Re: [PATCH net-next v3 2/9] dt-bindings: net: mediatek-dwmac: add support for MT8189 SoC

[Andrew Lunn <andrew@lunn.ch>]

On Wed, Oct 21, 2026 at 11:43:39AM +0200, Louis-Alexis Eyraud wrote:
> Hi Andrew,
> 
> On Fri, 2026-10-16 at 17:40 +0200, Andrew Lunn wrote:
> > > Finally, update "mediatek,rx-delay-ps" and "mediatek,tx-delay-ps"
> > > property
> > > description to add MT8189 RX/TX delay stage divider values, as they
> > > differ from existing supported SoC.
> > 
> > These should be mark deprecated and the standard rx-internal-delay-ps
> > and tx-internal-delay-ps used in there place for new devices.
> > 
> I agree, it makes sense to use them to describe the RGMII internal
> delay configurations rather than the two existing vendor properties
> (mediatek,rx-delay-ps and mediatek,tx-delay-ps), and deprecate them, by
> adding a deprecated flag and the comments you've suggested. It would
> also need an additional patch for the driver to handle them (while
> keeping backward compatibility).
> 
> But what is bothering me is these two vendor properties description
> says they can be used in MII and RMII phy mode cases. The driver uses
> them indeed also to fine tune the RX/TX_CLK (in MII) or the REF_CLK (in
> RMII) configuration.
> rx/tx-internal-delay-ps properties are limited to RGMII only ([1]).
> So they can not substitute them as it and with their description it
> would wrong to use them for MII/RMII.
> 
> At the moment, in the upstream devicetrees, there is no "mediatek,rx-
> delay-ps" use and the "mediatek,tx-delay-ps" use are very few and only
> for RGMII phy modes (either rgmii-rxid or rmgii-id), so no MII/RMII
> real use cases.
> 
> So, how to proceed then to make these bindings right?
> Use rx/tx-internal-delay-ps properties for RGMII and mediatek,rx/tx-
> delay-ps for MII/RMII or do you see another solution ?

If you say "mediatek,rx-delay-ps" is not actually used, then it is not
an issue. As for "mediatek,tx-delay-ps", i would probably ignore it
until somebody actually requires it for MII or RMII. At that point, we
can add specific properties for MII and RMII, and not mix them up with
RGMII delays.

      Andrew