Re: [PATCH 4/5] ASoC: codecs: Add WCD939x Soundwire slave driver

[Neil Armstrong]

Hi,

On 23/11/2023 18:43, Mark Brown wrote:
> On Thu, Nov 23, 2023 at 03:49:14PM +0100, Neil Armstrong wrote:
> > Add Soundwire Slave driver for the WCD9390/WCD9395 Audio Codec.
>
> > The WCD9390/WCD9395 Soundwire Slaves will be used by the
>
> Please avoid using outdated terminology, "device" is probably a good
> alternative here.  There are some usages in APIs that need to be updated
> but still, good to avoid where possible.

Sure, I simply kept the current Soundwire kernel terminologies,
I'll change it to Device.

> > +static struct wcd939x_sdw_ch_info wcd939x_sdw_tx_ch_info[] = {
> > +	WCD_SDW_CH(WCD939X_ADC1, WCD939X_ADC_1_4_PORT, BIT(0)),
> > +	WCD_SDW_CH(WCD939X_ADC2, WCD939X_ADC_1_4_PORT, BIT(1)),
> > +	WCD_SDW_CH(WCD939X_ADC3, WCD939X_ADC_1_4_PORT, BIT(2)),
> > +	WCD_SDW_CH(WCD939X_ADC4, WCD939X_ADC_1_4_PORT, BIT(3)),
> > +	// TOFIX support ADC3/4 & DMIC0/1 on port 2
> > +	//WCD_SDW_CH(WCD939X_ADC3, WCD939X_ADC_DMIC_1_2_PORT, BIT(0)),
> > +	//WCD_SDW_CH(WCD939X_ADC4, WCD939X_ADC_DMIC_1_2_PORT, BIT(1)),
> > +	//WCD_SDW_CH(WCD939X_DMIC0, WCD939X_ADC_DMIC_1_2_PORT, BIT(2)),
> > +	//WCD_SDW_CH(WCD939X_DMIC1, WCD939X_ADC_DMIC_1_2_PORT, BIT(3)),
>
> Why are these commented out?

Future feature support, will remove and add those on my internal TODO list...

> > +static int wcd9390_interrupt_callback(struct sdw_slave *slave,
> > +				      struct sdw_slave_intr_status *status)
> > +{
> > +	struct wcd939x_sdw_priv *wcd = dev_get_drvdata(&slave->dev);
> > +	struct irq_domain *slave_irq = wcd->slave_irq;
> > +	u32 sts1, sts2, sts3;
> > +
> > +	do {
> > +		handle_nested_irq(irq_find_mapping(slave_irq, 0));
> > +		regmap_read(wcd->regmap, WCD939X_DIGITAL_INTR_STATUS_0, &sts1);
> > +		regmap_read(wcd->regmap, WCD939X_DIGITAL_INTR_STATUS_1, &sts2);
> > +		regmap_read(wcd->regmap, WCD939X_DIGITAL_INTR_STATUS_2, &sts3);
> > +
> > +	} while (sts1 || sts2 || sts3);
> > +
> > +	return IRQ_HANDLED;
> > +}
>
> We do this in the other Qualcomm drivers but it doesn't seem ideal to
> just ignore the interrupts.

It seems we simply ignore IRQs that are not mapped in the regmap_irq,
what would be the ideal way to handle this ?

> > +static int wcd939x_sdw_component_bind(struct device *dev, struct device *master,
> > +				      void *data)
> > +{
> > +	return 0;
> > +}
> > +
> > +static void wcd939x_sdw_component_unbind(struct device *dev,
> > +					 struct device *master, void *data)
> > +{
> > +}
> > +
> > +static const struct component_ops wcd939x_sdw_component_ops = {
> > +	.bind = wcd939x_sdw_component_bind,
> > +	.unbind = wcd939x_sdw_component_unbind,
> > +};
>
> Do these need to be provided if they can legitimately be empty?

AFAIK yes, component code will crash if those are not defined.
I'll add a comment explaining whey they are no-op.

Thanks,
Neil