Re: [PATCH v2 1/5] regulator: dt-bindings: Add Linear Technology LTM8054 regulator

[David Lechner <dlechner@baylibre.com>]

On Wed, Oct 1, 2025 at 8:40 PM Conor Dooley <conor@kernel.org> wrote:
>
> On Wed, Oct 01, 2025 at 01:18:51PM +0200, David Lechner wrote:
> > On Wed, Oct 1, 2025 at 9:12 AM Romain Gantois
> > <romain.gantois@bootlin.com> wrote:
> > >
> > > On Sunday, 28 September 2025 00:31:05 CEST Conor Dooley wrote:
> > > ...
> > > > > >
> > > > > > > +  lltc,fb-voltage-divider:
> > > > > > Why does this property have a ?linear? vendor prefix?
> > > > > > Shouldn't it be adi to match the other property and compatible?
> > > > >
> > > > > This component was originally from Linear Technology, before it was
> > > > > acquired by Analog Devices. The new properties and compatibles have the
> > > > > Analog Devices prefix, but the "fb-voltage-divider" property is already
> > > > > used by the LTC3676 and LTC3589 regulators, so I left the Linear
> > > > > Technology prefix for this one to avoid introducing a new property just
> > > > > to specify a vendor prefix change.
> > > > >
> > > > > I don't have a strong opinion about this though.
> > > >
> > > > Do they share the same driver?
> > >
> > > They do not. However, they use it in the exact same way, and I would've
> > > liked to factor out the handling of this property in a future patch. This
> > > would also make it easier to handle other types of feedback pin circuits
> > > and have a generic binding for "regulators using a feedback pin connected
> > > to some kind of analog circuit".
> > >
> > > For example:
> > >
> > > Vout----+
> > >         |
> > >         |
> > >        +++
> > >        | |
> > >        | | Rtop
> > >        | |
> > >        +++
> > >         |
> > >         |
> > >  FB ----+
> > >         |
> > >      +--+--+
> > >      |  |  |
> > >      |  |  |CCS
> > >      +--v--+
> > >         |
> > >         |
> > >        -+-
> > >         -
> > >
> > > This is all speculation at this point though, so I don't mind changing the
> > > property to "adi,fb-voltage-divider" and handling the different compatibles
> > > when it comes to it.
> > >
> >
> > Could we just make it `fb-voltage-divider-ohms`? The -ohms suffix
> > makes it match the standard property-units suffix which already has
> > the uint32-array type. There are a couple of bindings that have
> > `vout-voltage-divider` without a vendor prefix, so it sounds like this
> > pattern is considered somewhat of a standard property already. But I
> > think it would be better with the -ohms suffix. For example, there is
> > already `gw,voltage-divider-ohms` as well. But there are so many
> > similar properties without the suffix, it is kind of the defacto
> > standard already, so might be better to stick with that rather than
> > making it even more different variants than there already are.
>
> Ye, by all means standardise it. I suppose that means insertion into
> regulator.yaml, which usually also means a regulator- prefix - unless
> you're eyeing something wider than that?

Yes, there are also hwmon and iio bindings that are already already
using variations of ([a-z]+,)?(([a-x]+-)voltage-divider(-ohms)?.
Although `fb-voltage-divider` specifically seems to only be common for
regulators, so could make sense to just have
`regulator-fb-voltage-divider-ohms`.