[ubuntu-x] Intel GPU hangs and batchbuffer dumps

[Geir Ove Myhr]

On Fri, Feb 19, 2010 at 1:48 AM, Bryce Harrington <bryce at canonical.com> wrote:
> This is a pretty simple little python script, so should be
> straightforward to add to.  If there is any information missing that its
> not collecting please send me a patch (or describe what else it needs to
> do).

Could you attach /sys/kernel/debug/dri/*/i915_error_state if it
exists? I have 0 and 64 for * on my system, and I don't really know
the difference, but we have been asking for files in 0 before. I don't
trust my python skills enough to send a patch.

I've got the self-compiled kernel with the record-batchbuffer patch to
boot after an update of Lucid.
With that patch the relevant information is captured in
i915_error_state after a GPU hang.

I'm also not familiar with the timing of apport in this case. If the
bug is filed to LP only after next reboot, dmesg will have no trace of
the GPU hang. For example, I can't see any GPU hang messages in the
logs of bug 524211 (gem_hang test). If so, we should also save dmesg
and possibly Xorg.0.log a the time of the hang (just like the
procedure on Troubleshooting/Freezes used to do).

Geir Ove