intel gpu freeze (dumps/logs attached to upstream bug)

[Martin Olsson]

Filing downstream bug for tracking purposes only (maybe upstream bug is fixed right before karmic and we want to cherry pick).

https://bugs.freedesktop.org/show_bug.cgi?id=23897

[Martin Olsson (mnemo)]

Created an attachment (id=29463)
gdm.log (captured using ssh while frozen)

[Martin Olsson (mnemo)]

Created an attachment (id=29464)
i915_batchbuffers (captured using ssh while frozen)

[Martin Olsson (mnemo)]

Created an attachment (id=29465)
i915_error_state.txt

[Martin Olsson (mnemo)]

Created an attachment (id=29466)
gem_active

[Martin Olsson (mnemo)]

Created an attachment (id=29467)
fence_regs

[Martin Olsson (mnemo)]

Created an attachment (id=29468)
flushing

[Martin Olsson (mnemo)]

Created an attachment (id=29469)
hws

[Martin Olsson (mnemo)]

Created an attachment (id=29470)
inactive

[Martin Olsson (mnemo)]

Created an attachment (id=29471)
interrupt

[Martin Olsson (mnemo)]

Created an attachment (id=29472)
request

[Martin Olsson (mnemo)]

Created an attachment (id=29473)
seqno

[Martin Olsson (mnemo)]

Created an attachment (id=29474)
ringbuf_data

[Martin Olsson (mnemo)]

Created an attachment (id=29475)
ringbuf_info

[Martin Olsson (mnemo)]

Created an attachment (id=29476)
intel_gpu_dump.txt.gz

[Martin Olsson (mnemo)]

Created an attachment (id=29477)
xorg.log

[Martin Olsson (mnemo)]

Created an attachment (id=29482)
dri_debug.tgz (the freeze just happened again, here is another set of logs)

[Martin Olsson (mnemo)]

Created an attachment (id=29483)
dri_debug.tgz (after the second freeze the machine froze a third time right after I restored my firefox session)

[Martin Olsson (mnemo)]

To my untrained eye the freeze dumps I've posted in this bug looks somewhat similar to the GPU freeze I in the end of August (also using karmic):
https://bugs.freedesktop.org/show_bug.cgi?id=23589

[Martin Olsson (mnemo)]

One additional thing; when the machine froze the third time (just 1-2 minutes after rebooting from the second freeze) I had just restored my GNOME session which started gedit and a gnome-terminal and it also re-launched my Firefox session with around 10 TABs in one window. One of the TABs in Firefox had a 6507x1787 pixel PNG in it (I guess that's not huge but it's bigger than the average texture maybe?).

[Martin Olsson (mnemo)]

Filing downstream bug for tracking purposes only (maybe upstream bug is fixed right before karmic and we want to cherry pick).

https://bugs.freedesktop.org/show_bug.cgi?id=23897

[Gordon Jin (gordon-jin)]

Does the patch in http://bugs.freedesktop.org/show_bug.cgi?id=22336#c18 help?

[Eric Anholt (eric-anholt)]

commit 5604b27b9326ac542069a49ed9650c4b0d3e939a
Author: Eric Anholt <email address hidden>
Date: Wed Sep 9 12:35:30 2009 -0700

    i965: Fix relocation delta for WM surfaces.

Please retest with this patch in Mesa, as it fixed issues when the aperture was near full (big images being a cause)

[Martin Olsson (mnemo)]

Ubuntu karmic cherry picked that "Fix relocation delta for WM surfaces" patch a few days ago so it's getting some nice testing right now before the upcoming Ubuntu beta.

I've been running with compiz off for the last few days (because I was previously getting _very_ frequent crashes while frequently generating and viewing huge PNG file for a project I'm working on).

Right now I've been running with compiz on, plus the patch applied and so far things looks great. No crashes or hangs and I've generated several big graphs, opened them in Firefox and Chrome, tried to stress it a bit with the compiz ring switcher and a few different games. So far, it feels pretty solid to me.

I'll report it here if the GPU hangs again with the patch applied.

[Martin Olsson (mnemo)]

I suspect this GPU hang might have been fixed by:

mesa (7.6.0~git20090817.7c422387-0ubuntu4) karmic; urgency=low

  * Add 109_fix_relocation_delta_for_wm_surfaces.patch: Fix relocation
    delta for WM surfaces. This was a regression introduced in
    0f328c90dbc893e15005f2ab441d309c1c176245 (LP: #429241).

 -- Alberto Milone <email address hidden> Tue, 15 Sep 2009 18:14:48 +0200

[Eric Anholt (eric-anholt)]

OK, I'll go ahead and close it as fixed for now. Feel free to reopen if you see the same problem again. Thanks for testing!

[Bryce Harrington (bryce)]

Great, sounds like it's no longer an issue so closing bug. Feel free to reopen if it continues to be a problem.