Bug 63666 - [3.2-stable 3.5-stable regression SNB] TLB invalidate strikes back
Summary: [3.2-stable 3.5-stable regression SNB] TLB invalidate strikes back
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-04-18 04:50 UTC by petrelharp
Modified: 2017-07-24 22:58 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
backtrace from gdb and accompanying Xorg bt. (14.49 KB, text/plain)
2013-04-18 04:50 UTC, petrelharp
no flags Details
hung i915 error state (2.06 MB, text/plain)
2013-04-18 15:42 UTC, petrelharp
no flags Details
Xorg.0.log that accompanies i915_error_state (33.25 KB, text/plain)
2013-04-18 15:54 UTC, petrelharp
no flags Details

Description petrelharp 2013-04-18 04:50:28 UTC
Created attachment 78163 [details]
backtrace from gdb and accompanying Xorg bt.

Hardware: lenovo E530c ("Intel Corporation 2nd Generation Core Processor Family Integrated Graphics Controller")
driver: i915
OS: debian 3.2.0-4-amd64
xserver-xorg-video-intel: 2:2.20.14-1

Symptom:

1) Open libreoffice writer; new document.
2) Make a few new pages (ctl-enter).
3) Scroll up and down rapidly with the touchpad a few times.
4) X freezes, with two outcomes: (a) either a hung GPU and all windows black (but, mouse still moves); or else (b) totally frozen X with "[mi] EQ overflow continuing." and backtrace in Xorg.0.log.

This doesn't occur on a similar lenovo machine (T420i).  It doesn't seem to happen in other applications; libreoffice folks determined it to be not their problem ( https://bugs.freedesktop.org/show_bug.cgi?id=62946 ) and suggested the intel drivers.

Logs for case (a) are at http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=10;att=0;bug=703749 and for case (b) are at http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=15;att=0;bug=703749 .

I've also attached a debugger to Xorg, as described here ( http://wiki.x.org/wiki/Development/Documentation/ServerDebugging , although it need ctrl-c to get it to backtrace) -- I'm attaching the backtrace from case (b).
Comment 1 Chris Wilson 2013-04-18 07:14:23 UTC
It's a GPU hang. Please attach /sys/kernel/debug/dri/0/i915_error_state.
Comment 2 petrelharp 2013-04-18 15:42:14 UTC
Created attachment 78188 [details]
hung i915 error state

Sorry; should have said that this was also in one of those links.

Note that the GPU only hangs about half the time; the other half the time the error state is "no error state collected".
Comment 3 petrelharp 2013-04-18 15:54:25 UTC
Created attachment 78189 [details]
Xorg.0.log that accompanies i915_error_state

The backtrace from Xorg.0.log entered previously occurs when there is no i915_error_state; in the case that there is a i915_error_state, Xorg.0.log looks like this.
Comment 4 Chris Wilson 2013-05-08 15:48:35 UTC
Upstream stable reverted the chicken bit removal rather than backport the correct driver code.


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.