Re: [PATCH v2] ASoC: pcm3168a: Add option to force clock consumer

[Kuninori Morimoto <kuninori.morimoto.gx@xxxxxxxxxxx>]

Hi Stephen, Mark

> >  From "Sound Card" point of view, setup directly Bx_Fx base instead of
> > CBx_CFx base is not a big deal. But if we do it, we need to have such
> > setting for each DAI, instead of common dai_link.
> > 
> > Now dai_link->dai_fmt is including 4 type of info
> > 
> > 	1. DAI hardware audio formats
> > 	2. DAI Clock gating
> > 	3. DAI hardware signal polarity
> > 	4. DAI hardware clock providers/consumers
> > 
> > I think we want to separate 4 (and maybe 2 too ?) from it to support
> > more flexible system ?
> > 
> > Legacy system
> > 	dai_link->dai_fmt		// include all 1, 2, 3, 4
> > 
> > New system
> > 	dai_link->dai_fmt		// include 1, 2, 3 (or 1, 3)
> > 	dai_link->cpus[idx].fmt		// include 4       (or 2, 4)
> > 
> > 	fmt = dai_link->dai_fmt | dai_link->cpus[i].fmt
> > 
> > "flip" has effect to 4 only, so New system has no issue with flipping
> > on dai_link->dai_fmt.
> 
> Hi Morimoto-san,
> 
> This is a great idea. I'd add that #4 could still be in 
> dai_link->dai_fmt, but the new field in CPU/codec 
> (snd_soc_dai_link_component.fmt) should override the setting (not a 
> simple OR). This would preserve backward compatibility (I think). It 
> greatly improves support for multiple codecs sharing clock lines, where 
> 1 codec is the producer and others are consumers. The bit/frame logic in 
> simple/graph cards should be able to be changed to handle this properly too.
> 
> Without this type of override, the original patch is needed (or the card 
> driver needs to write registers directly, which seems bad).
> 
> I will wait to hear Mark's opinion.

I'm now preparing the patch for it.
I want to more test it. I think I can post it next week if Mark is OK for it.


Thank you for your help !!

Best regards
---
Kuninori Morimoto