Re: [PATCH v4 15/23] ASoC: soc-core: Identify 'no_pcm' DAI links for DPCM

[Sameer Pujar]

On 7/2/2020 6:22 AM, Kuninori Morimoto wrote:
> External email: Use caution opening links or attachments
>
>
> Hi Sameer
>
> > > At least my CPU driver doesn't use component:pcm_construct
> > > but is using DAI:pcm_new for some reasons.
> > > I'm not sure checking DAI:pcm here is enough, or not...
> > OK. If adding DAI:pcm_new above here is not sufficient, then a flag
> > can be used to describe FE component? or is there a better
> > alternative?

> soc_component_is_pcm() is called from simple_dai_link_of_dpcm :: "FE" side.

Yes I had to use this on "FE" side only because I wanted to find a real 
"FE" in FE<->BE and BE<->BE links. Please have a look at patch [23/23] 
for the sound DT binding I am using. Basically I want to connect 
multiple components in a chained fashion (FE <-> BE1 <-> BE2 ... <BEx>). 
Some of these BEs can be SoC components like resampler/mixer/mux/de-mux 
etc., the HW I am using has a cross bar which allows me to 
select/connect BEs in any order and I am trying to have the same 
flexibility here. Hence I only want to create PCM devices for real "FE" 
and trying to use simple-card as much as possible. More details about 
the HW and problems were discussed in [0].

[0] https://lkml.org/lkml/2020/4/30/519

> I wonder component->driver->non_legacy_dai_naming can't work for you ?

I see currently in simple-card driver that, BE<->BE link would be 
treated as CODEC<->CODEC link if 'non_legacy_dai_naming' flag is set at 
both ends of BE. Do we need to set this flag for all BE?
However I am not sure how this will work out for a BE<->BE DPCM DAI link 
considering the fact that I want to use chain of components and I guess 
routing map would get complicated. Also going by the flag name it was 
not meant to differentiate between a FE and BE?
> Thank you for your help !!
>
> Best regards
> ---
> Kuninori Morimoto