Re: [PATCH 2/3] RFC: dt-bindings: marvell: Rewrite in schema

[Rob Herring <robh@kernel.org>]

On Fri, Oct 13, 2023 at 03:04:10PM +0200, Linus Walleij wrote:
> Hi Andrew,
> 
> thanks for reviewing!
> 
> On Fri, Oct 13, 2023 at 2:43 PM Andrew Lunn <andrew@lunn.ch> wrote:
> 
> > > +properties:
> > > +  compatible:
> > > +    oneOf:
> > > +      - enum:
> > > +          - marvell,mv88e6060
> >
> > The 6060 is a separate driver. Its not part of mv88e6xxx. So it should
> > have a binding document of its own.
> 
> It really doesn't matter to the DT bindings.
> It is not the job of DT to reflect the state of Linux.
> 
> In another operating system they might all be the same driver.
> Or all four variants have their own driver.
> 
> If the hardware is distinctly different so a lot of the properties
> are unique then it may be warranted with a separate DT
> binding, for the sake of keeping bindings simpler and
> coherent.

Exactly.

> 
> > > +  '#interrupt-cells':
> > > +    description: The internal interrupt controller only supports triggering
> > > +      on IRQ_TYPE_LEVEL_HIGH
> > > +      # FIXME: what is this? this should be one cell should it not?
> > > +      # the Linux mv88e6xxx driver does not implement .irq_set_type in its irq_chip
> > > +      # so at least in that implementation the type is flat out ignored.
> > > +    const: 2
> >
> > This interrupt controller is for the embedded PHYs. Its is hard wired
> > active high.
> 
> Hmm.... I need feedback from the DT people here. It does have a
> polarity, but the polarity cannot be changed. So shall we encode this
> always the same polarity in the flags cell or skip it altogether?
> 
> I'm uncertain. The currens scheme does reflect a reality.

Either way is fine. If users are already doing 2 cells, then I'd 
probably just keep that and state that the flags cell is ignored/unused.

Rob