Carlos Carvalho wrote:
Lachlan McIlroy (lachlan@xxxxxxx) wrote on 25 January 2008 13:00:
>Carlos Carvalho wrote:
>> I compiled the kernel with Ingo's CONFIG_PROVE_LOCKING and got the
>> below at boot. Is it a problem?
>It was a problem - it has since been fixed in 2.6.23.
>Patch is attached in case you're interested.
[patch removed]
Thanks. I applied it to 22.16 without rejects but then the machine
doesn't even mount the partition:
Filesystem "md0": Disabling barriers, not supported by the underlying device
XFS mounting filesystem md0
Unable to handle kernel paging request at ffff8110e325e8d0 RIP:
The patch shouldn't have caused that. Did you keep a stack trace from this?
Then I ran 23.14, no errors. However the machine locks hard a few
seconds after we start the heavy IO tests...
Not good either - again, any stack traces from the stuck threads?
So we used .24. Now the machine stood some 16h of heavy IO load. Good!
However we got the log below after about 15 minutes of running the test.
It didn't repeat. Is it still a problem?
If it didn't cause a deadlock then it's probably not an issue. We probably
just need some more lockdep annotations to tell lockdep to ignore this
locking scenario. I can't see where we acquire nested qh_lock's in
xfs_qm_dqget(). It might actually be caused by acquiring the xfs_qm_freelist
lock in xfs_qm_dqget while holding the hash lock because both locks are
initialised with the same stack and lockdep will think they are locks for
the same object type.
Jan 26 19:25:43 mumm kernel: Filesystem "md0": Disabling barriers, not supported by the underlying device
Jan 26 19:25:43 mumm kernel: XFS mounting filesystem md0
Jan 26 19:25:43 mumm kernel: Ending clean XFS mount for filesystem: md0
15min later...
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: =============================================
Jan 26 19:41:40 mumm kernel: [ INFO: possible recursive locking detected ]
Jan 26 19:41:40 mumm kernel: 2.6.24 #3
Jan 26 19:41:40 mumm kernel: ---------------------------------------------
Jan 26 19:41:40 mumm kernel: rm/1589 is trying to acquire lock:
Jan 26 19:41:40 mumm kernel: (&list->qh_lock){--..}, at: [<ffffffff802c58e2>] xfs_qm_dqget+0x297/0x355
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: but task is already holding lock:
Jan 26 19:41:40 mumm kernel: (&list->qh_lock){--..}, at: [<ffffffff802c5841>] xfs_qm_dqget+0x1f6/0x355
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: other info that might help us debug this:
Jan 26 19:41:40 mumm kernel: 4 locks held by rm/1589:
Jan 26 19:41:40 mumm kernel: #0: (&type->i_mutex_dir_key#2/1){--..}, at: [<ffffffff8027c400>] do_unlinkat+0x6b/0x133
Jan 26 19:41:40 mumm kernel: #1: (&sb->s_type->i_mutex_key#2){--..}, at: [<ffffffff8027a59d>] vfs_unlink+0x42/0x8c
Jan 26 19:41:40 mumm kernel: #2: (&(&ip->i_lock)->mr_lock){----}, at: [<ffffffff802f4093>] xfs_ilock+0x63/0x8d
Jan 26 19:41:40 mumm kernel: #3: (&list->qh_lock){--..}, at: [<ffffffff802c5841>] xfs_qm_dqget+0x1f6/0x355
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: stack backtrace:
Jan 26 19:41:40 mumm kernel: Pid: 1589, comm: rm Not tainted 2.6.24 #3
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: Call Trace:
Jan 26 19:41:40 mumm kernel: [<ffffffff802446b2>] __lock_acquire+0x1b7/0xcd0
Jan 26 19:41:40 mumm kernel: [<ffffffff802c58e2>] xfs_qm_dqget+0x297/0x355
Jan 26 19:41:40 mumm kernel: [<ffffffff802455f4>] lock_acquire+0x51/0x6c
Jan 26 19:41:40 mumm kernel: [<ffffffff802c58e2>] xfs_qm_dqget+0x297/0x355
Jan 26 19:41:40 mumm kernel: [<ffffffff80241ef9>] debug_mutex_lock_common+0x16/0x23
Jan 26 19:41:40 mumm kernel: other info that might help us debug this:
Jan 26 19:41:40 mumm kernel: 4 locks held by rm/1589:
Jan 26 19:41:40 mumm kernel: #0: (&type->i_mutex_dir_key#2/1){--..}, at: [<ffffffff8027c400>] do_unlinkat+0x6b/0x133
Jan 26 19:41:40 mumm kernel: #1: (&sb->s_type->i_mutex_key#2){--..}, at: [<ffffffff8027a59d>] vfs_unlink+0x42/0x8c
Jan 26 19:41:40 mumm kernel: #2: (&(&ip->i_lock)->mr_lock){----}, at: [<ffffffff802f4093>] xfs_ilock+0x63/0x8d
Jan 26 19:41:40 mumm kernel: #3: (&list->qh_lock){--..}, at: [<ffffffff802c5841>] xfs_qm_dqget+0x1f6/0x355
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: stack backtrace:
Jan 26 19:41:40 mumm kernel: Pid: 1589, comm: rm Not tainted 2.6.24 #3
Jan 26 19:41:40 mumm kernel:
Jan 26 19:41:40 mumm kernel: Call Trace:
Jan 26 19:41:40 mumm kernel: [<ffffffff802446b2>] __lock_acquire+0x1b7/0xcd0
Jan 26 19:41:40 mumm kernel: [<ffffffff802c58e2>] xfs_qm_dqget+0x297/0x355
Jan 26 19:41:40 mumm kernel: [<ffffffff802455f4>] lock_acquire+0x51/0x6c
Jan 26 19:41:40 mumm kernel: [<ffffffff802c58e2>] xfs_qm_dqget+0x297/0x355
Jan 26 19:41:40 mumm kernel: [<ffffffff80241ef9>] debug_mutex_lock_common+0x16/0x23
Jan 26 19:41:40 mumm kernel: [<ffffffff803f8c7f>] mutex_lock_nested+0xec/0x286
Jan 26 19:41:40 mumm kernel: [<ffffffff802c58e2>] xfs_qm_dqget+0x297/0x355
Jan 26 19:41:40 mumm kernel: [<ffffffff802c8c2c>] xfs_qm_dqattach_one+0xe2/0x150
Jan 26 19:41:40 mumm kernel: [<ffffffff802c8d51>] xfs_qm_dqattach+0xb7/0x175
Jan 26 19:41:40 mumm kernel: [<ffffffff80311ac5>] xfs_remove+0x8f/0x364
Jan 26 19:41:40 mumm kernel: [<ffffffff803f8dff>] mutex_lock_nested+0x26c/0x286
Jan 26 19:41:40 mumm kernel: [<ffffffff80244082>] trace_hardirqs_on+0x115/0x139
Jan 26 19:41:40 mumm kernel: [<ffffffff80318e6f>] xfs_vn_unlink+0x17/0x36
Jan 26 19:41:40 mumm kernel: [<ffffffff8027a5b7>] vfs_unlink+0x5c/0x8c
Jan 26 19:41:40 mumm kernel: [<ffffffff8027c449>] do_unlinkat+0xb4/0x133
Jan 26 19:41:40 mumm kernel: [<ffffffff803f997a>] trace_hardirqs_on_thunk+0x35/0x3a
Jan 26 19:41:40 mumm kernel: [<ffffffff80244082>] trace_hardirqs_on+0x115/0x139
Jan 26 19:41:40 mumm kernel: [<ffffffff803f997a>] trace_hardirqs_on_thunk+0x35/0x3a
Jan 26 19:41:40 mumm kernel: [<ffffffff8020b66e>] system_call+0x7e/0x83
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
