Recently, I needed to change some network settings. I disabled Docker, made my changes, and re-enabled Docker. Now any Docker containers that utilize my GPU (Nvidia RTX 2000 Ada 16gb) refuse to start. I get an error Execution error: Bad parameter pop up. I've done a few Google searches and even asked (cautiously) chatGPT for assistance. ChatGPT suggested to temporarily try starting Plex with the "--privileged" next to the already populated "--runtime=nvidia" but this also doesn't work (and has been removed). When attempting to change that variable I received the following error at the end of the docker run:

docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: initialization error: open failed: /proc/sys/kernel/overflowuid: permission denied: unknown.

The command failed.

Again, various Google searches and more chatGPT lead me to do the following, all to no avail:
1:
a-remove nvidia driver
b-reboot server
c-reinstall nvidia driver

2: (figured the order could matter here but no)

a-remove nvidia driver
b-reinstall nvidia driver

c-reboot server

3: I have also followed spaceinvaderone's corrupt docker video that lead me to delete and recreate my docker image. Initially this provided errors stating that the 20gb vdisk size was full. There are probably some changes I could make to correct this but I assume they're unrelated to my issue at the moment. I increased the vdisk size limit to 55gb for now and those space errors went away but after re-installing all my containers, all the gpu-needy ones failed to start still

4: Rebooting the server, some of the gpu-needy containers are running (due to having auto-start selected) but not all of them. Plex, Jellyfin, Tdarr for example are all running. Ollama is not and trying to start it results in the same Execution error: Bad parameter pop up. The same error comes up if it try to restart or even stop/start any of the others I mentioned. Similarly after rebooting, the docker run still gives the same error about the nvidia-container-cli and a permission denied on /proc/sys/kernel/overflowuid.

5: Running nvidia-smi via ssh:

6: Viewing the permissions of the file thats mentioned to be denying permissions:

Diagnostics attached.

bananas-diagnostics-20250805-1459.zip

Exact same issue on my side since this morning

I continued looking into this today and came across https://github.com/NVIDIA/nvidia-container-toolkit/issues/102

I saw that Krusader was an issue previously. I use krusader about once a year and the initial reason I downloaded it (unrar abilities) I no longer need thanks to un-get. I removed the krusader container and no longer have any issues passing my GPU through to any other containers.