Bug#433187: linux-2.6 - [sparc64-smp] produces unkillable processes

[Bernd Zeimetz <bernd@bzed.de>]

Hi,

> We (David Miller and I) are already working on this. We finally got some info
> dump from a debugging patched kernel and I expect we will have a fix within the
> next 3/4 weeks.
>>From our first look it seems like a futex bug and some users have reported that
> the latest 2.6.23-rcX do not show this behavior. Clearly we also want to figure
> out a fix for .22.

Last night an Ultrasparc II machine here shot itself again, it was
running fine for days with a .22 kernel which has was patched with the
diff from 179c85ea53bef807621f335767e41e23f86f01df
Seems this doesn't fix the problem completely, but it is not as bad as
before, it just takes much longer to produce hanging processes.

Behaviour is still much worse on Ultrasparc III machines like lebrun.d.o
and another machine here. Last kernel I've tried was 2.6.23-rc6-git7.
Since 179c85ea53bef807621f335767e41e23f86f01df behaviour became better,
the processes are only dead, before that patch the machine was pretty
much frozen and didn't even react on stop+a anymore.

Is there any new patch we can give a try?

Thanks,

Bernd

-- 
Bernd Zeimetz
<bernd@bzed.de>                         <http://bzed.de/>