Re: [PATCH v2 2/2] hwmon: iio: Add labels from IIO channels

[Sean Anderson <sean.anderson@linux.dev>]

On 6/24/24 14:47, Guenter Roeck wrote:
> On 6/24/24 10:46, Sean Anderson wrote:
>> Add labels from IIO channels to our channels. This allows userspace to
>> display more meaningful names instead of "in0" or "temp5".
>>
>> Although lm-sensors gracefully handles errors when reading channel
>> labels, the ABI says the label attribute
>>
>>> Should only be created if the driver has hints about what this voltage
>>> channel is being used for, and user-space doesn't.
>>
>> Therefore, we test to see if the channel has a label before
>> creating the attribute.
>>
> 
> FWIW, complaining about an ABI really does not belong into a commit
> message. Maybe you and lm-sensors don't care about error returns when
> reading a label, but there are other userspace applications which may
> expect drivers to follow the ABI. Last time I checked, the basic rule
> was still "Don't break userspace", and that doesn't mean "it's ok to
> violate / break an ABI as long as no one notices".

This isn't complaining about the ABI, just documenting the reason it was
done this way...

>> Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
>> ---
>>
>> Changes in v2:
>> - Check if the label exists before creating the attribute
>>
>>   drivers/hwmon/iio_hwmon.c | 45 ++++++++++++++++++++++++++++++++++-----
>>   1 file changed, 40 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/hwmon/iio_hwmon.c b/drivers/hwmon/iio_hwmon.c
>> index 4c8a80847891..5722cb9d81f9 100644
>> --- a/drivers/hwmon/iio_hwmon.c
>> +++ b/drivers/hwmon/iio_hwmon.c
>> @@ -33,6 +33,17 @@ struct iio_hwmon_state {
>>       struct attribute **attrs;
>>   };
>>   +static ssize_t iio_hwmon_read_label(struct device *dev,
>> +                  struct device_attribute *attr,
>> +                  char *buf)
>> +{
>> +    struct sensor_device_attribute *sattr = to_sensor_dev_attr(attr);
>> +    struct iio_hwmon_state *state = dev_get_drvdata(dev);
>> +    struct iio_channel *chan = &state->channels[sattr->index];
>> +
>> +    return iio_read_channel_label(chan, buf);
>> +}
>> +
> 
> I personally find it a bit kludgy that an in-kernel API would do a
> sysfs write like this and expect a page-aligned buffer as parameter,
> but since Jonathan is fine with it:

This is also how the in-kernel iio_read_channel_ext_info API works.
I agree that it is a strange API, but I do not want to rewrite all
the implementing drivers.

--Sean

> Acked-by: Guenter Roeck <linux@roeck-us.net>
> 
> Jonathan, please apply through your tree.
> 
> Thanks,
> Guenter
>