f2fs: should cover i_xattr_nid with its xattr node page lock
authorJaegeuk Kim <jaegeuk.kim@samsung.com>
Mon, 12 Aug 2013 07:04:53 +0000 (16:04 +0900)
committerJaegeuk Kim <jaegeuk.kim@samsung.com>
Mon, 12 Aug 2013 07:04:53 +0000 (16:04 +0900)
commit479bd73ac425ff117efeea051077b4277baab52e
treea078a20d6d657f4b77140392be5a6316c96110a3
parent9c02740c0174932162531a28ba8593e82884a9d7
f2fs: should cover i_xattr_nid with its xattr node page lock

Previously, f2fs_setxattr assigns i_xattr_nid in the inode page inconsistently.

The scenario is:

= Thread 1 =         = Thread 2 =     = fi->i_xattr_nid =  = on-disk nid =

f2fs_setxattr                                   0                 0
  new_node_page                                 X                 0
                   sync_inode_page              X                 X
                   checkpoint                   X                 X -.
    grab_cache_page                             X                 X  |
--> allocate a new xattr node block or -ENOSPC      <----------------'

At this moment, the checkpoint stores inconsistent data where the inode has
i_xattr_nid but actual xattr node block is not allocated yet.

So, we should assign the real i_xattr_nid only after its xattr node block is
allocated.

Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
fs/f2fs/node.c
fs/f2fs/xattr.c