nvidia-driver

Salt formulas for deployment of NVIDIA hardware-accelerated qubes with cuda support

Contents

• nvidia-driver/:
  • default.jinja - configuration variables
  • d12.sls - deploy debian-12 qube
  • d13m - states for deployment of debian-13 qubes
    • template.sls - prepare hvm-capable minimal template
    • standalone.sls - create a standalone based on an hvm-capable template. Will work with debian-13-xfce as well as with debian-13-minimal-hvm created by template.sls
  • f41.sls - deploy fedora-41 qube
  • f41-disable-nouveau.sls - disables nouveau (use if nvidia driver isn't prioritized upon installation and reboot)
  • any-minimize-swap.sls - set swappiness to 0
  • any-prime.sls - install prime script into .local/bin/ of the user. Execute nvrun myprogram to accelerate myprogram with prime.

How to use

1. Upload nvidia-driver to your salt environment
2. Review config in default.jinja
3. Use states:

• Directly: sudo qubesctl --show-output --targets <target_name> state.sls nvidia-driver.f41 saltenv=user
• In another formula {% include 'nvidia-driver/f41.sls' %}

These states also support providing your own variables (as opposed to the ones in default.jinja). If jinja sees appropriately named variable in the context it won't import the same variable from default.jinja.

Current tasks

• optimus prime
• debian 13
• fedora 42
• generalize the installation of drivers (pkg.installed and pkgrepo.managed)

far future:

• rpm repository

Mirrors

• https://github.com/RandyTheOtter/nvidia-driver
• https://codeberg.org/otter2/nvidia-driver