Re: [PATCH] regulator: add new regulator driver for lp872x

[Mark Brown <broonie@opensource.wolfsonmicro.com>]

[-- Attachment #1: Type: text/plain, Size: 3366 bytes --]

On Wed, May 23, 2012 at 07:50:59AM +0000, Kim, Milo wrote:

> +static int lp872x_read_byte(struct lp872x *lp, u8 addr, u8 *data)
> +{
> +	int ret;
> +	unsigned int val;
> +
> +	mutex_lock(&lp->xfer_lock);
> +	ret = regmap_read(lp->regmap, addr, &val);
> +	if (ret < 0) {
> +		mutex_unlock(&lp->xfer_lock);
> +		dev_err(lp->dev, "failed to read 0x%.2x\n", addr);
> +		return ret;
> +	}
> +	mutex_unlock(&lp->xfer_lock);
> +
> +	*data = (u8)val;
> +	return 0;
> +}

This should just be a trivial wrapper around regmap_read, the regmap API
handles concurrency for you.  Similarly for all your other I/O
functions.

> +static int lp872x_ldo_list_voltage(struct regulator_dev *rdev,
> +				unsigned selector)

regulator_list_voltage_table() will appear in -next just after the merge
window.

> +static int lp872x_ldo_set_voltage_sel(struct regulator_dev *rdev,
> +				unsigned selector)

regulator_set_voltage_sel_regmap().

> +static int lp872x_ldo_get_voltage_sel(struct regulator_dev *rdev)

regulator_get_voltage_sel_regmap().

> +static int lp872x_regulator_enable(struct regulator_dev *rdev)

Your ENABLE define made the code a bit obscure but this is
regulator_enable_regmap()

> +static int lp872x_regualtor_disable(struct regulator_dev *rdev)

regulator_disable_regmap()

> +static int lp872x_regulator_is_enabled(struct regulator_dev *rdev)

regulator_is_enabled_regmap()

> +static int lp872x_buck_list_voltage(struct regulator_dev *rdev,
> +				unsigned selector)

regulator_list_voltage_table()

> +	case LP8720_ID_BUCK:
> +		if (val & LP8720_EXT_DVS_M) {
> +			pinstate = pdata->get_dvs_pin_state ?
> +				pdata->get_dvs_pin_state() : DVS_LOW;

This is a really odd implementation.  Why isn't the driver managing the
pin and what happens to the consumers that expected to set a voltage
when the pin gets changed?

> +static int lp8725_buck_enable(struct regulator_dev *rdev)

regulator_enable_regmap()

> +static int lp8725_buck_disable(struct regulator_dev *rdev)

regulator_disable_regmap()

> +static int lp8725_buck_is_enabled(struct regulator_dev *rdev)

regulator_is_enabled_regmap()

> +	for (i = 0 ; i < num ; i++) {
> +		regulator_data = lp->pdata->regulator_data + i;
> +		desc = &lp872x_regulator_desc[regulator_data->id];
> +		init_data = regulator_data->init_data;
> +
> +		rdev = regulator_register(desc, lp->dev, init_data, lp, NULL);
> +		if (IS_ERR(rdev)) {
> +			dev_err(lp->dev, "regulator register err");
> +			ret =  PTR_ERR(rdev);
> +			goto err;
> +		}

You should unconditionally register all the regulators the chip has and
match the init_data to them rather than only registering regulators you
have init_data.  This allows users to inspect unconfigured regulators
and allows the core to support things like powering off unused
regulators.

> +static const struct regmap_config lp872x_regmap_config = {
> +	.reg_bits = 8,
> +	.val_bits = 8,
> +};

Setting max_register would get you debugfs inspection of the regmap.

> +	lp->regmap = regmap_init_i2c(cl, &lp872x_regmap_config);

devm_regmap_init_i2c().

> +/**
> + * lp872x_platform_data
> + * @general_config    (mandatory) : the value of LP872X_GENERAL_CFG register

Is the chip default never valid?

> + * @regulator_data    (mandatory) : platform regulator id and init data
> + * @num_regulators    (mandatory) : total number of platform regulators

These should be optional.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]