Intel ARC PCI passthrough issues on Proxmox

Software & Operating Systems VFIO
, ,
1

I’m htiting some issues with a newly-purchased A750 card: Proxmox doesn’t allow it to be passed through (but allows it to be specified). Here are some relevant systemd logs when it attempts to do so (looks like it’s dropping out after some power state transition)

Mar 29 15:42:30 pve kernel: VFIO - User Level meta-driver version: 0.3
Mar 29 15:42:30 pve kernel: Console: switching to colour dummy device 80x25
Mar 29 15:42:30 pve kernel: vfio-pci 0000:03:00.0: vgaarb: deactivate vga console
Mar 29 15:42:30 pve kernel: vfio-pci 0000:03:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=none
Mar 29 15:42:30 pve kernel: vfio_pci: add [8086:56a1[ffffffff:ffffffff]] class 0x000000/00000000
Mar 29 15:42:30 pve kernel: vfio_pci: add [8086:4f90[ffffffff:ffffffff]] class 0x000000/00000000
Mar 29 15:42:32 pve kernel: pcieport 0000:00:01.0: broken device, retraining non-functional downstream link at 2.5GT/s
Mar 29 15:42:33 pve kernel: pcieport 0000:00:01.0: retraining failed
Mar 29 15:42:34 pve kernel: pcieport 0000:00:01.0: broken device, retraining non-functional downstream link at 2.5GT/s
Mar 29 15:42:35 pve kernel: pcieport 0000:00:01.0: retraining failed
Mar 29 15:42:35 pve kernel: pcieport 0000:01:00.0: not ready 1023ms after resume; waiting
Mar 29 15:42:36 pve kernel: pcieport 0000:01:00.0: not ready 2047ms after resume; waiting
Mar 29 15:42:38 pve kernel: pcieport 0000:01:00.0: not ready 4095ms after resume; waiting
Mar 29 15:42:43 pve kernel: pcieport 0000:01:00.0: not ready 8191ms after resume; waiting
Mar 29 15:42:51 pve kernel: pcieport 0000:01:00.0: not ready 16383ms after resume; waiting
Mar 29 15:43:08 pve kernel: pcieport 0000:01:00.0: not ready 32767ms after resume; waiting
Mar 29 15:43:43 pve kernel: pcieport 0000:01:00.0: not ready 65535ms after resume; giving up
Mar 29 15:43:43 pve kernel: pcieport 0000:01:00.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:43 pve kernel: pcieport 0000:02:04.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:43 pve kernel: pcieport 0000:02:01.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:43 pve kernel: vfio-pci 0000:03:00.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:43 pve kernel: vfio-pci 0000:04:00.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:43 pve kernel: vfio-pci 0000:03:00.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:43 pve kernel: vfio-pci 0000:04:00.0: Unable to change power state from D3cold to D0, device inaccessible
Mar 29 15:43:44 pve kernel: vfio-pci 0000:04:00.0: timed out waiting for pending transaction; performing function level reset anyway
Mar 29 15:43:45 pve kernel: vfio-pci 0000:04:00.0: not ready 1023ms after FLR; waiting
Mar 29 15:43:46 pve kernel: vfio-pci 0000:04:00.0: not ready 2047ms after FLR; waiting
Mar 29 15:43:48 pve kernel: vfio-pci 0000:04:00.0: not ready 4095ms after FLR; waiting
Mar 29 15:43:53 pve kernel: vfio-pci 0000:04:00.0: not ready 8191ms after FLR; waiting

I can provide more information if needed. Thanks

2

Are you also passing thorough the Arc Audio device? If you do, please try removing it.

If the guest is Windows, you may also need to use a dummy HDMI adapter for the card to properly initialized.

3

Tried without the audio device and the same issue still exists. The guest is Linux

4

I have a770 that passthough to windows on proxmox more less it works. To get were I got I had to blacklist i915 driver from loaded. This is partly because it was only gpu on box. Downside you get not direct console/