System freezes after updating Nvidia driver

[JamesNZ]

Hi everyone,
For past 2-ish days fedora has been freezing on me About 2-3 days ago I did an update and yum pulled in a Nvidia update (can't remember the version). The next time I booted up there was a red "beta driver" warning beneath the Nvidia splash screen (still there).

Then yesterday I was doing something and all of a sudden everything froze, just like a windows-style crash; no VT and no mouse action would work. I had did a poweroff, rebooted and everything worked fine. Then today fedora froze on me twice, only the last time I could still move the mouse around (no VT though)

I had a look at /var/log/messages and nothing really stood out to me. This is the only portion I could find that was around the time of the last freeze:

Aug 6 19:32:47 ParanoiaInc kernel: [ 9404.548894] NVRM: Xid (0000:01:00): 13, 0003 00000000 00008297 00000f10 44480000 00000040
Aug 6 19:33:00 ParanoiaInc kernel: [ 9417.388090] NVRM: Xid (0000:01:00): 8, Channel 00000001
Aug 6 19:33:02 ParanoiaInc kernel: [ 9419.388051] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context

The driver version is 304.32. Any help is appreciated, let me know if I can give any more info....

Thanks, JamesNZ

EDIT: BTW the rpmfusion updates-testing repo isn't enabled, so I don't know how I could end up with a beta driver?

[glennzo]

Hello JamesNZ. What kernel version are you using? I only ask because I believe that I had issues like yours with a few of the 3.4 kernels, although I attributed them to a hot GPU. Once kernel 3.5.0-2.fc17.x86_64 was available and installed the issues seem to have gone away.

I thought the lockups were caused by the hot GPU and thought I solved the issue by adding a small fan on the GPU heatsink. Then I did a fresh Fedora 17 install on another partition. At some point while using the new install (I think I was doing the first major update) the system again froze. This made me think that maybe the lockups weren't heat related at all. It hasn't frozen with the 3.5 kernel.

[meine]

unfortunately the same on my Fedora 17 64-bit Gnome3 box. 2 days ago there only were a few nvidia updates, last software added was akmod-virtualbox

system starting up terribly slow, GUI is crippled (no window-topbars, no/partial mouse and keyboard input possible), BSOD-like screen with looking monitor saying there is a serious problem that could not be resolved

temperature seems no issue -- didn't check any gauges, but the fan works normally booting on a room-temp cooled down machine

my hope lies in coming updates that wil bring back functionality as it is something that happens once or twice a year -- Fedora Bleedin' Edge

[JamesNZ]

Originally Posted by glennzo

Hello JamesNZ. What kernel version are you using? I only ask because I believe that I had issues like yours with a few of the 3.4 kernels, although I attributed them to a hot GPU. Once kernel 3.5.0-2.fc17.x86_64 was available and installed the issues seem to have gone away.

I thought the lockups were caused by the hot GPU and thought I solved the issue by adding a small fan on the GPU heatsink. Then I did a fresh Fedora 17 install on another partition. At some point while using the new install (I think I was doing the first major update) the system again froze. This made me think that maybe the lockups weren't heat related at all. It hasn't frozen with the 3.5 kernel.

I'm running the same: 3.5.0-2.fc17.x86_64. Oddly enough, I think my problems started after I got the 3.5x kernel. However I think that the drivers were updated during the same period so I'm not sure if they're related. At the time of the last freeze I think the GPU temp was around 58C, though I've noticed that the freezes tend to occur when I'm doing something rather strenuous e.g running Blender or Nuke.

Originally Posted by meine

unfortunately the same on my Fedora 17 64-bit Gnome3 box. 2 days ago there only were a few nvidia updates, last software added was akmod-virtualbox

system starting up terribly slow, GUI is crippled (no window-topbars, no/partial mouse and keyboard input possible), BSOD-like screen with looking monitor saying there is a serious problem that could not be resolved

temperature seems no issue -- didn't check any gauges, but the fan works normally booting on a room-temp cooled down machine

Thankfully I don't have any of those symptoms It's just the freezes...

[Penguinclaw]

I'm not a great expert on nvidia but I was wondering if akmod-nvidia was updated to the latest version (if you use that method of installing the nvidia drivers) after the kernel update as I believe this can otherwise cause issues [Akmod-nvidia will be removed after each kernel update]. Mine is 1:304.32-1 relating to the latest driver and I can confirm that after this was updated my system has been very stable.

[meine]

Originally Posted by meine

my hope lies in coming updates that wil bring back functionality as it is something that happens once or twice a year -- Fedora Bleedin' Edge ;)

as indeed yesterday evening all things normal... luckily I'm no coder and don't have to understand ;)

[mikee]

Are the Xorg nvidia drivers at the same level as the akmod/kmod?

I have F16, and there is an update for the kmod/akmod for 304.32-2, but I am getting a dependancy error.
Looks like the corresponding xorg driver is not there yet.

All my stuffs are at the same level right now:

xorg-x11-drv-nvidia-libs-295.59-1.fc16.i686
kmod-nvidia-3.4.6-1.fc16.i686.PAE-295.59-1.fc16.5.i686
kmod-nvidia-3.4.7-1.fc16.i686.PAE-295.59-1.fc16.i686
xorg-x11-drv-nvidia-295.59-1.fc16.i686
kmod-nvidia-3.4.4-4.fc16.i686.PAE-295.59-1.fc16.3.i686
akmod-nvidia-295.59-1.fc16.i686

[JamesNZ]

@ mikee, here's the output of rpm -qa nvidia:

nvidia-settings-1.0-19.fc17.x86_64
akmod-nvidia-304.32-1.fc17.x86_64
nvidia-xconfig-1.0-17.fc17.x86_64
kmod-nvidia-3.4.6-2.fc17.x86_64-304.30-1.fc17.x86_64
xorg-x11-drv-nvidia-libs-304.32-1.fc17.x86_64
xorg-x11-drv-nvidia-libs-304.32-1.fc17.i686
xorg-x11-drv-nvidia-304.32-1.fc17.x86_64
kmod-nvidia-3.4.5-2.fc17.x86_64-295.59-1.fc17.5.x86_64
kmod-nvidia-3.5.0-2.fc17.x86_64-304.32-1.fc17.x86_64

Looks ok to me, except for kmod-nvidia-3.4.5-2.fc17. Should that be there?

@ Penguinclaw, yes I've got akmod-nvidia-304.32 as well.

Happily, no more freezes have occurred since I posted the thread OTOH, the nvidia splash screen is still saying that it's a beta driver Not sure how seriously I should take it.