On Wed, Mar 12, 2014 at 05:33:54PM +0800, Nicolin Chen wrote:
I'm not sure if there's any defect from this idea, but as long as we keep DAIFMT settings identical for both dai-link ends, it'll be a neater way.
So I'd like to pend this patch and wait for further solution from that topic.
My understanding of the binding had been that exactly one of the devices on the link would have each master flag set and the code was parsing both devices together in order to come up with the setting. Obviously I didn't check the code closely enough here. It seems that the Renesas devices actually do this, though just due to a bug in the master/slave handling. The patch you posted would essentially implement that (without validating the settings on the CPU side, but that could be added).
The above suggestion would also work though it *is* a change in the binding and right now we don't have code for it. We need something here very soon, the merge window is going to open shortly and we ought to avoid having the problematic code in v3.15 if we can - indeed I'd really prefer to get something into v3.14 if possible (though this patch doesn't apply cleanly there and it's *very* late).
Thoughts?