Integrated Intel GPU does not work under Proxmox

[ChriZ982]

Hello,
I am using Proxmox VE 8.3.2 (no-subscription) to run a Talos 1.9.2 VM with a custom nocloud image (including extensions i915 20241210-v1.9.2, intel-ucode 20241112, qemu-guest-agent 9.2.0).
My Proxmox Host uses an Intel N100 CPU with SR-IOV enabled for the integrated GPU:

root@prox-3:~# lspci | grep VGA
00:02.0 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.1 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.2 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.3 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.4 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.5 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.6 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:02.7 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]

Generally, I am able to use GPU passthrough. I tested it with a Windows VM. However, when using Talos I can't get it to work properly.
dmesg shows the following:

❯ talosctl dmesg --nodes kube-03 | grep -E "(drm|i915|vga)"
kube-03: kern:    info: [2025-01-24T19:25:38.032388861Z]: pci 0000:00:01.0: vgaarb: setting as boot VGA device
kube-03: kern:    info: [2025-01-24T19:25:38.032429861Z]: pci 0000:00:01.0: vgaarb: bridge control possible
kube-03: kern:    info: [2025-01-24T19:25:38.033202861Z]: pci 0000:00:01.0: vgaarb: VGA device added: decodes=io+mem,owns=io+mem,locks=none
kube-03: kern:    info: [2025-01-24T19:25:38.034428861Z]: pci 0000:01:00.0: vgaarb: bridge control possible
kube-03: kern:    info: [2025-01-24T19:25:38.035228861Z]: pci 0000:01:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none
kube-03: kern:    info: [2025-01-24T19:25:38.035530861Z]: vgaarb: loaded
kube-03: kern:    info: [2025-01-24T19:25:38.332137861Z]: ACPI: bus type drm_connector registered
kube-03: user: warning: [2025-01-24T19:25:38.503460861Z]: [talos] [initramfs] enabling system extension i915 20241210-v1.9.2
kube-03: kern:    info: [2025-01-24T19:25:40.151670861Z]: i915 0000:01:00.0: [drm] Found ALDERLAKE_P/ADL-N (device ID 46d1) display version 13.00 stepping D0
kube-03: kern:     err: [2025-01-24T19:25:42.154105861Z]: i915 0000:01:00.0: [drm] *ERROR* Device is non-operational; MMIO access returns 0xFFFFFFFF!
kube-03: kern:     err: [2025-01-24T19:25:42.155799861Z]: i915 0000:01:00.0: [drm] *ERROR* Device initialization failed (-5)
kube-03: kern:     err: [2025-01-24T19:25:42.156940861Z]: i915 0000:01:00.0: probe with driver i915 failed with error -5

Also, /dev/dri does not seem to exist:

❯ talosctl list /dev/dri --nodes kube-03
NODE   NAME
1 error occurred:
 rpc error: code = Unknown desc = lstat /dev/dri: no such file or directory

If I understood it correctly, /dev/dri needs to exist to be able to pass the GPU to Kubernetes Pods.

I have troubleshooted this for hours now, but did not get anywhere. I would really appreciate if someone could give me some hints on how to debug this further.

Thanks in advance!

[ChriZ982]

https://www.michaelstinkerings.org/gpu-virtualization-with-intel-12th-gen-igpu-uhd-730/
This post states that the DKMS modules need to be installed inside the VM which is not possible with Talos I guess. Closing this discussion then...