Re: [alsa-devel] [PATCH v2 4/4] ASoC: codecs: add wsa881x amplifier support

[Srinivas Kandagatla]

Thanks for taking time to review,

On 08/08/2019 16:18, Pierre-Louis Bossart wrote:
> > +/* 4 ports */
> > +static struct sdw_dpn_prop wsa_sink_dpn_prop[WSA881X_MAX_SWR_PORTS] = {
> > +ÂÂÂ {
> > +ÂÂÂÂÂÂÂ /* DAC */
> > +ÂÂÂÂÂÂÂ .num = 1,
> > +ÂÂÂÂÂÂÂ .type = SDW_DPN_SIMPLE,
>
> IIRC we added the REDUCED type in SoundWire 1.1 to cover the PDM case 
> with channel packing (or was it grouping) used by Qualcomm. I am not 
> sure the SIMPLE type works?
grouping I guess.

This is a simplified data port as there is no DPn_OffsetCtrl2 register 
implemented.

Having said below channel count looks incorrect, i should fix that.

> > +ÂÂÂÂÂÂÂ .min_ch = 1,
> > +ÂÂÂÂÂÂÂ .max_ch = 8,
> > +ÂÂÂÂÂÂÂ .simple_ch_prep_sm = true,
> > +ÂÂÂ }, {
> > +ÂÂÂÂÂÂÂ /* COMP */
> > +ÂÂÂÂÂÂÂ .num = 2,
> > +ÂÂÂÂÂÂÂ .type = SDW_DPN_SIMPLE,
> > +ÂÂÂÂÂÂÂ .min_ch = 1,
> > +ÂÂÂÂÂÂÂ .max_ch = 8,
> > +ÂÂÂÂÂÂÂ .simple_ch_prep_sm = true,
> > +ÂÂÂ }, {
> > +ÂÂÂÂÂÂÂ /* BOOST */
> > +ÂÂÂÂÂÂÂ .num = 3,
> > +ÂÂÂÂÂÂÂ .type = SDW_DPN_SIMPLE,
> > +ÂÂÂÂÂÂÂ .min_ch = 1,
> > +ÂÂÂÂÂÂÂ .max_ch = 8,
> > +ÂÂÂÂÂÂÂ .simple_ch_prep_sm = true,
> > +ÂÂÂ }, {
> > +ÂÂÂÂÂÂÂ /* VISENSE */
> > +ÂÂÂÂÂÂÂ .num = 4,
> > +ÂÂÂÂÂÂÂ .type = SDW_DPN_SIMPLE,
> > +ÂÂÂÂÂÂÂ .min_ch = 1,
> > +ÂÂÂÂÂÂÂ .max_ch = 8,
> > +ÂÂÂÂÂÂÂ .simple_ch_prep_sm = true,
> > +ÂÂÂ }
> > +};
>
> > +static int wsa881x_update_status(struct sdw_slave *slave,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ enum sdw_slave_status status)
> > +{
> > +ÂÂÂ struct wsa881x_priv *wsa881x = dev_get_drvdata(&slave->dev);
> > +
> > +ÂÂÂ if (status == SDW_SLAVE_ATTACHED) {
>
> there is an ambiguity here, the Slave can be attached but as device0 
> (not enumerated). We should check dev_num > 0
Thanks for point that! will add a check!

> > +ÂÂÂÂÂÂÂ if (!wsa881x->regmap) {
> > +ÂÂÂÂÂÂÂÂÂÂÂ wsa881x->regmap = devm_regmap_init_sdw(slave,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ &wsa881x_regmap_config);
> > +ÂÂÂÂÂÂÂÂÂÂÂ if (IS_ERR(wsa881x->regmap)) {
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ dev_err(&slave->dev, "regmap_init failed\n");
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ return PTR_ERR(wsa881x->regmap);
> > +ÂÂÂÂÂÂÂÂÂÂÂ }
> > +ÂÂÂÂÂÂÂ }
> > +
> > +ÂÂÂÂÂÂÂ return snd_soc_register_component(&slave->dev,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ &wsa881x_component_drv,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ NULL, 0);
> > +ÂÂÂ } else if (status == SDW_SLAVE_UNATTACHED) {
> > +ÂÂÂÂÂÂÂ snd_soc_unregister_component(&slave->dev);
>
> the update_status() is supposed to be called based on bus events, it'd 
> be very odd to register/unregister the component itself dynamically. In 
> our existing Realtek-based solutions the register_component() is called 
> in the probe function (and unregister_component() in remove). We do the 
> inits when the Slave becomes attached but the component is already 
> registered.
looks less intrusive!  I will give that a try!

> > +ÂÂÂ }
> > +
> > +ÂÂÂ return 0;
> > +}
> > +
> > +
> > +static int wsa881x_remove(struct sdw_slave *sdw)
> > +{
> > +ÂÂÂ return 0;
> > +}
> > +
> > +static const struct sdw_device_id wsa881x_slave_id[] = {
> > +ÂÂÂ SDW_SLAVE_ENTRY(0x0217, 0x2010, 0),
> > +ÂÂÂ {},
> > +};
> > +MODULE_DEVICE_TABLE(sdw, wsa881x_slave_id);
> > +
> > +static struct sdw_driver wsa881x_codec_driver = {
> > +ÂÂÂ .probeÂÂÂ = wsa881x_probe,
> > +ÂÂÂ .remove = wsa881x_remove,
>
> is this needed since you do nothing in that function?

yes, it can be removed! will do that in next version.

--srini