Bug 21514 - [i945] X freezes with corrupted screen on login with UXA enabled
Summary: [i945] X freezes with corrupted screen on login with UXA enabled
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.3 (2007.09)
Hardware: x86 (IA32) Linux (All)
: high critical
Assignee: Jesse Barnes
QA Contact: Xorg Project Team
URL: https://bugs.edge.launchpad.net/ubunt...
Keywords: NEEDINFO
Depends on:
Reported: 2009-05-01 16:15 UTC by Bryce Harrington
Modified: 2009-06-15 18:09 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:

Xorg.0.log (21.61 KB, text/plain)
2009-05-01 16:16 UTC, Bryce Harrington
no flags Details
dri_debug.tgz (276.67 KB, application/x-compressed-tar)
2009-05-01 16:17 UTC, Bryce Harrington
no flags Details

Description Bryce Harrington 2009-05-01 16:15:31 UTC
On X start, the system displays a corrupted login screen when Option "AccelMethod" "UXA" is specified and KDE desktop effects is enabled.  After entering credentials, X freezes.

System: Intel Mac Mini Core Solo, 2M RAM,
Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)
Software: Kubuntu 9.04 updated on 4 April 2009.

I am using a xorg.conf containing Option "AccelMethod" "UXA" and no driver is specified.

After the last (large) update [to -intel 2.6.3 and mesa 7.4 presumably], at login I get a corrupt screen that is divided into 5 vertical bands, each band containing hundreds of light and dark horizontal stripes in a cyclic pattern. The login window, and text, is visible behind this but has many dark spots where the rendering has failed. I can enter the login password but soon after that, when showing the last icon in the middle of the screen, X freezes, apart from mouse movement. No response from keyboard. Have to power down.

I can get round this by switching to tty1 immediately after entering the password (so I don't see the icons on the foreground screen) then waiting until I hear the logon sounds. Then I know it is safe to switch to tty7 and view the graphics sceen. This is a repeatable workaround but obviously not ideal.

I found that turning off desktop effects got rid of the problem completely.

Also I tried regenerating xorg.conf using dpkg-reconfigure. This produced a vesa driver version of xorg.conf. This vesa version works without problem and there is no noticeable difference in performance compared with the UXA version I was using before.

This problem did not exist using desktop effects before the last update.

The corrupt login screen and graphics hang only occurs on initial login when that tty is on screen.
Other logins seem to behave normally.
A forced logout fails to get back to the login screen.

(Sometimes the system completely locks up (mouse too) when opening new pages in Firefox or running a card game.  But these may be separate problems.)

00:00.0 Host bridge [0600]: Intel Corporation Mobile 945GM/PM/GMS, 943/940GML and 945GT Express Memory Controller Hub [8086:27a0] (rev 03)
     Subsystem: Intel Corporation Device [8086:7270]
00:02.0 VGA compatible controller [0300]: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller [8086:27a2] (rev 03)
     Subsystem: Intel Corporation Device [8086:7270]
Comment 1 Bryce Harrington 2009-05-01 16:16:49 UTC
Created attachment 25357 [details]
Comment 2 Bryce Harrington 2009-05-01 16:17:15 UTC
Created attachment 25358 [details]
Comment 3 Bryce Harrington 2009-05-01 16:33:54 UTC
Forgot to mention, this is a Ubuntu bug report I'm forwarding from Peter T Hayward.

Comment 4 Jesse Barnes 2009-05-04 09:57:12 UTC
I just saw this recently too, on my Eee PC after an update.  It looked like the stride was wrong on the blits to the frontbuffer, though I don't know of any changes in that area that might be responsible.  I'll check it out.
Comment 5 Jesse Barnes 2009-05-11 11:21:23 UTC
Adjusting severity: crashes & hangs should be marked critical.
Comment 6 Jesse Barnes 2009-05-27 03:47:39 UTC
Does this still happen with current bits (i.e. from git)?  Some fixes related to bad sharing of the front buffer went in recently, along with some others that might help.
Comment 7 Jesse Barnes 2009-06-15 18:09:22 UTC
This one is fixed for me; I assume due to the libdrm fixes.

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.