UBUNTU: ubuntu: AUFS -- update to c5021514085a5d96364e096dbd34cadb2251abfd

UBUNTU: ubuntu: AUFS -- update to c5021514085a5d96364e096dbd34cadb2251abfd

    commit c5021514085a5d96364e096dbd34cadb2251abfd
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Dec 16 01:29:04 2010 +0900

        aufs: bugfix, missign test for branch management

        In aufs_lookup(), the parent dir MAY be removed logically by branch
        management.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 5df24d5e0dc992ab991d0113703b9c90416ce527
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 15 22:55:09 2010 +0900

        aufs: debugging, stop dcsub in sysrq handler

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 3f141020200c0e39c41c039eb261f65db0b3f18e
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 15 22:50:15 2010 +0900

        aufs: tiny, revised simply

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 8f46271b6cd7521da44a087dcbf3d783f4af24c7
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 15 22:18:33 2010 +0900

        aufs: debugging, false positive around NOPLMW

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 788f93f121fcafecaca33dafb824495db02edfe7
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 15 21:19:26 2010 +0900

        aufs: tiny, just an indentation

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 9a8dc61df4b4b722f9a9ef8a568cfcb78a4eb0aa
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 15 21:19:09 2010 +0900

        aufs: minor optimization, skip file-refresh for root

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 9784a821611537ad9d0d4593ba775bb816ace3e9
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Mon Dec 13 17:13:59 2010 +0900

        aufs: possible bugfix, protect branch from deleting in nfsd fh_to_dentry()

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 8c017b90207d3e0914f53f7e15ff9fb30d33ffac
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Mon Dec 13 17:00:20 2010 +0900

        aufs: possible bugfix, br_count in async rmdir

        There is a missing decrement br_count for 'br' member in struct
        au_whtmp_rmdir in an error path. Introduce a larger and bigger but more
        generic and simple solution.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 29c19e545f53894ca89a9525685572b9c99f34c3
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sun Dec 12 10:25:26 2010 +0900

        aufs: move missing declaration

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 0de8fbc346a9bc5cf314ea82063a65c08ca9e41b
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Dec 10 13:25:09 2010 +0900

        aufs: possible bugfix, walk in dcache limited to aufs

        In some cases, au_dcsub_pages_rev() should not handle out of aufs or
        broken (but still cached) aufs dentry.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 52f9727243bf644db05168e6c77d2f063bcc01ba
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Dec 10 13:20:06 2010 +0900

        aufs: d_instantiate in link(2)

        Historically there was(were?) filesystem which calls d_drop() after the
        success of link(2). In order to support such filesystem, old aufs1
        follows it if h_dentry is dropped without d_instantiate().
        Recently I looked whether there still exists such filesystem or not, and
        could not find. So aufs_link() uncondionally calls d_instantiate() after
        success of vfs_link() for branch fs.
        But, as a fallback, d_drop() call with a d_unhashed() test are left.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 6b0be66a68b95ca0bab3068133c8730882a38385
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Dec 10 02:29:50 2010 +0900

        aufs: bugfix, lock subclass in copying-up a dir hierarchy

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit e1d13b721d930c72217df0a9b6e8a87e51ed1810
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Dec 10 02:27:55 2010 +0900

        aufs: bugfix, restore the internal array after special copyup

        When copyup a file which exists on the lower readonly branch and already
        removed logically, au_do_cpup_wh() operates the internal array in dinfo
        before copyup. After that, regardless an error happened in copyup,
        restore the internal array.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit d539fca6887a013c6fbd1b579898aa78ba7c65c9
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Dec 9 02:19:42 2010 +0900

        aufs: minor, make inode bad before unlock

        If an error happens in settting a new inode, we call iget_failed() and
        make it "bad inode." Since someone else is wating for the inode is
        unlocked, call iget_failed() before unlock it.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit b69574acb527171af24eddb51ac414664d70ddc4
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Dec 9 02:17:00 2010 +0900

        aufs: bugfix, assign inode number for hardlink

        If the cached inode doesn't match the dentry in operation, never combine
        them (instatiate). Find another inode number.
        Basically the known inode number is stored in XINO file. But UDBA or
        branch management easily break it.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit efa578c51e3132442f8cd65e6efe4e344969449f
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:28:32 2010 +0900

        aufs: tiny, a comment

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit ff7babf5257fa57b341eb7230817a9ec0b45767c
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:28:16 2010 +0900

        aufs: debug by lockdep_depth()

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit f44393848b3d0ab046de672752d4c23cf14c6b62
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:27:04 2010 +0900

        aufs: minor, sysrq+A prints all

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 8fa5bfe552fdab2c900309f73f8a4174f705282d
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:26:03 2010 +0900

        aufs: limit of rdcache

        MAX_SEC_IN_JIFFIES is less meaningful and hard to handle.
        Introduce a new reasonable upper limit AUFS_RDCACHE_MAX.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit f85351d0360f40a2595c99f1182ddf1de0eed8a1
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:24:40 2010 +0900

        aufs: limit of MFS seconds

        MAX_SEC_IN_JIFFIES is less meaningful and hard to handle.
        Introduce a new reasonable upper limit AUFS_MFS_MAX_SEC.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit b64f629ab85e8d0641704fea2f2787d53713afd9
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:21:00 2010 +0900

        aufs: tiny, revise error messages

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 3e17bf45f4ab40d6516c9de073417942661cd31f
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Dec 7 01:19:32 2010 +0900

        aufs: bugfix, hfsnotify for multiple aufs 2/2

        Implement new operations, ->fin_br(), ->init_br() and ->reset_br().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit fec63cfd00c69fe89ecc60fdf22e610b836e0785
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Dec 7 01:18:47 2010 +0900

        aufs: bugfix, hfsnotify for multiple aufs 1/2

        Since fsnotify doesn't allow registering a same fsnotify_ops to the same
        fsnotify_group, aufs should make them per branch.
        Theoritically it is enough to make them per aufs super_block, but users
        MAY move branch into the same filesystem where other branch exists. To
        support such case, make them per branch.
        Introduce new operations, ->fin_br(), ->init_br() and ->reset_br().

        Reported-by: Joonwoo Park <joonwpark81@gmail.com>
        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 3da80541bb8d42a8f3cd2e2caf04b3037a0040bf
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Mon Dec 6 23:53:49 2010 +0900

        aufs: tiny, revise the parameter for hinotify

        - convert au_hnotify to au_hinode, remove h_inode from au_hn_alloc() and
          ->alloc().
        - convert au_hnotify to au_hinode, ->free().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit b632f916658d019669d393975bdc66644b333269
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Dec 10 13:29:13 2010 +0900

        aufs: debugging dinfo

        New debug function to verify the correctness of dinfo.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 1341643c5de13471e781f9c0bcd8d987023494b6
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Mon Dec 6 23:41:42 2010 +0900

        aufs: bugfix, return value of au_do_refresh_d()

        Intialize it as success (0).
        Add some debug messgages.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit a4935871b43ea40fa73a67f228da6ef8c3dd0b14
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Dec 8 02:27:39 2010 +0900

        aufs stdalone: export debug_locks

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 09853aaa34cca5eeedb9277591668258de081813
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Dec 7 21:53:59 2010 +0900

        aufs: support for 2.6.34, protect f_mode

        By introducing FMODE_RANDOM, file->f_mode may be modified out of aufs.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit f1617ce6112b85ed5a686be5e0cb8f48794ea2c4
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Dec 2 03:48:57 2010 +0900

        aufs: revise how to build

        Reported-by: wanyanqing <udknight@gmail.com>
        Reported-by: Alex Vogt <elias@linexa.de>
        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 543820e74625523e263faad047928266dc672cfd
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Nov 30 15:39:35 2010 +0900

        aufs: tiny, remove unnecesary var

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 38fe8516196e35e6fab39a49f650dd5dd24deb63
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Nov 30 06:27:32 2010 +0900

        aufs: refreshing, replace functions for remount time

        (This is a part of a patch series "new strategy for refreshing")

        In remounting (branch management), refresh all dentries and inodes
        regardless their status (unhashed, unlinked, etc).

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit f822af6506e20cd2bd328dfc678ffbcf31c6f910
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Nov 30 06:25:27 2010 +0900

        aufs: refreshing, new functions for remount time

        (This is a part of a patch series "new strategy for refreshing")

        Introduce new functions do_refresh, au_do_refresh_d, au_refresh_d, and
        au_refresh_i which will replace current functions.
        Since they are large and to keep single commit simple, this commit just
        define these functions. The behaviour won't change.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 85a3a786f9fb483aea021caf66d71a930a327974
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sat Nov 27 14:32:52 2010 +0900

        aufs: refreshing, replace old function by new one

        (This is a part of a patch series "new strategy for refreshing")

        Replace au_refresh_hinode() by au_refresh_dentry() which was committed
        previously.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 872472221dc4771c29010fb89609ffc4ec31d9f9
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sat Nov 27 03:32:18 2010 +0900

        aufs: refreshing, new functions to refresh dentry

        (This is a part of a patch series "new strategy for refreshing")

        Introduce new functions au_do_hide, au_hide_children, au_hide,
        au_refresh_by_dinfo, and au_refresh_dentry which will replace current
        au_refresh_hdentry.
        Since they are large and to keep single commit simple, this commit just
        define these functions. The behaviour won't change.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 92dcda8894818f43b7a627d4a3062996552306b3
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 22:10:47 2010 +0900

        aufs: refreshing, refine au_do_refresh_hdentry

        (This is a part of a patch series "new strategy for refreshing")

        Now dinfo sets -1 to dbstart and dbend to represent invalid.
        Currently its return value is ignored. It will be used in later commits.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit db0c0d602004fe2884225624af19e9ac97d8b19c
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 16:59:17 2010 +0900

        aufs: refreshing, stop updating iigen in test_inode_busy()

        (This is a part of a patch series "new strategy for refreshing")

        In order to force refreshing in the later au_remount_refresh().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 48c759bfc43e795092bf0c0f31c6a8047cc6c9b1
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 03:19:02 2010 +0900

        aufs: refreshing, consolidate REFRESH macros

        (This is a part of a patch series "new strategy for refreshing")

        At remounting (branch management), aufs refreshes some internal arrays.
        In this refreshing, objects for a dir and non-dir are separatedly
        handled for performance. This commit forces always refreshing both.
        And rename a macro AuSi_FAILED_REFRESH_DIRS to AuSiFAILED_REFRESH_DIR.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 40058d7c0ee7d8f449f5eef9ed178335303924a5
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Nov 25 02:40:13 2010 +0900

        aufs: new strategy for refreshing, refresh negative dentries

        (This is a part of a patch series "new strategy for refreshing")

        In branch management at remount, refresh negative dentries too, since
        they may hold a refrence to an object in the deleting branch.
        Also they may be in freeing, so call dget_locked() instead of simple
        dget().
        The permission of the branch who has the negative dentry may be changed
        to readonly, so confirm it in au_pin_and_icpup().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 97f1369508bf1e71935363adc85c5fd52695a915
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sun Nov 28 11:24:12 2010 +0900

        aufs: keep dinfo valid by temp dinfo

        To decide creating a new whiteout is necessary or not, aufs tries
        finding whiteout on the lower branches. In this lookup, use another
        dinfo temporary to keep the original dinfo.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 381ba01af8e781512742858854a510ebb7667b1f
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 00:52:57 2010 +0900

        aufs: temporary dinfo

        Introduce some functions to handle dinfo which will be used to make a
        temporary dinfo. The existing interfaces (di_init() and di_fin()) are
        not changed.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit ca7dc2240a3b787761d31dc096fe95f3beb9a18a
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 21:59:45 2010 +0900

        aufs: store br_id in dinfo

        In order to identify the index of branch at refreshing, store br_id in
        dinfo.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 447344925adad5f89950e409d3de501e7431468e
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sun Nov 28 22:47:25 2010 +0900

        aufs: bugfix, write to a removed file more than once

        In file write, the file may be already removed and au_ready_to_write()
        supports such case. But the write to the removed file, eg, the file
        had been already handled by au_ready_to_write(), then the function
        doesn't detect it and a bug appear.
        - when the file is removed, au_cpup_dirs() is unnecessary.
        - when the file is removevd and already handled by au_ready_to_write(),
          just re-open without copyup.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 24dddc56d3cab8671e30ad2213436b405ca715bc
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 16:55:26 2010 +0900

        aufs: split au_refresh_hinode_self()

        Extract a part of au_refresh_hinode_self() and create a new function
        au_ii_refresh(). The 'do_attr' parameter of au_refresh_hinode_self()
        becomes unnecessary.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit d954281d7bd5f2523804290b120fd8a8bc7bb7fe
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 15:46:32 2010 +0900

        aufs: refine au_update_ibrange()

        Now iinfo sets -1 to ibstart and ibend to represent invalid.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit bbc392af26ace27fc5e41e68f9437ad57ec2b85e
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 01:45:43 2010 +0900

        aufs: keep dinfo valid

        Keep dinfo valid even if an error happened.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 49b3e86c8b8337fc7db9093ec98906de4cd4acb3
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Nov 25 17:00:04 2010 +0900

        aufs: use a generic warpper dbrange_test()

        Call au_dbrange_test() near au_digen_test() call.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit e3d0c8ea457f3192caacaaeb83a01e9befb5176c
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Nov 25 16:59:11 2010 +0900

        aufs: use a generic warpper [di]i_gen_test()

        Replace some direct test codes by au_[di]igen_test().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 6680a2cc71b37511097df27149ae072034e89491
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Tue Nov 30 13:35:07 2010 +0900

        aufs: bugfix, missing unlock in an error path

        File revalidating in au_reval_and_lock_fdi().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit f1abf5b5d55ffeae0d3850a62853c99c02286228
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Mon Nov 29 03:22:07 2010 +0900

        aufs: possible bugfix, the generation of dentry 3/3

        Test the generation of a dentry in the beginning of many operations.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 69e55ee7c6f87380df71b37dca94f6023407e1c2
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Mon Nov 29 03:21:36 2010 +0900

        aufs: possible bugfix, the generation of dentry 2/3

        Introduce AuLock_GEN flag to test the generation of a dentry in the
        beginning of many operations.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 8755a096f8cc12a00b5f4826ac6cf8a3f5014535
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Nov 25 01:25:57 2010 +0900

        aufs: possible bugfix, the generation of dentry 1/3

        Introduce new functions au_[di]igen_test() and au_dbrange_test().

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 49bfb76c10c75603561fcd6e9d79d5ccf000e7ce
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sun Nov 28 23:33:50 2010 +0900

        aufs: possible bugfix, test valid dentry 2/2

        Test the validity of a dentry in the beginning of many operations.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 2ecf50571003219f9d7ab257420c570764332469
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sun Nov 28 23:05:57 2010 +0900

        aufs: possible bugfix, test valid dentry 1/2

        Introduce some functions to test the validity of a dentry.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 952cec591bffdb1d470286eeea4ae27ed196bf94
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sun Nov 28 22:51:06 2010 +0900

        aufs: tiny, debug code for i_mode

        To detect setting type unmatching h_inode to iinfo, initialize i_mode.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit c00655e4c8f66871decf0d79a2d3b4c1fb6207d7
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Nov 25 00:00:44 2010 +0900

        aufs: tiny, comment around d_drop

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 426ed6061a419cb1c7a5e1504052a50c5198d3a7
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Nov 24 23:59:08 2010 +0900

        aufs: possible bugfix, keep dentry of renamed target valid

        Basically the renamed target is removed, but it may be still referenced
        and alive. Call d_drop in aufs and keep its dinfo valid.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 6f00617abf56f74429ab06751df6519b957fc325
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sat Nov 27 22:06:48 2010 +0900

        aufs: possible bugfix, decrement the link count of non-dir rename target

        Fix a missing drop_nlink() for the non-dir rename target.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit b8ee2bd5b62c5dece464571a83f4142181a67a72
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Fri Nov 26 17:28:15 2010 +0900

        aufs: set S_DEAD to i_flags in aufs

        The S_DEAD flag is set in VFS, but we need to do it by ourselves because
        a race condition may happen with aufs remount/branch-mgmt operation.
        Set the flag within the aufs lock region.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 68fd4a243e4b9bbf397040c68ccc0f11da12c304
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Nov 24 23:58:35 2010 +0900

        aufs: minor optimization, d_drop after failure of adding an entry

        After the failure of creating an entry, the aufs dentry private data may
        still refer the dentry in the branch fs's or may have invalid info.
        Discard it by d_drop() in error case.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 6f7ebf0d44bacda459a8ae6235c59f47c9715808
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Nov 24 15:57:22 2010 +0900

        aufs: possible bugfix, d_drop() job for hnotify

        Current d_drop() calls may have unexpected side effects.
        In order to force re-lookup, aufs should use di_gen instead of testing
        unhashed. When the test (in d_revalidate) fails, make it obsoleted.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit ade2eb21fc942caa1c48bd931b5b8e2411eaadde
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Thu Nov 25 02:52:39 2010 +0900

        aufs: tiny, remove an empty line

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 16514eeea2d4b3ddcdd01b6cf54c8ae60cdc9bdf
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Nov 24 15:09:23 2010 +0900

        aufs: tiny, use a wrapper for fs private data

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 8823675f5fd00ef30d6fea5f6779b45b3eb3bfd1
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Sat Nov 27 04:05:07 2010 +0900

        aufs: possible bugfix, test d_fsdata in revalidating

        The passed dentry may not be created correctly, error case in previous
        lookup.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

    commit 33d73de5ae0c44312ce6fbce59819cf04b9de9fa
    Author: J. R. Okajima <hooanon05@yahoo.co.jp>
    Date:   Wed Nov 24 13:22:02 2010 +0900

        aufs: possible bugfix, release all before au_nwt_done()

        In racing unmount process, iput() after au_nwt_done() MAY cause a problem.

        Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>

Signed-off-by: Andy Whitcroft <apw@canonical.com>