Bug#433187: linux-2.6 - [sparc64-smp] produces unkillable processes
Josip Rodin wrote:
> On Tue, Sep 04, 2007 at 06:16:05AM +0200, Fabio Massimo Di Nitto wrote:
>>>> #433187 is the bug that has killed the buildds on lebrun and spontini, right?
>>> AIUC, yes. at least i can reproduce that on my buildd.
>> Hi guys,
>> 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.
> I should mention that lebrun.d.o is still dead since the last attempt
> (ssh unresponsive since 2007-08-30 ~21:25), when it was running a 188.8.131.52
> with one davem patch applied (one line in kernel/futex_compat.c). If you
> need something more done to lebrun, such as kicking it back to life,
> just tell me...
If you have console access, it would be good to get a processor dump by break + p.
anyway we were able to reproduce the problem by doing some fancy building on
Niagara and that already isolate the problems to a more generic bit of the code
rather than CPU specific.
I personally have no say on how buildds should be managed.. i guess it's up to
you guys if you want to kick it back. If you do so just make sure you can grab
CPU register dumps from console.
I'm going to make him an offer he can't refuse.