[PATCH] dt-bindings: usb: switch: split out ports definition

[Dmitry Baryshkov]

On Tue, Sep 30, 2025 at 08:48:13PM +0200, Neil Armstrong wrote:
> On 9/30/25 20:43, Dmitry Baryshkov wrote:
> > On Tue, Sep 30, 2025 at 07:17:21PM +0200, Neil Armstrong wrote:
> > > The ports definition currently defined in the usb-switch.yaml
> > > fits standards devices which are either recipient of altmode
> > > muxing and orientation switching events or an element of the
> > > USB Super Speed data lanes.
> > > 
> > > This doesn't necessarely fit combo PHYs like the Qualcomm
> > > USB3/DP Combo which has a different ports representation.
> > > 
> > > Move the ports definition to a separate usb-switch-ports.yaml
> > > and reference it next to the usb-switch.yaml, except for
> > > the Qualcomm USB3/DP Combo PHY bindings.
> > 
> > Isn't it easier to make QMP PHY use $ref for port nodes instead of allOf
> > and keep ports definitions inside the usb-switch schema?
> 
> Rob asked to not do that... see https://lore.kernel.org/all/20250905175533.GA1000951-robh@kernel.org/

(Just for the sake of keeping it here too): your current patch allows
using mode-switch and retimer-switch for the fsl,imx95-usb-phy device,
which it should not (because it can only be an orientation-switch).
Likewise our Combo QMP PHYs should only be orientation and/or mux, but
not retimer. I am not sure how to solve it without actually allowing
properties to reference nodes from another schema.


-- 
With best wishes
Dmitry