I'm trying to get vWS set up so that I can use WDDM mode on a Windows VM. The system is running PVE 8.4.5 and has two L4 GPUs. I've followed the guide here https://pve.proxmox.com/wiki/NVIDIA_vGPU_on_Proxmox_VE as well as referencing this video from NVIDIA https://www.youtube.com/watch?v=Xp_TPHnCOWw. I'm using the vGPU 18.4 package for Linux KVM, the installer is NVIDIA-Linux-x86_64-570.172.07-vgpu-kvm.run. The system is using secure boot, but I followed the additional steps for that with dkms and enrolling the MOK. Running lspci shows Kernel driver in use: nvidia
Everything looked correct to how it showed up in the video, in Proxmox I was able to create the resource mapping and assign the L4-24Q profile to an instance. When I launch the instance, I get:
kvm: -device vfio-pci,host=0000:61:00.4,id=hostpci0,bus=pci.0,addr=0x10: vfio 0000:61:00.4: error getting device info: Input/output error
Running /usr/lib/nvidia/sriov-manage -e ALL shows that I have VFs enabled on both GPUs and running nvidia-smi vgpu --creatable --verbose lists all the profiles, including the L4-24Q that I am trying to use.
If I run dmesg | grep -i vfio, I see the message of no vGPU devices found when the system boots and then the other messages will appear every time I try to start an instance with the GPU mapped.
[ 49.930960] VFIO - User Level meta-driver version: 0.3
[ 53.426246] [nvidia-vgpu-vfio] vGPU device data not available. No vGPU devices found for GPU 0000:41:00.0
[ 56.005107] [nvidia-vgpu-vfio] vGPU device data not available. No vGPU devices found for GPU 0000:61:00.0
[ 2442.861126] [nvidia-vgpu-vfio] 0000:61:00.4: vGPU migration enabled with upstream V2 migration protocol
[ 2442.861261] nvidia 0000:61:00.4: vfio: runtime resume failed -13
Everything looked correct to how it showed up in the video, in Proxmox I was able to create the resource mapping and assign the L4-24Q profile to an instance. When I launch the instance, I get:
kvm: -device vfio-pci,host=0000:61:00.4,id=hostpci0,bus=pci.0,addr=0x10: vfio 0000:61:00.4: error getting device info: Input/output error
Running /usr/lib/nvidia/sriov-manage -e ALL shows that I have VFs enabled on both GPUs and running nvidia-smi vgpu --creatable --verbose lists all the profiles, including the L4-24Q that I am trying to use.
If I run dmesg | grep -i vfio, I see the message of no vGPU devices found when the system boots and then the other messages will appear every time I try to start an instance with the GPU mapped.
[ 49.930960] VFIO - User Level meta-driver version: 0.3
[ 53.426246] [nvidia-vgpu-vfio] vGPU device data not available. No vGPU devices found for GPU 0000:41:00.0
[ 56.005107] [nvidia-vgpu-vfio] vGPU device data not available. No vGPU devices found for GPU 0000:61:00.0
[ 2442.861126] [nvidia-vgpu-vfio] 0000:61:00.4: vGPU migration enabled with upstream V2 migration protocol
[ 2442.861261] nvidia 0000:61:00.4: vfio: runtime resume failed -13
