The Embedded Linux GPU Mess & How It Can Be Fixed

Earlier this week Qualcomm released an open-source 2D/3D kernel driver for their Snapdragon SoC that's found within the Nexus One, Dell Streak, and many other mobile phones. However, it was just the kernel driver that leveraged their own driver design and no open-source user-space driver, which leads to a dirty mess. David Airlie, the DRM maintainer within the Linux kernel, will not accept open-source kernel drivers that is only used by a closed-source component and as such there's been a lengthy mailing list discussion over the past few days.

Various users and developers have expressed their views on the matter within this discussion thread (along with the usual bickering between David and Luc) but as it stands right now there is no user-space Linux graphics driver for Qualcomm's Snapdragon graphics core that is open-source. Nor is it likely we will see a complete open-source Qualcomm Linux driver in the immediate future.

While David has already clarified his position on rejecting open-source kernel DRM (Direct Rendering Manager) drivers that are only used by closed-source clients, a few hours ago he wrote another blog post in which he briefly talks about the embedded systems / GPU market and how he feels they should play the Linux game.

In this blog post, David answers the following questions: What does the embedded industry get from Linux? So what are they actually hiding in user-space? So why do they think it's valuable? Is the value of this IP more valuable than what the receive from Linux? Isn't it up to them what they do? So shouldn't we give a little? What will make them change their minds? So are you saying they should drop all their in-house developed solutions? So why should they re-write drivers? What would be nice to happen? What would be most likely negative solution?

The best solution would be for Qualcomm and/or the other embedded ARM players to provide a fully open-source graphics driver stack that takes advantage of DRI2, KMS, GEM/TTM, Gallium3D, etc, but it may be some time before we actually see such a solution. Within the desktop space, VIA has been working on their "open-source strategy" for over two years now yet they haven't even released their whole documentation set publicly and are still failing to produce results and it will still be a while.