Re: 2.6.9-rc1-mm4 kjournald oops (repeatable)

[Richard A Nelson]

On Wed, 8 Sep 2004, Lee Revell wrote:

> On Wed, 2004-09-08 at 19:07, Richard A Nelson wrote:
> > On Wed, 8 Sep 2004, Richard A Nelson wrote:
> >
> > > On Wed, 8 Sep 2004, Stephen C. Tweedie wrote:
> > >
> > > > On Wed, 2004-09-08 at 10:04, Andrew Morton wrote:
> > > >
> > > > > >   Unable to handle kernel paging request at virtual address 6b6b6b93
> > > > > > ...
> > > > > >   EIP: 0060:[__journal_clean_checkpoint_list+199/240]    Not tainted VLI
> > > > >
> > > > > This might have been caused by a fishy latency-reduction patch.  I today
> > > > > dropped that patch so could you please test next -mm and let me know?
> > > >
> > > > That, or preempt.  If the next -mm still breaks, time to hunt for the
> > > > preempt problem, I guess.
> > >
> > > Ok, if it still fails (I'll have to wait until this afternoon for the
> > > true test - dpkg breaks it everytime), I'll check out preempt.
> >
> > Well, it looks like backing out the patch was sufficient, I've made it
> > through the torture that is a dpkg install (70+meg).
> >
> > So we needn't (at this time) look to preempt.
>
> Hmm, I have been running this patch for weeks as part of the voluntary
> preemption patches, and put it through every torture test I can think
> of, with nary an Oops.  None of the other VP testers have reported
> problems either.  Maybe this is some interaction between that patch and
> something else in -mm.

Interestingly, I notice Zwane had a very similiar oops, posted on the
7th: Oops in __journal_clean_checkpoint_list
He also had preempt enabled...

I've found upgrading my Debian system using dselect to be a *very* good
stress test of the filesystem...

If you have candidates, I'll try to test them - I've typically had no
problem reproducing the issue :)

-- 
Rick Nelson
 * Equivalent code is available from RSA Data Security, Inc.
 * This code has been tested against that, and is equivalent,
 * except that you don't need to include two pages of legalese
 * with every copy.
        -- public domain MD5 source
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/