OMV kvm Plugin: pcie coral tpu and iGPU passthrough

Hi! I have a coral tpu pci card that I want to use with frigate for image processing. So far, I used the coral with frigate installed as a docker container directly on the OMV system. Now I would like to isolate more, and I'm thinking about using the OMV KVM plugin for that. The only thing I can't figure out is how I could passthrough the pci card and the igpu from my intel n100 cpu.

I found this post which explains how to do exactly this, but with proxmox.

Do you have any hind or advice on how to accomplish the same results in OMV? Thank you!

I have also found this, but I don't know how to insert it in my VM configuration. I tried to add the example snippet (properly adapted for my PCI card) into the VM XML, but I get errors from the plugin:

500 - Internal Server Error

Unable to undefine VM :: HomeAssistant error: failed to get

domain 'HomeAssistant' virsh undefine -- domain

"HomeAssistant"

And after the error log, the VM disappears.

In case anyone needed this: I used this guide to edit the xml from the terminal and it worked very well. Now the coral pci card is detected in the VM. Next step will be to pass the iGPU.

Zitat von ryecoaaron

I will see if I can get some kind of device passthru working in the plugin.

Thank you very much. Do you have any hint about the sharing of iGPU?

Zitat von ryecoaaron

Most libvirt guides (not proxmox) will work with the plugin. Otherwise, you can wait for the plugin to have the feature. I am almost done with it.

https://wiki.gentoo.org/wiki/G…ugh_with_libvirt_qemu_kvm

This is AWESOME! Thank you.

Zitat von ryecoaaron

Most libvirt guides (not proxmox) will work with the plugin. Otherwise, you can wait for the plugin to have the feature. I am almost done with it.

https://wiki.gentoo.org/wiki/G…ugh_with_libvirt_qemu_kvm

Hi, sorry to bother you again...
I'm trying to passthrough the iGPU of my intel N100 (in order to have hardware acceleration for Jellyfin and Frigate), but I'm not suer what I can passthrough and what I should not.

These are my pci devices:

lspci
00:00.0 Host bridge: Intel Corporation Device 461c
00:02.0 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics]
00:14.0 USB controller: Intel Corporation Device 54ed
00:14.2 RAM memory: Intel Corporation Device 54ef
00:14.3 Network controller: Intel Corporation CNVi: Wi-Fi
00:15.0 Serial bus controller: Intel Corporation Device 54e8
00:15.1 Serial bus controller: Intel Corporation Device 54e9
00:16.0 Communication controller: Intel Corporation Device 54e0
00:17.0 SATA controller: Intel Corporation Device 54d3
00:19.0 Serial bus controller: Intel Corporation Device 54c5
00:19.1 Serial bus controller: Intel Corporation Device 54c6
00:1c.0 PCI bridge: Intel Corporation Device 54be
00:1d.0 PCI bridge: Intel Corporation Device 54b2
00:1e.0 Communication controller: Intel Corporation Device 54a8
00:1e.3 Serial bus controller: Intel Corporation Device 54ab
00:1f.0 ISA bridge: Intel Corporation Device 5481
00:1f.3 Audio device: Intel Corporation Device 54c8
00:1f.4 SMBus: Intel Corporation Device 54a3
00:1f.5 Serial bus controller: Intel Corporation Device 54a4
01:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 15)
02:00.0 System peripheral: Global Unichip Corp. Coral Edge TPU

I have OMV 7 installed on bare metal, and I want to give access to my iGPU to one VM only. When I see this message

"Be very careful not to passthrough a device being used by the host operating system."

I'm not sure what I can select: I want OMV to be able to display its cli via HDMI, but that's all I can think of in terms of gpu usage. How should I proceed? Thank you!

Zitat von ryecoaaron

You will need to two graphics adapters and your output only has one. You can't passthrough a device and expect the host to still be able to use it.

I asked because I've seen this post, where someone managed to create a partial passthrough of the igpu, so I was hoping it was possible for me too

Zitat von ryecoaaron

Never heard of partial passthrough or whether it is possible with libvirt.

Ok, I think he used an old intel technology that allowed the share of the gpu instead of an actual passthrough.
Anyways... Do you think it would be stupid to passthrough the only gpu I have?

If not, I think what I should do is:

1. disable gpu drivers on host

2. enable passthrough of the gpu in the kvm plugin

3. enable gpu drivers on guest

Is this correct? Do you happen to have any trusted guide or reference on how to do these steps (or the correct ones)?

Thank you again