Message-ID: <20100203225232.GI5068@nowhere>
Date:	Wed, 3 Feb 2010 23:52:34 +0100
From:	Frederic Weisbecker <fweisbec@...il.com>
To:	Alexander Beregalov <a.beregalov@...il.com>
Cc:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: reiserfs deadlock

On Thu, Feb 04, 2010 at 01:43:53AM +0300, Alexander Beregalov wrote:
> On 3 February 2010 23:29, Frederic Weisbecker <fweisbec@...il.com> wrote:
> > On Wed, Feb 03, 2010 at 10:08:57PM +0300, Alexander Beregalov wrote:
> >> On 3 February 2010 22:03, Alexander Beregalov <a.beregalov@...il.com> wrote:
> >> > Hi Frederic
> >> >
> >> > I do not have previous messages and do not know how to reproduce it.
> >> > Kernel was 2.6.33-rc5-00237-g9a3cbe3
> >> >
> >>
> >> Hm, I have the same after reboot.
> >>
> >> Do you need me to do anything before I try to fsck ?
> >
> >
> > Yeah. Rebooting again makes your kernel soft lockup?
> Yes, reboot does not help. I even can't login, agetty and sshd are frozen.
> 
> INFO: task sshd:1863 blocked for more than 120 seconds.
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> sshd          D 6f60ec44  6576  1863   1810 0x00000000
>  f633dd78 00000046 ffffffff 6f60ec44 0000000f f7306b30 f73068b0 00000000
>  f7306d84 7fffffff 00000000 f633de70 f633dde8 c134da45 00000000 f633dd8c
>  c104ca3b 00000000 7fffffff 0000000f 6f618f50 f73068b0 00000000 00000000
> Call Trace:
>  [<c134da45>] schedule_timeout+0x125/0x1b0
>  [<c104ca3b>] ? trace_hardirqs_off+0xb/0x10
>  [<c1350152>] ? _raw_spin_unlock_irq+0x22/0x30
>  [<c104e4c4>] ? trace_hardirqs_on_caller+0x124/0x170
>  [<c104e51b>] ? trace_hardirqs_on+0xb/0x10
>  [<c134d7d0>] wait_for_common+0xd0/0x130
>  [<c1024850>] ? default_wake_function+0x0/0x10
>  [<c134d8c2>] wait_for_completion+0x12/0x20
>  [<c1039709>] call_usermodehelper_exec+0x89/0xb0
>  [<c1039471>] ? call_usermodehelper_setup+0x71/0xb0
>  [<c134d730>] ? wait_for_common+0x30/0x130
>  [<c10398e2>] __request_module+0xa2/0xf0
>  [<c10a6136>] ? new_inode+0x76/0x80
>  [<c13501cd>] ? _raw_spin_unlock+0x1d/0x20
>  [<c12cc89f>] __sock_create+0x18f/0x1f0
>  [<c107b22a>] ? might_fault+0x4a/0xa0
>  [<c12cc967>] sock_create+0x37/0x40
>  [<c12ccb1e>] sys_socket+0x3e/0x70
>  [<c12ccbb0>] sys_socketcall+0x60/0x270
>  [<c1002b43>] ? sysenter_exit+0xf/0x18
>  [<c11d5eb4>] ? trace_hardirqs_on_thunk+0xc/0x10
>  [<c1002b10>] sysenter_do_call+0x12/0x36
> no locks held by sshd/1863.
> 
> No locks - what does it mean?



This is the call_usermodehelper_exec path, so probably
the kernel tries to ask userspace to load a module, but
since the filesystem is locked up, this can't happen.



> >
> > Usually such softlockup happens because we have a lock
> > inversion, in which case you should have a lockdep report
> > before the softlockup.
> 
> No, I do not have it. 120 seconds after boot I see these messages on
> the console,
> no lockdep reports (lockdep is enabled).



So this is probably this event waited thing.



> >
> > Otherwise this can also happen when we wait for an event
> > that needs the lock to complete but
> > that can not happen because we already have the lock.
> >
> > Task A hold reiserfs lock and wait for event 1
> > Task B wants to complete event 1 but it need the reisers lock
> > for that => deadlock.
> >
> > This can usually be found in a softlockup report: lots of
> > tasks are blocked on reiserfs_write_lock/mutex_lock
> > except one, and this one is important as it is probably
> > the waiter: the task that holds the lock and that is waiting
> > for another event (that in turn needs the lock to complete).
> >
> > Having more reports could probably help us:
> >
> > echo 100 > /proc/sys/kernel/hung_task_warnings
> 
> Ok, I will modify rc scripts to do it, as I can't login.


Thanks!

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists