Re: [PATCH v2 4/8] power: supply: core: tabularize HWMON temperature labels

[Guenter Roeck <linux@roeck-us.net>]

On 4/3/20 9:23 AM, Michał Mirosław wrote:
> Rework power_supply_hwmon_read_string() to check it's parameters.
> This allows to extend it later with labels for other types of
> measurements.
> 
> Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> ---
> v2: split from fix temperature labels
> ---
>  drivers/power/supply/power_supply_hwmon.c | 21 +++++++++++++++++++--
>  1 file changed, 19 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/power/supply/power_supply_hwmon.c b/drivers/power/supply/power_supply_hwmon.c
> index 67b6ee60085e..48c73994732c 100644
> --- a/drivers/power/supply/power_supply_hwmon.c
> +++ b/drivers/power/supply/power_supply_hwmon.c
> @@ -43,6 +43,11 @@ static int power_supply_hwmon_curr_to_property(u32 attr)
>  	}
>  }
>  
> +static const char *const ps_temp_label[] = {
> +	"temp",
> +	"ambient temp",
> +};
> +
>  static int power_supply_hwmon_temp_to_property(u32 attr, int channel)
>  {
>  	if (channel) {
> @@ -144,8 +149,20 @@ static int power_supply_hwmon_read_string(struct device *dev,
>  					  u32 attr, int channel,
>  					  const char **str)
>  {
> -	*str = channel ? "temp ambient" : "temp";
> -	return 0;
> +	if (channel < 0)
> +		return -EINVAL;
> +

This is unnecessary.

> +	switch (type) {
> +	case hwmon_temp:
> +		if (channel >= ARRAY_SIZE(ps_temp_label))
> +			return -EINVAL;

As is this. We don't usually check boundaries like this for
in-kernel APIs, and I personally would not want to have it
introduced in the kernel more than necessary. This just increases
kernel image size with zero benefit.

Guenter

> +		*str = ps_temp_label[channel];
> +		return 0;
> +	default:
> +		break;
> +	}
> +
> +	return -EINVAL;
>  }
>  
>  static int
>