On Oct 17, 2012 5:36 PM, "Peter Hutterer" <<a href="mailto:peter.hutterer@who-t.net">peter.hutterer@who-t.net</a>> wrote: > > On Thu, Oct 18, 2012 at 09:40:11AM +1000, Dave Airlie wrote: > > On Thu, Oct 18, 2012 at 9:15 AM, Aaron Plattner <<a href="mailto:aplattner@nvidia.com">aplattner@nvidia.com</a>> wrote: > > > On 10/17/2012 04:12 PM, Peter Hutterer wrote: > > >> > > >> On Wed, Oct 17, 2012 at 03:02:53PM -0700, Aaron Plattner wrote: > > >>> > > >>> If the udev device corresponding to a platform drm device has an > > >>> "xorg_drivers" > > >>> property associated with it, treat that as a comma-separated list of > > >>> driver > > >>> suggestions to use. Otherwise, fall back to the existing PCI ID table. > > >>> > > >>> This lets people create a udev rule to associate an X driver with a given > > >>> drm > > >>> device using rules that look like this: > > >>> > > >>> SUBSYSTEM=="drm", DRIVERS=="i915", ENV{xorg_drivers}="intel" > > >>> > > >>> Signed-off-by: Aaron Plattner <<a href="mailto:aplattner@nvidia.com">aplattner@nvidia.com</a>> > > >> > > >> > > >> we had a similar suggestion for input drivers when we added udev support. > > >> but to keep the xorg configuration in X, we then added InputClass and > > >> xorg.conf.d snippets. With this patch, you'd have xorg configuration in > > >> udev. Is this really what you want, or is it worth investigating > > >> VideoClass > > >> support? > > > > > > > > > Hmm, maybe. That would be a lot more work, but might be more flexible in > > > the long run. I'll see if I can find some time to work on that. > > > > Though I'm not sure we really VideoClass type situations, we generally > > have a real driver or fallback to modesetting per set of devices. > > > > we don't really have the synaptics situation where one driver can > > drive loads of non-synaptics. > > Except evdev, we generally don't use "catchall classes". wacom and synaptics > only get applied for such devices, the rest of the InputClass features is > overloading configurations. e.g. for a product of this name, set that > option. I think this is quite similar to what would apply here. > The above example Aaron gave would be something like > > Section "VideoClass" > Identifier "blah" > MatchDRMDriver "i915" > Driver "intel" > EndSection This was something I always intended to work on after InputClass but never got started. At the very least it would be far nicer than the hardcoded PCI table. It would be a good project for somebody. Dan