From Funtoo
Revision as of 03:42, October 27, 2019 by Drobbins (talk | contribs) (Host Validation)
Jump to: navigation, search

Steam is a content delivery system and ecosystem for gaming, developed by Valve Software. It offers hundreds of games, from popular ones such as DOTA 2 to many other less popular and even obscure community games. Steam runs on Microsoft Windows, MacOS, and also Linux.

History on Funtoo

Through Funtoo Linux 1.2, it was possible to play Steam directly under Funtoo. However, with the move to Funtoo Linux 1.3, 32-bit support was dropped from Funtoo, and Steam is currently dependent on a host of 32-bit libraries. Thus, Steam no longer worked under Funtoo directly. Howevever, it is still possible to run Steam under Funtoo via use of containerization technology, but much of the setup was left to users to figure out.

Now, not only is Steam supported, but we now have officially-tested Docker images to allow you to do this easily. Currently we have just an NVIDIA docker image, but this is actually the more challenging one to build and one for Open Source graphics will be appearing shortly.

Funtoo's official Steam container for use with NVIDIA graphics is based on Ubuntu 18, since Steam and Ubuntu have an unholy alliance and this is the officially-supported platform for Steam. That's fine -- we can still run Steam on Ubuntu -- on Funtoo!

The base Ubuntu container image upon which our container is an official Ubuntu container from NVIDIA Corporation and has been configured to work perfectly with container GPU acceleration and libglvnd.

Steam in Docker


This page will currently document the setup of Steam running on NVIDIA hardware, with documentation for non-NVIDIA hardware to follow shortly.

To run Steam on NVIDIA GPUs, you must meet the following prerequisites:

  • You must be using Funtoo Linux 1.4 or later, x86-64bit.
  • You must have an NVIDIA graphics card.
  • Already have NVIDIA drivers (x11-drivers/nvidia-drivers) installed and a desktop environment running that uses them.
  • A regular, non-root user created from which you will be launching Steam.
  • PulseAudio working.

If you are using the Funtoo Linux GNOME stage3 as a starting point, you should be in good shape regarding all these requirements.

Host Setup

To get your host ready to run Steam, emerge the following packages:

root # emerge -av --jobs docker nvidia-container-runtime nvidia-docker

Please ensure that this pulls in libnvidia-container-bin-1.0.2-r1 or later.

You will also want to ensure that you have NVIDIA proprietary graphics running in a graphical environment on your host. Please ensure that you have x11-drivers/nvidia-kernel-modules-435.21-r1 or later installed on your host, and that you don't have any special permissions settings in /etc/modprobe.d (See the Device Node Permissions section below for details.)

Next, you will want to add docker and nvidia-container to your default runlevel, and start them:

root # rc-update add docker default
 * service docker added to runlevel default
root # rc-update add nvidia-container default
 * nvidia-container added to runlevel default.
root # rc

Device Node Permissions

If you are using x11-drivers/nvidia-kernel-modules-435.21-r1 or later, and have run etc-update, then your device nodes should have correct permissions for GPU acceleration to work:

user $ ls /dev/nvidia* -l
crw-rw-rw- 1 root root 195, 254 Oct 23 10:09 /dev/nvidia-modeset
crw-rw-rw- 1 root root 239,   0 Oct 23 10:09 /dev/nvidia-uvm
crw-rw-rw- 1 root root 239,   1 Oct 23 10:09 /dev/nvidia-uvm-tools
crw-rw-rw- 1 root root 195,   0 Oct 23 10:08 /dev/nvidia0
crw-rw-rw- 1 root root 195, 255 Oct 23 10:08 /dev/nvidiactl

Otherwise, you will need to follow these steps. To allow all local users to access your GPU, not just the video user, by setting NVreg_DeviceFileMode=0666 in /etc/modprobe.d/nvidia.conf and rebooting.

Host Validation

With the host and device node permissions set up properly, you should be able to run nvidia-container-cli info as your regular user and see details about your GPU and not receive any error messages:

user $ nvidia-container-cli info
NVRM version:   435.21
CUDA version:   10.1

Device Index:   0
Device Minor:   0
Model:          Quadro P400
Brand:          Quadro
GPU UUID:       GPU-9bb98f5f-7d9d-ee0a-64a1-154c46934f45
Bus Location:   00000000:01:00.0
Architecture:   6.1

If you receive an error message, this indicates that your /dev/nvidia* device node permissions are too restrictive or that /etc/init.d/nvidia-container has not been started properly. Try rebooting or restarting Docker to resolve the issue.

User Setup

When using docker, you will be starting the Steam container as a regular user account so the container can inherit the connection to your X server. You will want to make sure your user account is in the docker group:

root # gpasswd -a drobbins docker
Adding user drobbins to group docker

You will need to log out and log back in for this group change to take effect.

PulseAudio Setup

In order to allow the container to connect to PulseAudio, you will of course need to be using PulseAudio, and then you will also need to enable support for UNIX socket connectivity in PulseAudio. This can be done by adding the following to /etc/pulse/

load-module module-native-protocol-unix auth-anonymous=1

You should restart your user's pulseaudio daemon for this to take immediate effect:

user $ killall pulseaudio

Once this has been done, you should be able to see a native UNIX socket in the PulseAudio run directory. This socket will get mapped into the container:

user $ ls /run/user/$UID/pulse/
native  pid

Docker Container Setup

We now have a fairly sophisticated launcher script called steam-nvidia-launcher which you should use to launch Steam. Download and install as follows:

user $ wget
user $ chmod +x steam-nvidia-launcher

Container Validation

To ensure that PulseAudio is functioning properly from within the container, the following command can be run to play back white noise via PulseAudio:

steam-container # pacat -vvvv /dev/urandom

To ensure that OpenGL is working properly from within the container, and that the container is properly connecting to your X server, you can run glxgears:

steam-container # glxgears

You should see glxgears running in a window on your desktop.

At this point, you can exit the container by pressing Control-D or typing exit. Make a mental note that the container name is steam-nvidia-yourusername, and is specifically configured to be launched by this user and no other.