On Sat, Mar 07, 2015 at 02:58:53PM +0100, Maciej S. Szmigiero wrote:
W dniu 07.03.2015 11:52, Mark Brown pisze:
On Fri, Mar 06, 2015 at 07:55:26PM +0100, Maciej S. Szmigiero wrote:
AC'97 shouldn't need DT bindings for the CODEC, it's an enumerable bus.
What this driver did / does is essentially to allow attaching AC'97 bus to ASoC AC'97 controllers which don't do it on its own (from what I can see only board files do it).
That's true, and it's something it's easy to get away with with board files, but that doesn't mean it's something we should be doing for device tree where we're not just finding an expedient way to load things but rather defining an ABI.
This driver originally constrained rates to 8000, 11025, 22050, 44100, 48000.
An alternative would be to remove this constraint at all from this driver and leave such constraining for controller and AC'97 bus code.
That's a much better approach.
But since this driver was originally limited to these rates I think it would be safer to do it only when instantiated via OF.
Again, DT is defining an ABI.