[Bug]: Installing NVIDIA drivers fails: /tmp is not executable

[t4cmyk]

Describe the Bug

I tried to setup steam-headless on my system but get an error after "Installing NVIDIA driver v525.105.17 to match what is running on the host".

Looking into the log from the nvidia script it says ERROR: Temporary directory /tmp is not executable - use the --tmpdir option to specify a different one.

Steps to Reproduce

1. Preare docker compose setup as described
2. docker compose up

Expected Behavior

No response

Screenshots

No response

Relevant Settings

No response

Version

Build: [2023-05-22 09:57:47] [master] [d4e70487453b7107bd73cff98f2fdfe23d668df0]

Platform

Distribution: Ubuntu - 22.04.2 LTS (Jammy Jellyfish)
Linux Kernel: 5.15.0-72-generic x86_64 x86_64 GNU/Linux
GPU Driver versions: | NVIDIA-SMI 525.105.17   Driver Version: 525.105.17   CUDA Version: 12.0     |
Docker: Docker version 24.0.1, build 6802122
Docker-compose: Docker Compose version v2.18.1

Relevant log output

steam-headless-steam-headless-1  | [ /etc/cont-init.d/60-configure_gpu_driver.sh: executing... ]
steam-headless-steam-headless-1  | **** Found NVIDIA device 'NVIDIA GeForce GTX 970' ****
steam-headless-steam-headless-1  | Installing NVIDIA driver v525.105.17 to match what is running on the host
steam-headless-steam-headless-1 exited with code 1

[t4cmyk]

When I replace the installer with this simple bash script

#/bin/bash

touch /home/default/Downloads/test

It'll create the test file with 0000 permission. That's probably the reason why the installation doesn't work. But I've no idea why. This only happens when running the entrypoint.sh script.

[t4cmyk]

NVM I found the problem. I changed umask to 777 when it should be 000 in the .env file.