Hi all, apologies if this was already covered in another thread, but I can't find anyone with this exact issue (which usually makes me think I've done something really stupid).

 

I'm trying to passthrough an m.2 Coral TPU to a Home Assistant VM. Everything seems to work fine, but when I try and start a VM (Home Assistant, Debian, whatever) with the Coral passed through, it crashes with the same error:

 

internal error: QEMU unexpectedly closed the monitor (vm='Home Assistant'): qxl_send_events: spice-server bug: guest stopped, ignoring qemu-system-x86_64: ../hw/pci/pci.c:1633: pci_irq_handler: Assertion 0 <= irq_num && irq_num < PCI_NUM_PINS' failed.

 

Virtualization and VT-d are enabled, it's bound to its own IOMMU group with nothing else there (I also tried with ACS override but no change), I've tried blacklisting apex and gasket and confirmed they aren't running. I've tried VMs with Q35, I've messed around with the default hostdev block within the XML based on similar issues I've seen people have, but I keep coming back to that error. I've attached the end of the VM log, a snipped of the Unraid log, IOMMU groupings and one of the XML blocks I tried. I'd really appreciate some advice!

 

Cheers

 

 

Debian VM log

 

Unraid log

 

Hello I'm having a similar issue with a Samsung SSD in a 4x4 PCI-e bifurcation card. I've used this in 2 other OS's on the same hardware with no issue so it is something to do with QEMU or Unraid. I cannot figure out how to view the logs unfortunately. I've gotten the card to work a few times, but it wont work after rebooting the VM. I can usually boot the VM by trying again after the error, but there is no disk visible. I swapped the NVME card to another slot on the bifurcation card, after which the SSD was visible in the VM, but again after rebooting it dissapeared.

I've tried reinstalling Unraid, several different VM configs, and chanting in Latin while booting up but no luck :/

I'm as novice as they come, but have now had some limited experience trying to pass through devices in Unraid. Are you getting the same QEMU error as mine, or similar, when you try to boot the VM? I wouldn't read too much into the VM starting properly the second time around. I get different errors if I try to run the VM a second time. Something goes wonky with the handling of the PCI slot - see the last lines of my Unraid log. A full restart of Unraid will always give me the same QEMU error on first VM start.

Back to your problem: have you tried the things I've tried first up? Particularly these ones?

"Virtualization and VT-d are enabled, it's bound to its own IOMMU group with nothing else there (I also tried with ACS override but no change), I've tried blacklisting apex and gasket and confirmed they aren't running."

From all the bits and pieces I've read, these steps solve most problems. For what it's worth, I gave up on the VM and used a docker container instead. No concerns since.