Kernel 6.2.x and NVIDIA 390.xx (problem with Nvidia GPU passthrough in LXC)

[ekzidua]

I am trying to install Linux x64 (AMD64/EM64T) Display Driver Version: 390.157 (the last supported by my video card) on my freshly installed pve-manager/8.0.3/bbf3993334bfa916 (running kernel: 6.2.16-3-pve)in order to do GPU passthrough to my LXC container.

• Installed all kernel headers pve-headers package - done

apt list --installed | grep pve-kernel

pve-kernel-6.2.16-3-pve/now 6.2.16-3 amd64 [installed,local]

pve-kernel-6.2/now 8.0.2 all [installed,local]

• Blacklist - done

:~# cat /etc/modprobe.d/blacklist.conf

blacklist vfio-pci

blacklist radeon

blacklist nouveau

options nouveau modeset=0

• Installed Dependencies - done

:~# apt install build-essential

Reading package lists... Done

Building dependency tree... Done

Reading state information... Done

build-essential is already the newest version (12.9).

0 upgraded, 0 newly installed, 0 to remove and 2 not upgraded.

• My video card is NVIDIA Quadro K4100M

01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GK104GLM [Quadro K4100M] [10de:11b7] (rev a1)

Subsystem: Dell GK104GLM [Quadro K4100M] [1028:15cd]

Kernel modules: nvidiafb, nouveau

• /var/log/nvidia-installer.log

It seems the NVDIA driver 390.157 doesn’t work with this version of proxmox.

Any suggestions on how to proceed with this?

[ibr.exg]

Maybe this helps:

If you are using NVIDIA’s GRID/vGPU technology, its driver must be compatible with the kernel you are using. Make sure you use at least GRID version 16.0 (driver version 535.54.06 - current as of July 2023) on the host before upgrading, since older versions (e.g. 15.x) are not compatible with kernel versions >= 6.0 and Proxmox VE 8.0 ships with at least 6.2.

[ekzidua]

Hello @ibr.exg , thanks for the advice.

NVIDIA Display Driver Version 535.54.06 is not supported for my Quadro K4100M
390.157 is the last driver supported by my video card.

I hope the new version of the NVIDIA driver 390 will be released soon that’s why I’m writing here. Only NVIDIA can fix their closed-source driver, or give a time-line.

Please someone from NVIDIA team to answer. Thanks

[ekzidua]

Any other suggestions ?

[ekzidua]

Please someone from NVIDIA team to answer.
Thanks

[aplattner]

The legacy release 390 series of drivers is out of support, but the Quadro K4100M was actually supported by the legacy 470 series. Official support for the Quadro K4000 series ended on January 31 of this year but you could give that driver a try anyway.

Please refer to this page for our official support timelines: Support timeframes for Unix legacy GPU releases | NVIDIA

[ekzidua]

Hello @aplattner , thanks for the advice.

Unfortunately the output of 470.199 driver is not good.

[ekzidua]

• /var/log/nvidia-installer.log

The kernel problem remains.

When is the new version of 470 expected to be released?

[ekzidua]

Any other suggestions ?

[generix]

Error “no such device” points to the nvidia gpu not being available.
Please run nvidia-bug-report.sh as root and attach the resulting nvidia-bug-report.log.gz file to your post.

[ekzidua]

Hi @generix, thanks for your suggestion but the output from the “nvidia-bug-report.sh” script is not needed. The error is expected and the output is self-explanatory.

The result is normal because the video card has no driver installed (this is the problem i am trying to solve)

Тhe actual nvidia driver which works with my video card /NVIDIA Quadro K4100M/ is 390.xx which is оut of support , but the Quadro K4100M was actually supported by the legacy 470 series.
For reference: Support timeframes for Unix legacy GPU releases | NVIDIA
Unfortunately the output of 470.199 driver is not good.

[ekzidua]

the output from /var/log/nvidia-installer.log

When is the new version of 470 expected to be released?

[generix]

I have no idea what you’re talking about. What do you mean with " new version of 470". According to the snippet of log you made a screenshot from, the modules compiled fine so no updated driver is required.