X Windows Freezes (100% CPU), Cursor Active
I've been trying to track down a bug on my computer for a few weeks now, to no avail. The behavior is this - every once in a while (from a few hours to a week), X windows will 'freeze', meaning the display doesn't change and it's unresponsive to the keyboard, but the cursor can still move around. These crashes don't happen during any particular application or task, though commonly while watching full screen video or using firefox. The machine is otherwise fine - music continues to play, and I can ssh in remotely, to see that X is at 100% CPU. kill -9 on X gets things moving again. I've been looking around on a lot of forums to find more info about this bug, and have found that on NVIDIA's linux forum, as well as ATI's linux forum, there are extensive discussions going on about this very problem:
(note: I had links here, but the forum won't let me post them because I haven't made enough posts yet. They're easy enough to find, from nV News dot net for NVIDIA and Rage3D dot com for ATI.)
So, that of course implies that it's not a problem particular to the graphics card or driver. I also suspected X.org or the kernel, but there are reports in those threads of users with both X.org and XFree86, as well as both kernel 2.4 and 2.6. Most people seem to be trying to address this issue by lowering their AGP / X settings. One thing interesting suggestion on one of the threads is that it may be a mobo chipset problem, that the people who have trouble seem to have VIA KT400 / KT600 chipsets frequently (there isn't as much info here, as it's not commonly reported), but there's been no followup so far.
A common theme among the bug reports is that people who dual boot linux and windows don't have problems in windows. This implies that it's not a hardware problem in general, but I've recently been thinking that it's possible windows disables certain hardware features automatically if they're known to be problematic (e.g., windows may detect a certain chipset and set AGP to 4x instead of 8x, or it may disable fast writes or something else, who knows? it may also automatically limit bus traffic on its own). I don't dual boot, and even if I did, I know nothing about debugging hardware problems in windows, so I would be particularly good at such an investigation.
Anyway here's my configuration, in all it's gory detail. Values are extracted from various commands and files.
Software (these are all the latest versions available):
Distribution: Fedora Core 3
Kernel: 2.6.11-1.27_FC3
X.org: 6.8.2-1.FC3.13
gcc: 3.4.3 20050227
NVIDIA driver: 1.0-7664
Hardware:
CPU: AMD Athlon(tm) XP 2600+
Mobo: ABIT KD7A
Chipset: VIA KT400A and VT8235CE
Graphics: GeForce FX 5600XT, AGP
Graphics BIOS: 04.31.20.69.11
/proc/driver/nvidia/agp/status:
Status: Enabled
Driver: AGPGART
AGP Rate: 8x
Fast Writes: Disabled
SBA: Enabled
Soooo, if anyone has any suggestions or experience, I'd love to hear it.
|