Bug 61668

Summary: [sna] intel 2.21.2 crashes with Bus error in libpixmann-1.so
Product: xorg Reporter: Michael Stapelberg <michael+freedesktop>
Component: Driver/intelAssignee: Chris Wilson <chris>
Status: RESOLVED FIXED QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: medium    
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
See Also: https://bugs.freedesktop.org/show_bug.cgi?id=63921
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
full Xorg log containing the crash none

Description Michael Stapelberg 2013-03-01 17:35:51 UTC
Created attachment 75746 [details]
full Xorg log containing the crash

I am using the intel driver, version 2.21.2, with Linux 3.5. X just crashed for me, AFAIR it was when switching to the Chrome window, this triggering some rendering operation.

I am using libpixman-1-0 0.26.0. Is there a minimum version required for recent intel drivers maybe?

My machine has an Intel DH67GD mainboard with an Intel Core i7-2600K CPU.

The full Xorg log is attached.
Comment 1 Chris Wilson 2013-03-01 19:48:20 UTC
v3.5 can generate spurious SIGBUS.
Comment 2 Michael Stapelberg 2013-03-02 10:13:45 UTC
(In reply to comment #1)
> v3.5 can generate spurious SIGBUS.
Thanks for clarifying this.

Which kernel version is the minimum to use intel 2.21.2? I was using 3.8-rc7 before, but encountered a hard-lockup of my machine while watching a video in MPlayer and thus downgraded to 3.5.
Comment 3 Chris Wilson 2013-03-02 10:24:28 UTC
The theory is that the kernels become more stable with time... If you have a regression in v3.8, we need to get that resolved asap.

Is this using mplayer -vo xv or -vo gl?

Sorry, there is a known issue in 2.21.2 (I thought was fixed in that release, but it was only fixed in 2.21.3) with a stale mapping being kept around after a window is reparented. So try 2.21.3 / v3.5 and see if that is stable for you. And please do file a regression report for the hang if it persists with 3.8.
Comment 4 Michael Stapelberg 2013-03-02 11:19:54 UTC
(In reply to comment #3)
> The theory is that the kernels become more stable with time... If you have a
> regression in v3.8, we need to get that resolved asap.
> 
> Is this using mplayer -vo xv or -vo gl?
> 
> Sorry, there is a known issue in 2.21.2 (I thought was fixed in that
> release, but it was only fixed in 2.21.3) with a stale mapping being kept
> around after a window is reparented. So try 2.21.3 / v3.5 and see if that is
> stable for you. And please do file a regression report for the hang if it
> persists with 3.8.
I was using -vo xv, but it occurs to me that -vo gl might be better? Can you comment on that?

I will switch to 3.8 with 2.21.3 tomorrow and run a few tests/report when I encounter the issue again.
Comment 5 Chris Wilson 2013-03-02 20:45:27 UTC
Experience says that -xv vo uses less CPU and has fewer artifacts than -xv gl.

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.