Re: [PATCH v1 03/11] arm64: dts: freescale: imx8mm-var-som: Update FEC support with MaxLinear PHY

[Andrew Lunn]

On Wed, Feb 18, 2026 at 09:59:31AM +0100, Stefano Radaelli wrote:
> Hi Andrew,
> 
> On Tue, Feb 17, 2026 at 09:43:16PM +0100, Andrew Lunn wrote:
> > 
> > And here you loose me. RGMII is between the MAC and the PHY. The PHY
> > is on the SOM, so there is no need for RGMII to leave the SOM.
> > 
> > MDIO is can understand going to the carrier, because there can be
> > other devices on the bus, such as an Ethernet switch.
> > 
> > You say the LEDs are not on the SOM. So they should not be listed in
> > the .dtsi file for the SOM. It could be the RJ45 is a cheap one,
> > without LEDs. The colours will depend on the RJ45 socket. It could be
> > a single LED is connected to the front panel, which is typical of STB
> > etc. So the LEDs are a board property, not a SOM property.
> > 
> > > > >   Two LEDs are defined to match the VAR-SOM carrier design:
> > > > >     * LED@0: Yellow, netdev trigger.
> > > > >     * LED@1: Green, netdev trigger.
> > > > 
> > > > Where are the LEDs? You say "carrier design", so are they on the
> > > > carrier? The DT properties should then be in the .dts file for the
> > > > carrier.
> > > >
> > > 
> > > The LED signals are driven directly by the PHY and originate on the SOM.
> > > They are exposed on the SOM connector and are typically routed straight
> > > to the RJ45 integrated LEDs on the carrier.
> > 
> > The problem here is 'typically', not absolutely required, otherwise
> > the magic smoke will escape making the box useless.
> > 
> 
> You are right sorry for the confusion: RGMII is internal between the MAC
> and the PHY and does not leave the SOM. What is exposed on the SOM
> connector are the MDI pairs towards the magnetics/RJ45 and the PHY LED
> output signals.

O.K, that is what i expected.

> Given that the actual LEDs are implemented on the carrier side (RJ45
> with integrated LEDs, pull-ups and LED supply), would you prefer that
> the LED description is moved out of the SOM .dtsi and into the carrier
> .dts?

Yes please.

> My idea would be to keep the Ethernet controller/PHY node fully
> described in the SOM .dtsi (since the PHY is on-module), and add only
> the LED subnode/definition in the carrier .dts where the LEDs are
> actually present.

Yes. That is good.

Thanks
     Andrew