Re: [PATCH v2 07/10] dt-bindings: adv748x: add information about serial audio interface (I2S/TDM)

[Alex Riesen]

Geert Uytterhoeven, Fri, Mar 20, 2020 10:15:17 +0100:
> On Fri, Mar 20, 2020 at 10:03 AM Alex Riesen <alexander.riesen@xxxxxxxxxxx> wrote:
> > Geert Uytterhoeven, Fri, Mar 20, 2020 09:48:14 +0100:
> > > On Fri, Mar 20, 2020 at 9:44 AM Alex Riesen <alexander.riesen@xxxxxxxxxxx> wrote:
> > > > Laurent Pinchart, Thu, Mar 19, 2020 19:01:25 +0100:
> > > > > On Thu, Mar 19, 2020 at 06:42:36PM +0100, Alex Riesen wrote:
> > > > > > As the driver has some support for the audio interface of the device,
> > > > > > the bindings file should mention it.
> > >
> > > > > > @@ -16,6 +18,8 @@ Required Properties:
> > > > > >      slave device on the I2C bus. The main address is mandatory, others are
> > > > > >      optional and remain at default values if not specified.
> > > > > >
> > > > > > +  - #clock-cells: must be <0> if the I2S port is used
> > > > >
> > > > > Wouldn't it be simpler to set it to 0 unconditionally ?
> > > >
> > > > Would it? If the port itself is optional, shouldn't the clock be an option
> > > > too?
> > >
> > > You'd be surprised how many board designers would consider this a cheap
> > > 12.288 MHz clock source, without using the I2S port ;-)
> >
> > Well, I am :-)
> >
> > Especially considering that the driver will not switch the MCLK pin aktive
> > (all I2S-related pins are tristate by default).
> 
> OK, didn't consider that.  But that still won't stop the hardware designer.
> E.g. on Lager, the clock input of the PMIC is tied to the clock line of an SPI
> bus, so to use that feature, the SPI clock must be kept running all the time,
> not just when doing a transfer.

Well... Maybe there is a convention to spell out the default state of the
clock lines?

> > And how do I require it to be set unconditionally? By just removing the
> > "if ..." part of the statement?
> 
> Indeed.  This is still the plain text binding, not yaml.

Conversion to YAML is on the list :)

Regards,
Alex