Bug 94893 - Segfault in pixman_region_contains_rectangle
Summary: Segfault in pixman_region_contains_rectangle
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-04-11 11:40 UTC by j.oberlaender
Modified: 2016-09-16 22:01 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Xorg log showing segfault/backtrace (40.33 KB, text/plain)
2016-04-11 11:40 UTC, j.oberlaender
no flags Details
Relevant software packages (5.49 KB, text/plain)
2016-04-11 11:43 UTC, j.oberlaender
no flags Details
Two plain LibreOffice Writer documents (11.21 KB, application/octet-stream)
2016-04-11 12:40 UTC, j.oberlaender
no flags Details
Another Xorg log, different crash location (40.25 KB, text/plain)
2016-04-11 13:18 UTC, j.oberlaender
no flags Details

Description j.oberlaender 2016-04-11 11:40:44 UTC
Created attachment 122861 [details]
Xorg log showing segfault/backtrace

Hi all,

I'm not 100% sure this is actually a bug in pixman, but it seems to be the most likely culprit.  This segfault crashes my X session fairly regularly (but not 100% reproducible) when I have LibreOffice running with an open document, and I try to open a second document.

Please see the attached logs.
Comment 1 j.oberlaender 2016-04-11 11:42:19 UTC
The OS is Ubuntu 14.04.4 LTS.
Comment 2 j.oberlaender 2016-04-11 11:43:41 UTC
Created attachment 122862 [details]
Relevant software packages
Comment 3 j.oberlaender 2016-04-11 11:47:46 UTC
CPU: Intel(R) Core(TM) i7-4710MQ CPU @ 2.50GHz
uname -a: 4.2.0-19-generic #23~14.04.1-Ubuntu SMP Thu Nov 12 12:33:30 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

I am not sure whether it is really pixman at fault here, or X, the Intel driver using it.  Let me know what you think.
Comment 4 j.oberlaender 2016-04-11 12:40:00 UTC
Created attachment 122864 [details]
Two plain LibreOffice Writer documents

After some more tests, I can confirm that it is enough to open/close/reopen some simple documents in LibreOffice Writer a few times to trigger the segfault.  I open one of them through the LibreOffice Open or Recent Files dialog, and the second one by double-clicking in Nautilus. This does not always trigger the segfault, but fairly often.

The testcase documents are trivial, but I'm attaching them anyway in case it might help.
Comment 5 j.oberlaender 2016-04-11 13:18:00 UTC
Created attachment 122866 [details]
Another Xorg log, different crash location

This time I had a different crash location, probably indicating that it is not pixman at fault, but rather something further up in the backtrace.

In case it is useful, this is on a ThinkPad T540p with an external monitor connected via the analog VGA output on the docking station.  I have not been able to reproduce the issue without the external monitor (yet).

For what it's worth, I'm running the latest BIOS:

$ sudo dmidecode -s bios-version && sudo dmidecode -s bios-release-date
GMET75WW (2.23 )
03/16/2016
Comment 6 j.oberlaender 2016-04-12 07:25:32 UTC
After some more tests, I can only reproduce the crash if I am using two monitors (laptop display + external monitor).  As soon as I switch to a single display, I can no longer reproduce it.
Comment 7 Oded Gabbay 2016-04-17 12:32:30 UTC
Probably related to x server's intel driver. pixman is a "do what you say" library, so it just fills/copies pixels according to the parameters given to it. If those are not correct, then a crash might occur.

So moving this bug to a different component
Comment 8 Chris Wilson 2016-04-17 12:40:43 UTC
This is very likely to be fixed already in xf86-video-intel.git.
Comment 9 j.oberlaender 2016-04-17 16:47:36 UTC
(In reply to Chris Wilson from comment #8)
> This is very likely to be fixed already in xf86-video-intel.git.

Thanks for the information (and thanks to Oded for moving the ticket).  Do you know roughly the commit (or a range of commits) where this was probably fixed?  Since I am observing this on my work laptop, I do not have the time to try out my own driver build from git, but I could check Ubuntu (or the PPAs) for versions which include the commit, and see if they make a difference.


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.