Re: [PATCH v2 5/8] dt-bindings: iio: add sigma delta modulator backend

[Olivier MOYSAN]

Hi Conor,

On 6/27/24 18:13, Conor Dooley wrote:
> On Wed, Jun 26, 2024 at 06:40:58PM +0200, Olivier MOYSAN wrote:
> > Hi Conor,
> >
> > On 6/25/24 17:34, Conor Dooley wrote:
> > > On Tue, Jun 25, 2024 at 05:07:13PM +0200, Olivier Moysan wrote:
> > > > Add documentation of device tree bindings to support
> > > > sigma delta modulator backend in IIO framework.
> > > >
> > > > Signed-off-by: Olivier Moysan <olivier.moysan@xxxxxxxxxxx>
> > >
> > > I don't review bindings for a job, I can only reliably get to look at
> > > my mail queue in the evenings, please give me a chance to reply to you
> > > before you submit a new version.
> >
> > Sorry, the short review delay.
> >
> > > > +$id: http://devicetree.org/schemas/iio/adc/sd-modulator-backend.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Sigma delta modulator backend
> > >
> > > Same comments about filename and title apply here as the previous
> > > version. "TI $foo Sigma Delta Modulator" and drop the reference to back
> > > ends or the pretence of being generic.
> >
> > The logic here is the same as for the former sigma delta modulator driver.
> > (see discussion [1])
> > I mean introducing a generic and minimalist driver to support sd modulators,
> > but not dedicated to a specific modulator. The ads1201 is chosen as a basic
> > modulator here. But it is rather an arbitrary choice.
> >
> > I agree "backend" reference is not really relevant here. I have to think
> > about a way to manage the coexistence of this sigma delta modulator driver
> > with its former version.
>
> To be blunt, I don't care about drivers! Well I do, but not in this
> particular context. You can absolutely have a driver that supports
> multiple backends or sigma delta modulators, but right now we are
> talking about a binding and this binding supports exactly one sigma
> delta modulator - and with an explicit compatible. In that context,
> presenting the binding as generic makes little sense.
>
> > [1] https://lore.kernel.org/all/6943aaf5-b580-0fd1-7a2e-b99f7a266388@xxxxxx/
>
> Looking at this though, I question the binding more... The programming
> model of the device is identical as a backend or otherwise, so it
> shouldn't be getting a new compatible. Isn't this actually as simple as
> adding #io-backend-cells to the existing binding and using that to
> determine whether the device is being used as a backend or in isolation?

For sure. I came to the same conclusion. My first idea was to isolate 
the deprecated binding. However, I agree that the best approach is to 
adapt the existing binding. I prepared a v3 like this.

BRs
Olivier

> Thanks,
> Conor.