Re: [PATCH v2] Fix FSP-3Y YH-5151E non-compliant vout encoding

[Guenter Roeck]

On 4/28/21 1:38 AM, Václav Kubernát wrote:
> I didn't properly test the driver for YH-5151E, so it was completely
> broken. Firstly, the log/real mapping was incorrect in one case.
> Secondly, PMBus specifies that output voltages should be in the linear16
> encoding. However, the PDU is non-compliant and uses linear11. YM-2151E
> isn't affected by this. Fix this by converting the values inside the
> read functions. linear16 gets the exponent from the VOUT_MODE command.
> The device doesn't support it, so I have to manually supply the value
> for it.
> 
> Both supported devices have now been tested to report correct vout
> values.
> 

Fixes: 1734b4135a62 ("hwmon: Add driver for fsp-3y PSUs and PDUs")

> Signed-off-by: Václav Kubernát <kubernat@xxxxxxxxx>
> ---
>  drivers/hwmon/pmbus/fsp-3y.c | 26 ++++++++++++++++++++++++--
>  1 file changed, 24 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/hwmon/pmbus/fsp-3y.c b/drivers/hwmon/pmbus/fsp-3y.c
> index 564649e87e34..dc147a79d6aa 100644
> --- a/drivers/hwmon/pmbus/fsp-3y.c
> +++ b/drivers/hwmon/pmbus/fsp-3y.c
> @@ -57,7 +57,7 @@ static int page_log_to_page_real(int page_log, enum chips chip)
>  		case YH5151E_PAGE_12V_LOG:
>  			return YH5151E_PAGE_12V_REAL;
>  		case YH5151E_PAGE_5V_LOG:
> -			return YH5151E_PAGE_5V_LOG;
> +			return YH5151E_PAGE_5V_REAL;
>  		case YH5151E_PAGE_3V3_LOG:
>  			return YH5151E_PAGE_3V3_REAL;
>  		}
> @@ -103,17 +103,28 @@ static int set_page(struct i2c_client *client, int page_log)
>  
>  static int fsp3y_read_byte_data(struct i2c_client *client, int page, int reg)
>  {
> +	const struct pmbus_driver_info *info = pmbus_get_driver_info(client);
> +	struct fsp3y_data *data = to_fsp3y_data(info);
>  	int rv;
>  
>  	rv = set_page(client, page);
>  	if (rv < 0)
>  		return rv;
>  
> +	/*
> +	 * YH5151-E outputs vout in linear11. The conversion is done later, but
> +	 * I have to inject pmbus_core with the correct exponent.

s/I/we/

Also, it might be useful to mention the actual exponent (-6 ?) in the comment.

> +	 */
> +	if (data->chip == yh5151e && reg == PMBUS_VOUT_MODE)
> +		return 0x1A;

Can this code be moved ahead of set_page() ?

> +
>  	return i2c_smbus_read_byte_data(client, reg);
>  }
>  
>  static int fsp3y_read_word_data(struct i2c_client *client, int page, int phase, int reg)
>  {
> +	const struct pmbus_driver_info *info = pmbus_get_driver_info(client);
> +	struct fsp3y_data *data = to_fsp3y_data(info);
>  	int rv;
>  
>  	/*
> @@ -144,7 +155,18 @@ static int fsp3y_read_word_data(struct i2c_client *client, int page, int phase,
>  	if (rv < 0)
>  		return rv;
>  
> -	return i2c_smbus_read_word_data(client, reg);
> +	rv = i2c_smbus_read_word_data(client, reg);
> +	if (rv < 0)
> +		return rv;
> +
> +	/*
> +	 * The PDU is non-compliant and outputs output voltages in linear11

s/The PDU/YH-5151E/

> +	 * instead of linear16.
> +	 */
> +	if (data->chip == yh5151e && reg == PMBUS_READ_VOUT)
> +		rv = ((s16)((rv & 0x7ff) << 5)) >> 5;

Problem with this is that it auto-converts to int and will return a
negative value if bit 10 is set (I wrote some test code and confirmed it,
just to be sure).
		rv = sign_extend32(rv, 10) & 0xffff;
should work here.

Thanks,
Guenter

> +
> +	return rv;
>  }
>  
>  struct pmbus_driver_info fsp3y_info[] = {
>