Bugzilla – Bug 22597
[855GM KMS] X fatal freeze on startup under kernel 2.6.29/2.6.30
Last modified: 2009-10-05 11:32:56 UTC
Created attachment 27332 [details]
dmesg up to before X start with kernel-2.6.30-gentoo-r1
I am running Gentoo testing, currently using kernel-2.6.30-gentoo-r1, xorg-server-188.8.131.521-r4, xf86-video-intel-2.7.1. When attempting to use kernel modesetting, X when launched with KDM crashes fatally and the machine (Toshiba Satellite A30-104 with Intel 82852/82855GM/GME graphics chipset) can only be shutdown via magic SysReq.
The symptoms were slightly different under 2.6.29 in that KDM would start and I could login, but it would then hang immediately after submitting the login dialogue. Another difference was that an ACPI-controlled shutdown using the machine's power-button was possible then, which it isn't under 2.6.30.
With either kernel version, it's possible to start X/KDE normally with KMS disabled and using uvesafb instead, though enabling desktop compositing causes a similar freeze, and there is severe corruption on all native KDE applications (separate issue presumably).
dmesg attached; Xorg.0.log to follow.
Created attachment 27333 [details]
Xorg.0.log showing the crash
Please attach intel_gpu_dump output at the time of the freeze (preferably with KMS)
I can't really do anything when the freeze occurs. Can you clarify?
(In reply to comment #3)
> I can't really do anything when the freeze occurs. Can you clarify?
Is it possible for you to remote login?
Yeah a lot of times you can login with ssh and collect some info, but only if your machine hasn't locked up hard.
But I'm hoping this bug is fixed now; there have been some 8xx related fixes and several "random hang" fixes committed recently. Is there any way for you to test recent git bits?
Well, I now have another distro version (or two) of intel-driver and xorg-server to upgrade, so let me get on that (takes some time on this machine) and if there's still the same issue I'll look into git builds.
Sorry, my update process is taking even longer due to library breakage. Give it another day or 2 at this rate :(
In the meantime, I have a question: my machine is a Taiwanese-built Toshiba Satellite A30-104 with Phoenix BIOS, often referred to as not a "true" Toshiba. I have read reports that the video BIOS on these boxen is very buggy (S-video out has never worked properly as long as I've been using Gentoo, and barely worked on Mandrake). If this is correct, can this be a factor here and is there much I can do about that?
Once the machine is up and running we don't depend much on the BIOS. The video driver just needs panel and a few other bits of info, but beyond that I wouldn't think it would have an effect.
No activity, assuming this is fixed. Please reopen if not.