Re: [PATCH 09/18] ASoC: sti: Update DT example to match the driver code

[Arnd Bergmann]

On Wednesday 04 May 2016 09:52:19 Arnaud Pouliquen wrote:
> hello Arnd, peter,
> 
> On 04/26/2016 01:44 PM, Arnd Bergmann wrote:
> > On Tuesday 26 April 2016 12:15:32 Peter Griffin wrote:
> >>>
> >>>> If not what would you recommend instead? 
> >>>
> >>> It's still not clear to me what that bit in the syscfg register
> >>> is for. Given the error message about "sti-audio-clk-glue",
> >>> I suspect that this is actually a clock controller and that
> >>> it should be using the clock binding with a separate driver
> >>> instead of manipulating the regmap directly from the audio driver.
> >>
> >> Luckily I do have the datasheet for the audio-glue sysconf register.
> >>
> >> It says: -
> >>
> >> [11:8] PCM_CLK_SEL: Selects the frequency synthesizer clock or the external
> >> PCM clock for each channel.
> >>
> >> The driver only ever sets this to 1 which selects the frequency synthesizer
> >> clock. So the bitfield of the register which the driver is using (PCM_CLK_SEL)
> >> is a clock mux.
> > 
> > Ok, that sounds like it could be either a really simple clock driver
> > with just a few lines, or integrated into an existing clock driver
> > if you already have one for this syscon node.
> > 
> >       Arnd
> > 
> FYI, Name of this glue is related to the register name. But it does not
> concern only clock...
>  This glue register is used to :
> - select clock source ( clock framework or external clock from GPIO)
> => one bit field per IP instance (player->clk_sel)
> - select uniperiph player IP instance for PCM out.
> (http://www.spinics.net/lists/alsa-devel/msg49034.html)

Ok, I see. This is of course again the STi platform being a bit different
from everyone else, and whatever we do to hide it won't give us a nice
abstraction. Having just a clock driver for the register won't do the
job here as you say, so I guess the original patch was already the
least awkward way to handle it.

	Arnd