On Mon, 9 Dec 2024 01:58:59 +0000
Kuninori Morimoto <kuninori.morimoto.gx@renesas.com> wrote:

> Hi Stephen
> 
> > > You should use one of the audio-graph-card bindings for anything
> > > new.  
> 
> Using audio-graph-card is good idea, but all
> simple_card/audio-graph-card/audio-graph-card2 are using same logic
> around here. So, you will have same issue on audio-graph-card too.
> 
> > > > Basically, simple_card appears to set the CPU as producer if you
> > > > don't specify a producer. I am not sure whether this is a bug.
> > > >   
> > > 
> > > Well, if nothing is configured it's got to pick a default?  
> 
> If my understand was correct, your issue can be solved...
> 
> 	dailink_out_master: simple-audio-card,dai-link@0 {
> 		...
> =>		pcm3168_playback: codec {  
> 			...
> 		};
> 	};
> 	dailink_in_slave: simple-audio-card,dai-link@1 {
> =>		bitclock-master = <&pcm3168_playback>;
> =>		frame-master    = <&pcm3168_playback>;  
> 		...
> 	};
> 
> asoc_simple_parse_daifmt() is just checking where the node was codec
> node or not. So, if bitclock-master/frame-master were produced, but
> was not codec, both CPU/Codec can be consumer ?
> 
> # Clock producer/consumer settings is very confusable, because it was
> # Codec base, and has flip, etc...
> 
> Thank you for your help !!
> 
> Best regards
> ---
> Kuninori Morimoto

Hi Morimoto-san,

This is one of the things I tried initially, but I wasn't sure if it
was a valid configuration. 

I just tried it again with debug enabled (see dt_snippet.txt) and I get
the attached (see dbgout.txt) in dmesg.

It is trying to set the CPU end as producer. I think it's because
asoc_simple_parse_daifmt() checks whether the bit/frame phandles passed
match the codec phandle - since it doesn't (it's a different codec),
it sets the DAI link as "consumer mode" (i.e. clocks come from CPU).
Therefore, the CPU side gets configured as producer.

I am using the  version from 6.12.3 as that is the latest I can
build.

Regards
Stephen