kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

[Ferris McCormick]

[-- Attachment #1: Type: text/plain, Size: 634 bytes --]

A few days ago, I indicated that I was testing subject kernel on various
systems (U2-SMP, U60-SMP, SB1000-MP) for stability.  U60/SB1000 appear
to be stable.

However, U2-SMP(2x450) is not.  As is the case with ALL 2.6.xx kernels,
2.6.15-r4 cannot handle the sort of disk activity which portage requires
of it on a daily basis (although the daily slocate cron task no linger
kills it).  Symptom is as with all kernels in the 2.6.xx series ---
silent death with system completely non-responsive to anything.

Regards,
Ferris 
-- 
Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
Developer, Gentoo Linux (Sparc, Devrel)

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-sparc] kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

[Ferris McCormick]

[-- Attachment #1: Type: text/plain, Size: 1035 bytes --]

On Wed, 2006-02-15 at 08:28 -0500, Jameel Akari wrote:
> On Wed, 15 Feb 2006, Ferris McCormick wrote:
> 
> > However, U2-SMP(2x450) is not.  As is the case with ALL 2.6.xx kernels,
> > 2.6.15-r4 cannot handle the sort of disk activity which portage requires
> 
> Just as a data point, you do realize that the 450MHz CPU was never 
> actually supported in the Ultra-2?  While it may boot, it wasn't made to 
> work together nor was it sold as an upgrade. You may be tripping on some 
> marginal timing issue.
> 
> If you have 400MHz/2MB CPUs sitting around it may be worth testing with 
> them to isolate that as a cause.
> 

As is too often the case, I mistyped.  The stable system is U60(2x450);
the unstable system is U2(2x400).  Thanks for catching this.

> 
> -- 
> #!/jameel/akari
> sleep 4800;
> make clean && make breakfast
Regards,
Ferris (who sometimes expects way too much ESP support from his readers)
-- 
Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
Developer, Gentoo Linux (Sparc, Devrel)

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

[simoneau]

[-- Attachment #1: Type: text/plain, Size: 787 bytes --]

Quoting Ferris McCormick <fmccor@gentoo.org>:

> A few days ago, I indicated that I was testing subject kernel on various
> systems (U2-SMP, U60-SMP, SB1000-MP) for stability.  U60/SB1000 appear
> to be stable.
>
> However, U2-SMP(2x450) is not.  As is the case with ALL 2.6.xx kernels,
> 2.6.15-r4 cannot handle the sort of disk activity which portage requires
> of it on a daily basis (although the daily slocate cron task no linger
> kills it).  Symptom is as with all kernels in the 2.6.xx series ---
> silent death with system completely non-responsive to anything.
>
> Regards,
> Ferris
> --
> Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
> Developer, Gentoo Linux (Sparc, Devrel)
>

2.6.15-rc5 worked for me (over 100 iterations of crashme with no problems) on
U2-SMP 2x248.

[-- Attachment #2: PGP Digital Signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

RE: [gentoo-sparc] kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

[Leif Sawyer]

Ferris McCormick wrote in response to
> On Wed, 2006-02-15 at 08:28 -0500, Jameel Akari wrote:
>> On Wed, 15 Feb 2006, Ferris McCormick wrote:
>> 
>>> However, U2-SMP(2x450) is not.  As is the case with ALL 2.6.xx 
>>> kernels,2.6.15-r4 cannot handle the sort of disk activity which
>>> portage requires
>> 
>> If you have 400MHz/2MB CPUs sitting around it may be worth testing 
>> with them to isolate that as a cause.
>  
> As is too often the case, I mistyped.  The stable system is 
> U60(2x450); the unstable system is U2(2x400).  Thanks for 
> catching this.

I can also add that my U2 (2x300) is also having the same lockup
on high disk activity.  I've been able to 'nice' my portage related
activity down and avoid it crashing on the last few package emerges,
but it's not a cure-all.

RE: [gentoo-sparc] kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

[Jurij Smakov]

On Fri, 17 Feb 2006, Leif Sawyer wrote:

> I can also add that my U2 (2x300) is also having the same lockup
> on high disk activity.  I've been able to 'nice' my portage related
> activity down and avoid it crashing on the last few package emerges,
> but it's not a cure-all.

If you can easily reproduce this problem, could you try to get some useful 
debugging info? If you can drop into prom after machine hangs, what 
backtrace does the ctrace command give you? Could you try compiling the 
kernel with spinlock debugging ang magic sysrq key support and see if you 
can get some information on where it hangs that way?

We are hitting some kind of SMP lockup issues in Debian, however it is 
much harder to reproduce. The automatic sparc build machines (running 
2.6.8) occasionally hang when trying to build large packages, like 
openoffice.org. All attempts to reproduce this failure locally failed so 
far. As it might be caused by the same bug you are hitting, we would be 
*very* interested in any progress on this, since these hangs currently 
are the main obstacle which might prevent the inclusion of sparc port into 
the next official Debian release.

Best regards,

Jurij Smakov                                        jurij@wooyd.org
Key: http://www.wooyd.org/pgpkey/                   KeyID: C99E03CC

Re: kernel-2.6.15-r4 (in)stability on U2-SMP, etc.

[Christopher Zimmermann]

I'd like to try reproducing this on my U2-SMP(2x300). Could you give me
some instructions? I use Debian, so I just tried

find / -exec dd if='{}' of=/dev/null bs\x1000 count 0 \;

This should generate some heavy disk load, shouldn't it?
I also tried starting two such processes in parallel, but I did not
experience any crashes so far.
Maybe this is of some importance: I use the deadline scheduler.

Christopher

On Wed, Feb 15, 2006 at 01:09:15PM +0000, Ferris McCormick wrote:
> A few days ago, I indicated that I was testing subject kernel on various
> systems (U2-SMP, U60-SMP, SB1000-MP) for stability.  U60/SB1000 appear
> to be stable.
> 
> However, U2-SMP(2x450) is not.  As is the case with ALL 2.6.xx kernels,
> 2.6.15-r4 cannot handle the sort of disk activity which portage requires
> of it on a daily basis (although the daily slocate cron task no linger
> kills it).  Symptom is as with all kernels in the 2.6.xx series ---
> silent death with system completely non-responsive to anything.
> 
> Regards,
> Ferris 
> -- 
> Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
> Developer, Gentoo Linux (Sparc, Devrel)