Re: [PATCH 04/11] dt-bindings: usb: Add NVIDIA LSTP USB adapter

[Adrian Ambrozewicz]

On Tue, Oct 13, 2026 at 08:17:18AM +0200, Krzysztof Kozlowski wrote:
> No, compatible, not presence of properties defines whether device has
> SPI or not.
[...]
> What's more, your top level compatible EXACTLY defines all
> this, thus none of your properties are needed as they are implied by the
> top-level compatible.

It cannot here: firmware decides the channel set, so one VID:PID
carries different channels per board, several of one type.

On folding into the parent [1]: one device can have several I2C buses
as well as SPI and GPIO channels, each with its own devices, line names
and hogs, and other nodes reference the GPIO channels by phandle. Each
needs a node.

Is a node per I2C, SPI and GPIO channel acceptable? If not, how would
you describe them? The interface node I'll answer separately. Why DT
at all is in my reply to Greg [2], the other open questions in the
RFC [3].

[1] https://lore.kernel.org/r/58cbdffa-7574-4daa-851b-dbb50012f158@xxxxxxxxxx
[2] https://lore.kernel.org/r/20261016084443.1972351-1-aambrozewicz@xxxxxxxxxx
[3] https://lore.kernel.org/r/20261012-lstp-dt-rfc-v1-aambrozewicz@xxxxxxxxxx

Thanks,
Adrian