Re: [PATCH 1/2] hwmon: pmbus: Add support for ltc2971

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

On 10/26/24 02:08, Patryk Biel wrote:
> LTC2971 is a power manager similar to already supported LTC2972,
> it uses the same register set but supports a different voltage range.
> 
> Signed-off-by: Patryk Biel <pbiel7@gmail.com>
> ---
>   drivers/hwmon/pmbus/ltc2978.c | 34 +++++++++++++++++++++++++++++++---

Please also update Documentation/hwmon/ltc2978.rst and drivers/hwmon/pmbus/Kconfig.

>   1 file changed, 31 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/hwmon/pmbus/ltc2978.c b/drivers/hwmon/pmbus/ltc2978.c
> index 73a86f4d647288be97615f19a5c6314e4036e6a3..681e7b811dbcc263fe8f5e4f9da30fcbc7e019ad 100644
> --- a/drivers/hwmon/pmbus/ltc2978.c
> +++ b/drivers/hwmon/pmbus/ltc2978.c
> @@ -21,7 +21,8 @@
>   
>   enum chips {
>   	/* Managers */
> -	ltc2972, ltc2974, ltc2975, ltc2977, ltc2978, ltc2979, ltc2980,
> +	ltc2971, ltc2972, ltc2974, ltc2975, ltc2977, ltc2978,
> +	ltc2979, ltc2980,
>   	/* Controllers */
>   	ltc3880, ltc3882, ltc3883, ltc3884, ltc3886, ltc3887, ltc3889, ltc7132, ltc7880,
>   	/* Modules */
> @@ -61,6 +62,7 @@ enum chips {
>   
>   #define LTC2978_ID_MASK			0xfff0
>   
> +#define LTC2971_ID			0x0320
>   #define LTC2972_ID			0x0310
>   #define LTC2974_ID			0x0210
>   #define LTC2975_ID			0x0220
> @@ -533,6 +535,7 @@ static int ltc2978_write_word_data(struct i2c_client *client, int page,
>   }
>   
>   static const struct i2c_device_id ltc2978_id[] = {
> +	{"ltc2971", ltc2971},
>   	{"ltc2972", ltc2972},
>   	{"ltc2974", ltc2974},
>   	{"ltc2975", ltc2975},
> @@ -564,11 +567,19 @@ MODULE_DEVICE_TABLE(i2c, ltc2978_id);
>   
>   #if IS_ENABLED(CONFIG_SENSORS_LTC2978_REGULATOR)
>   #define LTC2978_ADC_RES	0xFFFF
> +#define LTC2978_UV_STEP	1000
> +
> +/* Common for most chips */
>   #define LTC2978_N_ADC	122
>   #define LTC2978_MAX_UV	(LTC2978_ADC_RES * LTC2978_N_ADC)
> -#define LTC2978_UV_STEP	1000
>   #define LTC2978_N_VOLTAGES	((LTC2978_MAX_UV / LTC2978_UV_STEP) + 1)
>   

Please just add the chip specific defines as you do below, and don't move
the defines above around. It happens a lot that there are common definitions
followed by chip specific ones. Spelling that out explicitly doesn't really
add much if any value and just results in rearranged code whenever a new
chip with other parameters is added.

> +/* LTC2971 */
> +#define LTC2971_N_ADC	4500
> +#define LTC2971_MAX_UV	(LTC2978_ADC_RES * LTC2971_N_ADC)

I think something is wrong here.

4500 * 65535 uV ~= 295V.

However, the datasheet says that the output voltage is calculated by
(register value / 1024), where the register value is a 16-bit value.
65535 / 1024 ~= 64V. Where does the 4,500 come from ?

> +#define LTC2971_N_VOLTAGES	((LTC2971_MAX_UV / LTC2978_UV_STEP) + 1)

Does the chip really support 294,908 voltages ? That seems impossible since
the VOUT_COMMAND register is only 16 bit wide.

Unless I misunderstand the datasheet, the voltage resolution should be 1/1024V
or 976 uV, and there should be 65,536 voltages. This is from the chip datasheet,
PMBus Command Description, Note 1: Data Formats, for L16 data.

If my understanding is wrong, please explain and document your numbers.

Thanks,
Guenter