Re: [PATCH 1/4] dt-bindings: usb: generic-ehci: Allow describing the root hub ports
From: Chen-Yu Tsai
Date: Thu Nov 12 2026 - 03:16:06 EST
On Thu, Nov 12, 2026 at 5:00 PM Arun Iyengar <aiyengar1990@xxxxxxxxx> wrote:
>
> Hi Chen-Yu,
>
> On Wed, Nov 11, 2026 at 02:35 UTC, Chen-Yu Tsai wrote:
> > Would this patch covering all HCDs be better?
> >
> > https://lore.kernel.org/all/20261014082037.1774672-2-wenst@xxxxxxxxxxxx/
> >
> > Also see Rob's comment about covering the port numbers.
>
> Yes, yours is better: it covers every host controller, and
> generic-ehci.yaml gets it through usb-hcd.yaml. Rob's point about the
> port numbers applies to my patch too.
>
> v2 drops this patch and builds on your patch 1 instead. Only the
> Orange Pi 5/5B DT patch, which now goes to the Rockchip tree on its
> own, needs it; that DT passes dtbs_check with your patch applied, and
> also with 'reg' limited to 1 and up as Rob suggested.
>
> The v2 cover has one question for you: patch 1 there looks the graph
> ports up by root hub port number, while you have said root hub graph
> ports should use the raw xHCI port number. Which one do you want?
I don't see a v2 yet. So that's patch 2 of v1.
It should definitely use the raw port number, since that is what is
supposed to be used in DT (and what is used in ACPI).
FWIW I'm also pushing a series to wire up USB on M.2 connectors [1].
That also has some code that does OF graph lookup. I still need to
push a v10. And I have a few other patches that fix up port lookup
in other bits of USB OF code.
ChenYu
[1] https://lore.kernel.org/all/20260916075745.3549953-1-wenst@xxxxxxxxxxxx/
> Thanks,
> Arun
>
> On Tue, Nov 10, 2026 at 6:35 PM Chen-Yu Tsai <wens@xxxxxxxxxx> wrote:
> >
> > On Tue, Nov 10, 2026 at 4:20 PM Arun Iyengar via B4 Relay
> > <devnull+aiyengar1990.gmail.com@xxxxxxxxxx> wrote:
> > >
> > > From: Arun Iyengar <aiyengar1990@xxxxxxxxx>
> > >
> > > The USB 2.0 lines and the SuperSpeed lanes of a USB 3 connector don't
> > > always reach the same host controller. On the Rockchip RK3588, the
> > > SuperSpeed-only USB 3.0 host 2 (xHCI) is paired on the board with one
> > > of the separate USB 2.0 hosts (EHCI/OHCI). Only the device tree knows
> > > which EHCI root hub port shares a connector with which SuperSpeed port.
> > >
> > > The USB core already reads a hub's OF graph, with one graph port per
> > > downstream port, and uses it for the port connect type. Allow the same
> > > "ports" node on EHCI controllers, so that a root hub port can be linked
> > > to its usb-connector node.
> > >
> > > Signed-off-by: Arun Iyengar <aiyengar1990@xxxxxxxxx>
> > > ---
> > > Documentation/devicetree/bindings/usb/generic-ehci.yaml | 13 +++++++++++++
> > > 1 file changed, 13 insertions(+)
> > >
> > > diff --git a/Documentation/devicetree/bindings/usb/generic-ehci.yaml b/Documentation/devicetree/bindings/usb/generic-ehci.yaml
> > > index 908b02d9175e..9dcba539d60d 100644
> > > --- a/Documentation/devicetree/bindings/usb/generic-ehci.yaml
> > > +++ b/Documentation/devicetree/bindings/usb/generic-ehci.yaml
> > > @@ -160,6 +160,19 @@ properties:
> > > - host
> > > - otg
> > >
> > > + ports:
> > > + $ref: /schemas/graph.yaml#/properties/ports
> > > + description:
> > > + The root hub's downstream facing ports, linked to the connectors they
> > > + are wired to. Needed when a connector's SuperSpeed lanes go to a
> > > + different host controller, which then pairs with the root hub port
> > > + through the connector.
> > > +
> > > + patternProperties:
> > > + "^port@[1-9a-f][0-9a-f]*$":
> > > + $ref: /schemas/graph.yaml#/properties/port
> > > + description: Root hub port, numbered from 1.
> > > +
> >
> > Would this patch covering all HCDs be better?
> >
> > https://lore.kernel.org/all/20261014082037.1774672-2-wenst@xxxxxxxxxxxx/
> >
> > Also see Rob's comment about covering the port numbers.
> >
> > ChenYu
> >
> > > required:
> > > - compatible
> > > - reg
> > >
> > > --
> > > 2.55.0
> > >
> > >
> > >