Lockups with 302.07 x86_64 8800GT

[marcnarc]

Hi all,

I have an 8800GT, x86_64. I've been getting hard freezes ever since I upgraded from Kubuntu 11.10 to 12.04. I don't recall what driver version 11.10 was using, but in 12.04 I've tried 295.33, 295.40 and 302.07 and they all lockup randomly.

Usually the freeze happens while I'm away from my PC, which is when a random OpenGL xscreensaver app is running. But lockups occasionally happen when I'm working on the machine. Most of the time the freeze is complete -- the kernel doesn't even respond to any SysRq magic keys.

For now I've returned to 295.40, just to keep the Ubuntu upgrade path clean (also, when I installed 302.07 I was getting "API mismatch" log messages). I've also disabled all the OpenGL and DPMS stuff I can find, as I need this machine to be stable so I can get some work done! If I get another not-too-deep freeze I'll post a bug report.

[brebs]

Originally Posted by marcnarc

"API mismatch" log messages

Well, what did they say? You're supposed to read them properly. Sounds like you didn't quit X & "modprobe -r nvidia", to actually use the new driver.

[marcnarc]

Hi brebs, thanks for replying.

I quit all X and other graphical applications before installing a new driver, then rebooted afterwards to restart everything cleanly.

Here're the log messages:

May  3 10:38:12 rincewind kernel: [ 2297.542330] NVRM: API mismatch: the client has the version 295.40, but
May  3 10:38:12 rincewind kernel: [ 2297.542333] NVRM: this kernel module has the version 302.07.  Please
May  3 10:38:12 rincewind kernel: [ 2297.542335] NVRM: make sure that this kernel module and all NVIDIA driver
May  3 10:38:12 rincewind kernel: [ 2297.542337] NVRM: components have the same version.
May  3 10:38:12 rincewind kdm[1222]: X server died during startup
May  3 10:38:12 rincewind kdm[1222]: Failed to start X server. Starting failsafe X server.
May  3 10:38:12 rincewind kdm[2572]: Cannot execute "/etc/gdm/failsafeXServer": No such file or directory

Any insights?

[Plagman]

If I had to guess, you still have the Ubuntu modprobe.conf file that aliases 'nvidia' to 'nvidia-current' for their own packaging, which means you're loading the old Ubuntu module. Try finding such a file in /etc/modprobe.d/ and removing it, then trying again.

[marcnarc]

Thanks Plagman!

I've installed the 302.07 and removed an nvidia-graphics-drivers.conf symlink from /etc/modprobe.d. X started up just fine with no complaints about API mistmatches.

I'll leave my OpenGL screensavers running over the weekend and see what happens.

[Milos_SD]

I had 2 lockups with this driver on 7600GT and Ubuntu 12.04 in 2 days. First time it was when scrolling the map in "Lord of Ultima" web browser game, and second was when I tried to fast forward a youtube video. Here is the log of that:

http://pastebin.com/VUvLcj8M

[marcnarc]

My X session froze over the weekend. I could still SSH into it though. Here's the bug report:

nvidia-bug-report.log.gz