NVIDIA GTX 980 Driver problem

Hello

Todays update of the driver broke the display, using safe mode and I tried to return to a previous verison of the driver but it’s not listed.

Currently I have installed:

nvidia-uvm-gfxG04-kmp-desktop ver. 352.79.k3.16.6_2.19.1 (in red)

nvidia-gfxG04-kmp-desktop ver. 361.28.23.1

nvidia-computeG04 ver. 361.28.23.1

nvidia-glG04 ver. 361.28.23.1

x11-video-nvidiaG04 ver. 361.28.23.1

Display goes black when it’s suppose to display the login screen.

I use the NVIDIA repository.

kernel: 3.16.7-36 64 bit
Suse 13.2

Sugestions ?

Thank you

In case this helps finding the problem…

rpm -qa | grep nvidia

nvidia-gfxG04-kmp-desktop-346.59_k3.16.6_2-5.1.x86_64
nvidia-gfxG04-kmp-desktop-352.63_k3.16.6_2-18.1.x86_64
nvidia-gfxG04-kmp-desktop-352.41_k3.16.6_2-16.1.x86_64
nvidia-gfxG04-kmp-desktop-352.55_k3.16.6_2-17.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-352.79_k3.16.6_2-19.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-352.21_k3.16.6_2-14.1.x86_64
nvidia-gfxG04-kmp-desktop-352.21_k3.16.6_2-14.1.x86_64
x11-video-nvidiaG04-361.28-23.1.x86_64
nvidia-gfxG04-kmp-desktop-346.35_k3.16.6_2-4.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-346.72_k3.16.6_2-6.1.x86_64
nvidia-gfxG04-kmp-desktop-361.28_k3.16.6_2-23.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-352.41_k3.16.6_2-16.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-352.55_k3.16.6_2-17.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-346.59_k3.16.6_2-5.1.x86_64
nvidia-gfxG04-kmp-desktop-352.30_k3.16.6_2-15.1.x86_64
nvidia-computeG04-361.28-23.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-352.63_k3.16.6_2-18.1.x86_64
nvidia-gfxG04-kmp-desktop-352.79_k3.16.6_2-19.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-346.35_k3.16.6_2-4.1.x86_64
nvidia-glG04-361.28-23.1.x86_64
nvidia-uvm-gfxG04-kmp-desktop-352.30_k3.16.6_2-15.1.x86_64
nvidia-gfxG04-kmp-desktop-346.72_k3.16.6_2-6.1.x86_64

well… uninstalled the nvidia driver and installed again, but it doesn’t work, it won’t load kde.

The command: rpm -qa | grep nvidia
now gives:

x11-video-nvidiaG04-361.28-23.1.x86_64
nvidia-gfxG04-kmp-desktop-361.28_k3.16.6_2-23.1.x86_64
nvidia-computeG04-361.28-23.1.x86_64
nvidia-glG04-361.28-23.1.x86_64

notice there is no "“nvidia-uvm-gfxG04-kmp-desktop”

which was the one in red. when installing from YAST, it didnt’ ask for it either.

Does it needs it ? do I have to wait until Suse releases the apropiate version since that file does not comes form the the nvidia repository ?

No luck finding a solution, it’s the latest driver, no question about it. I read another thread about it on the ‘installation’ topic. Posted a thread on the NVIDIA forum asking for the previous version to be uploaded to the repositories. There’s a -run file there but I always use the repository so I hope they upload the old files.

The reason it was red was that not the right version for some reason

Also you have multiple kernels thus multiple drivers one set for each kernel installed

you must have the uvm (unified memory module) force the install be sure it it the same version as the rest

the UVM doesn’t exists in the NVIDIA repository anymore and I uninstalled the previous version when trying to install the driver again. There is no uvm for current version of the driver.

Same here for two users: no login prompt after last Nvidia update - Install/Boot/Login - openSUSE Forums

But so far nothing has helped to solve the problem.
Just now I’m trying to restore the oldest snapshot with snapper.

OK I see I’m still using GO3 (630 card) and it still has a uvm package but G04 does not. Don’t know if this is deliberate or a package mistake.

It is deliberate. From the package changelog:

Wed Feb 17 14:27:53 UTC 2016 - sndirsch@suse.com
 
 - update to latest long lived branch version 361.28
 - new kernel module source layout (now 3 modules); facilitated by
   merging UVM kernel module into nvidia-gfxG04 KMP
 - added new GLVND libs required for EGL support, but keep non-GLVND
   installation for now 

And even if uvm was missing, the driver itself should still work, only CUDA will not AFAIK.

I would recommend to explicitly uninstall it though if it is still installed. And then reinstall the standard kmp package:

sudo zypper rm nvidia-gfxG04-uvm-kmp-desktop
sudo zypper in -f nvidia-gfxG04-kmp-desktop

(on Leap, replace “desktop” with “default”)

Maybe that would help, but I’m only guessing…

… I tried a manual install of the previous version of the driver (a .run file) and now desktop won’t even display in safe mode, guess I’ll have to do a re-installation and wait until a driver update is posted, use the default driver (noveau) that is installed so I can have a functional pc.

Try to run “NVIDIA-xxx.run --uninstall”.

You do not have to reinstall just get rid of the newest package before installing the older one

I did, it doesn’t shows it listed. Don’t know what I did wrong, using the XP (game) machine to gather info but it appears th instalation broke further even if I only tried the manual install. Previous to that, when ‘safe mode’ was working, I used Yast to remove the installed NVidia package completely, after trying the manual install of the previous (working) version of the NVidia driver trough a .run file the intaller ends, no error reported but desktop didn’t load not even in safe mode, tried to check what was installed with

rpm -qa | grep nvidia

and nothing is listed, tried to install again the .run file (in command line mode, as root) and it returns an error saying that the versions don’t match or that another drive is blocking, closed te installer and… now neither keyboard or mouse works. -.-

yep… about to start re-installation now.

What is your status with the driver now? Did you do the manual uninstall as suggested by wolfi323? The thing is if you want to make your X server work, you cannot mix the nVidia-provided installer with packages from repos. This only leads to breakage, because zypper will never see (register) the *.run installer (it’s not an .rpm package) and the installer will throw an error when it detects existing openGL libraries from an unknown source.

I think your problem may come from the fact that new nVidia drivers require a higher Xorg, X11 and/or kernel version. If possible, I would simply go for openSUSE Leap 42.1 and leave all of the trouble behind :).

Yes, I tried that with the manual install, had some practice back with 10.3 when I had a legacy nvidia card.

I haven’t read that the problem is not presetn with Leap… as far as the G04 drivers go.

POSSIBLE FIX
I’ve had this same problem with my Quadro K2000D card - I think it’s a G04 general problem.
Stefan Dirsch, the packager, suggested a fix which worked for me. Details are in the other thread about this problem (follow this link) but the executive summary is:

Weak updates linking is not working properly.
**DELETE **any older installed kernel versions.
Make sure you are running the (single) installed kernel.
Re-install nvidia-gfxG04-kmp-desktop-361.28_k3.16.6_2-23.1.x86_64.rpm

This forces a re-build of the relevant modules. I should add that I had also deleted /lib/modules/3.16.6-2-desktop as part of my cleanup. This was re-created by the installer. My system now works perfectly again (apart from only having a single kernel version available).

Try this in last resort (https://forums.opensuse.org/showthread.php/514015-no-login-prompt-after-last-Nvidia-update?p=2757202#post2757202). I’ve also reached a point where I couldn’t start the recovery mode neither. I had to edit the xorg.conf file.
Then I could uninstall the driver properly then I could reboot and log in a normal session.

FYI, I just updated nvidia-bumblebee to version 361.28 on Leap without a glitch. Don’t know what happens with non-Optimus systems with the “regular” Nvidia G04 though…

thinking about it… a bit scare after 2 re-install yesterday… I may try after day’s work is done since at least I have a working system with a low-res screen.

The procedure described on the install thread by laurence_tyler worked for me too :slight_smile:

  • Use Yast to delete any other kernel different than the one you use/active

  • go to Lib/Modules and delete any folder with a kernel version different than the active.

  • re-install NVidia drivers (it will recreate the folder you deleted but only with nvidia related files)

  • reboot

Hopefully it’ll work for other like it did to me.

Thanks to laurence_tyler for the solution.

https://forums.opensuse.org/images/misc/quote_icon.png Originally Posted by laurence_tyler https://forums.opensuse.org/images/buttons/viewpost-right.png](https://forums.opensuse.org/showthread.php?p=2757233#post2757233)**
**POSSIBLE FIX
I contacted Stefan Dirsch who maintains the openSUSE packages for the nvidia driva. He very kindly suggested a fix that worked perfectly for me with the G04 driver:

There seems to be a problem with setting up the weak update module links properly in the install script. Stefan’s suggestion was to **delete any old kernel versions then **re-install nvidia-gfxG04-kmp-desktop-361.28_k3.16.6_2-23.1.x86_64.rpm

Make sure that the one kernel installed is in fact the one that is running, then re-install the driver.
This procedure forces the installer to compile fresh modules and gets the links right as well.

After doing this, my system is back to running with its full 4K graphics glory! lol! The only downside is that now I only have the one kernel available. Normally I keep two versions in case an update has unexpected effects that mean I have to revert.

Before the re-installation step I visited /lib/modules and deleted any stuff from the previous kernel versions. I noticed that after re-installation, the directory /lib/modules/3.16.6-2-desktop re-appeared, and this is where the actual nvidia modules seem to be kept. There are symlinks from the current kernel weak-updates/update directory to these modules********