Troubleshooting
vfio-pci 0000:08:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
error getting device from group 22: No such device
AMD Navi (RDNA 2/3) ROM BAR quirk. Supply a VBIOS dump via rom_file in vm.yaml. See AMD Navi ROM BAR Quirk.
User not in the vfio group:
sudo usermod -aG vfio $USER
Log out and back in.
The memlock limit is too low:
sudo tee /etc/security/limits.d/vee-vfio.conf <<'EOF'
* - memlock unlimited
EOF
Log out and back in. Verify with ulimit -l.
Multiple VFIO devices with colliding PCIe root port slots. This is a vee bug — update to the latest version. vee assigns unique slots automatically.
The GPU was left in D3cold by an unclean exit. vee attempts a PCI function-level reset automatically before each start. If it fails, cold reboot the host.
The GPU display engine did not initialize:
video=kernel param not set → add it and reboot (see Gaming Setup)- First boot after adding the param → plug a physical monitor in for the first boot
- GPU in D3cold → cold reboot the host
Sunshine started before Xorg was ready. The systemd service override (polling xrandr) prevents this. If it still occurs:
systemctl --user restart sunshine
Symptom: Fatal: Hang detected! Session failed to terminate in 10 seconds + core dump.
Cause: deadlock in session teardown when AV1 or HEVC is used with vaapi on some AMD GPU + Sunshine nightly build combinations.
Fix: set av1_mode = 0 and hevc_mode = 0 in sunshine.conf to force H.264 only.
Firewall blocking Sunshine’s stream ports:
sudo ufw disable
- Check client-side bitrate setting in Moonlight preferences
- Set bitrate cap via Sunshine web UI at
https://<vm-ip>:47990 - Ensure both host and client are on wired ethernet
ARP table may not have an IPv4 entry yet:
ping -c1 <vm-ip> && vee ssh <name>
With guest_agent: true and qemu-guest-agent installed, vee ssh resolves the IP via QGA without needing ARP.