Ask Your Question
1

NVIDIA Driver Stability After Kernel Updates?

asked 2019-01-11 13:47:18 -0500

ckneale gravatar image

Hey everyone, (If TLDR please skip to the last paragraph)

Been using Fedora OS's for ~5yrs exclusively. Nearly the perfect OS for me. Little history with me and NVIDIA cards/drivers...

On Fedora 27 I bought a NVIDIA card for scientific computing. I installed it through the usual process, but once a new kernel came along everything broke, but I had backups in triplicate. At that point I made a decision not to update my kernel ever again as the work on my computer was very important to me (sounds ironic right?), and it was a time sink getting everything back in working order (lost about 2 days to install the specialized soft-ware I needed).

Now, I am rocking Fedora29, and wow I loved the 3rd part NVIDIA driver installation tool (excellent job!). I had my first glimpse of a linux OS doing everything I wanted. I was able to update the kernel a week ago or so, with those drivers, and it was absolutely glorious. Then this week, I do my usual DNF update and get black screen on boot. No startup services, modemmanager, can't get to console, you name it. I was able to recover the disk, through a live image after sinking a few hours into debugging the actual problem, but there I was again; a kernel update shattered my work computer.

I'm not sure what I mind more: not updating my kernel and being vulnerable to fashionable exploits, or trying everything I can to recover important files as days are taken off my life from stress. Is there a more practical way other people have managed to keep their system up to date with NVIDIA cards? This type of hardware assurance is something I'd happily pay for, does CentOS/RedHat have good 3rd party support for graphics cards that don't get mangled after updates? Any advice appreciated

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted
0

answered 2019-01-12 04:44:54 -0500

johanh gravatar image

Dkms has been unreliable lately. I don't know if this is a fault of dkms, the nvidia driver or something else. Be prepared to learn how to manually update dkms modules from command line after installing a new kernel or a new nvidia driver version.

Anyway, I don't recommend using Nvidia's own installer. Use the package from rpmfusion or negativo17. Then you also have a choice to uninstall dkms and use akmod/kmod instead. It tends to be a bit more reliable.

edit flag offensive delete link more

Comments

If you learn how dkms works, it only takes a couple of minutes to fix. It is annoying when it happens, however.

johanh gravatar imagejohanh ( 2019-01-12 04:46:43 -0500 )edit

you will get it, just follow the command and it will help you.

duckfern gravatar imageduckfern ( 2019-01-12 06:26:10 -0500 )edit

Thanks for the comments everyone. Yea, I've had to do the DKMS routine before, but, this last time I couldn't even get to a console due to no log-in services.

ckneale gravatar imageckneale ( 2019-01-12 08:10:51 -0500 )edit

usually, when dkms has issues, it is due to cross version of the compiler that was used to compile the kernel. You can run the nvidia installer from nvidia.com with an option to ignore compiler version mismatch, and dkms will update just fine. I have an nvidia driver with beta support for vulkan, and it updates perfectly every time the kernel is upgraded.

SteveEbey73701 gravatar imageSteveEbey73701 ( 2019-01-12 08:30:54 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

Stats

Asked: 2019-01-11 13:34:04 -0500

Seen: 97 times

Last updated: Jan 12