Nvidia X && zypper dup (conflict)

I have a nVidia GTX 1080 graphics card, and I install the drivers “the hard way.”

Each time I get system updates (by zypper dup) and reboot, I can only boot into shell/command line interface; startx gives some error like: xinit failed, possibly incorrect permissions, or something about Xauthority missing…

If I reinstall my graphics card drivers from the shell, and choose “use nvidia X” during installation, the problem goes away and I can boot to desktop with GUI as normal.

I think this could have something to do with the repositories I have on my system ( zypper dup might be updating some other X Server which interferes with the nvidia X server ? )… but I’m hesitant to make any major adjustments until I’m sure.

If anyone could give some insight on this sort of problem, or potential ways to avoid it in the future, I’d greatly appreciate it.

If you use the hard way you must reinstall the driver after any change in kernel or the xstack

If you get the driver from the NVIDIA repository it will auto reinstall

But in TW the kernels may get ahead of the NVIDIA repos. So many prefer to just use the hardway

If you want to stick to “the hard way”, there’s a good method to automate the rebuild of the kernel modules: dkms ( dynamic kernel module support ). To use that:


sudo zypper in dkms
sudo systemctl enable dkms.service
sudo systemctl start dkms.service

Next, run the NVIDIA…run installer with the ( apart from your other parameters ) " --dkms" option. This will trigger a rebuild of the kernel modules on a kernel update.

Yet another option ( a friend does this ) is to lock the kernel-default package. Once you feel like having a newer kernel for what ever reason, you unlock the kernel-default package, then run the zypper dup and rebuild the driver after reboot.

so if i have a card that absolutely REQUIRES with no other OS or repository and they are older than all heck, 367.130 that setup will rebuild the new kernels? i wonder if this may be related to why my driver currently has tons of issues finding it things it needs to install. How stable would a setup like this be?

Hi
There isn’t a 367.130 (367.57 is all I see) version? Or is that a cuda version number? Anyway an old driver like that would need patching and creating a custom run file… if it would even run with the latest kernel. I had an older GT880 and gave up and switched to later cards in my desktop.

This is the newest one i was able to find, Linux x64 (AMD64/EM64T) Display Driver | 367.130 | Linux 64-bit | NVIDIA i can’t afford the newer cards at least the GOOD ones at $4000.00 - $10,000.00 USD. i might look for a slightly newer card but, at $800 for a minimum upgrade, times 2 or 3 the cost as ineed 1 one for each server… meh. maybe i should learn to patch linux and just not update it?

i was kind of hopping to not have to run a windows server for everythig but, windows truly shine when it comes to outdated VGA drivers. sorry linux that’s your one akeles heel.

EDIT: found a slightly newer driver: Linux x64 (AMD64/EM64T) Display Driver | 367.133 | Linux 64-bit | NVIDIA

Hi
Ahh ok, a dedicated one, so you can’t use the standard repo drivers, install the hardway.

Can you show the output from;


/sbin/lspci -nnk | egrep -A3 "VGA|Display|3D"

sure!


I also have this issue too



Hi
The path is incorrect, if your running the current kernel it’s just /usr/src/linux which is a softlink to the running kernel source.


ls -la /usr/src/linux
lrwxrwxrwx 1 root root 13 Nov  6 09:37 /usr/src/linux -> linux-5.3.8-1

Does it really need the path? I normally run with just the -aq switch…

I didn’t know that, nor do i know how to properly use that argument. my brain dumped recently i forgot how to use that file path too as it complains with /usr/src/linux but, may have not properly argumented commands

EDIT: Imgur: The magic of the Internet

Hi
Your system is way out of date? It’s only /usr/src/linux which will be a softlink to the running kernel… suggest you look in /usr/src to see what is there… the uname won’t work either as -default is added…

I installed it a few weeks ago and did updates a week or so ago. here is the LS


if stuff is missing i wonder if the various build and linux headers packages i’ve tried installing have moved things or changed things from standard?

EDIT:


EDIT:

Hi
You need to use zypper dup, kernel is now at 5.3.12, release date 20191124.

Is kernel-default-devel installed?


ls /usr/src

fortran.c         fortran.h           fortran_thunking.h  linux-5.3.11-1      linux-5.3.12-1      linux-obj      packages
fortran_common.h  fortran_thunking.c  linux               linux-5.3.11-1-obj  linux-5.3.12-1-obj  nvidia-440.36

…huh… that’s wierd… …here’s what i get, just installed devel

Hi
That looks fine, next kernel update it will increase to two kernels unless you have configured zypper for more.

Now, drop to runlevel 3 at a tty and run the nvidia installer script, no need for a path, just -aq


EDIT: I checked for updates the other day and it said it was up to date or didn’t find any.

Hi
It looks like that version may need some patching… if you head over to the Nvidia forums and browse or ask there if there are patches…