Re: [PATCH v1 2/2] hwmon: (amc6821) Add PWM polarity configuration with OF

[Quentin Schulz <quentin.schulz@cherry.de>]

Hi Francesco,

On 2/18/25 5:56 PM, Francesco Dolcini wrote:
> From: Francesco Dolcini <francesco.dolcini@toradex.com>
> 
> Add support to configure the PWM-Out pin polarity based on a device
> tree property.
> 
> Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
> ---
>   drivers/hwmon/amc6821.c | 7 +++++--
>   1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/hwmon/amc6821.c b/drivers/hwmon/amc6821.c
> index 1e3c6acd8974..1ea2d97eebca 100644
> --- a/drivers/hwmon/amc6821.c
> +++ b/drivers/hwmon/amc6821.c
> @@ -845,7 +845,7 @@ static int amc6821_detect(struct i2c_client *client, struct i2c_board_info *info
>   	return 0;
>   }
>   
> -static int amc6821_init_client(struct amc6821_data *data)
> +static int amc6821_init_client(struct i2c_client *client, struct amc6821_data *data)
>   {
>   	struct regmap *regmap = data->regmap;
>   	int err;
> @@ -864,6 +864,9 @@ static int amc6821_init_client(struct amc6821_data *data)
>   		if (err)
>   			return err;
>   
> +		if (of_property_read_bool(client->dev.of_node, "ti,pwm-inverted"))

I know that the AMC6821 is doing a lot of smart things, but this really 
tickled me. PWM controllers actually do support that already via 
PWM_POLARITY_INVERTED flag for example. See 
Documentation/devicetree/bindings/hwmon/adt7475.yaml which seems to be 
another HWMON driver which acts as a PWM controller. I'm not sure this 
is relevant, applicable or desired but I wanted to highlight this.

> +			pwminv = 1;
> +

This is silently overriding the module parameter.

I don't think this is a good idea, at the very least not silently.

I would suggest to add some logic in the probe function to set this 
value and check its consistency.

Something like:

"""
diff --git a/drivers/hwmon/amc6821.c b/drivers/hwmon/amc6821.c
index 1e3c6acd89740..3a13a914e2bbb 100644
--- a/drivers/hwmon/amc6821.c
+++ b/drivers/hwmon/amc6821.c
@@ -37,7 +37,7 @@ static const unsigned short normal_i2c[] = {0x18, 
0x19, 0x1a, 0x2c, 0x2d, 0x2e,
   * Insmod parameters
   */

-static int pwminv;	/*Inverted PWM output. */
+static int pwminv = -1;	/* -1 not modified by the user, 0 default PWM 
output, 1 inverted PWM output */
  module_param(pwminv, int, 0444);

  static int init = 1; /*Power-on initialization.*/
@@ -904,6 +904,7 @@ static int amc6821_probe(struct i2c_client *client)
  	struct amc6821_data *data;
  	struct device *hwmon_dev;
  	struct regmap *regmap;
+	bool pwminv_dt;
  	int err;

  	data = devm_kzalloc(dev, sizeof(struct amc6821_data), GFP_KERNEL);
@@ -916,6 +917,18 @@ static int amc6821_probe(struct i2c_client *client)
  				     "Failed to initialize regmap\n");
  	data->regmap = regmap;

+	pwminv_dt = of_property_read_bool(client->dev.of_node, "ti,pwm-inverted");
+
+	if (pwminv == -1) {
+		pwminv = pwminv_dt;
+	} else if (is_of_node(client->dev.fwnode)) {
+		if ((!pwminv_dt && pwminv) || (pwminv_dt && pwminv == 0)) {
+			dev_err(dev,
+				"Polarity of PWM output passed by module parameter (pwminv=%d) 
differs from the one provided through the Device Tree, ignoring Device 
Tree value\n",
+				pwminv);
+		}
+	}
+
  	err = amc6821_init_client(data);
  	if (err)
  		return err;
"""

maybe? Note that I have neither compiled nor tested this code.

This also changes the precedence compared to the patch you sent, I think 
we may want the module param to override the DT property if there's a 
conflict.

What do you think?

Cheers,
Quentin