I am attaching two nvidia bug report logs, one for 410.78, and one for 415.18.
I also have a third system running the same level of X server and kernel with an HD display connected via HDMI that is working correctly with 410.73. 410.73 also had the same problem on the other two systems.
First of all, this has nothing to do with your display/monitor, ‘screen’ means just the useable surface of your desktop.
The problem is that the Xserver is incorrectly trying to enable AIGLX using mesa which obviously doesn’t work since you’re using the nvidia driver.
I don’t know if this is connected to the issue but on system boot, the nvidia drivers get loaded, then unloaded and loaded again, maybe due to the in the logs mentioned nvidia-docker, don’t know, looks bogus, might be a timing issue which is surfacing with the newer drivers.
Two things to try to work around the issue:
Thanks for the reply. Your suggestions didn’t solve the problem, so I did a bit of digging on the system that is working. It turns out that I was overwriting libglvnd on that system. As a result the ABI level didn’t match libglx.so, and so the library wasn’t loaded. As a quick, and klugey, test, I moved /usr/lib64/modules/extentions/libglx.so so that it wasn’t found at load time. The systems came up after that. However, if you look at the attached Xorg log from 396.54, you’ll see that it loads libglx.so, and that recognizes that it wants the nvidia version, and does not load glx twice.
The driver layout canged with v410, the nvidia glx module is now a sub-module getting loaded by the DDX in order to make server-side glxvnd work in Xserver 1.20.
Thank you !
Removing /usr/lib64/xorg/modules/extensions/libglx.so before already made it to run all drivers.
That explains the segfault from xinit as well.