Re: [PATCH v7 7/7] arm64: dts: amlogic: t7: khadas-vim4: enable the ethernet port

From: Lucas Tanure

Date: Sat Nov 21 2026 - 16:26:34 EST


On Sun, Nov 15, 2026 at 6:08 PM Martin Blumenstingl
<martin.blumenstingl@xxxxxxxxxxxxxx> wrote:
>
> Hi Lucas,
>
> On Fri, Nov 13, 2026 at 1:19 PM Lucas Tanure <tanure@xxxxxxxxx> wrote:
> [...]
> > +&ext_mdio {
> > + external_phy: ethernet-phy@0 {
> > + /* Realtek RTL8211FD-CG */
> > + reg = <0>;
> We've been following this for a long time until someone found out that
> RTL8211F has a broadcast feature.
> Can you check the strapping registers (or simply see if the PHY
> answers on address 1 as well)?

Checked both. The schematic (v13a, page 9) straps PHYAD[2:0] to 001:
RXCTL/PHYAD2 and RXCLK/PHYAD1 have 4.7K to GND, RXD3/PHYAD0 has 4.7K
to EVDDIO. So the PHY sits at address 1.

A scan from the running kernel agrees. It reads PHY ID registers 2
and 3 at all 32 addresses through SIOCGMIIREG, which goes through the
mux to the external bus:

addr 0: 0x001c 0xc916
addr 1: 0x001c 0xc916
addr 2..31: 0xffff 0xffff

Address 1 is the strap. Address 0 is the RTL8211F broadcast address
(PHYAD0_EN in PHYCR1, on after reset). The vendor tree uses 0 and
only works because of that.

> I'd appreciate it if you could update this patch so we can use this
> approach as reference from now on.

v8 uses reg = <1>. The PHY attaches as mdio_mux-0.0:01, link at
1Gbps/Full, 941/944Mbit/s with iperf3, no error. The commit message
says where the address comes from and why 0 also worked.

Thanks,
Lucas

>
>
> Best regards,
> Martin
>
>
> [0] https://lore.kernel.org/all/bc8c7c6a-5d5f-4f7c-a1e2-e10a6a82d50e@xxxxxxx/