Re: [PATCH 2/2] iio: frequency: adf4377: add adf4378 support

[Jonathan Cameron <jic23@kernel.org>]

On Wed, 17 Jul 2024 11:38:30 +0200
Krzysztof Kozlowski <krzk@kernel.org> wrote:

> On 17/07/2024 11:30, Antoniu Miclaus wrote:
> > Add separate handling for adf4378 within the driver.
> > 
> > The main difference between adf4377 and adf4378 is that adf4378 has only
> > one output which is handled by only one gpio.
> > 
> > Signed-off-by: Antoniu Miclaus <antoniu.miclaus@analog.com>

Replying on top of Krzysztof's review as he is raising very similar
points to those I was going to make.

> > ---
> >  drivers/iio/frequency/adf4377.c | 25 ++++++++++++++++++-------
> >  1 file changed, 18 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/iio/frequency/adf4377.c b/drivers/iio/frequency/adf4377.c
> > index 9284c13f1abb..e02298a8b47f 100644
> > --- a/drivers/iio/frequency/adf4377.c
> > +++ b/drivers/iio/frequency/adf4377.c
> > @@ -387,6 +387,11 @@
> >  #define ADF4377_FREQ_PFD_250MHZ			(250 * HZ_PER_MHZ)
> >  #define ADF4377_FREQ_PFD_320MHZ			(320 * HZ_PER_MHZ)
> >  
> > +enum adf4377_dev_type {
> > +	ADF4377,
> > +	ADF4378,
> > +};
See below - but using an enum for device type is normally a bad sign.
It means you are adding a bunch of code paths that will need continual
extension as new chips are added.

Much better to add a description of chip features in a const structure.
> > +
> >  enum {
> >  	ADF4377_FREQ,
> >  };
> > @@ -402,6 +407,7 @@ enum muxout_select_mode {
> >  
> >  struct adf4377_state {
> >  	struct spi_device	*spi;
> > +	enum adf4377_dev_type	type;
> >  	struct regmap		*regmap;
> >  	struct clk		*clkin;
> >  	/* Protect against concurrent accesses to the device and data content */
> > @@ -687,7 +693,7 @@ static void adf4377_gpio_init(struct adf4377_state *st)
> >  	if (st->gpio_enclk1)
> >  		gpiod_set_value(st->gpio_enclk1, 1);
> >  
> > -	if (st->gpio_enclk2)
> > +	if (st->gpio_enclk2 && st->type == ADF4377)  
> 
> Why? Isn't everything correct for NULL?
> 
> >  		gpiod_set_value(st->gpio_enclk2, 1);
> >  }
> >  
> > @@ -889,11 +895,13 @@ static int adf4377_properties_parse(struct adf4377_state *st)
> >  		return dev_err_probe(&spi->dev, PTR_ERR(st->gpio_enclk1),
> >  				     "failed to get the CE GPIO\n");
> >  
> > -	st->gpio_enclk2 = devm_gpiod_get_optional(&st->spi->dev, "clk2-enable",
> > -						  GPIOD_OUT_LOW);
> > -	if (IS_ERR(st->gpio_enclk2))
> > -		return dev_err_probe(&spi->dev, PTR_ERR(st->gpio_enclk2),
> > -				     "failed to get the CE GPIO\n");
> > +	if (st->type == ADF4377) {  
> 
> So the device does not have this pin? Then you should express it in the
> bindings.
Agreed: That binding needs to ensure that there isn't a second pin expressed
for a chip where it makes no sense.

> 
> > +		st->gpio_enclk2 = devm_gpiod_get_optional(&st->spi->dev, "clk2-enable",
> > +							  GPIOD_OUT_LOW);
> > +		if (IS_ERR(st->gpio_enclk2))
> > +			return dev_err_probe(&spi->dev, PTR_ERR(st->gpio_enclk2),
> > +					"failed to get the CE GPIO\n");
> > +	}
> >  
> >  	ret = device_property_match_property_string(&spi->dev, "adi,muxout-select",
> >  						    adf4377_muxout_modes,
> > @@ -945,6 +953,7 @@ static int adf4377_probe(struct spi_device *spi)
> >  
> >  	st->regmap = regmap;
> >  	st->spi = spi;
> > +	st->type = spi_get_device_id(spi)->driver_data;  
> 
> 
> spi_get_device_match_data()
> 
> >  	mutex_init(&st->lock);
> >  
> >  	ret = adf4377_properties_parse(st);
> > @@ -964,13 +973,15 @@ static int adf4377_probe(struct spi_device *spi)
> >  }
> >  
> >  static const struct spi_device_id adf4377_id[] = {
> > -	{ "adf4377", 0 },
> > +	{ "adf4377", ADF4377 },
> > +	{ "adf4378", ADF4378 },
> >  	{}
> >  };
> >  MODULE_DEVICE_TABLE(spi, adf4377_id);
> >  
> >  static const struct of_device_id adf4377_of_match[] = {
> >  	{ .compatible = "adi,adf4377" },
> > +	{ .compatible = "adi,adf4378" },  
> 
> Your device ID tables have incoherent match data. Considering that one
> type is 0, this is error-prone and discouraged.
Agreed.  Much better to use a pointer to a chip specific structure for these
thus avoiding the accidental NULL value and turning chip differences into
data, not code.

> 
> Best regards,
> Krzysztof
>