1facf9fc 31192+ mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
31193+ rerr = au_diropq_remove(dentry, bdst);
31194+ mutex_unlock(&h_inode->i_mutex);
31195+ if (unlikely(rerr)) {
523b37e3
AM
31196+ AuIOErr("failed removing diropq for %pd b%d (%d)\n",
31197+ dentry, bdst, rerr);
1facf9fc 31198+ err = -EIO;
31199+ goto out;
31200+ }
31201+ }
4f0767ce 31202+out_dir:
c2b27bf2 31203+ if (au_ftest_cpdown(*flags, MADE_DIR)) {
1facf9fc 31204+ rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
31205+ if (unlikely(rerr)) {
523b37e3
AM
31206+ AuIOErr("failed removing %pd b%d (%d)\n",
31207+ dentry, bdst, rerr);
1facf9fc 31208+ err = -EIO;
31209+ }
31210+ }
4f0767ce 31211+out_put:
1facf9fc 31212+ au_set_h_dptr(dentry, bdst, NULL);
31213+ if (au_dbend(dentry) == bdst)
31214+ au_update_dbend(dentry);
4f0767ce 31215+out:
1facf9fc 31216+ dput(parent);
31217+ return err;
31218+}
31219+
31220+int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
31221+{
31222+ int err;
c2b27bf2 31223+ unsigned int flags;
1facf9fc 31224+
c2b27bf2
AM
31225+ flags = 0;
31226+ err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &flags);
1facf9fc 31227+
31228+ return err;
31229+}
31230+
31231+/* ---------------------------------------------------------------------- */
31232+
31233+/* policies for create */
31234+
c2b27bf2 31235+int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
4a4d8108
AM
31236+{
31237+ int err, i, j, ndentry;
31238+ aufs_bindex_t bopq;
31239+ struct au_dcsub_pages dpages;
31240+ struct au_dpage *dpage;
31241+ struct dentry **dentries, *parent, *d;
31242+
31243+ err = au_dpages_init(&dpages, GFP_NOFS);
31244+ if (unlikely(err))
31245+ goto out;
31246+ parent = dget_parent(dentry);
027c5e7a 31247+ err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
4a4d8108
AM
31248+ if (unlikely(err))
31249+ goto out_free;
31250+
31251+ err = bindex;
31252+ for (i = 0; i < dpages.ndpage; i++) {
31253+ dpage = dpages.dpages + i;
31254+ dentries = dpage->dentries;
31255+ ndentry = dpage->ndentry;
31256+ for (j = 0; j < ndentry; j++) {
31257+ d = dentries[j];
31258+ di_read_lock_parent2(d, !AuLock_IR);
31259+ bopq = au_dbdiropq(d);
31260+ di_read_unlock(d, !AuLock_IR);
31261+ if (bopq >= 0 && bopq < err)
31262+ err = bopq;
31263+ }
31264+ }
31265+
31266+out_free:
31267+ dput(parent);
31268+ au_dpages_free(&dpages);
31269+out:
31270+ return err;
31271+}
31272+
1facf9fc 31273+static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
31274+{
31275+ for (; bindex >= 0; bindex--)
31276+ if (!au_br_rdonly(au_sbr(sb, bindex)))
31277+ return bindex;
31278+ return -EROFS;
31279+}
31280+
31281+/* top down parent */
392086de
AM
31282+static int au_wbr_create_tdp(struct dentry *dentry,
31283+ unsigned int flags __maybe_unused)
1facf9fc 31284+{
31285+ int err;
31286+ aufs_bindex_t bstart, bindex;
31287+ struct super_block *sb;
31288+ struct dentry *parent, *h_parent;
31289+
31290+ sb = dentry->d_sb;
31291+ bstart = au_dbstart(dentry);
31292+ err = bstart;
31293+ if (!au_br_rdonly(au_sbr(sb, bstart)))
31294+ goto out;
31295+
31296+ err = -EROFS;
31297+ parent = dget_parent(dentry);
31298+ for (bindex = au_dbstart(parent); bindex < bstart; bindex++) {
31299+ h_parent = au_h_dptr(parent, bindex);
5527c038 31300+ if (!h_parent || d_is_negative(h_parent))
1facf9fc 31301+ continue;
31302+
31303+ if (!au_br_rdonly(au_sbr(sb, bindex))) {
31304+ err = bindex;
31305+ break;
31306+ }
31307+ }
31308+ dput(parent);
31309+
31310+ /* bottom up here */
4a4d8108 31311+ if (unlikely(err < 0)) {
1facf9fc 31312+ err = au_wbr_bu(sb, bstart - 1);
4a4d8108
AM
31313+ if (err >= 0)
31314+ err = au_wbr_nonopq(dentry, err);
31315+ }
1facf9fc 31316+
4f0767ce 31317+out:
1facf9fc 31318+ AuDbg("b%d\n", err);
31319+ return err;
31320+}
31321+
31322+/* ---------------------------------------------------------------------- */
31323+
31324+/* an exception for the policy other than tdp */
31325+static int au_wbr_create_exp(struct dentry *dentry)
31326+{
31327+ int err;
31328+ aufs_bindex_t bwh, bdiropq;
31329+ struct dentry *parent;
31330+
31331+ err = -1;
31332+ bwh = au_dbwh(dentry);
31333+ parent = dget_parent(dentry);
31334+ bdiropq = au_dbdiropq(parent);
31335+ if (bwh >= 0) {
31336+ if (bdiropq >= 0)
31337+ err = min(bdiropq, bwh);
31338+ else
31339+ err = bwh;
31340+ AuDbg("%d\n", err);
31341+ } else if (bdiropq >= 0) {
31342+ err = bdiropq;
31343+ AuDbg("%d\n", err);
31344+ }
31345+ dput(parent);
31346+
4a4d8108
AM
31347+ if (err >= 0)
31348+ err = au_wbr_nonopq(dentry, err);
31349+
1facf9fc 31350+ if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
31351+ err = -1;
31352+
31353+ AuDbg("%d\n", err);
31354+ return err;
31355+}
31356+
31357+/* ---------------------------------------------------------------------- */
31358+
31359+/* round robin */
31360+static int au_wbr_create_init_rr(struct super_block *sb)
31361+{
31362+ int err;
31363+
31364+ err = au_wbr_bu(sb, au_sbend(sb));
31365+ atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
dece6358 31366+ /* smp_mb(); */
1facf9fc 31367+
31368+ AuDbg("b%d\n", err);
31369+ return err;
31370+}
31371+
392086de 31372+static int au_wbr_create_rr(struct dentry *dentry, unsigned int flags)
1facf9fc 31373+{
31374+ int err, nbr;
31375+ unsigned int u;
31376+ aufs_bindex_t bindex, bend;
31377+ struct super_block *sb;
31378+ atomic_t *next;
31379+
31380+ err = au_wbr_create_exp(dentry);
31381+ if (err >= 0)
31382+ goto out;
31383+
31384+ sb = dentry->d_sb;
31385+ next = &au_sbi(sb)->si_wbr_rr_next;
31386+ bend = au_sbend(sb);
31387+ nbr = bend + 1;
31388+ for (bindex = 0; bindex <= bend; bindex++) {
392086de 31389+ if (!au_ftest_wbr(flags, DIR)) {
1facf9fc 31390+ err = atomic_dec_return(next) + 1;
31391+ /* modulo for 0 is meaningless */
31392+ if (unlikely(!err))
31393+ err = atomic_dec_return(next) + 1;
31394+ } else
31395+ err = atomic_read(next);
31396+ AuDbg("%d\n", err);
31397+ u = err;
31398+ err = u % nbr;
31399+ AuDbg("%d\n", err);
31400+ if (!au_br_rdonly(au_sbr(sb, err)))
31401+ break;
31402+ err = -EROFS;
31403+ }
31404+
4a4d8108
AM
31405+ if (err >= 0)
31406+ err = au_wbr_nonopq(dentry, err);
31407+
4f0767ce 31408+out:
1facf9fc 31409+ AuDbg("%d\n", err);
31410+ return err;
31411+}
31412+
31413+/* ---------------------------------------------------------------------- */
31414+
31415+/* most free space */
392086de 31416+static void au_mfs(struct dentry *dentry, struct dentry *parent)
1facf9fc 31417+{
31418+ struct super_block *sb;
31419+ struct au_branch *br;
31420+ struct au_wbr_mfs *mfs;
392086de 31421+ struct dentry *h_parent;
1facf9fc 31422+ aufs_bindex_t bindex, bend;
31423+ int err;
31424+ unsigned long long b, bavail;
7f207e10 31425+ struct path h_path;
1facf9fc 31426+ /* reduce the stack usage */
31427+ struct kstatfs *st;
31428+
31429+ st = kmalloc(sizeof(*st), GFP_NOFS);
31430+ if (unlikely(!st)) {
31431+ AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
31432+ return;
31433+ }
31434+
31435+ bavail = 0;
31436+ sb = dentry->d_sb;
31437+ mfs = &au_sbi(sb)->si_wbr_mfs;
dece6358 31438+ MtxMustLock(&mfs->mfs_lock);
1facf9fc 31439+ mfs->mfs_bindex = -EROFS;
31440+ mfs->mfsrr_bytes = 0;
392086de
AM
31441+ if (!parent) {
31442+ bindex = 0;
31443+ bend = au_sbend(sb);
31444+ } else {
31445+ bindex = au_dbstart(parent);
31446+ bend = au_dbtaildir(parent);
31447+ }
31448+
31449+ for (; bindex <= bend; bindex++) {
31450+ if (parent) {
31451+ h_parent = au_h_dptr(parent, bindex);
5527c038 31452+ if (!h_parent || d_is_negative(h_parent))
392086de
AM
31453+ continue;
31454+ }
1facf9fc 31455+ br = au_sbr(sb, bindex);
31456+ if (au_br_rdonly(br))
31457+ continue;
31458+
31459+ /* sb->s_root for NFS is unreliable */
86dc4139 31460+ h_path.mnt = au_br_mnt(br);
7f207e10
AM
31461+ h_path.dentry = h_path.mnt->mnt_root;
31462+ err = vfs_statfs(&h_path, st);
1facf9fc 31463+ if (unlikely(err)) {
31464+ AuWarn1("failed statfs, b%d, %d\n", bindex, err);
31465+ continue;
31466+ }
31467+
31468+ /* when the available size is equal, select the lower one */
31469+ BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
31470+ || sizeof(b) < sizeof(st->f_bsize));
31471+ b = st->f_bavail * st->f_bsize;
31472+ br->br_wbr->wbr_bytes = b;
31473+ if (b >= bavail) {
31474+ bavail = b;
31475+ mfs->mfs_bindex = bindex;
31476+ mfs->mfs_jiffy = jiffies;
31477+ }
31478+ }
31479+
31480+ mfs->mfsrr_bytes = bavail;
31481+ AuDbg("b%d\n", mfs->mfs_bindex);
31482+ kfree(st);
31483+}
31484+
392086de 31485+static int au_wbr_create_mfs(struct dentry *dentry, unsigned int flags)
1facf9fc 31486+{
31487+ int err;
392086de 31488+ struct dentry *parent;
1facf9fc 31489+ struct super_block *sb;
31490+ struct au_wbr_mfs *mfs;
31491+
31492+ err = au_wbr_create_exp(dentry);
31493+ if (err >= 0)
31494+ goto out;
31495+
31496+ sb = dentry->d_sb;
392086de
AM
31497+ parent = NULL;
31498+ if (au_ftest_wbr(flags, PARENT))
31499+ parent = dget_parent(dentry);
1facf9fc 31500+ mfs = &au_sbi(sb)->si_wbr_mfs;
31501+ mutex_lock(&mfs->mfs_lock);
31502+ if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
31503+ || mfs->mfs_bindex < 0
31504+ || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
392086de 31505+ au_mfs(dentry, parent);
1facf9fc 31506+ mutex_unlock(&mfs->mfs_lock);
31507+ err = mfs->mfs_bindex;
392086de 31508+ dput(parent);
1facf9fc 31509+
4a4d8108
AM
31510+ if (err >= 0)
31511+ err = au_wbr_nonopq(dentry, err);
31512+
4f0767ce 31513+out:
1facf9fc 31514+ AuDbg("b%d\n", err);
31515+ return err;
31516+}
31517+
31518+static int au_wbr_create_init_mfs(struct super_block *sb)
31519+{
31520+ struct au_wbr_mfs *mfs;
31521+
31522+ mfs = &au_sbi(sb)->si_wbr_mfs;
31523+ mutex_init(&mfs->mfs_lock);
31524+ mfs->mfs_jiffy = 0;
31525+ mfs->mfs_bindex = -EROFS;
31526+
31527+ return 0;
31528+}
31529+
31530+static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
31531+{
31532+ mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
31533+ return 0;
31534+}
31535+
31536+/* ---------------------------------------------------------------------- */
31537+
31538+/* most free space and then round robin */
392086de 31539+static int au_wbr_create_mfsrr(struct dentry *dentry, unsigned int flags)
1facf9fc 31540+{
31541+ int err;
31542+ struct au_wbr_mfs *mfs;
31543+
392086de 31544+ err = au_wbr_create_mfs(dentry, flags);
1facf9fc 31545+ if (err >= 0) {
31546+ mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
dece6358 31547+ mutex_lock(&mfs->mfs_lock);
1facf9fc 31548+ if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
392086de 31549+ err = au_wbr_create_rr(dentry, flags);
dece6358 31550+ mutex_unlock(&mfs->mfs_lock);
1facf9fc 31551+ }
31552+
31553+ AuDbg("b%d\n", err);
31554+ return err;
31555+}
31556+
31557+static int au_wbr_create_init_mfsrr(struct super_block *sb)
31558+{
31559+ int err;
31560+
31561+ au_wbr_create_init_mfs(sb); /* ignore */
31562+ err = au_wbr_create_init_rr(sb);
31563+
31564+ return err;
31565+}
31566+
31567+/* ---------------------------------------------------------------------- */
31568+
31569+/* top down parent and most free space */
392086de 31570+static int au_wbr_create_pmfs(struct dentry *dentry, unsigned int flags)
1facf9fc 31571+{
31572+ int err, e2;
31573+ unsigned long long b;
31574+ aufs_bindex_t bindex, bstart, bend;
31575+ struct super_block *sb;
31576+ struct dentry *parent, *h_parent;
31577+ struct au_branch *br;
31578+
392086de 31579+ err = au_wbr_create_tdp(dentry, flags);
1facf9fc 31580+ if (unlikely(err < 0))
31581+ goto out;
31582+ parent = dget_parent(dentry);
31583+ bstart = au_dbstart(parent);
31584+ bend = au_dbtaildir(parent);
31585+ if (bstart == bend)
31586+ goto out_parent; /* success */
31587+
392086de 31588+ e2 = au_wbr_create_mfs(dentry, flags);
1facf9fc 31589+ if (e2 < 0)
31590+ goto out_parent; /* success */
31591+
31592+ /* when the available size is equal, select upper one */
31593+ sb = dentry->d_sb;
31594+ br = au_sbr(sb, err);
31595+ b = br->br_wbr->wbr_bytes;
31596+ AuDbg("b%d, %llu\n", err, b);
31597+
31598+ for (bindex = bstart; bindex <= bend; bindex++) {
31599+ h_parent = au_h_dptr(parent, bindex);
5527c038 31600+ if (!h_parent || d_is_negative(h_parent))
1facf9fc 31601+ continue;
31602+
31603+ br = au_sbr(sb, bindex);
31604+ if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
31605+ b = br->br_wbr->wbr_bytes;
31606+ err = bindex;
31607+ AuDbg("b%d, %llu\n", err, b);
31608+ }
31609+ }
31610+
4a4d8108
AM
31611+ if (err >= 0)
31612+ err = au_wbr_nonopq(dentry, err);
31613+
4f0767ce 31614+out_parent:
1facf9fc 31615+ dput(parent);
4f0767ce 31616+out:
1facf9fc 31617+ AuDbg("b%d\n", err);
31618+ return err;
31619+}
31620+
31621+/* ---------------------------------------------------------------------- */
31622+
392086de
AM
31623+/*
31624+ * - top down parent
31625+ * - most free space with parent
31626+ * - most free space round-robin regardless parent
31627+ */
31628+static int au_wbr_create_pmfsrr(struct dentry *dentry, unsigned int flags)
31629+{
31630+ int err;
31631+ unsigned long long watermark;
31632+ struct super_block *sb;
31633+ struct au_branch *br;
31634+ struct au_wbr_mfs *mfs;
31635+
31636+ err = au_wbr_create_pmfs(dentry, flags | AuWbr_PARENT);
31637+ if (unlikely(err < 0))
31638+ goto out;
31639+
31640+ sb = dentry->d_sb;
31641+ br = au_sbr(sb, err);
31642+ mfs = &au_sbi(sb)->si_wbr_mfs;
31643+ mutex_lock(&mfs->mfs_lock);
31644+ watermark = mfs->mfsrr_watermark;
31645+ mutex_unlock(&mfs->mfs_lock);
31646+ if (br->br_wbr->wbr_bytes < watermark)
31647+ /* regardless the parent dir */
31648+ err = au_wbr_create_mfsrr(dentry, flags);
31649+
31650+out:
31651+ AuDbg("b%d\n", err);
31652+ return err;
31653+}
31654+
31655+/* ---------------------------------------------------------------------- */
31656+
1facf9fc 31657+/* policies for copyup */
31658+
31659+/* top down parent */
31660+static int au_wbr_copyup_tdp(struct dentry *dentry)
31661+{
392086de 31662+ return au_wbr_create_tdp(dentry, /*flags, anything is ok*/0);
1facf9fc 31663+}
31664+
31665+/* bottom up parent */
31666+static int au_wbr_copyup_bup(struct dentry *dentry)
31667+{
31668+ int err;
31669+ aufs_bindex_t bindex, bstart;
31670+ struct dentry *parent, *h_parent;
31671+ struct super_block *sb;
31672+
31673+ err = -EROFS;
31674+ sb = dentry->d_sb;
31675+ parent = dget_parent(dentry);
31676+ bstart = au_dbstart(parent);
31677+ for (bindex = au_dbstart(dentry); bindex >= bstart; bindex--) {
31678+ h_parent = au_h_dptr(parent, bindex);
5527c038 31679+ if (!h_parent || d_is_negative(h_parent))
1facf9fc 31680+ continue;
31681+
31682+ if (!au_br_rdonly(au_sbr(sb, bindex))) {
31683+ err = bindex;
31684+ break;
31685+ }
31686+ }
31687+ dput(parent);
31688+
31689+ /* bottom up here */
31690+ if (unlikely(err < 0))
31691+ err = au_wbr_bu(sb, bstart - 1);
31692+
31693+ AuDbg("b%d\n", err);
31694+ return err;
31695+}
31696+
31697+/* bottom up */
076b876e 31698+int au_wbr_do_copyup_bu(struct dentry *dentry, aufs_bindex_t bstart)
1facf9fc 31699+{
31700+ int err;
31701+
4a4d8108
AM
31702+ err = au_wbr_bu(dentry->d_sb, bstart);
31703+ AuDbg("b%d\n", err);
31704+ if (err > bstart)
31705+ err = au_wbr_nonopq(dentry, err);
1facf9fc 31706+
31707+ AuDbg("b%d\n", err);
31708+ return err;
31709+}
31710+
076b876e
AM
31711+static int au_wbr_copyup_bu(struct dentry *dentry)
31712+{
31713+ int err;
31714+ aufs_bindex_t bstart;
31715+
31716+ bstart = au_dbstart(dentry);
31717+ err = au_wbr_do_copyup_bu(dentry, bstart);
31718+ return err;
31719+}
31720+
1facf9fc 31721+/* ---------------------------------------------------------------------- */
31722+
31723+struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
31724+ [AuWbrCopyup_TDP] = {
31725+ .copyup = au_wbr_copyup_tdp
31726+ },
31727+ [AuWbrCopyup_BUP] = {
31728+ .copyup = au_wbr_copyup_bup
31729+ },
31730+ [AuWbrCopyup_BU] = {
31731+ .copyup = au_wbr_copyup_bu
31732+ }
31733+};
31734+
31735+struct au_wbr_create_operations au_wbr_create_ops[] = {
31736+ [AuWbrCreate_TDP] = {
31737+ .create = au_wbr_create_tdp
31738+ },
31739+ [AuWbrCreate_RR] = {
31740+ .create = au_wbr_create_rr,
31741+ .init = au_wbr_create_init_rr
31742+ },
31743+ [AuWbrCreate_MFS] = {
31744+ .create = au_wbr_create_mfs,
31745+ .init = au_wbr_create_init_mfs,
31746+ .fin = au_wbr_create_fin_mfs
31747+ },
31748+ [AuWbrCreate_MFSV] = {
31749+ .create = au_wbr_create_mfs,
31750+ .init = au_wbr_create_init_mfs,
31751+ .fin = au_wbr_create_fin_mfs
31752+ },
31753+ [AuWbrCreate_MFSRR] = {
31754+ .create = au_wbr_create_mfsrr,
31755+ .init = au_wbr_create_init_mfsrr,
31756+ .fin = au_wbr_create_fin_mfs
31757+ },
31758+ [AuWbrCreate_MFSRRV] = {
31759+ .create = au_wbr_create_mfsrr,
31760+ .init = au_wbr_create_init_mfsrr,
31761+ .fin = au_wbr_create_fin_mfs
31762+ },
31763+ [AuWbrCreate_PMFS] = {
31764+ .create = au_wbr_create_pmfs,
31765+ .init = au_wbr_create_init_mfs,
31766+ .fin = au_wbr_create_fin_mfs
31767+ },
31768+ [AuWbrCreate_PMFSV] = {
31769+ .create = au_wbr_create_pmfs,
31770+ .init = au_wbr_create_init_mfs,
31771+ .fin = au_wbr_create_fin_mfs
392086de
AM
31772+ },
31773+ [AuWbrCreate_PMFSRR] = {
31774+ .create = au_wbr_create_pmfsrr,
31775+ .init = au_wbr_create_init_mfsrr,
31776+ .fin = au_wbr_create_fin_mfs
31777+ },
31778+ [AuWbrCreate_PMFSRRV] = {
31779+ .create = au_wbr_create_pmfsrr,
31780+ .init = au_wbr_create_init_mfsrr,
31781+ .fin = au_wbr_create_fin_mfs
1facf9fc 31782+ }
31783+};
7f207e10
AM
31784diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
31785--- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 31786+++ linux/fs/aufs/whout.c 2015-09-24 10:47:58.258053165 +0200
5527c038 31787@@ -0,0 +1,1063 @@
1facf9fc 31788+/*
2000de60 31789+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 31790+ *
31791+ * This program, aufs is free software; you can redistribute it and/or modify
31792+ * it under the terms of the GNU General Public License as published by
31793+ * the Free Software Foundation; either version 2 of the License, or
31794+ * (at your option) any later version.
dece6358
AM
31795+ *
31796+ * This program is distributed in the hope that it will be useful,
31797+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
31798+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31799+ * GNU General Public License for more details.
31800+ *
31801+ * You should have received a copy of the GNU General Public License
523b37e3 31802+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 31803+ */
31804+
31805+/*
31806+ * whiteout for logical deletion and opaque directory
31807+ */
31808+
1facf9fc 31809+#include "aufs.h"
31810+
31811+#define WH_MASK S_IRUGO
31812+
31813+/*
31814+ * If a directory contains this file, then it is opaque. We start with the
31815+ * .wh. flag so that it is blocked by lookup.
31816+ */
0c3ec466
AM
31817+static struct qstr diropq_name = QSTR_INIT(AUFS_WH_DIROPQ,
31818+ sizeof(AUFS_WH_DIROPQ) - 1);
1facf9fc 31819+
31820+/*
31821+ * generate whiteout name, which is NOT terminated by NULL.
31822+ * @name: original d_name.name
31823+ * @len: original d_name.len
31824+ * @wh: whiteout qstr
31825+ * returns zero when succeeds, otherwise error.
31826+ * succeeded value as wh->name should be freed by kfree().
31827+ */
31828+int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
31829+{
31830+ char *p;
31831+
31832+ if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
31833+ return -ENAMETOOLONG;
31834+
31835+ wh->len = name->len + AUFS_WH_PFX_LEN;
31836+ p = kmalloc(wh->len, GFP_NOFS);
31837+ wh->name = p;
31838+ if (p) {
31839+ memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
31840+ memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
31841+ /* smp_mb(); */
31842+ return 0;
31843+ }
31844+ return -ENOMEM;
31845+}
31846+
31847+/* ---------------------------------------------------------------------- */
31848+
31849+/*
31850+ * test if the @wh_name exists under @h_parent.
31851+ * @try_sio specifies the necessary of super-io.
31852+ */
076b876e 31853+int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio)
1facf9fc 31854+{
31855+ int err;
31856+ struct dentry *wh_dentry;
1facf9fc 31857+
1facf9fc 31858+ if (!try_sio)
b4510431 31859+ wh_dentry = vfsub_lkup_one(wh_name, h_parent);
1facf9fc 31860+ else
076b876e 31861+ wh_dentry = au_sio_lkup_one(wh_name, h_parent);
1facf9fc 31862+ err = PTR_ERR(wh_dentry);
2000de60
JR
31863+ if (IS_ERR(wh_dentry)) {
31864+ if (err == -ENAMETOOLONG)
31865+ err = 0;
1facf9fc 31866+ goto out;
2000de60 31867+ }
1facf9fc 31868+
31869+ err = 0;
5527c038 31870+ if (d_is_negative(wh_dentry))
1facf9fc 31871+ goto out_wh; /* success */
31872+
31873+ err = 1;
7e9cd9fe 31874+ if (d_is_reg(wh_dentry))
1facf9fc 31875+ goto out_wh; /* success */
31876+
31877+ err = -EIO;
523b37e3 31878+ AuIOErr("%pd Invalid whiteout entry type 0%o.\n",
5527c038 31879+ wh_dentry, d_inode(wh_dentry)->i_mode);
1facf9fc 31880+
4f0767ce 31881+out_wh:
1facf9fc 31882+ dput(wh_dentry);
4f0767ce 31883+out:
1facf9fc 31884+ return err;
31885+}
31886+
31887+/*
31888+ * test if the @h_dentry sets opaque or not.
31889+ */
076b876e 31890+int au_diropq_test(struct dentry *h_dentry)
1facf9fc 31891+{
31892+ int err;
31893+ struct inode *h_dir;
31894+
5527c038 31895+ h_dir = d_inode(h_dentry);
076b876e 31896+ err = au_wh_test(h_dentry, &diropq_name,
1facf9fc 31897+ au_test_h_perm_sio(h_dir, MAY_EXEC));
31898+ return err;
31899+}
31900+
31901+/*
31902+ * returns a negative dentry whose name is unique and temporary.
31903+ */
31904+struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
31905+ struct qstr *prefix)
31906+{
1facf9fc 31907+ struct dentry *dentry;
31908+ int i;
027c5e7a 31909+ char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
4a4d8108 31910+ *name, *p;
027c5e7a 31911+ /* strict atomic_t is unnecessary here */
1facf9fc 31912+ static unsigned short cnt;
31913+ struct qstr qs;
31914+
4a4d8108
AM
31915+ BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
31916+
1facf9fc 31917+ name = defname;
027c5e7a
AM
31918+ qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
31919+ if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
1facf9fc 31920+ dentry = ERR_PTR(-ENAMETOOLONG);
4a4d8108 31921+ if (unlikely(qs.len > NAME_MAX))
1facf9fc 31922+ goto out;
31923+ dentry = ERR_PTR(-ENOMEM);
31924+ name = kmalloc(qs.len + 1, GFP_NOFS);
31925+ if (unlikely(!name))
31926+ goto out;
31927+ }
31928+
31929+ /* doubly whiteout-ed */
31930+ memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
31931+ p = name + AUFS_WH_PFX_LEN * 2;
31932+ memcpy(p, prefix->name, prefix->len);
31933+ p += prefix->len;
31934+ *p++ = '.';
4a4d8108 31935+ AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
1facf9fc 31936+
31937+ qs.name = name;
31938+ for (i = 0; i < 3; i++) {
b752ccd1 31939+ sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
076b876e 31940+ dentry = au_sio_lkup_one(&qs, h_parent);
5527c038 31941+ if (IS_ERR(dentry) || d_is_negative(dentry))
1facf9fc 31942+ goto out_name;
31943+ dput(dentry);
31944+ }
0c3ec466 31945+ /* pr_warn("could not get random name\n"); */
1facf9fc 31946+ dentry = ERR_PTR(-EEXIST);
31947+ AuDbg("%.*s\n", AuLNPair(&qs));
31948+ BUG();
31949+
4f0767ce 31950+out_name:
1facf9fc 31951+ if (name != defname)
31952+ kfree(name);
4f0767ce 31953+out:
4a4d8108 31954+ AuTraceErrPtr(dentry);
1facf9fc 31955+ return dentry;
1facf9fc 31956+}
31957+
31958+/*
31959+ * rename the @h_dentry on @br to the whiteouted temporary name.
31960+ */
31961+int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
31962+{
31963+ int err;
31964+ struct path h_path = {
86dc4139 31965+ .mnt = au_br_mnt(br)
1facf9fc 31966+ };
523b37e3 31967+ struct inode *h_dir, *delegated;
1facf9fc 31968+ struct dentry *h_parent;
31969+
31970+ h_parent = h_dentry->d_parent; /* dir inode is locked */
5527c038 31971+ h_dir = d_inode(h_parent);
1facf9fc 31972+ IMustLock(h_dir);
31973+
31974+ h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
31975+ err = PTR_ERR(h_path.dentry);
31976+ if (IS_ERR(h_path.dentry))
31977+ goto out;
31978+
31979+ /* under the same dir, no need to lock_rename() */
523b37e3
AM
31980+ delegated = NULL;
31981+ err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path, &delegated);
1facf9fc 31982+ AuTraceErr(err);
523b37e3
AM
31983+ if (unlikely(err == -EWOULDBLOCK)) {
31984+ pr_warn("cannot retry for NFSv4 delegation"
31985+ " for an internal rename\n");
31986+ iput(delegated);
31987+ }
1facf9fc 31988+ dput(h_path.dentry);
31989+
4f0767ce 31990+out:
4a4d8108 31991+ AuTraceErr(err);
1facf9fc 31992+ return err;
31993+}
31994+
31995+/* ---------------------------------------------------------------------- */
31996+/*
31997+ * functions for removing a whiteout
31998+ */
31999+
32000+static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
32001+{
523b37e3
AM
32002+ int err, force;
32003+ struct inode *delegated;
1facf9fc 32004+
32005+ /*
32006+ * forces superio when the dir has a sticky bit.
32007+ * this may be a violation of unix fs semantics.
32008+ */
32009+ force = (h_dir->i_mode & S_ISVTX)
5527c038 32010+ && !uid_eq(current_fsuid(), d_inode(h_path->dentry)->i_uid);
523b37e3
AM
32011+ delegated = NULL;
32012+ err = vfsub_unlink(h_dir, h_path, &delegated, force);
32013+ if (unlikely(err == -EWOULDBLOCK)) {
32014+ pr_warn("cannot retry for NFSv4 delegation"
32015+ " for an internal unlink\n");
32016+ iput(delegated);
32017+ }
32018+ return err;
1facf9fc 32019+}
32020+
32021+int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
32022+ struct dentry *dentry)
32023+{
32024+ int err;
32025+
32026+ err = do_unlink_wh(h_dir, h_path);
32027+ if (!err && dentry)
32028+ au_set_dbwh(dentry, -1);
32029+
32030+ return err;
32031+}
32032+
32033+static int unlink_wh_name(struct dentry *h_parent, struct qstr *wh,
32034+ struct au_branch *br)
32035+{
32036+ int err;
32037+ struct path h_path = {
86dc4139 32038+ .mnt = au_br_mnt(br)
1facf9fc 32039+ };
32040+
32041+ err = 0;
b4510431 32042+ h_path.dentry = vfsub_lkup_one(wh, h_parent);
1facf9fc 32043+ if (IS_ERR(h_path.dentry))
32044+ err = PTR_ERR(h_path.dentry);
32045+ else {
5527c038
JR
32046+ if (d_is_reg(h_path.dentry))
32047+ err = do_unlink_wh(d_inode(h_parent), &h_path);
1facf9fc 32048+ dput(h_path.dentry);
32049+ }
32050+
32051+ return err;
32052+}
32053+
32054+/* ---------------------------------------------------------------------- */
32055+/*
32056+ * initialize/clean whiteout for a branch
32057+ */
32058+
32059+static void au_wh_clean(struct inode *h_dir, struct path *whpath,
32060+ const int isdir)
32061+{
32062+ int err;
523b37e3 32063+ struct inode *delegated;
1facf9fc 32064+
5527c038 32065+ if (d_is_negative(whpath->dentry))
1facf9fc 32066+ return;
32067+
86dc4139
AM
32068+ if (isdir)
32069+ err = vfsub_rmdir(h_dir, whpath);
523b37e3
AM
32070+ else {
32071+ delegated = NULL;
32072+ err = vfsub_unlink(h_dir, whpath, &delegated, /*force*/0);
32073+ if (unlikely(err == -EWOULDBLOCK)) {
32074+ pr_warn("cannot retry for NFSv4 delegation"
32075+ " for an internal unlink\n");
32076+ iput(delegated);
32077+ }
32078+ }
1facf9fc 32079+ if (unlikely(err))
523b37e3
AM
32080+ pr_warn("failed removing %pd (%d), ignored.\n",
32081+ whpath->dentry, err);
1facf9fc 32082+}
32083+
32084+static int test_linkable(struct dentry *h_root)
32085+{
5527c038 32086+ struct inode *h_dir = d_inode(h_root);
1facf9fc 32087+
32088+ if (h_dir->i_op->link)
32089+ return 0;
32090+
523b37e3
AM
32091+ pr_err("%pd (%s) doesn't support link(2), use noplink and rw+nolwh\n",
32092+ h_root, au_sbtype(h_root->d_sb));
1facf9fc 32093+ return -ENOSYS;
32094+}
32095+
32096+/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
32097+static int au_whdir(struct inode *h_dir, struct path *path)
32098+{
32099+ int err;
32100+
32101+ err = -EEXIST;
5527c038 32102+ if (d_is_negative(path->dentry)) {
1facf9fc 32103+ int mode = S_IRWXU;
32104+
32105+ if (au_test_nfs(path->dentry->d_sb))
32106+ mode |= S_IXUGO;
86dc4139 32107+ err = vfsub_mkdir(h_dir, path, mode);
2000de60 32108+ } else if (d_is_dir(path->dentry))
1facf9fc 32109+ err = 0;
32110+ else
523b37e3 32111+ pr_err("unknown %pd exists\n", path->dentry);
1facf9fc 32112+
32113+ return err;
32114+}
32115+
32116+struct au_wh_base {
32117+ const struct qstr *name;
32118+ struct dentry *dentry;
32119+};
32120+
32121+static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
32122+ struct path *h_path)
32123+{
32124+ h_path->dentry = base[AuBrWh_BASE].dentry;
32125+ au_wh_clean(h_dir, h_path, /*isdir*/0);
32126+ h_path->dentry = base[AuBrWh_PLINK].dentry;
32127+ au_wh_clean(h_dir, h_path, /*isdir*/1);
32128+ h_path->dentry = base[AuBrWh_ORPH].dentry;
32129+ au_wh_clean(h_dir, h_path, /*isdir*/1);
32130+}
32131+
32132+/*
32133+ * returns tri-state,
c1595e42 32134+ * minus: error, caller should print the message
1facf9fc 32135+ * zero: succuess
c1595e42 32136+ * plus: error, caller should NOT print the message
1facf9fc 32137+ */
32138+static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
32139+ int do_plink, struct au_wh_base base[],
32140+ struct path *h_path)
32141+{
32142+ int err;
32143+ struct inode *h_dir;
32144+
5527c038 32145+ h_dir = d_inode(h_root);
1facf9fc 32146+ h_path->dentry = base[AuBrWh_BASE].dentry;
32147+ au_wh_clean(h_dir, h_path, /*isdir*/0);
32148+ h_path->dentry = base[AuBrWh_PLINK].dentry;
32149+ if (do_plink) {
32150+ err = test_linkable(h_root);
32151+ if (unlikely(err)) {
32152+ err = 1;
32153+ goto out;
32154+ }
32155+
32156+ err = au_whdir(h_dir, h_path);
32157+ if (unlikely(err))
32158+ goto out;
32159+ wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
32160+ } else
32161+ au_wh_clean(h_dir, h_path, /*isdir*/1);
32162+ h_path->dentry = base[AuBrWh_ORPH].dentry;
32163+ err = au_whdir(h_dir, h_path);
32164+ if (unlikely(err))
32165+ goto out;
32166+ wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
32167+
4f0767ce 32168+out:
1facf9fc 32169+ return err;
32170+}
32171+
32172+/*
32173+ * for the moment, aufs supports the branch filesystem which does not support
32174+ * link(2). testing on FAT which does not support i_op->setattr() fully either,
32175+ * copyup failed. finally, such filesystem will not be used as the writable
32176+ * branch.
32177+ *
32178+ * returns tri-state, see above.
32179+ */
32180+static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
32181+ int do_plink, struct au_wh_base base[],
32182+ struct path *h_path)
32183+{
32184+ int err;
32185+ struct inode *h_dir;
32186+
1308ab2a 32187+ WbrWhMustWriteLock(wbr);
32188+
1facf9fc 32189+ err = test_linkable(h_root);
32190+ if (unlikely(err)) {
32191+ err = 1;
32192+ goto out;
32193+ }
32194+
32195+ /*
32196+ * todo: should this create be done in /sbin/mount.aufs helper?
32197+ */
32198+ err = -EEXIST;
5527c038
JR
32199+ h_dir = d_inode(h_root);
32200+ if (d_is_negative(base[AuBrWh_BASE].dentry)) {
86dc4139
AM
32201+ h_path->dentry = base[AuBrWh_BASE].dentry;
32202+ err = vfsub_create(h_dir, h_path, WH_MASK, /*want_excl*/true);
7e9cd9fe 32203+ } else if (d_is_reg(base[AuBrWh_BASE].dentry))
1facf9fc 32204+ err = 0;
32205+ else
523b37e3 32206+ pr_err("unknown %pd2 exists\n", base[AuBrWh_BASE].dentry);
1facf9fc 32207+ if (unlikely(err))
32208+ goto out;
32209+
32210+ h_path->dentry = base[AuBrWh_PLINK].dentry;
32211+ if (do_plink) {
32212+ err = au_whdir(h_dir, h_path);
32213+ if (unlikely(err))
32214+ goto out;
32215+ wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
32216+ } else
32217+ au_wh_clean(h_dir, h_path, /*isdir*/1);
32218+ wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
32219+
32220+ h_path->dentry = base[AuBrWh_ORPH].dentry;
32221+ err = au_whdir(h_dir, h_path);
32222+ if (unlikely(err))
32223+ goto out;
32224+ wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
32225+
4f0767ce 32226+out:
1facf9fc 32227+ return err;
32228+}
32229+
32230+/*
32231+ * initialize the whiteout base file/dir for @br.
32232+ */
86dc4139 32233+int au_wh_init(struct au_branch *br, struct super_block *sb)
1facf9fc 32234+{
32235+ int err, i;
32236+ const unsigned char do_plink
32237+ = !!au_opt_test(au_mntflags(sb), PLINK);
1facf9fc 32238+ struct inode *h_dir;
86dc4139
AM
32239+ struct path path = br->br_path;
32240+ struct dentry *h_root = path.dentry;
1facf9fc 32241+ struct au_wbr *wbr = br->br_wbr;
32242+ static const struct qstr base_name[] = {
0c3ec466
AM
32243+ [AuBrWh_BASE] = QSTR_INIT(AUFS_BASE_NAME,
32244+ sizeof(AUFS_BASE_NAME) - 1),
32245+ [AuBrWh_PLINK] = QSTR_INIT(AUFS_PLINKDIR_NAME,
32246+ sizeof(AUFS_PLINKDIR_NAME) - 1),
32247+ [AuBrWh_ORPH] = QSTR_INIT(AUFS_ORPHDIR_NAME,
32248+ sizeof(AUFS_ORPHDIR_NAME) - 1)
1facf9fc 32249+ };
32250+ struct au_wh_base base[] = {
32251+ [AuBrWh_BASE] = {
32252+ .name = base_name + AuBrWh_BASE,
32253+ .dentry = NULL
32254+ },
32255+ [AuBrWh_PLINK] = {
32256+ .name = base_name + AuBrWh_PLINK,
32257+ .dentry = NULL
32258+ },
32259+ [AuBrWh_ORPH] = {
32260+ .name = base_name + AuBrWh_ORPH,
32261+ .dentry = NULL
32262+ }
32263+ };
32264+
1308ab2a 32265+ if (wbr)
32266+ WbrWhMustWriteLock(wbr);
1facf9fc 32267+
1facf9fc 32268+ for (i = 0; i < AuBrWh_Last; i++) {
32269+ /* doubly whiteouted */
32270+ struct dentry *d;
32271+
32272+ d = au_wh_lkup(h_root, (void *)base[i].name, br);
32273+ err = PTR_ERR(d);
32274+ if (IS_ERR(d))
32275+ goto out;
32276+
32277+ base[i].dentry = d;
32278+ AuDebugOn(wbr
32279+ && wbr->wbr_wh[i]
32280+ && wbr->wbr_wh[i] != base[i].dentry);
32281+ }
32282+
32283+ if (wbr)
32284+ for (i = 0; i < AuBrWh_Last; i++) {
32285+ dput(wbr->wbr_wh[i]);
32286+ wbr->wbr_wh[i] = NULL;
32287+ }
32288+
32289+ err = 0;
1e00d052 32290+ if (!au_br_writable(br->br_perm)) {
5527c038 32291+ h_dir = d_inode(h_root);
1facf9fc 32292+ au_wh_init_ro(h_dir, base, &path);
1e00d052 32293+ } else if (!au_br_wh_linkable(br->br_perm)) {
1facf9fc 32294+ err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
32295+ if (err > 0)
32296+ goto out;
32297+ else if (err)
32298+ goto out_err;
1e00d052 32299+ } else {
1facf9fc 32300+ err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
32301+ if (err > 0)
32302+ goto out;
32303+ else if (err)
32304+ goto out_err;
1facf9fc 32305+ }
32306+ goto out; /* success */
32307+
4f0767ce 32308+out_err:
523b37e3
AM
32309+ pr_err("an error(%d) on the writable branch %pd(%s)\n",
32310+ err, h_root, au_sbtype(h_root->d_sb));
4f0767ce 32311+out:
1facf9fc 32312+ for (i = 0; i < AuBrWh_Last; i++)
32313+ dput(base[i].dentry);
32314+ return err;
32315+}
32316+
32317+/* ---------------------------------------------------------------------- */
32318+/*
32319+ * whiteouts are all hard-linked usually.
32320+ * when its link count reaches a ceiling, we create a new whiteout base
32321+ * asynchronously.
32322+ */
32323+
32324+struct reinit_br_wh {
32325+ struct super_block *sb;
32326+ struct au_branch *br;
32327+};
32328+
32329+static void reinit_br_wh(void *arg)
32330+{
32331+ int err;
32332+ aufs_bindex_t bindex;
32333+ struct path h_path;
32334+ struct reinit_br_wh *a = arg;
32335+ struct au_wbr *wbr;
523b37e3 32336+ struct inode *dir, *delegated;
1facf9fc 32337+ struct dentry *h_root;
32338+ struct au_hinode *hdir;
32339+
32340+ err = 0;
32341+ wbr = a->br->br_wbr;
32342+ /* big aufs lock */
32343+ si_noflush_write_lock(a->sb);
32344+ if (!au_br_writable(a->br->br_perm))
32345+ goto out;
32346+ bindex = au_br_index(a->sb, a->br->br_id);
32347+ if (unlikely(bindex < 0))
32348+ goto out;
32349+
1308ab2a 32350+ di_read_lock_parent(a->sb->s_root, AuLock_IR);
5527c038 32351+ dir = d_inode(a->sb->s_root);
1facf9fc 32352+ hdir = au_hi(dir, bindex);
32353+ h_root = au_h_dptr(a->sb->s_root, bindex);
86dc4139 32354+ AuDebugOn(h_root != au_br_dentry(a->br));
1facf9fc 32355+
4a4d8108 32356+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
1facf9fc 32357+ wbr_wh_write_lock(wbr);
32358+ err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
32359+ h_root, a->br);
32360+ if (!err) {
86dc4139
AM
32361+ h_path.dentry = wbr->wbr_whbase;
32362+ h_path.mnt = au_br_mnt(a->br);
523b37e3
AM
32363+ delegated = NULL;
32364+ err = vfsub_unlink(hdir->hi_inode, &h_path, &delegated,
32365+ /*force*/0);
32366+ if (unlikely(err == -EWOULDBLOCK)) {
32367+ pr_warn("cannot retry for NFSv4 delegation"
32368+ " for an internal unlink\n");
32369+ iput(delegated);
32370+ }
1facf9fc 32371+ } else {
523b37e3 32372+ pr_warn("%pd is moved, ignored\n", wbr->wbr_whbase);
1facf9fc 32373+ err = 0;
32374+ }
32375+ dput(wbr->wbr_whbase);
32376+ wbr->wbr_whbase = NULL;
32377+ if (!err)
86dc4139 32378+ err = au_wh_init(a->br, a->sb);
1facf9fc 32379+ wbr_wh_write_unlock(wbr);
4a4d8108 32380+ au_hn_imtx_unlock(hdir);
1308ab2a 32381+ di_read_unlock(a->sb->s_root, AuLock_IR);
076b876e
AM
32382+ if (!err)
32383+ au_fhsm_wrote(a->sb, bindex, /*force*/0);
1facf9fc 32384+
4f0767ce 32385+out:
1facf9fc 32386+ if (wbr)
32387+ atomic_dec(&wbr->wbr_wh_running);
32388+ atomic_dec(&a->br->br_count);
1facf9fc 32389+ si_write_unlock(a->sb);
027c5e7a 32390+ au_nwt_done(&au_sbi(a->sb)->si_nowait);
1facf9fc 32391+ kfree(arg);
32392+ if (unlikely(err))
32393+ AuIOErr("err %d\n", err);
32394+}
32395+
32396+static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
32397+{
32398+ int do_dec, wkq_err;
32399+ struct reinit_br_wh *arg;
32400+
32401+ do_dec = 1;
32402+ if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
32403+ goto out;
32404+
32405+ /* ignore ENOMEM */
32406+ arg = kmalloc(sizeof(*arg), GFP_NOFS);
32407+ if (arg) {
32408+ /*
32409+ * dec(wh_running), kfree(arg) and dec(br_count)
32410+ * in reinit function
32411+ */
32412+ arg->sb = sb;
32413+ arg->br = br;
32414+ atomic_inc(&br->br_count);
53392da6 32415+ wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
1facf9fc 32416+ if (unlikely(wkq_err)) {
32417+ atomic_dec(&br->br_wbr->wbr_wh_running);
32418+ atomic_dec(&br->br_count);
32419+ kfree(arg);
32420+ }
32421+ do_dec = 0;
32422+ }
32423+
4f0767ce 32424+out:
1facf9fc 32425+ if (do_dec)
32426+ atomic_dec(&br->br_wbr->wbr_wh_running);
32427+}
32428+
32429+/* ---------------------------------------------------------------------- */
32430+
32431+/*
32432+ * create the whiteout @wh.
32433+ */
32434+static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
32435+ struct dentry *wh)
32436+{
32437+ int err;
32438+ struct path h_path = {
32439+ .dentry = wh
32440+ };
32441+ struct au_branch *br;
32442+ struct au_wbr *wbr;
32443+ struct dentry *h_parent;
523b37e3 32444+ struct inode *h_dir, *delegated;
1facf9fc 32445+
32446+ h_parent = wh->d_parent; /* dir inode is locked */
5527c038 32447+ h_dir = d_inode(h_parent);
1facf9fc 32448+ IMustLock(h_dir);
32449+
32450+ br = au_sbr(sb, bindex);
86dc4139 32451+ h_path.mnt = au_br_mnt(br);
1facf9fc 32452+ wbr = br->br_wbr;
32453+ wbr_wh_read_lock(wbr);
32454+ if (wbr->wbr_whbase) {
523b37e3
AM
32455+ delegated = NULL;
32456+ err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path, &delegated);
32457+ if (unlikely(err == -EWOULDBLOCK)) {
32458+ pr_warn("cannot retry for NFSv4 delegation"
32459+ " for an internal link\n");
32460+ iput(delegated);
32461+ }
1facf9fc 32462+ if (!err || err != -EMLINK)
32463+ goto out;
32464+
32465+ /* link count full. re-initialize br_whbase. */
32466+ kick_reinit_br_wh(sb, br);
32467+ }
32468+
32469+ /* return this error in this context */
b4510431 32470+ err = vfsub_create(h_dir, &h_path, WH_MASK, /*want_excl*/true);
076b876e
AM
32471+ if (!err)
32472+ au_fhsm_wrote(sb, bindex, /*force*/0);
1facf9fc 32473+
4f0767ce 32474+out:
1facf9fc 32475+ wbr_wh_read_unlock(wbr);
32476+ return err;
32477+}
32478+
32479+/* ---------------------------------------------------------------------- */
32480+
32481+/*
32482+ * create or remove the diropq.
32483+ */
32484+static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
32485+ unsigned int flags)
32486+{
32487+ struct dentry *opq_dentry, *h_dentry;
32488+ struct super_block *sb;
32489+ struct au_branch *br;
32490+ int err;
32491+
32492+ sb = dentry->d_sb;
32493+ br = au_sbr(sb, bindex);
32494+ h_dentry = au_h_dptr(dentry, bindex);
b4510431 32495+ opq_dentry = vfsub_lkup_one(&diropq_name, h_dentry);
1facf9fc 32496+ if (IS_ERR(opq_dentry))
32497+ goto out;
32498+
32499+ if (au_ftest_diropq(flags, CREATE)) {
32500+ err = link_or_create_wh(sb, bindex, opq_dentry);
32501+ if (!err) {
32502+ au_set_dbdiropq(dentry, bindex);
32503+ goto out; /* success */
32504+ }
32505+ } else {
32506+ struct path tmp = {
32507+ .dentry = opq_dentry,
86dc4139 32508+ .mnt = au_br_mnt(br)
1facf9fc 32509+ };
5527c038 32510+ err = do_unlink_wh(au_h_iptr(d_inode(dentry), bindex), &tmp);
1facf9fc 32511+ if (!err)
32512+ au_set_dbdiropq(dentry, -1);
32513+ }
32514+ dput(opq_dentry);
32515+ opq_dentry = ERR_PTR(err);
32516+
4f0767ce 32517+out:
1facf9fc 32518+ return opq_dentry;
32519+}
32520+
32521+struct do_diropq_args {
32522+ struct dentry **errp;
32523+ struct dentry *dentry;
32524+ aufs_bindex_t bindex;
32525+ unsigned int flags;
32526+};
32527+
32528+static void call_do_diropq(void *args)
32529+{
32530+ struct do_diropq_args *a = args;
32531+ *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
32532+}
32533+
32534+struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
32535+ unsigned int flags)
32536+{
32537+ struct dentry *diropq, *h_dentry;
32538+
32539+ h_dentry = au_h_dptr(dentry, bindex);
5527c038 32540+ if (!au_test_h_perm_sio(d_inode(h_dentry), MAY_EXEC | MAY_WRITE))
1facf9fc 32541+ diropq = do_diropq(dentry, bindex, flags);
32542+ else {
32543+ int wkq_err;
32544+ struct do_diropq_args args = {
32545+ .errp = &diropq,
32546+ .dentry = dentry,
32547+ .bindex = bindex,
32548+ .flags = flags
32549+ };
32550+
32551+ wkq_err = au_wkq_wait(call_do_diropq, &args);
32552+ if (unlikely(wkq_err))
32553+ diropq = ERR_PTR(wkq_err);
32554+ }
32555+
32556+ return diropq;
32557+}
32558+
32559+/* ---------------------------------------------------------------------- */
32560+
32561+/*
32562+ * lookup whiteout dentry.
32563+ * @h_parent: lower parent dentry which must exist and be locked
32564+ * @base_name: name of dentry which will be whiteouted
32565+ * returns dentry for whiteout.
32566+ */
32567+struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
32568+ struct au_branch *br)
32569+{
32570+ int err;
32571+ struct qstr wh_name;
32572+ struct dentry *wh_dentry;
32573+
32574+ err = au_wh_name_alloc(&wh_name, base_name);
32575+ wh_dentry = ERR_PTR(err);
32576+ if (!err) {
b4510431 32577+ wh_dentry = vfsub_lkup_one(&wh_name, h_parent);
1facf9fc 32578+ kfree(wh_name.name);
32579+ }
32580+ return wh_dentry;
32581+}
32582+
32583+/*
32584+ * link/create a whiteout for @dentry on @bindex.
32585+ */
32586+struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
32587+ struct dentry *h_parent)
32588+{
32589+ struct dentry *wh_dentry;
32590+ struct super_block *sb;
32591+ int err;
32592+
32593+ sb = dentry->d_sb;
32594+ wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
5527c038 32595+ if (!IS_ERR(wh_dentry) && d_is_negative(wh_dentry)) {
1facf9fc 32596+ err = link_or_create_wh(sb, bindex, wh_dentry);
076b876e 32597+ if (!err) {
1facf9fc 32598+ au_set_dbwh(dentry, bindex);
076b876e
AM
32599+ au_fhsm_wrote(sb, bindex, /*force*/0);
32600+ } else {
1facf9fc 32601+ dput(wh_dentry);
32602+ wh_dentry = ERR_PTR(err);
32603+ }
32604+ }
32605+
32606+ return wh_dentry;
32607+}
32608+
32609+/* ---------------------------------------------------------------------- */
32610+
32611+/* Delete all whiteouts in this directory on branch bindex. */
32612+static int del_wh_children(struct dentry *h_dentry, struct au_nhash *whlist,
32613+ aufs_bindex_t bindex, struct au_branch *br)
32614+{
32615+ int err;
32616+ unsigned long ul, n;
32617+ struct qstr wh_name;
32618+ char *p;
32619+ struct hlist_head *head;
c06a8ce3 32620+ struct au_vdir_wh *pos;
1facf9fc 32621+ struct au_vdir_destr *str;
32622+
32623+ err = -ENOMEM;
537831f9 32624+ p = (void *)__get_free_page(GFP_NOFS);
1facf9fc 32625+ wh_name.name = p;
32626+ if (unlikely(!wh_name.name))
32627+ goto out;
32628+
32629+ err = 0;
32630+ memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
32631+ p += AUFS_WH_PFX_LEN;
32632+ n = whlist->nh_num;
32633+ head = whlist->nh_head;
32634+ for (ul = 0; !err && ul < n; ul++, head++) {
c06a8ce3
AM
32635+ hlist_for_each_entry(pos, head, wh_hash) {
32636+ if (pos->wh_bindex != bindex)
1facf9fc 32637+ continue;
32638+
c06a8ce3 32639+ str = &pos->wh_str;
1facf9fc 32640+ if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
32641+ memcpy(p, str->name, str->len);
32642+ wh_name.len = AUFS_WH_PFX_LEN + str->len;
32643+ err = unlink_wh_name(h_dentry, &wh_name, br);
32644+ if (!err)
32645+ continue;
32646+ break;
32647+ }
32648+ AuIOErr("whiteout name too long %.*s\n",
32649+ str->len, str->name);
32650+ err = -EIO;
32651+ break;
32652+ }
32653+ }
537831f9 32654+ free_page((unsigned long)wh_name.name);
1facf9fc 32655+
4f0767ce 32656+out:
1facf9fc 32657+ return err;
32658+}
32659+
32660+struct del_wh_children_args {
32661+ int *errp;
32662+ struct dentry *h_dentry;
1308ab2a 32663+ struct au_nhash *whlist;
1facf9fc 32664+ aufs_bindex_t bindex;
32665+ struct au_branch *br;
32666+};
32667+
32668+static void call_del_wh_children(void *args)
32669+{
32670+ struct del_wh_children_args *a = args;
1308ab2a 32671+ *a->errp = del_wh_children(a->h_dentry, a->whlist, a->bindex, a->br);
1facf9fc 32672+}
32673+
32674+/* ---------------------------------------------------------------------- */
32675+
32676+struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
32677+{
32678+ struct au_whtmp_rmdir *whtmp;
dece6358 32679+ int err;
1308ab2a 32680+ unsigned int rdhash;
dece6358
AM
32681+
32682+ SiMustAnyLock(sb);
1facf9fc 32683+
32684+ whtmp = kmalloc(sizeof(*whtmp), gfp);
dece6358
AM
32685+ if (unlikely(!whtmp)) {
32686+ whtmp = ERR_PTR(-ENOMEM);
1facf9fc 32687+ goto out;
dece6358 32688+ }
1facf9fc 32689+
32690+ whtmp->dir = NULL;
027c5e7a 32691+ whtmp->br = NULL;
1facf9fc 32692+ whtmp->wh_dentry = NULL;
1308ab2a 32693+ /* no estimation for dir size */
32694+ rdhash = au_sbi(sb)->si_rdhash;
32695+ if (!rdhash)
32696+ rdhash = AUFS_RDHASH_DEF;
32697+ err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
32698+ if (unlikely(err)) {
32699+ kfree(whtmp);
32700+ whtmp = ERR_PTR(err);
32701+ }
dece6358 32702+
4f0767ce 32703+out:
dece6358 32704+ return whtmp;
1facf9fc 32705+}
32706+
32707+void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
32708+{
027c5e7a
AM
32709+ if (whtmp->br)
32710+ atomic_dec(&whtmp->br->br_count);
1facf9fc 32711+ dput(whtmp->wh_dentry);
32712+ iput(whtmp->dir);
dece6358 32713+ au_nhash_wh_free(&whtmp->whlist);
1facf9fc 32714+ kfree(whtmp);
32715+}
32716+
32717+/*
32718+ * rmdir the whiteouted temporary named dir @h_dentry.
32719+ * @whlist: whiteouted children.
32720+ */
32721+int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
32722+ struct dentry *wh_dentry, struct au_nhash *whlist)
32723+{
32724+ int err;
2000de60 32725+ unsigned int h_nlink;
1facf9fc 32726+ struct path h_tmp;
32727+ struct inode *wh_inode, *h_dir;
32728+ struct au_branch *br;
32729+
5527c038 32730+ h_dir = d_inode(wh_dentry->d_parent); /* dir inode is locked */
1facf9fc 32731+ IMustLock(h_dir);
32732+
32733+ br = au_sbr(dir->i_sb, bindex);
5527c038 32734+ wh_inode = d_inode(wh_dentry);
1facf9fc 32735+ mutex_lock_nested(&wh_inode->i_mutex, AuLsc_I_CHILD);
32736+
32737+ /*
32738+ * someone else might change some whiteouts while we were sleeping.
32739+ * it means this whlist may have an obsoleted entry.
32740+ */
32741+ if (!au_test_h_perm_sio(wh_inode, MAY_EXEC | MAY_WRITE))
32742+ err = del_wh_children(wh_dentry, whlist, bindex, br);
32743+ else {
32744+ int wkq_err;
32745+ struct del_wh_children_args args = {
32746+ .errp = &err,
32747+ .h_dentry = wh_dentry,
1308ab2a 32748+ .whlist = whlist,
1facf9fc 32749+ .bindex = bindex,
32750+ .br = br
32751+ };
32752+
32753+ wkq_err = au_wkq_wait(call_del_wh_children, &args);
32754+ if (unlikely(wkq_err))
32755+ err = wkq_err;
32756+ }
32757+ mutex_unlock(&wh_inode->i_mutex);
32758+
32759+ if (!err) {
32760+ h_tmp.dentry = wh_dentry;
86dc4139 32761+ h_tmp.mnt = au_br_mnt(br);
2000de60 32762+ h_nlink = h_dir->i_nlink;
1facf9fc 32763+ err = vfsub_rmdir(h_dir, &h_tmp);
2000de60
JR
32764+ /* some fs doesn't change the parent nlink in some cases */
32765+ h_nlink -= h_dir->i_nlink;
1facf9fc 32766+ }
32767+
32768+ if (!err) {
32769+ if (au_ibstart(dir) == bindex) {
7f207e10 32770+ /* todo: dir->i_mutex is necessary */
1facf9fc 32771+ au_cpup_attr_timesizes(dir);
2000de60
JR
32772+ if (h_nlink)
32773+ vfsub_drop_nlink(dir);
1facf9fc 32774+ }
32775+ return 0; /* success */
32776+ }
32777+
523b37e3 32778+ pr_warn("failed removing %pd(%d), ignored\n", wh_dentry, err);
1facf9fc 32779+ return err;
32780+}
32781+
32782+static void call_rmdir_whtmp(void *args)
32783+{
32784+ int err;
e49829fe 32785+ aufs_bindex_t bindex;
1facf9fc 32786+ struct au_whtmp_rmdir *a = args;
32787+ struct super_block *sb;
32788+ struct dentry *h_parent;
32789+ struct inode *h_dir;
1facf9fc 32790+ struct au_hinode *hdir;
32791+
32792+ /* rmdir by nfsd may cause deadlock with this i_mutex */
32793+ /* mutex_lock(&a->dir->i_mutex); */
e49829fe 32794+ err = -EROFS;
1facf9fc 32795+ sb = a->dir->i_sb;
e49829fe
JR
32796+ si_read_lock(sb, !AuLock_FLUSH);
32797+ if (!au_br_writable(a->br->br_perm))
32798+ goto out;
32799+ bindex = au_br_index(sb, a->br->br_id);
32800+ if (unlikely(bindex < 0))
1facf9fc 32801+ goto out;
32802+
32803+ err = -EIO;
1facf9fc 32804+ ii_write_lock_parent(a->dir);
32805+ h_parent = dget_parent(a->wh_dentry);
5527c038 32806+ h_dir = d_inode(h_parent);
e49829fe 32807+ hdir = au_hi(a->dir, bindex);
86dc4139
AM
32808+ err = vfsub_mnt_want_write(au_br_mnt(a->br));
32809+ if (unlikely(err))
32810+ goto out_mnt;
4a4d8108 32811+ au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
e49829fe
JR
32812+ err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
32813+ a->br);
86dc4139
AM
32814+ if (!err)
32815+ err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry, &a->whlist);
4a4d8108 32816+ au_hn_imtx_unlock(hdir);
86dc4139
AM
32817+ vfsub_mnt_drop_write(au_br_mnt(a->br));
32818+
32819+out_mnt:
1facf9fc 32820+ dput(h_parent);
32821+ ii_write_unlock(a->dir);
4f0767ce 32822+out:
1facf9fc 32823+ /* mutex_unlock(&a->dir->i_mutex); */
1facf9fc 32824+ au_whtmp_rmdir_free(a);
027c5e7a
AM
32825+ si_read_unlock(sb);
32826+ au_nwt_done(&au_sbi(sb)->si_nowait);
1facf9fc 32827+ if (unlikely(err))
32828+ AuIOErr("err %d\n", err);
32829+}
32830+
32831+void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
32832+ struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
32833+{
32834+ int wkq_err;
e49829fe 32835+ struct super_block *sb;
1facf9fc 32836+
32837+ IMustLock(dir);
32838+
32839+ /* all post-process will be done in do_rmdir_whtmp(). */
e49829fe 32840+ sb = dir->i_sb;
1facf9fc 32841+ args->dir = au_igrab(dir);
e49829fe
JR
32842+ args->br = au_sbr(sb, bindex);
32843+ atomic_inc(&args->br->br_count);
1facf9fc 32844+ args->wh_dentry = dget(wh_dentry);
53392da6 32845+ wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
1facf9fc 32846+ if (unlikely(wkq_err)) {
523b37e3 32847+ pr_warn("rmdir error %pd (%d), ignored\n", wh_dentry, wkq_err);
1facf9fc 32848+ au_whtmp_rmdir_free(args);
32849+ }
32850+}
7f207e10
AM
32851diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
32852--- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 32853+++ linux/fs/aufs/whout.h 2015-09-24 10:47:58.258053165 +0200
076b876e 32854@@ -0,0 +1,85 @@
1facf9fc 32855+/*
2000de60 32856+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 32857+ *
32858+ * This program, aufs is free software; you can redistribute it and/or modify
32859+ * it under the terms of the GNU General Public License as published by
32860+ * the Free Software Foundation; either version 2 of the License, or
32861+ * (at your option) any later version.
dece6358
AM
32862+ *
32863+ * This program is distributed in the hope that it will be useful,
32864+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
32865+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32866+ * GNU General Public License for more details.
32867+ *
32868+ * You should have received a copy of the GNU General Public License
523b37e3 32869+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 32870+ */
32871+
32872+/*
32873+ * whiteout for logical deletion and opaque directory
32874+ */
32875+
32876+#ifndef __AUFS_WHOUT_H__
32877+#define __AUFS_WHOUT_H__
32878+
32879+#ifdef __KERNEL__
32880+
1facf9fc 32881+#include "dir.h"
32882+
32883+/* whout.c */
32884+int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
076b876e
AM
32885+int au_wh_test(struct dentry *h_parent, struct qstr *wh_name, int try_sio);
32886+int au_diropq_test(struct dentry *h_dentry);
7e9cd9fe 32887+struct au_branch;
1facf9fc 32888+struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
32889+ struct qstr *prefix);
32890+int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
32891+int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
32892+ struct dentry *dentry);
86dc4139 32893+int au_wh_init(struct au_branch *br, struct super_block *sb);
1facf9fc 32894+
32895+/* diropq flags */
32896+#define AuDiropq_CREATE 1
32897+#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
7f207e10
AM
32898+#define au_fset_diropq(flags, name) \
32899+ do { (flags) |= AuDiropq_##name; } while (0)
32900+#define au_fclr_diropq(flags, name) \
32901+ do { (flags) &= ~AuDiropq_##name; } while (0)
1facf9fc 32902+
32903+struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
32904+ unsigned int flags);
32905+struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
32906+ struct au_branch *br);
32907+struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
32908+ struct dentry *h_parent);
32909+
32910+/* real rmdir for the whiteout-ed dir */
32911+struct au_whtmp_rmdir {
32912+ struct inode *dir;
e49829fe 32913+ struct au_branch *br;
1facf9fc 32914+ struct dentry *wh_dentry;
dece6358 32915+ struct au_nhash whlist;
1facf9fc 32916+};
32917+
32918+struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
32919+void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
32920+int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
32921+ struct dentry *wh_dentry, struct au_nhash *whlist);
32922+void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
32923+ struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
32924+
32925+/* ---------------------------------------------------------------------- */
32926+
32927+static inline struct dentry *au_diropq_create(struct dentry *dentry,
32928+ aufs_bindex_t bindex)
32929+{
32930+ return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
32931+}
32932+
32933+static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
32934+{
32935+ return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
32936+}
32937+
32938+#endif /* __KERNEL__ */
32939+#endif /* __AUFS_WHOUT_H__ */
7f207e10
AM
32940diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
32941--- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 32942+++ linux/fs/aufs/wkq.c 2015-09-24 10:47:58.258053165 +0200
38d290e6 32943@@ -0,0 +1,213 @@
1facf9fc 32944+/*
2000de60 32945+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 32946+ *
32947+ * This program, aufs is free software; you can redistribute it and/or modify
32948+ * it under the terms of the GNU General Public License as published by
32949+ * the Free Software Foundation; either version 2 of the License, or
32950+ * (at your option) any later version.
dece6358
AM
32951+ *
32952+ * This program is distributed in the hope that it will be useful,
32953+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
32954+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
32955+ * GNU General Public License for more details.
32956+ *
32957+ * You should have received a copy of the GNU General Public License
523b37e3 32958+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 32959+ */
32960+
32961+/*
32962+ * workqueue for asynchronous/super-io operations
32963+ * todo: try new dredential scheme
32964+ */
32965+
dece6358 32966+#include <linux/module.h>
1facf9fc 32967+#include "aufs.h"
32968+
9dbd164d 32969+/* internal workqueue named AUFS_WKQ_NAME */
b752ccd1 32970+
9dbd164d 32971+static struct workqueue_struct *au_wkq;
1facf9fc 32972+
32973+struct au_wkinfo {
32974+ struct work_struct wk;
7f207e10 32975+ struct kobject *kobj;
1facf9fc 32976+
32977+ unsigned int flags; /* see wkq.h */
32978+
32979+ au_wkq_func_t func;
32980+ void *args;
32981+
1facf9fc 32982+ struct completion *comp;
32983+};
32984+
32985+/* ---------------------------------------------------------------------- */
32986+
1facf9fc 32987+static void wkq_func(struct work_struct *wk)
32988+{
32989+ struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
32990+
2dfbb274 32991+ AuDebugOn(!uid_eq(current_fsuid(), GLOBAL_ROOT_UID));
7f207e10
AM
32992+ AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
32993+
1facf9fc 32994+ wkinfo->func(wkinfo->args);
1facf9fc 32995+ if (au_ftest_wkq(wkinfo->flags, WAIT))
32996+ complete(wkinfo->comp);
32997+ else {
7f207e10 32998+ kobject_put(wkinfo->kobj);
9dbd164d 32999+ module_put(THIS_MODULE); /* todo: ?? */
1facf9fc 33000+ kfree(wkinfo);
33001+ }
33002+}
33003+
33004+/*
33005+ * Since struct completion is large, try allocating it dynamically.
33006+ */
c2b27bf2 33007+#if 1 /* defined(CONFIG_4KSTACKS) || defined(AuTest4KSTACKS) */
1facf9fc 33008+#define AuWkqCompDeclare(name) struct completion *comp = NULL
33009+
33010+static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
33011+{
33012+ *comp = kmalloc(sizeof(**comp), GFP_NOFS);
33013+ if (*comp) {
33014+ init_completion(*comp);
33015+ wkinfo->comp = *comp;
33016+ return 0;
33017+ }
33018+ return -ENOMEM;
33019+}
33020+
33021+static void au_wkq_comp_free(struct completion *comp)
33022+{
33023+ kfree(comp);
33024+}
33025+
33026+#else
33027+
33028+/* no braces */
33029+#define AuWkqCompDeclare(name) \
33030+ DECLARE_COMPLETION_ONSTACK(_ ## name); \
33031+ struct completion *comp = &_ ## name
33032+
33033+static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
33034+{
33035+ wkinfo->comp = *comp;
33036+ return 0;
33037+}
33038+
33039+static void au_wkq_comp_free(struct completion *comp __maybe_unused)
33040+{
33041+ /* empty */
33042+}
33043+#endif /* 4KSTACKS */
33044+
53392da6 33045+static void au_wkq_run(struct au_wkinfo *wkinfo)
1facf9fc 33046+{
53392da6
AM
33047+ if (au_ftest_wkq(wkinfo->flags, NEST)) {
33048+ if (au_wkq_test()) {
38d290e6
JR
33049+ AuWarn1("wkq from wkq, unless silly-rename on NFS,"
33050+ " due to a dead dir by UDBA?\n");
53392da6
AM
33051+ AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
33052+ }
33053+ } else
33054+ au_dbg_verify_kthread();
33055+
33056+ if (au_ftest_wkq(wkinfo->flags, WAIT)) {
a1f66529 33057+ INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
9dbd164d 33058+ queue_work(au_wkq, &wkinfo->wk);
4a4d8108
AM
33059+ } else {
33060+ INIT_WORK(&wkinfo->wk, wkq_func);
33061+ schedule_work(&wkinfo->wk);
33062+ }
1facf9fc 33063+}
33064+
7f207e10
AM
33065+/*
33066+ * Be careful. It is easy to make deadlock happen.
33067+ * processA: lock, wkq and wait
33068+ * processB: wkq and wait, lock in wkq
33069+ * --> deadlock
33070+ */
b752ccd1 33071+int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
1facf9fc 33072+{
33073+ int err;
33074+ AuWkqCompDeclare(comp);
33075+ struct au_wkinfo wkinfo = {
b752ccd1 33076+ .flags = flags,
1facf9fc 33077+ .func = func,
33078+ .args = args
33079+ };
33080+
33081+ err = au_wkq_comp_alloc(&wkinfo, &comp);
33082+ if (!err) {
53392da6 33083+ au_wkq_run(&wkinfo);
1facf9fc 33084+ /* no timeout, no interrupt */
33085+ wait_for_completion(wkinfo.comp);
33086+ au_wkq_comp_free(comp);
4a4d8108 33087+ destroy_work_on_stack(&wkinfo.wk);
1facf9fc 33088+ }
33089+
33090+ return err;
33091+
33092+}
33093+
027c5e7a
AM
33094+/*
33095+ * Note: dget/dput() in func for aufs dentries are not supported. It will be a
33096+ * problem in a concurrent umounting.
33097+ */
53392da6
AM
33098+int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
33099+ unsigned int flags)
1facf9fc 33100+{
33101+ int err;
33102+ struct au_wkinfo *wkinfo;
33103+
33104+ atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
33105+
33106+ /*
33107+ * wkq_func() must free this wkinfo.
33108+ * it highly depends upon the implementation of workqueue.
33109+ */
33110+ err = 0;
33111+ wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
33112+ if (wkinfo) {
7f207e10 33113+ wkinfo->kobj = &au_sbi(sb)->si_kobj;
53392da6 33114+ wkinfo->flags = flags & ~AuWkq_WAIT;
1facf9fc 33115+ wkinfo->func = func;
33116+ wkinfo->args = args;
33117+ wkinfo->comp = NULL;
7f207e10 33118+ kobject_get(wkinfo->kobj);
9dbd164d 33119+ __module_get(THIS_MODULE); /* todo: ?? */
1facf9fc 33120+
53392da6 33121+ au_wkq_run(wkinfo);
1facf9fc 33122+ } else {
33123+ err = -ENOMEM;
e49829fe 33124+ au_nwt_done(&au_sbi(sb)->si_nowait);
1facf9fc 33125+ }
33126+
33127+ return err;
33128+}
33129+
33130+/* ---------------------------------------------------------------------- */
33131+
33132+void au_nwt_init(struct au_nowait_tasks *nwt)
33133+{
33134+ atomic_set(&nwt->nw_len, 0);
4a4d8108 33135+ /* smp_mb(); */ /* atomic_set */
1facf9fc 33136+ init_waitqueue_head(&nwt->nw_wq);
33137+}
33138+
33139+void au_wkq_fin(void)
33140+{
9dbd164d 33141+ destroy_workqueue(au_wkq);
1facf9fc 33142+}
33143+
33144+int __init au_wkq_init(void)
33145+{
9dbd164d 33146+ int err;
b752ccd1
AM
33147+
33148+ err = 0;
86dc4139 33149+ au_wkq = alloc_workqueue(AUFS_WKQ_NAME, 0, WQ_DFL_ACTIVE);
9dbd164d
AM
33150+ if (IS_ERR(au_wkq))
33151+ err = PTR_ERR(au_wkq);
33152+ else if (!au_wkq)
33153+ err = -ENOMEM;
b752ccd1
AM
33154+
33155+ return err;
1facf9fc 33156+}
7f207e10
AM
33157diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
33158--- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 33159+++ linux/fs/aufs/wkq.h 2015-09-24 10:47:58.258053165 +0200
523b37e3 33160@@ -0,0 +1,91 @@
1facf9fc 33161+/*
2000de60 33162+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 33163+ *
33164+ * This program, aufs is free software; you can redistribute it and/or modify
33165+ * it under the terms of the GNU General Public License as published by
33166+ * the Free Software Foundation; either version 2 of the License, or
33167+ * (at your option) any later version.
dece6358
AM
33168+ *
33169+ * This program is distributed in the hope that it will be useful,
33170+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
33171+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33172+ * GNU General Public License for more details.
33173+ *
33174+ * You should have received a copy of the GNU General Public License
523b37e3 33175+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 33176+ */
33177+
33178+/*
33179+ * workqueue for asynchronous/super-io operations
33180+ * todo: try new credentials management scheme
33181+ */
33182+
33183+#ifndef __AUFS_WKQ_H__
33184+#define __AUFS_WKQ_H__
33185+
33186+#ifdef __KERNEL__
33187+
dece6358
AM
33188+struct super_block;
33189+
1facf9fc 33190+/* ---------------------------------------------------------------------- */
33191+
33192+/*
33193+ * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
33194+ */
33195+struct au_nowait_tasks {
33196+ atomic_t nw_len;
33197+ wait_queue_head_t nw_wq;
33198+};
33199+
33200+/* ---------------------------------------------------------------------- */
33201+
33202+typedef void (*au_wkq_func_t)(void *args);
33203+
33204+/* wkq flags */
33205+#define AuWkq_WAIT 1
9dbd164d 33206+#define AuWkq_NEST (1 << 1)
1facf9fc 33207+#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
7f207e10
AM
33208+#define au_fset_wkq(flags, name) \
33209+ do { (flags) |= AuWkq_##name; } while (0)
33210+#define au_fclr_wkq(flags, name) \
33211+ do { (flags) &= ~AuWkq_##name; } while (0)
1facf9fc 33212+
9dbd164d
AM
33213+#ifndef CONFIG_AUFS_HNOTIFY
33214+#undef AuWkq_NEST
33215+#define AuWkq_NEST 0
33216+#endif
33217+
1facf9fc 33218+/* wkq.c */
b752ccd1 33219+int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
53392da6
AM
33220+int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
33221+ unsigned int flags);
1facf9fc 33222+void au_nwt_init(struct au_nowait_tasks *nwt);
33223+int __init au_wkq_init(void);
33224+void au_wkq_fin(void);
33225+
33226+/* ---------------------------------------------------------------------- */
33227+
53392da6
AM
33228+static inline int au_wkq_test(void)
33229+{
33230+ return current->flags & PF_WQ_WORKER;
33231+}
33232+
b752ccd1 33233+static inline int au_wkq_wait(au_wkq_func_t func, void *args)
1facf9fc 33234+{
b752ccd1 33235+ return au_wkq_do_wait(AuWkq_WAIT, func, args);
1facf9fc 33236+}
33237+
33238+static inline void au_nwt_done(struct au_nowait_tasks *nwt)
33239+{
e49829fe 33240+ if (atomic_dec_and_test(&nwt->nw_len))
1facf9fc 33241+ wake_up_all(&nwt->nw_wq);
33242+}
33243+
33244+static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
33245+{
33246+ wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
33247+ return 0;
33248+}
33249+
33250+#endif /* __KERNEL__ */
33251+#endif /* __AUFS_WKQ_H__ */
c1595e42
JR
33252diff -urN /usr/share/empty/fs/aufs/xattr.c linux/fs/aufs/xattr.c
33253--- /usr/share/empty/fs/aufs/xattr.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 33254+++ linux/fs/aufs/xattr.c 2015-09-24 10:47:58.258053165 +0200
b912730e 33255@@ -0,0 +1,344 @@
c1595e42 33256+/*
2000de60 33257+ * Copyright (C) 2014-2015 Junjiro R. Okajima
c1595e42
JR
33258+ *
33259+ * This program, aufs is free software; you can redistribute it and/or modify
33260+ * it under the terms of the GNU General Public License as published by
33261+ * the Free Software Foundation; either version 2 of the License, or
33262+ * (at your option) any later version.
33263+ *
33264+ * This program is distributed in the hope that it will be useful,
33265+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
33266+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33267+ * GNU General Public License for more details.
33268+ *
33269+ * You should have received a copy of the GNU General Public License
33270+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
33271+ */
33272+
33273+/*
33274+ * handling xattr functions
33275+ */
33276+
33277+#include <linux/xattr.h>
33278+#include "aufs.h"
33279+
33280+static int au_xattr_ignore(int err, char *name, unsigned int ignore_flags)
33281+{
33282+ if (!ignore_flags)
33283+ goto out;
33284+ switch (err) {
33285+ case -ENOMEM:
33286+ case -EDQUOT:
33287+ goto out;
33288+ }
33289+
33290+ if ((ignore_flags & AuBrAttr_ICEX) == AuBrAttr_ICEX) {
33291+ err = 0;
33292+ goto out;
33293+ }
33294+
33295+#define cmp(brattr, prefix) do { \
33296+ if (!strncmp(name, XATTR_##prefix##_PREFIX, \
33297+ XATTR_##prefix##_PREFIX_LEN)) { \
33298+ if (ignore_flags & AuBrAttr_ICEX_##brattr) \
33299+ err = 0; \
33300+ goto out; \
33301+ } \
33302+ } while (0)
33303+
33304+ cmp(SEC, SECURITY);
33305+ cmp(SYS, SYSTEM);
33306+ cmp(TR, TRUSTED);
33307+ cmp(USR, USER);
33308+#undef cmp
33309+
33310+ if (ignore_flags & AuBrAttr_ICEX_OTH)
33311+ err = 0;
33312+
33313+out:
33314+ return err;
33315+}
33316+
33317+static const int au_xattr_out_of_list = AuBrAttr_ICEX_OTH << 1;
33318+
33319+static int au_do_cpup_xattr(struct dentry *h_dst, struct dentry *h_src,
7e9cd9fe
AM
33320+ char *name, char **buf, unsigned int ignore_flags,
33321+ unsigned int verbose)
c1595e42
JR
33322+{
33323+ int err;
33324+ ssize_t ssz;
33325+ struct inode *h_idst;
33326+
33327+ ssz = vfs_getxattr_alloc(h_src, name, buf, 0, GFP_NOFS);
33328+ err = ssz;
33329+ if (unlikely(err <= 0)) {
c1595e42
JR
33330+ if (err == -ENODATA
33331+ || (err == -EOPNOTSUPP
b912730e 33332+ && ((ignore_flags & au_xattr_out_of_list)
5527c038 33333+ || (au_test_nfs_noacl(d_inode(h_src))
b912730e
AM
33334+ && (!strcmp(name, XATTR_NAME_POSIX_ACL_ACCESS)
33335+ || !strcmp(name,
33336+ XATTR_NAME_POSIX_ACL_DEFAULT))))
33337+ ))
c1595e42 33338+ err = 0;
b912730e
AM
33339+ if (err && (verbose || au_debug_test()))
33340+ pr_err("%s, err %d\n", name, err);
c1595e42
JR
33341+ goto out;
33342+ }
33343+
33344+ /* unlock it temporary */
5527c038 33345+ h_idst = d_inode(h_dst);
c1595e42
JR
33346+ mutex_unlock(&h_idst->i_mutex);
33347+ err = vfsub_setxattr(h_dst, name, *buf, ssz, /*flags*/0);
33348+ mutex_lock_nested(&h_idst->i_mutex, AuLsc_I_CHILD2);
33349+ if (unlikely(err)) {
7e9cd9fe
AM
33350+ if (verbose || au_debug_test())
33351+ pr_err("%s, err %d\n", name, err);
c1595e42
JR
33352+ err = au_xattr_ignore(err, name, ignore_flags);
33353+ }
33354+
33355+out:
33356+ return err;
33357+}
33358+
7e9cd9fe
AM
33359+int au_cpup_xattr(struct dentry *h_dst, struct dentry *h_src, int ignore_flags,
33360+ unsigned int verbose)
c1595e42
JR
33361+{
33362+ int err, unlocked, acl_access, acl_default;
33363+ ssize_t ssz;
33364+ struct inode *h_isrc, *h_idst;
33365+ char *value, *p, *o, *e;
33366+
33367+ /* try stopping to update the source inode while we are referencing */
7e9cd9fe 33368+ /* there should not be the parent-child relationship between them */
5527c038
JR
33369+ h_isrc = d_inode(h_src);
33370+ h_idst = d_inode(h_dst);
c1595e42
JR
33371+ mutex_unlock(&h_idst->i_mutex);
33372+ mutex_lock_nested(&h_isrc->i_mutex, AuLsc_I_CHILD);
33373+ mutex_lock_nested(&h_idst->i_mutex, AuLsc_I_CHILD2);
33374+ unlocked = 0;
33375+
33376+ /* some filesystems don't list POSIX ACL, for example tmpfs */
33377+ ssz = vfs_listxattr(h_src, NULL, 0);
33378+ err = ssz;
33379+ if (unlikely(err < 0)) {
33380+ AuTraceErr(err);
33381+ if (err == -ENODATA
33382+ || err == -EOPNOTSUPP)
33383+ err = 0; /* ignore */
33384+ goto out;
33385+ }
33386+
33387+ err = 0;
33388+ p = NULL;
33389+ o = NULL;
33390+ if (ssz) {
33391+ err = -ENOMEM;
33392+ p = kmalloc(ssz, GFP_NOFS);
33393+ o = p;
33394+ if (unlikely(!p))
33395+ goto out;
33396+ err = vfs_listxattr(h_src, p, ssz);
33397+ }
33398+ mutex_unlock(&h_isrc->i_mutex);
33399+ unlocked = 1;
33400+ AuDbg("err %d, ssz %zd\n", err, ssz);
33401+ if (unlikely(err < 0))
33402+ goto out_free;
33403+
33404+ err = 0;
33405+ e = p + ssz;
33406+ value = NULL;
33407+ acl_access = 0;
33408+ acl_default = 0;
33409+ while (!err && p < e) {
33410+ acl_access |= !strncmp(p, XATTR_NAME_POSIX_ACL_ACCESS,
33411+ sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1);
33412+ acl_default |= !strncmp(p, XATTR_NAME_POSIX_ACL_DEFAULT,
33413+ sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)
33414+ - 1);
7e9cd9fe
AM
33415+ err = au_do_cpup_xattr(h_dst, h_src, p, &value, ignore_flags,
33416+ verbose);
c1595e42
JR
33417+ p += strlen(p) + 1;
33418+ }
33419+ AuTraceErr(err);
33420+ ignore_flags |= au_xattr_out_of_list;
33421+ if (!err && !acl_access) {
33422+ err = au_do_cpup_xattr(h_dst, h_src,
33423+ XATTR_NAME_POSIX_ACL_ACCESS, &value,
7e9cd9fe 33424+ ignore_flags, verbose);
c1595e42
JR
33425+ AuTraceErr(err);
33426+ }
33427+ if (!err && !acl_default) {
33428+ err = au_do_cpup_xattr(h_dst, h_src,
33429+ XATTR_NAME_POSIX_ACL_DEFAULT, &value,
7e9cd9fe 33430+ ignore_flags, verbose);
c1595e42
JR
33431+ AuTraceErr(err);
33432+ }
33433+
33434+ kfree(value);
33435+
33436+out_free:
33437+ kfree(o);
33438+out:
33439+ if (!unlocked)
33440+ mutex_unlock(&h_isrc->i_mutex);
33441+ AuTraceErr(err);
33442+ return err;
33443+}
33444+
33445+/* ---------------------------------------------------------------------- */
33446+
33447+enum {
33448+ AU_XATTR_LIST,
33449+ AU_XATTR_GET
33450+};
33451+
33452+struct au_lgxattr {
33453+ int type;
33454+ union {
33455+ struct {
33456+ char *list;
33457+ size_t size;
33458+ } list;
33459+ struct {
33460+ const char *name;
33461+ void *value;
33462+ size_t size;
33463+ } get;
33464+ } u;
33465+};
33466+
33467+static ssize_t au_lgxattr(struct dentry *dentry, struct au_lgxattr *arg)
33468+{
33469+ ssize_t err;
33470+ struct path h_path;
33471+ struct super_block *sb;
33472+
33473+ sb = dentry->d_sb;
33474+ err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
33475+ if (unlikely(err))
33476+ goto out;
33477+ err = au_h_path_getattr(dentry, /*force*/1, &h_path);
33478+ if (unlikely(err))
33479+ goto out_si;
33480+ if (unlikely(!h_path.dentry))
33481+ /* illegally overlapped or something */
33482+ goto out_di; /* pretending success */
33483+
33484+ /* always topmost entry only */
33485+ switch (arg->type) {
33486+ case AU_XATTR_LIST:
33487+ err = vfs_listxattr(h_path.dentry,
33488+ arg->u.list.list, arg->u.list.size);
33489+ break;
33490+ case AU_XATTR_GET:
33491+ err = vfs_getxattr(h_path.dentry,
33492+ arg->u.get.name, arg->u.get.value,
33493+ arg->u.get.size);
33494+ break;
33495+ }
33496+
33497+out_di:
33498+ di_read_unlock(dentry, AuLock_IR);
33499+out_si:
33500+ si_read_unlock(sb);
33501+out:
33502+ AuTraceErr(err);
33503+ return err;
33504+}
33505+
33506+ssize_t aufs_listxattr(struct dentry *dentry, char *list, size_t size)
33507+{
33508+ struct au_lgxattr arg = {
33509+ .type = AU_XATTR_LIST,
33510+ .u.list = {
33511+ .list = list,
33512+ .size = size
33513+ },
33514+ };
33515+
33516+ return au_lgxattr(dentry, &arg);
33517+}
33518+
33519+ssize_t aufs_getxattr(struct dentry *dentry, const char *name, void *value,
33520+ size_t size)
33521+{
33522+ struct au_lgxattr arg = {
33523+ .type = AU_XATTR_GET,
33524+ .u.get = {
33525+ .name = name,
33526+ .value = value,
33527+ .size = size
33528+ },
33529+ };
33530+
33531+ return au_lgxattr(dentry, &arg);
33532+}
33533+
33534+int aufs_setxattr(struct dentry *dentry, const char *name, const void *value,
33535+ size_t size, int flags)
33536+{
33537+ struct au_srxattr arg = {
33538+ .type = AU_XATTR_SET,
33539+ .u.set = {
33540+ .name = name,
33541+ .value = value,
33542+ .size = size,
33543+ .flags = flags
33544+ },
33545+ };
33546+
33547+ return au_srxattr(dentry, &arg);
33548+}
33549+
33550+int aufs_removexattr(struct dentry *dentry, const char *name)
33551+{
33552+ struct au_srxattr arg = {
33553+ .type = AU_XATTR_REMOVE,
33554+ .u.remove = {
33555+ .name = name
33556+ },
33557+ };
33558+
33559+ return au_srxattr(dentry, &arg);
33560+}
33561+
33562+/* ---------------------------------------------------------------------- */
33563+
33564+#if 0
33565+static size_t au_xattr_list(struct dentry *dentry, char *list, size_t list_size,
33566+ const char *name, size_t name_len, int type)
33567+{
33568+ return aufs_listxattr(dentry, list, list_size);
33569+}
33570+
33571+static int au_xattr_get(struct dentry *dentry, const char *name, void *buffer,
33572+ size_t size, int type)
33573+{
33574+ return aufs_getxattr(dentry, name, buffer, size);
33575+}
33576+
33577+static int au_xattr_set(struct dentry *dentry, const char *name,
33578+ const void *value, size_t size, int flags, int type)
33579+{
33580+ return aufs_setxattr(dentry, name, value, size, flags);
33581+}
33582+
33583+static const struct xattr_handler au_xattr_handler = {
33584+ /* no prefix, no flags */
33585+ .list = au_xattr_list,
33586+ .get = au_xattr_get,
33587+ .set = au_xattr_set
33588+ /* why no remove? */
33589+};
33590+
33591+static const struct xattr_handler *au_xattr_handlers[] = {
33592+ &au_xattr_handler
33593+};
33594+
33595+void au_xattr_init(struct super_block *sb)
33596+{
33597+ /* sb->s_xattr = au_xattr_handlers; */
33598+}
33599+#endif
7f207e10
AM
33600diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
33601--- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 33602+++ linux/fs/aufs/xino.c 2015-09-24 10:47:58.258053165 +0200
5527c038 33603@@ -0,0 +1,1297 @@
1facf9fc 33604+/*
2000de60 33605+ * Copyright (C) 2005-2015 Junjiro R. Okajima
1facf9fc 33606+ *
33607+ * This program, aufs is free software; you can redistribute it and/or modify
33608+ * it under the terms of the GNU General Public License as published by
33609+ * the Free Software Foundation; either version 2 of the License, or
33610+ * (at your option) any later version.
dece6358
AM
33611+ *
33612+ * This program is distributed in the hope that it will be useful,
33613+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
33614+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
33615+ * GNU General Public License for more details.
33616+ *
33617+ * You should have received a copy of the GNU General Public License
523b37e3 33618+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
1facf9fc 33619+ */
33620+
33621+/*
33622+ * external inode number translation table and bitmap
33623+ */
33624+
33625+#include <linux/seq_file.h>
392086de 33626+#include <linux/statfs.h>
1facf9fc 33627+#include "aufs.h"
33628+
9dbd164d 33629+/* todo: unnecessary to support mmap_sem since kernel-space? */
5527c038 33630+ssize_t xino_fread(vfs_readf_t func, struct file *file, void *kbuf, size_t size,
1facf9fc 33631+ loff_t *pos)
33632+{
33633+ ssize_t err;
33634+ mm_segment_t oldfs;
b752ccd1
AM
33635+ union {
33636+ void *k;
33637+ char __user *u;
33638+ } buf;
1facf9fc 33639+
b752ccd1 33640+ buf.k = kbuf;
1facf9fc 33641+ oldfs = get_fs();
33642+ set_fs(KERNEL_DS);
33643+ do {
33644+ /* todo: signal_pending? */
b752ccd1 33645+ err = func(file, buf.u, size, pos);
1facf9fc 33646+ } while (err == -EAGAIN || err == -EINTR);
33647+ set_fs(oldfs);
33648+
33649+#if 0 /* reserved for future use */
33650+ if (err > 0)
2000de60 33651+ fsnotify_access(file->f_path.dentry);
1facf9fc 33652+#endif
33653+
33654+ return err;
33655+}
33656+
33657+/* ---------------------------------------------------------------------- */
33658+
5527c038 33659+static ssize_t do_xino_fwrite(vfs_writef_t func, struct file *file, void *kbuf,
1facf9fc 33660+ size_t size, loff_t *pos)
33661+{
33662+ ssize_t err;
33663+ mm_segment_t oldfs;
b752ccd1
AM
33664+ union {
33665+ void *k;
33666+ const char __user *u;
33667+ } buf;
1facf9fc 33668+
b752ccd1 33669+ buf.k = kbuf;
1facf9fc 33670+ oldfs = get_fs();
33671+ set_fs(KERNEL_DS);
1facf9fc 33672+ do {
33673+ /* todo: signal_pending? */
b752ccd1 33674+ err = func(file, buf.u, size, pos);
1facf9fc 33675+ } while (err == -EAGAIN || err == -EINTR);
1facf9fc 33676+ set_fs(oldfs);
33677+
33678+#if 0 /* reserved for future use */
33679+ if (err > 0)
2000de60 33680+ fsnotify_modify(file->f_path.dentry);
1facf9fc 33681+#endif
33682+
33683+ return err;
33684+}
33685+
33686+struct do_xino_fwrite_args {
33687+ ssize_t *errp;
5527c038 33688+ vfs_writef_t func;
1facf9fc 33689+ struct file *file;
33690+ void *buf;
33691+ size_t size;
33692+ loff_t *pos;
33693+};
33694+
33695+static void call_do_xino_fwrite(void *args)
33696+{
33697+ struct do_xino_fwrite_args *a = args;
33698+ *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos);
33699+}
33700+
5527c038
JR
33701+ssize_t xino_fwrite(vfs_writef_t func, struct file *file, void *buf,
33702+ size_t size, loff_t *pos)
1facf9fc 33703+{
33704+ ssize_t err;
33705+
33706+ /* todo: signal block and no wkq? */
b752ccd1
AM
33707+ if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
33708+ lockdep_off();
33709+ err = do_xino_fwrite(func, file, buf, size, pos);
33710+ lockdep_on();
33711+ } else {
33712+ /*
33713+ * it breaks RLIMIT_FSIZE and normal user's limit,
33714+ * users should care about quota and real 'filesystem full.'
33715+ */
1facf9fc 33716+ int wkq_err;
33717+ struct do_xino_fwrite_args args = {
33718+ .errp = &err,
33719+ .func = func,
33720+ .file = file,
33721+ .buf = buf,
33722+ .size = size,
33723+ .pos = pos
33724+ };
33725+
33726+ wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
33727+ if (unlikely(wkq_err))
33728+ err = wkq_err;
b752ccd1 33729+ }
1facf9fc 33730+
33731+ return err;
33732+}
33733+
33734+/* ---------------------------------------------------------------------- */
33735+
33736+/*
33737+ * create a new xinofile at the same place/path as @base_file.
33738+ */
33739+struct file *au_xino_create2(struct file *base_file, struct file *copy_src)
33740+{
33741+ struct file *file;
4a4d8108 33742+ struct dentry *base, *parent;
523b37e3 33743+ struct inode *dir, *delegated;
1facf9fc 33744+ struct qstr *name;
1308ab2a 33745+ struct path path;
4a4d8108 33746+ int err;
1facf9fc 33747+
2000de60 33748+ base = base_file->f_path.dentry;
1facf9fc 33749+ parent = base->d_parent; /* dir inode is locked */
5527c038 33750+ dir = d_inode(parent);
1facf9fc 33751+ IMustLock(dir);
33752+
33753+ file = ERR_PTR(-EINVAL);
33754+ name = &base->d_name;
4a4d8108
AM
33755+ path.dentry = vfsub_lookup_one_len(name->name, parent, name->len);
33756+ if (IS_ERR(path.dentry)) {
33757+ file = (void *)path.dentry;
523b37e3
AM
33758+ pr_err("%pd lookup err %ld\n",
33759+ base, PTR_ERR(path.dentry));
1facf9fc 33760+ goto out;
33761+ }
33762+
33763+ /* no need to mnt_want_write() since we call dentry_open() later */
4a4d8108 33764+ err = vfs_create(dir, path.dentry, S_IRUGO | S_IWUGO, NULL);
1facf9fc 33765+ if (unlikely(err)) {
33766+ file = ERR_PTR(err);
523b37e3 33767+ pr_err("%pd create err %d\n", base, err);
1facf9fc 33768+ goto out_dput;
33769+ }
33770+
c06a8ce3 33771+ path.mnt = base_file->f_path.mnt;
4a4d8108 33772+ file = vfsub_dentry_open(&path,
7f207e10 33773+ O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
2cbb1c4b 33774+ /* | __FMODE_NONOTIFY */);
1facf9fc 33775+ if (IS_ERR(file)) {
523b37e3 33776+ pr_err("%pd open err %ld\n", base, PTR_ERR(file));
1facf9fc 33777+ goto out_dput;
33778+ }
33779+
523b37e3
AM
33780+ delegated = NULL;
33781+ err = vfsub_unlink(dir, &file->f_path, &delegated, /*force*/0);
33782+ if (unlikely(err == -EWOULDBLOCK)) {
33783+ pr_warn("cannot retry for NFSv4 delegation"
33784+ " for an internal unlink\n");
33785+ iput(delegated);
33786+ }
1facf9fc 33787+ if (unlikely(err)) {
523b37e3 33788+ pr_err("%pd unlink err %d\n", base, err);
1facf9fc 33789+ goto out_fput;
33790+ }
33791+
33792+ if (copy_src) {
33793+ /* no one can touch copy_src xino */
c06a8ce3 33794+ err = au_copy_file(file, copy_src, vfsub_f_size_read(copy_src));
1facf9fc 33795+ if (unlikely(err)) {
523b37e3 33796+ pr_err("%pd copy err %d\n", base, err);
1facf9fc 33797+ goto out_fput;
33798+ }
33799+ }
33800+ goto out_dput; /* success */
33801+
4f0767ce 33802+out_fput:
1facf9fc 33803+ fput(file);
33804+ file = ERR_PTR(err);
4f0767ce 33805+out_dput:
4a4d8108 33806+ dput(path.dentry);
4f0767ce 33807+out:
1facf9fc 33808+ return file;
33809+}
33810+
33811+struct au_xino_lock_dir {
33812+ struct au_hinode *hdir;
33813+ struct dentry *parent;
33814+ struct mutex *mtx;
33815+};
33816+
33817+static void au_xino_lock_dir(struct super_block *sb, struct file *xino,
33818+ struct au_xino_lock_dir *ldir)
33819+{
33820+ aufs_bindex_t brid, bindex;
33821+
33822+ ldir->hdir = NULL;
33823+ bindex = -1;
33824+ brid = au_xino_brid(sb);
33825+ if (brid >= 0)
33826+ bindex = au_br_index(sb, brid);
33827+ if (bindex >= 0) {
5527c038 33828+ ldir->hdir = au_hi(d_inode(sb->s_root), bindex);
4a4d8108 33829+ au_hn_imtx_lock_nested(ldir->hdir, AuLsc_I_PARENT);
1facf9fc 33830+ } else {
2000de60 33831+ ldir->parent = dget_parent(xino->f_path.dentry);
5527c038 33832+ ldir->mtx = &d_inode(ldir->parent)->i_mutex;
1facf9fc 33833+ mutex_lock_nested(ldir->mtx, AuLsc_I_PARENT);
33834+ }
33835+}
33836+
33837+static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
33838+{
33839+ if (ldir->hdir)
4a4d8108 33840+ au_hn_imtx_unlock(ldir->hdir);
1facf9fc 33841+ else {
33842+ mutex_unlock(ldir->mtx);
33843+ dput(ldir->parent);
33844+ }
33845+}
33846+
33847+/* ---------------------------------------------------------------------- */
33848+
33849+/* trucate xino files asynchronously */
33850+
33851+int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex)
33852+{
33853+ int err;
392086de
AM
33854+ unsigned long jiffy;
33855+ blkcnt_t blocks;
1facf9fc 33856+ aufs_bindex_t bi, bend;
392086de 33857+ struct kstatfs *st;
1facf9fc 33858+ struct au_branch *br;
33859+ struct file *new_xino, *file;
33860+ struct super_block *h_sb;
33861+ struct au_xino_lock_dir ldir;
33862+
392086de
AM
33863+ err = -ENOMEM;
33864+ st = kzalloc(sizeof(*st), GFP_NOFS);
33865+ if (unlikely(!st))
33866+ goto out;
33867+
1facf9fc 33868+ err = -EINVAL;
33869+ bend = au_sbend(sb);
33870+ if (unlikely(bindex < 0 || bend < bindex))
392086de 33871+ goto out_st;
1facf9fc 33872+ br = au_sbr(sb, bindex);
33873+ file = br->br_xino.xi_file;
33874+ if (!file)
392086de
AM
33875+ goto out_st;
33876+
33877+ err = vfs_statfs(&file->f_path, st);
33878+ if (unlikely(err))
33879+ AuErr1("statfs err %d, ignored\n", err);
33880+ jiffy = jiffies;
33881+ blocks = file_inode(file)->i_blocks;
33882+ pr_info("begin truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
33883+ bindex, (u64)blocks, st->f_bfree, st->f_blocks);
1facf9fc 33884+
33885+ au_xino_lock_dir(sb, file, &ldir);
33886+ /* mnt_want_write() is unnecessary here */
33887+ new_xino = au_xino_create2(file, file);
33888+ au_xino_unlock_dir(&ldir);
33889+ err = PTR_ERR(new_xino);
392086de
AM
33890+ if (IS_ERR(new_xino)) {
33891+ pr_err("err %d, ignored\n", err);
33892+ goto out_st;
33893+ }
1facf9fc 33894+ err = 0;
33895+ fput(file);
33896+ br->br_xino.xi_file = new_xino;
33897+
86dc4139 33898+ h_sb = au_br_sb(br);
1facf9fc 33899+ for (bi = 0; bi <= bend; bi++) {
33900+ if (unlikely(bi == bindex))
33901+ continue;
33902+ br = au_sbr(sb, bi);
86dc4139 33903+ if (au_br_sb(br) != h_sb)
1facf9fc 33904+ continue;
33905+
33906+ fput(br->br_xino.xi_file);
33907+ br->br_xino.xi_file = new_xino;
33908+ get_file(new_xino);
33909+ }
33910+
392086de
AM
33911+ err = vfs_statfs(&new_xino->f_path, st);
33912+ if (!err) {
33913+ pr_info("end truncating xino(b%d), ib%llu, %llu/%llu free blks\n",
33914+ bindex, (u64)file_inode(new_xino)->i_blocks,
33915+ st->f_bfree, st->f_blocks);
33916+ if (file_inode(new_xino)->i_blocks < blocks)
33917+ au_sbi(sb)->si_xino_jiffy = jiffy;
33918+ } else
33919+ AuErr1("statfs err %d, ignored\n", err);
33920+
33921+out_st:
33922+ kfree(st);
4f0767ce 33923+out:
1facf9fc 33924+ return err;
33925+}
33926+
33927+struct xino_do_trunc_args {
33928+ struct super_block *sb;
33929+ struct au_branch *br;
33930+};
33931+
33932+static void xino_do_trunc(void *_args)
33933+{
33934+ struct xino_do_trunc_args *args = _args;
33935+ struct super_block *sb;
33936+ struct au_branch *br;
33937+ struct inode *dir;
33938+ int err;
33939+ aufs_bindex_t bindex;
33940+
33941+ err = 0;
33942+ sb = args->sb;
5527c038 33943+ dir = d_inode(sb->s_root);
1facf9fc 33944+ br = args->br;
33945+
33946+ si_noflush_write_lock(sb);
33947+ ii_read_lock_parent(dir);
33948+ bindex = au_br_index(sb, br->br_id);
33949+ err = au_xino_trunc(sb, bindex);
1facf9fc 33950+ ii_read_unlock(dir);
33951+ if (unlikely(err))
392086de 33952+ pr_warn("err b%d, (%d)\n", bindex, err);
1facf9fc 33953+ atomic_dec(&br->br_xino_running);
33954+ atomic_dec(&br->br_count);
1facf9fc 33955+ si_write_unlock(sb);
027c5e7a 33956+ au_nwt_done(&au_sbi(sb)->si_nowait);
1facf9fc 33957+ kfree(args);
33958+}
33959+
392086de
AM
33960+static int xino_trunc_test(struct super_block *sb, struct au_branch *br)
33961+{
33962+ int err;
33963+ struct kstatfs st;
33964+ struct au_sbinfo *sbinfo;
33965+
33966+ /* todo: si_xino_expire and the ratio should be customizable */
33967+ sbinfo = au_sbi(sb);
33968+ if (time_before(jiffies,
33969+ sbinfo->si_xino_jiffy + sbinfo->si_xino_expire))
33970+ return 0;
33971+
33972+ /* truncation border */
33973+ err = vfs_statfs(&br->br_xino.xi_file->f_path, &st);
33974+ if (unlikely(err)) {
33975+ AuErr1("statfs err %d, ignored\n", err);
33976+ return 0;
33977+ }
33978+ if (div64_u64(st.f_bfree * 100, st.f_blocks) >= AUFS_XINO_DEF_TRUNC)
33979+ return 0;
33980+
33981+ return 1;
33982+}
33983+
1facf9fc 33984+static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
33985+{
33986+ struct xino_do_trunc_args *args;
33987+ int wkq_err;
33988+
392086de 33989+ if (!xino_trunc_test(sb, br))
1facf9fc 33990+ return;
33991+
33992+ if (atomic_inc_return(&br->br_xino_running) > 1)
33993+ goto out;
33994+
33995+ /* lock and kfree() will be called in trunc_xino() */
33996+ args = kmalloc(sizeof(*args), GFP_NOFS);
33997+ if (unlikely(!args)) {
33998+ AuErr1("no memory\n");
33999+ goto out_args;
34000+ }
34001+
e49829fe 34002+ atomic_inc(&br->br_count);
1facf9fc 34003+ args->sb = sb;
34004+ args->br = br;
53392da6 34005+ wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
1facf9fc 34006+ if (!wkq_err)
34007+ return; /* success */
34008+
4a4d8108 34009+ pr_err("wkq %d\n", wkq_err);
e49829fe 34010+ atomic_dec(&br->br_count);
1facf9fc 34011+
4f0767ce 34012+out_args:
1facf9fc 34013+ kfree(args);
4f0767ce 34014+out:
e49829fe 34015+ atomic_dec(&br->br_xino_running);
1facf9fc 34016+}
34017+
34018+/* ---------------------------------------------------------------------- */
34019+
5527c038 34020+static int au_xino_do_write(vfs_writef_t write, struct file *file,
1facf9fc 34021+ ino_t h_ino, ino_t ino)
34022+{
34023+ loff_t pos;
34024+ ssize_t sz;
34025+
34026+ pos = h_ino;
34027+ if (unlikely(au_loff_max / sizeof(ino) - 1 < pos)) {
34028+ AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
34029+ return -EFBIG;
34030+ }
34031+ pos *= sizeof(ino);
34032+ sz = xino_fwrite(write, file, &ino, sizeof(ino), &pos);
34033+ if (sz == sizeof(ino))
34034+ return 0; /* success */
34035+
34036+ AuIOErr("write failed (%zd)\n", sz);
34037+ return -EIO;
34038+}
34039+
34040+/*
34041+ * write @ino to the xinofile for the specified branch{@sb, @bindex}
34042+ * at the position of @h_ino.
34043+ * even if @ino is zero, it is written to the xinofile and means no entry.
34044+ * if the size of the xino file on a specific filesystem exceeds the watermark,
34045+ * try truncating it.
34046+ */
34047+int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
34048+ ino_t ino)
34049+{
34050+ int err;
34051+ unsigned int mnt_flags;
34052+ struct au_branch *br;
34053+
34054+ BUILD_BUG_ON(sizeof(long long) != sizeof(au_loff_max)
34055+ || ((loff_t)-1) > 0);
dece6358 34056+ SiMustAnyLock(sb);
1facf9fc 34057+
34058+ mnt_flags = au_mntflags(sb);
34059+ if (!au_opt_test(mnt_flags, XINO))
34060+ return 0;
34061+
34062+ br = au_sbr(sb, bindex);
34063+ err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
34064+ h_ino, ino);
34065+ if (!err) {
34066+ if (au_opt_test(mnt_flags, TRUNC_XINO)
86dc4139 34067+ && au_test_fs_trunc_xino(au_br_sb(br)))
1facf9fc 34068+ xino_try_trunc(sb, br);
34069+ return 0; /* success */
34070+ }
34071+
34072+ AuIOErr("write failed (%d)\n", err);
34073+ return -EIO;
34074+}
34075+
34076+/* ---------------------------------------------------------------------- */
34077+
34078+/* aufs inode number bitmap */
34079+
34080+static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
34081+static ino_t xib_calc_ino(unsigned long pindex, int bit)
34082+{
34083+ ino_t ino;
34084+
34085+ AuDebugOn(bit < 0 || page_bits <= bit);
34086+ ino = AUFS_FIRST_INO + pindex * page_bits + bit;
34087+ return ino;
34088+}
34089+
34090+static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
34091+{
34092+ AuDebugOn(ino < AUFS_FIRST_INO);
34093+ ino -= AUFS_FIRST_INO;
34094+ *pindex = ino / page_bits;
34095+ *bit = ino % page_bits;
34096+}
34097+
34098+static int xib_pindex(struct super_block *sb, unsigned long pindex)
34099+{
34100+ int err;
34101+ loff_t pos;
34102+ ssize_t sz;
34103+ struct au_sbinfo *sbinfo;
34104+ struct file *xib;
34105+ unsigned long *p;
34106+
34107+ sbinfo = au_sbi(sb);
34108+ MtxMustLock(&sbinfo->si_xib_mtx);
34109+ AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
34110+ || !au_opt_test(sbinfo->si_mntflags, XINO));
34111+
34112+ if (pindex == sbinfo->si_xib_last_pindex)
34113+ return 0;
34114+
34115+ xib = sbinfo->si_xib;
34116+ p = sbinfo->si_xib_buf;
34117+ pos = sbinfo->si_xib_last_pindex;
34118+ pos *= PAGE_SIZE;
34119+ sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
34120+ if (unlikely(sz != PAGE_SIZE))
34121+ goto out;
34122+
34123+ pos = pindex;
34124+ pos *= PAGE_SIZE;
c06a8ce3 34125+ if (vfsub_f_size_read(xib) >= pos + PAGE_SIZE)
1facf9fc 34126+ sz = xino_fread(sbinfo->si_xread, xib, p, PAGE_SIZE, &pos);
34127+ else {
34128+ memset(p, 0, PAGE_SIZE);
34129+ sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
34130+ }
34131+ if (sz == PAGE_SIZE) {
34132+ sbinfo->si_xib_last_pindex = pindex;
34133+ return 0; /* success */
34134+ }
34135+
4f0767ce 34136+out:
b752ccd1
AM
34137+ AuIOErr1("write failed (%zd)\n", sz);
34138+ err = sz;
34139+ if (sz >= 0)
34140+ err = -EIO;
34141+ return err;
34142+}
34143+
34144+/* ---------------------------------------------------------------------- */
34145+
34146+static void au_xib_clear_bit(struct inode *inode)
34147+{
34148+ int err, bit;
34149+ unsigned long pindex;
34150+ struct super_block *sb;
34151+ struct au_sbinfo *sbinfo;
34152+
34153+ AuDebugOn(inode->i_nlink);
34154+
34155+ sb = inode->i_sb;
34156+ xib_calc_bit(inode->i_ino, &pindex, &bit);
34157+ AuDebugOn(page_bits <= bit);
34158+ sbinfo = au_sbi(sb);
34159+ mutex_lock(&sbinfo->si_xib_mtx);
34160+ err = xib_pindex(sb, pindex);
34161+ if (!err) {
34162+ clear_bit(bit, sbinfo->si_xib_buf);
34163+ sbinfo->si_xib_next_bit = bit;
34164+ }
34165+ mutex_unlock(&sbinfo->si_xib_mtx);
34166+}
34167+
34168+/* for s_op->delete_inode() */
34169+void au_xino_delete_inode(struct inode *inode, const int unlinked)
34170+{
34171+ int err;
34172+ unsigned int mnt_flags;
34173+ aufs_bindex_t bindex, bend, bi;
34174+ unsigned char try_trunc;
34175+ struct au_iinfo *iinfo;
34176+ struct super_block *sb;
34177+ struct au_hinode *hi;
34178+ struct inode *h_inode;
34179+ struct au_branch *br;
5527c038 34180+ vfs_writef_t xwrite;
b752ccd1
AM
34181+
34182+ sb = inode->i_sb;
34183+ mnt_flags = au_mntflags(sb);
34184+ if (!au_opt_test(mnt_flags, XINO)
34185+ || inode->i_ino == AUFS_ROOT_INO)
34186+ return;
34187+
34188+ if (unlinked) {
34189+ au_xigen_inc(inode);
34190+ au_xib_clear_bit(inode);
34191+ }
34192+
34193+ iinfo = au_ii(inode);
34194+ if (!iinfo)
34195+ return;
1facf9fc 34196+
b752ccd1
AM
34197+ bindex = iinfo->ii_bstart;
34198+ if (bindex < 0)
34199+ return;
1facf9fc 34200+
b752ccd1
AM
34201+ xwrite = au_sbi(sb)->si_xwrite;
34202+ try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
34203+ hi = iinfo->ii_hinode + bindex;
34204+ bend = iinfo->ii_bend;
34205+ for (; bindex <= bend; bindex++, hi++) {
34206+ h_inode = hi->hi_inode;
34207+ if (!h_inode
34208+ || (!unlinked && h_inode->i_nlink))
34209+ continue;
1facf9fc 34210+
b752ccd1
AM
34211+ /* inode may not be revalidated */
34212+ bi = au_br_index(sb, hi->hi_id);
34213+ if (bi < 0)
34214+ continue;
1facf9fc 34215+
b752ccd1
AM
34216+ br = au_sbr(sb, bi);
34217+ err = au_xino_do_write(xwrite, br->br_xino.xi_file,
34218+ h_inode->i_ino, /*ino*/0);
34219+ if (!err && try_trunc
86dc4139 34220+ && au_test_fs_trunc_xino(au_br_sb(br)))
b752ccd1 34221+ xino_try_trunc(sb, br);
1facf9fc 34222+ }
1facf9fc 34223+}
34224+
34225+/* get an unused inode number from bitmap */
34226+ino_t au_xino_new_ino(struct super_block *sb)
34227+{
34228+ ino_t ino;
34229+ unsigned long *p, pindex, ul, pend;
34230+ struct au_sbinfo *sbinfo;
34231+ struct file *file;
34232+ int free_bit, err;
34233+
34234+ if (!au_opt_test(au_mntflags(sb), XINO))
34235+ return iunique(sb, AUFS_FIRST_INO);
34236+
34237+ sbinfo = au_sbi(sb);
34238+ mutex_lock(&sbinfo->si_xib_mtx);
34239+ p = sbinfo->si_xib_buf;
34240+ free_bit = sbinfo->si_xib_next_bit;
34241+ if (free_bit < page_bits && !test_bit(free_bit, p))
34242+ goto out; /* success */
34243+ free_bit = find_first_zero_bit(p, page_bits);
34244+ if (free_bit < page_bits)
34245+ goto out; /* success */
34246+
34247+ pindex = sbinfo->si_xib_last_pindex;
34248+ for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
34249+ err = xib_pindex(sb, ul);
34250+ if (unlikely(err))
34251+ goto out_err;
34252+ free_bit = find_first_zero_bit(p, page_bits);
34253+ if (free_bit < page_bits)
34254+ goto out; /* success */
34255+ }
34256+
34257+ file = sbinfo->si_xib;
c06a8ce3 34258+ pend = vfsub_f_size_read(file) / PAGE_SIZE;
1facf9fc 34259+ for (ul = pindex + 1; ul <= pend; ul++) {
34260+ err = xib_pindex(sb, ul);
34261+ if (unlikely(err))
34262+ goto out_err;
34263+ free_bit = find_first_zero_bit(p, page_bits);
34264+ if (free_bit < page_bits)
34265+ goto out; /* success */
34266+ }
34267+ BUG();
34268+
4f0767ce 34269+out:
1facf9fc 34270+ set_bit(free_bit, p);
7f207e10 34271+ sbinfo->si_xib_next_bit = free_bit + 1;
1facf9fc 34272+ pindex = sbinfo->si_xib_last_pindex;
34273+ mutex_unlock(&sbinfo->si_xib_mtx);
34274+ ino = xib_calc_ino(pindex, free_bit);
34275+ AuDbg("i%lu\n", (unsigned long)ino);
34276+ return ino;
4f0767ce 34277+out_err:
1facf9fc 34278+ mutex_unlock(&sbinfo->si_xib_mtx);
34279+ AuDbg("i0\n");
34280+ return 0;
34281+}
34282+
34283+/*
34284+ * read @ino from xinofile for the specified branch{@sb, @bindex}
34285+ * at the position of @h_ino.
34286+ * if @ino does not exist and @do_new is true, get new one.
34287+ */
34288+int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
34289+ ino_t *ino)
34290+{
34291+ int err;
34292+ ssize_t sz;
34293+ loff_t pos;
34294+ struct file *file;
34295+ struct au_sbinfo *sbinfo;
34296+
34297+ *ino = 0;
34298+ if (!au_opt_test(au_mntflags(sb), XINO))
34299+ return 0; /* no xino */
34300+
34301+ err = 0;
34302+ sbinfo = au_sbi(sb);
34303+ pos = h_ino;
34304+ if (unlikely(au_loff_max / sizeof(*ino) - 1 < pos)) {
34305+ AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
34306+ return -EFBIG;
34307+ }
34308+ pos *= sizeof(*ino);
34309+
34310+ file = au_sbr(sb, bindex)->br_xino.xi_file;
c06a8ce3 34311+ if (vfsub_f_size_read(file) < pos + sizeof(*ino))
1facf9fc 34312+ return 0; /* no ino */
34313+
34314+ sz = xino_fread(sbinfo->si_xread, file, ino, sizeof(*ino), &pos);
34315+ if (sz == sizeof(*ino))
34316+ return 0; /* success */
34317+
34318+ err = sz;
34319+ if (unlikely(sz >= 0)) {
34320+ err = -EIO;
34321+ AuIOErr("xino read error (%zd)\n", sz);
34322+ }
34323+
34324+ return err;
34325+}
34326+
34327+/* ---------------------------------------------------------------------- */
34328+
34329+/* create and set a new xino file */
34330+
34331+struct file *au_xino_create(struct super_block *sb, char *fname, int silent)
34332+{
34333+ struct file *file;
34334+ struct dentry *h_parent, *d;
b912730e 34335+ struct inode *h_dir, *inode;
1facf9fc 34336+ int err;
34337+
34338+ /*
34339+ * at mount-time, and the xino file is the default path,
4a4d8108 34340+ * hnotify is disabled so we have no notify events to ignore.
1facf9fc 34341+ * when a user specified the xino, we cannot get au_hdir to be ignored.
34342+ */
7f207e10 34343+ file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
2cbb1c4b 34344+ /* | __FMODE_NONOTIFY */,
1facf9fc 34345+ S_IRUGO | S_IWUGO);
34346+ if (IS_ERR(file)) {
34347+ if (!silent)
4a4d8108 34348+ pr_err("open %s(%ld)\n", fname, PTR_ERR(file));
1facf9fc 34349+ return file;
34350+ }
34351+
34352+ /* keep file count */
b912730e
AM
34353+ err = 0;
34354+ inode = file_inode(file);
2000de60 34355+ h_parent = dget_parent(file->f_path.dentry);
5527c038 34356+ h_dir = d_inode(h_parent);
1facf9fc 34357+ mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
34358+ /* mnt_want_write() is unnecessary here */
523b37e3 34359+ /* no delegation since it is just created */
b912730e
AM
34360+ if (inode->i_nlink)
34361+ err = vfsub_unlink(h_dir, &file->f_path, /*delegated*/NULL,
34362+ /*force*/0);
1facf9fc 34363+ mutex_unlock(&h_dir->i_mutex);
34364+ dput(h_parent);
34365+ if (unlikely(err)) {
34366+ if (!silent)
4a4d8108 34367+ pr_err("unlink %s(%d)\n", fname, err);
1facf9fc 34368+ goto out;
34369+ }
34370+
34371+ err = -EINVAL;
2000de60 34372+ d = file->f_path.dentry;
1facf9fc 34373+ if (unlikely(sb == d->d_sb)) {
34374+ if (!silent)
4a4d8108 34375+ pr_err("%s must be outside\n", fname);
1facf9fc 34376+ goto out;
34377+ }
34378+ if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
34379+ if (!silent)
4a4d8108
AM
34380+ pr_err("xino doesn't support %s(%s)\n",
34381+ fname, au_sbtype(d->d_sb));
1facf9fc 34382+ goto out;
34383+ }
34384+ return file; /* success */
34385+
4f0767ce 34386+out:
1facf9fc 34387+ fput(file);
34388+ file = ERR_PTR(err);
34389+ return file;
34390+}
34391+
34392+/*
34393+ * find another branch who is on the same filesystem of the specified
34394+ * branch{@btgt}. search until @bend.
34395+ */
34396+static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
34397+ aufs_bindex_t bend)
34398+{
34399+ aufs_bindex_t bindex;
34400+ struct super_block *tgt_sb = au_sbr_sb(sb, btgt);
34401+
34402+ for (bindex = 0; bindex < btgt; bindex++)
34403+ if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
34404+ return bindex;
34405+ for (bindex++; bindex <= bend; bindex++)
34406+ if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
34407+ return bindex;
34408+ return -1;
34409+}
34410+
34411+/* ---------------------------------------------------------------------- */
34412+
34413+/*
34414+ * initialize the xinofile for the specified branch @br
34415+ * at the place/path where @base_file indicates.
34416+ * test whether another branch is on the same filesystem or not,
34417+ * if @do_test is true.
34418+ */
34419+int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
34420+ struct file *base_file, int do_test)
34421+{
34422+ int err;
34423+ ino_t ino;
34424+ aufs_bindex_t bend, bindex;
34425+ struct au_branch *shared_br, *b;
34426+ struct file *file;
34427+ struct super_block *tgt_sb;
34428+
34429+ shared_br = NULL;
34430+ bend = au_sbend(sb);
34431+ if (do_test) {
86dc4139 34432+ tgt_sb = au_br_sb(br);
1facf9fc 34433+ for (bindex = 0; bindex <= bend; bindex++) {
34434+ b = au_sbr(sb, bindex);
86dc4139 34435+ if (tgt_sb == au_br_sb(b)) {
1facf9fc 34436+ shared_br = b;
34437+ break;
34438+ }
34439+ }
34440+ }
34441+
34442+ if (!shared_br || !shared_br->br_xino.xi_file) {
34443+ struct au_xino_lock_dir ldir;
34444+
34445+ au_xino_lock_dir(sb, base_file, &ldir);
34446+ /* mnt_want_write() is unnecessary here */
34447+ file = au_xino_create2(base_file, NULL);
34448+ au_xino_unlock_dir(&ldir);
34449+ err = PTR_ERR(file);
34450+ if (IS_ERR(file))
34451+ goto out;
34452+ br->br_xino.xi_file = file;
34453+ } else {
34454+ br->br_xino.xi_file = shared_br->br_xino.xi_file;
34455+ get_file(br->br_xino.xi_file);
34456+ }
34457+
34458+ ino = AUFS_ROOT_INO;
34459+ err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
34460+ h_ino, ino);
b752ccd1
AM
34461+ if (unlikely(err)) {
34462+ fput(br->br_xino.xi_file);
34463+ br->br_xino.xi_file = NULL;
34464+ }
1facf9fc 34465+
4f0767ce 34466+out:
1facf9fc 34467+ return err;
34468+}
34469+
34470+/* ---------------------------------------------------------------------- */
34471+
34472+/* trucate a xino bitmap file */
34473+
34474+/* todo: slow */
34475+static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
34476+{
34477+ int err, bit;
34478+ ssize_t sz;
34479+ unsigned long pindex;
34480+ loff_t pos, pend;
34481+ struct au_sbinfo *sbinfo;
5527c038 34482+ vfs_readf_t func;
1facf9fc 34483+ ino_t *ino;
34484+ unsigned long *p;
34485+
34486+ err = 0;
34487+ sbinfo = au_sbi(sb);
dece6358 34488+ MtxMustLock(&sbinfo->si_xib_mtx);
1facf9fc 34489+ p = sbinfo->si_xib_buf;
34490+ func = sbinfo->si_xread;
c06a8ce3 34491+ pend = vfsub_f_size_read(file);
1facf9fc 34492+ pos = 0;
34493+ while (pos < pend) {
34494+ sz = xino_fread(func, file, page, PAGE_SIZE, &pos);
34495+ err = sz;
34496+ if (unlikely(sz <= 0))
34497+ goto out;
34498+
34499+ err = 0;
34500+ for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
34501+ if (unlikely(*ino < AUFS_FIRST_INO))
34502+ continue;
34503+
34504+ xib_calc_bit(*ino, &pindex, &bit);
34505+ AuDebugOn(page_bits <= bit);
34506+ err = xib_pindex(sb, pindex);
34507+ if (!err)
34508+ set_bit(bit, p);
34509+ else
34510+ goto out;
34511+ }
34512+ }
34513+
4f0767ce 34514+out:
1facf9fc 34515+ return err;
34516+}
34517+
34518+static int xib_restore(struct super_block *sb)
34519+{
34520+ int err;
34521+ aufs_bindex_t bindex, bend;
34522+ void *page;
34523+
34524+ err = -ENOMEM;
34525+ page = (void *)__get_free_page(GFP_NOFS);
34526+ if (unlikely(!page))
34527+ goto out;
34528+
34529+ err = 0;
34530+ bend = au_sbend(sb);
34531+ for (bindex = 0; !err && bindex <= bend; bindex++)
34532+ if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0)
34533+ err = do_xib_restore
34534+ (sb, au_sbr(sb, bindex)->br_xino.xi_file, page);
34535+ else
34536+ AuDbg("b%d\n", bindex);
34537+ free_page((unsigned long)page);
34538+
4f0767ce 34539+out:
1facf9fc 34540+ return err;
34541+}
34542+
34543+int au_xib_trunc(struct super_block *sb)
34544+{
34545+ int err;
34546+ ssize_t sz;
34547+ loff_t pos;
34548+ struct au_xino_lock_dir ldir;
34549+ struct au_sbinfo *sbinfo;
34550+ unsigned long *p;
34551+ struct file *file;
34552+
dece6358
AM
34553+ SiMustWriteLock(sb);
34554+
1facf9fc 34555+ err = 0;
34556+ sbinfo = au_sbi(sb);
34557+ if (!au_opt_test(sbinfo->si_mntflags, XINO))
34558+ goto out;
34559+
34560+ file = sbinfo->si_xib;
c06a8ce3 34561+ if (vfsub_f_size_read(file) <= PAGE_SIZE)
1facf9fc 34562+ goto out;
34563+
34564+ au_xino_lock_dir(sb, file, &ldir);
34565+ /* mnt_want_write() is unnecessary here */
34566+ file = au_xino_create2(sbinfo->si_xib, NULL);
34567+ au_xino_unlock_dir(&ldir);
34568+ err = PTR_ERR(file);
34569+ if (IS_ERR(file))
34570+ goto out;
34571+ fput(sbinfo->si_xib);
34572+ sbinfo->si_xib = file;
34573+
34574+ p = sbinfo->si_xib_buf;
34575+ memset(p, 0, PAGE_SIZE);
34576+ pos = 0;
34577+ sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xib, p, PAGE_SIZE, &pos);
34578+ if (unlikely(sz != PAGE_SIZE)) {
34579+ err = sz;
34580+ AuIOErr("err %d\n", err);
34581+ if (sz >= 0)
34582+ err = -EIO;
34583+ goto out;
34584+ }
34585+
34586+ mutex_lock(&sbinfo->si_xib_mtx);
34587+ /* mnt_want_write() is unnecessary here */
34588+ err = xib_restore(sb);
34589+ mutex_unlock(&sbinfo->si_xib_mtx);
34590+
34591+out:
34592+ return err;
34593+}
34594+
34595+/* ---------------------------------------------------------------------- */
34596+
34597+/*
34598+ * xino mount option handlers
34599+ */
1facf9fc 34600+
34601+/* xino bitmap */
34602+static void xino_clear_xib(struct super_block *sb)
34603+{
34604+ struct au_sbinfo *sbinfo;
34605+
dece6358
AM
34606+ SiMustWriteLock(sb);
34607+
1facf9fc 34608+ sbinfo = au_sbi(sb);
34609+ sbinfo->si_xread = NULL;
34610+ sbinfo->si_xwrite = NULL;
34611+ if (sbinfo->si_xib)
34612+ fput(sbinfo->si_xib);
34613+ sbinfo->si_xib = NULL;
34614+ free_page((unsigned long)sbinfo->si_xib_buf);
34615+ sbinfo->si_xib_buf = NULL;
34616+}
34617+
34618+static int au_xino_set_xib(struct super_block *sb, struct file *base)
34619+{
34620+ int err;
34621+ loff_t pos;
34622+ struct au_sbinfo *sbinfo;
34623+ struct file *file;
34624+
dece6358
AM
34625+ SiMustWriteLock(sb);
34626+
1facf9fc 34627+ sbinfo = au_sbi(sb);
34628+ file = au_xino_create2(base, sbinfo->si_xib);
34629+ err = PTR_ERR(file);
34630+ if (IS_ERR(file))
34631+ goto out;
34632+ if (sbinfo->si_xib)
34633+ fput(sbinfo->si_xib);
34634+ sbinfo->si_xib = file;
5527c038
JR
34635+ sbinfo->si_xread = vfs_readf(file);
34636+ sbinfo->si_xwrite = vfs_writef(file);
1facf9fc 34637+
34638+ err = -ENOMEM;
34639+ if (!sbinfo->si_xib_buf)
34640+ sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
34641+ if (unlikely(!sbinfo->si_xib_buf))
34642+ goto out_unset;
34643+
34644+ sbinfo->si_xib_last_pindex = 0;
34645+ sbinfo->si_xib_next_bit = 0;
c06a8ce3 34646+ if (vfsub_f_size_read(file) < PAGE_SIZE) {
1facf9fc 34647+ pos = 0;
34648+ err = xino_fwrite(sbinfo->si_xwrite, file, sbinfo->si_xib_buf,
34649+ PAGE_SIZE, &pos);
34650+ if (unlikely(err != PAGE_SIZE))
34651+ goto out_free;
34652+ }
34653+ err = 0;
34654+ goto out; /* success */
34655+
4f0767ce 34656+out_free:
1facf9fc 34657+ free_page((unsigned long)sbinfo->si_xib_buf);
b752ccd1
AM
34658+ sbinfo->si_xib_buf = NULL;
34659+ if (err >= 0)
34660+ err = -EIO;
4f0767ce 34661+out_unset:
b752ccd1
AM
34662+ fput(sbinfo->si_xib);
34663+ sbinfo->si_xib = NULL;
34664+ sbinfo->si_xread = NULL;
34665+ sbinfo->si_xwrite = NULL;
4f0767ce 34666+out:
b752ccd1 34667+ return err;
1facf9fc 34668+}
34669+
b752ccd1
AM
34670+/* xino for each branch */
34671+static void xino_clear_br(struct super_block *sb)
34672+{
34673+ aufs_bindex_t bindex, bend;
34674+ struct au_branch *br;
1facf9fc 34675+
b752ccd1
AM
34676+ bend = au_sbend(sb);
34677+ for (bindex = 0; bindex <= bend; bindex++) {
34678+ br = au_sbr(sb, bindex);
34679+ if (!br || !br->br_xino.xi_file)
34680+ continue;
34681+
34682+ fput(br->br_xino.xi_file);
34683+ br->br_xino.xi_file = NULL;
34684+ }
34685+}
34686+
34687+static int au_xino_set_br(struct super_block *sb, struct file *base)
1facf9fc 34688+{
34689+ int err;
b752ccd1
AM
34690+ ino_t ino;
34691+ aufs_bindex_t bindex, bend, bshared;
34692+ struct {
34693+ struct file *old, *new;
34694+ } *fpair, *p;
34695+ struct au_branch *br;
34696+ struct inode *inode;
5527c038 34697+ vfs_writef_t writef;
1facf9fc 34698+
b752ccd1
AM
34699+ SiMustWriteLock(sb);
34700+
34701+ err = -ENOMEM;
34702+ bend = au_sbend(sb);
34703+ fpair = kcalloc(bend + 1, sizeof(*fpair), GFP_NOFS);
34704+ if (unlikely(!fpair))
1facf9fc 34705+ goto out;
34706+
5527c038 34707+ inode = d_inode(sb->s_root);
b752ccd1
AM
34708+ ino = AUFS_ROOT_INO;
34709+ writef = au_sbi(sb)->si_xwrite;
34710+ for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
34711+ br = au_sbr(sb, bindex);
34712+ bshared = is_sb_shared(sb, bindex, bindex - 1);
34713+ if (bshared >= 0) {
34714+ /* shared xino */
34715+ *p = fpair[bshared];
34716+ get_file(p->new);
34717+ }
34718+
34719+ if (!p->new) {
34720+ /* new xino */
34721+ p->old = br->br_xino.xi_file;
34722+ p->new = au_xino_create2(base, br->br_xino.xi_file);
34723+ err = PTR_ERR(p->new);
34724+ if (IS_ERR(p->new)) {
34725+ p->new = NULL;
34726+ goto out_pair;
34727+ }
34728+ }
34729+
34730+ err = au_xino_do_write(writef, p->new,
34731+ au_h_iptr(inode, bindex)->i_ino, ino);
34732+ if (unlikely(err))
34733+ goto out_pair;
34734+ }
34735+
34736+ for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
34737+ br = au_sbr(sb, bindex);
34738+ if (br->br_xino.xi_file)
34739+ fput(br->br_xino.xi_file);
34740+ get_file(p->new);
34741+ br->br_xino.xi_file = p->new;
34742+ }
1facf9fc 34743+
4f0767ce 34744+out_pair:
b752ccd1
AM
34745+ for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++)
34746+ if (p->new)
34747+ fput(p->new);
34748+ else
34749+ break;
34750+ kfree(fpair);
4f0767ce 34751+out:
1facf9fc 34752+ return err;
34753+}
b752ccd1
AM
34754+
34755+void au_xino_clr(struct super_block *sb)
34756+{
34757+ struct au_sbinfo *sbinfo;
34758+
34759+ au_xigen_clr(sb);
34760+ xino_clear_xib(sb);
34761+ xino_clear_br(sb);
34762+ sbinfo = au_sbi(sb);
34763+ /* lvalue, do not call au_mntflags() */
34764+ au_opt_clr(sbinfo->si_mntflags, XINO);
34765+}
34766+
34767+int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount)
34768+{
34769+ int err, skip;
34770+ struct dentry *parent, *cur_parent;
34771+ struct qstr *dname, *cur_name;
34772+ struct file *cur_xino;
34773+ struct inode *dir;
34774+ struct au_sbinfo *sbinfo;
34775+
34776+ SiMustWriteLock(sb);
34777+
34778+ err = 0;
34779+ sbinfo = au_sbi(sb);
2000de60 34780+ parent = dget_parent(xino->file->f_path.dentry);
b752ccd1
AM
34781+ if (remount) {
34782+ skip = 0;
2000de60 34783+ dname = &xino->file->f_path.dentry->d_name;
b752ccd1
AM
34784+ cur_xino = sbinfo->si_xib;
34785+ if (cur_xino) {
2000de60
JR
34786+ cur_parent = dget_parent(cur_xino->f_path.dentry);
34787+ cur_name = &cur_xino->f_path.dentry->d_name;
b752ccd1 34788+ skip = (cur_parent == parent
38d290e6 34789+ && au_qstreq(dname, cur_name));
b752ccd1
AM
34790+ dput(cur_parent);
34791+ }
34792+ if (skip)
34793+ goto out;
34794+ }
34795+
34796+ au_opt_set(sbinfo->si_mntflags, XINO);
5527c038 34797+ dir = d_inode(parent);
b752ccd1
AM
34798+ mutex_lock_nested(&dir->i_mutex, AuLsc_I_PARENT);
34799+ /* mnt_want_write() is unnecessary here */
34800+ err = au_xino_set_xib(sb, xino->file);
34801+ if (!err)
34802+ err = au_xigen_set(sb, xino->file);
34803+ if (!err)
34804+ err = au_xino_set_br(sb, xino->file);
34805+ mutex_unlock(&dir->i_mutex);
34806+ if (!err)
34807+ goto out; /* success */
34808+
34809+ /* reset all */
34810+ AuIOErr("failed creating xino(%d).\n", err);
c1595e42
JR
34811+ au_xigen_clr(sb);
34812+ xino_clear_xib(sb);
b752ccd1 34813+
4f0767ce 34814+out:
b752ccd1
AM
34815+ dput(parent);
34816+ return err;
34817+}
34818+
34819+/* ---------------------------------------------------------------------- */
34820+
34821+/*
34822+ * create a xinofile at the default place/path.
34823+ */
34824+struct file *au_xino_def(struct super_block *sb)
34825+{
34826+ struct file *file;
34827+ char *page, *p;
34828+ struct au_branch *br;
34829+ struct super_block *h_sb;
34830+ struct path path;
34831+ aufs_bindex_t bend, bindex, bwr;
34832+
34833+ br = NULL;
34834+ bend = au_sbend(sb);
34835+ bwr = -1;
34836+ for (bindex = 0; bindex <= bend; bindex++) {
34837+ br = au_sbr(sb, bindex);
34838+ if (au_br_writable(br->br_perm)
86dc4139 34839+ && !au_test_fs_bad_xino(au_br_sb(br))) {
b752ccd1
AM
34840+ bwr = bindex;
34841+ break;
34842+ }
34843+ }
34844+
7f207e10
AM
34845+ if (bwr >= 0) {
34846+ file = ERR_PTR(-ENOMEM);
537831f9 34847+ page = (void *)__get_free_page(GFP_NOFS);
7f207e10
AM
34848+ if (unlikely(!page))
34849+ goto out;
86dc4139 34850+ path.mnt = au_br_mnt(br);
7f207e10
AM
34851+ path.dentry = au_h_dptr(sb->s_root, bwr);
34852+ p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
34853+ file = (void *)p;
34854+ if (!IS_ERR(p)) {
34855+ strcat(p, "/" AUFS_XINO_FNAME);
34856+ AuDbg("%s\n", p);
34857+ file = au_xino_create(sb, p, /*silent*/0);
34858+ if (!IS_ERR(file))
34859+ au_xino_brid_set(sb, br->br_id);
34860+ }
537831f9 34861+ free_page((unsigned long)page);
7f207e10
AM
34862+ } else {
34863+ file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0);
34864+ if (IS_ERR(file))
34865+ goto out;
2000de60 34866+ h_sb = file->f_path.dentry->d_sb;
7f207e10
AM
34867+ if (unlikely(au_test_fs_bad_xino(h_sb))) {
34868+ pr_err("xino doesn't support %s(%s)\n",
34869+ AUFS_XINO_DEFPATH, au_sbtype(h_sb));
34870+ fput(file);
34871+ file = ERR_PTR(-EINVAL);
34872+ }
34873+ if (!IS_ERR(file))
34874+ au_xino_brid_set(sb, -1);
34875+ }
0c5527e5 34876+
7f207e10
AM
34877+out:
34878+ return file;
34879+}
34880+
34881+/* ---------------------------------------------------------------------- */
34882+
34883+int au_xino_path(struct seq_file *seq, struct file *file)
34884+{
34885+ int err;
34886+
34887+ err = au_seq_path(seq, &file->f_path);
34888+ if (unlikely(err < 0))
34889+ goto out;
34890+
34891+ err = 0;
34892+#define Deleted "\\040(deleted)"
34893+ seq->count -= sizeof(Deleted) - 1;
34894+ AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
34895+ sizeof(Deleted) - 1));
34896+#undef Deleted
34897+
34898+out:
34899+ return err;
34900+}
537831f9
AM
34901diff -urN /usr/share/empty/include/uapi/linux/aufs_type.h linux/include/uapi/linux/aufs_type.h
34902--- /usr/share/empty/include/uapi/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
c2c0f25c 34903+++ linux/include/uapi/linux/aufs_type.h 2015-09-24 10:47:58.258053165 +0200
c1595e42 34904@@ -0,0 +1,419 @@
7f207e10 34905+/*
2000de60 34906+ * Copyright (C) 2005-2015 Junjiro R. Okajima
7f207e10
AM
34907+ *
34908+ * This program, aufs is free software; you can redistribute it and/or modify
34909+ * it under the terms of the GNU General Public License as published by
34910+ * the Free Software Foundation; either version 2 of the License, or
34911+ * (at your option) any later version.
34912+ *
34913+ * This program is distributed in the hope that it will be useful,
34914+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
34915+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
34916+ * GNU General Public License for more details.
34917+ *
34918+ * You should have received a copy of the GNU General Public License
523b37e3 34919+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
7f207e10
AM
34920+ */
34921+
34922+#ifndef __AUFS_TYPE_H__
34923+#define __AUFS_TYPE_H__
34924+
f6c5ef8b
AM
34925+#define AUFS_NAME "aufs"
34926+
9dbd164d 34927+#ifdef __KERNEL__
f6c5ef8b
AM
34928+/*
34929+ * define it before including all other headers.
34930+ * sched.h may use pr_* macros before defining "current", so define the
34931+ * no-current version first, and re-define later.
34932+ */
34933+#define pr_fmt(fmt) AUFS_NAME " %s:%d: " fmt, __func__, __LINE__
34934+#include <linux/sched.h>
34935+#undef pr_fmt
a2a7ad62
AM
34936+#define pr_fmt(fmt) \
34937+ AUFS_NAME " %s:%d:%.*s[%d]: " fmt, __func__, __LINE__, \
34938+ (int)sizeof(current->comm), current->comm, current->pid
9dbd164d
AM
34939+#else
34940+#include <stdint.h>
34941+#include <sys/types.h>
f6c5ef8b 34942+#endif /* __KERNEL__ */
7f207e10 34943+
f6c5ef8b
AM
34944+#include <linux/limits.h>
34945+
c2c0f25c 34946+#define AUFS_VERSION "4.x-rcN-20150921"
7f207e10
AM
34947+
34948+/* todo? move this to linux-2.6.19/include/magic.h */
34949+#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
34950+
34951+/* ---------------------------------------------------------------------- */
34952+
34953+#ifdef CONFIG_AUFS_BRANCH_MAX_127
9dbd164d 34954+typedef int8_t aufs_bindex_t;
7f207e10
AM
34955+#define AUFS_BRANCH_MAX 127
34956+#else
9dbd164d 34957+typedef int16_t aufs_bindex_t;
7f207e10
AM
34958+#ifdef CONFIG_AUFS_BRANCH_MAX_511
34959+#define AUFS_BRANCH_MAX 511
34960+#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
34961+#define AUFS_BRANCH_MAX 1023
34962+#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
34963+#define AUFS_BRANCH_MAX 32767
34964+#endif
34965+#endif
34966+
34967+#ifdef __KERNEL__
34968+#ifndef AUFS_BRANCH_MAX
34969+#error unknown CONFIG_AUFS_BRANCH_MAX value
34970+#endif
34971+#endif /* __KERNEL__ */
34972+
34973+/* ---------------------------------------------------------------------- */
34974+
7f207e10
AM
34975+#define AUFS_FSTYPE AUFS_NAME
34976+
34977+#define AUFS_ROOT_INO 2
34978+#define AUFS_FIRST_INO 11
34979+
34980+#define AUFS_WH_PFX ".wh."
34981+#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
34982+#define AUFS_WH_TMP_LEN 4
86dc4139 34983+/* a limit for rmdir/rename a dir and copyup */
7f207e10
AM
34984+#define AUFS_MAX_NAMELEN (NAME_MAX \
34985+ - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
34986+ - 1 /* dot */\
34987+ - AUFS_WH_TMP_LEN) /* hex */
34988+#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
34989+#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
392086de
AM
34990+#define AUFS_XINO_DEF_SEC 30 /* seconds */
34991+#define AUFS_XINO_DEF_TRUNC 45 /* percentage */
7f207e10
AM
34992+#define AUFS_DIRWH_DEF 3
34993+#define AUFS_RDCACHE_DEF 10 /* seconds */
027c5e7a 34994+#define AUFS_RDCACHE_MAX 3600 /* seconds */
7f207e10
AM
34995+#define AUFS_RDBLK_DEF 512 /* bytes */
34996+#define AUFS_RDHASH_DEF 32
34997+#define AUFS_WKQ_NAME AUFS_NAME "d"
027c5e7a
AM
34998+#define AUFS_MFS_DEF_SEC 30 /* seconds */
34999+#define AUFS_MFS_MAX_SEC 3600 /* seconds */
076b876e 35000+#define AUFS_FHSM_CACHE_DEF_SEC 30 /* seconds */
86dc4139 35001+#define AUFS_PLINK_WARN 50 /* number of plinks in a single bucket */
7f207e10
AM
35002+
35003+/* pseudo-link maintenace under /proc */
35004+#define AUFS_PLINK_MAINT_NAME "plink_maint"
35005+#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
35006+#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
35007+
35008+#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
35009+#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
35010+
35011+#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
35012+#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
35013+#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
35014+
35015+/* doubly whiteouted */
35016+#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
35017+#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
35018+#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
35019+
1e00d052 35020+/* branch permissions and attributes */
7f207e10
AM
35021+#define AUFS_BRPERM_RW "rw"
35022+#define AUFS_BRPERM_RO "ro"
35023+#define AUFS_BRPERM_RR "rr"
076b876e
AM
35024+#define AUFS_BRATTR_COO_REG "coo_reg"
35025+#define AUFS_BRATTR_COO_ALL "coo_all"
35026+#define AUFS_BRATTR_FHSM "fhsm"
35027+#define AUFS_BRATTR_UNPIN "unpin"
c1595e42
JR
35028+#define AUFS_BRATTR_ICEX "icex"
35029+#define AUFS_BRATTR_ICEX_SEC "icexsec"
35030+#define AUFS_BRATTR_ICEX_SYS "icexsys"
35031+#define AUFS_BRATTR_ICEX_TR "icextr"
35032+#define AUFS_BRATTR_ICEX_USR "icexusr"
35033+#define AUFS_BRATTR_ICEX_OTH "icexoth"
1e00d052
AM
35034+#define AUFS_BRRATTR_WH "wh"
35035+#define AUFS_BRWATTR_NLWH "nolwh"
076b876e
AM
35036+#define AUFS_BRWATTR_MOO "moo"
35037+
35038+#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
35039+#define AuBrPerm_RO (1 << 1) /* readonly */
35040+#define AuBrPerm_RR (1 << 2) /* natively readonly */
35041+#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
35042+
35043+#define AuBrAttr_COO_REG (1 << 3) /* copy-up on open */
35044+#define AuBrAttr_COO_ALL (1 << 4)
35045+#define AuBrAttr_COO_Mask (AuBrAttr_COO_REG | AuBrAttr_COO_ALL)
35046+
35047+#define AuBrAttr_FHSM (1 << 5) /* file-based hsm */
35048+#define AuBrAttr_UNPIN (1 << 6) /* rename-able top dir of
c1595e42
JR
35049+ branch. meaningless since
35050+ linux-3.18-rc1 */
35051+
35052+/* ignore error in copying XATTR */
35053+#define AuBrAttr_ICEX_SEC (1 << 7)
35054+#define AuBrAttr_ICEX_SYS (1 << 8)
35055+#define AuBrAttr_ICEX_TR (1 << 9)
35056+#define AuBrAttr_ICEX_USR (1 << 10)
35057+#define AuBrAttr_ICEX_OTH (1 << 11)
35058+#define AuBrAttr_ICEX (AuBrAttr_ICEX_SEC \
35059+ | AuBrAttr_ICEX_SYS \
35060+ | AuBrAttr_ICEX_TR \
35061+ | AuBrAttr_ICEX_USR \
35062+ | AuBrAttr_ICEX_OTH)
35063+
35064+#define AuBrRAttr_WH (1 << 12) /* whiteout-able */
076b876e
AM
35065+#define AuBrRAttr_Mask AuBrRAttr_WH
35066+
c1595e42
JR
35067+#define AuBrWAttr_NoLinkWH (1 << 13) /* un-hardlinkable whiteouts */
35068+#define AuBrWAttr_MOO (1 << 14) /* move-up on open */
076b876e
AM
35069+#define AuBrWAttr_Mask (AuBrWAttr_NoLinkWH | AuBrWAttr_MOO)
35070+
35071+#define AuBrAttr_CMOO_Mask (AuBrAttr_COO_Mask | AuBrWAttr_MOO)
35072+
c1595e42 35073+/* #warning test userspace */
076b876e
AM
35074+#ifdef __KERNEL__
35075+#ifndef CONFIG_AUFS_FHSM
35076+#undef AuBrAttr_FHSM
35077+#define AuBrAttr_FHSM 0
35078+#endif
c1595e42
JR
35079+#ifndef CONFIG_AUFS_XATTR
35080+#undef AuBrAttr_ICEX
35081+#define AuBrAttr_ICEX 0
35082+#undef AuBrAttr_ICEX_SEC
35083+#define AuBrAttr_ICEX_SEC 0
35084+#undef AuBrAttr_ICEX_SYS
35085+#define AuBrAttr_ICEX_SYS 0
35086+#undef AuBrAttr_ICEX_TR
35087+#define AuBrAttr_ICEX_TR 0
35088+#undef AuBrAttr_ICEX_USR
35089+#define AuBrAttr_ICEX_USR 0
35090+#undef AuBrAttr_ICEX_OTH
35091+#define AuBrAttr_ICEX_OTH 0
35092+#endif
076b876e
AM
35093+#endif
35094+
35095+/* the longest combination */
c1595e42
JR
35096+/* AUFS_BRATTR_ICEX and AUFS_BRATTR_ICEX_TR don't affect here */
35097+#define AuBrPermStrSz sizeof(AUFS_BRPERM_RW \
35098+ "+" AUFS_BRATTR_COO_REG \
35099+ "+" AUFS_BRATTR_FHSM \
35100+ "+" AUFS_BRATTR_UNPIN \
7e9cd9fe
AM
35101+ "+" AUFS_BRATTR_ICEX_SEC \
35102+ "+" AUFS_BRATTR_ICEX_SYS \
35103+ "+" AUFS_BRATTR_ICEX_USR \
35104+ "+" AUFS_BRATTR_ICEX_OTH \
076b876e
AM
35105+ "+" AUFS_BRWATTR_NLWH)
35106+
35107+typedef struct {
35108+ char a[AuBrPermStrSz];
35109+} au_br_perm_str_t;
35110+
35111+static inline int au_br_writable(int brperm)
35112+{
35113+ return brperm & AuBrPerm_RW;
35114+}
35115+
35116+static inline int au_br_whable(int brperm)
35117+{
35118+ return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
35119+}
35120+
35121+static inline int au_br_wh_linkable(int brperm)
35122+{
35123+ return !(brperm & AuBrWAttr_NoLinkWH);
35124+}
35125+
35126+static inline int au_br_cmoo(int brperm)
35127+{
35128+ return brperm & AuBrAttr_CMOO_Mask;
35129+}
35130+
35131+static inline int au_br_fhsm(int brperm)
35132+{
35133+ return brperm & AuBrAttr_FHSM;
35134+}
7f207e10
AM
35135+
35136+/* ---------------------------------------------------------------------- */
35137+
35138+/* ioctl */
35139+enum {
35140+ /* readdir in userspace */
35141+ AuCtl_RDU,
35142+ AuCtl_RDU_INO,
35143+
076b876e
AM
35144+ AuCtl_WBR_FD, /* pathconf wrapper */
35145+ AuCtl_IBUSY, /* busy inode */
35146+ AuCtl_MVDOWN, /* move-down */
35147+ AuCtl_BR, /* info about branches */
35148+ AuCtl_FHSM_FD /* connection for fhsm */
7f207e10
AM
35149+};
35150+
35151+/* borrowed from linux/include/linux/kernel.h */
35152+#ifndef ALIGN
35153+#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
35154+#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
35155+#endif
35156+
35157+/* borrowed from linux/include/linux/compiler-gcc3.h */
35158+#ifndef __aligned
35159+#define __aligned(x) __attribute__((aligned(x)))
53392da6
AM
35160+#endif
35161+
35162+#ifdef __KERNEL__
35163+#ifndef __packed
7f207e10
AM
35164+#define __packed __attribute__((packed))
35165+#endif
53392da6 35166+#endif
7f207e10
AM
35167+
35168+struct au_rdu_cookie {
9dbd164d
AM
35169+ uint64_t h_pos;
35170+ int16_t bindex;
35171+ uint8_t flags;
35172+ uint8_t pad;
35173+ uint32_t generation;
7f207e10
AM
35174+} __aligned(8);
35175+
35176+struct au_rdu_ent {
9dbd164d
AM
35177+ uint64_t ino;
35178+ int16_t bindex;
35179+ uint8_t type;
35180+ uint8_t nlen;
35181+ uint8_t wh;
7f207e10
AM
35182+ char name[0];
35183+} __aligned(8);
35184+
35185+static inline int au_rdu_len(int nlen)
35186+{
35187+ /* include the terminating NULL */
35188+ return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
9dbd164d 35189+ sizeof(uint64_t));
7f207e10
AM
35190+}
35191+
35192+union au_rdu_ent_ul {
35193+ struct au_rdu_ent __user *e;
9dbd164d 35194+ uint64_t ul;
7f207e10
AM
35195+};
35196+
35197+enum {
35198+ AufsCtlRduV_SZ,
35199+ AufsCtlRduV_End
35200+};
35201+
35202+struct aufs_rdu {
35203+ /* input */
35204+ union {
9dbd164d
AM
35205+ uint64_t sz; /* AuCtl_RDU */
35206+ uint64_t nent; /* AuCtl_RDU_INO */
7f207e10
AM
35207+ };
35208+ union au_rdu_ent_ul ent;
9dbd164d 35209+ uint16_t verify[AufsCtlRduV_End];
7f207e10
AM
35210+
35211+ /* input/output */
9dbd164d 35212+ uint32_t blk;
7f207e10
AM
35213+
35214+ /* output */
35215+ union au_rdu_ent_ul tail;
35216+ /* number of entries which were added in a single call */
9dbd164d
AM
35217+ uint64_t rent;
35218+ uint8_t full;
35219+ uint8_t shwh;
7f207e10
AM
35220+
35221+ struct au_rdu_cookie cookie;
35222+} __aligned(8);
35223+
1e00d052
AM
35224+/* ---------------------------------------------------------------------- */
35225+
35226+struct aufs_wbr_fd {
9dbd164d
AM
35227+ uint32_t oflags;
35228+ int16_t brid;
1e00d052
AM
35229+} __aligned(8);
35230+
35231+/* ---------------------------------------------------------------------- */
35232+
027c5e7a 35233+struct aufs_ibusy {
9dbd164d
AM
35234+ uint64_t ino, h_ino;
35235+ int16_t bindex;
027c5e7a
AM
35236+} __aligned(8);
35237+
1e00d052
AM
35238+/* ---------------------------------------------------------------------- */
35239+
392086de
AM
35240+/* error code for move-down */
35241+/* the actual message strings are implemented in aufs-util.git */
35242+enum {
35243+ EAU_MVDOWN_OPAQUE = 1,
35244+ EAU_MVDOWN_WHITEOUT,
35245+ EAU_MVDOWN_UPPER,
35246+ EAU_MVDOWN_BOTTOM,
35247+ EAU_MVDOWN_NOUPPER,
35248+ EAU_MVDOWN_NOLOWERBR,
35249+ EAU_Last
35250+};
35251+
c2b27bf2 35252+/* flags for move-down */
392086de
AM
35253+#define AUFS_MVDOWN_DMSG 1
35254+#define AUFS_MVDOWN_OWLOWER (1 << 1) /* overwrite lower */
35255+#define AUFS_MVDOWN_KUPPER (1 << 2) /* keep upper */
35256+#define AUFS_MVDOWN_ROLOWER (1 << 3) /* do even if lower is RO */
35257+#define AUFS_MVDOWN_ROLOWER_R (1 << 4) /* did on lower RO */
35258+#define AUFS_MVDOWN_ROUPPER (1 << 5) /* do even if upper is RO */
35259+#define AUFS_MVDOWN_ROUPPER_R (1 << 6) /* did on upper RO */
35260+#define AUFS_MVDOWN_BRID_UPPER (1 << 7) /* upper brid */
35261+#define AUFS_MVDOWN_BRID_LOWER (1 << 8) /* lower brid */
076b876e
AM
35262+#define AUFS_MVDOWN_FHSM_LOWER (1 << 9) /* find fhsm attr for lower */
35263+#define AUFS_MVDOWN_STFS (1 << 10) /* req. stfs */
35264+#define AUFS_MVDOWN_STFS_FAILED (1 << 11) /* output: stfs is unusable */
35265+#define AUFS_MVDOWN_BOTTOM (1 << 12) /* output: no more lowers */
c2b27bf2 35266+
076b876e 35267+/* index for move-down */
392086de
AM
35268+enum {
35269+ AUFS_MVDOWN_UPPER,
35270+ AUFS_MVDOWN_LOWER,
35271+ AUFS_MVDOWN_NARRAY
35272+};
35273+
076b876e
AM
35274+/*
35275+ * additional info of move-down
35276+ * number of free blocks and inodes.
35277+ * subset of struct kstatfs, but smaller and always 64bit.
35278+ */
35279+struct aufs_stfs {
35280+ uint64_t f_blocks;
35281+ uint64_t f_bavail;
35282+ uint64_t f_files;
35283+ uint64_t f_ffree;
35284+};
35285+
35286+struct aufs_stbr {
35287+ int16_t brid; /* optional input */
35288+ int16_t bindex; /* output */
35289+ struct aufs_stfs stfs; /* output when AUFS_MVDOWN_STFS set */
35290+} __aligned(8);
35291+
c2b27bf2 35292+struct aufs_mvdown {
076b876e
AM
35293+ uint32_t flags; /* input/output */
35294+ struct aufs_stbr stbr[AUFS_MVDOWN_NARRAY]; /* input/output */
35295+ int8_t au_errno; /* output */
35296+} __aligned(8);
35297+
35298+/* ---------------------------------------------------------------------- */
35299+
35300+union aufs_brinfo {
35301+ /* PATH_MAX may differ between kernel-space and user-space */
35302+ char _spacer[4096];
392086de 35303+ struct {
076b876e
AM
35304+ int16_t id;
35305+ int perm;
35306+ char path[0];
35307+ };
c2b27bf2
AM
35308+} __aligned(8);
35309+
35310+/* ---------------------------------------------------------------------- */
35311+
7f207e10
AM
35312+#define AuCtlType 'A'
35313+#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
35314+#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
1e00d052
AM
35315+#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
35316+ struct aufs_wbr_fd)
027c5e7a 35317+#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
392086de
AM
35318+#define AUFS_CTL_MVDOWN _IOWR(AuCtlType, AuCtl_MVDOWN, \
35319+ struct aufs_mvdown)
076b876e
AM
35320+#define AUFS_CTL_BRINFO _IOW(AuCtlType, AuCtl_BR, union aufs_brinfo)
35321+#define AUFS_CTL_FHSM_FD _IOW(AuCtlType, AuCtl_FHSM_FD, int)
7f207e10
AM
35322+
35323+#endif /* __AUFS_TYPE_H__ */
5527c038
JR
35324aufs4.x-rcN loopback patch
35325
35326diff --git a/drivers/block/loop.c b/drivers/block/loop.c
c2c0f25c 35327index 23103ad..5c3263f 100644
5527c038
JR
35328--- a/drivers/block/loop.c
35329+++ b/drivers/block/loop.c
c2c0f25c 35330@@ -417,7 +417,7 @@ static int do_req_filebacked(struct loop_device *lo, struct request *rq)
5527c038
JR
35331 }
35332
35333 struct switch_request {
35334- struct file *file;
35335+ struct file *file, *virt_file;
35336 struct completion wait;
35337 };
35338
c2c0f25c 35339@@ -437,6 +437,7 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
5527c038
JR
35340 mapping = file->f_mapping;
35341 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
35342 lo->lo_backing_file = file;
35343+ lo->lo_backing_virt_file = p->virt_file;
35344 lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ?
35345 mapping->host->i_bdev->bd_block_size : PAGE_SIZE;
35346 lo->old_gfp_mask = mapping_gfp_mask(mapping);
c2c0f25c 35347@@ -448,11 +449,13 @@ static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
5527c038
JR
35348 * First it needs to flush existing IO, it does this by sending a magic
35349 * BIO down the pipe. The completion of this BIO does the actual switch.
35350 */
35351-static int loop_switch(struct loop_device *lo, struct file *file)
35352+static int loop_switch(struct loop_device *lo, struct file *file,
35353+ struct file *virt_file)
35354 {
35355 struct switch_request w;
35356
35357 w.file = file;
35358+ w.virt_file = virt_file;
35359
35360 /* freeze queue and wait for completion of scheduled requests */
35361 blk_mq_freeze_queue(lo->lo_queue);
c2c0f25c 35362@@ -471,7 +474,16 @@ static int loop_switch(struct loop_device *lo, struct file *file)
5527c038
JR
35363 */
35364 static int loop_flush(struct loop_device *lo)
35365 {
35366- return loop_switch(lo, NULL);
35367+ return loop_switch(lo, NULL, NULL);
35368+}
35369+
35370+static struct file *loop_real_file(struct file *file)
35371+{
35372+ struct file *f = NULL;
35373+
35374+ if (file->f_path.dentry->d_sb->s_op->real_loop)
35375+ f = file->f_path.dentry->d_sb->s_op->real_loop(file);
35376+ return f;
35377 }
35378
c2c0f25c
AM
35379 static void loop_reread_partitions(struct loop_device *lo,
35380@@ -508,6 +520,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
5527c038
JR
35381 unsigned int arg)
35382 {
35383 struct file *file, *old_file;
35384+ struct file *f, *virt_file = NULL, *old_virt_file;
35385 struct inode *inode;
35386 int error;
35387
c2c0f25c 35388@@ -524,9 +537,16 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
5527c038
JR
35389 file = fget(arg);
35390 if (!file)
35391 goto out;
35392+ f = loop_real_file(file);
35393+ if (f) {
35394+ virt_file = file;
35395+ file = f;
35396+ get_file(file);
35397+ }
35398
35399 inode = file->f_mapping->host;
35400 old_file = lo->lo_backing_file;
35401+ old_virt_file = lo->lo_backing_virt_file;
35402
35403 error = -EINVAL;
35404
c2c0f25c 35405@@ -538,17 +558,21 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
5527c038
JR
35406 goto out_putf;
35407
35408 /* and ... switch */
35409- error = loop_switch(lo, file);
35410+ error = loop_switch(lo, file, virt_file);
35411 if (error)
35412 goto out_putf;
35413
35414 fput(old_file);
35415+ if (old_virt_file)
35416+ fput(old_virt_file);
35417 if (lo->lo_flags & LO_FLAGS_PARTSCAN)
c2c0f25c 35418 loop_reread_partitions(lo, bdev);
5527c038
JR
35419 return 0;
35420
35421 out_putf:
35422 fput(file);
35423+ if (virt_file)
35424+ fput(virt_file);
35425 out:
35426 return error;
35427 }
c2c0f25c 35428@@ -709,7 +733,7 @@ static void loop_config_discard(struct loop_device *lo)
5527c038
JR
35429 static int loop_set_fd(struct loop_device *lo, fmode_t mode,
35430 struct block_device *bdev, unsigned int arg)
35431 {
35432- struct file *file, *f;
35433+ struct file *file, *f, *virt_file = NULL;
35434 struct inode *inode;
35435 struct address_space *mapping;
35436 unsigned lo_blocksize;
c2c0f25c 35437@@ -724,6 +748,12 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
5527c038
JR
35438 file = fget(arg);
35439 if (!file)
35440 goto out;
35441+ f = loop_real_file(file);
35442+ if (f) {
35443+ virt_file = file;
35444+ file = f;
35445+ get_file(file);
35446+ }
35447
35448 error = -EBUSY;
35449 if (lo->lo_state != Lo_unbound)
c2c0f25c 35450@@ -778,6 +808,7 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
5527c038
JR
35451 lo->lo_device = bdev;
35452 lo->lo_flags = lo_flags;
35453 lo->lo_backing_file = file;
35454+ lo->lo_backing_virt_file = virt_file;
35455 lo->transfer = NULL;
35456 lo->ioctl = NULL;
35457 lo->lo_sizelimit = 0;
c2c0f25c 35458@@ -809,6 +840,8 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
5527c038
JR
35459
35460 out_putf:
35461 fput(file);
35462+ if (virt_file)
35463+ fput(virt_file);
35464 out:
35465 /* This is safe: open() is still holding a reference. */
35466 module_put(THIS_MODULE);
c2c0f25c 35467@@ -855,6 +888,7 @@ loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
5527c038
JR
35468 static int loop_clr_fd(struct loop_device *lo)
35469 {
35470 struct file *filp = lo->lo_backing_file;
35471+ struct file *virt_filp = lo->lo_backing_virt_file;
35472 gfp_t gfp = lo->old_gfp_mask;
35473 struct block_device *bdev = lo->lo_device;
35474
c2c0f25c 35475@@ -886,6 +920,7 @@ static int loop_clr_fd(struct loop_device *lo)
5527c038
JR
35476 spin_lock_irq(&lo->lo_lock);
35477 lo->lo_state = Lo_rundown;
35478 lo->lo_backing_file = NULL;
35479+ lo->lo_backing_virt_file = NULL;
35480 spin_unlock_irq(&lo->lo_lock);
35481
35482 loop_release_xfer(lo);
c2c0f25c 35483@@ -931,6 +966,8 @@ static int loop_clr_fd(struct loop_device *lo)
5527c038
JR
35484 * bd_mutex which is usually taken before lo_ctl_mutex.
35485 */
35486 fput(filp);
35487+ if (virt_filp)
35488+ fput(virt_filp);
35489 return 0;
35490 }
35491
35492diff --git a/drivers/block/loop.h b/drivers/block/loop.h
c2c0f25c 35493index 25e8997..93b6fce 100644
5527c038
JR
35494--- a/drivers/block/loop.h
35495+++ b/drivers/block/loop.h
35496@@ -46,7 +46,7 @@ struct loop_device {
35497 int (*ioctl)(struct loop_device *, int cmd,
35498 unsigned long arg);
35499
35500- struct file * lo_backing_file;
35501+ struct file * lo_backing_file, *lo_backing_virt_file;
35502 struct block_device *lo_device;
35503 unsigned lo_blocksize;
35504 void *key_data;
35505diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c
35506index 91c2ce7..d4ee5a7 100644
35507--- a/fs/aufs/f_op.c
35508+++ b/fs/aufs/f_op.c
35509@@ -389,7 +389,7 @@ static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
35510 if (IS_ERR(h_file))
35511 goto out;
35512
35513- if (au_test_loopback_kthread()) {
35514+ if (0 && au_test_loopback_kthread()) {
35515 au_warn_loopback(h_file->f_path.dentry->d_sb);
35516 if (file->f_mapping != h_file->f_mapping) {
35517 file->f_mapping = h_file->f_mapping;
35518diff --git a/fs/aufs/loop.c b/fs/aufs/loop.c
35519index 69f7e96..7941063 100644
35520--- a/fs/aufs/loop.c
35521+++ b/fs/aufs/loop.c
35522@@ -130,3 +130,19 @@ void au_loopback_fin(void)
35523 symbol_put(loop_backing_file);
35524 kfree(au_warn_loopback_array);
35525 }
35526+
35527+/* ---------------------------------------------------------------------- */
35528+
35529+/* support the loopback block device insude aufs */
35530+
35531+struct file *aufs_real_loop(struct file *file)
35532+{
35533+ struct file *f;
35534+
35535+ BUG_ON(!au_test_aufs(file->f_path.dentry->d_sb));
35536+ fi_read_lock(file);
35537+ f = au_hf_top(file);
35538+ fi_read_unlock(file);
35539+ AuDebugOn(!f);
35540+ return f;
35541+}
35542diff --git a/fs/aufs/loop.h b/fs/aufs/loop.h
35543index 6d9864d..3322557 100644
35544--- a/fs/aufs/loop.h
35545+++ b/fs/aufs/loop.h
35546@@ -25,7 +25,11 @@ void au_warn_loopback(struct super_block *h_sb);
35547
35548 int au_loopback_init(void);
35549 void au_loopback_fin(void);
35550+
35551+struct file *aufs_real_loop(struct file *file);
35552 #else
35553+AuStub(struct file *, loop_backing_file, return NULL)
35554+
35555 AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
35556 struct dentry *h_adding)
35557 AuStubInt0(au_test_loopback_kthread, void)
35558@@ -33,6 +37,8 @@ AuStubVoid(au_warn_loopback, struct super_block *h_sb)
35559
35560 AuStubInt0(au_loopback_init, void)
35561 AuStubVoid(au_loopback_fin, void)
35562+
35563+AuStub(struct file *, aufs_real_loop, return NULL, struct file *file)
35564 #endif /* BLK_DEV_LOOP */
35565
35566 #endif /* __KERNEL__ */
35567diff --git a/fs/aufs/super.c b/fs/aufs/super.c
35568index ee5780d..da35759 100644
35569--- a/fs/aufs/super.c
35570+++ b/fs/aufs/super.c
35571@@ -807,7 +807,10 @@ static const struct super_operations aufs_sop = {
35572 .statfs = aufs_statfs,
35573 .put_super = aufs_put_super,
35574 .sync_fs = aufs_sync_fs,
35575- .remount_fs = aufs_remount_fs
35576+ .remount_fs = aufs_remount_fs,
35577+#ifdef CONFIG_AUFS_BDEV_LOOP
35578+ .real_loop = aufs_real_loop
35579+#endif
35580 };
35581
35582 /* ---------------------------------------------------------------------- */
35583diff --git a/include/linux/fs.h b/include/linux/fs.h
c2c0f25c 35584index 86080ea..5f8e0f2 100644
5527c038
JR
35585--- a/include/linux/fs.h
35586+++ b/include/linux/fs.h
c2c0f25c 35587@@ -1708,6 +1708,10 @@ struct super_operations {
5527c038
JR
35588 struct shrink_control *);
35589 long (*free_cached_objects)(struct super_block *,
35590 struct shrink_control *);
35591+#if defined(CONFIG_BLK_DEV_LOOP) || defined(CONFIG_BLK_DEV_LOOP_MODULE)
35592+ /* and aufs */
35593+ struct file *(*real_loop)(struct file *);
35594+#endif
35595 };
35596
35597 /*
This page took 6.124075 seconds and 4 git commands to generate.