Red Hat Bugzilla – Bug 67431
Kernel's after 2.4.7-10 cause random X glitches making X unusable
Last modified: 2007-04-18 12:43:37 EDT
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR
Description of problem:
I have an ABit SE6 with Intel 815E integrated graphics & audio. Initially, I
thought this was a RH 7.3 issue since it happened immediately after installing
and didn't happen on RH 7.2 or Mandrake 8.2. However, after more testing, it
also happens under RH 7.2 if I update the kernel past the base 2.4.7-10.
What happens: Upon logging into KDE, all seems well at first. After a short
period of time, the screen will simple go black andc I'll be back at the login
prompt for X. Under GNOME, application and bus errors will popup for anything
running and I'll be lucky if I can get it to log out.
I currently have both the 2.4.7-10 and 2.4.9-34 kernel's installed and can run
flawlessly under the 2.4.7-10. If I boot into the 2.4.9-34, it will
immediately start going crazy.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Load into Kernel 2.4.9-34
2. Run Theme Selector under GNOME and select a theme
3. Use the arrow keys to flip through all the themes and within 1 minute,
everything goes haywire.
Actual Results: GNOME: Window title bars disappeared, any opened programs I
could get to appear would appear in upper left w/o titlebar, panel sometimes
disappears, taskbar flashes programs on and off, bus and application errors
Expected Results: Nothing - it should work without issue as it does under
What XFree86 version-release are you using? Please attach your X server
log and config file. If you're not already using all released errata
updates, please update your system with up2date and provide new information
in the bug report also.
Ok... no problem. I'll post the logs shortly... but as for the errata
updates, all are installed.
You never did attach those logs...
At any rate, the Intel i8x0 kernel DRM code was changed by Intel in an
ABI incompatible way, which broke the driver. It's released and in
production, so there is no way to properly "fix" it. Many great thanks
to Intel for this wonderfulness.
Anyway... Our latest kernel has an ugly hack workaround for this issue
by plopping some crap in /etc/modules.conf which instructs the kernel
which ABI to use. The kernel keeps both ABI's present internally, and
if modules.conf is set correctly, it should work properly. This same
issue occurs for Red Hat Linux 7.3 and 8.0 also.
DRI developers have since discovered another hackish way they can
autodetect this at runtime somehow and XFree86 4.3.0 should come with
this code by default. We might backport the fix as well, but since
it is quite nasty, we might just leave it with our existing hack that
seems to work.
If you still experience these problems even after applying the latest
erratum to 7. or 8.0, please reopen the bug report if you like.
I apologize for not following up on this... I ended up dropping Linux because
of the issue and haven't touched it since... I may in the future pick it up
again and explore some more but right now, I do not have the time.
Ok, if you start using Linux again (c'mon, you know you want
to <grin>) and the problem manages to pop up for some reason, please
reopen this report if you like, or file a new one against the newest