[Mesa-dev] gpu hung with mesa 9.1.2~git

Martin Steigerwald Martin at lichtvoll.de
Sun May 19 15:44:21 PDT 2013


Am Sonntag, 19. Mai 2013, 14:14:40 schrieb Kenneth Graunke:
> On 05/19/2013 03:59 AM, Martin Steigerwald wrote:
> > Hi Kenneth, Brian, everyone,
> > 
> > Am Samstag, 11. Mai 2013, 21:07:55 schrieb Martin Steigerwald:
> >> I got some hangs tough:
> >> 
> >> [151564.311483] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer
> >> elapsed... GPU hung [151564.311488] [drm] capturing error event; look
> >> for more information in/sys/kernel/debug/dri/0/i915_error_state
> >> [151664.323926] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer
> >> elapsed... GPU hung
> >> 
> >> [152099.395463] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer
> >> elapsed... GPU hung [152121.399082] [drm:i915_hangcheck_hung] *ERROR*
> >> Hangcheck timer elapsed... GPU hung
> >> 
> >> Could be quite easy to bisect, cause previous state of mesa 9.1 git
> >> I compiled didn´t have these hangs. I am right now using mesa 9.1
> >> branch as of 1e043ebe034b6e33b4c5228770db397b22fd26e0.
> >> 
> >> Kernel is 3.9.
> > 
> > Ok, I bisected this meanwhile:
> > 
> > martin at merkaba:~/Spielen/Mesa/mesa.git> git bisect good
> > 9f66038b5bac7f3574a00beda850be31712f9610 is the first bad commit
> > commit 9f66038b5bac7f3574a00beda850be31712f9610
> > Author: Kenneth Graunke <kenneth at whitecape.org>
> > Date:   Mon Apr 22 23:13:47 2013 -0700
> > 
> >      mesa: Add an unpack function for ARGB2101010_UINT.
> >      
> >      v2: Remove extra parenthesis (suggested by Brian).
> >      
> >      NOTE: This is a candidate for stable branches.
> >      
> >      Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=63569
> >      Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> >      Reviewed-by: Brian Paul <brianp at vmware.com>
> >      (cherry picked from commit 531be501de7530ede3c0ea7a1d94516df37e4ae5)
> 
> That's odd.  I didn't think PlaneShift used ARGB2101010_UINT.  It looks
> like CS may have support for it though, so I guess it's possible...
> 
> I'm still a bit skeptical of the bisect though.

Well I went back 9548c93768ec1cf5b9d589408d047559b463c2bb which should be one 
commit before the first bad one. I didn´t have a hang… well until I switched to 
compose mode in Kwin where it displays all windows in order to switch to 
ksnapshot to grab a screenie of game. That may the the same issue or it is 
unrelated. I tend to think it is unrelated, cause it happening during compose 
and I had no hang during regular game play. But it may still be the same hang 
and then the git-bisect result would be wrong.

I think I will keep playing with that one commit before the first bad one to 
verify whether the gpu hang is gone during normal play. If it is then I try 
again with the first bad commit.

Or do you have any other suggestion? I could also go to 9.1 branch HEAD and 
just revert this one commit…

It takes some time till a hang happens, about an hour or so, but sometimes 
sooner, somtimes later, so the bisect could be erratic.

Thanks,
-- 
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7


More information about the mesa-dev mailing list