Splitting DRM/KMS device nodes*
One Line Summary
Splitting DRM and KMS nodes to allow advanced access-control and sharing between subsystems.
For GPUs we currently put DRM and KMS APIs into a single device node even though they’re mostly independent of each other. With render-nodes and modeset-nodes, we already have pending proposals to split DRM and KMS. These allow fine-grained access-control and shared nodes across multiple GPUs.
Furthermore, many devices provide modesetting functionality but no other advanced DRM features. Splitting KMS and DRM allows reusing KMS for these devices without requiring all the DRM code.
This discussion will highlight the ongoing work, show the benefits of multiple device nodes and allow for broad discussion how to proceed.
An outline of this discussion can be found on my blog:
David Herrmann is a student from Germany working on the linux kernel input and graphics subsystems to improve the core user-space interface between graphics servers and the kernel.
David Herrmann studies mathematics and computer science at University of Tübingen and works on several open-source software projects in his spare time. Since 2011 he spent most of his time on projects around the HID and DRM kernel layers to replace the ancient linux VT system and improve the way graphics servers configure GPUs and input devices.
He got involved in open-source development while working on server-daemons for IRC-networks in 2007. Since then he has contributed to many linux open-source software projects including: Wii-Remote driver, Bluetooth and HID kernel subsystems, Wayland+DRM and kmscon.