Re: [PATCH v2 1/2] dt-bindings: usb: Introduce GPIO-based SBU mux

[Rob Herring <robh@kernel.org>]

On Thu, Jan 19, 2023 at 11:40 AM Bjorn Andersson
<quic_bjorande@quicinc.com> wrote:
>
> On Thu, Jan 19, 2023 at 10:11:32AM -0600, Rob Herring wrote:
> > On Wed, Jan 18, 2023 at 10:08:11AM -0800, Bjorn Andersson wrote:
> > > On Tue, Jan 17, 2023 at 11:56:57AM -0600, Rob Herring wrote:
> > > > On Thu, Jan 12, 2023 at 08:11:14PM -0800, Bjorn Andersson wrote:
> > > > > From: Bjorn Andersson <bjorn.andersson@linaro.org>
> > > > >
> > > > > Introduce a binding for GPIO-based mux hardware used for connecting,
> > > > > disconnecting and switching orientation of the SBU lines in USB Type-C
> > > > > applications.
> > > > >
> > > > > Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
> > > > > Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
> > > > > ---
> >
> >
> > > > > +    tcpm {
> > > > > +        connector {
> > > > > +            compatible = "usb-c-connector";
> > > > > +
> > > > > +            ports {
> > > > > +                #address-cells = <1>;
> > > > > +                #size-cells = <0>;
> > > > > +
> > > > > +                port@0 {
> > > > > +                    reg = <0>;
> > > > > +                    tcpm_hs_out: endpoint {
> > > > > +                        remote-endpoint = <&usb_hs_phy_in>;
> > > > > +                    };
> > > > > +                };
> > > > > +
> > > > > +                port@1 {
> > > > > +                    reg = <1>;
> > > > > +                    tcpm_ss_out: endpoint {
> > > > > +                        remote-endpoint = <&usb_ss_phy_in>;
> > > > > +                    };
> > > > > +                };
> > > > > +
> > > > > +                port@2 {
> > > > > +                    reg = <2>;
> > > > > +                    tcpm_sbu_out: endpoint {
> > > > > +                        remote-endpoint = <&sbu_mux_in>;
> > > > > +                    };
> > > > > +                };
> > > > > +            };
> > > > > +        };
> > > > > +    };
> > > > > +
> > > > > +    sbu-mux {
> > > > > +        compatible = "pericom,pi3usb102", "gpio-sbu-mux";
> > > > > +
> > > > > +        enable-gpios = <&tlmm 101 GPIO_ACTIVE_LOW>;
> > > > > +        select-gpios = <&tlmm 164 GPIO_ACTIVE_HIGH>;
> > > > > +
> > > > > +        mode-switch;
> > > > > +        orientation-switch;
> > > > > +
> > > > > +        port {
> > > > > +            sbu_mux_in: endpoint {
> > > > > +                remote-endpoint = <&tcpm_sbu_out>;
> > > > > +            };
> > > >
> > > > Don't you need a connection to whatever drives SBU? Maybe your case is
> > > > fixed because the phy does the DP/USB muxing? But the binding needs to
> > > > support the worst case which I guess would be all the muxing/switching
> > > > is done by separate board level components.
> > > >
> > >
> > > Perhaps I'm misunderstanding your request, but I think this is the worst
> > > case you're talking about.
> > >
> > > &usb_ss_phy_in is a reference to the PHY, which does switching/muxing of
> > > the SuperSpeed lanes in the connector, but the PHY provides no control
> > > over the SBU signals.
> > >
> > > So this sbu-mux is a separate component between the SBU-pads on the SoC
> > > and the usb-c-connector, referenced through he &sbu_mux_in reference.
> > >
> > >
> > > So upon e.g. a orientation switch, the typec_switch_set() call the tcpm
> > > implementation will request orientation switching from port@1 and port@2
> > > (no orientation-switch on port@0/HS pins).
> >
> > 'port@2' is supposed to define the connection to what controls SBU. The
> > mux here switches the signals, but it doesn't control them.
>
> The SBU signals are driven by the SS PHY, on behalf of the DisplayPort
> controller. These signals are  turned on/off as a result of the TCPM
> indicating the HPD state to the DisplayPort controller.
>
> There's a such not really a direct representation today of the entity
> that drives the SBU lines. It happens to be a sub-block in
> &usb_ss_phy_in, but I don't envision that we need/want any signaling
> between the TCPM and the SBU-"driver".
>
>
> I see that I missed that in the example above, your suggestion on how to
> model that relationship (TCPM - DP controller) was to add an additional
> endpoint in port@1. So that's the current design (but neither ports nor
> endpoints are significant from an implementation point of view).
>
> > The mux should sit in the middle, but the graph terminates at the mux.
> > You don't have a connection presumably because you know what the
> > connection.
>
> But do you suggest that the graph should reference the entity that
> drives the SBU signals?

Yes, that was the original intent.

> What about the discrete mux?

You mean the mux in this binding, right? That should be in the middle:

DPaux --> SBUmux --> connector

Maybe the SS phy is in there too.

>
> > Perhaps because there is only 1 connector and controller.
> >
>
> There is one SBU mux, one DP controller and one SS PHY per
> usb-c-connector.
>
> > Suppose you have 2 connectors and 2 controllers which drive SBU
> > signals. Also assume that the SBU signals are completely independent
> > from what's driving the altmode SS signals. How would you describe that?
> >
>
> This is the setup we have on e.g. SC8280XP CRD; where the TCPM has two
> usb-c-connectors defined, each with their graph referencing the SS PHY,
> DP controller and respective sbu-mux.
>
> There's an incomplete example of this published at [1] (where the SS phy
> isn't represented yet - and hence there's no control over the SS lanes,
> nor is the HS lanes connected to the dwc3 for role switching).
>
> Perhaps I'm misunderstanding your concerns though?

That looks like you can assume who drives SBU based on the DP
controller. Probably a safe assumption for DP (that DP-aux is part of
the DP controller), but I was more worried about if you can't assume
that relationship. Take HDMI for example where the DDC signals can
come from anywhere. They could be part of the HDMI bridge, a general
purpose I2C bus off the SoC, or bitbanged GPIOs. Though from what I've
read, HDMI Altmode is dead. I don't know if the need to describe the
SBU connection would apply to anything else.

I guess this all boils down to whether the SBU mux should have a 2nd
optional port as the input for what drives it.

Rob