2005-12-26 11:58:00

by Tarkan Erimer

[permalink] [raw]
Subject: [BUG]: Hard lockups continue with linux-2.6.15-rc1-rc7

Hi all,

I'm having hard lockups with all the RCs of linux-2.6.15. I,
previously, mentioned this with the subject "[BUG]: Software compiling
occasionlly hangs under 2.6.15-rc1/rc2 and 2.6.15-rc1-mm2" in the
list. I investigated a bit at found these interesting things.

-- Always reproducable. To reproduce:
- in console 1, issueing "updatedb"
- in console 2, issueing "find / -name "blahblah" -print
- in console 3, issueing "emerge -uDp world" (BTW, I'm using Gentoo.)
- in console 4, X started.
- a few minutes later, system completely freezes. No Alt+SysRq+t
works. (Normally, it does)

When the system freezes, there is nothing in logs. But hardly, I
captured an Alt+SysRq+t. A few seconds (15-20 seconds) before hang. I
attached this Alt+SysRq+t and lsmod output. Hope this helps to solve
this.

PS: These problems never occured in 2.6.14.xx and downwards.

Regards.


Attachments:
(No filename) (895.00 B)
lsmod.out (1.99 kB)
trace.out.bz2 (4.53 kB)
Download all attachments

2005-12-26 12:48:08

by Pekka Enberg

[permalink] [raw]
Subject: Re: [BUG]: Hard lockups continue with linux-2.6.15-rc1-rc7

Hi Tarkan,

On 12/26/05, Tarkan Erimer <[email protected]> wrote:
> I'm having hard lockups with all the RCs of linux-2.6.15. I,
> previously, mentioned this with the subject "[BUG]: Software compiling
> occasionlly hangs under 2.6.15-rc1/rc2 and 2.6.15-rc1-mm2" in the
> list. I investigated a bit at found these interesting things.
>
> -- Always reproducable. To reproduce:
> - in console 1, issueing "updatedb"
> - in console 2, issueing "find / -name "blahblah" -print
> - in console 3, issueing "emerge -uDp world" (BTW, I'm using Gentoo.)
> - in console 4, X started.
> - a few minutes later, system completely freezes. No Alt+SysRq+t
> works. (Normally, it does)
>
> When the system freezes, there is nothing in logs. But hardly, I
> captured an Alt+SysRq+t. A few seconds (15-20 seconds) before hang. I
> attached this Alt+SysRq+t and lsmod output. Hope this helps to solve
> this.

You can use git bisect to narrow down the changeset that introduced
the bug. Please refer to the following URL for details:
http://www.kernel.org/pub/software/scm/git/docs/howto/isolate-bugs-with-bisect.txt

Pekka

2005-12-26 21:28:08

by Sid Boyce

[permalink] [raw]
Subject: RE: [BUG]: Hard lockups continue with linux-2.6.15-rc1-rc7

Tarkan Erimer wrote:
> Hi all,
>
> I'm having hard lockups with all the RCs of linux-2.6.15. I,
> previously, mentioned this with the subject "[BUG]: Software compiling
> occasionlly hangs under 2.6.15-rc1/rc2 and 2.6.15-rc1-mm2" in the
> list. I investigated a bit at found these interesting things.
>
> -- Always reproducable. To reproduce:
> - in console 1, issueing "updatedb"
> - in console 2, issueing "find / -name "blahblah" -print
> - in console 3, issueing "emerge -uDp world" (BTW, I'm using Gentoo.)
> - in console 4, X started.
> - a few minutes later, system completely freezes. No Alt+SysRq+t
> works. (Normally, it does)
>
> When the system freezes, there is nothing in logs. But hardly, I
> captured an Alt+SysRq+t. A few seconds (15-20 seconds) before hang. I
> attached this Alt+SysRq+t and lsmod output. Hope this helps to solve
> this.
>
> PS: These problems never occured in 2.6.14.xx and downwards.
>
> Regards.

Don't rule out hardware. This SuSE 10.0 x86 box worked without problems
on kernels up to 2.6.15-rc6-git2, but I experienced strange apparent
filesystem corruptions/compile failures running normally and hard
lockups when running mythtv with 2.6.15-rc6-git6 and 2.6.15-rc7, while
on the Mandriva 2006 x86 box and the SuSE x86_64 there were no problems.
Until I found the suspect SDRAM, on some occasions I had to run
reiserfsck before 2.6.15-rc6-git2 would boot again correctly after
trying rc6-git6 or -rc7. Finally I got a corruption again with
2.6.15-rc7, replaced the SDRAM stick with the one taken out previously,
booted up on 2.6.15-rc7 with no problems. I had run memtest some days
earlier, but only for a couple of hours. (current uptime 1 day 1.04hrs).
Regards
Sid.
--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Retired IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

2005-12-29 00:06:12

by Tarkan Erimer

[permalink] [raw]
Subject: Re: [BUG]: Hard lockups continue with linux-2.6.15-rc1-rc7

Hi Sid,

On 12/26/05, Sid Boyce <[email protected]> wrote:
> Don't rule out hardware. This SuSE 10.0 x86 box worked without problems
> on kernels up to 2.6.15-rc6-git2, but I experienced strange apparent
> filesystem corruptions/compile failures running normally and hard
> lockups when running mythtv with 2.6.15-rc6-git6 and 2.6.15-rc7, while
> on the Mandriva 2006 x86 box and the SuSE x86_64 there were no problems.
> Until I found the suspect SDRAM, on some occasions I had to run
> reiserfsck before 2.6.15-rc6-git2 would boot again correctly after
> trying rc6-git6 or -rc7. Finally I got a corruption again with
> 2.6.15-rc7, replaced the SDRAM stick with the one taken out previously,
> booted up on 2.6.15-rc7 with no problems. I had run memtest some days
> earlier, but only for a couple of hours. (current uptime 1 day 1.04hrs).

Hmmm.. It looks, it is time to run memtest on my box. Thanks for the tip.

Regards,

2005-12-29 00:29:43

by Sid Boyce

[permalink] [raw]
Subject: Re: [BUG]: Hard lockups continue with linux-2.6.15-rc1-rc7

Tarkan Erimer wrote:
> Hi Sid,
>
> On 12/26/05, Sid Boyce <[email protected]> wrote:
>> Don't rule out hardware. This SuSE 10.0 x86 box worked without problems
>> on kernels up to 2.6.15-rc6-git2, but I experienced strange apparent
>> filesystem corruptions/compile failures running normally and hard
>> lockups when running mythtv with 2.6.15-rc6-git6 and 2.6.15-rc7, while
>> on the Mandriva 2006 x86 box and the SuSE x86_64 there were no problems.
>> Until I found the suspect SDRAM, on some occasions I had to run
>> reiserfsck before 2.6.15-rc6-git2 would boot again correctly after
>> trying rc6-git6 or -rc7. Finally I got a corruption again with
>> 2.6.15-rc7, replaced the SDRAM stick with the one taken out previously,
>> booted up on 2.6.15-rc7 with no problems. I had run memtest some days
>> earlier, but only for a couple of hours. (current uptime 1 day 1.04hrs).
>
> Hmmm.. It looks, it is time to run memtest on my box. Thanks for the tip.
>
> Regards,
>
>
It's strongly recommended. I've now found that both SDRAMS with the same
batch number are experiencing the same error with memtest, that's after
I had a lockup with the one I thought was good. They will be replaced
under lifetime warranty by Kingston memory. I've temporarily stuck a
PC2700 512M stick in and it's been solid under memtest and with a full
workload.
Regards
Sid.
--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Retired IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks