Re: [PATCH v2 1/4] dt-bindings: connector: pcie-m2-e: Add vendor LGA connector compatible

[Loic Poulain]

On Fri, Jul 3, 2026 at 9:11 AM Krzysztof Kozlowski <krzk@xxxxxxxxxx> wrote:
>
> On Thu, Jul 02, 2026 at 12:46:13PM +0200, Loic Poulain wrote:
> > Some modules (e.g. the Qualcomm QCA2066/QCNFA765) expose the same M.2
> > Key E interface signals but are physically soldered as an LGA package
>
> Why do we need device nodes for soldered connectors? Soldered stuff is
> afixed one, not connectable connector, thus it has no point in DT
> representation.

The LGA module is soldered, yes, but the node here isn't really
modeling a mateable/removable connector, it's the DT description of
the M.2 Key E electrical interface that the module exposes: the
W_DISABLE1 and #/W_DISABLE2# enable signals, the shared PCIe/UART
regulators, and the graph routing PCIe->WiFi and UART->BT. The
pwrseq-pcie-m2 driver and the graph binding rely on exactly this
description to sequence PCIe and UART bring-up, and that's independent
of whether the module is socketed or soldered.

Electrically the QCA2066 LGA on that  part is an M.2 Key E design,
only the physical attachment differs. That's what the vendor-specific
qcom,pcie-m2-1418-lga-connector compatible is meant to capture, the
soldered/LGA variant of that interface. So reusing the connector
binding is a pragmatic way to model and drive it.

It also keeps the enumerable-slot semantics: since M.2 devices are
discovered via PCIe IDs, a board variant with a different chipset
soldered to the same LGA grid would be handled automatically, without
DT changes, exactly as it would be for a physical M.2 slot.

If the concern is purely the "connector" naming for a soldered part, I
can reframe the binding.

Regards,
Loic