How to use the intel iGPU ?

[xh116]

Current situation

I am running flatcar in homelab, works well. recently planning to use it as media server, seems it doesn't support intel iGPU.

Impact

for Plex/Emby/Jellyfin, which require GPU for transcoding, but intel iGPU driver not installed.

Any idea to install that?

[ader1990]

Hello,

can you please provide the output of this command lspci -vvv |grep -i graphic?

I have a Raptor Lake CPU, and in my case the output is "UHD Graphics 770", which needs CONFIG_DRM_I915=m.
https://cateee.net/lkddb/web-lkddb/DRM_I915.html

Thanks,
Adrian

[xh116]

lspci -vvv |grep -i graphic
03:00.0 VGA compatible controller: Intel Corporation Celeron N3350/Pentium N4200/Atom E3900 Series Integrated Graphics Controller (rev 0b) (prog-if 00 [VGA controller])
	Subsystem: Intel Corporation Apollo Lake [HD Graphics 505]

I have older one, It is recognized by default.
but folder of /dev/dri is empty, which means it has no driver.

[Superpiffer]

I have the same problerm with a different CPU:

lspci -vvv |grep -i graphic
00:02.0 VGA compatible controller: Intel Corporation HD Graphics 500 (rev 0b) (prog-if 00 [VGA controller])
        Subsystem: Intel Corporation HD Graphics 500

No i915 module and no /dev/dri. I prefer not to build the module...

[r0bb10]

Same here with Alder Lake iGPU.

lspci -vvv |grep -i graphic
06:10.0 VGA compatible controller: Intel Corporation Alder Lake-N [UHD Graphics] (prog-if 00 [VGA controller])
        Subsystem: Intel Corporation Alder Lake-N [UHD Graphics]

[ader1990]

Hello,

I will make a PR with the required changes, which will trigger a github action that will build a Flatcar image with the new kernel.
Would it be possible that you download and test the resulting Flatcar image on your environments?

Thanks.

[xh116]

Hello,

I will make a PR with the required changes, which will trigger a github action that will build a Flatcar image with the new kernel. Would it be possible that you download and test the resulting Flatcar image on your environments?

Thanks.

Thank you for the effort. I did a test on ESXi 8.0 update 2 (install by flatcar-install script ).

It is running.

But when I add pci deivce (iGPU passthrough), It is stucked on booting.

It maybe problem with ESXi , if someone else could do a test on a physical machine will be better.

[ader1990]

Hello @xh116,

Currently I do not have a hardware env to test it either, help is welcome on that part.

Saw a similar issue documented here -> https://forum.proxmox.com/threads/vm-with-mediated-i915-gpu-fails-to-start-failed-to-find-vbios-tables.101756/ with this linked document that might bring some more clarity:
https://3os.org/infrastructure/proxmox/gpu-passthrough/igpu-split-passthrough/#linux-virtual-machine-igpu-passthrough-configuration

Most likely, in the virtualized scenario, there are extra steps to be done, like enabling iommu.

Thanks.

[r0bb10]

Hello,

I will make a PR with the required changes, which will trigger a github action that will build a Flatcar image with the new kernel. Would it be possible that you download and test the resulting Flatcar image on your environments?

Thanks.

Hi!

I have to build the dev image and test it (on proxmox).

Alder Lake supports VF up to 7 virtual displays, just the host (in this case proxmox) should have iommu enabled and the kernel module loaded, the vm should have the same driver as the host. Like described here https://github.com/strongtz/i915-sriov-dkms the flatcar vm should only have a grub parameter set i915.enable_guc=3 but it's only needed in a scenario with SRIOV enabled. Maybe build the module from the github repo and include it in flatcar? Shoul be backward compatible with everything older anyway.

As soon as i can i will test the dev image and report back.

[xh116]

Hello @xh116,

Currently I do not have a hardware env to test it either, help is welcome on that part.

Saw a similar issue documented here -> https://forum.proxmox.com/threads/vm-with-mediated-i915-gpu-fails-to-start-failed-to-find-vbios-tables.101756/ with this linked document that might bring some more clarity: https://3os.org/infrastructure/proxmox/gpu-passthrough/igpu-split-passthrough/#linux-virtual-machine-igpu-passthrough-configuration

Most likely, in the virtualized scenario, there are extra steps to be done, like enabling iommu.

Thanks.

Hi @ader1990
I just tested on PVE 8.2.2.
It is working now !

When it will be available in Beta or Alpha Channel to get updated every month ?