2003-07-12 16:55:34

by Marc Heckmann

[permalink] [raw]
Subject: 2.4.21 smp: system lockup

Hi,

I recently experienced a lockup on an SMP 2.4.21 kernel (vanilla).

The machine was under very heavy IO at the time. (a full backup was in
progress). gzipped Alt+sysrq+t output is attached. I forgot
Alt+sysrq+P unfortunatly.

iptables was still functional, it still responded to pings, but
everything else was locked up. obviously sysrq was still working. I
managed to reboot it with sysrq+b over the serial line.

the machine has been in production for a while and has run in the past
for almost 2 months w/o any problems, so I don't believe it's hardware
related.

A similar lockup happened once with the redhat 2.4.20-13.9 kernel. I
blamed that on all the extra patches present and switched to 2.4.21
hoping that it would alleviate the problem.

PS: I'm not an linux-kernel so please CC me directly.

-m


Attachments:
(No filename) (812.00 B)
loging.txt.gz (4.26 kB)
Download all attachments

2003-07-14 20:27:32

by Marc Heckmann

[permalink] [raw]
Subject: Re: 2.4.21 smp: system lockup

forgot the attachement...

-m

On Mon, Jul 14, 2003 at 04:32:58PM -0400, Marc Heckmann wrote:
> Hi there,
>
> just realised that I forgot to run ksymoops over the sysrq+t output.
>
> the resolved output is attached. (it's big, so I had to gzip it.i sorry ).
>
> Also, I was wondering if the deadlock I encountered might whave been the
> IO-pausing under SMP systems that people have been talking about?
>
> thanks in advance.
>
> -m
>
> On Sat, Jul 12, 2003 at 01:10:02PM -0400, Marc Heckmann wrote:
> > Hi,
> >
> > I recently experienced a lockup on an SMP 2.4.21 kernel (vanilla).
> >
> > The machine was under very heavy IO at the time. (a full backup was in
> > progress). gzipped Alt+sysrq+t output is attached. I forgot
> > Alt+sysrq+P unfortunatly.
> >
> > iptables was still functional, it still responded to pings, but
> > everything else was locked up. obviously sysrq was still working. I
> > managed to reboot it with sysrq+b over the serial line.
> >
> > the machine has been in production for a while and has run in the past
> > for almost 2 months w/o any problems, so I don't believe it's hardware
> > related.
> >
> > A similar lockup happened once with the redhat 2.4.20-13.9 kernel. I
> > blamed that on all the extra patches present and switched to 2.4.21
> > hoping that it would alleviate the problem.
> >
> > PS: I'm not an linux-kernel so please CC me directly.
> >
> > -m
>
>


Attachments:
(No filename) (1.39 kB)
2.4.21-out.txt.gz (8.92 kB)
Download all attachments

2003-07-14 20:23:21

by Marc Heckmann

[permalink] [raw]
Subject: Re: 2.4.21 smp: system lockup

Hi there,

just realised that I forgot to run ksymoops over the sysrq+t output.

the resolved output is attached. (it's big, so I had to gzip it.i sorry ).

Also, I was wondering if the deadlock I encountered might whave been the
IO-pausing under SMP systems that people have been talking about?

thanks in advance.

-m

On Sat, Jul 12, 2003 at 01:10:02PM -0400, Marc Heckmann wrote:
> Hi,
>
> I recently experienced a lockup on an SMP 2.4.21 kernel (vanilla).
>
> The machine was under very heavy IO at the time. (a full backup was in
> progress). gzipped Alt+sysrq+t output is attached. I forgot
> Alt+sysrq+P unfortunatly.
>
> iptables was still functional, it still responded to pings, but
> everything else was locked up. obviously sysrq was still working. I
> managed to reboot it with sysrq+b over the serial line.
>
> the machine has been in production for a while and has run in the past
> for almost 2 months w/o any problems, so I don't believe it's hardware
> related.
>
> A similar lockup happened once with the redhat 2.4.20-13.9 kernel. I
> blamed that on all the extra patches present and switched to 2.4.21
> hoping that it would alleviate the problem.
>
> PS: I'm not an linux-kernel so please CC me directly.
>
> -m