'BUG: scheduling with irqs disabled' when umounting NFS volume

'BUG: scheduling with irqs disabled' when umounting NFS volume

[Lee Revell]

Kernel is 2.6.12-rc1-RT-V0.7.43-05.

BUG: scheduling with irqs disabled: umount/0x00000000/20612
caller is schedule_timeout+0x63/0xc0
 [<c01033d3>] dump_stack+0x23/0x30 (20)
 [<c02b4f5a>] schedule+0xea/0x140 (36)
 [<c02b5b23>] schedule_timeout+0x63/0xc0 (64)
 [<c02b5744>] interruptible_sleep_on_timeout+0x74/0xe0 (64)
 [<c01cf898>] lockd_down+0xb8/0x140 (24)
 [<c01c2137>] nfs_kill_super+0x77/0x80 (16)
 [<c016033c>] deactivate_super+0x8c/0xb0 (28)
 [<c0178bc1>] sys_umount+0x41/0x90 (88)
 [<c0178c2e>] sys_oldumount+0x1e/0x20 (16)
 [<c0102dee>] syscall_call+0x7/0xb (-8124)
---------------------------
| preempt count: 00000001 ]
| 1-level deep critical section nesting:
----------------------------------------
.. [<c01320ad>] .... print_traces+0x1d/0x60
.....[<c01033d3>] ..   ( <= dump_stack+0x23/0x30)

Lee

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Daniel Walker]

I submitted a fix for this a while ago, I think ..
interruptible_sleep_on()'s are broken .. 


Daniel

On Fri, 2005-04-08 at 13:15, Lee Revell wrote:
> Kernel is 2.6.12-rc1-RT-V0.7.43-05.
> 
> BUG: scheduling with irqs disabled: umount/0x00000000/20612
> caller is schedule_timeout+0x63/0xc0
>  [<c01033d3>] dump_stack+0x23/0x30 (20)
>  [<c02b4f5a>] schedule+0xea/0x140 (36)
>  [<c02b5b23>] schedule_timeout+0x63/0xc0 (64)
>  [<c02b5744>] interruptible_sleep_on_timeout+0x74/0xe0 (64)
>  [<c01cf898>] lockd_down+0xb8/0x140 (24)
>  [<c01c2137>] nfs_kill_super+0x77/0x80 (16)
>  [<c016033c>] deactivate_super+0x8c/0xb0 (28)
>  [<c0178bc1>] sys_umount+0x41/0x90 (88)
>  [<c0178c2e>] sys_oldumount+0x1e/0x20 (16)
>  [<c0102dee>] syscall_call+0x7/0xb (-8124)
> ---------------------------
> | preempt count: 00000001 ]
> | 1-level deep critical section nesting:
> ----------------------------------------
> .. [<c01320ad>] .... print_traces+0x1d/0x60
> .....[<c01033d3>] ..   ( <= dump_stack+0x23/0x30)
> 
> Lee
> 
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Lee Revell]

On Fri, 2005-04-08 at 13:38 -0700, Daniel Walker wrote:
> I submitted a fix for this a while ago, I think ..
> interruptible_sleep_on()'s are broken .. 

I saw the fix in -stable, but it does not seem to be in 2.6.12-rc2.

Lee

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Ingo Molnar]

* Daniel Walker <dwalker@mvista.com> wrote:

> I submitted a fix for this a while ago, I think ..
> interruptible_sleep_on()'s are broken .. 

sleep_on() is a fundamentally broken interface, it only works on UP - 
but there it _does_ rely on the behavior your patch removes. (i.e.  
disabled interrupts until we hit schedule())

the PREEMPT_RT kernel makes the limitations of sleep_on() even more 
apparent. The patch only removes the warning, it doesnt remove the race.  
To remove the race, sleep_on() usage should be converted to something 
else. (e.g. one of the wait_event() variants)

	Ingo

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Ingo Molnar]

* Lee Revell <rlrevell@joe-job.com> wrote:

> Kernel is 2.6.12-rc1-RT-V0.7.43-05.
> 
> BUG: scheduling with irqs disabled: umount/0x00000000/20612
> caller is schedule_timeout+0x63/0xc0
>  [<c01033d3>] dump_stack+0x23/0x30 (20)
>  [<c02b4f5a>] schedule+0xea/0x140 (36)
>  [<c02b5b23>] schedule_timeout+0x63/0xc0 (64)
>  [<c02b5744>] interruptible_sleep_on_timeout+0x74/0xe0 (64)
>  [<c01cf898>] lockd_down+0xb8/0x140 (24)
>  [<c01c2137>] nfs_kill_super+0x77/0x80 (16)
>  [<c016033c>] deactivate_super+0x8c/0xb0 (28)
>  [<c0178bc1>] sys_umount+0x41/0x90 (88)
>  [<c0178c2e>] sys_oldumount+0x1e/0x20 (16)
>  [<c0102dee>] syscall_call+0x7/0xb (-8124)

i've uploaded the -44-04 kernel, which should fix this bug.

	Ingo

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Daniel Walker]

On Fri, 2005-04-08 at 14:51, Lee Revell wrote:
> On Fri, 2005-04-08 at 13:38 -0700, Daniel Walker wrote:
> > I submitted a fix for this a while ago, I think ..
> > interruptible_sleep_on()'s are broken .. 
> 
> I saw the fix in -stable, but it does not seem to be in 2.6.12-rc2.


I didn't know it was in any of the kernels. Do I need to submit it to
Linus or something?

Daniel

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Lee Revell]

On Mon, 2005-04-11 at 09:29 -0700, Daniel Walker wrote:
> On Fri, 2005-04-08 at 14:51, Lee Revell wrote:
> > On Fri, 2005-04-08 at 13:38 -0700, Daniel Walker wrote:
> > > I submitted a fix for this a while ago, I think ..
> > > interruptible_sleep_on()'s are broken .. 
> > 
> > I saw the fix in -stable, but it does not seem to be in 2.6.12-rc2.
> 
> 
> I didn't know it was in any of the kernels. Do I need to submit it to
> Linus or something?

I must be thinking of a different bug then.  Anyway, Ingo said this was
fixed in the latest RT kernels.

Lee

Re: 'BUG: scheduling with irqs disabled' when umounting NFS volume

[Daniel Walker]

On Fri, 2005-04-08 at 21:44, Ingo Molnar wrote:
> * Daniel Walker <dwalker@mvista.com> wrote:
> 
> > I submitted a fix for this a while ago, I think ..
> > interruptible_sleep_on()'s are broken .. 
> 
> sleep_on() is a fundamentally broken interface, it only works on UP - 
> but there it _does_ rely on the behavior your patch removes. (i.e.  
> disabled interrupts until we hit schedule())
> 
> the PREEMPT_RT kernel makes the limitations of sleep_on() even more 
> apparent. The patch only removes the warning, it doesnt remove the race.  
> To remove the race, sleep_on() usage should be converted to something 
> else. (e.g. one of the wait_event() variants)

I know they aren't suppose to be used any more. However, there are 100+
of these calls in the kernel right now ..

Daniel