Re: [PATCH v2 4/7] ASoC: apple: mca: Support FEs being clock consumers

[Joshua Peisach]

On Sun Jun 28, 2026 at 7:42 AM EDT, James Calligeros wrote:
> From: Martin Povišer <povik+lin@xxxxxxxxxxx>
>
> Support FEs being I2S clock consumers. This does not mean we support
> accepting clocks from outside the SoC (although it paves the way for
> that support in the future), but it means multiple FEs can attach to one
> BE, one being clock producer and the rest clock consumers.
>
> This is useful for grabbing I/V sense data on some machines, since in
> such a scenario the format of the sense data on the I2S bus differs
> from that of the audio data (the two formats differing in slot width).
> With two FEs attached to the bus, we can split the responsibilities and
> command different slot widths to the two.
>
> Signed-off-by: Martin Povišer <povik+lin@xxxxxxxxxxx>
> Signed-off-by: James Calligeros <jcalligeros99@xxxxxxxxx>
> ---
>  sound/soc/apple/mca.c | 129 ++++++++++++++++++++-----
>  1 file changed, 106 insertions(+), 23 deletions(-)
>
> diff --git a/sound/soc/apple/mca.c b/sound/soc/apple/mca.c
> index 730ffe0805e1..0c1fc8242907 100644
> --- a/sound/soc/apple/mca.c
> +++ b/sound/soc/apple/mca.c
> @@ -380,6 +406,57 @@ static int mca_be_hw_free(struct snd_pcm_substream *substream,
>  	return 0;
>  }
>  
> +static int mca_fe_prepare(struct snd_pcm_substream *substream,
> +			  struct snd_soc_dai *dai)
> +{
> +	struct mca_cluster *cl = mca_dai_to_cluster(dai);
> +	struct mca_data *mca = cl->host;
> +
> +	if (cl->clk_provider)
> +		return 0;
> +
> +	/* Turn on the cluster power domain if not already in use */
> +	if (!cl->syncgen_in_use) {
> +		int port = mca_fe_get_port(substream);
> +
> +		cl->pd_link = device_link_add(mca->dev, cl->pd_dev,
> +					      DL_FLAG_STATELESS | DL_FLAG_PM_RUNTIME |
> +						DL_FLAG_RPM_ACTIVE);
> +		if (!cl->pd_link) {
> +			dev_err(mca->dev,
> +				"cluster %d: unable to prop-up power domain\n", cl->no);
> +			return -EINVAL;
> +		}
> +
> +		writel_relaxed(port + 6 + 1,
> +			       cl->base + REG_SYNCGEN_MCLK_SEL);

Where does the "6 + 1" come from, and should it have its own #define?

-Josh