Re: [PATCH v3] ASoC: simple-card-utils: add sysclk ordering support

[Stefano Radaelli]

Hi Kuninori,
> 
> > (snip)
> > > +enum simple_util_sysclk_order {
> > > +	SIMPLE_SYSCLK_ORDER_UNSPEC = 0,
> > > +	SIMPLE_SYSCLK_ORDER_CODEC_FIRST,
> > > +	SIMPLE_SYSCLK_ORDER_CPU_FIRST,
> > > +};
> > 
> > I think UNSPEC is not needed ?
> > 
> > 	enum simple_util_sysclk_order {
> > 		SIMPLE_SYSCLK_ORDER_CODEC_FIRST = 0,
> > 		SIMPLE_SYSCLK_ORDER_CPU_FIRST,
> > 	};
> > 
> > Do you set this "order" via custome driver if you want ot use CPU_FIRST ?
> 
> Or, not use flag, but check whether "clock master" ?
> 

thanks for the feedback!

Yes, my initial idea was to set CPU_FIRST from our downstream/custom
kernel (e.g. via a small local change) when needed.

Also thanks for the suggestion about checking the clock master.
I had a look at it, but unfortunately the information we can reliably
derive is about BCLK/FS (bitclock-master and frame-master).
These properties describe who provides the serial bus clocks, but they
do not imply anything about MCLK.

In my case, for example, the codec is both frame-master and bitclock-master,
yet I still need to call CPU DAI .set_sysclk() first because the CPU DAI may
finalize/adjust the actual MCLK rate (e.g. via clk_set_rate()) and the codec
may otherwise configure its internal clocking based on a non-final MCLK value.

So selecting the sysclk ordering based on BCLK/FS master would not
cover this case.

So, do you think the solution I proposed would be acceptable if I remove
SIMPLE_SYSCLK_ORDER_UNSPEC and keep only:

enum simple_util_sysclk_order {
        SIMPLE_SYSCLK_ORDER_CODEC_FIRST = 0,
        SIMPLE_SYSCLK_ORDER_CPU_FIRST,
};

with codec-first as the default behaviour?

Many thanks and best regards,
Stefano Radaelli