[net-next PATCH RFC v3 1/8] dt-bindings: net: document ethernet PHY package nodes

[Andrew Lunn]

> On one hand it makes sense and looks useful for software development. On
> another, it looks like a violation of the main DT designing rule, when DT
> should be used to describe that hardware properties, which can not be learnt
> from other sources.
> 
> As far as I understand this specific chip, each of embedded PHYs has its own
> MDIO bus address and not an offset from a main common address. Correct me
> please, if I am got it wrong.

I don't have the datasheet for this specific PHY. But the concept of a
quad PHY in one package is well known. Take for example the
VSC8584. The datasheet is open on Microchips website. It has four
strapping pins to determine the addresses of the PHYs in the
package. The PHY number, 0 - 3 determine bits 0-1 of the MDIO
address. The 4 strapping pins then determine bit 2-5 of the address.

In theory, each PHY could have its own strapping pins, allowing it to
be set to any address, and two PHYs could even have the same
address. But i doubt anybody actually builds hardware like that. I
expect the base addresses is set at the package level, and then PHYs
are just offsets from this.

So to me, a range property does seem reasonable. However, I agree, we
need acceptance from the DT Maintainers.

     Andrew