and, in advance, sorry about another NVIDIA driver post… but so far I did not find any hints on my individual issue. I recently switched to QubesOS on a Dell XPS 15 9530 with NVIDIA GeForce RTX 4050. Using the newest QubesOS 4.2.3 with the default kernel version 6.6.48-1.qubes.fc37.x86_64.
Currently, I am trying to use the RPMFusion packages and I am facing problems at step when building the kernel packages, here:
echo ‘kmodtool error: Kernel uname from /lib/modules/6.6.48-1.qubes.fc37.x86_64/source/Makefile doesn’'‘t match 6.6.48-1.qubes.fc37.x86_64 from --for-kernels argument.’
kmodtool error: Kernel uname from /lib/modules/6.6.48-1.qubes.fc37.x86_64/source/Makefile doesn’t match 6.6.48-1.qubes.fc37.x86_64 from --for-kernels argument.
exit 2
For ‘uname -r’ I inserted 6.6.48-1.qubes.fc37.x86_64, which is exactly the name of the default kernel version. Tried many other variants but then the path does not match and he can’t find the sources respectively the makefile…
The makefile contains only insertions to generate the kernel version “6.6.48”. So, I tried also to extend the missing name of the full kernel version name but the system refuses any modifications on that file.
So far, I tried to install it in dom0 by using a fedora37 template for building the packages… but I am open to simpler approaches if available! It is not important to have the driver working generally, it would be sufficient to be able to use that within 2 or 3 AppVm qubes.
The GPU in dom0 is not used for rendering in the qubes, it’s only used for the simple tasks in dom0 itself so the integrated Intel GPU is more than enough for dom0 and using NVIDIA GPU in dom0 don’t have any advantages: Frequently asked questions (FAQ) | Qubes OS
You can use Intel GPU in dom0 and passthrough the NVIDIA GPU to the qubes to use it there, but you can only use the NVIDIA GPU in a single qube at the same time.
If you want to use the NVIDIA GPU in the qubes then you can try to use in-VM kernel instead of kernel provided by dom0: Managing qube kernels | Qubes OS
And install NVIDIA proprietary drivers for this kernel.
It is sufficient to me to use it only in one app qube at the same time
So, how does this procedure look like in detail (since I am still pretty new to QubesOS)? Do I create a regular AppVm qube and pass the NVIDIA GPU under settings-devices? Because I do not see the graphic card adapter on this list of available devices…
I haven’t found time so far. Now I have started reading in detail and just stumbled upon the requirements. Second graphic adapter check , separate mouse , lot of patience … but I actually do not have a second keyboard and a second monitor since I am living in a camper van… is it possible to get this done without second monitor and keyboard? Can I share the built-in keyboard of my notebook, and can I grant that qube access to use regular built-in display with fullscreen?
You can use GPU in the qube without using separate display/keyboard/mouse.
In the app qube based on a template with GPU attached to it:
if the app that you’re running in it supports Vulkan then it’ll use the GPU to render the graphics by default without any additional configuration required
if the app that you’re running in it doesn’t support Vulkan then for it use the GPU to render the graphics you need to use the VurtualGL to run the app, check the second link in my previous post