> > > +
> > > +static int al3000a_read_raw(struct iio_dev *indio_dev,
> > > + struct iio_chan_spec const *chan, int *val,
> > > + int *val2, long mask)
> > > +{
> > > + struct al3000a_data *data = iio_priv(indio_dev);
> > > + int ret, gain;
> > > +
> > > + switch (mask) {
> > > + case IIO_CHAN_INFO_RAW:
> > > + ret = regmap_read(data->regmap, AL3000A_REG_DATA, &gain);
> > > + if (ret < 0)
> > > + return ret;
> > > +
> > > + *val = lux_table[gain & AL3000A_GAIN_MASK];
> >
> > Why did you chosen to do post-processing in the RAW channel instead
> > doing it in INFO_SCALE (same as al3010 does)?
> >
>
> From my observation INFO_SCALE will just perform multiplication of RAW
> to SCALE. In this case values which are read are not actual raw values
> of illumination. Next is my assumption (since there is no datasheet),
> but values obtained from register are similar to values from adc
> thermal sensors, they need be converted via reference table to get
> actual data.
Please add a comment somewhere here to say that we don't know the
relationship of these values to illuminance hence providing
_RAW and _SCALE would not be helpful.
>
> > Except this, LGTM.
> >
> > Documentation and DT patch:
> >
> > Reviewed-by: David Heidelberg <david@xxxxxxx>
> > > +
> > > + return IIO_VAL_INT;
> > > + case IIO_CHAN_INFO_SCALE:
> > > + *val = 1;
> > > +
Don't do this. The above lack of known relationship has to be
expressed by not providing the _scale attribute.
> > > + return IIO_VAL_INT;
> > > + default:
> > > + return -EINVAL;
> > > + }
> > > +}
> > > +
> > > +static const struct iio_info al3000a_info = {
> > > + .read_raw = al3000a_read_raw,
> > > +};
