ubuntu-x-swat team mailing list archive

[Bryce Harrington <bryce@xxxxxxxxxxxxx>]

On Tue, Mar 15, 2011 at 01:41:06PM -0000, Benjamin Redelings wrote:
> On 03/15/2011 05:45 AM, Bryce Harrington wrote:
> > Hi Benjamin, unfortunately in this case the automatic bug collection
> > didn't grab valid id's from your GPU.  Often times with these kinds of
> > bugs you'll encounter them multiple times, and sometimes apport has a
> > better chance of collecting the right info.  So, I'm going to close this
> > one out since it's (probably) not actionable, but can I ask that
> > whenever you encounter this gpu lockup and get shown a dialog to report
> > it, please do.  Hopefully in one of those will be proper error codes.
> >
> > ** Changed in: xserver-xorg-video-intel (Ubuntu)
> >         Status: New =>  Invalid
> >
> Thanks for the feedback, Bryce.
> 
> Yup, this happens about every two hours when using compiz.
> 
> Usually when this happens, I get some kind of message saying something 
> like "gpu_apport_intel.py failed".  (I probably have not exactly 
> remembered the name of the *.py script.)  But, given that this script 
> has failed so many times, is there any reason to suppose it will start 
> working next time?
> 
> However, I will try again to upload a report, if you think that there is 
> a chance of this working.

Btw, on another bug report I see the reason the hook fails is because
the system runs out of memory, so the utility simply can't allocate the
memory it needs to operate.  My best guess is that's what's happening in
this case too.

If you want, you could try sshing into the locked up box and collecting
the data manually.  You'll need to:

Kill X and anything else that might be consuming memory to make some
space:

  sudo kill -9 `pidof X`

copy this file:

  sudo cp /sys/kernel/debug/dri/0/i915_error_state ~/i915_error_state.txt

and run this command:

  sudo intel_gpu_dump > ~/intel_gpu_dump.txt

Attach both to this bug report.