Re: [PATCH] hwmon: adm1275: Make sure we are reading enough data for different chips

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

On 7/8/20 4:53 PM, Chu Lin wrote:
> Issue:
> When binding adm1272 to the adm1275 driver, PEC error is reported.
> See blow:

s/blow/below/

Also, unless I am missing something, the error is only seen if PEC is
enabled.

> adm1275: probe of xxxx failed with error -74
> 
> Diagnosis:
> Per the datasheet of adm1272 and adm1278 (www.analog.com),
> PMON_CONFIG (0xd4) is 16bits wide. On the other hand,
> PMON_CONFIG (0xd4) for adm1275 is 8bits wide.
> The driver should not assume everything is 8bits wide and read only
> 8bits from it.
> 
> Solution:
> if it is adm1272 or adm1278, use i2c_read_word. Else, use i2c_read_byte
> 

See below for more chips with 16-bit configuration registers.

> Signed-off-by: Chu Lin <linchuyuan@google.com>
> ---
>  drivers/hwmon/pmbus/adm1275.c | 24 ++++++++++++++++--------
>  1 file changed, 16 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/hwmon/pmbus/adm1275.c b/drivers/hwmon/pmbus/adm1275.c
> index e25f541227da..a0d547d20358 100644
> --- a/drivers/hwmon/pmbus/adm1275.c
> +++ b/drivers/hwmon/pmbus/adm1275.c
> @@ -472,6 +472,7 @@ static int adm1275_probe(struct i2c_client *client,
>  	struct adm1275_data *data;
>  	const struct i2c_device_id *mid;
>  	const struct coefficients *coefficients;
> +	s32 (*config_read_fn_ptr)(const struct i2c_client *client, u8 reg);

_ptr is really not needed here and redundant. A variable pointing to a function
is always a pointer.

>  	int vindex = -1, voindex = -1, cindex = -1, pindex = -1;
>  	int tindex = -1;
>  	u32 shunt;
> @@ -510,14 +511,6 @@ static int adm1275_probe(struct i2c_client *client,
>  			   "Device mismatch: Configured %s, detected %s\n",
>  			   id->name, mid->name);
>  
> -	config = i2c_smbus_read_byte_data(client, ADM1275_PMON_CONFIG);
> -	if (config < 0)
> -		return config;
> -
> -	device_config = i2c_smbus_read_byte_data(client, ADM1275_DEVICE_CONFIG);
> -	if (device_config < 0)
> -		return device_config;
> -

The reason for doing this here was to avoid the memory allocation failure
on error. I don't really see the point of moving the code block.

>  	data = devm_kzalloc(&client->dev, sizeof(struct adm1275_data),
>  			    GFP_KERNEL);
>  	if (!data)
> @@ -532,6 +525,21 @@ static int adm1275_probe(struct i2c_client *client,
>  
>  	data->id = mid->driver_data;
>  
> +	/* adm1272 and adm1278 supports temperature monitoring.  The config and device config
> +	 * are 16bits wide for these two devices. On the other hand, other devices are 8 bits wide
> +	 */

Standard multi-line comments, please. Also, the destinction is really
based on have_power_sampling, though that is set only later.

> +	if (data->id == adm1272 || data->id == adm1278)
> +		config_read_fn_ptr = &i2c_smbus_read_word_data;

Function pointers don't need &.
For ADM1293 and ADM1294, the registers are also 16 bit wide
(again, see use of have_power_sampling).

Thanks,
Guenter

> +	else
> +		config_read_fn_ptr = &i2c_smbus_read_byte_data;
> +	config = config_read_fn_ptr(client, ADM1275_PMON_CONFIG);
> +	if (config < 0)
> +		return config;
> +
> +	device_config = config_read_fn_ptr(client, ADM1275_DEVICE_CONFIG);
> +	if (device_config < 0)
> +		return device_config;
> +
>  	info = &data->info;
>  
>  	info->pages = 1;
>