Re: [OOPS] 2.6.21-rc6-git5 in cfq_dispatch_insert

From: Roland Kuhn
Date: Wed Apr 25 2007 - 09:59:32 EST


Hi Jens!

On 25 Apr 2007, at 12:18, Jens Axboe wrote:

On Wed, Apr 25 2007, Brad Campbell wrote:
Jens Axboe wrote:

It looks to be extremely rare. Aliases are extremely rare, front merges
are rare. And you need both to happen with the details you outlined. But
it's a large user base, and we've had 3-4 reports on this in the past
months. So it obviously does happen. I could not make it trigger without
doctoring the unplug code when I used aio.

Well, not that rare on this particular machine (I had a case where I could
reproduce it in less than an hour of normal use previously on this box),
and I've had it occur a number of times on my servers, I just never
reported it before as I never took the time to set up a serial console and
capture the oops.

Extremely rare in the sense that it takes md and some certain conditions
to happen for it to trigger. So for most people it'll be extremely rare,
and for others (such as yourself) that hit it, it wont be so rare :-)

Here's a fix for it, confirmed.

Shall I leave the other debugging in, apply this and run it for a few hard
days?

Yes, that would be perfect!

Okay, I left all debugging patches in, disabled all kernel debugging .config stuff and gave it a spin with our usual "killer" workload (as far as batch systems are repeatable anyway) and so far there was not a single glitch or message, so I preliminarily conclude that the bug is squashed. The final word will come once my 1800 batch jobs are processed and I have my particle physics result ;-)

Ciao,
Roland

--
TU Muenchen, Physik-Department E18, James-Franck-Str., 85748 Garching
Telefon 089/289-12575; Telefax 089/289-12570
--
CERN office: 892-1-D23 phone: +41 22 7676540 mobile: +41 76 487 4482
--
Any society that would give up a little liberty to gain a little
security will deserve neither and lose both. - Benjamin Franklin
-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GS/CS/M/MU d-(++) s:+ a-> C+++ UL++++ P+++ L+++ E(+) W+ !N K- w--- M + !V Y+
PGP++ t+(++) 5 R+ tv-- b+ DI++ e+++>++++ h---- y+++
------END GEEK CODE BLOCK------


Attachment: smime.p7s
Description: S/MIME cryptographic signature

Attachment: PGP.sig
Description: This is a digitally signed message part