RE: [PATCH V7] iio: light: isl29018: add vcc regulator operation support

[Anson Huang]

Hi, Jonathan

Best Regards!
Anson Huang

> -----Original Message-----
> From: Jonathan Cameron [mailto:jic23@xxxxxxxxxx]
> Sent: 2018年12月23日 1:15
> To: Anson Huang <anson.huang@xxxxxxx>
> Cc: knaack.h@xxxxxx; lars@xxxxxxxxxx; pmeerw@xxxxxxxxxx;
> linux-iio@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; festevam@xxxxxxxxx;
> preid@xxxxxxxxxxxxxxxxx; dl-linux-imx <linux-imx@xxxxxxx>
> Subject: Re: [PATCH V7] iio: light: isl29018: add vcc regulator operation
> support
> 
> On Mon, 17 Dec 2018 03:25:17 +0000
> Anson Huang <anson.huang@xxxxxxx> wrote:
> 
> > The light sensor's power supply could be controllable by regulator on
> > some platforms, such as i.MX6Q-SABRESD board, the light sensor
> > isl29023's power supply is controlled by a GPIO fixed regulator, need
> > to make sure the regulator is enabled before any operation of sensor,
> > this patch adds vcc regulator operation support.
> >
> > Signed-off-by: Anson Huang <Anson.Huang@xxxxxxx>
> Hi Anson
> 
> See below.
> 
> > ---
> > ChangeLog since V6
> >     - using devm_regulator_get() instead of devm_regulator_get_optional()
> since the regulator is
> >       there anyway, if dtb does NOT specify one, regulator framework will
> assign dummy regulator for it;
> >     - Setup devm action for cleaning up regulator resource for error
> handling.
> > ---
> >  drivers/iio/light/isl29018.c | 58
> > ++++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 58 insertions(+)
> >
> > diff --git a/drivers/iio/light/isl29018.c
> > b/drivers/iio/light/isl29018.c index b45400f..63f7b9d 100644
> > --- a/drivers/iio/light/isl29018.c
> > +++ b/drivers/iio/light/isl29018.c
> > @@ -23,6 +23,7 @@
> >  #include <linux/mutex.h>
> >  #include <linux/delay.h>
> >  #include <linux/regmap.h>
> > +#include <linux/regulator/consumer.h>
> >  #include <linux/slab.h>
> >  #include <linux/iio/iio.h>
> >  #include <linux/iio/sysfs.h>
> > @@ -95,6 +96,7 @@ struct isl29018_chip {
> >  	struct isl29018_scale	scale;
> >  	int			prox_scheme;
> >  	bool			suspended;
> > +	struct regulator	*vcc_reg;
> >  };
> >
> >  static int isl29018_set_integration_time(struct isl29018_chip *chip,
> > @@ -708,6 +710,17 @@ static const char
> *isl29018_match_acpi_device(struct device *dev, int *data)
> >  	return dev_name(dev);
> >  }
> >
> > +static void isl29018_disable_regulator_action(void *_data) {
> > +	struct isl29018_chip *chip = _data;
> > +	int err;
> > +
> > +	err = regulator_disable(chip->vcc_reg);
> > +	if (err)
> > +		dev_err(regmap_get_device(chip->regmap),
> > +			"failed to disable VCC regulator!\n"); }
> > +
> >  static int isl29018_probe(struct i2c_client *client,
> >  			  const struct i2c_device_id *id)
> >  {
> > @@ -742,6 +755,37 @@ static int isl29018_probe(struct i2c_client *client,
> >  	chip->scale = isl29018_scales[chip->int_time][0];
> >  	chip->suspended = false;
> >
> > +	chip->vcc_reg = devm_regulator_get(&client->dev, "vcc");
> > +	if (IS_ERR(chip->vcc_reg)) {
> > +		err = PTR_ERR(chip->vcc_reg);
> > +		if (err != -EPROBE_DEFER)
> > +			dev_err(&client->dev, "failed to get VCC regulator!\n");
> > +		return err;
> > +	}
> > +
> > +	err = regulator_enable(chip->vcc_reg);
> > +	if (err) {
> > +		dev_err(&client->dev, "failed to enable VCC regulator!\n");
> > +		return err;
> > +	}
> > +
> > +	chip->regmap = devm_regmap_init_i2c(client,
> > +				isl29018_chip_info_tbl[dev_id].regmap_cfg);
> > +	if (IS_ERR(chip->regmap)) {
> > +		err = PTR_ERR(chip->regmap);
> > +		dev_err(&client->dev, "regmap initialization fails: %d\n", err);
> > +		regulator_disable(chip->vcc_reg);
> > +		return err;
> > +	}
> > +
> > +	err = devm_add_action(&client->dev, isl29018_disable_regulator_action,
> > +				 chip);
> > +	if (err) {
> 
> I'm a little confused, why not do this before devm_regmap_init_i2c.
> That way you won't have to disable the regulator in that one error path.
> Also, devm_add_action_or_reset will call isl29018_disable_regulator_action
> for you on error.

It is because I used dev_err() in isl29018_disable_regulator_action which need regmap
to get "dev" by regmap_get_device(chip->regmap), if it is accepted by just using
pr_err() instead of dev_err, then I can do the devm_add_action before devm_regmap_init_i2c.

I think using pr_err should be OK, I will use it in V8 patch.

> 
> > +		isl29018_disable_regulator_action(chip);
> > +		dev_err(&client->dev, "failed to setup regulator cleanup action!\n");
> > +		return err;
> > +	}
> > +
> >  	chip->regmap = devm_regmap_init_i2c(client,
> >  				isl29018_chip_info_tbl[dev_id].regmap_cfg);
> >  	if (IS_ERR(chip->regmap)) {
> > @@ -768,6 +812,7 @@ static int isl29018_probe(struct i2c_client
> > *client,  static int isl29018_suspend(struct device *dev)  {
> >  	struct isl29018_chip *chip = iio_priv(dev_get_drvdata(dev));
> > +	int ret;
> >
> >  	mutex_lock(&chip->lock);
> >
> > @@ -777,6 +822,12 @@ static int isl29018_suspend(struct device *dev)
> >  	 * So we do not have much to do here.
> >  	 */
> >  	chip->suspended = true;
> > +	ret = regulator_disable(chip->vcc_reg);
> > +	if (ret) {
> > +		dev_err(dev, "failed to disable VCC regulator\n");
> > +		mutex_unlock(&chip->lock);
> > +		return ret;
> 
> Given you are about to unlock anyway a common pattern is to not check ret
> until after the unlock, thus simplifying the code.

I will improve this in V8 patch, thanks.

Anson.

> 
> > +	}
> >
> >  	mutex_unlock(&chip->lock);
> >
> > @@ -790,6 +841,13 @@ static int isl29018_resume(struct device *dev)
> >
> >  	mutex_lock(&chip->lock);
> >
> > +	err = regulator_enable(chip->vcc_reg);
> > +	if (err) {
> > +		dev_err(dev, "failed to enable VCC regulator\n");
> > +		mutex_unlock(&chip->lock);
> > +		return err;
> > +	}
> > +
> >  	err = isl29018_chip_init(chip);
> >  	if (!err)
> >  		chip->suspended = false;