Bug 84615 - [SNB gt1] GPU HANG: ecode 0:0x85fffffc on Intel P3000 (Xeon E3 V1) on first X startup
Summary: [SNB gt1] GPU HANG: ecode 0:0x85fffffc on Intel P3000 (Xeon E3 V1) on first X...
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: Other All
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-10-03 01:13 UTC by mauorrizze
Modified: 2017-07-24 22:51 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
GPU hang crash dump after X startup from /sys/class/drm/card0/error (2.04 MB, text/plain)
2014-10-03 01:13 UTC, mauorrizze
no flags Details
last dmesg lines after GPU hang (711 bytes, text/plain)
2014-10-03 01:15 UTC, mauorrizze
no flags Details
dmesg | grep drm: kernel 3.14 (working) vs. 3.15 (hang) (2.26 KB, text/plain)
2014-10-03 21:31 UTC, mauorrizze
no flags Details
crash dump in Arch with 3.15.8 kernel (2.00 MB, text/plain)
2014-10-03 21:32 UTC, mauorrizze
no flags Details
crash dump in fedora 21 (2.13 MB, text/plain)
2014-10-03 21:33 UTC, mauorrizze
no flags Details
crash dump in Ubuntu 14.10 (2.13 MB, text/plain)
2014-10-03 21:34 UTC, mauorrizze
no flags Details

Description mauorrizze 2014-10-03 01:13:56 UTC
Created attachment 107241 [details]
GPU hang crash dump after X startup from /sys/class/drm/card0/error

When I start X, dmesg throws an GPU ring hang / reset message, similar to #76368. But it isn't fixed for me and setting i915.enable_ppgtt=0 changes nothing.
When starting, Awesome WM draws its menus with a little delay and not correctly, a quick restart solves this. Starting gnome 3, there's a longer delay (~5-8 sec), showing only a black screen but after that it starts (or restarts?) normally.
Used kernels: Arch Linux default 3.16 and mainline 3.17-rc6.
Comment 1 mauorrizze 2014-10-03 01:15:19 UTC
Created attachment 107242 [details]
last dmesg lines after GPU hang
Comment 2 Chris Wilson 2014-10-03 06:19:10 UTC
Can you try disabling systemd-logind and starting X (or gdm) by hand from the vt?
Comment 3 mauorrizze 2014-10-03 13:23:18 UTC
It seems starting X without logind became quite hard in arch. I've already been using startx from vt, but even after disabling and stopping systemd-logind and reducing .xinitrc to "exec awesome", it's still executed. I think I had to replace systemd completely.
But, I've tested several live images to verify the problem and get more kernel/systemd combinations.

Fedora 21 alpha (3.16.something, systemd): 
GPU HANG: ecode 0:0x85fffffc, in Xorg.bin [1238]

Ubuntu 14.10 daily image (3.16.0-17, kind of beta systemd?):
GPU HANG: ecode 0:0x85fffff8, in Xorg [2438]

Ubuntu 14.04 (3.13.something, upstart): NO gpu hang

Fedora 20 (3.11.10, systemd): NO gpu hang

What's missing is a current 3.16+ kernel without systemd, if you know a distro or easy method of configuring that combination, I'm willing to test. But from my point of view it seems the problem starts somewhere between kernel 3.13 and 3.16. Should I compile older mainline kernels for my arch linux installation? Crash dumps from fedora 21 and Ubuntu 14.10 available, if interested.
Comment 4 Chris Wilson 2014-10-03 19:43:53 UTC
Trying an older kernel or two would be useful indeed. And please do attach a couple more error states. The first implies some GPU state was left stale, I think, but it is not clear, maybe another state would have something more obviously amiss.
Comment 5 mauorrizze 2014-10-03 21:29:43 UTC
The problem starts in 3.15. In 3.14.6-1 (downgraded Arch default kernel) no hang occurs.
Comment 6 mauorrizze 2014-10-03 21:31:15 UTC
Created attachment 107283 [details]
dmesg | grep drm: kernel 3.14 (working) vs. 3.15 (hang)
Comment 7 mauorrizze 2014-10-03 21:32:45 UTC
Created attachment 107284 [details]
crash dump in Arch with 3.15.8 kernel
Comment 8 mauorrizze 2014-10-03 21:33:29 UTC
Created attachment 107285 [details]
crash dump in fedora 21
Comment 9 mauorrizze 2014-10-03 21:34:02 UTC
Created attachment 107286 [details]
crash dump in Ubuntu 14.10
Comment 10 Mika Kuoppala 2014-11-05 16:00:55 UTC
mauorrizze, could you please try with:

https://bugs.freedesktop.org/attachment.cgi?id=108894
Comment 11 mauorrizze 2014-11-16 01:33:46 UTC
Lots of thanks Mika, the patch works for me (tested with mainline 3.17.3).
Awesome WM doesn't need a restart after X's first startup and there is no "GPU HANG" in dmesg.
This bug then might be a duplicate of 79996 and seems to be fixed.
Comment 12 Jani Nikula 2014-11-17 08:43:40 UTC
commit 2208d655a91f9879bd9a39ff9df05dd668b3512c
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Nov 14 09:25:29 2014 +0100

    drm/i915: drop WaSetupGtModeTdRowDispatch:snb


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.