[PATCH v4 0/3] usb: dwc3: support two connectors sharing one HS data bus

[Qianhui Xu]

This is a formal submission, not an RFC: it supersedes the two-patch v2
series (<20261017081320.220525-1-xiaqinghuan@xxxxxxxxxxxxxxxx>) and the
shape-only v3 RFC (<20261019131848.344689-1-xiaqinghuan@xxxxxxxxxxxxxxxx>),
and it contains the driver that uses the binding it describes.  Sorry for the confusion caused by sending v3 as an
RFC threaded under v2; the intent was to settle the binding shape first,
but that is not how it should have been posted.  The series is complete
below, with the disposition of every review comment listed.

Two connectors on one DWC3 controller
=====================================

The board (Lenovo Legion Y700 Gen4, SM8750) wires the Super Speed pairs of
two USB-C receptacles to separate PHYs, while the High Speed D+/D- pair is
shared through a 2:1 mux controlled by a single GPIO:

	usb@a600000 {
		port-select-gpios = <&tlmm 29 GPIO_ACTIVE_HIGH>;

		ports {
			#address-cells = <1>;
			#size-cells = <0>;

			port@0 {
				reg = <0>;
				#address-cells = <1>;
				#size-cells = <0>;

				endpoint@0 {
					reg = <0>;
					remote-endpoint = <&usb_0_dwc3_hs>;
					usb-role-switch;
					mux-state = <0>;
				};

				endpoint@1 {
					reg = <1>;
					remote-endpoint = <&usb_1_dwc3_hs>;
					usb-role-switch;
					mux-state = <1>;
				};
			};

			port@1 {
				/* Super Speed of connector 0, separate PHY */
			};

			port@2 {
				/* Super Speed of connector 1, separate PHY */
			};
		};
	};

Both connectors are data-role dual and hand their role to the SoC through
pmic-glink, so the controller has to arbitrate between two independent role
requests and move the shared data path before it is reinitialised.

Series layout
=============

1/3 documents the two-connector HS data path: "port-select-gpios" plus two
    numbered endpoints of "port@0" carrying "mux-state" (which GPIO level
    connects that connector) and "usb-role-switch" (this connector has its
    own role-switch input).

2/3 adds usb_role_switch_at_endpoint(), so a role switch registered against
    an OF graph endpoint can be found by the consumer that owns the other
    end of that link.  Without it the endpoint registration in 1/3 would
    have no lookup path.

3/3 implements the arbitration in DWC3: one role switch per connector, the
    oldest pending request wins, the GPIO is flipped inside __dwc3_set_mode()
    after the previous host/gadget instance is torn down, and a connector
    only reports a role while it owns the data path.

Disposition of the review comments
==================================

sashiko, binding (v2 1/2):

 - "model the board mux with the standard mux subsystem": not adopted.
   drivers/usb/ has no mux_control consumer, mux only selects a channel and
   cannot express "oldest requester wins" role arbitration, and gpio-mux
   drives its idle level at probe, which would override the selection left
   by the bootloader.  The mux is therefore described where it is used, as
   a GPIO plus per-endpoint level.

 - "port@2 models the second connector as a controller bus": fixed.
   The second connector is now endpoint@1 of port@0, so "ports" keeps
   describing the controller's own buses.

 - "usb-role-switch does not belong on a graph endpoint": kept, but the
   missing lookup path is fixed by 2/3 instead of relying on a downstream
   only helper.

 - "conditionals do not require ports": fixed, "ports" and "port@0" are
   required when "port-select-gpios" is present, and the endpoint
   properties are rejected when it is absent.

 - "two ports may carry the same mux-state": fixed, the values are
   constrained to be complementary.

sashiko, driver (v2 2/2):

 - Critical, "the init error path leaves drd_work queued": fixed.
   Both switches are now registered before the initial dwc3_set_mode(),
   and the error path cancels the queued role work.

 - High, "dwc3_set_mode() is called after dropping the lock": fixed.
   The request order and the resulting mode request are taken in one
   critical section; dwc3_set_mode() only queues work, so holding the mux
   lock across it is safe.

 - High, "an in-place role swap overwrites the request timestamp": fixed.
   A request ticket is only taken on a NONE -> role transition, so a
   connector that is already using the bus cannot steal it.

 - High, "registering on the endpoint breaks the lookup": addressed by 2/3.

 - Medium, "fwnode leak on the error path": fixed, the endpoint reference
   is dropped on every exit path.

 - Medium, "error and desired_role are never read": fixed, desired_role is
   gone and the sticky mux error is reported to the connector on its next
   role request.

 - Medium, "add dwc3_pre_set_role()": not adopted in this series; the role
   is set from the same place as the single-connector switch, and adding a
   second entry point would duplicate that path.  If you prefer the helper,
   I will restructure the series that way.

Testing
=======

Mechanical, re-run on v4:

 - checkpatch --strict: 0 errors, 0 warnings, 0 checks on all three patches.
 - dt-doc-validate on snps,dwc3-common.yaml: clean.
 - W=1 compiles clean for drivers/usb/dwc3/ and drivers/usb/roles/.
 - Full arm64 Image.gz + dtbs build of a v7.3-rc7 tree with the series
   applied.

Board behaviour: this board's daily kernel carries this series, and it
registers both per-connector switches, visible as
a600000.usb-port0-role-switch and a600000.usb-port1-role-switch.  In the
earlier revisions the mux logic was exercised plug-by-plug: host mode on
either receptacle, moving the cable from one receptacle to the other hands
the controller over, and a charging-only connector next to an active one
does not disturb it.

The v4 changes are the review fixes listed above: registering the switches
before the initial mode request, taking the request order and the mode
request in one critical section, ticket handling on NONE -> role only, the
monotonic counter instead of jiffies, and dropping the unused field.  Those
were reviewed and built but not re-exercised plug-by-plug, because the
author's test setup talks to this board through the very USB link being
switched; I would rather say that than claim a test I did not run.  I can
re-test on request if a maintainer wants the swap sequence confirmed before
this goes in.

Qianhui Xu (3):
  dt-bindings: usb: snps,dwc3: describe two connectors sharing one HS bus
  usb: roles: look up role switches attached to a graph endpoint
  usb: dwc3: support two connectors sharing one HS data bus


Qianhui Xu (3):
  dt-bindings: usb: snps,dwc3: describe two connectors sharing one HS
    bus
  usb: roles: look up role switches attached to a graph endpoint
  usb: dwc3: support two connectors sharing one HS data bus

 .../bindings/usb/snps,dwc3-common.yaml        | 109 ++++++++-
 drivers/usb/dwc3/core.c                       |  29 ++-
 drivers/usb/dwc3/core.h                       |  39 ++++
 drivers/usb/dwc3/debugfs.c                    |   4 +
 drivers/usb/dwc3/drd.c                        | 219 +++++++++++++++++-
 drivers/usb/roles/class.c                     |  40 ++++
 6 files changed, 437 insertions(+), 3 deletions(-)

-- 
2.53.0