On Fri, Jun 17, 2016 at 01:35:23PM +0100, Mark Brown wrote:
> On Fri, Jun 17, 2016 at 05:44:00PM +0530, Vinod Koul wrote:
> > On Fri, Jun 17, 2016 at 12:14:07PM +0100, Mark Brown wrote:
> 
> > > Just use snd_soc_resume() directly.
> 
> > Nope, that leads to warnings.
> 
> > .complete is the only callback that expects void return whereas
> > snd_soc_resume like other returns int.
> 
> This is possibly an indication that you're abusing things then - this is
> a very unusual interface to use as well...  the patch just looks very
> fishy as is, and as we were talking about in the other thread Lars
> started it seems like this is bodging around something rather than
> solving the right problem.

In the hindsight, I do agree that instead of this, we should move platform
to do suspend_late and resume_early. That way the dependency between card
and platform is taken care in a much bteer fashion.

The problem is two components racing and need of the resume being in serial
fashion.

Thanks
-- 
~Vinod