So, while trying to go back to my old single-GPU passthrough VM that I used daily for about a year then stopped in 2022, I realized that something had broken.
I spent days trying to troubleshoot it, and then I realized that the actual problem is the nvidia kernel module. It will NOT unload.
Even when I reboot, SSH into the system, kill SDDM, unload the other nvidia kernel modules (nvidia-modeset, nvidia-drm, and nvidia-uvm), and then try to unload nvidia, I always get the “module is in use” error.
Even trying to use modprobe -rf or rmmod -f.
I do NOT have it in my initramfs/mkinitcpio.conf. I completely removed it.
Even checking lsof /dev/nvidia0 gives zero output once I’ve killed everything.
And yes, I’m even unbinding the vtconsoles and efi framebuffer. I’ve done everything.
lsmod | grep nvidia shows that no other modules are depending on it:
nvidia 62525440 5
I created an nvidia-bug-report.log.gz while in a TTY after unloading the other modules and killing SDDM, and running lsof /dev/nvidia0 and getting no output:
nvidia-bug-report.log.gz (2.5 MB)