Hi folks
I’m trying to enable NVIDIA GPU access for an Immich LXC on Proxmox VE 9 and I’m currently blocked by an NVIDIA DKMS build failure on the host.
It fails in nvidia-drm with DRM API changes (e.g. fb_create / drm_helper_mode_fill_fb_struct signature mismatch), ending with:
So it looks like the current Debian NVIDIA DKMS doesn’t build against the PVE 6.17 kernel DRM changes yet.
Happy to paste full logs or provide more details if needed. Thanks!
I’m trying to enable NVIDIA GPU access for an Immich LXC on Proxmox VE 9 and I’m currently blocked by an NVIDIA DKMS build failure on the host.
Setup
- Host: Proxmox VE 9.x (Debian trixie base)
- Kernel: 6.17.4-1-pve
- GPU: NVIDIA Quadro P2000 (GP106GL)
- Goal: expose /dev/nvidia* devices to an LXC container (not VFIO / not PCI passthrough to a VM)
What happens
- GPU is detected via lspci, but there are no /dev/nvidia* devices.
- nouveau is present by default, so I added non-free and tried the Debian NVIDIA packages:
- apt-get install --no-install-recommends nvidia-kernel-dkms nvidia-driver-bin nvidia-smi nvidia-modprobe
- DKMS fails when building nvidia-current/550.163.01 for 6.17.4-1-pve.
DKMS error snippet
It fails in nvidia-drm with DRM API changes (e.g. fb_create / drm_helper_mode_fill_fb_struct signature mismatch), ending with:
- “initialization from incompatible pointer type…”
- “too few arguments to function ‘drm_helper_mode_fill_fb_struct’”
- DKMS “bad exit status: 2”
So it looks like the current Debian NVIDIA DKMS doesn’t build against the PVE 6.17 kernel DRM changes yet.
Questions
- Is there a known workaround / recommended driver path for PVE 9 + 6.17 right now?
- Is there an ETA on a kernel update or package update that resolves this DKMS build issue?
- Is the recommended approach simply to stick to 6.14 for now if you need NVIDIA on the host?
Happy to paste full logs or provide more details if needed. Thanks!
