Graphics error after kernel update to 3.12.53-40-desktop

Running openSUSE 13.1 (Bottle) (x86_64) KDE Platform Version 4.14.10
Graphics: GF114 [GeForce GTX 560]

Patch update today included kernel update from standard Update-Oss repository. After rebooting the installed nVidia driver [x11-video-nvidiaG03 340.96-41.1] is no longer loaded and screen defaults to 1,024 x 768 pixels.

The updated kernel is reported as:
~ > uname -r
3.12.53-40-desktop
~ > rpm -qa kernel-*
kernel-devel-3.12.53-40.1.noarch
kernel-macros-3.12.53-40.1.noarch
kernel-desktop-devel-3.12.53-40.1.x86_64
kernel-firmware-20140807git-2.24.2.noarch
kernel-source-3.12.53-40.1.noarch
kernel-desktop-3.12.53-40.1.x86_64

What action is recommended to remedy this?

I note that this recent post may be related.

Progress update … and thanks for reading
This is my daily desktop and there is work to be done.

  1. tried updating nvidia drivers to G04 version - problem persists
  2. tried downgrading to previous kernel version - problem persists
  3. tried uninstalling all nvidia drivers, re-booting and reinstalling - problem persists
  4. uninstalled all nvidia drivers, reverted to nouveau - screen resolution 1920 x 1200 (normal)
    so back to work …
    I will stay with this until technical advice on the correct procedure is available
    FWIW the installation logs for nvidia G04 drivers reported more than 40 errors like this:
    [snip] make: Leaving directory `/usr/src/kernel-modules/nvidia-340.96-desktop’
    WARNING: /lib/modules/3.12.53-40-desktop/weak-updates/updates/nvidia-uvm.ko needs unknown symbol nvUvmInterfaceChannelDestroy [snip]
    google search on this sort of error did not clarify the issue for me

Same here:
landed in login prompt after update and reboot (3.12.53-40)
dont know right now to fix it.
workaround: booting 3.11.10-34 desktop

Try removing all nvidia packages completely and reinstall the driver from scratch (not upgrade or downgrade, that’s known to cause problems anyway).

same here 13.1, had to revert my booted kernel:


lrwxrwxrwx 1 root root       26 Mar 10 20:32 vmlinuz ->  vmlinuz-3.11.10-34-desktop                                                 
-rw-r--r-- 1 root root  5186696 Feb  3 14:49 vmlinuz-3.11.10-34-desktop                                                            
-rw-r--r-- 1 root root  5428568 Feb 26 02:24 vmlinuz-3.12.53-40-desktop                                                            
lrwxrwxrwx 1 root root       26 Mar 10 02:23 vmlinuz.bad -> vmlinuz-3.12.53-40-desktop      
lrwxrwxrwx 1 root root       25 Mar 10 20:33 initrd -> initrd-3.11.10-34-desktop
-rw------- 1 root root 31637097 Mar 10 09:43 initrd-3.11.10-34-desktop
-rw------- 1 root root 31682305 Mar 10 09:44 initrd-3.12.53-40-desktop
lrwxrwxrwx 1 root root       25 Mar 10 09:44 initrd.bad -> initrd-3.12.53-40-desktop

are we sure the current nvidia packages are compatible with the new kernel ?


> rpm -qa | grep nvidia
nvidia-gfxG02-kmp-default-304.125_k3.11.6_4-33.1.x86_64
nvidia-gfxG02-kmp-desktop-304.108_k3.11.6_4-31.2.x86_64
nvidia-gfxG02-kmp-default-304.125_k3.11.6_4-38.1.x86_64
nvidia-gfxG02-kmp-desktop-304.125_k3.11.6_4-34.1.x86_64
nvidia-gfxG02-kmp-pae-304.119_k3.11.6_4-33.1.i586
nvidia-gfxG02-kmp-pae-304.128_k3.11.6_4-33.1.i586
nvidia-gfxG02-kmp-desktop-304.123_k3.11.6_4-31.1.x86_64
nvidia-gfxG02-kmp-desktop-304.125_k3.11.6_4-35.1.x86_64
nvidia-gfxG02-kmp-desktop-304.117_k3.11.6_4-28.1.x86_64
nvidia-gfxG02-kmp-desktop-304.125_k3.11.6_4-32.1.x86_64
nvidia-gfxG02-kmp-pae-304.131_k3.11.6_4-38.1.i586
nvidia-gfxG02-kmp-pae-304.121_k3.11.6_4-31.1.i586
nvidia-computeG02-304.131-39.1.x86_64
nvidia-gfxG02-kmp-pae-304.121_k3.11.6_4-39.1.i586
nvidia-gfxG02-kmp-pae-304.117_k3.11.6_4-28.1.i586
nvidia-gfxG02-kmp-pae-304.119_k3.11.6_4-30.1.i586
nvidia-gfxG02-kmp-desktop-304.125_k3.11.6_4-33.1.x86_64
nvidia-gfxG02-kmp-pae-304.125_k3.11.6_4-32.1.i586
nvidia-gfxG02-kmp-default-304.123_k3.11.6_4-31.1.x86_64
nvidia-gfxG02-kmp-default-304.125_k3.11.6_4-35.1.x86_64
nvidia-gfxG02-kmp-pae-304.121_k3.11.6_4-37.1.i586
nvidia-gfxG02-kmp-default-304.125_k3.11.6_4-34.1.x86_64
nvidia-gfxG02-kmp-desktop-304.119_k3.11.6_4-33.1.x86_64
nvidia-gfxG02-kmp-default-304.108_k3.11.6_4-31.2.x86_64
nvidia-gfxG02-kmp-pae-304.108_k3.11.6_4-31.2.i586
nvidia-gfxG02-kmp-default-304.131_k3.11.6_4-36.1.x86_64
nvidia-gfxG02-kmp-pae-304.125_k3.11.6_4-38.1.i586
nvidia-gfxG02-kmp-default-304.131_k3.11.6_4-38.1.x86_64
nvidia-gfxG02-kmp-pae-304.131_k3.11.6_4-36.1.i586
nvidia-gfxG02-kmp-default-304.121_k3.11.6_4-31.1.x86_64
nvidia-gfxG02-kmp-pae-304.125_k3.11.6_4-34.1.i586
nvidia-gfxG02-kmp-desktop-304.121_k3.11.6_4-37.1.x86_64
x11-video-nvidiaG02-304.131-39.1.x86_64
nvidia-gfxG02-kmp-desktop-304.128_k3.11.6_4-33.1.x86_64
nvidia-gfxG02-kmp-desktop-304.119_k3.11.6_4-30.1.x86_64
nvidia-gfxG02-kmp-pae-304.125_k3.11.6_4-33.1.i586
nvidia-gfxG02-kmp-default-304.119_k3.11.6_4-30.1.x86_64
nvidia-gfxG02-kmp-default-304.125_k3.11.6_4-32.1.x86_64
nvidia-gfxG02-kmp-pae-304.123_k3.11.6_4-31.1.i586
nvidia-gfxG02-kmp-pae-304.125_k3.11.6_4-35.1.i586
nvidia-gfxG02-kmp-desktop-304.131_k3.11.6_4-36.1.x86_64
nvidia-gfxG02-kmp-desktop-304.131_k3.11.6_4-38.1.x86_64
nvidia-gfxG02-kmp-desktop-304.121_k3.11.6_4-31.1.x86_64
nvidia-gfxG02-kmp-default-304.117_k3.11.6_4-28.1.x86_64
nvidia-gfxG02-kmp-default-304.128_k3.11.6_4-33.1.x86_64
nvidia-gfxG02-kmp-default-304.119_k3.11.6_4-33.1.x86_64
nvidia-gfxG02-kmp-desktop-304.125_k3.11.6_4-38.1.x86_64
nvidia-gfxG02-kmp-default-304.121_k3.11.6_4-37.1.x86_64

anybody had succes reinstalling “all nvidia” from scratch ?

Tried that (item 3 in list above, using yast2 software management), the re-installed drivers still failed to load. The installation log showed many similar warnings as reported above. I found some pertinent comments related to these warnings on CentOS forum here but I lack skill to relate that case to openSUSE 13.1.

Well, they are compatible with Leap 42.1’s much newer 4.1.x kernel, so yes they should be.

But: you need to build the kernel module for the 3.12 kernel, 3.12 is not compatible to 3.11, a module built for 3.11 will not work.

So again, remove all nvidia packages, then boot to kernel 3.12, and then install the driver again while running on 3.12. The module is being compiled for the running kernel when you install the kmp package.

simply replacing symbolic links will do it? Just to be sure, before rebooting.
Anybody done that?

Changing the symlinks in /boot will do absolutely nothing.

Again, you need to compile the nvidia kernel module for Kernel 3.12 to be able to use it with Kernel 3.12. Re-installing it while using Kernel 3.12 will do that.

In case you wanted to know how to boot a different kernel, you can select it in “Advanced Options” in the boot menu.
And you can uninstall specific kernels in YaST via the “Versions” tab e.g.

Ahem, if your grub config points to the symlinks, like mine, this will make you boot the old kernel, of course.
I just posted this to show the old and new version numbers we are talking about…
Sorry if I confused people.

Confirming that this works, removed all G02 packages plus xf86-video-nv (not sure which ones were really required…), rebooted, reinstalled, rebooted, all fine. (and re-linked my symlinks, ahem :wink:
Many thanks!

Just a question, in many years of using SuSE this has not been required so far (at least for me).
Is this just “bad luck”, a bug, or a limitation of the Evergreen project ? (no offense meant, really, just curious).

Again, Thanks, Tom.

Yes, but this is not the case normally.

I just posted this to show the old and new version numbers we are talking about…
Sorry if I confused people.

Ok, I thought you were suggesting this as solution or asking whether this will help.

You’re welcome.

Just a question, in many years of using SuSE this has not been required so far (at least for me).
Is this just “bad luck”, a bug, or a limitation of the Evergreen project ? (no offense meant, really, just curious).

Well, it’s the latter. The Evergreen team updated to Kernel 3.12 (the kernel used by SLES12, to not having to maintain Kernel 3.11 themselves) and ignored this.

See also the discussion here:
https://lists.rosenauer.org/pipermail/evergreen/2016-February/001656.html

PS: you wouldn’t have had to uninstall xf86-video-nv.
That’s the old nv driver that only supports 2D and also only older cards AFAIK.

Although you probably won’t miss it either (especially if using nvidia anyway), for the above reasons… :wink:

I can also confirm that uninstall/reinstall of nvidia packages after the kernel update resolved the issuse.
Can you confirm whether, or not, there may be other modules, similarly affected, which might be less obvious than a failed graphics driver?

Most kernel modules are included in the kernel package itself, and should not have problems.

You should know yourself whether you installed 3rd party modules. The most common ones are probably VirtualBox and broadcom-wl.
The openSUSE packages of those should have been updated as well, if you installed them manually (or used Oracle’s package in the case of VirtualBox) you have to rebuild or reinstall them manually but that applies for every kernel update.

AFAIK it is just the NVIDA package

That make sense, I know I uninstalled Virtualbox a little while back, because I hardly ever used it anyway.
I’ll watch my logs for failed modules.

Thank you

Thanks for your technical advice for Evergreen users on this issue. The recommended procedure resolved the issue for me once I manually selected the correct drivers to install - yast2 kept auto selecting the -pae driver but -desktop kernel is installed. It is also good to know why this happened.