Created attachment 92822 [details]
gzipped output of /sys/class/drm/card0/error after gpu crash
Lately, from time to time my screen turns suddenly black. The system still works, I can access it remotely and shutdown cleanly. This happens quite often when I try to run youtube videos with flash, so this may be related.
In the kernel log I can read this:
Jan 26 20:25:46 pc kernel: [drm] stuck on render ring
Jan 26 20:25:46 pc kernel: [drm] GPU crash dump saved to /sys/class/drm/card0/error
Jan 26 20:25:46 pc kernel: [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan 26 20:25:46 pc kernel: [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan 26 20:25:46 pc kernel: [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan 26 20:25:46 pc kernel: [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Jan 26 20:25:46 pc kernel: [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x2e66000 ctx 0) at 0x2e66f34
Jan 26 20:25:46 pc kernel: [drm:i915_reset] *ERROR* Failed to reset chip.
I will attach the output of
xorg-server is 1.15.0, kernel 3.13, xf86-video-intel 2.99.907, mesa 9.2.5, libdrm 2.4.52. Distribution is Gentoo Linux.
Forgot to mention: My graphics card (lspci output) is
00:02.0 VGA compatible controller: Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (primary) (rev 0c)
In a Thinkpad T61 laptop.
*** This bug has been marked as a duplicate of bug 73348 ***
Too fast. Different bug.
Erm, the error state is broken. GRRR.
That error state is unusable for bug diagnosis.
> That error state is unusable for bug diagnosis.
Okay, so how do I get one that is usable? You've marked this bug "needinfo", but you haven't told me what kind of additional info you need.
An error state from 3.12 should do fine. And if it doesn't hang then we know it is a kernel regression and not the mesa batchbuffer.
What works for me when my GM965 GPU hangs is to suspend and resume. That might be less inconvenient than rebooting your machine.
Created attachment 92843 [details]
gzipped output of /sys/class/drm/card0/error with kernel 3.12.9
Attached is the error-output with kernel 3.12.9.
One additional information: I was wrong above, this isn't triggered by a flash-video, but by a html5-video on youtube. Flash videos work, so it seems it's something in chromiums internal video playback that triggers this error. I also now have a video that always triggers this bug, so I have reproducibility, which should make further investigation easier.
Is there any more info from me you need to investigate this bug?
I have tried a number of things:
* Downgrading to the latest stable xf86-video-intel 2.21.15
* Upgrading mesa to version 10.0.0.2
* Upgrading both xf86-video-intel and mesa to latest git code
* Switching from sna to uxa
None of this changed anything, the bug persists. This is pretty severe, as it can happen all the time when I surf the net that some video crashes my system.
*** This bug has been marked as a duplicate of bug 80568 ***