Re: [PATCH 3/9] net: mdio: realtek-rtl9300: Add ports to info structure

[Daniel Golle <daniel@makrotopia.org>]

On Wed, May 20, 2026 at 03:42:44AM +0200, Andrew Lunn wrote:
> > > Do any of these MDIO busses appear on the 
> > > outside? Somewhere i can connect an 
> > > external PHY, or another switch in cascade?
> > 
> > Yes. See 6.14 in datasheet [1]
> > We even can disable/enable each on RTL931x
> 
> So how do you handle that this external PHY is not for a port of the
> switch? It is for an i210 hanging off the PCIe bus, or a USB-Ethernet
> dongle? Linux itself would have no problem with that.

On the *primary* MDIO busses of this class of SoCs (ie. which this
driver deals with) it only possible to access MDIO devices which are
Ethernet PHYs mapped to a switch port.

Theoretically you could maybe use pinctrl/pinmux to switch the MDC and
MDIO pins to simple GPIOs and bit-bang the MDIO bus. However, as
configuring the MACs depends on the hardware PHY polling, that
bit-banged bus could then no longer be used to connect Ethernet PHYs
used for switch ports of that SoC.

That being said, many of those SoCs also got an auxilary MDIO bus
(which require another, completely different and much more simple
driver[1]) to connect things (such as RealTek's own LED/GPIO
controllers) which aren't directly mapped to a switch port.

Obviously all existing hardware (ie. low-cost switches) using those
SoCs obey to those restrictions.

(plus: there simply aren't any PCIe or USB busses; it's an Ethernet
switch; the CPU is a low-end MIPS and itself hardly capably of moving
more than a few megabits, it's meant to manage and monitor the
Ethernet switch fabric doing all the actual work)

[1]: https://git.openwrt.org/openwrt/openwrt/tree/target/linux/realtek/patches-6.18/723-net-mdio-Add-Realtek-Otto-auxiliary-controller.patch