On Tuesday 13 of August 2013 21:35:36 Mark Brown wrote:
On Tue, Aug 13, 2013 at 01:10:02PM -0700, Tomasz Figa wrote:
On Tuesday 13 of August 2013 18:52:51 Mark Brown wrote:
On Sun, Aug 11, 2013 at 07:59:24PM +0200, Tomasz Figa wrote:
+#define pl330_filter pl08x_filter_id
So, this is pretty fun and probably not going to play too nicely for multiplatform kernels...
Yes, this is really funny and only shows the brokenness of current channel filtering mechanism used in DMA engine subsystem. Fortunately this is quite irrelevant to multiplatform kernels, because no users of it will be compiled in, since it's used only for legacy (non-DT) channel filtering.
Well, the thing here is more about the fact that the match function is silently being hijacked - it should be possible to do this in the device registrations. I'd expect you could get a non-DT kernel which was affected, but you'd probably have to try.
Let me see what I can do to improve things for now.
Anyway, I would happily redesign and rewrite the DMA engine channel matching completely to make it much more sane, but I guess I would need an ack from Vinod and Dan to do it.
I'd be really keen to get this code merged sooner rather than later for ASoC development so if there's a way to do it separately to fixing the filtering that'd be good.
I fully agree on this. I still have other things to work on, so the possible redesign of channel filtering would not probably happen before 3.12 merge window closes.
Best regards, Tomasz