ocfs2/lockdep: Move ip_xattr_sem out of ocfs2_xattr_get_nolock.
As the name shows, we shouldn't have any lock in
ocfs2_xattr_get_nolock. so lift ip_xattr_sem to the caller.
This should be safe for us since the only 2 callers are:
1. ocfs2_xattr_get which will lock the resources.
2. ocfs2_mknod which don't need this locking.
And this also resolves the following lockdep warning.
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.35+ #5
-------------------------------------------------------
reflink/30027 is trying to acquire lock:
(&oi->ip_alloc_sem){+.+.+.}, at: [<
ffffffffa0673b67>] ocfs2_reflink_ioctl+0x69a/0x1226 [ocfs2]
but task is already holding lock:
(&oi->ip_xattr_sem){++++..}, at: [<
ffffffffa0673b58>] ocfs2_reflink_ioctl+0x68b/0x1226 [ocfs2]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 (&oi->ip_xattr_sem){++++..}:
[<
ffffffff82064d6d>] __lock_acquire+0x79a/0x7f1
[<
ffffffff82065a81>] lock_acquire+0xc6/0xed
[<
ffffffff82339650>] down_read+0x34/0x47
[<
ffffffffa0691cb8>] ocfs2_xattr_get_nolock+0xa0/0x4e6 [ocfs2]
[<
ffffffffa069d64f>] ocfs2_get_acl_nolock+0x5c/0x132 [ocfs2]
[<
ffffffffa069d9c7>] ocfs2_init_acl+0x60/0x243 [ocfs2]
[<
ffffffffa066499d>] ocfs2_mknod+0xae8/0xfea [ocfs2]
[<
ffffffffa0665041>] ocfs2_create+0x9d/0x105 [ocfs2]
[<
ffffffff820e1c83>] vfs_create+0x9b/0xf4
[<
ffffffff820e20bb>] do_last+0x2fd/0x5be
[<
ffffffff820e31c0>] do_filp_open+0x1fb/0x572
[<
ffffffff820d6cf6>] do_sys_open+0x5a/0xe7
[<
ffffffff820d6dac>] sys_open+0x1b/0x1d
[<
ffffffff8200296b>] system_call_fastpath+0x16/0x1b
-> #2 (jbd2_handle){+.+...}:
[<
ffffffff82064d6d>] __lock_acquire+0x79a/0x7f1
[<
ffffffff82065a81>] lock_acquire+0xc6/0xed
[<
ffffffffa0604ff8>] start_this_handle+0x4a3/0x4bc [jbd2]
[<
ffffffffa06051d6>] jbd2__journal_start+0xba/0xee [jbd2]
[<
ffffffffa0605218>] jbd2_journal_start+0xe/0x10 [jbd2]
[<
ffffffffa065ca34>] ocfs2_start_trans+0xb7/0x19b [ocfs2]
[<
ffffffffa06645f3>] ocfs2_mknod+0x73e/0xfea [ocfs2]
[<
ffffffffa0665041>] ocfs2_create+0x9d/0x105 [ocfs2]
[<
ffffffff820e1c83>] vfs_create+0x9b/0xf4
[<
ffffffff820e20bb>] do_last+0x2fd/0x5be
[<
ffffffff820e31c0>] do_filp_open+0x1fb/0x572
[<
ffffffff820d6cf6>] do_sys_open+0x5a/0xe7
[<
ffffffff820d6dac>] sys_open+0x1b/0x1d
[<
ffffffff8200296b>] system_call_fastpath+0x16/0x1b
-> #1 (&journal->j_trans_barrier){.+.+..}:
[<
ffffffff82064d6d>] __lock_acquire+0x79a/0x7f1
[<
ffffffff82064fa9>] lock_release_non_nested+0x1e5/0x24b
[<
ffffffff82065999>] lock_release+0x158/0x17a
[<
ffffffff823389f6>] __mutex_unlock_slowpath+0xbf/0x11b
[<
ffffffff82338a5b>] mutex_unlock+0x9/0xb
[<
ffffffffa0679673>] ocfs2_free_ac_resource+0x31/0x67 [ocfs2]
[<
ffffffffa067c6bc>] ocfs2_free_alloc_context+0x11/0x1d [ocfs2]
[<
ffffffffa0633de0>] ocfs2_write_begin_nolock+0x141e/0x159b [ocfs2]
[<
ffffffffa0635523>] ocfs2_write_begin+0x11e/0x1e7 [ocfs2]
[<
ffffffff820a1297>] generic_file_buffered_write+0x10c/0x210
[<
ffffffffa0653624>] ocfs2_file_aio_write+0x4cc/0x6d3 [ocfs2]
[<
ffffffff820d822d>] do_sync_write+0xc2/0x106
[<
ffffffff820d897b>] vfs_write+0xae/0x131
[<
ffffffff820d8e55>] sys_write+0x47/0x6f
[<
ffffffff8200296b>] system_call_fastpath+0x16/0x1b
-> #0 (&oi->ip_alloc_sem){+.+.+.}:
[<
ffffffff82063f92>] validate_chain+0x727/0xd68
[<
ffffffff82064d6d>] __lock_acquire+0x79a/0x7f1
[<
ffffffff82065a81>] lock_acquire+0xc6/0xed
[<
ffffffff82339694>] down_write+0x31/0x52
[<
ffffffffa0673b67>] ocfs2_reflink_ioctl+0x69a/0x1226 [ocfs2]
[<
ffffffffa06599f6>] ocfs2_ioctl+0x61a/0x656 [ocfs2]
[<
ffffffff820e53ac>] vfs_ioctl+0x2a/0x9d
[<
ffffffff820e5903>] do_vfs_ioctl+0x45d/0x4ae
[<
ffffffff820e59ab>] sys_ioctl+0x57/0x7a
[<
ffffffff8200296b>] system_call_fastpath+0x16/0x1b
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>