On Monday, 1 September 2025 17:35:23 Central European Summer Time Andy
Shevchenko wrote:
> On Sun, Aug 31, 2025 at 12:33:05PM +0200, Duje Mihanović wrote:
> > +static const struct regmap_config pm886_gpadc_regmap_config = {
> > + .reg_bits = 8,
> > + .val_bits = 8,
> > + .max_register = PM886_REG_GPADC_VBAT_SLP + 1,
>
> What is this + 1 register? Why is it not defined / documented?
It is the second field of the vbat_slp channel.
> > +static int
> > +gpadc_get_resistance_ohm(struct iio_dev *iio, struct iio_chan_spec const
> > *chan) +{
> > + struct pm886_gpadc *gpadc = iio_priv(iio);
> > + unsigned int raw_uv, raw_ua;
> > + int ret;
> > +
> > + ret = gpadc_set_bias(gpadc, chan->channel, true);
> > + if (ret)
> > + goto err;
> > +
> > + ret = gpadc_find_bias_current(iio, chan, &raw_uv, &raw_ua);
> > + if (ret)
> > + goto err;
> > +
> > + ret = gpadc_set_bias(gpadc, chan->channel, false);
> > + if (ret)
> > + return ret;
> > +
> > + return DIV_ROUND_CLOSEST(raw_uv, raw_ua);
> > +err:
> > + gpadc_set_bias(gpadc, chan->channel, false);
>
> You do the same in the other branch and checking there for an error. Why this
> one is so special?
My rationale here was to not override the error from either the first
gpadc_set_bias() call or the subsequent gpadc_find_bias_current() call.
> > + dev_dbg(&iio->dev, "chan: %d, raw: %d\n", chan->channel, *val);
>
> How is this useful? The userspace gets the same value. Do you expect problems
> on its way to the user space?
That's a leftover from debugging v1, will drop.
> > + iio->dev.of_node = dev->parent->of_node;
>
> No, use device_set_node() with the respective parameters.
>
> But rather debug why firmware node (or OF in your case) is not propagated from
> the parent device.
I guess it is because the IIO device is registered as a child of the
GPADC platform device, which does not have a node unlike the PMIC
device (GPADC pdev's parent). It seems that the regulator cell
registers its regulators directly under the PMIC dev, so maybe I should
do the same here with the IIO dev?
Regards,
--
Duje
