Re: open issues with the hppa port
On Mon, Sep 14, 2009 at 3:56 PM, dann frazier<firstname.lastname@example.org> wrote:
> On Mon, Sep 14, 2009 at 03:36:48PM -0400, Carlos O'Donell wrote:
>> On Mon, Sep 14, 2009 at 1:26 PM, dann frazier <email@example.com> wrote:
>> > After about a week of loaded uptime, penalosa returned to instability
>> > over the weekend.
>> > It began with an Illegal instruction leading to a panic:
>> > http://people.debian.org/~dannf/penalosa/penalosa.0
>> > I then rebooted it, and the console was hung immediately after boot:
>> > http://people.debian.org/~dannf/penalosa/penalosa.1
>> > Note the rc.local segfault there.
>> > I rebooted it again, and it crashed after only a few minutes:
>> > http://people.debian.org/~dannf/penalosa/penalosa.2
>> Exactly how did you install the kernel and what sources were used to build it?
> I used dpkg -i to install it, sources are here:
> To create an unpacked source tree, you can:
> dget http://people.debian.org/~dannf/tmp/linux-2.6_2.6.31~rc6-1~experimental.1~snapshot.14160.dsc
> cd linux-2.6-2.6.31~rc6
> ./debian/rules setup
> It is a pretty vanilla 2.6.31-rc6.
>> The trace looks familiar (crash in unwind_once), and I thought it had
>> been fixed by Kyle already. I will put together a kernel for you to
>> boot on penalosa.
> I don't see any obvious fixes between 2.6.31-rc6 & 2.6.31 - is the fix in kyle's tree?
I don't know, that will be part of my analysis.
>> > peri (identical hardware/kernel) has been up for about 20 days now.
>> Please issue a sysrq-t (ctrl^c, send break, t) on peri's console and
>> tell me if it crashes the kernel?
> Yes, it does. (echo t > /proc/sysrq-trigger didn't though, for whatever reason).
Good, it's got the same bug then :-)
Please note that fixing the kernel crash will not fix the userspace
stability issues, but will prevent the kernel from crashing.