Re: [PATCH RFC 1/3] arm64: dts: qcom: nord: Add Ethernet support

[Shawn Guo]

On Wed, Nov 04, 2026 at 10:04:00AM +0100, Konrad Dybcio wrote:
> On 11/4/26 9:42 AM, Shawn Guo wrote:
> > On Wed, Nov 04, 2026 at 09:04:33AM +0100, Konrad Dybcio wrote:
> >> On 11/4/26 8:59 AM, Shawn Guo wrote:
> >>> On Mon, Nov 02, 2026 at 04:01:03PM +0100, Konrad Dybcio wrote:
> >>>> On 11/1/26 9:04 AM, Shawn Guo wrote:
> >>>>> From: Mohd Ayaan Anwar <mohd.anwar@xxxxxxxxxxxxxxxx>
> >>>>>
> >>>>> Describe both Nord 10-Gbps Ethernet controllers, including their XPCS and
> >>>>> SerDes blocks, clocks, resets, power domains, interconnects, SMMU streams
> >>>>> and DMA queues.
> >>>>>
> >>>>> Signed-off-by: Mohd Ayaan Anwar <mohd.anwar@xxxxxxxxxxxxxxxx>
> >>>>> Signed-off-by: Shawn Guo <shengchao.guo@xxxxxxxxxxxxxxxx>
> >>>>> ---
> >>>>
> >>>> [...]
> >>>>
> >>>>> +	phys = <&serdes0>;
> >>>>> +	phy-names = "serdes";
> >>>>
> >>>> I believe this holds for both GearVM and embedded, so you can move
> >>>> it to the common DTSI
> >>>
> >>> No, this doesn't apply to GearVM. The SerDes is owned entirely by the SCMI
> >>> server firmware for GearVM variant. LDOs, clkref, reset, and the per-speed
> >>> PLL/RX/TX programming driven by an SCMI perf domain whose level is
> >>> the link speed.
> >>
> >> Right, but I meant the assignment to the EMACs specifically
> > 
> > The EMAC0/SerDes0 and EMAC1/SerDes1 pairing is indeed the same on both
> > variants. But the serdes nodes only exist in nord-embedded.dtsi, since
> > GearVM doesn't expose the PHY to Linux. Or am I missing your point
> > completely?
> 
> No, I was just completely sure that the nodes exist in the gvm one too
> because it seems like they do on lemans:
> 
> https://lore.kernel.org/linux-arm-msm/20261103-sa8255p-eth-phy-v18-1-b00cc2fa0404@xxxxxxxxxxxxxxxx/

Ah, right! We will likely have it for Nord GearVM variant as well,
so your suggestion makes sense. I will update.

Shawn