Debugging tools/approach for GPU hangs?

[Bryce Harrington]

On Tue, Oct 20, 2009 at 04:04:46PM +0100, Matt Zimmerman wrote:
> On Fri, Sep 04, 2009 at 10:17:02AM +0100, Chris Jones wrote:
> > Hi
> > 
> > Matt Zimmerman wrote:
> > >    Bryce (and Jesse Barnes at Intel) mentioned that the kernel is now
> > >    supposed to log an error message when this happens, but I've never seen
> > >    evidence of that happening.
> > 
> > I had an X wedge on resuming my laptop the other day and after SSHing in
> > I found this in dmesg:
> > 
> > http://launchpadlibrarian.net/31253955/kernel.txt
> > 
> > (attached to https://bugs.launchpad.net/ubuntu/+source/linux/+bug/423362)
> 
> That's a symptom of this class of problem, yes.
> 
> > Is that the thing you're referring to? Having that kernel event trigger
> > an automatic dump into /var/crash seems like it would be a massive win
> > (cf my essentially useless bug report in the above case)
> 
> Yep, that's the idea.  We just need to add the necessary glue code to listen
> for the uevent and run the hook.  I've prototyped this but haven't been able
> to test it yet.

This is now live in lucid.  I uploaded last week before Thanksgiving.
However, I'm not spotting any bug reports in launchpad so far of -intel
freeze bugs.  So either it isn't kicking in, or no one is having freezes
on lucid with -intel, or hardly anyone is brave enough to upgrade to
lucid yet.

Anyway, I'd love to see a confirmed report of this working in the wild,
so we know the packaging is hooked up right and that the data is being
collected appropriately.  So if you have an X freeze and apport pops its
head up offering to report a bug about it, please do!

Matt/Jesse - if you know how to synthetically create a freeze, that
would be good to know for testing it.

Bryce