Debian-12-xfce with nvidia often breaks & redoing the drivers big hassle

User Support
1

I am finding myself often having my template with nvidia drivers break.

My solution is to purge nvidia-kernel-open-dkms and then reinstall sudo apt install nvidia-driver linux-image-amd64 and nvidia-kernel-open-dkms

This is a big pain because nvidia often blocks every tor node and any IP that doesn’t have a sterling IP reputation score (any VPN). They want to log real IP addresses, compromising privacy.

Any small update to the debian-12-xfce results in needing to go through this process again to get the drivers usable in GPT4ALL. It’s such a waste of time and it takes many efforts to find a tor node that is new enough that’s not blocked.

Is there any way around this problem? This is wasting many hours of time. I am not expecting Pop! OS, but can’t we find a better way?

2

Check the debian-12-xfce update log to see what happens during update. Maybe new kernel is installed but the nvidia driver is failing to build using dkms for new kernel.

3

I don’t know how to check this update log or what I would be looking for.

I am sure that every time I install something new on debian-12-xfce that I end up having to rebuild the nvidia drivers. I end up purging different nvidia things and then reinstalling them and it redoes things.

I don’t know how to explain exactly what is happening or what to even look for

4

this is because the kernel driver was installed manually, so every time an update installs a new kernel (quite often), the nvidia kernel module is missing for this new kernel (even if it is a semantic minor release, it is a different kernel system wise).

A solution could be to lock the kernel package to not be updated, but it is not really good for security.

There is a “akmod” tool whose purpose is to automatic rebuild this kind of drivers on each kernel change, there should be some documentation somewhere to use akmod with nvidia on debian :slight_smile:

5

@solene or @apparatus, so in order to avoid this every time, I have to learn how to use a tool called akmod? that seems like it could be hard

is this hard for someone with a low skill level? or is this something i’ll be able to do?

6

Did you configure your template and qube based on this template to use in-VM kernel?

The nvidia kernel module should be automatically built using dkms on every kernel update in your template.
Post the template update log when it fails to see what’s happening.