Re: [PATCH net-next v2 07/10] dt-bindings: net: enforce phylink bindings on certain ethernet controllers

[Arınç ÜNAL]

On 18.09.2023 21:13, Rob Herring wrote:
> On Sat, Sep 16, 2023 at 02:08:59PM +0300, Arınç ÜNAL wrote:
> > Phylink bindings are required for ethernet controllers that utilise
> > phylink_fwnode_phy_connect() directly or through phylink_of_phy_connect(),
> > and register OF-based only MDIO buses, if they register any.
>
> What is phylink?
>
> Don't describe/justify binding changes based on some Linux functions.

I'd like to discuss the influence of the Linux drivers on the devicetree
bindings. Before that let me clarify these "phylink" bindings. They are
certain rules to properly describe the link of the ethernet controller. Any
of the phy-handle, pcs-handle, sfp, and fixed-link properties describe the
link. They are already defined on ethernet-controller.yaml, just not
enforced.

Why I called them phylink bindings, is because when phylink is used on the
ethernet controller driver, only then we can be sure that the link
descriptions on the ethernet controller are needed. Because, some drivers
don't need the link descriptions on the ethernet controller as they can
search a certain MDIO bus to find and connect a MAC to a PHY or set up a
fixed link without looking at the devicetree.

I don't like this approach because, as you said, devicetree bindings should
describe the hardware, not driver policies. So I propose that we don't let
Linux drivers interfere with dt-bindings. To that extent:

- Link descriptions must be required on ethernet controllers. We don't care
  whether some Linux driver can or cannot find the PHY or set up a fixed
  link without looking at the devicetree.

- The MDIO bus description on an ethernet controller describes that the
  hardware comes with an MDIO bus. The MDIO property is optional not
  because some Linux driver can find and map the PHYs to MACs without DT
  descriptions but because the hardware doesn't have to use that or any
  MDIO bus.

Let's also talk about the elephant in the room: DSA bindings. Distributed
Switch Architecture is a Linux subsystem, not a hardware property.
Devicetree bindings are supposed to describe the hardware, yet we have
bindings specifically for DSA so that the DSA driver can properly control
the hardware.

Although I see dsa.yaml and dsa-port.yaml mostly consist of describing an
ethernet switch with CPU port(s), there're properties that are specific to
DSA, such as dsa,member on dsa.yaml and dsa-tag-protocol and label on
dsa-port.yaml.

Depending on how I get responses, I will address my patches. Then I can
further discuss the DSA bindings.

Arınç