Prepare for GPU Passthrough

iGPU as Host Device Passthrough

This will pass the whole iGPU to a single VM. This is useful if you want that VM to pass slices of iGPU to docker containers for example.

Tested on a Lenovo M70q Gen3

SSH into the Proxmox server.

lspci -nn | grep VGA

Example output:

root@pve:~# lspci -nn | grep VGA
00:02.0 VGA compatible controller [0300]: Intel Corporation Device [8086:4692] (rev 0c)

Grab the ID, for me it was 8086:4692.
Add the ID into a .conf file in /etc/modprobe.d
Terminal window
```
nano /etc/modprobe.d/intel-gpu.conf
```
Paste:
Terminal window
```
options vfio-pci ids=8086:4692
```
Press CTRL + S to save
Press CTRL + X to exit

Get GPU driver name

lspci -nnk | grep VGA -A5

Example output:

root@pve:~# lspci -nnk | grep VGA -A5
00:02.0 VGA compatible controller [0300]: Intel Corporation Device [8086:4692] (rev 0c)
        DeviceName: Onboard - Video
        Subsystem: Lenovo Device [17aa:330b]
        Kernel driver in use: i915
        Kernel modules: i915

Grab driver, for me it was i915
Blacklist driver
Terminal window
```
nano /etc/modprobe.d/intel-gpu.conf
```
Paste:
Terminal window
```
blacklist i915
```
Press CTRL + S to save
Press CTRL + X to exit

Update initramfs

update-initramfs -u -k all

Example output:

root@pve:~# update-initramfs -u -k all
update-initramfs: Generating /boot/initrd.img-5.15.74-1-pve
Running hook script 'zz-proxmox-boot'..
Re-executing '/etc/kernel/postinst.d/zz-proxmox-boot' in new private mount namespace..
No /etc/kernel/proxmox-boot-uuids found, skipping ESP sync.

Once you are done with the above, Reboot

iGPU with SR-IOV

It needs at least kernel 5.19 and a modified driver I’ll be waiting for it to hit mainline kernel. Tracking this issue