[i965] intel driver seems to get stuck in an EDID-fetching loop

[Colin Watson]

Binary package hint: xserver-xorg-video-intel

This bug seems to be new on upgrade to jaunty.

After a period of use, X gets incredibly slow - scrolling a terminal window takes seconds to redraw, the mouse pointer is jerky, top shows X taking 40% of CPU, etc. Xorg.0.log shows a very long repeated sequence of:

(II) intel(0): EDID vendor "LPL", prod id 50432
(II) intel(0): DDCModeFromDetailedTiming: 1280x800 Warning: We only handle separate sync.
(II) intel(0): EDID vendor "LPL", prod id 50432
(II) intel(0): DDCModeFromDetailedTiming: 1280x800 Warning: We only handle separate sync.
(II) intel(0): Printing DDC gathered Modelines:
(II) intel(0): Modeline "1280x800"x0.0 75.50 1280 1344 1392 1512 800 803 809 832 -hsync -vsync (49.9 kHz)
(II) intel(0): Modeline "1280x800"x0.0 75.50 1280 1344 1392 1512 800 803 809 832 -hsync -vsync (49.9 kHz)

... and it's clear from 'tail -f' that it's repeating this sequence quite rapidly. The full log also shows these messages before it really gets stuck into the loop:

(EE) intel(0): underrun on pipe B!
...
exaCopyDirty: Pending damage region empty!

The Xorg.0.log that ubuntu-bug should attach is from a more or less working system, before X slows down (but all the same package versions); I'll attach a much larger (but gzipped) log from a broken system in a moment.

ProblemType: Bug
Architecture: i386
DistroRelease: Ubuntu 9.04
NonfreeKernelModules: wl
Package: xserver-xorg-video-intel 2:2.5.1-1ubuntu7
ProcEnviron:
 LC_COLLATE=C
 PATH: custom, user
 LANG=en_GB.UTF-8
 SHELL=/bin/bash
ProcVersion: Linux version 2.6.28-3-generic (buildd@palmer) (gcc version 4.3.3 20081210 (prerelease) (Ubuntu 4.3.2-2ubuntu8) ) #4-Ubuntu SMP Fri Dec 12 22:48:15 UTC 2008

SourcePackage: xserver-xorg-video-intel
Uname: Linux 2.6.28-3-generic i686
xkbcomp:

[Colin Watson (cjwatson)]

While the EDID messages appear a few times in the log after startup, this isn't immediately followed by severe performance degradation (although, subjectively, this version of X seems rather "jerkier" than that in Ubuntu 8.10). It's only when X goes into a real death spiral that I see these messages appearing in rapid succession in the log.

Anecdotally, this only seems to happen after removing the laptop from my desk and going to use it elsewhere, which is accompanied by removing mains power and the Ethernet cable. I have a hard time seeing how either of those could be directly related to X, but, of the two, being on battery power seems the more likely!

I tried xf86-video-intel.git 5f347020a52b583e19d333361ea7bf2591255899 (much newer than that required a newer libdrm than is currently in jaunty), with no obvious difference. As long as I build it myself (I know the server ABI has changed), is it possible for me to roll back to the Intel driver from 8.10 without having to roll back the server and everything else too? This bug is having a serious negative effect on my ability to use my laptop, and I'd like to try to figure out when this was introduced.

[Timo Aaltonen (tjaalton)]

can you try just downgrading the libdrm to the version on intrepid? That should be possible using the current versions. IIRC wgrant had this issue when we were at UDS, and downgrading libdrm helped.

[Timo Aaltonen (tjaalton)]

sorry, it was libxrandr and pitti, found the dupe.