[RFC PATCH v3 0/2] USB: describe two connectors sharing one DWC3 HS data bus

[Qianhui Xu]

This is v3 of the series describing two USB-C receptacles that share one
DWC3 High Speed data bus.  It is deliberately an RFC for the shape: the
driver part is not included yet.

Changes since v2:
- Drop port@2.  The shared HS bus is now port@0 with two numbered
  endpoints, endpoint@0 and endpoint@1, so "ports" keeps describing the
  controller's own buses.
- Require "ports" when port-select-gpios is present, and reject the
  endpoint properties when it is not.
- Constrain the two mux-state values to be complementary.
- Move the per-connector role-switch lookup into a generic helper,
  usb_role_switch_at_endpoint() (new patch 2/2).  v2 relied on the
  lookup walking from the endpoint up to the parent device, which is not
  what usb_role_switch_get() does upstream, so a consumer would never
  find the switch.  With the helper the switch is registered against the
  remote endpoint and found from the connector's endpoint.
- Do not describe driver switching policy in the binding.

Two questions for you:

1. Why not the standard mux bindings?
   The GPIO only routes the HS D+/D- pair.  What the driver needs is role
   arbitration between two connectors (the oldest active request owns the
   data path, and a connector reports its role only while it owns the
   path).  The mux subsystem can select a state but has no notion of
   ownership or of USB roles, and drivers/usb/ has no mux_control
   consumer today.  gpio-mux also drives its mux-gpio to the idle state
   at probe time, which would override the level left by the bootloader
   on this board.  So the arbitration stays in the controller and the
   binding only describes the wiring.  If you would still prefer the mux
   bindings we will rework it -- that is why this is RFC.

2. Is "usb-role-switch" on a graph endpoint acceptable, given that the
   property normally means "this device registers a role switch"?  Here
   it declares that this connector endpoint has its own role switch, and
   the new helper looks it up from the connector's endpoint.  A separate
   property name is possible if you prefer.

The board is the same as in v2: two USB-C receptacles, Super Speed pairs
independent, only the HS D+/D- pair goes through a 2:1 mux controlled by
TLMM GPIO 29; both connectors have data-role = "dual" and are handled by
pmic-glink.

Once the shape is acked I will send the DWC3 driver part as v4, including
the fixes from the review of v2.

Qianhui Xu (2):
  dt-bindings: usb: snps,dwc3: describe two connectors sharing one HS
    bus
  usb: roles: look up role switches attached to a graph endpoint

 .../bindings/usb/snps,dwc3-common.yaml        | 109 +++++++++++++++++-
 drivers/usb/roles/class.c                     |  40 +++++++
 2 files changed, 148 insertions(+), 1 deletion(-)

-- 
2.53.0