Possible Regression with GPU Passthrough on Qubes OS 4.2 rc4

I got a Rog Flow X13 (GV302XI), and after updating the BIOS I was able to install 4.3 rc3 with kernel latest without an issue. I was also able to pass-through the Nvidia dGPU to a Windows 10 HVM using the stubroot method and changing the max-ram-below-4g variable to 2G instead of 3.5G. The VM would allow me to assign as much ram as possible.

However, when I tried passing through the dGPU on 4.2 rc4, the VM would crash as soon as it would start. I also installed 4.2 rc4 with kernel latest, but I otherwise followed the exact same process as I did with 4.2 rc3. I also tried installing 4.2 rc3 and upgrading to 4.2 rc4 which gave me the same result, so I assume that it is something within Qubes or Xen that changed.

Does anyone have a clue as to what may have changed between 4.2 rc3 and 4.2 rc4 that no longer allows me to pass through the dGPU? Do you have any advice on how I can begin troubleshooting this?

Did you try

Also, try to patch xen.xml only, not the stub.

Yes, I tried both methods, stubroot and xen.xml, independently by re-installing rc4 after trying each of the methods.
They both worked on 4.2 rc3, although the xen.xml method introduced other issues where PVH VMs would not start, so I abandoned that method after a couple of tries.
Neither of the methods worked on 4.2 rc4, however, even when I installed rc3 and upgraded to rc4.

1 Like

This worked for some reason: