1 aufs3.x-rcN kbuild patch
3 diff --git a/fs/Kconfig b/fs/Kconfig
4 index 5f4c45d..357a8a6 100644
7 @@ -215,6 +215,7 @@ source "fs/pstore/Kconfig"
8 source "fs/sysv/Kconfig"
9 source "fs/ufs/Kconfig"
10 source "fs/exofs/Kconfig"
11 +source "fs/aufs/Kconfig"
13 endif # MISC_FILESYSTEMS
15 diff --git a/fs/Makefile b/fs/Makefile
16 index d2c3353..680ad8a 100644
19 @@ -123,3 +123,4 @@ obj-$(CONFIG_GFS2_FS) += gfs2/
20 obj-y += exofs/ # Multiple modules
21 obj-$(CONFIG_CEPH_FS) += ceph/
22 obj-$(CONFIG_PSTORE) += pstore/
23 +obj-$(CONFIG_AUFS_FS) += aufs/
24 diff --git a/include/linux/Kbuild b/include/linux/Kbuild
25 index 619b565..29f386b 100644
26 --- a/include/linux/Kbuild
27 +++ b/include/linux/Kbuild
28 @@ -65,6 +65,7 @@ header-y += atmppp.h
32 +header-y += aufs_type.h
34 header-y += auto_fs4.h
36 aufs3.x-rcN base patch
38 diff --git a/fs/namei.c b/fs/namei.c
39 index 5008f01..4cc94cf 100644
42 @@ -1753,7 +1753,7 @@ static struct dentry *__lookup_hash(struct qstr *name,
43 * needs parent already locked. Doesn't follow mounts.
46 -static struct dentry *lookup_hash(struct nameidata *nd)
47 +struct dentry *lookup_hash(struct nameidata *nd)
49 return __lookup_hash(&nd->last, nd->path.dentry, nd);
51 diff --git a/fs/splice.c b/fs/splice.c
52 index fa2defa..e3569b0 100644
55 @@ -1085,8 +1085,8 @@ EXPORT_SYMBOL(generic_splice_sendpage);
57 * Attempt to initiate a splice from pipe to file.
59 -static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
60 - loff_t *ppos, size_t len, unsigned int flags)
61 +long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
62 + loff_t *ppos, size_t len, unsigned int flags)
64 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *,
65 loff_t *, size_t, unsigned int);
66 @@ -1113,9 +1113,9 @@ static long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
68 * Attempt to initiate a splice from a file to a pipe.
70 -static long do_splice_to(struct file *in, loff_t *ppos,
71 - struct pipe_inode_info *pipe, size_t len,
73 +long do_splice_to(struct file *in, loff_t *ppos,
74 + struct pipe_inode_info *pipe, size_t len,
77 ssize_t (*splice_read)(struct file *, loff_t *,
78 struct pipe_inode_info *, size_t, unsigned int);
79 diff --git a/include/linux/namei.h b/include/linux/namei.h
80 index ffc0213..ef35a31 100644
81 --- a/include/linux/namei.h
82 +++ b/include/linux/namei.h
83 @@ -85,6 +85,7 @@ extern int vfs_path_lookup(struct dentry *, struct vfsmount *,
84 extern struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
85 int (*open)(struct inode *, struct file *));
87 +extern struct dentry *lookup_hash(struct nameidata *nd);
88 extern struct dentry *lookup_one_len(const char *, struct dentry *, int);
90 extern int follow_down_one(struct path *);
91 diff --git a/include/linux/splice.h b/include/linux/splice.h
92 index 26e5b61..3ffef2f 100644
93 --- a/include/linux/splice.h
94 +++ b/include/linux/splice.h
95 @@ -91,4 +91,10 @@ extern void splice_shrink_spd(struct pipe_inode_info *,
96 extern void spd_release_page(struct splice_pipe_desc *, unsigned int);
98 extern const struct pipe_buf_operations page_cache_pipe_buf_ops;
100 +extern long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
101 + loff_t *ppos, size_t len, unsigned int flags);
102 +extern long do_splice_to(struct file *in, loff_t *ppos,
103 + struct pipe_inode_info *pipe, size_t len,
104 + unsigned int flags);
106 aufs3.x-rcN standalone patch
108 diff --git a/fs/file_table.c b/fs/file_table.c
109 index c322794..2aad244 100644
110 --- a/fs/file_table.c
111 +++ b/fs/file_table.c
112 @@ -443,6 +443,8 @@ void file_sb_list_del(struct file *file)
116 +EXPORT_SYMBOL(file_sb_list_del);
121 diff --git a/fs/inode.c b/fs/inode.c
122 index ee4e66b..728042b 100644
125 @@ -65,6 +65,7 @@ static struct hlist_head *inode_hashtable __read_mostly;
126 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_hash_lock);
128 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
129 +EXPORT_SYMBOL(inode_sb_list_lock);
132 * Empty aops. Can be used for the cases where the user does not
133 diff --git a/fs/namei.c b/fs/namei.c
134 index 4cc94cf..af19e30 100644
137 @@ -1757,6 +1757,7 @@ struct dentry *lookup_hash(struct nameidata *nd)
139 return __lookup_hash(&nd->last, nd->path.dentry, nd);
141 +EXPORT_SYMBOL(lookup_hash);
144 * lookup_one_len - filesystem helper to lookup single pathname component
145 diff --git a/fs/namespace.c b/fs/namespace.c
146 index 6d3a196..d832654 100644
149 @@ -1509,6 +1509,7 @@ int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
153 +EXPORT_SYMBOL(iterate_mounts);
155 static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end)
157 diff --git a/fs/notify/group.c b/fs/notify/group.c
158 index 63fc294..6f4adca 100644
159 --- a/fs/notify/group.c
160 +++ b/fs/notify/group.c
162 #include <linux/srcu.h>
163 #include <linux/rculist.h>
164 #include <linux/wait.h>
165 +#include <linux/module.h>
167 #include <linux/fsnotify_backend.h>
168 #include "fsnotify.h"
169 @@ -70,6 +71,7 @@ void fsnotify_put_group(struct fsnotify_group *group)
170 if (atomic_dec_and_test(&group->refcnt))
171 fsnotify_destroy_group(group);
173 +EXPORT_SYMBOL(fsnotify_put_group);
176 * Create a new fsnotify_group and hold a reference for the group returned.
177 @@ -102,3 +104,4 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
181 +EXPORT_SYMBOL(fsnotify_alloc_group);
182 diff --git a/fs/notify/mark.c b/fs/notify/mark.c
183 index e14587d..be6533b 100644
184 --- a/fs/notify/mark.c
185 +++ b/fs/notify/mark.c
186 @@ -112,6 +112,7 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
187 if (atomic_dec_and_test(&mark->refcnt))
188 mark->free_mark(mark);
190 +EXPORT_SYMBOL(fsnotify_put_mark);
193 * Any time a mark is getting freed we end up here.
194 @@ -189,6 +190,7 @@ void fsnotify_destroy_mark(struct fsnotify_mark *mark)
195 if (unlikely(atomic_dec_and_test(&group->num_marks)))
196 fsnotify_final_destroy_group(group);
198 +EXPORT_SYMBOL(fsnotify_destroy_mark);
200 void fsnotify_set_mark_mask_locked(struct fsnotify_mark *mark, __u32 mask)
202 @@ -276,6 +278,7 @@ err:
206 +EXPORT_SYMBOL(fsnotify_add_mark);
209 * clear any marks in a group in which mark->flags & flags is true
210 @@ -331,6 +334,7 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
211 atomic_set(&mark->refcnt, 1);
212 mark->free_mark = free_mark;
214 +EXPORT_SYMBOL(fsnotify_init_mark);
216 static int fsnotify_mark_destroy(void *ignored)
218 diff --git a/fs/open.c b/fs/open.c
219 index 22c41b5..33b4033 100644
222 @@ -60,6 +60,7 @@ int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
223 mutex_unlock(&dentry->d_inode->i_mutex);
226 +EXPORT_SYMBOL(do_truncate);
228 static long do_sys_truncate(const char __user *pathname, loff_t length)
230 diff --git a/fs/splice.c b/fs/splice.c
231 index e3569b0..9dc07b7 100644
234 @@ -1109,6 +1109,7 @@ long do_splice_from(struct pipe_inode_info *pipe, struct file *out,
236 return splice_write(pipe, out, ppos, len, flags);
238 +EXPORT_SYMBOL(do_splice_from);
241 * Attempt to initiate a splice from a file to a pipe.
242 @@ -1135,6 +1136,7 @@ long do_splice_to(struct file *in, loff_t *ppos,
244 return splice_read(in, ppos, pipe, len, flags);
246 +EXPORT_SYMBOL(do_splice_to);
249 * splice_direct_to_actor - splices data directly between two non-pipes
250 diff --git a/security/commoncap.c b/security/commoncap.c
251 index ee4f848..611fd70 100644
252 --- a/security/commoncap.c
253 +++ b/security/commoncap.c
254 @@ -975,3 +975,4 @@ int cap_file_mmap(struct file *file, unsigned long reqprot,
258 +EXPORT_SYMBOL(cap_file_mmap);
259 diff --git a/security/device_cgroup.c b/security/device_cgroup.c
260 index 4450fbe..2c437e5 100644
261 --- a/security/device_cgroup.c
262 +++ b/security/device_cgroup.c
263 @@ -500,6 +500,7 @@ found:
267 +EXPORT_SYMBOL(__devcgroup_inode_permission);
269 int devcgroup_inode_mknod(int mode, dev_t dev)
271 diff --git a/security/security.c b/security/security.c
272 index 0c6cc69..d58ba1b 100644
273 --- a/security/security.c
274 +++ b/security/security.c
275 @@ -411,6 +411,7 @@ int security_path_rmdir(struct path *dir, struct dentry *dentry)
277 return security_ops->path_rmdir(dir, dentry);
279 +EXPORT_SYMBOL(security_path_rmdir);
281 int security_path_unlink(struct path *dir, struct dentry *dentry)
283 @@ -427,6 +428,7 @@ int security_path_symlink(struct path *dir, struct dentry *dentry,
285 return security_ops->path_symlink(dir, dentry, old_name);
287 +EXPORT_SYMBOL(security_path_symlink);
289 int security_path_link(struct dentry *old_dentry, struct path *new_dir,
290 struct dentry *new_dentry)
291 @@ -435,6 +437,7 @@ int security_path_link(struct dentry *old_dentry, struct path *new_dir,
293 return security_ops->path_link(old_dentry, new_dir, new_dentry);
295 +EXPORT_SYMBOL(security_path_link);
297 int security_path_rename(struct path *old_dir, struct dentry *old_dentry,
298 struct path *new_dir, struct dentry *new_dentry)
299 @@ -453,6 +456,7 @@ int security_path_truncate(struct path *path)
301 return security_ops->path_truncate(path);
303 +EXPORT_SYMBOL(security_path_truncate);
305 int security_path_chmod(struct dentry *dentry, struct vfsmount *mnt,
307 @@ -461,6 +465,7 @@ int security_path_chmod(struct dentry *dentry, struct vfsmount *mnt,
309 return security_ops->path_chmod(dentry, mnt, mode);
311 +EXPORT_SYMBOL(security_path_chmod);
313 int security_path_chown(struct path *path, uid_t uid, gid_t gid)
315 @@ -468,6 +473,7 @@ int security_path_chown(struct path *path, uid_t uid, gid_t gid)
317 return security_ops->path_chown(path, uid, gid);
319 +EXPORT_SYMBOL(security_path_chown);
321 int security_path_chroot(struct path *path)
323 @@ -544,6 +550,7 @@ int security_inode_readlink(struct dentry *dentry)
325 return security_ops->inode_readlink(dentry);
327 +EXPORT_SYMBOL(security_inode_readlink);
329 int security_inode_follow_link(struct dentry *dentry, struct nameidata *nd)
331 @@ -558,6 +565,7 @@ int security_inode_permission(struct inode *inode, int mask)
333 return security_ops->inode_permission(inode, mask);
335 +EXPORT_SYMBOL(security_inode_permission);
337 int security_inode_setattr(struct dentry *dentry, struct iattr *attr)
339 @@ -673,6 +681,7 @@ int security_file_permission(struct file *file, int mask)
341 return fsnotify_perm(file, mask);
343 +EXPORT_SYMBOL(security_file_permission);
345 int security_file_alloc(struct file *file)
347 @@ -700,6 +709,7 @@ int security_file_mmap(struct file *file, unsigned long reqprot,
349 return ima_file_mmap(file, prot);
351 +EXPORT_SYMBOL(security_file_mmap);
353 int security_file_mprotect(struct vm_area_struct *vma, unsigned long reqprot,
355 diff -urN /usr/share/empty/Documentation/ABI/testing/debugfs-aufs linux/Documentation/ABI/testing/debugfs-aufs
356 --- /usr/share/empty/Documentation/ABI/testing/debugfs-aufs 1970-01-01 01:00:00.000000000 +0100
357 +++ linux/Documentation/ABI/testing/debugfs-aufs 2011-08-24 13:30:24.727980364 +0200
359 +What: /debug/aufs/si_<id>/
361 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
363 + Under /debug/aufs, a directory named si_<id> is created
364 + per aufs mount, where <id> is a unique id generated
367 +What: /debug/aufs/si_<id>/xib
369 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
371 + It shows the consumed blocks by xib (External Inode Number
372 + Bitmap), its block size and file size.
373 + When the aufs mount option 'noxino' is specified, it
374 + will be empty. About XINO files, see the aufs manual.
376 +What: /debug/aufs/si_<id>/xino0, xino1 ... xinoN
378 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
380 + It shows the consumed blocks by xino (External Inode Number
381 + Translation Table), its link count, block size and file
383 + When the aufs mount option 'noxino' is specified, it
384 + will be empty. About XINO files, see the aufs manual.
386 +What: /debug/aufs/si_<id>/xigen
388 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
390 + It shows the consumed blocks by xigen (External Inode
391 + Generation Table), its block size and file size.
392 + If CONFIG_AUFS_EXPORT is disabled, this entry will not
394 + When the aufs mount option 'noxino' is specified, it
395 + will be empty. About XINO files, see the aufs manual.
396 diff -urN /usr/share/empty/Documentation/ABI/testing/sysfs-aufs linux/Documentation/ABI/testing/sysfs-aufs
397 --- /usr/share/empty/Documentation/ABI/testing/sysfs-aufs 1970-01-01 01:00:00.000000000 +0100
398 +++ linux/Documentation/ABI/testing/sysfs-aufs 2011-08-24 13:30:24.727980364 +0200
400 +What: /sys/fs/aufs/si_<id>/
402 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
404 + Under /sys/fs/aufs, a directory named si_<id> is created
405 + per aufs mount, where <id> is a unique id generated
408 +What: /sys/fs/aufs/si_<id>/br0, br1 ... brN
410 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
412 + It shows the abolute path of a member directory (which
413 + is called branch) in aufs, and its permission.
415 +What: /sys/fs/aufs/si_<id>/xi_path
417 +Contact: J. R. Okajima <hooanon05@yahoo.co.jp>
419 + It shows the abolute path of XINO (External Inode Number
420 + Bitmap, Translation Table and Generation Table) file
421 + even if it is the default path.
422 + When the aufs mount option 'noxino' is specified, it
423 + will be empty. About XINO files, see the aufs manual.
424 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt linux/Documentation/filesystems/aufs/design/01intro.txt
425 --- /usr/share/empty/Documentation/filesystems/aufs/design/01intro.txt 1970-01-01 01:00:00.000000000 +0100
426 +++ linux/Documentation/filesystems/aufs/design/01intro.txt 2011-08-24 13:30:24.727980364 +0200
429 +# Copyright (C) 2005-2011 Junjiro R. Okajima
431 +# This program is free software; you can redistribute it and/or modify
432 +# it under the terms of the GNU General Public License as published by
433 +# the Free Software Foundation; either version 2 of the License, or
434 +# (at your option) any later version.
436 +# This program is distributed in the hope that it will be useful,
437 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
438 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
439 +# GNU General Public License for more details.
441 +# You should have received a copy of the GNU General Public License
442 +# along with this program; if not, write to the Free Software
443 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
446 +----------------------------------------
448 +aufs [ei ju: ef es] | [a u f s]
449 +1. abbrev. for "advanced multi-layered unification filesystem".
450 +2. abbrev. for "another unionfs".
451 +3. abbrev. for "auf das" in German which means "on the" in English.
452 + Ex. "Butter aufs Brot"(G) means "butter onto bread"(E).
453 + But "Filesystem aufs Filesystem" is hard to understand.
455 +AUFS is a filesystem with features:
456 +- multi layered stackable unification filesystem, the member directory
457 + is called as a branch.
458 +- branch permission and attribute, 'readonly', 'real-readonly',
459 + 'readwrite', 'whiteout-able', 'link-able whiteout' and their
461 +- internal "file copy-on-write".
462 +- logical deletion, whiteout.
463 +- dynamic branch manipulation, adding, deleting and changing permission.
464 +- allow bypassing aufs, user's direct branch access.
465 +- external inode number translation table and bitmap which maintains the
466 + persistent aufs inode number.
467 +- seekable directory, including NFS readdir.
468 +- file mapping, mmap and sharing pages.
469 +- pseudo-link, hardlink over branches.
470 +- loopback mounted filesystem as a branch.
471 +- several policies to select one among multiple writable branches.
472 +- revert a single systemcall when an error occurs in aufs.
476 +Multi Layered Stackable Unification Filesystem
477 +----------------------------------------------------------------------
478 +Most people already knows what it is.
479 +It is a filesystem which unifies several directories and provides a
480 +merged single directory. When users access a file, the access will be
481 +passed/re-directed/converted (sorry, I am not sure which English word is
482 +correct) to the real file on the member filesystem. The member
483 +filesystem is called 'lower filesystem' or 'branch' and has a mode
484 +'readonly' and 'readwrite.' And the deletion for a file on the lower
485 +readonly branch is handled by creating 'whiteout' on the upper writable
488 +On LKML, there have been discussions about UnionMount (Jan Blunck,
489 +Bharata B Rao and Valerie Aurora) and Unionfs (Erez Zadok). They took
490 +different approaches to implement the merged-view.
491 +The former tries putting it into VFS, and the latter implements as a
492 +separate filesystem.
493 +(If I misunderstand about these implementations, please let me know and
494 +I shall correct it. Because it is a long time ago when I read their
495 +source files last time).
497 +UnionMount's approach will be able to small, but may be hard to share
498 +branches between several UnionMount since the whiteout in it is
499 +implemented in the inode on branch filesystem and always
500 +shared. According to Bharata's post, readdir does not seems to be
502 +There are several missing features known in this implementations such as
503 +- for users, the inode number may change silently. eg. copy-up.
504 +- link(2) may break by copy-up.
505 +- read(2) may get an obsoleted filedata (fstat(2) too).
506 +- fcntl(F_SETLK) may be broken by copy-up.
507 +- unnecessary copy-up may happen, for example mmap(MAP_PRIVATE) after
510 +Unionfs has a longer history. When I started implementing a stacking filesystem
511 +(Aug 2005), it already existed. It has virtual super_block, inode,
512 +dentry and file objects and they have an array pointing lower same kind
513 +objects. After contributing many patches for Unionfs, I re-started my
514 +project AUFS (Jun 2006).
516 +In AUFS, the structure of filesystem resembles to Unionfs, but I
517 +implemented my own ideas, approaches and enhancements and it became
518 +totally different one.
520 +Comparing DM snapshot and fs based implementation
521 +- the number of bytes to be copied between devices is much smaller.
522 +- the type of filesystem must be one and only.
523 +- the fs must be writable, no readonly fs, even for the lower original
524 + device. so the compression fs will not be usable. but if we use
525 + loopback mount, we may address this issue.
527 + mount /cdrom/squashfs.img /sq
528 + losetup /sq/ext2.img
529 + losetup /somewhere/cow
530 + dmsetup "snapshot /dev/loop0 /dev/loop1 ..."
531 +- it will be difficult (or needs more operations) to extract the
532 + difference between the original device and COW.
533 +- DM snapshot-merge may help a lot when users try merging. in the
534 + fs-layer union, users will use rsync(1).
537 +Several characters/aspects of aufs
538 +----------------------------------------------------------------------
540 +Aufs has several characters or aspects.
541 +1. a filesystem, callee of VFS helper
542 +2. sub-VFS, caller of VFS helper for branches
543 +3. a virtual filesystem which maintains persistent inode number
544 +4. reader/writer of files on branches such like an application
546 +1. Callee of VFS Helper
547 +As an ordinary linux filesystem, aufs is a callee of VFS. For instance,
548 +unlink(2) from an application reaches sys_unlink() kernel function and
549 +then vfs_unlink() is called. vfs_unlink() is one of VFS helper and it
550 +calls filesystem specific unlink operation. Actually aufs implements the
551 +unlink operation but it behaves like a redirector.
553 +2. Caller of VFS Helper for Branches
554 +aufs_unlink() passes the unlink request to the branch filesystem as if
555 +it were called from VFS. So the called unlink operation of the branch
556 +filesystem acts as usual. As a caller of VFS helper, aufs should handle
557 +every necessary pre/post operation for the branch filesystem.
558 +- acquire the lock for the parent dir on a branch
559 +- lookup in a branch
560 +- revalidate dentry on a branch
561 +- mnt_want_write() for a branch
562 +- vfs_unlink() for a branch
563 +- mnt_drop_write() for a branch
564 +- release the lock on a branch
566 +3. Persistent Inode Number
567 +One of the most important issue for a filesystem is to maintain inode
568 +numbers. This is particularly important to support exporting a
569 +filesystem via NFS. Aufs is a virtual filesystem which doesn't have a
570 +backend block device for its own. But some storage is necessary to
571 +maintain inode number. It may be a large space and may not suit to keep
572 +in memory. Aufs rents some space from its first writable branch
573 +filesystem (by default) and creates file(s) on it. These files are
574 +created by aufs internally and removed soon (currently) keeping opened.
575 +Note: Because these files are removed, they are totally gone after
576 + unmounting aufs. It means the inode numbers are not persistent
577 + across unmount or reboot. I have a plan to make them really
578 + persistent which will be important for aufs on NFS server.
580 +4. Read/Write Files Internally (copy-on-write)
581 +Because a branch can be readonly, when you write a file on it, aufs will
582 +"copy-up" it to the upper writable branch internally. And then write the
583 +originally requested thing to the file. Generally kernel doesn't
584 +open/read/write file actively. In aufs, even a single write may cause a
585 +internal "file copy". This behaviour is very similar to cp(1) command.
587 +Some people may think it is better to pass such work to user space
588 +helper, instead of doing in kernel space. Actually I am still thinking
589 +about it. But currently I have implemented it in kernel space.
590 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt linux/Documentation/filesystems/aufs/design/02struct.txt
591 --- /usr/share/empty/Documentation/filesystems/aufs/design/02struct.txt 1970-01-01 01:00:00.000000000 +0100
592 +++ linux/Documentation/filesystems/aufs/design/02struct.txt 2011-08-24 13:30:24.727980364 +0200
595 +# Copyright (C) 2005-2011 Junjiro R. Okajima
597 +# This program is free software; you can redistribute it and/or modify
598 +# it under the terms of the GNU General Public License as published by
599 +# the Free Software Foundation; either version 2 of the License, or
600 +# (at your option) any later version.
602 +# This program is distributed in the hope that it will be useful,
603 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
604 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
605 +# GNU General Public License for more details.
607 +# You should have received a copy of the GNU General Public License
608 +# along with this program; if not, write to the Free Software
609 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
611 +Basic Aufs Internal Structure
613 +Superblock/Inode/Dentry/File Objects
614 +----------------------------------------------------------------------
615 +As like an ordinary filesystem, aufs has its own
616 +superblock/inode/dentry/file objects. All these objects have a
617 +dynamically allocated array and store the same kind of pointers to the
618 +lower filesystem, branch.
619 +For example, when you build a union with one readwrite branch and one
620 +readonly, mounted /au, /rw and /ro respectively.
622 +- /ro/fileA exists but /rw/fileA
624 +Aufs lookup operation finds /ro/fileA and gets dentry for that. These
625 +pointers are stored in a aufs dentry. The array in aufs dentry will be,
629 +This style of an array is essentially same to the aufs
630 +superblock/inode/dentry/file objects.
632 +Because aufs supports manipulating branches, ie. add/delete/change
633 +dynamically, these objects has its own generation. When branches are
634 +changed, the generation in aufs superblock is incremented. And a
635 +generation in other object are compared when it is accessed.
636 +When a generation in other objects are obsoleted, aufs refreshes the
641 +----------------------------------------------------------------------
642 +Additionally aufs superblock has some data for policies to select one
643 +among multiple writable branches, XIB files, pseudo-links and kobject.
644 +See below in detail.
645 +About the policies which supports copy-down a directory, see policy.txt
649 +Branch and XINO(External Inode Number Translation Table)
650 +----------------------------------------------------------------------
651 +Every branch has its own xino (external inode number translation table)
652 +file. The xino file is created and unlinked by aufs internally. When two
653 +members of a union exist on the same filesystem, they share the single
655 +The struct of a xino file is simple, just a sequence of aufs inode
656 +numbers which is indexed by the lower inode number.
657 +In the above sample, assume the inode number of /ro/fileA is i111 and
658 +aufs assigns the inode number i999 for fileA. Then aufs writes 999 as
659 +4(8) bytes at 111 * 4(8) bytes offset in the xino file.
661 +When the inode numbers are not contiguous, the xino file will be sparse
662 +which has a hole in it and doesn't consume as much disk space as it
663 +might appear. If your branch filesystem consumes disk space for such
664 +holes, then you should specify 'xino=' option at mounting aufs.
666 +Also a writable branch has three kinds of "whiteout bases". All these
667 +are existed when the branch is joined to aufs and the names are
668 +whiteout-ed doubly, so that users will never see their names in aufs
670 +1. a regular file which will be linked to all whiteouts.
671 +2. a directory to store a pseudo-link.
672 +3. a directory to store an "orphan-ed" file temporary.
675 + When you remove a file on a readonly branch, aufs handles it as a
676 + logical deletion and creates a whiteout on the upper writable branch
677 + as a hardlink of this file in order not to consume inode on the
680 + See below, Pseudo-link.
682 + When "fileC" exists on the lower readonly branch only and it is
683 + opened and removed with its parent dir, and then user writes
684 + something into it, then aufs copies-up fileC to this
685 + directory. Because there is no other dir to store fileC. After
686 + creating a file under this dir, the file is unlinked.
688 +Because aufs supports manipulating branches, ie. add/delete/change
689 +dynamically, a branch has its own id. When the branch order changes, aufs
690 +finds the new index by searching the branch id.
694 +----------------------------------------------------------------------
695 +Assume "fileA" exists on the lower readonly branch only and it is
696 +hardlinked to "fileB" on the branch. When you write something to fileA,
697 +aufs copies-up it to the upper writable branch. Additionally aufs
698 +creates a hardlink under the Pseudo-link Directory of the writable
699 +branch. The inode of a pseudo-link is kept in aufs super_block as a
700 +simple list. If fileB is read after unlinking fileA, aufs returns
701 +filedata from the pseudo-link instead of the lower readonly
702 +branch. Because the pseudo-link is based upon the inode, to keep the
703 +inode number by xino (see above) is important.
705 +All the hardlinks under the Pseudo-link Directory of the writable branch
706 +should be restored in a proper location later. Aufs provides a utility
707 +to do this. The userspace helpers executed at remounting and unmounting
709 +During this utility is running, it puts aufs into the pseudo-link
710 +maintenance mode. In this mode, only the process which began the
711 +maintenance mode (and its child processes) is allowed to operate in
712 +aufs. Some other processes which are not related to the pseudo-link will
713 +be allowed to run too, but the rest have to return an error or wait
714 +until the maintenance mode ends. If a process already acquires an inode
715 +mutex (in VFS), it has to return an error.
718 +XIB(external inode number bitmap)
719 +----------------------------------------------------------------------
720 +Addition to the xino file per a branch, aufs has an external inode number
721 +bitmap in a superblock object. It is also a file such like a xino file.
722 +It is a simple bitmap to mark whether the aufs inode number is in-use or
724 +To reduce the file I/O, aufs prepares a single memory page to cache xib.
726 +Aufs implements a feature to truncate/refresh both of xino and xib to
727 +reduce the number of consumed disk blocks for these files.
730 +Virtual or Vertical Dir, and Readdir in Userspace
731 +----------------------------------------------------------------------
732 +In order to support multiple layers (branches), aufs readdir operation
733 +constructs a virtual dir block on memory. For readdir, aufs calls
734 +vfs_readdir() internally for each dir on branches, merges their entries
735 +with eliminating the whiteout-ed ones, and sets it to file (dir)
736 +object. So the file object has its entry list until it is closed. The
737 +entry list will be updated when the file position is zero and becomes
738 +old. This decision is made in aufs automatically.
740 +The dynamically allocated memory block for the name of entries has a
741 +unit of 512 bytes (by default) and stores the names contiguously (no
742 +padding). Another block for each entry is handled by kmem_cache too.
743 +During building dir blocks, aufs creates hash list and judging whether
744 +the entry is whiteouted by its upper branch or already listed.
745 +The merged result is cached in the corresponding inode object and
746 +maintained by a customizable life-time option.
748 +Some people may call it can be a security hole or invite DoS attack
749 +since the opened and once readdir-ed dir (file object) holds its entry
750 +list and becomes a pressure for system memory. But I'd say it is similar
751 +to files under /proc or /sys. The virtual files in them also holds a
752 +memory page (generally) while they are opened. When an idea to reduce
753 +memory for them is introduced, it will be applied to aufs too.
754 +For those who really hate this situation, I've developed readdir(3)
755 +library which operates this merging in userspace. You just need to set
756 +LD_PRELOAD environment variable, and aufs will not consume no memory in
757 +kernel space for readdir(3).
761 +----------------------------------------------------------------------
762 +Aufs sometimes requires privilege access to a branch. For instance,
763 +in copy-up/down operation. When a user process is going to make changes
764 +to a file which exists in the lower readonly branch only, and the mode
765 +of one of ancestor directories may not be writable by a user
766 +process. Here aufs copy-up the file with its ancestors and they may
767 +require privilege to set its owner/group/mode/etc.
768 +This is a typical case of a application character of aufs (see
771 +Aufs uses workqueue synchronously for this case. It creates its own
772 +workqueue. The workqueue is a kernel thread and has privilege. Aufs
773 +passes the request to call mkdir or write (for example), and wait for
774 +its completion. This approach solves a problem of a signal handler
776 +If aufs didn't adopt the workqueue and changed the privilege of the
777 +process, and if the mkdir/write call arises SIGXFSZ or other signal,
778 +then the user process might gain a privilege or the generated core file
779 +was owned by a superuser.
781 +Also aufs uses the system global workqueue ("events" kernel thread) too
782 +for asynchronous tasks, such like handling inotify/fsnotify, re-creating a
783 +whiteout base and etc. This is unrelated to a privilege.
784 +Most of aufs operation tries acquiring a rw_semaphore for aufs
785 +superblock at the beginning, at the same time waits for the completion
786 +of all queued asynchronous tasks.
790 +----------------------------------------------------------------------
791 +The whiteout in aufs is very similar to Unionfs's. That is represented
792 +by its filename. UnionMount takes an approach of a file mode, but I am
793 +afraid several utilities (find(1) or something) will have to support it.
795 +Basically the whiteout represents "logical deletion" which stops aufs to
796 +lookup further, but also it represents "dir is opaque" which also stop
799 +In aufs, rmdir(2) and rename(2) for dir uses whiteout alternatively.
800 +In order to make several functions in a single systemcall to be
801 +revertible, aufs adopts an approach to rename a directory to a temporary
802 +unique whiteouted name.
803 +For example, in rename(2) dir where the target dir already existed, aufs
804 +renames the target dir to a temporary unique whiteouted name before the
805 +actual rename on a branch and then handles other actions (make it opaque,
806 +update the attributes, etc). If an error happens in these actions, aufs
807 +simply renames the whiteouted name back and returns an error. If all are
808 +succeeded, aufs registers a function to remove the whiteouted unique
809 +temporary name completely and asynchronously to the system global
814 +----------------------------------------------------------------------
815 +It is a well-known feature or concept.
816 +When user modifies a file on a readonly branch, aufs operate "copy-up"
817 +internally and makes change to the new file on the upper writable branch.
818 +When the trigger systemcall does not update the timestamps of the parent
819 +dir, aufs reverts it after copy-up.
820 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt linux/Documentation/filesystems/aufs/design/03lookup.txt
821 --- /usr/share/empty/Documentation/filesystems/aufs/design/03lookup.txt 1970-01-01 01:00:00.000000000 +0100
822 +++ linux/Documentation/filesystems/aufs/design/03lookup.txt 2011-08-24 13:30:24.727980364 +0200
825 +# Copyright (C) 2005-2011 Junjiro R. Okajima
827 +# This program is free software; you can redistribute it and/or modify
828 +# it under the terms of the GNU General Public License as published by
829 +# the Free Software Foundation; either version 2 of the License, or
830 +# (at your option) any later version.
832 +# This program is distributed in the hope that it will be useful,
833 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
834 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
835 +# GNU General Public License for more details.
837 +# You should have received a copy of the GNU General Public License
838 +# along with this program; if not, write to the Free Software
839 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
842 +----------------------------------------------------------------------
843 +Since aufs has a character of sub-VFS (see Introduction), it operates
844 +lookup for branches as VFS does. It may be a heavy work. Generally
845 +speaking struct nameidata is a bigger structure and includes many
846 +information. But almost all lookup operation in aufs is the simplest
847 +case, ie. lookup only an entry directly connected to its parent. Digging
848 +down the directory hierarchy is unnecessary.
850 +VFS has a function lookup_one_len() for that use, but it is not usable
851 +for a branch filesystem which requires struct nameidata. So aufs
852 +implements a simple lookup wrapper function. When a branch filesystem
853 +allows NULL as nameidata, it calls lookup_one_len(). Otherwise it builds
854 +a simplest nameidata and calls lookup_hash().
855 +Here aufs applies "a principle in NFSD", ie. if the filesystem supports
856 +NFS-export, then it has to support NULL as a nameidata parameter for
857 +->create(), ->lookup() and ->d_revalidate(). So the lookup wrapper in
858 +aufs tests if ->s_export_op in the branch is NULL or not.
860 +When a branch is a remote filesystem, aufs basically trusts its
861 +->d_revalidate(), also aufs forces the hardest revalidate tests for
863 +For d_revalidate, aufs implements three levels of revalidate tests. See
864 +"Revalidate Dentry and UDBA" in detail.
868 +----------------------------------------------------------------------
869 +Basically aufs supports any type of filesystem and block device for a
870 +branch (actually there are some exceptions). But it is prohibited to add
871 +a loopback mounted one whose backend file exists in a filesystem which is
872 +already added to aufs. The reason is to protect aufs from a recursive
873 +lookup. If it was allowed, the aufs lookup operation might re-enter a
874 +lookup for the loopback mounted branch in the same context, and will
878 +Revalidate Dentry and UDBA (User's Direct Branch Access)
879 +----------------------------------------------------------------------
880 +Generally VFS helpers re-validate a dentry as a part of lookup.
881 +0. digging down the directory hierarchy.
882 +1. lock the parent dir by its i_mutex.
883 +2. lookup the final (child) entry.
885 +4. call the actual operation (create, unlink, etc.)
886 +5. unlock the parent dir
888 +If the filesystem implements its ->d_revalidate() (step 3), then it is
889 +called. Actually aufs implements it and checks the dentry on a branch is
891 +But it is not enough. Because aufs has to release the lock for the
892 +parent dir on a branch at the end of ->lookup() (step 2) and
893 +->d_revalidate() (step 3) while the i_mutex of the aufs dir is still
895 +If the file on a branch is changed directly, eg. bypassing aufs, after
896 +aufs released the lock, then the subsequent operation may cause
897 +something unpleasant result.
899 +This situation is a result of VFS architecture, ->lookup() and
900 +->d_revalidate() is separated. But I never say it is wrong. It is a good
901 +design from VFS's point of view. It is just not suitable for sub-VFS
904 +Aufs supports such case by three level of revalidation which is
906 +1. Simple Revalidate
907 + Addition to the native flow in VFS's, confirm the child-parent
908 + relationship on the branch just after locking the parent dir on the
909 + branch in the "actual operation" (step 4). When this validation
910 + fails, aufs returns EBUSY. ->d_revalidate() (step 3) in aufs still
911 + checks the validation of the dentry on branches.
912 +2. Monitor Changes Internally by Inotify/Fsnotify
913 + Addition to above, in the "actual operation" (step 4) aufs re-lookup
914 + the dentry on the branch, and returns EBUSY if it finds different
916 + Additionally, aufs sets the inotify/fsnotify watch for every dir on branches
917 + during it is in cache. When the event is notified, aufs registers a
918 + function to kernel 'events' thread by schedule_work(). And the
919 + function sets some special status to the cached aufs dentry and inode
920 + private data. If they are not cached, then aufs has nothing to
921 + do. When the same file is accessed through aufs (step 0-3) later,
922 + aufs will detect the status and refresh all necessary data.
923 + In this mode, aufs has to ignore the event which is fired by aufs
925 +3. No Extra Validation
926 + This is the simplest test and doesn't add any additional revalidation
927 + test, and skip therevalidatin in step 4. It is useful and improves
928 + aufs performance when system surely hide the aufs branches from user,
929 + by over-mounting something (or another method).
930 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt linux/Documentation/filesystems/aufs/design/04branch.txt
931 --- /usr/share/empty/Documentation/filesystems/aufs/design/04branch.txt 1970-01-01 01:00:00.000000000 +0100
932 +++ linux/Documentation/filesystems/aufs/design/04branch.txt 2011-08-24 13:30:24.727980364 +0200
935 +# Copyright (C) 2005-2011 Junjiro R. Okajima
937 +# This program is free software; you can redistribute it and/or modify
938 +# it under the terms of the GNU General Public License as published by
939 +# the Free Software Foundation; either version 2 of the License, or
940 +# (at your option) any later version.
942 +# This program is distributed in the hope that it will be useful,
943 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
944 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
945 +# GNU General Public License for more details.
947 +# You should have received a copy of the GNU General Public License
948 +# along with this program; if not, write to the Free Software
949 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
953 +Since aufs supports dynamic branch manipulation, ie. add/remove a branch
954 +and changing its permission/attribute, there are a lot of works to do.
958 +----------------------------------------------------------------------
959 +o Confirm the adding dir exists outside of aufs, including loopback
961 +- and other various attributes...
962 +o Initialize the xino file and whiteout bases if necessary.
965 +o Check the owner/group/mode of the directory
966 + When the owner/group/mode of the adding directory differs from the
967 + existing branch, aufs issues a warning because it may impose a
969 + For example, when a upper writable branch has a world writable empty
970 + top directory, a malicious user can create any files on the writable
971 + branch directly, like copy-up and modify manually. If something like
972 + /etc/{passwd,shadow} exists on the lower readonly branch but the upper
973 + writable branch, and the writable branch is world-writable, then a
974 + malicious guy may create /etc/passwd on the writable branch directly
975 + and the infected file will be valid in aufs.
976 + I am afraid it can be a security issue, but nothing to do except
977 + producing a warning.
981 +----------------------------------------------------------------------
982 +o Confirm the deleting branch is not busy
983 + To be general, there is one merit to adopt "remount" interface to
984 + manipulate branches. It is to discard caches. At deleting a branch,
985 + aufs checks the still cached (and connected) dentries and inodes. If
986 + there are any, then they are all in-use. An inode without its
987 + corresponding dentry can be alive alone (for example, inotify/fsnotify case).
989 + For the cached one, aufs checks whether the same named entry exists on
991 + If the cached one is a directory, because aufs provides a merged view
992 + to users, as long as one dir is left on any branch aufs can show the
993 + dir to users. In this case, the branch can be removed from aufs.
994 + Otherwise aufs rejects deleting the branch.
996 + If any file on the deleting branch is opened by aufs, then aufs
1000 +Modify the Permission of a Branch
1001 +----------------------------------------------------------------------
1002 +o Re-initialize or remove the xino file and whiteout bases if necessary.
1005 +o rw --> ro: Confirm the modifying branch is not busy
1006 + Aufs rejects the request if any of these conditions are true.
1007 + - a file on the branch is mmap-ed.
1008 + - a regular file on the branch is opened for write and there is no
1009 + same named entry on the upper branch.
1010 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt linux/Documentation/filesystems/aufs/design/05wbr_policy.txt
1011 --- /usr/share/empty/Documentation/filesystems/aufs/design/05wbr_policy.txt 1970-01-01 01:00:00.000000000 +0100
1012 +++ linux/Documentation/filesystems/aufs/design/05wbr_policy.txt 2011-08-24 13:30:24.727980364 +0200
1015 +# Copyright (C) 2005-2011 Junjiro R. Okajima
1017 +# This program is free software; you can redistribute it and/or modify
1018 +# it under the terms of the GNU General Public License as published by
1019 +# the Free Software Foundation; either version 2 of the License, or
1020 +# (at your option) any later version.
1022 +# This program is distributed in the hope that it will be useful,
1023 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1024 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1025 +# GNU General Public License for more details.
1027 +# You should have received a copy of the GNU General Public License
1028 +# along with this program; if not, write to the Free Software
1029 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1031 +Policies to Select One among Multiple Writable Branches
1032 +----------------------------------------------------------------------
1033 +When the number of writable branch is more than one, aufs has to decide
1034 +the target branch for file creation or copy-up. By default, the highest
1035 +writable branch which has the parent (or ancestor) dir of the target
1036 +file is chosen (top-down-parent policy).
1037 +By user's request, aufs implements some other policies to select the
1038 +writable branch, for file creation two policies, round-robin and
1039 +most-free-space policies. For copy-up three policies, top-down-parent,
1040 +bottom-up-parent and bottom-up policies.
1042 +As expected, the round-robin policy selects the branch in circular. When
1043 +you have two writable branches and creates 10 new files, 5 files will be
1044 +created for each branch. mkdir(2) systemcall is an exception. When you
1045 +create 10 new directories, all will be created on the same branch.
1046 +And the most-free-space policy selects the one which has most free
1047 +space among the writable branches. The amount of free space will be
1048 +checked by aufs internally, and users can specify its time interval.
1050 +The policies for copy-up is more simple,
1051 +top-down-parent is equivalent to the same named on in create policy,
1052 +bottom-up-parent selects the writable branch where the parent dir
1053 +exists and the nearest upper one from the copyup-source,
1054 +bottom-up selects the nearest upper writable branch from the
1055 +copyup-source, regardless the existence of the parent dir.
1057 +There are some rules or exceptions to apply these policies.
1058 +- If there is a readonly branch above the policy-selected branch and
1059 + the parent dir is marked as opaque (a variation of whiteout), or the
1060 + target (creating) file is whiteout-ed on the upper readonly branch,
1061 + then the result of the policy is ignored and the target file will be
1062 + created on the nearest upper writable branch than the readonly branch.
1063 +- If there is a writable branch above the policy-selected branch and
1064 + the parent dir is marked as opaque or the target file is whiteouted
1065 + on the branch, then the result of the policy is ignored and the target
1066 + file will be created on the highest one among the upper writable
1067 + branches who has diropq or whiteout. In case of whiteout, aufs removes
1069 +- link(2) and rename(2) systemcalls are exceptions in every policy.
1070 + They try selecting the branch where the source exists as possible
1071 + since copyup a large file will take long time. If it can't be,
1072 + ie. the branch where the source exists is readonly, then they will
1073 + follow the copyup policy.
1074 +- There is an exception for rename(2) when the target exists.
1075 + If the rename target exists, aufs compares the index of the branches
1076 + where the source and the target exists and selects the higher
1077 + one. If the selected branch is readonly, then aufs follows the
1079 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt linux/Documentation/filesystems/aufs/design/06mmap.txt
1080 --- /usr/share/empty/Documentation/filesystems/aufs/design/06mmap.txt 1970-01-01 01:00:00.000000000 +0100
1081 +++ linux/Documentation/filesystems/aufs/design/06mmap.txt 2011-08-24 13:30:24.727980364 +0200
1084 +# Copyright (C) 2005-2011 Junjiro R. Okajima
1086 +# This program is free software; you can redistribute it and/or modify
1087 +# it under the terms of the GNU General Public License as published by
1088 +# the Free Software Foundation; either version 2 of the License, or
1089 +# (at your option) any later version.
1091 +# This program is distributed in the hope that it will be useful,
1092 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1093 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1094 +# GNU General Public License for more details.
1096 +# You should have received a copy of the GNU General Public License
1097 +# along with this program; if not, write to the Free Software
1098 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1100 +mmap(2) -- File Memory Mapping
1101 +----------------------------------------------------------------------
1102 +In aufs, the file-mapped pages are handled by a branch fs directly, no
1103 +interaction with aufs. It means aufs_mmap() calls the branch fs's
1105 +This approach is simple and good, but there is one problem.
1106 +Under /proc, several entries show the mmap-ped files by its path (with
1107 +device and inode number), and the printed path will be the path on the
1108 +branch fs's instead of virtual aufs's.
1109 +This is not a problem in most cases, but some utilities lsof(1) (and its
1110 +user) may expect the path on aufs.
1112 +To address this issue, aufs adds a new member called vm_prfile in struct
1113 +vm_area_struct (and struct vm_region). The original vm_file points to
1114 +the file on the branch fs in order to handle everything correctly as
1115 +usual. The new vm_prfile points to a virtual file in aufs, and the
1116 +show-functions in procfs refers to vm_prfile if it is set.
1117 +Also we need to maintain several other places where touching vm_file
1119 +- fork()/clone() copies vma and the reference count of vm_file is
1121 +- merging vma maintains the ref count too.
1123 +This is not a good approach. It just faking the printed path. But it
1124 +leaves all behaviour around f_mapping unchanged. This is surely an
1126 +Actually aufs had adopted another complicated approach which calls
1127 +generic_file_mmap() and handles struct vm_operations_struct. In this
1128 +approach, aufs met a hard problem and I could not solve it without
1129 +switching the approach.
1130 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt linux/Documentation/filesystems/aufs/design/07export.txt
1131 --- /usr/share/empty/Documentation/filesystems/aufs/design/07export.txt 1970-01-01 01:00:00.000000000 +0100
1132 +++ linux/Documentation/filesystems/aufs/design/07export.txt 2011-08-24 13:30:24.727980364 +0200
1135 +# Copyright (C) 2005-2011 Junjiro R. Okajima
1137 +# This program is free software; you can redistribute it and/or modify
1138 +# it under the terms of the GNU General Public License as published by
1139 +# the Free Software Foundation; either version 2 of the License, or
1140 +# (at your option) any later version.
1142 +# This program is distributed in the hope that it will be useful,
1143 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1144 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1145 +# GNU General Public License for more details.
1147 +# You should have received a copy of the GNU General Public License
1148 +# along with this program; if not, write to the Free Software
1149 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1151 +Export Aufs via NFS
1152 +----------------------------------------------------------------------
1153 +Here is an approach.
1154 +- like xino/xib, add a new file 'xigen' which stores aufs inode
1156 +- iget_locked(): initialize aufs inode generation for a new inode, and
1157 + store it in xigen file.
1158 +- destroy_inode(): increment aufs inode generation and store it in xigen
1159 + file. it is necessary even if it is not unlinked, because any data of
1160 + inode may be changed by UDBA.
1161 +- encode_fh(): for a root dir, simply return FILEID_ROOT. otherwise
1162 + build file handle by
1163 + + branch id (4 bytes)
1164 + + superblock generation (4 bytes)
1165 + + inode number (4 or 8 bytes)
1166 + + parent dir inode number (4 or 8 bytes)
1167 + + inode generation (4 bytes))
1168 + + return value of exportfs_encode_fh() for the parent on a branch (4
1170 + + file handle for a branch (by exportfs_encode_fh())
1172 + + find the index of a branch from its id in handle, and check it is
1173 + still exist in aufs.
1174 + + 1st level: get the inode number from handle and search it in cache.
1175 + + 2nd level: if not found, get the parent inode number from handle and
1176 + search it in cache. and then open the parent dir, find the matching
1177 + inode number by vfs_readdir() and get its name, and call
1178 + lookup_one_len() for the target dentry.
1179 + + 3rd level: if the parent dir is not cached, call
1180 + exportfs_decode_fh() for a branch and get the parent on a branch,
1181 + build a pathname of it, convert it a pathname in aufs, call
1182 + path_lookup(). now aufs gets a parent dir dentry, then handle it as
1184 + + to open the dir, aufs needs struct vfsmount. aufs keeps vfsmount
1185 + for every branch, but not itself. to get this, (currently) aufs
1186 + searches in current->nsproxy->mnt_ns list. it may not be a good
1187 + idea, but I didn't get other approach.
1188 + + test the generation of the gotten inode.
1189 +- every inode operation: they may get EBUSY due to UDBA. in this case,
1190 + convert it into ESTALE for NFSD.
1191 +- readdir(): call lockdep_on/off() because filldir in NFSD calls
1192 + lookup_one_len(), vfs_getattr(), encode_fh() and others.
1193 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt linux/Documentation/filesystems/aufs/design/08shwh.txt
1194 --- /usr/share/empty/Documentation/filesystems/aufs/design/08shwh.txt 1970-01-01 01:00:00.000000000 +0100
1195 +++ linux/Documentation/filesystems/aufs/design/08shwh.txt 2011-08-24 13:30:24.727980364 +0200
1198 +# Copyright (C) 2005-2011 Junjiro R. Okajima
1200 +# This program is free software; you can redistribute it and/or modify
1201 +# it under the terms of the GNU General Public License as published by
1202 +# the Free Software Foundation; either version 2 of the License, or
1203 +# (at your option) any later version.
1205 +# This program is distributed in the hope that it will be useful,
1206 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1207 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1208 +# GNU General Public License for more details.
1210 +# You should have received a copy of the GNU General Public License
1211 +# along with this program; if not, write to the Free Software
1212 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1214 +Show Whiteout Mode (shwh)
1215 +----------------------------------------------------------------------
1216 +Generally aufs hides the name of whiteouts. But in some cases, to show
1217 +them is very useful for users. For instance, creating a new middle layer
1218 +(branch) by merging existing layers.
1220 +(borrowing aufs1 HOW-TO from a user, Michael Towers)
1221 +When you have three branches,
1222 +- Bottom: 'system', squashfs (underlying base system), read-only
1223 +- Middle: 'mods', squashfs, read-only
1224 +- Top: 'overlay', ram (tmpfs), read-write
1226 +The top layer is loaded at boot time and saved at shutdown, to preserve
1227 +the changes made to the system during the session.
1228 +When larger changes have been made, or smaller changes have accumulated,
1229 +the size of the saved top layer data grows. At this point, it would be
1230 +nice to be able to merge the two overlay branches ('mods' and 'overlay')
1231 +and rewrite the 'mods' squashfs, clearing the top layer and thus
1232 +restoring save and load speed.
1234 +This merging is simplified by the use of another aufs mount, of just the
1235 +two overlay branches using the 'shwh' option.
1236 +# mount -t aufs -o ro,shwh,br:/livesys/overlay=ro+wh:/livesys/mods=rr+wh \
1237 + aufs /livesys/merge_union
1239 +A merged view of these two branches is then available at
1240 +/livesys/merge_union, and the new feature is that the whiteouts are
1242 +Note that in 'shwh' mode the aufs mount must be 'ro', which will disable
1243 +writing to all branches. Also the default mode for all branches is 'ro'.
1244 +It is now possible to save the combined contents of the two overlay
1245 +branches to a new squashfs, e.g.:
1246 +# mksquashfs /livesys/merge_union /path/to/newmods.squash
1248 +This new squashfs archive can be stored on the boot device and the
1249 +initramfs will use it to replace the old one at the next boot.
1250 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt linux/Documentation/filesystems/aufs/design/10dynop.txt
1251 --- /usr/share/empty/Documentation/filesystems/aufs/design/10dynop.txt 1970-01-01 01:00:00.000000000 +0100
1252 +++ linux/Documentation/filesystems/aufs/design/10dynop.txt 2011-08-24 13:30:24.727980364 +0200
1255 +# Copyright (C) 2010-2011 Junjiro R. Okajima
1257 +# This program is free software; you can redistribute it and/or modify
1258 +# it under the terms of the GNU General Public License as published by
1259 +# the Free Software Foundation; either version 2 of the License, or
1260 +# (at your option) any later version.
1262 +# This program is distributed in the hope that it will be useful,
1263 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1264 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1265 +# GNU General Public License for more details.
1267 +# You should have received a copy of the GNU General Public License
1268 +# along with this program; if not, write to the Free Software
1269 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1271 +Dynamically customizable FS operations
1272 +----------------------------------------------------------------------
1273 +Generally FS operations (struct inode_operations, struct
1274 +address_space_operations, struct file_operations, etc.) are defined as
1275 +"static const", but it never means that FS have only one set of
1276 +operation. Some FS have multiple sets of them. For instance, ext2 has
1277 +three sets, one for XIP, for NOBH, and for normal.
1278 +Since aufs overrides and redirects these operations, sometimes aufs has
1279 +to change its behaviour according to the branch FS type. More imporantly
1280 +VFS acts differently if a function (member in the struct) is set or
1281 +not. It means aufs should have several sets of operations and select one
1282 +among them according to the branch FS definition.
1284 +In order to solve this problem and not to affect the behavour of VFS,
1285 +aufs defines these operations dynamically. For instance, aufs defines
1286 +aio_read function for struct file_operations, but it may not be set to
1287 +the file_operations. When the branch FS doesn't have it, aufs doesn't
1288 +set it to its file_operations while the function definition itself is
1289 +still alive. So the behaviour of io_submit(2) will not change, and it
1290 +will return an error when aio_read is not defined.
1292 +The lifetime of these dynamically generated operation object is
1293 +maintained by aufs branch object. When the branch is removed from aufs,
1294 +the reference counter of the object is decremented. When it reaches
1295 +zero, the dynamically generated operation object will be freed.
1297 +This approach is designed to support AIO (io_submit), Direcit I/O and
1299 +Currently this approach is applied to file_operations and
1300 +vm_operations_struct for regular files only.
1301 diff -urN /usr/share/empty/Documentation/filesystems/aufs/design/99plan.txt linux/Documentation/filesystems/aufs/design/99plan.txt
1302 --- /usr/share/empty/Documentation/filesystems/aufs/design/99plan.txt 1970-01-01 01:00:00.000000000 +0100
1303 +++ linux/Documentation/filesystems/aufs/design/99plan.txt 2011-08-24 13:30:24.727980364 +0200
1306 +# Copyright (C) 2005-2011 Junjiro R. Okajima
1308 +# This program is free software; you can redistribute it and/or modify
1309 +# it under the terms of the GNU General Public License as published by
1310 +# the Free Software Foundation; either version 2 of the License, or
1311 +# (at your option) any later version.
1313 +# This program is distributed in the hope that it will be useful,
1314 +# but WITHOUT ANY WARRANTY; without even the implied warranty of
1315 +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1316 +# GNU General Public License for more details.
1318 +# You should have received a copy of the GNU General Public License
1319 +# along with this program; if not, write to the Free Software
1320 +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1324 +Restoring some features which was implemented in aufs1.
1325 +They were dropped in aufs2 in order to make source files simpler and
1326 +easier to be reviewed.
1329 +Test Only the Highest One for the Directory Permission (dirperm1 option)
1330 +----------------------------------------------------------------------
1331 +Let's try case study.
1332 +- aufs has two branches, upper readwrite and lower readonly.
1334 +- "dirA" exists under /ro, but /rw. and its mode is 0700.
1335 +- user invoked "chmod a+rx /au/dirA"
1336 +- then "dirA" becomes world readable?
1338 +In this case, /ro/dirA is still 0700 since it exists in readonly branch,
1339 +or it may be a natively readonly filesystem. If aufs respects the lower
1340 +branch, it should not respond readdir request from other users. But user
1341 +allowed it by chmod. Should really aufs rejects showing the entries
1344 +To be honest, I don't have a best solution for this case. So I
1345 +implemented 'dirperm1' and 'nodirperm1' option in aufs1, and leave it to
1347 +When dirperm1 is specified, aufs checks only the highest one for the
1348 +directory permission, and shows the entries. Otherwise, as usual, checks
1349 +every dir existing on all branches and rejects the request.
1351 +As a side effect, dirperm1 option improves the performance of aufs
1352 +because the number of permission check is reduced.
1355 +Being Another Aufs's Readonly Branch (robr)
1356 +----------------------------------------------------------------------
1357 +Aufs1 allows aufs to be another aufs's readonly branch.
1358 +This feature was developed by a user's request. But it may not be used
1362 +Copy-up on Open (coo=)
1363 +----------------------------------------------------------------------
1364 +By default the internal copy-up is executed when it is really necessary.
1365 +It is not done when a file is opened for writing, but when write(2) is
1366 +done. Users who have many (over 100) branches want to know and analyse
1367 +when and what file is copied-up. To insert a new upper branch which
1368 +contains such files only may improve the performance of aufs.
1370 +Aufs1 implemented "coo=none | leaf | all" option.
1373 +Refresh the Opened File (refrof)
1374 +----------------------------------------------------------------------
1375 +This option is implemented in aufs1 but incomplete.
1377 +When user reads from a file, he expects to get its latest filedata
1378 +generally. If the file is removed and a new same named file is created,
1379 +the content he gets is unchanged, ie. the unlinked filedata.
1381 +Let's try case study again.
1382 +- aufs has two branches.
1384 +- "fileA" exists under /ro, but /rw.
1385 +- user opened "/au/fileA".
1386 +- he or someone else inserts a branch (/new) between /rw and /ro.
1387 + /au = /rw + /new + /ro
1388 +- the new branch has "fileA".
1389 +- user reads from the opened "fileA"
1390 +- which filedata should aufs return, from /ro or /new?
1392 +Some people says it has to be "from /ro" and it is a semantics of Unix.
1393 +The others say it should be "from /new" because the file is not removed
1394 +and it is equivalent to the case of someone else modifies the file.
1396 +Here again I don't have a best and final answer. I got an idea to
1397 +implement 'refrof' and 'norefrof' option. When 'refrof' (REFResh the
1398 +Opened File) is specified (by default), aufs returns the filedata from
1400 +Otherwise from /new.
1401 diff -urN /usr/share/empty/Documentation/filesystems/aufs/README linux/Documentation/filesystems/aufs/README
1402 --- /usr/share/empty/Documentation/filesystems/aufs/README 1970-01-01 01:00:00.000000000 +0100
1403 +++ linux/Documentation/filesystems/aufs/README 2012-01-05 11:00:59.215501779 +0100
1406 +Aufs3 -- advanced multi layered unification filesystem version 3.x
1412 +----------------------------------------
1413 +In the early days, aufs was entirely re-designed and re-implemented
1414 +Unionfs Version 1.x series. After many original ideas, approaches,
1415 +improvements and implementations, it becomes totally different from
1416 +Unionfs while keeping the basic features.
1417 +Recently, Unionfs Version 2.x series begin taking some of the same
1418 +approaches to aufs1's.
1419 +Unionfs is being developed by Professor Erez Zadok at Stony Brook
1420 +University and his team.
1422 +Aufs3 supports linux-3.0 and later.
1423 +If you want older kernel version support, try aufs2-2.6.git or
1424 +aufs2-standalone.git repository, aufs1 from CVS on SourceForge.
1426 +Note: it becomes clear that "Aufs was rejected. Let's give it up."
1427 +According to Christoph Hellwig, linux rejects all union-type filesystems
1429 +<http://marc.info/?l=linux-kernel&m=123938533724484&w=2>
1433 +----------------------------------------
1434 +- unite several directories into a single virtual filesystem. The member
1435 + directory is called as a branch.
1436 +- you can specify the permission flags to the branch, which are 'readonly',
1437 + 'readwrite' and 'whiteout-able.'
1438 +- by upper writable branch, internal copyup and whiteout, files/dirs on
1439 + readonly branch are modifiable logically.
1440 +- dynamic branch manipulation, add, del.
1443 +Also there are many enhancements in aufs1, such as:
1444 +- readdir(3) in userspace.
1445 +- keep inode number by external inode number table
1446 +- keep the timestamps of file/dir in internal copyup operation
1447 +- seekable directory, supporting NFS readdir.
1448 +- whiteout is hardlinked in order to reduce the consumption of inodes
1450 +- do not copyup, nor create a whiteout when it is unnecessary
1451 +- revert a single systemcall when an error occurs in aufs
1452 +- remount interface instead of ioctl
1453 +- maintain /etc/mtab by an external command, /sbin/mount.aufs.
1454 +- loopback mounted filesystem as a branch
1455 +- kernel thread for removing the dir who has a plenty of whiteouts
1456 +- support copyup sparse file (a file which has a 'hole' in it)
1457 +- default permission flags for branches
1458 +- selectable permission flags for ro branch, whether whiteout can
1461 +- support <sysfs>/fs/aufs and <debugfs>/aufs.
1462 +- support multiple writable branches, some policies to select one
1463 + among multiple writable branches.
1464 +- a new semantics for link(2) and rename(2) to support multiple
1465 + writable branches.
1466 +- no glibc changes are required.
1467 +- pseudo hardlink (hardlink over branches)
1468 +- allow a direct access manually to a file on branch, e.g. bypassing aufs.
1469 + including NFS or remote filesystem branch.
1470 +- userspace wrapper for pathconf(3)/fpathconf(3) with _PC_LINK_MAX.
1473 +Currently these features are dropped temporary from aufs3.
1474 +See design/08plan.txt in detail.
1475 +- test only the highest one for the directory permission (dirperm1)
1476 +- copyup on open (coo=)
1477 +- nested mount, i.e. aufs as readonly no-whiteout branch of another aufs
1479 +- statistics of aufs thread (/sys/fs/aufs/stat)
1480 +- delegation mode (dlgt)
1481 + a delegation of the internal branch access to support task I/O
1482 + accounting, which also supports Linux Security Modules (LSM) mainly
1483 + for Suse AppArmor.
1484 +- intent.open/create (file open in a single lookup)
1486 +Features or just an idea in the future (see also design/*.txt),
1487 +- reorder the branch index without del/re-add.
1488 +- permanent xino files for NFSD
1489 +- an option for refreshing the opened files after add/del branches
1490 +- 'move' policy for copy-up between two writable branches, after
1491 + checking free space.
1492 +- light version, without branch manipulation. (unnecessary?)
1493 +- copyup in userspace
1494 +- inotify in userspace
1500 +----------------------------------------
1501 +There were three GIT trees for aufs3, aufs3-linux.git,
1502 +aufs3-standalone.git, and aufs-util.git. Note that there is no "3" in
1504 +While the aufs-util is always necessary, you need either of aufs3-linux
1505 +or aufs3-standalone.
1507 +The aufs3-linux tree includes the whole linux mainline GIT tree,
1508 +git://git.kernel.org/.../torvalds/linux.git.
1509 +And you cannot select CONFIG_AUFS_FS=m for this version, eg. you cannot
1510 +build aufs3 as an externel kernel module.
1512 +On the other hand, the aufs3-standalone tree has only aufs source files
1513 +and necessary patches, and you can select CONFIG_AUFS_FS=m.
1515 +You will find GIT branches whose name is in form of "aufs3.x" where "x"
1516 +represents the linux kernel version, "linux-3.x". For instance,
1517 +"aufs3.0" is for linux-3.0. For latest "linux-3.x-rcN", use
1518 +"aufs3.x-rcN" branch.
1521 +$ git clone --reference /your/linux/git/tree \
1522 + git://aufs.git.sourceforge.net/gitroot/aufs/aufs3-linux.git \
1524 +- if you don't have linux GIT tree, then remove "--reference ..."
1525 +$ cd aufs3-linux.git
1526 +$ git checkout origin/aufs3.0
1528 +o aufs3-standalone tree
1529 +$ git clone git://aufs.git.sourceforge.net/gitroot/aufs/aufs3-standalone.git \
1530 + aufs3-standalone.git
1531 +$ cd aufs3-standalone.git
1532 +$ git checkout origin/aufs3.0
1535 +$ git clone git://aufs.git.sourceforge.net/gitroot/aufs/aufs-util.git \
1538 +$ git checkout origin/aufs3.0
1540 +Note: The 3.x-rcN branch is to be used with `rc' kernel versions ONLY.
1541 +The minor version number, 'x' in '3.x', of aufs may not always
1542 +follow the minor version number of the kernel.
1543 +Because changes in the kernel that cause the use of a new
1544 +minor version number do not always require changes to aufs-util.
1546 +Since aufs-util has its own minor version number, you may not be
1547 +able to find a GIT branch in aufs-util for your kernel's
1548 +exact minor version number.
1549 +In this case, you should git-checkout the branch for the
1550 +nearest lower number.
1552 +For (an unreleased) example:
1553 +If you are using "linux-3.10" and the "aufs3.10" branch
1554 +does not exit in aufs-util repository, then "aufs3.9", "aufs3.8"
1555 +or something numerically smaller is the branch for your kernel.
1557 +Also you can view all branches by
1561 +3. Configuration and Compilation
1562 +----------------------------------------
1563 +Make sure you have git-checkout'ed the correct branch.
1565 +For aufs3-linux tree,
1566 +- enable CONFIG_EXPERIMENTAL and CONFIG_AUFS_FS.
1567 +- set other aufs configurations if necessary.
1569 +For aufs3-standalone tree,
1570 +There are several ways to build.
1573 +- apply ./aufs3-kbuild.patch to your kernel source files.
1574 +- apply ./aufs3-base.patch too.
1575 +- apply ./aufs3-proc_map.patch too, if you want to make /proc/PID/maps (and
1576 + others including lsof(1)) show the file path on aufs instead of the
1577 + path on the branch fs.
1578 +- apply ./aufs3-standalone.patch too, if you have a plan to set
1579 + CONFIG_AUFS_FS=m. otherwise you don't need ./aufs3-standalone.patch.
1580 +- copy ./{Documentation,fs,include/linux/aufs_type.h} files to your
1581 + kernel source tree. Never copy ./include/linux/Kbuild.
1582 +- enable CONFIG_EXPERIMENTAL and CONFIG_AUFS_FS, you can select either
1584 +- and build your kernel as usual.
1585 +- install the built kernel.
1586 +- install the header files too by "make headers_install".
1587 +- and reboot your system.
1590 +- module only (CONFIG_AUFS_FS=m).
1591 +- apply ./aufs3-base.patch to your kernel source files.
1592 +- apply ./aufs3-proc_map.patch too to your kernel source files,
1593 + if you want to make /proc/PID/maps (and others including lsof(1)) show
1594 + the file path on aufs instead of the path on the branch fs.
1595 +- apply ./aufs3-standalone.patch too.
1596 +- build your kernel, don't forget "make headers_install", and reboot.
1597 +- edit ./config.mk and set other aufs configurations if necessary.
1598 + Note: You should read ./fs/aufs/Kconfig carefully which describes
1599 + every aufs configurations.
1600 +- build the module by simple "make".
1601 +- you can specify ${KDIR} make variable which points to your kernel
1603 +- install the files
1604 + + run "make install" to install the aufs module, or copy the built
1605 + ./aufs.ko to /lib/modules/... and run depmod -a (or reboot simply).
1606 + + run "make headers_install" to install the aufs header file (you can
1607 + specify DESTDIR), or copty ./usr/include/linux/aufs_type.h to
1608 + /usr/include/linux or wherever you like.
1609 +- no need to apply aufs3-kbuild.patch, nor copying source files to your
1610 + kernel source tree.
1612 +Note: The haeder file aufs_type.h is necessary to build aufs-util
1613 + as well as "make headers_install" in the kernel source tree.
1614 + headers_install is subject to be forgotten, but it is essentially
1615 + necessary, not only for building aufs-util.
1616 + You may not meet problems without headers_install in some older
1620 +- read README in aufs-util, build and install it
1621 +- note that your distribution may contain an obsoleted version of
1622 + aufs_type.h in /usr/include/linux or something. When you build aufs
1623 + utilities, make sure that your compiler refers the correct aufs header
1624 + file which is built by "make headers_install."
1625 +- if you want to use readdir(3) in userspace or pathconf(3) wrapper,
1626 + then run "make install_ulib" too. And refer to the aufs manual in
1631 +----------------------------------------
1632 +At first, make sure aufs-util are installed, and please read the aufs
1633 +manual, aufs.5 in aufs-util.git tree.
1637 +$ mkdir /tmp/rw /tmp/aufs
1638 +# mount -t aufs -o br=/tmp/rw:${HOME} none /tmp/aufs
1640 +Here is another example. The result is equivalent.
1641 +# mount -t aufs -o br=/tmp/rw=rw:${HOME}=ro none /tmp/aufs
1643 +# mount -t aufs -o br:/tmp/rw none /tmp/aufs
1644 +# mount -o remount,append:${HOME} /tmp/aufs
1646 +Then, you can see whole tree of your home dir through /tmp/aufs. If
1647 +you modify a file under /tmp/aufs, the one on your home directory is
1648 +not affected, instead the same named file will be newly created under
1649 +/tmp/rw. And all of your modification to a file will be applied to
1650 +the one under /tmp/rw. This is called the file based Copy on Write
1652 +Aufs mount options are described in aufs.5.
1653 +If you run chroot or something and make your aufs as a root directory,
1654 +then you need to customize the shutdown script. See the aufs manual in
1657 +Additionally, there are some sample usages of aufs which are a
1658 +diskless system with network booting, and LiveCD over NFS.
1659 +See sample dir in CVS tree on SourceForge.
1663 +----------------------------------------
1664 +When you have any problems or strange behaviour in aufs, please let me
1666 +- /proc/mounts (instead of the output of mount(8))
1667 +- /sys/module/aufs/*
1668 +- /sys/fs/aufs/* (if you have them)
1669 +- /debug/aufs/* (if you have them)
1670 +- linux kernel version
1671 + if your kernel is not plain, for example modified by distributor,
1672 + the url where i can download its source is necessary too.
1673 +- aufs version which was printed at loading the module or booting the
1674 + system, instead of the date you downloaded.
1675 +- configuration (define/undefine CONFIG_AUFS_xxx)
1676 +- kernel configuration or /proc/config.gz (if you have it)
1677 +- behaviour which you think to be incorrect
1678 +- actual operation, reproducible one is better
1679 +- mailto: aufs-users at lists.sourceforge.net
1681 +Usually, I don't watch the Public Areas(Bugs, Support Requests, Patches,
1682 +and Feature Requests) on SourceForge. Please join and write to
1686 +6. Acknowledgements
1687 +----------------------------------------
1688 +Thanks to everyone who have tried and are using aufs, whoever
1689 +have reported a bug or any feedback.
1691 +Especially donators:
1692 +Tomas Matejicek(slax.org) made a donation (much more than once).
1693 + Since Apr 2010, Tomas M (the author of Slax and Linux Live
1694 + scripts) is making "doubling" donations.
1695 + Unfortunately I cannot list all of the donators, but I really
1697 + It ends Aug 2010, but the ordinary donation URL is still available.
1698 + <http://sourceforge.net/donate/index.php?group_id=167503>
1699 +Dai Itasaka made a donation (2007/8).
1700 +Chuck Smith made a donation (2008/4, 10 and 12).
1701 +Henk Schoneveld made a donation (2008/9).
1702 +Chih-Wei Huang, ASUS, CTC donated Eee PC 4G (2008/10).
1703 +Francois Dupoux made a donation (2008/11).
1704 +Bruno Cesar Ribas and Luis Carlos Erpen de Bona, C3SL serves public
1705 + aufs2 GIT tree (2009/2).
1706 +William Grant made a donation (2009/3).
1707 +Patrick Lane made a donation (2009/4).
1708 +The Mail Archive (mail-archive.com) made donations (2009/5).
1709 +Nippy Networks (Ed Wildgoose) made a donation (2009/7).
1710 +New Dream Network, LLC (www.dreamhost.com) made a donation (2009/11).
1711 +Pavel Pronskiy made a donation (2011/2).
1712 +Iridium and Inmarsat satellite phone retailer (www.mailasail.com), Nippy
1713 + Networks (Ed Wildgoose) made a donation for hardware (2011/3).
1714 +Max Lekomcev (DOM-TV project) made a donation (2011/7).
1715 +Sam Liddicott made a donation (2011/9).
1717 +Thank you very much.
1718 +Donations are always, including future donations, very important and
1719 +helpful for me to keep on developing aufs.
1723 +----------------------------------------
1724 +If you are an experienced user, no explanation is needed. Aufs is
1725 +just a linux filesystem.
1730 +# Local variables: ;
1733 diff -urN /usr/share/empty/fs/aufs/aufs.h linux/fs/aufs/aufs.h
1734 --- /usr/share/empty/fs/aufs/aufs.h 1970-01-01 01:00:00.000000000 +0100
1735 +++ linux/fs/aufs/aufs.h 2011-08-24 13:30:24.731313534 +0200
1738 + * Copyright (C) 2005-2011 Junjiro R. Okajima
1740 + * This program, aufs is free software; you can redistribute it and/or modify
1741 + * it under the terms of the GNU General Public License as published by
1742 + * the Free Software Foundation; either version 2 of the License, or
1743 + * (at your option) any later version.
1745 + * This program is distributed in the hope that it will be useful,
1746 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
1747 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1748 + * GNU General Public License for more details.
1750 + * You should have received a copy of the GNU General Public License
1751 + * along with this program; if not, write to the Free Software
1752 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1756 + * all header files
1764 +#define AuStub(type, name, body, ...) \
1765 + static inline type name(__VA_ARGS__) { body; }
1767 +#define AuStubVoid(name, ...) \
1768 + AuStub(void, name, , __VA_ARGS__)
1769 +#define AuStubInt0(name, ...) \
1770 + AuStub(int, name, return 0, __VA_ARGS__)
1774 +#include "branch.h"
1777 +#include "dbgaufs.h"
1778 +#include "dentry.h"
1782 +#include "fstype.h"
1785 +#include "module.h"
1790 +#include "sysaufs.h"
1795 +#endif /* __KERNEL__ */
1796 +#endif /* __AUFS_H__ */
1797 diff -urN /usr/share/empty/fs/aufs/branch.c linux/fs/aufs/branch.c
1798 --- /usr/share/empty/fs/aufs/branch.c 1970-01-01 01:00:00.000000000 +0100
1799 +++ linux/fs/aufs/branch.c 2011-08-24 13:30:24.731313534 +0200
1802 + * Copyright (C) 2005-2011 Junjiro R. Okajima
1804 + * This program, aufs is free software; you can redistribute it and/or modify
1805 + * it under the terms of the GNU General Public License as published by
1806 + * the Free Software Foundation; either version 2 of the License, or
1807 + * (at your option) any later version.
1809 + * This program is distributed in the hope that it will be useful,
1810 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
1811 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1812 + * GNU General Public License for more details.
1814 + * You should have received a copy of the GNU General Public License
1815 + * along with this program; if not, write to the Free Software
1816 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1820 + * branch management
1823 +#include <linux/compat.h>
1824 +#include <linux/file.h>
1825 +#include <linux/statfs.h>
1829 + * free a single branch
1831 +static void au_br_do_free(struct au_branch *br)
1834 + struct au_wbr *wbr;
1835 + struct au_dykey **key;
1837 + au_hnotify_fin_br(br);
1839 + if (br->br_xino.xi_file)
1840 + fput(br->br_xino.xi_file);
1841 + mutex_destroy(&br->br_xino.xi_nondir_mtx);
1843 + AuDebugOn(atomic_read(&br->br_count));
1847 + for (i = 0; i < AuBrWh_Last; i++)
1848 + dput(wbr->wbr_wh[i]);
1849 + AuDebugOn(atomic_read(&wbr->wbr_wh_running));
1850 + AuRwDestroy(&wbr->wbr_wh_rwsem);
1853 + key = br->br_dykey;
1854 + for (i = 0; i < AuBrDynOp; i++, key++)
1860 + mntput(br->br_mnt);
1866 + * frees all branches
1868 +void au_br_free(struct au_sbinfo *sbinfo)
1870 + aufs_bindex_t bmax;
1871 + struct au_branch **br;
1873 + AuRwMustWriteLock(&sbinfo->si_rwsem);
1875 + bmax = sbinfo->si_bend + 1;
1876 + br = sbinfo->si_branch;
1878 + au_br_do_free(*br++);
1882 + * find the index of a branch which is specified by @br_id.
1884 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id)
1886 + aufs_bindex_t bindex, bend;
1888 + bend = au_sbend(sb);
1889 + for (bindex = 0; bindex <= bend; bindex++)
1890 + if (au_sbr_id(sb, bindex) == br_id)
1895 +/* ---------------------------------------------------------------------- */
1901 +static int test_overlap(struct super_block *sb, struct dentry *h_adding,
1902 + struct dentry *h_root)
1904 + if (unlikely(h_adding == h_root
1905 + || au_test_loopback_overlap(sb, h_adding)))
1907 + if (h_adding->d_sb != h_root->d_sb)
1909 + return au_test_subdir(h_adding, h_root)
1910 + || au_test_subdir(h_root, h_adding);
1914 + * returns a newly allocated branch. @new_nbranch is a number of branches
1915 + * after adding a branch.
1917 +static struct au_branch *au_br_alloc(struct super_block *sb, int new_nbranch,
1920 + struct au_branch *add_branch;
1921 + struct dentry *root;
1925 + root = sb->s_root;
1926 + add_branch = kmalloc(sizeof(*add_branch), GFP_NOFS);
1927 + if (unlikely(!add_branch))
1930 + err = au_hnotify_init_br(add_branch, perm);
1931 + if (unlikely(err))
1934 + add_branch->br_wbr = NULL;
1935 + if (au_br_writable(perm)) {
1936 + /* may be freed separately at changing the branch permission */
1937 + add_branch->br_wbr = kmalloc(sizeof(*add_branch->br_wbr),
1939 + if (unlikely(!add_branch->br_wbr))
1943 + err = au_sbr_realloc(au_sbi(sb), new_nbranch);
1945 + err = au_di_realloc(au_di(root), new_nbranch);
1947 + err = au_ii_realloc(au_ii(root->d_inode), new_nbranch);
1949 + return add_branch; /* success */
1951 + kfree(add_branch->br_wbr);
1954 + au_hnotify_fin_br(add_branch);
1956 + kfree(add_branch);
1958 + return ERR_PTR(err);
1962 + * test if the branch permission is legal or not.
1964 +static int test_br(struct inode *inode, int brperm, char *path)
1968 + err = (au_br_writable(brperm) && IS_RDONLY(inode));
1973 + pr_err("write permission for readonly mount or inode, %s\n", path);
1981 + * 0: success, the caller will add it
1982 + * plus: success, it is already unified, the caller should ignore it
1985 +static int test_add(struct super_block *sb, struct au_opt_add *add, int remount)
1988 + aufs_bindex_t bend, bindex;
1989 + struct dentry *root;
1990 + struct inode *inode, *h_inode;
1992 + root = sb->s_root;
1993 + bend = au_sbend(sb);
1994 + if (unlikely(bend >= 0
1995 + && au_find_dbindex(root, add->path.dentry) >= 0)) {
1999 + pr_err("%s duplicated\n", add->pathname);
2004 + err = -ENOSPC; /* -E2BIG; */
2005 + if (unlikely(AUFS_BRANCH_MAX <= add->bindex
2006 + || AUFS_BRANCH_MAX - 1 <= bend)) {
2007 + pr_err("number of branches exceeded %s\n", add->pathname);
2012 + if (unlikely(add->bindex < 0 || bend + 1 < add->bindex)) {
2013 + pr_err("bad index %d\n", add->bindex);
2017 + inode = add->path.dentry->d_inode;
2019 + if (unlikely(!inode->i_nlink)) {
2020 + pr_err("no existence %s\n", add->pathname);
2025 + if (unlikely(inode->i_sb == sb)) {
2026 + pr_err("%s must be outside\n", add->pathname);
2030 + if (unlikely(au_test_fs_unsuppoted(inode->i_sb))) {
2031 + pr_err("unsupported filesystem, %s (%s)\n",
2032 + add->pathname, au_sbtype(inode->i_sb));
2036 + err = test_br(add->path.dentry->d_inode, add->perm, add->pathname);
2037 + if (unlikely(err))
2041 + return 0; /* success */
2044 + for (bindex = 0; bindex <= bend; bindex++)
2045 + if (unlikely(test_overlap(sb, add->path.dentry,
2046 + au_h_dptr(root, bindex)))) {
2047 + pr_err("%s is overlapped\n", add->pathname);
2052 + if (au_opt_test(au_mntflags(sb), WARN_PERM)) {
2053 + h_inode = au_h_dptr(root, 0)->d_inode;
2054 + if ((h_inode->i_mode & S_IALLUGO) != (inode->i_mode & S_IALLUGO)
2055 + || h_inode->i_uid != inode->i_uid
2056 + || h_inode->i_gid != inode->i_gid)
2057 + pr_warning("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n",
2059 + inode->i_uid, inode->i_gid,
2060 + (inode->i_mode & S_IALLUGO),
2061 + h_inode->i_uid, h_inode->i_gid,
2062 + (h_inode->i_mode & S_IALLUGO));
2070 + * initialize or clean the whiteouts for an adding branch
2072 +static int au_br_init_wh(struct super_block *sb, struct au_branch *br,
2073 + int new_perm, struct dentry *h_root)
2075 + int err, old_perm;
2076 + aufs_bindex_t bindex;
2077 + struct mutex *h_mtx;
2078 + struct au_wbr *wbr;
2079 + struct au_hinode *hdir;
2082 + old_perm = br->br_perm;
2083 + br->br_perm = new_perm;
2086 + bindex = au_br_index(sb, br->br_id);
2087 + if (0 <= bindex) {
2088 + hdir = au_hi(sb->s_root->d_inode, bindex);
2089 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
2091 + h_mtx = &h_root->d_inode->i_mutex;
2092 + mutex_lock_nested(h_mtx, AuLsc_I_PARENT);
2095 + err = au_wh_init(h_root, br, sb);
2097 + wbr_wh_write_lock(wbr);
2098 + err = au_wh_init(h_root, br, sb);
2099 + wbr_wh_write_unlock(wbr);
2102 + au_hn_imtx_unlock(hdir);
2104 + mutex_unlock(h_mtx);
2105 + br->br_perm = old_perm;
2107 + if (!err && wbr && !au_br_writable(new_perm)) {
2109 + br->br_wbr = NULL;
2115 +static int au_wbr_init(struct au_branch *br, struct super_block *sb,
2116 + int perm, struct path *path)
2119 + struct kstatfs kst;
2120 + struct au_wbr *wbr;
2121 + struct dentry *h_dentry;
2124 + au_rw_init(&wbr->wbr_wh_rwsem);
2125 + memset(wbr->wbr_wh, 0, sizeof(wbr->wbr_wh));
2126 + atomic_set(&wbr->wbr_wh_running, 0);
2127 + wbr->wbr_bytes = 0;
2130 + * a limit for rmdir/rename a dir
2131 + * cf. AUFS_MAX_NAMELEN in include/linux/aufs_type.h
2133 + err = vfs_statfs(path, &kst);
2134 + if (unlikely(err))
2137 + h_dentry = path->dentry;
2138 + if (kst.f_namelen >= NAME_MAX)
2139 + err = au_br_init_wh(sb, br, perm, h_dentry);
2141 + pr_err("%.*s(%s), unsupported namelen %ld\n",
2142 + AuDLNPair(h_dentry), au_sbtype(h_dentry->d_sb),
2149 +/* intialize a new branch */
2150 +static int au_br_init(struct au_branch *br, struct super_block *sb,
2151 + struct au_opt_add *add)
2156 + memset(&br->br_xino, 0, sizeof(br->br_xino));
2157 + mutex_init(&br->br_xino.xi_nondir_mtx);
2158 + br->br_perm = add->perm;
2159 + br->br_mnt = add->path.mnt; /* set first, mntget() later */
2160 + spin_lock_init(&br->br_dykey_lock);
2161 + memset(br->br_dykey, 0, sizeof(br->br_dykey));
2162 + atomic_set(&br->br_count, 0);
2163 + br->br_xino_upper = AUFS_XINO_TRUNC_INIT;
2164 + atomic_set(&br->br_xino_running, 0);
2165 + br->br_id = au_new_br_id(sb);
2166 + AuDebugOn(br->br_id < 0);
2168 + if (au_br_writable(add->perm)) {
2169 + err = au_wbr_init(br, sb, add->perm, &add->path);
2170 + if (unlikely(err))
2174 + if (au_opt_test(au_mntflags(sb), XINO)) {
2175 + err = au_xino_br(sb, br, add->path.dentry->d_inode->i_ino,
2176 + au_sbr(sb, 0)->br_xino.xi_file, /*do_test*/1);
2177 + if (unlikely(err)) {
2178 + AuDebugOn(br->br_xino.xi_file);
2183 + sysaufs_br_init(br);
2184 + mntget(add->path.mnt);
2185 + goto out; /* success */
2188 + br->br_mnt = NULL;
2193 +static void au_br_do_add_brp(struct au_sbinfo *sbinfo, aufs_bindex_t bindex,
2194 + struct au_branch *br, aufs_bindex_t bend,
2195 + aufs_bindex_t amount)
2197 + struct au_branch **brp;
2199 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2201 + brp = sbinfo->si_branch + bindex;
2202 + memmove(brp + 1, brp, sizeof(*brp) * amount);
2204 + sbinfo->si_bend++;
2205 + if (unlikely(bend < 0))
2206 + sbinfo->si_bend = 0;
2209 +static void au_br_do_add_hdp(struct au_dinfo *dinfo, aufs_bindex_t bindex,
2210 + aufs_bindex_t bend, aufs_bindex_t amount)
2212 + struct au_hdentry *hdp;
2214 + AuRwMustWriteLock(&dinfo->di_rwsem);
2216 + hdp = dinfo->di_hdentry + bindex;
2217 + memmove(hdp + 1, hdp, sizeof(*hdp) * amount);
2218 + au_h_dentry_init(hdp);
2220 + if (unlikely(bend < 0))
2221 + dinfo->di_bstart = 0;
2224 +static void au_br_do_add_hip(struct au_iinfo *iinfo, aufs_bindex_t bindex,
2225 + aufs_bindex_t bend, aufs_bindex_t amount)
2227 + struct au_hinode *hip;
2229 + AuRwMustWriteLock(&iinfo->ii_rwsem);
2231 + hip = iinfo->ii_hinode + bindex;
2232 + memmove(hip + 1, hip, sizeof(*hip) * amount);
2233 + hip->hi_inode = NULL;
2236 + if (unlikely(bend < 0))
2237 + iinfo->ii_bstart = 0;
2240 +static void au_br_do_add(struct super_block *sb, struct dentry *h_dentry,
2241 + struct au_branch *br, aufs_bindex_t bindex)
2243 + struct dentry *root;
2244 + struct inode *root_inode;
2245 + aufs_bindex_t bend, amount;
2247 + root = sb->s_root;
2248 + root_inode = root->d_inode;
2249 + bend = au_sbend(sb);
2250 + amount = bend + 1 - bindex;
2251 + au_sbilist_lock();
2252 + au_br_do_add_brp(au_sbi(sb), bindex, br, bend, amount);
2253 + au_br_do_add_hdp(au_di(root), bindex, bend, amount);
2254 + au_br_do_add_hip(au_ii(root_inode), bindex, bend, amount);
2255 + au_set_h_dptr(root, bindex, dget(h_dentry));
2256 + au_set_h_iptr(root_inode, bindex, au_igrab(h_dentry->d_inode),
2258 + au_sbilist_unlock();
2261 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount)
2264 + aufs_bindex_t bend, add_bindex;
2265 + struct dentry *root, *h_dentry;
2266 + struct inode *root_inode;
2267 + struct au_branch *add_branch;
2269 + root = sb->s_root;
2270 + root_inode = root->d_inode;
2271 + IMustLock(root_inode);
2272 + err = test_add(sb, add, remount);
2273 + if (unlikely(err < 0))
2277 + goto out; /* success */
2280 + bend = au_sbend(sb);
2281 + add_branch = au_br_alloc(sb, bend + 2, add->perm);
2282 + err = PTR_ERR(add_branch);
2283 + if (IS_ERR(add_branch))
2286 + err = au_br_init(add_branch, sb, add);
2287 + if (unlikely(err)) {
2288 + au_br_do_free(add_branch);
2292 + add_bindex = add->bindex;
2293 + h_dentry = add->path.dentry;
2295 + au_br_do_add(sb, h_dentry, add_branch, add_bindex);
2297 + sysaufs_brs_del(sb, add_bindex);
2298 + au_br_do_add(sb, h_dentry, add_branch, add_bindex);
2299 + sysaufs_brs_add(sb, add_bindex);
2302 + if (!add_bindex) {
2303 + au_cpup_attr_all(root_inode, /*force*/1);
2304 + sb->s_maxbytes = h_dentry->d_sb->s_maxbytes;
2306 + au_add_nlink(root_inode, h_dentry->d_inode);
2309 + * this test/set prevents aufs from handling unnecesary notify events
2310 + * of xino files, in case of re-adding a writable branch which was
2311 + * once detached from aufs.
2313 + if (au_xino_brid(sb) < 0
2314 + && au_br_writable(add_branch->br_perm)
2315 + && !au_test_fs_bad_xino(h_dentry->d_sb)
2316 + && add_branch->br_xino.xi_file
2317 + && add_branch->br_xino.xi_file->f_dentry->d_parent == h_dentry)
2318 + au_xino_brid_set(sb, add_branch->br_id);
2324 +/* ---------------------------------------------------------------------- */
2330 +/* to show the line number, do not make it inlined function */
2331 +#define AuVerbose(do_info, fmt, ...) do { \
2333 + pr_info(fmt, ##__VA_ARGS__); \
2336 +static int au_test_ibusy(struct inode *inode, aufs_bindex_t bstart,
2337 + aufs_bindex_t bend)
2339 + return (inode && !S_ISDIR(inode->i_mode)) || bstart == bend;
2342 +static int au_test_dbusy(struct dentry *dentry, aufs_bindex_t bstart,
2343 + aufs_bindex_t bend)
2345 + return au_test_ibusy(dentry->d_inode, bstart, bend);
2349 + * test if the branch is deletable or not.
2351 +static int test_dentry_busy(struct dentry *root, aufs_bindex_t bindex,
2352 + unsigned int sigen, const unsigned int verbose)
2354 + int err, i, j, ndentry;
2355 + aufs_bindex_t bstart, bend;
2356 + struct au_dcsub_pages dpages;
2357 + struct au_dpage *dpage;
2360 + err = au_dpages_init(&dpages, GFP_NOFS);
2361 + if (unlikely(err))
2363 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
2364 + if (unlikely(err))
2367 + for (i = 0; !err && i < dpages.ndpage; i++) {
2368 + dpage = dpages.dpages + i;
2369 + ndentry = dpage->ndentry;
2370 + for (j = 0; !err && j < ndentry; j++) {
2371 + d = dpage->dentries[j];
2372 + AuDebugOn(!d->d_count);
2373 + if (!au_digen_test(d, sigen)) {
2374 + di_read_lock_child(d, AuLock_IR);
2375 + if (unlikely(au_dbrange_test(d))) {
2376 + di_read_unlock(d, AuLock_IR);
2380 + di_write_lock_child(d);
2381 + if (unlikely(au_dbrange_test(d))) {
2382 + di_write_unlock(d);
2385 + err = au_reval_dpath(d, sigen);
2387 + di_downgrade_lock(d, AuLock_IR);
2389 + di_write_unlock(d);
2394 + /* AuDbgDentry(d); */
2395 + bstart = au_dbstart(d);
2396 + bend = au_dbend(d);
2397 + if (bstart <= bindex
2399 + && au_h_dptr(d, bindex)
2400 + && au_test_dbusy(d, bstart, bend)) {
2402 + AuVerbose(verbose, "busy %.*s\n", AuDLNPair(d));
2405 + di_read_unlock(d, AuLock_IR);
2410 + au_dpages_free(&dpages);
2415 +static int test_inode_busy(struct super_block *sb, aufs_bindex_t bindex,
2416 + unsigned int sigen, const unsigned int verbose)
2419 + unsigned long long max, ull;
2420 + struct inode *i, **array;
2421 + aufs_bindex_t bstart, bend;
2423 + array = au_iarray_alloc(sb, &max);
2424 + err = PTR_ERR(array);
2425 + if (IS_ERR(array))
2429 + AuDbg("b%d\n", bindex);
2430 + for (ull = 0; !err && ull < max; ull++) {
2432 + if (i->i_ino == AUFS_ROOT_INO)
2435 + /* AuDbgInode(i); */
2436 + if (au_iigen(i) == sigen)
2437 + ii_read_lock_child(i);
2439 + ii_write_lock_child(i);
2440 + err = au_refresh_hinode_self(i);
2443 + ii_downgrade_lock(i);
2445 + ii_write_unlock(i);
2450 + bstart = au_ibstart(i);
2451 + bend = au_ibend(i);
2452 + if (bstart <= bindex
2454 + && au_h_iptr(i, bindex)
2455 + && au_test_ibusy(i, bstart, bend)) {
2457 + AuVerbose(verbose, "busy i%lu\n", i->i_ino);
2460 + ii_read_unlock(i);
2462 + au_iarray_free(array, max);
2468 +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex,
2469 + const unsigned int verbose)
2472 + unsigned int sigen;
2474 + sigen = au_sigen(root->d_sb);
2475 + DiMustNoWaiters(root);
2476 + IiMustNoWaiters(root->d_inode);
2477 + di_write_unlock(root);
2478 + err = test_dentry_busy(root, bindex, sigen, verbose);
2480 + err = test_inode_busy(root->d_sb, bindex, sigen, verbose);
2481 + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */
2486 +static void au_br_do_del_brp(struct au_sbinfo *sbinfo,
2487 + const aufs_bindex_t bindex,
2488 + const aufs_bindex_t bend)
2490 + struct au_branch **brp, **p;
2492 + AuRwMustWriteLock(&sbinfo->si_rwsem);
2494 + brp = sbinfo->si_branch + bindex;
2495 + if (bindex < bend)
2496 + memmove(brp, brp + 1, sizeof(*brp) * (bend - bindex));
2497 + sbinfo->si_branch[0 + bend] = NULL;
2498 + sbinfo->si_bend--;
2500 + p = krealloc(sbinfo->si_branch, sizeof(*p) * bend, AuGFP_SBILIST);
2502 + sbinfo->si_branch = p;
2503 + /* harmless error */
2506 +static void au_br_do_del_hdp(struct au_dinfo *dinfo, const aufs_bindex_t bindex,
2507 + const aufs_bindex_t bend)
2509 + struct au_hdentry *hdp, *p;
2511 + AuRwMustWriteLock(&dinfo->di_rwsem);
2513 + hdp = dinfo->di_hdentry;
2514 + if (bindex < bend)
2515 + memmove(hdp + bindex, hdp + bindex + 1,
2516 + sizeof(*hdp) * (bend - bindex));
2517 + hdp[0 + bend].hd_dentry = NULL;
2520 + p = krealloc(hdp, sizeof(*p) * bend, AuGFP_SBILIST);
2522 + dinfo->di_hdentry = p;
2523 + /* harmless error */
2526 +static void au_br_do_del_hip(struct au_iinfo *iinfo, const aufs_bindex_t bindex,
2527 + const aufs_bindex_t bend)
2529 + struct au_hinode *hip, *p;
2531 + AuRwMustWriteLock(&iinfo->ii_rwsem);
2533 + hip = iinfo->ii_hinode + bindex;
2534 + if (bindex < bend)
2535 + memmove(hip, hip + 1, sizeof(*hip) * (bend - bindex));
2536 + iinfo->ii_hinode[0 + bend].hi_inode = NULL;
2537 + au_hn_init(iinfo->ii_hinode + bend);
2540 + p = krealloc(iinfo->ii_hinode, sizeof(*p) * bend, AuGFP_SBILIST);
2542 + iinfo->ii_hinode = p;
2543 + /* harmless error */
2546 +static void au_br_do_del(struct super_block *sb, aufs_bindex_t bindex,
2547 + struct au_branch *br)
2549 + aufs_bindex_t bend;
2550 + struct au_sbinfo *sbinfo;
2551 + struct dentry *root, *h_root;
2552 + struct inode *inode, *h_inode;
2553 + struct au_hinode *hinode;
2555 + SiMustWriteLock(sb);
2557 + root = sb->s_root;
2558 + inode = root->d_inode;
2559 + sbinfo = au_sbi(sb);
2560 + bend = sbinfo->si_bend;
2562 + h_root = au_h_dptr(root, bindex);
2563 + hinode = au_hi(inode, bindex);
2564 + h_inode = au_igrab(hinode->hi_inode);
2567 + au_sbilist_lock();
2568 + au_br_do_del_brp(sbinfo, bindex, bend);
2569 + au_br_do_del_hdp(au_di(root), bindex, bend);
2570 + au_br_do_del_hip(au_ii(inode), bindex, bend);
2571 + au_sbilist_unlock();
2575 + au_br_do_free(br);
2578 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount)
2581 + unsigned int mnt_flags;
2582 + aufs_bindex_t bindex, bend, br_id;
2583 + unsigned char do_wh, verbose;
2584 + struct au_branch *br;
2585 + struct au_wbr *wbr;
2588 + bindex = au_find_dbindex(sb->s_root, del->h_path.dentry);
2591 + goto out; /* success */
2593 + pr_err("%s no such branch\n", del->pathname);
2596 + AuDbg("bindex b%d\n", bindex);
2599 + mnt_flags = au_mntflags(sb);
2600 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
2601 + bend = au_sbend(sb);
2602 + if (unlikely(!bend)) {
2603 + AuVerbose(verbose, "no more branches left\n");
2606 + br = au_sbr(sb, bindex);
2607 + i = atomic_read(&br->br_count);
2608 + if (unlikely(i)) {
2609 + AuVerbose(verbose, "%d file(s) opened\n", i);
2614 + do_wh = wbr && (wbr->wbr_whbase || wbr->wbr_plink || wbr->wbr_orph);
2616 + /* instead of WbrWhMustWriteLock(wbr) */
2617 + SiMustWriteLock(sb);
2618 + for (i = 0; i < AuBrWh_Last; i++) {
2619 + dput(wbr->wbr_wh[i]);
2620 + wbr->wbr_wh[i] = NULL;
2624 + err = test_children_busy(sb->s_root, bindex, verbose);
2625 + if (unlikely(err)) {
2632 + br_id = br->br_id;
2634 + au_br_do_del(sb, bindex, br);
2636 + sysaufs_brs_del(sb, bindex);
2637 + au_br_do_del(sb, bindex, br);
2638 + sysaufs_brs_add(sb, bindex);
2642 + au_cpup_attr_all(sb->s_root->d_inode, /*force*/1);
2643 + sb->s_maxbytes = au_sbr_sb(sb, 0)->s_maxbytes;
2645 + au_sub_nlink(sb->s_root->d_inode, del->h_path.dentry->d_inode);
2646 + if (au_opt_test(mnt_flags, PLINK))
2647 + au_plink_half_refresh(sb, br_id);
2649 + if (au_xino_brid(sb) == br_id)
2650 + au_xino_brid_set(sb, -1);
2651 + goto out; /* success */
2655 + rerr = au_br_init_wh(sb, br, br->br_perm, del->h_path.dentry);
2657 + pr_warning("failed re-creating base whiteout, %s. (%d)\n",
2658 + del->pathname, rerr);
2663 +/* ---------------------------------------------------------------------- */
2665 +static int au_ibusy(struct super_block *sb, struct aufs_ibusy __user *arg)
2668 + aufs_bindex_t bstart, bend;
2669 + struct aufs_ibusy ibusy;
2670 + struct inode *inode, *h_inode;
2673 + if (unlikely(!capable(CAP_SYS_ADMIN)))
2676 + err = copy_from_user(&ibusy, arg, sizeof(ibusy));
2678 + err = !access_ok(VERIFY_WRITE, &arg->h_ino, sizeof(arg->h_ino));
2679 + if (unlikely(err)) {
2686 + si_read_lock(sb, AuLock_FLUSH);
2687 + if (unlikely(ibusy.bindex < 0 || ibusy.bindex > au_sbend(sb)))
2691 + ibusy.h_ino = 0; /* invalid */
2692 + inode = ilookup(sb, ibusy.ino);
2694 + || inode->i_ino == AUFS_ROOT_INO
2695 + || is_bad_inode(inode))
2698 + ii_read_lock_child(inode);
2699 + bstart = au_ibstart(inode);
2700 + bend = au_ibend(inode);
2701 + if (bstart <= ibusy.bindex && ibusy.bindex <= bend) {
2702 + h_inode = au_h_iptr(inode, ibusy.bindex);
2703 + if (h_inode && au_test_ibusy(inode, bstart, bend))
2704 + ibusy.h_ino = h_inode->i_ino;
2706 + ii_read_unlock(inode);
2710 + si_read_unlock(sb);
2712 + err = __put_user(ibusy.h_ino, &arg->h_ino);
2713 + if (unlikely(err)) {
2722 +long au_ibusy_ioctl(struct file *file, unsigned long arg)
2724 + return au_ibusy(file->f_dentry->d_sb, (void __user *)arg);
2727 +#ifdef CONFIG_COMPAT
2728 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg)
2730 + return au_ibusy(file->f_dentry->d_sb, compat_ptr(arg));
2734 +/* ---------------------------------------------------------------------- */
2737 + * change a branch permission
2740 +static void au_warn_ima(void)
2743 + /* since it doesn't support mark_files_ro() */
2744 + AuWarn1("RW -> RO makes IMA to produce wrong message\n");
2748 +static int do_need_sigen_inc(int a, int b)
2750 + return au_br_whable(a) && !au_br_whable(b);
2753 +static int need_sigen_inc(int old, int new)
2755 + return do_need_sigen_inc(old, new)
2756 + || do_need_sigen_inc(new, old);
2759 +static unsigned long long au_farray_cb(void *a,
2760 + unsigned long long max __maybe_unused,
2763 + unsigned long long n;
2764 + struct file **p, *f;
2765 + struct super_block *sb = arg;
2769 + lg_global_lock(files_lglock);
2770 + do_file_list_for_each_entry(sb, f) {
2773 + && !special_file(f->f_dentry->d_inode->i_mode)) {
2777 + AuDebugOn(n > max);
2779 + } while_file_list_for_each_entry;
2780 + lg_global_unlock(files_lglock);
2785 +static struct file **au_farray_alloc(struct super_block *sb,
2786 + unsigned long long *max)
2788 + *max = atomic_long_read(&au_sbi(sb)->si_nfiles);
2789 + return au_array_alloc(max, au_farray_cb, sb);
2792 +static void au_farray_free(struct file **a, unsigned long long max)
2794 + unsigned long long ull;
2796 + for (ull = 0; ull < max; ull++)
2802 +static int au_br_mod_files_ro(struct super_block *sb, aufs_bindex_t bindex)
2805 + unsigned int mnt_flags;
2806 + unsigned long long ull, max;
2807 + aufs_bindex_t br_id;
2808 + unsigned char verbose;
2809 + struct file *file, *hf, **array;
2810 + struct inode *inode;
2811 + struct au_hfile *hfile;
2813 + mnt_flags = au_mntflags(sb);
2814 + verbose = !!au_opt_test(mnt_flags, VERBOSE);
2816 + array = au_farray_alloc(sb, &max);
2817 + err = PTR_ERR(array);
2818 + if (IS_ERR(array))
2822 + br_id = au_sbr_id(sb, bindex);
2823 + for (ull = 0; ull < max; ull++) {
2824 + file = array[ull];
2826 + /* AuDbg("%.*s\n", AuDLNPair(file->f_dentry)); */
2827 + fi_read_lock(file);
2828 + if (unlikely(au_test_mmapped(file))) {
2830 + AuVerbose(verbose, "mmapped %.*s\n",
2831 + AuDLNPair(file->f_dentry));
2833 + FiMustNoWaiters(file);
2834 + fi_read_unlock(file);
2838 + inode = file->f_dentry->d_inode;
2839 + hfile = &au_fi(file)->fi_htop;
2840 + hf = hfile->hf_file;
2841 + if (!S_ISREG(inode->i_mode)
2842 + || !(file->f_mode & FMODE_WRITE)
2843 + || hfile->hf_br->br_id != br_id
2844 + || !(hf->f_mode & FMODE_WRITE))
2845 + array[ull] = NULL;
2851 + FiMustNoWaiters(file);
2852 + fi_read_unlock(file);
2860 + for (ull = 0; ull < max; ull++) {
2861 + file = array[ull];
2865 + /* todo: already flushed? */
2866 + /* cf. fs/super.c:mark_files_ro() */
2867 + /* fi_read_lock(file); */
2868 + hfile = &au_fi(file)->fi_htop;
2869 + hf = hfile->hf_file;
2870 + /* fi_read_unlock(file); */
2871 + spin_lock(&hf->f_lock);
2872 + hf->f_mode &= ~FMODE_WRITE;
2873 + spin_unlock(&hf->f_lock);
2874 + if (!file_check_writeable(hf)) {
2875 + file_release_write(hf);
2876 + mnt_drop_write(hf->f_vfsmnt);
2881 + au_farray_free(array, max);
2887 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
2891 + aufs_bindex_t bindex;
2893 + struct dentry *root;
2894 + struct au_branch *br;
2896 + root = sb->s_root;
2897 + bindex = au_find_dbindex(root, mod->h_root);
2900 + return 0; /* success */
2902 + pr_err("%s no such branch\n", mod->path);
2905 + AuDbg("bindex b%d\n", bindex);
2907 + err = test_br(mod->h_root->d_inode, mod->perm, mod->path);
2908 + if (unlikely(err))
2911 + br = au_sbr(sb, bindex);
2912 + if (br->br_perm == mod->perm)
2913 + return 0; /* success */
2915 + if (au_br_writable(br->br_perm)) {
2916 + /* remove whiteout base */
2917 + err = au_br_init_wh(sb, br, mod->perm, mod->h_root);
2918 + if (unlikely(err))
2921 + if (!au_br_writable(mod->perm)) {
2922 + /* rw --> ro, file might be mmapped */
2923 + DiMustNoWaiters(root);
2924 + IiMustNoWaiters(root->d_inode);
2925 + di_write_unlock(root);
2926 + err = au_br_mod_files_ro(sb, bindex);
2927 + /* aufs_write_lock() calls ..._child() */
2928 + di_write_lock_child(root);
2930 + if (unlikely(err)) {
2932 + br->br_wbr = kmalloc(sizeof(*br->br_wbr),
2935 + path.mnt = br->br_mnt;
2936 + path.dentry = mod->h_root;
2937 + rerr = au_wbr_init(br, sb, br->br_perm,
2940 + if (unlikely(rerr)) {
2941 + AuIOErr("nested error %d (%d)\n",
2943 + br->br_perm = mod->perm;
2947 + } else if (au_br_writable(mod->perm)) {
2950 + br->br_wbr = kmalloc(sizeof(*br->br_wbr), GFP_NOFS);
2952 + path.mnt = br->br_mnt;
2953 + path.dentry = mod->h_root;
2954 + err = au_wbr_init(br, sb, mod->perm, &path);
2955 + if (unlikely(err)) {
2956 + kfree(br->br_wbr);
2957 + br->br_wbr = NULL;
2963 + *do_refresh |= need_sigen_inc(br->br_perm, mod->perm);
2964 + br->br_perm = mod->perm;
2971 diff -urN /usr/share/empty/fs/aufs/branch.h linux/fs/aufs/branch.h
2972 --- /usr/share/empty/fs/aufs/branch.h 1970-01-01 01:00:00.000000000 +0100
2973 +++ linux/fs/aufs/branch.h 2011-10-24 20:51:51.580466925 +0200
2976 + * Copyright (C) 2005-2011 Junjiro R. Okajima
2978 + * This program, aufs is free software; you can redistribute it and/or modify
2979 + * it under the terms of the GNU General Public License as published by
2980 + * the Free Software Foundation; either version 2 of the License, or
2981 + * (at your option) any later version.
2983 + * This program is distributed in the hope that it will be useful,
2984 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
2985 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2986 + * GNU General Public License for more details.
2988 + * You should have received a copy of the GNU General Public License
2989 + * along with this program; if not, write to the Free Software
2990 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2994 + * branch filesystems and xino for them
2997 +#ifndef __AUFS_BRANCH_H__
2998 +#define __AUFS_BRANCH_H__
3002 +#include <linux/fs.h>
3003 +#include <linux/mount.h>
3004 +#include <linux/aufs_type.h>
3009 +/* ---------------------------------------------------------------------- */
3012 +struct au_xino_file {
3013 + struct file *xi_file;
3014 + struct mutex xi_nondir_mtx;
3016 + /* todo: make xino files an array to support huge inode number */
3018 +#ifdef CONFIG_DEBUG_FS
3019 + struct dentry *xi_dbgaufs;
3023 +/* members for writable branch only */
3024 +enum {AuBrWh_BASE, AuBrWh_PLINK, AuBrWh_ORPH, AuBrWh_Last};
3026 + struct au_rwsem wbr_wh_rwsem;
3027 + struct dentry *wbr_wh[AuBrWh_Last];
3028 + atomic_t wbr_wh_running;
3029 +#define wbr_whbase wbr_wh[AuBrWh_BASE] /* whiteout base */
3030 +#define wbr_plink wbr_wh[AuBrWh_PLINK] /* pseudo-link dir */
3031 +#define wbr_orph wbr_wh[AuBrWh_ORPH] /* dir for orphans */
3034 + unsigned long long wbr_bytes;
3037 +/* ext2 has 3 types of operations at least, ext3 has 4 */
3038 +#define AuBrDynOp (AuDyLast * 4)
3040 +/* protected by superblock rwsem */
3042 + struct au_xino_file br_xino;
3044 + aufs_bindex_t br_id;
3047 + struct vfsmount *br_mnt;
3048 + spinlock_t br_dykey_lock;
3049 + struct au_dykey *br_dykey[AuBrDynOp];
3050 + atomic_t br_count;
3052 + struct au_wbr *br_wbr;
3054 + /* xino truncation */
3055 + blkcnt_t br_xino_upper; /* watermark in blocks */
3056 + atomic_t br_xino_running;
3058 +#ifdef CONFIG_AUFS_HFSNOTIFY
3059 + struct fsnotify_group *br_hfsn_group;
3060 + struct fsnotify_ops br_hfsn_ops;
3063 +#ifdef CONFIG_SYSFS
3064 + /* an entry under sysfs per mount-point */
3066 + struct attribute br_attr;
3070 +/* ---------------------------------------------------------------------- */
3072 +/* branch permissions and attributes */
3073 +#define AuBrPerm_RW 1 /* writable, hardlinkable wh */
3074 +#define AuBrPerm_RO (1 << 1) /* readonly */
3075 +#define AuBrPerm_RR (1 << 2) /* natively readonly */
3076 +#define AuBrPerm_Mask (AuBrPerm_RW | AuBrPerm_RO | AuBrPerm_RR)
3078 +#define AuBrRAttr_WH (1 << 3) /* whiteout-able */
3080 +#define AuBrWAttr_NoLinkWH (1 << 4) /* un-hardlinkable whiteouts */
3082 +static inline int au_br_writable(int brperm)
3084 + return brperm & AuBrPerm_RW;
3087 +static inline int au_br_whable(int brperm)
3089 + return brperm & (AuBrPerm_RW | AuBrRAttr_WH);
3092 +static inline int au_br_wh_linkable(int brperm)
3094 + return !(brperm & AuBrWAttr_NoLinkWH);
3097 +static inline int au_br_rdonly(struct au_branch *br)
3099 + return ((br->br_mnt->mnt_sb->s_flags & MS_RDONLY)
3100 + || !au_br_writable(br->br_perm))
3104 +static inline int au_br_hnotifyable(int brperm __maybe_unused)
3106 +#ifdef CONFIG_AUFS_HNOTIFY
3107 + return !(brperm & AuBrPerm_RR);
3113 +/* ---------------------------------------------------------------------- */
3117 +void au_br_free(struct au_sbinfo *sinfo);
3118 +int au_br_index(struct super_block *sb, aufs_bindex_t br_id);
3120 +int au_br_add(struct super_block *sb, struct au_opt_add *add, int remount);
3122 +int au_br_del(struct super_block *sb, struct au_opt_del *del, int remount);
3123 +long au_ibusy_ioctl(struct file *file, unsigned long arg);
3124 +#ifdef CONFIG_COMPAT
3125 +long au_ibusy_compat_ioctl(struct file *file, unsigned long arg);
3128 +int au_br_mod(struct super_block *sb, struct au_opt_mod *mod, int remount,
3132 +static const loff_t au_loff_max = LLONG_MAX;
3134 +int au_xib_trunc(struct super_block *sb);
3135 +ssize_t xino_fread(au_readf_t func, struct file *file, void *buf, size_t size,
3137 +ssize_t xino_fwrite(au_writef_t func, struct file *file, void *buf, size_t size,
3139 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src);
3140 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent);
3141 +ino_t au_xino_new_ino(struct super_block *sb);
3142 +void au_xino_delete_inode(struct inode *inode, const int unlinked);
3143 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
3145 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
3147 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t hino,
3148 + struct file *base_file, int do_test);
3149 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex);
3151 +struct au_opt_xino;
3152 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount);
3153 +void au_xino_clr(struct super_block *sb);
3154 +struct file *au_xino_def(struct super_block *sb);
3155 +int au_xino_path(struct seq_file *seq, struct file *file);
3157 +/* ---------------------------------------------------------------------- */
3159 +/* Superblock to branch */
3161 +aufs_bindex_t au_sbr_id(struct super_block *sb, aufs_bindex_t bindex)
3163 + return au_sbr(sb, bindex)->br_id;
3167 +struct vfsmount *au_sbr_mnt(struct super_block *sb, aufs_bindex_t bindex)
3169 + return au_sbr(sb, bindex)->br_mnt;
3173 +struct super_block *au_sbr_sb(struct super_block *sb, aufs_bindex_t bindex)
3175 + return au_sbr_mnt(sb, bindex)->mnt_sb;
3178 +static inline void au_sbr_put(struct super_block *sb, aufs_bindex_t bindex)
3180 + atomic_dec(&au_sbr(sb, bindex)->br_count);
3183 +static inline int au_sbr_perm(struct super_block *sb, aufs_bindex_t bindex)
3185 + return au_sbr(sb, bindex)->br_perm;
3188 +static inline int au_sbr_whable(struct super_block *sb, aufs_bindex_t bindex)
3190 + return au_br_whable(au_sbr_perm(sb, bindex));
3193 +/* ---------------------------------------------------------------------- */
3196 + * wbr_wh_read_lock, wbr_wh_write_lock
3197 + * wbr_wh_read_unlock, wbr_wh_write_unlock, wbr_wh_downgrade_lock
3199 +AuSimpleRwsemFuncs(wbr_wh, struct au_wbr *wbr, &wbr->wbr_wh_rwsem);
3201 +#define WbrWhMustNoWaiters(wbr) AuRwMustNoWaiters(&wbr->wbr_wh_rwsem)
3202 +#define WbrWhMustAnyLock(wbr) AuRwMustAnyLock(&wbr->wbr_wh_rwsem)
3203 +#define WbrWhMustWriteLock(wbr) AuRwMustWriteLock(&wbr->wbr_wh_rwsem)
3205 +#endif /* __KERNEL__ */
3206 +#endif /* __AUFS_BRANCH_H__ */
3207 diff -urN /usr/share/empty/fs/aufs/conf.mk linux/fs/aufs/conf.mk
3208 --- /usr/share/empty/fs/aufs/conf.mk 1970-01-01 01:00:00.000000000 +0100
3209 +++ linux/fs/aufs/conf.mk 2011-08-24 13:30:24.731313534 +0200
3212 +AuConfStr = CONFIG_AUFS_FS=${CONFIG_AUFS_FS}
3216 +AuConfStr += ${1}=${${1}}
3220 +AuConfAll = BRANCH_MAX_127 BRANCH_MAX_511 BRANCH_MAX_1023 BRANCH_MAX_32767 \
3222 + HNOTIFY HFSNOTIFY \
3233 +$(foreach i, ${AuConfAll}, \
3234 + $(eval $(call AuConf,CONFIG_AUFS_${i})))
3236 +AuConfName = ${obj}/conf.str
3237 +${AuConfName}.tmp: FORCE
3238 + @echo ${AuConfStr} | tr ' ' '\n' | sed -e 's/^/"/' -e 's/$$/\\n"/' > $@
3239 +${AuConfName}: ${AuConfName}.tmp
3240 + @diff -q $< $@ > /dev/null 2>&1 || { \
3241 + echo ' GEN ' $@; \
3245 +clean-files += ${AuConfName} ${AuConfName}.tmp
3246 +${obj}/sysfs.o: ${AuConfName}
3248 +-include ${srctree}/${src}/conf_priv.mk
3249 diff -urN /usr/share/empty/fs/aufs/cpup.c linux/fs/aufs/cpup.c
3250 --- /usr/share/empty/fs/aufs/cpup.c 1970-01-01 01:00:00.000000000 +0100
3251 +++ linux/fs/aufs/cpup.c 2012-01-05 11:00:59.215501779 +0100
3254 + * Copyright (C) 2005-2011 Junjiro R. Okajima
3256 + * This program, aufs is free software; you can redistribute it and/or modify
3257 + * it under the terms of the GNU General Public License as published by
3258 + * the Free Software Foundation; either version 2 of the License, or
3259 + * (at your option) any later version.
3261 + * This program is distributed in the hope that it will be useful,
3262 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
3263 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
3264 + * GNU General Public License for more details.
3266 + * You should have received a copy of the GNU General Public License
3267 + * along with this program; if not, write to the Free Software
3268 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
3272 + * copy-up functions, see wbr_policy.c for copy-down
3275 +#include <linux/file.h>
3276 +#include <linux/fs_stack.h>
3277 +#include <linux/mm.h>
3278 +#include <linux/uaccess.h>
3281 +void au_cpup_attr_flags(struct inode *dst, struct inode *src)
3283 + const unsigned int mask = S_DEAD | S_SWAPFILE | S_PRIVATE
3284 + | S_NOATIME | S_NOCMTIME;
3286 + dst->i_flags |= src->i_flags & ~mask;
3287 + if (au_test_fs_notime(dst->i_sb))
3288 + dst->i_flags |= S_NOATIME | S_NOCMTIME;
3291 +void au_cpup_attr_timesizes(struct inode *inode)
3293 + struct inode *h_inode;
3295 + h_inode = au_h_iptr(inode, au_ibstart(inode));
3296 + fsstack_copy_attr_times(inode, h_inode);
3297 + fsstack_copy_inode_size(inode, h_inode);
3300 +void au_cpup_attr_nlink(struct inode *inode, int force)
3302 + struct inode *h_inode;
3303 + struct super_block *sb;
3304 + aufs_bindex_t bindex, bend;
3307 + bindex = au_ibstart(inode);
3308 + h_inode = au_h_iptr(inode, bindex);
3310 + && !S_ISDIR(h_inode->i_mode)
3311 + && au_opt_test(au_mntflags(sb), PLINK)
3312 + && au_plink_test(inode))
3315 + set_nlink(inode, h_inode->i_nlink);
3318 + * fewer nlink makes find(1) noisy, but larger nlink doesn't.
3319 + * it may includes whplink directory.
3321 + if (S_ISDIR(h_inode->i_mode)) {
3322 + bend = au_ibend(inode);
3323 + for (bindex++; bindex <= bend; bindex++) {
3324 + h_inode = au_h_iptr(inode, bindex);
3326 + au_add_nlink(inode, h_inode);
3331 +void au_cpup_attr_changeable(struct inode *inode)
3333 + struct inode *h_inode;
3335 + h_inode = au_h_iptr(inode, au_ibstart(inode));
3336 + inode->i_mode = h_inode->i_mode;
3337 + inode->i_uid = h_inode->i_uid;
3338 + inode->i_gid = h_inode->i_gid;
3339 + au_cpup_attr_timesizes(inode);
3340 + au_cpup_attr_flags(inode, h_inode);
3343 +void au_cpup_igen(struct inode *inode, struct inode *h_inode)
3345 + struct au_iinfo *iinfo = au_ii(inode);
3347 + IiMustWriteLock(inode);
3349 + iinfo->ii_higen = h_inode->i_generation;
3350 + iinfo->ii_hsb1 = h_inode->i_sb;
3353 +void au_cpup_attr_all(struct inode *inode, int force)
3355 + struct inode *h_inode;
3357 + h_inode = au_h_iptr(inode, au_ibstart(inode));
3358 + au_cpup_attr_changeable(inode);
3359 + if (inode->i_nlink > 0)
3360 + au_cpup_attr_nlink(inode, force);
3361 + inode->i_rdev = h_inode->i_rdev;
3362 + inode->i_blkbits = h_inode->i_blkbits;
3363 + au_cpup_igen(inode, h_inode);
3366 +/* ---------------------------------------------------------------------- */
3368 +/* Note: dt_dentry and dt_h_dentry are not dget/dput-ed */
3370 +/* keep the timestamps of the parent dir when cpup */
3371 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
3372 + struct path *h_path)
3374 + struct inode *h_inode;
3376 + dt->dt_dentry = dentry;
3377 + dt->dt_h_path = *h_path;
3378 + h_inode = h_path->dentry->d_inode;
3379 + dt->dt_atime = h_inode->i_atime;
3380 + dt->dt_mtime = h_inode->i_mtime;
3384 +void au_dtime_revert(struct au_dtime *dt)
3386 + struct iattr attr;
3389 + attr.ia_atime = dt->dt_atime;
3390 + attr.ia_mtime = dt->dt_mtime;
3391 + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET
3392 + | ATTR_ATIME | ATTR_ATIME_SET;
3394 + err = vfsub_notify_change(&dt->dt_h_path, &attr);
3395 + if (unlikely(err))
3396 + pr_warning("restoring timestamps failed(%d). ignored\n", err);
3399 +/* ---------------------------------------------------------------------- */
3401 +static noinline_for_stack
3402 +int cpup_iattr(struct dentry *dst, aufs_bindex_t bindex, struct dentry *h_src)
3406 + struct path h_path;
3407 + struct inode *h_isrc, *h_idst;
3409 + h_path.dentry = au_h_dptr(dst, bindex);
3410 + h_idst = h_path.dentry->d_inode;
3411 + h_path.mnt = au_sbr_mnt(dst->d_sb, bindex);
3412 + h_isrc = h_src->d_inode;
3413 + ia.ia_valid = ATTR_FORCE | ATTR_UID | ATTR_GID
3414 + | ATTR_ATIME | ATTR_MTIME
3415 + | ATTR_ATIME_SET | ATTR_MTIME_SET;
3416 + ia.ia_uid = h_isrc->i_uid;
3417 + ia.ia_gid = h_isrc->i_gid;
3418 + ia.ia_atime = h_isrc->i_atime;
3419 + ia.ia_mtime = h_isrc->i_mtime;
3420 + if (h_idst->i_mode != h_isrc->i_mode
3421 + && !S_ISLNK(h_idst->i_mode)) {
3422 + ia.ia_valid |= ATTR_MODE;
3423 + ia.ia_mode = h_isrc->i_mode;
3425 + sbits = !!(h_isrc->i_mode & (S_ISUID | S_ISGID));
3426 + au_cpup_attr_flags(h_idst, h_isrc);
3427 + err = vfsub_notify_change(&h_path, &ia);
3429 + /* is this nfs only? */
3430 + if (!err && sbits && au_test_nfs(h_path.dentry->d_sb)) {
3431 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
3432 + ia.ia_mode = h_isrc->i_mode;
3433 + err = vfsub_notify_change(&h_path, &ia);
3439 +/* ---------------------------------------------------------------------- */
3441 +static int au_do_copy_file(struct file *dst, struct file *src, loff_t len,
3442 + char *buf, unsigned long blksize)
3445 + size_t sz, rbytes, wbytes;
3446 + unsigned char all_zero;
3448 + struct mutex *h_mtx;
3449 + /* reduce stack usage */
3452 + zp = page_address(ZERO_PAGE(0));
3453 + if (unlikely(!zp))
3454 + return -ENOMEM; /* possible? */
3459 + AuDbg("len %lld\n", len);
3461 + if (len < blksize)
3465 + /* todo: signal_pending? */
3466 + while (!rbytes || err == -EAGAIN || err == -EINTR) {
3467 + rbytes = vfsub_read_k(src, buf, sz, &src->f_pos);
3470 + if (unlikely(err < 0))
3474 + if (len >= rbytes && rbytes == blksize)
3475 + all_zero = !memcmp(buf, zp, rbytes);
3482 + b = vfsub_write_k(dst, p, wbytes, &dst->f_pos);
3484 + /* todo: signal_pending? */
3485 + if (unlikely(err == -EAGAIN || err == -EINTR))
3487 + if (unlikely(err < 0))
3496 + res = vfsub_llseek(dst, rbytes, SEEK_CUR);
3498 + if (unlikely(res < 0))
3505 + /* the last block may be a hole */
3506 + if (!err && all_zero) {
3507 + AuLabel(last hole);
3510 + if (au_test_nfs(dst->f_dentry->d_sb)) {
3511 + /* nfs requires this step to make last hole */
3512 + /* is this only nfs? */
3514 + /* todo: signal_pending? */
3515 + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos);
3516 + } while (err == -EAGAIN || err == -EINTR);
3523 + ia->ia_size = dst->f_pos;
3524 + ia->ia_valid = ATTR_SIZE | ATTR_FILE;
3525 + ia->ia_file = dst;
3526 + h_mtx = &dst->f_dentry->d_inode->i_mutex;
3527 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
3528 + err = vfsub_notify_change(&dst->f_path, ia);
3529 + mutex_unlock(h_mtx);
3536 +int au_copy_file(struct file *dst, struct file *src, loff_t len)
3539 + unsigned long blksize;
3540 + unsigned char do_kfree;
3544 + blksize = dst->f_dentry->d_sb->s_blocksize;
3545 + if (!blksize || PAGE_SIZE < blksize)
3546 + blksize = PAGE_SIZE;
3547 + AuDbg("blksize %lu\n", blksize);
3548 + do_kfree = (blksize != PAGE_SIZE && blksize >= sizeof(struct iattr *));
3550 + buf = kmalloc(blksize, GFP_NOFS);
3552 + buf = (void *)__get_free_page(GFP_NOFS);
3553 + if (unlikely(!buf))
3556 + if (len > (1 << 22))
3557 + AuDbg("copying a large file %lld\n", (long long)len);
3561 + err = au_do_copy_file(dst, src, len, buf, blksize);
3565 + free_page((unsigned long)buf);
3572 + * to support a sparse file which is opened with O_APPEND,
3573 + * we need to close the file.
3575 +static int au_cp_regular(struct dentry *dentry, aufs_bindex_t bdst,
3576 + aufs_bindex_t bsrc, loff_t len)
3579 + enum { SRC, DST };
3581 + aufs_bindex_t bindex;
3582 + unsigned int flags;
3583 + struct dentry *dentry;
3584 + struct file *file;
3585 + void *label, *label_file;
3589 + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE,
3592 + .label_file = &&out_src
3596 + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE,
3598 + .label = &&out_src,
3599 + .label_file = &&out_dst
3602 + struct super_block *sb;
3604 + /* bsrc branch can be ro/rw. */
3605 + sb = dentry->d_sb;
3607 + for (i = 0; i < 2; i++, f++) {
3608 + f->dentry = au_h_dptr(dentry, f->bindex);
3609 + f->file = au_h_open(dentry, f->bindex, f->flags, /*file*/NULL);
3610 + err = PTR_ERR(f->file);
3611 + if (IS_ERR(f->file))
3614 + if (unlikely(!f->file->f_op))
3615 + goto *f->label_file;
3618 + /* try stopping to update while we copyup */
3619 + IMustLock(file[SRC].dentry->d_inode);
3620 + err = au_copy_file(file[DST].file, file[SRC].file, len);
3623 + fput(file[DST].file);
3624 + au_sbr_put(sb, file[DST].bindex);
3626 + fput(file[SRC].file);
3627 + au_sbr_put(sb, file[SRC].bindex);
3632 +static int au_do_cpup_regular(struct dentry *dentry, aufs_bindex_t bdst,
3633 + aufs_bindex_t bsrc, loff_t len,
3634 + struct inode *h_dir, struct path *h_path)
3640 + l = i_size_read(au_h_iptr(dentry->d_inode, bsrc));
3641 + if (len == -1 || l < len)
3644 + err = au_cp_regular(dentry, bdst, bsrc, len);
3646 + goto out; /* success */
3648 + rerr = vfsub_unlink(h_dir, h_path, /*force*/0);
3650 + AuIOErr("failed unlinking cpup-ed %.*s(%d, %d)\n",
3651 + AuDLNPair(h_path->dentry), err, rerr);
3659 +static int au_do_cpup_symlink(struct path *h_path, struct dentry *h_src,
3660 + struct inode *h_dir)
3663 + mm_segment_t old_fs;
3670 + if (unlikely(!h_src->d_inode->i_op->readlink))
3674 + sym.k = __getname_gfp(GFP_NOFS);
3675 + if (unlikely(!sym.k))
3678 + /* unnecessary to support mmap_sem since symlink is not mmap-able */
3679 + old_fs = get_fs();
3680 + set_fs(KERNEL_DS);
3681 + symlen = h_src->d_inode->i_op->readlink(h_src, sym.u, PATH_MAX);
3686 + sym.k[symlen] = 0;
3687 + err = vfsub_symlink(h_dir, h_path, sym.k);
3695 +/* return with the lower dst inode is locked */
3696 +static noinline_for_stack
3697 +int cpup_entry(struct dentry *dentry, aufs_bindex_t bdst,
3698 + aufs_bindex_t bsrc, loff_t len, unsigned int flags,
3699 + struct dentry *dst_parent)
3703 + unsigned int mnt_flags;
3704 + unsigned char isdir;
3705 + const unsigned char do_dt = !!au_ftest_cpup(flags, DTIME);
3706 + struct au_dtime dt;
3707 + struct path h_path;
3708 + struct dentry *h_src, *h_dst, *h_parent;
3709 + struct inode *h_inode, *h_dir;
3710 + struct super_block *sb;
3712 + /* bsrc branch can be ro/rw. */
3713 + h_src = au_h_dptr(dentry, bsrc);
3714 + h_inode = h_src->d_inode;
3715 + AuDebugOn(h_inode != au_h_iptr(dentry->d_inode, bsrc));
3717 + /* try stopping to be referenced while we are creating */
3718 + h_dst = au_h_dptr(dentry, bdst);
3719 + h_parent = h_dst->d_parent; /* dir inode is locked */
3720 + h_dir = h_parent->d_inode;
3722 + AuDebugOn(h_parent != h_dst->d_parent);
3724 + sb = dentry->d_sb;
3725 + h_path.mnt = au_sbr_mnt(sb, bdst);
3727 + h_path.dentry = h_parent;
3728 + au_dtime_store(&dt, dst_parent, &h_path);
3730 + h_path.dentry = h_dst;
3733 + mode = h_inode->i_mode;
3734 + switch (mode & S_IFMT) {
3736 + /* try stopping to update while we are referencing */
3737 + IMustLock(h_inode);
3738 + err = vfsub_create(h_dir, &h_path, mode | S_IWUSR);
3740 + err = au_do_cpup_regular
3741 + (dentry, bdst, bsrc, len,
3742 + au_h_iptr(dst_parent->d_inode, bdst), &h_path);
3746 + err = vfsub_mkdir(h_dir, &h_path, mode);
3749 + * strange behaviour from the users view,
3750 + * particularry setattr case
3752 + if (au_ibstart(dst_parent->d_inode) == bdst)
3753 + au_cpup_attr_nlink(dst_parent->d_inode,
3755 + au_cpup_attr_nlink(dentry->d_inode, /*force*/1);
3759 + err = au_do_cpup_symlink(&h_path, h_src, h_dir);
3763 + AuDebugOn(!capable(CAP_MKNOD));
3767 + err = vfsub_mknod(h_dir, &h_path, mode, h_inode->i_rdev);
3770 + AuIOErr("Unknown inode type 0%o\n", mode);
3774 + mnt_flags = au_mntflags(sb);
3775 + if (!au_opt_test(mnt_flags, UDBA_NONE)
3777 + && au_opt_test(mnt_flags, XINO)
3778 + && h_inode->i_nlink == 1
3779 + /* todo: unnecessary? */
3780 + /* && dentry->d_inode->i_nlink == 1 */
3782 + && !au_ftest_cpup(flags, KEEPLINO))
3783 + au_xino_write(sb, bsrc, h_inode->i_ino, /*ino*/0);
3784 + /* ignore this error */
3787 + au_dtime_revert(&dt);
3792 + * copyup the @dentry from @bsrc to @bdst.
3793 + * the caller must set the both of lower dentries.
3794 + * @len is for truncating when it is -1 copyup the entire file.
3795 + * in link/rename cases, @dst_parent may be different from the real one.
3797 +static int au_cpup_single(struct dentry *dentry, aufs_bindex_t bdst,
3798 + aufs_bindex_t bsrc, loff_t len, unsigned int flags,
3799 + struct dentry *dst_parent)
3802 + aufs_bindex_t old_ibstart;
3803 + unsigned char isdir, plink;
3804 + struct au_dtime dt;
3805 + struct path h_path;
3806 + struct dentry *h_src, *h_dst, *h_parent;
3807 + struct inode *dst_inode, *h_dir, *inode;
3808 + struct super_block *sb;
3810 + AuDebugOn(bsrc <= bdst);
3812 + sb = dentry->d_sb;
3813 + h_path.mnt = au_sbr_mnt(sb, bdst);
3814 + h_dst = au_h_dptr(dentry, bdst);
3815 + h_parent = h_dst->d_parent; /* dir inode is locked */
3816 + h_dir = h_parent->d_inode;
3819 + h_src = au_h_dptr(dentry, bsrc);
3820 + inode = dentry->d_inode;
3823 + dst_parent = dget_parent(dentry);
3827 + plink = !!au_opt_test(au_mntflags(sb), PLINK);
3828 + dst_inode = au_h_iptr(inode, bdst);
3830 + if (unlikely(!plink)) {
3832 + AuIOErr("hi%lu(i%lu) exists on b%d "
3833 + "but plink is disabled\n",
3834 + dst_inode->i_ino, inode->i_ino, bdst);
3838 + if (dst_inode->i_nlink) {
3839 + const int do_dt = au_ftest_cpup(flags, DTIME);
3841 + h_src = au_plink_lkup(inode, bdst);
3842 + err = PTR_ERR(h_src);
3843 + if (IS_ERR(h_src))
3845 + if (unlikely(!h_src->d_inode)) {
3847 + AuIOErr("i%lu exists on a upper branch "
3848 + "but not pseudo-linked\n",
3855 + h_path.dentry = h_parent;
3856 + au_dtime_store(&dt, dst_parent, &h_path);
3858 + h_path.dentry = h_dst;
3859 + err = vfsub_link(h_src, h_dir, &h_path);
3861 + au_dtime_revert(&dt);
3865 + /* todo: cpup_wh_file? */
3867 + au_update_ibrange(inode, /*do_put_zero*/1);
3870 + old_ibstart = au_ibstart(inode);
3871 + err = cpup_entry(dentry, bdst, bsrc, len, flags, dst_parent);
3872 + if (unlikely(err))
3874 + dst_inode = h_dst->d_inode;
3875 + mutex_lock_nested(&dst_inode->i_mutex, AuLsc_I_CHILD2);
3877 + err = cpup_iattr(dentry, bdst, h_src);
3878 + isdir = S_ISDIR(dst_inode->i_mode);
3880 + if (bdst < old_ibstart) {
3881 + if (S_ISREG(inode->i_mode)) {
3882 + err = au_dy_iaop(inode, bdst, dst_inode);
3883 + if (unlikely(err))
3886 + au_set_ibstart(inode, bdst);
3888 + au_set_h_iptr(inode, bdst, au_igrab(dst_inode),
3889 + au_hi_flags(inode, isdir));
3890 + mutex_unlock(&dst_inode->i_mutex);
3892 + && h_src->d_inode->i_nlink > 1
3894 + au_plink_append(inode, bdst, h_dst);
3895 + goto out; /* success */
3900 + h_path.dentry = h_parent;
3901 + mutex_unlock(&dst_inode->i_mutex);
3902 + au_dtime_store(&dt, dst_parent, &h_path);
3903 + h_path.dentry = h_dst;
3905 + rerr = vfsub_unlink(h_dir, &h_path, /*force*/0);
3907 + rerr = vfsub_rmdir(h_dir, &h_path);
3908 + au_dtime_revert(&dt);
3910 + AuIOErr("failed removing broken entry(%d, %d)\n", err, rerr);
3919 +struct au_cpup_single_args {
3921 + struct dentry *dentry;
3922 + aufs_bindex_t bdst, bsrc;
3924 + unsigned int flags;
3925 + struct dentry *dst_parent;
3928 +static void au_call_cpup_single(void *args)
3930 + struct au_cpup_single_args *a = args;
3931 + *a->errp = au_cpup_single(a->dentry, a->bdst, a->bsrc, a->len,
3932 + a->flags, a->dst_parent);
3936 + * prevent SIGXFSZ in copy-up.
3937 + * testing CAP_MKNOD is for generic fs,
3938 + * but CAP_FSETID is for xfs only, currently.
3940 +static int au_cpup_sio_test(struct super_block *sb, umode_t mode)
3945 + if (!au_wkq_test()
3946 + && (!au_sbi(sb)->si_plink_maint_pid
3947 + || au_plink_maint(sb, AuLock_NOPLM))) {
3948 + switch (mode & S_IFMT) {
3950 + /* no condition about RLIMIT_FSIZE and the file size */
3955 + do_sio = !capable(CAP_MKNOD);
3959 + do_sio = ((mode & (S_ISUID | S_ISGID))
3960 + && !capable(CAP_FSETID));
3966 +int au_sio_cpup_single(struct dentry *dentry, aufs_bindex_t bdst,
3967 + aufs_bindex_t bsrc, loff_t len, unsigned int flags,
3968 + struct dentry *dst_parent)
3971 + struct dentry *h_dentry;
3973 + h_dentry = au_h_dptr(dentry, bsrc);
3974 + if (!au_cpup_sio_test(dentry->d_sb, h_dentry->d_inode->i_mode))
3975 + err = au_cpup_single(dentry, bdst, bsrc, len, flags,
3978 + struct au_cpup_single_args args = {
3985 + .dst_parent = dst_parent
3987 + wkq_err = au_wkq_wait(au_call_cpup_single, &args);
3988 + if (unlikely(wkq_err))
3996 + * copyup the @dentry from the first active lower branch to @bdst,
3997 + * using au_cpup_single().
3999 +static int au_cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len,
4000 + unsigned int flags)
4003 + aufs_bindex_t bsrc, bend;
4005 + bend = au_dbend(dentry);
4006 + for (bsrc = bdst + 1; bsrc <= bend; bsrc++)
4007 + if (au_h_dptr(dentry, bsrc))
4010 + err = au_lkup_neg(dentry, bdst);
4012 + err = au_cpup_single(dentry, bdst, bsrc, len, flags, NULL);
4014 + return 0; /* success */
4017 + au_set_h_dptr(dentry, bdst, NULL);
4018 + au_set_dbstart(dentry, bsrc);
4024 +struct au_cpup_simple_args {
4026 + struct dentry *dentry;
4027 + aufs_bindex_t bdst;
4029 + unsigned int flags;
4032 +static void au_call_cpup_simple(void *args)
4034 + struct au_cpup_simple_args *a = args;
4035 + *a->errp = au_cpup_simple(a->dentry, a->bdst, a->len, a->flags);
4038 +int au_sio_cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len,
4039 + unsigned int flags)
4042 + struct dentry *parent;
4043 + struct inode *h_dir;
4045 + parent = dget_parent(dentry);
4046 + h_dir = au_h_iptr(parent->d_inode, bdst);
4047 + if (!au_test_h_perm_sio(h_dir, MAY_EXEC | MAY_WRITE)
4048 + && !au_cpup_sio_test(dentry->d_sb, dentry->d_inode->i_mode))
4049 + err = au_cpup_simple(dentry, bdst, len, flags);
4051 + struct au_cpup_simple_args args = {
4058 + wkq_err = au_wkq_wait(au_call_cpup_simple, &args);
4059 + if (unlikely(wkq_err))
4067 +/* ---------------------------------------------------------------------- */
4070 + * copyup the deleted file for writing.
4072 +static int au_do_cpup_wh(struct dentry *dentry, aufs_bindex_t bdst,
4073 + struct dentry *wh_dentry, struct file *file,
4077 + aufs_bindex_t bstart;
4078 + struct au_dinfo *dinfo;
4079 + struct dentry *h_d_dst, *h_d_start;
4080 + struct au_hdentry *hdp;
4082 + dinfo = au_di(dentry);
4083 + AuRwMustWriteLock(&dinfo->di_rwsem);
4085 + bstart = dinfo->di_bstart;
4086 + hdp = dinfo->di_hdentry;
4087 + h_d_dst = hdp[0 + bdst].hd_dentry;
4088 + dinfo->di_bstart = bdst;
4089 + hdp[0 + bdst].hd_dentry = wh_dentry;
4091 + h_d_start = hdp[0 + bstart].hd_dentry;
4092 + hdp[0 + bstart].hd_dentry = au_hf_top(file)->f_dentry;
4094 + err = au_cpup_single(dentry, bdst, bstart, len, !AuCpup_DTIME,
4095 + /*h_parent*/NULL);
4098 + err = au_reopen_nondir(file);
4099 + hdp[0 + bstart].hd_dentry = h_d_start;
4101 + hdp[0 + bdst].hd_dentry = h_d_dst;
4102 + dinfo->di_bstart = bstart;
4107 +static int au_cpup_wh(struct dentry *dentry, aufs_bindex_t bdst, loff_t len,
4108 + struct file *file)
4111 + struct au_dtime dt;
4112 + struct dentry *parent, *h_parent, *wh_dentry;
4113 + struct au_branch *br;
4114 + struct path h_path;
4116 + br = au_sbr(dentry->d_sb, bdst);
4117 + parent = dget_parent(dentry);
4118 + h_parent = au_h_dptr(parent, bdst);
4119 + wh_dentry = au_whtmp_lkup(h_parent, br, &dentry->d_name);
4120 + err = PTR_ERR(wh_dentry);
4121 + if (IS_ERR(wh_dentry))
4124 + h_path.dentry = h_parent;
4125 + h_path.mnt = br->br_mnt;
4126 + au_dtime_store(&dt, parent, &h_path);
4127 + err = au_do_cpup_wh(dentry, bdst, wh_dentry, file, len);
4128 + if (unlikely(err))
4132 + h_path.dentry = wh_dentry;
4133 + if (!S_ISDIR(wh_dentry->d_inode->i_mode))
4134 + err = vfsub_unlink(h_parent->d_inode, &h_path, /*force*/0);
4136 + err = vfsub_rmdir(h_parent->d_inode, &h_path);
4137 + if (unlikely(err)) {
4138 + AuIOErr("failed remove copied-up tmp file %.*s(%d)\n",
4139 + AuDLNPair(wh_dentry), err);
4142 + au_dtime_revert(&dt);
4143 + au_set_hi_wh(dentry->d_inode, bdst, wh_dentry);
4152 +struct au_cpup_wh_args {
4154 + struct dentry *dentry;
4155 + aufs_bindex_t bdst;
4157 + struct file *file;
4160 +static void au_call_cpup_wh(void *args)
4162 + struct au_cpup_wh_args *a = args;
4163 + *a->errp = au_cpup_wh(a->dentry, a->bdst, a->len, a->file);
4166 +int au_sio_cpup_wh(struct dentry *dentry, aufs_bindex_t bdst, loff_t len,
4167 + struct file *file)
4170 + struct dentry *parent, *h_orph, *h_parent, *h_dentry;
4171 + struct inode *dir, *h_dir, *h_tmpdir, *h_inode;
4172 + struct au_wbr *wbr;
4174 + parent = dget_parent(dentry);
4175 + dir = parent->d_inode;
4178 + h_dir = au_igrab(au_h_iptr(dir, bdst));
4180 + if (!h_dir->i_nlink) {
4181 + wbr = au_sbr(dentry->d_sb, bdst)->br_wbr;
4182 + h_orph = wbr->wbr_orph;
4184 + h_parent = dget(au_h_dptr(parent, bdst));
4185 + au_set_h_dptr(parent, bdst, dget(h_orph));
4186 + h_tmpdir = h_orph->d_inode;
4187 + au_set_h_iptr(dir, bdst, au_igrab(h_tmpdir), /*flags*/0);
4189 + /* this temporary unlock is safe */
4191 + h_dentry = au_hf_top(file)->f_dentry;
4193 + h_dentry = au_h_dptr(dentry, au_dbstart(dentry));
4194 + h_inode = h_dentry->d_inode;
4195 + IMustLock(h_inode);
4196 + mutex_unlock(&h_inode->i_mutex);
4197 + mutex_lock_nested(&h_tmpdir->i_mutex, AuLsc_I_PARENT3);
4198 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
4199 + /* todo: au_h_open_pre()? */
4202 + if (!au_test_h_perm_sio(h_tmpdir, MAY_EXEC | MAY_WRITE)
4203 + && !au_cpup_sio_test(dentry->d_sb, dentry->d_inode->i_mode))
4204 + err = au_cpup_wh(dentry, bdst, len, file);
4206 + struct au_cpup_wh_args args = {
4213 + wkq_err = au_wkq_wait(au_call_cpup_wh, &args);
4214 + if (unlikely(wkq_err))
4219 + mutex_unlock(&h_tmpdir->i_mutex);
4220 + /* todo: au_h_open_post()? */
4221 + au_set_h_iptr(dir, bdst, au_igrab(h_dir), /*flags*/0);
4222 + au_set_h_dptr(parent, bdst, h_parent);
4230 +/* ---------------------------------------------------------------------- */
4233 + * generic routine for both of copy-up and copy-down.
4235 +/* cf. revalidate function in file.c */
4236 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
4237 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
4238 + struct dentry *h_parent, void *arg),
4242 + struct au_pin pin;
4243 + struct dentry *d, *parent, *h_parent, *real_parent;
4246 + parent = dget_parent(dentry);
4247 + if (IS_ROOT(parent))
4250 + au_pin_init(&pin, dentry, bdst, AuLsc_DI_PARENT2, AuLsc_I_PARENT2,
4251 + au_opt_udba(dentry->d_sb), AuPin_MNT_WRITE);
4253 + /* do not use au_dpage */
4254 + real_parent = parent;
4257 + parent = dget_parent(dentry);
4258 + h_parent = au_h_dptr(parent, bdst);
4260 + goto out; /* success */
4262 + /* find top dir which is necessary to cpup */
4266 + parent = dget_parent(d);
4267 + di_read_lock_parent3(parent, !AuLock_IR);
4268 + h_parent = au_h_dptr(parent, bdst);
4269 + di_read_unlock(parent, !AuLock_IR);
4270 + } while (!h_parent);
4272 + if (d != real_parent)
4273 + di_write_lock_child3(d);
4275 + /* somebody else might create while we were sleeping */
4276 + if (!au_h_dptr(d, bdst) || !au_h_dptr(d, bdst)->d_inode) {
4277 + if (au_h_dptr(d, bdst))
4278 + au_update_dbstart(d);
4280 + au_pin_set_dentry(&pin, d);
4281 + err = au_do_pin(&pin);
4283 + err = cp(d, bdst, h_parent, arg);
4288 + if (d != real_parent)
4289 + di_write_unlock(d);
4290 + if (unlikely(err))
4299 +static int au_cpup_dir(struct dentry *dentry, aufs_bindex_t bdst,
4300 + struct dentry *h_parent __maybe_unused ,
4301 + void *arg __maybe_unused)
4303 + return au_sio_cpup_simple(dentry, bdst, -1, AuCpup_DTIME);
4306 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
4308 + return au_cp_dirs(dentry, bdst, au_cpup_dir, NULL);
4311 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst)
4314 + struct dentry *parent;
4315 + struct inode *dir;
4317 + parent = dget_parent(dentry);
4318 + dir = parent->d_inode;
4320 + if (au_h_iptr(dir, bdst))
4323 + di_read_unlock(parent, AuLock_IR);
4324 + di_write_lock_parent(parent);
4325 + /* someone else might change our inode while we were sleeping */
4326 + if (!au_h_iptr(dir, bdst))
4327 + err = au_cpup_dirs(dentry, bdst);
4328 + di_downgrade_lock(parent, AuLock_IR);
4334 diff -urN /usr/share/empty/fs/aufs/cpup.h linux/fs/aufs/cpup.h
4335 --- /usr/share/empty/fs/aufs/cpup.h 1970-01-01 01:00:00.000000000 +0100
4336 +++ linux/fs/aufs/cpup.h 2011-08-24 13:30:24.731313534 +0200
4339 + * Copyright (C) 2005-2011 Junjiro R. Okajima
4341 + * This program, aufs is free software; you can redistribute it and/or modify
4342 + * it under the terms of the GNU General Public License as published by
4343 + * the Free Software Foundation; either version 2 of the License, or
4344 + * (at your option) any later version.
4346 + * This program is distributed in the hope that it will be useful,
4347 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4348 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4349 + * GNU General Public License for more details.
4351 + * You should have received a copy of the GNU General Public License
4352 + * along with this program; if not, write to the Free Software
4353 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
4357 + * copy-up/down functions
4360 +#ifndef __AUFS_CPUP_H__
4361 +#define __AUFS_CPUP_H__
4365 +#include <linux/path.h>
4366 +#include <linux/time.h>
4367 +#include <linux/aufs_type.h>
4372 +void au_cpup_attr_flags(struct inode *dst, struct inode *src);
4373 +void au_cpup_attr_timesizes(struct inode *inode);
4374 +void au_cpup_attr_nlink(struct inode *inode, int force);
4375 +void au_cpup_attr_changeable(struct inode *inode);
4376 +void au_cpup_igen(struct inode *inode, struct inode *h_inode);
4377 +void au_cpup_attr_all(struct inode *inode, int force);
4379 +/* ---------------------------------------------------------------------- */
4382 +#define AuCpup_DTIME 1 /* do dtime_store/revert */
4383 +#define AuCpup_KEEPLINO (1 << 1) /* do not clear the lower xino,
4385 +#define au_ftest_cpup(flags, name) ((flags) & AuCpup_##name)
4386 +#define au_fset_cpup(flags, name) \
4387 + do { (flags) |= AuCpup_##name; } while (0)
4388 +#define au_fclr_cpup(flags, name) \
4389 + do { (flags) &= ~AuCpup_##name; } while (0)
4391 +int au_copy_file(struct file *dst, struct file *src, loff_t len);
4392 +int au_sio_cpup_single(struct dentry *dentry, aufs_bindex_t bdst,
4393 + aufs_bindex_t bsrc, loff_t len, unsigned int flags,
4394 + struct dentry *dst_parent);
4395 +int au_sio_cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len,
4396 + unsigned int flags);
4397 +int au_sio_cpup_wh(struct dentry *dentry, aufs_bindex_t bdst, loff_t len,
4398 + struct file *file);
4400 +int au_cp_dirs(struct dentry *dentry, aufs_bindex_t bdst,
4401 + int (*cp)(struct dentry *dentry, aufs_bindex_t bdst,
4402 + struct dentry *h_parent, void *arg),
4404 +int au_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
4405 +int au_test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst);
4407 +/* ---------------------------------------------------------------------- */
4409 +/* keep timestamps when copyup */
4411 + struct dentry *dt_dentry;
4412 + struct path dt_h_path;
4413 + struct timespec dt_atime, dt_mtime;
4415 +void au_dtime_store(struct au_dtime *dt, struct dentry *dentry,
4416 + struct path *h_path);
4417 +void au_dtime_revert(struct au_dtime *dt);
4419 +#endif /* __KERNEL__ */
4420 +#endif /* __AUFS_CPUP_H__ */
4421 diff -urN /usr/share/empty/fs/aufs/dbgaufs.c linux/fs/aufs/dbgaufs.c
4422 --- /usr/share/empty/fs/aufs/dbgaufs.c 1970-01-01 01:00:00.000000000 +0100
4423 +++ linux/fs/aufs/dbgaufs.c 2012-01-05 11:00:59.215501779 +0100
4426 + * Copyright (C) 2005-2011 Junjiro R. Okajima
4428 + * This program, aufs is free software; you can redistribute it and/or modify
4429 + * it under the terms of the GNU General Public License as published by
4430 + * the Free Software Foundation; either version 2 of the License, or
4431 + * (at your option) any later version.
4433 + * This program is distributed in the hope that it will be useful,
4434 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4435 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4436 + * GNU General Public License for more details.
4438 + * You should have received a copy of the GNU General Public License
4439 + * along with this program; if not, write to the Free Software
4440 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
4444 + * debugfs interface
4447 +#include <linux/debugfs.h>
4450 +#ifndef CONFIG_SYSFS
4451 +#error DEBUG_FS depends upon SYSFS
4454 +static struct dentry *dbgaufs;
4455 +static const mode_t dbgaufs_mode = S_IRUSR | S_IRGRP | S_IROTH;
4457 +/* 20 is max digits length of ulong 64 */
4458 +struct dbgaufs_arg {
4464 + * common function for all XINO files
4466 +static int dbgaufs_xi_release(struct inode *inode __maybe_unused,
4467 + struct file *file)
4469 + kfree(file->private_data);
4473 +static int dbgaufs_xi_open(struct file *xf, struct file *file, int do_fcnt)
4477 + struct dbgaufs_arg *p;
4480 + p = kmalloc(sizeof(*p), GFP_NOFS);
4486 + file->private_data = p;
4490 + err = vfs_getattr(xf->f_vfsmnt, xf->f_dentry, &st);
4494 + (p->a, sizeof(p->a), "%ld, %llux%lu %lld\n",
4495 + (long)file_count(xf), st.blocks, st.blksize,
4496 + (long long)st.size);
4498 + p->n = snprintf(p->a, sizeof(p->a), "%llux%lu %lld\n",
4499 + st.blocks, st.blksize,
4500 + (long long)st.size);
4501 + AuDebugOn(p->n >= sizeof(p->a));
4503 + p->n = snprintf(p->a, sizeof(p->a), "err %d\n", err);
4512 +static ssize_t dbgaufs_xi_read(struct file *file, char __user *buf,
4513 + size_t count, loff_t *ppos)
4515 + struct dbgaufs_arg *p;
4517 + p = file->private_data;
4518 + return simple_read_from_buffer(buf, count, ppos, p->a, p->n);
4521 +/* ---------------------------------------------------------------------- */
4523 +static int dbgaufs_xib_open(struct inode *inode, struct file *file)
4526 + struct au_sbinfo *sbinfo;
4527 + struct super_block *sb;
4529 + sbinfo = inode->i_private;
4530 + sb = sbinfo->si_sb;
4531 + si_noflush_read_lock(sb);
4532 + err = dbgaufs_xi_open(sbinfo->si_xib, file, /*do_fcnt*/0);
4533 + si_read_unlock(sb);
4537 +static const struct file_operations dbgaufs_xib_fop = {
4538 + .owner = THIS_MODULE,
4539 + .open = dbgaufs_xib_open,
4540 + .release = dbgaufs_xi_release,
4541 + .read = dbgaufs_xi_read
4544 +/* ---------------------------------------------------------------------- */
4546 +#define DbgaufsXi_PREFIX "xi"
4548 +static int dbgaufs_xino_open(struct inode *inode, struct file *file)
4552 + struct au_sbinfo *sbinfo;
4553 + struct super_block *sb;
4555 + struct qstr *name;
4559 + name = &file->f_dentry->d_name;
4560 + if (unlikely(name->len < sizeof(DbgaufsXi_PREFIX)
4561 + || memcmp(name->name, DbgaufsXi_PREFIX,
4562 + sizeof(DbgaufsXi_PREFIX) - 1)))
4564 + err = kstrtol(name->name + sizeof(DbgaufsXi_PREFIX) - 1, 10, &l);
4565 + if (unlikely(err))
4568 + sbinfo = inode->i_private;
4569 + sb = sbinfo->si_sb;
4570 + si_noflush_read_lock(sb);
4571 + if (l <= au_sbend(sb)) {
4572 + xf = au_sbr(sb, (aufs_bindex_t)l)->br_xino.xi_file;
4573 + err = dbgaufs_xi_open(xf, file, /*do_fcnt*/1);
4576 + si_read_unlock(sb);
4582 +static const struct file_operations dbgaufs_xino_fop = {
4583 + .owner = THIS_MODULE,
4584 + .open = dbgaufs_xino_open,
4585 + .release = dbgaufs_xi_release,
4586 + .read = dbgaufs_xi_read
4589 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
4591 + aufs_bindex_t bend;
4592 + struct au_branch *br;
4593 + struct au_xino_file *xi;
4595 + if (!au_sbi(sb)->si_dbgaufs)
4598 + bend = au_sbend(sb);
4599 + for (; bindex <= bend; bindex++) {
4600 + br = au_sbr(sb, bindex);
4601 + xi = &br->br_xino;
4602 + if (xi->xi_dbgaufs) {
4603 + debugfs_remove(xi->xi_dbgaufs);
4604 + xi->xi_dbgaufs = NULL;
4609 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
4611 + struct au_sbinfo *sbinfo;
4612 + struct dentry *parent;
4613 + struct au_branch *br;
4614 + struct au_xino_file *xi;
4615 + aufs_bindex_t bend;
4616 + char name[sizeof(DbgaufsXi_PREFIX) + 5]; /* "xi" bindex NULL */
4618 + sbinfo = au_sbi(sb);
4619 + parent = sbinfo->si_dbgaufs;
4623 + bend = au_sbend(sb);
4624 + for (; bindex <= bend; bindex++) {
4625 + snprintf(name, sizeof(name), DbgaufsXi_PREFIX "%d", bindex);
4626 + br = au_sbr(sb, bindex);
4627 + xi = &br->br_xino;
4628 + AuDebugOn(xi->xi_dbgaufs);
4629 + xi->xi_dbgaufs = debugfs_create_file(name, dbgaufs_mode, parent,
4630 + sbinfo, &dbgaufs_xino_fop);
4631 + /* ignore an error */
4632 + if (unlikely(!xi->xi_dbgaufs))
4633 + AuWarn1("failed %s under debugfs\n", name);
4637 +/* ---------------------------------------------------------------------- */
4639 +#ifdef CONFIG_AUFS_EXPORT
4640 +static int dbgaufs_xigen_open(struct inode *inode, struct file *file)
4643 + struct au_sbinfo *sbinfo;
4644 + struct super_block *sb;
4646 + sbinfo = inode->i_private;
4647 + sb = sbinfo->si_sb;
4648 + si_noflush_read_lock(sb);
4649 + err = dbgaufs_xi_open(sbinfo->si_xigen, file, /*do_fcnt*/0);
4650 + si_read_unlock(sb);
4654 +static const struct file_operations dbgaufs_xigen_fop = {
4655 + .owner = THIS_MODULE,
4656 + .open = dbgaufs_xigen_open,
4657 + .release = dbgaufs_xi_release,
4658 + .read = dbgaufs_xi_read
4661 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
4666 + * This function is a dynamic '__init' fucntion actually,
4667 + * so the tiny check for si_rwsem is unnecessary.
4669 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
4672 + sbinfo->si_dbgaufs_xigen = debugfs_create_file
4673 + ("xigen", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
4674 + &dbgaufs_xigen_fop);
4675 + if (sbinfo->si_dbgaufs_xigen)
4681 +static int dbgaufs_xigen_init(struct au_sbinfo *sbinfo)
4685 +#endif /* CONFIG_AUFS_EXPORT */
4687 +/* ---------------------------------------------------------------------- */
4689 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo)
4692 + * This function is a dynamic '__init' fucntion actually,
4693 + * so the tiny check for si_rwsem is unnecessary.
4695 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
4697 + debugfs_remove_recursive(sbinfo->si_dbgaufs);
4698 + sbinfo->si_dbgaufs = NULL;
4699 + kobject_put(&sbinfo->si_kobj);
4702 +int dbgaufs_si_init(struct au_sbinfo *sbinfo)
4705 + char name[SysaufsSiNameLen];
4708 + * This function is a dynamic '__init' fucntion actually,
4709 + * so the tiny check for si_rwsem is unnecessary.
4711 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
4715 + AuErr1("/debug/aufs is uninitialized\n");
4720 + sysaufs_name(sbinfo, name);
4721 + sbinfo->si_dbgaufs = debugfs_create_dir(name, dbgaufs);
4722 + if (unlikely(!sbinfo->si_dbgaufs))
4724 + kobject_get(&sbinfo->si_kobj);
4726 + sbinfo->si_dbgaufs_xib = debugfs_create_file
4727 + ("xib", dbgaufs_mode, sbinfo->si_dbgaufs, sbinfo,
4728 + &dbgaufs_xib_fop);
4729 + if (unlikely(!sbinfo->si_dbgaufs_xib))
4732 + err = dbgaufs_xigen_init(sbinfo);
4734 + goto out; /* success */
4737 + dbgaufs_si_fin(sbinfo);
4742 +/* ---------------------------------------------------------------------- */
4744 +void dbgaufs_fin(void)
4746 + debugfs_remove(dbgaufs);
4749 +int __init dbgaufs_init(void)
4754 + dbgaufs = debugfs_create_dir(AUFS_NAME, NULL);
4759 diff -urN /usr/share/empty/fs/aufs/dbgaufs.h linux/fs/aufs/dbgaufs.h
4760 --- /usr/share/empty/fs/aufs/dbgaufs.h 1970-01-01 01:00:00.000000000 +0100
4761 +++ linux/fs/aufs/dbgaufs.h 2011-08-24 13:30:24.731313534 +0200
4764 + * Copyright (C) 2005-2011 Junjiro R. Okajima
4766 + * This program, aufs is free software; you can redistribute it and/or modify
4767 + * it under the terms of the GNU General Public License as published by
4768 + * the Free Software Foundation; either version 2 of the License, or
4769 + * (at your option) any later version.
4771 + * This program is distributed in the hope that it will be useful,
4772 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4773 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4774 + * GNU General Public License for more details.
4776 + * You should have received a copy of the GNU General Public License
4777 + * along with this program; if not, write to the Free Software
4778 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
4782 + * debugfs interface
4785 +#ifndef __DBGAUFS_H__
4786 +#define __DBGAUFS_H__
4790 +#include <linux/init.h>
4791 +#include <linux/aufs_type.h>
4793 +struct super_block;
4796 +#ifdef CONFIG_DEBUG_FS
4798 +void dbgaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
4799 +void dbgaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
4800 +void dbgaufs_si_fin(struct au_sbinfo *sbinfo);
4801 +int dbgaufs_si_init(struct au_sbinfo *sbinfo);
4802 +void dbgaufs_fin(void);
4803 +int __init dbgaufs_init(void);
4805 +AuStubVoid(dbgaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
4806 +AuStubVoid(dbgaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
4807 +AuStubVoid(dbgaufs_si_fin, struct au_sbinfo *sbinfo)
4808 +AuStubInt0(dbgaufs_si_init, struct au_sbinfo *sbinfo)
4809 +AuStubVoid(dbgaufs_fin, void)
4810 +AuStubInt0(__init dbgaufs_init, void)
4811 +#endif /* CONFIG_DEBUG_FS */
4813 +#endif /* __KERNEL__ */
4814 +#endif /* __DBGAUFS_H__ */
4815 diff -urN /usr/share/empty/fs/aufs/dcsub.c linux/fs/aufs/dcsub.c
4816 --- /usr/share/empty/fs/aufs/dcsub.c 1970-01-01 01:00:00.000000000 +0100
4817 +++ linux/fs/aufs/dcsub.c 2011-08-24 13:30:24.731313534 +0200
4820 + * Copyright (C) 2005-2011 Junjiro R. Okajima
4822 + * This program, aufs is free software; you can redistribute it and/or modify
4823 + * it under the terms of the GNU General Public License as published by
4824 + * the Free Software Foundation; either version 2 of the License, or
4825 + * (at your option) any later version.
4827 + * This program is distributed in the hope that it will be useful,
4828 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
4829 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4830 + * GNU General Public License for more details.
4832 + * You should have received a copy of the GNU General Public License
4833 + * along with this program; if not, write to the Free Software
4834 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
4838 + * sub-routines for dentry cache
4843 +static void au_dpage_free(struct au_dpage *dpage)
4846 + struct dentry **p;
4848 + p = dpage->dentries;
4849 + for (i = 0; i < dpage->ndentry; i++)
4851 + free_page((unsigned long)dpage->dentries);
4854 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp)
4860 + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp);
4861 + if (unlikely(!dpages->dpages))
4864 + p = (void *)__get_free_page(gfp);
4868 + dpages->dpages[0].ndentry = 0;
4869 + dpages->dpages[0].dentries = p;
4870 + dpages->ndpage = 1;
4871 + return 0; /* success */
4874 + kfree(dpages->dpages);
4879 +void au_dpages_free(struct au_dcsub_pages *dpages)
4882 + struct au_dpage *p;
4884 + p = dpages->dpages;
4885 + for (i = 0; i < dpages->ndpage; i++)
4886 + au_dpage_free(p++);
4887 + kfree(dpages->dpages);
4890 +static int au_dpages_append(struct au_dcsub_pages *dpages,
4891 + struct dentry *dentry, gfp_t gfp)
4894 + struct au_dpage *dpage;
4897 + dpage = dpages->dpages + dpages->ndpage - 1;
4898 + sz = PAGE_SIZE / sizeof(dentry);
4899 + if (unlikely(dpage->ndentry >= sz)) {
4900 + AuLabel(new dpage);
4902 + sz = dpages->ndpage * sizeof(*dpages->dpages);
4903 + p = au_kzrealloc(dpages->dpages, sz,
4904 + sz + sizeof(*dpages->dpages), gfp);
4908 + dpages->dpages = p;
4909 + dpage = dpages->dpages + dpages->ndpage;
4910 + p = (void *)__get_free_page(gfp);
4914 + dpage->ndentry = 0;
4915 + dpage->dentries = p;
4919 + AuDebugOn(!dentry->d_count);
4920 + dpage->dentries[dpage->ndentry++] = dget_dlock(dentry);
4921 + return 0; /* success */
4927 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
4928 + au_dpages_test test, void *arg)
4931 + struct dentry *this_parent;
4932 + struct list_head *next;
4933 + struct super_block *sb = root->d_sb;
4936 + write_seqlock(&rename_lock);
4937 + this_parent = root;
4938 + spin_lock(&this_parent->d_lock);
4940 + next = this_parent->d_subdirs.next;
4942 + if (this_parent->d_sb == sb
4943 + && !IS_ROOT(this_parent)
4944 + && au_di(this_parent)
4945 + && this_parent->d_count
4946 + && (!test || test(this_parent, arg))) {
4947 + err = au_dpages_append(dpages, this_parent, GFP_ATOMIC);
4948 + if (unlikely(err))
4952 + while (next != &this_parent->d_subdirs) {
4953 + struct list_head *tmp = next;
4954 + struct dentry *dentry = list_entry(tmp, struct dentry,
4958 + spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
4959 + if (dentry->d_count) {
4960 + if (!list_empty(&dentry->d_subdirs)) {
4961 + spin_unlock(&this_parent->d_lock);
4962 + spin_release(&dentry->d_lock.dep_map, 1,
4964 + this_parent = dentry;
4965 + spin_acquire(&this_parent->d_lock.dep_map, 0, 1,
4969 + if (dentry->d_sb == sb
4971 + && (!test || test(dentry, arg)))
4972 + err = au_dpages_append(dpages, dentry,
4975 + spin_unlock(&dentry->d_lock);
4976 + if (unlikely(err))
4980 + if (this_parent != root) {
4981 + struct dentry *tmp;
4982 + struct dentry *child;
4984 + tmp = this_parent->d_parent;
4986 + spin_unlock(&this_parent->d_lock);
4987 + child = this_parent;
4988 + this_parent = tmp;
4989 + spin_lock(&this_parent->d_lock);
4990 + rcu_read_unlock();
4991 + next = child->d_u.d_child.next;
4996 + spin_unlock(&this_parent->d_lock);
4997 + write_sequnlock(&rename_lock);
5001 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
5002 + int do_include, au_dpages_test test, void *arg)
5007 + write_seqlock(&rename_lock);
5008 + spin_lock(&dentry->d_lock);
5010 + && dentry->d_count
5011 + && (!test || test(dentry, arg)))
5012 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
5013 + spin_unlock(&dentry->d_lock);
5014 + if (unlikely(err))
5018 + * vfsmount_lock is unnecessary since this is a traverse in a single
5021 + while (!IS_ROOT(dentry)) {
5022 + dentry = dentry->d_parent; /* rename_lock is locked */
5023 + spin_lock(&dentry->d_lock);
5024 + if (dentry->d_count
5025 + && (!test || test(dentry, arg)))
5026 + err = au_dpages_append(dpages, dentry, GFP_ATOMIC);
5027 + spin_unlock(&dentry->d_lock);
5028 + if (unlikely(err))
5033 + write_sequnlock(&rename_lock);
5037 +static inline int au_dcsub_dpages_aufs(struct dentry *dentry, void *arg)
5039 + return au_di(dentry) && dentry->d_sb == arg;
5042 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
5043 + struct dentry *dentry, int do_include)
5045 + return au_dcsub_pages_rev(dpages, dentry, do_include,
5046 + au_dcsub_dpages_aufs, dentry->d_sb);
5049 +int au_test_subdir(struct dentry *d1, struct dentry *d2)
5051 + struct path path[2] = {
5060 + return path_is_under(path + 0, path + 1);
5062 diff -urN /usr/share/empty/fs/aufs/dcsub.h linux/fs/aufs/dcsub.h
5063 --- /usr/share/empty/fs/aufs/dcsub.h 1970-01-01 01:00:00.000000000 +0100
5064 +++ linux/fs/aufs/dcsub.h 2011-08-24 13:30:24.731313534 +0200
5067 + * Copyright (C) 2005-2011 Junjiro R. Okajima
5069 + * This program, aufs is free software; you can redistribute it and/or modify
5070 + * it under the terms of the GNU General Public License as published by
5071 + * the Free Software Foundation; either version 2 of the License, or
5072 + * (at your option) any later version.
5074 + * This program is distributed in the hope that it will be useful,
5075 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5076 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5077 + * GNU General Public License for more details.
5079 + * You should have received a copy of the GNU General Public License
5080 + * along with this program; if not, write to the Free Software
5081 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
5085 + * sub-routines for dentry cache
5088 +#ifndef __AUFS_DCSUB_H__
5089 +#define __AUFS_DCSUB_H__
5093 +#include <linux/dcache.h>
5094 +#include <linux/fs.h>
5095 +#include <linux/types.h>
5101 + struct dentry **dentries;
5104 +struct au_dcsub_pages {
5106 + struct au_dpage *dpages;
5109 +/* ---------------------------------------------------------------------- */
5112 +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp);
5113 +void au_dpages_free(struct au_dcsub_pages *dpages);
5114 +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg);
5115 +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root,
5116 + au_dpages_test test, void *arg);
5117 +int au_dcsub_pages_rev(struct au_dcsub_pages *dpages, struct dentry *dentry,
5118 + int do_include, au_dpages_test test, void *arg);
5119 +int au_dcsub_pages_rev_aufs(struct au_dcsub_pages *dpages,
5120 + struct dentry *dentry, int do_include);
5121 +int au_test_subdir(struct dentry *d1, struct dentry *d2);
5123 +/* ---------------------------------------------------------------------- */
5125 +static inline int au_d_hashed_positive(struct dentry *d)
5128 + struct inode *inode = d->d_inode;
5130 + if (unlikely(d_unhashed(d) || !inode || !inode->i_nlink))
5135 +static inline int au_d_alive(struct dentry *d)
5138 + struct inode *inode;
5141 + err = au_d_hashed_positive(d);
5143 + inode = d->d_inode;
5144 + if (unlikely(d_unlinked(d) || !inode || !inode->i_nlink))
5150 +static inline int au_alive_dir(struct dentry *d)
5153 + err = au_d_alive(d);
5154 + if (unlikely(err || IS_DEADDIR(d->d_inode)))
5159 +#endif /* __KERNEL__ */
5160 +#endif /* __AUFS_DCSUB_H__ */
5161 diff -urN /usr/share/empty/fs/aufs/debug.c linux/fs/aufs/debug.c
5162 --- /usr/share/empty/fs/aufs/debug.c 1970-01-01 01:00:00.000000000 +0100
5163 +++ linux/fs/aufs/debug.c 2011-10-24 20:51:51.580466925 +0200
5166 + * Copyright (C) 2005-2011 Junjiro R. Okajima
5168 + * This program, aufs is free software; you can redistribute it and/or modify
5169 + * it under the terms of the GNU General Public License as published by
5170 + * the Free Software Foundation; either version 2 of the License, or
5171 + * (at your option) any later version.
5173 + * This program is distributed in the hope that it will be useful,
5174 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5175 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5176 + * GNU General Public License for more details.
5178 + * You should have received a copy of the GNU General Public License
5179 + * along with this program; if not, write to the Free Software
5180 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
5184 + * debug print functions
5187 +#include <linux/module.h>
5188 +#include <linux/vt_kern.h>
5192 +MODULE_PARM_DESC(debug, "debug print");
5193 +module_param_named(debug, aufs_debug, int, S_IRUGO | S_IWUSR | S_IWGRP);
5195 +char *au_plevel = KERN_DEBUG;
5196 +#define dpri(fmt, ...) do { \
5198 + && strcmp(au_plevel, KERN_DEBUG)) \
5199 + || au_debug_test()) \
5200 + printk("%s" fmt, au_plevel, ##__VA_ARGS__); \
5203 +/* ---------------------------------------------------------------------- */
5205 +void au_dpri_whlist(struct au_nhash *whlist)
5207 + unsigned long ul, n;
5208 + struct hlist_head *head;
5209 + struct au_vdir_wh *tpos;
5210 + struct hlist_node *pos;
5212 + n = whlist->nh_num;
5213 + head = whlist->nh_head;
5214 + for (ul = 0; ul < n; ul++) {
5215 + hlist_for_each_entry(tpos, pos, head, wh_hash)
5216 + dpri("b%d, %.*s, %d\n",
5218 + tpos->wh_str.len, tpos->wh_str.name,
5219 + tpos->wh_str.len);
5224 +void au_dpri_vdir(struct au_vdir *vdir)
5227 + union au_vdir_deblk_p p;
5230 + if (!vdir || IS_ERR(vdir)) {
5231 + dpri("err %ld\n", PTR_ERR(vdir));
5235 + dpri("deblk %u, nblk %lu, deblk %p, last{%lu, %p}, ver %lu\n",
5236 + vdir->vd_deblk_sz, vdir->vd_nblk, vdir->vd_deblk,
5237 + vdir->vd_last.ul, vdir->vd_last.p.deblk, vdir->vd_version);
5238 + for (ul = 0; ul < vdir->vd_nblk; ul++) {
5239 + p.deblk = vdir->vd_deblk[ul];
5241 + dpri("[%lu]: %p\n", ul, o);
5245 +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode, int hn,
5246 + struct dentry *wh)
5251 + if (!inode || IS_ERR(inode)) {
5252 + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode));
5256 + /* the type of i_blocks depends upon CONFIG_LSF */
5257 + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long)
5258 + && sizeof(inode->i_blocks) != sizeof(u64));
5260 + n = (void *)wh->d_name.name;
5261 + l = wh->d_name.len;
5264 + dpri("i%d: %p, i%lu, %s, cnt %d, nl %u, 0%o, sz %llu, blk %llu,"
5265 + " hn %d, ct %lld, np %lu, st 0x%lx, f 0x%x, v %llu, g %x%s%.*s\n",
5267 + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??",
5268 + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode,
5269 + i_size_read(inode), (unsigned long long)inode->i_blocks,
5270 + hn, (long long)timespec_to_ns(&inode->i_ctime) & 0x0ffff,
5271 + inode->i_mapping ? inode->i_mapping->nrpages : 0,
5272 + inode->i_state, inode->i_flags, inode->i_version,
5273 + inode->i_generation,
5274 + l ? ", wh " : "", l, n);
5278 +void au_dpri_inode(struct inode *inode)
5280 + struct au_iinfo *iinfo;
5281 + aufs_bindex_t bindex;
5284 + err = do_pri_inode(-1, inode, -1, NULL);
5285 + if (err || !au_test_aufs(inode->i_sb))
5288 + iinfo = au_ii(inode);
5291 + dpri("i-1: bstart %d, bend %d, gen %d\n",
5292 + iinfo->ii_bstart, iinfo->ii_bend, au_iigen(inode));
5293 + if (iinfo->ii_bstart < 0)
5296 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; bindex++) {
5297 + hn = !!au_hn(iinfo->ii_hinode + bindex);
5298 + do_pri_inode(bindex, iinfo->ii_hinode[0 + bindex].hi_inode, hn,
5299 + iinfo->ii_hinode[0 + bindex].hi_whdentry);
5303 +void au_dpri_dalias(struct inode *inode)
5307 + spin_lock(&inode->i_lock);
5308 + list_for_each_entry(d, &inode->i_dentry, d_alias)
5309 + au_dpri_dentry(d);
5310 + spin_unlock(&inode->i_lock);
5313 +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry)
5315 + struct dentry *wh = NULL;
5318 + if (!dentry || IS_ERR(dentry)) {
5319 + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry));
5322 + /* do not call dget_parent() here */
5323 + /* note: access d_xxx without d_lock */
5324 + dpri("d%d: %.*s?/%.*s, %s, cnt %d, flags 0x%x\n",
5326 + AuDLNPair(dentry->d_parent), AuDLNPair(dentry),
5327 + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??",
5328 + dentry->d_count, dentry->d_flags);
5330 + if (bindex >= 0 && dentry->d_inode && au_test_aufs(dentry->d_sb)) {
5331 + struct au_iinfo *iinfo = au_ii(dentry->d_inode);
5333 + hn = !!au_hn(iinfo->ii_hinode + bindex);
5334 + wh = iinfo->ii_hinode[0 + bindex].hi_whdentry;
5337 + do_pri_inode(bindex, dentry->d_inode, hn, wh);
5341 +void au_dpri_dentry(struct dentry *dentry)
5343 + struct au_dinfo *dinfo;
5344 + aufs_bindex_t bindex;
5346 + struct au_hdentry *hdp;
5348 + err = do_pri_dentry(-1, dentry);
5349 + if (err || !au_test_aufs(dentry->d_sb))
5352 + dinfo = au_di(dentry);
5355 + dpri("d-1: bstart %d, bend %d, bwh %d, bdiropq %d, gen %d\n",
5356 + dinfo->di_bstart, dinfo->di_bend,
5357 + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry));
5358 + if (dinfo->di_bstart < 0)
5360 + hdp = dinfo->di_hdentry;
5361 + for (bindex = dinfo->di_bstart; bindex <= dinfo->di_bend; bindex++)
5362 + do_pri_dentry(bindex, hdp[0 + bindex].hd_dentry);
5365 +static int do_pri_file(aufs_bindex_t bindex, struct file *file)
5369 + if (!file || IS_ERR(file)) {
5370 + dpri("f%d: err %ld\n", bindex, PTR_ERR(file));
5376 + && au_test_aufs(file->f_dentry->d_sb)
5378 + snprintf(a, sizeof(a), ", gen %d, mmapped %d",
5379 + au_figen(file), atomic_read(&au_fi(file)->fi_mmapped));
5380 + dpri("f%d: mode 0x%x, flags 0%o, cnt %ld, v %llu, pos %llu%s\n",
5381 + bindex, file->f_mode, file->f_flags, (long)file_count(file),
5382 + file->f_version, file->f_pos, a);
5383 + if (file->f_dentry)
5384 + do_pri_dentry(bindex, file->f_dentry);
5388 +void au_dpri_file(struct file *file)
5390 + struct au_finfo *finfo;
5391 + struct au_fidir *fidir;
5392 + struct au_hfile *hfile;
5393 + aufs_bindex_t bindex;
5396 + err = do_pri_file(-1, file);
5397 + if (err || !file->f_dentry || !au_test_aufs(file->f_dentry->d_sb))
5400 + finfo = au_fi(file);
5403 + if (finfo->fi_btop < 0)
5405 + fidir = finfo->fi_hdir;
5407 + do_pri_file(finfo->fi_btop, finfo->fi_htop.hf_file);
5409 + for (bindex = finfo->fi_btop;
5410 + bindex >= 0 && bindex <= fidir->fd_bbot;
5412 + hfile = fidir->fd_hfile + bindex;
5413 + do_pri_file(bindex, hfile ? hfile->hf_file : NULL);
5417 +static int do_pri_br(aufs_bindex_t bindex, struct au_branch *br)
5419 + struct vfsmount *mnt;
5420 + struct super_block *sb;
5422 + if (!br || IS_ERR(br))
5425 + if (!mnt || IS_ERR(mnt))
5428 + if (!sb || IS_ERR(sb))
5431 + dpri("s%d: {perm 0x%x, id %d, cnt %d, wbr %p}, "
5432 + "%s, dev 0x%02x%02x, flags 0x%lx, cnt %d, active %d, "
5434 + bindex, br->br_perm, br->br_id, atomic_read(&br->br_count),
5435 + br->br_wbr, au_sbtype(sb), MAJOR(sb->s_dev), MINOR(sb->s_dev),
5436 + sb->s_flags, sb->s_count,
5437 + atomic_read(&sb->s_active), !!br->br_xino.xi_file);
5441 + dpri("s%d: err %ld\n", bindex, PTR_ERR(br));
5445 +void au_dpri_sb(struct super_block *sb)
5447 + struct au_sbinfo *sbinfo;
5448 + aufs_bindex_t bindex;
5450 + /* to reuduce stack size */
5452 + struct vfsmount mnt;
5453 + struct au_branch fake;
5456 + /* this function can be called from magic sysrq */
5457 + a = kzalloc(sizeof(*a), GFP_ATOMIC);
5458 + if (unlikely(!a)) {
5459 + dpri("no memory\n");
5463 + a->mnt.mnt_sb = sb;
5464 + a->fake.br_perm = 0;
5465 + a->fake.br_mnt = &a->mnt;
5466 + a->fake.br_xino.xi_file = NULL;
5467 + atomic_set(&a->fake.br_count, 0);
5468 + smp_mb(); /* atomic_set */
5469 + err = do_pri_br(-1, &a->fake);
5471 + dpri("dev 0x%x\n", sb->s_dev);
5472 + if (err || !au_test_aufs(sb))
5475 + sbinfo = au_sbi(sb);
5478 + dpri("nw %d, gen %u, kobj %d\n",
5479 + atomic_read(&sbinfo->si_nowait.nw_len), sbinfo->si_generation,
5480 + atomic_read(&sbinfo->si_kobj.kref.refcount));
5481 + for (bindex = 0; bindex <= sbinfo->si_bend; bindex++)
5482 + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]);
5485 +/* ---------------------------------------------------------------------- */
5487 +void au_dbg_sleep_jiffy(int jiffy)
5490 + jiffy = schedule_timeout_uninterruptible(jiffy);
5493 +void au_dbg_iattr(struct iattr *ia)
5495 +#define AuBit(name) if (ia->ia_valid & ATTR_ ## name) \
5515 + dpri("ia_file %p\n", ia->ia_file);
5518 +/* ---------------------------------------------------------------------- */
5520 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line)
5522 + struct inode *h_inode, *inode = dentry->d_inode;
5523 + struct dentry *h_dentry;
5524 + aufs_bindex_t bindex, bend, bi;
5526 + if (!inode /* || au_di(dentry)->di_lsc == AuLsc_DI_TMP */)
5529 + bend = au_dbend(dentry);
5530 + bi = au_ibend(inode);
5533 + bindex = au_dbstart(dentry);
5534 + bi = au_ibstart(inode);
5538 + for (; bindex <= bend; bindex++) {
5539 + h_dentry = au_h_dptr(dentry, bindex);
5542 + h_inode = au_h_iptr(inode, bindex);
5543 + if (unlikely(h_inode != h_dentry->d_inode)) {
5544 + int old = au_debug_test();
5547 + AuDbg("b%d, %s:%d\n", bindex, func, line);
5548 + AuDbgDentry(dentry);
5549 + AuDbgInode(inode);
5557 +void au_dbg_verify_dir_parent(struct dentry *dentry, unsigned int sigen)
5559 + struct dentry *parent;
5561 + parent = dget_parent(dentry);
5562 + AuDebugOn(!S_ISDIR(dentry->d_inode->i_mode));
5563 + AuDebugOn(IS_ROOT(dentry));
5564 + AuDebugOn(au_digen_test(parent, sigen));
5568 +void au_dbg_verify_nondir_parent(struct dentry *dentry, unsigned int sigen)
5570 + struct dentry *parent;
5571 + struct inode *inode;
5573 + parent = dget_parent(dentry);
5574 + inode = dentry->d_inode;
5575 + AuDebugOn(inode && S_ISDIR(dentry->d_inode->i_mode));
5576 + AuDebugOn(au_digen_test(parent, sigen));
5580 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen)
5583 + struct au_dcsub_pages dpages;
5584 + struct au_dpage *dpage;
5585 + struct dentry **dentries;
5587 + err = au_dpages_init(&dpages, GFP_NOFS);
5589 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/1);
5591 + for (i = dpages.ndpage - 1; !err && i >= 0; i--) {
5592 + dpage = dpages.dpages + i;
5593 + dentries = dpage->dentries;
5594 + for (j = dpage->ndentry - 1; !err && j >= 0; j--)
5595 + AuDebugOn(au_digen_test(dentries[j], sigen));
5597 + au_dpages_free(&dpages);
5600 +void au_dbg_verify_kthread(void)
5602 + if (au_wkq_test()) {
5605 + * It may be recursive, but udba=notify between two aufs mounts,
5606 + * where a single ro branch is shared, is not a problem.
5612 +/* ---------------------------------------------------------------------- */
5614 +void au_debug_sbinfo_init(struct au_sbinfo *sbinfo __maybe_unused)
5616 +#ifdef AuForceNoPlink
5617 + au_opt_clr(sbinfo->si_mntflags, PLINK);
5619 +#ifdef AuForceNoXino
5620 + au_opt_clr(sbinfo->si_mntflags, XINO);
5622 +#ifdef AuForceNoRefrof
5623 + au_opt_clr(sbinfo->si_mntflags, REFROF);
5625 +#ifdef AuForceHnotify
5626 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_HNOTIFY);
5629 + sbinfo->si_rdblk = 0;
5630 + sbinfo->si_rdhash = 0;
5634 +int __init au_debug_init(void)
5636 + aufs_bindex_t bindex;
5637 + struct au_vdir_destr destr;
5640 + AuDebugOn(bindex >= 0);
5643 + AuDebugOn(destr.len < NAME_MAX);
5645 +#ifdef CONFIG_4KSTACKS
5646 + pr_warning("CONFIG_4KSTACKS is defined.\n");
5649 +#ifdef AuForceNoBrs
5655 diff -urN /usr/share/empty/fs/aufs/debug.h linux/fs/aufs/debug.h
5656 --- /usr/share/empty/fs/aufs/debug.h 1970-01-01 01:00:00.000000000 +0100
5657 +++ linux/fs/aufs/debug.h 2011-08-24 13:30:24.731313534 +0200
5660 + * Copyright (C) 2005-2011 Junjiro R. Okajima
5662 + * This program, aufs is free software; you can redistribute it and/or modify
5663 + * it under the terms of the GNU General Public License as published by
5664 + * the Free Software Foundation; either version 2 of the License, or
5665 + * (at your option) any later version.
5667 + * This program is distributed in the hope that it will be useful,
5668 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5669 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5670 + * GNU General Public License for more details.
5672 + * You should have received a copy of the GNU General Public License
5673 + * along with this program; if not, write to the Free Software
5674 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
5678 + * debug print functions
5681 +#ifndef __AUFS_DEBUG_H__
5682 +#define __AUFS_DEBUG_H__
5686 +#include <asm/system.h>
5687 +#include <linux/bug.h>
5688 +/* #include <linux/err.h> */
5689 +#include <linux/init.h>
5690 +#include <linux/module.h>
5691 +#include <linux/kallsyms.h>
5692 +/* #include <linux/kernel.h> */
5693 +#include <linux/delay.h>
5694 +/* #include <linux/kd.h> */
5695 +#include <linux/sysrq.h>
5696 +#include <linux/aufs_type.h>
5698 +#include <asm/system.h>
5700 +#ifdef CONFIG_AUFS_DEBUG
5701 +#define AuDebugOn(a) BUG_ON(a)
5703 +/* module parameter */
5704 +extern int aufs_debug;
5705 +static inline void au_debug(int n)
5711 +static inline int au_debug_test(void)
5713 + return aufs_debug;
5716 +#define AuDebugOn(a) do {} while (0)
5717 +AuStubVoid(au_debug, int n)
5718 +AuStubInt0(au_debug_test, void)
5719 +#endif /* CONFIG_AUFS_DEBUG */
5721 +/* ---------------------------------------------------------------------- */
5725 +#define AuDbg(fmt, ...) do { \
5726 + if (au_debug_test()) \
5727 + pr_debug("DEBUG: " fmt, ##__VA_ARGS__); \
5729 +#define AuLabel(l) AuDbg(#l "\n")
5730 +#define AuIOErr(fmt, ...) pr_err("I/O Error, " fmt, ##__VA_ARGS__)
5731 +#define AuWarn1(fmt, ...) do { \
5732 + static unsigned char _c; \
5734 + pr_warning(fmt, ##__VA_ARGS__); \
5737 +#define AuErr1(fmt, ...) do { \
5738 + static unsigned char _c; \
5740 + pr_err(fmt, ##__VA_ARGS__); \
5743 +#define AuIOErr1(fmt, ...) do { \
5744 + static unsigned char _c; \
5746 + AuIOErr(fmt, ##__VA_ARGS__); \
5749 +#define AuUnsupportMsg "This operation is not supported." \
5750 + " Please report this application to aufs-users ML."
5751 +#define AuUnsupport(fmt, ...) do { \
5752 + pr_err(AuUnsupportMsg "\n" fmt, ##__VA_ARGS__); \
5756 +#define AuTraceErr(e) do { \
5757 + if (unlikely((e) < 0)) \
5758 + AuDbg("err %d\n", (int)(e)); \
5761 +#define AuTraceErrPtr(p) do { \
5763 + AuDbg("err %ld\n", PTR_ERR(p)); \
5766 +/* dirty macros for debug print, use with "%.*s" and caution */
5767 +#define AuLNPair(qstr) (qstr)->len, (qstr)->name
5768 +#define AuDLNPair(d) AuLNPair(&(d)->d_name)
5770 +/* ---------------------------------------------------------------------- */
5775 +#ifdef CONFIG_AUFS_DEBUG
5776 +extern char *au_plevel;
5778 +void au_dpri_whlist(struct au_nhash *whlist);
5780 +void au_dpri_vdir(struct au_vdir *vdir);
5782 +void au_dpri_inode(struct inode *inode);
5783 +void au_dpri_dalias(struct inode *inode);
5784 +void au_dpri_dentry(struct dentry *dentry);
5786 +void au_dpri_file(struct file *filp);
5787 +struct super_block;
5788 +void au_dpri_sb(struct super_block *sb);
5790 +void au_dbg_sleep_jiffy(int jiffy);
5792 +void au_dbg_iattr(struct iattr *ia);
5794 +#define au_dbg_verify_dinode(d) __au_dbg_verify_dinode(d, __func__, __LINE__)
5795 +void __au_dbg_verify_dinode(struct dentry *dentry, const char *func, int line);
5796 +void au_dbg_verify_dir_parent(struct dentry *dentry, unsigned int sigen);
5797 +void au_dbg_verify_nondir_parent(struct dentry *dentry, unsigned int sigen);
5798 +void au_dbg_verify_gen(struct dentry *parent, unsigned int sigen);
5799 +void au_dbg_verify_kthread(void);
5801 +int __init au_debug_init(void);
5802 +void au_debug_sbinfo_init(struct au_sbinfo *sbinfo);
5803 +#define AuDbgWhlist(w) do { \
5805 + au_dpri_whlist(w); \
5808 +#define AuDbgVdir(v) do { \
5810 + au_dpri_vdir(v); \
5813 +#define AuDbgInode(i) do { \
5815 + au_dpri_inode(i); \
5818 +#define AuDbgDAlias(i) do { \
5820 + au_dpri_dalias(i); \
5823 +#define AuDbgDentry(d) do { \
5825 + au_dpri_dentry(d); \
5828 +#define AuDbgFile(f) do { \
5830 + au_dpri_file(f); \
5833 +#define AuDbgSb(sb) do { \
5834 + AuDbg(#sb "\n"); \
5838 +#define AuDbgSleep(sec) do { \
5839 + AuDbg("sleep %d sec\n", sec); \
5843 +#define AuDbgSleepJiffy(jiffy) do { \
5844 + AuDbg("sleep %d jiffies\n", jiffy); \
5845 + au_dbg_sleep_jiffy(jiffy); \
5848 +#define AuDbgIAttr(ia) do { \
5849 + AuDbg("ia_valid 0x%x\n", (ia)->ia_valid); \
5850 + au_dbg_iattr(ia); \
5853 +#define AuDbgSym(addr) do { \
5854 + char sym[KSYM_SYMBOL_LEN]; \
5855 + sprint_symbol(sym, (unsigned long)addr); \
5856 + AuDbg("%s\n", sym); \
5859 +#define AuInfoSym(addr) do { \
5860 + char sym[KSYM_SYMBOL_LEN]; \
5861 + sprint_symbol(sym, (unsigned long)addr); \
5862 + AuInfo("%s\n", sym); \
5865 +AuStubVoid(au_dbg_verify_dinode, struct dentry *dentry)
5866 +AuStubVoid(au_dbg_verify_dir_parent, struct dentry *dentry, unsigned int sigen)
5867 +AuStubVoid(au_dbg_verify_nondir_parent, struct dentry *dentry,
5868 + unsigned int sigen)
5869 +AuStubVoid(au_dbg_verify_gen, struct dentry *parent, unsigned int sigen)
5870 +AuStubVoid(au_dbg_verify_kthread, void)
5871 +AuStubInt0(__init au_debug_init, void)
5872 +AuStubVoid(au_debug_sbinfo_init, struct au_sbinfo *sbinfo)
5874 +#define AuDbgWhlist(w) do {} while (0)
5875 +#define AuDbgVdir(v) do {} while (0)
5876 +#define AuDbgInode(i) do {} while (0)
5877 +#define AuDbgDAlias(i) do {} while (0)
5878 +#define AuDbgDentry(d) do {} while (0)
5879 +#define AuDbgFile(f) do {} while (0)
5880 +#define AuDbgSb(sb) do {} while (0)
5881 +#define AuDbgSleep(sec) do {} while (0)
5882 +#define AuDbgSleepJiffy(jiffy) do {} while (0)
5883 +#define AuDbgIAttr(ia) do {} while (0)
5884 +#define AuDbgSym(addr) do {} while (0)
5885 +#define AuInfoSym(addr) do {} while (0)
5886 +#endif /* CONFIG_AUFS_DEBUG */
5888 +/* ---------------------------------------------------------------------- */
5890 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
5891 +int __init au_sysrq_init(void);
5892 +void au_sysrq_fin(void);
5894 +#ifdef CONFIG_HW_CONSOLE
5895 +#define au_dbg_blocked() do { \
5897 + handle_sysrq('w'); \
5900 +AuStubVoid(au_dbg_blocked, void)
5904 +AuStubInt0(__init au_sysrq_init, void)
5905 +AuStubVoid(au_sysrq_fin, void)
5906 +AuStubVoid(au_dbg_blocked, void)
5907 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
5909 +#endif /* __KERNEL__ */
5910 +#endif /* __AUFS_DEBUG_H__ */
5911 diff -urN /usr/share/empty/fs/aufs/dentry.c linux/fs/aufs/dentry.c
5912 --- /usr/share/empty/fs/aufs/dentry.c 1970-01-01 01:00:00.000000000 +0100
5913 +++ linux/fs/aufs/dentry.c 2011-08-24 13:30:24.731313534 +0200
5916 + * Copyright (C) 2005-2011 Junjiro R. Okajima
5918 + * This program, aufs is free software; you can redistribute it and/or modify
5919 + * it under the terms of the GNU General Public License as published by
5920 + * the Free Software Foundation; either version 2 of the License, or
5921 + * (at your option) any later version.
5923 + * This program is distributed in the hope that it will be useful,
5924 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
5925 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
5926 + * GNU General Public License for more details.
5928 + * You should have received a copy of the GNU General Public License
5929 + * along with this program; if not, write to the Free Software
5930 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
5934 + * lookup and dentry operations
5937 +#include <linux/namei.h>
5940 +static void au_h_nd(struct nameidata *h_nd, struct nameidata *nd)
5946 + * gave up supporting LOOKUP_CREATE/OPEN for lower fs,
5947 + * due to whiteout and branch permission.
5949 + h_nd->flags &= ~(/*LOOKUP_PARENT |*/ LOOKUP_OPEN | LOOKUP_CREATE
5950 + | LOOKUP_FOLLOW | LOOKUP_EXCL);
5951 + /* unnecessary? */
5952 + h_nd->intent.open.file = NULL;
5954 + memset(h_nd, 0, sizeof(*h_nd));
5957 +struct au_lkup_one_args {
5958 + struct dentry **errp;
5959 + struct qstr *name;
5960 + struct dentry *h_parent;
5961 + struct au_branch *br;
5962 + struct nameidata *nd;
5965 +struct dentry *au_lkup_one(struct qstr *name, struct dentry *h_parent,
5966 + struct au_branch *br, struct nameidata *nd)
5968 + struct dentry *h_dentry;
5970 + struct nameidata h_nd;
5972 + if (au_test_fs_null_nd(h_parent->d_sb))
5973 + return vfsub_lookup_one_len(name->name, h_parent, name->len);
5975 + au_h_nd(&h_nd, nd);
5976 + h_nd.path.dentry = h_parent;
5977 + h_nd.path.mnt = br->br_mnt;
5979 + err = vfsub_name_hash(name->name, &h_nd.last, name->len);
5980 + h_dentry = ERR_PTR(err);
5982 + path_get(&h_nd.path);
5983 + h_dentry = vfsub_lookup_hash(&h_nd);
5984 + path_put(&h_nd.path);
5987 + AuTraceErrPtr(h_dentry);
5991 +static void au_call_lkup_one(void *args)
5993 + struct au_lkup_one_args *a = args;
5994 + *a->errp = au_lkup_one(a->name, a->h_parent, a->br, a->nd);
5997 +#define AuLkup_ALLOW_NEG 1
5998 +#define au_ftest_lkup(flags, name) ((flags) & AuLkup_##name)
5999 +#define au_fset_lkup(flags, name) \
6000 + do { (flags) |= AuLkup_##name; } while (0)
6001 +#define au_fclr_lkup(flags, name) \
6002 + do { (flags) &= ~AuLkup_##name; } while (0)
6004 +struct au_do_lookup_args {
6005 + unsigned int flags;
6007 + struct nameidata *nd;
6011 + * returns positive/negative dentry, NULL or an error.
6012 + * NULL means whiteout-ed or not-found.
6014 +static struct dentry*
6015 +au_do_lookup(struct dentry *h_parent, struct dentry *dentry,
6016 + aufs_bindex_t bindex, struct qstr *wh_name,
6017 + struct au_do_lookup_args *args)
6019 + struct dentry *h_dentry;
6020 + struct inode *h_inode, *inode;
6021 + struct au_branch *br;
6022 + int wh_found, opq;
6023 + unsigned char wh_able;
6024 + const unsigned char allow_neg = !!au_ftest_lkup(args->flags, ALLOW_NEG);
6027 + br = au_sbr(dentry->d_sb, bindex);
6028 + wh_able = !!au_br_whable(br->br_perm);
6030 + wh_found = au_wh_test(h_parent, wh_name, br, /*try_sio*/0);
6031 + h_dentry = ERR_PTR(wh_found);
6034 + if (unlikely(wh_found < 0))
6037 + /* We found a whiteout */
6038 + /* au_set_dbend(dentry, bindex); */
6039 + au_set_dbwh(dentry, bindex);
6041 + return NULL; /* success */
6044 + h_dentry = au_lkup_one(&dentry->d_name, h_parent, br, args->nd);
6045 + if (IS_ERR(h_dentry))
6048 + h_inode = h_dentry->d_inode;
6052 + } else if (wh_found
6053 + || (args->type && args->type != (h_inode->i_mode & S_IFMT)))
6056 + if (au_dbend(dentry) <= bindex)
6057 + au_set_dbend(dentry, bindex);
6058 + if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
6059 + au_set_dbstart(dentry, bindex);
6060 + au_set_h_dptr(dentry, bindex, h_dentry);
6062 + inode = dentry->d_inode;
6063 + if (!h_inode || !S_ISDIR(h_inode->i_mode) || !wh_able
6064 + || (inode && !S_ISDIR(inode->i_mode)))
6065 + goto out; /* success */
6067 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
6068 + opq = au_diropq_test(h_dentry, br);
6069 + mutex_unlock(&h_inode->i_mutex);
6071 + au_set_dbdiropq(dentry, bindex);
6072 + else if (unlikely(opq < 0)) {
6073 + au_set_h_dptr(dentry, bindex, NULL);
6074 + h_dentry = ERR_PTR(opq);
6085 +static int au_test_shwh(struct super_block *sb, const struct qstr *name)
6087 + if (unlikely(!au_opt_test(au_mntflags(sb), SHWH)
6088 + && !strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)))
6094 + * returns the number of lower positive dentries,
6095 + * otherwise an error.
6096 + * can be called at unlinking with @type is zero.
6098 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type,
6099 + struct nameidata *nd)
6101 + int npositive, err;
6102 + aufs_bindex_t bindex, btail, bdiropq;
6103 + unsigned char isdir;
6104 + struct qstr whname;
6105 + struct au_do_lookup_args args = {
6110 + const struct qstr *name = &dentry->d_name;
6111 + struct dentry *parent;
6112 + struct inode *inode;
6114 + err = au_test_shwh(dentry->d_sb, name);
6115 + if (unlikely(err))
6118 + err = au_wh_name_alloc(&whname, name);
6119 + if (unlikely(err))
6122 + inode = dentry->d_inode;
6123 + isdir = !!(inode && S_ISDIR(inode->i_mode));
6125 + au_fset_lkup(args.flags, ALLOW_NEG);
6128 + parent = dget_parent(dentry);
6129 + btail = au_dbtaildir(parent);
6130 + for (bindex = bstart; bindex <= btail; bindex++) {
6131 + struct dentry *h_parent, *h_dentry;
6132 + struct inode *h_inode, *h_dir;
6134 + h_dentry = au_h_dptr(dentry, bindex);
6136 + if (h_dentry->d_inode)
6138 + if (type != S_IFDIR)
6142 + h_parent = au_h_dptr(parent, bindex);
6145 + h_dir = h_parent->d_inode;
6146 + if (!h_dir || !S_ISDIR(h_dir->i_mode))
6149 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
6150 + h_dentry = au_do_lookup(h_parent, dentry, bindex, &whname,
6152 + mutex_unlock(&h_dir->i_mutex);
6153 + err = PTR_ERR(h_dentry);
6154 + if (IS_ERR(h_dentry))
6156 + au_fclr_lkup(args.flags, ALLOW_NEG);
6158 + if (au_dbwh(dentry) >= 0)
6162 + h_inode = h_dentry->d_inode;
6167 + args.type = h_inode->i_mode & S_IFMT;
6168 + if (args.type != S_IFDIR)
6171 + /* the type of lower may be different */
6172 + bdiropq = au_dbdiropq(dentry);
6173 + if (bdiropq >= 0 && bdiropq <= bindex)
6179 + AuLabel(positive);
6180 + au_update_dbstart(dentry);
6183 + if (unlikely(!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
6184 + && au_dbstart(dentry) < 0)) {
6186 + AuIOErr("both of real entry and whiteout found, %.*s, err %d\n",
6187 + AuDLNPair(dentry), err);
6192 + kfree(whname.name);
6197 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent,
6198 + struct au_branch *br)
6200 + struct dentry *dentry;
6203 + if (!au_test_h_perm_sio(parent->d_inode, MAY_EXEC))
6204 + dentry = au_lkup_one(name, parent, br, /*nd*/NULL);
6206 + struct au_lkup_one_args args = {
6209 + .h_parent = parent,
6214 + wkq_err = au_wkq_wait(au_call_lkup_one, &args);
6215 + if (unlikely(wkq_err))
6216 + dentry = ERR_PTR(wkq_err);
6223 + * lookup @dentry on @bindex which should be negative.
6225 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex)
6228 + struct dentry *parent, *h_parent, *h_dentry;
6230 + parent = dget_parent(dentry);
6231 + h_parent = au_h_dptr(parent, bindex);
6232 + h_dentry = au_sio_lkup_one(&dentry->d_name, h_parent,
6233 + au_sbr(dentry->d_sb, bindex));
6234 + err = PTR_ERR(h_dentry);
6235 + if (IS_ERR(h_dentry))
6237 + if (unlikely(h_dentry->d_inode)) {
6239 + AuIOErr("%.*s should be negative on b%d.\n",
6240 + AuDLNPair(h_dentry), bindex);
6246 + if (bindex < au_dbstart(dentry))
6247 + au_set_dbstart(dentry, bindex);
6248 + if (au_dbend(dentry) < bindex)
6249 + au_set_dbend(dentry, bindex);
6250 + au_set_h_dptr(dentry, bindex, h_dentry);
6257 +/* ---------------------------------------------------------------------- */
6259 +/* subset of struct inode */
6261 + unsigned long i_ino;
6262 + /* unsigned int i_nlink; */
6268 + blkcnt_t i_blocks;
6273 +static void au_iattr_save(struct au_iattr *ia, struct inode *h_inode)
6275 + ia->i_ino = h_inode->i_ino;
6276 + /* ia->i_nlink = h_inode->i_nlink; */
6277 + ia->i_uid = h_inode->i_uid;
6278 + ia->i_gid = h_inode->i_gid;
6279 + ia->i_version = h_inode->i_version;
6281 + ia->i_size = h_inode->i_size;
6282 + ia->i_blocks = h_inode->i_blocks;
6284 + ia->i_mode = (h_inode->i_mode & S_IFMT);
6287 +static int au_iattr_test(struct au_iattr *ia, struct inode *h_inode)
6289 + return ia->i_ino != h_inode->i_ino
6290 + /* || ia->i_nlink != h_inode->i_nlink */
6291 + || ia->i_uid != h_inode->i_uid
6292 + || ia->i_gid != h_inode->i_gid
6293 + || ia->i_version != h_inode->i_version
6295 + || ia->i_size != h_inode->i_size
6296 + || ia->i_blocks != h_inode->i_blocks
6298 + || ia->i_mode != (h_inode->i_mode & S_IFMT);
6301 +static int au_h_verify_dentry(struct dentry *h_dentry, struct dentry *h_parent,
6302 + struct au_branch *br)
6305 + struct au_iattr ia;
6306 + struct inode *h_inode;
6307 + struct dentry *h_d;
6308 + struct super_block *h_sb;
6311 + memset(&ia, -1, sizeof(ia));
6312 + h_sb = h_dentry->d_sb;
6313 + h_inode = h_dentry->d_inode;
6315 + au_iattr_save(&ia, h_inode);
6316 + else if (au_test_nfs(h_sb) || au_test_fuse(h_sb))
6317 + /* nfs d_revalidate may return 0 for negative dentry */
6318 + /* fuse d_revalidate always return 0 for negative dentry */
6321 + /* main purpose is namei.c:cached_lookup() and d_revalidate */
6322 + h_d = au_lkup_one(&h_dentry->d_name, h_parent, br, /*nd*/NULL);
6323 + err = PTR_ERR(h_d);
6328 + if (unlikely(h_d != h_dentry
6329 + || h_d->d_inode != h_inode
6330 + || (h_inode && au_iattr_test(&ia, h_inode))))
6331 + err = au_busy_or_stale();
6339 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
6340 + struct dentry *h_parent, struct au_branch *br)
6345 + if (udba == AuOpt_UDBA_REVAL
6346 + && !au_test_fs_remote(h_dentry->d_sb)) {
6348 + err = (h_dentry->d_parent->d_inode != h_dir);
6349 + } else if (udba != AuOpt_UDBA_NONE)
6350 + err = au_h_verify_dentry(h_dentry, h_parent, br);
6355 +/* ---------------------------------------------------------------------- */
6357 +static int au_do_refresh_hdentry(struct dentry *dentry, struct dentry *parent)
6360 + aufs_bindex_t new_bindex, bindex, bend, bwh, bdiropq;
6361 + struct au_hdentry tmp, *p, *q;
6362 + struct au_dinfo *dinfo;
6363 + struct super_block *sb;
6365 + DiMustWriteLock(dentry);
6367 + sb = dentry->d_sb;
6368 + dinfo = au_di(dentry);
6369 + bend = dinfo->di_bend;
6370 + bwh = dinfo->di_bwh;
6371 + bdiropq = dinfo->di_bdiropq;
6372 + p = dinfo->di_hdentry + dinfo->di_bstart;
6373 + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++, p++) {
6374 + if (!p->hd_dentry)
6377 + new_bindex = au_br_index(sb, p->hd_id);
6378 + if (new_bindex == bindex)
6381 + if (dinfo->di_bwh == bindex)
6383 + if (dinfo->di_bdiropq == bindex)
6384 + bdiropq = new_bindex;
6385 + if (new_bindex < 0) {
6387 + p->hd_dentry = NULL;
6391 + /* swap two lower dentries, and loop again */
6392 + q = dinfo->di_hdentry + new_bindex;
6396 + if (tmp.hd_dentry) {
6402 + dinfo->di_bwh = -1;
6403 + if (bwh >= 0 && bwh <= au_sbend(sb) && au_sbr_whable(sb, bwh))
6404 + dinfo->di_bwh = bwh;
6406 + dinfo->di_bdiropq = -1;
6408 + && bdiropq <= au_sbend(sb)
6409 + && au_sbr_whable(sb, bdiropq))
6410 + dinfo->di_bdiropq = bdiropq;
6413 + dinfo->di_bstart = -1;
6414 + dinfo->di_bend = -1;
6415 + bend = au_dbend(parent);
6416 + p = dinfo->di_hdentry;
6417 + for (bindex = 0; bindex <= bend; bindex++, p++)
6418 + if (p->hd_dentry) {
6419 + dinfo->di_bstart = bindex;
6423 + if (dinfo->di_bstart >= 0) {
6424 + p = dinfo->di_hdentry + bend;
6425 + for (bindex = bend; bindex >= 0; bindex--, p--)
6426 + if (p->hd_dentry) {
6427 + dinfo->di_bend = bindex;
6436 +static void au_do_hide(struct dentry *dentry)
6438 + struct inode *inode;
6440 + inode = dentry->d_inode;
6442 + if (!S_ISDIR(inode->i_mode)) {
6443 + if (inode->i_nlink && !d_unhashed(dentry))
6444 + drop_nlink(inode);
6446 + clear_nlink(inode);
6447 + /* stop next lookup */
6448 + inode->i_flags |= S_DEAD;
6450 + smp_mb(); /* necessary? */
6455 +static int au_hide_children(struct dentry *parent)
6457 + int err, i, j, ndentry;
6458 + struct au_dcsub_pages dpages;
6459 + struct au_dpage *dpage;
6460 + struct dentry *dentry;
6462 + err = au_dpages_init(&dpages, GFP_NOFS);
6463 + if (unlikely(err))
6465 + err = au_dcsub_pages(&dpages, parent, NULL, NULL);
6466 + if (unlikely(err))
6469 + /* in reverse order */
6470 + for (i = dpages.ndpage - 1; i >= 0; i--) {
6471 + dpage = dpages.dpages + i;
6472 + ndentry = dpage->ndentry;
6473 + for (j = ndentry - 1; j >= 0; j--) {
6474 + dentry = dpage->dentries[j];
6475 + if (dentry != parent)
6476 + au_do_hide(dentry);
6481 + au_dpages_free(&dpages);
6486 +static void au_hide(struct dentry *dentry)
6489 + struct inode *inode;
6491 + AuDbgDentry(dentry);
6492 + inode = dentry->d_inode;
6493 + if (inode && S_ISDIR(inode->i_mode)) {
6494 + /* shrink_dcache_parent(dentry); */
6495 + err = au_hide_children(dentry);
6496 + if (unlikely(err))
6497 + AuIOErr("%.*s, failed hiding children, ignored %d\n",
6498 + AuDLNPair(dentry), err);
6500 + au_do_hide(dentry);
6504 + * By adding a dirty branch, a cached dentry may be affected in various ways.
6506 + * a dirty branch is added
6507 + * - on the top of layers
6508 + * - in the middle of layers
6509 + * - to the bottom of layers
6511 + * on the added branch there exists
6514 + * - a same named entry
6516 + * * negative --> positive
6517 + * * positive --> positive
6518 + * - type is unchanged
6519 + * - type is changed
6521 + * * negative --> negative
6522 + * * positive --> negative (rejected by au_br_del() for non-dir case)
6525 +static int au_refresh_by_dinfo(struct dentry *dentry, struct au_dinfo *dinfo,
6526 + struct au_dinfo *tmp)
6529 + aufs_bindex_t bindex, bend;
6531 + struct dentry *dentry;
6532 + struct inode *inode;
6535 + struct au_hdentry *hd;
6536 + struct inode *inode, *h_inode;
6537 + struct dentry *h_dentry;
6540 + AuDebugOn(dinfo->di_bstart < 0);
6541 + orig_h.dentry = dinfo->di_hdentry[dinfo->di_bstart].hd_dentry;
6542 + orig_h.inode = orig_h.dentry->d_inode;
6545 + orig_h.mode = orig_h.inode->i_mode & S_IFMT;
6546 + memset(&tmp_h, 0, sizeof(tmp_h));
6547 + if (tmp->di_bstart >= 0) {
6548 + tmp_h.dentry = tmp->di_hdentry[tmp->di_bstart].hd_dentry;
6549 + tmp_h.inode = tmp_h.dentry->d_inode;
6551 + tmp_h.mode = tmp_h.inode->i_mode & S_IFMT;
6554 + inode = dentry->d_inode;
6555 + if (!orig_h.inode) {
6556 + AuDbg("nagative originally\n");
6562 + AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
6563 + AuDebugOn(dinfo->di_bdiropq != -1);
6565 + if (!tmp_h.inode) {
6566 + AuDbg("negative --> negative\n");
6567 + /* should have only one negative lower */
6568 + if (tmp->di_bstart >= 0
6569 + && tmp->di_bstart < dinfo->di_bstart) {
6570 + AuDebugOn(tmp->di_bstart != tmp->di_bend);
6571 + AuDebugOn(dinfo->di_bstart != dinfo->di_bend);
6572 + au_set_h_dptr(dentry, dinfo->di_bstart, NULL);
6573 + au_di_cp(dinfo, tmp);
6574 + hd = tmp->di_hdentry + tmp->di_bstart;
6575 + au_set_h_dptr(dentry, tmp->di_bstart,
6576 + dget(hd->hd_dentry));
6578 + au_dbg_verify_dinode(dentry);
6580 + AuDbg("negative --> positive\n");
6582 + * similar to the behaviour of creating with bypassing
6584 + * unhash it in order to force an error in the
6585 + * succeeding create operation.
6586 + * we should not set S_DEAD here.
6589 + /* au_di_swap(tmp, dinfo); */
6590 + au_dbg_verify_dinode(dentry);
6593 + AuDbg("positive originally\n");
6594 + /* inode may be NULL */
6595 + AuDebugOn(inode && (inode->i_mode & S_IFMT) != orig_h.mode);
6596 + if (!tmp_h.inode) {
6597 + AuDbg("positive --> negative\n");
6598 + /* or bypassing aufs */
6600 + if (tmp->di_bwh >= 0 && tmp->di_bwh <= dinfo->di_bstart)
6601 + dinfo->di_bwh = tmp->di_bwh;
6603 + err = au_refresh_hinode_self(inode);
6604 + au_dbg_verify_dinode(dentry);
6605 + } else if (orig_h.mode == tmp_h.mode) {
6606 + AuDbg("positive --> positive, same type\n");
6607 + if (!S_ISDIR(orig_h.mode)
6608 + && dinfo->di_bstart > tmp->di_bstart) {
6610 + * similar to the behaviour of removing and
6615 + err = au_refresh_hinode_self(inode);
6616 + au_dbg_verify_dinode(dentry);
6618 + /* fill empty slots */
6619 + if (dinfo->di_bstart > tmp->di_bstart)
6620 + dinfo->di_bstart = tmp->di_bstart;
6621 + if (dinfo->di_bend < tmp->di_bend)
6622 + dinfo->di_bend = tmp->di_bend;
6623 + dinfo->di_bwh = tmp->di_bwh;
6624 + dinfo->di_bdiropq = tmp->di_bdiropq;
6625 + hd = tmp->di_hdentry;
6626 + bend = dinfo->di_bend;
6627 + for (bindex = tmp->di_bstart; bindex <= bend;
6629 + if (au_h_dptr(dentry, bindex))
6631 + h_dentry = hd[bindex].hd_dentry;
6634 + h_inode = h_dentry->d_inode;
6635 + AuDebugOn(!h_inode);
6636 + AuDebugOn(orig_h.mode
6637 + != (h_inode->i_mode
6639 + au_set_h_dptr(dentry, bindex,
6642 + err = au_refresh_hinode(inode, dentry);
6643 + au_dbg_verify_dinode(dentry);
6646 + AuDbg("positive --> positive, different type\n");
6647 + /* similar to the behaviour of removing and creating */
6650 + err = au_refresh_hinode_self(inode);
6651 + au_dbg_verify_dinode(dentry);
6659 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent)
6662 + unsigned int sigen;
6663 + struct au_dinfo *dinfo, *tmp;
6664 + struct super_block *sb;
6665 + struct inode *inode;
6667 + DiMustWriteLock(dentry);
6668 + AuDebugOn(IS_ROOT(dentry));
6669 + AuDebugOn(!parent->d_inode);
6671 + sb = dentry->d_sb;
6672 + inode = dentry->d_inode;
6673 + sigen = au_sigen(sb);
6674 + err = au_digen_test(parent, sigen);
6675 + if (unlikely(err))
6678 + dinfo = au_di(dentry);
6679 + err = au_di_realloc(dinfo, au_sbend(sb) + 1);
6680 + if (unlikely(err))
6682 + ebrange = au_dbrange_test(dentry);
6684 + ebrange = au_do_refresh_hdentry(dentry, parent);
6686 + if (d_unhashed(dentry) || ebrange) {
6687 + AuDebugOn(au_dbstart(dentry) < 0 && au_dbend(dentry) >= 0);
6689 + err = au_refresh_hinode_self(inode);
6690 + au_dbg_verify_dinode(dentry);
6692 + goto out_dgen; /* success */
6696 + /* temporary dinfo */
6697 + AuDbgDentry(dentry);
6699 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
6700 + if (unlikely(!tmp))
6702 + au_di_swap(tmp, dinfo);
6703 + /* returns the number of positive dentries */
6705 + * if current working dir is removed, it returns an error.
6706 + * but the dentry is legal.
6708 + err = au_lkup_dentry(dentry, /*bstart*/0, /*type*/0, /*nd*/NULL);
6709 + AuDbgDentry(dentry);
6710 + au_di_swap(tmp, dinfo);
6711 + if (err == -ENOENT)
6714 + /* compare/refresh by dinfo */
6715 + AuDbgDentry(dentry);
6716 + err = au_refresh_by_dinfo(dentry, dinfo, tmp);
6717 + au_dbg_verify_dinode(dentry);
6720 + au_rw_write_unlock(&tmp->di_rwsem);
6722 + if (unlikely(err))
6726 + au_update_digen(dentry);
6728 + if (unlikely(err && !(dentry->d_flags & DCACHE_NFSFS_RENAMED))) {
6729 + AuIOErr("failed refreshing %.*s, %d\n",
6730 + AuDLNPair(dentry), err);
6731 + AuDbgDentry(dentry);
6737 +static noinline_for_stack
6738 +int au_do_h_d_reval(struct dentry *h_dentry, struct nameidata *nd,
6739 + struct dentry *dentry, aufs_bindex_t bindex)
6742 + int (*reval)(struct dentry *, struct nameidata *);
6745 + if (!(h_dentry->d_flags & DCACHE_OP_REVALIDATE))
6747 + reval = h_dentry->d_op->d_revalidate;
6749 + AuDbg("b%d\n", bindex);
6750 + if (au_test_fs_null_nd(h_dentry->d_sb))
6751 + /* it may return tri-state */
6752 + valid = reval(h_dentry, NULL);
6754 + struct nameidata h_nd;
6756 + struct dentry *parent;
6758 + au_h_nd(&h_nd, nd);
6759 + parent = nd->path.dentry;
6760 + locked = (nd && nd->path.dentry != dentry);
6762 + di_read_lock_parent(parent, AuLock_IR);
6763 + BUG_ON(bindex > au_dbend(parent));
6764 + h_nd.path.dentry = au_h_dptr(parent, bindex);
6765 + BUG_ON(!h_nd.path.dentry);
6766 + h_nd.path.mnt = au_sbr(parent->d_sb, bindex)->br_mnt;
6767 + path_get(&h_nd.path);
6768 + valid = reval(h_dentry, &h_nd);
6769 + path_put(&h_nd.path);
6771 + di_read_unlock(parent, AuLock_IR);
6774 + if (unlikely(valid < 0))
6784 +/* todo: remove this */
6785 +static int h_d_revalidate(struct dentry *dentry, struct inode *inode,
6786 + struct nameidata *nd, int do_udba)
6789 + umode_t mode, h_mode;
6790 + aufs_bindex_t bindex, btail, bstart, ibs, ibe;
6791 + unsigned char plus, unhashed, is_root, h_plus;
6792 + struct inode *h_inode, *h_cached_inode;
6793 + struct dentry *h_dentry;
6794 + struct qstr *name, *h_name;
6801 + unhashed = !!d_unhashed(dentry);
6802 + is_root = !!IS_ROOT(dentry);
6803 + name = &dentry->d_name;
6806 + * Theoretically, REVAL test should be unnecessary in case of
6808 + * But {fs,i}notify doesn't fire some necessary events,
6809 + * IN_ATTRIB for atime/nlink/pageio
6810 + * IN_DELETE for NFS dentry
6811 + * Let's do REVAL test too.
6813 + if (do_udba && inode) {
6814 + mode = (inode->i_mode & S_IFMT);
6815 + plus = (inode->i_nlink > 0);
6816 + ibs = au_ibstart(inode);
6817 + ibe = au_ibend(inode);
6820 + bstart = au_dbstart(dentry);
6822 + if (inode && S_ISDIR(inode->i_mode))
6823 + btail = au_dbtaildir(dentry);
6824 + for (bindex = bstart; bindex <= btail; bindex++) {
6825 + h_dentry = au_h_dptr(dentry, bindex);
6829 + AuDbg("b%d, %.*s\n", bindex, AuDLNPair(h_dentry));
6830 + spin_lock(&h_dentry->d_lock);
6831 + h_name = &h_dentry->d_name;
6832 + if (unlikely(do_udba
6834 + && (unhashed != !!d_unhashed(h_dentry)
6835 + || name->len != h_name->len
6836 + || memcmp(name->name, h_name->name, name->len))
6838 + AuDbg("unhash 0x%x 0x%x, %.*s %.*s\n",
6839 + unhashed, d_unhashed(h_dentry),
6840 + AuDLNPair(dentry), AuDLNPair(h_dentry));
6841 + spin_unlock(&h_dentry->d_lock);
6844 + spin_unlock(&h_dentry->d_lock);
6846 + err = au_do_h_d_reval(h_dentry, nd, dentry, bindex);
6847 + if (unlikely(err))
6848 + /* do not goto err, to keep the errno */
6851 + /* todo: plink too? */
6856 + h_inode = h_dentry->d_inode;
6857 + if (unlikely(!!inode != !!h_inode))
6862 + h_cached_inode = h_inode;
6864 + h_mode = (h_inode->i_mode & S_IFMT);
6865 + h_plus = (h_inode->i_nlink > 0);
6867 + if (inode && ibs <= bindex && bindex <= ibe)
6868 + h_cached_inode = au_h_iptr(inode, bindex);
6870 + if (unlikely(plus != h_plus
6872 + || h_cached_inode != h_inode))
6884 +/* todo: consolidate with do_refresh() and au_reval_for_attr() */
6885 +static int simple_reval_dpath(struct dentry *dentry, unsigned int sigen)
6888 + struct dentry *parent;
6890 + if (!au_digen_test(dentry, sigen))
6893 + parent = dget_parent(dentry);
6894 + di_read_lock_parent(parent, AuLock_IR);
6895 + AuDebugOn(au_digen_test(parent, sigen));
6896 + au_dbg_verify_gen(parent, sigen);
6897 + err = au_refresh_dentry(dentry, parent);
6898 + di_read_unlock(parent, AuLock_IR);
6904 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen)
6907 + struct dentry *d, *parent;
6908 + struct inode *inode;
6910 + if (!au_ftest_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR))
6911 + return simple_reval_dpath(dentry, sigen);
6913 + /* slow loop, keep it simple and stupid */
6914 + /* cf: au_cpup_dirs() */
6917 + while (au_digen_test(dentry, sigen)) {
6921 + parent = dget_parent(d);
6922 + if (!au_digen_test(parent, sigen))
6927 + inode = d->d_inode;
6929 + di_write_lock_child2(d);
6931 + /* someone might update our dentry while we were sleeping */
6932 + if (au_digen_test(d, sigen)) {
6934 + * todo: consolidate with simple_reval_dpath(),
6935 + * do_refresh() and au_reval_for_attr().
6937 + di_read_lock_parent(parent, AuLock_IR);
6938 + err = au_refresh_dentry(d, parent);
6939 + di_read_unlock(parent, AuLock_IR);
6943 + di_write_unlock(d);
6945 + if (unlikely(err))
6953 + * if valid returns 1, otherwise 0.
6955 +static int aufs_d_revalidate(struct dentry *dentry, struct nameidata *nd)
6958 + unsigned int sigen;
6959 + unsigned char do_udba;
6960 + struct super_block *sb;
6961 + struct inode *inode;
6963 + /* todo: support rcu-walk? */
6964 + if (nd && (nd->flags & LOOKUP_RCU))
6968 + if (unlikely(!au_di(dentry)))
6971 + inode = dentry->d_inode;
6972 + if (inode && is_bad_inode(inode))
6976 + sb = dentry->d_sb;
6979 + * i_mutex of parent dir may be held,
6980 + * but we should not return 'invalid' due to busy.
6982 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_DW | AuLock_NOPLM);
6983 + if (unlikely(err)) {
6988 + if (unlikely(au_dbrange_test(dentry))) {
6994 + sigen = au_sigen(sb);
6995 + if (au_digen_test(dentry, sigen)) {
6996 + AuDebugOn(IS_ROOT(dentry));
6997 + err = au_reval_dpath(dentry, sigen);
6998 + if (unlikely(err)) {
7003 + di_downgrade_lock(dentry, AuLock_IR);
7006 + if (inode && (IS_DEADDIR(inode) || !inode->i_nlink))
7009 + do_udba = !au_opt_test(au_mntflags(sb), UDBA_NONE);
7010 + if (do_udba && inode) {
7011 + aufs_bindex_t bstart = au_ibstart(inode);
7012 + struct inode *h_inode;
7014 + if (bstart >= 0) {
7015 + h_inode = au_h_iptr(inode, bstart);
7016 + if (h_inode && au_test_higen(inode, h_inode))
7021 + err = h_d_revalidate(dentry, inode, nd, do_udba);
7022 + if (unlikely(!err && do_udba && au_dbstart(dentry) < 0)) {
7024 + AuDbg("both of real entry and whiteout found, %.*s, err %d\n",
7025 + AuDLNPair(dentry), err);
7030 + di_downgrade_lock(dentry, AuLock_IR);
7032 + aufs_read_unlock(dentry, AuLock_IR);
7037 + AuDbg("%.*s invalid, %d\n", AuDLNPair(dentry), valid);
7043 +static void aufs_d_release(struct dentry *dentry)
7045 + if (au_di(dentry)) {
7046 + au_di_fin(dentry);
7047 + au_hn_di_reinit(dentry);
7051 +const struct dentry_operations aufs_dop = {
7052 + .d_revalidate = aufs_d_revalidate,
7053 + .d_release = aufs_d_release
7055 diff -urN /usr/share/empty/fs/aufs/dentry.h linux/fs/aufs/dentry.h
7056 --- /usr/share/empty/fs/aufs/dentry.h 1970-01-01 01:00:00.000000000 +0100
7057 +++ linux/fs/aufs/dentry.h 2011-08-24 13:30:24.731313534 +0200
7060 + * Copyright (C) 2005-2011 Junjiro R. Okajima
7062 + * This program, aufs is free software; you can redistribute it and/or modify
7063 + * it under the terms of the GNU General Public License as published by
7064 + * the Free Software Foundation; either version 2 of the License, or
7065 + * (at your option) any later version.
7067 + * This program is distributed in the hope that it will be useful,
7068 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7069 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7070 + * GNU General Public License for more details.
7072 + * You should have received a copy of the GNU General Public License
7073 + * along with this program; if not, write to the Free Software
7074 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
7078 + * lookup and dentry operations
7081 +#ifndef __AUFS_DENTRY_H__
7082 +#define __AUFS_DENTRY_H__
7086 +#include <linux/dcache.h>
7087 +#include <linux/aufs_type.h>
7090 +struct au_hdentry {
7091 + struct dentry *hd_dentry;
7092 + aufs_bindex_t hd_id;
7096 + atomic_t di_generation;
7098 + struct au_rwsem di_rwsem;
7099 + aufs_bindex_t di_bstart, di_bend, di_bwh, di_bdiropq;
7100 + struct au_hdentry *di_hdentry;
7101 +} ____cacheline_aligned_in_smp;
7103 +/* ---------------------------------------------------------------------- */
7106 +extern const struct dentry_operations aufs_dop;
7108 +struct dentry *au_lkup_one(struct qstr *name, struct dentry *h_parent,
7109 + struct au_branch *br, struct nameidata *nd);
7110 +struct dentry *au_sio_lkup_one(struct qstr *name, struct dentry *parent,
7111 + struct au_branch *br);
7112 +int au_h_verify(struct dentry *h_dentry, unsigned int udba, struct inode *h_dir,
7113 + struct dentry *h_parent, struct au_branch *br);
7115 +int au_lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type,
7116 + struct nameidata *nd);
7117 +int au_lkup_neg(struct dentry *dentry, aufs_bindex_t bindex);
7118 +int au_refresh_dentry(struct dentry *dentry, struct dentry *parent);
7119 +int au_reval_dpath(struct dentry *dentry, unsigned int sigen);
7122 +void au_di_init_once(void *_di);
7123 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc);
7124 +void au_di_free(struct au_dinfo *dinfo);
7125 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b);
7126 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src);
7127 +int au_di_init(struct dentry *dentry);
7128 +void au_di_fin(struct dentry *dentry);
7129 +int au_di_realloc(struct au_dinfo *dinfo, int nbr);
7131 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc);
7132 +void di_read_unlock(struct dentry *d, int flags);
7133 +void di_downgrade_lock(struct dentry *d, int flags);
7134 +void di_write_lock(struct dentry *d, unsigned int lsc);
7135 +void di_write_unlock(struct dentry *d);
7136 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir);
7137 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir);
7138 +void di_write_unlock2(struct dentry *d1, struct dentry *d2);
7140 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex);
7141 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex);
7142 +aufs_bindex_t au_dbtail(struct dentry *dentry);
7143 +aufs_bindex_t au_dbtaildir(struct dentry *dentry);
7145 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
7146 + struct dentry *h_dentry);
7147 +int au_digen_test(struct dentry *dentry, unsigned int sigen);
7148 +int au_dbrange_test(struct dentry *dentry);
7149 +void au_update_digen(struct dentry *dentry);
7150 +void au_update_dbrange(struct dentry *dentry, int do_put_zero);
7151 +void au_update_dbstart(struct dentry *dentry);
7152 +void au_update_dbend(struct dentry *dentry);
7153 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry);
7155 +/* ---------------------------------------------------------------------- */
7157 +static inline struct au_dinfo *au_di(struct dentry *dentry)
7159 + return dentry->d_fsdata;
7162 +/* ---------------------------------------------------------------------- */
7164 +/* lock subclass for dinfo */
7166 + AuLsc_DI_CHILD, /* child first */
7167 + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hnotify */
7168 + AuLsc_DI_CHILD3, /* copyup dirs */
7172 + AuLsc_DI_TMP /* temp for replacing dinfo */
7176 + * di_read_lock_child, di_write_lock_child,
7177 + * di_read_lock_child2, di_write_lock_child2,
7178 + * di_read_lock_child3, di_write_lock_child3,
7179 + * di_read_lock_parent, di_write_lock_parent,
7180 + * di_read_lock_parent2, di_write_lock_parent2,
7181 + * di_read_lock_parent3, di_write_lock_parent3,
7183 +#define AuReadLockFunc(name, lsc) \
7184 +static inline void di_read_lock_##name(struct dentry *d, int flags) \
7185 +{ di_read_lock(d, flags, AuLsc_DI_##lsc); }
7187 +#define AuWriteLockFunc(name, lsc) \
7188 +static inline void di_write_lock_##name(struct dentry *d) \
7189 +{ di_write_lock(d, AuLsc_DI_##lsc); }
7191 +#define AuRWLockFuncs(name, lsc) \
7192 + AuReadLockFunc(name, lsc) \
7193 + AuWriteLockFunc(name, lsc)
7195 +AuRWLockFuncs(child, CHILD);
7196 +AuRWLockFuncs(child2, CHILD2);
7197 +AuRWLockFuncs(child3, CHILD3);
7198 +AuRWLockFuncs(parent, PARENT);
7199 +AuRWLockFuncs(parent2, PARENT2);
7200 +AuRWLockFuncs(parent3, PARENT3);
7202 +#undef AuReadLockFunc
7203 +#undef AuWriteLockFunc
7204 +#undef AuRWLockFuncs
7206 +#define DiMustNoWaiters(d) AuRwMustNoWaiters(&au_di(d)->di_rwsem)
7207 +#define DiMustAnyLock(d) AuRwMustAnyLock(&au_di(d)->di_rwsem)
7208 +#define DiMustWriteLock(d) AuRwMustWriteLock(&au_di(d)->di_rwsem)
7210 +/* ---------------------------------------------------------------------- */
7212 +/* todo: memory barrier? */
7213 +static inline unsigned int au_digen(struct dentry *d)
7215 + return atomic_read(&au_di(d)->di_generation);
7218 +static inline void au_h_dentry_init(struct au_hdentry *hdentry)
7220 + hdentry->hd_dentry = NULL;
7223 +static inline void au_hdput(struct au_hdentry *hd)
7226 + dput(hd->hd_dentry);
7229 +static inline aufs_bindex_t au_dbstart(struct dentry *dentry)
7231 + DiMustAnyLock(dentry);
7232 + return au_di(dentry)->di_bstart;
7235 +static inline aufs_bindex_t au_dbend(struct dentry *dentry)
7237 + DiMustAnyLock(dentry);
7238 + return au_di(dentry)->di_bend;
7241 +static inline aufs_bindex_t au_dbwh(struct dentry *dentry)
7243 + DiMustAnyLock(dentry);
7244 + return au_di(dentry)->di_bwh;
7247 +static inline aufs_bindex_t au_dbdiropq(struct dentry *dentry)
7249 + DiMustAnyLock(dentry);
7250 + return au_di(dentry)->di_bdiropq;
7253 +/* todo: hard/soft set? */
7254 +static inline void au_set_dbstart(struct dentry *dentry, aufs_bindex_t bindex)
7256 + DiMustWriteLock(dentry);
7257 + au_di(dentry)->di_bstart = bindex;
7260 +static inline void au_set_dbend(struct dentry *dentry, aufs_bindex_t bindex)
7262 + DiMustWriteLock(dentry);
7263 + au_di(dentry)->di_bend = bindex;
7266 +static inline void au_set_dbwh(struct dentry *dentry, aufs_bindex_t bindex)
7268 + DiMustWriteLock(dentry);
7269 + /* dbwh can be outside of bstart - bend range */
7270 + au_di(dentry)->di_bwh = bindex;
7273 +static inline void au_set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex)
7275 + DiMustWriteLock(dentry);
7276 + au_di(dentry)->di_bdiropq = bindex;
7279 +/* ---------------------------------------------------------------------- */
7281 +#ifdef CONFIG_AUFS_HNOTIFY
7282 +static inline void au_digen_dec(struct dentry *d)
7284 + atomic_dec(&au_di(d)->di_generation);
7287 +static inline void au_hn_di_reinit(struct dentry *dentry)
7289 + dentry->d_fsdata = NULL;
7292 +AuStubVoid(au_hn_di_reinit, struct dentry *dentry __maybe_unused)
7293 +#endif /* CONFIG_AUFS_HNOTIFY */
7295 +#endif /* __KERNEL__ */
7296 +#endif /* __AUFS_DENTRY_H__ */
7297 diff -urN /usr/share/empty/fs/aufs/dinfo.c linux/fs/aufs/dinfo.c
7298 --- /usr/share/empty/fs/aufs/dinfo.c 1970-01-01 01:00:00.000000000 +0100
7299 +++ linux/fs/aufs/dinfo.c 2011-08-24 13:30:24.731313534 +0200
7302 + * Copyright (C) 2005-2011 Junjiro R. Okajima
7304 + * This program, aufs is free software; you can redistribute it and/or modify
7305 + * it under the terms of the GNU General Public License as published by
7306 + * the Free Software Foundation; either version 2 of the License, or
7307 + * (at your option) any later version.
7309 + * This program is distributed in the hope that it will be useful,
7310 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7311 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7312 + * GNU General Public License for more details.
7314 + * You should have received a copy of the GNU General Public License
7315 + * along with this program; if not, write to the Free Software
7316 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
7320 + * dentry private data
7325 +void au_di_init_once(void *_dinfo)
7327 + struct au_dinfo *dinfo = _dinfo;
7328 + static struct lock_class_key aufs_di;
7330 + au_rw_init(&dinfo->di_rwsem);
7331 + au_rw_class(&dinfo->di_rwsem, &aufs_di);
7334 +struct au_dinfo *au_di_alloc(struct super_block *sb, unsigned int lsc)
7336 + struct au_dinfo *dinfo;
7339 + dinfo = au_cache_alloc_dinfo();
7340 + if (unlikely(!dinfo))
7343 + nbr = au_sbend(sb) + 1;
7346 + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), GFP_NOFS);
7347 + if (dinfo->di_hdentry) {
7348 + au_rw_write_lock_nested(&dinfo->di_rwsem, lsc);
7349 + dinfo->di_bstart = -1;
7350 + dinfo->di_bend = -1;
7351 + dinfo->di_bwh = -1;
7352 + dinfo->di_bdiropq = -1;
7353 + for (i = 0; i < nbr; i++)
7354 + dinfo->di_hdentry[i].hd_id = -1;
7358 + au_cache_free_dinfo(dinfo);
7365 +void au_di_free(struct au_dinfo *dinfo)
7367 + struct au_hdentry *p;
7368 + aufs_bindex_t bend, bindex;
7370 + /* dentry may not be revalidated */
7371 + bindex = dinfo->di_bstart;
7372 + if (bindex >= 0) {
7373 + bend = dinfo->di_bend;
7374 + p = dinfo->di_hdentry + bindex;
7375 + while (bindex++ <= bend)
7378 + kfree(dinfo->di_hdentry);
7379 + au_cache_free_dinfo(dinfo);
7382 +void au_di_swap(struct au_dinfo *a, struct au_dinfo *b)
7384 + struct au_hdentry *p;
7387 + AuRwMustWriteLock(&a->di_rwsem);
7388 + AuRwMustWriteLock(&b->di_rwsem);
7390 +#define DiSwap(v, name) \
7392 + v = a->di_##name; \
7393 + a->di_##name = b->di_##name; \
7394 + b->di_##name = v; \
7397 + DiSwap(p, hdentry);
7398 + DiSwap(bi, bstart);
7401 + DiSwap(bi, bdiropq);
7407 +void au_di_cp(struct au_dinfo *dst, struct au_dinfo *src)
7409 + AuRwMustWriteLock(&dst->di_rwsem);
7410 + AuRwMustWriteLock(&src->di_rwsem);
7412 + dst->di_bstart = src->di_bstart;
7413 + dst->di_bend = src->di_bend;
7414 + dst->di_bwh = src->di_bwh;
7415 + dst->di_bdiropq = src->di_bdiropq;
7419 +int au_di_init(struct dentry *dentry)
7422 + struct super_block *sb;
7423 + struct au_dinfo *dinfo;
7426 + sb = dentry->d_sb;
7427 + dinfo = au_di_alloc(sb, AuLsc_DI_CHILD);
7429 + atomic_set(&dinfo->di_generation, au_sigen(sb));
7430 + /* smp_mb(); */ /* atomic_set */
7431 + dentry->d_fsdata = dinfo;
7438 +void au_di_fin(struct dentry *dentry)
7440 + struct au_dinfo *dinfo;
7442 + dinfo = au_di(dentry);
7443 + AuRwDestroy(&dinfo->di_rwsem);
7444 + au_di_free(dinfo);
7447 +int au_di_realloc(struct au_dinfo *dinfo, int nbr)
7450 + struct au_hdentry *hdp;
7452 + AuRwMustWriteLock(&dinfo->di_rwsem);
7455 + sz = sizeof(*hdp) * (dinfo->di_bend + 1);
7457 + sz = sizeof(*hdp);
7458 + hdp = au_kzrealloc(dinfo->di_hdentry, sz, sizeof(*hdp) * nbr, GFP_NOFS);
7460 + dinfo->di_hdentry = hdp;
7467 +/* ---------------------------------------------------------------------- */
7469 +static void do_ii_write_lock(struct inode *inode, unsigned int lsc)
7472 + case AuLsc_DI_CHILD:
7473 + ii_write_lock_child(inode);
7475 + case AuLsc_DI_CHILD2:
7476 + ii_write_lock_child2(inode);
7478 + case AuLsc_DI_CHILD3:
7479 + ii_write_lock_child3(inode);
7481 + case AuLsc_DI_PARENT:
7482 + ii_write_lock_parent(inode);
7484 + case AuLsc_DI_PARENT2:
7485 + ii_write_lock_parent2(inode);
7487 + case AuLsc_DI_PARENT3:
7488 + ii_write_lock_parent3(inode);
7495 +static void do_ii_read_lock(struct inode *inode, unsigned int lsc)
7498 + case AuLsc_DI_CHILD:
7499 + ii_read_lock_child(inode);
7501 + case AuLsc_DI_CHILD2:
7502 + ii_read_lock_child2(inode);
7504 + case AuLsc_DI_CHILD3:
7505 + ii_read_lock_child3(inode);
7507 + case AuLsc_DI_PARENT:
7508 + ii_read_lock_parent(inode);
7510 + case AuLsc_DI_PARENT2:
7511 + ii_read_lock_parent2(inode);
7513 + case AuLsc_DI_PARENT3:
7514 + ii_read_lock_parent3(inode);
7521 +void di_read_lock(struct dentry *d, int flags, unsigned int lsc)
7523 + au_rw_read_lock_nested(&au_di(d)->di_rwsem, lsc);
7525 + if (au_ftest_lock(flags, IW))
7526 + do_ii_write_lock(d->d_inode, lsc);
7527 + else if (au_ftest_lock(flags, IR))
7528 + do_ii_read_lock(d->d_inode, lsc);
7532 +void di_read_unlock(struct dentry *d, int flags)
7535 + if (au_ftest_lock(flags, IW)) {
7536 + au_dbg_verify_dinode(d);
7537 + ii_write_unlock(d->d_inode);
7538 + } else if (au_ftest_lock(flags, IR)) {
7539 + au_dbg_verify_dinode(d);
7540 + ii_read_unlock(d->d_inode);
7543 + au_rw_read_unlock(&au_di(d)->di_rwsem);
7546 +void di_downgrade_lock(struct dentry *d, int flags)
7548 + if (d->d_inode && au_ftest_lock(flags, IR))
7549 + ii_downgrade_lock(d->d_inode);
7550 + au_rw_dgrade_lock(&au_di(d)->di_rwsem);
7553 +void di_write_lock(struct dentry *d, unsigned int lsc)
7555 + au_rw_write_lock_nested(&au_di(d)->di_rwsem, lsc);
7557 + do_ii_write_lock(d->d_inode, lsc);
7560 +void di_write_unlock(struct dentry *d)
7562 + au_dbg_verify_dinode(d);
7564 + ii_write_unlock(d->d_inode);
7565 + au_rw_write_unlock(&au_di(d)->di_rwsem);
7568 +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir)
7570 + AuDebugOn(d1 == d2
7571 + || d1->d_inode == d2->d_inode
7572 + || d1->d_sb != d2->d_sb);
7574 + if (isdir && au_test_subdir(d1, d2)) {
7575 + di_write_lock_child(d1);
7576 + di_write_lock_child2(d2);
7578 + /* there should be no races */
7579 + di_write_lock_child(d2);
7580 + di_write_lock_child2(d1);
7584 +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir)
7586 + AuDebugOn(d1 == d2
7587 + || d1->d_inode == d2->d_inode
7588 + || d1->d_sb != d2->d_sb);
7590 + if (isdir && au_test_subdir(d1, d2)) {
7591 + di_write_lock_parent(d1);
7592 + di_write_lock_parent2(d2);
7594 + /* there should be no races */
7595 + di_write_lock_parent(d2);
7596 + di_write_lock_parent2(d1);
7600 +void di_write_unlock2(struct dentry *d1, struct dentry *d2)
7602 + di_write_unlock(d1);
7603 + if (d1->d_inode == d2->d_inode)
7604 + au_rw_write_unlock(&au_di(d2)->di_rwsem);
7606 + di_write_unlock(d2);
7609 +/* ---------------------------------------------------------------------- */
7611 +struct dentry *au_h_dptr(struct dentry *dentry, aufs_bindex_t bindex)
7615 + DiMustAnyLock(dentry);
7617 + if (au_dbstart(dentry) < 0 || bindex < au_dbstart(dentry))
7619 + AuDebugOn(bindex < 0);
7620 + d = au_di(dentry)->di_hdentry[0 + bindex].hd_dentry;
7621 + AuDebugOn(d && d->d_count <= 0);
7626 + * extended version of au_h_dptr().
7627 + * returns a hashed and positive h_dentry in bindex, NULL, or error.
7629 +struct dentry *au_h_d_alias(struct dentry *dentry, aufs_bindex_t bindex)
7631 + struct dentry *h_dentry;
7632 + struct inode *inode, *h_inode;
7634 + inode = dentry->d_inode;
7635 + AuDebugOn(!inode);
7638 + if (au_dbstart(dentry) <= bindex
7639 + && bindex <= au_dbend(dentry))
7640 + h_dentry = au_h_dptr(dentry, bindex);
7641 + if (h_dentry && !au_d_hashed_positive(h_dentry)) {
7643 + goto out; /* success */
7646 + AuDebugOn(bindex < au_ibstart(inode));
7647 + AuDebugOn(au_ibend(inode) < bindex);
7648 + h_inode = au_h_iptr(inode, bindex);
7649 + h_dentry = d_find_alias(h_inode);
7651 + if (!IS_ERR(h_dentry)) {
7652 + if (!au_d_hashed_positive(h_dentry))
7653 + goto out; /* success */
7659 + if (au_opt_test(au_mntflags(dentry->d_sb), PLINK)) {
7660 + h_dentry = au_plink_lkup(inode, bindex);
7661 + AuDebugOn(!h_dentry);
7662 + if (!IS_ERR(h_dentry)) {
7663 + if (!au_d_hashed_positive(h_dentry))
7664 + goto out; /* success */
7671 + AuDbgDentry(h_dentry);
7675 +aufs_bindex_t au_dbtail(struct dentry *dentry)
7677 + aufs_bindex_t bend, bwh;
7679 + bend = au_dbend(dentry);
7681 + bwh = au_dbwh(dentry);
7684 + if (0 < bwh && bwh < bend)
7690 +aufs_bindex_t au_dbtaildir(struct dentry *dentry)
7692 + aufs_bindex_t bend, bopq;
7694 + bend = au_dbtail(dentry);
7696 + bopq = au_dbdiropq(dentry);
7697 + if (0 <= bopq && bopq < bend)
7703 +/* ---------------------------------------------------------------------- */
7705 +void au_set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex,
7706 + struct dentry *h_dentry)
7708 + struct au_hdentry *hd = au_di(dentry)->di_hdentry + bindex;
7709 + struct au_branch *br;
7711 + DiMustWriteLock(dentry);
7714 + hd->hd_dentry = h_dentry;
7716 + br = au_sbr(dentry->d_sb, bindex);
7717 + hd->hd_id = br->br_id;
7721 +int au_dbrange_test(struct dentry *dentry)
7724 + aufs_bindex_t bstart, bend;
7727 + bstart = au_dbstart(dentry);
7728 + bend = au_dbend(dentry);
7730 + AuDebugOn(bend < 0 && bstart > bend);
7733 + AuDebugOn(bend >= 0);
7739 +int au_digen_test(struct dentry *dentry, unsigned int sigen)
7744 + if (unlikely(au_digen(dentry) != sigen
7745 + || au_iigen_test(dentry->d_inode, sigen)))
7751 +void au_update_digen(struct dentry *dentry)
7753 + atomic_set(&au_di(dentry)->di_generation, au_sigen(dentry->d_sb));
7754 + /* smp_mb(); */ /* atomic_set */
7757 +void au_update_dbrange(struct dentry *dentry, int do_put_zero)
7759 + struct au_dinfo *dinfo;
7760 + struct dentry *h_d;
7761 + struct au_hdentry *hdp;
7763 + DiMustWriteLock(dentry);
7765 + dinfo = au_di(dentry);
7766 + if (!dinfo || dinfo->di_bstart < 0)
7769 + hdp = dinfo->di_hdentry;
7770 + if (do_put_zero) {
7771 + aufs_bindex_t bindex, bend;
7773 + bend = dinfo->di_bend;
7774 + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++) {
7775 + h_d = hdp[0 + bindex].hd_dentry;
7776 + if (h_d && !h_d->d_inode)
7777 + au_set_h_dptr(dentry, bindex, NULL);
7781 + dinfo->di_bstart = -1;
7782 + while (++dinfo->di_bstart <= dinfo->di_bend)
7783 + if (hdp[0 + dinfo->di_bstart].hd_dentry)
7785 + if (dinfo->di_bstart > dinfo->di_bend) {
7786 + dinfo->di_bstart = -1;
7787 + dinfo->di_bend = -1;
7792 + while (0 <= --dinfo->di_bend)
7793 + if (hdp[0 + dinfo->di_bend].hd_dentry)
7795 + AuDebugOn(dinfo->di_bstart > dinfo->di_bend || dinfo->di_bend < 0);
7798 +void au_update_dbstart(struct dentry *dentry)
7800 + aufs_bindex_t bindex, bend;
7801 + struct dentry *h_dentry;
7803 + bend = au_dbend(dentry);
7804 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
7805 + h_dentry = au_h_dptr(dentry, bindex);
7808 + if (h_dentry->d_inode) {
7809 + au_set_dbstart(dentry, bindex);
7812 + au_set_h_dptr(dentry, bindex, NULL);
7816 +void au_update_dbend(struct dentry *dentry)
7818 + aufs_bindex_t bindex, bstart;
7819 + struct dentry *h_dentry;
7821 + bstart = au_dbstart(dentry);
7822 + for (bindex = au_dbend(dentry); bindex >= bstart; bindex--) {
7823 + h_dentry = au_h_dptr(dentry, bindex);
7826 + if (h_dentry->d_inode) {
7827 + au_set_dbend(dentry, bindex);
7830 + au_set_h_dptr(dentry, bindex, NULL);
7834 +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry)
7836 + aufs_bindex_t bindex, bend;
7838 + bend = au_dbend(dentry);
7839 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++)
7840 + if (au_h_dptr(dentry, bindex) == h_dentry)
7844 diff -urN /usr/share/empty/fs/aufs/dir.c linux/fs/aufs/dir.c
7845 --- /usr/share/empty/fs/aufs/dir.c 1970-01-01 01:00:00.000000000 +0100
7846 +++ linux/fs/aufs/dir.c 2012-01-05 11:00:59.215501779 +0100
7849 + * Copyright (C) 2005-2011 Junjiro R. Okajima
7851 + * This program, aufs is free software; you can redistribute it and/or modify
7852 + * it under the terms of the GNU General Public License as published by
7853 + * the Free Software Foundation; either version 2 of the License, or
7854 + * (at your option) any later version.
7856 + * This program is distributed in the hope that it will be useful,
7857 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
7858 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
7859 + * GNU General Public License for more details.
7861 + * You should have received a copy of the GNU General Public License
7862 + * along with this program; if not, write to the Free Software
7863 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
7867 + * directory operations
7870 +#include <linux/file.h>
7871 +#include <linux/fs_stack.h>
7874 +void au_add_nlink(struct inode *dir, struct inode *h_dir)
7876 + unsigned int nlink;
7878 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
7880 + nlink = dir->i_nlink;
7881 + nlink += h_dir->i_nlink - 2;
7882 + if (h_dir->i_nlink < 2)
7884 + set_nlink(dir, nlink);
7887 +void au_sub_nlink(struct inode *dir, struct inode *h_dir)
7889 + unsigned int nlink;
7891 + AuDebugOn(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode));
7893 + nlink = dir->i_nlink;
7894 + nlink -= h_dir->i_nlink - 2;
7895 + if (h_dir->i_nlink < 2)
7897 + set_nlink(dir, nlink);
7900 +loff_t au_dir_size(struct file *file, struct dentry *dentry)
7903 + aufs_bindex_t bindex, bend;
7904 + struct file *h_file;
7905 + struct dentry *h_dentry;
7909 + AuDebugOn(!file->f_dentry);
7910 + AuDebugOn(!file->f_dentry->d_inode);
7911 + AuDebugOn(!S_ISDIR(file->f_dentry->d_inode->i_mode));
7913 + bend = au_fbend_dir(file);
7914 + for (bindex = au_fbstart(file);
7915 + bindex <= bend && sz < KMALLOC_MAX_SIZE;
7917 + h_file = au_hf_dir(file, bindex);
7919 + && h_file->f_dentry
7920 + && h_file->f_dentry->d_inode)
7921 + sz += i_size_read(h_file->f_dentry->d_inode);
7924 + AuDebugOn(!dentry);
7925 + AuDebugOn(!dentry->d_inode);
7926 + AuDebugOn(!S_ISDIR(dentry->d_inode->i_mode));
7928 + bend = au_dbtaildir(dentry);
7929 + for (bindex = au_dbstart(dentry);
7930 + bindex <= bend && sz < KMALLOC_MAX_SIZE;
7932 + h_dentry = au_h_dptr(dentry, bindex);
7933 + if (h_dentry && h_dentry->d_inode)
7934 + sz += i_size_read(h_dentry->d_inode);
7937 + if (sz < KMALLOC_MAX_SIZE)
7938 + sz = roundup_pow_of_two(sz);
7939 + if (sz > KMALLOC_MAX_SIZE)
7940 + sz = KMALLOC_MAX_SIZE;
7941 + else if (sz < NAME_MAX) {
7942 + BUILD_BUG_ON(AUFS_RDBLK_DEF < NAME_MAX);
7943 + sz = AUFS_RDBLK_DEF;
7948 +/* ---------------------------------------------------------------------- */
7950 +static int reopen_dir(struct file *file)
7953 + unsigned int flags;
7954 + aufs_bindex_t bindex, btail, bstart;
7955 + struct dentry *dentry, *h_dentry;
7956 + struct file *h_file;
7958 + /* open all lower dirs */
7959 + dentry = file->f_dentry;
7960 + bstart = au_dbstart(dentry);
7961 + for (bindex = au_fbstart(file); bindex < bstart; bindex++)
7962 + au_set_h_fptr(file, bindex, NULL);
7963 + au_set_fbstart(file, bstart);
7965 + btail = au_dbtaildir(dentry);
7966 + for (bindex = au_fbend_dir(file); btail < bindex; bindex--)
7967 + au_set_h_fptr(file, bindex, NULL);
7968 + au_set_fbend_dir(file, btail);
7970 + flags = vfsub_file_flags(file);
7971 + for (bindex = bstart; bindex <= btail; bindex++) {
7972 + h_dentry = au_h_dptr(dentry, bindex);
7975 + h_file = au_hf_dir(file, bindex);
7979 + h_file = au_h_open(dentry, bindex, flags, file);
7980 + err = PTR_ERR(h_file);
7981 + if (IS_ERR(h_file))
7982 + goto out; /* close all? */
7983 + au_set_h_fptr(file, bindex, h_file);
7985 + au_update_figen(file);
7986 + /* todo: necessary? */
7987 + /* file->f_ra = h_file->f_ra; */
7994 +static int do_open_dir(struct file *file, int flags)
7997 + aufs_bindex_t bindex, btail;
7998 + struct dentry *dentry, *h_dentry;
7999 + struct file *h_file;
8001 + FiMustWriteLock(file);
8003 + dentry = file->f_dentry;
8004 + err = au_alive_dir(dentry);
8005 + if (unlikely(err))
8008 + file->f_version = dentry->d_inode->i_version;
8009 + bindex = au_dbstart(dentry);
8010 + au_set_fbstart(file, bindex);
8011 + btail = au_dbtaildir(dentry);
8012 + au_set_fbend_dir(file, btail);
8013 + for (; !err && bindex <= btail; bindex++) {
8014 + h_dentry = au_h_dptr(dentry, bindex);
8018 + h_file = au_h_open(dentry, bindex, flags, file);
8019 + if (IS_ERR(h_file)) {
8020 + err = PTR_ERR(h_file);
8023 + au_set_h_fptr(file, bindex, h_file);
8025 + au_update_figen(file);
8026 + /* todo: necessary? */
8027 + /* file->f_ra = h_file->f_ra; */
8029 + return 0; /* success */
8032 + for (bindex = au_fbstart(file); bindex <= btail; bindex++)
8033 + au_set_h_fptr(file, bindex, NULL);
8034 + au_set_fbstart(file, -1);
8035 + au_set_fbend_dir(file, -1);
8041 +static int aufs_open_dir(struct inode *inode __maybe_unused,
8042 + struct file *file)
8045 + struct super_block *sb;
8046 + struct au_fidir *fidir;
8049 + sb = file->f_dentry->d_sb;
8050 + si_read_lock(sb, AuLock_FLUSH);
8051 + fidir = au_fidir_alloc(sb);
8053 + err = au_do_open(file, do_open_dir, fidir);
8054 + if (unlikely(err))
8057 + si_read_unlock(sb);
8061 +static int aufs_release_dir(struct inode *inode __maybe_unused,
8062 + struct file *file)
8064 + struct au_vdir *vdir_cache;
8065 + struct au_finfo *finfo;
8066 + struct au_fidir *fidir;
8067 + aufs_bindex_t bindex, bend;
8069 + finfo = au_fi(file);
8070 + fidir = finfo->fi_hdir;
8072 + /* remove me from sb->s_files */
8073 + file_sb_list_del(file);
8075 + vdir_cache = fidir->fd_vdir_cache; /* lock-free */
8077 + au_vdir_free(vdir_cache);
8079 + bindex = finfo->fi_btop;
8080 + if (bindex >= 0) {
8082 + * calls fput() instead of filp_close(),
8083 + * since no dnotify or lock for the lower file.
8085 + bend = fidir->fd_bbot;
8086 + for (; bindex <= bend; bindex++)
8087 + au_set_h_fptr(file, bindex, NULL);
8090 + finfo->fi_hdir = NULL;
8092 + au_finfo_fin(file);
8096 +/* ---------------------------------------------------------------------- */
8098 +static int au_do_flush_dir(struct file *file, fl_owner_t id)
8101 + aufs_bindex_t bindex, bend;
8102 + struct file *h_file;
8105 + bend = au_fbend_dir(file);
8106 + for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
8107 + h_file = au_hf_dir(file, bindex);
8109 + err = vfsub_flush(h_file, id);
8114 +static int aufs_flush_dir(struct file *file, fl_owner_t id)
8116 + return au_do_flush(file, id, au_do_flush_dir);
8119 +/* ---------------------------------------------------------------------- */
8121 +static int au_do_fsync_dir_no_file(struct dentry *dentry, int datasync)
8124 + aufs_bindex_t bend, bindex;
8125 + struct inode *inode;
8126 + struct super_block *sb;
8129 + sb = dentry->d_sb;
8130 + inode = dentry->d_inode;
8132 + bend = au_dbend(dentry);
8133 + for (bindex = au_dbstart(dentry); !err && bindex <= bend; bindex++) {
8134 + struct path h_path;
8136 + if (au_test_ro(sb, bindex, inode))
8138 + h_path.dentry = au_h_dptr(dentry, bindex);
8139 + if (!h_path.dentry)
8142 + h_path.mnt = au_sbr_mnt(sb, bindex);
8143 + err = vfsub_fsync(NULL, &h_path, datasync);
8149 +static int au_do_fsync_dir(struct file *file, int datasync)
8152 + aufs_bindex_t bend, bindex;
8153 + struct file *h_file;
8154 + struct super_block *sb;
8155 + struct inode *inode;
8157 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
8158 + if (unlikely(err))
8161 + sb = file->f_dentry->d_sb;
8162 + inode = file->f_dentry->d_inode;
8163 + bend = au_fbend_dir(file);
8164 + for (bindex = au_fbstart(file); !err && bindex <= bend; bindex++) {
8165 + h_file = au_hf_dir(file, bindex);
8166 + if (!h_file || au_test_ro(sb, bindex, inode))
8169 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
8177 + * @file may be NULL
8179 +static int aufs_fsync_dir(struct file *file, loff_t start, loff_t end,
8183 + struct dentry *dentry;
8184 + struct super_block *sb;
8185 + struct mutex *mtx;
8188 + dentry = file->f_dentry;
8189 + mtx = &dentry->d_inode->i_mutex;
8191 + sb = dentry->d_sb;
8192 + si_noflush_read_lock(sb);
8194 + err = au_do_fsync_dir(file, datasync);
8196 + di_write_lock_child(dentry);
8197 + err = au_do_fsync_dir_no_file(dentry, datasync);
8199 + au_cpup_attr_timesizes(dentry->d_inode);
8200 + di_write_unlock(dentry);
8202 + fi_write_unlock(file);
8204 + si_read_unlock(sb);
8205 + mutex_unlock(mtx);
8209 +/* ---------------------------------------------------------------------- */
8211 +static int aufs_readdir(struct file *file, void *dirent, filldir_t filldir)
8214 + struct dentry *dentry;
8215 + struct inode *inode, *h_inode;
8216 + struct super_block *sb;
8218 + dentry = file->f_dentry;
8219 + inode = dentry->d_inode;
8222 + sb = dentry->d_sb;
8223 + si_read_lock(sb, AuLock_FLUSH);
8224 + err = au_reval_and_lock_fdi(file, reopen_dir, /*wlock*/1);
8225 + if (unlikely(err))
8227 + err = au_alive_dir(dentry);
8229 + err = au_vdir_init(file);
8230 + di_downgrade_lock(dentry, AuLock_IR);
8231 + if (unlikely(err))
8234 + h_inode = au_h_iptr(inode, au_ibstart(inode));
8235 + if (!au_test_nfsd()) {
8236 + err = au_vdir_fill_de(file, dirent, filldir);
8237 + fsstack_copy_attr_atime(inode, h_inode);
8240 + * nfsd filldir may call lookup_one_len(), vfs_getattr(),
8241 + * encode_fh() and others.
8243 + atomic_inc(&h_inode->i_count);
8244 + di_read_unlock(dentry, AuLock_IR);
8245 + si_read_unlock(sb);
8246 + err = au_vdir_fill_de(file, dirent, filldir);
8247 + fsstack_copy_attr_atime(inode, h_inode);
8248 + fi_write_unlock(file);
8256 + di_read_unlock(dentry, AuLock_IR);
8257 + fi_write_unlock(file);
8259 + si_read_unlock(sb);
8263 +/* ---------------------------------------------------------------------- */
8265 +#define AuTestEmpty_WHONLY 1
8266 +#define AuTestEmpty_CALLED (1 << 1)
8267 +#define AuTestEmpty_SHWH (1 << 2)
8268 +#define au_ftest_testempty(flags, name) ((flags) & AuTestEmpty_##name)
8269 +#define au_fset_testempty(flags, name) \
8270 + do { (flags) |= AuTestEmpty_##name; } while (0)
8271 +#define au_fclr_testempty(flags, name) \
8272 + do { (flags) &= ~AuTestEmpty_##name; } while (0)
8274 +#ifndef CONFIG_AUFS_SHWH
8275 +#undef AuTestEmpty_SHWH
8276 +#define AuTestEmpty_SHWH 0
8279 +struct test_empty_arg {
8280 + struct au_nhash *whlist;
8281 + unsigned int flags;
8283 + aufs_bindex_t bindex;
8286 +static int test_empty_cb(void *__arg, const char *__name, int namelen,
8287 + loff_t offset __maybe_unused, u64 ino,
8288 + unsigned int d_type)
8290 + struct test_empty_arg *arg = __arg;
8291 + char *name = (void *)__name;
8294 + au_fset_testempty(arg->flags, CALLED);
8296 + if (name[0] == '.'
8297 + && (namelen == 1 || (name[1] == '.' && namelen == 2)))
8298 + goto out; /* success */
8300 + if (namelen <= AUFS_WH_PFX_LEN
8301 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
8302 + if (au_ftest_testempty(arg->flags, WHONLY)
8303 + && !au_nhash_test_known_wh(arg->whlist, name, namelen))
8304 + arg->err = -ENOTEMPTY;
8308 + name += AUFS_WH_PFX_LEN;
8309 + namelen -= AUFS_WH_PFX_LEN;
8310 + if (!au_nhash_test_known_wh(arg->whlist, name, namelen))
8311 + arg->err = au_nhash_append_wh
8312 + (arg->whlist, name, namelen, ino, d_type, arg->bindex,
8313 + au_ftest_testempty(arg->flags, SHWH));
8317 + AuTraceErr(arg->err);
8321 +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
8324 + struct file *h_file;
8326 + h_file = au_h_open(dentry, arg->bindex,
8327 + O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_LARGEFILE,
8329 + err = PTR_ERR(h_file);
8330 + if (IS_ERR(h_file))
8334 + if (!au_opt_test(au_mntflags(dentry->d_sb), UDBA_NONE)
8335 + && !h_file->f_dentry->d_inode->i_nlink)
8340 + au_fclr_testempty(arg->flags, CALLED);
8342 + err = vfsub_readdir(h_file, test_empty_cb, arg);
8345 + } while (!err && au_ftest_testempty(arg->flags, CALLED));
8349 + au_sbr_put(dentry->d_sb, arg->bindex);
8354 +struct do_test_empty_args {
8356 + struct dentry *dentry;
8357 + struct test_empty_arg *arg;
8360 +static void call_do_test_empty(void *args)
8362 + struct do_test_empty_args *a = args;
8363 + *a->errp = do_test_empty(a->dentry, a->arg);
8366 +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg)
8369 + struct dentry *h_dentry;
8370 + struct inode *h_inode;
8372 + h_dentry = au_h_dptr(dentry, arg->bindex);
8373 + h_inode = h_dentry->d_inode;
8374 + /* todo: i_mode changes anytime? */
8375 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
8376 + err = au_test_h_perm_sio(h_inode, MAY_EXEC | MAY_READ);
8377 + mutex_unlock(&h_inode->i_mutex);
8379 + err = do_test_empty(dentry, arg);
8381 + struct do_test_empty_args args = {
8386 + unsigned int flags = arg->flags;
8388 + wkq_err = au_wkq_wait(call_do_test_empty, &args);
8389 + if (unlikely(wkq_err))
8391 + arg->flags = flags;
8397 +int au_test_empty_lower(struct dentry *dentry)
8400 + unsigned int rdhash;
8401 + aufs_bindex_t bindex, bstart, btail;
8402 + struct au_nhash whlist;
8403 + struct test_empty_arg arg;
8405 + SiMustAnyLock(dentry->d_sb);
8407 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
8409 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, dentry));
8410 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
8411 + if (unlikely(err))
8415 + arg.whlist = &whlist;
8416 + bstart = au_dbstart(dentry);
8417 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
8418 + au_fset_testempty(arg.flags, SHWH);
8419 + arg.bindex = bstart;
8420 + err = do_test_empty(dentry, &arg);
8421 + if (unlikely(err))
8424 + au_fset_testempty(arg.flags, WHONLY);
8425 + btail = au_dbtaildir(dentry);
8426 + for (bindex = bstart + 1; !err && bindex <= btail; bindex++) {
8427 + struct dentry *h_dentry;
8429 + h_dentry = au_h_dptr(dentry, bindex);
8430 + if (h_dentry && h_dentry->d_inode) {
8431 + arg.bindex = bindex;
8432 + err = do_test_empty(dentry, &arg);
8437 + au_nhash_wh_free(&whlist);
8442 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist)
8445 + struct test_empty_arg arg;
8446 + aufs_bindex_t bindex, btail;
8449 + arg.whlist = whlist;
8450 + arg.flags = AuTestEmpty_WHONLY;
8451 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH))
8452 + au_fset_testempty(arg.flags, SHWH);
8453 + btail = au_dbtaildir(dentry);
8454 + for (bindex = au_dbstart(dentry); !err && bindex <= btail; bindex++) {
8455 + struct dentry *h_dentry;
8457 + h_dentry = au_h_dptr(dentry, bindex);
8458 + if (h_dentry && h_dentry->d_inode) {
8459 + arg.bindex = bindex;
8460 + err = sio_test_empty(dentry, &arg);
8467 +/* ---------------------------------------------------------------------- */
8469 +const struct file_operations aufs_dir_fop = {
8470 + .owner = THIS_MODULE,
8471 + .llseek = default_llseek,
8472 + .read = generic_read_dir,
8473 + .readdir = aufs_readdir,
8474 + .unlocked_ioctl = aufs_ioctl_dir,
8475 +#ifdef CONFIG_COMPAT
8476 + .compat_ioctl = aufs_compat_ioctl_dir,
8478 + .open = aufs_open_dir,
8479 + .release = aufs_release_dir,
8480 + .flush = aufs_flush_dir,
8481 + .fsync = aufs_fsync_dir
8483 diff -urN /usr/share/empty/fs/aufs/dir.h linux/fs/aufs/dir.h
8484 --- /usr/share/empty/fs/aufs/dir.h 1970-01-01 01:00:00.000000000 +0100
8485 +++ linux/fs/aufs/dir.h 2011-08-24 13:30:24.731313534 +0200
8488 + * Copyright (C) 2005-2011 Junjiro R. Okajima
8490 + * This program, aufs is free software; you can redistribute it and/or modify
8491 + * it under the terms of the GNU General Public License as published by
8492 + * the Free Software Foundation; either version 2 of the License, or
8493 + * (at your option) any later version.
8495 + * This program is distributed in the hope that it will be useful,
8496 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8497 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8498 + * GNU General Public License for more details.
8500 + * You should have received a copy of the GNU General Public License
8501 + * along with this program; if not, write to the Free Software
8502 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
8506 + * directory operations
8509 +#ifndef __AUFS_DIR_H__
8510 +#define __AUFS_DIR_H__
8514 +#include <linux/fs.h>
8515 +#include <linux/aufs_type.h>
8517 +/* ---------------------------------------------------------------------- */
8519 +/* need to be faster and smaller */
8522 + unsigned int nh_num;
8523 + struct hlist_head *nh_head;
8526 +struct au_vdir_destr {
8527 + unsigned char len;
8528 + unsigned char name[0];
8531 +struct au_vdir_dehstr {
8532 + struct hlist_node hash;
8533 + struct au_vdir_destr *str;
8534 +} ____cacheline_aligned_in_smp;
8536 +struct au_vdir_de {
8538 + unsigned char de_type;
8539 + /* caution: packed */
8540 + struct au_vdir_destr de_str;
8543 +struct au_vdir_wh {
8544 + struct hlist_node wh_hash;
8545 +#ifdef CONFIG_AUFS_SHWH
8547 + aufs_bindex_t wh_bindex;
8548 + unsigned char wh_type;
8550 + aufs_bindex_t wh_bindex;
8552 + /* caution: packed */
8553 + struct au_vdir_destr wh_str;
8556 +union au_vdir_deblk_p {
8557 + unsigned char *deblk;
8558 + struct au_vdir_de *de;
8562 + unsigned char **vd_deblk;
8563 + unsigned long vd_nblk;
8566 + union au_vdir_deblk_p p;
8569 + unsigned long vd_version;
8570 + unsigned int vd_deblk_sz;
8571 + unsigned long vd_jiffy;
8572 +} ____cacheline_aligned_in_smp;
8574 +/* ---------------------------------------------------------------------- */
8577 +extern const struct file_operations aufs_dir_fop;
8578 +void au_add_nlink(struct inode *dir, struct inode *h_dir);
8579 +void au_sub_nlink(struct inode *dir, struct inode *h_dir);
8580 +loff_t au_dir_size(struct file *file, struct dentry *dentry);
8581 +int au_test_empty_lower(struct dentry *dentry);
8582 +int au_test_empty(struct dentry *dentry, struct au_nhash *whlist);
8585 +unsigned int au_rdhash_est(loff_t sz);
8586 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp);
8587 +void au_nhash_wh_free(struct au_nhash *whlist);
8588 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
8590 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen);
8591 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
8592 + unsigned int d_type, aufs_bindex_t bindex,
8593 + unsigned char shwh);
8594 +void au_vdir_free(struct au_vdir *vdir);
8595 +int au_vdir_init(struct file *file);
8596 +int au_vdir_fill_de(struct file *file, void *dirent, filldir_t filldir);
8599 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg);
8601 +#ifdef CONFIG_AUFS_RDU
8603 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
8604 +#ifdef CONFIG_COMPAT
8605 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
8606 + unsigned long arg);
8609 +static inline long au_rdu_ioctl(struct file *file, unsigned int cmd,
8610 + unsigned long arg)
8614 +#ifdef CONFIG_COMPAT
8615 +static inline long au_rdu_compat_ioctl(struct file *file, unsigned int cmd,
8616 + unsigned long arg)
8623 +#endif /* __KERNEL__ */
8624 +#endif /* __AUFS_DIR_H__ */
8625 diff -urN /usr/share/empty/fs/aufs/dynop.c linux/fs/aufs/dynop.c
8626 --- /usr/share/empty/fs/aufs/dynop.c 1970-01-01 01:00:00.000000000 +0100
8627 +++ linux/fs/aufs/dynop.c 2011-08-24 13:30:24.731313534 +0200
8630 + * Copyright (C) 2010-2011 Junjiro R. Okajima
8632 + * This program, aufs is free software; you can redistribute it and/or modify
8633 + * it under the terms of the GNU General Public License as published by
8634 + * the Free Software Foundation; either version 2 of the License, or
8635 + * (at your option) any later version.
8637 + * This program is distributed in the hope that it will be useful,
8638 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
8639 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
8640 + * GNU General Public License for more details.
8642 + * You should have received a copy of the GNU General Public License
8643 + * along with this program; if not, write to the Free Software
8644 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
8648 + * dynamically customizable operations for regular files
8653 +#define DyPrSym(key) AuDbgSym(key->dk_op.dy_hop)
8656 + * How large will these lists be?
8657 + * Usually just a few elements, 20-30 at most for each, I guess.
8659 +static struct au_splhead dynop[AuDyLast];
8661 +static struct au_dykey *dy_gfind_get(struct au_splhead *spl, const void *h_op)
8663 + struct au_dykey *key, *tmp;
8664 + struct list_head *head;
8667 + head = &spl->head;
8669 + list_for_each_entry_rcu(tmp, head, dk_list)
8670 + if (tmp->dk_op.dy_hop == h_op) {
8672 + kref_get(&key->dk_kref);
8675 + rcu_read_unlock();
8680 +static struct au_dykey *dy_bradd(struct au_branch *br, struct au_dykey *key)
8682 + struct au_dykey **k, *found;
8683 + const void *h_op = key->dk_op.dy_hop;
8688 + for (i = 0; i < AuBrDynOp; i++)
8690 + if (k[i]->dk_op.dy_hop == h_op) {
8697 + spin_lock(&br->br_dykey_lock);
8698 + for (; i < AuBrDynOp; i++)
8700 + if (k[i]->dk_op.dy_hop == h_op) {
8708 + spin_unlock(&br->br_dykey_lock);
8709 + BUG_ON(i == AuBrDynOp); /* expand the array */
8715 +/* kref_get() if @key is already added */
8716 +static struct au_dykey *dy_gadd(struct au_splhead *spl, struct au_dykey *key)
8718 + struct au_dykey *tmp, *found;
8719 + struct list_head *head;
8720 + const void *h_op = key->dk_op.dy_hop;
8723 + head = &spl->head;
8724 + spin_lock(&spl->spin);
8725 + list_for_each_entry(tmp, head, dk_list)
8726 + if (tmp->dk_op.dy_hop == h_op) {
8727 + kref_get(&tmp->dk_kref);
8732 + list_add_rcu(&key->dk_list, head);
8733 + spin_unlock(&spl->spin);
8740 +static void dy_free_rcu(struct rcu_head *rcu)
8742 + struct au_dykey *key;
8744 + key = container_of(rcu, struct au_dykey, dk_rcu);
8749 +static void dy_free(struct kref *kref)
8751 + struct au_dykey *key;
8752 + struct au_splhead *spl;
8754 + key = container_of(kref, struct au_dykey, dk_kref);
8755 + spl = dynop + key->dk_op.dy_type;
8756 + au_spl_del_rcu(&key->dk_list, spl);
8757 + call_rcu(&key->dk_rcu, dy_free_rcu);
8760 +void au_dy_put(struct au_dykey *key)
8762 + kref_put(&key->dk_kref, dy_free);
8765 +/* ---------------------------------------------------------------------- */
8767 +#define DyDbgSize(cnt, op) AuDebugOn(cnt != sizeof(op)/sizeof(void *))
8769 +#ifdef CONFIG_AUFS_DEBUG
8770 +#define DyDbgDeclare(cnt) unsigned int cnt = 0
8771 +#define DyDbgInc(cnt) do { cnt++; } while (0)
8773 +#define DyDbgDeclare(cnt) do {} while (0)
8774 +#define DyDbgInc(cnt) do {} while (0)
8777 +#define DySet(func, dst, src, h_op, h_sb) do { \
8779 + if (h_op->func) { \
8781 + dst.func = src.func; \
8783 + AuDbg("%s %s\n", au_sbtype(h_sb), #func); \
8787 +#define DySetForce(func, dst, src) do { \
8788 + AuDebugOn(!src.func); \
8790 + dst.func = src.func; \
8793 +#define DySetAop(func) \
8794 + DySet(func, dyaop->da_op, aufs_aop, h_aop, h_sb)
8795 +#define DySetAopForce(func) \
8796 + DySetForce(func, dyaop->da_op, aufs_aop)
8798 +static void dy_aop(struct au_dykey *key, const void *h_op,
8799 + struct super_block *h_sb __maybe_unused)
8801 + struct au_dyaop *dyaop = (void *)key;
8802 + const struct address_space_operations *h_aop = h_op;
8803 + DyDbgDeclare(cnt);
8805 + AuDbg("%s\n", au_sbtype(h_sb));
8807 + DySetAop(writepage);
8808 + DySetAopForce(readpage); /* force */
8809 + DySetAop(writepages);
8810 + DySetAop(set_page_dirty);
8811 + DySetAop(readpages);
8812 + DySetAop(write_begin);
8813 + DySetAop(write_end);
8815 + DySetAop(invalidatepage);
8816 + DySetAop(releasepage);
8817 + DySetAop(freepage);
8818 + /* these two will be changed according to an aufs mount option */
8819 + DySetAop(direct_IO);
8820 + DySetAop(get_xip_mem);
8821 + DySetAop(migratepage);
8822 + DySetAop(launder_page);
8823 + DySetAop(is_partially_uptodate);
8824 + DySetAop(error_remove_page);
8826 + DyDbgSize(cnt, *h_aop);
8827 + dyaop->da_get_xip_mem = h_aop->get_xip_mem;
8830 +/* ---------------------------------------------------------------------- */
8832 +static void dy_bug(struct kref *kref)
8837 +static struct au_dykey *dy_get(struct au_dynop *op, struct au_branch *br)
8839 + struct au_dykey *key, *old;
8840 + struct au_splhead *spl;
8843 + void (*set)(struct au_dykey *key, const void *h_op,
8844 + struct super_block *h_sb __maybe_unused);
8846 + static const struct op a[] = {
8848 + .sz = sizeof(struct au_dyaop),
8852 + const struct op *p;
8854 + spl = dynop + op->dy_type;
8855 + key = dy_gfind_get(spl, op->dy_hop);
8857 + goto out_add; /* success */
8859 + p = a + op->dy_type;
8860 + key = kzalloc(p->sz, GFP_NOFS);
8861 + if (unlikely(!key)) {
8862 + key = ERR_PTR(-ENOMEM);
8866 + key->dk_op.dy_hop = op->dy_hop;
8867 + kref_init(&key->dk_kref);
8868 + p->set(key, op->dy_hop, br->br_mnt->mnt_sb);
8869 + old = dy_gadd(spl, key);
8876 + old = dy_bradd(br, key);
8878 + /* its ref-count should never be zero here */
8879 + kref_put(&key->dk_kref, dy_bug);
8884 +/* ---------------------------------------------------------------------- */
8886 + * Aufs prohibits O_DIRECT by defaut even if the branch supports it.
8887 + * This behaviour is neccessary to return an error from open(O_DIRECT) instead
8888 + * of the succeeding I/O. The dio mount option enables O_DIRECT and makes
8889 + * open(O_DIRECT) always succeed, but the succeeding I/O may return an error.
8890 + * See the aufs manual in detail.
8892 + * To keep this behaviour, aufs has to set NULL to ->get_xip_mem too, and the
8893 + * performance of fadvise() and madvise() may be affected.
8895 +static void dy_adx(struct au_dyaop *dyaop, int do_dx)
8898 + dyaop->da_op.direct_IO = NULL;
8899 + dyaop->da_op.get_xip_mem = NULL;
8901 + dyaop->da_op.direct_IO = aufs_aop.direct_IO;
8902 + dyaop->da_op.get_xip_mem = aufs_aop.get_xip_mem;
8903 + if (!dyaop->da_get_xip_mem)
8904 + dyaop->da_op.get_xip_mem = NULL;
8908 +static struct au_dyaop *dy_aget(struct au_branch *br,
8909 + const struct address_space_operations *h_aop,
8912 + struct au_dyaop *dyaop;
8913 + struct au_dynop op;
8915 + op.dy_type = AuDy_AOP;
8916 + op.dy_haop = h_aop;
8917 + dyaop = (void *)dy_get(&op, br);
8918 + if (IS_ERR(dyaop))
8920 + dy_adx(dyaop, do_dx);
8926 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
8927 + struct inode *h_inode)
8930 + struct super_block *sb;
8931 + struct au_branch *br;
8932 + struct au_dyaop *dyaop;
8934 + AuDebugOn(!S_ISREG(h_inode->i_mode));
8935 + IiMustWriteLock(inode);
8938 + br = au_sbr(sb, bindex);
8939 + do_dx = !!au_opt_test(au_mntflags(sb), DIO);
8940 + dyaop = dy_aget(br, h_inode->i_mapping->a_ops, do_dx);
8941 + err = PTR_ERR(dyaop);
8942 + if (IS_ERR(dyaop))
8943 + /* unnecessary to call dy_fput() */
8947 + inode->i_mapping->a_ops = &dyaop->da_op;
8954 + * Is it safe to replace a_ops during the inode/file is in operation?
8957 +int au_dy_irefresh(struct inode *inode)
8960 + aufs_bindex_t bstart;
8961 + struct inode *h_inode;
8964 + if (S_ISREG(inode->i_mode)) {
8965 + bstart = au_ibstart(inode);
8966 + h_inode = au_h_iptr(inode, bstart);
8967 + err = au_dy_iaop(inode, bstart, h_inode);
8972 +void au_dy_arefresh(int do_dx)
8974 + struct au_splhead *spl;
8975 + struct list_head *head;
8976 + struct au_dykey *key;
8978 + spl = dynop + AuDy_AOP;
8979 + head = &spl->head;
8980 + spin_lock(&spl->spin);
8981 + list_for_each_entry(key, head, dk_list)
8982 + dy_adx((void *)key, do_dx);
8983 + spin_unlock(&spl->spin);
8986 +/* ---------------------------------------------------------------------- */
8988 +void __init au_dy_init(void)
8992 + /* make sure that 'struct au_dykey *' can be any type */
8993 + BUILD_BUG_ON(offsetof(struct au_dyaop, da_key));
8995 + for (i = 0; i < AuDyLast; i++)
8996 + au_spl_init(dynop + i);
8999 +void au_dy_fin(void)
9003 + for (i = 0; i < AuDyLast; i++)
9004 + WARN_ON(!list_empty(&dynop[i].head));
9006 diff -urN /usr/share/empty/fs/aufs/dynop.h linux/fs/aufs/dynop.h
9007 --- /usr/share/empty/fs/aufs/dynop.h 1970-01-01 01:00:00.000000000 +0100
9008 +++ linux/fs/aufs/dynop.h 2011-08-24 13:30:24.731313534 +0200
9011 + * Copyright (C) 2010-2011 Junjiro R. Okajima
9013 + * This program, aufs is free software; you can redistribute it and/or modify
9014 + * it under the terms of the GNU General Public License as published by
9015 + * the Free Software Foundation; either version 2 of the License, or
9016 + * (at your option) any later version.
9018 + * This program is distributed in the hope that it will be useful,
9019 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9020 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9021 + * GNU General Public License for more details.
9023 + * You should have received a copy of the GNU General Public License
9024 + * along with this program; if not, write to the Free Software
9025 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
9029 + * dynamically customizable operations (for regular files only)
9032 +#ifndef __AUFS_DYNOP_H__
9033 +#define __AUFS_DYNOP_H__
9037 +#include <linux/fs.h>
9038 +#include <linux/mm.h>
9039 +#include <linux/rcupdate.h>
9040 +#include <linux/aufs_type.h>
9043 +enum {AuDy_AOP, AuDyLast};
9048 + const void *dy_hop;
9049 + const struct address_space_operations *dy_haop;
9055 + struct list_head dk_list;
9056 + struct rcu_head dk_rcu;
9058 + struct au_dynop dk_op;
9061 + * during I am in the branch local array, kref is gotten. when the
9062 + * branch is removed, kref is put.
9064 + struct kref dk_kref;
9067 +/* stop unioning since their sizes are very different from each other */
9069 + struct au_dykey da_key;
9070 + struct address_space_operations da_op; /* not const */
9071 + int (*da_get_xip_mem)(struct address_space *, pgoff_t, int,
9072 + void **, unsigned long *);
9075 +/* ---------------------------------------------------------------------- */
9079 +void au_dy_put(struct au_dykey *key);
9080 +int au_dy_iaop(struct inode *inode, aufs_bindex_t bindex,
9081 + struct inode *h_inode);
9082 +int au_dy_irefresh(struct inode *inode);
9083 +void au_dy_arefresh(int do_dio);
9085 +void __init au_dy_init(void);
9086 +void au_dy_fin(void);
9088 +#endif /* __KERNEL__ */
9089 +#endif /* __AUFS_DYNOP_H__ */
9090 diff -urN /usr/share/empty/fs/aufs/export.c linux/fs/aufs/export.c
9091 --- /usr/share/empty/fs/aufs/export.c 1970-01-01 01:00:00.000000000 +0100
9092 +++ linux/fs/aufs/export.c 2011-08-24 13:30:24.731313534 +0200
9095 + * Copyright (C) 2005-2011 Junjiro R. Okajima
9097 + * This program, aufs is free software; you can redistribute it and/or modify
9098 + * it under the terms of the GNU General Public License as published by
9099 + * the Free Software Foundation; either version 2 of the License, or
9100 + * (at your option) any later version.
9102 + * This program is distributed in the hope that it will be useful,
9103 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9104 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9105 + * GNU General Public License for more details.
9107 + * You should have received a copy of the GNU General Public License
9108 + * along with this program; if not, write to the Free Software
9109 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
9116 +#include <linux/exportfs.h>
9117 +#include <linux/file.h>
9118 +#include <linux/mnt_namespace.h>
9119 +#include <linux/namei.h>
9120 +#include <linux/nsproxy.h>
9121 +#include <linux/random.h>
9122 +#include <linux/writeback.h>
9126 +#ifdef CONFIG_AUFS_INO_T_64
9134 +static ino_t decode_ino(__u32 *a)
9138 + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a));
9140 +#ifdef CONFIG_AUFS_INO_T_64
9146 +static void encode_ino(__u32 *a, ino_t ino)
9152 +#ifdef CONFIG_AUFS_INO_T_64
9157 +/* NFS file handle */
9161 +#ifdef CONFIG_AUFS_INO_T_64
9162 + /* support 64bit inode number */
9176 + Fh_dir_ino = Fh_dir_ino1
9179 +static int au_test_anon(struct dentry *dentry)
9181 + /* note: read d_flags without d_lock */
9182 + return !!(dentry->d_flags & DCACHE_DISCONNECTED);
9185 +/* ---------------------------------------------------------------------- */
9186 +/* inode generation external table */
9188 +void au_xigen_inc(struct inode *inode)
9193 + struct super_block *sb;
9194 + struct au_sbinfo *sbinfo;
9197 + AuDebugOn(!au_opt_test(au_mntflags(sb), XINO));
9199 + sbinfo = au_sbi(sb);
9200 + pos = inode->i_ino;
9201 + pos *= sizeof(igen);
9202 + igen = inode->i_generation + 1;
9203 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xigen, &igen,
9204 + sizeof(igen), &pos);
9205 + if (sz == sizeof(igen))
9206 + return; /* success */
9208 + if (unlikely(sz >= 0))
9209 + AuIOErr("xigen error (%zd)\n", sz);
9212 +int au_xigen_new(struct inode *inode)
9217 + struct super_block *sb;
9218 + struct au_sbinfo *sbinfo;
9219 + struct file *file;
9222 + /* todo: dirty, at mount time */
9223 + if (inode->i_ino == AUFS_ROOT_INO)
9226 + SiMustAnyLock(sb);
9227 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
9231 + pos = inode->i_ino;
9232 + if (unlikely(au_loff_max / sizeof(inode->i_generation) - 1 < pos)) {
9233 + AuIOErr1("too large i%lld\n", pos);
9236 + pos *= sizeof(inode->i_generation);
9239 + sbinfo = au_sbi(sb);
9240 + file = sbinfo->si_xigen;
9243 + if (i_size_read(file->f_dentry->d_inode)
9244 + < pos + sizeof(inode->i_generation)) {
9245 + inode->i_generation = atomic_inc_return(&sbinfo->si_xigen_next);
9246 + sz = xino_fwrite(sbinfo->si_xwrite, file, &inode->i_generation,
9247 + sizeof(inode->i_generation), &pos);
9249 + sz = xino_fread(sbinfo->si_xread, file, &inode->i_generation,
9250 + sizeof(inode->i_generation), &pos);
9251 + if (sz == sizeof(inode->i_generation))
9252 + goto out; /* success */
9255 + if (unlikely(sz >= 0)) {
9257 + AuIOErr("xigen error (%zd)\n", sz);
9264 +int au_xigen_set(struct super_block *sb, struct file *base)
9267 + struct au_sbinfo *sbinfo;
9268 + struct file *file;
9270 + SiMustWriteLock(sb);
9272 + sbinfo = au_sbi(sb);
9273 + file = au_xino_create2(base, sbinfo->si_xigen);
9274 + err = PTR_ERR(file);
9278 + if (sbinfo->si_xigen)
9279 + fput(sbinfo->si_xigen);
9280 + sbinfo->si_xigen = file;
9286 +void au_xigen_clr(struct super_block *sb)
9288 + struct au_sbinfo *sbinfo;
9290 + SiMustWriteLock(sb);
9292 + sbinfo = au_sbi(sb);
9293 + if (sbinfo->si_xigen) {
9294 + fput(sbinfo->si_xigen);
9295 + sbinfo->si_xigen = NULL;
9299 +/* ---------------------------------------------------------------------- */
9301 +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino,
9304 + struct dentry *dentry, *d;
9305 + struct inode *inode;
9306 + unsigned int sigen;
9309 + inode = ilookup(sb, ino);
9313 + dentry = ERR_PTR(-ESTALE);
9314 + sigen = au_sigen(sb);
9315 + if (unlikely(is_bad_inode(inode)
9316 + || IS_DEADDIR(inode)
9317 + || sigen != au_iigen(inode)))
9321 + if (!dir_ino || S_ISDIR(inode->i_mode))
9322 + dentry = d_find_alias(inode);
9324 + spin_lock(&inode->i_lock);
9325 + list_for_each_entry(d, &inode->i_dentry, d_alias) {
9326 + spin_lock(&d->d_lock);
9327 + if (!au_test_anon(d)
9328 + && d->d_parent->d_inode->i_ino == dir_ino) {
9329 + dentry = dget_dlock(d);
9330 + spin_unlock(&d->d_lock);
9333 + spin_unlock(&d->d_lock);
9335 + spin_unlock(&inode->i_lock);
9337 + if (unlikely(dentry && au_digen_test(dentry, sigen))) {
9338 + /* need to refresh */
9346 + AuTraceErrPtr(dentry);
9350 +/* ---------------------------------------------------------------------- */
9353 +/* if exportfs_decode_fh() passed vfsmount*, we could be happy */
9355 +struct au_compare_mnt_args {
9357 + struct super_block *sb;
9360 + struct vfsmount *mnt;
9363 +static int au_compare_mnt(struct vfsmount *mnt, void *arg)
9365 + struct au_compare_mnt_args *a = arg;
9367 + if (mnt->mnt_sb != a->sb)
9369 + a->mnt = mntget(mnt);
9373 +static struct vfsmount *au_mnt_get(struct super_block *sb)
9376 + struct au_compare_mnt_args args = {
9379 + struct mnt_namespace *ns;
9381 + br_read_lock(vfsmount_lock);
9382 + /* no get/put ?? */
9383 + AuDebugOn(!current->nsproxy);
9384 + ns = current->nsproxy->mnt_ns;
9386 + err = iterate_mounts(au_compare_mnt, &args, ns->root);
9387 + br_read_unlock(vfsmount_lock);
9389 + AuDebugOn(!args.mnt);
9393 +struct au_nfsd_si_lock {
9394 + unsigned int sigen;
9395 + aufs_bindex_t bindex, br_id;
9396 + unsigned char force_lock;
9399 +static int si_nfsd_read_lock(struct super_block *sb,
9400 + struct au_nfsd_si_lock *nsi_lock)
9403 + aufs_bindex_t bindex;
9405 + si_read_lock(sb, AuLock_FLUSH);
9407 + /* branch id may be wrapped around */
9409 + bindex = au_br_index(sb, nsi_lock->br_id);
9410 + if (bindex >= 0 && nsi_lock->sigen + AUFS_BRANCH_MAX > au_sigen(sb))
9411 + goto out; /* success */
9415 + if (!nsi_lock->force_lock)
9416 + si_read_unlock(sb);
9419 + nsi_lock->bindex = bindex;
9423 +struct find_name_by_ino {
9424 + int called, found;
9431 +find_name_by_ino(void *arg, const char *name, int namelen, loff_t offset,
9432 + u64 ino, unsigned int d_type)
9434 + struct find_name_by_ino *a = arg;
9437 + if (a->ino != ino)
9440 + memcpy(a->name, name, namelen);
9441 + a->namelen = namelen;
9446 +static struct dentry *au_lkup_by_ino(struct path *path, ino_t ino,
9447 + struct au_nfsd_si_lock *nsi_lock)
9449 + struct dentry *dentry, *parent;
9450 + struct file *file;
9451 + struct inode *dir;
9452 + struct find_name_by_ino arg;
9455 + parent = path->dentry;
9457 + si_read_unlock(parent->d_sb);
9458 + file = vfsub_dentry_open(path, au_dir_roflags);
9459 + dentry = (void *)file;
9463 + dentry = ERR_PTR(-ENOMEM);
9464 + arg.name = __getname_gfp(GFP_NOFS);
9465 + if (unlikely(!arg.name))
9472 + err = vfsub_readdir(file, find_name_by_ino, &arg);
9473 + } while (!err && !arg.found && arg.called);
9474 + dentry = ERR_PTR(err);
9475 + if (unlikely(err))
9477 + dentry = ERR_PTR(-ENOENT);
9481 + /* do not call au_lkup_one() */
9482 + dir = parent->d_inode;
9483 + mutex_lock(&dir->i_mutex);
9484 + dentry = vfsub_lookup_one_len(arg.name, parent, arg.namelen);
9485 + mutex_unlock(&dir->i_mutex);
9486 + AuTraceErrPtr(dentry);
9487 + if (IS_ERR(dentry))
9489 + AuDebugOn(au_test_anon(dentry));
9490 + if (unlikely(!dentry->d_inode)) {
9492 + dentry = ERR_PTR(-ENOENT);
9496 + __putname(arg.name);
9500 + if (unlikely(nsi_lock
9501 + && si_nfsd_read_lock(parent->d_sb, nsi_lock) < 0))
9502 + if (!IS_ERR(dentry)) {
9504 + dentry = ERR_PTR(-ESTALE);
9506 + AuTraceErrPtr(dentry);
9510 +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino,
9512 + struct au_nfsd_si_lock *nsi_lock)
9514 + struct dentry *dentry;
9517 + if (dir_ino != AUFS_ROOT_INO) {
9518 + path.dentry = decode_by_ino(sb, dir_ino, 0);
9519 + dentry = path.dentry;
9520 + if (!path.dentry || IS_ERR(path.dentry))
9522 + AuDebugOn(au_test_anon(path.dentry));
9524 + path.dentry = dget(sb->s_root);
9526 + path.mnt = au_mnt_get(sb);
9527 + dentry = au_lkup_by_ino(&path, ino, nsi_lock);
9531 + AuTraceErrPtr(dentry);
9535 +/* ---------------------------------------------------------------------- */
9537 +static int h_acceptable(void *expv, struct dentry *dentry)
9542 +static char *au_build_path(struct dentry *h_parent, struct path *h_rootpath,
9543 + char *buf, int len, struct super_block *sb)
9549 + p = d_path(h_rootpath, buf, len);
9554 + path.mnt = h_rootpath->mnt;
9555 + path.dentry = h_parent;
9556 + p = d_path(&path, buf, len);
9562 + path.mnt = au_mnt_get(sb);
9563 + path.dentry = sb->s_root;
9564 + p = d_path(&path, buf, len - strlen(p));
9569 + p[strlen(p)] = '/';
9577 +struct dentry *decode_by_path(struct super_block *sb, ino_t ino, __u32 *fh,
9578 + int fh_len, struct au_nfsd_si_lock *nsi_lock)
9580 + struct dentry *dentry, *h_parent, *root;
9581 + struct super_block *h_sb;
9582 + char *pathname, *p;
9583 + struct vfsmount *h_mnt;
9584 + struct au_branch *br;
9588 + br = au_sbr(sb, nsi_lock->bindex);
9589 + h_mnt = br->br_mnt;
9590 + h_sb = h_mnt->mnt_sb;
9591 + /* todo: call lower fh_to_dentry()? fh_to_parent()? */
9592 + h_parent = exportfs_decode_fh(h_mnt, (void *)(fh + Fh_tail),
9593 + fh_len - Fh_tail, fh[Fh_h_type],
9594 + h_acceptable, /*context*/NULL);
9595 + dentry = h_parent;
9596 + if (unlikely(!h_parent || IS_ERR(h_parent))) {
9597 + AuWarn1("%s decode_fh failed, %ld\n",
9598 + au_sbtype(h_sb), PTR_ERR(h_parent));
9602 + if (unlikely(au_test_anon(h_parent))) {
9603 + AuWarn1("%s decode_fh returned a disconnected dentry\n",
9605 + goto out_h_parent;
9608 + dentry = ERR_PTR(-ENOMEM);
9609 + pathname = (void *)__get_free_page(GFP_NOFS);
9610 + if (unlikely(!pathname))
9611 + goto out_h_parent;
9613 + root = sb->s_root;
9615 + di_read_lock_parent(root, !AuLock_IR);
9616 + path.dentry = au_h_dptr(root, nsi_lock->bindex);
9617 + di_read_unlock(root, !AuLock_IR);
9618 + p = au_build_path(h_parent, &path, pathname, PAGE_SIZE, sb);
9619 + dentry = (void *)p;
9621 + goto out_pathname;
9623 + si_read_unlock(sb);
9624 + err = vfsub_kern_path(p, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &path);
9625 + dentry = ERR_PTR(err);
9626 + if (unlikely(err))
9629 + dentry = ERR_PTR(-ENOENT);
9630 + AuDebugOn(au_test_anon(path.dentry));
9631 + if (unlikely(!path.dentry->d_inode))
9634 + if (ino != path.dentry->d_inode->i_ino)
9635 + dentry = au_lkup_by_ino(&path, ino, /*nsi_lock*/NULL);
9637 + dentry = dget(path.dentry);
9642 + if (unlikely(si_nfsd_read_lock(sb, nsi_lock) < 0))
9643 + if (!IS_ERR(dentry)) {
9645 + dentry = ERR_PTR(-ESTALE);
9648 + free_page((unsigned long)pathname);
9652 + AuTraceErrPtr(dentry);
9656 +/* ---------------------------------------------------------------------- */
9658 +static struct dentry *
9659 +aufs_fh_to_dentry(struct super_block *sb, struct fid *fid, int fh_len,
9662 + struct dentry *dentry;
9663 + __u32 *fh = fid->raw;
9664 + struct au_branch *br;
9665 + ino_t ino, dir_ino;
9666 + struct au_nfsd_si_lock nsi_lock = {
9670 + dentry = ERR_PTR(-ESTALE);
9671 + /* it should never happen, but the file handle is unreliable */
9672 + if (unlikely(fh_len < Fh_tail))
9674 + nsi_lock.sigen = fh[Fh_sigen];
9675 + nsi_lock.br_id = fh[Fh_br_id];
9677 + /* branch id may be wrapped around */
9679 + if (unlikely(si_nfsd_read_lock(sb, &nsi_lock)))
9681 + nsi_lock.force_lock = 1;
9683 + /* is this inode still cached? */
9684 + ino = decode_ino(fh + Fh_ino);
9685 + /* it should never happen */
9686 + if (unlikely(ino == AUFS_ROOT_INO))
9689 + dir_ino = decode_ino(fh + Fh_dir_ino);
9690 + dentry = decode_by_ino(sb, ino, dir_ino);
9691 + if (IS_ERR(dentry))
9696 + /* is the parent dir cached? */
9697 + br = au_sbr(sb, nsi_lock.bindex);
9698 + atomic_inc(&br->br_count);
9699 + dentry = decode_by_dir_ino(sb, ino, dir_ino, &nsi_lock);
9700 + if (IS_ERR(dentry))
9706 + dentry = decode_by_path(sb, ino, fh, fh_len, &nsi_lock);
9707 + if (IS_ERR(dentry))
9709 + if (unlikely(!dentry))
9710 + /* todo?: make it ESTALE */
9714 + if (!au_digen_test(dentry, au_sigen(sb))
9715 + && dentry->d_inode->i_generation == fh[Fh_igen])
9716 + goto out_unlock; /* success */
9719 + dentry = ERR_PTR(-ESTALE);
9722 + atomic_dec(&br->br_count);
9723 + si_read_unlock(sb);
9725 + AuTraceErrPtr(dentry);
9729 +#if 0 /* reserved for future use */
9730 +/* support subtreecheck option */
9731 +static struct dentry *aufs_fh_to_parent(struct super_block *sb, struct fid *fid,
9732 + int fh_len, int fh_type)
9734 + struct dentry *parent;
9735 + __u32 *fh = fid->raw;
9738 + dir_ino = decode_ino(fh + Fh_dir_ino);
9739 + parent = decode_by_ino(sb, dir_ino, 0);
9740 + if (IS_ERR(parent))
9743 + parent = decode_by_path(sb, au_br_index(sb, fh[Fh_br_id]),
9744 + dir_ino, fh, fh_len);
9747 + AuTraceErrPtr(parent);
9752 +/* ---------------------------------------------------------------------- */
9754 +static int aufs_encode_fh(struct dentry *dentry, __u32 *fh, int *max_len,
9758 + aufs_bindex_t bindex, bend;
9759 + struct super_block *sb, *h_sb;
9760 + struct inode *inode;
9761 + struct dentry *parent, *h_parent;
9762 + struct au_branch *br;
9764 + AuDebugOn(au_test_anon(dentry));
9768 + if (unlikely(*max_len <= Fh_tail)) {
9769 + AuWarn1("NFSv2 client (max_len %d)?\n", *max_len);
9773 + err = FILEID_ROOT;
9774 + if (IS_ROOT(dentry)) {
9775 + AuDebugOn(dentry->d_inode->i_ino != AUFS_ROOT_INO);
9780 + err = aufs_read_lock(dentry, AuLock_FLUSH | AuLock_IR | AuLock_GEN);
9781 + if (unlikely(err))
9784 + inode = dentry->d_inode;
9785 + AuDebugOn(!inode);
9786 + sb = dentry->d_sb;
9787 +#ifdef CONFIG_AUFS_DEBUG
9788 + if (unlikely(!au_opt_test(au_mntflags(sb), XINO)))
9789 + AuWarn1("NFS-exporting requires xino\n");
9792 + parent = dget_parent(dentry);
9793 + di_read_lock_parent(parent, !AuLock_IR);
9794 + bend = au_dbtaildir(parent);
9795 + for (bindex = au_dbstart(parent); bindex <= bend; bindex++) {
9796 + h_parent = au_h_dptr(parent, bindex);
9802 + if (unlikely(!h_parent))
9806 + br = au_sbr(sb, bindex);
9807 + h_sb = br->br_mnt->mnt_sb;
9808 + if (unlikely(!h_sb->s_export_op)) {
9809 + AuErr1("%s branch is not exportable\n", au_sbtype(h_sb));
9813 + fh[Fh_br_id] = br->br_id;
9814 + fh[Fh_sigen] = au_sigen(sb);
9815 + encode_ino(fh + Fh_ino, inode->i_ino);
9816 + encode_ino(fh + Fh_dir_ino, parent->d_inode->i_ino);
9817 + fh[Fh_igen] = inode->i_generation;
9819 + *max_len -= Fh_tail;
9820 + fh[Fh_h_type] = exportfs_encode_fh(h_parent, (void *)(fh + Fh_tail),
9822 + /*connectable or subtreecheck*/0);
9823 + err = fh[Fh_h_type];
9824 + *max_len += Fh_tail;
9825 + /* todo: macros? */
9829 + AuWarn1("%s encode_fh failed\n", au_sbtype(h_sb));
9834 + di_read_unlock(parent, !AuLock_IR);
9836 + aufs_read_unlock(dentry, AuLock_IR);
9838 + if (unlikely(err < 0))
9843 +/* ---------------------------------------------------------------------- */
9845 +static int aufs_commit_metadata(struct inode *inode)
9848 + aufs_bindex_t bindex;
9849 + struct super_block *sb;
9850 + struct inode *h_inode;
9851 + int (*f)(struct inode *inode);
9854 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
9855 + ii_write_lock_child(inode);
9856 + bindex = au_ibstart(inode);
9857 + AuDebugOn(bindex < 0);
9858 + h_inode = au_h_iptr(inode, bindex);
9860 + f = h_inode->i_sb->s_export_op->commit_metadata;
9864 + struct writeback_control wbc = {
9865 + .sync_mode = WB_SYNC_ALL,
9866 + .nr_to_write = 0 /* metadata only */
9869 + err = sync_inode(h_inode, &wbc);
9872 + au_cpup_attr_timesizes(inode);
9873 + ii_write_unlock(inode);
9874 + si_read_unlock(sb);
9878 +/* ---------------------------------------------------------------------- */
9880 +static struct export_operations aufs_export_op = {
9881 + .fh_to_dentry = aufs_fh_to_dentry,
9882 + /* .fh_to_parent = aufs_fh_to_parent, */
9883 + .encode_fh = aufs_encode_fh,
9884 + .commit_metadata = aufs_commit_metadata
9887 +void au_export_init(struct super_block *sb)
9889 + struct au_sbinfo *sbinfo;
9892 + sb->s_export_op = &aufs_export_op;
9893 + sbinfo = au_sbi(sb);
9894 + sbinfo->si_xigen = NULL;
9895 + get_random_bytes(&u, sizeof(u));
9896 + BUILD_BUG_ON(sizeof(u) != sizeof(int));
9897 + atomic_set(&sbinfo->si_xigen_next, u);
9899 diff -urN /usr/share/empty/fs/aufs/file.c linux/fs/aufs/file.c
9900 --- /usr/share/empty/fs/aufs/file.c 1970-01-01 01:00:00.000000000 +0100
9901 +++ linux/fs/aufs/file.c 2011-08-24 13:30:24.731313534 +0200
9904 + * Copyright (C) 2005-2011 Junjiro R. Okajima
9906 + * This program, aufs is free software; you can redistribute it and/or modify
9907 + * it under the terms of the GNU General Public License as published by
9908 + * the Free Software Foundation; either version 2 of the License, or
9909 + * (at your option) any later version.
9911 + * This program is distributed in the hope that it will be useful,
9912 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
9913 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9914 + * GNU General Public License for more details.
9916 + * You should have received a copy of the GNU General Public License
9917 + * along with this program; if not, write to the Free Software
9918 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
9922 + * handling file/dir, and address_space operation
9925 +#include <linux/file.h>
9926 +#include <linux/fsnotify.h>
9927 +#include <linux/namei.h>
9928 +#include <linux/pagemap.h>
9931 +/* drop flags for writing */
9932 +unsigned int au_file_roflags(unsigned int flags)
9934 + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC);
9935 + flags |= O_RDONLY | O_NOATIME;
9939 +/* common functions to regular file and dir */
9940 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
9941 + struct file *file)
9943 + struct file *h_file;
9944 + struct dentry *h_dentry;
9945 + struct inode *h_inode;
9946 + struct super_block *sb;
9947 + struct au_branch *br;
9948 + struct path h_path;
9949 + int err, exec_flag;
9951 + /* a race condition can happen between open and unlink/rmdir */
9952 + h_file = ERR_PTR(-ENOENT);
9953 + h_dentry = au_h_dptr(dentry, bindex);
9954 + if (au_test_nfsd() && !h_dentry)
9956 + h_inode = h_dentry->d_inode;
9957 + if (au_test_nfsd() && !h_inode)
9959 + spin_lock(&h_dentry->d_lock);
9960 + err = (!d_unhashed(dentry) && d_unlinked(h_dentry))
9962 + /* || !dentry->d_inode->i_nlink */
9964 + spin_unlock(&h_dentry->d_lock);
9965 + if (unlikely(err))
9968 + sb = dentry->d_sb;
9969 + br = au_sbr(sb, bindex);
9970 + h_file = ERR_PTR(-EACCES);
9971 + exec_flag = flags & __FMODE_EXEC;
9972 + if (exec_flag && (br->br_mnt->mnt_flags & MNT_NOEXEC))
9975 + /* drop flags for writing */
9976 + if (au_test_ro(sb, bindex, dentry->d_inode))
9977 + flags = au_file_roflags(flags);
9978 + flags &= ~O_CREAT;
9979 + atomic_inc(&br->br_count);
9980 + h_path.dentry = h_dentry;
9981 + h_path.mnt = br->br_mnt;
9982 + if (!au_special_file(h_inode->i_mode))
9983 + h_file = vfsub_dentry_open(&h_path, flags);
9985 + /* this block depends upon the configuration */
9986 + di_read_unlock(dentry, AuLock_IR);
9987 + fi_write_unlock(file);
9988 + si_read_unlock(sb);
9989 + h_file = vfsub_dentry_open(&h_path, flags);
9990 + si_noflush_read_lock(sb);
9991 + fi_write_lock(file);
9992 + di_read_lock_child(dentry, AuLock_IR);
9994 + if (IS_ERR(h_file))
9998 + err = deny_write_access(h_file);
9999 + if (unlikely(err)) {
10001 + h_file = ERR_PTR(err);
10005 + fsnotify_open(h_file);
10006 + goto out; /* success */
10009 + atomic_dec(&br->br_count);
10014 +int au_do_open(struct file *file, int (*open)(struct file *file, int flags),
10015 + struct au_fidir *fidir)
10018 + struct dentry *dentry;
10020 + err = au_finfo_init(file, fidir);
10021 + if (unlikely(err))
10024 + dentry = file->f_dentry;
10025 + di_read_lock_child(dentry, AuLock_IR);
10026 + err = open(file, vfsub_file_flags(file));
10027 + di_read_unlock(dentry, AuLock_IR);
10029 + fi_write_unlock(file);
10030 + if (unlikely(err)) {
10031 + au_fi(file)->fi_hdir = NULL;
10032 + au_finfo_fin(file);
10039 +int au_reopen_nondir(struct file *file)
10042 + aufs_bindex_t bstart;
10043 + struct dentry *dentry;
10044 + struct file *h_file, *h_file_tmp;
10046 + dentry = file->f_dentry;
10047 + AuDebugOn(au_special_file(dentry->d_inode->i_mode));
10048 + bstart = au_dbstart(dentry);
10049 + h_file_tmp = NULL;
10050 + if (au_fbstart(file) == bstart) {
10051 + h_file = au_hf_top(file);
10052 + if (file->f_mode == h_file->f_mode)
10053 + return 0; /* success */
10054 + h_file_tmp = h_file;
10055 + get_file(h_file_tmp);
10056 + au_set_h_fptr(file, bstart, NULL);
10058 + AuDebugOn(au_fi(file)->fi_hdir);
10059 + AuDebugOn(au_fbstart(file) < bstart);
10061 + h_file = au_h_open(dentry, bstart, vfsub_file_flags(file) & ~O_TRUNC,
10063 + err = PTR_ERR(h_file);
10064 + if (IS_ERR(h_file))
10065 + goto out; /* todo: close all? */
10068 + au_set_fbstart(file, bstart);
10069 + au_set_h_fptr(file, bstart, h_file);
10070 + au_update_figen(file);
10071 + /* todo: necessary? */
10072 + /* file->f_ra = h_file->f_ra; */
10076 + fput(h_file_tmp);
10080 +/* ---------------------------------------------------------------------- */
10082 +static int au_reopen_wh(struct file *file, aufs_bindex_t btgt,
10083 + struct dentry *hi_wh)
10086 + aufs_bindex_t bstart;
10087 + struct au_dinfo *dinfo;
10088 + struct dentry *h_dentry;
10089 + struct au_hdentry *hdp;
10091 + dinfo = au_di(file->f_dentry);
10092 + AuRwMustWriteLock(&dinfo->di_rwsem);
10094 + bstart = dinfo->di_bstart;
10095 + dinfo->di_bstart = btgt;
10096 + hdp = dinfo->di_hdentry;
10097 + h_dentry = hdp[0 + btgt].hd_dentry;
10098 + hdp[0 + btgt].hd_dentry = hi_wh;
10099 + err = au_reopen_nondir(file);
10100 + hdp[0 + btgt].hd_dentry = h_dentry;
10101 + dinfo->di_bstart = bstart;
10106 +static int au_ready_to_write_wh(struct file *file, loff_t len,
10107 + aufs_bindex_t bcpup)
10110 + struct inode *inode, *h_inode;
10111 + struct dentry *dentry, *h_dentry, *hi_wh;
10113 + dentry = file->f_dentry;
10114 + au_update_dbstart(dentry);
10115 + inode = dentry->d_inode;
10117 + if (au_dbstart(dentry) <= bcpup && au_dbend(dentry) >= bcpup) {
10118 + h_dentry = au_h_dptr(dentry, bcpup);
10120 + h_inode = h_dentry->d_inode;
10122 + hi_wh = au_hi_wh(inode, bcpup);
10123 + if (!hi_wh && !h_inode)
10124 + err = au_sio_cpup_wh(dentry, bcpup, len, file);
10126 + /* already copied-up after unlink */
10127 + err = au_reopen_wh(file, bcpup, hi_wh);
10130 + && inode->i_nlink > 1
10131 + && au_opt_test(au_mntflags(dentry->d_sb), PLINK))
10132 + au_plink_append(inode, bcpup, au_h_dptr(dentry, bcpup));
10138 + * prepare the @file for writing.
10140 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin)
10143 + aufs_bindex_t bstart, bcpup, dbstart;
10144 + struct dentry *dentry, *parent, *h_dentry;
10145 + struct inode *h_inode, *inode;
10146 + struct super_block *sb;
10147 + struct file *h_file;
10149 + dentry = file->f_dentry;
10150 + sb = dentry->d_sb;
10151 + inode = dentry->d_inode;
10152 + AuDebugOn(au_special_file(inode->i_mode));
10153 + bstart = au_fbstart(file);
10154 + err = au_test_ro(sb, bstart, inode);
10155 + if (!err && (au_hf_top(file)->f_mode & FMODE_WRITE)) {
10156 + err = au_pin(pin, dentry, bstart, AuOpt_UDBA_NONE, /*flags*/0);
10160 + /* need to cpup or reopen */
10161 + parent = dget_parent(dentry);
10162 + di_write_lock_parent(parent);
10163 + err = AuWbrCopyup(au_sbi(sb), dentry);
10165 + if (unlikely(err < 0))
10169 + if (!d_unhashed(dentry) && !au_h_dptr(parent, bcpup)) {
10170 + err = au_cpup_dirs(dentry, bcpup);
10171 + if (unlikely(err))
10175 + err = au_pin(pin, dentry, bcpup, AuOpt_UDBA_NONE,
10176 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
10177 + if (unlikely(err))
10180 + h_dentry = au_hf_top(file)->f_dentry;
10181 + h_inode = h_dentry->d_inode;
10182 + dbstart = au_dbstart(dentry);
10183 + if (dbstart <= bcpup) {
10184 + h_dentry = au_h_dptr(dentry, bcpup);
10185 + AuDebugOn(!h_dentry);
10186 + h_inode = h_dentry->d_inode;
10187 + AuDebugOn(!h_inode);
10191 + if (dbstart <= bcpup /* just reopen */
10192 + || !d_unhashed(dentry) /* copyup and reopen */
10194 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
10195 + h_file = au_h_open_pre(dentry, bstart);
10196 + if (IS_ERR(h_file)) {
10197 + err = PTR_ERR(h_file);
10200 + di_downgrade_lock(parent, AuLock_IR);
10201 + if (dbstart > bcpup)
10202 + err = au_sio_cpup_simple(dentry, bcpup, len,
10205 + err = au_reopen_nondir(file);
10207 + mutex_unlock(&h_inode->i_mutex);
10208 + au_h_open_post(dentry, bstart, h_file);
10209 + } else { /* copyup as wh and reopen */
10211 + * since writable hfsplus branch is not supported,
10212 + * h_open_pre/post() are unnecessary.
10214 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
10215 + err = au_ready_to_write_wh(file, len, bcpup);
10216 + di_downgrade_lock(parent, AuLock_IR);
10217 + mutex_unlock(&h_inode->i_mutex);
10221 + au_pin_set_parent_lflag(pin, /*lflag*/0);
10222 + goto out_dput; /* success */
10228 + di_downgrade_lock(parent, AuLock_IR);
10230 + di_read_unlock(parent, AuLock_IR);
10237 +/* ---------------------------------------------------------------------- */
10239 +int au_do_flush(struct file *file, fl_owner_t id,
10240 + int (*flush)(struct file *file, fl_owner_t id))
10243 + struct dentry *dentry;
10244 + struct super_block *sb;
10245 + struct inode *inode;
10247 + dentry = file->f_dentry;
10248 + sb = dentry->d_sb;
10249 + inode = dentry->d_inode;
10250 + si_noflush_read_lock(sb);
10251 + fi_read_lock(file);
10252 + ii_read_lock_child(inode);
10254 + err = flush(file, id);
10255 + au_cpup_attr_timesizes(inode);
10257 + ii_read_unlock(inode);
10258 + fi_read_unlock(file);
10259 + si_read_unlock(sb);
10263 +/* ---------------------------------------------------------------------- */
10265 +static int au_file_refresh_by_inode(struct file *file, int *need_reopen)
10268 + aufs_bindex_t bstart;
10269 + struct au_pin pin;
10270 + struct au_finfo *finfo;
10271 + struct dentry *dentry, *parent, *hi_wh;
10272 + struct inode *inode;
10273 + struct super_block *sb;
10275 + FiMustWriteLock(file);
10278 + finfo = au_fi(file);
10279 + dentry = file->f_dentry;
10280 + sb = dentry->d_sb;
10281 + inode = dentry->d_inode;
10282 + bstart = au_ibstart(inode);
10283 + if (bstart == finfo->fi_btop || IS_ROOT(dentry))
10286 + parent = dget_parent(dentry);
10287 + if (au_test_ro(sb, bstart, inode)) {
10288 + di_read_lock_parent(parent, !AuLock_IR);
10289 + err = AuWbrCopyup(au_sbi(sb), dentry);
10291 + di_read_unlock(parent, !AuLock_IR);
10292 + if (unlikely(err < 0))
10297 + di_read_lock_parent(parent, AuLock_IR);
10298 + hi_wh = au_hi_wh(inode, bstart);
10299 + if (!S_ISDIR(inode->i_mode)
10300 + && au_opt_test(au_mntflags(sb), PLINK)
10301 + && au_plink_test(inode)
10302 + && !d_unhashed(dentry)) {
10303 + err = au_test_and_cpup_dirs(dentry, bstart);
10304 + if (unlikely(err))
10307 + /* always superio. */
10308 + err = au_pin(&pin, dentry, bstart, AuOpt_UDBA_NONE,
10309 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
10311 + err = au_sio_cpup_simple(dentry, bstart, -1,
10314 + } else if (hi_wh) {
10315 + /* already copied-up after unlink */
10316 + err = au_reopen_wh(file, bstart, hi_wh);
10317 + *need_reopen = 0;
10321 + di_read_unlock(parent, AuLock_IR);
10328 +static void au_do_refresh_dir(struct file *file)
10330 + aufs_bindex_t bindex, bend, new_bindex, brid;
10331 + struct au_hfile *p, tmp, *q;
10332 + struct au_finfo *finfo;
10333 + struct super_block *sb;
10334 + struct au_fidir *fidir;
10336 + FiMustWriteLock(file);
10338 + sb = file->f_dentry->d_sb;
10339 + finfo = au_fi(file);
10340 + fidir = finfo->fi_hdir;
10341 + AuDebugOn(!fidir);
10342 + p = fidir->fd_hfile + finfo->fi_btop;
10343 + brid = p->hf_br->br_id;
10344 + bend = fidir->fd_bbot;
10345 + for (bindex = finfo->fi_btop; bindex <= bend; bindex++, p++) {
10349 + new_bindex = au_br_index(sb, p->hf_br->br_id);
10350 + if (new_bindex == bindex)
10352 + if (new_bindex < 0) {
10353 + au_set_h_fptr(file, bindex, NULL);
10357 + /* swap two lower inode, and loop again */
10358 + q = fidir->fd_hfile + new_bindex;
10362 + if (tmp.hf_file) {
10368 + p = fidir->fd_hfile;
10369 + if (!au_test_mmapped(file) && !d_unlinked(file->f_dentry)) {
10370 + bend = au_sbend(sb);
10371 + for (finfo->fi_btop = 0; finfo->fi_btop <= bend;
10372 + finfo->fi_btop++, p++)
10373 + if (p->hf_file) {
10374 + if (p->hf_file->f_dentry
10375 + && p->hf_file->f_dentry->d_inode)
10378 + au_hfput(p, file);
10381 + bend = au_br_index(sb, brid);
10382 + for (finfo->fi_btop = 0; finfo->fi_btop < bend;
10383 + finfo->fi_btop++, p++)
10385 + au_hfput(p, file);
10386 + bend = au_sbend(sb);
10389 + p = fidir->fd_hfile + bend;
10390 + for (fidir->fd_bbot = bend; fidir->fd_bbot >= finfo->fi_btop;
10391 + fidir->fd_bbot--, p--)
10392 + if (p->hf_file) {
10393 + if (p->hf_file->f_dentry
10394 + && p->hf_file->f_dentry->d_inode)
10397 + au_hfput(p, file);
10399 + AuDebugOn(fidir->fd_bbot < finfo->fi_btop);
10403 + * after branch manipulating, refresh the file.
10405 +static int refresh_file(struct file *file, int (*reopen)(struct file *file))
10407 + int err, need_reopen;
10408 + aufs_bindex_t bend, bindex;
10409 + struct dentry *dentry;
10410 + struct au_finfo *finfo;
10411 + struct au_hfile *hfile;
10413 + dentry = file->f_dentry;
10414 + finfo = au_fi(file);
10415 + if (!finfo->fi_hdir) {
10416 + hfile = &finfo->fi_htop;
10417 + AuDebugOn(!hfile->hf_file);
10418 + bindex = au_br_index(dentry->d_sb, hfile->hf_br->br_id);
10419 + AuDebugOn(bindex < 0);
10420 + if (bindex != finfo->fi_btop)
10421 + au_set_fbstart(file, bindex);
10423 + err = au_fidir_realloc(finfo, au_sbend(dentry->d_sb) + 1);
10424 + if (unlikely(err))
10426 + au_do_refresh_dir(file);
10431 + if (!au_test_mmapped(file))
10432 + err = au_file_refresh_by_inode(file, &need_reopen);
10433 + if (!err && need_reopen && !d_unlinked(dentry))
10434 + err = reopen(file);
10436 + au_update_figen(file);
10437 + goto out; /* success */
10440 + /* error, close all lower files */
10441 + if (finfo->fi_hdir) {
10442 + bend = au_fbend_dir(file);
10443 + for (bindex = au_fbstart(file); bindex <= bend; bindex++)
10444 + au_set_h_fptr(file, bindex, NULL);
10451 +/* common function to regular file and dir */
10452 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
10456 + unsigned int sigen, figen;
10457 + aufs_bindex_t bstart;
10458 + unsigned char pseudo_link;
10459 + struct dentry *dentry;
10460 + struct inode *inode;
10463 + dentry = file->f_dentry;
10464 + inode = dentry->d_inode;
10465 + AuDebugOn(au_special_file(inode->i_mode));
10466 + sigen = au_sigen(dentry->d_sb);
10467 + fi_write_lock(file);
10468 + figen = au_figen(file);
10469 + di_write_lock_child(dentry);
10470 + bstart = au_dbstart(dentry);
10471 + pseudo_link = (bstart != au_ibstart(inode));
10472 + if (sigen == figen && !pseudo_link && au_fbstart(file) == bstart) {
10474 + di_downgrade_lock(dentry, AuLock_IR);
10475 + fi_downgrade_lock(file);
10477 + goto out; /* success */
10480 + AuDbg("sigen %d, figen %d\n", sigen, figen);
10481 + if (au_digen_test(dentry, sigen)) {
10482 + err = au_reval_dpath(dentry, sigen);
10483 + AuDebugOn(!err && au_digen_test(dentry, sigen));
10487 + err = refresh_file(file, reopen);
10490 + di_downgrade_lock(dentry, AuLock_IR);
10491 + fi_downgrade_lock(file);
10494 + di_write_unlock(dentry);
10495 + fi_write_unlock(file);
10502 +/* ---------------------------------------------------------------------- */
10504 +/* cf. aufs_nopage() */
10505 +/* for madvise(2) */
10506 +static int aufs_readpage(struct file *file __maybe_unused, struct page *page)
10508 + unlock_page(page);
10512 +/* it will never be called, but necessary to support O_DIRECT */
10513 +static ssize_t aufs_direct_IO(int rw, struct kiocb *iocb,
10514 + const struct iovec *iov, loff_t offset,
10515 + unsigned long nr_segs)
10516 +{ BUG(); return 0; }
10519 + * it will never be called, but madvise and fadvise behaves differently
10520 + * when get_xip_mem is defined
10522 +static int aufs_get_xip_mem(struct address_space *mapping, pgoff_t pgoff,
10523 + int create, void **kmem, unsigned long *pfn)
10524 +{ BUG(); return 0; }
10526 +/* they will never be called. */
10527 +#ifdef CONFIG_AUFS_DEBUG
10528 +static int aufs_write_begin(struct file *file, struct address_space *mapping,
10529 + loff_t pos, unsigned len, unsigned flags,
10530 + struct page **pagep, void **fsdata)
10531 +{ AuUnsupport(); return 0; }
10532 +static int aufs_write_end(struct file *file, struct address_space *mapping,
10533 + loff_t pos, unsigned len, unsigned copied,
10534 + struct page *page, void *fsdata)
10535 +{ AuUnsupport(); return 0; }
10536 +static int aufs_writepage(struct page *page, struct writeback_control *wbc)
10537 +{ AuUnsupport(); return 0; }
10539 +static int aufs_set_page_dirty(struct page *page)
10540 +{ AuUnsupport(); return 0; }
10541 +static void aufs_invalidatepage(struct page *page, unsigned long offset)
10542 +{ AuUnsupport(); }
10543 +static int aufs_releasepage(struct page *page, gfp_t gfp)
10544 +{ AuUnsupport(); return 0; }
10545 +static int aufs_migratepage(struct address_space *mapping, struct page *newpage,
10546 + struct page *page)
10547 +{ AuUnsupport(); return 0; }
10548 +static int aufs_launder_page(struct page *page)
10549 +{ AuUnsupport(); return 0; }
10550 +static int aufs_is_partially_uptodate(struct page *page,
10551 + read_descriptor_t *desc,
10552 + unsigned long from)
10553 +{ AuUnsupport(); return 0; }
10554 +static int aufs_error_remove_page(struct address_space *mapping,
10555 + struct page *page)
10556 +{ AuUnsupport(); return 0; }
10557 +#endif /* CONFIG_AUFS_DEBUG */
10559 +const struct address_space_operations aufs_aop = {
10560 + .readpage = aufs_readpage,
10561 + .direct_IO = aufs_direct_IO,
10562 + .get_xip_mem = aufs_get_xip_mem,
10563 +#ifdef CONFIG_AUFS_DEBUG
10564 + .writepage = aufs_writepage,
10565 + /* no writepages, because of writepage */
10566 + .set_page_dirty = aufs_set_page_dirty,
10567 + /* no readpages, because of readpage */
10568 + .write_begin = aufs_write_begin,
10569 + .write_end = aufs_write_end,
10570 + /* no bmap, no block device */
10571 + .invalidatepage = aufs_invalidatepage,
10572 + .releasepage = aufs_releasepage,
10573 + .migratepage = aufs_migratepage,
10574 + .launder_page = aufs_launder_page,
10575 + .is_partially_uptodate = aufs_is_partially_uptodate,
10576 + .error_remove_page = aufs_error_remove_page
10577 +#endif /* CONFIG_AUFS_DEBUG */
10579 diff -urN /usr/share/empty/fs/aufs/file.h linux/fs/aufs/file.h
10580 --- /usr/share/empty/fs/aufs/file.h 1970-01-01 01:00:00.000000000 +0100
10581 +++ linux/fs/aufs/file.h 2011-08-24 13:30:24.731313534 +0200
10584 + * Copyright (C) 2005-2011 Junjiro R. Okajima
10586 + * This program, aufs is free software; you can redistribute it and/or modify
10587 + * it under the terms of the GNU General Public License as published by
10588 + * the Free Software Foundation; either version 2 of the License, or
10589 + * (at your option) any later version.
10591 + * This program is distributed in the hope that it will be useful,
10592 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10593 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10594 + * GNU General Public License for more details.
10596 + * You should have received a copy of the GNU General Public License
10597 + * along with this program; if not, write to the Free Software
10598 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
10602 + * file operations
10605 +#ifndef __AUFS_FILE_H__
10606 +#define __AUFS_FILE_H__
10610 +#include <linux/file.h>
10611 +#include <linux/fs.h>
10612 +#include <linux/poll.h>
10613 +#include <linux/aufs_type.h>
10614 +#include "rwsem.h"
10618 + struct file *hf_file;
10619 + struct au_branch *hf_br;
10624 + aufs_bindex_t fd_bbot;
10625 + aufs_bindex_t fd_nent;
10626 + struct au_vdir *fd_vdir_cache;
10627 + struct au_hfile fd_hfile[];
10630 +static inline int au_fidir_sz(int nent)
10632 + AuDebugOn(nent < 0);
10633 + return sizeof(struct au_fidir) + sizeof(struct au_hfile) * nent;
10637 + atomic_t fi_generation;
10639 + struct au_rwsem fi_rwsem;
10640 + aufs_bindex_t fi_btop;
10642 + /* do not union them */
10643 + struct { /* for non-dir */
10644 + struct au_hfile fi_htop;
10645 + atomic_t fi_mmapped;
10647 + struct au_fidir *fi_hdir; /* for dir only */
10648 +} ____cacheline_aligned_in_smp;
10650 +/* ---------------------------------------------------------------------- */
10653 +extern const struct address_space_operations aufs_aop;
10654 +unsigned int au_file_roflags(unsigned int flags);
10655 +struct file *au_h_open(struct dentry *dentry, aufs_bindex_t bindex, int flags,
10656 + struct file *file);
10657 +int au_do_open(struct file *file, int (*open)(struct file *file, int flags),
10658 + struct au_fidir *fidir);
10659 +int au_reopen_nondir(struct file *file);
10661 +int au_ready_to_write(struct file *file, loff_t len, struct au_pin *pin);
10662 +int au_reval_and_lock_fdi(struct file *file, int (*reopen)(struct file *file),
10664 +int au_do_flush(struct file *file, fl_owner_t id,
10665 + int (*flush)(struct file *file, fl_owner_t id));
10668 +#ifdef CONFIG_AUFS_POLL
10669 +unsigned int aufs_poll(struct file *file, poll_table *wait);
10672 +#ifdef CONFIG_AUFS_BR_HFSPLUS
10674 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex);
10675 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
10676 + struct file *h_file);
10679 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex)
10684 +AuStubVoid(au_h_open_post, struct dentry *dentry, aufs_bindex_t bindex,
10685 + struct file *h_file);
10689 +extern const struct file_operations aufs_file_fop;
10690 +int au_do_open_nondir(struct file *file, int flags);
10691 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file);
10693 +#ifdef CONFIG_AUFS_SP_IATTR
10695 +int au_special_file(umode_t mode);
10696 +void au_init_special_fop(struct inode *inode, umode_t mode, dev_t rdev);
10698 +AuStubInt0(au_special_file, umode_t mode)
10699 +static inline void au_init_special_fop(struct inode *inode, umode_t mode,
10702 + init_special_inode(inode, mode, rdev);
10707 +void au_hfput(struct au_hfile *hf, struct file *file);
10708 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex,
10709 + struct file *h_file);
10711 +void au_update_figen(struct file *file);
10712 +struct au_fidir *au_fidir_alloc(struct super_block *sb);
10713 +int au_fidir_realloc(struct au_finfo *finfo, int nbr);
10715 +void au_fi_init_once(void *_fi);
10716 +void au_finfo_fin(struct file *file);
10717 +int au_finfo_init(struct file *file, struct au_fidir *fidir);
10720 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg);
10721 +#ifdef CONFIG_COMPAT
10722 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
10723 + unsigned long arg);
10726 +/* ---------------------------------------------------------------------- */
10728 +static inline struct au_finfo *au_fi(struct file *file)
10730 + return file->private_data;
10733 +/* ---------------------------------------------------------------------- */
10736 + * fi_read_lock, fi_write_lock,
10737 + * fi_read_unlock, fi_write_unlock, fi_downgrade_lock
10739 +AuSimpleRwsemFuncs(fi, struct file *f, &au_fi(f)->fi_rwsem);
10741 +#define FiMustNoWaiters(f) AuRwMustNoWaiters(&au_fi(f)->fi_rwsem)
10742 +#define FiMustAnyLock(f) AuRwMustAnyLock(&au_fi(f)->fi_rwsem)
10743 +#define FiMustWriteLock(f) AuRwMustWriteLock(&au_fi(f)->fi_rwsem)
10745 +/* ---------------------------------------------------------------------- */
10747 +/* todo: hard/soft set? */
10748 +static inline aufs_bindex_t au_fbstart(struct file *file)
10750 + FiMustAnyLock(file);
10751 + return au_fi(file)->fi_btop;
10754 +static inline aufs_bindex_t au_fbend_dir(struct file *file)
10756 + FiMustAnyLock(file);
10757 + AuDebugOn(!au_fi(file)->fi_hdir);
10758 + return au_fi(file)->fi_hdir->fd_bbot;
10761 +static inline struct au_vdir *au_fvdir_cache(struct file *file)
10763 + FiMustAnyLock(file);
10764 + AuDebugOn(!au_fi(file)->fi_hdir);
10765 + return au_fi(file)->fi_hdir->fd_vdir_cache;
10768 +static inline void au_set_fbstart(struct file *file, aufs_bindex_t bindex)
10770 + FiMustWriteLock(file);
10771 + au_fi(file)->fi_btop = bindex;
10774 +static inline void au_set_fbend_dir(struct file *file, aufs_bindex_t bindex)
10776 + FiMustWriteLock(file);
10777 + AuDebugOn(!au_fi(file)->fi_hdir);
10778 + au_fi(file)->fi_hdir->fd_bbot = bindex;
10781 +static inline void au_set_fvdir_cache(struct file *file,
10782 + struct au_vdir *vdir_cache)
10784 + FiMustWriteLock(file);
10785 + AuDebugOn(!au_fi(file)->fi_hdir);
10786 + au_fi(file)->fi_hdir->fd_vdir_cache = vdir_cache;
10789 +static inline struct file *au_hf_top(struct file *file)
10791 + FiMustAnyLock(file);
10792 + AuDebugOn(au_fi(file)->fi_hdir);
10793 + return au_fi(file)->fi_htop.hf_file;
10796 +static inline struct file *au_hf_dir(struct file *file, aufs_bindex_t bindex)
10798 + FiMustAnyLock(file);
10799 + AuDebugOn(!au_fi(file)->fi_hdir);
10800 + return au_fi(file)->fi_hdir->fd_hfile[0 + bindex].hf_file;
10803 +/* todo: memory barrier? */
10804 +static inline unsigned int au_figen(struct file *f)
10806 + return atomic_read(&au_fi(f)->fi_generation);
10809 +static inline void au_set_mmapped(struct file *f)
10811 + if (atomic_inc_return(&au_fi(f)->fi_mmapped))
10813 + pr_warning("fi_mmapped wrapped around\n");
10814 + while (!atomic_inc_return(&au_fi(f)->fi_mmapped))
10818 +static inline void au_unset_mmapped(struct file *f)
10820 + atomic_dec(&au_fi(f)->fi_mmapped);
10823 +static inline int au_test_mmapped(struct file *f)
10825 + return atomic_read(&au_fi(f)->fi_mmapped);
10828 +/* customize vma->vm_file */
10830 +static inline void au_do_vm_file_reset(struct vm_area_struct *vma,
10831 + struct file *file)
10835 + f = vma->vm_file;
10837 + vma->vm_file = file;
10842 +#define AuDbgVmRegion(file, vma) do {} while (0)
10844 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
10845 + struct file *file)
10847 + au_do_vm_file_reset(vma, file);
10850 +#define AuDbgVmRegion(file, vma) \
10851 + AuDebugOn((vma)->vm_region && (vma)->vm_region->vm_file != (file))
10853 +static inline void au_vm_file_reset(struct vm_area_struct *vma,
10854 + struct file *file)
10858 + au_do_vm_file_reset(vma, file);
10859 + f = vma->vm_region->vm_file;
10861 + vma->vm_region->vm_file = file;
10864 +#endif /* CONFIG_MMU */
10866 +/* handle vma->vm_prfile */
10867 +static inline void au_vm_prfile_set(struct vm_area_struct *vma,
10868 + struct file *file)
10870 +#ifdef CONFIG_AUFS_PROC_MAP
10872 + vma->vm_prfile = file;
10873 +#ifndef CONFIG_MMU
10875 + vma->vm_region->vm_prfile = file;
10880 +#endif /* __KERNEL__ */
10881 +#endif /* __AUFS_FILE_H__ */
10882 diff -urN /usr/share/empty/fs/aufs/finfo.c linux/fs/aufs/finfo.c
10883 --- /usr/share/empty/fs/aufs/finfo.c 1970-01-01 01:00:00.000000000 +0100
10884 +++ linux/fs/aufs/finfo.c 2012-01-05 11:00:59.215501779 +0100
10887 + * Copyright (C) 2005-2011 Junjiro R. Okajima
10889 + * This program, aufs is free software; you can redistribute it and/or modify
10890 + * it under the terms of the GNU General Public License as published by
10891 + * the Free Software Foundation; either version 2 of the License, or
10892 + * (at your option) any later version.
10894 + * This program is distributed in the hope that it will be useful,
10895 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
10896 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10897 + * GNU General Public License for more details.
10899 + * You should have received a copy of the GNU General Public License
10900 + * along with this program; if not, write to the Free Software
10901 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
10905 + * file private data
10908 +#include <linux/file.h>
10911 +void au_hfput(struct au_hfile *hf, struct file *file)
10913 + /* todo: direct access f_flags */
10914 + if (vfsub_file_flags(file) & __FMODE_EXEC)
10915 + allow_write_access(hf->hf_file);
10916 + fput(hf->hf_file);
10917 + hf->hf_file = NULL;
10918 + atomic_dec(&hf->hf_br->br_count);
10919 + hf->hf_br = NULL;
10922 +void au_set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val)
10924 + struct au_finfo *finfo = au_fi(file);
10925 + struct au_hfile *hf;
10926 + struct au_fidir *fidir;
10928 + fidir = finfo->fi_hdir;
10930 + AuDebugOn(finfo->fi_btop != bindex);
10931 + hf = &finfo->fi_htop;
10933 + hf = fidir->fd_hfile + bindex;
10935 + if (hf && hf->hf_file)
10936 + au_hfput(hf, file);
10938 + FiMustWriteLock(file);
10939 + hf->hf_file = val;
10940 + hf->hf_br = au_sbr(file->f_dentry->d_sb, bindex);
10944 +void au_update_figen(struct file *file)
10946 + atomic_set(&au_fi(file)->fi_generation, au_digen(file->f_dentry));
10947 + /* smp_mb(); */ /* atomic_set */
10950 +/* ---------------------------------------------------------------------- */
10952 +struct au_fidir *au_fidir_alloc(struct super_block *sb)
10954 + struct au_fidir *fidir;
10957 + nbr = au_sbend(sb) + 1;
10959 + nbr = 2; /* initial allocate for 2 branches */
10960 + fidir = kzalloc(au_fidir_sz(nbr), GFP_NOFS);
10962 + fidir->fd_bbot = -1;
10963 + fidir->fd_nent = nbr;
10964 + fidir->fd_vdir_cache = NULL;
10970 +int au_fidir_realloc(struct au_finfo *finfo, int nbr)
10973 + struct au_fidir *fidir, *p;
10975 + AuRwMustWriteLock(&finfo->fi_rwsem);
10976 + fidir = finfo->fi_hdir;
10977 + AuDebugOn(!fidir);
10980 + p = au_kzrealloc(fidir, au_fidir_sz(fidir->fd_nent), au_fidir_sz(nbr),
10983 + p->fd_nent = nbr;
10984 + finfo->fi_hdir = p;
10991 +/* ---------------------------------------------------------------------- */
10993 +void au_finfo_fin(struct file *file)
10995 + struct au_finfo *finfo;
10997 + au_nfiles_dec(file->f_dentry->d_sb);
10999 + finfo = au_fi(file);
11000 + AuDebugOn(finfo->fi_hdir);
11001 + AuRwDestroy(&finfo->fi_rwsem);
11002 + au_cache_free_finfo(finfo);
11005 +void au_fi_init_once(void *_finfo)
11007 + struct au_finfo *finfo = _finfo;
11008 + static struct lock_class_key aufs_fi;
11010 + au_rw_init(&finfo->fi_rwsem);
11011 + au_rw_class(&finfo->fi_rwsem, &aufs_fi);
11014 +int au_finfo_init(struct file *file, struct au_fidir *fidir)
11017 + struct au_finfo *finfo;
11018 + struct dentry *dentry;
11021 + dentry = file->f_dentry;
11022 + finfo = au_cache_alloc_finfo();
11023 + if (unlikely(!finfo))
11027 + au_nfiles_inc(dentry->d_sb);
11028 + lc_idx = AuLcNonDir_FIINFO;
11030 + lc_idx = AuLcDir_FIINFO;
11031 + au_rw_class(&finfo->fi_rwsem, au_lc_key + lc_idx);
11032 + au_rw_write_lock(&finfo->fi_rwsem);
11033 + finfo->fi_btop = -1;
11034 + finfo->fi_hdir = fidir;
11035 + atomic_set(&finfo->fi_generation, au_digen(dentry));
11036 + /* smp_mb(); */ /* atomic_set */
11038 + file->private_data = finfo;
11043 diff -urN /usr/share/empty/fs/aufs/f_op.c linux/fs/aufs/f_op.c
11044 --- /usr/share/empty/fs/aufs/f_op.c 1970-01-01 01:00:00.000000000 +0100
11045 +++ linux/fs/aufs/f_op.c 2012-01-05 11:00:59.215501779 +0100
11048 + * Copyright (C) 2005-2011 Junjiro R. Okajima
11050 + * This program, aufs is free software; you can redistribute it and/or modify
11051 + * it under the terms of the GNU General Public License as published by
11052 + * the Free Software Foundation; either version 2 of the License, or
11053 + * (at your option) any later version.
11055 + * This program is distributed in the hope that it will be useful,
11056 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
11057 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11058 + * GNU General Public License for more details.
11060 + * You should have received a copy of the GNU General Public License
11061 + * along with this program; if not, write to the Free Software
11062 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
11066 + * file and vm operations
11069 +#include <linux/file.h>
11070 +#include <linux/fs_stack.h>
11071 +#include <linux/mman.h>
11072 +#include <linux/mm.h>
11073 +#include <linux/security.h>
11076 +int au_do_open_nondir(struct file *file, int flags)
11079 + aufs_bindex_t bindex;
11080 + struct file *h_file;
11081 + struct dentry *dentry;
11082 + struct au_finfo *finfo;
11084 + FiMustWriteLock(file);
11086 + dentry = file->f_dentry;
11087 + err = au_d_alive(dentry);
11088 + if (unlikely(err))
11091 + finfo = au_fi(file);
11092 + memset(&finfo->fi_htop, 0, sizeof(finfo->fi_htop));
11093 + atomic_set(&finfo->fi_mmapped, 0);
11094 + bindex = au_dbstart(dentry);
11095 + h_file = au_h_open(dentry, bindex, flags, file);
11096 + if (IS_ERR(h_file))
11097 + err = PTR_ERR(h_file);
11099 + au_set_fbstart(file, bindex);
11100 + au_set_h_fptr(file, bindex, h_file);
11101 + au_update_figen(file);
11102 + /* todo: necessary? */
11103 + /* file->f_ra = h_file->f_ra; */
11110 +static int aufs_open_nondir(struct inode *inode __maybe_unused,
11111 + struct file *file)
11114 + struct super_block *sb;
11116 + AuDbg("%.*s, f_flags 0x%x, f_mode 0x%x\n",
11117 + AuDLNPair(file->f_dentry), vfsub_file_flags(file),
11120 + sb = file->f_dentry->d_sb;
11121 + si_read_lock(sb, AuLock_FLUSH);
11122 + err = au_do_open(file, au_do_open_nondir, /*fidir*/NULL);
11123 + si_read_unlock(sb);
11127 +int aufs_release_nondir(struct inode *inode __maybe_unused, struct file *file)
11129 + struct au_finfo *finfo;
11130 + aufs_bindex_t bindex;
11132 + finfo = au_fi(file);
11133 + bindex = finfo->fi_btop;
11134 + if (bindex >= 0) {
11135 + /* remove me from sb->s_files */
11136 + file_sb_list_del(file);
11137 + au_set_h_fptr(file, bindex, NULL);
11140 + au_finfo_fin(file);
11144 +/* ---------------------------------------------------------------------- */
11146 +static int au_do_flush_nondir(struct file *file, fl_owner_t id)
11149 + struct file *h_file;
11152 + h_file = au_hf_top(file);
11154 + err = vfsub_flush(h_file, id);
11158 +static int aufs_flush_nondir(struct file *file, fl_owner_t id)
11160 + return au_do_flush(file, id, au_do_flush_nondir);
11163 +/* ---------------------------------------------------------------------- */
11165 + * read and write functions acquire [fdi]_rwsem once, but release before
11166 + * mmap_sem. This is because to stop a race condition between mmap(2).
11167 + * Releasing these aufs-rwsem should be safe, no branch-mamagement (by keeping
11168 + * si_rwsem), no harmful copy-up should happen. Actually copy-up may happen in
11169 + * read functions after [fdi]_rwsem are released, but it should be harmless.
11172 +static ssize_t aufs_read(struct file *file, char __user *buf, size_t count,
11176 + struct dentry *dentry;
11177 + struct file *h_file;
11178 + struct super_block *sb;
11180 + dentry = file->f_dentry;
11181 + sb = dentry->d_sb;
11182 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
11183 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
11184 + if (unlikely(err))
11187 + h_file = au_hf_top(file);
11188 + get_file(h_file);
11189 + di_read_unlock(dentry, AuLock_IR);
11190 + fi_read_unlock(file);
11192 + /* filedata may be obsoleted by concurrent copyup, but no problem */
11193 + err = vfsub_read_u(h_file, buf, count, ppos);
11194 + /* todo: necessary? */
11195 + /* file->f_ra = h_file->f_ra; */
11196 + /* update without lock, I don't think it a problem */
11197 + fsstack_copy_attr_atime(dentry->d_inode, h_file->f_dentry->d_inode);
11201 + si_read_unlock(sb);
11206 + * todo: very ugly
11207 + * it locks both of i_mutex and si_rwsem for read in safe.
11208 + * if the plink maintenance mode continues forever (that is the problem),
11209 + * may loop forever.
11211 +static void au_mtx_and_read_lock(struct inode *inode)
11214 + struct super_block *sb = inode->i_sb;
11217 + mutex_lock(&inode->i_mutex);
11218 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
11221 + mutex_unlock(&inode->i_mutex);
11222 + si_read_lock(sb, AuLock_NOPLMW);
11223 + si_read_unlock(sb);
11227 +static ssize_t aufs_write(struct file *file, const char __user *ubuf,
11228 + size_t count, loff_t *ppos)
11231 + struct au_pin pin;
11232 + struct dentry *dentry;
11233 + struct super_block *sb;
11234 + struct inode *inode;
11235 + struct file *h_file;
11236 + char __user *buf = (char __user *)ubuf;
11238 + dentry = file->f_dentry;
11239 + sb = dentry->d_sb;
11240 + inode = dentry->d_inode;
11241 + au_mtx_and_read_lock(inode);
11243 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
11244 + if (unlikely(err))
11247 + err = au_ready_to_write(file, -1, &pin);
11248 + di_downgrade_lock(dentry, AuLock_IR);
11249 + if (unlikely(err)) {
11250 + di_read_unlock(dentry, AuLock_IR);
11251 + fi_write_unlock(file);
11255 + h_file = au_hf_top(file);
11256 + get_file(h_file);
11258 + di_read_unlock(dentry, AuLock_IR);
11259 + fi_write_unlock(file);
11261 + err = vfsub_write_u(h_file, buf, count, ppos);
11262 + ii_write_lock_child(inode);
11263 + au_cpup_attr_timesizes(inode);
11264 + inode->i_mode = h_file->f_dentry->d_inode->i_mode;
11265 + ii_write_unlock(inode);
11269 + si_read_unlock(sb);
11270 + mutex_unlock(&inode->i_mutex);
11274 +static ssize_t au_do_aio(struct file *h_file, int rw, struct kiocb *kio,
11275 + const struct iovec *iov, unsigned long nv, loff_t pos)
11278 + struct file *file;
11279 + ssize_t (*func)(struct kiocb *, const struct iovec *, unsigned long,
11282 + err = security_file_permission(h_file, rw);
11283 + if (unlikely(err))
11288 + if (rw == MAY_READ)
11289 + func = h_file->f_op->aio_read;
11290 + else if (rw == MAY_WRITE)
11291 + func = h_file->f_op->aio_write;
11293 + file = kio->ki_filp;
11294 + kio->ki_filp = h_file;
11296 + err = func(kio, iov, nv, pos);
11298 + kio->ki_filp = file;
11300 + /* currently there is no such fs */
11307 +static ssize_t aufs_aio_read(struct kiocb *kio, const struct iovec *iov,
11308 + unsigned long nv, loff_t pos)
11311 + struct file *file, *h_file;
11312 + struct dentry *dentry;
11313 + struct super_block *sb;
11315 + file = kio->ki_filp;
11316 + dentry = file->f_dentry;
11317 + sb = dentry->d_sb;
11318 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
11319 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
11320 + if (unlikely(err))
11323 + h_file = au_hf_top(file);
11324 + get_file(h_file);
11325 + di_read_unlock(dentry, AuLock_IR);
11326 + fi_read_unlock(file);
11328 + err = au_do_aio(h_file, MAY_READ, kio, iov, nv, pos);
11329 + /* todo: necessary? */
11330 + /* file->f_ra = h_file->f_ra; */
11331 + /* update without lock, I don't think it a problem */
11332 + fsstack_copy_attr_atime(dentry->d_inode, h_file->f_dentry->d_inode);
11336 + si_read_unlock(sb);
11340 +static ssize_t aufs_aio_write(struct kiocb *kio, const struct iovec *iov,
11341 + unsigned long nv, loff_t pos)
11344 + struct au_pin pin;
11345 + struct dentry *dentry;
11346 + struct inode *inode;
11347 + struct file *file, *h_file;
11348 + struct super_block *sb;
11350 + file = kio->ki_filp;
11351 + dentry = file->f_dentry;
11352 + sb = dentry->d_sb;
11353 + inode = dentry->d_inode;
11354 + au_mtx_and_read_lock(inode);
11356 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
11357 + if (unlikely(err))
11360 + err = au_ready_to_write(file, -1, &pin);
11361 + di_downgrade_lock(dentry, AuLock_IR);
11362 + if (unlikely(err)) {
11363 + di_read_unlock(dentry, AuLock_IR);
11364 + fi_write_unlock(file);
11368 + h_file = au_hf_top(file);
11369 + get_file(h_file);
11371 + di_read_unlock(dentry, AuLock_IR);
11372 + fi_write_unlock(file);
11374 + err = au_do_aio(h_file, MAY_WRITE, kio, iov, nv, pos);
11375 + ii_write_lock_child(inode);
11376 + au_cpup_attr_timesizes(inode);
11377 + inode->i_mode = h_file->f_dentry->d_inode->i_mode;
11378 + ii_write_unlock(inode);
11382 + si_read_unlock(sb);
11383 + mutex_unlock(&inode->i_mutex);
11387 +static ssize_t aufs_splice_read(struct file *file, loff_t *ppos,
11388 + struct pipe_inode_info *pipe, size_t len,
11389 + unsigned int flags)
11392 + struct file *h_file;
11393 + struct dentry *dentry;
11394 + struct super_block *sb;
11396 + dentry = file->f_dentry;
11397 + sb = dentry->d_sb;
11398 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
11399 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
11400 + if (unlikely(err))
11404 + h_file = au_hf_top(file);
11405 + get_file(h_file);
11406 + if (au_test_loopback_kthread()) {
11407 + au_warn_loopback(h_file->f_dentry->d_sb);
11408 + if (file->f_mapping != h_file->f_mapping) {
11409 + file->f_mapping = h_file->f_mapping;
11410 + smp_mb(); /* unnecessary? */
11413 + di_read_unlock(dentry, AuLock_IR);
11414 + fi_read_unlock(file);
11416 + err = vfsub_splice_to(h_file, ppos, pipe, len, flags);
11417 + /* todo: necessasry? */
11418 + /* file->f_ra = h_file->f_ra; */
11419 + /* update without lock, I don't think it a problem */
11420 + fsstack_copy_attr_atime(dentry->d_inode, h_file->f_dentry->d_inode);
11424 + si_read_unlock(sb);
11429 +aufs_splice_write(struct pipe_inode_info *pipe, struct file *file, loff_t *ppos,
11430 + size_t len, unsigned int flags)
11433 + struct au_pin pin;
11434 + struct dentry *dentry;
11435 + struct inode *inode;
11436 + struct file *h_file;
11437 + struct super_block *sb;
11439 + dentry = file->f_dentry;
11440 + sb = dentry->d_sb;
11441 + inode = dentry->d_inode;
11442 + au_mtx_and_read_lock(inode);
11444 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
11445 + if (unlikely(err))
11448 + err = au_ready_to_write(file, -1, &pin);
11449 + di_downgrade_lock(dentry, AuLock_IR);
11450 + if (unlikely(err)) {
11451 + di_read_unlock(dentry, AuLock_IR);
11452 + fi_write_unlock(file);
11456 + h_file = au_hf_top(file);
11457 + get_file(h_file);
11459 + di_read_unlock(dentry, AuLock_IR);
11460 + fi_write_unlock(file);
11462 + err = vfsub_splice_from(pipe, h_file, ppos, len, flags);
11463 + ii_write_lock_child(inode);
11464 + au_cpup_attr_timesizes(inode);
11465 + inode->i_mode = h_file->f_dentry->d_inode->i_mode;
11466 + ii_write_unlock(inode);
11470 + si_read_unlock(sb);
11471 + mutex_unlock(&inode->i_mutex);
11475 +/* ---------------------------------------------------------------------- */
11478 + * The locking order around current->mmap_sem.
11479 + * - in most and regular cases
11480 + * file I/O syscall -- aufs_read() or something
11481 + * -- si_rwsem for read -- mmap_sem
11482 + * (Note that [fdi]i_rwsem are released before mmap_sem).
11484 + * mmap(2) -- mmap_sem -- aufs_mmap() -- si_rwsem for read -- [fdi]i_rwsem
11485 + * This AB-BA order is definitly bad, but is not a problem since "si_rwsem for
11486 + * read" allows muliple processes to acquire it and [fdi]i_rwsem are not held in
11487 + * file I/O. Aufs needs to stop lockdep in aufs_mmap() though.
11488 + * It means that when aufs acquires si_rwsem for write, the process should never
11489 + * acquire mmap_sem.
11491 + * Actually aufs_readdir() holds [fdi]i_rwsem before mmap_sem, but this is not a
11492 + * problem either since any directory is not able to be mmap-ed.
11493 + * The similar scenario is applied to aufs_readlink() too.
11496 +/* cf. linux/include/linux/mman.h: calc_vm_prot_bits() */
11497 +#define AuConv_VM_PROT(f, b) _calc_vm_trans(f, VM_##b, PROT_##b)
11499 +static unsigned long au_arch_prot_conv(unsigned long flags)
11501 + /* currently ppc64 only */
11502 +#ifdef CONFIG_PPC64
11503 + /* cf. linux/arch/powerpc/include/asm/mman.h */
11504 + AuDebugOn(arch_calc_vm_prot_bits(-1) != VM_SAO);
11505 + return AuConv_VM_PROT(flags, SAO);
11507 + AuDebugOn(arch_calc_vm_prot_bits(-1));
11512 +static unsigned long au_prot_conv(unsigned long flags)
11514 + return AuConv_VM_PROT(flags, READ)
11515 + | AuConv_VM_PROT(flags, WRITE)
11516 + | AuConv_VM_PROT(flags, EXEC)
11517 + | au_arch_prot_conv(flags);
11520 +/* cf. linux/include/linux/mman.h: calc_vm_flag_bits() */
11521 +#define AuConv_VM_MAP(f, b) _calc_vm_trans(f, VM_##b, MAP_##b)
11523 +static unsigned long au_flag_conv(unsigned long flags)
11525 + return AuConv_VM_MAP(flags, GROWSDOWN)
11526 + | AuConv_VM_MAP(flags, DENYWRITE)
11527 + | AuConv_VM_MAP(flags, EXECUTABLE)
11528 + | AuConv_VM_MAP(flags, LOCKED);
11531 +static int aufs_mmap(struct file *file, struct vm_area_struct *vma)
11534 + unsigned long prot;
11535 + aufs_bindex_t bstart;
11536 + const unsigned char wlock
11537 + = (file->f_mode & FMODE_WRITE) && (vma->vm_flags & VM_SHARED);
11538 + struct dentry *dentry;
11539 + struct super_block *sb;
11540 + struct file *h_file;
11541 + struct au_branch *br;
11542 + struct au_pin pin;
11544 + AuDbgVmRegion(file, vma);
11546 + dentry = file->f_dentry;
11547 + sb = dentry->d_sb;
11549 + si_read_lock(sb, AuLock_NOPLMW);
11550 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
11551 + if (unlikely(err))
11555 + err = au_ready_to_write(file, -1, &pin);
11556 + di_write_unlock(dentry);
11557 + if (unlikely(err)) {
11558 + fi_write_unlock(file);
11563 + di_write_unlock(dentry);
11565 + bstart = au_fbstart(file);
11566 + br = au_sbr(sb, bstart);
11567 + h_file = au_hf_top(file);
11568 + get_file(h_file);
11569 + au_set_mmapped(file);
11570 + fi_write_unlock(file);
11573 + au_vm_file_reset(vma, h_file);
11574 + prot = au_prot_conv(vma->vm_flags);
11575 + err = security_file_mmap(h_file, /*reqprot*/prot, prot,
11576 + au_flag_conv(vma->vm_flags), vma->vm_start, 0);
11578 + err = h_file->f_op->mmap(h_file, vma);
11579 + if (unlikely(err))
11582 + au_vm_prfile_set(vma, file);
11583 + /* update without lock, I don't think it a problem */
11584 + fsstack_copy_attr_atime(file->f_dentry->d_inode,
11585 + h_file->f_dentry->d_inode);
11586 + goto out_fput; /* success */
11589 + au_unset_mmapped(file);
11590 + au_vm_file_reset(vma, file);
11595 + si_read_unlock(sb);
11601 +/* ---------------------------------------------------------------------- */
11603 +static int aufs_fsync_nondir(struct file *file, loff_t start, loff_t end,
11607 + struct au_pin pin;
11608 + struct dentry *dentry;
11609 + struct inode *inode;
11610 + struct file *h_file;
11611 + struct super_block *sb;
11613 + dentry = file->f_dentry;
11614 + inode = dentry->d_inode;
11615 + sb = dentry->d_sb;
11616 + mutex_lock(&inode->i_mutex);
11617 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
11618 + if (unlikely(err))
11621 + err = 0; /* -EBADF; */ /* posix? */
11622 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
11624 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
11625 + if (unlikely(err))
11628 + err = au_ready_to_write(file, -1, &pin);
11629 + di_downgrade_lock(dentry, AuLock_IR);
11630 + if (unlikely(err))
11635 + h_file = au_hf_top(file);
11636 + err = vfsub_fsync(h_file, &h_file->f_path, datasync);
11637 + au_cpup_attr_timesizes(inode);
11640 + di_read_unlock(dentry, AuLock_IR);
11641 + fi_write_unlock(file);
11643 + si_read_unlock(sb);
11645 + mutex_unlock(&inode->i_mutex);
11649 +/* no one supports this operation, currently */
11651 +static int aufs_aio_fsync_nondir(struct kiocb *kio, int datasync)
11654 + struct au_pin pin;
11655 + struct dentry *dentry;
11656 + struct inode *inode;
11657 + struct file *file, *h_file;
11659 + file = kio->ki_filp;
11660 + dentry = file->f_dentry;
11661 + inode = dentry->d_inode;
11662 + au_mtx_and_read_lock(inode);
11664 + err = 0; /* -EBADF; */ /* posix? */
11665 + if (unlikely(!(file->f_mode & FMODE_WRITE)))
11667 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
11668 + if (unlikely(err))
11671 + err = au_ready_to_write(file, -1, &pin);
11672 + di_downgrade_lock(dentry, AuLock_IR);
11673 + if (unlikely(err))
11678 + h_file = au_hf_top(file);
11679 + if (h_file->f_op && h_file->f_op->aio_fsync) {
11680 + struct dentry *h_d;
11681 + struct mutex *h_mtx;
11683 + h_d = h_file->f_dentry;
11684 + h_mtx = &h_d->d_inode->i_mutex;
11685 + if (!is_sync_kiocb(kio)) {
11686 + get_file(h_file);
11689 + kio->ki_filp = h_file;
11690 + err = h_file->f_op->aio_fsync(kio, datasync);
11691 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
11693 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL);
11695 + au_cpup_attr_timesizes(inode);
11696 + mutex_unlock(h_mtx);
11700 + di_read_unlock(dentry, AuLock_IR);
11701 + fi_write_unlock(file);
11703 + si_read_unlock(inode->sb);
11704 + mutex_unlock(&inode->i_mutex);
11709 +static int aufs_fasync(int fd, struct file *file, int flag)
11712 + struct file *h_file;
11713 + struct dentry *dentry;
11714 + struct super_block *sb;
11716 + dentry = file->f_dentry;
11717 + sb = dentry->d_sb;
11718 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
11719 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
11720 + if (unlikely(err))
11723 + h_file = au_hf_top(file);
11724 + if (h_file->f_op && h_file->f_op->fasync)
11725 + err = h_file->f_op->fasync(fd, h_file, flag);
11727 + di_read_unlock(dentry, AuLock_IR);
11728 + fi_read_unlock(file);
11731 + si_read_unlock(sb);
11735 +/* ---------------------------------------------------------------------- */
11737 +/* no one supports this operation, currently */
11739 +static ssize_t aufs_sendpage(struct file *file, struct page *page, int offset,
11740 + size_t len, loff_t *pos , int more)
11745 +/* ---------------------------------------------------------------------- */
11747 +const struct file_operations aufs_file_fop = {
11748 + .owner = THIS_MODULE,
11750 + .llseek = default_llseek,
11752 + .read = aufs_read,
11753 + .write = aufs_write,
11754 + .aio_read = aufs_aio_read,
11755 + .aio_write = aufs_aio_write,
11756 +#ifdef CONFIG_AUFS_POLL
11757 + .poll = aufs_poll,
11759 + .unlocked_ioctl = aufs_ioctl_nondir,
11760 +#ifdef CONFIG_COMPAT
11761 + .compat_ioctl = aufs_ioctl_nondir, /* same */
11763 + .mmap = aufs_mmap,
11764 + .open = aufs_open_nondir,
11765 + .flush = aufs_flush_nondir,
11766 + .release = aufs_release_nondir,
11767 + .fsync = aufs_fsync_nondir,
11768 + /* .aio_fsync = aufs_aio_fsync_nondir, */
11769 + .fasync = aufs_fasync,
11770 + /* .sendpage = aufs_sendpage, */
11771 + .splice_write = aufs_splice_write,
11772 + .splice_read = aufs_splice_read,
11774 + .aio_splice_write = aufs_aio_splice_write,
11775 + .aio_splice_read = aufs_aio_splice_read
11778 diff -urN /usr/share/empty/fs/aufs/f_op_sp.c linux/fs/aufs/f_op_sp.c
11779 --- /usr/share/empty/fs/aufs/f_op_sp.c 1970-01-01 01:00:00.000000000 +0100
11780 +++ linux/fs/aufs/f_op_sp.c 2011-08-24 13:30:24.731313534 +0200
11783 + * Copyright (C) 2005-2011 Junjiro R. Okajima
11785 + * This program, aufs is free software; you can redistribute it and/or modify
11786 + * it under the terms of the GNU General Public License as published by
11787 + * the Free Software Foundation; either version 2 of the License, or
11788 + * (at your option) any later version.
11790 + * This program is distributed in the hope that it will be useful,
11791 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
11792 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11793 + * GNU General Public License for more details.
11795 + * You should have received a copy of the GNU General Public License
11796 + * along with this program; if not, write to the Free Software
11797 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
11801 + * file operations for special files.
11802 + * while they exist in aufs virtually,
11803 + * their file I/O is handled out of aufs.
11806 +#include <linux/fs_stack.h>
11809 +static ssize_t aufs_aio_read_sp(struct kiocb *kio, const struct iovec *iov,
11810 + unsigned long nv, loff_t pos)
11813 + aufs_bindex_t bstart;
11814 + unsigned char wbr;
11815 + struct file *file, *h_file;
11816 + struct super_block *sb;
11818 + file = kio->ki_filp;
11819 + sb = file->f_dentry->d_sb;
11820 + si_read_lock(sb, AuLock_FLUSH);
11821 + fi_read_lock(file);
11822 + bstart = au_fbstart(file);
11823 + h_file = au_hf_top(file);
11824 + fi_read_unlock(file);
11825 + wbr = !!au_br_writable(au_sbr(sb, bstart)->br_perm);
11826 + si_read_unlock(sb);
11828 + /* do not change the file in kio */
11829 + AuDebugOn(!h_file->f_op || !h_file->f_op->aio_read);
11830 + err = h_file->f_op->aio_read(kio, iov, nv, pos);
11831 + if (err > 0 && wbr)
11832 + file_accessed(h_file);
11837 +static ssize_t aufs_aio_write_sp(struct kiocb *kio, const struct iovec *iov,
11838 + unsigned long nv, loff_t pos)
11841 + aufs_bindex_t bstart;
11842 + unsigned char wbr;
11843 + struct super_block *sb;
11844 + struct file *file, *h_file;
11846 + file = kio->ki_filp;
11847 + sb = file->f_dentry->d_sb;
11848 + si_read_lock(sb, AuLock_FLUSH);
11849 + fi_read_lock(file);
11850 + bstart = au_fbstart(file);
11851 + h_file = au_hf_top(file);
11852 + fi_read_unlock(file);
11853 + wbr = !!au_br_writable(au_sbr(sb, bstart)->br_perm);
11854 + si_read_unlock(sb);
11856 + /* do not change the file in kio */
11857 + AuDebugOn(!h_file->f_op || !h_file->f_op->aio_write);
11858 + err = h_file->f_op->aio_write(kio, iov, nv, pos);
11859 + if (err > 0 && wbr)
11860 + file_update_time(h_file);
11865 +/* ---------------------------------------------------------------------- */
11867 +static int aufs_release_sp(struct inode *inode, struct file *file)
11870 + struct file *h_file;
11872 + fi_read_lock(file);
11873 + h_file = au_hf_top(file);
11874 + fi_read_unlock(file);
11875 + /* close this fifo in aufs */
11876 + err = h_file->f_op->release(inode, file); /* ignore */
11877 + aufs_release_nondir(inode, file); /* ignore */
11881 +/* ---------------------------------------------------------------------- */
11883 +/* currently, support only FIFO */
11885 + AuSp_FIFO, AuSp_FIFO_R, AuSp_FIFO_W, AuSp_FIFO_RW,
11886 + /* AuSp_SOCK, AuSp_CHR, AuSp_BLK, */
11889 +static int aufs_open_sp(struct inode *inode, struct file *file);
11890 +static struct au_sp_fop {
11892 + struct file_operations fop; /* not 'const' */
11894 +} au_sp_fop[AuSp_Last] = {
11897 + .owner = THIS_MODULE,
11898 + .open = aufs_open_sp
11903 +static void au_init_fop_sp(struct file *file)
11905 + struct au_sp_fop *p;
11907 + struct file *h_file;
11910 + if (unlikely(!p->done)) {
11911 + /* initialize first time only */
11912 + static DEFINE_SPINLOCK(spin);
11914 + spin_lock(&spin);
11916 + BUILD_BUG_ON(sizeof(au_sp_fop)/sizeof(*au_sp_fop)
11918 + for (i = 0; i < AuSp_Last; i++)
11919 + spin_lock_init(&p[i].spin);
11922 + spin_unlock(&spin);
11925 + switch (file->f_mode & (FMODE_READ | FMODE_WRITE)) {
11929 + case FMODE_WRITE:
11932 + case FMODE_READ | FMODE_WRITE:
11933 + i = AuSp_FIFO_RW;
11940 + if (unlikely(!p->done)) {
11941 + /* initialize first time only */
11942 + h_file = au_hf_top(file);
11943 + spin_lock(&p->spin);
11945 + p->fop = *h_file->f_op;
11946 + p->fop.owner = THIS_MODULE;
11947 + if (p->fop.aio_read)
11948 + p->fop.aio_read = aufs_aio_read_sp;
11949 + if (p->fop.aio_write)
11950 + p->fop.aio_write = aufs_aio_write_sp;
11951 + p->fop.release = aufs_release_sp;
11954 + spin_unlock(&p->spin);
11956 + file->f_op = &p->fop;
11959 +static int au_cpup_sp(struct dentry *dentry)
11962 + aufs_bindex_t bcpup;
11963 + struct au_pin pin;
11964 + struct au_wr_dir_args wr_dir_args = {
11965 + .force_btgt = -1,
11969 + AuDbg("%.*s\n", AuDLNPair(dentry));
11971 + di_read_unlock(dentry, AuLock_IR);
11972 + di_write_lock_child(dentry);
11973 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
11974 + if (unlikely(err < 0))
11978 + if (bcpup == au_dbstart(dentry))
11979 + goto out; /* success */
11981 + err = au_pin(&pin, dentry, bcpup, au_opt_udba(dentry->d_sb),
11982 + AuPin_MNT_WRITE);
11984 + err = au_sio_cpup_simple(dentry, bcpup, -1, AuCpup_DTIME);
11989 + di_downgrade_lock(dentry, AuLock_IR);
11993 +static int au_do_open_sp(struct file *file, int flags)
11996 + struct dentry *dentry;
11997 + struct super_block *sb;
11998 + struct file *h_file;
11999 + struct inode *h_inode;
12001 + dentry = file->f_dentry;
12002 + AuDbg("%.*s\n", AuDLNPair(dentry));
12005 + * try copying-up.
12006 + * operate on the ro branch is not an error.
12008 + au_cpup_sp(dentry); /* ignore */
12010 + /* prepare h_file */
12011 + err = au_do_open_nondir(file, vfsub_file_flags(file));
12012 + if (unlikely(err))
12015 + sb = dentry->d_sb;
12016 + h_file = au_hf_top(file);
12017 + h_inode = h_file->f_dentry->d_inode;
12018 + di_read_unlock(dentry, AuLock_IR);
12019 + fi_write_unlock(file);
12020 + si_read_unlock(sb);
12021 + /* open this fifo in aufs */
12022 + err = h_inode->i_fop->open(file->f_dentry->d_inode, file);
12023 + si_noflush_read_lock(sb);
12024 + fi_write_lock(file);
12025 + di_read_lock_child(dentry, AuLock_IR);
12027 + au_init_fop_sp(file);
12033 +static int aufs_open_sp(struct inode *inode, struct file *file)
12036 + struct super_block *sb;
12038 + sb = file->f_dentry->d_sb;
12039 + si_read_lock(sb, AuLock_FLUSH);
12040 + err = au_do_open(file, au_do_open_sp, /*fidir*/NULL);
12041 + si_read_unlock(sb);
12045 +/* ---------------------------------------------------------------------- */
12047 +void au_init_special_fop(struct inode *inode, umode_t mode, dev_t rdev)
12049 + init_special_inode(inode, mode, rdev);
12051 + switch (mode & S_IFMT) {
12053 + inode->i_fop = &au_sp_fop[AuSp_FIFO].fop;
12064 +int au_special_file(umode_t mode)
12069 + switch (mode & S_IFMT) {
12081 diff -urN /usr/share/empty/fs/aufs/fstype.h linux/fs/aufs/fstype.h
12082 --- /usr/share/empty/fs/aufs/fstype.h 1970-01-01 01:00:00.000000000 +0100
12083 +++ linux/fs/aufs/fstype.h 2011-08-24 13:30:24.731313534 +0200
12086 + * Copyright (C) 2005-2011 Junjiro R. Okajima
12088 + * This program, aufs is free software; you can redistribute it and/or modify
12089 + * it under the terms of the GNU General Public License as published by
12090 + * the Free Software Foundation; either version 2 of the License, or
12091 + * (at your option) any later version.
12093 + * This program is distributed in the hope that it will be useful,
12094 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12095 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12096 + * GNU General Public License for more details.
12098 + * You should have received a copy of the GNU General Public License
12099 + * along with this program; if not, write to the Free Software
12100 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
12104 + * judging filesystem type
12107 +#ifndef __AUFS_FSTYPE_H__
12108 +#define __AUFS_FSTYPE_H__
12112 +#include <linux/fs.h>
12113 +#include <linux/magic.h>
12114 +#include <linux/romfs_fs.h>
12115 +#include <linux/aufs_type.h>
12117 +static inline int au_test_aufs(struct super_block *sb)
12119 + return sb->s_magic == AUFS_SUPER_MAGIC;
12122 +static inline const char *au_sbtype(struct super_block *sb)
12124 + return sb->s_type->name;
12127 +static inline int au_test_iso9660(struct super_block *sb __maybe_unused)
12129 +#if defined(CONFIG_ROMFS_FS) || defined(CONFIG_ROMFS_FS_MODULE)
12130 + return sb->s_magic == ROMFS_MAGIC;
12136 +static inline int au_test_romfs(struct super_block *sb __maybe_unused)
12138 +#if defined(CONFIG_ISO9660_FS) || defined(CONFIG_ISO9660_FS_MODULE)
12139 + return sb->s_magic == ISOFS_SUPER_MAGIC;
12145 +static inline int au_test_cramfs(struct super_block *sb __maybe_unused)
12147 +#if defined(CONFIG_CRAMFS) || defined(CONFIG_CRAMFS_MODULE)
12148 + return sb->s_magic == CRAMFS_MAGIC;
12153 +static inline int au_test_nfs(struct super_block *sb __maybe_unused)
12155 +#if defined(CONFIG_NFS_FS) || defined(CONFIG_NFS_FS_MODULE)
12156 + return sb->s_magic == NFS_SUPER_MAGIC;
12162 +static inline int au_test_fuse(struct super_block *sb __maybe_unused)
12164 +#if defined(CONFIG_FUSE_FS) || defined(CONFIG_FUSE_FS_MODULE)
12165 + return sb->s_magic == FUSE_SUPER_MAGIC;
12171 +static inline int au_test_xfs(struct super_block *sb __maybe_unused)
12173 +#if defined(CONFIG_XFS_FS) || defined(CONFIG_XFS_FS_MODULE)
12174 + return sb->s_magic == XFS_SB_MAGIC;
12180 +static inline int au_test_tmpfs(struct super_block *sb __maybe_unused)
12182 +#ifdef CONFIG_TMPFS
12183 + return sb->s_magic == TMPFS_MAGIC;
12189 +static inline int au_test_ecryptfs(struct super_block *sb __maybe_unused)
12191 +#if defined(CONFIG_ECRYPT_FS) || defined(CONFIG_ECRYPT_FS_MODULE)
12192 + return !strcmp(au_sbtype(sb), "ecryptfs");
12198 +static inline int au_test_smbfs(struct super_block *sb __maybe_unused)
12200 +#if defined(CONFIG_SMB_FS) || defined(CONFIG_SMB_FS_MODULE)
12201 + return sb->s_magic == SMB_SUPER_MAGIC;
12207 +static inline int au_test_ocfs2(struct super_block *sb __maybe_unused)
12209 +#if defined(CONFIG_OCFS2_FS) || defined(CONFIG_OCFS2_FS_MODULE)
12210 + return sb->s_magic == OCFS2_SUPER_MAGIC;
12216 +static inline int au_test_ocfs2_dlmfs(struct super_block *sb __maybe_unused)
12218 +#if defined(CONFIG_OCFS2_FS_O2CB) || defined(CONFIG_OCFS2_FS_O2CB_MODULE)
12219 + return sb->s_magic == DLMFS_MAGIC;
12225 +static inline int au_test_coda(struct super_block *sb __maybe_unused)
12227 +#if defined(CONFIG_CODA_FS) || defined(CONFIG_CODA_FS_MODULE)
12228 + return sb->s_magic == CODA_SUPER_MAGIC;
12234 +static inline int au_test_v9fs(struct super_block *sb __maybe_unused)
12236 +#if defined(CONFIG_9P_FS) || defined(CONFIG_9P_FS_MODULE)
12237 + return sb->s_magic == V9FS_MAGIC;
12243 +static inline int au_test_ext4(struct super_block *sb __maybe_unused)
12245 +#if defined(CONFIG_EXT4DEV_FS) || defined(CONFIG_EXT4DEV_FS_MODULE)
12246 + return sb->s_magic == EXT4_SUPER_MAGIC;
12252 +static inline int au_test_sysv(struct super_block *sb __maybe_unused)
12254 +#if defined(CONFIG_SYSV_FS) || defined(CONFIG_SYSV_FS_MODULE)
12255 + return !strcmp(au_sbtype(sb), "sysv");
12261 +static inline int au_test_ramfs(struct super_block *sb)
12263 + return sb->s_magic == RAMFS_MAGIC;
12266 +static inline int au_test_ubifs(struct super_block *sb __maybe_unused)
12268 +#if defined(CONFIG_UBIFS_FS) || defined(CONFIG_UBIFS_FS_MODULE)
12269 + return sb->s_magic == UBIFS_SUPER_MAGIC;
12275 +static inline int au_test_procfs(struct super_block *sb __maybe_unused)
12277 +#ifdef CONFIG_PROC_FS
12278 + return sb->s_magic == PROC_SUPER_MAGIC;
12284 +static inline int au_test_sysfs(struct super_block *sb __maybe_unused)
12286 +#ifdef CONFIG_SYSFS
12287 + return sb->s_magic == SYSFS_MAGIC;
12293 +static inline int au_test_configfs(struct super_block *sb __maybe_unused)
12295 +#if defined(CONFIG_CONFIGFS_FS) || defined(CONFIG_CONFIGFS_FS_MODULE)
12296 + return sb->s_magic == CONFIGFS_MAGIC;
12302 +static inline int au_test_minix(struct super_block *sb __maybe_unused)
12304 +#if defined(CONFIG_MINIX_FS) || defined(CONFIG_MINIX_FS_MODULE)
12305 + return sb->s_magic == MINIX3_SUPER_MAGIC
12306 + || sb->s_magic == MINIX2_SUPER_MAGIC
12307 + || sb->s_magic == MINIX2_SUPER_MAGIC2
12308 + || sb->s_magic == MINIX_SUPER_MAGIC
12309 + || sb->s_magic == MINIX_SUPER_MAGIC2;
12315 +static inline int au_test_cifs(struct super_block *sb __maybe_unused)
12317 +#if defined(CONFIG_CIFS_FS) || defined(CONFIGCIFS_FS_MODULE)
12318 + return sb->s_magic == CIFS_MAGIC_NUMBER;
12324 +static inline int au_test_fat(struct super_block *sb __maybe_unused)
12326 +#if defined(CONFIG_FAT_FS) || defined(CONFIG_FAT_FS_MODULE)
12327 + return sb->s_magic == MSDOS_SUPER_MAGIC;
12333 +static inline int au_test_msdos(struct super_block *sb)
12335 + return au_test_fat(sb);
12338 +static inline int au_test_vfat(struct super_block *sb)
12340 + return au_test_fat(sb);
12343 +static inline int au_test_securityfs(struct super_block *sb __maybe_unused)
12345 +#ifdef CONFIG_SECURITYFS
12346 + return sb->s_magic == SECURITYFS_MAGIC;
12352 +static inline int au_test_squashfs(struct super_block *sb __maybe_unused)
12354 +#if defined(CONFIG_SQUASHFS) || defined(CONFIG_SQUASHFS_MODULE)
12355 + return sb->s_magic == SQUASHFS_MAGIC;
12361 +static inline int au_test_btrfs(struct super_block *sb __maybe_unused)
12363 +#if defined(CONFIG_BTRFS_FS) || defined(CONFIG_BTRFS_FS_MODULE)
12364 + return sb->s_magic == BTRFS_SUPER_MAGIC;
12370 +static inline int au_test_xenfs(struct super_block *sb __maybe_unused)
12372 +#if defined(CONFIG_XENFS) || defined(CONFIG_XENFS_MODULE)
12373 + return sb->s_magic == XENFS_SUPER_MAGIC;
12379 +static inline int au_test_debugfs(struct super_block *sb __maybe_unused)
12381 +#ifdef CONFIG_DEBUG_FS
12382 + return sb->s_magic == DEBUGFS_MAGIC;
12388 +static inline int au_test_nilfs(struct super_block *sb __maybe_unused)
12390 +#if defined(CONFIG_NILFS) || defined(CONFIG_NILFS_MODULE)
12391 + return sb->s_magic == NILFS_SUPER_MAGIC;
12397 +static inline int au_test_hfsplus(struct super_block *sb __maybe_unused)
12399 +#if defined(CONFIG_HFSPLUS_FS) || defined(CONFIG_HFSPLUS_FS_MODULE)
12400 + return sb->s_magic == HFSPLUS_SUPER_MAGIC;
12406 +/* ---------------------------------------------------------------------- */
12408 + * they can't be an aufs branch.
12410 +static inline int au_test_fs_unsuppoted(struct super_block *sb)
12413 +#ifndef CONFIG_AUFS_BR_RAMFS
12414 + au_test_ramfs(sb) ||
12416 + au_test_procfs(sb)
12417 + || au_test_sysfs(sb)
12418 + || au_test_configfs(sb)
12419 + || au_test_debugfs(sb)
12420 + || au_test_securityfs(sb)
12421 + || au_test_xenfs(sb)
12422 + || au_test_ecryptfs(sb)
12423 + /* || !strcmp(au_sbtype(sb), "unionfs") */
12424 + || au_test_aufs(sb); /* will be supported in next version */
12428 + * If the filesystem supports NFS-export, then it has to support NULL as
12429 + * a nameidata parameter for ->create(), ->lookup() and ->d_revalidate().
12430 + * We can apply this principle when we handle a lower filesystem.
12432 +static inline int au_test_fs_null_nd(struct super_block *sb)
12434 + return !!sb->s_export_op;
12437 +static inline int au_test_fs_remote(struct super_block *sb)
12439 + return !au_test_tmpfs(sb)
12440 +#ifdef CONFIG_AUFS_BR_RAMFS
12441 + && !au_test_ramfs(sb)
12443 + && !(sb->s_type->fs_flags & FS_REQUIRES_DEV);
12446 +/* ---------------------------------------------------------------------- */
12449 + * Note: these functions (below) are created after reading ->getattr() in all
12450 + * filesystems under linux/fs. it means we have to do so in every update...
12454 + * some filesystems require getattr to refresh the inode attributes before
12456 + * in most cases, we can rely on the inode attribute in NFS (or every remote fs)
12457 + * and leave the work for d_revalidate()
12459 +static inline int au_test_fs_refresh_iattr(struct super_block *sb)
12461 + return au_test_nfs(sb)
12462 + || au_test_fuse(sb)
12463 + /* || au_test_smbfs(sb) */ /* untested */
12464 + /* || au_test_ocfs2(sb) */ /* untested */
12465 + /* || au_test_btrfs(sb) */ /* untested */
12466 + /* || au_test_coda(sb) */ /* untested */
12467 + /* || au_test_v9fs(sb) */ /* untested */
12472 + * filesystems which don't maintain i_size or i_blocks.
12474 +static inline int au_test_fs_bad_iattr_size(struct super_block *sb)
12476 + return au_test_xfs(sb)
12477 + || au_test_btrfs(sb)
12478 + || au_test_ubifs(sb)
12479 + || au_test_hfsplus(sb) /* maintained, but incorrect */
12480 + /* || au_test_ext4(sb) */ /* untested */
12481 + /* || au_test_ocfs2(sb) */ /* untested */
12482 + /* || au_test_ocfs2_dlmfs(sb) */ /* untested */
12483 + /* || au_test_sysv(sb) */ /* untested */
12484 + /* || au_test_minix(sb) */ /* untested */
12489 + * filesystems which don't store the correct value in some of their inode
12492 +static inline int au_test_fs_bad_iattr(struct super_block *sb)
12494 + return au_test_fs_bad_iattr_size(sb)
12495 + /* || au_test_cifs(sb) */ /* untested */
12496 + || au_test_fat(sb)
12497 + || au_test_msdos(sb)
12498 + || au_test_vfat(sb);
12501 +/* they don't check i_nlink in link(2) */
12502 +static inline int au_test_fs_no_limit_nlink(struct super_block *sb)
12504 + return au_test_tmpfs(sb)
12505 +#ifdef CONFIG_AUFS_BR_RAMFS
12506 + || au_test_ramfs(sb)
12508 + || au_test_ubifs(sb)
12509 + || au_test_btrfs(sb)
12510 + || au_test_hfsplus(sb);
12514 + * filesystems which sets S_NOATIME and S_NOCMTIME.
12516 +static inline int au_test_fs_notime(struct super_block *sb)
12518 + return au_test_nfs(sb)
12519 + || au_test_fuse(sb)
12520 + || au_test_ubifs(sb)
12521 + /* || au_test_cifs(sb) */ /* untested */
12526 + * filesystems which requires replacing i_mapping.
12528 +static inline int au_test_fs_bad_mapping(struct super_block *sb)
12530 + return au_test_fuse(sb)
12531 + || au_test_ubifs(sb);
12534 +/* temporary support for i#1 in cramfs */
12535 +static inline int au_test_fs_unique_ino(struct inode *inode)
12537 + if (au_test_cramfs(inode->i_sb))
12538 + return inode->i_ino != 1;
12542 +/* ---------------------------------------------------------------------- */
12545 + * the filesystem where the xino files placed must support i/o after unlink and
12546 + * maintain i_size and i_blocks.
12548 +static inline int au_test_fs_bad_xino(struct super_block *sb)
12550 + return au_test_fs_remote(sb)
12551 + || au_test_fs_bad_iattr_size(sb)
12552 +#ifdef CONFIG_AUFS_BR_RAMFS
12553 + || !(au_test_ramfs(sb) || au_test_fs_null_nd(sb))
12555 + || !au_test_fs_null_nd(sb) /* to keep xino code simple */
12557 + /* don't want unnecessary work for xino */
12558 + || au_test_aufs(sb)
12559 + || au_test_ecryptfs(sb)
12560 + || au_test_nilfs(sb);
12563 +static inline int au_test_fs_trunc_xino(struct super_block *sb)
12565 + return au_test_tmpfs(sb)
12566 + || au_test_ramfs(sb);
12570 + * test if the @sb is real-readonly.
12572 +static inline int au_test_fs_rr(struct super_block *sb)
12574 + return au_test_squashfs(sb)
12575 + || au_test_iso9660(sb)
12576 + || au_test_cramfs(sb)
12577 + || au_test_romfs(sb);
12580 +#endif /* __KERNEL__ */
12581 +#endif /* __AUFS_FSTYPE_H__ */
12582 diff -urN /usr/share/empty/fs/aufs/hfsnotify.c linux/fs/aufs/hfsnotify.c
12583 --- /usr/share/empty/fs/aufs/hfsnotify.c 1970-01-01 01:00:00.000000000 +0100
12584 +++ linux/fs/aufs/hfsnotify.c 2011-08-24 13:30:24.731313534 +0200
12587 + * Copyright (C) 2005-2011 Junjiro R. Okajima
12589 + * This program, aufs is free software; you can redistribute it and/or modify
12590 + * it under the terms of the GNU General Public License as published by
12591 + * the Free Software Foundation; either version 2 of the License, or
12592 + * (at your option) any later version.
12594 + * This program is distributed in the hope that it will be useful,
12595 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12596 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12597 + * GNU General Public License for more details.
12599 + * You should have received a copy of the GNU General Public License
12600 + * along with this program; if not, write to the Free Software
12601 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
12605 + * fsnotify for the lower directories
12610 +/* FS_IN_IGNORED is unnecessary */
12611 +static const __u32 AuHfsnMask = (FS_MOVED_TO | FS_MOVED_FROM | FS_DELETE
12612 + | FS_CREATE | FS_EVENT_ON_CHILD);
12613 +static DECLARE_WAIT_QUEUE_HEAD(au_hfsn_wq);
12615 +static void au_hfsn_free_mark(struct fsnotify_mark *mark)
12617 + struct au_hnotify *hn = container_of(mark, struct au_hnotify,
12620 + hn->hn_mark_dead = 1;
12622 + wake_up_all(&au_hfsn_wq);
12625 +static int au_hfsn_alloc(struct au_hinode *hinode)
12627 + struct au_hnotify *hn;
12628 + struct super_block *sb;
12629 + struct au_branch *br;
12630 + struct fsnotify_mark *mark;
12631 + aufs_bindex_t bindex;
12633 + hn = hinode->hi_notify;
12634 + sb = hn->hn_aufs_inode->i_sb;
12635 + bindex = au_br_index(sb, hinode->hi_id);
12636 + br = au_sbr(sb, bindex);
12637 + hn->hn_mark_dead = 0;
12638 + mark = &hn->hn_mark;
12639 + fsnotify_init_mark(mark, au_hfsn_free_mark);
12640 + mark->mask = AuHfsnMask;
12642 + * by udba rename or rmdir, aufs assign a new inode to the known
12643 + * h_inode, so specify 1 to allow dups.
12645 + return fsnotify_add_mark(mark, br->br_hfsn_group, hinode->hi_inode,
12646 + /*mnt*/NULL, /*allow_dups*/1);
12649 +static void au_hfsn_free(struct au_hinode *hinode)
12651 + struct au_hnotify *hn;
12652 + struct fsnotify_mark *mark;
12654 + hn = hinode->hi_notify;
12655 + mark = &hn->hn_mark;
12656 + fsnotify_destroy_mark(mark);
12657 + fsnotify_put_mark(mark);
12659 + /* TODO: bad approach */
12660 + wait_event(au_hfsn_wq, hn->hn_mark_dead);
12663 +/* ---------------------------------------------------------------------- */
12665 +static void au_hfsn_ctl(struct au_hinode *hinode, int do_set)
12667 + struct fsnotify_mark *mark;
12669 + mark = &hinode->hi_notify->hn_mark;
12670 + spin_lock(&mark->lock);
12672 + AuDebugOn(mark->mask & AuHfsnMask);
12673 + mark->mask |= AuHfsnMask;
12675 + AuDebugOn(!(mark->mask & AuHfsnMask));
12676 + mark->mask &= ~AuHfsnMask;
12678 + spin_unlock(&mark->lock);
12679 + /* fsnotify_recalc_inode_mask(hinode->hi_inode); */
12682 +/* ---------------------------------------------------------------------- */
12684 +/* #define AuDbgHnotify */
12685 +#ifdef AuDbgHnotify
12686 +static char *au_hfsn_name(u32 mask)
12688 +#ifdef CONFIG_AUFS_DEBUG
12689 +#define test_ret(flag) if (mask & flag) \
12691 + test_ret(FS_ACCESS);
12692 + test_ret(FS_MODIFY);
12693 + test_ret(FS_ATTRIB);
12694 + test_ret(FS_CLOSE_WRITE);
12695 + test_ret(FS_CLOSE_NOWRITE);
12696 + test_ret(FS_OPEN);
12697 + test_ret(FS_MOVED_FROM);
12698 + test_ret(FS_MOVED_TO);
12699 + test_ret(FS_CREATE);
12700 + test_ret(FS_DELETE);
12701 + test_ret(FS_DELETE_SELF);
12702 + test_ret(FS_MOVE_SELF);
12703 + test_ret(FS_UNMOUNT);
12704 + test_ret(FS_Q_OVERFLOW);
12705 + test_ret(FS_IN_IGNORED);
12706 + test_ret(FS_IN_ISDIR);
12707 + test_ret(FS_IN_ONESHOT);
12708 + test_ret(FS_EVENT_ON_CHILD);
12717 +/* ---------------------------------------------------------------------- */
12719 +static int au_hfsn_handle_event(struct fsnotify_group *group,
12720 + struct fsnotify_mark *inode_mark,
12721 + struct fsnotify_mark *vfsmount_mark,
12722 + struct fsnotify_event *event)
12725 + struct au_hnotify *hnotify;
12726 + struct inode *h_dir, *h_inode;
12728 + struct qstr h_child_qstr = {
12729 + .name = event->file_name,
12730 + .len = event->name_len
12733 + AuDebugOn(event->data_type != FSNOTIFY_EVENT_INODE);
12736 + /* if FS_UNMOUNT happens, there must be another bug */
12737 + mask = event->mask;
12738 + AuDebugOn(mask & FS_UNMOUNT);
12739 + if (mask & (FS_IN_IGNORED | FS_UNMOUNT))
12742 + h_dir = event->to_tell;
12743 + h_inode = event->inode;
12744 +#ifdef AuDbgHnotify
12746 + if (1 || h_child_qstr.len != sizeof(AUFS_XINO_FNAME) - 1
12747 + || strncmp(h_child_qstr.name, AUFS_XINO_FNAME, h_child_qstr.len)) {
12748 + AuDbg("i%lu, mask 0x%x %s, hcname %.*s, hi%lu\n",
12749 + h_dir->i_ino, mask, au_hfsn_name(mask),
12750 + AuLNPair(&h_child_qstr), h_inode ? h_inode->i_ino : 0);
12751 + /* WARN_ON(1); */
12756 + AuDebugOn(!inode_mark);
12757 + hnotify = container_of(inode_mark, struct au_hnotify, hn_mark);
12758 + err = au_hnotify(h_dir, hnotify, mask, &h_child_qstr, h_inode);
12764 +/* isn't it waste to ask every registered 'group'? */
12765 +/* copied from linux/fs/notify/inotify/inotify_fsnotiry.c */
12766 +/* it should be exported to modules */
12767 +static bool au_hfsn_should_send_event(struct fsnotify_group *group,
12768 + struct inode *h_inode,
12769 + struct fsnotify_mark *inode_mark,
12770 + struct fsnotify_mark *vfsmount_mark,
12771 + __u32 mask, void *data, int data_type)
12773 + mask = (mask & ~FS_EVENT_ON_CHILD);
12774 + return inode_mark->mask & mask;
12777 +static struct fsnotify_ops au_hfsn_ops = {
12778 + .should_send_event = au_hfsn_should_send_event,
12779 + .handle_event = au_hfsn_handle_event
12782 +/* ---------------------------------------------------------------------- */
12784 +static void au_hfsn_fin_br(struct au_branch *br)
12786 + if (br->br_hfsn_group)
12787 + fsnotify_put_group(br->br_hfsn_group);
12790 +static int au_hfsn_init_br(struct au_branch *br, int perm)
12792 + br->br_hfsn_group = NULL;
12793 + br->br_hfsn_ops = au_hfsn_ops;
12797 +static int au_hfsn_reset_br(unsigned int udba, struct au_branch *br, int perm)
12802 + if (udba != AuOpt_UDBA_HNOTIFY
12803 + || !au_br_hnotifyable(perm)) {
12804 + au_hfsn_fin_br(br);
12805 + br->br_hfsn_group = NULL;
12809 + if (br->br_hfsn_group)
12812 + br->br_hfsn_group = fsnotify_alloc_group(&br->br_hfsn_ops);
12813 + if (IS_ERR(br->br_hfsn_group)) {
12814 + err = PTR_ERR(br->br_hfsn_group);
12815 + pr_err("fsnotify_alloc_group() failed, %d\n", err);
12816 + br->br_hfsn_group = NULL;
12824 +const struct au_hnotify_op au_hnotify_op = {
12825 + .ctl = au_hfsn_ctl,
12826 + .alloc = au_hfsn_alloc,
12827 + .free = au_hfsn_free,
12829 + .reset_br = au_hfsn_reset_br,
12830 + .fin_br = au_hfsn_fin_br,
12831 + .init_br = au_hfsn_init_br
12833 diff -urN /usr/share/empty/fs/aufs/hfsplus.c linux/fs/aufs/hfsplus.c
12834 --- /usr/share/empty/fs/aufs/hfsplus.c 1970-01-01 01:00:00.000000000 +0100
12835 +++ linux/fs/aufs/hfsplus.c 2011-08-24 13:30:24.731313534 +0200
12838 + * Copyright (C) 2010-2011 Junjiro R. Okajima
12840 + * This program, aufs is free software; you can redistribute it and/or modify
12841 + * it under the terms of the GNU General Public License as published by
12842 + * the Free Software Foundation; either version 2 of the License, or
12843 + * (at your option) any later version.
12845 + * This program is distributed in the hope that it will be useful,
12846 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12847 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12848 + * GNU General Public License for more details.
12850 + * You should have received a copy of the GNU General Public License
12851 + * along with this program; if not, write to the Free Software
12852 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
12856 + * special support for filesystems which aqucires an inode mutex
12857 + * at final closing a file, eg, hfsplus.
12859 + * This trick is very simple and stupid, just to open the file before really
12860 + * neceeary open to tell hfsplus that this is not the final closing.
12861 + * The caller should call au_h_open_pre() after acquiring the inode mutex,
12862 + * and au_h_open_post() after releasing it.
12865 +#include <linux/file.h>
12868 +struct file *au_h_open_pre(struct dentry *dentry, aufs_bindex_t bindex)
12870 + struct file *h_file;
12871 + struct dentry *h_dentry;
12873 + h_dentry = au_h_dptr(dentry, bindex);
12874 + AuDebugOn(!h_dentry);
12875 + AuDebugOn(!h_dentry->d_inode);
12876 + IMustLock(h_dentry->d_inode);
12879 + if (au_test_hfsplus(h_dentry->d_sb)
12880 + && S_ISREG(h_dentry->d_inode->i_mode))
12881 + h_file = au_h_open(dentry, bindex,
12882 + O_RDONLY | O_NOATIME | O_LARGEFILE,
12887 +void au_h_open_post(struct dentry *dentry, aufs_bindex_t bindex,
12888 + struct file *h_file)
12892 + au_sbr_put(dentry->d_sb, bindex);
12895 diff -urN /usr/share/empty/fs/aufs/hnotify.c linux/fs/aufs/hnotify.c
12896 --- /usr/share/empty/fs/aufs/hnotify.c 1970-01-01 01:00:00.000000000 +0100
12897 +++ linux/fs/aufs/hnotify.c 2011-08-24 13:30:24.731313534 +0200
12900 + * Copyright (C) 2005-2011 Junjiro R. Okajima
12902 + * This program, aufs is free software; you can redistribute it and/or modify
12903 + * it under the terms of the GNU General Public License as published by
12904 + * the Free Software Foundation; either version 2 of the License, or
12905 + * (at your option) any later version.
12907 + * This program is distributed in the hope that it will be useful,
12908 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
12909 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12910 + * GNU General Public License for more details.
12912 + * You should have received a copy of the GNU General Public License
12913 + * along with this program; if not, write to the Free Software
12914 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
12918 + * abstraction to notify the direct changes on lower directories
12923 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode)
12926 + struct au_hnotify *hn;
12929 + hn = au_cache_alloc_hnotify();
12931 + hn->hn_aufs_inode = inode;
12932 + hinode->hi_notify = hn;
12933 + err = au_hnotify_op.alloc(hinode);
12935 + if (unlikely(err)) {
12936 + hinode->hi_notify = NULL;
12937 + au_cache_free_hnotify(hn);
12939 + * The upper dir was removed by udba, but the same named
12940 + * dir left. In this case, aufs assignes a new inode
12941 + * number and set the monitor again.
12942 + * For the lower dir, the old monitnor is still left.
12944 + if (err == -EEXIST)
12953 +void au_hn_free(struct au_hinode *hinode)
12955 + struct au_hnotify *hn;
12957 + hn = hinode->hi_notify;
12959 + au_hnotify_op.free(hinode);
12960 + au_cache_free_hnotify(hn);
12961 + hinode->hi_notify = NULL;
12965 +/* ---------------------------------------------------------------------- */
12967 +void au_hn_ctl(struct au_hinode *hinode, int do_set)
12969 + if (hinode->hi_notify)
12970 + au_hnotify_op.ctl(hinode, do_set);
12973 +void au_hn_reset(struct inode *inode, unsigned int flags)
12975 + aufs_bindex_t bindex, bend;
12976 + struct inode *hi;
12977 + struct dentry *iwhdentry;
12979 + bend = au_ibend(inode);
12980 + for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
12981 + hi = au_h_iptr(inode, bindex);
12985 + /* mutex_lock_nested(&hi->i_mutex, AuLsc_I_CHILD); */
12986 + iwhdentry = au_hi_wh(inode, bindex);
12990 + au_set_h_iptr(inode, bindex, NULL, 0);
12991 + au_set_h_iptr(inode, bindex, au_igrab(hi),
12992 + flags & ~AuHi_XINO);
12995 + /* mutex_unlock(&hi->i_mutex); */
12999 +/* ---------------------------------------------------------------------- */
13001 +static int hn_xino(struct inode *inode, struct inode *h_inode)
13004 + aufs_bindex_t bindex, bend, bfound, bstart;
13005 + struct inode *h_i;
13008 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
13009 + pr_warning("branch root dir was changed\n");
13014 + bend = au_ibend(inode);
13015 + bstart = au_ibstart(inode);
13016 +#if 0 /* reserved for future use */
13017 + if (bindex == bend) {
13018 + /* keep this ino in rename case */
13022 + for (bindex = bstart; bindex <= bend; bindex++)
13023 + if (au_h_iptr(inode, bindex) == h_inode) {
13030 + for (bindex = bstart; bindex <= bend; bindex++) {
13031 + h_i = au_h_iptr(inode, bindex);
13035 + err = au_xino_write(inode->i_sb, bindex, h_i->i_ino, /*ino*/0);
13036 + /* ignore this error */
13037 + /* bad action? */
13040 + /* children inode number will be broken */
13047 +static int hn_gen_tree(struct dentry *dentry)
13049 + int err, i, j, ndentry;
13050 + struct au_dcsub_pages dpages;
13051 + struct au_dpage *dpage;
13052 + struct dentry **dentries;
13054 + err = au_dpages_init(&dpages, GFP_NOFS);
13055 + if (unlikely(err))
13057 + err = au_dcsub_pages(&dpages, dentry, NULL, NULL);
13058 + if (unlikely(err))
13061 + for (i = 0; i < dpages.ndpage; i++) {
13062 + dpage = dpages.dpages + i;
13063 + dentries = dpage->dentries;
13064 + ndentry = dpage->ndentry;
13065 + for (j = 0; j < ndentry; j++) {
13066 + struct dentry *d;
13074 + /* todo: reset children xino?
13075 + cached children only? */
13076 + au_iigen_dec(d->d_inode);
13081 + au_dpages_free(&dpages);
13084 + /* discard children */
13085 + dentry_unhash(dentry);
13093 + * return 0 if processed.
13095 +static int hn_gen_by_inode(char *name, unsigned int nlen, struct inode *inode,
13096 + const unsigned int isdir)
13099 + struct dentry *d;
13100 + struct qstr *dname;
13103 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
13104 + pr_warning("branch root dir was changed\n");
13110 + AuDebugOn(!name);
13111 + au_iigen_dec(inode);
13112 + spin_lock(&inode->i_lock);
13113 + list_for_each_entry(d, &inode->i_dentry, d_alias) {
13114 + spin_lock(&d->d_lock);
13115 + dname = &d->d_name;
13116 + if (dname->len != nlen
13117 + && memcmp(dname->name, name, nlen)) {
13118 + spin_unlock(&d->d_lock);
13123 + spin_unlock(&d->d_lock);
13126 + spin_unlock(&inode->i_lock);
13128 + au_fset_si(au_sbi(inode->i_sb), FAILED_REFRESH_DIR);
13129 + d = d_find_alias(inode);
13131 + au_iigen_dec(inode);
13135 + spin_lock(&d->d_lock);
13136 + dname = &d->d_name;
13137 + if (dname->len == nlen && !memcmp(dname->name, name, nlen)) {
13138 + spin_unlock(&d->d_lock);
13139 + err = hn_gen_tree(d);
13140 + spin_lock(&d->d_lock);
13142 + spin_unlock(&d->d_lock);
13151 +static int hn_gen_by_name(struct dentry *dentry, const unsigned int isdir)
13154 + struct inode *inode;
13156 + inode = dentry->d_inode;
13157 + if (IS_ROOT(dentry)
13158 + /* || (inode && inode->i_ino == AUFS_ROOT_INO) */
13160 + pr_warning("branch root dir was changed\n");
13166 + au_digen_dec(dentry);
13168 + au_iigen_dec(inode);
13170 + au_fset_si(au_sbi(dentry->d_sb), FAILED_REFRESH_DIR);
13172 + err = hn_gen_tree(dentry);
13179 +/* ---------------------------------------------------------------------- */
13181 +/* hnotify job flags */
13182 +#define AuHnJob_XINO0 1
13183 +#define AuHnJob_GEN (1 << 1)
13184 +#define AuHnJob_DIRENT (1 << 2)
13185 +#define AuHnJob_ISDIR (1 << 3)
13186 +#define AuHnJob_TRYXINO0 (1 << 4)
13187 +#define AuHnJob_MNTPNT (1 << 5)
13188 +#define au_ftest_hnjob(flags, name) ((flags) & AuHnJob_##name)
13189 +#define au_fset_hnjob(flags, name) \
13190 + do { (flags) |= AuHnJob_##name; } while (0)
13191 +#define au_fclr_hnjob(flags, name) \
13192 + do { (flags) &= ~AuHnJob_##name; } while (0)
13200 +struct au_hnotify_args {
13201 + struct inode *h_dir, *dir, *h_child_inode;
13203 + unsigned int flags[AuHnLast];
13204 + unsigned int h_child_nlen;
13205 + char h_child_name[];
13208 +struct hn_job_args {
13209 + unsigned int flags;
13210 + struct inode *inode, *h_inode, *dir, *h_dir;
13211 + struct dentry *dentry;
13216 +static int hn_job(struct hn_job_args *a)
13218 + const unsigned int isdir = au_ftest_hnjob(a->flags, ISDIR);
13221 + if (au_ftest_hnjob(a->flags, XINO0) && a->inode)
13222 + hn_xino(a->inode, a->h_inode); /* ignore this error */
13224 + if (au_ftest_hnjob(a->flags, TRYXINO0)
13227 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
13228 + if (!a->h_inode->i_nlink)
13229 + hn_xino(a->inode, a->h_inode); /* ignore this error */
13230 + mutex_unlock(&a->h_inode->i_mutex);
13233 + /* make the generation obsolete */
13234 + if (au_ftest_hnjob(a->flags, GEN)) {
13237 + err = hn_gen_by_inode(a->h_name, a->h_nlen, a->inode,
13239 + if (err && a->dentry)
13240 + hn_gen_by_name(a->dentry, isdir);
13241 + /* ignore this error */
13244 + /* make dir entries obsolete */
13245 + if (au_ftest_hnjob(a->flags, DIRENT) && a->inode) {
13246 + struct au_vdir *vdir;
13248 + vdir = au_ivdir(a->inode);
13250 + vdir->vd_jiffy = 0;
13251 + /* IMustLock(a->inode); */
13252 + /* a->inode->i_version++; */
13255 + /* can do nothing but warn */
13256 + if (au_ftest_hnjob(a->flags, MNTPNT)
13258 + && d_mountpoint(a->dentry))
13259 + pr_warning("mount-point %.*s is removed or renamed\n",
13260 + AuDLNPair(a->dentry));
13265 +/* ---------------------------------------------------------------------- */
13267 +static struct dentry *lookup_wlock_by_name(char *name, unsigned int nlen,
13268 + struct inode *dir)
13270 + struct dentry *dentry, *d, *parent;
13271 + struct qstr *dname;
13273 + parent = d_find_alias(dir);
13278 + spin_lock(&parent->d_lock);
13279 + list_for_each_entry(d, &parent->d_subdirs, d_u.d_child) {
13280 + /* AuDbg("%.*s\n", AuDLNPair(d)); */
13281 + spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
13282 + dname = &d->d_name;
13283 + if (dname->len != nlen || memcmp(dname->name, name, nlen))
13284 + goto cont_unlock;
13288 + goto cont_unlock;
13289 + if (d->d_count) {
13290 + dentry = dget_dlock(d);
13291 + spin_unlock(&d->d_lock);
13296 + spin_unlock(&d->d_lock);
13298 + spin_unlock(&parent->d_lock);
13302 + di_write_lock_child(dentry);
13307 +static struct inode *lookup_wlock_by_ino(struct super_block *sb,
13308 + aufs_bindex_t bindex, ino_t h_ino)
13310 + struct inode *inode;
13315 + err = au_xino_read(sb, bindex, h_ino, &ino);
13317 + inode = ilookup(sb, ino);
13321 + if (unlikely(inode->i_ino == AUFS_ROOT_INO)) {
13322 + pr_warning("wrong root branch\n");
13328 + ii_write_lock_child(inode);
13334 +static void au_hn_bh(void *_args)
13336 + struct au_hnotify_args *a = _args;
13337 + struct super_block *sb;
13338 + aufs_bindex_t bindex, bend, bfound;
13339 + unsigned char xino, try_iput;
13341 + struct inode *inode;
13343 + struct hn_job_args args;
13344 + struct dentry *dentry;
13345 + struct au_sbinfo *sbinfo;
13347 + AuDebugOn(!_args);
13348 + AuDebugOn(!a->h_dir);
13349 + AuDebugOn(!a->dir);
13350 + AuDebugOn(!a->mask);
13351 + AuDbg("mask 0x%x, i%lu, hi%lu, hci%lu\n",
13352 + a->mask, a->dir->i_ino, a->h_dir->i_ino,
13353 + a->h_child_inode ? a->h_child_inode->i_ino : 0);
13358 + * do not lock a->dir->i_mutex here
13359 + * because of d_revalidate() may cause a deadlock.
13361 + sb = a->dir->i_sb;
13363 + sbinfo = au_sbi(sb);
13364 + AuDebugOn(!sbinfo);
13365 + si_write_lock(sb, AuLock_NOPLMW);
13367 + ii_read_lock_parent(a->dir);
13369 + bend = au_ibend(a->dir);
13370 + for (bindex = au_ibstart(a->dir); bindex <= bend; bindex++)
13371 + if (au_h_iptr(a->dir, bindex) == a->h_dir) {
13375 + ii_read_unlock(a->dir);
13376 + if (unlikely(bfound < 0))
13379 + xino = !!au_opt_test(au_mntflags(sb), XINO);
13381 + if (a->h_child_inode)
13382 + h_ino = a->h_child_inode->i_ino;
13384 + if (a->h_child_nlen
13385 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], GEN)
13386 + || au_ftest_hnjob(a->flags[AuHn_CHILD], MNTPNT)))
13387 + dentry = lookup_wlock_by_name(a->h_child_name, a->h_child_nlen,
13391 + inode = dentry->d_inode;
13392 + if (xino && !inode && h_ino
13393 + && (au_ftest_hnjob(a->flags[AuHn_CHILD], XINO0)
13394 + || au_ftest_hnjob(a->flags[AuHn_CHILD], TRYXINO0)
13395 + || au_ftest_hnjob(a->flags[AuHn_CHILD], GEN))) {
13396 + inode = lookup_wlock_by_ino(sb, bfound, h_ino);
13400 + args.flags = a->flags[AuHn_CHILD];
13401 + args.dentry = dentry;
13402 + args.inode = inode;
13403 + args.h_inode = a->h_child_inode;
13404 + args.dir = a->dir;
13405 + args.h_dir = a->h_dir;
13406 + args.h_name = a->h_child_name;
13407 + args.h_nlen = a->h_child_nlen;
13408 + err = hn_job(&args);
13410 + if (au_di(dentry))
13411 + di_write_unlock(dentry);
13414 + if (inode && try_iput) {
13415 + ii_write_unlock(inode);
13419 + ii_write_lock_parent(a->dir);
13420 + args.flags = a->flags[AuHn_PARENT];
13421 + args.dentry = NULL;
13422 + args.inode = a->dir;
13423 + args.h_inode = a->h_dir;
13425 + args.h_dir = NULL;
13426 + args.h_name = NULL;
13428 + err = hn_job(&args);
13429 + ii_write_unlock(a->dir);
13432 + iput(a->h_child_inode);
13435 + si_write_unlock(sb);
13436 + au_nwt_done(&sbinfo->si_nowait);
13440 +/* ---------------------------------------------------------------------- */
13442 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
13443 + struct qstr *h_child_qstr, struct inode *h_child_inode)
13446 + unsigned int flags[AuHnLast], f;
13447 + unsigned char isdir, isroot, wh;
13448 + struct inode *dir;
13449 + struct au_hnotify_args *args;
13450 + char *p, *h_child_name;
13453 + AuDebugOn(!hnotify || !hnotify->hn_aufs_inode);
13454 + dir = igrab(hnotify->hn_aufs_inode);
13458 + isroot = (dir->i_ino == AUFS_ROOT_INO);
13460 + h_child_name = (void *)h_child_qstr->name;
13461 + len = h_child_qstr->len;
13462 + if (h_child_name) {
13463 + if (len > AUFS_WH_PFX_LEN
13464 + && !memcmp(h_child_name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
13465 + h_child_name += AUFS_WH_PFX_LEN;
13466 + len -= AUFS_WH_PFX_LEN;
13472 + if (h_child_inode)
13473 + isdir = !!S_ISDIR(h_child_inode->i_mode);
13474 + flags[AuHn_PARENT] = AuHnJob_ISDIR;
13475 + flags[AuHn_CHILD] = 0;
13477 + flags[AuHn_CHILD] = AuHnJob_ISDIR;
13478 + au_fset_hnjob(flags[AuHn_PARENT], DIRENT);
13479 + au_fset_hnjob(flags[AuHn_CHILD], GEN);
13480 + switch (mask & FS_EVENTS_POSS_ON_CHILD) {
13481 + case FS_MOVED_FROM:
13482 + case FS_MOVED_TO:
13483 + au_fset_hnjob(flags[AuHn_CHILD], XINO0);
13484 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
13487 + AuDebugOn(!h_child_name || !h_child_inode);
13492 + * aufs never be able to get this child inode.
13493 + * revalidation should be in d_revalidate()
13494 + * by checking i_nlink, i_generation or d_unhashed().
13496 + AuDebugOn(!h_child_name);
13497 + au_fset_hnjob(flags[AuHn_CHILD], TRYXINO0);
13498 + au_fset_hnjob(flags[AuHn_CHILD], MNTPNT);
13506 + h_child_inode = NULL;
13509 + /* iput() and kfree() will be called in au_hnotify() */
13510 + args = kmalloc(sizeof(*args) + len + 1, GFP_NOFS);
13511 + if (unlikely(!args)) {
13512 + AuErr1("no memory\n");
13516 + args->flags[AuHn_PARENT] = flags[AuHn_PARENT];
13517 + args->flags[AuHn_CHILD] = flags[AuHn_CHILD];
13518 + args->mask = mask;
13520 + args->h_dir = igrab(h_dir);
13521 + if (h_child_inode)
13522 + h_child_inode = igrab(h_child_inode); /* can be NULL */
13523 + args->h_child_inode = h_child_inode;
13524 + args->h_child_nlen = len;
13526 + p = (void *)args;
13527 + p += sizeof(*args);
13528 + memcpy(p, h_child_name, len);
13533 + if (!dir->i_nlink)
13535 + err = au_wkq_nowait(au_hn_bh, args, dir->i_sb, f);
13536 + if (unlikely(err)) {
13537 + pr_err("wkq %d\n", err);
13538 + iput(args->h_child_inode);
13539 + iput(args->h_dir);
13548 +/* ---------------------------------------------------------------------- */
13550 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm)
13554 + AuDebugOn(!(udba & AuOptMask_UDBA));
13557 + if (au_hnotify_op.reset_br)
13558 + err = au_hnotify_op.reset_br(udba, br, perm);
13563 +int au_hnotify_init_br(struct au_branch *br, int perm)
13568 + if (au_hnotify_op.init_br)
13569 + err = au_hnotify_op.init_br(br, perm);
13574 +void au_hnotify_fin_br(struct au_branch *br)
13576 + if (au_hnotify_op.fin_br)
13577 + au_hnotify_op.fin_br(br);
13580 +static void au_hn_destroy_cache(void)
13582 + kmem_cache_destroy(au_cachep[AuCache_HNOTIFY]);
13583 + au_cachep[AuCache_HNOTIFY] = NULL;
13586 +int __init au_hnotify_init(void)
13591 + au_cachep[AuCache_HNOTIFY] = AuCache(au_hnotify);
13592 + if (au_cachep[AuCache_HNOTIFY]) {
13594 + if (au_hnotify_op.init)
13595 + err = au_hnotify_op.init();
13596 + if (unlikely(err))
13597 + au_hn_destroy_cache();
13603 +void au_hnotify_fin(void)
13605 + if (au_hnotify_op.fin)
13606 + au_hnotify_op.fin();
13607 + /* cf. au_cache_fin() */
13608 + if (au_cachep[AuCache_HNOTIFY])
13609 + au_hn_destroy_cache();
13611 diff -urN /usr/share/empty/fs/aufs/iinfo.c linux/fs/aufs/iinfo.c
13612 --- /usr/share/empty/fs/aufs/iinfo.c 1970-01-01 01:00:00.000000000 +0100
13613 +++ linux/fs/aufs/iinfo.c 2011-08-24 13:30:24.734646739 +0200
13616 + * Copyright (C) 2005-2011 Junjiro R. Okajima
13618 + * This program, aufs is free software; you can redistribute it and/or modify
13619 + * it under the terms of the GNU General Public License as published by
13620 + * the Free Software Foundation; either version 2 of the License, or
13621 + * (at your option) any later version.
13623 + * This program is distributed in the hope that it will be useful,
13624 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13625 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13626 + * GNU General Public License for more details.
13628 + * You should have received a copy of the GNU General Public License
13629 + * along with this program; if not, write to the Free Software
13630 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
13634 + * inode private data
13639 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex)
13641 + struct inode *h_inode;
13643 + IiMustAnyLock(inode);
13645 + h_inode = au_ii(inode)->ii_hinode[0 + bindex].hi_inode;
13646 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
13650 +/* todo: hard/soft set? */
13651 +void au_hiput(struct au_hinode *hinode)
13653 + au_hn_free(hinode);
13654 + dput(hinode->hi_whdentry);
13655 + iput(hinode->hi_inode);
13658 +unsigned int au_hi_flags(struct inode *inode, int isdir)
13660 + unsigned int flags;
13661 + const unsigned int mnt_flags = au_mntflags(inode->i_sb);
13664 + if (au_opt_test(mnt_flags, XINO))
13665 + au_fset_hi(flags, XINO);
13666 + if (isdir && au_opt_test(mnt_flags, UDBA_HNOTIFY))
13667 + au_fset_hi(flags, HNOTIFY);
13671 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
13672 + struct inode *h_inode, unsigned int flags)
13674 + struct au_hinode *hinode;
13675 + struct inode *hi;
13676 + struct au_iinfo *iinfo = au_ii(inode);
13678 + IiMustWriteLock(inode);
13680 + hinode = iinfo->ii_hinode + bindex;
13681 + hi = hinode->hi_inode;
13682 + AuDebugOn(h_inode && atomic_read(&h_inode->i_count) <= 0);
13685 + au_hiput(hinode);
13686 + hinode->hi_inode = h_inode;
13689 + struct super_block *sb = inode->i_sb;
13690 + struct au_branch *br;
13692 + AuDebugOn(inode->i_mode
13693 + && (h_inode->i_mode & S_IFMT)
13694 + != (inode->i_mode & S_IFMT));
13695 + if (bindex == iinfo->ii_bstart)
13696 + au_cpup_igen(inode, h_inode);
13697 + br = au_sbr(sb, bindex);
13698 + hinode->hi_id = br->br_id;
13699 + if (au_ftest_hi(flags, XINO)) {
13700 + err = au_xino_write(sb, bindex, h_inode->i_ino,
13702 + if (unlikely(err))
13703 + AuIOErr1("failed au_xino_write() %d\n", err);
13706 + if (au_ftest_hi(flags, HNOTIFY)
13707 + && au_br_hnotifyable(br->br_perm)) {
13708 + err = au_hn_alloc(hinode, inode);
13709 + if (unlikely(err))
13710 + AuIOErr1("au_hn_alloc() %d\n", err);
13715 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
13716 + struct dentry *h_wh)
13718 + struct au_hinode *hinode;
13720 + IiMustWriteLock(inode);
13722 + hinode = au_ii(inode)->ii_hinode + bindex;
13723 + AuDebugOn(hinode->hi_whdentry);
13724 + hinode->hi_whdentry = h_wh;
13727 +void au_update_iigen(struct inode *inode)
13729 + atomic_set(&au_ii(inode)->ii_generation, au_sigen(inode->i_sb));
13730 + /* smp_mb(); */ /* atomic_set */
13733 +/* it may be called at remount time, too */
13734 +void au_update_ibrange(struct inode *inode, int do_put_zero)
13736 + struct au_iinfo *iinfo;
13737 + aufs_bindex_t bindex, bend;
13739 + iinfo = au_ii(inode);
13743 + IiMustWriteLock(inode);
13745 + if (do_put_zero && iinfo->ii_bstart >= 0) {
13746 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
13748 + struct inode *h_i;
13750 + h_i = iinfo->ii_hinode[0 + bindex].hi_inode;
13751 + if (h_i && !h_i->i_nlink)
13752 + au_set_h_iptr(inode, bindex, NULL, 0);
13756 + iinfo->ii_bstart = -1;
13757 + iinfo->ii_bend = -1;
13758 + bend = au_sbend(inode->i_sb);
13759 + for (bindex = 0; bindex <= bend; bindex++)
13760 + if (iinfo->ii_hinode[0 + bindex].hi_inode) {
13761 + iinfo->ii_bstart = bindex;
13764 + if (iinfo->ii_bstart >= 0)
13765 + for (bindex = bend; bindex >= iinfo->ii_bstart; bindex--)
13766 + if (iinfo->ii_hinode[0 + bindex].hi_inode) {
13767 + iinfo->ii_bend = bindex;
13770 + AuDebugOn(iinfo->ii_bstart > iinfo->ii_bend);
13773 +/* ---------------------------------------------------------------------- */
13775 +void au_icntnr_init_once(void *_c)
13777 + struct au_icntnr *c = _c;
13778 + struct au_iinfo *iinfo = &c->iinfo;
13779 + static struct lock_class_key aufs_ii;
13781 + au_rw_init(&iinfo->ii_rwsem);
13782 + au_rw_class(&iinfo->ii_rwsem, &aufs_ii);
13783 + inode_init_once(&c->vfs_inode);
13786 +int au_iinfo_init(struct inode *inode)
13788 + struct au_iinfo *iinfo;
13789 + struct super_block *sb;
13792 + sb = inode->i_sb;
13793 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
13794 + nbr = au_sbend(sb) + 1;
13795 + if (unlikely(nbr <= 0))
13797 + iinfo->ii_hinode = kcalloc(nbr, sizeof(*iinfo->ii_hinode), GFP_NOFS);
13798 + if (iinfo->ii_hinode) {
13799 + au_ninodes_inc(sb);
13800 + for (i = 0; i < nbr; i++)
13801 + iinfo->ii_hinode[i].hi_id = -1;
13803 + atomic_set(&iinfo->ii_generation, au_sigen(sb));
13804 + /* smp_mb(); */ /* atomic_set */
13805 + iinfo->ii_bstart = -1;
13806 + iinfo->ii_bend = -1;
13807 + iinfo->ii_vdir = NULL;
13813 +int au_ii_realloc(struct au_iinfo *iinfo, int nbr)
13816 + struct au_hinode *hip;
13818 + AuRwMustWriteLock(&iinfo->ii_rwsem);
13821 + sz = sizeof(*hip) * (iinfo->ii_bend + 1);
13823 + sz = sizeof(*hip);
13824 + hip = au_kzrealloc(iinfo->ii_hinode, sz, sizeof(*hip) * nbr, GFP_NOFS);
13826 + iinfo->ii_hinode = hip;
13833 +void au_iinfo_fin(struct inode *inode)
13835 + struct au_iinfo *iinfo;
13836 + struct au_hinode *hi;
13837 + struct super_block *sb;
13838 + aufs_bindex_t bindex, bend;
13839 + const unsigned char unlinked = !inode->i_nlink;
13841 + iinfo = au_ii(inode);
13842 + /* bad_inode case */
13846 + sb = inode->i_sb;
13847 + au_ninodes_dec(sb);
13848 + if (si_pid_test(sb))
13849 + au_xino_delete_inode(inode, unlinked);
13852 + * it is safe to hide the dependency between sbinfo and
13856 + si_noflush_read_lock(sb);
13857 + au_xino_delete_inode(inode, unlinked);
13858 + si_read_unlock(sb);
13862 + if (iinfo->ii_vdir)
13863 + au_vdir_free(iinfo->ii_vdir);
13865 + bindex = iinfo->ii_bstart;
13866 + if (bindex >= 0) {
13867 + hi = iinfo->ii_hinode + bindex;
13868 + bend = iinfo->ii_bend;
13869 + while (bindex++ <= bend) {
13870 + if (hi->hi_inode)
13875 + kfree(iinfo->ii_hinode);
13876 + iinfo->ii_hinode = NULL;
13877 + AuRwDestroy(&iinfo->ii_rwsem);
13879 diff -urN /usr/share/empty/fs/aufs/inode.c linux/fs/aufs/inode.c
13880 --- /usr/share/empty/fs/aufs/inode.c 1970-01-01 01:00:00.000000000 +0100
13881 +++ linux/fs/aufs/inode.c 2011-08-24 13:30:24.734646739 +0200
13884 + * Copyright (C) 2005-2011 Junjiro R. Okajima
13886 + * This program, aufs is free software; you can redistribute it and/or modify
13887 + * it under the terms of the GNU General Public License as published by
13888 + * the Free Software Foundation; either version 2 of the License, or
13889 + * (at your option) any later version.
13891 + * This program is distributed in the hope that it will be useful,
13892 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
13893 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13894 + * GNU General Public License for more details.
13896 + * You should have received a copy of the GNU General Public License
13897 + * along with this program; if not, write to the Free Software
13898 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
13902 + * inode functions
13907 +struct inode *au_igrab(struct inode *inode)
13910 + AuDebugOn(!atomic_read(&inode->i_count));
13916 +static void au_refresh_hinode_attr(struct inode *inode, int do_version)
13918 + au_cpup_attr_all(inode, /*force*/0);
13919 + au_update_iigen(inode);
13921 + inode->i_version++;
13924 +static int au_ii_refresh(struct inode *inode, int *update)
13928 + aufs_bindex_t bindex, new_bindex;
13929 + struct super_block *sb;
13930 + struct au_iinfo *iinfo;
13931 + struct au_hinode *p, *q, tmp;
13933 + IiMustWriteLock(inode);
13936 + sb = inode->i_sb;
13937 + type = inode->i_mode & S_IFMT;
13938 + iinfo = au_ii(inode);
13939 + err = au_ii_realloc(iinfo, au_sbend(sb) + 1);
13940 + if (unlikely(err))
13943 + AuDebugOn(iinfo->ii_bstart < 0);
13944 + p = iinfo->ii_hinode + iinfo->ii_bstart;
13945 + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend;
13947 + if (!p->hi_inode)
13950 + AuDebugOn(type != (p->hi_inode->i_mode & S_IFMT));
13951 + new_bindex = au_br_index(sb, p->hi_id);
13952 + if (new_bindex == bindex)
13955 + if (new_bindex < 0) {
13958 + p->hi_inode = NULL;
13962 + if (new_bindex < iinfo->ii_bstart)
13963 + iinfo->ii_bstart = new_bindex;
13964 + if (iinfo->ii_bend < new_bindex)
13965 + iinfo->ii_bend = new_bindex;
13966 + /* swap two lower inode, and loop again */
13967 + q = iinfo->ii_hinode + new_bindex;
13971 + if (tmp.hi_inode) {
13976 + au_update_ibrange(inode, /*do_put_zero*/0);
13977 + e = au_dy_irefresh(inode);
13978 + if (unlikely(e && !err))
13986 +int au_refresh_hinode_self(struct inode *inode)
13990 + err = au_ii_refresh(inode, &update);
13992 + au_refresh_hinode_attr(inode, update && S_ISDIR(inode->i_mode));
13998 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry)
14000 + int err, e, update;
14001 + unsigned int flags;
14003 + aufs_bindex_t bindex, bend;
14004 + unsigned char isdir;
14005 + struct au_hinode *p;
14006 + struct au_iinfo *iinfo;
14008 + err = au_ii_refresh(inode, &update);
14009 + if (unlikely(err))
14013 + iinfo = au_ii(inode);
14014 + p = iinfo->ii_hinode + iinfo->ii_bstart;
14015 + mode = (inode->i_mode & S_IFMT);
14016 + isdir = S_ISDIR(mode);
14017 + flags = au_hi_flags(inode, isdir);
14018 + bend = au_dbend(dentry);
14019 + for (bindex = au_dbstart(dentry); bindex <= bend; bindex++) {
14020 + struct inode *h_i;
14021 + struct dentry *h_d;
14023 + h_d = au_h_dptr(dentry, bindex);
14024 + if (!h_d || !h_d->d_inode)
14027 + AuDebugOn(mode != (h_d->d_inode->i_mode & S_IFMT));
14028 + if (iinfo->ii_bstart <= bindex && bindex <= iinfo->ii_bend) {
14029 + h_i = au_h_iptr(inode, bindex);
14031 + if (h_i == h_d->d_inode)
14037 + if (bindex < iinfo->ii_bstart)
14038 + iinfo->ii_bstart = bindex;
14039 + if (iinfo->ii_bend < bindex)
14040 + iinfo->ii_bend = bindex;
14041 + au_set_h_iptr(inode, bindex, au_igrab(h_d->d_inode), flags);
14044 + au_update_ibrange(inode, /*do_put_zero*/0);
14045 + e = au_dy_irefresh(inode);
14046 + if (unlikely(e && !err))
14049 + au_refresh_hinode_attr(inode, update && isdir);
14056 +static int set_inode(struct inode *inode, struct dentry *dentry)
14059 + unsigned int flags;
14061 + aufs_bindex_t bindex, bstart, btail;
14062 + unsigned char isdir;
14063 + struct dentry *h_dentry;
14064 + struct inode *h_inode;
14065 + struct au_iinfo *iinfo;
14067 + IiMustWriteLock(inode);
14071 + bstart = au_dbstart(dentry);
14072 + h_inode = au_h_dptr(dentry, bstart)->d_inode;
14073 + mode = h_inode->i_mode;
14074 + switch (mode & S_IFMT) {
14076 + btail = au_dbtail(dentry);
14077 + inode->i_op = &aufs_iop;
14078 + inode->i_fop = &aufs_file_fop;
14079 + err = au_dy_iaop(inode, bstart, h_inode);
14080 + if (unlikely(err))
14085 + btail = au_dbtaildir(dentry);
14086 + inode->i_op = &aufs_dir_iop;
14087 + inode->i_fop = &aufs_dir_fop;
14090 + btail = au_dbtail(dentry);
14091 + inode->i_op = &aufs_symlink_iop;
14097 + btail = au_dbtail(dentry);
14098 + inode->i_op = &aufs_iop;
14099 + au_init_special_fop(inode, mode, h_inode->i_rdev);
14102 + AuIOErr("Unknown file type 0%o\n", mode);
14107 + /* do not set hnotify for whiteouted dirs (SHWH mode) */
14108 + flags = au_hi_flags(inode, isdir);
14109 + if (au_opt_test(au_mntflags(dentry->d_sb), SHWH)
14110 + && au_ftest_hi(flags, HNOTIFY)
14111 + && dentry->d_name.len > AUFS_WH_PFX_LEN
14112 + && !memcmp(dentry->d_name.name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))
14113 + au_fclr_hi(flags, HNOTIFY);
14114 + iinfo = au_ii(inode);
14115 + iinfo->ii_bstart = bstart;
14116 + iinfo->ii_bend = btail;
14117 + for (bindex = bstart; bindex <= btail; bindex++) {
14118 + h_dentry = au_h_dptr(dentry, bindex);
14120 + au_set_h_iptr(inode, bindex,
14121 + au_igrab(h_dentry->d_inode), flags);
14123 + au_cpup_attr_all(inode, /*force*/1);
14130 + * successful returns with iinfo write_locked
14132 + * zero: success, matched
14133 + * plus: no error, but unmatched
14135 +static int reval_inode(struct inode *inode, struct dentry *dentry)
14138 + aufs_bindex_t bindex, bend;
14139 + struct inode *h_inode, *h_dinode;
14142 + * before this function, if aufs got any iinfo lock, it must be only
14143 + * one, the parent dir.
14144 + * it can happen by UDBA and the obsoleted inode number.
14147 + if (unlikely(inode->i_ino == parent_ino(dentry)))
14151 + ii_write_lock_new_child(inode);
14152 + h_dinode = au_h_dptr(dentry, au_dbstart(dentry))->d_inode;
14153 + bend = au_ibend(inode);
14154 + for (bindex = au_ibstart(inode); bindex <= bend; bindex++) {
14155 + h_inode = au_h_iptr(inode, bindex);
14156 + if (h_inode && h_inode == h_dinode) {
14158 + if (au_iigen_test(inode, au_digen(dentry)))
14159 + err = au_refresh_hinode(inode, dentry);
14164 + if (unlikely(err))
14165 + ii_write_unlock(inode);
14170 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
14171 + unsigned int d_type, ino_t *ino)
14174 + struct mutex *mtx;
14176 + /* prevent hardlinked inode number from race condition */
14178 + if (d_type != DT_DIR) {
14179 + mtx = &au_sbr(sb, bindex)->br_xino.xi_nondir_mtx;
14182 + err = au_xino_read(sb, bindex, h_ino, ino);
14183 + if (unlikely(err))
14188 + *ino = au_xino_new_ino(sb);
14189 + if (unlikely(!*ino))
14191 + err = au_xino_write(sb, bindex, h_ino, *ino);
14192 + if (unlikely(err))
14198 + mutex_unlock(mtx);
14202 +/* successful returns with iinfo write_locked */
14203 +/* todo: return with unlocked? */
14204 +struct inode *au_new_inode(struct dentry *dentry, int must_new)
14206 + struct inode *inode, *h_inode;
14207 + struct dentry *h_dentry;
14208 + struct super_block *sb;
14209 + struct mutex *mtx;
14210 + ino_t h_ino, ino;
14212 + aufs_bindex_t bstart;
14214 + sb = dentry->d_sb;
14215 + bstart = au_dbstart(dentry);
14216 + h_dentry = au_h_dptr(dentry, bstart);
14217 + h_inode = h_dentry->d_inode;
14218 + h_ino = h_inode->i_ino;
14221 + * stop 'race'-ing between hardlinks under different
14225 + if (!S_ISDIR(h_inode->i_mode))
14226 + mtx = &au_sbr(sb, bstart)->br_xino.xi_nondir_mtx;
14231 + err = au_xino_read(sb, bstart, h_ino, &ino);
14232 + inode = ERR_PTR(err);
14233 + if (unlikely(err))
14237 + ino = au_xino_new_ino(sb);
14238 + if (unlikely(!ino)) {
14239 + inode = ERR_PTR(-EIO);
14244 + AuDbg("i%lu\n", (unsigned long)ino);
14245 + inode = au_iget_locked(sb, ino);
14246 + err = PTR_ERR(inode);
14247 + if (IS_ERR(inode))
14250 + AuDbg("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW));
14251 + if (inode->i_state & I_NEW) {
14252 + ii_write_lock_new_child(inode);
14253 + err = set_inode(inode, dentry);
14255 + unlock_new_inode(inode);
14256 + goto out; /* success */
14260 + * iget_failed() calls iput(), but we need to call
14261 + * ii_write_unlock() after iget_failed(). so dirty hack for
14264 + atomic_inc(&inode->i_count);
14265 + iget_failed(inode);
14266 + ii_write_unlock(inode);
14267 + au_xino_write(sb, bstart, h_ino, /*ino*/0);
14268 + /* ignore this error */
14270 + } else if (!must_new && !IS_DEADDIR(inode) && inode->i_nlink) {
14272 + * horrible race condition between lookup, readdir and copyup
14273 + * (or something).
14276 + mutex_unlock(mtx);
14277 + err = reval_inode(inode, dentry);
14278 + if (unlikely(err < 0)) {
14285 + goto out; /* success */
14290 + if (unlikely(au_test_fs_unique_ino(h_dentry->d_inode)))
14291 + AuWarn1("Warning: Un-notified UDBA or repeatedly renamed dir,"
14292 + " b%d, %s, %.*s, hi%lu, i%lu.\n",
14293 + bstart, au_sbtype(h_dentry->d_sb), AuDLNPair(dentry),
14294 + (unsigned long)h_ino, (unsigned long)ino);
14296 + err = au_xino_write(sb, bstart, h_ino, /*ino*/0);
14300 + mutex_unlock(mtx);
14306 + inode = ERR_PTR(err);
14309 + mutex_unlock(mtx);
14313 +/* ---------------------------------------------------------------------- */
14315 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
14316 + struct inode *inode)
14320 + err = au_br_rdonly(au_sbr(sb, bindex));
14322 + /* pseudo-link after flushed may happen out of bounds */
14325 + && au_ibstart(inode) <= bindex
14326 + && bindex <= au_ibend(inode)) {
14328 + * permission check is unnecessary since vfsub routine
14329 + * will be called later
14331 + struct inode *hi = au_h_iptr(inode, bindex);
14333 + err = IS_IMMUTABLE(hi) ? -EROFS : 0;
14339 +int au_test_h_perm(struct inode *h_inode, int mask)
14341 + if (!current_fsuid())
14343 + return inode_permission(h_inode, mask);
14346 +int au_test_h_perm_sio(struct inode *h_inode, int mask)
14348 + if (au_test_nfs(h_inode->i_sb)
14349 + && (mask & MAY_WRITE)
14350 + && S_ISDIR(h_inode->i_mode))
14351 + mask |= MAY_READ; /* force permission check */
14352 + return au_test_h_perm(h_inode, mask);
14354 diff -urN /usr/share/empty/fs/aufs/inode.h linux/fs/aufs/inode.h
14355 --- /usr/share/empty/fs/aufs/inode.h 1970-01-01 01:00:00.000000000 +0100
14356 +++ linux/fs/aufs/inode.h 2011-08-24 13:30:24.734646739 +0200
14359 + * Copyright (C) 2005-2011 Junjiro R. Okajima
14361 + * This program, aufs is free software; you can redistribute it and/or modify
14362 + * it under the terms of the GNU General Public License as published by
14363 + * the Free Software Foundation; either version 2 of the License, or
14364 + * (at your option) any later version.
14366 + * This program is distributed in the hope that it will be useful,
14367 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14368 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14369 + * GNU General Public License for more details.
14371 + * You should have received a copy of the GNU General Public License
14372 + * along with this program; if not, write to the Free Software
14373 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
14377 + * inode operations
14380 +#ifndef __AUFS_INODE_H__
14381 +#define __AUFS_INODE_H__
14385 +#include <linux/fs.h>
14386 +#include <linux/fsnotify.h>
14387 +#include <linux/aufs_type.h>
14388 +#include "rwsem.h"
14392 +struct au_hnotify {
14393 +#ifdef CONFIG_AUFS_HNOTIFY
14394 +#ifdef CONFIG_AUFS_HFSNOTIFY
14395 + /* never use fsnotify_add_vfsmount_mark() */
14396 + struct fsnotify_mark hn_mark;
14397 + int hn_mark_dead;
14399 + struct inode *hn_aufs_inode; /* no get/put */
14401 +} ____cacheline_aligned_in_smp;
14403 +struct au_hinode {
14404 + struct inode *hi_inode;
14405 + aufs_bindex_t hi_id;
14406 +#ifdef CONFIG_AUFS_HNOTIFY
14407 + struct au_hnotify *hi_notify;
14410 + /* reference to the copied-up whiteout with get/put */
14411 + struct dentry *hi_whdentry;
14416 + atomic_t ii_generation;
14417 + struct super_block *ii_hsb1; /* no get/put */
14419 + struct au_rwsem ii_rwsem;
14420 + aufs_bindex_t ii_bstart, ii_bend;
14422 + struct au_hinode *ii_hinode;
14423 + struct au_vdir *ii_vdir;
14426 +struct au_icntnr {
14427 + struct au_iinfo iinfo;
14428 + struct inode vfs_inode;
14429 +} ____cacheline_aligned_in_smp;
14431 +/* au_pin flags */
14432 +#define AuPin_DI_LOCKED 1
14433 +#define AuPin_MNT_WRITE (1 << 1)
14434 +#define au_ftest_pin(flags, name) ((flags) & AuPin_##name)
14435 +#define au_fset_pin(flags, name) \
14436 + do { (flags) |= AuPin_##name; } while (0)
14437 +#define au_fclr_pin(flags, name) \
14438 + do { (flags) &= ~AuPin_##name; } while (0)
14442 + struct dentry *dentry;
14443 + unsigned int udba;
14444 + unsigned char lsc_di, lsc_hi, flags;
14445 + aufs_bindex_t bindex;
14448 + struct dentry *parent;
14449 + struct au_hinode *hdir;
14450 + struct vfsmount *h_mnt;
14453 +/* ---------------------------------------------------------------------- */
14455 +static inline struct au_iinfo *au_ii(struct inode *inode)
14457 + struct au_iinfo *iinfo;
14459 + iinfo = &(container_of(inode, struct au_icntnr, vfs_inode)->iinfo);
14460 + if (iinfo->ii_hinode)
14462 + return NULL; /* debugging bad_inode case */
14465 +/* ---------------------------------------------------------------------- */
14468 +struct inode *au_igrab(struct inode *inode);
14469 +int au_refresh_hinode_self(struct inode *inode);
14470 +int au_refresh_hinode(struct inode *inode, struct dentry *dentry);
14471 +int au_ino(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
14472 + unsigned int d_type, ino_t *ino);
14473 +struct inode *au_new_inode(struct dentry *dentry, int must_new);
14474 +int au_test_ro(struct super_block *sb, aufs_bindex_t bindex,
14475 + struct inode *inode);
14476 +int au_test_h_perm(struct inode *h_inode, int mask);
14477 +int au_test_h_perm_sio(struct inode *h_inode, int mask);
14479 +static inline int au_wh_ino(struct super_block *sb, aufs_bindex_t bindex,
14480 + ino_t h_ino, unsigned int d_type, ino_t *ino)
14482 +#ifdef CONFIG_AUFS_SHWH
14483 + return au_ino(sb, bindex, h_ino, d_type, ino);
14490 +extern struct inode_operations aufs_iop, aufs_symlink_iop, aufs_dir_iop;
14492 +/* au_wr_dir flags */
14493 +#define AuWrDir_ADD_ENTRY 1
14494 +#define AuWrDir_ISDIR (1 << 1)
14495 +#define au_ftest_wrdir(flags, name) ((flags) & AuWrDir_##name)
14496 +#define au_fset_wrdir(flags, name) \
14497 + do { (flags) |= AuWrDir_##name; } while (0)
14498 +#define au_fclr_wrdir(flags, name) \
14499 + do { (flags) &= ~AuWrDir_##name; } while (0)
14501 +struct au_wr_dir_args {
14502 + aufs_bindex_t force_btgt;
14503 + unsigned char flags;
14505 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
14506 + struct au_wr_dir_args *args);
14508 +struct dentry *au_pinned_h_parent(struct au_pin *pin);
14509 +void au_pin_init(struct au_pin *pin, struct dentry *dentry,
14510 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
14511 + unsigned int udba, unsigned char flags);
14512 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
14513 + unsigned int udba, unsigned char flags) __must_check;
14514 +int au_do_pin(struct au_pin *pin) __must_check;
14515 +void au_unpin(struct au_pin *pin);
14518 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
14519 + struct dentry *h_parent, int isdir);
14520 +int aufs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev);
14521 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname);
14522 +int aufs_create(struct inode *dir, struct dentry *dentry, int mode,
14523 + struct nameidata *nd);
14524 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
14525 + struct dentry *dentry);
14526 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, int mode);
14529 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup);
14530 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
14531 + struct dentry *h_parent, int isdir);
14532 +int aufs_unlink(struct inode *dir, struct dentry *dentry);
14533 +int aufs_rmdir(struct inode *dir, struct dentry *dentry);
14536 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt);
14537 +int aufs_rename(struct inode *src_dir, struct dentry *src_dentry,
14538 + struct inode *dir, struct dentry *dentry);
14541 +struct inode *au_h_iptr(struct inode *inode, aufs_bindex_t bindex);
14542 +void au_hiput(struct au_hinode *hinode);
14543 +void au_set_hi_wh(struct inode *inode, aufs_bindex_t bindex,
14544 + struct dentry *h_wh);
14545 +unsigned int au_hi_flags(struct inode *inode, int isdir);
14547 +/* hinode flags */
14548 +#define AuHi_XINO 1
14549 +#define AuHi_HNOTIFY (1 << 1)
14550 +#define au_ftest_hi(flags, name) ((flags) & AuHi_##name)
14551 +#define au_fset_hi(flags, name) \
14552 + do { (flags) |= AuHi_##name; } while (0)
14553 +#define au_fclr_hi(flags, name) \
14554 + do { (flags) &= ~AuHi_##name; } while (0)
14556 +#ifndef CONFIG_AUFS_HNOTIFY
14557 +#undef AuHi_HNOTIFY
14558 +#define AuHi_HNOTIFY 0
14561 +void au_set_h_iptr(struct inode *inode, aufs_bindex_t bindex,
14562 + struct inode *h_inode, unsigned int flags);
14564 +void au_update_iigen(struct inode *inode);
14565 +void au_update_ibrange(struct inode *inode, int do_put_zero);
14567 +void au_icntnr_init_once(void *_c);
14568 +int au_iinfo_init(struct inode *inode);
14569 +void au_iinfo_fin(struct inode *inode);
14570 +int au_ii_realloc(struct au_iinfo *iinfo, int nbr);
14572 +#ifdef CONFIG_PROC_FS
14574 +int au_plink_maint(struct super_block *sb, int flags);
14575 +void au_plink_maint_leave(struct au_sbinfo *sbinfo);
14576 +int au_plink_maint_enter(struct super_block *sb);
14577 +#ifdef CONFIG_AUFS_DEBUG
14578 +void au_plink_list(struct super_block *sb);
14580 +AuStubVoid(au_plink_list, struct super_block *sb)
14582 +int au_plink_test(struct inode *inode);
14583 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex);
14584 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
14585 + struct dentry *h_dentry);
14586 +void au_plink_put(struct super_block *sb, int verbose);
14587 +void au_plink_clean(struct super_block *sb, int verbose);
14588 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id);
14590 +AuStubInt0(au_plink_maint, struct super_block *sb, int flags);
14591 +AuStubVoid(au_plink_maint_leave, struct au_sbinfo *sbinfo);
14592 +AuStubInt0(au_plink_maint_enter, struct super_block *sb);
14593 +AuStubVoid(au_plink_list, struct super_block *sb);
14594 +AuStubInt0(au_plink_test, struct inode *inode);
14595 +AuStub(struct dentry *, au_plink_lkup, return NULL,
14596 + struct inode *inode, aufs_bindex_t bindex);
14597 +AuStubVoid(au_plink_append, struct inode *inode, aufs_bindex_t bindex,
14598 + struct dentry *h_dentry);
14599 +AuStubVoid(au_plink_put, struct super_block *sb, int verbose);
14600 +AuStubVoid(au_plink_clean, struct super_block *sb, int verbose);
14601 +AuStubVoid(au_plink_half_refresh, struct super_block *sb, aufs_bindex_t br_id);
14602 +#endif /* CONFIG_PROC_FS */
14604 +/* ---------------------------------------------------------------------- */
14606 +/* lock subclass for iinfo */
14608 + AuLsc_II_CHILD, /* child first */
14609 + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hnotify */
14610 + AuLsc_II_CHILD3, /* copyup dirs */
14611 + AuLsc_II_PARENT, /* see AuLsc_I_PARENT in vfsub.h */
14612 + AuLsc_II_PARENT2,
14613 + AuLsc_II_PARENT3, /* copyup dirs */
14614 + AuLsc_II_NEW_CHILD
14618 + * ii_read_lock_child, ii_write_lock_child,
14619 + * ii_read_lock_child2, ii_write_lock_child2,
14620 + * ii_read_lock_child3, ii_write_lock_child3,
14621 + * ii_read_lock_parent, ii_write_lock_parent,
14622 + * ii_read_lock_parent2, ii_write_lock_parent2,
14623 + * ii_read_lock_parent3, ii_write_lock_parent3,
14624 + * ii_read_lock_new_child, ii_write_lock_new_child,
14626 +#define AuReadLockFunc(name, lsc) \
14627 +static inline void ii_read_lock_##name(struct inode *i) \
14629 + au_rw_read_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
14632 +#define AuWriteLockFunc(name, lsc) \
14633 +static inline void ii_write_lock_##name(struct inode *i) \
14635 + au_rw_write_lock_nested(&au_ii(i)->ii_rwsem, AuLsc_II_##lsc); \
14638 +#define AuRWLockFuncs(name, lsc) \
14639 + AuReadLockFunc(name, lsc) \
14640 + AuWriteLockFunc(name, lsc)
14642 +AuRWLockFuncs(child, CHILD);
14643 +AuRWLockFuncs(child2, CHILD2);
14644 +AuRWLockFuncs(child3, CHILD3);
14645 +AuRWLockFuncs(parent, PARENT);
14646 +AuRWLockFuncs(parent2, PARENT2);
14647 +AuRWLockFuncs(parent3, PARENT3);
14648 +AuRWLockFuncs(new_child, NEW_CHILD);
14650 +#undef AuReadLockFunc
14651 +#undef AuWriteLockFunc
14652 +#undef AuRWLockFuncs
14655 + * ii_read_unlock, ii_write_unlock, ii_downgrade_lock
14657 +AuSimpleUnlockRwsemFuncs(ii, struct inode *i, &au_ii(i)->ii_rwsem);
14659 +#define IiMustNoWaiters(i) AuRwMustNoWaiters(&au_ii(i)->ii_rwsem)
14660 +#define IiMustAnyLock(i) AuRwMustAnyLock(&au_ii(i)->ii_rwsem)
14661 +#define IiMustWriteLock(i) AuRwMustWriteLock(&au_ii(i)->ii_rwsem)
14663 +/* ---------------------------------------------------------------------- */
14665 +static inline void au_icntnr_init(struct au_icntnr *c)
14667 +#ifdef CONFIG_AUFS_DEBUG
14668 + c->vfs_inode.i_mode = 0;
14672 +static inline unsigned int au_iigen(struct inode *inode)
14674 + return atomic_read(&au_ii(inode)->ii_generation);
14677 +/* tiny test for inode number */
14678 +/* tmpfs generation is too rough */
14679 +static inline int au_test_higen(struct inode *inode, struct inode *h_inode)
14681 + struct au_iinfo *iinfo;
14683 + iinfo = au_ii(inode);
14684 + AuRwMustAnyLock(&iinfo->ii_rwsem);
14685 + return !(iinfo->ii_hsb1 == h_inode->i_sb
14686 + && iinfo->ii_higen == h_inode->i_generation);
14689 +static inline void au_iigen_dec(struct inode *inode)
14691 + atomic_dec(&au_ii(inode)->ii_generation);
14694 +static inline int au_iigen_test(struct inode *inode, unsigned int sigen)
14699 + if (unlikely(inode && au_iigen(inode) != sigen))
14705 +/* ---------------------------------------------------------------------- */
14707 +static inline aufs_bindex_t au_ii_br_id(struct inode *inode,
14708 + aufs_bindex_t bindex)
14710 + IiMustAnyLock(inode);
14711 + return au_ii(inode)->ii_hinode[0 + bindex].hi_id;
14714 +static inline aufs_bindex_t au_ibstart(struct inode *inode)
14716 + IiMustAnyLock(inode);
14717 + return au_ii(inode)->ii_bstart;
14720 +static inline aufs_bindex_t au_ibend(struct inode *inode)
14722 + IiMustAnyLock(inode);
14723 + return au_ii(inode)->ii_bend;
14726 +static inline struct au_vdir *au_ivdir(struct inode *inode)
14728 + IiMustAnyLock(inode);
14729 + return au_ii(inode)->ii_vdir;
14732 +static inline struct dentry *au_hi_wh(struct inode *inode, aufs_bindex_t bindex)
14734 + IiMustAnyLock(inode);
14735 + return au_ii(inode)->ii_hinode[0 + bindex].hi_whdentry;
14738 +static inline void au_set_ibstart(struct inode *inode, aufs_bindex_t bindex)
14740 + IiMustWriteLock(inode);
14741 + au_ii(inode)->ii_bstart = bindex;
14744 +static inline void au_set_ibend(struct inode *inode, aufs_bindex_t bindex)
14746 + IiMustWriteLock(inode);
14747 + au_ii(inode)->ii_bend = bindex;
14750 +static inline void au_set_ivdir(struct inode *inode, struct au_vdir *vdir)
14752 + IiMustWriteLock(inode);
14753 + au_ii(inode)->ii_vdir = vdir;
14756 +static inline struct au_hinode *au_hi(struct inode *inode, aufs_bindex_t bindex)
14758 + IiMustAnyLock(inode);
14759 + return au_ii(inode)->ii_hinode + bindex;
14762 +/* ---------------------------------------------------------------------- */
14764 +static inline struct dentry *au_pinned_parent(struct au_pin *pin)
14767 + return pin->parent;
14771 +static inline struct inode *au_pinned_h_dir(struct au_pin *pin)
14773 + if (pin && pin->hdir)
14774 + return pin->hdir->hi_inode;
14778 +static inline struct au_hinode *au_pinned_hdir(struct au_pin *pin)
14781 + return pin->hdir;
14785 +static inline void au_pin_set_dentry(struct au_pin *pin, struct dentry *dentry)
14788 + pin->dentry = dentry;
14791 +static inline void au_pin_set_parent_lflag(struct au_pin *pin,
14792 + unsigned char lflag)
14796 + au_fset_pin(pin->flags, DI_LOCKED);
14798 + au_fclr_pin(pin->flags, DI_LOCKED);
14802 +static inline void au_pin_set_parent(struct au_pin *pin, struct dentry *parent)
14805 + dput(pin->parent);
14806 + pin->parent = dget(parent);
14810 +/* ---------------------------------------------------------------------- */
14813 +#ifdef CONFIG_AUFS_HNOTIFY
14814 +struct au_hnotify_op {
14815 + void (*ctl)(struct au_hinode *hinode, int do_set);
14816 + int (*alloc)(struct au_hinode *hinode);
14817 + void (*free)(struct au_hinode *hinode);
14819 + void (*fin)(void);
14820 + int (*init)(void);
14822 + int (*reset_br)(unsigned int udba, struct au_branch *br, int perm);
14823 + void (*fin_br)(struct au_branch *br);
14824 + int (*init_br)(struct au_branch *br, int perm);
14828 +int au_hn_alloc(struct au_hinode *hinode, struct inode *inode);
14829 +void au_hn_free(struct au_hinode *hinode);
14830 +void au_hn_ctl(struct au_hinode *hinode, int do_set);
14831 +void au_hn_reset(struct inode *inode, unsigned int flags);
14832 +int au_hnotify(struct inode *h_dir, struct au_hnotify *hnotify, u32 mask,
14833 + struct qstr *h_child_qstr, struct inode *h_child_inode);
14834 +int au_hnotify_reset_br(unsigned int udba, struct au_branch *br, int perm);
14835 +int au_hnotify_init_br(struct au_branch *br, int perm);
14836 +void au_hnotify_fin_br(struct au_branch *br);
14837 +int __init au_hnotify_init(void);
14838 +void au_hnotify_fin(void);
14841 +extern const struct au_hnotify_op au_hnotify_op;
14844 +void au_hn_init(struct au_hinode *hinode)
14846 + hinode->hi_notify = NULL;
14849 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
14851 + return hinode->hi_notify;
14856 +int au_hn_alloc(struct au_hinode *hinode __maybe_unused,
14857 + struct inode *inode __maybe_unused)
14859 + return -EOPNOTSUPP;
14862 +static inline struct au_hnotify *au_hn(struct au_hinode *hinode)
14867 +AuStubVoid(au_hn_free, struct au_hinode *hinode __maybe_unused)
14868 +AuStubVoid(au_hn_ctl, struct au_hinode *hinode __maybe_unused,
14869 + int do_set __maybe_unused)
14870 +AuStubVoid(au_hn_reset, struct inode *inode __maybe_unused,
14871 + unsigned int flags __maybe_unused)
14872 +AuStubInt0(au_hnotify_reset_br, unsigned int udba __maybe_unused,
14873 + struct au_branch *br __maybe_unused,
14874 + int perm __maybe_unused)
14875 +AuStubInt0(au_hnotify_init_br, struct au_branch *br __maybe_unused,
14876 + int perm __maybe_unused)
14877 +AuStubVoid(au_hnotify_fin_br, struct au_branch *br __maybe_unused)
14878 +AuStubInt0(__init au_hnotify_init, void)
14879 +AuStubVoid(au_hnotify_fin, void)
14880 +AuStubVoid(au_hn_init, struct au_hinode *hinode __maybe_unused)
14881 +#endif /* CONFIG_AUFS_HNOTIFY */
14883 +static inline void au_hn_suspend(struct au_hinode *hdir)
14885 + au_hn_ctl(hdir, /*do_set*/0);
14888 +static inline void au_hn_resume(struct au_hinode *hdir)
14890 + au_hn_ctl(hdir, /*do_set*/1);
14893 +static inline void au_hn_imtx_lock(struct au_hinode *hdir)
14895 + mutex_lock(&hdir->hi_inode->i_mutex);
14896 + au_hn_suspend(hdir);
14899 +static inline void au_hn_imtx_lock_nested(struct au_hinode *hdir,
14900 + unsigned int sc __maybe_unused)
14902 + mutex_lock_nested(&hdir->hi_inode->i_mutex, sc);
14903 + au_hn_suspend(hdir);
14906 +static inline void au_hn_imtx_unlock(struct au_hinode *hdir)
14908 + au_hn_resume(hdir);
14909 + mutex_unlock(&hdir->hi_inode->i_mutex);
14912 +#endif /* __KERNEL__ */
14913 +#endif /* __AUFS_INODE_H__ */
14914 diff -urN /usr/share/empty/fs/aufs/ioctl.c linux/fs/aufs/ioctl.c
14915 --- /usr/share/empty/fs/aufs/ioctl.c 1970-01-01 01:00:00.000000000 +0100
14916 +++ linux/fs/aufs/ioctl.c 2011-10-24 20:51:51.580466925 +0200
14919 + * Copyright (C) 2005-2011 Junjiro R. Okajima
14921 + * This program, aufs is free software; you can redistribute it and/or modify
14922 + * it under the terms of the GNU General Public License as published by
14923 + * the Free Software Foundation; either version 2 of the License, or
14924 + * (at your option) any later version.
14926 + * This program is distributed in the hope that it will be useful,
14927 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
14928 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14929 + * GNU General Public License for more details.
14931 + * You should have received a copy of the GNU General Public License
14932 + * along with this program; if not, write to the Free Software
14933 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
14938 + * plink-management and readdir in userspace.
14939 + * assist the pathconf(3) wrapper library.
14942 +#include <linux/file.h>
14945 +static int au_wbr_fd(struct path *path, struct aufs_wbr_fd __user *arg)
14948 + aufs_bindex_t wbi, bindex, bend;
14949 + struct file *h_file;
14950 + struct super_block *sb;
14951 + struct dentry *root;
14952 + struct au_branch *br;
14953 + struct aufs_wbr_fd wbrfd = {
14954 + .oflags = au_dir_roflags,
14957 + const int valid = O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_DIRECTORY
14958 + | O_NOATIME | O_CLOEXEC;
14960 + AuDebugOn(wbrfd.oflags & ~valid);
14963 + err = copy_from_user(&wbrfd, arg, sizeof(wbrfd));
14964 + if (unlikely(err)) {
14970 + AuDbg("wbrfd{0%o, %d}\n", wbrfd.oflags, wbrfd.brid);
14971 + wbrfd.oflags |= au_dir_roflags;
14972 + AuDbg("0%o\n", wbrfd.oflags);
14973 + if (unlikely(wbrfd.oflags & ~valid))
14977 + fd = get_unused_fd();
14979 + if (unlikely(fd < 0))
14982 + h_file = ERR_PTR(-EINVAL);
14985 + sb = path->dentry->d_sb;
14986 + root = sb->s_root;
14987 + aufs_read_lock(root, AuLock_IR);
14988 + bend = au_sbend(sb);
14989 + if (wbrfd.brid >= 0) {
14990 + wbi = au_br_index(sb, wbrfd.brid);
14991 + if (unlikely(wbi < 0 || wbi > bend))
14995 + h_file = ERR_PTR(-ENOENT);
14996 + br = au_sbr(sb, wbi);
14997 + if (!au_br_writable(br->br_perm)) {
15001 + bindex = wbi + 1;
15003 + for (; bindex <= bend; bindex++) {
15004 + br = au_sbr(sb, bindex);
15005 + if (au_br_writable(br->br_perm)) {
15007 + br = au_sbr(sb, wbi);
15012 + AuDbg("wbi %d\n", wbi);
15014 + h_file = au_h_open(root, wbi, wbrfd.oflags, NULL);
15017 + aufs_read_unlock(root, AuLock_IR);
15018 + err = PTR_ERR(h_file);
15019 + if (IS_ERR(h_file))
15022 + atomic_dec(&br->br_count); /* cf. au_h_open() */
15023 + fd_install(fd, h_file);
15025 + goto out; /* success */
15028 + put_unused_fd(fd);
15034 +/* ---------------------------------------------------------------------- */
15036 +long aufs_ioctl_dir(struct file *file, unsigned int cmd, unsigned long arg)
15041 + case AUFS_CTL_RDU:
15042 + case AUFS_CTL_RDU_INO:
15043 + err = au_rdu_ioctl(file, cmd, arg);
15046 + case AUFS_CTL_WBR_FD:
15047 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
15050 + case AUFS_CTL_IBUSY:
15051 + err = au_ibusy_ioctl(file, arg);
15055 + /* do not call the lower */
15056 + AuDbg("0x%x\n", cmd);
15064 +long aufs_ioctl_nondir(struct file *file, unsigned int cmd, unsigned long arg)
15069 + case AUFS_CTL_WBR_FD:
15070 + err = au_wbr_fd(&file->f_path, (void __user *)arg);
15074 + /* do not call the lower */
15075 + AuDbg("0x%x\n", cmd);
15083 +#ifdef CONFIG_COMPAT
15084 +long aufs_compat_ioctl_dir(struct file *file, unsigned int cmd,
15085 + unsigned long arg)
15090 + case AUFS_CTL_RDU:
15091 + case AUFS_CTL_RDU_INO:
15092 + err = au_rdu_compat_ioctl(file, cmd, arg);
15095 + case AUFS_CTL_IBUSY:
15096 + err = au_ibusy_compat_ioctl(file, arg);
15100 + err = aufs_ioctl_dir(file, cmd, arg);
15107 +#if 0 /* unused yet */
15108 +long aufs_compat_ioctl_nondir(struct file *file, unsigned int cmd,
15109 + unsigned long arg)
15111 + return aufs_ioctl_nondir(file, cmd, (unsigned long)compat_ptr(arg));
15115 diff -urN /usr/share/empty/fs/aufs/i_op_add.c linux/fs/aufs/i_op_add.c
15116 --- /usr/share/empty/fs/aufs/i_op_add.c 1970-01-01 01:00:00.000000000 +0100
15117 +++ linux/fs/aufs/i_op_add.c 2011-10-24 20:51:51.580466925 +0200
15120 + * Copyright (C) 2005-2011 Junjiro R. Okajima
15122 + * This program, aufs is free software; you can redistribute it and/or modify
15123 + * it under the terms of the GNU General Public License as published by
15124 + * the Free Software Foundation; either version 2 of the License, or
15125 + * (at your option) any later version.
15127 + * This program is distributed in the hope that it will be useful,
15128 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15129 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15130 + * GNU General Public License for more details.
15132 + * You should have received a copy of the GNU General Public License
15133 + * along with this program; if not, write to the Free Software
15134 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
15138 + * inode operations (add entry)
15144 + * final procedure of adding a new entry, except link(2).
15145 + * remove whiteout, instantiate, copyup the parent dir's times and size
15146 + * and update version.
15147 + * if it failed, re-create the removed whiteout.
15149 +static int epilog(struct inode *dir, aufs_bindex_t bindex,
15150 + struct dentry *wh_dentry, struct dentry *dentry)
15153 + aufs_bindex_t bwh;
15154 + struct path h_path;
15155 + struct inode *inode, *h_dir;
15156 + struct dentry *wh;
15160 + h_dir = wh_dentry->d_parent->d_inode; /* dir inode is locked */
15161 + IMustLock(h_dir);
15162 + AuDebugOn(au_h_iptr(dir, bindex) != h_dir);
15163 + bwh = au_dbwh(dentry);
15164 + h_path.dentry = wh_dentry;
15165 + h_path.mnt = au_sbr_mnt(dir->i_sb, bindex);
15166 + err = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path,
15168 + if (unlikely(err))
15172 + inode = au_new_inode(dentry, /*must_new*/1);
15173 + if (!IS_ERR(inode)) {
15174 + d_instantiate(dentry, inode);
15175 + dir = dentry->d_parent->d_inode; /* dir inode is locked */
15177 + if (au_ibstart(dir) == au_dbstart(dentry))
15178 + au_cpup_attr_timesizes(dir);
15179 + dir->i_version++;
15180 + return 0; /* success */
15183 + err = PTR_ERR(inode);
15188 + /* dir inode is locked */
15189 + wh = au_wh_create(dentry, bwh, wh_dentry->d_parent);
15190 + rerr = PTR_ERR(wh);
15191 + if (IS_ERR(wh)) {
15192 + AuIOErr("%.*s reverting whiteout failed(%d, %d)\n",
15193 + AuDLNPair(dentry), err, rerr);
15202 +static int au_d_may_add(struct dentry *dentry)
15207 + if (unlikely(d_unhashed(dentry)))
15209 + if (unlikely(dentry->d_inode))
15215 + * simple tests for the adding inode operations.
15216 + * following the checks in vfs, plus the parent-child relationship.
15218 +int au_may_add(struct dentry *dentry, aufs_bindex_t bindex,
15219 + struct dentry *h_parent, int isdir)
15223 + struct dentry *h_dentry;
15224 + struct inode *h_inode;
15226 + err = -ENAMETOOLONG;
15227 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
15230 + h_dentry = au_h_dptr(dentry, bindex);
15231 + h_inode = h_dentry->d_inode;
15232 + if (!dentry->d_inode) {
15234 + if (unlikely(h_inode))
15237 + /* rename(2) case */
15239 + if (unlikely(!h_inode || !h_inode->i_nlink))
15242 + h_mode = h_inode->i_mode;
15245 + if (unlikely(S_ISDIR(h_mode)))
15247 + } else if (unlikely(!S_ISDIR(h_mode))) {
15254 + /* expected parent dir is locked */
15255 + if (unlikely(h_parent != h_dentry->d_parent))
15264 + * initial procedure of adding a new entry.
15265 + * prepare writable branch and the parent dir, lock it,
15266 + * and lookup whiteout for the new entry.
15268 +static struct dentry*
15269 +lock_hdir_lkup_wh(struct dentry *dentry, struct au_dtime *dt,
15270 + struct dentry *src_dentry, struct au_pin *pin,
15271 + struct au_wr_dir_args *wr_dir_args)
15273 + struct dentry *wh_dentry, *h_parent;
15274 + struct super_block *sb;
15275 + struct au_branch *br;
15277 + unsigned int udba;
15278 + aufs_bindex_t bcpup;
15280 + AuDbg("%.*s\n", AuDLNPair(dentry));
15282 + err = au_wr_dir(dentry, src_dentry, wr_dir_args);
15284 + wh_dentry = ERR_PTR(err);
15285 + if (unlikely(err < 0))
15288 + sb = dentry->d_sb;
15289 + udba = au_opt_udba(sb);
15290 + err = au_pin(pin, dentry, bcpup, udba,
15291 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
15292 + wh_dentry = ERR_PTR(err);
15293 + if (unlikely(err))
15296 + h_parent = au_pinned_h_parent(pin);
15297 + if (udba != AuOpt_UDBA_NONE
15298 + && au_dbstart(dentry) == bcpup)
15299 + err = au_may_add(dentry, bcpup, h_parent,
15300 + au_ftest_wrdir(wr_dir_args->flags, ISDIR));
15301 + else if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
15302 + err = -ENAMETOOLONG;
15303 + wh_dentry = ERR_PTR(err);
15304 + if (unlikely(err))
15307 + br = au_sbr(sb, bcpup);
15309 + struct path tmp = {
15310 + .dentry = h_parent,
15311 + .mnt = br->br_mnt
15313 + au_dtime_store(dt, au_pinned_parent(pin), &tmp);
15316 + wh_dentry = NULL;
15317 + if (bcpup != au_dbwh(dentry))
15318 + goto out; /* success */
15320 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
15323 + if (IS_ERR(wh_dentry))
15326 + return wh_dentry;
15329 +/* ---------------------------------------------------------------------- */
15331 +enum { Mknod, Symlink, Creat };
15332 +struct simple_arg {
15337 + struct nameidata *nd;
15340 + const char *symname;
15349 +static int add_simple(struct inode *dir, struct dentry *dentry,
15350 + struct simple_arg *arg)
15353 + aufs_bindex_t bstart;
15354 + unsigned char created;
15355 + struct au_dtime dt;
15356 + struct au_pin pin;
15357 + struct path h_path;
15358 + struct dentry *wh_dentry, *parent;
15359 + struct inode *h_dir;
15360 + struct au_wr_dir_args wr_dir_args = {
15361 + .force_btgt = -1,
15362 + .flags = AuWrDir_ADD_ENTRY
15365 + AuDbg("%.*s\n", AuDLNPair(dentry));
15368 + parent = dentry->d_parent; /* dir inode is locked */
15369 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
15370 + if (unlikely(err))
15372 + err = au_d_may_add(dentry);
15373 + if (unlikely(err))
15375 + di_write_lock_parent(parent);
15376 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, /*src_dentry*/NULL, &pin,
15378 + err = PTR_ERR(wh_dentry);
15379 + if (IS_ERR(wh_dentry))
15382 + bstart = au_dbstart(dentry);
15383 + h_path.dentry = au_h_dptr(dentry, bstart);
15384 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
15385 + h_dir = au_pinned_h_dir(&pin);
15386 + switch (arg->type) {
15388 + err = vfsub_create(h_dir, &h_path, arg->u.c.mode);
15391 + err = vfsub_symlink(h_dir, &h_path, arg->u.s.symname);
15394 + err = vfsub_mknod(h_dir, &h_path, arg->u.m.mode, arg->u.m.dev);
15401 + err = epilog(dir, bstart, wh_dentry, dentry);
15404 + if (unlikely(created && err && h_path.dentry->d_inode)) {
15406 + rerr = vfsub_unlink(h_dir, &h_path, /*force*/0);
15408 + AuIOErr("%.*s revert failure(%d, %d)\n",
15409 + AuDLNPair(dentry), err, rerr);
15412 + au_dtime_revert(&dt);
15419 + di_write_unlock(parent);
15421 + if (unlikely(err)) {
15422 + au_update_dbstart(dentry);
15425 + aufs_read_unlock(dentry, AuLock_DW);
15430 +int aufs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
15432 + struct simple_arg arg = {
15439 + return add_simple(dir, dentry, &arg);
15442 +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname)
15444 + struct simple_arg arg = {
15446 + .u.s.symname = symname
15448 + return add_simple(dir, dentry, &arg);
15451 +int aufs_create(struct inode *dir, struct dentry *dentry, int mode,
15452 + struct nameidata *nd)
15454 + struct simple_arg arg = {
15461 + return add_simple(dir, dentry, &arg);
15464 +/* ---------------------------------------------------------------------- */
15466 +struct au_link_args {
15467 + aufs_bindex_t bdst, bsrc;
15468 + struct au_pin pin;
15469 + struct path h_path;
15470 + struct dentry *src_parent, *parent;
15473 +static int au_cpup_before_link(struct dentry *src_dentry,
15474 + struct au_link_args *a)
15477 + struct dentry *h_src_dentry;
15478 + struct mutex *h_mtx;
15479 + struct file *h_file;
15481 + di_read_lock_parent(a->src_parent, AuLock_IR);
15482 + err = au_test_and_cpup_dirs(src_dentry, a->bdst);
15483 + if (unlikely(err))
15486 + h_src_dentry = au_h_dptr(src_dentry, a->bsrc);
15487 + h_mtx = &h_src_dentry->d_inode->i_mutex;
15488 + err = au_pin(&a->pin, src_dentry, a->bdst,
15489 + au_opt_udba(src_dentry->d_sb),
15490 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
15491 + if (unlikely(err))
15493 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
15494 + h_file = au_h_open_pre(src_dentry, a->bsrc);
15495 + if (IS_ERR(h_file)) {
15496 + err = PTR_ERR(h_file);
15499 + err = au_sio_cpup_simple(src_dentry, a->bdst, -1,
15500 + AuCpup_DTIME /* | AuCpup_KEEPLINO */);
15501 + mutex_unlock(h_mtx);
15502 + au_h_open_post(src_dentry, a->bsrc, h_file);
15503 + au_unpin(&a->pin);
15506 + di_read_unlock(a->src_parent, AuLock_IR);
15510 +static int au_cpup_or_link(struct dentry *src_dentry, struct au_link_args *a)
15513 + unsigned char plink;
15514 + struct inode *h_inode, *inode;
15515 + struct dentry *h_src_dentry;
15516 + struct super_block *sb;
15517 + struct file *h_file;
15521 + sb = src_dentry->d_sb;
15522 + inode = src_dentry->d_inode;
15523 + if (au_ibstart(inode) <= a->bdst)
15524 + h_inode = au_h_iptr(inode, a->bdst);
15525 + if (!h_inode || !h_inode->i_nlink) {
15526 + /* copyup src_dentry as the name of dentry. */
15527 + au_set_dbstart(src_dentry, a->bdst);
15528 + au_set_h_dptr(src_dentry, a->bdst, dget(a->h_path.dentry));
15529 + h_inode = au_h_dptr(src_dentry, a->bsrc)->d_inode;
15530 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
15531 + h_file = au_h_open_pre(src_dentry, a->bsrc);
15532 + if (IS_ERR(h_file)) {
15533 + err = PTR_ERR(h_file);
15536 + err = au_sio_cpup_single(src_dentry, a->bdst, a->bsrc,
15537 + -1, AuCpup_KEEPLINO,
15539 + mutex_unlock(&h_inode->i_mutex);
15540 + au_h_open_post(src_dentry, a->bsrc, h_file);
15541 + au_set_h_dptr(src_dentry, a->bdst, NULL);
15542 + au_set_dbstart(src_dentry, a->bsrc);
15544 + /* the inode of src_dentry already exists on a.bdst branch */
15545 + h_src_dentry = d_find_alias(h_inode);
15546 + if (!h_src_dentry && au_plink_test(inode)) {
15548 + h_src_dentry = au_plink_lkup(inode, a->bdst);
15549 + err = PTR_ERR(h_src_dentry);
15550 + if (IS_ERR(h_src_dentry))
15553 + if (unlikely(!h_src_dentry->d_inode)) {
15554 + dput(h_src_dentry);
15555 + h_src_dentry = NULL;
15559 + if (h_src_dentry) {
15560 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
15562 + dput(h_src_dentry);
15564 + AuIOErr("no dentry found for hi%lu on b%d\n",
15565 + h_inode->i_ino, a->bdst);
15570 + if (!err && !plink)
15571 + au_plink_append(inode, a->bdst, a->h_path.dentry);
15578 +int aufs_link(struct dentry *src_dentry, struct inode *dir,
15579 + struct dentry *dentry)
15582 + struct au_dtime dt;
15583 + struct au_link_args *a;
15584 + struct dentry *wh_dentry, *h_src_dentry;
15585 + struct inode *inode;
15586 + struct super_block *sb;
15587 + struct au_wr_dir_args wr_dir_args = {
15588 + /* .force_btgt = -1, */
15589 + .flags = AuWrDir_ADD_ENTRY
15593 + inode = src_dentry->d_inode;
15594 + IMustLock(inode);
15597 + a = kzalloc(sizeof(*a), GFP_NOFS);
15598 + if (unlikely(!a))
15601 + a->parent = dentry->d_parent; /* dir inode is locked */
15602 + err = aufs_read_and_write_lock2(dentry, src_dentry,
15603 + AuLock_NOPLM | AuLock_GEN);
15604 + if (unlikely(err))
15606 + err = au_d_hashed_positive(src_dentry);
15607 + if (unlikely(err))
15609 + err = au_d_may_add(dentry);
15610 + if (unlikely(err))
15613 + a->src_parent = dget_parent(src_dentry);
15614 + wr_dir_args.force_btgt = au_ibstart(inode);
15616 + di_write_lock_parent(a->parent);
15617 + wr_dir_args.force_btgt = au_wbr(dentry, wr_dir_args.force_btgt);
15618 + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, &a->pin,
15620 + err = PTR_ERR(wh_dentry);
15621 + if (IS_ERR(wh_dentry))
15625 + sb = dentry->d_sb;
15626 + a->bdst = au_dbstart(dentry);
15627 + a->h_path.dentry = au_h_dptr(dentry, a->bdst);
15628 + a->h_path.mnt = au_sbr_mnt(sb, a->bdst);
15629 + a->bsrc = au_ibstart(inode);
15630 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
15631 + if (!h_src_dentry) {
15632 + a->bsrc = au_dbstart(src_dentry);
15633 + h_src_dentry = au_h_d_alias(src_dentry, a->bsrc);
15634 + AuDebugOn(!h_src_dentry);
15635 + } else if (IS_ERR(h_src_dentry))
15638 + if (au_opt_test(au_mntflags(sb), PLINK)) {
15639 + if (a->bdst < a->bsrc
15640 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */)
15641 + err = au_cpup_or_link(src_dentry, a);
15643 + err = vfsub_link(h_src_dentry, au_pinned_h_dir(&a->pin),
15645 + dput(h_src_dentry);
15648 + * copyup src_dentry to the branch we process,
15649 + * and then link(2) to it.
15651 + dput(h_src_dentry);
15652 + if (a->bdst < a->bsrc
15653 + /* && h_src_dentry->d_sb != a->h_path.dentry->d_sb */) {
15654 + au_unpin(&a->pin);
15655 + di_write_unlock(a->parent);
15656 + err = au_cpup_before_link(src_dentry, a);
15657 + di_write_lock_parent(a->parent);
15659 + err = au_pin(&a->pin, dentry, a->bdst,
15661 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
15662 + if (unlikely(err))
15666 + h_src_dentry = au_h_dptr(src_dentry, a->bdst);
15668 + if (h_src_dentry && h_src_dentry->d_inode)
15669 + err = vfsub_link(h_src_dentry,
15670 + au_pinned_h_dir(&a->pin),
15674 + if (unlikely(err))
15678 + a->h_path.dentry = wh_dentry;
15679 + err = au_wh_unlink_dentry(au_pinned_h_dir(&a->pin), &a->h_path,
15681 + if (unlikely(err))
15685 + dir->i_version++;
15686 + if (au_ibstart(dir) == au_dbstart(dentry))
15687 + au_cpup_attr_timesizes(dir);
15688 + inc_nlink(inode);
15689 + inode->i_ctime = dir->i_ctime;
15690 + d_instantiate(dentry, au_igrab(inode));
15691 + if (d_unhashed(a->h_path.dentry))
15692 + /* some filesystem calls d_drop() */
15694 + goto out_unpin; /* success */
15697 + rerr = vfsub_unlink(au_pinned_h_dir(&a->pin), &a->h_path, /*force*/0);
15698 + if (unlikely(rerr)) {
15699 + AuIOErr("%.*s reverting failed(%d, %d)\n",
15700 + AuDLNPair(dentry), err, rerr);
15703 + au_dtime_revert(&dt);
15705 + au_unpin(&a->pin);
15709 + di_write_unlock(a->parent);
15710 + dput(a->src_parent);
15712 + if (unlikely(err)) {
15713 + au_update_dbstart(dentry);
15716 + aufs_read_and_write_unlock2(dentry, src_dentry);
15723 +int aufs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
15726 + aufs_bindex_t bindex;
15727 + unsigned char diropq;
15728 + struct path h_path;
15729 + struct dentry *wh_dentry, *parent, *opq_dentry;
15730 + struct mutex *h_mtx;
15731 + struct super_block *sb;
15733 + struct au_pin pin;
15734 + struct au_dtime dt;
15735 + } *a; /* reduce the stack usage */
15736 + struct au_wr_dir_args wr_dir_args = {
15737 + .force_btgt = -1,
15738 + .flags = AuWrDir_ADD_ENTRY | AuWrDir_ISDIR
15744 + a = kmalloc(sizeof(*a), GFP_NOFS);
15745 + if (unlikely(!a))
15748 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
15749 + if (unlikely(err))
15751 + err = au_d_may_add(dentry);
15752 + if (unlikely(err))
15755 + parent = dentry->d_parent; /* dir inode is locked */
15756 + di_write_lock_parent(parent);
15757 + wh_dentry = lock_hdir_lkup_wh(dentry, &a->dt, /*src_dentry*/NULL,
15758 + &a->pin, &wr_dir_args);
15759 + err = PTR_ERR(wh_dentry);
15760 + if (IS_ERR(wh_dentry))
15763 + sb = dentry->d_sb;
15764 + bindex = au_dbstart(dentry);
15765 + h_path.dentry = au_h_dptr(dentry, bindex);
15766 + h_path.mnt = au_sbr_mnt(sb, bindex);
15767 + err = vfsub_mkdir(au_pinned_h_dir(&a->pin), &h_path, mode);
15768 + if (unlikely(err))
15771 + /* make the dir opaque */
15773 + h_mtx = &h_path.dentry->d_inode->i_mutex;
15775 + || au_opt_test(au_mntflags(sb), ALWAYS_DIROPQ)) {
15776 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
15777 + opq_dentry = au_diropq_create(dentry, bindex);
15778 + mutex_unlock(h_mtx);
15779 + err = PTR_ERR(opq_dentry);
15780 + if (IS_ERR(opq_dentry))
15782 + dput(opq_dentry);
15786 + err = epilog(dir, bindex, wh_dentry, dentry);
15789 + goto out_unpin; /* success */
15794 + AuLabel(revert opq);
15795 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
15796 + rerr = au_diropq_remove(dentry, bindex);
15797 + mutex_unlock(h_mtx);
15799 + AuIOErr("%.*s reverting diropq failed(%d, %d)\n",
15800 + AuDLNPair(dentry), err, rerr);
15806 + AuLabel(revert dir);
15807 + rerr = vfsub_rmdir(au_pinned_h_dir(&a->pin), &h_path);
15809 + AuIOErr("%.*s reverting dir failed(%d, %d)\n",
15810 + AuDLNPair(dentry), err, rerr);
15813 + au_dtime_revert(&a->dt);
15815 + au_unpin(&a->pin);
15818 + di_write_unlock(parent);
15820 + if (unlikely(err)) {
15821 + au_update_dbstart(dentry);
15824 + aufs_read_unlock(dentry, AuLock_DW);
15830 diff -urN /usr/share/empty/fs/aufs/i_op.c linux/fs/aufs/i_op.c
15831 --- /usr/share/empty/fs/aufs/i_op.c 1970-01-01 01:00:00.000000000 +0100
15832 +++ linux/fs/aufs/i_op.c 2012-01-05 11:00:59.215501779 +0100
15835 + * Copyright (C) 2005-2011 Junjiro R. Okajima
15837 + * This program, aufs is free software; you can redistribute it and/or modify
15838 + * it under the terms of the GNU General Public License as published by
15839 + * the Free Software Foundation; either version 2 of the License, or
15840 + * (at your option) any later version.
15842 + * This program is distributed in the hope that it will be useful,
15843 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
15844 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15845 + * GNU General Public License for more details.
15847 + * You should have received a copy of the GNU General Public License
15848 + * along with this program; if not, write to the Free Software
15849 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
15853 + * inode operations (except add/del/rename)
15856 +#include <linux/device_cgroup.h>
15857 +#include <linux/fs_stack.h>
15858 +#include <linux/mm.h>
15859 +#include <linux/namei.h>
15860 +#include <linux/security.h>
15861 +#include <linux/uaccess.h>
15864 +static int h_permission(struct inode *h_inode, int mask,
15865 + struct vfsmount *h_mnt, int brperm)
15868 + const unsigned char write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
15871 + if ((write_mask && IS_IMMUTABLE(h_inode))
15872 + || ((mask & MAY_EXEC)
15873 + && S_ISREG(h_inode->i_mode)
15874 + && ((h_mnt->mnt_flags & MNT_NOEXEC)
15875 + || !(h_inode->i_mode & S_IXUGO))))
15879 + * - skip the lower fs test in the case of write to ro branch.
15880 + * - nfs dir permission write check is optimized, but a policy for
15881 + * link/rename requires a real check.
15883 + if ((write_mask && !au_br_writable(brperm))
15884 + || (au_test_nfs(h_inode->i_sb) && S_ISDIR(h_inode->i_mode)
15885 + && write_mask && !(mask & MAY_READ))
15886 + || !h_inode->i_op->permission) {
15887 + /* AuLabel(generic_permission); */
15888 + err = generic_permission(h_inode, mask);
15890 + /* AuLabel(h_inode->permission); */
15891 + err = h_inode->i_op->permission(h_inode, mask);
15896 + err = devcgroup_inode_permission(h_inode, mask);
15898 + err = security_inode_permission(h_inode, mask);
15902 + /* todo: do we need to call ima_path_check()? */
15903 + struct path h_path = {
15907 + err = ima_path_check(&h_path,
15908 + mask & (MAY_READ | MAY_WRITE | MAY_EXEC),
15909 + IMA_COUNT_LEAVE);
15917 +static int aufs_permission(struct inode *inode, int mask)
15920 + aufs_bindex_t bindex, bend;
15921 + const unsigned char isdir = !!S_ISDIR(inode->i_mode),
15922 + write_mask = !!(mask & (MAY_WRITE | MAY_APPEND));
15923 + struct inode *h_inode;
15924 + struct super_block *sb;
15925 + struct au_branch *br;
15927 + /* todo: support rcu-walk? */
15928 + if (mask & MAY_NOT_BLOCK)
15931 + sb = inode->i_sb;
15932 + si_read_lock(sb, AuLock_FLUSH);
15933 + ii_read_lock_child(inode);
15935 + err = au_iigen_test(inode, au_sigen(sb));
15936 + if (unlikely(err))
15940 + if (!isdir || write_mask) {
15941 + err = au_busy_or_stale();
15942 + h_inode = au_h_iptr(inode, au_ibstart(inode));
15943 + if (unlikely(!h_inode
15944 + || (h_inode->i_mode & S_IFMT)
15945 + != (inode->i_mode & S_IFMT)))
15949 + bindex = au_ibstart(inode);
15950 + br = au_sbr(sb, bindex);
15951 + err = h_permission(h_inode, mask, br->br_mnt, br->br_perm);
15954 + && !special_file(h_inode->i_mode)) {
15955 + /* test whether the upper writable branch exists */
15957 + for (; bindex >= 0; bindex--)
15958 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
15966 + /* non-write to dir */
15968 + bend = au_ibend(inode);
15969 + for (bindex = au_ibstart(inode); !err && bindex <= bend; bindex++) {
15970 + h_inode = au_h_iptr(inode, bindex);
15972 + err = au_busy_or_stale();
15973 + if (unlikely(!S_ISDIR(h_inode->i_mode)))
15976 + br = au_sbr(sb, bindex);
15977 + err = h_permission(h_inode, mask, br->br_mnt,
15983 + ii_read_unlock(inode);
15984 + si_read_unlock(sb);
15988 +/* ---------------------------------------------------------------------- */
15990 +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry,
15991 + struct nameidata *nd)
15993 + struct dentry *ret, *parent;
15994 + struct inode *inode;
15995 + struct super_block *sb;
15996 + int err, npositive, lc_idx;
16001 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16002 + ret = ERR_PTR(err);
16003 + if (unlikely(err))
16006 + ret = ERR_PTR(-ENAMETOOLONG);
16007 + if (unlikely(dentry->d_name.len > AUFS_MAX_NAMELEN))
16009 + err = au_di_init(dentry);
16010 + ret = ERR_PTR(err);
16011 + if (unlikely(err))
16015 + npositive = 0; /* suppress a warning */
16016 + parent = dentry->d_parent; /* dir inode is locked */
16017 + di_read_lock_parent(parent, AuLock_IR);
16018 + err = au_alive_dir(parent);
16020 + err = au_digen_test(parent, au_sigen(sb));
16022 + npositive = au_lkup_dentry(dentry, au_dbstart(parent),
16026 + di_read_unlock(parent, AuLock_IR);
16027 + ret = ERR_PTR(err);
16028 + if (unlikely(err < 0))
16032 + inode = au_new_inode(dentry, /*must_new*/0);
16033 + ret = (void *)inode;
16035 + if (IS_ERR(inode)) {
16040 + ret = d_splice_alias(inode, dentry);
16041 + if (unlikely(IS_ERR(ret) && inode)) {
16042 + ii_write_unlock(inode);
16043 + lc_idx = AuLcNonDir_IIINFO;
16044 + if (S_ISLNK(inode->i_mode))
16045 + lc_idx = AuLcSymlink_IIINFO;
16046 + else if (S_ISDIR(inode->i_mode))
16047 + lc_idx = AuLcDir_IIINFO;
16048 + au_rw_class(&au_ii(inode)->ii_rwsem, au_lc_key + lc_idx);
16053 + di_write_unlock(dentry);
16054 + if (unlikely(IS_ERR(ret) && inode)) {
16055 + lc_idx = AuLcNonDir_DIINFO;
16056 + if (S_ISLNK(inode->i_mode))
16057 + lc_idx = AuLcSymlink_DIINFO;
16058 + else if (S_ISDIR(inode->i_mode))
16059 + lc_idx = AuLcDir_DIINFO;
16060 + au_rw_class(&au_di(dentry)->di_rwsem, au_lc_key + lc_idx);
16063 + si_read_unlock(sb);
16068 +/* ---------------------------------------------------------------------- */
16070 +static int au_wr_dir_cpup(struct dentry *dentry, struct dentry *parent,
16071 + const unsigned char add_entry, aufs_bindex_t bcpup,
16072 + aufs_bindex_t bstart)
16075 + struct dentry *h_parent;
16076 + struct inode *h_dir;
16079 + IMustLock(parent->d_inode);
16081 + di_write_lock_parent(parent);
16084 + if (!au_h_dptr(parent, bcpup)) {
16085 + if (bstart < bcpup)
16086 + err = au_cpdown_dirs(dentry, bcpup);
16088 + err = au_cpup_dirs(dentry, bcpup);
16090 + if (!err && add_entry) {
16091 + h_parent = au_h_dptr(parent, bcpup);
16092 + h_dir = h_parent->d_inode;
16093 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
16094 + err = au_lkup_neg(dentry, bcpup);
16095 + /* todo: no unlock here */
16096 + mutex_unlock(&h_dir->i_mutex);
16098 + AuDbg("bcpup %d\n", bcpup);
16100 + if (!dentry->d_inode)
16101 + au_set_h_dptr(dentry, bstart, NULL);
16102 + au_update_dbrange(dentry, /*do_put_zero*/0);
16107 + di_write_unlock(parent);
16109 + err = bcpup; /* success */
16116 + * decide the branch and the parent dir where we will create a new entry.
16117 + * returns new bindex or an error.
16118 + * copyup the parent dir if needed.
16120 +int au_wr_dir(struct dentry *dentry, struct dentry *src_dentry,
16121 + struct au_wr_dir_args *args)
16124 + aufs_bindex_t bcpup, bstart, src_bstart;
16125 + const unsigned char add_entry = !!au_ftest_wrdir(args->flags,
16127 + struct super_block *sb;
16128 + struct dentry *parent;
16129 + struct au_sbinfo *sbinfo;
16131 + sb = dentry->d_sb;
16132 + sbinfo = au_sbi(sb);
16133 + parent = dget_parent(dentry);
16134 + bstart = au_dbstart(dentry);
16136 + if (args->force_btgt < 0) {
16137 + if (src_dentry) {
16138 + src_bstart = au_dbstart(src_dentry);
16139 + if (src_bstart < bstart)
16140 + bcpup = src_bstart;
16141 + } else if (add_entry) {
16142 + err = AuWbrCreate(sbinfo, dentry,
16143 + au_ftest_wrdir(args->flags, ISDIR));
16147 + if (bcpup < 0 || au_test_ro(sb, bcpup, dentry->d_inode)) {
16149 + err = AuWbrCopyup(sbinfo, dentry);
16151 + if (!IS_ROOT(dentry)) {
16152 + di_read_lock_parent(parent, !AuLock_IR);
16153 + err = AuWbrCopyup(sbinfo, dentry);
16154 + di_read_unlock(parent, !AuLock_IR);
16156 + err = AuWbrCopyup(sbinfo, dentry);
16159 + if (unlikely(err < 0))
16163 + bcpup = args->force_btgt;
16164 + AuDebugOn(au_test_ro(sb, bcpup, dentry->d_inode));
16167 + AuDbg("bstart %d, bcpup %d\n", bstart, bcpup);
16169 + if (bcpup == bstart)
16170 + goto out; /* success */
16172 + /* copyup the new parent into the branch we process */
16173 + err = au_wr_dir_cpup(dentry, parent, add_entry, bcpup, bstart);
16175 + if (!dentry->d_inode) {
16176 + au_set_h_dptr(dentry, bstart, NULL);
16177 + au_set_dbstart(dentry, bcpup);
16178 + au_set_dbend(dentry, bcpup);
16180 + AuDebugOn(add_entry && !au_h_dptr(dentry, bcpup));
16188 +/* ---------------------------------------------------------------------- */
16190 +struct dentry *au_pinned_h_parent(struct au_pin *pin)
16192 + if (pin && pin->parent)
16193 + return au_h_dptr(pin->parent, pin->bindex);
16197 +void au_unpin(struct au_pin *p)
16199 + if (p->h_mnt && au_ftest_pin(p->flags, MNT_WRITE))
16200 + mnt_drop_write(p->h_mnt);
16204 + au_hn_imtx_unlock(p->hdir);
16205 + if (!au_ftest_pin(p->flags, DI_LOCKED))
16206 + di_read_unlock(p->parent, AuLock_IR);
16207 + iput(p->hdir->hi_inode);
16209 + p->parent = NULL;
16214 +int au_do_pin(struct au_pin *p)
16217 + struct super_block *sb;
16218 + struct dentry *h_dentry, *h_parent;
16219 + struct au_branch *br;
16220 + struct inode *h_dir;
16223 + sb = p->dentry->d_sb;
16224 + br = au_sbr(sb, p->bindex);
16225 + if (IS_ROOT(p->dentry)) {
16226 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
16227 + p->h_mnt = br->br_mnt;
16228 + err = mnt_want_write(p->h_mnt);
16229 + if (unlikely(err)) {
16230 + au_fclr_pin(p->flags, MNT_WRITE);
16238 + if (p->bindex <= au_dbend(p->dentry))
16239 + h_dentry = au_h_dptr(p->dentry, p->bindex);
16241 + p->parent = dget_parent(p->dentry);
16242 + if (!au_ftest_pin(p->flags, DI_LOCKED))
16243 + di_read_lock(p->parent, AuLock_IR, p->lsc_di);
16246 + h_parent = au_h_dptr(p->parent, p->bindex);
16247 + p->hdir = au_hi(p->parent->d_inode, p->bindex);
16249 + h_dir = p->hdir->hi_inode;
16253 + * if DI_LOCKED is not set, then p->parent may be different
16254 + * and h_parent can be NULL.
16256 + if (unlikely(!p->hdir || !h_dir || !h_parent)) {
16258 + if (!au_ftest_pin(p->flags, DI_LOCKED))
16259 + di_read_unlock(p->parent, AuLock_IR);
16261 + p->parent = NULL;
16266 + au_hn_imtx_lock_nested(p->hdir, p->lsc_hi);
16268 + if (unlikely(p->hdir->hi_inode != h_parent->d_inode)) {
16273 + err = au_h_verify(h_dentry, p->udba, h_dir, h_parent, br);
16274 + if (unlikely(err)) {
16275 + au_fclr_pin(p->flags, MNT_WRITE);
16280 + if (au_ftest_pin(p->flags, MNT_WRITE)) {
16281 + p->h_mnt = br->br_mnt;
16282 + err = mnt_want_write(p->h_mnt);
16283 + if (unlikely(err)) {
16284 + au_fclr_pin(p->flags, MNT_WRITE);
16288 + goto out; /* success */
16293 + pr_err("err %d\n", err);
16294 + err = au_busy_or_stale();
16299 +void au_pin_init(struct au_pin *p, struct dentry *dentry,
16300 + aufs_bindex_t bindex, int lsc_di, int lsc_hi,
16301 + unsigned int udba, unsigned char flags)
16303 + p->dentry = dentry;
16305 + p->lsc_di = lsc_di;
16306 + p->lsc_hi = lsc_hi;
16307 + p->flags = flags;
16308 + p->bindex = bindex;
16310 + p->parent = NULL;
16315 +int au_pin(struct au_pin *pin, struct dentry *dentry, aufs_bindex_t bindex,
16316 + unsigned int udba, unsigned char flags)
16318 + au_pin_init(pin, dentry, bindex, AuLsc_DI_PARENT, AuLsc_I_PARENT2,
16320 + return au_do_pin(pin);
16323 +/* ---------------------------------------------------------------------- */
16326 + * ->setattr() and ->getattr() are called in various cases.
16327 + * chmod, stat: dentry is revalidated.
16328 + * fchmod, fstat: file and dentry are not revalidated, additionally they may be
16330 + * for ->setattr(), ia->ia_file is passed from ftruncate only.
16332 +/* todo: consolidate with do_refresh() and simple_reval_dpath() */
16333 +static int au_reval_for_attr(struct dentry *dentry, unsigned int sigen)
16336 + struct inode *inode;
16337 + struct dentry *parent;
16340 + inode = dentry->d_inode;
16341 + if (au_digen_test(dentry, sigen)) {
16342 + parent = dget_parent(dentry);
16343 + di_read_lock_parent(parent, AuLock_IR);
16344 + err = au_refresh_dentry(dentry, parent);
16345 + di_read_unlock(parent, AuLock_IR);
16353 +#define AuIcpup_DID_CPUP 1
16354 +#define au_ftest_icpup(flags, name) ((flags) & AuIcpup_##name)
16355 +#define au_fset_icpup(flags, name) \
16356 + do { (flags) |= AuIcpup_##name; } while (0)
16357 +#define au_fclr_icpup(flags, name) \
16358 + do { (flags) &= ~AuIcpup_##name; } while (0)
16360 +struct au_icpup_args {
16361 + unsigned char flags;
16362 + unsigned char pin_flags;
16363 + aufs_bindex_t btgt;
16364 + unsigned int udba;
16365 + struct au_pin pin;
16366 + struct path h_path;
16367 + struct inode *h_inode;
16370 +static int au_pin_and_icpup(struct dentry *dentry, struct iattr *ia,
16371 + struct au_icpup_args *a)
16375 + aufs_bindex_t bstart, ibstart;
16376 + struct dentry *hi_wh, *parent;
16377 + struct inode *inode;
16378 + struct file *h_file;
16379 + struct au_wr_dir_args wr_dir_args = {
16380 + .force_btgt = -1,
16384 + bstart = au_dbstart(dentry);
16385 + inode = dentry->d_inode;
16386 + if (S_ISDIR(inode->i_mode))
16387 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
16388 + /* plink or hi_wh() case */
16389 + ibstart = au_ibstart(inode);
16390 + if (bstart != ibstart && !au_test_ro(inode->i_sb, ibstart, inode))
16391 + wr_dir_args.force_btgt = ibstart;
16392 + err = au_wr_dir(dentry, /*src_dentry*/NULL, &wr_dir_args);
16393 + if (unlikely(err < 0))
16396 + if (err != bstart)
16397 + au_fset_icpup(a->flags, DID_CPUP);
16400 + a->pin_flags = AuPin_MNT_WRITE;
16402 + if (!IS_ROOT(dentry)) {
16403 + au_fset_pin(a->pin_flags, DI_LOCKED);
16404 + parent = dget_parent(dentry);
16405 + di_write_lock_parent(parent);
16408 + err = au_pin(&a->pin, dentry, a->btgt, a->udba, a->pin_flags);
16409 + if (unlikely(err))
16412 + a->h_path.dentry = au_h_dptr(dentry, bstart);
16413 + a->h_inode = a->h_path.dentry->d_inode;
16414 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
16416 + if ((ia->ia_valid & ATTR_SIZE) && ia->ia_size < i_size_read(a->h_inode))
16417 + sz = ia->ia_size;
16421 + if (au_ftest_icpup(a->flags, DID_CPUP) && d_unlinked(dentry)) {
16422 + hi_wh = au_hi_wh(inode, a->btgt);
16424 + err = au_sio_cpup_wh(dentry, a->btgt, sz, /*file*/NULL);
16425 + if (unlikely(err))
16427 + hi_wh = au_hi_wh(inode, a->btgt);
16428 + /* todo: revalidate hi_wh? */
16433 + au_pin_set_parent_lflag(&a->pin, /*lflag*/0);
16434 + di_downgrade_lock(parent, AuLock_IR);
16438 + if (!au_ftest_icpup(a->flags, DID_CPUP))
16439 + goto out; /* success */
16441 + if (!d_unhashed(dentry)) {
16442 + h_file = au_h_open_pre(dentry, bstart);
16443 + if (IS_ERR(h_file)) {
16444 + err = PTR_ERR(h_file);
16447 + err = au_sio_cpup_simple(dentry, a->btgt, sz,
16450 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
16451 + } else if (!hi_wh)
16452 + a->h_path.dentry = au_h_dptr(dentry, a->btgt);
16454 + a->h_path.dentry = hi_wh; /* do not dget here */
16457 + mutex_unlock(&a->h_inode->i_mutex);
16458 + au_h_open_post(dentry, bstart, h_file);
16459 + a->h_inode = a->h_path.dentry->d_inode;
16461 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
16462 + goto out; /* success */
16465 + au_unpin(&a->pin);
16468 + di_write_unlock(parent);
16475 +static int aufs_setattr(struct dentry *dentry, struct iattr *ia)
16478 + struct inode *inode;
16479 + struct super_block *sb;
16480 + struct file *file;
16481 + struct au_icpup_args *a;
16483 + inode = dentry->d_inode;
16484 + IMustLock(inode);
16487 + a = kzalloc(sizeof(*a), GFP_NOFS);
16488 + if (unlikely(!a))
16491 + if (ia->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
16492 + ia->ia_valid &= ~ATTR_MODE;
16495 + sb = dentry->d_sb;
16496 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16497 + if (unlikely(err))
16500 + if (ia->ia_valid & ATTR_FILE) {
16501 + /* currently ftruncate(2) only */
16502 + AuDebugOn(!S_ISREG(inode->i_mode));
16503 + file = ia->ia_file;
16504 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/1);
16505 + if (unlikely(err))
16507 + ia->ia_file = au_hf_top(file);
16508 + a->udba = AuOpt_UDBA_NONE;
16510 + /* fchmod() doesn't pass ia_file */
16511 + a->udba = au_opt_udba(sb);
16512 + di_write_lock_child(dentry);
16513 + /* no d_unlinked(), to set UDBA_NONE for root */
16514 + if (d_unhashed(dentry))
16515 + a->udba = AuOpt_UDBA_NONE;
16516 + if (a->udba != AuOpt_UDBA_NONE) {
16517 + AuDebugOn(IS_ROOT(dentry));
16518 + err = au_reval_for_attr(dentry, au_sigen(sb));
16519 + if (unlikely(err))
16524 + err = au_pin_and_icpup(dentry, ia, a);
16525 + if (unlikely(err < 0))
16527 + if (au_ftest_icpup(a->flags, DID_CPUP)) {
16528 + ia->ia_file = NULL;
16529 + ia->ia_valid &= ~ATTR_FILE;
16532 + a->h_path.mnt = au_sbr_mnt(sb, a->btgt);
16533 + if ((ia->ia_valid & (ATTR_MODE | ATTR_CTIME))
16534 + == (ATTR_MODE | ATTR_CTIME)) {
16535 + err = security_path_chmod(a->h_path.dentry, a->h_path.mnt,
16537 + if (unlikely(err))
16539 + } else if ((ia->ia_valid & (ATTR_UID | ATTR_GID))
16540 + && (ia->ia_valid & ATTR_CTIME)) {
16541 + err = security_path_chown(&a->h_path, ia->ia_uid, ia->ia_gid);
16542 + if (unlikely(err))
16546 + if (ia->ia_valid & ATTR_SIZE) {
16549 + if (ia->ia_size < i_size_read(inode))
16551 + truncate_setsize(inode, ia->ia_size);
16554 + if (ia->ia_valid & ATTR_FILE)
16556 + mutex_unlock(&a->h_inode->i_mutex);
16557 + err = vfsub_trunc(&a->h_path, ia->ia_size, ia->ia_valid, f);
16558 + mutex_lock_nested(&a->h_inode->i_mutex, AuLsc_I_CHILD);
16560 + err = vfsub_notify_change(&a->h_path, ia);
16562 + au_cpup_attr_changeable(inode);
16565 + mutex_unlock(&a->h_inode->i_mutex);
16566 + au_unpin(&a->pin);
16567 + if (unlikely(err))
16568 + au_update_dbstart(dentry);
16570 + di_write_unlock(dentry);
16572 + fi_write_unlock(file);
16573 + ia->ia_file = file;
16574 + ia->ia_valid |= ATTR_FILE;
16577 + si_read_unlock(sb);
16585 +static void au_refresh_iattr(struct inode *inode, struct kstat *st,
16586 + unsigned int nlink)
16590 + inode->i_mode = st->mode;
16591 + inode->i_uid = st->uid;
16592 + inode->i_gid = st->gid;
16593 + inode->i_atime = st->atime;
16594 + inode->i_mtime = st->mtime;
16595 + inode->i_ctime = st->ctime;
16597 + au_cpup_attr_nlink(inode, /*force*/0);
16598 + if (S_ISDIR(inode->i_mode)) {
16599 + n = inode->i_nlink;
16602 + set_nlink(inode, n);
16605 + spin_lock(&inode->i_lock);
16606 + inode->i_blocks = st->blocks;
16607 + i_size_write(inode, st->size);
16608 + spin_unlock(&inode->i_lock);
16611 +static int aufs_getattr(struct vfsmount *mnt __maybe_unused,
16612 + struct dentry *dentry, struct kstat *st)
16615 + unsigned int mnt_flags;
16616 + aufs_bindex_t bindex;
16617 + unsigned char udba_none, positive;
16618 + struct super_block *sb, *h_sb;
16619 + struct inode *inode;
16620 + struct vfsmount *h_mnt;
16621 + struct dentry *h_dentry;
16623 + sb = dentry->d_sb;
16624 + inode = dentry->d_inode;
16625 + err = si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
16626 + if (unlikely(err))
16628 + mnt_flags = au_mntflags(sb);
16629 + udba_none = !!au_opt_test(mnt_flags, UDBA_NONE);
16631 + /* support fstat(2) */
16632 + if (!d_unlinked(dentry) && !udba_none) {
16633 + unsigned int sigen = au_sigen(sb);
16634 + err = au_digen_test(dentry, sigen);
16636 + di_read_lock_child(dentry, AuLock_IR);
16637 + err = au_dbrange_test(dentry);
16638 + if (unlikely(err))
16641 + AuDebugOn(IS_ROOT(dentry));
16642 + di_write_lock_child(dentry);
16643 + err = au_dbrange_test(dentry);
16645 + err = au_reval_for_attr(dentry, sigen);
16646 + di_downgrade_lock(dentry, AuLock_IR);
16647 + if (unlikely(err))
16651 + di_read_lock_child(dentry, AuLock_IR);
16653 + bindex = au_ibstart(inode);
16654 + h_mnt = au_sbr_mnt(sb, bindex);
16655 + h_sb = h_mnt->mnt_sb;
16656 + if (!au_test_fs_bad_iattr(h_sb) && udba_none)
16657 + goto out_fill; /* success */
16660 + if (au_dbstart(dentry) == bindex)
16661 + h_dentry = dget(au_h_dptr(dentry, bindex));
16662 + else if (au_opt_test(mnt_flags, PLINK) && au_plink_test(inode)) {
16663 + h_dentry = au_plink_lkup(inode, bindex);
16664 + if (IS_ERR(h_dentry))
16665 + goto out_fill; /* pretending success */
16667 + /* illegally overlapped or something */
16668 + if (unlikely(!h_dentry))
16669 + goto out_fill; /* pretending success */
16671 + positive = !!h_dentry->d_inode;
16673 + err = vfs_getattr(h_mnt, h_dentry, st);
16677 + au_refresh_iattr(inode, st, h_dentry->d_inode->i_nlink);
16678 + goto out_fill; /* success */
16684 + generic_fillattr(inode, st);
16686 + di_read_unlock(dentry, AuLock_IR);
16687 + si_read_unlock(sb);
16693 +/* ---------------------------------------------------------------------- */
16695 +static int h_readlink(struct dentry *dentry, int bindex, char __user *buf,
16699 + struct super_block *sb;
16700 + struct dentry *h_dentry;
16703 + h_dentry = au_h_dptr(dentry, bindex);
16704 + if (unlikely(!h_dentry->d_inode->i_op->readlink))
16707 + err = security_inode_readlink(h_dentry);
16708 + if (unlikely(err))
16711 + sb = dentry->d_sb;
16712 + if (!au_test_ro(sb, bindex, dentry->d_inode)) {
16713 + vfsub_touch_atime(au_sbr_mnt(sb, bindex), h_dentry);
16714 + fsstack_copy_attr_atime(dentry->d_inode, h_dentry->d_inode);
16716 + err = h_dentry->d_inode->i_op->readlink(h_dentry, buf, bufsiz);
16722 +static int aufs_readlink(struct dentry *dentry, char __user *buf, int bufsiz)
16726 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
16727 + if (unlikely(err))
16729 + err = au_d_hashed_positive(dentry);
16731 + err = h_readlink(dentry, au_dbstart(dentry), buf, bufsiz);
16732 + aufs_read_unlock(dentry, AuLock_IR);
16738 +static void *aufs_follow_link(struct dentry *dentry, struct nameidata *nd)
16741 + mm_segment_t old_fs;
16748 + buf.k = __getname_gfp(GFP_NOFS);
16749 + if (unlikely(!buf.k))
16752 + err = aufs_read_lock(dentry, AuLock_IR | AuLock_GEN);
16753 + if (unlikely(err))
16756 + err = au_d_hashed_positive(dentry);
16758 + old_fs = get_fs();
16759 + set_fs(KERNEL_DS);
16760 + err = h_readlink(dentry, au_dbstart(dentry), buf.u, PATH_MAX);
16763 + aufs_read_unlock(dentry, AuLock_IR);
16767 + /* will be freed by put_link */
16768 + nd_set_link(nd, buf.k);
16769 + return NULL; /* success */
16773 + __putname(buf.k);
16775 + path_put(&nd->path);
16777 + return ERR_PTR(err);
16780 +static void aufs_put_link(struct dentry *dentry __maybe_unused,
16781 + struct nameidata *nd, void *cookie __maybe_unused)
16783 + __putname(nd_get_link(nd));
16786 +/* ---------------------------------------------------------------------- */
16788 +static void aufs_truncate_range(struct inode *inode __maybe_unused,
16789 + loff_t start __maybe_unused,
16790 + loff_t end __maybe_unused)
16795 +/* ---------------------------------------------------------------------- */
16797 +struct inode_operations aufs_symlink_iop = {
16798 + .permission = aufs_permission,
16799 + .setattr = aufs_setattr,
16800 + .getattr = aufs_getattr,
16801 + .readlink = aufs_readlink,
16802 + .follow_link = aufs_follow_link,
16803 + .put_link = aufs_put_link
16806 +struct inode_operations aufs_dir_iop = {
16807 + .create = aufs_create,
16808 + .lookup = aufs_lookup,
16809 + .link = aufs_link,
16810 + .unlink = aufs_unlink,
16811 + .symlink = aufs_symlink,
16812 + .mkdir = aufs_mkdir,
16813 + .rmdir = aufs_rmdir,
16814 + .mknod = aufs_mknod,
16815 + .rename = aufs_rename,
16817 + .permission = aufs_permission,
16818 + .setattr = aufs_setattr,
16819 + .getattr = aufs_getattr
16822 +struct inode_operations aufs_iop = {
16823 + .permission = aufs_permission,
16824 + .setattr = aufs_setattr,
16825 + .getattr = aufs_getattr,
16826 + .truncate_range = aufs_truncate_range
16828 diff -urN /usr/share/empty/fs/aufs/i_op_del.c linux/fs/aufs/i_op_del.c
16829 --- /usr/share/empty/fs/aufs/i_op_del.c 1970-01-01 01:00:00.000000000 +0100
16830 +++ linux/fs/aufs/i_op_del.c 2011-08-24 13:30:24.731313534 +0200
16833 + * Copyright (C) 2005-2011 Junjiro R. Okajima
16835 + * This program, aufs is free software; you can redistribute it and/or modify
16836 + * it under the terms of the GNU General Public License as published by
16837 + * the Free Software Foundation; either version 2 of the License, or
16838 + * (at your option) any later version.
16840 + * This program is distributed in the hope that it will be useful,
16841 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
16842 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16843 + * GNU General Public License for more details.
16845 + * You should have received a copy of the GNU General Public License
16846 + * along with this program; if not, write to the Free Software
16847 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
16851 + * inode operations (del entry)
16857 + * decide if a new whiteout for @dentry is necessary or not.
16858 + * when it is necessary, prepare the parent dir for the upper branch whose
16859 + * branch index is @bcpup for creation. the actual creation of the whiteout will
16860 + * be done by caller.
16862 + * 0: wh is unnecessary
16863 + * plus: wh is necessary
16866 +int au_wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup)
16868 + int need_wh, err;
16869 + aufs_bindex_t bstart;
16870 + struct super_block *sb;
16872 + sb = dentry->d_sb;
16873 + bstart = au_dbstart(dentry);
16874 + if (*bcpup < 0) {
16876 + if (au_test_ro(sb, bstart, dentry->d_inode)) {
16877 + err = AuWbrCopyup(au_sbi(sb), dentry);
16879 + if (unlikely(err < 0))
16883 + AuDebugOn(bstart < *bcpup
16884 + || au_test_ro(sb, *bcpup, dentry->d_inode));
16885 + AuDbg("bcpup %d, bstart %d\n", *bcpup, bstart);
16887 + if (*bcpup != bstart) {
16888 + err = au_cpup_dirs(dentry, *bcpup);
16889 + if (unlikely(err))
16893 + struct au_dinfo *dinfo, *tmp;
16895 + need_wh = -ENOMEM;
16896 + dinfo = au_di(dentry);
16897 + tmp = au_di_alloc(sb, AuLsc_DI_TMP);
16899 + au_di_cp(tmp, dinfo);
16900 + au_di_swap(tmp, dinfo);
16901 + /* returns the number of positive dentries */
16902 + need_wh = au_lkup_dentry(dentry, bstart + 1, /*type*/0,
16904 + au_di_swap(tmp, dinfo);
16905 + au_rw_write_unlock(&tmp->di_rwsem);
16909 + AuDbg("need_wh %d\n", need_wh);
16917 + * simple tests for the del-entry operations.
16918 + * following the checks in vfs, plus the parent-child relationship.
16920 +int au_may_del(struct dentry *dentry, aufs_bindex_t bindex,
16921 + struct dentry *h_parent, int isdir)
16925 + struct dentry *h_dentry, *h_latest;
16926 + struct inode *h_inode;
16928 + h_dentry = au_h_dptr(dentry, bindex);
16929 + h_inode = h_dentry->d_inode;
16930 + if (dentry->d_inode) {
16932 + if (unlikely(!h_inode || !h_inode->i_nlink))
16935 + h_mode = h_inode->i_mode;
16938 + if (unlikely(S_ISDIR(h_mode)))
16940 + } else if (unlikely(!S_ISDIR(h_mode))) {
16945 + /* rename(2) case */
16947 + if (unlikely(h_inode))
16952 + /* expected parent dir is locked */
16953 + if (unlikely(h_parent != h_dentry->d_parent))
16958 + * rmdir a dir may break the consistency on some filesystem.
16959 + * let's try heavy test.
16962 + if (unlikely(au_test_h_perm(h_parent->d_inode, MAY_EXEC | MAY_WRITE)))
16965 + h_latest = au_sio_lkup_one(&dentry->d_name, h_parent,
16966 + au_sbr(dentry->d_sb, bindex));
16968 + if (IS_ERR(h_latest))
16970 + if (h_latest == h_dentry)
16979 + * decide the branch where we operate for @dentry. the branch index will be set
16980 + * @rbcpup. after diciding it, 'pin' it and store the timestamps of the parent
16981 + * dir for reverting.
16982 + * when a new whiteout is necessary, create it.
16984 +static struct dentry*
16985 +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *rbcpup,
16986 + struct au_dtime *dt, struct au_pin *pin)
16988 + struct dentry *wh_dentry;
16989 + struct super_block *sb;
16990 + struct path h_path;
16991 + int err, need_wh;
16992 + unsigned int udba;
16993 + aufs_bindex_t bcpup;
16995 + need_wh = au_wr_dir_need_wh(dentry, isdir, rbcpup);
16996 + wh_dentry = ERR_PTR(need_wh);
16997 + if (unlikely(need_wh < 0))
17000 + sb = dentry->d_sb;
17001 + udba = au_opt_udba(sb);
17003 + err = au_pin(pin, dentry, bcpup, udba,
17004 + AuPin_DI_LOCKED | AuPin_MNT_WRITE);
17005 + wh_dentry = ERR_PTR(err);
17006 + if (unlikely(err))
17009 + h_path.dentry = au_pinned_h_parent(pin);
17010 + if (udba != AuOpt_UDBA_NONE
17011 + && au_dbstart(dentry) == bcpup) {
17012 + err = au_may_del(dentry, bcpup, h_path.dentry, isdir);
17013 + wh_dentry = ERR_PTR(err);
17014 + if (unlikely(err))
17018 + h_path.mnt = au_sbr_mnt(sb, bcpup);
17019 + au_dtime_store(dt, au_pinned_parent(pin), &h_path);
17020 + wh_dentry = NULL;
17022 + goto out; /* success, no need to create whiteout */
17024 + wh_dentry = au_wh_create(dentry, bcpup, h_path.dentry);
17025 + if (IS_ERR(wh_dentry))
17028 + /* returns with the parent is locked and wh_dentry is dget-ed */
17029 + goto out; /* success */
17034 + return wh_dentry;
17038 + * when removing a dir, rename it to a unique temporary whiteout-ed name first
17039 + * in order to be revertible and save time for removing many child whiteouts
17041 + * returns 1 when there are too many child whiteout and caller should remove
17042 + * them asynchronously. returns 0 when the number of children is enough small to
17043 + * remove now or the branch fs is a remote fs.
17044 + * otherwise return an error.
17046 +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex,
17047 + struct au_nhash *whlist, struct inode *dir)
17049 + int rmdir_later, err, dirwh;
17050 + struct dentry *h_dentry;
17051 + struct super_block *sb;
17053 + sb = dentry->d_sb;
17054 + SiMustAnyLock(sb);
17055 + h_dentry = au_h_dptr(dentry, bindex);
17056 + err = au_whtmp_ren(h_dentry, au_sbr(sb, bindex));
17057 + if (unlikely(err))
17060 + /* stop monitoring */
17061 + au_hn_free(au_hi(dentry->d_inode, bindex));
17063 + if (!au_test_fs_remote(h_dentry->d_sb)) {
17064 + dirwh = au_sbi(sb)->si_dirwh;
17065 + rmdir_later = (dirwh <= 1);
17066 + if (!rmdir_later)
17067 + rmdir_later = au_nhash_test_longer_wh(whlist, bindex,
17070 + return rmdir_later;
17073 + err = au_whtmp_rmdir(dir, bindex, h_dentry, whlist);
17074 + if (unlikely(err)) {
17075 + AuIOErr("rmdir %.*s, b%d failed, %d. ignored\n",
17076 + AuDLNPair(h_dentry), bindex, err);
17086 + * final procedure for deleting a entry.
17087 + * maintain dentry and iattr.
17089 +static void epilog(struct inode *dir, struct dentry *dentry,
17090 + aufs_bindex_t bindex)
17092 + struct inode *inode;
17094 + inode = dentry->d_inode;
17096 + inode->i_ctime = dir->i_ctime;
17098 + if (au_ibstart(dir) == bindex)
17099 + au_cpup_attr_timesizes(dir);
17100 + dir->i_version++;
17104 + * when an error happened, remove the created whiteout and revert everything.
17106 +static int do_revert(int err, struct inode *dir, aufs_bindex_t bindex,
17107 + aufs_bindex_t bwh, struct dentry *wh_dentry,
17108 + struct dentry *dentry, struct au_dtime *dt)
17111 + struct path h_path = {
17112 + .dentry = wh_dentry,
17113 + .mnt = au_sbr_mnt(dir->i_sb, bindex)
17116 + rerr = au_wh_unlink_dentry(au_h_iptr(dir, bindex), &h_path, dentry);
17118 + au_set_dbwh(dentry, bwh);
17119 + au_dtime_revert(dt);
17123 + AuIOErr("%.*s reverting whiteout failed(%d, %d)\n",
17124 + AuDLNPair(dentry), err, rerr);
17128 +/* ---------------------------------------------------------------------- */
17130 +int aufs_unlink(struct inode *dir, struct dentry *dentry)
17133 + aufs_bindex_t bwh, bindex, bstart;
17134 + struct au_dtime dt;
17135 + struct au_pin pin;
17136 + struct path h_path;
17137 + struct inode *inode, *h_dir;
17138 + struct dentry *parent, *wh_dentry;
17142 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_GEN);
17143 + if (unlikely(err))
17145 + err = au_d_hashed_positive(dentry);
17146 + if (unlikely(err))
17148 + inode = dentry->d_inode;
17149 + IMustLock(inode);
17151 + if (unlikely(S_ISDIR(inode->i_mode)))
17152 + goto out_unlock; /* possible? */
17154 + bstart = au_dbstart(dentry);
17155 + bwh = au_dbwh(dentry);
17157 + parent = dentry->d_parent; /* dir inode is locked */
17158 + di_write_lock_parent(parent);
17159 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &dt, &pin);
17160 + err = PTR_ERR(wh_dentry);
17161 + if (IS_ERR(wh_dentry))
17164 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bstart);
17165 + h_path.dentry = au_h_dptr(dentry, bstart);
17166 + dget(h_path.dentry);
17167 + if (bindex == bstart) {
17168 + h_dir = au_pinned_h_dir(&pin);
17169 + err = vfsub_unlink(h_dir, &h_path, /*force*/0);
17171 + /* dir inode is locked */
17172 + h_dir = wh_dentry->d_parent->d_inode;
17173 + IMustLock(h_dir);
17178 + vfsub_drop_nlink(inode);
17179 + epilog(dir, dentry, bindex);
17181 + /* update target timestamps */
17182 + if (bindex == bstart) {
17183 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
17184 + inode->i_ctime = h_path.dentry->d_inode->i_ctime;
17186 + /* todo: this timestamp may be reverted later */
17187 + inode->i_ctime = h_dir->i_ctime;
17188 + goto out_unpin; /* success */
17195 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry, &dt);
17203 + dput(h_path.dentry);
17205 + di_write_unlock(parent);
17207 + aufs_read_unlock(dentry, AuLock_DW);
17212 +int aufs_rmdir(struct inode *dir, struct dentry *dentry)
17214 + int err, rmdir_later;
17215 + aufs_bindex_t bwh, bindex, bstart;
17216 + struct au_dtime dt;
17217 + struct au_pin pin;
17218 + struct inode *inode;
17219 + struct dentry *parent, *wh_dentry, *h_dentry;
17220 + struct au_whtmp_rmdir *args;
17224 + err = aufs_read_lock(dentry, AuLock_DW | AuLock_FLUSH | AuLock_GEN);
17225 + if (unlikely(err))
17227 + err = au_alive_dir(dentry);
17228 + if (unlikely(err))
17230 + inode = dentry->d_inode;
17231 + IMustLock(inode);
17233 + if (unlikely(!S_ISDIR(inode->i_mode)))
17234 + goto out_unlock; /* possible? */
17237 + args = au_whtmp_rmdir_alloc(dir->i_sb, GFP_NOFS);
17238 + if (unlikely(!args))
17241 + parent = dentry->d_parent; /* dir inode is locked */
17242 + di_write_lock_parent(parent);
17243 + err = au_test_empty(dentry, &args->whlist);
17244 + if (unlikely(err))
17247 + bstart = au_dbstart(dentry);
17248 + bwh = au_dbwh(dentry);
17250 + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/1, &bindex, &dt, &pin);
17251 + err = PTR_ERR(wh_dentry);
17252 + if (IS_ERR(wh_dentry))
17255 + h_dentry = au_h_dptr(dentry, bstart);
17258 + if (bindex == bstart) {
17259 + err = renwh_and_rmdir(dentry, bstart, &args->whlist, dir);
17261 + rmdir_later = err;
17265 + /* stop monitoring */
17266 + au_hn_free(au_hi(inode, bstart));
17268 + /* dir inode is locked */
17269 + IMustLock(wh_dentry->d_parent->d_inode);
17274 + vfsub_dead_dir(inode);
17275 + au_set_dbdiropq(dentry, -1);
17276 + epilog(dir, dentry, bindex);
17278 + if (rmdir_later) {
17279 + au_whtmp_kick_rmdir(dir, bstart, h_dentry, args);
17283 + goto out_unpin; /* success */
17291 + rerr = do_revert(err, dir, bindex, bwh, wh_dentry, dentry, &dt);
17301 + di_write_unlock(parent);
17303 + au_whtmp_rmdir_free(args);
17305 + aufs_read_unlock(dentry, AuLock_DW);
17310 diff -urN /usr/share/empty/fs/aufs/i_op_ren.c linux/fs/aufs/i_op_ren.c
17311 --- /usr/share/empty/fs/aufs/i_op_ren.c 1970-01-01 01:00:00.000000000 +0100
17312 +++ linux/fs/aufs/i_op_ren.c 2011-08-24 13:30:24.734646739 +0200
17315 + * Copyright (C) 2005-2011 Junjiro R. Okajima
17317 + * This program, aufs is free software; you can redistribute it and/or modify
17318 + * it under the terms of the GNU General Public License as published by
17319 + * the Free Software Foundation; either version 2 of the License, or
17320 + * (at your option) any later version.
17322 + * This program is distributed in the hope that it will be useful,
17323 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
17324 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17325 + * GNU General Public License for more details.
17327 + * You should have received a copy of the GNU General Public License
17328 + * along with this program; if not, write to the Free Software
17329 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17333 + * inode operation (rename entry)
17334 + * todo: this is crazy monster
17339 +enum { AuSRC, AuDST, AuSrcDst };
17340 +enum { AuPARENT, AuCHILD, AuParentChild };
17342 +#define AuRen_ISDIR 1
17343 +#define AuRen_ISSAMEDIR (1 << 1)
17344 +#define AuRen_WHSRC (1 << 2)
17345 +#define AuRen_WHDST (1 << 3)
17346 +#define AuRen_MNT_WRITE (1 << 4)
17347 +#define AuRen_DT_DSTDIR (1 << 5)
17348 +#define AuRen_DIROPQ (1 << 6)
17349 +#define AuRen_CPUP (1 << 7)
17350 +#define au_ftest_ren(flags, name) ((flags) & AuRen_##name)
17351 +#define au_fset_ren(flags, name) \
17352 + do { (flags) |= AuRen_##name; } while (0)
17353 +#define au_fclr_ren(flags, name) \
17354 + do { (flags) &= ~AuRen_##name; } while (0)
17356 +struct au_ren_args {
17358 + struct dentry *dentry, *h_dentry, *parent, *h_parent,
17360 + struct inode *dir, *inode;
17361 + struct au_hinode *hdir;
17362 + struct au_dtime dt[AuParentChild];
17363 + aufs_bindex_t bstart;
17366 +#define src_dentry sd[AuSRC].dentry
17367 +#define src_dir sd[AuSRC].dir
17368 +#define src_inode sd[AuSRC].inode
17369 +#define src_h_dentry sd[AuSRC].h_dentry
17370 +#define src_parent sd[AuSRC].parent
17371 +#define src_h_parent sd[AuSRC].h_parent
17372 +#define src_wh_dentry sd[AuSRC].wh_dentry
17373 +#define src_hdir sd[AuSRC].hdir
17374 +#define src_h_dir sd[AuSRC].hdir->hi_inode
17375 +#define src_dt sd[AuSRC].dt
17376 +#define src_bstart sd[AuSRC].bstart
17378 +#define dst_dentry sd[AuDST].dentry
17379 +#define dst_dir sd[AuDST].dir
17380 +#define dst_inode sd[AuDST].inode
17381 +#define dst_h_dentry sd[AuDST].h_dentry
17382 +#define dst_parent sd[AuDST].parent
17383 +#define dst_h_parent sd[AuDST].h_parent
17384 +#define dst_wh_dentry sd[AuDST].wh_dentry
17385 +#define dst_hdir sd[AuDST].hdir
17386 +#define dst_h_dir sd[AuDST].hdir->hi_inode
17387 +#define dst_dt sd[AuDST].dt
17388 +#define dst_bstart sd[AuDST].bstart
17390 + struct dentry *h_trap;
17391 + struct au_branch *br;
17392 + struct au_hinode *src_hinode;
17393 + struct path h_path;
17394 + struct au_nhash whlist;
17395 + aufs_bindex_t btgt, src_bwh, src_bdiropq;
17397 + unsigned int flags;
17399 + struct au_whtmp_rmdir *thargs;
17400 + struct dentry *h_dst;
17403 +/* ---------------------------------------------------------------------- */
17406 + * functions for reverting.
17407 + * when an error happened in a single rename systemcall, we should revert
17408 + * everything as if nothing happend.
17409 + * we don't need to revert the copied-up/down the parent dir since they are
17413 +#define RevertFailure(fmt, ...) do { \
17414 + AuIOErr("revert failure: " fmt " (%d, %d)\n", \
17415 + ##__VA_ARGS__, err, rerr); \
17419 +static void au_ren_rev_diropq(int err, struct au_ren_args *a)
17423 + au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
17424 + rerr = au_diropq_remove(a->src_dentry, a->btgt);
17425 + au_hn_imtx_unlock(a->src_hinode);
17426 + au_set_dbdiropq(a->src_dentry, a->src_bdiropq);
17428 + RevertFailure("remove diropq %.*s", AuDLNPair(a->src_dentry));
17431 +static void au_ren_rev_rename(int err, struct au_ren_args *a)
17435 + a->h_path.dentry = au_lkup_one(&a->src_dentry->d_name, a->src_h_parent,
17436 + a->br, /*nd*/NULL);
17437 + rerr = PTR_ERR(a->h_path.dentry);
17438 + if (IS_ERR(a->h_path.dentry)) {
17439 + RevertFailure("au_lkup_one %.*s", AuDLNPair(a->src_dentry));
17443 + rerr = vfsub_rename(a->dst_h_dir,
17444 + au_h_dptr(a->src_dentry, a->btgt),
17445 + a->src_h_dir, &a->h_path);
17446 + d_drop(a->h_path.dentry);
17447 + dput(a->h_path.dentry);
17448 + /* au_set_h_dptr(a->src_dentry, a->btgt, NULL); */
17450 + RevertFailure("rename %.*s", AuDLNPair(a->src_dentry));
17453 +static void au_ren_rev_cpup(int err, struct au_ren_args *a)
17457 + a->h_path.dentry = a->dst_h_dentry;
17458 + rerr = vfsub_unlink(a->dst_h_dir, &a->h_path, /*force*/0);
17459 + au_set_h_dptr(a->src_dentry, a->btgt, NULL);
17460 + au_set_dbstart(a->src_dentry, a->src_bstart);
17462 + RevertFailure("unlink %.*s", AuDLNPair(a->dst_h_dentry));
17465 +static void au_ren_rev_whtmp(int err, struct au_ren_args *a)
17469 + a->h_path.dentry = au_lkup_one(&a->dst_dentry->d_name, a->dst_h_parent,
17470 + a->br, /*nd*/NULL);
17471 + rerr = PTR_ERR(a->h_path.dentry);
17472 + if (IS_ERR(a->h_path.dentry)) {
17473 + RevertFailure("lookup %.*s", AuDLNPair(a->dst_dentry));
17476 + if (a->h_path.dentry->d_inode) {
17477 + d_drop(a->h_path.dentry);
17478 + dput(a->h_path.dentry);
17482 + rerr = vfsub_rename(a->dst_h_dir, a->h_dst, a->dst_h_dir, &a->h_path);
17483 + d_drop(a->h_path.dentry);
17484 + dput(a->h_path.dentry);
17486 + au_set_h_dptr(a->dst_dentry, a->btgt, dget(a->h_dst));
17488 + RevertFailure("rename %.*s", AuDLNPair(a->h_dst));
17491 +static void au_ren_rev_whsrc(int err, struct au_ren_args *a)
17495 + a->h_path.dentry = a->src_wh_dentry;
17496 + rerr = au_wh_unlink_dentry(a->src_h_dir, &a->h_path, a->src_dentry);
17497 + au_set_dbwh(a->src_dentry, a->src_bwh);
17499 + RevertFailure("unlink %.*s", AuDLNPair(a->src_wh_dentry));
17501 +#undef RevertFailure
17503 +/* ---------------------------------------------------------------------- */
17506 + * when we have to copyup the renaming entry, do it with the rename-target name
17507 + * in order to minimize the cost (the later actual rename is unnecessary).
17508 + * otherwise rename it on the target branch.
17510 +static int au_ren_or_cpup(struct au_ren_args *a)
17513 + struct dentry *d;
17515 + d = a->src_dentry;
17516 + if (au_dbstart(d) == a->btgt) {
17517 + a->h_path.dentry = a->dst_h_dentry;
17518 + if (au_ftest_ren(a->flags, DIROPQ)
17519 + && au_dbdiropq(d) == a->btgt)
17520 + au_fclr_ren(a->flags, DIROPQ);
17521 + AuDebugOn(au_dbstart(d) != a->btgt);
17522 + err = vfsub_rename(a->src_h_dir, au_h_dptr(d, a->btgt),
17523 + a->dst_h_dir, &a->h_path);
17525 + struct mutex *h_mtx = &a->src_h_dentry->d_inode->i_mutex;
17526 + struct file *h_file;
17528 + au_fset_ren(a->flags, CPUP);
17529 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
17530 + au_set_dbstart(d, a->btgt);
17531 + au_set_h_dptr(d, a->btgt, dget(a->dst_h_dentry));
17532 + h_file = au_h_open_pre(d, a->src_bstart);
17533 + if (IS_ERR(h_file)) {
17534 + err = PTR_ERR(h_file);
17537 + err = au_sio_cpup_single(d, a->btgt, a->src_bstart, -1,
17538 + !AuCpup_DTIME, a->dst_parent);
17539 + mutex_unlock(h_mtx);
17540 + au_h_open_post(d, a->src_bstart, h_file);
17542 + d = a->dst_dentry;
17543 + au_set_h_dptr(d, a->btgt, NULL);
17544 + au_update_dbstart(d);
17546 + au_set_h_dptr(d, a->btgt, NULL);
17547 + au_set_dbstart(d, a->src_bstart);
17550 + if (!err && a->h_dst)
17551 + /* it will be set to dinfo later */
17557 +/* cf. aufs_rmdir() */
17558 +static int au_ren_del_whtmp(struct au_ren_args *a)
17561 + struct inode *dir;
17563 + dir = a->dst_dir;
17564 + SiMustAnyLock(dir->i_sb);
17565 + if (!au_nhash_test_longer_wh(&a->whlist, a->btgt,
17566 + au_sbi(dir->i_sb)->si_dirwh)
17567 + || au_test_fs_remote(a->h_dst->d_sb)) {
17568 + err = au_whtmp_rmdir(dir, a->btgt, a->h_dst, &a->whlist);
17569 + if (unlikely(err))
17570 + pr_warning("failed removing whtmp dir %.*s (%d), "
17571 + "ignored.\n", AuDLNPair(a->h_dst), err);
17573 + au_nhash_wh_free(&a->thargs->whlist);
17574 + a->thargs->whlist = a->whlist;
17575 + a->whlist.nh_num = 0;
17576 + au_whtmp_kick_rmdir(dir, a->btgt, a->h_dst, a->thargs);
17578 + a->thargs = NULL;
17584 +/* make it 'opaque' dir. */
17585 +static int au_ren_diropq(struct au_ren_args *a)
17588 + struct dentry *diropq;
17591 + a->src_bdiropq = au_dbdiropq(a->src_dentry);
17592 + a->src_hinode = au_hi(a->src_inode, a->btgt);
17593 + au_hn_imtx_lock_nested(a->src_hinode, AuLsc_I_CHILD);
17594 + diropq = au_diropq_create(a->src_dentry, a->btgt);
17595 + au_hn_imtx_unlock(a->src_hinode);
17596 + if (IS_ERR(diropq))
17597 + err = PTR_ERR(diropq);
17603 +static int do_rename(struct au_ren_args *a)
17606 + struct dentry *d, *h_d;
17608 + /* prepare workqueue args for asynchronous rmdir */
17609 + h_d = a->dst_h_dentry;
17610 + if (au_ftest_ren(a->flags, ISDIR) && h_d->d_inode) {
17612 + a->thargs = au_whtmp_rmdir_alloc(a->src_dentry->d_sb, GFP_NOFS);
17613 + if (unlikely(!a->thargs))
17615 + a->h_dst = dget(h_d);
17618 + /* create whiteout for src_dentry */
17619 + if (au_ftest_ren(a->flags, WHSRC)) {
17620 + a->src_bwh = au_dbwh(a->src_dentry);
17621 + AuDebugOn(a->src_bwh >= 0);
17623 + = au_wh_create(a->src_dentry, a->btgt, a->src_h_parent);
17624 + err = PTR_ERR(a->src_wh_dentry);
17625 + if (IS_ERR(a->src_wh_dentry))
17629 + /* lookup whiteout for dentry */
17630 + if (au_ftest_ren(a->flags, WHDST)) {
17631 + h_d = au_wh_lkup(a->dst_h_parent, &a->dst_dentry->d_name,
17633 + err = PTR_ERR(h_d);
17636 + if (!h_d->d_inode)
17639 + a->dst_wh_dentry = h_d;
17642 + /* rename dentry to tmpwh */
17644 + err = au_whtmp_ren(a->dst_h_dentry, a->br);
17645 + if (unlikely(err))
17648 + d = a->dst_dentry;
17649 + au_set_h_dptr(d, a->btgt, NULL);
17650 + err = au_lkup_neg(d, a->btgt);
17651 + if (unlikely(err))
17653 + a->dst_h_dentry = au_h_dptr(d, a->btgt);
17657 + if (a->dst_h_dentry->d_inode && a->src_bstart != a->btgt) {
17658 + struct mutex *h_mtx = &a->src_h_dentry->d_inode->i_mutex;
17659 + struct file *h_file;
17661 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
17662 + AuDebugOn(au_dbstart(a->src_dentry) != a->src_bstart);
17663 + h_file = au_h_open_pre(a->src_dentry, a->src_bstart);
17664 + if (IS_ERR(h_file)) {
17665 + err = PTR_ERR(h_file);
17668 + err = au_sio_cpup_simple(a->src_dentry, a->btgt, -1,
17670 + mutex_unlock(h_mtx);
17671 + au_h_open_post(a->src_dentry, a->src_bstart, h_file);
17672 + if (unlikely(err))
17676 + /* rename by vfs_rename or cpup */
17677 + d = a->dst_dentry;
17678 + if (au_ftest_ren(a->flags, ISDIR)
17679 + && (a->dst_wh_dentry
17680 + || au_dbdiropq(d) == a->btgt
17681 + /* hide the lower to keep xino */
17682 + || a->btgt < au_dbend(d)
17683 + || au_opt_test(au_mntflags(d->d_sb), ALWAYS_DIROPQ)))
17684 + au_fset_ren(a->flags, DIROPQ);
17685 + err = au_ren_or_cpup(a);
17686 + if (unlikely(err))
17687 + /* leave the copied-up one */
17690 + /* make dir opaque */
17691 + if (au_ftest_ren(a->flags, DIROPQ)) {
17692 + err = au_ren_diropq(a);
17693 + if (unlikely(err))
17697 + /* update target timestamps */
17698 + AuDebugOn(au_dbstart(a->src_dentry) != a->btgt);
17699 + a->h_path.dentry = au_h_dptr(a->src_dentry, a->btgt);
17700 + vfsub_update_h_iattr(&a->h_path, /*did*/NULL); /*ignore*/
17701 + a->src_inode->i_ctime = a->h_path.dentry->d_inode->i_ctime;
17703 + /* remove whiteout for dentry */
17704 + if (a->dst_wh_dentry) {
17705 + a->h_path.dentry = a->dst_wh_dentry;
17706 + err = au_wh_unlink_dentry(a->dst_h_dir, &a->h_path,
17708 + if (unlikely(err))
17712 + /* remove whtmp */
17714 + au_ren_del_whtmp(a); /* ignore this error */
17717 + goto out_success;
17720 + if (au_ftest_ren(a->flags, DIROPQ))
17721 + au_ren_rev_diropq(err, a);
17723 + if (!au_ftest_ren(a->flags, CPUP))
17724 + au_ren_rev_rename(err, a);
17726 + au_ren_rev_cpup(err, a);
17730 + au_ren_rev_whtmp(err, a);
17732 + dput(a->dst_wh_dentry);
17733 + a->dst_wh_dentry = NULL;
17735 + if (a->src_wh_dentry)
17736 + au_ren_rev_whsrc(err, a);
17738 + dput(a->src_wh_dentry);
17739 + dput(a->dst_wh_dentry);
17743 + au_whtmp_rmdir_free(a->thargs);
17744 + a->thargs = NULL;
17750 +/* ---------------------------------------------------------------------- */
17753 + * test if @dentry dir can be rename destination or not.
17754 + * success means, it is a logically empty dir.
17756 +static int may_rename_dstdir(struct dentry *dentry, struct au_nhash *whlist)
17758 + return au_test_empty(dentry, whlist);
17762 + * test if @dentry dir can be rename source or not.
17763 + * if it can, return 0 and @children is filled.
17765 + * - it is a logically empty dir.
17766 + * - or, it exists on writable branch and has no children including whiteouts
17767 + * on the lower branch.
17769 +static int may_rename_srcdir(struct dentry *dentry, aufs_bindex_t btgt)
17772 + unsigned int rdhash;
17773 + aufs_bindex_t bstart;
17775 + bstart = au_dbstart(dentry);
17776 + if (bstart != btgt) {
17777 + struct au_nhash whlist;
17779 + SiMustAnyLock(dentry->d_sb);
17780 + rdhash = au_sbi(dentry->d_sb)->si_rdhash;
17782 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL,
17784 + err = au_nhash_alloc(&whlist, rdhash, GFP_NOFS);
17785 + if (unlikely(err))
17787 + err = au_test_empty(dentry, &whlist);
17788 + au_nhash_wh_free(&whlist);
17792 + if (bstart == au_dbtaildir(dentry))
17793 + return 0; /* success */
17795 + err = au_test_empty_lower(dentry);
17798 + if (err == -ENOTEMPTY) {
17799 + AuWarn1("renaming dir who has child(ren) on multiple branches,"
17800 + " is not supported\n");
17806 +/* side effect: sets whlist and h_dentry */
17807 +static int au_ren_may_dir(struct au_ren_args *a)
17810 + unsigned int rdhash;
17811 + struct dentry *d;
17813 + d = a->dst_dentry;
17814 + SiMustAnyLock(d->d_sb);
17817 + if (au_ftest_ren(a->flags, ISDIR) && a->dst_inode) {
17818 + rdhash = au_sbi(d->d_sb)->si_rdhash;
17820 + rdhash = au_rdhash_est(au_dir_size(/*file*/NULL, d));
17821 + err = au_nhash_alloc(&a->whlist, rdhash, GFP_NOFS);
17822 + if (unlikely(err))
17825 + au_set_dbstart(d, a->dst_bstart);
17826 + err = may_rename_dstdir(d, &a->whlist);
17827 + au_set_dbstart(d, a->btgt);
17829 + a->dst_h_dentry = au_h_dptr(d, au_dbstart(d));
17830 + if (unlikely(err))
17833 + d = a->src_dentry;
17834 + a->src_h_dentry = au_h_dptr(d, au_dbstart(d));
17835 + if (au_ftest_ren(a->flags, ISDIR)) {
17836 + err = may_rename_srcdir(d, a->btgt);
17837 + if (unlikely(err)) {
17838 + au_nhash_wh_free(&a->whlist);
17839 + a->whlist.nh_num = 0;
17846 +/* ---------------------------------------------------------------------- */
17849 + * simple tests for rename.
17850 + * following the checks in vfs, plus the parent-child relationship.
17852 +static int au_may_ren(struct au_ren_args *a)
17855 + struct inode *h_inode;
17857 + if (a->src_bstart == a->btgt) {
17858 + err = au_may_del(a->src_dentry, a->btgt, a->src_h_parent,
17859 + au_ftest_ren(a->flags, ISDIR));
17860 + if (unlikely(err))
17863 + if (unlikely(a->src_h_dentry == a->h_trap))
17868 + if (a->dst_bstart != a->btgt)
17871 + err = -ENOTEMPTY;
17872 + if (unlikely(a->dst_h_dentry == a->h_trap))
17876 + h_inode = a->dst_h_dentry->d_inode;
17877 + isdir = !!au_ftest_ren(a->flags, ISDIR);
17878 + if (!a->dst_dentry->d_inode) {
17879 + if (unlikely(h_inode))
17881 + err = au_may_add(a->dst_dentry, a->btgt, a->dst_h_parent,
17884 + if (unlikely(!h_inode || !h_inode->i_nlink))
17886 + err = au_may_del(a->dst_dentry, a->btgt, a->dst_h_parent,
17888 + if (unlikely(err))
17893 + if (unlikely(err == -ENOENT || err == -EEXIST))
17899 +/* ---------------------------------------------------------------------- */
17904 + * - src_dir and dir by lock_rename()
17905 + * - inode if exitsts
17908 + * + src_dentry and dentry by aufs_read_and_write_lock2() which calls,
17910 + * + di_write_lock2_child()
17911 + * + di_write_lock_child()
17912 + * + ii_write_lock_child()
17913 + * + di_write_lock_child2()
17914 + * + ii_write_lock_child2()
17915 + * + src_parent and parent
17916 + * + di_write_lock_parent()
17917 + * + ii_write_lock_parent()
17918 + * + di_write_lock_parent2()
17919 + * + ii_write_lock_parent2()
17920 + * + lower src_dir and dir by vfsub_lock_rename()
17921 + * + verify the every relationships between child and parent. if any
17922 + * of them failed, unlock all and return -EBUSY.
17924 +static void au_ren_unlock(struct au_ren_args *a)
17926 + struct super_block *sb;
17928 + sb = a->dst_dentry->d_sb;
17929 + if (au_ftest_ren(a->flags, MNT_WRITE))
17930 + mnt_drop_write(a->br->br_mnt);
17931 + vfsub_unlock_rename(a->src_h_parent, a->src_hdir,
17932 + a->dst_h_parent, a->dst_hdir);
17935 +static int au_ren_lock(struct au_ren_args *a)
17938 + unsigned int udba;
17941 + a->src_h_parent = au_h_dptr(a->src_parent, a->btgt);
17942 + a->src_hdir = au_hi(a->src_dir, a->btgt);
17943 + a->dst_h_parent = au_h_dptr(a->dst_parent, a->btgt);
17944 + a->dst_hdir = au_hi(a->dst_dir, a->btgt);
17945 + a->h_trap = vfsub_lock_rename(a->src_h_parent, a->src_hdir,
17946 + a->dst_h_parent, a->dst_hdir);
17947 + udba = au_opt_udba(a->src_dentry->d_sb);
17948 + if (unlikely(a->src_hdir->hi_inode != a->src_h_parent->d_inode
17949 + || a->dst_hdir->hi_inode != a->dst_h_parent->d_inode))
17950 + err = au_busy_or_stale();
17951 + if (!err && au_dbstart(a->src_dentry) == a->btgt)
17952 + err = au_h_verify(a->src_h_dentry, udba,
17953 + a->src_h_parent->d_inode, a->src_h_parent,
17955 + if (!err && au_dbstart(a->dst_dentry) == a->btgt)
17956 + err = au_h_verify(a->dst_h_dentry, udba,
17957 + a->dst_h_parent->d_inode, a->dst_h_parent,
17960 + err = mnt_want_write(a->br->br_mnt);
17961 + if (unlikely(err))
17963 + au_fset_ren(a->flags, MNT_WRITE);
17964 + goto out; /* success */
17967 + err = au_busy_or_stale();
17970 + au_ren_unlock(a);
17975 +/* ---------------------------------------------------------------------- */
17977 +static void au_ren_refresh_dir(struct au_ren_args *a)
17979 + struct inode *dir;
17981 + dir = a->dst_dir;
17982 + dir->i_version++;
17983 + if (au_ftest_ren(a->flags, ISDIR)) {
17984 + /* is this updating defined in POSIX? */
17985 + au_cpup_attr_timesizes(a->src_inode);
17986 + au_cpup_attr_nlink(dir, /*force*/1);
17989 + if (au_ibstart(dir) == a->btgt)
17990 + au_cpup_attr_timesizes(dir);
17992 + if (au_ftest_ren(a->flags, ISSAMEDIR))
17995 + dir = a->src_dir;
17996 + dir->i_version++;
17997 + if (au_ftest_ren(a->flags, ISDIR))
17998 + au_cpup_attr_nlink(dir, /*force*/1);
17999 + if (au_ibstart(dir) == a->btgt)
18000 + au_cpup_attr_timesizes(dir);
18003 +static void au_ren_refresh(struct au_ren_args *a)
18005 + aufs_bindex_t bend, bindex;
18006 + struct dentry *d, *h_d;
18007 + struct inode *i, *h_i;
18008 + struct super_block *sb;
18010 + d = a->dst_dentry;
18013 + /* already dget-ed by au_ren_or_cpup() */
18014 + au_set_h_dptr(d, a->btgt, a->h_dst);
18016 + i = a->dst_inode;
18018 + if (!au_ftest_ren(a->flags, ISDIR))
18019 + vfsub_drop_nlink(i);
18021 + vfsub_dead_dir(i);
18022 + au_cpup_attr_timesizes(i);
18024 + au_update_dbrange(d, /*do_put_zero*/1);
18027 + for (bindex = au_dbstart(d); bindex < bend; bindex++)
18028 + au_set_h_dptr(d, bindex, NULL);
18029 + bend = au_dbend(d);
18030 + for (bindex = a->btgt + 1; bindex <= bend; bindex++)
18031 + au_set_h_dptr(d, bindex, NULL);
18032 + au_update_dbrange(d, /*do_put_zero*/0);
18035 + d = a->src_dentry;
18036 + au_set_dbwh(d, -1);
18037 + bend = au_dbend(d);
18038 + for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
18039 + h_d = au_h_dptr(d, bindex);
18041 + au_set_h_dptr(d, bindex, NULL);
18043 + au_set_dbend(d, a->btgt);
18046 + i = a->src_inode;
18047 + if (au_opt_test(au_mntflags(sb), PLINK) && au_plink_test(i))
18048 + return; /* success */
18050 + bend = au_ibend(i);
18051 + for (bindex = a->btgt + 1; bindex <= bend; bindex++) {
18052 + h_i = au_h_iptr(i, bindex);
18054 + au_xino_write(sb, bindex, h_i->i_ino, /*ino*/0);
18055 + /* ignore this error */
18056 + au_set_h_iptr(i, bindex, NULL, 0);
18059 + au_set_ibend(i, a->btgt);
18062 +/* ---------------------------------------------------------------------- */
18064 +/* mainly for link(2) and rename(2) */
18065 +int au_wbr(struct dentry *dentry, aufs_bindex_t btgt)
18067 + aufs_bindex_t bdiropq, bwh;
18068 + struct dentry *parent;
18069 + struct au_branch *br;
18071 + parent = dentry->d_parent;
18072 + IMustLock(parent->d_inode); /* dir is locked */
18074 + bdiropq = au_dbdiropq(parent);
18075 + bwh = au_dbwh(dentry);
18076 + br = au_sbr(dentry->d_sb, btgt);
18077 + if (au_br_rdonly(br)
18078 + || (0 <= bdiropq && bdiropq < btgt)
18079 + || (0 <= bwh && bwh < btgt))
18082 + AuDbg("btgt %d\n", btgt);
18086 +/* sets src_bstart, dst_bstart and btgt */
18087 +static int au_ren_wbr(struct au_ren_args *a)
18090 + struct au_wr_dir_args wr_dir_args = {
18091 + /* .force_btgt = -1, */
18092 + .flags = AuWrDir_ADD_ENTRY
18095 + a->src_bstart = au_dbstart(a->src_dentry);
18096 + a->dst_bstart = au_dbstart(a->dst_dentry);
18097 + if (au_ftest_ren(a->flags, ISDIR))
18098 + au_fset_wrdir(wr_dir_args.flags, ISDIR);
18099 + wr_dir_args.force_btgt = a->src_bstart;
18100 + if (a->dst_inode && a->dst_bstart < a->src_bstart)
18101 + wr_dir_args.force_btgt = a->dst_bstart;
18102 + wr_dir_args.force_btgt = au_wbr(a->dst_dentry, wr_dir_args.force_btgt);
18103 + err = au_wr_dir(a->dst_dentry, a->src_dentry, &wr_dir_args);
18109 +static void au_ren_dt(struct au_ren_args *a)
18111 + a->h_path.dentry = a->src_h_parent;
18112 + au_dtime_store(a->src_dt + AuPARENT, a->src_parent, &a->h_path);
18113 + if (!au_ftest_ren(a->flags, ISSAMEDIR)) {
18114 + a->h_path.dentry = a->dst_h_parent;
18115 + au_dtime_store(a->dst_dt + AuPARENT, a->dst_parent, &a->h_path);
18118 + au_fclr_ren(a->flags, DT_DSTDIR);
18119 + if (!au_ftest_ren(a->flags, ISDIR))
18122 + a->h_path.dentry = a->src_h_dentry;
18123 + au_dtime_store(a->src_dt + AuCHILD, a->src_dentry, &a->h_path);
18124 + if (a->dst_h_dentry->d_inode) {
18125 + au_fset_ren(a->flags, DT_DSTDIR);
18126 + a->h_path.dentry = a->dst_h_dentry;
18127 + au_dtime_store(a->dst_dt + AuCHILD, a->dst_dentry, &a->h_path);
18131 +static void au_ren_rev_dt(int err, struct au_ren_args *a)
18133 + struct dentry *h_d;
18134 + struct mutex *h_mtx;
18136 + au_dtime_revert(a->src_dt + AuPARENT);
18137 + if (!au_ftest_ren(a->flags, ISSAMEDIR))
18138 + au_dtime_revert(a->dst_dt + AuPARENT);
18140 + if (au_ftest_ren(a->flags, ISDIR) && err != -EIO) {
18141 + h_d = a->src_dt[AuCHILD].dt_h_path.dentry;
18142 + h_mtx = &h_d->d_inode->i_mutex;
18143 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18144 + au_dtime_revert(a->src_dt + AuCHILD);
18145 + mutex_unlock(h_mtx);
18147 + if (au_ftest_ren(a->flags, DT_DSTDIR)) {
18148 + h_d = a->dst_dt[AuCHILD].dt_h_path.dentry;
18149 + h_mtx = &h_d->d_inode->i_mutex;
18150 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD);
18151 + au_dtime_revert(a->dst_dt + AuCHILD);
18152 + mutex_unlock(h_mtx);
18157 +/* ---------------------------------------------------------------------- */
18159 +int aufs_rename(struct inode *_src_dir, struct dentry *_src_dentry,
18160 + struct inode *_dst_dir, struct dentry *_dst_dentry)
18163 + /* reduce stack space */
18164 + struct au_ren_args *a;
18166 + AuDbg("%.*s, %.*s\n", AuDLNPair(_src_dentry), AuDLNPair(_dst_dentry));
18167 + IMustLock(_src_dir);
18168 + IMustLock(_dst_dir);
18171 + BUILD_BUG_ON(sizeof(*a) > PAGE_SIZE);
18172 + a = kzalloc(sizeof(*a), GFP_NOFS);
18173 + if (unlikely(!a))
18176 + a->src_dir = _src_dir;
18177 + a->src_dentry = _src_dentry;
18178 + a->src_inode = a->src_dentry->d_inode;
18179 + a->src_parent = a->src_dentry->d_parent; /* dir inode is locked */
18180 + a->dst_dir = _dst_dir;
18181 + a->dst_dentry = _dst_dentry;
18182 + a->dst_inode = a->dst_dentry->d_inode;
18183 + a->dst_parent = a->dst_dentry->d_parent; /* dir inode is locked */
18184 + if (a->dst_inode) {
18185 + IMustLock(a->dst_inode);
18186 + au_igrab(a->dst_inode);
18190 + flags = AuLock_FLUSH | AuLock_NOPLM | AuLock_GEN;
18191 + if (S_ISDIR(a->src_inode->i_mode)) {
18192 + au_fset_ren(a->flags, ISDIR);
18193 + if (unlikely(a->dst_inode && !S_ISDIR(a->dst_inode->i_mode)))
18195 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
18196 + AuLock_DIR | flags);
18198 + err = aufs_read_and_write_lock2(a->dst_dentry, a->src_dentry,
18200 + if (unlikely(err))
18203 + err = au_d_hashed_positive(a->src_dentry);
18204 + if (unlikely(err))
18207 + if (a->dst_inode) {
18209 + * If it is a dir, VFS unhash dst_dentry before this
18210 + * function. It means we cannot rely upon d_unhashed().
18212 + if (unlikely(!a->dst_inode->i_nlink))
18214 + if (!S_ISDIR(a->dst_inode->i_mode)) {
18215 + err = au_d_hashed_positive(a->dst_dentry);
18216 + if (unlikely(err))
18218 + } else if (unlikely(IS_DEADDIR(a->dst_inode)))
18220 + } else if (unlikely(d_unhashed(a->dst_dentry)))
18223 + au_fset_ren(a->flags, ISSAMEDIR); /* temporary */
18224 + di_write_lock_parent(a->dst_parent);
18226 + /* which branch we process */
18227 + err = au_ren_wbr(a);
18228 + if (unlikely(err < 0))
18230 + a->br = au_sbr(a->dst_dentry->d_sb, a->btgt);
18231 + a->h_path.mnt = a->br->br_mnt;
18233 + /* are they available to be renamed */
18234 + err = au_ren_may_dir(a);
18235 + if (unlikely(err))
18236 + goto out_children;
18238 + /* prepare the writable parent dir on the same branch */
18239 + if (a->dst_bstart == a->btgt) {
18240 + au_fset_ren(a->flags, WHDST);
18242 + err = au_cpup_dirs(a->dst_dentry, a->btgt);
18243 + if (unlikely(err))
18244 + goto out_children;
18247 + if (a->src_dir != a->dst_dir) {
18249 + * this temporary unlock is safe,
18250 + * because both dir->i_mutex are locked.
18252 + di_write_unlock(a->dst_parent);
18253 + di_write_lock_parent(a->src_parent);
18254 + err = au_wr_dir_need_wh(a->src_dentry,
18255 + au_ftest_ren(a->flags, ISDIR),
18257 + di_write_unlock(a->src_parent);
18258 + di_write_lock2_parent(a->src_parent, a->dst_parent, /*isdir*/1);
18259 + au_fclr_ren(a->flags, ISSAMEDIR);
18261 + err = au_wr_dir_need_wh(a->src_dentry,
18262 + au_ftest_ren(a->flags, ISDIR),
18264 + if (unlikely(err < 0))
18265 + goto out_children;
18267 + au_fset_ren(a->flags, WHSRC);
18269 + /* lock them all */
18270 + err = au_ren_lock(a);
18271 + if (unlikely(err))
18272 + goto out_children;
18274 + if (!au_opt_test(au_mntflags(a->dst_dir->i_sb), UDBA_NONE))
18275 + err = au_may_ren(a);
18276 + else if (unlikely(a->dst_dentry->d_name.len > AUFS_MAX_NAMELEN))
18277 + err = -ENAMETOOLONG;
18278 + if (unlikely(err))
18281 + /* store timestamps to be revertible */
18285 + err = do_rename(a);
18286 + if (unlikely(err))
18289 + /* update dir attributes */
18290 + au_ren_refresh_dir(a);
18292 + /* dput/iput all lower dentries */
18293 + au_ren_refresh(a);
18295 + goto out_hdir; /* success */
18298 + au_ren_rev_dt(err, a);
18300 + au_ren_unlock(a);
18302 + au_nhash_wh_free(&a->whlist);
18303 + if (err && a->dst_inode && a->dst_bstart != a->btgt) {
18304 + AuDbg("bstart %d, btgt %d\n", a->dst_bstart, a->btgt);
18305 + au_set_h_dptr(a->dst_dentry, a->btgt, NULL);
18306 + au_set_dbstart(a->dst_dentry, a->dst_bstart);
18310 + d_move(a->src_dentry, a->dst_dentry);
18312 + au_update_dbstart(a->dst_dentry);
18313 + if (!a->dst_inode)
18314 + d_drop(a->dst_dentry);
18316 + if (au_ftest_ren(a->flags, ISSAMEDIR))
18317 + di_write_unlock(a->dst_parent);
18319 + di_write_unlock2(a->src_parent, a->dst_parent);
18321 + aufs_read_and_write_unlock2(a->dst_dentry, a->src_dentry);
18323 + iput(a->dst_inode);
18325 + au_whtmp_rmdir_free(a->thargs);
18331 diff -urN /usr/share/empty/fs/aufs/Kconfig linux/fs/aufs/Kconfig
18332 --- /usr/share/empty/fs/aufs/Kconfig 1970-01-01 01:00:00.000000000 +0100
18333 +++ linux/fs/aufs/Kconfig 2011-08-24 13:30:24.727980364 +0200
18336 + tristate "Aufs (Advanced multi layered unification filesystem) support"
18337 + depends on EXPERIMENTAL
18339 + Aufs is a stackable unification filesystem such as Unionfs,
18340 + which unifies several directories and provides a merged single
18342 + In the early days, aufs was entirely re-designed and
18343 + re-implemented Unionfs Version 1.x series. Introducing many
18344 + original ideas, approaches and improvements, it becomes totally
18345 + different from Unionfs while keeping the basic features.
18349 + prompt "Maximum number of branches"
18350 + default AUFS_BRANCH_MAX_127
18352 + Specifies the maximum number of branches (or member directories)
18353 + in a single aufs. The larger value consumes more system
18354 + resources and has a minor impact to performance.
18355 +config AUFS_BRANCH_MAX_127
18358 + Specifies the maximum number of branches (or member directories)
18359 + in a single aufs. The larger value consumes more system
18360 + resources and has a minor impact to performance.
18361 +config AUFS_BRANCH_MAX_511
18364 + Specifies the maximum number of branches (or member directories)
18365 + in a single aufs. The larger value consumes more system
18366 + resources and has a minor impact to performance.
18367 +config AUFS_BRANCH_MAX_1023
18370 + Specifies the maximum number of branches (or member directories)
18371 + in a single aufs. The larger value consumes more system
18372 + resources and has a minor impact to performance.
18373 +config AUFS_BRANCH_MAX_32767
18376 + Specifies the maximum number of branches (or member directories)
18377 + in a single aufs. The larger value consumes more system
18378 + resources and has a minor impact to performance.
18381 +config AUFS_SBILIST
18383 + depends on AUFS_MAGIC_SYSRQ || PROC_FS
18386 + Automatic configuration for internal use.
18387 + When aufs supports Magic SysRq or /proc, enabled automatically.
18389 +config AUFS_HNOTIFY
18390 + bool "Detect direct branch access (bypassing aufs)"
18392 + If you want to modify files on branches directly, eg. bypassing aufs,
18393 + and want aufs to detect the changes of them fully, then enable this
18394 + option and use 'udba=notify' mount option.
18395 + Currently there is only one available configuration, "fsnotify".
18396 + It will have a negative impact to the performance.
18397 + See detail in aufs.5.
18400 + prompt "method" if AUFS_HNOTIFY
18401 + default AUFS_HFSNOTIFY
18402 +config AUFS_HFSNOTIFY
18407 +config AUFS_EXPORT
18408 + bool "NFS-exportable aufs"
18409 + depends on EXPORTFS
18411 + If you want to export your mounted aufs via NFS, then enable this
18412 + option. There are several requirements for this configuration.
18413 + See detail in aufs.5.
18415 +config AUFS_INO_T_64
18417 + depends on AUFS_EXPORT
18418 + depends on 64BIT && !(ALPHA || S390)
18421 + Automatic configuration for internal use.
18422 + /* typedef unsigned long/int __kernel_ino_t */
18423 + /* alpha and s390x are int */
18426 + bool "Readdir in userspace"
18428 + Aufs has two methods to provide a merged view for a directory,
18429 + by a user-space library and by kernel-space natively. The latter
18430 + is always enabled but sometimes large and slow.
18431 + If you enable this option, install the library in aufs2-util
18432 + package, and set some environment variables for your readdir(3),
18433 + then the work will be handled in user-space which generally
18434 + shows better performance in most cases.
18435 + See detail in aufs.5.
18437 +config AUFS_PROC_MAP
18438 + bool "support for /proc/maps and lsof(1)"
18439 + depends on PROC_FS
18441 + When you issue mmap(2) in aufs, it is actually a direct mmap(2)
18442 + call to the file on the branch fs since the file in aufs is
18443 + purely virtual. And the file path printed in /proc/maps (and
18444 + others) will be the path on the branch fs. In most cases, it
18445 + does no harm. But some utilities like lsof(1) may confuse since
18446 + the utility or user may expect the file path in aufs to be
18448 + To address this issue, aufs provides a patch which introduces a
18449 + new member called vm_prfile into struct vm_are_struct. The patch
18450 + is meaningless without enabling this configuration since nobody
18451 + sets the new vm_prfile member.
18452 + If you don't apply the patch, then enabling this configuration
18453 + will cause a compile error.
18454 + This approach is fragile since if someone else make some changes
18455 + around vm_file, then vm_prfile may not work anymore. As a
18456 + workaround such case, aufs provides this configuration. If you
18457 + disable it, then lsof(1) may produce incorrect result but the
18458 + problem will be gone even if the aufs patch is applied (I hope).
18460 +config AUFS_SP_IATTR
18461 + bool "Respect the attributes (mtime/ctime mainly) of special files"
18463 + When you write something to a special file, some attributes of it
18464 + (mtime/ctime mainly) may be updated. Generally such updates are
18465 + less important (actually some device drivers and NFS ignore
18466 + it). But some applications (such like test program) requires
18467 + such updates. If you need these updates, then enable this
18468 + configuration which introduces some overhead.
18469 + Currently this configuration handles FIFO only.
18472 + bool "Show whiteouts"
18474 + If you want to make the whiteouts in aufs visible, then enable
18475 + this option and specify 'shwh' mount option. Although it may
18476 + sounds like philosophy or something, but in technically it
18477 + simply shows the name of whiteout with keeping its behaviour.
18479 +config AUFS_BR_RAMFS
18480 + bool "Ramfs (initramfs/rootfs) as an aufs branch"
18482 + If you want to use ramfs as an aufs branch fs, then enable this
18483 + option. Generally tmpfs is recommended.
18484 + Aufs prohibited them to be a branch fs by default, because
18485 + initramfs becomes unusable after switch_root or something
18486 + generally. If you sets initramfs as an aufs branch and boot your
18487 + system by switch_root, you will meet a problem easily since the
18488 + files in initramfs may be inaccessible.
18489 + Unless you are going to use ramfs as an aufs branch fs without
18490 + switch_root or something, leave it N.
18492 +config AUFS_BR_FUSE
18493 + bool "Fuse fs as an aufs branch"
18494 + depends on FUSE_FS
18497 + If you want to use fuse-based userspace filesystem as an aufs
18498 + branch fs, then enable this option.
18499 + It implements the internal poll(2) operation which is
18500 + implemented by fuse only (curretnly).
18505 + Automatic configuration for internal use.
18507 +config AUFS_BR_HFSPLUS
18508 + bool "Hfsplus as an aufs branch"
18509 + depends on HFSPLUS_FS
18512 + If you want to use hfsplus fs as an aufs branch fs, then enable
18513 + this option. This option introduces a small overhead at
18514 + copying-up a file on hfsplus.
18516 +config AUFS_BDEV_LOOP
18518 + depends on BLK_DEV_LOOP
18521 + Automatic configuration for internal use.
18522 + Convert =[ym] into =y.
18525 + bool "Debug aufs"
18527 + Enable this to compile aufs internal debug code.
18528 + It will have a negative impact to the performance.
18530 +config AUFS_MAGIC_SYSRQ
18532 + depends on AUFS_DEBUG && MAGIC_SYSRQ
18535 + Automatic configuration for internal use.
18536 + When aufs supports Magic SysRq, enabled automatically.
18538 diff -urN /usr/share/empty/fs/aufs/loop.c linux/fs/aufs/loop.c
18539 --- /usr/share/empty/fs/aufs/loop.c 1970-01-01 01:00:00.000000000 +0100
18540 +++ linux/fs/aufs/loop.c 2011-08-24 13:30:24.734646739 +0200
18543 + * Copyright (C) 2005-2011 Junjiro R. Okajima
18545 + * This program, aufs is free software; you can redistribute it and/or modify
18546 + * it under the terms of the GNU General Public License as published by
18547 + * the Free Software Foundation; either version 2 of the License, or
18548 + * (at your option) any later version.
18550 + * This program is distributed in the hope that it will be useful,
18551 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18552 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18553 + * GNU General Public License for more details.
18555 + * You should have received a copy of the GNU General Public License
18556 + * along with this program; if not, write to the Free Software
18557 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18561 + * support for loopback block device as a branch
18564 +#include <linux/loop.h>
18568 + * test if two lower dentries have overlapping branches.
18570 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding)
18572 + struct super_block *h_sb;
18573 + struct loop_device *l;
18575 + h_sb = h_adding->d_sb;
18576 + if (MAJOR(h_sb->s_dev) != LOOP_MAJOR)
18579 + l = h_sb->s_bdev->bd_disk->private_data;
18580 + h_adding = l->lo_backing_file->f_dentry;
18582 + * h_adding can be local NFS.
18583 + * in this case aufs cannot detect the loop.
18585 + if (unlikely(h_adding->d_sb == sb))
18587 + return !!au_test_subdir(h_adding, sb->s_root);
18590 +/* true if a kernel thread named 'loop[0-9].*' accesses a file */
18591 +int au_test_loopback_kthread(void)
18594 + struct task_struct *tsk = current;
18597 + if (tsk->flags & PF_KTHREAD) {
18598 + const char c = tsk->comm[4];
18599 + ret = ('0' <= c && c <= '9'
18600 + && !strncmp(tsk->comm, "loop", 4));
18606 +/* ---------------------------------------------------------------------- */
18608 +#define au_warn_loopback_step 16
18609 +static int au_warn_loopback_nelem = au_warn_loopback_step;
18610 +static unsigned long *au_warn_loopback_array;
18612 +void au_warn_loopback(struct super_block *h_sb)
18614 + int i, new_nelem;
18615 + unsigned long *a, magic;
18616 + static DEFINE_SPINLOCK(spin);
18618 + magic = h_sb->s_magic;
18619 + spin_lock(&spin);
18620 + a = au_warn_loopback_array;
18621 + for (i = 0; i < au_warn_loopback_nelem && *a; i++)
18622 + if (a[i] == magic) {
18623 + spin_unlock(&spin);
18627 + /* h_sb is new to us, print it */
18628 + if (i < au_warn_loopback_nelem) {
18633 + /* expand the array */
18634 + new_nelem = au_warn_loopback_nelem + au_warn_loopback_step;
18635 + a = au_kzrealloc(au_warn_loopback_array,
18636 + au_warn_loopback_nelem * sizeof(unsigned long),
18637 + new_nelem * sizeof(unsigned long), GFP_ATOMIC);
18639 + au_warn_loopback_nelem = new_nelem;
18640 + au_warn_loopback_array = a;
18645 + spin_unlock(&spin);
18646 + AuWarn1("realloc failed, ignored\n");
18650 + spin_unlock(&spin);
18651 + pr_warning("you may want to try another patch for loopback file "
18652 + "on %s(0x%lx) branch\n", au_sbtype(h_sb), magic);
18655 +int au_loopback_init(void)
18658 + struct super_block *sb __maybe_unused;
18660 + AuDebugOn(sizeof(sb->s_magic) != sizeof(unsigned long));
18663 + au_warn_loopback_array = kcalloc(au_warn_loopback_step,
18664 + sizeof(unsigned long), GFP_NOFS);
18665 + if (unlikely(!au_warn_loopback_array))
18671 +void au_loopback_fin(void)
18673 + kfree(au_warn_loopback_array);
18675 diff -urN /usr/share/empty/fs/aufs/loop.h linux/fs/aufs/loop.h
18676 --- /usr/share/empty/fs/aufs/loop.h 1970-01-01 01:00:00.000000000 +0100
18677 +++ linux/fs/aufs/loop.h 2011-08-24 13:30:24.734646739 +0200
18680 + * Copyright (C) 2005-2011 Junjiro R. Okajima
18682 + * This program, aufs is free software; you can redistribute it and/or modify
18683 + * it under the terms of the GNU General Public License as published by
18684 + * the Free Software Foundation; either version 2 of the License, or
18685 + * (at your option) any later version.
18687 + * This program is distributed in the hope that it will be useful,
18688 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18689 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18690 + * GNU General Public License for more details.
18692 + * You should have received a copy of the GNU General Public License
18693 + * along with this program; if not, write to the Free Software
18694 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18698 + * support for loopback mount as a branch
18701 +#ifndef __AUFS_LOOP_H__
18702 +#define __AUFS_LOOP_H__
18707 +struct super_block;
18709 +#ifdef CONFIG_AUFS_BDEV_LOOP
18711 +int au_test_loopback_overlap(struct super_block *sb, struct dentry *h_adding);
18712 +int au_test_loopback_kthread(void);
18713 +void au_warn_loopback(struct super_block *h_sb);
18715 +int au_loopback_init(void);
18716 +void au_loopback_fin(void);
18718 +AuStubInt0(au_test_loopback_overlap, struct super_block *sb,
18719 + struct dentry *h_adding)
18720 +AuStubInt0(au_test_loopback_kthread, void)
18721 +AuStubVoid(au_warn_loopback, struct super_block *h_sb)
18723 +AuStubInt0(au_loopback_init, void)
18724 +AuStubVoid(au_loopback_fin, void)
18725 +#endif /* BLK_DEV_LOOP */
18727 +#endif /* __KERNEL__ */
18728 +#endif /* __AUFS_LOOP_H__ */
18729 diff -urN /usr/share/empty/fs/aufs/magic.mk linux/fs/aufs/magic.mk
18730 --- /usr/share/empty/fs/aufs/magic.mk 1970-01-01 01:00:00.000000000 +0100
18731 +++ linux/fs/aufs/magic.mk 2011-08-24 13:30:24.734646739 +0200
18734 +# defined in ${srctree}/fs/fuse/inode.c
18736 +ifdef CONFIG_FUSE_FS
18737 +ccflags-y += -DFUSE_SUPER_MAGIC=0x65735546
18740 +# defined in ${srctree}/fs/ocfs2/ocfs2_fs.h
18742 +ifdef CONFIG_OCFS2_FS
18743 +ccflags-y += -DOCFS2_SUPER_MAGIC=0x7461636f
18746 +# defined in ${srctree}/fs/ocfs2/dlm/userdlm.h
18748 +ifdef CONFIG_OCFS2_FS_O2CB
18749 +ccflags-y += -DDLMFS_MAGIC=0x76a9f425
18752 +# defined in ${srctree}/fs/cifs/cifsfs.c
18754 +ifdef CONFIG_CIFS_FS
18755 +ccflags-y += -DCIFS_MAGIC_NUMBER=0xFF534D42
18758 +# defined in ${srctree}/fs/xfs/xfs_sb.h
18760 +ifdef CONFIG_XFS_FS
18761 +ccflags-y += -DXFS_SB_MAGIC=0x58465342
18764 +# defined in ${srctree}/fs/configfs/mount.c
18766 +ifdef CONFIG_CONFIGFS_FS
18767 +ccflags-y += -DCONFIGFS_MAGIC=0x62656570
18770 +# defined in ${srctree}/fs/9p/v9fs.h
18772 +ifdef CONFIG_9P_FS
18773 +ccflags-y += -DV9FS_MAGIC=0x01021997
18776 +# defined in ${srctree}/fs/ubifs/ubifs.h
18778 +ifdef CONFIG_UBIFS_FS
18779 +ccflags-y += -DUBIFS_SUPER_MAGIC=0x24051905
18782 +# defined in ${srctree}/fs/hfsplus/hfsplus_raw.h
18784 +ifdef CONFIG_HFSPLUS_FS
18785 +ccflags-y += -DHFSPLUS_SUPER_MAGIC=0x482b
18787 diff -urN /usr/share/empty/fs/aufs/Makefile linux/fs/aufs/Makefile
18788 --- /usr/share/empty/fs/aufs/Makefile 1970-01-01 01:00:00.000000000 +0100
18789 +++ linux/fs/aufs/Makefile 2011-08-24 13:30:24.727980364 +0200
18792 +include ${src}/magic.mk
18793 +ifeq (${CONFIG_AUFS_FS},m)
18794 +include ${src}/conf.mk
18796 +-include ${src}/priv_def.mk
18798 +# cf. include/linux/kernel.h
18800 +ccflags-y += -DDEBUG
18801 +# sparse doesn't allow spaces
18802 +ccflags-y += -D'pr_fmt(fmt)=AUFS_NAME"\040%s:%d:%s[%d]:\040"fmt,__func__,__LINE__,current->comm,current->pid'
18804 +obj-$(CONFIG_AUFS_FS) += aufs.o
18805 +aufs-y := module.o sbinfo.o super.o branch.o xino.o sysaufs.o opts.o \
18806 + wkq.o vfsub.o dcsub.o \
18807 + cpup.o whout.o wbr_policy.o \
18808 + dinfo.o dentry.o \
18810 + finfo.o file.o f_op.o \
18812 + iinfo.o inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o \
18816 +aufs-$(CONFIG_PROC_FS) += procfs.o plink.o
18817 +aufs-$(CONFIG_SYSFS) += sysfs.o
18818 +aufs-$(CONFIG_DEBUG_FS) += dbgaufs.o
18819 +aufs-$(CONFIG_AUFS_BDEV_LOOP) += loop.o
18820 +aufs-$(CONFIG_AUFS_HNOTIFY) += hnotify.o
18821 +aufs-$(CONFIG_AUFS_HFSNOTIFY) += hfsnotify.o
18822 +aufs-$(CONFIG_AUFS_EXPORT) += export.o
18823 +aufs-$(CONFIG_AUFS_POLL) += poll.o
18824 +aufs-$(CONFIG_AUFS_RDU) += rdu.o
18825 +aufs-$(CONFIG_AUFS_SP_IATTR) += f_op_sp.o
18826 +aufs-$(CONFIG_AUFS_BR_HFSPLUS) += hfsplus.o
18827 +aufs-$(CONFIG_AUFS_DEBUG) += debug.o
18828 +aufs-$(CONFIG_AUFS_MAGIC_SYSRQ) += sysrq.o
18829 diff -urN /usr/share/empty/fs/aufs/module.c linux/fs/aufs/module.c
18830 --- /usr/share/empty/fs/aufs/module.c 1970-01-01 01:00:00.000000000 +0100
18831 +++ linux/fs/aufs/module.c 2012-01-05 11:00:59.218835185 +0100
18834 + * Copyright (C) 2005-2011 Junjiro R. Okajima
18836 + * This program, aufs is free software; you can redistribute it and/or modify
18837 + * it under the terms of the GNU General Public License as published by
18838 + * the Free Software Foundation; either version 2 of the License, or
18839 + * (at your option) any later version.
18841 + * This program is distributed in the hope that it will be useful,
18842 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
18843 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18844 + * GNU General Public License for more details.
18846 + * You should have received a copy of the GNU General Public License
18847 + * along with this program; if not, write to the Free Software
18848 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18852 + * module global variables and operations
18855 +#include <linux/module.h>
18856 +#include <linux/seq_file.h>
18859 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp)
18861 + if (new_sz <= nused)
18864 + p = krealloc(p, new_sz, gfp);
18866 + memset(p + nused, 0, new_sz - nused);
18870 +/* ---------------------------------------------------------------------- */
18875 +struct kmem_cache *au_cachep[AuCache_Last];
18876 +static int __init au_cache_init(void)
18878 + au_cachep[AuCache_DINFO] = AuCacheCtor(au_dinfo, au_di_init_once);
18879 + if (au_cachep[AuCache_DINFO])
18880 + /* SLAB_DESTROY_BY_RCU */
18881 + au_cachep[AuCache_ICNTNR] = AuCacheCtor(au_icntnr,
18882 + au_icntnr_init_once);
18883 + if (au_cachep[AuCache_ICNTNR])
18884 + au_cachep[AuCache_FINFO] = AuCacheCtor(au_finfo,
18885 + au_fi_init_once);
18886 + if (au_cachep[AuCache_FINFO])
18887 + au_cachep[AuCache_VDIR] = AuCache(au_vdir);
18888 + if (au_cachep[AuCache_VDIR])
18889 + au_cachep[AuCache_DEHSTR] = AuCache(au_vdir_dehstr);
18890 + if (au_cachep[AuCache_DEHSTR])
18896 +static void au_cache_fin(void)
18900 + /* including AuCache_HNOTIFY */
18901 + for (i = 0; i < AuCache_Last; i++)
18902 + if (au_cachep[i]) {
18903 + kmem_cache_destroy(au_cachep[i]);
18904 + au_cachep[i] = NULL;
18908 +/* ---------------------------------------------------------------------- */
18910 +int au_dir_roflags;
18912 +#ifdef CONFIG_AUFS_SBILIST
18914 + * iterate_supers_type() doesn't protect us from
18915 + * remounting (branch management)
18917 +struct au_splhead au_sbilist;
18920 +struct lock_class_key au_lc_key[AuLcKey_Last];
18923 + * functions for module interface.
18925 +MODULE_LICENSE("GPL");
18926 +/* MODULE_LICENSE("GPL v2"); */
18927 +MODULE_AUTHOR("Junjiro R. Okajima <aufs-users@lists.sourceforge.net>");
18928 +MODULE_DESCRIPTION(AUFS_NAME
18929 + " -- Advanced multi layered unification filesystem");
18930 +MODULE_VERSION(AUFS_VERSION);
18932 +/* this module parameter has no meaning when SYSFS is disabled */
18933 +int sysaufs_brs = 1;
18934 +MODULE_PARM_DESC(brs, "use <sysfs>/fs/aufs/si_*/brN");
18935 +module_param_named(brs, sysaufs_brs, int, S_IRUGO);
18937 +/* ---------------------------------------------------------------------- */
18939 +static char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */
18941 +int au_seq_path(struct seq_file *seq, struct path *path)
18943 + return seq_path(seq, path, au_esc_chars);
18946 +/* ---------------------------------------------------------------------- */
18948 +static int __init aufs_init(void)
18953 + p = au_esc_chars;
18954 + for (i = 1; i <= ' '; i++)
18960 + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE);
18962 + au_sbilist_init();
18963 + sysaufs_brs_init();
18966 + err = sysaufs_init();
18967 + if (unlikely(err))
18969 + err = au_procfs_init();
18970 + if (unlikely(err))
18971 + goto out_sysaufs;
18972 + err = au_wkq_init();
18973 + if (unlikely(err))
18975 + err = au_loopback_init();
18976 + if (unlikely(err))
18978 + err = au_hnotify_init();
18979 + if (unlikely(err))
18980 + goto out_loopback;
18981 + err = au_sysrq_init();
18982 + if (unlikely(err))
18984 + err = au_cache_init();
18985 + if (unlikely(err))
18987 + err = register_filesystem(&aufs_fs_type);
18988 + if (unlikely(err))
18990 + /* since we define pr_fmt, call printk directly */
18991 + printk(KERN_INFO AUFS_NAME " " AUFS_VERSION "\n");
18992 + goto out; /* success */
18999 + au_hnotify_fin();
19001 + au_loopback_fin();
19013 +static void __exit aufs_exit(void)
19015 + unregister_filesystem(&aufs_fs_type);
19018 + au_hnotify_fin();
19019 + au_loopback_fin();
19026 +module_init(aufs_init);
19027 +module_exit(aufs_exit);
19028 diff -urN /usr/share/empty/fs/aufs/module.h linux/fs/aufs/module.h
19029 --- /usr/share/empty/fs/aufs/module.h 1970-01-01 01:00:00.000000000 +0100
19030 +++ linux/fs/aufs/module.h 2012-01-05 11:00:59.218835185 +0100
19033 + * Copyright (C) 2005-2011 Junjiro R. Okajima
19035 + * This program, aufs is free software; you can redistribute it and/or modify
19036 + * it under the terms of the GNU General Public License as published by
19037 + * the Free Software Foundation; either version 2 of the License, or
19038 + * (at your option) any later version.
19040 + * This program is distributed in the hope that it will be useful,
19041 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19042 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19043 + * GNU General Public License for more details.
19045 + * You should have received a copy of the GNU General Public License
19046 + * along with this program; if not, write to the Free Software
19047 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19051 + * module initialization and module-global
19054 +#ifndef __AUFS_MODULE_H__
19055 +#define __AUFS_MODULE_H__
19059 +#include <linux/slab.h>
19064 +/* module parameters */
19065 +extern int sysaufs_brs;
19067 +/* ---------------------------------------------------------------------- */
19069 +extern int au_dir_roflags;
19072 + AuLcNonDir_FIINFO,
19073 + AuLcNonDir_DIINFO,
19074 + AuLcNonDir_IIINFO,
19080 + AuLcSymlink_DIINFO,
19081 + AuLcSymlink_IIINFO,
19085 +extern struct lock_class_key au_lc_key[AuLcKey_Last];
19087 +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp);
19088 +int au_seq_path(struct seq_file *seq, struct path *path);
19090 +#ifdef CONFIG_PROC_FS
19092 +int __init au_procfs_init(void);
19093 +void au_procfs_fin(void);
19095 +AuStubInt0(au_procfs_init, void);
19096 +AuStubVoid(au_procfs_fin, void);
19099 +/* ---------------------------------------------------------------------- */
19108 +#ifdef CONFIG_AUFS_HNOTIFY
19114 +#define AuCacheFlags (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD)
19115 +#define AuCache(type) KMEM_CACHE(type, AuCacheFlags)
19116 +#define AuCacheCtor(type, ctor) \
19117 + kmem_cache_create(#type, sizeof(struct type), \
19118 + __alignof__(struct type), AuCacheFlags, ctor)
19120 +extern struct kmem_cache *au_cachep[];
19122 +#define AuCacheFuncs(name, index) \
19123 +static inline struct au_##name *au_cache_alloc_##name(void) \
19124 +{ return kmem_cache_alloc(au_cachep[AuCache_##index], GFP_NOFS); } \
19125 +static inline void au_cache_free_##name(struct au_##name *p) \
19126 +{ kmem_cache_free(au_cachep[AuCache_##index], p); }
19128 +AuCacheFuncs(dinfo, DINFO);
19129 +AuCacheFuncs(icntnr, ICNTNR);
19130 +AuCacheFuncs(finfo, FINFO);
19131 +AuCacheFuncs(vdir, VDIR);
19132 +AuCacheFuncs(vdir_dehstr, DEHSTR);
19133 +#ifdef CONFIG_AUFS_HNOTIFY
19134 +AuCacheFuncs(hnotify, HNOTIFY);
19137 +#endif /* __KERNEL__ */
19138 +#endif /* __AUFS_MODULE_H__ */
19139 diff -urN /usr/share/empty/fs/aufs/opts.c linux/fs/aufs/opts.c
19140 --- /usr/share/empty/fs/aufs/opts.c 1970-01-01 01:00:00.000000000 +0100
19141 +++ linux/fs/aufs/opts.c 2012-01-05 11:00:59.218835185 +0100
19144 + * Copyright (C) 2005-2011 Junjiro R. Okajima
19146 + * This program, aufs is free software; you can redistribute it and/or modify
19147 + * it under the terms of the GNU General Public License as published by
19148 + * the Free Software Foundation; either version 2 of the License, or
19149 + * (at your option) any later version.
19151 + * This program is distributed in the hope that it will be useful,
19152 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
19153 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19154 + * GNU General Public License for more details.
19156 + * You should have received a copy of the GNU General Public License
19157 + * along with this program; if not, write to the Free Software
19158 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19162 + * mount options/flags
19165 +#include <linux/file.h>
19166 +#include <linux/jiffies.h>
19167 +#include <linux/namei.h>
19168 +#include <linux/types.h> /* a distribution requires */
19169 +#include <linux/parser.h>
19172 +/* ---------------------------------------------------------------------- */
19176 + Opt_add, Opt_del, Opt_mod, Opt_reorder, Opt_append, Opt_prepend,
19177 + Opt_idel, Opt_imod, Opt_ireorder,
19178 + Opt_dirwh, Opt_rdcache, Opt_rdblk, Opt_rdhash, Opt_rendir,
19179 + Opt_rdblk_def, Opt_rdhash_def,
19180 + Opt_xino, Opt_zxino, Opt_noxino,
19181 + Opt_trunc_xino, Opt_trunc_xino_v, Opt_notrunc_xino,
19182 + Opt_trunc_xino_path, Opt_itrunc_xino,
19183 + Opt_trunc_xib, Opt_notrunc_xib,
19184 + Opt_shwh, Opt_noshwh,
19185 + Opt_plink, Opt_noplink, Opt_list_plink,
19187 + Opt_dio, Opt_nodio,
19188 + /* Opt_lock, Opt_unlock, */
19189 + Opt_cmd, Opt_cmd_args,
19190 + Opt_diropq_a, Opt_diropq_w,
19191 + Opt_warn_perm, Opt_nowarn_perm,
19192 + Opt_wbr_copyup, Opt_wbr_create,
19193 + Opt_refrof, Opt_norefrof,
19194 + Opt_verbose, Opt_noverbose,
19195 + Opt_sum, Opt_nosum, Opt_wsum,
19196 + Opt_tail, Opt_ignore, Opt_ignore_silent, Opt_err
19199 +static match_table_t options = {
19200 + {Opt_br, "br=%s"},
19201 + {Opt_br, "br:%s"},
19203 + {Opt_add, "add=%d:%s"},
19204 + {Opt_add, "add:%d:%s"},
19205 + {Opt_add, "ins=%d:%s"},
19206 + {Opt_add, "ins:%d:%s"},
19207 + {Opt_append, "append=%s"},
19208 + {Opt_append, "append:%s"},
19209 + {Opt_prepend, "prepend=%s"},
19210 + {Opt_prepend, "prepend:%s"},
19212 + {Opt_del, "del=%s"},
19213 + {Opt_del, "del:%s"},
19214 + /* {Opt_idel, "idel:%d"}, */
19215 + {Opt_mod, "mod=%s"},
19216 + {Opt_mod, "mod:%s"},
19217 + /* {Opt_imod, "imod:%d:%s"}, */
19219 + {Opt_dirwh, "dirwh=%d"},
19221 + {Opt_xino, "xino=%s"},
19222 + {Opt_noxino, "noxino"},
19223 + {Opt_trunc_xino, "trunc_xino"},
19224 + {Opt_trunc_xino_v, "trunc_xino_v=%d:%d"},
19225 + {Opt_notrunc_xino, "notrunc_xino"},
19226 + {Opt_trunc_xino_path, "trunc_xino=%s"},
19227 + {Opt_itrunc_xino, "itrunc_xino=%d"},
19228 + /* {Opt_zxino, "zxino=%s"}, */
19229 + {Opt_trunc_xib, "trunc_xib"},
19230 + {Opt_notrunc_xib, "notrunc_xib"},
19232 +#ifdef CONFIG_PROC_FS
19233 + {Opt_plink, "plink"},
19235 + {Opt_ignore_silent, "plink"},
19238 + {Opt_noplink, "noplink"},
19240 +#ifdef CONFIG_AUFS_DEBUG
19241 + {Opt_list_plink, "list_plink"},
19244 + {Opt_udba, "udba=%s"},
19246 + {Opt_dio, "dio"},
19247 + {Opt_nodio, "nodio"},
19249 + {Opt_diropq_a, "diropq=always"},
19250 + {Opt_diropq_a, "diropq=a"},
19251 + {Opt_diropq_w, "diropq=whiteouted"},
19252 + {Opt_diropq_w, "diropq=w"},
19254 + {Opt_warn_perm, "warn_perm"},
19255 + {Opt_nowarn_perm, "nowarn_perm"},
19257 + /* keep them temporary */
19258 + {Opt_ignore_silent, "coo=%s"},
19259 + {Opt_ignore_silent, "nodlgt"},
19260 + {Opt_ignore_silent, "nodirperm1"},
19261 + {Opt_ignore_silent, "clean_plink"},
19263 +#ifdef CONFIG_AUFS_SHWH
19264 + {Opt_shwh, "shwh"},
19266 + {Opt_noshwh, "noshwh"},
19268 + {Opt_rendir, "rendir=%d"},
19270 + {Opt_refrof, "refrof"},
19271 + {Opt_norefrof, "norefrof"},
19273 + {Opt_verbose, "verbose"},
19274 + {Opt_verbose, "v"},
19275 + {Opt_noverbose, "noverbose"},
19276 + {Opt_noverbose, "quiet"},
19277 + {Opt_noverbose, "q"},
19278 + {Opt_noverbose, "silent"},
19280 + {Opt_sum, "sum"},
19281 + {Opt_nosum, "nosum"},
19282 + {Opt_wsum, "wsum"},
19284 + {Opt_rdcache, "rdcache=%d"},
19285 + {Opt_rdblk, "rdblk=%d"},
19286 + {Opt_rdblk_def, "rdblk=def"},
19287 + {Opt_rdhash, "rdhash=%d"},
19288 + {Opt_rdhash_def, "rdhash=def"},
19290 + {Opt_wbr_create, "create=%s"},
19291 + {Opt_wbr_create, "create_policy=%s"},
19292 + {Opt_wbr_copyup, "cpup=%s"},
19293 + {Opt_wbr_copyup, "copyup=%s"},
19294 + {Opt_wbr_copyup, "copyup_policy=%s"},
19296 + /* internal use for the scripts */
19297 + {Opt_ignore_silent, "si=%s"},
19299 + {Opt_br, "dirs=%s"},
19300 + {Opt_ignore, "debug=%d"},
19301 + {Opt_ignore, "delete=whiteout"},
19302 + {Opt_ignore, "delete=all"},
19303 + {Opt_ignore, "imap=%s"},
19305 + /* temporary workaround, due to old mount(8)? */
19306 + {Opt_ignore_silent, "relatime"},
19311 +/* ---------------------------------------------------------------------- */
19313 +static const char *au_parser_pattern(int val, struct match_token *token)
19315 + while (token->pattern) {
19316 + if (token->token == val)
19317 + return token->pattern;
19324 +/* ---------------------------------------------------------------------- */
19326 +static match_table_t brperm = {
19327 + {AuBrPerm_RO, AUFS_BRPERM_RO},
19328 + {AuBrPerm_RR, AUFS_BRPERM_RR},
19329 + {AuBrPerm_RW, AUFS_BRPERM_RW},
19333 +static match_table_t brrattr = {
19334 + {AuBrRAttr_WH, AUFS_BRRATTR_WH},
19338 +static match_table_t brwattr = {
19339 + {AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH},
19343 +#define AuBrStr_LONGEST AUFS_BRPERM_RW "+" AUFS_BRWATTR_NLWH
19345 +static int br_attr_val(char *str, match_table_t table, substring_t args[])
19352 + p = strchr(str, '+');
19355 + v = match_token(str, table, args);
19361 + pr_warning("ignored branch attribute %s\n", str);
19371 +static int noinline_for_stack br_perm_val(char *perm)
19375 + substring_t args[MAX_OPT_ARGS];
19377 + p = strchr(perm, '+');
19380 + val = match_token(perm, brperm, args);
19384 + pr_warning("ignored branch permission %s\n", perm);
19385 + val = AuBrPerm_RO;
19392 + case AuBrPerm_RO:
19393 + case AuBrPerm_RR:
19394 + val |= br_attr_val(p + 1, brrattr, args);
19396 + case AuBrPerm_RW:
19397 + val |= br_attr_val(p + 1, brwattr, args);
19405 +/* Caller should free the return value */
19406 +char *au_optstr_br_perm(int brperm)
19408 + char *p, a[sizeof(AuBrStr_LONGEST)];
19411 +#define SetPerm(str) do { \
19412 + sz = sizeof(str); \
19413 + memcpy(a, str, sz); \
19414 + p = a + sz - 1; \
19417 +#define AppendAttr(flag, str) do { \
19418 + if (brperm & flag) { \
19419 + sz = sizeof(str); \
19421 + memcpy(p, str, sz); \
19426 + switch (brperm & AuBrPerm_Mask) {
19427 + case AuBrPerm_RO:
19428 + SetPerm(AUFS_BRPERM_RO);
19430 + case AuBrPerm_RR:
19431 + SetPerm(AUFS_BRPERM_RR);
19433 + case AuBrPerm_RW:
19434 + SetPerm(AUFS_BRPERM_RW);
19440 + AppendAttr(AuBrRAttr_WH, AUFS_BRRATTR_WH);
19441 + AppendAttr(AuBrWAttr_NoLinkWH, AUFS_BRWATTR_NLWH);
19443 + AuDebugOn(strlen(a) >= sizeof(a));
19444 + return kstrdup(a, GFP_NOFS);
19449 +/* ---------------------------------------------------------------------- */
19451 +static match_table_t udbalevel = {
19452 + {AuOpt_UDBA_REVAL, "reval"},
19453 + {AuOpt_UDBA_NONE, "none"},
19454 +#ifdef CONFIG_AUFS_HNOTIFY
19455 + {AuOpt_UDBA_HNOTIFY, "notify"}, /* abstraction */
19456 +#ifdef CONFIG_AUFS_HFSNOTIFY
19457 + {AuOpt_UDBA_HNOTIFY, "fsnotify"},
19463 +static int noinline_for_stack udba_val(char *str)
19465 + substring_t args[MAX_OPT_ARGS];
19467 + return match_token(str, udbalevel, args);
19470 +const char *au_optstr_udba(int udba)
19472 + return au_parser_pattern(udba, (void *)udbalevel);
19475 +/* ---------------------------------------------------------------------- */
19477 +static match_table_t au_wbr_create_policy = {
19478 + {AuWbrCreate_TDP, "tdp"},
19479 + {AuWbrCreate_TDP, "top-down-parent"},
19480 + {AuWbrCreate_RR, "rr"},
19481 + {AuWbrCreate_RR, "round-robin"},
19482 + {AuWbrCreate_MFS, "mfs"},
19483 + {AuWbrCreate_MFS, "most-free-space"},
19484 + {AuWbrCreate_MFSV, "mfs:%d"},
19485 + {AuWbrCreate_MFSV, "most-free-space:%d"},
19487 + {AuWbrCreate_MFSRR, "mfsrr:%d"},
19488 + {AuWbrCreate_MFSRRV, "mfsrr:%d:%d"},
19489 + {AuWbrCreate_PMFS, "pmfs"},
19490 + {AuWbrCreate_PMFSV, "pmfs:%d"},
19496 + * cf. linux/lib/parser.c and cmdline.c
19497 + * gave up calling memparse() since it uses simple_strtoull() instead of
19500 +static int noinline_for_stack
19501 +au_match_ull(substring_t *s, unsigned long long *result)
19504 + unsigned int len;
19508 + len = s->to - s->from;
19509 + if (len + 1 <= sizeof(a)) {
19510 + memcpy(a, s->from, len);
19512 + err = kstrtoull(a, 0, result);
19517 +static int au_wbr_mfs_wmark(substring_t *arg, char *str,
19518 + struct au_opt_wbr_create *create)
19521 + unsigned long long ull;
19524 + if (!au_match_ull(arg, &ull))
19525 + create->mfsrr_watermark = ull;
19527 + pr_err("bad integer in %s\n", str);
19534 +static int au_wbr_mfs_sec(substring_t *arg, char *str,
19535 + struct au_opt_wbr_create *create)
19540 + if (!match_int(arg, &n) && 0 <= n && n <= AUFS_MFS_MAX_SEC)
19541 + create->mfs_second = n;
19543 + pr_err("bad integer in %s\n", str);
19550 +static int noinline_for_stack
19551 +au_wbr_create_val(char *str, struct au_opt_wbr_create *create)
19554 + substring_t args[MAX_OPT_ARGS];
19556 + err = match_token(str, au_wbr_create_policy, args);
19557 + create->wbr_create = err;
19559 + case AuWbrCreate_MFSRRV:
19560 + e = au_wbr_mfs_wmark(&args[0], str, create);
19562 + e = au_wbr_mfs_sec(&args[1], str, create);
19566 + case AuWbrCreate_MFSRR:
19567 + e = au_wbr_mfs_wmark(&args[0], str, create);
19568 + if (unlikely(e)) {
19573 + case AuWbrCreate_MFS:
19574 + case AuWbrCreate_PMFS:
19575 + create->mfs_second = AUFS_MFS_DEF_SEC;
19577 + case AuWbrCreate_MFSV:
19578 + case AuWbrCreate_PMFSV:
19579 + e = au_wbr_mfs_sec(&args[0], str, create);
19588 +const char *au_optstr_wbr_create(int wbr_create)
19590 + return au_parser_pattern(wbr_create, (void *)au_wbr_create_policy);
19593 +static match_table_t au_wbr_copyup_policy = {
19594 + {AuWbrCopyup_TDP, "tdp"},
19595 + {AuWbrCopyup_TDP, "top-down-parent"},
19596 + {AuWbrCopyup_BUP, "bup"},
19597 + {AuWbrCopyup_BUP, "bottom-up-parent"},
19598 + {AuWbrCopyup_BU, "bu"},
19599 + {AuWbrCopyup_BU, "bottom-up"},
19603 +static int noinline_for_stack au_wbr_copyup_val(char *str)
19605 + substring_t args[MAX_OPT_ARGS];
19607 + return match_token(str, au_wbr_copyup_policy, args);
19610 +const char *au_optstr_wbr_copyup(int wbr_copyup)
19612 + return au_parser_pattern(wbr_copyup, (void *)au_wbr_copyup_policy);
19615 +/* ---------------------------------------------------------------------- */
19617 +static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
19619 +static void dump_opts(struct au_opts *opts)
19621 +#ifdef CONFIG_AUFS_DEBUG
19622 + /* reduce stack space */
19624 + struct au_opt_add *add;
19625 + struct au_opt_del *del;
19626 + struct au_opt_mod *mod;
19627 + struct au_opt_xino *xino;
19628 + struct au_opt_xino_itrunc *xino_itrunc;
19629 + struct au_opt_wbr_create *create;
19631 + struct au_opt *opt;
19634 + while (opt->type != Opt_tail) {
19635 + switch (opt->type) {
19637 + u.add = &opt->add;
19638 + AuDbg("add {b%d, %s, 0x%x, %p}\n",
19639 + u.add->bindex, u.add->pathname, u.add->perm,
19640 + u.add->path.dentry);
19644 + u.del = &opt->del;
19645 + AuDbg("del {%s, %p}\n",
19646 + u.del->pathname, u.del->h_path.dentry);
19650 + u.mod = &opt->mod;
19651 + AuDbg("mod {%s, 0x%x, %p}\n",
19652 + u.mod->path, u.mod->perm, u.mod->h_root);
19655 + u.add = &opt->add;
19656 + AuDbg("append {b%d, %s, 0x%x, %p}\n",
19657 + u.add->bindex, u.add->pathname, u.add->perm,
19658 + u.add->path.dentry);
19660 + case Opt_prepend:
19661 + u.add = &opt->add;
19662 + AuDbg("prepend {b%d, %s, 0x%x, %p}\n",
19663 + u.add->bindex, u.add->pathname, u.add->perm,
19664 + u.add->path.dentry);
19667 + AuDbg("dirwh %d\n", opt->dirwh);
19669 + case Opt_rdcache:
19670 + AuDbg("rdcache %d\n", opt->rdcache);
19673 + AuDbg("rdblk %u\n", opt->rdblk);
19675 + case Opt_rdblk_def:
19676 + AuDbg("rdblk_def\n");
19679 + AuDbg("rdhash %u\n", opt->rdhash);
19681 + case Opt_rdhash_def:
19682 + AuDbg("rdhash_def\n");
19685 + u.xino = &opt->xino;
19686 + AuDbg("xino {%s %.*s}\n",
19688 + AuDLNPair(u.xino->file->f_dentry));
19690 + case Opt_trunc_xino:
19691 + AuLabel(trunc_xino);
19693 + case Opt_notrunc_xino:
19694 + AuLabel(notrunc_xino);
19696 + case Opt_trunc_xino_path:
19697 + case Opt_itrunc_xino:
19698 + u.xino_itrunc = &opt->xino_itrunc;
19699 + AuDbg("trunc_xino %d\n", u.xino_itrunc->bindex);
19705 + case Opt_trunc_xib:
19706 + AuLabel(trunc_xib);
19708 + case Opt_notrunc_xib:
19709 + AuLabel(notrunc_xib);
19720 + case Opt_noplink:
19721 + AuLabel(noplink);
19723 + case Opt_list_plink:
19724 + AuLabel(list_plink);
19727 + AuDbg("udba %d, %s\n",
19728 + opt->udba, au_optstr_udba(opt->udba));
19736 + case Opt_diropq_a:
19737 + AuLabel(diropq_a);
19739 + case Opt_diropq_w:
19740 + AuLabel(diropq_w);
19742 + case Opt_warn_perm:
19743 + AuLabel(warn_perm);
19745 + case Opt_nowarn_perm:
19746 + AuLabel(nowarn_perm);
19751 + case Opt_norefrof:
19752 + AuLabel(norefrof);
19754 + case Opt_verbose:
19755 + AuLabel(verbose);
19757 + case Opt_noverbose:
19758 + AuLabel(noverbose);
19769 + case Opt_wbr_create:
19770 + u.create = &opt->wbr_create;
19771 + AuDbg("create %d, %s\n", u.create->wbr_create,
19772 + au_optstr_wbr_create(u.create->wbr_create));
19773 + switch (u.create->wbr_create) {
19774 + case AuWbrCreate_MFSV:
19775 + case AuWbrCreate_PMFSV:
19776 + AuDbg("%d sec\n", u.create->mfs_second);
19778 + case AuWbrCreate_MFSRR:
19779 + AuDbg("%llu watermark\n",
19780 + u.create->mfsrr_watermark);
19782 + case AuWbrCreate_MFSRRV:
19783 + AuDbg("%llu watermark, %d sec\n",
19784 + u.create->mfsrr_watermark,
19785 + u.create->mfs_second);
19789 + case Opt_wbr_copyup:
19790 + AuDbg("copyup %d, %s\n", opt->wbr_copyup,
19791 + au_optstr_wbr_copyup(opt->wbr_copyup));
19801 +void au_opts_free(struct au_opts *opts)
19803 + struct au_opt *opt;
19806 + while (opt->type != Opt_tail) {
19807 + switch (opt->type) {
19810 + case Opt_prepend:
19811 + path_put(&opt->add.path);
19815 + path_put(&opt->del.h_path);
19819 + dput(opt->mod.h_root);
19822 + fput(opt->xino.file);
19829 +static int opt_add(struct au_opt *opt, char *opt_str, unsigned long sb_flags,
19830 + aufs_bindex_t bindex)
19833 + struct au_opt_add *add = &opt->add;
19836 + add->bindex = bindex;
19837 + add->perm = AuBrPerm_RO;
19838 + add->pathname = opt_str;
19839 + p = strchr(opt_str, '=');
19843 + add->perm = br_perm_val(p);
19846 + err = vfsub_kern_path(add->pathname, lkup_dirflags, &add->path);
19849 + add->perm = AuBrPerm_RO;
19850 + if (au_test_fs_rr(add->path.dentry->d_sb))
19851 + add->perm = AuBrPerm_RR;
19852 + else if (!bindex && !(sb_flags & MS_RDONLY))
19853 + add->perm = AuBrPerm_RW;
19855 + opt->type = Opt_add;
19858 + pr_err("lookup failed %s (%d)\n", add->pathname, err);
19865 +static int au_opts_parse_del(struct au_opt_del *del, substring_t args[])
19869 + del->pathname = args[0].from;
19870 + AuDbg("del path %s\n", del->pathname);
19872 + err = vfsub_kern_path(del->pathname, lkup_dirflags, &del->h_path);
19873 + if (unlikely(err))
19874 + pr_err("lookup failed %s (%d)\n", del->pathname, err);
19879 +#if 0 /* reserved for future use */
19880 +static int au_opts_parse_idel(struct super_block *sb, aufs_bindex_t bindex,
19881 + struct au_opt_del *del, substring_t args[])
19884 + struct dentry *root;
19887 + root = sb->s_root;
19888 + aufs_read_lock(root, AuLock_FLUSH);
19889 + if (bindex < 0 || au_sbend(sb) < bindex) {
19890 + pr_err("out of bounds, %d\n", bindex);
19895 + del->h_path.dentry = dget(au_h_dptr(root, bindex));
19896 + del->h_path.mnt = mntget(au_sbr_mnt(sb, bindex));
19899 + aufs_read_unlock(root, !AuLock_IR);
19904 +static int noinline_for_stack
19905 +au_opts_parse_mod(struct au_opt_mod *mod, substring_t args[])
19908 + struct path path;
19912 + mod->path = args[0].from;
19913 + p = strchr(mod->path, '=');
19914 + if (unlikely(!p)) {
19915 + pr_err("no permssion %s\n", args[0].from);
19920 + err = vfsub_kern_path(mod->path, lkup_dirflags, &path);
19921 + if (unlikely(err)) {
19922 + pr_err("lookup failed %s (%d)\n", mod->path, err);
19926 + mod->perm = br_perm_val(p);
19927 + AuDbg("mod path %s, perm 0x%x, %s\n", mod->path, mod->perm, p);
19928 + mod->h_root = dget(path.dentry);
19935 +#if 0 /* reserved for future use */
19936 +static int au_opts_parse_imod(struct super_block *sb, aufs_bindex_t bindex,
19937 + struct au_opt_mod *mod, substring_t args[])
19940 + struct dentry *root;
19943 + root = sb->s_root;
19944 + aufs_read_lock(root, AuLock_FLUSH);
19945 + if (bindex < 0 || au_sbend(sb) < bindex) {
19946 + pr_err("out of bounds, %d\n", bindex);
19951 + mod->perm = br_perm_val(args[1].from);
19952 + AuDbg("mod path %s, perm 0x%x, %s\n",
19953 + mod->path, mod->perm, args[1].from);
19954 + mod->h_root = dget(au_h_dptr(root, bindex));
19957 + aufs_read_unlock(root, !AuLock_IR);
19962 +static int au_opts_parse_xino(struct super_block *sb, struct au_opt_xino *xino,
19963 + substring_t args[])
19966 + struct file *file;
19968 + file = au_xino_create(sb, args[0].from, /*silent*/0);
19969 + err = PTR_ERR(file);
19970 + if (IS_ERR(file))
19974 + if (unlikely(file->f_dentry->d_sb == sb)) {
19976 + pr_err("%s must be outside\n", args[0].from);
19981 + xino->file = file;
19982 + xino->path = args[0].from;
19988 +static int noinline_for_stack
19989 +au_opts_parse_xino_itrunc_path(struct super_block *sb,
19990 + struct au_opt_xino_itrunc *xino_itrunc,
19991 + substring_t args[])
19994 + aufs_bindex_t bend, bindex;
19995 + struct path path;
19996 + struct dentry *root;
19998 + err = vfsub_kern_path(args[0].from, lkup_dirflags, &path);
19999 + if (unlikely(err)) {
20000 + pr_err("lookup failed %s (%d)\n", args[0].from, err);
20004 + xino_itrunc->bindex = -1;
20005 + root = sb->s_root;
20006 + aufs_read_lock(root, AuLock_FLUSH);
20007 + bend = au_sbend(sb);
20008 + for (bindex = 0; bindex <= bend; bindex++) {
20009 + if (au_h_dptr(root, bindex) == path.dentry) {
20010 + xino_itrunc->bindex = bindex;
20014 + aufs_read_unlock(root, !AuLock_IR);
20017 + if (unlikely(xino_itrunc->bindex < 0)) {
20018 + pr_err("no such branch %s\n", args[0].from);
20026 +/* called without aufs lock */
20027 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts)
20029 + int err, n, token;
20030 + aufs_bindex_t bindex;
20031 + unsigned char skipped;
20032 + struct dentry *root;
20033 + struct au_opt *opt, *opt_tail;
20035 + /* reduce the stack space */
20037 + struct au_opt_xino_itrunc *xino_itrunc;
20038 + struct au_opt_wbr_create *create;
20041 + substring_t args[MAX_OPT_ARGS];
20045 + a = kmalloc(sizeof(*a), GFP_NOFS);
20046 + if (unlikely(!a))
20049 + root = sb->s_root;
20053 + opt_tail = opt + opts->max_opt - 1;
20054 + opt->type = Opt_tail;
20055 + while (!err && (opt_str = strsep(&str, ",")) && *opt_str) {
20058 + token = match_token(opt_str, options, a->args);
20062 + while (!err && (opt_str = strsep(&a->args[0].from, ":"))
20064 + err = opt_add(opt, opt_str, opts->sb_flags,
20066 + if (unlikely(!err && ++opt > opt_tail)) {
20070 + opt->type = Opt_tail;
20075 + if (unlikely(match_int(&a->args[0], &n))) {
20076 + pr_err("bad integer in %s\n", opt_str);
20080 + err = opt_add(opt, a->args[1].from, opts->sb_flags,
20083 + opt->type = token;
20086 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
20087 + /*dummy bindex*/1);
20089 + opt->type = token;
20091 + case Opt_prepend:
20092 + err = opt_add(opt, a->args[0].from, opts->sb_flags,
20095 + opt->type = token;
20098 + err = au_opts_parse_del(&opt->del, a->args);
20100 + opt->type = token;
20102 +#if 0 /* reserved for future use */
20104 + del->pathname = "(indexed)";
20105 + if (unlikely(match_int(&args[0], &n))) {
20106 + pr_err("bad integer in %s\n", opt_str);
20109 + err = au_opts_parse_idel(sb, n, &opt->del, a->args);
20111 + opt->type = token;
20115 + err = au_opts_parse_mod(&opt->mod, a->args);
20117 + opt->type = token;
20119 +#ifdef IMOD /* reserved for future use */
20121 + u.mod->path = "(indexed)";
20122 + if (unlikely(match_int(&a->args[0], &n))) {
20123 + pr_err("bad integer in %s\n", opt_str);
20126 + err = au_opts_parse_imod(sb, n, &opt->mod, a->args);
20128 + opt->type = token;
20132 + err = au_opts_parse_xino(sb, &opt->xino, a->args);
20134 + opt->type = token;
20137 + case Opt_trunc_xino_path:
20138 + err = au_opts_parse_xino_itrunc_path
20139 + (sb, &opt->xino_itrunc, a->args);
20141 + opt->type = token;
20144 + case Opt_itrunc_xino:
20145 + u.xino_itrunc = &opt->xino_itrunc;
20146 + if (unlikely(match_int(&a->args[0], &n))) {
20147 + pr_err("bad integer in %s\n", opt_str);
20150 + u.xino_itrunc->bindex = n;
20151 + aufs_read_lock(root, AuLock_FLUSH);
20152 + if (n < 0 || au_sbend(sb) < n) {
20153 + pr_err("out of bounds, %d\n", n);
20154 + aufs_read_unlock(root, !AuLock_IR);
20157 + aufs_read_unlock(root, !AuLock_IR);
20159 + opt->type = token;
20163 + if (unlikely(match_int(&a->args[0], &opt->dirwh)))
20166 + opt->type = token;
20169 + case Opt_rdcache:
20170 + if (unlikely(match_int(&a->args[0], &n))) {
20171 + pr_err("bad integer in %s\n", opt_str);
20174 + if (unlikely(n > AUFS_RDCACHE_MAX)) {
20175 + pr_err("rdcache must be smaller than %d\n",
20176 + AUFS_RDCACHE_MAX);
20179 + opt->rdcache = n;
20181 + opt->type = token;
20184 + if (unlikely(match_int(&a->args[0], &n)
20186 + || n > KMALLOC_MAX_SIZE)) {
20187 + pr_err("bad integer in %s\n", opt_str);
20190 + if (unlikely(n && n < NAME_MAX)) {
20191 + pr_err("rdblk must be larger than %d\n",
20197 + opt->type = token;
20200 + if (unlikely(match_int(&a->args[0], &n)
20202 + || n * sizeof(struct hlist_head)
20203 + > KMALLOC_MAX_SIZE)) {
20204 + pr_err("bad integer in %s\n", opt_str);
20209 + opt->type = token;
20212 + case Opt_trunc_xino:
20213 + case Opt_notrunc_xino:
20215 + case Opt_trunc_xib:
20216 + case Opt_notrunc_xib:
20220 + case Opt_noplink:
20221 + case Opt_list_plink:
20224 + case Opt_diropq_a:
20225 + case Opt_diropq_w:
20226 + case Opt_warn_perm:
20227 + case Opt_nowarn_perm:
20229 + case Opt_norefrof:
20230 + case Opt_verbose:
20231 + case Opt_noverbose:
20235 + case Opt_rdblk_def:
20236 + case Opt_rdhash_def:
20238 + opt->type = token;
20242 + opt->udba = udba_val(a->args[0].from);
20243 + if (opt->udba >= 0) {
20245 + opt->type = token;
20247 + pr_err("wrong value, %s\n", opt_str);
20250 + case Opt_wbr_create:
20251 + u.create = &opt->wbr_create;
20252 + u.create->wbr_create
20253 + = au_wbr_create_val(a->args[0].from, u.create);
20254 + if (u.create->wbr_create >= 0) {
20256 + opt->type = token;
20258 + pr_err("wrong value, %s\n", opt_str);
20260 + case Opt_wbr_copyup:
20261 + opt->wbr_copyup = au_wbr_copyup_val(a->args[0].from);
20262 + if (opt->wbr_copyup >= 0) {
20264 + opt->type = token;
20266 + pr_err("wrong value, %s\n", opt_str);
20270 + pr_warning("ignored %s\n", opt_str);
20272 + case Opt_ignore_silent:
20277 + pr_err("unknown option %s\n", opt_str);
20281 + if (!err && !skipped) {
20282 + if (unlikely(++opt > opt_tail)) {
20285 + opt->type = Opt_tail;
20288 + opt->type = Opt_tail;
20294 + if (unlikely(err))
20295 + au_opts_free(opts);
20301 +static int au_opt_wbr_create(struct super_block *sb,
20302 + struct au_opt_wbr_create *create)
20305 + struct au_sbinfo *sbinfo;
20307 + SiMustWriteLock(sb);
20309 + err = 1; /* handled */
20310 + sbinfo = au_sbi(sb);
20311 + if (sbinfo->si_wbr_create_ops->fin) {
20312 + err = sbinfo->si_wbr_create_ops->fin(sb);
20317 + sbinfo->si_wbr_create = create->wbr_create;
20318 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + create->wbr_create;
20319 + switch (create->wbr_create) {
20320 + case AuWbrCreate_MFSRRV:
20321 + case AuWbrCreate_MFSRR:
20322 + sbinfo->si_wbr_mfs.mfsrr_watermark = create->mfsrr_watermark;
20324 + case AuWbrCreate_MFS:
20325 + case AuWbrCreate_MFSV:
20326 + case AuWbrCreate_PMFS:
20327 + case AuWbrCreate_PMFSV:
20328 + sbinfo->si_wbr_mfs.mfs_expire
20329 + = msecs_to_jiffies(create->mfs_second * MSEC_PER_SEC);
20333 + if (sbinfo->si_wbr_create_ops->init)
20334 + sbinfo->si_wbr_create_ops->init(sb); /* ignore */
20341 + * plus: processed without an error
20342 + * zero: unprocessed
20344 +static int au_opt_simple(struct super_block *sb, struct au_opt *opt,
20345 + struct au_opts *opts)
20348 + struct au_sbinfo *sbinfo;
20350 + SiMustWriteLock(sb);
20352 + err = 1; /* handled */
20353 + sbinfo = au_sbi(sb);
20354 + switch (opt->type) {
20356 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
20357 + sbinfo->si_mntflags |= opt->udba;
20358 + opts->given_udba |= opt->udba;
20362 + au_opt_set(sbinfo->si_mntflags, PLINK);
20364 + case Opt_noplink:
20365 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
20366 + au_plink_put(sb, /*verbose*/1);
20367 + au_opt_clr(sbinfo->si_mntflags, PLINK);
20369 + case Opt_list_plink:
20370 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
20371 + au_plink_list(sb);
20375 + au_opt_set(sbinfo->si_mntflags, DIO);
20376 + au_fset_opts(opts->flags, REFRESH_DYAOP);
20379 + au_opt_clr(sbinfo->si_mntflags, DIO);
20380 + au_fset_opts(opts->flags, REFRESH_DYAOP);
20383 + case Opt_diropq_a:
20384 + au_opt_set(sbinfo->si_mntflags, ALWAYS_DIROPQ);
20386 + case Opt_diropq_w:
20387 + au_opt_clr(sbinfo->si_mntflags, ALWAYS_DIROPQ);
20390 + case Opt_warn_perm:
20391 + au_opt_set(sbinfo->si_mntflags, WARN_PERM);
20393 + case Opt_nowarn_perm:
20394 + au_opt_clr(sbinfo->si_mntflags, WARN_PERM);
20398 + au_opt_set(sbinfo->si_mntflags, REFROF);
20400 + case Opt_norefrof:
20401 + au_opt_clr(sbinfo->si_mntflags, REFROF);
20404 + case Opt_verbose:
20405 + au_opt_set(sbinfo->si_mntflags, VERBOSE);
20407 + case Opt_noverbose:
20408 + au_opt_clr(sbinfo->si_mntflags, VERBOSE);
20412 + au_opt_set(sbinfo->si_mntflags, SUM);
20415 + au_opt_clr(sbinfo->si_mntflags, SUM);
20416 + au_opt_set(sbinfo->si_mntflags, SUM_W);
20418 + au_opt_clr(sbinfo->si_mntflags, SUM);
20419 + au_opt_clr(sbinfo->si_mntflags, SUM_W);
20422 + case Opt_wbr_create:
20423 + err = au_opt_wbr_create(sb, &opt->wbr_create);
20425 + case Opt_wbr_copyup:
20426 + sbinfo->si_wbr_copyup = opt->wbr_copyup;
20427 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + opt->wbr_copyup;
20431 + sbinfo->si_dirwh = opt->dirwh;
20434 + case Opt_rdcache:
20435 + sbinfo->si_rdcache
20436 + = msecs_to_jiffies(opt->rdcache * MSEC_PER_SEC);
20439 + sbinfo->si_rdblk = opt->rdblk;
20441 + case Opt_rdblk_def:
20442 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
20445 + sbinfo->si_rdhash = opt->rdhash;
20447 + case Opt_rdhash_def:
20448 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
20452 + au_opt_set(sbinfo->si_mntflags, SHWH);
20455 + au_opt_clr(sbinfo->si_mntflags, SHWH);
20458 + case Opt_trunc_xino:
20459 + au_opt_set(sbinfo->si_mntflags, TRUNC_XINO);
20461 + case Opt_notrunc_xino:
20462 + au_opt_clr(sbinfo->si_mntflags, TRUNC_XINO);
20465 + case Opt_trunc_xino_path:
20466 + case Opt_itrunc_xino:
20467 + err = au_xino_trunc(sb, opt->xino_itrunc.bindex);
20472 + case Opt_trunc_xib:
20473 + au_fset_opts(opts->flags, TRUNC_XIB);
20475 + case Opt_notrunc_xib:
20476 + au_fclr_opts(opts->flags, TRUNC_XIB);
20488 + * returns tri-state.
20489 + * plus: processed without an error
20490 + * zero: unprocessed
20493 +static int au_opt_br(struct super_block *sb, struct au_opt *opt,
20494 + struct au_opts *opts)
20496 + int err, do_refresh;
20499 + switch (opt->type) {
20501 + opt->add.bindex = au_sbend(sb) + 1;
20502 + if (opt->add.bindex < 0)
20503 + opt->add.bindex = 0;
20505 + case Opt_prepend:
20506 + opt->add.bindex = 0;
20509 + err = au_br_add(sb, &opt->add,
20510 + au_ftest_opts(opts->flags, REMOUNT));
20513 + au_fset_opts(opts->flags, REFRESH);
20519 + err = au_br_del(sb, &opt->del,
20520 + au_ftest_opts(opts->flags, REMOUNT));
20523 + au_fset_opts(opts->flags, TRUNC_XIB);
20524 + au_fset_opts(opts->flags, REFRESH);
20530 + err = au_br_mod(sb, &opt->mod,
20531 + au_ftest_opts(opts->flags, REMOUNT),
20536 + au_fset_opts(opts->flags, REFRESH);
20544 +static int au_opt_xino(struct super_block *sb, struct au_opt *opt,
20545 + struct au_opt_xino **opt_xino,
20546 + struct au_opts *opts)
20549 + aufs_bindex_t bend, bindex;
20550 + struct dentry *root, *parent, *h_root;
20553 + switch (opt->type) {
20555 + err = au_xino_set(sb, &opt->xino,
20556 + !!au_ftest_opts(opts->flags, REMOUNT));
20557 + if (unlikely(err))
20560 + *opt_xino = &opt->xino;
20561 + au_xino_brid_set(sb, -1);
20563 + /* safe d_parent access */
20564 + parent = opt->xino.file->f_dentry->d_parent;
20565 + root = sb->s_root;
20566 + bend = au_sbend(sb);
20567 + for (bindex = 0; bindex <= bend; bindex++) {
20568 + h_root = au_h_dptr(root, bindex);
20569 + if (h_root == parent) {
20570 + au_xino_brid_set(sb, au_sbr_id(sb, bindex));
20578 + au_xino_brid_set(sb, -1);
20579 + *opt_xino = (void *)-1;
20586 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
20587 + unsigned int pending)
20590 + aufs_bindex_t bindex, bend;
20591 + unsigned char do_plink, skip, do_free;
20592 + struct au_branch *br;
20593 + struct au_wbr *wbr;
20594 + struct dentry *root;
20595 + struct inode *dir, *h_dir;
20596 + struct au_sbinfo *sbinfo;
20597 + struct au_hinode *hdir;
20599 + SiMustAnyLock(sb);
20601 + sbinfo = au_sbi(sb);
20602 + AuDebugOn(!(sbinfo->si_mntflags & AuOptMask_UDBA));
20604 + if (!(sb_flags & MS_RDONLY)) {
20605 + if (unlikely(!au_br_writable(au_sbr_perm(sb, 0))))
20606 + pr_warning("first branch should be rw\n");
20607 + if (unlikely(au_opt_test(sbinfo->si_mntflags, SHWH)))
20608 + pr_warning("shwh should be used with ro\n");
20611 + if (au_opt_test((sbinfo->si_mntflags | pending), UDBA_HNOTIFY)
20612 + && !au_opt_test(sbinfo->si_mntflags, XINO))
20613 + pr_warning("udba=*notify requires xino\n");
20616 + root = sb->s_root;
20617 + dir = root->d_inode;
20618 + do_plink = !!au_opt_test(sbinfo->si_mntflags, PLINK);
20619 + bend = au_sbend(sb);
20620 + for (bindex = 0; !err && bindex <= bend; bindex++) {
20622 + h_dir = au_h_iptr(dir, bindex);
20623 + br = au_sbr(sb, bindex);
20626 + wbr = br->br_wbr;
20628 + wbr_wh_read_lock(wbr);
20630 + if (!au_br_writable(br->br_perm)) {
20633 + || (!wbr->wbr_whbase
20634 + && !wbr->wbr_plink
20635 + && !wbr->wbr_orph));
20636 + } else if (!au_br_wh_linkable(br->br_perm)) {
20637 + /* skip = (!br->br_whbase && !br->br_orph); */
20638 + skip = (!wbr || !wbr->wbr_whbase);
20639 + if (skip && wbr) {
20641 + skip = !!wbr->wbr_plink;
20643 + skip = !wbr->wbr_plink;
20646 + /* skip = (br->br_whbase && br->br_ohph); */
20647 + skip = (wbr && wbr->wbr_whbase);
20650 + skip = !!wbr->wbr_plink;
20652 + skip = !wbr->wbr_plink;
20656 + wbr_wh_read_unlock(wbr);
20661 + hdir = au_hi(dir, bindex);
20662 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
20664 + wbr_wh_write_lock(wbr);
20665 + err = au_wh_init(au_h_dptr(root, bindex), br, sb);
20667 + wbr_wh_write_unlock(wbr);
20668 + au_hn_imtx_unlock(hdir);
20670 + if (!err && do_free) {
20672 + br->br_wbr = NULL;
20679 +int au_opts_mount(struct super_block *sb, struct au_opts *opts)
20682 + unsigned int tmp;
20683 + aufs_bindex_t bindex, bend;
20684 + struct au_opt *opt;
20685 + struct au_opt_xino *opt_xino, xino;
20686 + struct au_sbinfo *sbinfo;
20687 + struct au_branch *br;
20689 + SiMustWriteLock(sb);
20694 + while (err >= 0 && opt->type != Opt_tail)
20695 + err = au_opt_simple(sb, opt++, opts);
20698 + else if (unlikely(err < 0))
20701 + /* disable xino and udba temporary */
20702 + sbinfo = au_sbi(sb);
20703 + tmp = sbinfo->si_mntflags;
20704 + au_opt_clr(sbinfo->si_mntflags, XINO);
20705 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_REVAL);
20708 + while (err >= 0 && opt->type != Opt_tail)
20709 + err = au_opt_br(sb, opt++, opts);
20712 + else if (unlikely(err < 0))
20715 + bend = au_sbend(sb);
20716 + if (unlikely(bend < 0)) {
20718 + pr_err("no branches\n");
20722 + if (au_opt_test(tmp, XINO))
20723 + au_opt_set(sbinfo->si_mntflags, XINO);
20725 + while (!err && opt->type != Opt_tail)
20726 + err = au_opt_xino(sb, opt++, &opt_xino, opts);
20727 + if (unlikely(err))
20730 + err = au_opts_verify(sb, sb->s_flags, tmp);
20731 + if (unlikely(err))
20734 + /* restore xino */
20735 + if (au_opt_test(tmp, XINO) && !opt_xino) {
20736 + xino.file = au_xino_def(sb);
20737 + err = PTR_ERR(xino.file);
20738 + if (IS_ERR(xino.file))
20741 + err = au_xino_set(sb, &xino, /*remount*/0);
20743 + if (unlikely(err))
20747 + /* restore udba */
20748 + tmp &= AuOptMask_UDBA;
20749 + sbinfo->si_mntflags &= ~AuOptMask_UDBA;
20750 + sbinfo->si_mntflags |= tmp;
20751 + bend = au_sbend(sb);
20752 + for (bindex = 0; bindex <= bend; bindex++) {
20753 + br = au_sbr(sb, bindex);
20754 + err = au_hnotify_reset_br(tmp, br, br->br_perm);
20755 + if (unlikely(err))
20756 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
20758 + /* go on even if err */
20760 + if (au_opt_test(tmp, UDBA_HNOTIFY)) {
20761 + struct inode *dir = sb->s_root->d_inode;
20762 + au_hn_reset(dir, au_hi_flags(dir, /*isdir*/1) & ~AuHi_XINO);
20769 +int au_opts_remount(struct super_block *sb, struct au_opts *opts)
20772 + struct inode *dir;
20773 + struct au_opt_xino *opt_xino;
20774 + struct au_opt *opt;
20775 + struct au_sbinfo *sbinfo;
20777 + SiMustWriteLock(sb);
20779 + dir = sb->s_root->d_inode;
20780 + sbinfo = au_sbi(sb);
20784 + while (err >= 0 && opt->type != Opt_tail) {
20785 + err = au_opt_simple(sb, opt, opts);
20787 + err = au_opt_br(sb, opt, opts);
20789 + err = au_opt_xino(sb, opt, &opt_xino, opts);
20795 + /* go on even err */
20797 + rerr = au_opts_verify(sb, opts->sb_flags, /*pending*/0);
20798 + if (unlikely(rerr && !err))
20801 + if (au_ftest_opts(opts->flags, TRUNC_XIB)) {
20802 + rerr = au_xib_trunc(sb);
20803 + if (unlikely(rerr && !err))
20807 + /* will be handled by the caller */
20808 + if (!au_ftest_opts(opts->flags, REFRESH)
20809 + && (opts->given_udba || au_opt_test(sbinfo->si_mntflags, XINO)))
20810 + au_fset_opts(opts->flags, REFRESH);
20812 + AuDbg("status 0x%x\n", opts->flags);
20816 +/* ---------------------------------------------------------------------- */
20818 +unsigned int au_opt_udba(struct super_block *sb)
20820 + return au_mntflags(sb) & AuOptMask_UDBA;
20822 diff -urN /usr/share/empty/fs/aufs/opts.h linux/fs/aufs/opts.h
20823 --- /usr/share/empty/fs/aufs/opts.h 1970-01-01 01:00:00.000000000 +0100
20824 +++ linux/fs/aufs/opts.h 2011-10-24 20:51:51.580466925 +0200
20827 + * Copyright (C) 2005-2011 Junjiro R. Okajima
20829 + * This program, aufs is free software; you can redistribute it and/or modify
20830 + * it under the terms of the GNU General Public License as published by
20831 + * the Free Software Foundation; either version 2 of the License, or
20832 + * (at your option) any later version.
20834 + * This program is distributed in the hope that it will be useful,
20835 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
20836 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20837 + * GNU General Public License for more details.
20839 + * You should have received a copy of the GNU General Public License
20840 + * along with this program; if not, write to the Free Software
20841 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20845 + * mount options/flags
20848 +#ifndef __AUFS_OPTS_H__
20849 +#define __AUFS_OPTS_H__
20853 +#include <linux/path.h>
20854 +#include <linux/aufs_type.h>
20857 +struct super_block;
20859 +/* ---------------------------------------------------------------------- */
20862 +#define AuOpt_XINO 1 /* external inode number bitmap
20863 + and translation table */
20864 +#define AuOpt_TRUNC_XINO (1 << 1) /* truncate xino files */
20865 +#define AuOpt_UDBA_NONE (1 << 2) /* users direct branch access */
20866 +#define AuOpt_UDBA_REVAL (1 << 3)
20867 +#define AuOpt_UDBA_HNOTIFY (1 << 4)
20868 +#define AuOpt_SHWH (1 << 5) /* show whiteout */
20869 +#define AuOpt_PLINK (1 << 6) /* pseudo-link */
20870 +#define AuOpt_DIRPERM1 (1 << 7) /* unimplemented */
20871 +#define AuOpt_REFROF (1 << 8) /* unimplemented */
20872 +#define AuOpt_ALWAYS_DIROPQ (1 << 9) /* policy to creating diropq */
20873 +#define AuOpt_SUM (1 << 10) /* summation for statfs(2) */
20874 +#define AuOpt_SUM_W (1 << 11) /* unimplemented */
20875 +#define AuOpt_WARN_PERM (1 << 12) /* warn when add-branch */
20876 +#define AuOpt_VERBOSE (1 << 13) /* busy inode when del-branch */
20877 +#define AuOpt_DIO (1 << 14) /* direct io */
20879 +#ifndef CONFIG_AUFS_HNOTIFY
20880 +#undef AuOpt_UDBA_HNOTIFY
20881 +#define AuOpt_UDBA_HNOTIFY 0
20883 +#ifndef CONFIG_AUFS_SHWH
20885 +#define AuOpt_SHWH 0
20888 +#define AuOpt_Def (AuOpt_XINO \
20889 + | AuOpt_UDBA_REVAL \
20891 + /* | AuOpt_DIRPERM1 */ \
20892 + | AuOpt_WARN_PERM)
20893 +#define AuOptMask_UDBA (AuOpt_UDBA_NONE \
20894 + | AuOpt_UDBA_REVAL \
20895 + | AuOpt_UDBA_HNOTIFY)
20897 +#define au_opt_test(flags, name) (flags & AuOpt_##name)
20898 +#define au_opt_set(flags, name) do { \
20899 + BUILD_BUG_ON(AuOpt_##name & AuOptMask_UDBA); \
20900 + ((flags) |= AuOpt_##name); \
20902 +#define au_opt_set_udba(flags, name) do { \
20903 + (flags) &= ~AuOptMask_UDBA; \
20904 + ((flags) |= AuOpt_##name); \
20906 +#define au_opt_clr(flags, name) do { \
20907 + ((flags) &= ~AuOpt_##name); \
20910 +static inline unsigned int au_opts_plink(unsigned int mntflags)
20912 +#ifdef CONFIG_PROC_FS
20915 + return mntflags & ~AuOpt_PLINK;
20919 +/* ---------------------------------------------------------------------- */
20921 +/* policies to select one among multiple writable branches */
20923 + AuWbrCreate_TDP, /* top down parent */
20924 + AuWbrCreate_RR, /* round robin */
20925 + AuWbrCreate_MFS, /* most free space */
20926 + AuWbrCreate_MFSV, /* mfs with seconds */
20927 + AuWbrCreate_MFSRR, /* mfs then rr */
20928 + AuWbrCreate_MFSRRV, /* mfs then rr with seconds */
20929 + AuWbrCreate_PMFS, /* parent and mfs */
20930 + AuWbrCreate_PMFSV, /* parent and mfs with seconds */
20932 + AuWbrCreate_Def = AuWbrCreate_TDP
20936 + AuWbrCopyup_TDP, /* top down parent */
20937 + AuWbrCopyup_BUP, /* bottom up parent */
20938 + AuWbrCopyup_BU, /* bottom up */
20940 + AuWbrCopyup_Def = AuWbrCopyup_TDP
20943 +/* ---------------------------------------------------------------------- */
20945 +struct au_opt_add {
20946 + aufs_bindex_t bindex;
20949 + struct path path;
20952 +struct au_opt_del {
20954 + struct path h_path;
20957 +struct au_opt_mod {
20960 + struct dentry *h_root;
20963 +struct au_opt_xino {
20965 + struct file *file;
20968 +struct au_opt_xino_itrunc {
20969 + aufs_bindex_t bindex;
20972 +struct au_opt_wbr_create {
20975 + unsigned long long mfsrr_watermark;
20981 + struct au_opt_xino xino;
20982 + struct au_opt_xino_itrunc xino_itrunc;
20983 + struct au_opt_add add;
20984 + struct au_opt_del del;
20985 + struct au_opt_mod mod;
20988 + unsigned int rdblk;
20989 + unsigned int rdhash;
20991 + struct au_opt_wbr_create wbr_create;
20997 +#define AuOpts_REMOUNT 1
20998 +#define AuOpts_REFRESH (1 << 1)
20999 +#define AuOpts_TRUNC_XIB (1 << 2)
21000 +#define AuOpts_REFRESH_DYAOP (1 << 3)
21001 +#define au_ftest_opts(flags, name) ((flags) & AuOpts_##name)
21002 +#define au_fset_opts(flags, name) \
21003 + do { (flags) |= AuOpts_##name; } while (0)
21004 +#define au_fclr_opts(flags, name) \
21005 + do { (flags) &= ~AuOpts_##name; } while (0)
21008 + struct au_opt *opt;
21011 + unsigned int given_udba;
21012 + unsigned int flags;
21013 + unsigned long sb_flags;
21016 +/* ---------------------------------------------------------------------- */
21018 +char *au_optstr_br_perm(int brperm);
21019 +const char *au_optstr_udba(int udba);
21020 +const char *au_optstr_wbr_copyup(int wbr_copyup);
21021 +const char *au_optstr_wbr_create(int wbr_create);
21023 +void au_opts_free(struct au_opts *opts);
21024 +int au_opts_parse(struct super_block *sb, char *str, struct au_opts *opts);
21025 +int au_opts_verify(struct super_block *sb, unsigned long sb_flags,
21026 + unsigned int pending);
21027 +int au_opts_mount(struct super_block *sb, struct au_opts *opts);
21028 +int au_opts_remount(struct super_block *sb, struct au_opts *opts);
21030 +unsigned int au_opt_udba(struct super_block *sb);
21032 +/* ---------------------------------------------------------------------- */
21034 +#endif /* __KERNEL__ */
21035 +#endif /* __AUFS_OPTS_H__ */
21036 diff -urN /usr/share/empty/fs/aufs/plink.c linux/fs/aufs/plink.c
21037 --- /usr/share/empty/fs/aufs/plink.c 1970-01-01 01:00:00.000000000 +0100
21038 +++ linux/fs/aufs/plink.c 2011-08-24 13:30:24.734646739 +0200
21041 + * Copyright (C) 2005-2011 Junjiro R. Okajima
21043 + * This program, aufs is free software; you can redistribute it and/or modify
21044 + * it under the terms of the GNU General Public License as published by
21045 + * the Free Software Foundation; either version 2 of the License, or
21046 + * (at your option) any later version.
21048 + * This program is distributed in the hope that it will be useful,
21049 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21050 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21051 + * GNU General Public License for more details.
21053 + * You should have received a copy of the GNU General Public License
21054 + * along with this program; if not, write to the Free Software
21055 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21065 + * the pseudo-link maintenance mode.
21066 + * during a user process maintains the pseudo-links,
21067 + * prohibit adding a new plink and branch manipulation.
21071 + * For entry functions which will handle plink, and i_mutex is already held
21073 + * They cannot wait and should return an error at once.
21074 + * Callers has to check the error.
21076 + * For entry functions which will handle plink, but i_mutex is not held
21078 + * They can wait the plink maintenance mode to finish.
21080 + * They behave like F_SETLK and F_SETLKW.
21081 + * If the caller never handle plink, then both flags are unnecessary.
21084 +int au_plink_maint(struct super_block *sb, int flags)
21088 + struct au_sbinfo *sbi;
21090 + SiMustAnyLock(sb);
21093 + if (!au_opt_test(au_mntflags(sb), PLINK))
21096 + sbi = au_sbi(sb);
21097 + pid = sbi->si_plink_maint_pid;
21098 + if (!pid || pid == current->pid)
21101 + /* todo: it highly depends upon /sbin/mount.aufs */
21103 + ppid = task_pid_vnr(rcu_dereference(current->real_parent));
21104 + rcu_read_unlock();
21108 + if (au_ftest_lock(flags, NOPLMW)) {
21109 + /* if there is no i_mutex lock in VFS, we don't need to wait */
21110 + /* AuDebugOn(!lockdep_depth(current)); */
21111 + while (sbi->si_plink_maint_pid) {
21112 + si_read_unlock(sb);
21113 + /* gave up wake_up_bit() */
21114 + wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid);
21116 + if (au_ftest_lock(flags, FLUSH))
21117 + au_nwt_flush(&sbi->si_nowait);
21118 + si_noflush_read_lock(sb);
21120 + } else if (au_ftest_lock(flags, NOPLM)) {
21121 + AuDbg("ppid %d, pid %d\n", ppid, pid);
21129 +void au_plink_maint_leave(struct au_sbinfo *sbinfo)
21131 + spin_lock(&sbinfo->si_plink_maint_lock);
21132 + sbinfo->si_plink_maint_pid = 0;
21133 + spin_unlock(&sbinfo->si_plink_maint_lock);
21134 + wake_up_all(&sbinfo->si_plink_wq);
21137 +int au_plink_maint_enter(struct super_block *sb)
21140 + struct au_sbinfo *sbinfo;
21143 + sbinfo = au_sbi(sb);
21144 + /* make sure i am the only one in this fs */
21145 + si_write_lock(sb, AuLock_FLUSH);
21146 + if (au_opt_test(au_mntflags(sb), PLINK)) {
21147 + spin_lock(&sbinfo->si_plink_maint_lock);
21148 + if (!sbinfo->si_plink_maint_pid)
21149 + sbinfo->si_plink_maint_pid = current->pid;
21152 + spin_unlock(&sbinfo->si_plink_maint_lock);
21154 + si_write_unlock(sb);
21159 +/* ---------------------------------------------------------------------- */
21161 +struct pseudo_link {
21163 + struct list_head list;
21164 + struct rcu_head rcu;
21166 + struct inode *inode;
21169 +#ifdef CONFIG_AUFS_DEBUG
21170 +void au_plink_list(struct super_block *sb)
21172 + struct au_sbinfo *sbinfo;
21173 + struct list_head *plink_list;
21174 + struct pseudo_link *plink;
21176 + SiMustAnyLock(sb);
21178 + sbinfo = au_sbi(sb);
21179 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
21180 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
21182 + plink_list = &sbinfo->si_plink.head;
21184 + list_for_each_entry_rcu(plink, plink_list, list)
21185 + AuDbg("%lu\n", plink->inode->i_ino);
21186 + rcu_read_unlock();
21190 +/* is the inode pseudo-linked? */
21191 +int au_plink_test(struct inode *inode)
21194 + struct au_sbinfo *sbinfo;
21195 + struct list_head *plink_list;
21196 + struct pseudo_link *plink;
21198 + sbinfo = au_sbi(inode->i_sb);
21199 + AuRwMustAnyLock(&sbinfo->si_rwsem);
21200 + AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK));
21201 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
21204 + plink_list = &sbinfo->si_plink.head;
21206 + list_for_each_entry_rcu(plink, plink_list, list)
21207 + if (plink->inode == inode) {
21211 + rcu_read_unlock();
21215 +/* ---------------------------------------------------------------------- */
21218 + * generate a name for plink.
21219 + * the file will be stored under AUFS_WH_PLINKDIR.
21221 +/* 20 is max digits length of ulong 64 */
21222 +#define PLINK_NAME_LEN ((20 + 1) * 2)
21224 +static int plink_name(char *name, int len, struct inode *inode,
21225 + aufs_bindex_t bindex)
21228 + struct inode *h_inode;
21230 + h_inode = au_h_iptr(inode, bindex);
21231 + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino);
21235 +struct au_do_plink_lkup_args {
21236 + struct dentry **errp;
21237 + struct qstr *tgtname;
21238 + struct dentry *h_parent;
21239 + struct au_branch *br;
21242 +static struct dentry *au_do_plink_lkup(struct qstr *tgtname,
21243 + struct dentry *h_parent,
21244 + struct au_branch *br)
21246 + struct dentry *h_dentry;
21247 + struct mutex *h_mtx;
21249 + h_mtx = &h_parent->d_inode->i_mutex;
21250 + mutex_lock_nested(h_mtx, AuLsc_I_CHILD2);
21251 + h_dentry = au_lkup_one(tgtname, h_parent, br, /*nd*/NULL);
21252 + mutex_unlock(h_mtx);
21256 +static void au_call_do_plink_lkup(void *args)
21258 + struct au_do_plink_lkup_args *a = args;
21259 + *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br);
21262 +/* lookup the plink-ed @inode under the branch at @bindex */
21263 +struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex)
21265 + struct dentry *h_dentry, *h_parent;
21266 + struct au_branch *br;
21267 + struct inode *h_dir;
21269 + char a[PLINK_NAME_LEN];
21270 + struct qstr tgtname = {
21274 + AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM));
21276 + br = au_sbr(inode->i_sb, bindex);
21277 + h_parent = br->br_wbr->wbr_plink;
21278 + h_dir = h_parent->d_inode;
21279 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
21281 + if (current_fsuid()) {
21282 + struct au_do_plink_lkup_args args = {
21283 + .errp = &h_dentry,
21284 + .tgtname = &tgtname,
21285 + .h_parent = h_parent,
21289 + wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args);
21290 + if (unlikely(wkq_err))
21291 + h_dentry = ERR_PTR(wkq_err);
21293 + h_dentry = au_do_plink_lkup(&tgtname, h_parent, br);
21298 +/* create a pseudo-link */
21299 +static int do_whplink(struct qstr *tgt, struct dentry *h_parent,
21300 + struct dentry *h_dentry, struct au_branch *br)
21303 + struct path h_path = {
21304 + .mnt = br->br_mnt
21306 + struct inode *h_dir;
21308 + h_dir = h_parent->d_inode;
21309 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_CHILD2);
21311 + h_path.dentry = au_lkup_one(tgt, h_parent, br, /*nd*/NULL);
21312 + err = PTR_ERR(h_path.dentry);
21313 + if (IS_ERR(h_path.dentry))
21317 + /* wh.plink dir is not monitored */
21318 + /* todo: is it really safe? */
21319 + if (h_path.dentry->d_inode
21320 + && h_path.dentry->d_inode != h_dentry->d_inode) {
21321 + err = vfsub_unlink(h_dir, &h_path, /*force*/0);
21322 + dput(h_path.dentry);
21323 + h_path.dentry = NULL;
21327 + if (!err && !h_path.dentry->d_inode)
21328 + err = vfsub_link(h_dentry, h_dir, &h_path);
21329 + dput(h_path.dentry);
21332 + mutex_unlock(&h_dir->i_mutex);
21336 +struct do_whplink_args {
21338 + struct qstr *tgt;
21339 + struct dentry *h_parent;
21340 + struct dentry *h_dentry;
21341 + struct au_branch *br;
21344 +static void call_do_whplink(void *args)
21346 + struct do_whplink_args *a = args;
21347 + *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br);
21350 +static int whplink(struct dentry *h_dentry, struct inode *inode,
21351 + aufs_bindex_t bindex, struct au_branch *br)
21353 + int err, wkq_err;
21354 + struct au_wbr *wbr;
21355 + struct dentry *h_parent;
21356 + struct inode *h_dir;
21357 + char a[PLINK_NAME_LEN];
21358 + struct qstr tgtname = {
21362 + wbr = au_sbr(inode->i_sb, bindex)->br_wbr;
21363 + h_parent = wbr->wbr_plink;
21364 + h_dir = h_parent->d_inode;
21365 + tgtname.len = plink_name(a, sizeof(a), inode, bindex);
21367 + /* always superio. */
21368 + if (current_fsuid()) {
21369 + struct do_whplink_args args = {
21372 + .h_parent = h_parent,
21373 + .h_dentry = h_dentry,
21376 + wkq_err = au_wkq_wait(call_do_whplink, &args);
21377 + if (unlikely(wkq_err))
21380 + err = do_whplink(&tgtname, h_parent, h_dentry, br);
21385 +/* free a single plink */
21386 +static void do_put_plink(struct pseudo_link *plink, int do_del)
21389 + list_del(&plink->list);
21390 + iput(plink->inode);
21394 +static void do_put_plink_rcu(struct rcu_head *rcu)
21396 + struct pseudo_link *plink;
21398 + plink = container_of(rcu, struct pseudo_link, rcu);
21399 + iput(plink->inode);
21404 + * create a new pseudo-link for @h_dentry on @bindex.
21405 + * the linked inode is held in aufs @inode.
21407 +void au_plink_append(struct inode *inode, aufs_bindex_t bindex,
21408 + struct dentry *h_dentry)
21410 + struct super_block *sb;
21411 + struct au_sbinfo *sbinfo;
21412 + struct list_head *plink_list;
21413 + struct pseudo_link *plink, *tmp;
21414 + int found, err, cnt;
21416 + sb = inode->i_sb;
21417 + sbinfo = au_sbi(sb);
21418 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
21419 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
21423 + plink_list = &sbinfo->si_plink.head;
21425 + list_for_each_entry_rcu(plink, plink_list, list) {
21427 + if (plink->inode == inode) {
21432 + rcu_read_unlock();
21436 + tmp = kmalloc(sizeof(*plink), GFP_NOFS);
21438 + tmp->inode = au_igrab(inode);
21444 + spin_lock(&sbinfo->si_plink.spin);
21445 + list_for_each_entry(plink, plink_list, list) {
21446 + if (plink->inode == inode) {
21452 + list_add_rcu(&tmp->list, plink_list);
21453 + spin_unlock(&sbinfo->si_plink.spin);
21456 + WARN_ONCE(cnt > AUFS_PLINK_WARN,
21457 + "unexpectedly many pseudo links, %d\n", cnt);
21458 + err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex));
21460 + do_put_plink(tmp, 0);
21465 + if (unlikely(err)) {
21466 + pr_warning("err %d, damaged pseudo link.\n", err);
21468 + au_spl_del_rcu(&tmp->list, &sbinfo->si_plink);
21469 + call_rcu(&tmp->rcu, do_put_plink_rcu);
21474 +/* free all plinks */
21475 +void au_plink_put(struct super_block *sb, int verbose)
21477 + struct au_sbinfo *sbinfo;
21478 + struct list_head *plink_list;
21479 + struct pseudo_link *plink, *tmp;
21481 + SiMustWriteLock(sb);
21483 + sbinfo = au_sbi(sb);
21484 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
21485 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
21487 + plink_list = &sbinfo->si_plink.head;
21488 + /* no spin_lock since sbinfo is write-locked */
21489 + WARN(verbose && !list_empty(plink_list), "pseudo-link is not flushed");
21490 + list_for_each_entry_safe(plink, tmp, plink_list, list)
21491 + do_put_plink(plink, 0);
21492 + INIT_LIST_HEAD(plink_list);
21495 +void au_plink_clean(struct super_block *sb, int verbose)
21497 + struct dentry *root;
21499 + root = sb->s_root;
21500 + aufs_write_lock(root);
21501 + if (au_opt_test(au_mntflags(sb), PLINK))
21502 + au_plink_put(sb, verbose);
21503 + aufs_write_unlock(root);
21506 +/* free the plinks on a branch specified by @br_id */
21507 +void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id)
21509 + struct au_sbinfo *sbinfo;
21510 + struct list_head *plink_list;
21511 + struct pseudo_link *plink, *tmp;
21512 + struct inode *inode;
21513 + aufs_bindex_t bstart, bend, bindex;
21514 + unsigned char do_put;
21516 + SiMustWriteLock(sb);
21518 + sbinfo = au_sbi(sb);
21519 + AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK));
21520 + AuDebugOn(au_plink_maint(sb, AuLock_NOPLM));
21522 + plink_list = &sbinfo->si_plink.head;
21523 + /* no spin_lock since sbinfo is write-locked */
21524 + list_for_each_entry_safe(plink, tmp, plink_list, list) {
21526 + inode = au_igrab(plink->inode);
21527 + ii_write_lock_child(inode);
21528 + bstart = au_ibstart(inode);
21529 + bend = au_ibend(inode);
21530 + if (bstart >= 0) {
21531 + for (bindex = bstart; bindex <= bend; bindex++) {
21532 + if (!au_h_iptr(inode, bindex)
21533 + || au_ii_br_id(inode, bindex) != br_id)
21535 + au_set_h_iptr(inode, bindex, NULL, 0);
21540 + do_put_plink(plink, 1);
21543 + for (bindex = bstart; bindex <= bend; bindex++)
21544 + if (au_h_iptr(inode, bindex)) {
21549 + do_put_plink(plink, 1);
21551 + ii_write_unlock(inode);
21555 diff -urN /usr/share/empty/fs/aufs/poll.c linux/fs/aufs/poll.c
21556 --- /usr/share/empty/fs/aufs/poll.c 1970-01-01 01:00:00.000000000 +0100
21557 +++ linux/fs/aufs/poll.c 2011-08-24 13:30:24.734646739 +0200
21560 + * Copyright (C) 2005-2011 Junjiro R. Okajima
21562 + * This program, aufs is free software; you can redistribute it and/or modify
21563 + * it under the terms of the GNU General Public License as published by
21564 + * the Free Software Foundation; either version 2 of the License, or
21565 + * (at your option) any later version.
21567 + * This program is distributed in the hope that it will be useful,
21568 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21569 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21570 + * GNU General Public License for more details.
21572 + * You should have received a copy of the GNU General Public License
21573 + * along with this program; if not, write to the Free Software
21574 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21579 + * There is only one filesystem which implements ->poll operation, currently.
21584 +unsigned int aufs_poll(struct file *file, poll_table *wait)
21586 + unsigned int mask;
21588 + struct file *h_file;
21589 + struct dentry *dentry;
21590 + struct super_block *sb;
21592 + /* We should pretend an error happened. */
21593 + mask = POLLERR /* | POLLIN | POLLOUT */;
21594 + dentry = file->f_dentry;
21595 + sb = dentry->d_sb;
21596 + si_read_lock(sb, AuLock_FLUSH | AuLock_NOPLMW);
21597 + err = au_reval_and_lock_fdi(file, au_reopen_nondir, /*wlock*/0);
21598 + if (unlikely(err))
21601 + /* it is not an error if h_file has no operation */
21602 + mask = DEFAULT_POLLMASK;
21603 + h_file = au_hf_top(file);
21604 + if (h_file->f_op && h_file->f_op->poll)
21605 + mask = h_file->f_op->poll(h_file, wait);
21607 + di_read_unlock(dentry, AuLock_IR);
21608 + fi_read_unlock(file);
21611 + si_read_unlock(sb);
21612 + AuTraceErr((int)mask);
21615 diff -urN /usr/share/empty/fs/aufs/procfs.c linux/fs/aufs/procfs.c
21616 --- /usr/share/empty/fs/aufs/procfs.c 1970-01-01 01:00:00.000000000 +0100
21617 +++ linux/fs/aufs/procfs.c 2012-01-05 11:00:59.218835185 +0100
21620 + * Copyright (C) 2010-2011 Junjiro R. Okajima
21622 + * This program, aufs is free software; you can redistribute it and/or modify
21623 + * it under the terms of the GNU General Public License as published by
21624 + * the Free Software Foundation; either version 2 of the License, or
21625 + * (at your option) any later version.
21627 + * This program is distributed in the hope that it will be useful,
21628 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21629 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21630 + * GNU General Public License for more details.
21632 + * You should have received a copy of the GNU General Public License
21633 + * along with this program; if not, write to the Free Software
21634 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21638 + * procfs interfaces
21641 +#include <linux/proc_fs.h>
21644 +static int au_procfs_plm_release(struct inode *inode, struct file *file)
21646 + struct au_sbinfo *sbinfo;
21648 + sbinfo = file->private_data;
21650 + au_plink_maint_leave(sbinfo);
21651 + kobject_put(&sbinfo->si_kobj);
21657 +static void au_procfs_plm_write_clean(struct file *file)
21659 + struct au_sbinfo *sbinfo;
21661 + sbinfo = file->private_data;
21663 + au_plink_clean(sbinfo->si_sb, /*verbose*/0);
21666 +static int au_procfs_plm_write_si(struct file *file, unsigned long id)
21669 + struct super_block *sb;
21670 + struct au_sbinfo *sbinfo;
21673 + if (unlikely(file->private_data))
21677 + /* don't use au_sbilist_lock() here */
21678 + spin_lock(&au_sbilist.spin);
21679 + list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
21680 + if (id == sysaufs_si_id(sbinfo)) {
21681 + kobject_get(&sbinfo->si_kobj);
21682 + sb = sbinfo->si_sb;
21685 + spin_unlock(&au_sbilist.spin);
21688 + if (unlikely(!sb))
21691 + err = au_plink_maint_enter(sb);
21693 + /* keep kobject_get() */
21694 + file->private_data = sbinfo;
21696 + kobject_put(&sbinfo->si_kobj);
21702 + * Accept a valid "si=xxxx" only.
21703 + * Once it is accepted successfully, accept "clean" too.
21705 +static ssize_t au_procfs_plm_write(struct file *file, const char __user *ubuf,
21706 + size_t count, loff_t *ppos)
21709 + unsigned long id;
21710 + /* last newline is allowed */
21711 + char buf[3 + sizeof(unsigned long) * 2 + 1];
21714 + if (unlikely(!capable(CAP_SYS_ADMIN)))
21718 + if (unlikely(count > sizeof(buf)))
21721 + err = copy_from_user(buf, ubuf, count);
21722 + if (unlikely(err)) {
21729 + if (!strcmp("clean", buf)) {
21730 + au_procfs_plm_write_clean(file);
21731 + goto out_success;
21732 + } else if (unlikely(strncmp("si=", buf, 3)))
21735 + err = kstrtoul(buf + 3, 16, &id);
21736 + if (unlikely(err))
21739 + err = au_procfs_plm_write_si(file, id);
21740 + if (unlikely(err))
21744 + err = count; /* success */
21749 +static const struct file_operations au_procfs_plm_fop = {
21750 + .write = au_procfs_plm_write,
21751 + .release = au_procfs_plm_release,
21752 + .owner = THIS_MODULE
21755 +/* ---------------------------------------------------------------------- */
21757 +static struct proc_dir_entry *au_procfs_dir;
21759 +void au_procfs_fin(void)
21761 + remove_proc_entry(AUFS_PLINK_MAINT_NAME, au_procfs_dir);
21762 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
21765 +int __init au_procfs_init(void)
21768 + struct proc_dir_entry *entry;
21771 + au_procfs_dir = proc_mkdir(AUFS_PLINK_MAINT_DIR, NULL);
21772 + if (unlikely(!au_procfs_dir))
21775 + entry = proc_create(AUFS_PLINK_MAINT_NAME, S_IFREG | S_IWUSR,
21776 + au_procfs_dir, &au_procfs_plm_fop);
21777 + if (unlikely(!entry))
21781 + goto out; /* success */
21785 + remove_proc_entry(AUFS_PLINK_MAINT_DIR, NULL);
21789 diff -urN /usr/share/empty/fs/aufs/rdu.c linux/fs/aufs/rdu.c
21790 --- /usr/share/empty/fs/aufs/rdu.c 1970-01-01 01:00:00.000000000 +0100
21791 +++ linux/fs/aufs/rdu.c 2012-01-05 11:00:59.218835185 +0100
21794 + * Copyright (C) 2005-2011 Junjiro R. Okajima
21796 + * This program, aufs is free software; you can redistribute it and/or modify
21797 + * it under the terms of the GNU General Public License as published by
21798 + * the Free Software Foundation; either version 2 of the License, or
21799 + * (at your option) any later version.
21801 + * This program is distributed in the hope that it will be useful,
21802 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
21803 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21804 + * GNU General Public License for more details.
21806 + * You should have received a copy of the GNU General Public License
21807 + * along with this program; if not, write to the Free Software
21808 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21812 + * readdir in userspace.
21815 +#include <linux/compat.h>
21816 +#include <linux/fs_stack.h>
21817 +#include <linux/security.h>
21818 +#include <linux/uaccess.h>
21819 +#include <linux/aufs_type.h>
21822 +/* bits for struct aufs_rdu.flags */
21823 +#define AuRdu_CALLED 1
21824 +#define AuRdu_CONT (1 << 1)
21825 +#define AuRdu_FULL (1 << 2)
21826 +#define au_ftest_rdu(flags, name) ((flags) & AuRdu_##name)
21827 +#define au_fset_rdu(flags, name) \
21828 + do { (flags) |= AuRdu_##name; } while (0)
21829 +#define au_fclr_rdu(flags, name) \
21830 + do { (flags) &= ~AuRdu_##name; } while (0)
21832 +struct au_rdu_arg {
21833 + struct aufs_rdu *rdu;
21834 + union au_rdu_ent_ul ent;
21835 + unsigned long end;
21837 + struct super_block *sb;
21841 +static int au_rdu_fill(void *__arg, const char *name, int nlen,
21842 + loff_t offset, u64 h_ino, unsigned int d_type)
21845 + struct au_rdu_arg *arg = __arg;
21846 + struct aufs_rdu *rdu = arg->rdu;
21847 + struct au_rdu_ent ent;
21851 + au_fset_rdu(rdu->cookie.flags, CALLED);
21852 + len = au_rdu_len(nlen);
21853 + if (arg->ent.ul + len < arg->end) {
21855 + ent.bindex = rdu->cookie.bindex;
21856 + ent.type = d_type;
21858 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
21859 + ent.type = DT_UNKNOWN;
21861 + /* unnecessary to support mmap_sem since this is a dir */
21863 + if (copy_to_user(arg->ent.e, &ent, sizeof(ent)))
21865 + if (copy_to_user(arg->ent.e->name, name, nlen))
21867 + /* the terminating NULL */
21868 + if (__put_user(0, arg->ent.e->name + nlen))
21871 + /* AuDbg("%p, %.*s\n", arg->ent.p, nlen, name); */
21872 + arg->ent.ul += len;
21876 + au_fset_rdu(rdu->cookie.flags, FULL);
21878 + rdu->tail = arg->ent;
21882 + /* AuTraceErr(err); */
21886 +static int au_rdu_do(struct file *h_file, struct au_rdu_arg *arg)
21890 + struct au_rdu_cookie *cookie = &arg->rdu->cookie;
21892 + offset = vfsub_llseek(h_file, cookie->h_pos, SEEK_SET);
21894 + if (unlikely(offset != cookie->h_pos))
21900 + au_fclr_rdu(cookie->flags, CALLED);
21902 + err = vfsub_readdir(h_file, au_rdu_fill, arg);
21906 + && au_ftest_rdu(cookie->flags, CALLED)
21907 + && !au_ftest_rdu(cookie->flags, FULL));
21908 + cookie->h_pos = h_file->f_pos;
21915 +static int au_rdu(struct file *file, struct aufs_rdu *rdu)
21918 + aufs_bindex_t bend;
21919 + struct au_rdu_arg arg;
21920 + struct dentry *dentry;
21921 + struct inode *inode;
21922 + struct file *h_file;
21923 + struct au_rdu_cookie *cookie = &rdu->cookie;
21925 + err = !access_ok(VERIFY_WRITE, rdu->ent.e, rdu->sz);
21926 + if (unlikely(err)) {
21932 + rdu->tail = rdu->ent;
21935 + arg.ent = rdu->ent;
21936 + arg.end = arg.ent.ul;
21937 + arg.end += rdu->sz;
21940 + if (unlikely(!file->f_op || !file->f_op->readdir))
21943 + err = security_file_permission(file, MAY_READ);
21945 + if (unlikely(err))
21948 + dentry = file->f_dentry;
21949 + inode = dentry->d_inode;
21951 + mutex_lock(&inode->i_mutex);
21953 + err = mutex_lock_killable(&inode->i_mutex);
21955 + if (unlikely(err))
21959 + arg.sb = inode->i_sb;
21960 + err = si_read_lock(arg.sb, AuLock_FLUSH | AuLock_NOPLM);
21961 + if (unlikely(err))
21963 + err = au_alive_dir(dentry);
21964 + if (unlikely(err))
21966 + /* todo: reval? */
21967 + fi_read_lock(file);
21970 + if (unlikely(au_ftest_rdu(cookie->flags, CONT)
21971 + && cookie->generation != au_figen(file)))
21976 + rdu->blk = au_sbi(arg.sb)->si_rdblk;
21978 + rdu->blk = au_dir_size(file, /*dentry*/NULL);
21980 + bend = au_fbstart(file);
21981 + if (cookie->bindex < bend)
21982 + cookie->bindex = bend;
21983 + bend = au_fbend_dir(file);
21984 + /* AuDbg("b%d, b%d\n", cookie->bindex, bend); */
21985 + for (; !err && cookie->bindex <= bend;
21986 + cookie->bindex++, cookie->h_pos = 0) {
21987 + h_file = au_hf_dir(file, cookie->bindex);
21991 + au_fclr_rdu(cookie->flags, FULL);
21992 + err = au_rdu_do(h_file, &arg);
21994 + if (unlikely(au_ftest_rdu(cookie->flags, FULL) || err))
21997 + AuDbg("rent %llu\n", rdu->rent);
21999 + if (!err && !au_ftest_rdu(cookie->flags, CONT)) {
22000 + rdu->shwh = !!au_opt_test(au_sbi(arg.sb)->si_mntflags, SHWH);
22001 + au_fset_rdu(cookie->flags, CONT);
22002 + cookie->generation = au_figen(file);
22005 + ii_read_lock_child(inode);
22006 + fsstack_copy_attr_atime(inode, au_h_iptr(inode, au_ibstart(inode)));
22007 + ii_read_unlock(inode);
22010 + fi_read_unlock(file);
22012 + si_read_unlock(arg.sb);
22014 + mutex_unlock(&inode->i_mutex);
22020 +static int au_rdu_ino(struct file *file, struct aufs_rdu *rdu)
22024 + unsigned long long nent;
22025 + union au_rdu_ent_ul *u;
22026 + struct au_rdu_ent ent;
22027 + struct super_block *sb;
22030 + nent = rdu->nent;
22032 + sb = file->f_dentry->d_sb;
22033 + si_read_lock(sb, AuLock_FLUSH);
22034 + while (nent-- > 0) {
22035 + /* unnecessary to support mmap_sem since this is a dir */
22036 + err = copy_from_user(&ent, u->e, sizeof(ent));
22038 + err = !access_ok(VERIFY_WRITE, &u->e->ino, sizeof(ino));
22039 + if (unlikely(err)) {
22045 + /* AuDbg("b%d, i%llu\n", ent.bindex, ent.ino); */
22047 + err = au_ino(sb, ent.bindex, ent.ino, ent.type, &ino);
22049 + err = au_wh_ino(sb, ent.bindex, ent.ino, ent.type,
22051 + if (unlikely(err)) {
22056 + err = __put_user(ino, &u->e->ino);
22057 + if (unlikely(err)) {
22062 + u->ul += au_rdu_len(ent.nlen);
22064 + si_read_unlock(sb);
22069 +/* ---------------------------------------------------------------------- */
22071 +static int au_rdu_verify(struct aufs_rdu *rdu)
22073 + AuDbg("rdu{%llu, %p, %u | %u | %llu, %u, %u | "
22074 + "%llu, b%d, 0x%x, g%u}\n",
22075 + rdu->sz, rdu->ent.e, rdu->verify[AufsCtlRduV_SZ],
22077 + rdu->rent, rdu->shwh, rdu->full,
22078 + rdu->cookie.h_pos, rdu->cookie.bindex, rdu->cookie.flags,
22079 + rdu->cookie.generation);
22081 + if (rdu->verify[AufsCtlRduV_SZ] == sizeof(*rdu))
22085 + rdu->verify[AufsCtlRduV_SZ], (unsigned int)sizeof(*rdu));
22089 +long au_rdu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
22092 + struct aufs_rdu rdu;
22093 + void __user *p = (void __user *)arg;
22095 + err = copy_from_user(&rdu, p, sizeof(rdu));
22096 + if (unlikely(err)) {
22101 + err = au_rdu_verify(&rdu);
22102 + if (unlikely(err))
22106 + case AUFS_CTL_RDU:
22107 + err = au_rdu(file, &rdu);
22108 + if (unlikely(err))
22111 + e = copy_to_user(p, &rdu, sizeof(rdu));
22112 + if (unlikely(e)) {
22117 + case AUFS_CTL_RDU_INO:
22118 + err = au_rdu_ino(file, &rdu);
22122 + /* err = -ENOTTY; */
22131 +#ifdef CONFIG_COMPAT
22132 +long au_rdu_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
22135 + struct aufs_rdu rdu;
22136 + void __user *p = compat_ptr(arg);
22138 + /* todo: get_user()? */
22139 + err = copy_from_user(&rdu, p, sizeof(rdu));
22140 + if (unlikely(err)) {
22145 + rdu.ent.e = compat_ptr(rdu.ent.ul);
22146 + err = au_rdu_verify(&rdu);
22147 + if (unlikely(err))
22151 + case AUFS_CTL_RDU:
22152 + err = au_rdu(file, &rdu);
22153 + if (unlikely(err))
22156 + rdu.ent.ul = ptr_to_compat(rdu.ent.e);
22157 + rdu.tail.ul = ptr_to_compat(rdu.tail.e);
22158 + e = copy_to_user(p, &rdu, sizeof(rdu));
22159 + if (unlikely(e)) {
22164 + case AUFS_CTL_RDU_INO:
22165 + err = au_rdu_ino(file, &rdu);
22169 + /* err = -ENOTTY; */
22178 diff -urN /usr/share/empty/fs/aufs/rwsem.h linux/fs/aufs/rwsem.h
22179 --- /usr/share/empty/fs/aufs/rwsem.h 1970-01-01 01:00:00.000000000 +0100
22180 +++ linux/fs/aufs/rwsem.h 2011-08-24 13:30:24.734646739 +0200
22183 + * Copyright (C) 2005-2011 Junjiro R. Okajima
22185 + * This program, aufs is free software; you can redistribute it and/or modify
22186 + * it under the terms of the GNU General Public License as published by
22187 + * the Free Software Foundation; either version 2 of the License, or
22188 + * (at your option) any later version.
22190 + * This program is distributed in the hope that it will be useful,
22191 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22192 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22193 + * GNU General Public License for more details.
22195 + * You should have received a copy of the GNU General Public License
22196 + * along with this program; if not, write to the Free Software
22197 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22201 + * simple read-write semaphore wrappers
22204 +#ifndef __AUFS_RWSEM_H__
22205 +#define __AUFS_RWSEM_H__
22209 +#include <linux/rwsem.h>
22210 +#include "debug.h"
22213 + struct rw_semaphore rwsem;
22214 +#ifdef CONFIG_AUFS_DEBUG
22215 + /* just for debugging, not almighty counter */
22216 + atomic_t rcnt, wcnt;
22220 +#ifdef CONFIG_AUFS_DEBUG
22221 +#define AuDbgCntInit(rw) do { \
22222 + atomic_set(&(rw)->rcnt, 0); \
22223 + atomic_set(&(rw)->wcnt, 0); \
22224 + smp_mb(); /* atomic set */ \
22227 +#define AuDbgRcntInc(rw) atomic_inc(&(rw)->rcnt)
22228 +#define AuDbgRcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->rcnt) < 0)
22229 +#define AuDbgWcntInc(rw) atomic_inc(&(rw)->wcnt)
22230 +#define AuDbgWcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->wcnt) < 0)
22232 +#define AuDbgCntInit(rw) do {} while (0)
22233 +#define AuDbgRcntInc(rw) do {} while (0)
22234 +#define AuDbgRcntDec(rw) do {} while (0)
22235 +#define AuDbgWcntInc(rw) do {} while (0)
22236 +#define AuDbgWcntDec(rw) do {} while (0)
22237 +#endif /* CONFIG_AUFS_DEBUG */
22239 +/* to debug easier, do not make them inlined functions */
22240 +#define AuRwMustNoWaiters(rw) AuDebugOn(!list_empty(&(rw)->rwsem.wait_list))
22241 +/* rwsem_is_locked() is unusable */
22242 +#define AuRwMustReadLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0)
22243 +#define AuRwMustWriteLock(rw) AuDebugOn(atomic_read(&(rw)->wcnt) <= 0)
22244 +#define AuRwMustAnyLock(rw) AuDebugOn(atomic_read(&(rw)->rcnt) <= 0 \
22245 + && atomic_read(&(rw)->wcnt) <= 0)
22246 +#define AuRwDestroy(rw) AuDebugOn(atomic_read(&(rw)->rcnt) \
22247 + || atomic_read(&(rw)->wcnt))
22249 +#define au_rw_class(rw, key) lockdep_set_class(&(rw)->rwsem, key)
22251 +static inline void au_rw_init(struct au_rwsem *rw)
22253 + AuDbgCntInit(rw);
22254 + init_rwsem(&rw->rwsem);
22257 +static inline void au_rw_init_wlock(struct au_rwsem *rw)
22260 + down_write(&rw->rwsem);
22261 + AuDbgWcntInc(rw);
22264 +static inline void au_rw_init_wlock_nested(struct au_rwsem *rw,
22265 + unsigned int lsc)
22268 + down_write_nested(&rw->rwsem, lsc);
22269 + AuDbgWcntInc(rw);
22272 +static inline void au_rw_read_lock(struct au_rwsem *rw)
22274 + down_read(&rw->rwsem);
22275 + AuDbgRcntInc(rw);
22278 +static inline void au_rw_read_lock_nested(struct au_rwsem *rw, unsigned int lsc)
22280 + down_read_nested(&rw->rwsem, lsc);
22281 + AuDbgRcntInc(rw);
22284 +static inline void au_rw_read_unlock(struct au_rwsem *rw)
22286 + AuRwMustReadLock(rw);
22287 + AuDbgRcntDec(rw);
22288 + up_read(&rw->rwsem);
22291 +static inline void au_rw_dgrade_lock(struct au_rwsem *rw)
22293 + AuRwMustWriteLock(rw);
22294 + AuDbgRcntInc(rw);
22295 + AuDbgWcntDec(rw);
22296 + downgrade_write(&rw->rwsem);
22299 +static inline void au_rw_write_lock(struct au_rwsem *rw)
22301 + down_write(&rw->rwsem);
22302 + AuDbgWcntInc(rw);
22305 +static inline void au_rw_write_lock_nested(struct au_rwsem *rw,
22306 + unsigned int lsc)
22308 + down_write_nested(&rw->rwsem, lsc);
22309 + AuDbgWcntInc(rw);
22312 +static inline void au_rw_write_unlock(struct au_rwsem *rw)
22314 + AuRwMustWriteLock(rw);
22315 + AuDbgWcntDec(rw);
22316 + up_write(&rw->rwsem);
22319 +/* why is not _nested version defined */
22320 +static inline int au_rw_read_trylock(struct au_rwsem *rw)
22322 + int ret = down_read_trylock(&rw->rwsem);
22324 + AuDbgRcntInc(rw);
22328 +static inline int au_rw_write_trylock(struct au_rwsem *rw)
22330 + int ret = down_write_trylock(&rw->rwsem);
22332 + AuDbgWcntInc(rw);
22336 +#undef AuDbgCntInit
22337 +#undef AuDbgRcntInc
22338 +#undef AuDbgRcntDec
22339 +#undef AuDbgWcntInc
22340 +#undef AuDbgWcntDec
22342 +#define AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
22343 +static inline void prefix##_read_lock(param) \
22344 +{ au_rw_read_lock(rwsem); } \
22345 +static inline void prefix##_write_lock(param) \
22346 +{ au_rw_write_lock(rwsem); } \
22347 +static inline int prefix##_read_trylock(param) \
22348 +{ return au_rw_read_trylock(rwsem); } \
22349 +static inline int prefix##_write_trylock(param) \
22350 +{ return au_rw_write_trylock(rwsem); }
22351 +/* why is not _nested version defined */
22352 +/* static inline void prefix##_read_trylock_nested(param, lsc)
22353 +{ au_rw_read_trylock_nested(rwsem, lsc)); }
22354 +static inline void prefix##_write_trylock_nestd(param, lsc)
22355 +{ au_rw_write_trylock_nested(rwsem, lsc); } */
22357 +#define AuSimpleUnlockRwsemFuncs(prefix, param, rwsem) \
22358 +static inline void prefix##_read_unlock(param) \
22359 +{ au_rw_read_unlock(rwsem); } \
22360 +static inline void prefix##_write_unlock(param) \
22361 +{ au_rw_write_unlock(rwsem); } \
22362 +static inline void prefix##_downgrade_lock(param) \
22363 +{ au_rw_dgrade_lock(rwsem); }
22365 +#define AuSimpleRwsemFuncs(prefix, param, rwsem) \
22366 + AuSimpleLockRwsemFuncs(prefix, param, rwsem) \
22367 + AuSimpleUnlockRwsemFuncs(prefix, param, rwsem)
22369 +#endif /* __KERNEL__ */
22370 +#endif /* __AUFS_RWSEM_H__ */
22371 diff -urN /usr/share/empty/fs/aufs/sbinfo.c linux/fs/aufs/sbinfo.c
22372 --- /usr/share/empty/fs/aufs/sbinfo.c 1970-01-01 01:00:00.000000000 +0100
22373 +++ linux/fs/aufs/sbinfo.c 2011-08-24 13:30:24.734646739 +0200
22376 + * Copyright (C) 2005-2011 Junjiro R. Okajima
22378 + * This program, aufs is free software; you can redistribute it and/or modify
22379 + * it under the terms of the GNU General Public License as published by
22380 + * the Free Software Foundation; either version 2 of the License, or
22381 + * (at your option) any later version.
22383 + * This program is distributed in the hope that it will be useful,
22384 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22385 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22386 + * GNU General Public License for more details.
22388 + * You should have received a copy of the GNU General Public License
22389 + * along with this program; if not, write to the Free Software
22390 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22394 + * superblock private data
22397 +#include <linux/jiffies.h>
22401 + * they are necessary regardless sysfs is disabled.
22403 +void au_si_free(struct kobject *kobj)
22405 + struct au_sbinfo *sbinfo;
22406 + char *locked __maybe_unused; /* debug only */
22408 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
22409 + AuDebugOn(!list_empty(&sbinfo->si_plink.head));
22410 + AuDebugOn(atomic_read(&sbinfo->si_nowait.nw_len));
22412 + au_rw_write_lock(&sbinfo->si_rwsem);
22413 + au_br_free(sbinfo);
22414 + au_rw_write_unlock(&sbinfo->si_rwsem);
22416 + AuDebugOn(radix_tree_gang_lookup
22417 + (&sbinfo->au_si_pid.tree, (void **)&locked,
22418 + /*first_index*/PID_MAX_DEFAULT - 1,
22419 + /*max_items*/sizeof(locked)/sizeof(*locked)));
22421 + kfree(sbinfo->si_branch);
22422 + kfree(sbinfo->au_si_pid.bitmap);
22423 + mutex_destroy(&sbinfo->si_xib_mtx);
22424 + AuRwDestroy(&sbinfo->si_rwsem);
22429 +int au_si_alloc(struct super_block *sb)
22432 + struct au_sbinfo *sbinfo;
22433 + static struct lock_class_key aufs_si;
22436 + sbinfo = kzalloc(sizeof(*sbinfo), GFP_NOFS);
22437 + if (unlikely(!sbinfo))
22440 + BUILD_BUG_ON(sizeof(unsigned long) !=
22441 + sizeof(*sbinfo->au_si_pid.bitmap));
22442 + sbinfo->au_si_pid.bitmap = kcalloc(BITS_TO_LONGS(PID_MAX_DEFAULT),
22443 + sizeof(*sbinfo->au_si_pid.bitmap),
22445 + if (unlikely(!sbinfo->au_si_pid.bitmap))
22448 + /* will be reallocated separately */
22449 + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_NOFS);
22450 + if (unlikely(!sbinfo->si_branch))
22453 + err = sysaufs_si_init(sbinfo);
22454 + if (unlikely(err))
22457 + au_nwt_init(&sbinfo->si_nowait);
22458 + au_rw_init_wlock(&sbinfo->si_rwsem);
22459 + au_rw_class(&sbinfo->si_rwsem, &aufs_si);
22460 + spin_lock_init(&sbinfo->au_si_pid.tree_lock);
22461 + INIT_RADIX_TREE(&sbinfo->au_si_pid.tree, GFP_ATOMIC | __GFP_NOFAIL);
22463 + atomic_long_set(&sbinfo->si_ninodes, 0);
22464 + atomic_long_set(&sbinfo->si_nfiles, 0);
22466 + sbinfo->si_bend = -1;
22468 + sbinfo->si_wbr_copyup = AuWbrCopyup_Def;
22469 + sbinfo->si_wbr_create = AuWbrCreate_Def;
22470 + sbinfo->si_wbr_copyup_ops = au_wbr_copyup_ops + sbinfo->si_wbr_copyup;
22471 + sbinfo->si_wbr_create_ops = au_wbr_create_ops + sbinfo->si_wbr_create;
22473 + sbinfo->si_mntflags = au_opts_plink(AuOpt_Def);
22475 + mutex_init(&sbinfo->si_xib_mtx);
22476 + sbinfo->si_xino_brid = -1;
22477 + /* leave si_xib_last_pindex and si_xib_next_bit */
22479 + sbinfo->si_rdcache = msecs_to_jiffies(AUFS_RDCACHE_DEF * MSEC_PER_SEC);
22480 + sbinfo->si_rdblk = AUFS_RDBLK_DEF;
22481 + sbinfo->si_rdhash = AUFS_RDHASH_DEF;
22482 + sbinfo->si_dirwh = AUFS_DIRWH_DEF;
22484 + au_spl_init(&sbinfo->si_plink);
22485 + init_waitqueue_head(&sbinfo->si_plink_wq);
22486 + spin_lock_init(&sbinfo->si_plink_maint_lock);
22488 + /* leave other members for sysaufs and si_mnt. */
22489 + sbinfo->si_sb = sb;
22490 + sb->s_fs_info = sbinfo;
22492 + au_debug_sbinfo_init(sbinfo);
22493 + return 0; /* success */
22496 + kfree(sbinfo->si_branch);
22498 + kfree(sbinfo->au_si_pid.bitmap);
22505 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr)
22508 + struct au_branch **brp;
22510 + AuRwMustWriteLock(&sbinfo->si_rwsem);
22513 + sz = sizeof(*brp) * (sbinfo->si_bend + 1);
22514 + if (unlikely(!sz))
22515 + sz = sizeof(*brp);
22516 + brp = au_kzrealloc(sbinfo->si_branch, sz, sizeof(*brp) * nbr, GFP_NOFS);
22518 + sbinfo->si_branch = brp;
22525 +/* ---------------------------------------------------------------------- */
22527 +unsigned int au_sigen_inc(struct super_block *sb)
22529 + unsigned int gen;
22531 + SiMustWriteLock(sb);
22533 + gen = ++au_sbi(sb)->si_generation;
22534 + au_update_digen(sb->s_root);
22535 + au_update_iigen(sb->s_root->d_inode);
22536 + sb->s_root->d_inode->i_version++;
22540 +aufs_bindex_t au_new_br_id(struct super_block *sb)
22542 + aufs_bindex_t br_id;
22544 + struct au_sbinfo *sbinfo;
22546 + SiMustWriteLock(sb);
22548 + sbinfo = au_sbi(sb);
22549 + for (i = 0; i <= AUFS_BRANCH_MAX; i++) {
22550 + br_id = ++sbinfo->si_last_br_id;
22551 + AuDebugOn(br_id < 0);
22552 + if (br_id && au_br_index(sb, br_id) < 0)
22559 +/* ---------------------------------------------------------------------- */
22561 +/* it is ok that new 'nwt' tasks are appended while we are sleeping */
22562 +int si_read_lock(struct super_block *sb, int flags)
22567 + if (au_ftest_lock(flags, FLUSH))
22568 + au_nwt_flush(&au_sbi(sb)->si_nowait);
22570 + si_noflush_read_lock(sb);
22571 + err = au_plink_maint(sb, flags);
22572 + if (unlikely(err))
22573 + si_read_unlock(sb);
22578 +int si_write_lock(struct super_block *sb, int flags)
22582 + if (au_ftest_lock(flags, FLUSH))
22583 + au_nwt_flush(&au_sbi(sb)->si_nowait);
22585 + si_noflush_write_lock(sb);
22586 + err = au_plink_maint(sb, flags);
22587 + if (unlikely(err))
22588 + si_write_unlock(sb);
22593 +/* dentry and super_block lock. call at entry point */
22594 +int aufs_read_lock(struct dentry *dentry, int flags)
22597 + struct super_block *sb;
22599 + sb = dentry->d_sb;
22600 + err = si_read_lock(sb, flags);
22601 + if (unlikely(err))
22604 + if (au_ftest_lock(flags, DW))
22605 + di_write_lock_child(dentry);
22607 + di_read_lock_child(dentry, flags);
22609 + if (au_ftest_lock(flags, GEN)) {
22610 + err = au_digen_test(dentry, au_sigen(sb));
22611 + AuDebugOn(!err && au_dbrange_test(dentry));
22612 + if (unlikely(err))
22613 + aufs_read_unlock(dentry, flags);
22620 +void aufs_read_unlock(struct dentry *dentry, int flags)
22622 + if (au_ftest_lock(flags, DW))
22623 + di_write_unlock(dentry);
22625 + di_read_unlock(dentry, flags);
22626 + si_read_unlock(dentry->d_sb);
22629 +void aufs_write_lock(struct dentry *dentry)
22631 + si_write_lock(dentry->d_sb, AuLock_FLUSH | AuLock_NOPLMW);
22632 + di_write_lock_child(dentry);
22635 +void aufs_write_unlock(struct dentry *dentry)
22637 + di_write_unlock(dentry);
22638 + si_write_unlock(dentry->d_sb);
22641 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags)
22644 + unsigned int sigen;
22645 + struct super_block *sb;
22648 + err = si_read_lock(sb, flags);
22649 + if (unlikely(err))
22652 + di_write_lock2_child(d1, d2, au_ftest_lock(flags, DIR));
22654 + if (au_ftest_lock(flags, GEN)) {
22655 + sigen = au_sigen(sb);
22656 + err = au_digen_test(d1, sigen);
22657 + AuDebugOn(!err && au_dbrange_test(d1));
22659 + err = au_digen_test(d2, sigen);
22660 + AuDebugOn(!err && au_dbrange_test(d2));
22662 + if (unlikely(err))
22663 + aufs_read_and_write_unlock2(d1, d2);
22670 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2)
22672 + di_write_unlock2(d1, d2);
22673 + si_read_unlock(d1->d_sb);
22676 +/* ---------------------------------------------------------------------- */
22678 +int si_pid_test_slow(struct super_block *sb)
22683 + p = radix_tree_lookup(&au_sbi(sb)->au_si_pid.tree, current->pid);
22684 + rcu_read_unlock();
22686 + return (long)!!p;
22689 +void si_pid_set_slow(struct super_block *sb)
22692 + struct au_sbinfo *sbinfo;
22694 + AuDebugOn(si_pid_test_slow(sb));
22696 + sbinfo = au_sbi(sb);
22697 + err = radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
22699 + spin_lock(&sbinfo->au_si_pid.tree_lock);
22700 + err = radix_tree_insert(&sbinfo->au_si_pid.tree, current->pid,
22701 + /*any valid ptr*/sb);
22702 + spin_unlock(&sbinfo->au_si_pid.tree_lock);
22704 + radix_tree_preload_end();
22707 +void si_pid_clr_slow(struct super_block *sb)
22710 + struct au_sbinfo *sbinfo;
22712 + AuDebugOn(!si_pid_test_slow(sb));
22714 + sbinfo = au_sbi(sb);
22715 + spin_lock(&sbinfo->au_si_pid.tree_lock);
22716 + p = radix_tree_delete(&sbinfo->au_si_pid.tree, current->pid);
22717 + spin_unlock(&sbinfo->au_si_pid.tree_lock);
22719 diff -urN /usr/share/empty/fs/aufs/spl.h linux/fs/aufs/spl.h
22720 --- /usr/share/empty/fs/aufs/spl.h 1970-01-01 01:00:00.000000000 +0100
22721 +++ linux/fs/aufs/spl.h 2011-08-24 13:30:24.734646739 +0200
22724 + * Copyright (C) 2005-2011 Junjiro R. Okajima
22726 + * This program, aufs is free software; you can redistribute it and/or modify
22727 + * it under the terms of the GNU General Public License as published by
22728 + * the Free Software Foundation; either version 2 of the License, or
22729 + * (at your option) any later version.
22731 + * This program is distributed in the hope that it will be useful,
22732 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22733 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22734 + * GNU General Public License for more details.
22736 + * You should have received a copy of the GNU General Public License
22737 + * along with this program; if not, write to the Free Software
22738 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22742 + * simple list protected by a spinlock
22745 +#ifndef __AUFS_SPL_H__
22746 +#define __AUFS_SPL_H__
22750 +#include <linux/spinlock.h>
22751 +#include <linux/list.h>
22752 +#include <linux/rculist.h>
22754 +struct au_splhead {
22756 + struct list_head head;
22759 +static inline void au_spl_init(struct au_splhead *spl)
22761 + spin_lock_init(&spl->spin);
22762 + INIT_LIST_HEAD(&spl->head);
22765 +static inline void au_spl_add(struct list_head *list, struct au_splhead *spl)
22767 + spin_lock(&spl->spin);
22768 + list_add(list, &spl->head);
22769 + spin_unlock(&spl->spin);
22772 +static inline void au_spl_del(struct list_head *list, struct au_splhead *spl)
22774 + spin_lock(&spl->spin);
22776 + spin_unlock(&spl->spin);
22779 +static inline void au_spl_del_rcu(struct list_head *list,
22780 + struct au_splhead *spl)
22782 + spin_lock(&spl->spin);
22783 + list_del_rcu(list);
22784 + spin_unlock(&spl->spin);
22787 +#endif /* __KERNEL__ */
22788 +#endif /* __AUFS_SPL_H__ */
22789 diff -urN /usr/share/empty/fs/aufs/super.c linux/fs/aufs/super.c
22790 --- /usr/share/empty/fs/aufs/super.c 1970-01-01 01:00:00.000000000 +0100
22791 +++ linux/fs/aufs/super.c 2012-01-05 11:00:59.218835185 +0100
22794 + * Copyright (C) 2005-2011 Junjiro R. Okajima
22796 + * This program, aufs is free software; you can redistribute it and/or modify
22797 + * it under the terms of the GNU General Public License as published by
22798 + * the Free Software Foundation; either version 2 of the License, or
22799 + * (at your option) any later version.
22801 + * This program is distributed in the hope that it will be useful,
22802 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
22803 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22804 + * GNU General Public License for more details.
22806 + * You should have received a copy of the GNU General Public License
22807 + * along with this program; if not, write to the Free Software
22808 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22812 + * mount and super_block operations
22815 +#include <linux/buffer_head.h>
22816 +#include <linux/jiffies.h>
22817 +#include <linux/module.h>
22818 +#include <linux/seq_file.h>
22819 +#include <linux/statfs.h>
22820 +#include <linux/vmalloc.h>
22821 +#include <linux/writeback.h>
22825 + * super_operations
22827 +static struct inode *aufs_alloc_inode(struct super_block *sb __maybe_unused)
22829 + struct au_icntnr *c;
22831 + c = au_cache_alloc_icntnr();
22833 + au_icntnr_init(c);
22834 + c->vfs_inode.i_version = 1; /* sigen(sb); */
22835 + c->iinfo.ii_hinode = NULL;
22836 + return &c->vfs_inode;
22841 +static void aufs_destroy_inode_cb(struct rcu_head *head)
22843 + struct inode *inode = container_of(head, struct inode, i_rcu);
22845 + INIT_LIST_HEAD(&inode->i_dentry);
22846 + au_cache_free_icntnr(container_of(inode, struct au_icntnr, vfs_inode));
22849 +static void aufs_destroy_inode(struct inode *inode)
22851 + au_iinfo_fin(inode);
22852 + call_rcu(&inode->i_rcu, aufs_destroy_inode_cb);
22855 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino)
22857 + struct inode *inode;
22860 + inode = iget_locked(sb, ino);
22861 + if (unlikely(!inode)) {
22862 + inode = ERR_PTR(-ENOMEM);
22865 + if (!(inode->i_state & I_NEW))
22868 + err = au_xigen_new(inode);
22870 + err = au_iinfo_init(inode);
22872 + inode->i_version++;
22874 + iget_failed(inode);
22875 + inode = ERR_PTR(err);
22879 + /* never return NULL */
22880 + AuDebugOn(!inode);
22881 + AuTraceErrPtr(inode);
22885 +/* lock free root dinfo */
22886 +static int au_show_brs(struct seq_file *seq, struct super_block *sb)
22889 + aufs_bindex_t bindex, bend;
22890 + struct path path;
22891 + struct au_hdentry *hdp;
22892 + struct au_branch *br;
22896 + bend = au_sbend(sb);
22897 + hdp = au_di(sb->s_root)->di_hdentry;
22898 + for (bindex = 0; !err && bindex <= bend; bindex++) {
22899 + br = au_sbr(sb, bindex);
22900 + path.mnt = br->br_mnt;
22901 + path.dentry = hdp[bindex].hd_dentry;
22902 + err = au_seq_path(seq, &path);
22904 + perm = au_optstr_br_perm(br->br_perm);
22906 + err = seq_printf(seq, "=%s", perm);
22913 + if (!err && bindex != bend)
22914 + err = seq_putc(seq, ':');
22920 +static void au_show_wbr_create(struct seq_file *m, int v,
22921 + struct au_sbinfo *sbinfo)
22925 + AuRwMustAnyLock(&sbinfo->si_rwsem);
22927 + seq_printf(m, ",create=");
22928 + pat = au_optstr_wbr_create(v);
22930 + case AuWbrCreate_TDP:
22931 + case AuWbrCreate_RR:
22932 + case AuWbrCreate_MFS:
22933 + case AuWbrCreate_PMFS:
22934 + seq_printf(m, pat);
22936 + case AuWbrCreate_MFSV:
22937 + seq_printf(m, /*pat*/"mfs:%lu",
22938 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
22941 + case AuWbrCreate_PMFSV:
22942 + seq_printf(m, /*pat*/"pmfs:%lu",
22943 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
22946 + case AuWbrCreate_MFSRR:
22947 + seq_printf(m, /*pat*/"mfsrr:%llu",
22948 + sbinfo->si_wbr_mfs.mfsrr_watermark);
22950 + case AuWbrCreate_MFSRRV:
22951 + seq_printf(m, /*pat*/"mfsrr:%llu:%lu",
22952 + sbinfo->si_wbr_mfs.mfsrr_watermark,
22953 + jiffies_to_msecs(sbinfo->si_wbr_mfs.mfs_expire)
22959 +static int au_show_xino(struct seq_file *seq, struct vfsmount *mnt)
22961 +#ifdef CONFIG_SYSFS
22965 + const int len = sizeof(AUFS_XINO_FNAME) - 1;
22966 + aufs_bindex_t bindex, brid;
22967 + struct super_block *sb;
22968 + struct qstr *name;
22970 + struct dentry *d, *h_root;
22971 + struct au_hdentry *hdp;
22973 + AuRwMustAnyLock(&sbinfo->si_rwsem);
22976 + sb = mnt->mnt_sb;
22977 + f = au_sbi(sb)->si_xib;
22981 + /* stop printing the default xino path on the first writable branch */
22983 + brid = au_xino_brid(sb);
22985 + bindex = au_br_index(sb, brid);
22986 + hdp = au_di(sb->s_root)->di_hdentry;
22987 + h_root = hdp[0 + bindex].hd_dentry;
22990 + name = &d->d_name;
22991 + /* safe ->d_parent because the file is unlinked */
22992 + if (d->d_parent == h_root
22993 + && name->len == len
22994 + && !memcmp(name->name, AUFS_XINO_FNAME, len))
22997 + seq_puts(seq, ",xino=");
22998 + err = au_xino_path(seq, f);
23005 +/* seq_file will re-call me in case of too long string */
23006 +static int aufs_show_options(struct seq_file *m, struct vfsmount *mnt)
23009 + unsigned int mnt_flags, v;
23010 + struct super_block *sb;
23011 + struct au_sbinfo *sbinfo;
23013 +#define AuBool(name, str) do { \
23014 + v = au_opt_test(mnt_flags, name); \
23015 + if (v != au_opt_test(AuOpt_Def, name)) \
23016 + seq_printf(m, ",%s" #str, v ? "" : "no"); \
23019 +#define AuStr(name, str) do { \
23020 + v = mnt_flags & AuOptMask_##name; \
23021 + if (v != (AuOpt_Def & AuOptMask_##name)) \
23022 + seq_printf(m, "," #str "=%s", au_optstr_##str(v)); \
23025 +#define AuUInt(name, str, val) do { \
23026 + if (val != AUFS_##name##_DEF) \
23027 + seq_printf(m, "," #str "=%u", val); \
23030 + /* lock free root dinfo */
23031 + sb = mnt->mnt_sb;
23032 + si_noflush_read_lock(sb);
23033 + sbinfo = au_sbi(sb);
23034 + seq_printf(m, ",si=%lx", sysaufs_si_id(sbinfo));
23036 + mnt_flags = au_mntflags(sb);
23037 + if (au_opt_test(mnt_flags, XINO)) {
23038 + err = au_show_xino(m, mnt);
23039 + if (unlikely(err))
23042 + seq_puts(m, ",noxino");
23044 + AuBool(TRUNC_XINO, trunc_xino);
23045 + AuStr(UDBA, udba);
23046 + AuBool(SHWH, shwh);
23047 + AuBool(PLINK, plink);
23048 + AuBool(DIO, dio);
23049 + /* AuBool(DIRPERM1, dirperm1); */
23050 + /* AuBool(REFROF, refrof); */
23052 + v = sbinfo->si_wbr_create;
23053 + if (v != AuWbrCreate_Def)
23054 + au_show_wbr_create(m, v, sbinfo);
23056 + v = sbinfo->si_wbr_copyup;
23057 + if (v != AuWbrCopyup_Def)
23058 + seq_printf(m, ",cpup=%s", au_optstr_wbr_copyup(v));
23060 + v = au_opt_test(mnt_flags, ALWAYS_DIROPQ);
23061 + if (v != au_opt_test(AuOpt_Def, ALWAYS_DIROPQ))
23062 + seq_printf(m, ",diropq=%c", v ? 'a' : 'w');
23064 + AuUInt(DIRWH, dirwh, sbinfo->si_dirwh);
23066 + v = jiffies_to_msecs(sbinfo->si_rdcache) / MSEC_PER_SEC;
23067 + AuUInt(RDCACHE, rdcache, v);
23069 + AuUInt(RDBLK, rdblk, sbinfo->si_rdblk);
23070 + AuUInt(RDHASH, rdhash, sbinfo->si_rdhash);
23072 + AuBool(SUM, sum);
23073 + /* AuBool(SUM_W, wsum); */
23074 + AuBool(WARN_PERM, warn_perm);
23075 + AuBool(VERBOSE, verbose);
23078 + /* be sure to print "br:" last */
23079 + if (!sysaufs_brs) {
23080 + seq_puts(m, ",br:");
23081 + au_show_brs(m, sb);
23083 + si_read_unlock(sb);
23091 +/* ---------------------------------------------------------------------- */
23093 +/* sum mode which returns the summation for statfs(2) */
23095 +static u64 au_add_till_max(u64 a, u64 b)
23103 + return ULLONG_MAX;
23106 +static int au_statfs_sum(struct super_block *sb, struct kstatfs *buf)
23109 + u64 blocks, bfree, bavail, files, ffree;
23110 + aufs_bindex_t bend, bindex, i;
23111 + unsigned char shared;
23112 + struct path h_path;
23113 + struct super_block *h_sb;
23122 + bend = au_sbend(sb);
23123 + for (bindex = bend; bindex >= 0; bindex--) {
23124 + h_path.mnt = au_sbr_mnt(sb, bindex);
23125 + h_sb = h_path.mnt->mnt_sb;
23127 + for (i = bindex + 1; !shared && i <= bend; i++)
23128 + shared = (au_sbr_sb(sb, i) == h_sb);
23132 + /* sb->s_root for NFS is unreliable */
23133 + h_path.dentry = h_path.mnt->mnt_root;
23134 + err = vfs_statfs(&h_path, buf);
23135 + if (unlikely(err))
23138 + blocks = au_add_till_max(blocks, buf->f_blocks);
23139 + bfree = au_add_till_max(bfree, buf->f_bfree);
23140 + bavail = au_add_till_max(bavail, buf->f_bavail);
23141 + files = au_add_till_max(files, buf->f_files);
23142 + ffree = au_add_till_max(ffree, buf->f_ffree);
23145 + buf->f_blocks = blocks;
23146 + buf->f_bfree = bfree;
23147 + buf->f_bavail = bavail;
23148 + buf->f_files = files;
23149 + buf->f_ffree = ffree;
23155 +static int aufs_statfs(struct dentry *dentry, struct kstatfs *buf)
23158 + struct path h_path;
23159 + struct super_block *sb;
23161 + /* lock free root dinfo */
23162 + sb = dentry->d_sb;
23163 + si_noflush_read_lock(sb);
23164 + if (!au_opt_test(au_mntflags(sb), SUM)) {
23165 + /* sb->s_root for NFS is unreliable */
23166 + h_path.mnt = au_sbr_mnt(sb, 0);
23167 + h_path.dentry = h_path.mnt->mnt_root;
23168 + err = vfs_statfs(&h_path, buf);
23170 + err = au_statfs_sum(sb, buf);
23171 + si_read_unlock(sb);
23174 + buf->f_type = AUFS_SUPER_MAGIC;
23175 + buf->f_namelen = AUFS_MAX_NAMELEN;
23176 + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid));
23178 + /* buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; */
23183 +/* ---------------------------------------------------------------------- */
23185 +/* final actions when unmounting a file system */
23186 +static void aufs_put_super(struct super_block *sb)
23188 + struct au_sbinfo *sbinfo;
23190 + sbinfo = au_sbi(sb);
23194 + dbgaufs_si_fin(sbinfo);
23195 + kobject_put(&sbinfo->si_kobj);
23198 +/* ---------------------------------------------------------------------- */
23200 +void au_array_free(void *array)
23203 + if (!is_vmalloc_addr(array))
23210 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb, void *arg)
23213 + unsigned long long n;
23220 + if (*hint > ULLONG_MAX / sizeof(array)) {
23221 + array = ERR_PTR(-EMFILE);
23222 + pr_err("hint %llu\n", *hint);
23226 + array = kmalloc(sizeof(array) * *hint, GFP_NOFS);
23227 + if (unlikely(!array))
23228 + array = vmalloc(sizeof(array) * *hint);
23229 + if (unlikely(!array)) {
23230 + array = ERR_PTR(-ENOMEM);
23234 + n = cb(array, *hint, arg);
23235 + AuDebugOn(n > *hint);
23242 +static unsigned long long au_iarray_cb(void *a,
23243 + unsigned long long max __maybe_unused,
23246 + unsigned long long n;
23247 + struct inode **p, *inode;
23248 + struct list_head *head;
23253 + spin_lock(&inode_sb_list_lock);
23254 + list_for_each_entry(inode, head, i_sb_list) {
23255 + if (!is_bad_inode(inode)
23256 + && au_ii(inode)->ii_bstart >= 0) {
23257 + spin_lock(&inode->i_lock);
23258 + if (atomic_read(&inode->i_count)) {
23262 + AuDebugOn(n > max);
23264 + spin_unlock(&inode->i_lock);
23267 + spin_unlock(&inode_sb_list_lock);
23272 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max)
23274 + *max = atomic_long_read(&au_sbi(sb)->si_ninodes);
23275 + return au_array_alloc(max, au_iarray_cb, &sb->s_inodes);
23278 +void au_iarray_free(struct inode **a, unsigned long long max)
23280 + unsigned long long ull;
23282 + for (ull = 0; ull < max; ull++)
23284 + au_array_free(a);
23287 +/* ---------------------------------------------------------------------- */
23290 + * refresh dentry and inode at remount time.
23292 +/* todo: consolidate with simple_reval_dpath() and au_reval_for_attr() */
23293 +static int au_do_refresh(struct dentry *dentry, unsigned int dir_flags,
23294 + struct dentry *parent)
23298 + di_write_lock_child(dentry);
23299 + di_read_lock_parent(parent, AuLock_IR);
23300 + err = au_refresh_dentry(dentry, parent);
23301 + if (!err && dir_flags)
23302 + au_hn_reset(dentry->d_inode, dir_flags);
23303 + di_read_unlock(parent, AuLock_IR);
23304 + di_write_unlock(dentry);
23309 +static int au_do_refresh_d(struct dentry *dentry, unsigned int sigen,
23310 + struct au_sbinfo *sbinfo,
23311 + const unsigned int dir_flags)
23314 + struct dentry *parent;
23315 + struct inode *inode;
23318 + parent = dget_parent(dentry);
23319 + if (!au_digen_test(parent, sigen) && au_digen_test(dentry, sigen)) {
23320 + inode = dentry->d_inode;
23322 + if (!S_ISDIR(inode->i_mode))
23323 + err = au_do_refresh(dentry, /*dir_flags*/0,
23326 + err = au_do_refresh(dentry, dir_flags, parent);
23327 + if (unlikely(err))
23328 + au_fset_si(sbinfo, FAILED_REFRESH_DIR);
23331 + err = au_do_refresh(dentry, /*dir_flags*/0, parent);
23332 + AuDbgDentry(dentry);
23340 +static int au_refresh_d(struct super_block *sb)
23342 + int err, i, j, ndentry, e;
23343 + unsigned int sigen;
23344 + struct au_dcsub_pages dpages;
23345 + struct au_dpage *dpage;
23346 + struct dentry **dentries, *d;
23347 + struct au_sbinfo *sbinfo;
23348 + struct dentry *root = sb->s_root;
23349 + const unsigned int dir_flags = au_hi_flags(root->d_inode, /*isdir*/1);
23351 + err = au_dpages_init(&dpages, GFP_NOFS);
23352 + if (unlikely(err))
23354 + err = au_dcsub_pages(&dpages, root, NULL, NULL);
23355 + if (unlikely(err))
23358 + sigen = au_sigen(sb);
23359 + sbinfo = au_sbi(sb);
23360 + for (i = 0; i < dpages.ndpage; i++) {
23361 + dpage = dpages.dpages + i;
23362 + dentries = dpage->dentries;
23363 + ndentry = dpage->ndentry;
23364 + for (j = 0; j < ndentry; j++) {
23366 + e = au_do_refresh_d(d, sigen, sbinfo, dir_flags);
23367 + if (unlikely(e && !err))
23369 + /* go on even err */
23374 + au_dpages_free(&dpages);
23379 +static int au_refresh_i(struct super_block *sb)
23382 + unsigned int sigen;
23383 + unsigned long long max, ull;
23384 + struct inode *inode, **array;
23386 + array = au_iarray_alloc(sb, &max);
23387 + err = PTR_ERR(array);
23388 + if (IS_ERR(array))
23392 + sigen = au_sigen(sb);
23393 + for (ull = 0; ull < max; ull++) {
23394 + inode = array[ull];
23395 + if (au_iigen(inode) != sigen) {
23396 + ii_write_lock_child(inode);
23397 + e = au_refresh_hinode_self(inode);
23398 + ii_write_unlock(inode);
23399 + if (unlikely(e)) {
23400 + pr_err("error %d, i%lu\n", e, inode->i_ino);
23403 + /* go on even if err */
23408 + au_iarray_free(array, max);
23414 +static void au_remount_refresh(struct super_block *sb)
23417 + unsigned int udba;
23418 + aufs_bindex_t bindex, bend;
23419 + struct dentry *root;
23420 + struct inode *inode;
23421 + struct au_branch *br;
23423 + au_sigen_inc(sb);
23424 + au_fclr_si(au_sbi(sb), FAILED_REFRESH_DIR);
23426 + root = sb->s_root;
23427 + DiMustNoWaiters(root);
23428 + inode = root->d_inode;
23429 + IiMustNoWaiters(inode);
23431 + udba = au_opt_udba(sb);
23432 + bend = au_sbend(sb);
23433 + for (bindex = 0; bindex <= bend; bindex++) {
23434 + br = au_sbr(sb, bindex);
23435 + err = au_hnotify_reset_br(udba, br, br->br_perm);
23436 + if (unlikely(err))
23437 + AuIOErr("hnotify failed on br %d, %d, ignored\n",
23439 + /* go on even if err */
23441 + au_hn_reset(inode, au_hi_flags(inode, /*isdir*/1));
23443 + di_write_unlock(root);
23444 + err = au_refresh_d(sb);
23445 + e = au_refresh_i(sb);
23446 + if (unlikely(e && !err))
23448 + /* aufs_write_lock() calls ..._child() */
23449 + di_write_lock_child(root);
23451 + au_cpup_attr_all(inode, /*force*/1);
23453 + if (unlikely(err))
23454 + AuIOErr("refresh failed, ignored, %d\n", err);
23457 +/* stop extra interpretation of errno in mount(8), and strange error messages */
23458 +static int cvt_err(int err)
23472 +static int aufs_remount_fs(struct super_block *sb, int *flags, char *data)
23475 + unsigned int mntflags;
23476 + struct au_opts opts;
23477 + struct dentry *root;
23478 + struct inode *inode;
23479 + struct au_sbinfo *sbinfo;
23482 + root = sb->s_root;
23483 + if (!data || !*data) {
23484 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
23486 + di_write_lock_child(root);
23487 + err = au_opts_verify(sb, *flags, /*pending*/0);
23488 + aufs_write_unlock(root);
23494 + memset(&opts, 0, sizeof(opts));
23495 + opts.opt = (void *)__get_free_page(GFP_NOFS);
23496 + if (unlikely(!opts.opt))
23498 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
23499 + opts.flags = AuOpts_REMOUNT;
23500 + opts.sb_flags = *flags;
23502 + /* parse it before aufs lock */
23503 + err = au_opts_parse(sb, data, &opts);
23504 + if (unlikely(err))
23507 + sbinfo = au_sbi(sb);
23508 + inode = root->d_inode;
23509 + mutex_lock(&inode->i_mutex);
23510 + err = si_write_lock(sb, AuLock_FLUSH | AuLock_NOPLM);
23511 + if (unlikely(err))
23513 + di_write_lock_child(root);
23515 + /* au_opts_remount() may return an error */
23516 + err = au_opts_remount(sb, &opts);
23517 + au_opts_free(&opts);
23519 + if (au_ftest_opts(opts.flags, REFRESH))
23520 + au_remount_refresh(sb);
23522 + if (au_ftest_opts(opts.flags, REFRESH_DYAOP)) {
23523 + mntflags = au_mntflags(sb);
23524 + do_dx = !!au_opt_test(mntflags, DIO);
23525 + au_dy_arefresh(do_dx);
23528 + aufs_write_unlock(root);
23531 + mutex_unlock(&inode->i_mutex);
23533 + free_page((unsigned long)opts.opt);
23535 + err = cvt_err(err);
23540 +static const struct super_operations aufs_sop = {
23541 + .alloc_inode = aufs_alloc_inode,
23542 + .destroy_inode = aufs_destroy_inode,
23543 + /* always deleting, no clearing */
23544 + .drop_inode = generic_delete_inode,
23545 + .show_options = aufs_show_options,
23546 + .statfs = aufs_statfs,
23547 + .put_super = aufs_put_super,
23548 + .remount_fs = aufs_remount_fs
23551 +/* ---------------------------------------------------------------------- */
23553 +static int alloc_root(struct super_block *sb)
23556 + struct inode *inode;
23557 + struct dentry *root;
23560 + inode = au_iget_locked(sb, AUFS_ROOT_INO);
23561 + err = PTR_ERR(inode);
23562 + if (IS_ERR(inode))
23565 + inode->i_op = &aufs_dir_iop;
23566 + inode->i_fop = &aufs_dir_fop;
23567 + inode->i_mode = S_IFDIR;
23568 + set_nlink(inode, 2);
23569 + unlock_new_inode(inode);
23571 + root = d_alloc_root(inode);
23572 + if (unlikely(!root))
23574 + err = PTR_ERR(root);
23575 + if (IS_ERR(root))
23578 + err = au_di_init(root);
23580 + sb->s_root = root;
23581 + return 0; /* success */
23584 + goto out; /* do not iput */
23587 + iget_failed(inode);
23593 +static int aufs_fill_super(struct super_block *sb, void *raw_data,
23594 + int silent __maybe_unused)
23597 + struct au_opts opts;
23598 + struct dentry *root;
23599 + struct inode *inode;
23600 + char *arg = raw_data;
23602 + if (unlikely(!arg || !*arg)) {
23604 + pr_err("no arg\n");
23609 + memset(&opts, 0, sizeof(opts));
23610 + opts.opt = (void *)__get_free_page(GFP_NOFS);
23611 + if (unlikely(!opts.opt))
23613 + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt);
23614 + opts.sb_flags = sb->s_flags;
23616 + err = au_si_alloc(sb);
23617 + if (unlikely(err))
23620 + /* all timestamps always follow the ones on the branch */
23621 + sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
23622 + sb->s_op = &aufs_sop;
23623 + sb->s_d_op = &aufs_dop;
23624 + sb->s_magic = AUFS_SUPER_MAGIC;
23625 + sb->s_maxbytes = 0;
23626 + au_export_init(sb);
23628 + err = alloc_root(sb);
23629 + if (unlikely(err)) {
23630 + si_write_unlock(sb);
23633 + root = sb->s_root;
23634 + inode = root->d_inode;
23637 + * actually we can parse options regardless aufs lock here.
23638 + * but at remount time, parsing must be done before aufs lock.
23639 + * so we follow the same rule.
23641 + ii_write_lock_parent(inode);
23642 + aufs_write_unlock(root);
23643 + err = au_opts_parse(sb, arg, &opts);
23644 + if (unlikely(err))
23647 + /* lock vfs_inode first, then aufs. */
23648 + mutex_lock(&inode->i_mutex);
23649 + aufs_write_lock(root);
23650 + err = au_opts_mount(sb, &opts);
23651 + au_opts_free(&opts);
23652 + aufs_write_unlock(root);
23653 + mutex_unlock(&inode->i_mutex);
23655 + goto out_opts; /* success */
23659 + sb->s_root = NULL;
23661 + dbgaufs_si_fin(au_sbi(sb));
23662 + kobject_put(&au_sbi(sb)->si_kobj);
23663 + sb->s_fs_info = NULL;
23665 + free_page((unsigned long)opts.opt);
23668 + err = cvt_err(err);
23673 +/* ---------------------------------------------------------------------- */
23675 +static struct dentry *aufs_mount(struct file_system_type *fs_type, int flags,
23676 + const char *dev_name __maybe_unused,
23679 + struct dentry *root;
23680 + struct super_block *sb;
23682 + /* all timestamps always follow the ones on the branch */
23683 + /* mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; */
23684 + root = mount_nodev(fs_type, flags, raw_data, aufs_fill_super);
23685 + if (IS_ERR(root))
23689 + si_write_lock(sb, !AuLock_FLUSH);
23690 + sysaufs_brs_add(sb, 0);
23691 + si_write_unlock(sb);
23692 + au_sbilist_add(sb);
23698 +static void aufs_kill_sb(struct super_block *sb)
23700 + struct au_sbinfo *sbinfo;
23702 + sbinfo = au_sbi(sb);
23704 + au_sbilist_del(sb);
23705 + aufs_write_lock(sb->s_root);
23706 + if (sbinfo->si_wbr_create_ops->fin)
23707 + sbinfo->si_wbr_create_ops->fin(sb);
23708 + if (au_opt_test(sbinfo->si_mntflags, UDBA_HNOTIFY)) {
23709 + au_opt_set_udba(sbinfo->si_mntflags, UDBA_NONE);
23710 + au_remount_refresh(sb);
23712 + if (au_opt_test(sbinfo->si_mntflags, PLINK))
23713 + au_plink_put(sb, /*verbose*/1);
23715 + sbinfo->si_sb = NULL;
23716 + aufs_write_unlock(sb->s_root);
23717 + au_nwt_flush(&sbinfo->si_nowait);
23719 + generic_shutdown_super(sb);
23722 +struct file_system_type aufs_fs_type = {
23723 + .name = AUFS_FSTYPE,
23725 + FS_RENAME_DOES_D_MOVE /* a race between rename and others */
23726 + | FS_REVAL_DOT, /* for NFS branch and udba */
23727 + .mount = aufs_mount,
23728 + .kill_sb = aufs_kill_sb,
23729 + /* no need to __module_get() and module_put(). */
23730 + .owner = THIS_MODULE,
23732 diff -urN /usr/share/empty/fs/aufs/super.h linux/fs/aufs/super.h
23733 --- /usr/share/empty/fs/aufs/super.h 1970-01-01 01:00:00.000000000 +0100
23734 +++ linux/fs/aufs/super.h 2011-08-24 13:30:24.734646739 +0200
23737 + * Copyright (C) 2005-2011 Junjiro R. Okajima
23739 + * This program, aufs is free software; you can redistribute it and/or modify
23740 + * it under the terms of the GNU General Public License as published by
23741 + * the Free Software Foundation; either version 2 of the License, or
23742 + * (at your option) any later version.
23744 + * This program is distributed in the hope that it will be useful,
23745 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
23746 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23747 + * GNU General Public License for more details.
23749 + * You should have received a copy of the GNU General Public License
23750 + * along with this program; if not, write to the Free Software
23751 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
23755 + * super_block operations
23758 +#ifndef __AUFS_SUPER_H__
23759 +#define __AUFS_SUPER_H__
23763 +#include <linux/fs.h>
23764 +#include <linux/aufs_type.h>
23765 +#include "rwsem.h"
23769 +typedef ssize_t (*au_readf_t)(struct file *, char __user *, size_t, loff_t *);
23770 +typedef ssize_t (*au_writef_t)(struct file *, const char __user *, size_t,
23773 +/* policies to select one among multiple writable branches */
23774 +struct au_wbr_copyup_operations {
23775 + int (*copyup)(struct dentry *dentry);
23778 +struct au_wbr_create_operations {
23779 + int (*create)(struct dentry *dentry, int isdir);
23780 + int (*init)(struct super_block *sb);
23781 + int (*fin)(struct super_block *sb);
23784 +struct au_wbr_mfs {
23785 + struct mutex mfs_lock; /* protect this structure */
23786 + unsigned long mfs_jiffy;
23787 + unsigned long mfs_expire;
23788 + aufs_bindex_t mfs_bindex;
23790 + unsigned long long mfsrr_bytes;
23791 + unsigned long long mfsrr_watermark;
23795 +struct au_sbinfo {
23796 + /* nowait tasks in the system-wide workqueue */
23797 + struct au_nowait_tasks si_nowait;
23800 + * tried sb->s_umount, but failed due to the dependecy between i_mutex.
23801 + * rwsem for au_sbinfo is necessary.
23803 + struct au_rwsem si_rwsem;
23805 + /* prevent recursive locking in deleting inode */
23807 + unsigned long *bitmap;
23808 + spinlock_t tree_lock;
23809 + struct radix_tree_root tree;
23813 + * dirty approach to protect sb->sb_inodes and ->s_files from remount.
23815 + atomic_long_t si_ninodes, si_nfiles;
23817 + /* branch management */
23818 + unsigned int si_generation;
23820 + /* see above flags */
23821 + unsigned char au_si_status;
23823 + aufs_bindex_t si_bend;
23825 + /* dirty trick to keep br_id plus */
23826 + unsigned int si_last_br_id :
23827 + sizeof(aufs_bindex_t) * BITS_PER_BYTE - 1;
23828 + struct au_branch **si_branch;
23830 + /* policy to select a writable branch */
23831 + unsigned char si_wbr_copyup;
23832 + unsigned char si_wbr_create;
23833 + struct au_wbr_copyup_operations *si_wbr_copyup_ops;
23834 + struct au_wbr_create_operations *si_wbr_create_ops;
23836 + /* round robin */
23837 + atomic_t si_wbr_rr_next;
23839 + /* most free space */
23840 + struct au_wbr_mfs si_wbr_mfs;
23842 + /* mount flags */
23843 + /* include/asm-ia64/siginfo.h defines a macro named si_flags */
23844 + unsigned int si_mntflags;
23846 + /* external inode number (bitmap and translation table) */
23847 + au_readf_t si_xread;
23848 + au_writef_t si_xwrite;
23849 + struct file *si_xib;
23850 + struct mutex si_xib_mtx; /* protect xib members */
23851 + unsigned long *si_xib_buf;
23852 + unsigned long si_xib_last_pindex;
23853 + int si_xib_next_bit;
23854 + aufs_bindex_t si_xino_brid;
23855 + /* reserved for future use */
23856 + /* unsigned long long si_xib_limit; */ /* Max xib file size */
23858 +#ifdef CONFIG_AUFS_EXPORT
23859 + /* i_generation */
23860 + struct file *si_xigen;
23861 + atomic_t si_xigen_next;
23864 + /* vdir parameters */
23865 + unsigned long si_rdcache; /* max cache time in jiffies */
23866 + unsigned int si_rdblk; /* deblk size */
23867 + unsigned int si_rdhash; /* hash size */
23870 + * If the number of whiteouts are larger than si_dirwh, leave all of
23871 + * them after au_whtmp_ren to reduce the cost of rmdir(2).
23872 + * future fsck.aufs or kernel thread will remove them later.
23873 + * Otherwise, remove all whiteouts and the dir in rmdir(2).
23875 + unsigned int si_dirwh;
23878 + * rename(2) a directory with all children.
23880 + /* reserved for future use */
23881 + /* int si_rendir; */
23883 + /* pseudo_link list */
23884 + struct au_splhead si_plink;
23885 + wait_queue_head_t si_plink_wq;
23886 + spinlock_t si_plink_maint_lock;
23887 + pid_t si_plink_maint_pid;
23890 + * sysfs and lifetime management.
23891 + * this is not a small structure and it may be a waste of memory in case
23892 + * of sysfs is disabled, particulary when many aufs-es are mounted.
23893 + * but using sysfs is majority.
23895 + struct kobject si_kobj;
23896 +#ifdef CONFIG_DEBUG_FS
23897 + struct dentry *si_dbgaufs, *si_dbgaufs_xib;
23898 +#ifdef CONFIG_AUFS_EXPORT
23899 + struct dentry *si_dbgaufs_xigen;
23903 +#ifdef CONFIG_AUFS_SBILIST
23904 + struct list_head si_list;
23907 + /* dirty, necessary for unmounting, sysfs and sysrq */
23908 + struct super_block *si_sb;
23911 +/* sbinfo status flags */
23913 + * set true when refresh_dirs() failed at remount time.
23914 + * then try refreshing dirs at access time again.
23915 + * if it is false, refreshing dirs at access time is unnecesary
23917 +#define AuSi_FAILED_REFRESH_DIR 1
23918 +static inline unsigned char au_do_ftest_si(struct au_sbinfo *sbi,
23919 + unsigned int flag)
23921 + AuRwMustAnyLock(&sbi->si_rwsem);
23922 + return sbi->au_si_status & flag;
23924 +#define au_ftest_si(sbinfo, name) au_do_ftest_si(sbinfo, AuSi_##name)
23925 +#define au_fset_si(sbinfo, name) do { \
23926 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
23927 + (sbinfo)->au_si_status |= AuSi_##name; \
23929 +#define au_fclr_si(sbinfo, name) do { \
23930 + AuRwMustWriteLock(&(sbinfo)->si_rwsem); \
23931 + (sbinfo)->au_si_status &= ~AuSi_##name; \
23934 +/* ---------------------------------------------------------------------- */
23936 +/* policy to select one among writable branches */
23937 +#define AuWbrCopyup(sbinfo, ...) \
23938 + ((sbinfo)->si_wbr_copyup_ops->copyup(__VA_ARGS__))
23939 +#define AuWbrCreate(sbinfo, ...) \
23940 + ((sbinfo)->si_wbr_create_ops->create(__VA_ARGS__))
23942 +/* flags for si_read_lock()/aufs_read_lock()/di_read_lock() */
23943 +#define AuLock_DW 1 /* write-lock dentry */
23944 +#define AuLock_IR (1 << 1) /* read-lock inode */
23945 +#define AuLock_IW (1 << 2) /* write-lock inode */
23946 +#define AuLock_FLUSH (1 << 3) /* wait for 'nowait' tasks */
23947 +#define AuLock_DIR (1 << 4) /* target is a dir */
23948 +#define AuLock_NOPLM (1 << 5) /* return err in plm mode */
23949 +#define AuLock_NOPLMW (1 << 6) /* wait for plm mode ends */
23950 +#define AuLock_GEN (1 << 7) /* test digen/iigen */
23951 +#define au_ftest_lock(flags, name) ((flags) & AuLock_##name)
23952 +#define au_fset_lock(flags, name) \
23953 + do { (flags) |= AuLock_##name; } while (0)
23954 +#define au_fclr_lock(flags, name) \
23955 + do { (flags) &= ~AuLock_##name; } while (0)
23957 +/* ---------------------------------------------------------------------- */
23960 +extern struct file_system_type aufs_fs_type;
23961 +struct inode *au_iget_locked(struct super_block *sb, ino_t ino);
23962 +typedef unsigned long long (*au_arraycb_t)(void *array, unsigned long long max,
23964 +void au_array_free(void *array);
23965 +void *au_array_alloc(unsigned long long *hint, au_arraycb_t cb, void *arg);
23966 +struct inode **au_iarray_alloc(struct super_block *sb, unsigned long long *max);
23967 +void au_iarray_free(struct inode **a, unsigned long long max);
23970 +void au_si_free(struct kobject *kobj);
23971 +int au_si_alloc(struct super_block *sb);
23972 +int au_sbr_realloc(struct au_sbinfo *sbinfo, int nbr);
23974 +unsigned int au_sigen_inc(struct super_block *sb);
23975 +aufs_bindex_t au_new_br_id(struct super_block *sb);
23977 +int si_read_lock(struct super_block *sb, int flags);
23978 +int si_write_lock(struct super_block *sb, int flags);
23979 +int aufs_read_lock(struct dentry *dentry, int flags);
23980 +void aufs_read_unlock(struct dentry *dentry, int flags);
23981 +void aufs_write_lock(struct dentry *dentry);
23982 +void aufs_write_unlock(struct dentry *dentry);
23983 +int aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int flags);
23984 +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2);
23986 +int si_pid_test_slow(struct super_block *sb);
23987 +void si_pid_set_slow(struct super_block *sb);
23988 +void si_pid_clr_slow(struct super_block *sb);
23990 +/* wbr_policy.c */
23991 +extern struct au_wbr_copyup_operations au_wbr_copyup_ops[];
23992 +extern struct au_wbr_create_operations au_wbr_create_ops[];
23993 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst);
23995 +/* ---------------------------------------------------------------------- */
23997 +static inline struct au_sbinfo *au_sbi(struct super_block *sb)
23999 + return sb->s_fs_info;
24002 +/* ---------------------------------------------------------------------- */
24004 +#ifdef CONFIG_AUFS_EXPORT
24005 +void au_export_init(struct super_block *sb);
24007 +static inline int au_test_nfsd(void)
24009 + struct task_struct *tsk = current;
24011 + return (tsk->flags & PF_KTHREAD)
24012 + && !strcmp(tsk->comm, "nfsd");
24015 +void au_xigen_inc(struct inode *inode);
24016 +int au_xigen_new(struct inode *inode);
24017 +int au_xigen_set(struct super_block *sb, struct file *base);
24018 +void au_xigen_clr(struct super_block *sb);
24020 +static inline int au_busy_or_stale(void)
24022 + if (!au_test_nfsd())
24027 +AuStubVoid(au_export_init, struct super_block *sb)
24028 +AuStubInt0(au_test_nfsd, void)
24029 +AuStubVoid(au_xigen_inc, struct inode *inode)
24030 +AuStubInt0(au_xigen_new, struct inode *inode)
24031 +AuStubInt0(au_xigen_set, struct super_block *sb, struct file *base)
24032 +AuStubVoid(au_xigen_clr, struct super_block *sb)
24033 +static inline int au_busy_or_stale(void)
24037 +#endif /* CONFIG_AUFS_EXPORT */
24039 +/* ---------------------------------------------------------------------- */
24041 +#ifdef CONFIG_AUFS_SBILIST
24043 +extern struct au_splhead au_sbilist;
24045 +static inline void au_sbilist_init(void)
24047 + au_spl_init(&au_sbilist);
24050 +static inline void au_sbilist_add(struct super_block *sb)
24052 + au_spl_add(&au_sbi(sb)->si_list, &au_sbilist);
24055 +static inline void au_sbilist_del(struct super_block *sb)
24057 + au_spl_del(&au_sbi(sb)->si_list, &au_sbilist);
24060 +#ifdef CONFIG_AUFS_MAGIC_SYSRQ
24061 +static inline void au_sbilist_lock(void)
24063 + spin_lock(&au_sbilist.spin);
24066 +static inline void au_sbilist_unlock(void)
24068 + spin_unlock(&au_sbilist.spin);
24070 +#define AuGFP_SBILIST GFP_ATOMIC
24072 +AuStubVoid(au_sbilist_lock, void)
24073 +AuStubVoid(au_sbilist_unlock, void)
24074 +#define AuGFP_SBILIST GFP_NOFS
24075 +#endif /* CONFIG_AUFS_MAGIC_SYSRQ */
24077 +AuStubVoid(au_sbilist_init, void)
24078 +AuStubVoid(au_sbilist_add, struct super_block*)
24079 +AuStubVoid(au_sbilist_del, struct super_block*)
24080 +AuStubVoid(au_sbilist_lock, void)
24081 +AuStubVoid(au_sbilist_unlock, void)
24082 +#define AuGFP_SBILIST GFP_NOFS
24085 +/* ---------------------------------------------------------------------- */
24087 +static inline void dbgaufs_si_null(struct au_sbinfo *sbinfo)
24090 + * This function is a dynamic '__init' fucntion actually,
24091 + * so the tiny check for si_rwsem is unnecessary.
24093 + /* AuRwMustWriteLock(&sbinfo->si_rwsem); */
24094 +#ifdef CONFIG_DEBUG_FS
24095 + sbinfo->si_dbgaufs = NULL;
24096 + sbinfo->si_dbgaufs_xib = NULL;
24097 +#ifdef CONFIG_AUFS_EXPORT
24098 + sbinfo->si_dbgaufs_xigen = NULL;
24103 +/* ---------------------------------------------------------------------- */
24105 +static inline pid_t si_pid_bit(void)
24107 + /* the origin of pid is 1, but the bitmap's is 0 */
24108 + return current->pid - 1;
24111 +static inline int si_pid_test(struct super_block *sb)
24113 + pid_t bit = si_pid_bit();
24114 + if (bit < PID_MAX_DEFAULT)
24115 + return test_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
24117 + return si_pid_test_slow(sb);
24120 +static inline void si_pid_set(struct super_block *sb)
24122 + pid_t bit = si_pid_bit();
24123 + if (bit < PID_MAX_DEFAULT) {
24124 + AuDebugOn(test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
24125 + set_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
24128 + si_pid_set_slow(sb);
24131 +static inline void si_pid_clr(struct super_block *sb)
24133 + pid_t bit = si_pid_bit();
24134 + if (bit < PID_MAX_DEFAULT) {
24135 + AuDebugOn(!test_bit(bit, au_sbi(sb)->au_si_pid.bitmap));
24136 + clear_bit(bit, au_sbi(sb)->au_si_pid.bitmap);
24139 + si_pid_clr_slow(sb);
24142 +/* ---------------------------------------------------------------------- */
24144 +/* lock superblock. mainly for entry point functions */
24146 + * __si_read_lock, __si_write_lock,
24147 + * __si_read_unlock, __si_write_unlock, __si_downgrade_lock
24149 +AuSimpleRwsemFuncs(__si, struct super_block *sb, &au_sbi(sb)->si_rwsem);
24151 +#define SiMustNoWaiters(sb) AuRwMustNoWaiters(&au_sbi(sb)->si_rwsem)
24152 +#define SiMustAnyLock(sb) AuRwMustAnyLock(&au_sbi(sb)->si_rwsem)
24153 +#define SiMustWriteLock(sb) AuRwMustWriteLock(&au_sbi(sb)->si_rwsem)
24155 +static inline void si_noflush_read_lock(struct super_block *sb)
24157 + __si_read_lock(sb);
24161 +static inline int si_noflush_read_trylock(struct super_block *sb)
24163 + int locked = __si_read_trylock(sb);
24169 +static inline void si_noflush_write_lock(struct super_block *sb)
24171 + __si_write_lock(sb);
24175 +static inline int si_noflush_write_trylock(struct super_block *sb)
24177 + int locked = __si_write_trylock(sb);
24183 +#if 0 /* unused */
24184 +static inline int si_read_trylock(struct super_block *sb, int flags)
24186 + if (au_ftest_lock(flags, FLUSH))
24187 + au_nwt_flush(&au_sbi(sb)->si_nowait);
24188 + return si_noflush_read_trylock(sb);
24192 +static inline void si_read_unlock(struct super_block *sb)
24195 + __si_read_unlock(sb);
24198 +#if 0 /* unused */
24199 +static inline int si_write_trylock(struct super_block *sb, int flags)
24201 + if (au_ftest_lock(flags, FLUSH))
24202 + au_nwt_flush(&au_sbi(sb)->si_nowait);
24203 + return si_noflush_write_trylock(sb);
24207 +static inline void si_write_unlock(struct super_block *sb)
24210 + __si_write_unlock(sb);
24213 +#if 0 /* unused */
24214 +static inline void si_downgrade_lock(struct super_block *sb)
24216 + __si_downgrade_lock(sb);
24220 +/* ---------------------------------------------------------------------- */
24222 +static inline aufs_bindex_t au_sbend(struct super_block *sb)
24224 + SiMustAnyLock(sb);
24225 + return au_sbi(sb)->si_bend;
24228 +static inline unsigned int au_mntflags(struct super_block *sb)
24230 + SiMustAnyLock(sb);
24231 + return au_sbi(sb)->si_mntflags;
24234 +static inline unsigned int au_sigen(struct super_block *sb)
24236 + SiMustAnyLock(sb);
24237 + return au_sbi(sb)->si_generation;
24240 +static inline void au_ninodes_inc(struct super_block *sb)
24242 + atomic_long_inc(&au_sbi(sb)->si_ninodes);
24245 +static inline void au_ninodes_dec(struct super_block *sb)
24247 + AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_ninodes));
24248 + atomic_long_dec(&au_sbi(sb)->si_ninodes);
24251 +static inline void au_nfiles_inc(struct super_block *sb)
24253 + atomic_long_inc(&au_sbi(sb)->si_nfiles);
24256 +static inline void au_nfiles_dec(struct super_block *sb)
24258 + AuDebugOn(!atomic_long_read(&au_sbi(sb)->si_nfiles));
24259 + atomic_long_dec(&au_sbi(sb)->si_nfiles);
24262 +static inline struct au_branch *au_sbr(struct super_block *sb,
24263 + aufs_bindex_t bindex)
24265 + SiMustAnyLock(sb);
24266 + return au_sbi(sb)->si_branch[0 + bindex];
24269 +static inline void au_xino_brid_set(struct super_block *sb, aufs_bindex_t brid)
24271 + SiMustWriteLock(sb);
24272 + au_sbi(sb)->si_xino_brid = brid;
24275 +static inline aufs_bindex_t au_xino_brid(struct super_block *sb)
24277 + SiMustAnyLock(sb);
24278 + return au_sbi(sb)->si_xino_brid;
24281 +#endif /* __KERNEL__ */
24282 +#endif /* __AUFS_SUPER_H__ */
24283 diff -urN /usr/share/empty/fs/aufs/sysaufs.c linux/fs/aufs/sysaufs.c
24284 --- /usr/share/empty/fs/aufs/sysaufs.c 1970-01-01 01:00:00.000000000 +0100
24285 +++ linux/fs/aufs/sysaufs.c 2011-08-24 13:30:24.734646739 +0200
24288 + * Copyright (C) 2005-2011 Junjiro R. Okajima
24290 + * This program, aufs is free software; you can redistribute it and/or modify
24291 + * it under the terms of the GNU General Public License as published by
24292 + * the Free Software Foundation; either version 2 of the License, or
24293 + * (at your option) any later version.
24295 + * This program is distributed in the hope that it will be useful,
24296 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24297 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24298 + * GNU General Public License for more details.
24300 + * You should have received a copy of the GNU General Public License
24301 + * along with this program; if not, write to the Free Software
24302 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
24306 + * sysfs interface and lifetime management
24307 + * they are necessary regardless sysfs is disabled.
24310 +#include <linux/fs.h>
24311 +#include <linux/random.h>
24312 +#include <linux/sysfs.h>
24315 +unsigned long sysaufs_si_mask;
24316 +struct kset *sysaufs_kset;
24318 +#define AuSiAttr(_name) { \
24319 + .attr = { .name = __stringify(_name), .mode = 0444 }, \
24320 + .show = sysaufs_si_##_name, \
24323 +static struct sysaufs_si_attr sysaufs_si_attr_xi_path = AuSiAttr(xi_path);
24324 +struct attribute *sysaufs_si_attrs[] = {
24325 + &sysaufs_si_attr_xi_path.attr,
24329 +static const struct sysfs_ops au_sbi_ops = {
24330 + .show = sysaufs_si_show
24333 +static struct kobj_type au_sbi_ktype = {
24334 + .release = au_si_free,
24335 + .sysfs_ops = &au_sbi_ops,
24336 + .default_attrs = sysaufs_si_attrs
24339 +/* ---------------------------------------------------------------------- */
24341 +int sysaufs_si_init(struct au_sbinfo *sbinfo)
24345 + sbinfo->si_kobj.kset = sysaufs_kset;
24346 + /* cf. sysaufs_name() */
24347 + err = kobject_init_and_add
24348 + (&sbinfo->si_kobj, &au_sbi_ktype, /*&sysaufs_kset->kobj*/NULL,
24349 + SysaufsSiNamePrefix "%lx", sysaufs_si_id(sbinfo));
24351 + dbgaufs_si_null(sbinfo);
24353 + err = dbgaufs_si_init(sbinfo);
24354 + if (unlikely(err))
24355 + kobject_put(&sbinfo->si_kobj);
24360 +void sysaufs_fin(void)
24363 + sysfs_remove_group(&sysaufs_kset->kobj, sysaufs_attr_group);
24364 + kset_unregister(sysaufs_kset);
24367 +int __init sysaufs_init(void)
24372 + get_random_bytes(&sysaufs_si_mask, sizeof(sysaufs_si_mask));
24373 + } while (!sysaufs_si_mask);
24376 + sysaufs_kset = kset_create_and_add(AUFS_NAME, NULL, fs_kobj);
24377 + if (unlikely(!sysaufs_kset))
24379 + err = PTR_ERR(sysaufs_kset);
24380 + if (IS_ERR(sysaufs_kset))
24382 + err = sysfs_create_group(&sysaufs_kset->kobj, sysaufs_attr_group);
24383 + if (unlikely(err)) {
24384 + kset_unregister(sysaufs_kset);
24388 + err = dbgaufs_init();
24389 + if (unlikely(err))
24394 diff -urN /usr/share/empty/fs/aufs/sysaufs.h linux/fs/aufs/sysaufs.h
24395 --- /usr/share/empty/fs/aufs/sysaufs.h 1970-01-01 01:00:00.000000000 +0100
24396 +++ linux/fs/aufs/sysaufs.h 2011-08-24 13:30:24.734646739 +0200
24399 + * Copyright (C) 2005-2011 Junjiro R. Okajima
24401 + * This program, aufs is free software; you can redistribute it and/or modify
24402 + * it under the terms of the GNU General Public License as published by
24403 + * the Free Software Foundation; either version 2 of the License, or
24404 + * (at your option) any later version.
24406 + * This program is distributed in the hope that it will be useful,
24407 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24408 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24409 + * GNU General Public License for more details.
24411 + * You should have received a copy of the GNU General Public License
24412 + * along with this program; if not, write to the Free Software
24413 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
24417 + * sysfs interface and mount lifetime management
24420 +#ifndef __SYSAUFS_H__
24421 +#define __SYSAUFS_H__
24425 +#include <linux/sysfs.h>
24426 +#include <linux/aufs_type.h>
24427 +#include "module.h"
24429 +struct super_block;
24432 +struct sysaufs_si_attr {
24433 + struct attribute attr;
24434 + int (*show)(struct seq_file *seq, struct super_block *sb);
24437 +/* ---------------------------------------------------------------------- */
24440 +extern unsigned long sysaufs_si_mask;
24441 +extern struct kset *sysaufs_kset;
24442 +extern struct attribute *sysaufs_si_attrs[];
24443 +int sysaufs_si_init(struct au_sbinfo *sbinfo);
24444 +int __init sysaufs_init(void);
24445 +void sysaufs_fin(void);
24447 +/* ---------------------------------------------------------------------- */
24449 +/* some people doesn't like to show a pointer in kernel */
24450 +static inline unsigned long sysaufs_si_id(struct au_sbinfo *sbinfo)
24452 + return sysaufs_si_mask ^ (unsigned long)sbinfo;
24455 +#define SysaufsSiNamePrefix "si_"
24456 +#define SysaufsSiNameLen (sizeof(SysaufsSiNamePrefix) + 16)
24457 +static inline void sysaufs_name(struct au_sbinfo *sbinfo, char *name)
24459 + snprintf(name, SysaufsSiNameLen, SysaufsSiNamePrefix "%lx",
24460 + sysaufs_si_id(sbinfo));
24464 +#ifdef CONFIG_SYSFS
24466 +extern struct attribute_group *sysaufs_attr_group;
24468 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb);
24469 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
24472 +void sysaufs_br_init(struct au_branch *br);
24473 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex);
24474 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex);
24476 +#define sysaufs_brs_init() do {} while (0)
24479 +#define sysaufs_attr_group NULL
24481 +AuStubInt0(sysaufs_si_xi_path, struct seq_file *seq, struct super_block *sb)
24484 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
24490 +AuStubVoid(sysaufs_br_init, struct au_branch *br)
24491 +AuStubVoid(sysaufs_brs_add, struct super_block *sb, aufs_bindex_t bindex)
24492 +AuStubVoid(sysaufs_brs_del, struct super_block *sb, aufs_bindex_t bindex)
24494 +static inline void sysaufs_brs_init(void)
24499 +#endif /* CONFIG_SYSFS */
24501 +#endif /* __KERNEL__ */
24502 +#endif /* __SYSAUFS_H__ */
24503 diff -urN /usr/share/empty/fs/aufs/sysfs.c linux/fs/aufs/sysfs.c
24504 --- /usr/share/empty/fs/aufs/sysfs.c 1970-01-01 01:00:00.000000000 +0100
24505 +++ linux/fs/aufs/sysfs.c 2012-01-05 11:00:59.218835185 +0100
24508 + * Copyright (C) 2005-2011 Junjiro R. Okajima
24510 + * This program, aufs is free software; you can redistribute it and/or modify
24511 + * it under the terms of the GNU General Public License as published by
24512 + * the Free Software Foundation; either version 2 of the License, or
24513 + * (at your option) any later version.
24515 + * This program is distributed in the hope that it will be useful,
24516 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24517 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24518 + * GNU General Public License for more details.
24520 + * You should have received a copy of the GNU General Public License
24521 + * along with this program; if not, write to the Free Software
24522 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
24526 + * sysfs interface
24529 +#include <linux/fs.h>
24530 +#include <linux/module.h>
24531 +#include <linux/seq_file.h>
24532 +#include <linux/sysfs.h>
24535 +#ifdef CONFIG_AUFS_FS_MODULE
24536 +/* this entry violates the "one line per file" policy of sysfs */
24537 +static ssize_t config_show(struct kobject *kobj, struct kobj_attribute *attr,
24541 + static char *conf =
24542 +/* this file is generated at compiling */
24543 +#include "conf.str"
24546 + err = snprintf(buf, PAGE_SIZE, conf);
24547 + if (unlikely(err >= PAGE_SIZE))
24552 +static struct kobj_attribute au_config_attr = __ATTR_RO(config);
24555 +static struct attribute *au_attr[] = {
24556 +#ifdef CONFIG_AUFS_FS_MODULE
24557 + &au_config_attr.attr,
24559 + NULL, /* need to NULL terminate the list of attributes */
24562 +static struct attribute_group sysaufs_attr_group_body = {
24566 +struct attribute_group *sysaufs_attr_group = &sysaufs_attr_group_body;
24568 +/* ---------------------------------------------------------------------- */
24570 +int sysaufs_si_xi_path(struct seq_file *seq, struct super_block *sb)
24574 + SiMustAnyLock(sb);
24577 + if (au_opt_test(au_mntflags(sb), XINO)) {
24578 + err = au_xino_path(seq, au_sbi(sb)->si_xib);
24579 + seq_putc(seq, '\n');
24585 + * the lifetime of branch is independent from the entry under sysfs.
24586 + * sysfs handles the lifetime of the entry, and never call ->show() after it is
24589 +static int sysaufs_si_br(struct seq_file *seq, struct super_block *sb,
24590 + aufs_bindex_t bindex)
24593 + struct path path;
24594 + struct dentry *root;
24595 + struct au_branch *br;
24598 + AuDbg("b%d\n", bindex);
24601 + root = sb->s_root;
24602 + di_read_lock_parent(root, !AuLock_IR);
24603 + br = au_sbr(sb, bindex);
24604 + path.mnt = br->br_mnt;
24605 + path.dentry = au_h_dptr(root, bindex);
24606 + au_seq_path(seq, &path);
24607 + di_read_unlock(root, !AuLock_IR);
24608 + perm = au_optstr_br_perm(br->br_perm);
24610 + err = seq_printf(seq, "=%s\n", perm);
24619 +/* ---------------------------------------------------------------------- */
24621 +static struct seq_file *au_seq(char *p, ssize_t len)
24623 + struct seq_file *seq;
24625 + seq = kzalloc(sizeof(*seq), GFP_NOFS);
24627 + /* mutex_init(&seq.lock); */
24630 + return seq; /* success */
24633 + seq = ERR_PTR(-ENOMEM);
24637 +#define SysaufsBr_PREFIX "br"
24639 +/* todo: file size may exceed PAGE_SIZE */
24640 +ssize_t sysaufs_si_show(struct kobject *kobj, struct attribute *attr,
24645 + aufs_bindex_t bend;
24646 + struct au_sbinfo *sbinfo;
24647 + struct super_block *sb;
24648 + struct seq_file *seq;
24650 + struct attribute **cattr;
24652 + sbinfo = container_of(kobj, struct au_sbinfo, si_kobj);
24653 + sb = sbinfo->si_sb;
24656 + * prevent a race condition between sysfs and aufs.
24657 + * for instance, sysfs_file_read() calls sysfs_get_active_two() which
24658 + * prohibits maintaining the sysfs entries.
24659 + * hew we acquire read lock after sysfs_get_active_two().
24660 + * on the other hand, the remount process may maintain the sysfs/aufs
24661 + * entries after acquiring write lock.
24662 + * it can cause a deadlock.
24663 + * simply we gave up processing read here.
24666 + if (unlikely(!si_noflush_read_trylock(sb)))
24669 + seq = au_seq(buf, PAGE_SIZE);
24670 + err = PTR_ERR(seq);
24674 + name = (void *)attr->name;
24675 + cattr = sysaufs_si_attrs;
24677 + if (!strcmp(name, (*cattr)->name)) {
24678 + err = container_of(*cattr, struct sysaufs_si_attr, attr)
24685 + bend = au_sbend(sb);
24686 + if (!strncmp(name, SysaufsBr_PREFIX, sizeof(SysaufsBr_PREFIX) - 1)) {
24687 + name += sizeof(SysaufsBr_PREFIX) - 1;
24688 + err = kstrtol(name, 10, &l);
24691 + err = sysaufs_si_br(seq, sb, (aufs_bindex_t)l);
24701 + err = seq->count;
24702 + /* sysfs limit */
24703 + if (unlikely(err == PAGE_SIZE))
24708 + si_read_unlock(sb);
24713 +/* ---------------------------------------------------------------------- */
24715 +void sysaufs_br_init(struct au_branch *br)
24717 + struct attribute *attr = &br->br_attr;
24719 + sysfs_attr_init(attr);
24720 + attr->name = br->br_name;
24721 + attr->mode = S_IRUGO;
24724 +void sysaufs_brs_del(struct super_block *sb, aufs_bindex_t bindex)
24726 + struct au_branch *br;
24727 + struct kobject *kobj;
24728 + aufs_bindex_t bend;
24730 + dbgaufs_brs_del(sb, bindex);
24732 + if (!sysaufs_brs)
24735 + kobj = &au_sbi(sb)->si_kobj;
24736 + bend = au_sbend(sb);
24737 + for (; bindex <= bend; bindex++) {
24738 + br = au_sbr(sb, bindex);
24739 + sysfs_remove_file(kobj, &br->br_attr);
24743 +void sysaufs_brs_add(struct super_block *sb, aufs_bindex_t bindex)
24746 + aufs_bindex_t bend;
24747 + struct kobject *kobj;
24748 + struct au_branch *br;
24750 + dbgaufs_brs_add(sb, bindex);
24752 + if (!sysaufs_brs)
24755 + kobj = &au_sbi(sb)->si_kobj;
24756 + bend = au_sbend(sb);
24757 + for (; bindex <= bend; bindex++) {
24758 + br = au_sbr(sb, bindex);
24759 + snprintf(br->br_name, sizeof(br->br_name), SysaufsBr_PREFIX
24761 + err = sysfs_create_file(kobj, &br->br_attr);
24762 + if (unlikely(err))
24763 + pr_warning("failed %s under sysfs(%d)\n",
24764 + br->br_name, err);
24767 diff -urN /usr/share/empty/fs/aufs/sysrq.c linux/fs/aufs/sysrq.c
24768 --- /usr/share/empty/fs/aufs/sysrq.c 1970-01-01 01:00:00.000000000 +0100
24769 +++ linux/fs/aufs/sysrq.c 2011-08-24 13:30:24.734646739 +0200
24772 + * Copyright (C) 2005-2011 Junjiro R. Okajima
24774 + * This program, aufs is free software; you can redistribute it and/or modify
24775 + * it under the terms of the GNU General Public License as published by
24776 + * the Free Software Foundation; either version 2 of the License, or
24777 + * (at your option) any later version.
24779 + * This program is distributed in the hope that it will be useful,
24780 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24781 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24782 + * GNU General Public License for more details.
24784 + * You should have received a copy of the GNU General Public License
24785 + * along with this program; if not, write to the Free Software
24786 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
24790 + * magic sysrq hanlder
24793 +#include <linux/fs.h>
24794 +#include <linux/module.h>
24795 +#include <linux/moduleparam.h>
24796 +/* #include <linux/sysrq.h> */
24797 +#include <linux/writeback.h>
24800 +/* ---------------------------------------------------------------------- */
24802 +static void sysrq_sb(struct super_block *sb)
24805 + struct au_sbinfo *sbinfo;
24806 + struct file *file;
24808 + plevel = au_plevel;
24809 + au_plevel = KERN_WARNING;
24811 + sbinfo = au_sbi(sb);
24812 + /* since we define pr_fmt, call printk directly */
24813 + printk(KERN_WARNING "si=%lx\n", sysaufs_si_id(sbinfo));
24814 + printk(KERN_WARNING AUFS_NAME ": superblock\n");
24818 + printk(KERN_WARNING AUFS_NAME ": root dentry\n");
24819 + au_dpri_dentry(sb->s_root);
24820 + printk(KERN_WARNING AUFS_NAME ": root inode\n");
24821 + au_dpri_inode(sb->s_root->d_inode);
24826 + int err, i, j, ndentry;
24827 + struct au_dcsub_pages dpages;
24828 + struct au_dpage *dpage;
24830 + err = au_dpages_init(&dpages, GFP_ATOMIC);
24831 + if (unlikely(err))
24833 + err = au_dcsub_pages(&dpages, sb->s_root, NULL, NULL);
24835 + for (i = 0; i < dpages.ndpage; i++) {
24836 + dpage = dpages.dpages + i;
24837 + ndentry = dpage->ndentry;
24838 + for (j = 0; j < ndentry; j++)
24839 + au_dpri_dentry(dpage->dentries[j]);
24841 + au_dpages_free(&dpages);
24848 + printk(KERN_WARNING AUFS_NAME ": isolated inode\n");
24849 + spin_lock(&inode_sb_list_lock);
24850 + list_for_each_entry(i, &sb->s_inodes, i_sb_list) {
24851 + spin_lock(&i->i_lock);
24852 + if (1 || list_empty(&i->i_dentry))
24853 + au_dpri_inode(i);
24854 + spin_unlock(&i->i_lock);
24856 + spin_unlock(&inode_sb_list_lock);
24859 + printk(KERN_WARNING AUFS_NAME ": files\n");
24860 + lg_global_lock(files_lglock);
24861 + do_file_list_for_each_entry(sb, file) {
24863 + mode = file->f_dentry->d_inode->i_mode;
24864 + if (!special_file(mode) || au_special_file(mode))
24865 + au_dpri_file(file);
24866 + } while_file_list_for_each_entry;
24867 + lg_global_unlock(files_lglock);
24868 + printk(KERN_WARNING AUFS_NAME ": done\n");
24870 + au_plevel = plevel;
24873 +/* ---------------------------------------------------------------------- */
24875 +/* module parameter */
24876 +static char *aufs_sysrq_key = "a";
24877 +module_param_named(sysrq, aufs_sysrq_key, charp, S_IRUGO);
24878 +MODULE_PARM_DESC(sysrq, "MagicSysRq key for " AUFS_NAME);
24880 +static void au_sysrq(int key __maybe_unused)
24882 + struct au_sbinfo *sbinfo;
24885 + au_sbilist_lock();
24886 + list_for_each_entry(sbinfo, &au_sbilist.head, si_list)
24887 + sysrq_sb(sbinfo->si_sb);
24888 + au_sbilist_unlock();
24892 +static struct sysrq_key_op au_sysrq_op = {
24893 + .handler = au_sysrq,
24894 + .help_msg = "Aufs",
24895 + .action_msg = "Aufs",
24896 + .enable_mask = SYSRQ_ENABLE_DUMP
24899 +/* ---------------------------------------------------------------------- */
24901 +int __init au_sysrq_init(void)
24907 + key = *aufs_sysrq_key;
24908 + if ('a' <= key && key <= 'z')
24909 + err = register_sysrq_key(key, &au_sysrq_op);
24910 + if (unlikely(err))
24911 + pr_err("err %d, sysrq=%c\n", err, key);
24915 +void au_sysrq_fin(void)
24918 + err = unregister_sysrq_key(*aufs_sysrq_key, &au_sysrq_op);
24919 + if (unlikely(err))
24920 + pr_err("err %d (ignored)\n", err);
24922 diff -urN /usr/share/empty/fs/aufs/vdir.c linux/fs/aufs/vdir.c
24923 --- /usr/share/empty/fs/aufs/vdir.c 1970-01-01 01:00:00.000000000 +0100
24924 +++ linux/fs/aufs/vdir.c 2011-08-24 13:30:24.734646739 +0200
24927 + * Copyright (C) 2005-2011 Junjiro R. Okajima
24929 + * This program, aufs is free software; you can redistribute it and/or modify
24930 + * it under the terms of the GNU General Public License as published by
24931 + * the Free Software Foundation; either version 2 of the License, or
24932 + * (at your option) any later version.
24934 + * This program is distributed in the hope that it will be useful,
24935 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
24936 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24937 + * GNU General Public License for more details.
24939 + * You should have received a copy of the GNU General Public License
24940 + * along with this program; if not, write to the Free Software
24941 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
24945 + * virtual or vertical directory
24948 +#include <linux/hash.h>
24951 +static unsigned int calc_size(int nlen)
24953 + return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
24956 +static int set_deblk_end(union au_vdir_deblk_p *p,
24957 + union au_vdir_deblk_p *deblk_end)
24959 + if (calc_size(0) <= deblk_end->deblk - p->deblk) {
24960 + p->de->de_str.len = 0;
24964 + return -1; /* error */
24967 +/* returns true or false */
24968 +static int is_deblk_end(union au_vdir_deblk_p *p,
24969 + union au_vdir_deblk_p *deblk_end)
24971 + if (calc_size(0) <= deblk_end->deblk - p->deblk)
24972 + return !p->de->de_str.len;
24976 +static unsigned char *last_deblk(struct au_vdir *vdir)
24978 + return vdir->vd_deblk[vdir->vd_nblk - 1];
24981 +/* ---------------------------------------------------------------------- */
24983 +/* estimate the apropriate size for name hash table */
24984 +unsigned int au_rdhash_est(loff_t sz)
24992 + if (sz < AUFS_RDHASH_DEF)
24993 + n = AUFS_RDHASH_DEF;
24994 + /* pr_info("n %u\n", n); */
24999 + * the allocated memory has to be freed by
25000 + * au_nhash_wh_free() or au_nhash_de_free().
25002 +int au_nhash_alloc(struct au_nhash *nhash, unsigned int num_hash, gfp_t gfp)
25004 + struct hlist_head *head;
25007 + head = kmalloc(sizeof(*nhash->nh_head) * num_hash, gfp);
25009 + nhash->nh_num = num_hash;
25010 + nhash->nh_head = head;
25011 + for (u = 0; u < num_hash; u++)
25012 + INIT_HLIST_HEAD(head++);
25013 + return 0; /* success */
25019 +static void nhash_count(struct hlist_head *head)
25023 + struct hlist_node *pos;
25026 + hlist_for_each(pos, head)
25028 + pr_info("%lu\n", n);
25032 +static void au_nhash_wh_do_free(struct hlist_head *head)
25034 + struct au_vdir_wh *tpos;
25035 + struct hlist_node *pos, *node;
25037 + hlist_for_each_entry_safe(tpos, pos, node, head, wh_hash) {
25038 + /* hlist_del(pos); */
25043 +static void au_nhash_de_do_free(struct hlist_head *head)
25045 + struct au_vdir_dehstr *tpos;
25046 + struct hlist_node *pos, *node;
25048 + hlist_for_each_entry_safe(tpos, pos, node, head, hash) {
25049 + /* hlist_del(pos); */
25050 + au_cache_free_vdir_dehstr(tpos);
25054 +static void au_nhash_do_free(struct au_nhash *nhash,
25055 + void (*free)(struct hlist_head *head))
25058 + struct hlist_head *head;
25060 + n = nhash->nh_num;
25064 + head = nhash->nh_head;
25065 + while (n-- > 0) {
25066 + nhash_count(head);
25069 + kfree(nhash->nh_head);
25072 +void au_nhash_wh_free(struct au_nhash *whlist)
25074 + au_nhash_do_free(whlist, au_nhash_wh_do_free);
25077 +static void au_nhash_de_free(struct au_nhash *delist)
25079 + au_nhash_do_free(delist, au_nhash_de_do_free);
25082 +/* ---------------------------------------------------------------------- */
25084 +int au_nhash_test_longer_wh(struct au_nhash *whlist, aufs_bindex_t btgt,
25088 + unsigned int u, n;
25089 + struct hlist_head *head;
25090 + struct au_vdir_wh *tpos;
25091 + struct hlist_node *pos;
25094 + n = whlist->nh_num;
25095 + head = whlist->nh_head;
25096 + for (u = 0; u < n; u++, head++)
25097 + hlist_for_each_entry(tpos, pos, head, wh_hash)
25098 + if (tpos->wh_bindex == btgt && ++num > limit)
25103 +static struct hlist_head *au_name_hash(struct au_nhash *nhash,
25104 + unsigned char *name,
25105 + unsigned int len)
25108 + /* const unsigned int magic_bit = 12; */
25110 + AuDebugOn(!nhash->nh_num || !nhash->nh_head);
25115 + /* v = hash_long(v, magic_bit); */
25116 + v %= nhash->nh_num;
25117 + return nhash->nh_head + v;
25120 +static int au_nhash_test_name(struct au_vdir_destr *str, const char *name,
25123 + return str->len == nlen && !memcmp(str->name, name, nlen);
25126 +/* returns found or not */
25127 +int au_nhash_test_known_wh(struct au_nhash *whlist, char *name, int nlen)
25129 + struct hlist_head *head;
25130 + struct au_vdir_wh *tpos;
25131 + struct hlist_node *pos;
25132 + struct au_vdir_destr *str;
25134 + head = au_name_hash(whlist, name, nlen);
25135 + hlist_for_each_entry(tpos, pos, head, wh_hash) {
25136 + str = &tpos->wh_str;
25137 + AuDbg("%.*s\n", str->len, str->name);
25138 + if (au_nhash_test_name(str, name, nlen))
25144 +/* returns found(true) or not */
25145 +static int test_known(struct au_nhash *delist, char *name, int nlen)
25147 + struct hlist_head *head;
25148 + struct au_vdir_dehstr *tpos;
25149 + struct hlist_node *pos;
25150 + struct au_vdir_destr *str;
25152 + head = au_name_hash(delist, name, nlen);
25153 + hlist_for_each_entry(tpos, pos, head, hash) {
25155 + AuDbg("%.*s\n", str->len, str->name);
25156 + if (au_nhash_test_name(str, name, nlen))
25162 +static void au_shwh_init_wh(struct au_vdir_wh *wh, ino_t ino,
25163 + unsigned char d_type)
25165 +#ifdef CONFIG_AUFS_SHWH
25166 + wh->wh_ino = ino;
25167 + wh->wh_type = d_type;
25171 +/* ---------------------------------------------------------------------- */
25173 +int au_nhash_append_wh(struct au_nhash *whlist, char *name, int nlen, ino_t ino,
25174 + unsigned int d_type, aufs_bindex_t bindex,
25175 + unsigned char shwh)
25178 + struct au_vdir_destr *str;
25179 + struct au_vdir_wh *wh;
25181 + AuDbg("%.*s\n", nlen, name);
25182 + AuDebugOn(!whlist->nh_num || !whlist->nh_head);
25185 + wh = kmalloc(sizeof(*wh) + nlen, GFP_NOFS);
25186 + if (unlikely(!wh))
25190 + wh->wh_bindex = bindex;
25192 + au_shwh_init_wh(wh, ino, d_type);
25193 + str = &wh->wh_str;
25195 + memcpy(str->name, name, nlen);
25196 + hlist_add_head(&wh->wh_hash, au_name_hash(whlist, name, nlen));
25203 +static int append_deblk(struct au_vdir *vdir)
25206 + unsigned long ul;
25207 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
25208 + union au_vdir_deblk_p p, deblk_end;
25209 + unsigned char **o;
25212 + o = krealloc(vdir->vd_deblk, sizeof(*o) * (vdir->vd_nblk + 1),
25214 + if (unlikely(!o))
25217 + vdir->vd_deblk = o;
25218 + p.deblk = kmalloc(deblk_sz, GFP_NOFS);
25220 + ul = vdir->vd_nblk++;
25221 + vdir->vd_deblk[ul] = p.deblk;
25222 + vdir->vd_last.ul = ul;
25223 + vdir->vd_last.p.deblk = p.deblk;
25224 + deblk_end.deblk = p.deblk + deblk_sz;
25225 + err = set_deblk_end(&p, &deblk_end);
25232 +static int append_de(struct au_vdir *vdir, char *name, int nlen, ino_t ino,
25233 + unsigned int d_type, struct au_nhash *delist)
25237 + const unsigned int deblk_sz = vdir->vd_deblk_sz;
25238 + union au_vdir_deblk_p p, *room, deblk_end;
25239 + struct au_vdir_dehstr *dehstr;
25241 + p.deblk = last_deblk(vdir);
25242 + deblk_end.deblk = p.deblk + deblk_sz;
25243 + room = &vdir->vd_last.p;
25244 + AuDebugOn(room->deblk < p.deblk || deblk_end.deblk <= room->deblk
25245 + || !is_deblk_end(room, &deblk_end));
25247 + sz = calc_size(nlen);
25248 + if (unlikely(sz > deblk_end.deblk - room->deblk)) {
25249 + err = append_deblk(vdir);
25250 + if (unlikely(err))
25253 + p.deblk = last_deblk(vdir);
25254 + deblk_end.deblk = p.deblk + deblk_sz;
25256 + AuDebugOn(room->deblk != p.deblk);
25260 + dehstr = au_cache_alloc_vdir_dehstr();
25261 + if (unlikely(!dehstr))
25264 + dehstr->str = &room->de->de_str;
25265 + hlist_add_head(&dehstr->hash, au_name_hash(delist, name, nlen));
25266 + room->de->de_ino = ino;
25267 + room->de->de_type = d_type;
25268 + room->de->de_str.len = nlen;
25269 + memcpy(room->de->de_str.name, name, nlen);
25272 + room->deblk += sz;
25273 + if (unlikely(set_deblk_end(room, &deblk_end)))
25274 + err = append_deblk(vdir);
25281 +/* ---------------------------------------------------------------------- */
25283 +void au_vdir_free(struct au_vdir *vdir)
25285 + unsigned char **deblk;
25287 + deblk = vdir->vd_deblk;
25288 + while (vdir->vd_nblk--)
25290 + kfree(vdir->vd_deblk);
25291 + au_cache_free_vdir(vdir);
25294 +static struct au_vdir *alloc_vdir(struct file *file)
25296 + struct au_vdir *vdir;
25297 + struct super_block *sb;
25300 + sb = file->f_dentry->d_sb;
25301 + SiMustAnyLock(sb);
25304 + vdir = au_cache_alloc_vdir();
25305 + if (unlikely(!vdir))
25308 + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_NOFS);
25309 + if (unlikely(!vdir->vd_deblk))
25312 + vdir->vd_deblk_sz = au_sbi(sb)->si_rdblk;
25313 + if (!vdir->vd_deblk_sz) {
25314 + /* estimate the apropriate size for deblk */
25315 + vdir->vd_deblk_sz = au_dir_size(file, /*dentry*/NULL);
25316 + /* pr_info("vd_deblk_sz %u\n", vdir->vd_deblk_sz); */
25318 + vdir->vd_nblk = 0;
25319 + vdir->vd_version = 0;
25320 + vdir->vd_jiffy = 0;
25321 + err = append_deblk(vdir);
25323 + return vdir; /* success */
25325 + kfree(vdir->vd_deblk);
25328 + au_cache_free_vdir(vdir);
25330 + vdir = ERR_PTR(err);
25334 +static int reinit_vdir(struct au_vdir *vdir)
25337 + union au_vdir_deblk_p p, deblk_end;
25339 + while (vdir->vd_nblk > 1) {
25340 + kfree(vdir->vd_deblk[vdir->vd_nblk - 1]);
25341 + /* vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; */
25344 + p.deblk = vdir->vd_deblk[0];
25345 + deblk_end.deblk = p.deblk + vdir->vd_deblk_sz;
25346 + err = set_deblk_end(&p, &deblk_end);
25347 + /* keep vd_dblk_sz */
25348 + vdir->vd_last.ul = 0;
25349 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
25350 + vdir->vd_version = 0;
25351 + vdir->vd_jiffy = 0;
25356 +/* ---------------------------------------------------------------------- */
25358 +#define AuFillVdir_CALLED 1
25359 +#define AuFillVdir_WHABLE (1 << 1)
25360 +#define AuFillVdir_SHWH (1 << 2)
25361 +#define au_ftest_fillvdir(flags, name) ((flags) & AuFillVdir_##name)
25362 +#define au_fset_fillvdir(flags, name) \
25363 + do { (flags) |= AuFillVdir_##name; } while (0)
25364 +#define au_fclr_fillvdir(flags, name) \
25365 + do { (flags) &= ~AuFillVdir_##name; } while (0)
25367 +#ifndef CONFIG_AUFS_SHWH
25368 +#undef AuFillVdir_SHWH
25369 +#define AuFillVdir_SHWH 0
25372 +struct fillvdir_arg {
25373 + struct file *file;
25374 + struct au_vdir *vdir;
25375 + struct au_nhash delist;
25376 + struct au_nhash whlist;
25377 + aufs_bindex_t bindex;
25378 + unsigned int flags;
25382 +static int fillvdir(void *__arg, const char *__name, int nlen,
25383 + loff_t offset __maybe_unused, u64 h_ino,
25384 + unsigned int d_type)
25386 + struct fillvdir_arg *arg = __arg;
25387 + char *name = (void *)__name;
25388 + struct super_block *sb;
25390 + const unsigned char shwh = !!au_ftest_fillvdir(arg->flags, SHWH);
25393 + sb = arg->file->f_dentry->d_sb;
25394 + au_fset_fillvdir(arg->flags, CALLED);
25396 + if (nlen <= AUFS_WH_PFX_LEN
25397 + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) {
25398 + if (test_known(&arg->delist, name, nlen)
25399 + || au_nhash_test_known_wh(&arg->whlist, name, nlen))
25400 + goto out; /* already exists or whiteouted */
25402 + sb = arg->file->f_dentry->d_sb;
25403 + arg->err = au_ino(sb, arg->bindex, h_ino, d_type, &ino);
25405 + if (unlikely(nlen > AUFS_MAX_NAMELEN))
25406 + d_type = DT_UNKNOWN;
25407 + arg->err = append_de(arg->vdir, name, nlen, ino,
25408 + d_type, &arg->delist);
25410 + } else if (au_ftest_fillvdir(arg->flags, WHABLE)) {
25411 + name += AUFS_WH_PFX_LEN;
25412 + nlen -= AUFS_WH_PFX_LEN;
25413 + if (au_nhash_test_known_wh(&arg->whlist, name, nlen))
25414 + goto out; /* already whiteouted */
25417 + arg->err = au_wh_ino(sb, arg->bindex, h_ino, d_type,
25420 + if (nlen <= AUFS_MAX_NAMELEN + AUFS_WH_PFX_LEN)
25421 + d_type = DT_UNKNOWN;
25422 + arg->err = au_nhash_append_wh
25423 + (&arg->whlist, name, nlen, ino, d_type,
25424 + arg->bindex, shwh);
25430 + arg->vdir->vd_jiffy = jiffies;
25432 + AuTraceErr(arg->err);
25436 +static int au_handle_shwh(struct super_block *sb, struct au_vdir *vdir,
25437 + struct au_nhash *whlist, struct au_nhash *delist)
25439 +#ifdef CONFIG_AUFS_SHWH
25441 + unsigned int nh, u;
25442 + struct hlist_head *head;
25443 + struct au_vdir_wh *tpos;
25444 + struct hlist_node *pos, *n;
25446 + struct au_vdir_destr *destr;
25448 + AuDebugOn(!au_opt_test(au_mntflags(sb), SHWH));
25451 + o = p = __getname_gfp(GFP_NOFS);
25452 + if (unlikely(!p))
25456 + nh = whlist->nh_num;
25457 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
25458 + p += AUFS_WH_PFX_LEN;
25459 + for (u = 0; u < nh; u++) {
25460 + head = whlist->nh_head + u;
25461 + hlist_for_each_entry_safe(tpos, pos, n, head, wh_hash) {
25462 + destr = &tpos->wh_str;
25463 + memcpy(p, destr->name, destr->len);
25464 + err = append_de(vdir, o, destr->len + AUFS_WH_PFX_LEN,
25465 + tpos->wh_ino, tpos->wh_type, delist);
25466 + if (unlikely(err))
25481 +static int au_do_read_vdir(struct fillvdir_arg *arg)
25484 + unsigned int rdhash;
25486 + aufs_bindex_t bend, bindex, bstart;
25487 + unsigned char shwh;
25488 + struct file *hf, *file;
25489 + struct super_block *sb;
25491 + file = arg->file;
25492 + sb = file->f_dentry->d_sb;
25493 + SiMustAnyLock(sb);
25495 + rdhash = au_sbi(sb)->si_rdhash;
25497 + rdhash = au_rdhash_est(au_dir_size(file, /*dentry*/NULL));
25498 + err = au_nhash_alloc(&arg->delist, rdhash, GFP_NOFS);
25499 + if (unlikely(err))
25501 + err = au_nhash_alloc(&arg->whlist, rdhash, GFP_NOFS);
25502 + if (unlikely(err))
25508 + if (au_opt_test(au_mntflags(sb), SHWH)) {
25510 + au_fset_fillvdir(arg->flags, SHWH);
25512 + bstart = au_fbstart(file);
25513 + bend = au_fbend_dir(file);
25514 + for (bindex = bstart; !err && bindex <= bend; bindex++) {
25515 + hf = au_hf_dir(file, bindex);
25519 + offset = vfsub_llseek(hf, 0, SEEK_SET);
25521 + if (unlikely(offset))
25524 + arg->bindex = bindex;
25525 + au_fclr_fillvdir(arg->flags, WHABLE);
25527 + || (bindex != bend
25528 + && au_br_whable(au_sbr_perm(sb, bindex))))
25529 + au_fset_fillvdir(arg->flags, WHABLE);
25532 + au_fclr_fillvdir(arg->flags, CALLED);
25534 + err = vfsub_readdir(hf, fillvdir, arg);
25537 + } while (!err && au_ftest_fillvdir(arg->flags, CALLED));
25540 + if (!err && shwh)
25541 + err = au_handle_shwh(sb, arg->vdir, &arg->whlist, &arg->delist);
25543 + au_nhash_wh_free(&arg->whlist);
25546 + au_nhash_de_free(&arg->delist);
25551 +static int read_vdir(struct file *file, int may_read)
25554 + unsigned long expire;
25555 + unsigned char do_read;
25556 + struct fillvdir_arg arg;
25557 + struct inode *inode;
25558 + struct au_vdir *vdir, *allocated;
25561 + inode = file->f_dentry->d_inode;
25562 + IMustLock(inode);
25563 + SiMustAnyLock(inode->i_sb);
25565 + allocated = NULL;
25567 + expire = au_sbi(inode->i_sb)->si_rdcache;
25568 + vdir = au_ivdir(inode);
25571 + vdir = alloc_vdir(file);
25572 + err = PTR_ERR(vdir);
25573 + if (IS_ERR(vdir))
25576 + allocated = vdir;
25577 + } else if (may_read
25578 + && (inode->i_version != vdir->vd_version
25579 + || time_after(jiffies, vdir->vd_jiffy + expire))) {
25581 + err = reinit_vdir(vdir);
25582 + if (unlikely(err))
25587 + return 0; /* success */
25591 + err = au_do_read_vdir(&arg);
25593 + /* file->f_pos = 0; */
25594 + vdir->vd_version = inode->i_version;
25595 + vdir->vd_last.ul = 0;
25596 + vdir->vd_last.p.deblk = vdir->vd_deblk[0];
25598 + au_set_ivdir(inode, allocated);
25599 + } else if (allocated)
25600 + au_vdir_free(allocated);
25606 +static int copy_vdir(struct au_vdir *tgt, struct au_vdir *src)
25609 + unsigned long ul, n;
25610 + const unsigned int deblk_sz = src->vd_deblk_sz;
25612 + AuDebugOn(tgt->vd_nblk != 1);
25615 + if (tgt->vd_nblk < src->vd_nblk) {
25616 + unsigned char **p;
25618 + p = krealloc(tgt->vd_deblk, sizeof(*p) * src->vd_nblk,
25620 + if (unlikely(!p))
25622 + tgt->vd_deblk = p;
25625 + if (tgt->vd_deblk_sz != deblk_sz) {
25626 + unsigned char *p;
25628 + tgt->vd_deblk_sz = deblk_sz;
25629 + p = krealloc(tgt->vd_deblk[0], deblk_sz, GFP_NOFS);
25630 + if (unlikely(!p))
25632 + tgt->vd_deblk[0] = p;
25634 + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], deblk_sz);
25635 + tgt->vd_version = src->vd_version;
25636 + tgt->vd_jiffy = src->vd_jiffy;
25638 + n = src->vd_nblk;
25639 + for (ul = 1; ul < n; ul++) {
25640 + tgt->vd_deblk[ul] = kmemdup(src->vd_deblk[ul], deblk_sz,
25642 + if (unlikely(!tgt->vd_deblk[ul]))
25646 + tgt->vd_nblk = n;
25647 + tgt->vd_last.ul = tgt->vd_last.ul;
25648 + tgt->vd_last.p.deblk = tgt->vd_deblk[tgt->vd_last.ul];
25649 + tgt->vd_last.p.deblk += src->vd_last.p.deblk
25650 + - src->vd_deblk[src->vd_last.ul];
25652 + return 0; /* success */
25655 + rerr = reinit_vdir(tgt);
25660 +int au_vdir_init(struct file *file)
25663 + struct inode *inode;
25664 + struct au_vdir *vdir_cache, *allocated;
25666 + err = read_vdir(file, !file->f_pos);
25667 + if (unlikely(err))
25670 + allocated = NULL;
25671 + vdir_cache = au_fvdir_cache(file);
25672 + if (!vdir_cache) {
25673 + vdir_cache = alloc_vdir(file);
25674 + err = PTR_ERR(vdir_cache);
25675 + if (IS_ERR(vdir_cache))
25677 + allocated = vdir_cache;
25678 + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) {
25679 + err = reinit_vdir(vdir_cache);
25680 + if (unlikely(err))
25683 + return 0; /* success */
25685 + inode = file->f_dentry->d_inode;
25686 + err = copy_vdir(vdir_cache, au_ivdir(inode));
25688 + file->f_version = inode->i_version;
25690 + au_set_fvdir_cache(file, allocated);
25691 + } else if (allocated)
25692 + au_vdir_free(allocated);
25698 +static loff_t calc_offset(struct au_vdir *vdir)
25701 + union au_vdir_deblk_p p;
25703 + p.deblk = vdir->vd_deblk[vdir->vd_last.ul];
25704 + offset = vdir->vd_last.p.deblk - p.deblk;
25705 + offset += vdir->vd_deblk_sz * vdir->vd_last.ul;
25709 +/* returns true or false */
25710 +static int seek_vdir(struct file *file)
25713 + unsigned int deblk_sz;
25714 + unsigned long ul, n;
25716 + union au_vdir_deblk_p p, deblk_end;
25717 + struct au_vdir *vdir_cache;
25720 + vdir_cache = au_fvdir_cache(file);
25721 + offset = calc_offset(vdir_cache);
25722 + AuDbg("offset %lld\n", offset);
25723 + if (file->f_pos == offset)
25726 + vdir_cache->vd_last.ul = 0;
25727 + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0];
25728 + if (!file->f_pos)
25732 + deblk_sz = vdir_cache->vd_deblk_sz;
25733 + ul = div64_u64(file->f_pos, deblk_sz);
25734 + AuDbg("ul %lu\n", ul);
25735 + if (ul >= vdir_cache->vd_nblk)
25738 + n = vdir_cache->vd_nblk;
25739 + for (; ul < n; ul++) {
25740 + p.deblk = vdir_cache->vd_deblk[ul];
25741 + deblk_end.deblk = p.deblk + deblk_sz;
25743 + offset *= deblk_sz;
25744 + while (!is_deblk_end(&p, &deblk_end) && offset < file->f_pos) {
25747 + l = calc_size(p.de->de_str.len);
25751 + if (!is_deblk_end(&p, &deblk_end)) {
25753 + vdir_cache->vd_last.ul = ul;
25754 + vdir_cache->vd_last.p = p;
25761 + AuTraceErr(!valid);
25765 +int au_vdir_fill_de(struct file *file, void *dirent, filldir_t filldir)
25768 + unsigned int l, deblk_sz;
25769 + union au_vdir_deblk_p deblk_end;
25770 + struct au_vdir *vdir_cache;
25771 + struct au_vdir_de *de;
25773 + vdir_cache = au_fvdir_cache(file);
25774 + if (!seek_vdir(file))
25777 + deblk_sz = vdir_cache->vd_deblk_sz;
25779 + deblk_end.deblk = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
25780 + deblk_end.deblk += deblk_sz;
25781 + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) {
25782 + de = vdir_cache->vd_last.p.de;
25783 + AuDbg("%.*s, off%lld, i%lu, dt%d\n",
25784 + de->de_str.len, de->de_str.name, file->f_pos,
25785 + (unsigned long)de->de_ino, de->de_type);
25786 + err = filldir(dirent, de->de_str.name, de->de_str.len,
25787 + file->f_pos, de->de_ino, de->de_type);
25788 + if (unlikely(err)) {
25790 + /* todo: ignore the error caused by udba? */
25791 + /* return err; */
25795 + l = calc_size(de->de_str.len);
25796 + vdir_cache->vd_last.p.deblk += l;
25797 + file->f_pos += l;
25799 + if (vdir_cache->vd_last.ul < vdir_cache->vd_nblk - 1) {
25800 + vdir_cache->vd_last.ul++;
25801 + vdir_cache->vd_last.p.deblk
25802 + = vdir_cache->vd_deblk[vdir_cache->vd_last.ul];
25803 + file->f_pos = deblk_sz * vdir_cache->vd_last.ul;
25812 diff -urN /usr/share/empty/fs/aufs/vfsub.c linux/fs/aufs/vfsub.c
25813 --- /usr/share/empty/fs/aufs/vfsub.c 1970-01-01 01:00:00.000000000 +0100
25814 +++ linux/fs/aufs/vfsub.c 2012-01-05 11:00:59.218835185 +0100
25817 + * Copyright (C) 2005-2011 Junjiro R. Okajima
25819 + * This program, aufs is free software; you can redistribute it and/or modify
25820 + * it under the terms of the GNU General Public License as published by
25821 + * the Free Software Foundation; either version 2 of the License, or
25822 + * (at your option) any later version.
25824 + * This program is distributed in the hope that it will be useful,
25825 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
25826 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25827 + * GNU General Public License for more details.
25829 + * You should have received a copy of the GNU General Public License
25830 + * along with this program; if not, write to the Free Software
25831 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
25835 + * sub-routines for VFS
25838 +#include <linux/file.h>
25839 +#include <linux/ima.h>
25840 +#include <linux/namei.h>
25841 +#include <linux/security.h>
25842 +#include <linux/splice.h>
25843 +#include <linux/uaccess.h>
25846 +int vfsub_update_h_iattr(struct path *h_path, int *did)
25850 + struct super_block *h_sb;
25852 + /* for remote fs, leave work for its getattr or d_revalidate */
25853 + /* for bad i_attr fs, handle them in aufs_getattr() */
25854 + /* still some fs may acquire i_mutex. we need to skip them */
25858 + h_sb = h_path->dentry->d_sb;
25859 + *did = (!au_test_fs_remote(h_sb) && au_test_fs_refresh_iattr(h_sb));
25861 + err = vfs_getattr(h_path->mnt, h_path->dentry, &st);
25866 +/* ---------------------------------------------------------------------- */
25868 +struct file *vfsub_dentry_open(struct path *path, int flags)
25870 + struct file *file;
25873 + file = dentry_open(path->dentry, path->mnt,
25874 + flags /* | __FMODE_NONOTIFY */,
25876 + if (!IS_ERR_OR_NULL(file)
25877 + && (file->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
25878 + i_readcount_inc(path->dentry->d_inode);
25883 +struct file *vfsub_filp_open(const char *path, int oflags, int mode)
25885 + struct file *file;
25888 + file = filp_open(path,
25889 + oflags /* | __FMODE_NONOTIFY */,
25892 + if (IS_ERR(file))
25894 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
25900 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path)
25904 + err = kern_path(name, flags, path);
25905 + if (!err && path->dentry->d_inode)
25906 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
25910 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
25913 + struct path path = {
25917 + /* VFS checks it too, but by WARN_ON_ONCE() */
25918 + IMustLock(parent->d_inode);
25920 + path.dentry = lookup_one_len(name, parent, len);
25921 + if (IS_ERR(path.dentry))
25923 + if (path.dentry->d_inode)
25924 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
25927 + AuTraceErrPtr(path.dentry);
25928 + return path.dentry;
25931 +struct dentry *vfsub_lookup_hash(struct nameidata *nd)
25933 + struct path path = {
25934 + .mnt = nd->path.mnt
25937 + IMustLock(nd->path.dentry->d_inode);
25939 + path.dentry = lookup_hash(nd);
25940 + if (IS_ERR(path.dentry))
25942 + if (path.dentry->d_inode)
25943 + vfsub_update_h_iattr(&path, /*did*/NULL); /*ignore*/
25946 + AuTraceErrPtr(path.dentry);
25947 + return path.dentry;
25951 + * this is "VFS:__lookup_one_len()" which was removed and merged into
25952 + * VFS:lookup_one_len() by the commit.
25953 + * 6a96ba5 2011-03-14 kill __lookup_one_len()
25954 + * this function should always be equivalent to the corresponding part in
25955 + * VFS:lookup_one_len().
25957 +int vfsub_name_hash(const char *name, struct qstr *this, int len)
25959 + unsigned long hash;
25962 + this->name = name;
25967 + hash = init_name_hash();
25969 + c = *(const unsigned char *)name++;
25970 + if (c == '/' || c == '\0')
25972 + hash = partial_name_hash(c, hash);
25974 + this->hash = end_name_hash(hash);
25978 +/* ---------------------------------------------------------------------- */
25980 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
25981 + struct dentry *d2, struct au_hinode *hdir2)
25983 + struct dentry *d;
25986 + d = lock_rename(d1, d2);
25988 + au_hn_suspend(hdir1);
25989 + if (hdir1 != hdir2)
25990 + au_hn_suspend(hdir2);
25995 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
25996 + struct dentry *d2, struct au_hinode *hdir2)
25998 + au_hn_resume(hdir1);
25999 + if (hdir1 != hdir2)
26000 + au_hn_resume(hdir2);
26002 + unlock_rename(d1, d2);
26006 +/* ---------------------------------------------------------------------- */
26008 +int vfsub_create(struct inode *dir, struct path *path, int mode)
26011 + struct dentry *d;
26015 + d = path->dentry;
26016 + path->dentry = d->d_parent;
26017 + err = security_path_mknod(path, d, mode, 0);
26018 + path->dentry = d;
26019 + if (unlikely(err))
26022 + if (au_test_fs_null_nd(dir->i_sb))
26023 + err = vfs_create(dir, path->dentry, mode, NULL);
26025 + struct nameidata h_nd;
26027 + memset(&h_nd, 0, sizeof(h_nd));
26028 + h_nd.flags = LOOKUP_CREATE;
26029 + h_nd.intent.open.flags = O_CREAT
26030 + | vfsub_fmode_to_uint(FMODE_READ);
26031 + h_nd.intent.open.create_mode = mode;
26032 + h_nd.path.dentry = path->dentry->d_parent;
26033 + h_nd.path.mnt = path->mnt;
26034 + path_get(&h_nd.path);
26035 + err = vfs_create(dir, path->dentry, mode, &h_nd);
26036 + path_put(&h_nd.path);
26040 + struct path tmp = *path;
26043 + vfsub_update_h_iattr(&tmp, &did);
26045 + tmp.dentry = path->dentry->d_parent;
26046 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26055 +int vfsub_symlink(struct inode *dir, struct path *path, const char *symname)
26058 + struct dentry *d;
26062 + d = path->dentry;
26063 + path->dentry = d->d_parent;
26064 + err = security_path_symlink(path, d, symname);
26065 + path->dentry = d;
26066 + if (unlikely(err))
26069 + err = vfs_symlink(dir, path->dentry, symname);
26071 + struct path tmp = *path;
26074 + vfsub_update_h_iattr(&tmp, &did);
26076 + tmp.dentry = path->dentry->d_parent;
26077 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26086 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev)
26089 + struct dentry *d;
26093 + d = path->dentry;
26094 + path->dentry = d->d_parent;
26095 + err = security_path_mknod(path, d, mode, new_encode_dev(dev));
26096 + path->dentry = d;
26097 + if (unlikely(err))
26100 + err = vfs_mknod(dir, path->dentry, mode, dev);
26102 + struct path tmp = *path;
26105 + vfsub_update_h_iattr(&tmp, &did);
26107 + tmp.dentry = path->dentry->d_parent;
26108 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26117 +static int au_test_nlink(struct inode *inode)
26119 + const unsigned int link_max = UINT_MAX >> 1; /* rough margin */
26121 + if (!au_test_fs_no_limit_nlink(inode->i_sb)
26122 + || inode->i_nlink < link_max)
26127 +int vfsub_link(struct dentry *src_dentry, struct inode *dir, struct path *path)
26130 + struct dentry *d;
26134 + err = au_test_nlink(src_dentry->d_inode);
26135 + if (unlikely(err))
26138 + d = path->dentry;
26139 + path->dentry = d->d_parent;
26140 + err = security_path_link(src_dentry, path, d);
26141 + path->dentry = d;
26142 + if (unlikely(err))
26146 + err = vfs_link(src_dentry, dir, path->dentry);
26149 + struct path tmp = *path;
26152 + /* fuse has different memory inode for the same inumber */
26153 + vfsub_update_h_iattr(&tmp, &did);
26155 + tmp.dentry = path->dentry->d_parent;
26156 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26157 + tmp.dentry = src_dentry;
26158 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26167 +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry,
26168 + struct inode *dir, struct path *path)
26171 + struct path tmp = {
26174 + struct dentry *d;
26177 + IMustLock(src_dir);
26179 + d = path->dentry;
26180 + path->dentry = d->d_parent;
26181 + tmp.dentry = src_dentry->d_parent;
26182 + err = security_path_rename(&tmp, src_dentry, path, d);
26183 + path->dentry = d;
26184 + if (unlikely(err))
26188 + err = vfs_rename(src_dir, src_dentry, dir, path->dentry);
26193 + tmp.dentry = d->d_parent;
26194 + vfsub_update_h_iattr(&tmp, &did);
26196 + tmp.dentry = src_dentry;
26197 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26198 + tmp.dentry = src_dentry->d_parent;
26199 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26208 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode)
26211 + struct dentry *d;
26215 + d = path->dentry;
26216 + path->dentry = d->d_parent;
26217 + err = security_path_mkdir(path, d, mode);
26218 + path->dentry = d;
26219 + if (unlikely(err))
26222 + err = vfs_mkdir(dir, path->dentry, mode);
26224 + struct path tmp = *path;
26227 + vfsub_update_h_iattr(&tmp, &did);
26229 + tmp.dentry = path->dentry->d_parent;
26230 + vfsub_update_h_iattr(&tmp, /*did*/NULL);
26239 +int vfsub_rmdir(struct inode *dir, struct path *path)
26242 + struct dentry *d;
26246 + d = path->dentry;
26247 + path->dentry = d->d_parent;
26248 + err = security_path_rmdir(path, d);
26249 + path->dentry = d;
26250 + if (unlikely(err))
26254 + err = vfs_rmdir(dir, path->dentry);
26257 + struct path tmp = {
26258 + .dentry = path->dentry->d_parent,
26262 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
26269 +/* ---------------------------------------------------------------------- */
26271 +/* todo: support mmap_sem? */
26272 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
26278 + err = vfs_read(file, ubuf, count, ppos);
26281 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
26285 +/* todo: kernel_read()? */
26286 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
26290 + mm_segment_t oldfs;
26297 + oldfs = get_fs();
26298 + set_fs(KERNEL_DS);
26299 + err = vfsub_read_u(file, buf.u, count, ppos);
26304 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
26310 + err = vfs_write(file, ubuf, count, ppos);
26313 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
26317 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos)
26320 + mm_segment_t oldfs;
26323 + const char __user *u;
26327 + oldfs = get_fs();
26328 + set_fs(KERNEL_DS);
26329 + err = vfsub_write_u(file, buf.u, count, ppos);
26334 +int vfsub_flush(struct file *file, fl_owner_t id)
26339 + if (file->f_op && file->f_op->flush) {
26340 + if (!au_test_nfs(file->f_dentry->d_sb))
26341 + err = file->f_op->flush(file, id);
26344 + err = file->f_op->flush(file, id);
26348 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL);
26354 +int vfsub_readdir(struct file *file, filldir_t filldir, void *arg)
26359 + err = vfs_readdir(file, filldir, arg);
26362 + vfsub_update_h_iattr(&file->f_path, /*did*/NULL); /*ignore*/
26366 +long vfsub_splice_to(struct file *in, loff_t *ppos,
26367 + struct pipe_inode_info *pipe, size_t len,
26368 + unsigned int flags)
26373 + err = do_splice_to(in, ppos, pipe, len, flags);
26375 + file_accessed(in);
26377 + vfsub_update_h_iattr(&in->f_path, /*did*/NULL); /*ignore*/
26381 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
26382 + loff_t *ppos, size_t len, unsigned int flags)
26387 + err = do_splice_from(pipe, out, ppos, len, flags);
26390 + vfsub_update_h_iattr(&out->f_path, /*did*/NULL); /*ignore*/
26394 +int vfsub_fsync(struct file *file, struct path *path, int datasync)
26398 + /* file can be NULL */
26400 + err = vfs_fsync(file, datasync);
26404 + AuDebugOn(!file);
26405 + path = &file->f_path;
26407 + vfsub_update_h_iattr(path, /*did*/NULL); /*ignore*/
26412 +/* cf. open.c:do_sys_truncate() and do_sys_ftruncate() */
26413 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
26414 + struct file *h_file)
26417 + struct inode *h_inode;
26419 + h_inode = h_path->dentry->d_inode;
26421 + err = mnt_want_write(h_path->mnt);
26424 + err = inode_permission(h_inode, MAY_WRITE);
26427 + err = get_write_access(h_inode);
26430 + err = break_lease(h_inode, O_WRONLY);
26435 + err = locks_verify_truncate(h_inode, h_file, length);
26437 + err = security_path_truncate(h_path);
26440 + err = do_truncate(h_path->dentry, length, attr, h_file);
26446 + put_write_access(h_inode);
26449 + mnt_drop_write(h_path->mnt);
26454 +/* ---------------------------------------------------------------------- */
26456 +struct au_vfsub_mkdir_args {
26458 + struct inode *dir;
26459 + struct path *path;
26463 +static void au_call_vfsub_mkdir(void *args)
26465 + struct au_vfsub_mkdir_args *a = args;
26466 + *a->errp = vfsub_mkdir(a->dir, a->path, a->mode);
26469 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode)
26471 + int err, do_sio, wkq_err;
26473 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
26475 + err = vfsub_mkdir(dir, path, mode);
26477 + struct au_vfsub_mkdir_args args = {
26483 + wkq_err = au_wkq_wait(au_call_vfsub_mkdir, &args);
26484 + if (unlikely(wkq_err))
26491 +struct au_vfsub_rmdir_args {
26493 + struct inode *dir;
26494 + struct path *path;
26497 +static void au_call_vfsub_rmdir(void *args)
26499 + struct au_vfsub_rmdir_args *a = args;
26500 + *a->errp = vfsub_rmdir(a->dir, a->path);
26503 +int vfsub_sio_rmdir(struct inode *dir, struct path *path)
26505 + int err, do_sio, wkq_err;
26507 + do_sio = au_test_h_perm_sio(dir, MAY_EXEC | MAY_WRITE);
26509 + err = vfsub_rmdir(dir, path);
26511 + struct au_vfsub_rmdir_args args = {
26516 + wkq_err = au_wkq_wait(au_call_vfsub_rmdir, &args);
26517 + if (unlikely(wkq_err))
26524 +/* ---------------------------------------------------------------------- */
26526 +struct notify_change_args {
26528 + struct path *path;
26529 + struct iattr *ia;
26532 +static void call_notify_change(void *args)
26534 + struct notify_change_args *a = args;
26535 + struct inode *h_inode;
26537 + h_inode = a->path->dentry->d_inode;
26538 + IMustLock(h_inode);
26540 + *a->errp = -EPERM;
26541 + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) {
26542 + *a->errp = notify_change(a->path->dentry, a->ia);
26544 + vfsub_update_h_iattr(a->path, /*did*/NULL); /*ignore*/
26546 + AuTraceErr(*a->errp);
26549 +int vfsub_notify_change(struct path *path, struct iattr *ia)
26552 + struct notify_change_args args = {
26558 + call_notify_change(&args);
26563 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia)
26565 + int err, wkq_err;
26566 + struct notify_change_args args = {
26572 + wkq_err = au_wkq_wait(call_notify_change, &args);
26573 + if (unlikely(wkq_err))
26579 +/* ---------------------------------------------------------------------- */
26581 +struct unlink_args {
26583 + struct inode *dir;
26584 + struct path *path;
26587 +static void call_unlink(void *args)
26589 + struct unlink_args *a = args;
26590 + struct dentry *d = a->path->dentry;
26591 + struct inode *h_inode;
26592 + const int stop_sillyrename = (au_test_nfs(d->d_sb)
26593 + && d->d_count == 1);
26595 + IMustLock(a->dir);
26597 + a->path->dentry = d->d_parent;
26598 + *a->errp = security_path_unlink(a->path, d);
26599 + a->path->dentry = d;
26600 + if (unlikely(*a->errp))
26603 + if (!stop_sillyrename)
26605 + h_inode = d->d_inode;
26610 + *a->errp = vfs_unlink(a->dir, d);
26613 + struct path tmp = {
26614 + .dentry = d->d_parent,
26615 + .mnt = a->path->mnt
26617 + vfsub_update_h_iattr(&tmp, /*did*/NULL); /*ignore*/
26620 + if (!stop_sillyrename)
26625 + AuTraceErr(*a->errp);
26629 + * @dir: must be locked.
26630 + * @dentry: target dentry.
26632 +int vfsub_unlink(struct inode *dir, struct path *path, int force)
26635 + struct unlink_args args = {
26642 + call_unlink(&args);
26646 + wkq_err = au_wkq_wait(call_unlink, &args);
26647 + if (unlikely(wkq_err))
26653 diff -urN /usr/share/empty/fs/aufs/vfsub.h linux/fs/aufs/vfsub.h
26654 --- /usr/share/empty/fs/aufs/vfsub.h 1970-01-01 01:00:00.000000000 +0100
26655 +++ linux/fs/aufs/vfsub.h 2011-08-24 13:30:24.734646739 +0200
26658 + * Copyright (C) 2005-2011 Junjiro R. Okajima
26660 + * This program, aufs is free software; you can redistribute it and/or modify
26661 + * it under the terms of the GNU General Public License as published by
26662 + * the Free Software Foundation; either version 2 of the License, or
26663 + * (at your option) any later version.
26665 + * This program is distributed in the hope that it will be useful,
26666 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26667 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26668 + * GNU General Public License for more details.
26670 + * You should have received a copy of the GNU General Public License
26671 + * along with this program; if not, write to the Free Software
26672 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
26676 + * sub-routines for VFS
26679 +#ifndef __AUFS_VFSUB_H__
26680 +#define __AUFS_VFSUB_H__
26684 +#include <linux/fs.h>
26685 +#include <linux/lglock.h>
26686 +#include "debug.h"
26688 +/* copied from linux/fs/internal.h */
26689 +/* todo: BAD approach!! */
26690 +DECLARE_BRLOCK(vfsmount_lock);
26691 +extern void file_sb_list_del(struct file *f);
26692 +extern spinlock_t inode_sb_list_lock;
26694 +/* copied from linux/fs/file_table.c */
26695 +DECLARE_LGLOCK(files_lglock);
26698 + * These macros iterate all files on all CPUs for a given superblock.
26699 + * files_lglock must be held globally.
26701 +#define do_file_list_for_each_entry(__sb, __file) \
26704 + for_each_possible_cpu(i) { \
26705 + struct list_head *list; \
26706 + list = per_cpu_ptr((__sb)->s_files, i); \
26707 + list_for_each_entry((__file), list, f_u.fu_list)
26709 +#define while_file_list_for_each_entry \
26715 +#define do_file_list_for_each_entry(__sb, __file) \
26717 + struct list_head *list; \
26718 + list = &(sb)->s_files; \
26719 + list_for_each_entry((__file), list, f_u.fu_list)
26721 +#define while_file_list_for_each_entry \
26725 +/* ---------------------------------------------------------------------- */
26727 +/* lock subclass for lower inode */
26728 +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */
26729 +/* reduce? gave up. */
26731 + AuLsc_I_Begin = I_MUTEX_QUOTA, /* 4 */
26732 + AuLsc_I_PARENT, /* lower inode, parent first */
26733 + AuLsc_I_PARENT2, /* copyup dirs */
26734 + AuLsc_I_PARENT3, /* copyup wh */
26740 +/* to debug easier, do not make them inlined functions */
26741 +#define MtxMustLock(mtx) AuDebugOn(!mutex_is_locked(mtx))
26742 +#define IMustLock(i) MtxMustLock(&(i)->i_mutex)
26744 +/* ---------------------------------------------------------------------- */
26746 +static inline void vfsub_drop_nlink(struct inode *inode)
26748 + AuDebugOn(!inode->i_nlink);
26749 + drop_nlink(inode);
26752 +static inline void vfsub_dead_dir(struct inode *inode)
26754 + AuDebugOn(!S_ISDIR(inode->i_mode));
26755 + inode->i_flags |= S_DEAD;
26756 + clear_nlink(inode);
26759 +/* ---------------------------------------------------------------------- */
26761 +int vfsub_update_h_iattr(struct path *h_path, int *did);
26762 +struct file *vfsub_dentry_open(struct path *path, int flags);
26763 +struct file *vfsub_filp_open(const char *path, int oflags, int mode);
26764 +int vfsub_kern_path(const char *name, unsigned int flags, struct path *path);
26765 +struct dentry *vfsub_lookup_one_len(const char *name, struct dentry *parent,
26767 +struct dentry *vfsub_lookup_hash(struct nameidata *nd);
26768 +int vfsub_name_hash(const char *name, struct qstr *this, int len);
26770 +/* ---------------------------------------------------------------------- */
26773 +struct dentry *vfsub_lock_rename(struct dentry *d1, struct au_hinode *hdir1,
26774 + struct dentry *d2, struct au_hinode *hdir2);
26775 +void vfsub_unlock_rename(struct dentry *d1, struct au_hinode *hdir1,
26776 + struct dentry *d2, struct au_hinode *hdir2);
26778 +int vfsub_create(struct inode *dir, struct path *path, int mode);
26779 +int vfsub_symlink(struct inode *dir, struct path *path,
26780 + const char *symname);
26781 +int vfsub_mknod(struct inode *dir, struct path *path, int mode, dev_t dev);
26782 +int vfsub_link(struct dentry *src_dentry, struct inode *dir,
26783 + struct path *path);
26784 +int vfsub_rename(struct inode *src_hdir, struct dentry *src_dentry,
26785 + struct inode *hdir, struct path *path);
26786 +int vfsub_mkdir(struct inode *dir, struct path *path, int mode);
26787 +int vfsub_rmdir(struct inode *dir, struct path *path);
26789 +/* ---------------------------------------------------------------------- */
26791 +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count,
26793 +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count,
26795 +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count,
26797 +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count,
26799 +int vfsub_flush(struct file *file, fl_owner_t id);
26800 +int vfsub_readdir(struct file *file, filldir_t filldir, void *arg);
26802 +static inline unsigned int vfsub_file_flags(struct file *file)
26804 + unsigned int flags;
26806 + spin_lock(&file->f_lock);
26807 + flags = file->f_flags;
26808 + spin_unlock(&file->f_lock);
26813 +static inline void vfsub_file_accessed(struct file *h_file)
26815 + file_accessed(h_file);
26816 + vfsub_update_h_iattr(&h_file->f_path, /*did*/NULL); /*ignore*/
26819 +static inline void vfsub_touch_atime(struct vfsmount *h_mnt,
26820 + struct dentry *h_dentry)
26822 + struct path h_path = {
26823 + .dentry = h_dentry,
26826 + touch_atime(h_mnt, h_dentry);
26827 + vfsub_update_h_iattr(&h_path, /*did*/NULL); /*ignore*/
26830 +long vfsub_splice_to(struct file *in, loff_t *ppos,
26831 + struct pipe_inode_info *pipe, size_t len,
26832 + unsigned int flags);
26833 +long vfsub_splice_from(struct pipe_inode_info *pipe, struct file *out,
26834 + loff_t *ppos, size_t len, unsigned int flags);
26835 +int vfsub_trunc(struct path *h_path, loff_t length, unsigned int attr,
26836 + struct file *h_file);
26837 +int vfsub_fsync(struct file *file, struct path *path, int datasync);
26839 +/* ---------------------------------------------------------------------- */
26841 +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin)
26846 + err = vfs_llseek(file, offset, origin);
26851 +/* ---------------------------------------------------------------------- */
26853 +/* dirty workaround for strict type of fmode_t */
26859 +static inline unsigned int vfsub_fmode_to_uint(fmode_t fm)
26861 + union vfsub_fmu u = {
26865 + BUILD_BUG_ON(sizeof(u.fm) != sizeof(u.ui));
26870 +static inline fmode_t vfsub_uint_to_fmode(unsigned int ui)
26872 + union vfsub_fmu u = {
26879 +/* ---------------------------------------------------------------------- */
26881 +int vfsub_sio_mkdir(struct inode *dir, struct path *path, int mode);
26882 +int vfsub_sio_rmdir(struct inode *dir, struct path *path);
26883 +int vfsub_sio_notify_change(struct path *path, struct iattr *ia);
26884 +int vfsub_notify_change(struct path *path, struct iattr *ia);
26885 +int vfsub_unlink(struct inode *dir, struct path *path, int force);
26887 +#endif /* __KERNEL__ */
26888 +#endif /* __AUFS_VFSUB_H__ */
26889 diff -urN /usr/share/empty/fs/aufs/wbr_policy.c linux/fs/aufs/wbr_policy.c
26890 --- /usr/share/empty/fs/aufs/wbr_policy.c 1970-01-01 01:00:00.000000000 +0100
26891 +++ linux/fs/aufs/wbr_policy.c 2011-08-24 13:30:24.734646739 +0200
26894 + * Copyright (C) 2005-2011 Junjiro R. Okajima
26896 + * This program, aufs is free software; you can redistribute it and/or modify
26897 + * it under the terms of the GNU General Public License as published by
26898 + * the Free Software Foundation; either version 2 of the License, or
26899 + * (at your option) any later version.
26901 + * This program is distributed in the hope that it will be useful,
26902 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
26903 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26904 + * GNU General Public License for more details.
26906 + * You should have received a copy of the GNU General Public License
26907 + * along with this program; if not, write to the Free Software
26908 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
26912 + * policies for selecting one among multiple writable branches
26915 +#include <linux/statfs.h>
26918 +/* subset of cpup_attr() */
26919 +static noinline_for_stack
26920 +int au_cpdown_attr(struct path *h_path, struct dentry *h_src)
26924 + struct inode *h_isrc;
26926 + h_isrc = h_src->d_inode;
26927 + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID;
26928 + ia.ia_mode = h_isrc->i_mode;
26929 + ia.ia_uid = h_isrc->i_uid;
26930 + ia.ia_gid = h_isrc->i_gid;
26931 + sbits = !!(ia.ia_mode & (S_ISUID | S_ISGID));
26932 + au_cpup_attr_flags(h_path->dentry->d_inode, h_isrc);
26933 + err = vfsub_sio_notify_change(h_path, &ia);
26935 + /* is this nfs only? */
26936 + if (!err && sbits && au_test_nfs(h_path->dentry->d_sb)) {
26937 + ia.ia_valid = ATTR_FORCE | ATTR_MODE;
26938 + ia.ia_mode = h_isrc->i_mode;
26939 + err = vfsub_sio_notify_change(h_path, &ia);
26945 +#define AuCpdown_PARENT_OPQ 1
26946 +#define AuCpdown_WHED (1 << 1)
26947 +#define AuCpdown_MADE_DIR (1 << 2)
26948 +#define AuCpdown_DIROPQ (1 << 3)
26949 +#define au_ftest_cpdown(flags, name) ((flags) & AuCpdown_##name)
26950 +#define au_fset_cpdown(flags, name) \
26951 + do { (flags) |= AuCpdown_##name; } while (0)
26952 +#define au_fclr_cpdown(flags, name) \
26953 + do { (flags) &= ~AuCpdown_##name; } while (0)
26955 +struct au_cpdown_dir_args {
26956 + struct dentry *parent;
26957 + unsigned int flags;
26960 +static int au_cpdown_dir_opq(struct dentry *dentry, aufs_bindex_t bdst,
26961 + struct au_cpdown_dir_args *a)
26964 + struct dentry *opq_dentry;
26966 + opq_dentry = au_diropq_create(dentry, bdst);
26967 + err = PTR_ERR(opq_dentry);
26968 + if (IS_ERR(opq_dentry))
26970 + dput(opq_dentry);
26971 + au_fset_cpdown(a->flags, DIROPQ);
26977 +static int au_cpdown_dir_wh(struct dentry *dentry, struct dentry *h_parent,
26978 + struct inode *dir, aufs_bindex_t bdst)
26981 + struct path h_path;
26982 + struct au_branch *br;
26984 + br = au_sbr(dentry->d_sb, bdst);
26985 + h_path.dentry = au_wh_lkup(h_parent, &dentry->d_name, br);
26986 + err = PTR_ERR(h_path.dentry);
26987 + if (IS_ERR(h_path.dentry))
26991 + if (h_path.dentry->d_inode) {
26992 + h_path.mnt = br->br_mnt;
26993 + err = au_wh_unlink_dentry(au_h_iptr(dir, bdst), &h_path,
26996 + dput(h_path.dentry);
27002 +static int au_cpdown_dir(struct dentry *dentry, aufs_bindex_t bdst,
27003 + struct dentry *h_parent, void *arg)
27006 + aufs_bindex_t bopq, bstart;
27007 + struct path h_path;
27008 + struct dentry *parent;
27009 + struct inode *h_dir, *h_inode, *inode, *dir;
27010 + struct au_cpdown_dir_args *args = arg;
27012 + bstart = au_dbstart(dentry);
27013 + /* dentry is di-locked */
27014 + parent = dget_parent(dentry);
27015 + dir = parent->d_inode;
27016 + h_dir = h_parent->d_inode;
27017 + AuDebugOn(h_dir != au_h_iptr(dir, bdst));
27018 + IMustLock(h_dir);
27020 + err = au_lkup_neg(dentry, bdst);
27021 + if (unlikely(err < 0))
27023 + h_path.dentry = au_h_dptr(dentry, bdst);
27024 + h_path.mnt = au_sbr_mnt(dentry->d_sb, bdst);
27025 + err = vfsub_sio_mkdir(au_h_iptr(dir, bdst), &h_path,
27026 + S_IRWXU | S_IRUGO | S_IXUGO);
27027 + if (unlikely(err))
27029 + au_fset_cpdown(args->flags, MADE_DIR);
27031 + bopq = au_dbdiropq(dentry);
27032 + au_fclr_cpdown(args->flags, WHED);
27033 + au_fclr_cpdown(args->flags, DIROPQ);
27034 + if (au_dbwh(dentry) == bdst)
27035 + au_fset_cpdown(args->flags, WHED);
27036 + if (!au_ftest_cpdown(args->flags, PARENT_OPQ) && bopq <= bdst)
27037 + au_fset_cpdown(args->flags, PARENT_OPQ);
27038 + h_inode = h_path.dentry->d_inode;
27039 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
27040 + if (au_ftest_cpdown(args->flags, WHED)) {
27041 + err = au_cpdown_dir_opq(dentry, bdst, args);
27042 + if (unlikely(err)) {
27043 + mutex_unlock(&h_inode->i_mutex);
27048 + err = au_cpdown_attr(&h_path, au_h_dptr(dentry, bstart));
27049 + mutex_unlock(&h_inode->i_mutex);
27050 + if (unlikely(err))
27053 + if (au_ftest_cpdown(args->flags, WHED)) {
27054 + err = au_cpdown_dir_wh(dentry, h_parent, dir, bdst);
27055 + if (unlikely(err))
27059 + inode = dentry->d_inode;
27060 + if (au_ibend(inode) < bdst)
27061 + au_set_ibend(inode, bdst);
27062 + au_set_h_iptr(inode, bdst, au_igrab(h_inode),
27063 + au_hi_flags(inode, /*isdir*/1));
27064 + goto out; /* success */
27068 + if (au_ftest_cpdown(args->flags, DIROPQ)) {
27069 + mutex_lock_nested(&h_inode->i_mutex, AuLsc_I_CHILD);
27070 + rerr = au_diropq_remove(dentry, bdst);
27071 + mutex_unlock(&h_inode->i_mutex);
27072 + if (unlikely(rerr)) {
27073 + AuIOErr("failed removing diropq for %.*s b%d (%d)\n",
27074 + AuDLNPair(dentry), bdst, rerr);
27080 + if (au_ftest_cpdown(args->flags, MADE_DIR)) {
27081 + rerr = vfsub_sio_rmdir(au_h_iptr(dir, bdst), &h_path);
27082 + if (unlikely(rerr)) {
27083 + AuIOErr("failed removing %.*s b%d (%d)\n",
27084 + AuDLNPair(dentry), bdst, rerr);
27089 + au_set_h_dptr(dentry, bdst, NULL);
27090 + if (au_dbend(dentry) == bdst)
27091 + au_update_dbend(dentry);
27097 +int au_cpdown_dirs(struct dentry *dentry, aufs_bindex_t bdst)
27100 + struct au_cpdown_dir_args args = {
27101 + .parent = dget_parent(dentry),
27105 + err = au_cp_dirs(dentry, bdst, au_cpdown_dir, &args);
27106 + dput(args.parent);
27111 +/* ---------------------------------------------------------------------- */
27113 +/* policies for create */
27115 +static int au_wbr_nonopq(struct dentry *dentry, aufs_bindex_t bindex)
27117 + int err, i, j, ndentry;
27118 + aufs_bindex_t bopq;
27119 + struct au_dcsub_pages dpages;
27120 + struct au_dpage *dpage;
27121 + struct dentry **dentries, *parent, *d;
27123 + err = au_dpages_init(&dpages, GFP_NOFS);
27124 + if (unlikely(err))
27126 + parent = dget_parent(dentry);
27127 + err = au_dcsub_pages_rev_aufs(&dpages, parent, /*do_include*/0);
27128 + if (unlikely(err))
27132 + for (i = 0; i < dpages.ndpage; i++) {
27133 + dpage = dpages.dpages + i;
27134 + dentries = dpage->dentries;
27135 + ndentry = dpage->ndentry;
27136 + for (j = 0; j < ndentry; j++) {
27138 + di_read_lock_parent2(d, !AuLock_IR);
27139 + bopq = au_dbdiropq(d);
27140 + di_read_unlock(d, !AuLock_IR);
27141 + if (bopq >= 0 && bopq < err)
27148 + au_dpages_free(&dpages);
27153 +static int au_wbr_bu(struct super_block *sb, aufs_bindex_t bindex)
27155 + for (; bindex >= 0; bindex--)
27156 + if (!au_br_rdonly(au_sbr(sb, bindex)))
27161 +/* top down parent */
27162 +static int au_wbr_create_tdp(struct dentry *dentry, int isdir __maybe_unused)
27165 + aufs_bindex_t bstart, bindex;
27166 + struct super_block *sb;
27167 + struct dentry *parent, *h_parent;
27169 + sb = dentry->d_sb;
27170 + bstart = au_dbstart(dentry);
27172 + if (!au_br_rdonly(au_sbr(sb, bstart)))
27176 + parent = dget_parent(dentry);
27177 + for (bindex = au_dbstart(parent); bindex < bstart; bindex++) {
27178 + h_parent = au_h_dptr(parent, bindex);
27179 + if (!h_parent || !h_parent->d_inode)
27182 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
27189 + /* bottom up here */
27190 + if (unlikely(err < 0)) {
27191 + err = au_wbr_bu(sb, bstart - 1);
27193 + err = au_wbr_nonopq(dentry, err);
27197 + AuDbg("b%d\n", err);
27201 +/* ---------------------------------------------------------------------- */
27203 +/* an exception for the policy other than tdp */
27204 +static int au_wbr_create_exp(struct dentry *dentry)
27207 + aufs_bindex_t bwh, bdiropq;
27208 + struct dentry *parent;
27211 + bwh = au_dbwh(dentry);
27212 + parent = dget_parent(dentry);
27213 + bdiropq = au_dbdiropq(parent);
27215 + if (bdiropq >= 0)
27216 + err = min(bdiropq, bwh);
27219 + AuDbg("%d\n", err);
27220 + } else if (bdiropq >= 0) {
27222 + AuDbg("%d\n", err);
27227 + err = au_wbr_nonopq(dentry, err);
27229 + if (err >= 0 && au_br_rdonly(au_sbr(dentry->d_sb, err)))
27232 + AuDbg("%d\n", err);
27236 +/* ---------------------------------------------------------------------- */
27239 +static int au_wbr_create_init_rr(struct super_block *sb)
27243 + err = au_wbr_bu(sb, au_sbend(sb));
27244 + atomic_set(&au_sbi(sb)->si_wbr_rr_next, -err); /* less important */
27247 + AuDbg("b%d\n", err);
27251 +static int au_wbr_create_rr(struct dentry *dentry, int isdir)
27255 + aufs_bindex_t bindex, bend;
27256 + struct super_block *sb;
27259 + err = au_wbr_create_exp(dentry);
27263 + sb = dentry->d_sb;
27264 + next = &au_sbi(sb)->si_wbr_rr_next;
27265 + bend = au_sbend(sb);
27267 + for (bindex = 0; bindex <= bend; bindex++) {
27269 + err = atomic_dec_return(next) + 1;
27270 + /* modulo for 0 is meaningless */
27271 + if (unlikely(!err))
27272 + err = atomic_dec_return(next) + 1;
27274 + err = atomic_read(next);
27275 + AuDbg("%d\n", err);
27278 + AuDbg("%d\n", err);
27279 + if (!au_br_rdonly(au_sbr(sb, err)))
27285 + err = au_wbr_nonopq(dentry, err);
27288 + AuDbg("%d\n", err);
27292 +/* ---------------------------------------------------------------------- */
27294 +/* most free space */
27295 +static void au_mfs(struct dentry *dentry)
27297 + struct super_block *sb;
27298 + struct au_branch *br;
27299 + struct au_wbr_mfs *mfs;
27300 + aufs_bindex_t bindex, bend;
27302 + unsigned long long b, bavail;
27303 + struct path h_path;
27304 + /* reduce the stack usage */
27305 + struct kstatfs *st;
27307 + st = kmalloc(sizeof(*st), GFP_NOFS);
27308 + if (unlikely(!st)) {
27309 + AuWarn1("failed updating mfs(%d), ignored\n", -ENOMEM);
27314 + sb = dentry->d_sb;
27315 + mfs = &au_sbi(sb)->si_wbr_mfs;
27316 + MtxMustLock(&mfs->mfs_lock);
27317 + mfs->mfs_bindex = -EROFS;
27318 + mfs->mfsrr_bytes = 0;
27319 + bend = au_sbend(sb);
27320 + for (bindex = 0; bindex <= bend; bindex++) {
27321 + br = au_sbr(sb, bindex);
27322 + if (au_br_rdonly(br))
27325 + /* sb->s_root for NFS is unreliable */
27326 + h_path.mnt = br->br_mnt;
27327 + h_path.dentry = h_path.mnt->mnt_root;
27328 + err = vfs_statfs(&h_path, st);
27329 + if (unlikely(err)) {
27330 + AuWarn1("failed statfs, b%d, %d\n", bindex, err);
27334 + /* when the available size is equal, select the lower one */
27335 + BUILD_BUG_ON(sizeof(b) < sizeof(st->f_bavail)
27336 + || sizeof(b) < sizeof(st->f_bsize));
27337 + b = st->f_bavail * st->f_bsize;
27338 + br->br_wbr->wbr_bytes = b;
27339 + if (b >= bavail) {
27341 + mfs->mfs_bindex = bindex;
27342 + mfs->mfs_jiffy = jiffies;
27346 + mfs->mfsrr_bytes = bavail;
27347 + AuDbg("b%d\n", mfs->mfs_bindex);
27351 +static int au_wbr_create_mfs(struct dentry *dentry, int isdir __maybe_unused)
27354 + struct super_block *sb;
27355 + struct au_wbr_mfs *mfs;
27357 + err = au_wbr_create_exp(dentry);
27361 + sb = dentry->d_sb;
27362 + mfs = &au_sbi(sb)->si_wbr_mfs;
27363 + mutex_lock(&mfs->mfs_lock);
27364 + if (time_after(jiffies, mfs->mfs_jiffy + mfs->mfs_expire)
27365 + || mfs->mfs_bindex < 0
27366 + || au_br_rdonly(au_sbr(sb, mfs->mfs_bindex)))
27368 + mutex_unlock(&mfs->mfs_lock);
27369 + err = mfs->mfs_bindex;
27372 + err = au_wbr_nonopq(dentry, err);
27375 + AuDbg("b%d\n", err);
27379 +static int au_wbr_create_init_mfs(struct super_block *sb)
27381 + struct au_wbr_mfs *mfs;
27383 + mfs = &au_sbi(sb)->si_wbr_mfs;
27384 + mutex_init(&mfs->mfs_lock);
27385 + mfs->mfs_jiffy = 0;
27386 + mfs->mfs_bindex = -EROFS;
27391 +static int au_wbr_create_fin_mfs(struct super_block *sb __maybe_unused)
27393 + mutex_destroy(&au_sbi(sb)->si_wbr_mfs.mfs_lock);
27397 +/* ---------------------------------------------------------------------- */
27399 +/* most free space and then round robin */
27400 +static int au_wbr_create_mfsrr(struct dentry *dentry, int isdir)
27403 + struct au_wbr_mfs *mfs;
27405 + err = au_wbr_create_mfs(dentry, isdir);
27407 + mfs = &au_sbi(dentry->d_sb)->si_wbr_mfs;
27408 + mutex_lock(&mfs->mfs_lock);
27409 + if (mfs->mfsrr_bytes < mfs->mfsrr_watermark)
27410 + err = au_wbr_create_rr(dentry, isdir);
27411 + mutex_unlock(&mfs->mfs_lock);
27414 + AuDbg("b%d\n", err);
27418 +static int au_wbr_create_init_mfsrr(struct super_block *sb)
27422 + au_wbr_create_init_mfs(sb); /* ignore */
27423 + err = au_wbr_create_init_rr(sb);
27428 +/* ---------------------------------------------------------------------- */
27430 +/* top down parent and most free space */
27431 +static int au_wbr_create_pmfs(struct dentry *dentry, int isdir)
27434 + unsigned long long b;
27435 + aufs_bindex_t bindex, bstart, bend;
27436 + struct super_block *sb;
27437 + struct dentry *parent, *h_parent;
27438 + struct au_branch *br;
27440 + err = au_wbr_create_tdp(dentry, isdir);
27441 + if (unlikely(err < 0))
27443 + parent = dget_parent(dentry);
27444 + bstart = au_dbstart(parent);
27445 + bend = au_dbtaildir(parent);
27446 + if (bstart == bend)
27447 + goto out_parent; /* success */
27449 + e2 = au_wbr_create_mfs(dentry, isdir);
27451 + goto out_parent; /* success */
27453 + /* when the available size is equal, select upper one */
27454 + sb = dentry->d_sb;
27455 + br = au_sbr(sb, err);
27456 + b = br->br_wbr->wbr_bytes;
27457 + AuDbg("b%d, %llu\n", err, b);
27459 + for (bindex = bstart; bindex <= bend; bindex++) {
27460 + h_parent = au_h_dptr(parent, bindex);
27461 + if (!h_parent || !h_parent->d_inode)
27464 + br = au_sbr(sb, bindex);
27465 + if (!au_br_rdonly(br) && br->br_wbr->wbr_bytes > b) {
27466 + b = br->br_wbr->wbr_bytes;
27468 + AuDbg("b%d, %llu\n", err, b);
27473 + err = au_wbr_nonopq(dentry, err);
27478 + AuDbg("b%d\n", err);
27482 +/* ---------------------------------------------------------------------- */
27484 +/* policies for copyup */
27486 +/* top down parent */
27487 +static int au_wbr_copyup_tdp(struct dentry *dentry)
27489 + return au_wbr_create_tdp(dentry, /*isdir, anything is ok*/0);
27492 +/* bottom up parent */
27493 +static int au_wbr_copyup_bup(struct dentry *dentry)
27496 + aufs_bindex_t bindex, bstart;
27497 + struct dentry *parent, *h_parent;
27498 + struct super_block *sb;
27501 + sb = dentry->d_sb;
27502 + parent = dget_parent(dentry);
27503 + bstart = au_dbstart(parent);
27504 + for (bindex = au_dbstart(dentry); bindex >= bstart; bindex--) {
27505 + h_parent = au_h_dptr(parent, bindex);
27506 + if (!h_parent || !h_parent->d_inode)
27509 + if (!au_br_rdonly(au_sbr(sb, bindex))) {
27516 + /* bottom up here */
27517 + if (unlikely(err < 0))
27518 + err = au_wbr_bu(sb, bstart - 1);
27520 + AuDbg("b%d\n", err);
27525 +static int au_wbr_copyup_bu(struct dentry *dentry)
27528 + aufs_bindex_t bstart;
27530 + bstart = au_dbstart(dentry);
27531 + err = au_wbr_bu(dentry->d_sb, bstart);
27532 + AuDbg("b%d\n", err);
27533 + if (err > bstart)
27534 + err = au_wbr_nonopq(dentry, err);
27536 + AuDbg("b%d\n", err);
27540 +/* ---------------------------------------------------------------------- */
27542 +struct au_wbr_copyup_operations au_wbr_copyup_ops[] = {
27543 + [AuWbrCopyup_TDP] = {
27544 + .copyup = au_wbr_copyup_tdp
27546 + [AuWbrCopyup_BUP] = {
27547 + .copyup = au_wbr_copyup_bup
27549 + [AuWbrCopyup_BU] = {
27550 + .copyup = au_wbr_copyup_bu
27554 +struct au_wbr_create_operations au_wbr_create_ops[] = {
27555 + [AuWbrCreate_TDP] = {
27556 + .create = au_wbr_create_tdp
27558 + [AuWbrCreate_RR] = {
27559 + .create = au_wbr_create_rr,
27560 + .init = au_wbr_create_init_rr
27562 + [AuWbrCreate_MFS] = {
27563 + .create = au_wbr_create_mfs,
27564 + .init = au_wbr_create_init_mfs,
27565 + .fin = au_wbr_create_fin_mfs
27567 + [AuWbrCreate_MFSV] = {
27568 + .create = au_wbr_create_mfs,
27569 + .init = au_wbr_create_init_mfs,
27570 + .fin = au_wbr_create_fin_mfs
27572 + [AuWbrCreate_MFSRR] = {
27573 + .create = au_wbr_create_mfsrr,
27574 + .init = au_wbr_create_init_mfsrr,
27575 + .fin = au_wbr_create_fin_mfs
27577 + [AuWbrCreate_MFSRRV] = {
27578 + .create = au_wbr_create_mfsrr,
27579 + .init = au_wbr_create_init_mfsrr,
27580 + .fin = au_wbr_create_fin_mfs
27582 + [AuWbrCreate_PMFS] = {
27583 + .create = au_wbr_create_pmfs,
27584 + .init = au_wbr_create_init_mfs,
27585 + .fin = au_wbr_create_fin_mfs
27587 + [AuWbrCreate_PMFSV] = {
27588 + .create = au_wbr_create_pmfs,
27589 + .init = au_wbr_create_init_mfs,
27590 + .fin = au_wbr_create_fin_mfs
27593 diff -urN /usr/share/empty/fs/aufs/whout.c linux/fs/aufs/whout.c
27594 --- /usr/share/empty/fs/aufs/whout.c 1970-01-01 01:00:00.000000000 +0100
27595 +++ linux/fs/aufs/whout.c 2011-10-24 20:51:51.583800333 +0200
27598 + * Copyright (C) 2005-2011 Junjiro R. Okajima
27600 + * This program, aufs is free software; you can redistribute it and/or modify
27601 + * it under the terms of the GNU General Public License as published by
27602 + * the Free Software Foundation; either version 2 of the License, or
27603 + * (at your option) any later version.
27605 + * This program is distributed in the hope that it will be useful,
27606 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
27607 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27608 + * GNU General Public License for more details.
27610 + * You should have received a copy of the GNU General Public License
27611 + * along with this program; if not, write to the Free Software
27612 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
27616 + * whiteout for logical deletion and opaque directory
27619 +#include <linux/fs.h>
27622 +#define WH_MASK S_IRUGO
27625 + * If a directory contains this file, then it is opaque. We start with the
27626 + * .wh. flag so that it is blocked by lookup.
27628 +static struct qstr diropq_name = {
27629 + .name = AUFS_WH_DIROPQ,
27630 + .len = sizeof(AUFS_WH_DIROPQ) - 1
27634 + * generate whiteout name, which is NOT terminated by NULL.
27635 + * @name: original d_name.name
27636 + * @len: original d_name.len
27637 + * @wh: whiteout qstr
27638 + * returns zero when succeeds, otherwise error.
27639 + * succeeded value as wh->name should be freed by kfree().
27641 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name)
27645 + if (unlikely(name->len > PATH_MAX - AUFS_WH_PFX_LEN))
27646 + return -ENAMETOOLONG;
27648 + wh->len = name->len + AUFS_WH_PFX_LEN;
27649 + p = kmalloc(wh->len, GFP_NOFS);
27652 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
27653 + memcpy(p + AUFS_WH_PFX_LEN, name->name, name->len);
27660 +/* ---------------------------------------------------------------------- */
27663 + * test if the @wh_name exists under @h_parent.
27664 + * @try_sio specifies the necessary of super-io.
27666 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name,
27667 + struct au_branch *br, int try_sio)
27670 + struct dentry *wh_dentry;
27673 + wh_dentry = au_lkup_one(wh_name, h_parent, br, /*nd*/NULL);
27675 + wh_dentry = au_sio_lkup_one(wh_name, h_parent, br);
27676 + err = PTR_ERR(wh_dentry);
27677 + if (IS_ERR(wh_dentry))
27681 + if (!wh_dentry->d_inode)
27682 + goto out_wh; /* success */
27685 + if (S_ISREG(wh_dentry->d_inode->i_mode))
27686 + goto out_wh; /* success */
27689 + AuIOErr("%.*s Invalid whiteout entry type 0%o.\n",
27690 + AuDLNPair(wh_dentry), wh_dentry->d_inode->i_mode);
27699 + * test if the @h_dentry sets opaque or not.
27701 +int au_diropq_test(struct dentry *h_dentry, struct au_branch *br)
27704 + struct inode *h_dir;
27706 + h_dir = h_dentry->d_inode;
27707 + err = au_wh_test(h_dentry, &diropq_name, br,
27708 + au_test_h_perm_sio(h_dir, MAY_EXEC));
27713 + * returns a negative dentry whose name is unique and temporary.
27715 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
27716 + struct qstr *prefix)
27718 + struct dentry *dentry;
27720 + char defname[NAME_MAX - AUFS_MAX_NAMELEN + DNAME_INLINE_LEN + 1],
27722 + /* strict atomic_t is unnecessary here */
27723 + static unsigned short cnt;
27726 + BUILD_BUG_ON(sizeof(cnt) * 2 > AUFS_WH_TMP_LEN);
27729 + qs.len = sizeof(defname) - DNAME_INLINE_LEN + prefix->len - 1;
27730 + if (unlikely(prefix->len > DNAME_INLINE_LEN)) {
27731 + dentry = ERR_PTR(-ENAMETOOLONG);
27732 + if (unlikely(qs.len > NAME_MAX))
27734 + dentry = ERR_PTR(-ENOMEM);
27735 + name = kmalloc(qs.len + 1, GFP_NOFS);
27736 + if (unlikely(!name))
27740 + /* doubly whiteout-ed */
27741 + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2);
27742 + p = name + AUFS_WH_PFX_LEN * 2;
27743 + memcpy(p, prefix->name, prefix->len);
27744 + p += prefix->len;
27746 + AuDebugOn(name + qs.len + 1 - p <= AUFS_WH_TMP_LEN);
27749 + for (i = 0; i < 3; i++) {
27750 + sprintf(p, "%.*x", AUFS_WH_TMP_LEN, cnt++);
27751 + dentry = au_sio_lkup_one(&qs, h_parent, br);
27752 + if (IS_ERR(dentry) || !dentry->d_inode)
27756 + /* pr_warning("could not get random name\n"); */
27757 + dentry = ERR_PTR(-EEXIST);
27758 + AuDbg("%.*s\n", AuLNPair(&qs));
27762 + if (name != defname)
27765 + AuTraceErrPtr(dentry);
27770 + * rename the @h_dentry on @br to the whiteouted temporary name.
27772 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br)
27775 + struct path h_path = {
27776 + .mnt = br->br_mnt
27778 + struct inode *h_dir;
27779 + struct dentry *h_parent;
27781 + h_parent = h_dentry->d_parent; /* dir inode is locked */
27782 + h_dir = h_parent->d_inode;
27783 + IMustLock(h_dir);
27785 + h_path.dentry = au_whtmp_lkup(h_parent, br, &h_dentry->d_name);
27786 + err = PTR_ERR(h_path.dentry);
27787 + if (IS_ERR(h_path.dentry))
27790 + /* under the same dir, no need to lock_rename() */
27791 + err = vfsub_rename(h_dir, h_dentry, h_dir, &h_path);
27793 + dput(h_path.dentry);
27800 +/* ---------------------------------------------------------------------- */
27802 + * functions for removing a whiteout
27805 +static int do_unlink_wh(struct inode *h_dir, struct path *h_path)
27810 + * forces superio when the dir has a sticky bit.
27811 + * this may be a violation of unix fs semantics.
27813 + force = (h_dir->i_mode & S_ISVTX)
27814 + && h_path->dentry->d_inode->i_uid != current_fsuid();
27815 + return vfsub_unlink(h_dir, h_path, force);
27818 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
27819 + struct dentry *dentry)
27823 + err = do_unlink_wh(h_dir, h_path);
27824 + if (!err && dentry)
27825 + au_set_dbwh(dentry, -1);
27830 +static int unlink_wh_name(struct dentry *h_parent, struct qstr *wh,
27831 + struct au_branch *br)
27834 + struct path h_path = {
27835 + .mnt = br->br_mnt
27839 + h_path.dentry = au_lkup_one(wh, h_parent, br, /*nd*/NULL);
27840 + if (IS_ERR(h_path.dentry))
27841 + err = PTR_ERR(h_path.dentry);
27843 + if (h_path.dentry->d_inode
27844 + && S_ISREG(h_path.dentry->d_inode->i_mode))
27845 + err = do_unlink_wh(h_parent->d_inode, &h_path);
27846 + dput(h_path.dentry);
27852 +/* ---------------------------------------------------------------------- */
27854 + * initialize/clean whiteout for a branch
27857 +static void au_wh_clean(struct inode *h_dir, struct path *whpath,
27862 + if (!whpath->dentry->d_inode)
27865 + err = mnt_want_write(whpath->mnt);
27868 + err = vfsub_rmdir(h_dir, whpath);
27870 + err = vfsub_unlink(h_dir, whpath, /*force*/0);
27871 + mnt_drop_write(whpath->mnt);
27873 + if (unlikely(err))
27874 + pr_warning("failed removing %.*s (%d), ignored.\n",
27875 + AuDLNPair(whpath->dentry), err);
27878 +static int test_linkable(struct dentry *h_root)
27880 + struct inode *h_dir = h_root->d_inode;
27882 + if (h_dir->i_op->link)
27885 + pr_err("%.*s (%s) doesn't support link(2), use noplink and rw+nolwh\n",
27886 + AuDLNPair(h_root), au_sbtype(h_root->d_sb));
27890 +/* todo: should this mkdir be done in /sbin/mount.aufs helper? */
27891 +static int au_whdir(struct inode *h_dir, struct path *path)
27896 + if (!path->dentry->d_inode) {
27897 + int mode = S_IRWXU;
27899 + if (au_test_nfs(path->dentry->d_sb))
27901 + err = mnt_want_write(path->mnt);
27903 + err = vfsub_mkdir(h_dir, path, mode);
27904 + mnt_drop_write(path->mnt);
27906 + } else if (S_ISDIR(path->dentry->d_inode->i_mode))
27909 + pr_err("unknown %.*s exists\n", AuDLNPair(path->dentry));
27914 +struct au_wh_base {
27915 + const struct qstr *name;
27916 + struct dentry *dentry;
27919 +static void au_wh_init_ro(struct inode *h_dir, struct au_wh_base base[],
27920 + struct path *h_path)
27922 + h_path->dentry = base[AuBrWh_BASE].dentry;
27923 + au_wh_clean(h_dir, h_path, /*isdir*/0);
27924 + h_path->dentry = base[AuBrWh_PLINK].dentry;
27925 + au_wh_clean(h_dir, h_path, /*isdir*/1);
27926 + h_path->dentry = base[AuBrWh_ORPH].dentry;
27927 + au_wh_clean(h_dir, h_path, /*isdir*/1);
27931 + * returns tri-state,
27932 + * minus: error, caller should print the mesage
27934 + * plus: error, caller should NOT print the mesage
27936 +static int au_wh_init_rw_nolink(struct dentry *h_root, struct au_wbr *wbr,
27937 + int do_plink, struct au_wh_base base[],
27938 + struct path *h_path)
27941 + struct inode *h_dir;
27943 + h_dir = h_root->d_inode;
27944 + h_path->dentry = base[AuBrWh_BASE].dentry;
27945 + au_wh_clean(h_dir, h_path, /*isdir*/0);
27946 + h_path->dentry = base[AuBrWh_PLINK].dentry;
27948 + err = test_linkable(h_root);
27949 + if (unlikely(err)) {
27954 + err = au_whdir(h_dir, h_path);
27955 + if (unlikely(err))
27957 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
27959 + au_wh_clean(h_dir, h_path, /*isdir*/1);
27960 + h_path->dentry = base[AuBrWh_ORPH].dentry;
27961 + err = au_whdir(h_dir, h_path);
27962 + if (unlikely(err))
27964 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
27971 + * for the moment, aufs supports the branch filesystem which does not support
27972 + * link(2). testing on FAT which does not support i_op->setattr() fully either,
27973 + * copyup failed. finally, such filesystem will not be used as the writable
27976 + * returns tri-state, see above.
27978 +static int au_wh_init_rw(struct dentry *h_root, struct au_wbr *wbr,
27979 + int do_plink, struct au_wh_base base[],
27980 + struct path *h_path)
27983 + struct inode *h_dir;
27985 + WbrWhMustWriteLock(wbr);
27987 + err = test_linkable(h_root);
27988 + if (unlikely(err)) {
27994 + * todo: should this create be done in /sbin/mount.aufs helper?
27997 + h_dir = h_root->d_inode;
27998 + if (!base[AuBrWh_BASE].dentry->d_inode) {
27999 + err = mnt_want_write(h_path->mnt);
28001 + h_path->dentry = base[AuBrWh_BASE].dentry;
28002 + err = vfsub_create(h_dir, h_path, WH_MASK);
28003 + mnt_drop_write(h_path->mnt);
28005 + } else if (S_ISREG(base[AuBrWh_BASE].dentry->d_inode->i_mode))
28008 + pr_err("unknown %.*s/%.*s exists\n",
28009 + AuDLNPair(h_root), AuDLNPair(base[AuBrWh_BASE].dentry));
28010 + if (unlikely(err))
28013 + h_path->dentry = base[AuBrWh_PLINK].dentry;
28015 + err = au_whdir(h_dir, h_path);
28016 + if (unlikely(err))
28018 + wbr->wbr_plink = dget(base[AuBrWh_PLINK].dentry);
28020 + au_wh_clean(h_dir, h_path, /*isdir*/1);
28021 + wbr->wbr_whbase = dget(base[AuBrWh_BASE].dentry);
28023 + h_path->dentry = base[AuBrWh_ORPH].dentry;
28024 + err = au_whdir(h_dir, h_path);
28025 + if (unlikely(err))
28027 + wbr->wbr_orph = dget(base[AuBrWh_ORPH].dentry);
28034 + * initialize the whiteout base file/dir for @br.
28036 +int au_wh_init(struct dentry *h_root, struct au_branch *br,
28037 + struct super_block *sb)
28040 + const unsigned char do_plink
28041 + = !!au_opt_test(au_mntflags(sb), PLINK);
28042 + struct path path = {
28043 + .mnt = br->br_mnt
28045 + struct inode *h_dir;
28046 + struct au_wbr *wbr = br->br_wbr;
28047 + static const struct qstr base_name[] = {
28048 + [AuBrWh_BASE] = {
28049 + .name = AUFS_BASE_NAME,
28050 + .len = sizeof(AUFS_BASE_NAME) - 1
28052 + [AuBrWh_PLINK] = {
28053 + .name = AUFS_PLINKDIR_NAME,
28054 + .len = sizeof(AUFS_PLINKDIR_NAME) - 1
28056 + [AuBrWh_ORPH] = {
28057 + .name = AUFS_ORPHDIR_NAME,
28058 + .len = sizeof(AUFS_ORPHDIR_NAME) - 1
28061 + struct au_wh_base base[] = {
28062 + [AuBrWh_BASE] = {
28063 + .name = base_name + AuBrWh_BASE,
28066 + [AuBrWh_PLINK] = {
28067 + .name = base_name + AuBrWh_PLINK,
28070 + [AuBrWh_ORPH] = {
28071 + .name = base_name + AuBrWh_ORPH,
28077 + WbrWhMustWriteLock(wbr);
28079 + for (i = 0; i < AuBrWh_Last; i++) {
28080 + /* doubly whiteouted */
28081 + struct dentry *d;
28083 + d = au_wh_lkup(h_root, (void *)base[i].name, br);
28084 + err = PTR_ERR(d);
28088 + base[i].dentry = d;
28090 + && wbr->wbr_wh[i]
28091 + && wbr->wbr_wh[i] != base[i].dentry);
28095 + for (i = 0; i < AuBrWh_Last; i++) {
28096 + dput(wbr->wbr_wh[i]);
28097 + wbr->wbr_wh[i] = NULL;
28101 + if (!au_br_writable(br->br_perm)) {
28102 + h_dir = h_root->d_inode;
28103 + au_wh_init_ro(h_dir, base, &path);
28104 + } else if (!au_br_wh_linkable(br->br_perm)) {
28105 + err = au_wh_init_rw_nolink(h_root, wbr, do_plink, base, &path);
28111 + err = au_wh_init_rw(h_root, wbr, do_plink, base, &path);
28117 + goto out; /* success */
28120 + pr_err("an error(%d) on the writable branch %.*s(%s)\n",
28121 + err, AuDLNPair(h_root), au_sbtype(h_root->d_sb));
28123 + for (i = 0; i < AuBrWh_Last; i++)
28124 + dput(base[i].dentry);
28128 +/* ---------------------------------------------------------------------- */
28130 + * whiteouts are all hard-linked usually.
28131 + * when its link count reaches a ceiling, we create a new whiteout base
28132 + * asynchronously.
28135 +struct reinit_br_wh {
28136 + struct super_block *sb;
28137 + struct au_branch *br;
28140 +static void reinit_br_wh(void *arg)
28143 + aufs_bindex_t bindex;
28144 + struct path h_path;
28145 + struct reinit_br_wh *a = arg;
28146 + struct au_wbr *wbr;
28147 + struct inode *dir;
28148 + struct dentry *h_root;
28149 + struct au_hinode *hdir;
28152 + wbr = a->br->br_wbr;
28153 + /* big aufs lock */
28154 + si_noflush_write_lock(a->sb);
28155 + if (!au_br_writable(a->br->br_perm))
28157 + bindex = au_br_index(a->sb, a->br->br_id);
28158 + if (unlikely(bindex < 0))
28161 + di_read_lock_parent(a->sb->s_root, AuLock_IR);
28162 + dir = a->sb->s_root->d_inode;
28163 + hdir = au_hi(dir, bindex);
28164 + h_root = au_h_dptr(a->sb->s_root, bindex);
28166 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
28167 + wbr_wh_write_lock(wbr);
28168 + err = au_h_verify(wbr->wbr_whbase, au_opt_udba(a->sb), hdir->hi_inode,
28171 + err = mnt_want_write(a->br->br_mnt);
28173 + h_path.dentry = wbr->wbr_whbase;
28174 + h_path.mnt = a->br->br_mnt;
28175 + err = vfsub_unlink(hdir->hi_inode, &h_path, /*force*/0);
28176 + mnt_drop_write(a->br->br_mnt);
28179 + pr_warning("%.*s is moved, ignored\n",
28180 + AuDLNPair(wbr->wbr_whbase));
28183 + dput(wbr->wbr_whbase);
28184 + wbr->wbr_whbase = NULL;
28186 + err = au_wh_init(h_root, a->br, a->sb);
28187 + wbr_wh_write_unlock(wbr);
28188 + au_hn_imtx_unlock(hdir);
28189 + di_read_unlock(a->sb->s_root, AuLock_IR);
28193 + atomic_dec(&wbr->wbr_wh_running);
28194 + atomic_dec(&a->br->br_count);
28195 + si_write_unlock(a->sb);
28196 + au_nwt_done(&au_sbi(a->sb)->si_nowait);
28198 + if (unlikely(err))
28199 + AuIOErr("err %d\n", err);
28202 +static void kick_reinit_br_wh(struct super_block *sb, struct au_branch *br)
28204 + int do_dec, wkq_err;
28205 + struct reinit_br_wh *arg;
28208 + if (atomic_inc_return(&br->br_wbr->wbr_wh_running) != 1)
28211 + /* ignore ENOMEM */
28212 + arg = kmalloc(sizeof(*arg), GFP_NOFS);
28215 + * dec(wh_running), kfree(arg) and dec(br_count)
28216 + * in reinit function
28220 + atomic_inc(&br->br_count);
28221 + wkq_err = au_wkq_nowait(reinit_br_wh, arg, sb, /*flags*/0);
28222 + if (unlikely(wkq_err)) {
28223 + atomic_dec(&br->br_wbr->wbr_wh_running);
28224 + atomic_dec(&br->br_count);
28232 + atomic_dec(&br->br_wbr->wbr_wh_running);
28235 +/* ---------------------------------------------------------------------- */
28238 + * create the whiteout @wh.
28240 +static int link_or_create_wh(struct super_block *sb, aufs_bindex_t bindex,
28241 + struct dentry *wh)
28244 + struct path h_path = {
28247 + struct au_branch *br;
28248 + struct au_wbr *wbr;
28249 + struct dentry *h_parent;
28250 + struct inode *h_dir;
28252 + h_parent = wh->d_parent; /* dir inode is locked */
28253 + h_dir = h_parent->d_inode;
28254 + IMustLock(h_dir);
28256 + br = au_sbr(sb, bindex);
28257 + h_path.mnt = br->br_mnt;
28258 + wbr = br->br_wbr;
28259 + wbr_wh_read_lock(wbr);
28260 + if (wbr->wbr_whbase) {
28261 + err = vfsub_link(wbr->wbr_whbase, h_dir, &h_path);
28262 + if (!err || err != -EMLINK)
28265 + /* link count full. re-initialize br_whbase. */
28266 + kick_reinit_br_wh(sb, br);
28269 + /* return this error in this context */
28270 + err = vfsub_create(h_dir, &h_path, WH_MASK);
28273 + wbr_wh_read_unlock(wbr);
28277 +/* ---------------------------------------------------------------------- */
28280 + * create or remove the diropq.
28282 +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex,
28283 + unsigned int flags)
28285 + struct dentry *opq_dentry, *h_dentry;
28286 + struct super_block *sb;
28287 + struct au_branch *br;
28290 + sb = dentry->d_sb;
28291 + br = au_sbr(sb, bindex);
28292 + h_dentry = au_h_dptr(dentry, bindex);
28293 + opq_dentry = au_lkup_one(&diropq_name, h_dentry, br, /*nd*/NULL);
28294 + if (IS_ERR(opq_dentry))
28297 + if (au_ftest_diropq(flags, CREATE)) {
28298 + err = link_or_create_wh(sb, bindex, opq_dentry);
28300 + au_set_dbdiropq(dentry, bindex);
28301 + goto out; /* success */
28304 + struct path tmp = {
28305 + .dentry = opq_dentry,
28306 + .mnt = br->br_mnt
28308 + err = do_unlink_wh(au_h_iptr(dentry->d_inode, bindex), &tmp);
28310 + au_set_dbdiropq(dentry, -1);
28312 + dput(opq_dentry);
28313 + opq_dentry = ERR_PTR(err);
28316 + return opq_dentry;
28319 +struct do_diropq_args {
28320 + struct dentry **errp;
28321 + struct dentry *dentry;
28322 + aufs_bindex_t bindex;
28323 + unsigned int flags;
28326 +static void call_do_diropq(void *args)
28328 + struct do_diropq_args *a = args;
28329 + *a->errp = do_diropq(a->dentry, a->bindex, a->flags);
28332 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
28333 + unsigned int flags)
28335 + struct dentry *diropq, *h_dentry;
28337 + h_dentry = au_h_dptr(dentry, bindex);
28338 + if (!au_test_h_perm_sio(h_dentry->d_inode, MAY_EXEC | MAY_WRITE))
28339 + diropq = do_diropq(dentry, bindex, flags);
28342 + struct do_diropq_args args = {
28344 + .dentry = dentry,
28345 + .bindex = bindex,
28349 + wkq_err = au_wkq_wait(call_do_diropq, &args);
28350 + if (unlikely(wkq_err))
28351 + diropq = ERR_PTR(wkq_err);
28357 +/* ---------------------------------------------------------------------- */
28360 + * lookup whiteout dentry.
28361 + * @h_parent: lower parent dentry which must exist and be locked
28362 + * @base_name: name of dentry which will be whiteouted
28363 + * returns dentry for whiteout.
28365 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
28366 + struct au_branch *br)
28369 + struct qstr wh_name;
28370 + struct dentry *wh_dentry;
28372 + err = au_wh_name_alloc(&wh_name, base_name);
28373 + wh_dentry = ERR_PTR(err);
28375 + wh_dentry = au_lkup_one(&wh_name, h_parent, br, /*nd*/NULL);
28376 + kfree(wh_name.name);
28378 + return wh_dentry;
28382 + * link/create a whiteout for @dentry on @bindex.
28384 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
28385 + struct dentry *h_parent)
28387 + struct dentry *wh_dentry;
28388 + struct super_block *sb;
28391 + sb = dentry->d_sb;
28392 + wh_dentry = au_wh_lkup(h_parent, &dentry->d_name, au_sbr(sb, bindex));
28393 + if (!IS_ERR(wh_dentry) && !wh_dentry->d_inode) {
28394 + err = link_or_create_wh(sb, bindex, wh_dentry);
28396 + au_set_dbwh(dentry, bindex);
28399 + wh_dentry = ERR_PTR(err);
28403 + return wh_dentry;
28406 +/* ---------------------------------------------------------------------- */
28408 +/* Delete all whiteouts in this directory on branch bindex. */
28409 +static int del_wh_children(struct dentry *h_dentry, struct au_nhash *whlist,
28410 + aufs_bindex_t bindex, struct au_branch *br)
28413 + unsigned long ul, n;
28414 + struct qstr wh_name;
28416 + struct hlist_head *head;
28417 + struct au_vdir_wh *tpos;
28418 + struct hlist_node *pos;
28419 + struct au_vdir_destr *str;
28422 + p = __getname_gfp(GFP_NOFS);
28423 + wh_name.name = p;
28424 + if (unlikely(!wh_name.name))
28428 + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN);
28429 + p += AUFS_WH_PFX_LEN;
28430 + n = whlist->nh_num;
28431 + head = whlist->nh_head;
28432 + for (ul = 0; !err && ul < n; ul++, head++) {
28433 + hlist_for_each_entry(tpos, pos, head, wh_hash) {
28434 + if (tpos->wh_bindex != bindex)
28437 + str = &tpos->wh_str;
28438 + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) {
28439 + memcpy(p, str->name, str->len);
28440 + wh_name.len = AUFS_WH_PFX_LEN + str->len;
28441 + err = unlink_wh_name(h_dentry, &wh_name, br);
28446 + AuIOErr("whiteout name too long %.*s\n",
28447 + str->len, str->name);
28452 + __putname(wh_name.name);
28458 +struct del_wh_children_args {
28460 + struct dentry *h_dentry;
28461 + struct au_nhash *whlist;
28462 + aufs_bindex_t bindex;
28463 + struct au_branch *br;
28466 +static void call_del_wh_children(void *args)
28468 + struct del_wh_children_args *a = args;
28469 + *a->errp = del_wh_children(a->h_dentry, a->whlist, a->bindex, a->br);
28472 +/* ---------------------------------------------------------------------- */
28474 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp)
28476 + struct au_whtmp_rmdir *whtmp;
28478 + unsigned int rdhash;
28480 + SiMustAnyLock(sb);
28482 + whtmp = kmalloc(sizeof(*whtmp), gfp);
28483 + if (unlikely(!whtmp)) {
28484 + whtmp = ERR_PTR(-ENOMEM);
28488 + whtmp->dir = NULL;
28489 + whtmp->br = NULL;
28490 + whtmp->wh_dentry = NULL;
28491 + /* no estimation for dir size */
28492 + rdhash = au_sbi(sb)->si_rdhash;
28494 + rdhash = AUFS_RDHASH_DEF;
28495 + err = au_nhash_alloc(&whtmp->whlist, rdhash, gfp);
28496 + if (unlikely(err)) {
28498 + whtmp = ERR_PTR(err);
28505 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp)
28508 + atomic_dec(&whtmp->br->br_count);
28509 + dput(whtmp->wh_dentry);
28510 + iput(whtmp->dir);
28511 + au_nhash_wh_free(&whtmp->whlist);
28516 + * rmdir the whiteouted temporary named dir @h_dentry.
28517 + * @whlist: whiteouted children.
28519 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
28520 + struct dentry *wh_dentry, struct au_nhash *whlist)
28523 + struct path h_tmp;
28524 + struct inode *wh_inode, *h_dir;
28525 + struct au_branch *br;
28527 + h_dir = wh_dentry->d_parent->d_inode; /* dir inode is locked */
28528 + IMustLock(h_dir);
28530 + br = au_sbr(dir->i_sb, bindex);
28531 + wh_inode = wh_dentry->d_inode;
28532 + mutex_lock_nested(&wh_inode->i_mutex, AuLsc_I_CHILD);
28535 + * someone else might change some whiteouts while we were sleeping.
28536 + * it means this whlist may have an obsoleted entry.
28538 + if (!au_test_h_perm_sio(wh_inode, MAY_EXEC | MAY_WRITE))
28539 + err = del_wh_children(wh_dentry, whlist, bindex, br);
28542 + struct del_wh_children_args args = {
28544 + .h_dentry = wh_dentry,
28545 + .whlist = whlist,
28546 + .bindex = bindex,
28550 + wkq_err = au_wkq_wait(call_del_wh_children, &args);
28551 + if (unlikely(wkq_err))
28554 + mutex_unlock(&wh_inode->i_mutex);
28557 + h_tmp.dentry = wh_dentry;
28558 + h_tmp.mnt = br->br_mnt;
28559 + err = vfsub_rmdir(h_dir, &h_tmp);
28563 + if (au_ibstart(dir) == bindex) {
28564 + /* todo: dir->i_mutex is necessary */
28565 + au_cpup_attr_timesizes(dir);
28566 + vfsub_drop_nlink(dir);
28568 + return 0; /* success */
28571 + pr_warning("failed removing %.*s(%d), ignored\n",
28572 + AuDLNPair(wh_dentry), err);
28576 +static void call_rmdir_whtmp(void *args)
28579 + aufs_bindex_t bindex;
28580 + struct au_whtmp_rmdir *a = args;
28581 + struct super_block *sb;
28582 + struct dentry *h_parent;
28583 + struct inode *h_dir;
28584 + struct au_hinode *hdir;
28586 + /* rmdir by nfsd may cause deadlock with this i_mutex */
28587 + /* mutex_lock(&a->dir->i_mutex); */
28589 + sb = a->dir->i_sb;
28590 + si_read_lock(sb, !AuLock_FLUSH);
28591 + if (!au_br_writable(a->br->br_perm))
28593 + bindex = au_br_index(sb, a->br->br_id);
28594 + if (unlikely(bindex < 0))
28598 + ii_write_lock_parent(a->dir);
28599 + h_parent = dget_parent(a->wh_dentry);
28600 + h_dir = h_parent->d_inode;
28601 + hdir = au_hi(a->dir, bindex);
28602 + au_hn_imtx_lock_nested(hdir, AuLsc_I_PARENT);
28603 + err = au_h_verify(a->wh_dentry, au_opt_udba(sb), h_dir, h_parent,
28606 + err = mnt_want_write(a->br->br_mnt);
28608 + err = au_whtmp_rmdir(a->dir, bindex, a->wh_dentry,
28610 + mnt_drop_write(a->br->br_mnt);
28613 + au_hn_imtx_unlock(hdir);
28615 + ii_write_unlock(a->dir);
28618 + /* mutex_unlock(&a->dir->i_mutex); */
28619 + au_whtmp_rmdir_free(a);
28620 + si_read_unlock(sb);
28621 + au_nwt_done(&au_sbi(sb)->si_nowait);
28622 + if (unlikely(err))
28623 + AuIOErr("err %d\n", err);
28626 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
28627 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args)
28630 + struct super_block *sb;
28634 + /* all post-process will be done in do_rmdir_whtmp(). */
28636 + args->dir = au_igrab(dir);
28637 + args->br = au_sbr(sb, bindex);
28638 + atomic_inc(&args->br->br_count);
28639 + args->wh_dentry = dget(wh_dentry);
28640 + wkq_err = au_wkq_nowait(call_rmdir_whtmp, args, sb, /*flags*/0);
28641 + if (unlikely(wkq_err)) {
28642 + pr_warning("rmdir error %.*s (%d), ignored\n",
28643 + AuDLNPair(wh_dentry), wkq_err);
28644 + au_whtmp_rmdir_free(args);
28647 diff -urN /usr/share/empty/fs/aufs/whout.h linux/fs/aufs/whout.h
28648 --- /usr/share/empty/fs/aufs/whout.h 1970-01-01 01:00:00.000000000 +0100
28649 +++ linux/fs/aufs/whout.h 2011-08-24 13:30:24.734646739 +0200
28652 + * Copyright (C) 2005-2011 Junjiro R. Okajima
28654 + * This program, aufs is free software; you can redistribute it and/or modify
28655 + * it under the terms of the GNU General Public License as published by
28656 + * the Free Software Foundation; either version 2 of the License, or
28657 + * (at your option) any later version.
28659 + * This program is distributed in the hope that it will be useful,
28660 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28661 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28662 + * GNU General Public License for more details.
28664 + * You should have received a copy of the GNU General Public License
28665 + * along with this program; if not, write to the Free Software
28666 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
28670 + * whiteout for logical deletion and opaque directory
28673 +#ifndef __AUFS_WHOUT_H__
28674 +#define __AUFS_WHOUT_H__
28678 +#include <linux/aufs_type.h>
28682 +int au_wh_name_alloc(struct qstr *wh, const struct qstr *name);
28684 +int au_wh_test(struct dentry *h_parent, struct qstr *wh_name,
28685 + struct au_branch *br, int try_sio);
28686 +int au_diropq_test(struct dentry *h_dentry, struct au_branch *br);
28687 +struct dentry *au_whtmp_lkup(struct dentry *h_parent, struct au_branch *br,
28688 + struct qstr *prefix);
28689 +int au_whtmp_ren(struct dentry *h_dentry, struct au_branch *br);
28690 +int au_wh_unlink_dentry(struct inode *h_dir, struct path *h_path,
28691 + struct dentry *dentry);
28692 +int au_wh_init(struct dentry *h_parent, struct au_branch *br,
28693 + struct super_block *sb);
28695 +/* diropq flags */
28696 +#define AuDiropq_CREATE 1
28697 +#define au_ftest_diropq(flags, name) ((flags) & AuDiropq_##name)
28698 +#define au_fset_diropq(flags, name) \
28699 + do { (flags) |= AuDiropq_##name; } while (0)
28700 +#define au_fclr_diropq(flags, name) \
28701 + do { (flags) &= ~AuDiropq_##name; } while (0)
28703 +struct dentry *au_diropq_sio(struct dentry *dentry, aufs_bindex_t bindex,
28704 + unsigned int flags);
28705 +struct dentry *au_wh_lkup(struct dentry *h_parent, struct qstr *base_name,
28706 + struct au_branch *br);
28707 +struct dentry *au_wh_create(struct dentry *dentry, aufs_bindex_t bindex,
28708 + struct dentry *h_parent);
28710 +/* real rmdir for the whiteout-ed dir */
28711 +struct au_whtmp_rmdir {
28712 + struct inode *dir;
28713 + struct au_branch *br;
28714 + struct dentry *wh_dentry;
28715 + struct au_nhash whlist;
28718 +struct au_whtmp_rmdir *au_whtmp_rmdir_alloc(struct super_block *sb, gfp_t gfp);
28719 +void au_whtmp_rmdir_free(struct au_whtmp_rmdir *whtmp);
28720 +int au_whtmp_rmdir(struct inode *dir, aufs_bindex_t bindex,
28721 + struct dentry *wh_dentry, struct au_nhash *whlist);
28722 +void au_whtmp_kick_rmdir(struct inode *dir, aufs_bindex_t bindex,
28723 + struct dentry *wh_dentry, struct au_whtmp_rmdir *args);
28725 +/* ---------------------------------------------------------------------- */
28727 +static inline struct dentry *au_diropq_create(struct dentry *dentry,
28728 + aufs_bindex_t bindex)
28730 + return au_diropq_sio(dentry, bindex, AuDiropq_CREATE);
28733 +static inline int au_diropq_remove(struct dentry *dentry, aufs_bindex_t bindex)
28735 + return PTR_ERR(au_diropq_sio(dentry, bindex, !AuDiropq_CREATE));
28738 +#endif /* __KERNEL__ */
28739 +#endif /* __AUFS_WHOUT_H__ */
28740 diff -urN /usr/share/empty/fs/aufs/wkq.c linux/fs/aufs/wkq.c
28741 --- /usr/share/empty/fs/aufs/wkq.c 1970-01-01 01:00:00.000000000 +0100
28742 +++ linux/fs/aufs/wkq.c 2012-01-05 11:00:59.218835185 +0100
28745 + * Copyright (C) 2005-2011 Junjiro R. Okajima
28747 + * This program, aufs is free software; you can redistribute it and/or modify
28748 + * it under the terms of the GNU General Public License as published by
28749 + * the Free Software Foundation; either version 2 of the License, or
28750 + * (at your option) any later version.
28752 + * This program is distributed in the hope that it will be useful,
28753 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28754 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28755 + * GNU General Public License for more details.
28757 + * You should have received a copy of the GNU General Public License
28758 + * along with this program; if not, write to the Free Software
28759 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
28763 + * workqueue for asynchronous/super-io operations
28764 + * todo: try new dredential scheme
28767 +#include <linux/module.h>
28770 +/* internal workqueue named AUFS_WKQ_NAME */
28772 +static struct workqueue_struct *au_wkq;
28774 +struct au_wkinfo {
28775 + struct work_struct wk;
28776 + struct kobject *kobj;
28778 + unsigned int flags; /* see wkq.h */
28780 + au_wkq_func_t func;
28783 + struct completion *comp;
28786 +/* ---------------------------------------------------------------------- */
28788 +static void wkq_func(struct work_struct *wk)
28790 + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk);
28792 + AuDebugOn(current_fsuid());
28793 + AuDebugOn(rlimit(RLIMIT_FSIZE) != RLIM_INFINITY);
28795 + wkinfo->func(wkinfo->args);
28796 + if (au_ftest_wkq(wkinfo->flags, WAIT))
28797 + complete(wkinfo->comp);
28799 + kobject_put(wkinfo->kobj);
28800 + module_put(THIS_MODULE); /* todo: ?? */
28806 + * Since struct completion is large, try allocating it dynamically.
28808 +#if defined(CONFIG_4KSTACKS) || defined(AuTest4KSTACKS)
28809 +#define AuWkqCompDeclare(name) struct completion *comp = NULL
28811 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
28813 + *comp = kmalloc(sizeof(**comp), GFP_NOFS);
28815 + init_completion(*comp);
28816 + wkinfo->comp = *comp;
28822 +static void au_wkq_comp_free(struct completion *comp)
28830 +#define AuWkqCompDeclare(name) \
28831 + DECLARE_COMPLETION_ONSTACK(_ ## name); \
28832 + struct completion *comp = &_ ## name
28834 +static int au_wkq_comp_alloc(struct au_wkinfo *wkinfo, struct completion **comp)
28836 + wkinfo->comp = *comp;
28840 +static void au_wkq_comp_free(struct completion *comp __maybe_unused)
28844 +#endif /* 4KSTACKS */
28846 +static void au_wkq_run(struct au_wkinfo *wkinfo)
28848 + if (au_ftest_wkq(wkinfo->flags, NEST)) {
28849 + if (au_wkq_test()) {
28850 + AuWarn1("wkq from wkq, due to a dead dir by UDBA?\n");
28851 + AuDebugOn(au_ftest_wkq(wkinfo->flags, WAIT));
28854 + au_dbg_verify_kthread();
28856 + if (au_ftest_wkq(wkinfo->flags, WAIT)) {
28857 + INIT_WORK_ONSTACK(&wkinfo->wk, wkq_func);
28858 + queue_work(au_wkq, &wkinfo->wk);
28860 + INIT_WORK(&wkinfo->wk, wkq_func);
28861 + schedule_work(&wkinfo->wk);
28866 + * Be careful. It is easy to make deadlock happen.
28867 + * processA: lock, wkq and wait
28868 + * processB: wkq and wait, lock in wkq
28871 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args)
28874 + AuWkqCompDeclare(comp);
28875 + struct au_wkinfo wkinfo = {
28881 + err = au_wkq_comp_alloc(&wkinfo, &comp);
28883 + au_wkq_run(&wkinfo);
28884 + /* no timeout, no interrupt */
28885 + wait_for_completion(wkinfo.comp);
28886 + au_wkq_comp_free(comp);
28887 + destroy_work_on_stack(&wkinfo.wk);
28895 + * Note: dget/dput() in func for aufs dentries are not supported. It will be a
28896 + * problem in a concurrent umounting.
28898 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
28899 + unsigned int flags)
28902 + struct au_wkinfo *wkinfo;
28904 + atomic_inc(&au_sbi(sb)->si_nowait.nw_len);
28907 + * wkq_func() must free this wkinfo.
28908 + * it highly depends upon the implementation of workqueue.
28911 + wkinfo = kmalloc(sizeof(*wkinfo), GFP_NOFS);
28913 + wkinfo->kobj = &au_sbi(sb)->si_kobj;
28914 + wkinfo->flags = flags & ~AuWkq_WAIT;
28915 + wkinfo->func = func;
28916 + wkinfo->args = args;
28917 + wkinfo->comp = NULL;
28918 + kobject_get(wkinfo->kobj);
28919 + __module_get(THIS_MODULE); /* todo: ?? */
28921 + au_wkq_run(wkinfo);
28924 + au_nwt_done(&au_sbi(sb)->si_nowait);
28930 +/* ---------------------------------------------------------------------- */
28932 +void au_nwt_init(struct au_nowait_tasks *nwt)
28934 + atomic_set(&nwt->nw_len, 0);
28935 + /* smp_mb(); */ /* atomic_set */
28936 + init_waitqueue_head(&nwt->nw_wq);
28939 +void au_wkq_fin(void)
28941 + destroy_workqueue(au_wkq);
28944 +int __init au_wkq_init(void)
28949 + BUILD_BUG_ON(!WQ_RESCUER);
28950 + au_wkq = alloc_workqueue(AUFS_WKQ_NAME, !WQ_RESCUER, WQ_DFL_ACTIVE);
28951 + if (IS_ERR(au_wkq))
28952 + err = PTR_ERR(au_wkq);
28953 + else if (!au_wkq)
28958 diff -urN /usr/share/empty/fs/aufs/wkq.h linux/fs/aufs/wkq.h
28959 --- /usr/share/empty/fs/aufs/wkq.h 1970-01-01 01:00:00.000000000 +0100
28960 +++ linux/fs/aufs/wkq.h 2012-01-05 11:00:59.218835185 +0100
28963 + * Copyright (C) 2005-2011 Junjiro R. Okajima
28965 + * This program, aufs is free software; you can redistribute it and/or modify
28966 + * it under the terms of the GNU General Public License as published by
28967 + * the Free Software Foundation; either version 2 of the License, or
28968 + * (at your option) any later version.
28970 + * This program is distributed in the hope that it will be useful,
28971 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
28972 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28973 + * GNU General Public License for more details.
28975 + * You should have received a copy of the GNU General Public License
28976 + * along with this program; if not, write to the Free Software
28977 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
28981 + * workqueue for asynchronous/super-io operations
28982 + * todo: try new credentials management scheme
28985 +#ifndef __AUFS_WKQ_H__
28986 +#define __AUFS_WKQ_H__
28990 +#include <linux/sched.h>
28991 +#include <linux/wait.h>
28992 +#include <linux/aufs_type.h>
28994 +struct super_block;
28996 +/* ---------------------------------------------------------------------- */
28999 + * in the next operation, wait for the 'nowait' tasks in system-wide workqueue
29001 +struct au_nowait_tasks {
29003 + wait_queue_head_t nw_wq;
29006 +/* ---------------------------------------------------------------------- */
29008 +typedef void (*au_wkq_func_t)(void *args);
29011 +#define AuWkq_WAIT 1
29012 +#define AuWkq_NEST (1 << 1)
29013 +#define au_ftest_wkq(flags, name) ((flags) & AuWkq_##name)
29014 +#define au_fset_wkq(flags, name) \
29015 + do { (flags) |= AuWkq_##name; } while (0)
29016 +#define au_fclr_wkq(flags, name) \
29017 + do { (flags) &= ~AuWkq_##name; } while (0)
29019 +#ifndef CONFIG_AUFS_HNOTIFY
29021 +#define AuWkq_NEST 0
29025 +int au_wkq_do_wait(unsigned int flags, au_wkq_func_t func, void *args);
29026 +int au_wkq_nowait(au_wkq_func_t func, void *args, struct super_block *sb,
29027 + unsigned int flags);
29028 +void au_nwt_init(struct au_nowait_tasks *nwt);
29029 +int __init au_wkq_init(void);
29030 +void au_wkq_fin(void);
29032 +/* ---------------------------------------------------------------------- */
29034 +static inline int au_wkq_test(void)
29036 + return current->flags & PF_WQ_WORKER;
29039 +static inline int au_wkq_wait(au_wkq_func_t func, void *args)
29041 + return au_wkq_do_wait(AuWkq_WAIT, func, args);
29044 +static inline void au_nwt_done(struct au_nowait_tasks *nwt)
29046 + if (atomic_dec_and_test(&nwt->nw_len))
29047 + wake_up_all(&nwt->nw_wq);
29050 +static inline int au_nwt_flush(struct au_nowait_tasks *nwt)
29052 + wait_event(nwt->nw_wq, !atomic_read(&nwt->nw_len));
29056 +#endif /* __KERNEL__ */
29057 +#endif /* __AUFS_WKQ_H__ */
29058 diff -urN /usr/share/empty/fs/aufs/xino.c linux/fs/aufs/xino.c
29059 --- /usr/share/empty/fs/aufs/xino.c 1970-01-01 01:00:00.000000000 +0100
29060 +++ linux/fs/aufs/xino.c 2012-01-05 11:00:59.218835185 +0100
29063 + * Copyright (C) 2005-2011 Junjiro R. Okajima
29065 + * This program, aufs is free software; you can redistribute it and/or modify
29066 + * it under the terms of the GNU General Public License as published by
29067 + * the Free Software Foundation; either version 2 of the License, or
29068 + * (at your option) any later version.
29070 + * This program is distributed in the hope that it will be useful,
29071 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
29072 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29073 + * GNU General Public License for more details.
29075 + * You should have received a copy of the GNU General Public License
29076 + * along with this program; if not, write to the Free Software
29077 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
29081 + * external inode number translation table and bitmap
29084 +#include <linux/file.h>
29085 +#include <linux/seq_file.h>
29086 +#include <linux/uaccess.h>
29089 +/* todo: unnecessary to support mmap_sem since kernel-space? */
29090 +ssize_t xino_fread(au_readf_t func, struct file *file, void *kbuf, size_t size,
29094 + mm_segment_t oldfs;
29101 + oldfs = get_fs();
29102 + set_fs(KERNEL_DS);
29104 + /* todo: signal_pending? */
29105 + err = func(file, buf.u, size, pos);
29106 + } while (err == -EAGAIN || err == -EINTR);
29109 +#if 0 /* reserved for future use */
29111 + fsnotify_access(file->f_dentry);
29117 +/* ---------------------------------------------------------------------- */
29119 +static ssize_t do_xino_fwrite(au_writef_t func, struct file *file, void *kbuf,
29120 + size_t size, loff_t *pos)
29123 + mm_segment_t oldfs;
29126 + const char __user *u;
29130 + oldfs = get_fs();
29131 + set_fs(KERNEL_DS);
29133 + /* todo: signal_pending? */
29134 + err = func(file, buf.u, size, pos);
29135 + } while (err == -EAGAIN || err == -EINTR);
29138 +#if 0 /* reserved for future use */
29140 + fsnotify_modify(file->f_dentry);
29146 +struct do_xino_fwrite_args {
29148 + au_writef_t func;
29149 + struct file *file;
29155 +static void call_do_xino_fwrite(void *args)
29157 + struct do_xino_fwrite_args *a = args;
29158 + *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos);
29161 +ssize_t xino_fwrite(au_writef_t func, struct file *file, void *buf, size_t size,
29166 + /* todo: signal block and no wkq? */
29167 + if (rlimit(RLIMIT_FSIZE) == RLIM_INFINITY) {
29169 + err = do_xino_fwrite(func, file, buf, size, pos);
29173 + * it breaks RLIMIT_FSIZE and normal user's limit,
29174 + * users should care about quota and real 'filesystem full.'
29177 + struct do_xino_fwrite_args args = {
29186 + wkq_err = au_wkq_wait(call_do_xino_fwrite, &args);
29187 + if (unlikely(wkq_err))
29194 +/* ---------------------------------------------------------------------- */
29197 + * create a new xinofile at the same place/path as @base_file.
29199 +struct file *au_xino_create2(struct file *base_file, struct file *copy_src)
29201 + struct file *file;
29202 + struct dentry *base, *parent;
29203 + struct inode *dir;
29204 + struct qstr *name;
29205 + struct path path;
29208 + base = base_file->f_dentry;
29209 + parent = base->d_parent; /* dir inode is locked */
29210 + dir = parent->d_inode;
29213 + file = ERR_PTR(-EINVAL);
29214 + name = &base->d_name;
29215 + path.dentry = vfsub_lookup_one_len(name->name, parent, name->len);
29216 + if (IS_ERR(path.dentry)) {
29217 + file = (void *)path.dentry;
29218 + pr_err("%.*s lookup err %ld\n",
29219 + AuLNPair(name), PTR_ERR(path.dentry));
29223 + /* no need to mnt_want_write() since we call dentry_open() later */
29224 + err = vfs_create(dir, path.dentry, S_IRUGO | S_IWUGO, NULL);
29225 + if (unlikely(err)) {
29226 + file = ERR_PTR(err);
29227 + pr_err("%.*s create err %d\n", AuLNPair(name), err);
29231 + path.mnt = base_file->f_vfsmnt;
29232 + file = vfsub_dentry_open(&path,
29233 + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
29234 + /* | __FMODE_NONOTIFY */);
29235 + if (IS_ERR(file)) {
29236 + pr_err("%.*s open err %ld\n", AuLNPair(name), PTR_ERR(file));
29240 + err = vfsub_unlink(dir, &file->f_path, /*force*/0);
29241 + if (unlikely(err)) {
29242 + pr_err("%.*s unlink err %d\n", AuLNPair(name), err);
29247 + /* no one can touch copy_src xino */
29248 + err = au_copy_file(file, copy_src,
29249 + i_size_read(copy_src->f_dentry->d_inode));
29250 + if (unlikely(err)) {
29251 + pr_err("%.*s copy err %d\n", AuLNPair(name), err);
29255 + goto out_dput; /* success */
29259 + file = ERR_PTR(err);
29261 + dput(path.dentry);
29266 +struct au_xino_lock_dir {
29267 + struct au_hinode *hdir;
29268 + struct dentry *parent;
29269 + struct mutex *mtx;
29272 +static void au_xino_lock_dir(struct super_block *sb, struct file *xino,
29273 + struct au_xino_lock_dir *ldir)
29275 + aufs_bindex_t brid, bindex;
29277 + ldir->hdir = NULL;
29279 + brid = au_xino_brid(sb);
29281 + bindex = au_br_index(sb, brid);
29282 + if (bindex >= 0) {
29283 + ldir->hdir = au_hi(sb->s_root->d_inode, bindex);
29284 + au_hn_imtx_lock_nested(ldir->hdir, AuLsc_I_PARENT);
29286 + ldir->parent = dget_parent(xino->f_dentry);
29287 + ldir->mtx = &ldir->parent->d_inode->i_mutex;
29288 + mutex_lock_nested(ldir->mtx, AuLsc_I_PARENT);
29292 +static void au_xino_unlock_dir(struct au_xino_lock_dir *ldir)
29295 + au_hn_imtx_unlock(ldir->hdir);
29297 + mutex_unlock(ldir->mtx);
29298 + dput(ldir->parent);
29302 +/* ---------------------------------------------------------------------- */
29304 +/* trucate xino files asynchronously */
29306 +int au_xino_trunc(struct super_block *sb, aufs_bindex_t bindex)
29309 + aufs_bindex_t bi, bend;
29310 + struct au_branch *br;
29311 + struct file *new_xino, *file;
29312 + struct super_block *h_sb;
29313 + struct au_xino_lock_dir ldir;
29316 + bend = au_sbend(sb);
29317 + if (unlikely(bindex < 0 || bend < bindex))
29319 + br = au_sbr(sb, bindex);
29320 + file = br->br_xino.xi_file;
29324 + au_xino_lock_dir(sb, file, &ldir);
29325 + /* mnt_want_write() is unnecessary here */
29326 + new_xino = au_xino_create2(file, file);
29327 + au_xino_unlock_dir(&ldir);
29328 + err = PTR_ERR(new_xino);
29329 + if (IS_ERR(new_xino))
29333 + br->br_xino.xi_file = new_xino;
29335 + h_sb = br->br_mnt->mnt_sb;
29336 + for (bi = 0; bi <= bend; bi++) {
29337 + if (unlikely(bi == bindex))
29339 + br = au_sbr(sb, bi);
29340 + if (br->br_mnt->mnt_sb != h_sb)
29343 + fput(br->br_xino.xi_file);
29344 + br->br_xino.xi_file = new_xino;
29345 + get_file(new_xino);
29352 +struct xino_do_trunc_args {
29353 + struct super_block *sb;
29354 + struct au_branch *br;
29357 +static void xino_do_trunc(void *_args)
29359 + struct xino_do_trunc_args *args = _args;
29360 + struct super_block *sb;
29361 + struct au_branch *br;
29362 + struct inode *dir;
29364 + aufs_bindex_t bindex;
29368 + dir = sb->s_root->d_inode;
29371 + si_noflush_write_lock(sb);
29372 + ii_read_lock_parent(dir);
29373 + bindex = au_br_index(sb, br->br_id);
29374 + err = au_xino_trunc(sb, bindex);
29376 + && br->br_xino.xi_file->f_dentry->d_inode->i_blocks
29377 + >= br->br_xino_upper)
29378 + br->br_xino_upper += AUFS_XINO_TRUNC_STEP;
29380 + ii_read_unlock(dir);
29381 + if (unlikely(err))
29382 + pr_warning("err b%d, (%d)\n", bindex, err);
29383 + atomic_dec(&br->br_xino_running);
29384 + atomic_dec(&br->br_count);
29385 + si_write_unlock(sb);
29386 + au_nwt_done(&au_sbi(sb)->si_nowait);
29390 +static void xino_try_trunc(struct super_block *sb, struct au_branch *br)
29392 + struct xino_do_trunc_args *args;
29395 + if (br->br_xino.xi_file->f_dentry->d_inode->i_blocks
29396 + < br->br_xino_upper)
29399 + if (atomic_inc_return(&br->br_xino_running) > 1)
29402 + /* lock and kfree() will be called in trunc_xino() */
29403 + args = kmalloc(sizeof(*args), GFP_NOFS);
29404 + if (unlikely(!args)) {
29405 + AuErr1("no memory\n");
29409 + atomic_inc(&br->br_count);
29412 + wkq_err = au_wkq_nowait(xino_do_trunc, args, sb, /*flags*/0);
29414 + return; /* success */
29416 + pr_err("wkq %d\n", wkq_err);
29417 + atomic_dec(&br->br_count);
29422 + atomic_dec(&br->br_xino_running);
29425 +/* ---------------------------------------------------------------------- */
29427 +static int au_xino_do_write(au_writef_t write, struct file *file,
29428 + ino_t h_ino, ino_t ino)
29434 + if (unlikely(au_loff_max / sizeof(ino) - 1 < pos)) {
29435 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
29438 + pos *= sizeof(ino);
29439 + sz = xino_fwrite(write, file, &ino, sizeof(ino), &pos);
29440 + if (sz == sizeof(ino))
29441 + return 0; /* success */
29443 + AuIOErr("write failed (%zd)\n", sz);
29448 + * write @ino to the xinofile for the specified branch{@sb, @bindex}
29449 + * at the position of @h_ino.
29450 + * even if @ino is zero, it is written to the xinofile and means no entry.
29451 + * if the size of the xino file on a specific filesystem exceeds the watermark,
29452 + * try truncating it.
29454 +int au_xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
29458 + unsigned int mnt_flags;
29459 + struct au_branch *br;
29461 + BUILD_BUG_ON(sizeof(long long) != sizeof(au_loff_max)
29462 + || ((loff_t)-1) > 0);
29463 + SiMustAnyLock(sb);
29465 + mnt_flags = au_mntflags(sb);
29466 + if (!au_opt_test(mnt_flags, XINO))
29469 + br = au_sbr(sb, bindex);
29470 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
29473 + if (au_opt_test(mnt_flags, TRUNC_XINO)
29474 + && au_test_fs_trunc_xino(br->br_mnt->mnt_sb))
29475 + xino_try_trunc(sb, br);
29476 + return 0; /* success */
29479 + AuIOErr("write failed (%d)\n", err);
29483 +/* ---------------------------------------------------------------------- */
29485 +/* aufs inode number bitmap */
29487 +static const int page_bits = (int)PAGE_SIZE * BITS_PER_BYTE;
29488 +static ino_t xib_calc_ino(unsigned long pindex, int bit)
29492 + AuDebugOn(bit < 0 || page_bits <= bit);
29493 + ino = AUFS_FIRST_INO + pindex * page_bits + bit;
29497 +static void xib_calc_bit(ino_t ino, unsigned long *pindex, int *bit)
29499 + AuDebugOn(ino < AUFS_FIRST_INO);
29500 + ino -= AUFS_FIRST_INO;
29501 + *pindex = ino / page_bits;
29502 + *bit = ino % page_bits;
29505 +static int xib_pindex(struct super_block *sb, unsigned long pindex)
29510 + struct au_sbinfo *sbinfo;
29511 + struct file *xib;
29512 + unsigned long *p;
29514 + sbinfo = au_sbi(sb);
29515 + MtxMustLock(&sbinfo->si_xib_mtx);
29516 + AuDebugOn(pindex > ULONG_MAX / PAGE_SIZE
29517 + || !au_opt_test(sbinfo->si_mntflags, XINO));
29519 + if (pindex == sbinfo->si_xib_last_pindex)
29522 + xib = sbinfo->si_xib;
29523 + p = sbinfo->si_xib_buf;
29524 + pos = sbinfo->si_xib_last_pindex;
29525 + pos *= PAGE_SIZE;
29526 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
29527 + if (unlikely(sz != PAGE_SIZE))
29531 + pos *= PAGE_SIZE;
29532 + if (i_size_read(xib->f_dentry->d_inode) >= pos + PAGE_SIZE)
29533 + sz = xino_fread(sbinfo->si_xread, xib, p, PAGE_SIZE, &pos);
29535 + memset(p, 0, PAGE_SIZE);
29536 + sz = xino_fwrite(sbinfo->si_xwrite, xib, p, PAGE_SIZE, &pos);
29538 + if (sz == PAGE_SIZE) {
29539 + sbinfo->si_xib_last_pindex = pindex;
29540 + return 0; /* success */
29544 + AuIOErr1("write failed (%zd)\n", sz);
29551 +/* ---------------------------------------------------------------------- */
29553 +static void au_xib_clear_bit(struct inode *inode)
29556 + unsigned long pindex;
29557 + struct super_block *sb;
29558 + struct au_sbinfo *sbinfo;
29560 + AuDebugOn(inode->i_nlink);
29562 + sb = inode->i_sb;
29563 + xib_calc_bit(inode->i_ino, &pindex, &bit);
29564 + AuDebugOn(page_bits <= bit);
29565 + sbinfo = au_sbi(sb);
29566 + mutex_lock(&sbinfo->si_xib_mtx);
29567 + err = xib_pindex(sb, pindex);
29569 + clear_bit(bit, sbinfo->si_xib_buf);
29570 + sbinfo->si_xib_next_bit = bit;
29572 + mutex_unlock(&sbinfo->si_xib_mtx);
29575 +/* for s_op->delete_inode() */
29576 +void au_xino_delete_inode(struct inode *inode, const int unlinked)
29579 + unsigned int mnt_flags;
29580 + aufs_bindex_t bindex, bend, bi;
29581 + unsigned char try_trunc;
29582 + struct au_iinfo *iinfo;
29583 + struct super_block *sb;
29584 + struct au_hinode *hi;
29585 + struct inode *h_inode;
29586 + struct au_branch *br;
29587 + au_writef_t xwrite;
29589 + sb = inode->i_sb;
29590 + mnt_flags = au_mntflags(sb);
29591 + if (!au_opt_test(mnt_flags, XINO)
29592 + || inode->i_ino == AUFS_ROOT_INO)
29596 + au_xigen_inc(inode);
29597 + au_xib_clear_bit(inode);
29600 + iinfo = au_ii(inode);
29604 + bindex = iinfo->ii_bstart;
29608 + xwrite = au_sbi(sb)->si_xwrite;
29609 + try_trunc = !!au_opt_test(mnt_flags, TRUNC_XINO);
29610 + hi = iinfo->ii_hinode + bindex;
29611 + bend = iinfo->ii_bend;
29612 + for (; bindex <= bend; bindex++, hi++) {
29613 + h_inode = hi->hi_inode;
29615 + || (!unlinked && h_inode->i_nlink))
29618 + /* inode may not be revalidated */
29619 + bi = au_br_index(sb, hi->hi_id);
29623 + br = au_sbr(sb, bi);
29624 + err = au_xino_do_write(xwrite, br->br_xino.xi_file,
29625 + h_inode->i_ino, /*ino*/0);
29626 + if (!err && try_trunc
29627 + && au_test_fs_trunc_xino(br->br_mnt->mnt_sb))
29628 + xino_try_trunc(sb, br);
29632 +/* get an unused inode number from bitmap */
29633 +ino_t au_xino_new_ino(struct super_block *sb)
29636 + unsigned long *p, pindex, ul, pend;
29637 + struct au_sbinfo *sbinfo;
29638 + struct file *file;
29639 + int free_bit, err;
29641 + if (!au_opt_test(au_mntflags(sb), XINO))
29642 + return iunique(sb, AUFS_FIRST_INO);
29644 + sbinfo = au_sbi(sb);
29645 + mutex_lock(&sbinfo->si_xib_mtx);
29646 + p = sbinfo->si_xib_buf;
29647 + free_bit = sbinfo->si_xib_next_bit;
29648 + if (free_bit < page_bits && !test_bit(free_bit, p))
29649 + goto out; /* success */
29650 + free_bit = find_first_zero_bit(p, page_bits);
29651 + if (free_bit < page_bits)
29652 + goto out; /* success */
29654 + pindex = sbinfo->si_xib_last_pindex;
29655 + for (ul = pindex - 1; ul < ULONG_MAX; ul--) {
29656 + err = xib_pindex(sb, ul);
29657 + if (unlikely(err))
29659 + free_bit = find_first_zero_bit(p, page_bits);
29660 + if (free_bit < page_bits)
29661 + goto out; /* success */
29664 + file = sbinfo->si_xib;
29665 + pend = i_size_read(file->f_dentry->d_inode) / PAGE_SIZE;
29666 + for (ul = pindex + 1; ul <= pend; ul++) {
29667 + err = xib_pindex(sb, ul);
29668 + if (unlikely(err))
29670 + free_bit = find_first_zero_bit(p, page_bits);
29671 + if (free_bit < page_bits)
29672 + goto out; /* success */
29677 + set_bit(free_bit, p);
29678 + sbinfo->si_xib_next_bit = free_bit + 1;
29679 + pindex = sbinfo->si_xib_last_pindex;
29680 + mutex_unlock(&sbinfo->si_xib_mtx);
29681 + ino = xib_calc_ino(pindex, free_bit);
29682 + AuDbg("i%lu\n", (unsigned long)ino);
29685 + mutex_unlock(&sbinfo->si_xib_mtx);
29691 + * read @ino from xinofile for the specified branch{@sb, @bindex}
29692 + * at the position of @h_ino.
29693 + * if @ino does not exist and @do_new is true, get new one.
29695 +int au_xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino,
29701 + struct file *file;
29702 + struct au_sbinfo *sbinfo;
29705 + if (!au_opt_test(au_mntflags(sb), XINO))
29706 + return 0; /* no xino */
29709 + sbinfo = au_sbi(sb);
29711 + if (unlikely(au_loff_max / sizeof(*ino) - 1 < pos)) {
29712 + AuIOErr1("too large hi%lu\n", (unsigned long)h_ino);
29715 + pos *= sizeof(*ino);
29717 + file = au_sbr(sb, bindex)->br_xino.xi_file;
29718 + if (i_size_read(file->f_dentry->d_inode) < pos + sizeof(*ino))
29719 + return 0; /* no ino */
29721 + sz = xino_fread(sbinfo->si_xread, file, ino, sizeof(*ino), &pos);
29722 + if (sz == sizeof(*ino))
29723 + return 0; /* success */
29726 + if (unlikely(sz >= 0)) {
29728 + AuIOErr("xino read error (%zd)\n", sz);
29734 +/* ---------------------------------------------------------------------- */
29736 +/* create and set a new xino file */
29738 +struct file *au_xino_create(struct super_block *sb, char *fname, int silent)
29740 + struct file *file;
29741 + struct dentry *h_parent, *d;
29742 + struct inode *h_dir;
29746 + * at mount-time, and the xino file is the default path,
29747 + * hnotify is disabled so we have no notify events to ignore.
29748 + * when a user specified the xino, we cannot get au_hdir to be ignored.
29750 + file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE
29751 + /* | __FMODE_NONOTIFY */,
29752 + S_IRUGO | S_IWUGO);
29753 + if (IS_ERR(file)) {
29755 + pr_err("open %s(%ld)\n", fname, PTR_ERR(file));
29759 + /* keep file count */
29760 + h_parent = dget_parent(file->f_dentry);
29761 + h_dir = h_parent->d_inode;
29762 + mutex_lock_nested(&h_dir->i_mutex, AuLsc_I_PARENT);
29763 + /* mnt_want_write() is unnecessary here */
29764 + err = vfsub_unlink(h_dir, &file->f_path, /*force*/0);
29765 + mutex_unlock(&h_dir->i_mutex);
29767 + if (unlikely(err)) {
29769 + pr_err("unlink %s(%d)\n", fname, err);
29774 + d = file->f_dentry;
29775 + if (unlikely(sb == d->d_sb)) {
29777 + pr_err("%s must be outside\n", fname);
29780 + if (unlikely(au_test_fs_bad_xino(d->d_sb))) {
29782 + pr_err("xino doesn't support %s(%s)\n",
29783 + fname, au_sbtype(d->d_sb));
29786 + return file; /* success */
29790 + file = ERR_PTR(err);
29795 + * find another branch who is on the same filesystem of the specified
29796 + * branch{@btgt}. search until @bend.
29798 +static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt,
29799 + aufs_bindex_t bend)
29801 + aufs_bindex_t bindex;
29802 + struct super_block *tgt_sb = au_sbr_sb(sb, btgt);
29804 + for (bindex = 0; bindex < btgt; bindex++)
29805 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
29807 + for (bindex++; bindex <= bend; bindex++)
29808 + if (unlikely(tgt_sb == au_sbr_sb(sb, bindex)))
29813 +/* ---------------------------------------------------------------------- */
29816 + * initialize the xinofile for the specified branch @br
29817 + * at the place/path where @base_file indicates.
29818 + * test whether another branch is on the same filesystem or not,
29819 + * if @do_test is true.
29821 +int au_xino_br(struct super_block *sb, struct au_branch *br, ino_t h_ino,
29822 + struct file *base_file, int do_test)
29826 + aufs_bindex_t bend, bindex;
29827 + struct au_branch *shared_br, *b;
29828 + struct file *file;
29829 + struct super_block *tgt_sb;
29831 + shared_br = NULL;
29832 + bend = au_sbend(sb);
29834 + tgt_sb = br->br_mnt->mnt_sb;
29835 + for (bindex = 0; bindex <= bend; bindex++) {
29836 + b = au_sbr(sb, bindex);
29837 + if (tgt_sb == b->br_mnt->mnt_sb) {
29844 + if (!shared_br || !shared_br->br_xino.xi_file) {
29845 + struct au_xino_lock_dir ldir;
29847 + au_xino_lock_dir(sb, base_file, &ldir);
29848 + /* mnt_want_write() is unnecessary here */
29849 + file = au_xino_create2(base_file, NULL);
29850 + au_xino_unlock_dir(&ldir);
29851 + err = PTR_ERR(file);
29852 + if (IS_ERR(file))
29854 + br->br_xino.xi_file = file;
29856 + br->br_xino.xi_file = shared_br->br_xino.xi_file;
29857 + get_file(br->br_xino.xi_file);
29860 + ino = AUFS_ROOT_INO;
29861 + err = au_xino_do_write(au_sbi(sb)->si_xwrite, br->br_xino.xi_file,
29863 + if (unlikely(err)) {
29864 + fput(br->br_xino.xi_file);
29865 + br->br_xino.xi_file = NULL;
29872 +/* ---------------------------------------------------------------------- */
29874 +/* trucate a xino bitmap file */
29877 +static int do_xib_restore(struct super_block *sb, struct file *file, void *page)
29881 + unsigned long pindex;
29882 + loff_t pos, pend;
29883 + struct au_sbinfo *sbinfo;
29886 + unsigned long *p;
29889 + sbinfo = au_sbi(sb);
29890 + MtxMustLock(&sbinfo->si_xib_mtx);
29891 + p = sbinfo->si_xib_buf;
29892 + func = sbinfo->si_xread;
29893 + pend = i_size_read(file->f_dentry->d_inode);
29895 + while (pos < pend) {
29896 + sz = xino_fread(func, file, page, PAGE_SIZE, &pos);
29898 + if (unlikely(sz <= 0))
29902 + for (ino = page; sz > 0; ino++, sz -= sizeof(ino)) {
29903 + if (unlikely(*ino < AUFS_FIRST_INO))
29906 + xib_calc_bit(*ino, &pindex, &bit);
29907 + AuDebugOn(page_bits <= bit);
29908 + err = xib_pindex(sb, pindex);
29920 +static int xib_restore(struct super_block *sb)
29923 + aufs_bindex_t bindex, bend;
29927 + page = (void *)__get_free_page(GFP_NOFS);
29928 + if (unlikely(!page))
29932 + bend = au_sbend(sb);
29933 + for (bindex = 0; !err && bindex <= bend; bindex++)
29934 + if (!bindex || is_sb_shared(sb, bindex, bindex - 1) < 0)
29935 + err = do_xib_restore
29936 + (sb, au_sbr(sb, bindex)->br_xino.xi_file, page);
29938 + AuDbg("b%d\n", bindex);
29939 + free_page((unsigned long)page);
29945 +int au_xib_trunc(struct super_block *sb)
29950 + struct au_xino_lock_dir ldir;
29951 + struct au_sbinfo *sbinfo;
29952 + unsigned long *p;
29953 + struct file *file;
29955 + SiMustWriteLock(sb);
29958 + sbinfo = au_sbi(sb);
29959 + if (!au_opt_test(sbinfo->si_mntflags, XINO))
29962 + file = sbinfo->si_xib;
29963 + if (i_size_read(file->f_dentry->d_inode) <= PAGE_SIZE)
29966 + au_xino_lock_dir(sb, file, &ldir);
29967 + /* mnt_want_write() is unnecessary here */
29968 + file = au_xino_create2(sbinfo->si_xib, NULL);
29969 + au_xino_unlock_dir(&ldir);
29970 + err = PTR_ERR(file);
29971 + if (IS_ERR(file))
29973 + fput(sbinfo->si_xib);
29974 + sbinfo->si_xib = file;
29976 + p = sbinfo->si_xib_buf;
29977 + memset(p, 0, PAGE_SIZE);
29979 + sz = xino_fwrite(sbinfo->si_xwrite, sbinfo->si_xib, p, PAGE_SIZE, &pos);
29980 + if (unlikely(sz != PAGE_SIZE)) {
29982 + AuIOErr("err %d\n", err);
29988 + mutex_lock(&sbinfo->si_xib_mtx);
29989 + /* mnt_want_write() is unnecessary here */
29990 + err = xib_restore(sb);
29991 + mutex_unlock(&sbinfo->si_xib_mtx);
29997 +/* ---------------------------------------------------------------------- */
30000 + * xino mount option handlers
30002 +static au_readf_t find_readf(struct file *h_file)
30004 + const struct file_operations *fop = h_file->f_op;
30008 + return fop->read;
30009 + if (fop->aio_read)
30010 + return do_sync_read;
30012 + return ERR_PTR(-ENOSYS);
30015 +static au_writef_t find_writef(struct file *h_file)
30017 + const struct file_operations *fop = h_file->f_op;
30021 + return fop->write;
30022 + if (fop->aio_write)
30023 + return do_sync_write;
30025 + return ERR_PTR(-ENOSYS);
30029 +static void xino_clear_xib(struct super_block *sb)
30031 + struct au_sbinfo *sbinfo;
30033 + SiMustWriteLock(sb);
30035 + sbinfo = au_sbi(sb);
30036 + sbinfo->si_xread = NULL;
30037 + sbinfo->si_xwrite = NULL;
30038 + if (sbinfo->si_xib)
30039 + fput(sbinfo->si_xib);
30040 + sbinfo->si_xib = NULL;
30041 + free_page((unsigned long)sbinfo->si_xib_buf);
30042 + sbinfo->si_xib_buf = NULL;
30045 +static int au_xino_set_xib(struct super_block *sb, struct file *base)
30049 + struct au_sbinfo *sbinfo;
30050 + struct file *file;
30052 + SiMustWriteLock(sb);
30054 + sbinfo = au_sbi(sb);
30055 + file = au_xino_create2(base, sbinfo->si_xib);
30056 + err = PTR_ERR(file);
30057 + if (IS_ERR(file))
30059 + if (sbinfo->si_xib)
30060 + fput(sbinfo->si_xib);
30061 + sbinfo->si_xib = file;
30062 + sbinfo->si_xread = find_readf(file);
30063 + sbinfo->si_xwrite = find_writef(file);
30066 + if (!sbinfo->si_xib_buf)
30067 + sbinfo->si_xib_buf = (void *)get_zeroed_page(GFP_NOFS);
30068 + if (unlikely(!sbinfo->si_xib_buf))
30071 + sbinfo->si_xib_last_pindex = 0;
30072 + sbinfo->si_xib_next_bit = 0;
30073 + if (i_size_read(file->f_dentry->d_inode) < PAGE_SIZE) {
30075 + err = xino_fwrite(sbinfo->si_xwrite, file, sbinfo->si_xib_buf,
30076 + PAGE_SIZE, &pos);
30077 + if (unlikely(err != PAGE_SIZE))
30081 + goto out; /* success */
30084 + free_page((unsigned long)sbinfo->si_xib_buf);
30085 + sbinfo->si_xib_buf = NULL;
30089 + fput(sbinfo->si_xib);
30090 + sbinfo->si_xib = NULL;
30091 + sbinfo->si_xread = NULL;
30092 + sbinfo->si_xwrite = NULL;
30097 +/* xino for each branch */
30098 +static void xino_clear_br(struct super_block *sb)
30100 + aufs_bindex_t bindex, bend;
30101 + struct au_branch *br;
30103 + bend = au_sbend(sb);
30104 + for (bindex = 0; bindex <= bend; bindex++) {
30105 + br = au_sbr(sb, bindex);
30106 + if (!br || !br->br_xino.xi_file)
30109 + fput(br->br_xino.xi_file);
30110 + br->br_xino.xi_file = NULL;
30114 +static int au_xino_set_br(struct super_block *sb, struct file *base)
30118 + aufs_bindex_t bindex, bend, bshared;
30120 + struct file *old, *new;
30122 + struct au_branch *br;
30123 + struct inode *inode;
30124 + au_writef_t writef;
30126 + SiMustWriteLock(sb);
30129 + bend = au_sbend(sb);
30130 + fpair = kcalloc(bend + 1, sizeof(*fpair), GFP_NOFS);
30131 + if (unlikely(!fpair))
30134 + inode = sb->s_root->d_inode;
30135 + ino = AUFS_ROOT_INO;
30136 + writef = au_sbi(sb)->si_xwrite;
30137 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
30138 + br = au_sbr(sb, bindex);
30139 + bshared = is_sb_shared(sb, bindex, bindex - 1);
30140 + if (bshared >= 0) {
30141 + /* shared xino */
30142 + *p = fpair[bshared];
30143 + get_file(p->new);
30148 + p->old = br->br_xino.xi_file;
30149 + p->new = au_xino_create2(base, br->br_xino.xi_file);
30150 + err = PTR_ERR(p->new);
30151 + if (IS_ERR(p->new)) {
30157 + err = au_xino_do_write(writef, p->new,
30158 + au_h_iptr(inode, bindex)->i_ino, ino);
30159 + if (unlikely(err))
30163 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++) {
30164 + br = au_sbr(sb, bindex);
30165 + if (br->br_xino.xi_file)
30166 + fput(br->br_xino.xi_file);
30167 + get_file(p->new);
30168 + br->br_xino.xi_file = p->new;
30172 + for (bindex = 0, p = fpair; bindex <= bend; bindex++, p++)
30182 +void au_xino_clr(struct super_block *sb)
30184 + struct au_sbinfo *sbinfo;
30186 + au_xigen_clr(sb);
30187 + xino_clear_xib(sb);
30188 + xino_clear_br(sb);
30189 + sbinfo = au_sbi(sb);
30190 + /* lvalue, do not call au_mntflags() */
30191 + au_opt_clr(sbinfo->si_mntflags, XINO);
30194 +int au_xino_set(struct super_block *sb, struct au_opt_xino *xino, int remount)
30197 + struct dentry *parent, *cur_parent;
30198 + struct qstr *dname, *cur_name;
30199 + struct file *cur_xino;
30200 + struct inode *dir;
30201 + struct au_sbinfo *sbinfo;
30203 + SiMustWriteLock(sb);
30206 + sbinfo = au_sbi(sb);
30207 + parent = dget_parent(xino->file->f_dentry);
30210 + dname = &xino->file->f_dentry->d_name;
30211 + cur_xino = sbinfo->si_xib;
30213 + cur_parent = dget_parent(cur_xino->f_dentry);
30214 + cur_name = &cur_xino->f_dentry->d_name;
30215 + skip = (cur_parent == parent
30216 + && dname->len == cur_name->len
30217 + && !memcmp(dname->name, cur_name->name,
30219 + dput(cur_parent);
30225 + au_opt_set(sbinfo->si_mntflags, XINO);
30226 + dir = parent->d_inode;
30227 + mutex_lock_nested(&dir->i_mutex, AuLsc_I_PARENT);
30228 + /* mnt_want_write() is unnecessary here */
30229 + err = au_xino_set_xib(sb, xino->file);
30231 + err = au_xigen_set(sb, xino->file);
30233 + err = au_xino_set_br(sb, xino->file);
30234 + mutex_unlock(&dir->i_mutex);
30236 + goto out; /* success */
30239 + AuIOErr("failed creating xino(%d).\n", err);
30246 +/* ---------------------------------------------------------------------- */
30249 + * create a xinofile at the default place/path.
30251 +struct file *au_xino_def(struct super_block *sb)
30253 + struct file *file;
30255 + struct au_branch *br;
30256 + struct super_block *h_sb;
30257 + struct path path;
30258 + aufs_bindex_t bend, bindex, bwr;
30261 + bend = au_sbend(sb);
30263 + for (bindex = 0; bindex <= bend; bindex++) {
30264 + br = au_sbr(sb, bindex);
30265 + if (au_br_writable(br->br_perm)
30266 + && !au_test_fs_bad_xino(br->br_mnt->mnt_sb)) {
30273 + file = ERR_PTR(-ENOMEM);
30274 + page = __getname_gfp(GFP_NOFS);
30275 + if (unlikely(!page))
30277 + path.mnt = br->br_mnt;
30278 + path.dentry = au_h_dptr(sb->s_root, bwr);
30279 + p = d_path(&path, page, PATH_MAX - sizeof(AUFS_XINO_FNAME));
30280 + file = (void *)p;
30281 + if (!IS_ERR(p)) {
30282 + strcat(p, "/" AUFS_XINO_FNAME);
30283 + AuDbg("%s\n", p);
30284 + file = au_xino_create(sb, p, /*silent*/0);
30285 + if (!IS_ERR(file))
30286 + au_xino_brid_set(sb, br->br_id);
30290 + file = au_xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0);
30291 + if (IS_ERR(file))
30293 + h_sb = file->f_dentry->d_sb;
30294 + if (unlikely(au_test_fs_bad_xino(h_sb))) {
30295 + pr_err("xino doesn't support %s(%s)\n",
30296 + AUFS_XINO_DEFPATH, au_sbtype(h_sb));
30298 + file = ERR_PTR(-EINVAL);
30300 + if (!IS_ERR(file))
30301 + au_xino_brid_set(sb, -1);
30308 +/* ---------------------------------------------------------------------- */
30310 +int au_xino_path(struct seq_file *seq, struct file *file)
30314 + err = au_seq_path(seq, &file->f_path);
30315 + if (unlikely(err < 0))
30319 +#define Deleted "\\040(deleted)"
30320 + seq->count -= sizeof(Deleted) - 1;
30321 + AuDebugOn(memcmp(seq->buf + seq->count, Deleted,
30322 + sizeof(Deleted) - 1));
30328 diff -urN /usr/share/empty/include/linux/aufs_type.h linux/include/linux/aufs_type.h
30329 --- /usr/share/empty/include/linux/aufs_type.h 1970-01-01 01:00:00.000000000 +0100
30330 +++ linux/include/linux/aufs_type.h 2012-01-05 11:00:59.218835185 +0100
30333 + * Copyright (C) 2005-2011 Junjiro R. Okajima
30335 + * This program, aufs is free software; you can redistribute it and/or modify
30336 + * it under the terms of the GNU General Public License as published by
30337 + * the Free Software Foundation; either version 2 of the License, or
30338 + * (at your option) any later version.
30340 + * This program is distributed in the hope that it will be useful,
30341 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
30342 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30343 + * GNU General Public License for more details.
30345 + * You should have received a copy of the GNU General Public License
30346 + * along with this program; if not, write to the Free Software
30347 + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
30350 +#ifndef __AUFS_TYPE_H__
30351 +#define __AUFS_TYPE_H__
30353 +#include <linux/ioctl.h>
30354 +#include <linux/kernel.h>
30355 +#include <linux/limits.h>
30357 +#include <linux/types.h>
30359 +#include <stdint.h>
30360 +#include <sys/types.h>
30363 +#define AUFS_VERSION "3.x-rcN-20111205"
30365 +/* todo? move this to linux-2.6.19/include/magic.h */
30366 +#define AUFS_SUPER_MAGIC ('a' << 24 | 'u' << 16 | 'f' << 8 | 's')
30368 +/* ---------------------------------------------------------------------- */
30370 +#ifdef CONFIG_AUFS_BRANCH_MAX_127
30371 +typedef int8_t aufs_bindex_t;
30372 +#define AUFS_BRANCH_MAX 127
30374 +typedef int16_t aufs_bindex_t;
30375 +#ifdef CONFIG_AUFS_BRANCH_MAX_511
30376 +#define AUFS_BRANCH_MAX 511
30377 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023)
30378 +#define AUFS_BRANCH_MAX 1023
30379 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767)
30380 +#define AUFS_BRANCH_MAX 32767
30385 +#ifndef AUFS_BRANCH_MAX
30386 +#error unknown CONFIG_AUFS_BRANCH_MAX value
30388 +#endif /* __KERNEL__ */
30390 +/* ---------------------------------------------------------------------- */
30392 +#define AUFS_NAME "aufs"
30393 +#define AUFS_FSTYPE AUFS_NAME
30395 +#define AUFS_ROOT_INO 2
30396 +#define AUFS_FIRST_INO 11
30398 +#define AUFS_WH_PFX ".wh."
30399 +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1)
30400 +#define AUFS_WH_TMP_LEN 4
30401 +/* a limit for rmdir/rename a dir */
30402 +#define AUFS_MAX_NAMELEN (NAME_MAX \
30403 + - AUFS_WH_PFX_LEN * 2 /* doubly whiteouted */\
30405 + - AUFS_WH_TMP_LEN) /* hex */
30406 +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino"
30407 +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME
30408 +#define AUFS_XINO_TRUNC_INIT 64 /* blocks */
30409 +#define AUFS_XINO_TRUNC_STEP 4 /* blocks */
30410 +#define AUFS_DIRWH_DEF 3
30411 +#define AUFS_RDCACHE_DEF 10 /* seconds */
30412 +#define AUFS_RDCACHE_MAX 3600 /* seconds */
30413 +#define AUFS_RDBLK_DEF 512 /* bytes */
30414 +#define AUFS_RDHASH_DEF 32
30415 +#define AUFS_WKQ_NAME AUFS_NAME "d"
30416 +#define AUFS_MFS_DEF_SEC 30 /* seconds */
30417 +#define AUFS_MFS_MAX_SEC 3600 /* seconds */
30418 +#define AUFS_PLINK_WARN 100 /* number of plinks */
30420 +/* pseudo-link maintenace under /proc */
30421 +#define AUFS_PLINK_MAINT_NAME "plink_maint"
30422 +#define AUFS_PLINK_MAINT_DIR "fs/" AUFS_NAME
30423 +#define AUFS_PLINK_MAINT_PATH AUFS_PLINK_MAINT_DIR "/" AUFS_PLINK_MAINT_NAME
30425 +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */
30426 +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME
30428 +#define AUFS_BASE_NAME AUFS_WH_PFX AUFS_NAME
30429 +#define AUFS_PLINKDIR_NAME AUFS_WH_PFX "plnk"
30430 +#define AUFS_ORPHDIR_NAME AUFS_WH_PFX "orph"
30432 +/* doubly whiteouted */
30433 +#define AUFS_WH_BASE AUFS_WH_PFX AUFS_BASE_NAME
30434 +#define AUFS_WH_PLINKDIR AUFS_WH_PFX AUFS_PLINKDIR_NAME
30435 +#define AUFS_WH_ORPHDIR AUFS_WH_PFX AUFS_ORPHDIR_NAME
30437 +/* branch permissions and attributes */
30438 +#define AUFS_BRPERM_RW "rw"
30439 +#define AUFS_BRPERM_RO "ro"
30440 +#define AUFS_BRPERM_RR "rr"
30441 +#define AUFS_BRRATTR_WH "wh"
30442 +#define AUFS_BRWATTR_NLWH "nolwh"
30444 +/* ---------------------------------------------------------------------- */
30448 + /* readdir in userspace */
30452 + /* pathconf wrapper */
30459 +/* borrowed from linux/include/linux/kernel.h */
30461 +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
30462 +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
30465 +/* borrowed from linux/include/linux/compiler-gcc3.h */
30467 +#define __aligned(x) __attribute__((aligned(x)))
30472 +#define __packed __attribute__((packed))
30476 +struct au_rdu_cookie {
30481 + uint32_t generation;
30484 +struct au_rdu_ent {
30493 +static inline int au_rdu_len(int nlen)
30495 + /* include the terminating NULL */
30496 + return ALIGN(sizeof(struct au_rdu_ent) + nlen + 1,
30497 + sizeof(uint64_t));
30500 +union au_rdu_ent_ul {
30501 + struct au_rdu_ent __user *e;
30513 + uint64_t sz; /* AuCtl_RDU */
30514 + uint64_t nent; /* AuCtl_RDU_INO */
30516 + union au_rdu_ent_ul ent;
30517 + uint16_t verify[AufsCtlRduV_End];
30519 + /* input/output */
30523 + union au_rdu_ent_ul tail;
30524 + /* number of entries which were added in a single call */
30529 + struct au_rdu_cookie cookie;
30532 +/* ---------------------------------------------------------------------- */
30534 +struct aufs_wbr_fd {
30539 +/* ---------------------------------------------------------------------- */
30541 +struct aufs_ibusy {
30542 + uint64_t ino, h_ino;
30546 +/* ---------------------------------------------------------------------- */
30548 +#define AuCtlType 'A'
30549 +#define AUFS_CTL_RDU _IOWR(AuCtlType, AuCtl_RDU, struct aufs_rdu)
30550 +#define AUFS_CTL_RDU_INO _IOWR(AuCtlType, AuCtl_RDU_INO, struct aufs_rdu)
30551 +#define AUFS_CTL_WBR_FD _IOW(AuCtlType, AuCtl_WBR_FD, \
30552 + struct aufs_wbr_fd)
30553 +#define AUFS_CTL_IBUSY _IOWR(AuCtlType, AuCtl_IBUSY, struct aufs_ibusy)
30555 +#endif /* __AUFS_TYPE_H__ */
30557 --- linux-3.2/security/device_cgroup.c~ 2012-01-05 13:48:24.000000000 +0100
30558 +++ linux-3.2/security/device_cgroup.c 2012-01-05 14:02:48.910843473 +0100
30560 #include <linux/slab.h>
30561 #include <linux/rcupdate.h>
30562 #include <linux/mutex.h>
30563 +#include <linux/module.h>
30565 #define ACC_MKNOD 1