For those having nvidia driver issues:
Boot into multi-user.target with the kernel boot option:
systemd.unit=multi-user.target
https://www.thegeekdiary.com/centos-rhel-7-how-to-boot-into-emergency-or-multi-user-mode-from-grub2/
Login as root. If you did not set a root password, login as a user that has sudo access.
Run:sudo dnf remove *nvidia*
sudo dnf update nobara-login --refresh
You should get nobara-login version 1.0-37.fc36
Reboot:sudo reboot
After reboot you should get your normal graphical login. Login and the Nvidia installer will pop up.
Let the Nvidia installer run (this is an updated version provided by the nobara-login package)
Tested and working on:
GTX 1650 Super
RTX 3070
What changed to cause the breakage?
We switched from the rpmfusion nvidia driver repositories to the negativo17 nvidia driver repositories because the Negativo17 module seemed to be more reliable in terms of loading in a timely manner after installation. Sometimes the rpmfusion module would take forever to load or not load at all.
The problem is that due to a recent change in the upstream kernel, ‘simpledrm’ was enabled, which is not compatible with the Nvidia drivers. rpmfusion has a patch that works around this issue, but it is not the best approach, so Negativo17 did not use the same patch.
Instead nvidia-drm.modeset=1 is needed as a kernel launch option, which bypasses simpledrm. Negativo17 has not updated this yet in their repository, so it was not added to the initial change when we moved to their drivers. For some users not having this option presents a black screen, for others it allows them to login, but gives a second “ghost” display/monitor due to simpledrm.
I’ve now corrected the installer so that it applies nvidia-drm.modeset=1 to grub and runs a grub update.