Bug 83408 - [82845G/GL] [drm:i915_report_and_clear_eir] GPU Hang *ERROR* EIR stuck: 0x00000010
Summary: [82845G/GL] [drm:i915_report_and_clear_eir] GPU Hang *ERROR* EIR stuck: 0x000...
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: XOrg git
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-09-02 17:46 UTC by bd
Modified: 2017-07-24 22:51 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
bzip2 < /sys/class/drm/card0/error > error.bz2 (90.04 KB, text/plain)
2014-09-02 17:46 UTC, bd
no flags Details
lspci -vvv (7.40 KB, text/plain)
2014-09-02 17:46 UTC, bd
no flags Details
X log (20.04 KB, text/plain)
2014-09-02 17:47 UTC, bd
no flags Details

Description bd 2014-09-02 17:46:32 UTC
Created attachment 105627 [details]
bzip2 < /sys/class/drm/card0/error > error.bz2

This is on a Dell Optiplex SX 260 with a fresh Debian Jessie install. Starting lightdm leads to screen becoming black with visible mouse pointer.

[ 2699.843537] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[ 2699.843546] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[ 2699.843548] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[ 2699.843550] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[ 2699.843552] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[ 2699.844007] i915: render error detected, EIR: 0x00000010
[ 2699.844007] [drm:i915_report_and_clear_eir] *ERROR* EIR stuck: 0x00000010, masking
[ 2699.844007] i915: render error detected, EIR: 0x00000010
Comment 1 bd 2014-09-02 17:46:53 UTC
Created attachment 105628 [details]
lspci -vvv
Comment 2 bd 2014-09-02 17:47:12 UTC
Created attachment 105629 [details]
X log
Comment 3 Chris Wilson 2014-09-02 18:34:37 UTC
That shouldn't be a fatal hang, and X looks quite happy. I think the blank screen is more indicative of lightdm failing. Are you able to run anything other than lightdm?
Comment 4 bd 2014-09-03 12:31:20 UTC
I tried xdm and that worked.

Also with xdm I haven't seen the "stuck" message in dmesg.
Comment 5 Jani Nikula 2014-09-11 16:19:36 UTC
Chris, further ideas?
Comment 6 Chris Wilson 2014-09-11 20:34:44 UTC
No, it could just be an invalid surface allocation inside mesa - but we should be protecting what we reference inside the GTT. The hardware rarely gives a good indication of the exact cause of the fault.
Comment 7 Rodrigo Vivi 2014-10-15 20:26:31 UTC
Coul you please test a more recent kernel and collect updated logs and error state?
Preferably use drm-intel-nightly branch from cgit.freedesktop.org/drm-intel
Comment 8 bd 2014-10-20 09:47:08 UTC
I tried with the debian 3.16-2-686-pae kernel and the screen still stays black with mouse cursor when I start lightdm.

However the cursor changes from pointer to text input depending on the position. It looks to me, that the login mask is, for the lack of a better word, rendered, but not displayed.

dmesg:

[   72.584020] [drm] GPU HANG: ecode -1:0x00000000, reason: Command parser error, iir 0x00008000, action: continue
[   72.584020] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[   72.584020] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[   72.584020] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[   72.584020] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[   72.584020] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[   72.584020] i915: render error detected, EIR: 0x00000010
[   72.584020] [drm:i915_report_and_clear_eir] *ERROR* EIR stuck: 0x00000010, masking
[   72.584020] [drm] GPU HANG: ecode -1:0x00000000, reason: Command parser error, iir 0x00008000, action: continue
[   72.584020] i915: render error detected, EIR: 0x00000010

FWIW, on a freshly started system with no display manager running, when I first start xdm, stop it and start lightdm the screen does not stay black but shows the login mask.
Comment 9 bd 2014-10-20 09:49:26 UTC
I also tried with 3.17.0+ from cgit.freedesktop.org/drm-intel branch drm-intel-nightly commit 1361e35 from Thu Oct 16 15:07:58 2014 +0300

This fixes the black screen problem with lightdm for me.

All I see in dmesg is:

[   12.931882] [drm] Initialized drm 1.1.0 20060810
[   16.027853] [drm] Memory usable by graphics device = 128M
[   16.027862] [drm] Replacing VGA console driver
[   16.030464] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[   16.030468] [drm] Driver supports precise vblank timestamp query.
[   16.040367] [drm] initialized overlay support
[   16.040428] [drm] Initialized i915 1.6.0 20141003 for 0000:00:02.0 on minor 0
[   16.226977] fbcon: inteldrmfb (fb0) is primary device
[   16.256457] i915 0000:00:02.0: fb0: inteldrmfb frame buffer device
Comment 10 Rodrigo Vivi 2015-01-15 00:15:04 UTC
So this seems a fixed bug for me.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.