On Tue, Sep 02, 2025 at 12:05:45PM +0200, Neil Armstrong wrote:
> On 02/09/2025 11:46, Dmitry Baryshkov wrote:
> > On Tue, Sep 02, 2025 at 11:35:25AM +0200, Neil Armstrong wrote:
> > > On 02/09/2025 11:30, Dmitry Baryshkov wrote:
> > > > On Tue, Sep 02, 2025 at 11:00:30AM +0200, Neil Armstrong wrote:
> > > > > The QMP USB3/DP Combo PHY hosts an USB3 phy and a DP PHY on top
> > > > > of a combo glue to route either lanes to the 4 shared physical lanes.
> > > > > 
> > > > > The routing of the lanes can be:
> > > > > - 2 DP + 2 USB3
> > > > > - 4 DP
> > > > > - 2 USB3
> > > > > 
> > > > > The layout of the lanes was designed to be mapped and swapped
> > > > > related to the USB-C Power Delivery negociation, so it supports
> > > > > a finite set of mappings inherited by the USB-C Altmode layouts.
> > > > > 
> > > > > Nevertheless those QMP Comby PHY can be statically used to
> > > > > drive a DisplayPort connector, DP->HDMI bridge, USB3 A Connector,
> > > > > etc... without an USB-C connector and no PD events.
> > > > > 
> > > > > Add a property that documents the static lanes mapping to
> > > > > each underlying PHY to allow supporting boards directly
> > > > > connecting USB3 and DisplayPort lanes to the QMP Combo
> > > > > lanes.
> > > > > 
> > > > > Signed-off-by: Neil Armstrong <[email protected]>
> > > > > ---
> > > > >    .../phy/qcom,sc8280xp-qmp-usb43dp-phy.yaml         | 29 
> > > > > ++++++++++++++++++++++
> > > > >    1 file changed, 29 insertions(+)
> > > > > 
> > > > > diff --git 
> > > > > a/Documentation/devicetree/bindings/phy/qcom,sc8280xp-qmp-usb43dp-phy.yaml
> > > > >  
> > > > > b/Documentation/devicetree/bindings/phy/qcom,sc8280xp-qmp-usb43dp-phy.yaml
> > > > > index 
> > > > > c8bc512df08b5694c8599f475de78679a4438449..12511a462bc6245e0b82726d053d8605148c5047
> > > > >  100644
> > > > > --- 
> > > > > a/Documentation/devicetree/bindings/phy/qcom,sc8280xp-qmp-usb43dp-phy.yaml
> > > > > +++ 
> > > > > b/Documentation/devicetree/bindings/phy/qcom,sc8280xp-qmp-usb43dp-phy.yaml
> > > > > @@ -76,6 +76,35 @@ properties:
> > > > >      mode-switch: true
> > > > >      orientation-switch: true
> > > > > +  qcom,static-lanes-mapping:
> > > > > +    $ref: /schemas/types.yaml#/definitions/uint32-array
> > > > > +    minItems: 4
> > > > > +    items:
> > > > > +      enum:
> > > > > +        - 0 # Unconnected (PHY_NONE)
> > > > > +        - 4 # USB3 (PHY_TYPE_USB3)
> > > > > +        - 6 # DisplayPort (PHY_TYPE_DP)
> > > > > +    description:
> > > > > +      Describes the static mapping of the Combo PHY lanes, when not 
> > > > > used
> > > > > +      a in a Type-C dynamic setup using USB-C PD Events to change 
> > > > > the mapping.
> > > > > +      The 4 lanes can either routed to the underlying DP PHY or the 
> > > > > USB3 PHY.
> > > > > +      Only 2 of the lanes can be connected to the USB3 PHY, but the 
> > > > > 4 lanes can
> > > > > +      be connected to the DP PHY.
> > > > 
> > > > It feels like this significantly duplicates existing data-lanes
> > > > definitions. Can we use that property to express the same semantics?
> > > 
> > > Well yes it has the same semantics, but not really the same meaning. 
> > > data-lanes is designed
> > > to describes the lanes layout/ordering, not the type/mapping.
> > > 
> > > Here, we do not describe the ordering, i.e which source lane is connected 
> > > to which endpoint splot,
> > > but which lane is supposed to connect to which internal PHY.
> > > 
> > > Anyway, I'm open to suggestions.
> > 
> > phy@abcdef {
> >     ports {
> >             port@1 {
> >                     endpoint {
> >                             remote-endpoint = <&&usb_1_dwc3_ss>;
> >                             data-lanes = <2 3>;
> >                     };
> >             };
> > 
> >             port@2 {
> >                     endpoint {
> >                             remote-endpoint = <&mdss_dp0_out>;
> >                             data-lanes = <1>;
> >                     };
> >             };
> >     };
> > };
> > 
> > phy@cafecafe {
> >     ports {
> >             port@1 {
> >                     endpoint {
> >                             remote-endpoint = <&&usb_1_dwc3_ss>;
> >                             status = "disabled";
> >                     };
> >             };
> > 
> >             port@2 {
> >                     endpoint {
> >                             remote-endpoint = <&mdss_dp0_out>;
> >                             data-lanes = <2 3 0 1>;
> >                     };
> >             };
> >     };
> > };
> 
> This is wrong, those are the internal connections to the controllers,
> those are fixed. I'm speaking about the external lanes, but there's only
> a single port.
> 
> So, following your suggestion, we should use the Output port 0, but as it's
> only a single port it would need to have 2 endpoints, one for USB3 and one for
> DP.
> 
> For example:
> 
> \{
>       dp-connector {
>               compatible = "dp-connector";
>               type = "a";
> 
>               port {
>                       dp_con: endpoint {
>                               remote-endpoint = <&usb_1_ss2_qmpphy_dp_out>;
>                       };
>               };
>       };
> 
>       usb-a-connector {
>               compatible = "usb-a-connector";
> 
>               ports {
>                       #address-cells = <1>;
>                       #size-cells = <0>;
> 
>                       port@0 {
>                               reg = <0>;
>                               usb_con_hs: endpoint {
>                                       remote-endpoint = <&usb_1_ss2_dwc3_hs>;
>                               };
>                       };
> 
>                       port@1 {
>                               reg = <1>;
>                               usb_con_ss: endpoint {
>                                       remote-endpoint = 
> <&usb_1_ss2_qmpphy_usb3_out>;
>                               };
>                       };
>               };
>       };
> 
> };
> 
> &usb_1_ss2_dwc3_hs {
>       remote-endpoint = <&usb_1_ss2_dwc3_hs>;
> };
> 
> &usb_1_ss2_qmpphy {
>       /delete-property/ mode-switch;
>       /delete-property/ orientation-switch;
> 
>       ports {
>               
>               port@0{
>                       #address-cells = <1>;
>                       #size-cells = <0>;
> 
>                       /delete-node/ endpoint;
> 
>                       usb_1_ss2_qmpphy_usb3_out: endpoint@0 {
>                               reg = <0>;
>                               
>                               remote-endpoint = <&usb_con_ss>;
> 
>                               data-lanes = <1 2 0 0>;
>                       };
> 
>                       usb_1_ss2_qmpphy_dp_out: endpoint@1 {
>                               reg = <1>;
>                               
>                               remote-endpoint = <&dp_con>;
> 
>                               data-lanes = <0 0 1 2>;
>                       };
>               };
>       };
> };
> 
> So the driver logic would need to look at the port0/endpoint0 and 
> port0/endpoint1
> data-lanes to figure out the mode.
> 
> Is this what you were thinking ?

No, I was really thinking about the data-lanes in the PHY parts, so I
was incorrect there (which is incrrect as you've pointed out).

The endpoints approach looks interesting though.

-- 
With best wishes
Dmitry

Reply via email to