Re: [PATCH 2/2] ALSA: hda/tas2563: Add tas2563 HDA driver

[Pierre-Louis Bossart]

> diff --git a/sound/pci/hda/patch_realtek.c b/sound/pci/hda/patch_realtek.c
> index 9677c09cf7a9..1d3e9f77c9d4 100644
> --- a/sound/pci/hda/patch_realtek.c
> +++ b/sound/pci/hda/patch_realtek.c
> @@ -6770,7 +6770,7 @@ static int comp_match_cs35l41_dev_name(struct device *dev, void *data)
>  	return !strcmp(d + n, tmp);
>  }
>  
> -static int comp_match_tas2781_dev_name(struct device *dev,
> +static int comp_match_tas2xxx_dev_name(struct device *dev,
>  	void *data)
>  {
>  	struct scodec_dev_name *p = data;
> @@ -6823,7 +6823,7 @@ static void cs35l41_generic_fixup(struct hda_codec *cdc, int action, const char
>  	}
>  }
>  
> -static void tas2781_generic_fixup(struct hda_codec *cdc, int action,
> +static void tas2xxx_generic_fixup(struct hda_codec *cdc, int action,
>  	const char *bus, const char *hid)
>  {
>  	struct device *dev = hda_codec_dev(cdc);
> @@ -6841,7 +6841,7 @@ static void tas2781_generic_fixup(struct hda_codec *cdc, int action,
>  		rec->index = 0;
>  		spec->comps[0].codec = cdc;
>  		component_match_add(dev, &spec->match,
> -			comp_match_tas2781_dev_name, rec);
> +			comp_match_tas2xxx_dev_name, rec);
>  		ret = component_master_add_with_match(dev, &comp_master_ops,
>  			spec->match);
>  		if (ret)
> @@ -6888,7 +6888,13 @@ static void alc287_fixup_legion_16ithg6_speakers(struct hda_codec *cdc, const st
>  static void tas2781_fixup_i2c(struct hda_codec *cdc,
>  	const struct hda_fixup *fix, int action)
>  {
> -	 tas2781_generic_fixup(cdc, action, "i2c", "TIAS2781");
> +	 tas2xxx_generic_fixup(cdc, action, "i2c", "TIAS2781");
> +}

this sort of rename should be part of a separate patch IMHO, it'd be
easier to review.

> +
> +static void tas2563_fixup_i2c(struct hda_codec *cdc,
> +	const struct hda_fixup *fix, int action)
> +{
> +	 tas2xxx_generic_fixup(cdc, action, "i2c", "INT8866");

Any specific reason to use an Intel ACPI identifier? Why not use
"TIAS2563" ?

> +#define TAS2563_REG_INIT_N 12

newline

> +static const struct reg_default tas2563_reg_init[TAS2563_MAX_CHANNELS]
> +	[TAS2563_REG_INIT_N] = {
> +	{
> +		{ TAS2562_TDM_CFG2, 0x5a },
> +		{ TAS2562_TDM_CFG4, 0xf3 },
> +		{ TAS2562_TDM_CFG5, 0x42 },
> +		{ TAS2562_TDM_CFG6, 0x40 },
> +		{ TAS2562_BOOST_CFG1, 0xd4 },
> +		{ TAS2562_BOOST_CFG3, 0xa4 },
> +		{ TAS2562_REG(0x00, 0x36), 0x0b },
> +		{ TAS2562_REG(0x00, 0x38), 0x21 },
> +		{ TAS2562_REG(0x00, 0x3c), 0x58 },
> +		{ TAS2562_BOOST_CFG4, 0xb6 },
> +		{ TAS2562_ASI_CONFIG3, 0x04},
> +		{ TAS2562_REG(0x00, 0x47), 0xb1 },

> +/* Update the calibrate data, including speaker impedance, f0, etc, into algo.

update the calibration data,

> + * Calibrate data is done by manufacturer in the factory. These data are used

The manufacturer calibrates the data in the factory.

> + * by Algo for calucating the speaker temperature, speaker membrance excursion

calculating

membrane


> +static int tas2563_hda_i2c_probe(struct i2c_client *client)
> +{
> +	struct tas2563_data *tas2563;
> +	int ret;
> +
> +	tas2563 = devm_kzalloc(&client->dev, sizeof(struct tas2563_data),
> +		GFP_KERNEL);
> +	if (!tas2563)
> +		return -ENOMEM;
> +	tas2563->dev = &client->dev;
> +	tas2563->client = client;
> +
> +	dev_set_drvdata(tas2563->dev, tas2563);
> +
> +	ret = tas2563_read_acpi(tas2563);
> +	if (ret)
> +		return dev_err_probe(tas2563->dev, ret,
> +			"Platform not supported\n");
> +
> +	for (int i = 0; i < tas2563->ndev; ++i) {
> +		struct tas2563_dev *tasdev = &tas2563->tasdevs[i];
> +
> +		ret = tas2563_tasdev_read_efi(tas2563, tasdev);
> +		if (ret)
> +			return dev_err_probe(tas2563->dev, ret,
> +				"Calibration data cannot be read from EFI\n");
> +
> +		ret = tas2563_tasdev_init_client(tas2563, tasdev);
> +		if (ret)
> +			return dev_err_probe(tas2563->dev, ret,
> +				"Failed to init i2c client\n");
> +
> +		ret = tas2563_tasdev_init_regmap(tas2563, tasdev);
> +		if (ret)
> +			return dev_err_probe(tas2563->dev, ret,
> +				"Failed to allocate register map\n");
> +	}
> +
> +	ret = component_add(tas2563->dev, &tas2563_hda_comp_ops);
> +	if (ret) {
> +		return dev_err_probe(tas2563->dev, ret,
> +			"Register component failed\n");
> +	}

I wonder how many of those tests actually depend on deferred probe, and
if this isn't a case of copy-paste "just in case"?

> +
> +	pm_runtime_set_autosuspend_delay(tas2563->dev, 3000);
> +	pm_runtime_use_autosuspend(tas2563->dev);
> +	pm_runtime_mark_last_busy(tas2563->dev);
> +	pm_runtime_set_active(tas2563->dev);
> +	pm_runtime_get_noresume(tas2563->dev);
> +	pm_runtime_enable(tas2563->dev);
> +
> +	pm_runtime_put_autosuspend(tas2563->dev);

the sequence get_noresume/enable/put_autosuspend makes no sense to me.
doing a get_noresume *before* enable should do exactly nothing, and
releasing the resource would already be handled with autosuspend based
on the last_busy mark.

> +
> +	return 0;
> +}
> +
> +static void tas2563_hda_i2c_remove(struct i2c_client *client)
> +{
> +	struct tas2563_data *tas2563 = dev_get_drvdata(&client->dev);
> +
> +	pm_runtime_get_sync(tas2563->dev);
> +	pm_runtime_disable(tas2563->dev);
> +
> +	component_del(tas2563->dev, &tas2563_hda_comp_ops);
> +
> +	pm_runtime_put_noidle(tas2563->dev);

that pm_runtime sequence also makes no sense to me, if you disable
pm_runtime the last command is useless/no-op.

> +}
> +
> +static int tas2563_system_suspend(struct device *dev)
> +{
> +	struct tas2563_data *tas2563 = dev_get_drvdata(dev);
> +	int ret;
> +
> +	dev_dbg(tas2563->dev, "System Suspend\n");
> +
> +	ret = pm_runtime_force_suspend(dev);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +static int tas2563_system_resume(struct device *dev)
> +{
> +	int ret;
> +	struct tas2563_data *tas2563 = dev_get_drvdata(dev);
> +
> +	dev_dbg(tas2563->dev, "System Resume\n");
> +
> +	ret = pm_runtime_force_resume(dev);
> +	if (ret)
> +		return ret;
> +
> +	for (int i = 0; i < tas2563->ndev; ++i)
> +		tas2563_tasdev_setup(tas2563, &tas2563->tasdevs[i]);
> +
> +	return 0;
> +}
> +
> +static const struct dev_pm_ops tas2563_hda_pm_ops = {
> +	SYSTEM_SLEEP_PM_OPS(tas2563_system_suspend, tas2563_system_resume)

where's the pm_runtime stuff?

> +};