The name of the parameter for not installing the kernel modules for the NVIDIA drivers changed between versions 470 and 535.
root# /tmp/NVIDIA.run -A | grep no-kernel && /tmp/NVIDIA.run -v | grep version
--no-kernel-modules
--no-kernel-module-source
nvidia-installer: version 535.98
root# ./NVIDIA_470.199.02.run -A | grep no-kernel && ./NVIDIA_470.199.02.run -v | grep version
--no-kernel-module
--no-kernel-module-source
nvidia-installer: version 470.199.0
The script has --no-kernel-modules hardcoded (here) which causes it to fail for drivers v470 with the following log message:
root# cat steam/Downloads/nvidia_gpu_install.log
Verifying archive integrity... OK
Uncompressing NVIDIA Accelerated Graphics Driver for Linux 470.199.02..........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
./nvidia-installer: unrecognized option: "--no-kernel-modules"
ERROR: Invalid commandline, please run `./nvidia-installer --help` for usage information.
Steps to Reproduce
Install the Nvidia Driver plugin for unraid. Selecting v470.199.02 as the preferred driver version.
Install the docker-steam-headless container leaving everything as the default other than the mounts and the extra parameters.
Adjust the extra parameters to add --runtime=nvidia
Run the container and obseve it crash looping with no logs. Review the logs in $(Appdata)/Downloads/nvidia_gpu_install.log and observe the error about the incorrect/unknown parameter.
Distribution: Slackware v15.0.
Kernel: 6.1.38-Unraid Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz GenuineIntel GNU/Linux
GPU Drivers: NVIDIA-SMI 470.199.02 Driver Version: 470.199.02 CUDA Version: 11.4
Docker version 20.10.24, build 297e128
Relevant log output
....
**** Found NVIDIA device 'NVIDIA 660' ****
Downloading driver v470.199.02 to match what is running on the host
# script then crashes with no further docker logs
Describe the Bug
The name of the parameter for not installing the kernel modules for the NVIDIA drivers changed between versions 470 and 535.
The script has
--no-kernel-modules
hardcoded (here) which causes it to fail for drivers v470 with the following log message:Steps to Reproduce
--runtime=nvidia
$(Appdata)/Downloads/nvidia_gpu_install.log
and observe the error about the incorrect/unknown parameter.Expected Behavior
No response
Screenshots
No response
Relevant Settings
No response
Version
Build: [2023-08-06 03:07:36] [master] [d2b4d2cae792a6446de1e0a88fc2d7ab1631354f] [debian]
Platform
Distribution: Slackware v15.0. Kernel: 6.1.38-Unraid Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz GenuineIntel GNU/Linux GPU Drivers: NVIDIA-SMI 470.199.02 Driver Version: 470.199.02 CUDA Version: 11.4 Docker version 20.10.24, build 297e128
Relevant log output
Related Reports
https://forums.unraid.net/topic/118390-support-josh5-steam-headless/?do=findComment&comment=1198242