/dev/dri/renderD128 group mismatch after update – best practice?

[IQIFireball]

Hello everyone,

I have a small but annoying issue with GPU passthrough (/dev/dri/renderD128) on my Proxmox host (Community Edition).

The problem appeared after a recent update. The device node somehow got assigned to the messagebus group:

bash
ls -all /dev/dri/renderD128
crw-rw---- 1 root messagebus 226, 128 Jun 16 18:10 /dev/dri/renderD128
On the host, render has GID 993, while messagebus has GID 992. My Debian Trixie containers use GID 992 for render by default.

So I'm stuck with a mismatch:

• Host expects GID 993 for render
• Containers expect GID 992
• The device is stuck on GID 992 (messagebus)

Options I see:

1. Change all Trixie containers to GID 993 (groupmod -g 993 render) and fix the host udev rule.
2. Keep containers on GID 992 and assign the device to GID 992 on the host (but that would mean it stays under messagebus, which feels wrong).

I'm leaning toward option 1, but I wanted to ask: What is the cleanest/best practice way to handle this?

Has anyone else seen this after a recent Community Edition update? Any advice would be much appreciated!

Thanks in advance!