[PATCH 2/4] dt-bindings: net: dsa: document internal MDIO bus

[Arınç ÜNAL]

On 10.09.2023 00:16, Andrew Lunn wrote:
> Please trim the text when replying.
> 
> 
>> I'm writing below as a mix of patch log and discussion.
>>
>> Phylink bindings are required for ports that are controlled by OF-based
>> buses. DSA, like any other driver utilising the Linux MDIO infrastructure,
>> can register a bus. If I understand correctly, non-OF-based registration of
>> OpenFirmware MDIO buses is a feature specific to DSA which certain DSA
>> subdrivers make use of.
> 
> This is not really DSA specific. Any MAC driver, or MDIO driver, can
> call mdiobus_regsiter(), or of_mdiobus_register() and pass a NULL
> pointer if there is no OF node available. It then requires that the
> MAC driver uses an function like phy_find_first(), or knows via other
> means what address the PHY uses on the bus. For DSA, that other means
> is assuming a 1:1 mapping between port number and bus address.

Understood. At least a phy-handle on the DSA user port for each PHY 
controlled by non-DSA drivers is always needed correct? Otherwise DSA 
wouldn't know which PHY to map the DSA switch port?

And that means DSA requires OF-based buses to map the ports controlled 
by non-DSA driver buses to PHYs?

I'm trying to understand if phylink bindings for DSA user ports that are 
controlled by non-DSA driver buses are always necessary.

Would this also apply to MAC drivers that control switches?

Arınç