I’ve been installing cuda-drivers on a RedHat 7.4 server with a Telsa K80 passthrough using the basic installation procedure:
sudo yum install cuda-driver
sudo reboot
nvidia-smi is been working properly, the card is correctly detected.
Now I’m moving to the installation of X Server with a headless X screen to run our OpenGL application in.
I took our basic xorg.conf configuration that worked with “CentOS x Quadro” and “CentOS x Tesla” but this time it gives me no good results.
startx fails with
xinit: connection to X server lost
journalctl records that line:
kernel: vgaarb: this pci device is not a vga device
The xserver is starting just fine and terminates because there’s no x client connected. You’re missing either the ~/.xinitrc or a general Xsession file/setup, i.e. if you want the xserver to keep alive, you have to run an application on it.