Bug 50455 - [snb uxa] Frequent hard lockups when using icedove (thunderbird)
Summary: [snb uxa] Frequent hard lockups when using icedove (thunderbird)
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: Other All
: medium major
Assignee: Daniel Vetter
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-05-29 04:32 UTC by Michael Biebl
Modified: 2017-07-24 23:01 UTC (History)
4 users (show)

See Also:
i915 platform:
i915 features:


Attachments
Xorg log (48.00 KB, text/plain)
2012-05-29 04:32 UTC, Michael Biebl
no flags Details
dmesg (62.26 KB, text/plain)
2012-05-29 05:27 UTC, Michael Biebl
no flags Details
i915_error_state (297.70 KB, application/x-gzip)
2012-05-29 05:29 UTC, Michael Biebl
no flags Details

Description Michael Biebl 2012-05-29 04:32:41 UTC
Created attachment 62204 [details]
Xorg log

Distro: Debian/sid
xserver-xorg-core: 2:1.12.1.902-1
xserver-xorg: 1:7.6+13
xserver-xorg-video-intel: 2:2.19.0-1
linux-image-3.2.0-2-amd64: 3.2.18-1

Since upgrading xserver-xorg-video-intel from 2:2.18.0-2 to 2:2.19.0-1 I get frequent lookups of X. So far this always happened when opening a large email folder in icedove(thunderbird). At this point the desktop is completely frozen and the system no longer reacts on keyboard input. I can still move the mouse, but X no longer reacts on clicks.

This both happened in a composited DE (gnome-shell) and a non-composited DE (gnome-fallback/metacity).
Downgrading to 2:2.18.0-2 reliably fixes the problem for me.

My system is a X220 sandy bridge laptop, i.e a HD 3000 integrated graphics card.

When the lockup happens I get a Xorg backtrace in Xorg.0.log (attached)
Comment 1 Chris Wilson 2012-05-29 04:39:58 UTC
Can you also attach your dmesg, and probably /sys/kernel/debug/dri/0/i915_error_state?
Comment 2 Michael Biebl 2012-05-29 04:57:23 UTC
(In reply to comment #1)
> Can you also attach your dmesg, and probably
> /sys/kernel/debug/dri/0/i915_error_state?

Atm it just prints "no error state collected".
I assume you want me to run that when the lockup happens?
As the machine is basically dead when that happens, i.e. I can't even switch to the console anymore, which makes this a bit more complicated.
Comment 3 Chris Wilson 2012-05-29 05:16:54 UTC
Ok, we shall just have to try the shotgun approach. Can you please install a 3.4 kernel? There were a number of SNB workarounds. The other approach would be to enable SNA.
Comment 4 Michael Biebl 2012-05-29 05:27:55 UTC
Created attachment 62217 [details]
dmesg
Comment 5 Michael Biebl 2012-05-29 05:29:25 UTC
Created attachment 62219 [details]
i915_error_state
Comment 6 Michael Biebl 2012-05-29 05:33:52 UTC
I was able to gather the requested information via SSH after the system had locked up again.
A very interesting thing happened: While gathering the information, the screensaver on the system had kicked in (after 10 min) and I was able to type again. The desktop session was back to normal.
I'm not sure if it was the screensaver which "unfroze" the hung system or if it was caused by something else.
Comment 7 Chris Wilson 2012-05-29 05:35:32 UTC
Can you please 'cat /sys/kernel/debug/dri/0/i915_fbc_status'?
Comment 8 Michael Biebl 2012-05-29 05:36:44 UTC
(In reply to comment #7)
> Can you please 'cat /sys/kernel/debug/dri/0/i915_fbc_status'?

# cat /sys/kernel/debug/dri/0/i915_fbc_status
FBC enabled
Comment 9 Chris Wilson 2012-05-29 05:39:25 UTC
Until debian update their kernel, you will need to append i915.i915_enable_fbc=0 to your command-line.
Comment 10 Michael Biebl 2012-05-29 05:57:36 UTC
(In reply to comment #9)
> Until debian update their kernel, you will need to append
> i915.i915_enable_fbc=0 to your command-line.

With 2:2.18.0-2 I also had FBC enabled, but I didn't have the lockups.
What changed between 2.18 and 2.19 in that regard that this makes a difference now?

I'll try to run 2.19 with FBC disabled to verify if the hangs are gone.

Thanks for your quick help so far!
Comment 11 Chris Wilson 2012-05-29 06:03:54 UTC
The FBC/BLT hang is principally a timing bug, it just requires the right sequence of commands to trigger. Likely that some of the bug fixes in 2.19.0 just altered your common workload sufficiently to trigger it. Though others have reported the FBC hang with earlier ddx.
Comment 12 Michael Biebl 2012-05-31 06:06:09 UTC
(In reply to comment #11)
> The FBC/BLT hang is principally a timing bug, it just requires the right
> sequence of commands to trigger. Likely that some of the bug fixes in 2.19.0
> just altered your common workload sufficiently to trigger it. Though others
> have reported the FBC hang with earlier ddx.

Just wanted to add, that disabling FBC did indeed avoid the GPU hangs 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.