lkml.org 
[lkml]   [2006]   [Dec]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: 2.6.16.32 stuck in generic_file_aio_write()

> > Done some more digging : isn't http://lkml.org/lkml/2006/10/13/139 somehow
> > related ? I do see pagefaults, and inode locks and mmap_locks.
> >
>
> I thought it was, but from my look through yout 8-billion-task backtrace,
> no task was stuck in D-state with the appropriate call trace.
>
> So I don't know what's causing this. In the first trace you have at least
> four D-state kjournalds and a lot of processes stuck on an i_mutex. I
> guess it's consistent with an IO system which is losing completion
> interrupts.

Hmm.. Is there any way to make sure ? I've got a second machine (almost
identical), which doesn't show this.

The main difference is the running kernel. I've had them at the same
kernel, at which bad machine still crashes.

/proc/interrupts

Bad machine : 18: 11160637 11235698 IO-APIC-fasteoi arcmsr
Good machine : 18: 61658630 79352227 IO-APIC-level arcmsr

Bad machine is running 2.6.19, good is running 2.6.14.7-grsec, which
probably accounts for these changes.

> AFAICT in the second trace all you have is a lot of processes
> stuck on i_mutex for no obvious reason - I don't know why that would
> happen.

It's consequent, also the traces.

> How long does it take for this to happen?

Days to a week tops. It does happen less frequent with the 2.6.19,
2.6.16.32 triggered it almost daily.

> Yes, lockdep might find something.

I've enabled most debug options. I'll boot the other kernel tomorrow.



Regards,


Igmar
-
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/

\
 
 \ /
  Last update: 2006-12-06 17:17    [W:0.143 / U:0.480 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site