Bug 20258

Summary: [945GM UXA] Rendering, VT switching hang
Product: DRI Reporter: matthew
Component: DRM/IntelAssignee: Eric Anholt <eric>
Status: CLOSED WORKSFORME QA Contact:
Severity: major    
Priority: high CC: hege, jian.j.zhao, john
Version: XOrg git   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Kernel log
none
X.org log - correct configuration, may be a different session
none
part of debugfs after the hang - /sys/kernel/debug/dri
none
result of dmesg after the the crash none

Description matthew 2009-02-22 18:03:31 UTC
Using Linux 2.6.29-rc5, X.org Intel 2.6.1, with UXA, sometimes rendering appears to hang. The cursor is still moves, but there is no update to the icon. There appears to be no other response. The following is reported in association with the above problems:
[drm:i915_get_vblank_counter] *ERROR* trying to get vblank count for disabled pipe 0
It does not appear that the modeset parameter to the Linux module i915 affects this. My hardware identifies itself thus:
00:02.0 VGA compatible controller [0300]: Intel Corporation Mobile 945GME Express Integrated Graphics Controller [8086:27ae] (rev 03)
00:02.1 Display controller [0380]: Intel Corporation Mobile 945GM/GMS/GME, 943/940GML Express Integrated Graphics Controller [8086:27a6] (rev 03)
Comment 1 Gordon Jin 2009-02-23 00:00:33 UTC
Please attach dmesg and Xorg.0.log.
Comment 2 matthew 2009-02-23 08:07:48 UTC
Created attachment 23211 [details]
Kernel log
Comment 3 matthew 2009-02-23 08:11:31 UTC
Created attachment 23212 [details]
X.org log - correct configuration, may be a different session
Comment 4 matthew 2009-02-25 16:20:20 UTC
I've just tried using linux 2.6.29-rc6 and am still experiencing the problem.

Further, I can most easily reproduce the failure by having mode setting enabled on boot, with gdm starting up. I logging in to Gnome happens successfully but on rendering any new window, things lock up as described. I.e., starting an application, a popup window from a panel applet, but not the panel window itself, oddly.
Comment 5 Gordon Jin 2009-02-25 17:18:52 UTC
What's your driver/kernel version combination when you had UXA working fine?
Comment 6 matthew 2009-02-26 15:35:11 UTC
Never. I was not reporting this specifically as a regression; just another data point.
Comment 7 Adam Lantos 2009-03-01 04:46:44 UTC
I confirm this bug on i915 - Intel Corporation Mobile 915GM/GMS/910GML Express Graphics Controller (rev 03)

linux-2.6.29-rc6; libdrm master (ad64ff30d0e85486bba8fb6f0e2fcb1caa06fd5c); mesa-intel-2008-q4; xorg-server-1.6.0; xf86-video-intel master (38a7683561cee7fffab174c2a166bfd51b51ba27)

VT switching -including sleep/hibernate- and x11perf triggers the issue for me, 'normal' 2D desktop usage doesn't.
Comment 8 Giovanni Pellerano 2009-03-05 13:14:28 UTC
I continue to have this using:

intel driver 2.6.2/2.6.3
mesa 7.3
libdrm 2.4.5
xorg 7.4
kernel 2.6.29-rc7-git1

the system locks on vt change on 2.6.2, and locks aftter kdm login in 2.6.3
Comment 9 matthew 2009-03-07 10:16:03 UTC
This still appears to be present on rc7, though it seems to trigger more consistently now (:)); still associated with window creation or application startup.

Which x11perf setting were being used, as I can't seem to trigger it with a default invocation?
Comment 10 Adam Lantos 2009-03-07 10:20:12 UTC
On x11perf 1.2 aa10text and shmput500 tests triggered the bug, although I didn't try other tests.
Comment 11 john 2009-05-01 05:11:56 UTC
I'm finding it hard not to reproduce this bug on my laptop atm! Wobbling wobbly windows in compiz a lot usually causes it. I've recorded the whole of /sys/kernel/debug/dri (I hope you dont mind that it's tarred), and dmesg. I have the result of intel_gpu_dump but it's big and I'm not sure how useful it would be, so havent posted it.

I have exactly the same graphics card, i945, I'm using:
ubuntu 9.04
kernel version 2.6.30-020630rc2-generic
"xserver-xorg-video-intel" 2:2.7.99.1+git20090427.e55d9431-0ubuntu0sarvatt

First bug report post!
Comment 12 john 2009-05-01 05:29:51 UTC
Created attachment 25331 [details]
part of debugfs after the hang - /sys/kernel/debug/dri
Comment 13 john 2009-05-01 05:31:19 UTC
Created attachment 25332 [details]
result of dmesg after the the crash
Comment 14 matthew 2009-05-01 15:44:52 UTC
I can no longer reproduce this issue using Linux 2.6.29 and the X.org Intel driver 2.7.99.1
Also, I have now changed to different hardware.
Comment 15 matthew 2009-05-01 15:49:26 UTC
I can no longer reproduce this issue using Linux 2.6.29 and the X.org Intel driver 2.7.99.1
Also, I have now changed to different hardware.
Comment 16 matthew 2009-05-01 15:50:09 UTC
I think that's the way to receive no more email. Not sure.

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.