Updates breaks nvidia kernel modules

hello all and glad o be here since been using Fedora workstation and silverblue for year now on/off and always coming back to fedora and now it has been my go to distro

lets start here he issues on Worksation 38 running on Hybrid optimus system Dell XPS 9510 and each app update and system update brakes Nvidia kernel module and it goes back to nouveau but integrated GPU still cant use chrome or some other apps that untill i force delete and rebuild nvidia to kernel and it loads again. Sometimes just wait 5 minutes and restart works.

example the chrome issue is it opens chrome, but wont render anything if nvidia is using nouveau and that should not be issue since intel igpu still should render everyhting since no need dedicated graphics on chrome

Are you using Silverblue or Workstation?

Workstation on this
silverblue on other laptop

How was nvidia installed? Using rpmfusion or directly from nvidia?

If nvidia drivers are installed directly from nvidia the modules seem to require manual updates for each kernel upgrade. If from rpmfusion using rpm-ostree and layering the packages then it requires that one wait for some time after an upgrade (5 minutes or more is often recommended) for the new compile to complete before rebooting, but the module rebuild should occur automatically.

1 Like

Installed from RPM-fusion and mostly can’t wait 5 minutes since restart and update comes so do it and then it updates then restart after update and then back to DE and then rebuild manually and restart then it works.

If app updates only no kernel updates it still brakes modules and Nvidia need to build again.

Even only chrome update it brakes Nvidia modules and yes I use mostly GUI since it is automatically updates and just notifying when updates need to install/reboot

And this is only workstation issue silver blue nothing it just works

This is your problem. The gui does its own thing and often does not wait for the ancillary functions to complete before doing the reboot. When a kernel upgrade occurs that also triggers running akmods to rebuild all the locally compiled drivers such as nvidia. If running of akmods has not completed before the reboot it usually corrupts the driver output and then requires a manual procedure to recover. In this case simply performing 2 steps usually solves it.

  1. sudo dnf remove kmod-nvidia-$(uname -r)
  2. sudo akmods --force
  3. reboot

When doing upgrades from the cli the user has control of timing and can wait for everything to properly complete before rebooting – thus avoiding the corruption and needing those 2 steps.

Silverblue seems to be better able to manage such upgrades in the way it is designed.

I wish they would get the gui update procedures fixed to take in account this issue since it affects a great number of users. A simple forced delay between the completion of the upgrade and the reboot should suffice. For me it affects virtualbox, nvidia, and v4l2loopback, hence I 100% avoid the gui upgrades.