Re: [PATCH 1/3] iio: adc: mt6359-auxadc: correct channel types to IIO_VOLTAGE

[Jonathan Cameron <jic23@kernel.org>]

On Tue, 10 Nov 2026 20:39:47 +0000
Roman Vivchar <rva333@protonmail.com> wrote:

> Hi Jonathan,
> 
> On Wednesday, November 4th, 2026 at 1:42 AM, Jonathan Cameron <jic23@kernel.org> wrote:
> 
> > On Tue, 03 Nov 2026 18:32:07 +0200
> > Roman Vivchar via B4 Relay <devnull+rva333.protonmail.com@kernel.org> wrote:  
> 
> ...
> 
> > >
> > > Correct all channel types to IIO_VOLTAGE.
> > >
> > > Fixes: 3587914bf61d ("iio: adc: Add support for MediaTek MT6357/8/9 Auxiliary ADC")
> > > Fixes: d6f49313272b ("iio: adc: mt6359: Add support for MediaTek MT6363 PMIC AUXADC")
> > > Signed-off-by: Roman Vivchar <rva333@protonmail.com>  
> > 
> > This one makes me very nervous, not because it is wrong but rather
> > because there may be a bunch of userspace relying on these wrong
> > channel types.
> > 
> > It is horrible but I think we can't touch that :(  What we can do is
> > put the right interface in place as well. Thankfully it looks like
> > the channel index is independent of the channel type so we can duplicate
> > the wrong channels with voltage ones + some comments to make sure this
> > doesn't propagate to any new supported devices.  
> 
> Sorry I forgot to include this in my previous reply, but do you suggest
> duplicating the old channels and adding the fixed ones as separate entries?

Yes.

> 
> That would cause issues for in-kernel consumers. In my tree I have a
> battery thermal zone driven by ntc_thermistor (which is actually how I
> found this bug, as it refused to probe due to the wrong channel type),
> and I will soon send a new revision of the PMIC thermal driver. They all
> require IIO_VOLTAGE channels to operate properly. If we duplicate the
> channels, consumers will get the wrong type unless we reorder the array
> or add a custom xlate callback.

Indeed, one of those two solutions should deal with this.
> 
> Existing userspace tools would get bogus values from the current ABI
> anyway. The driver applies a raw mV scale to all channels, while thermal
> needs efuse calibration with a special formula applied, and even plain
> voltage channels like bat_adc and cdt_v were mislabeled as resistance and
> temperature.

Ah.  If the values are effectively useless we can hopefully assume no
one is actually using them and hence the risk is low.
> 
> Given all this, I think it's better to simply fix channel types.
> 
> Best regards,
> Roman
>