Summary

Add an in-tree driver for the Fairchild (now ON Semiconductor)
FUSB302B USB Type-C port controller. The chip is an i2c slave
that implements CC1/CC2 sensing, BMC PHY for USB Power Delivery
messages, DRP toggle for role detection, and a VBUS measure ADC.
It is the Type-C controller used on the RockPro64, Pinebook Pro,
ROCK Pi 4, Khadas Edge, and several Chromebook designs around the
RK3399 era.

The driver supplies the struct usbc_tcpc_ops vtable consumed by
the usbc(4) policy framework (D3, #2211). All PD message
framing, policy state, VDM negotiation, and DisplayPort altmode
logic lives in usbc; this driver is the chip-specific I2C /
interrupt / FIFO half.

Functional coverage:

• SRC + SNK port roles, with DRP toggle for autonomous role
  discovery on cold plug.
• USB PD R3.2 messaging via the BMC TX/RX FIFOs (SOP only;
  SOP'/SOP'' cable messages scaffolded but not exercised).
• Spec rev autoselect with PD 2.0 / PD 3.0 fallback (sysctl-
  selectable for compatibility testing against picky sinks).
• Soft-reset and hard-reset paths.
• Polled DRP retoggle on lost partner -- recovers from sinks
  that momentarily drop CC during their own re-orientation.
• IRQ-driven; uses sx(9) for the soft state and never holds
  a spinlock across an I2C transaction (FUSB302's transfers can
  sleep on bus arbitration).
• Debug output gated behind dev.fusb302.<N>.debug, default 0.

Modeled functionally on the Linux
drivers/usb/typec/tcpm/fusb302.c driver but is a clean
reimplementation -- no code copied from Linux. Linux is the
reference for register sequencing only; the state machine,
locking discipline, and tunables are FreeBSD-native.

Includes a Rockchip-specific helper at
sys/arm64/rockchip/rk3399_fusb302_helper.c (built as a
separate module) that bridges FUSB302's role/orientation
detection to the RK3399 Type-C PHY mode-switch. This file is the
only piece in the review that is Rockchip-specific; the
fusb302(4) driver itself runs on any platform that exposes an
i2c bus and a level-sensitive GPIO for INT_N.

Dependencies

Depends on D3 (usbc framework, #2211) being committed first.
The vtable consumed (struct usbc_tcpc_ops, usbc_pd_policy_*
entry points) is defined there. Build will fail without D3
applied.

What's deliberately NOT in this review

• The rk_typec_phy driver itself -- D5 (rk_typec_phy: add USB-C DisplayPort altmode bring-up for RK3399 #2225).
• The Cadence MHDP DisplayPort framer -- D6 (forthcoming).
• Userspace control hooks (tcpmctl, sysctl-driven role
  override beyond debug knobs) -- reserved for a follow-up.
• Programmable-power-supply or fast-role-swap support.

Test plan

• Pine64 RockPro64 v2.1 + XYM W156F1 USB-C portable monitor:
  cold boot completes a full PD source contract (SRC_CAP ->
  REQUEST -> ACCEPT -> PS_RDY) and a DP altmode entry within
  ~1.5 s of probe attach. Verified across 20+ plug/unplug
  cycles; no leaked state, no IRQ storm (cf. D1, rk_gpio: implement PIC masking methods and mask unhandled IRQs #2197, the
  rk_gpio PIC fix that makes this even reachable on cold boot).
• Same board, USB-A type-C cable (no DP altmode advertised):
  PD source contract still completes; altmode entry is
  correctly declined; downstream rk_typec_phy stays in USB
  mode.
• Spec-rev forced via dev.fusb302.0.pd_spec_rev=20 (PD 2.0
  only) on three different PD sinks (XYM W156F1, Anker 535
  dock, a counterfeit Apple-pattern cable): all complete
  contract without falling back further. pd_spec_rev=30
  (PD 3.0) also completes on the W156F1 and the Anker but is
  rejected by the counterfeit cable, which is expected.
• Hard-reset recovery: forced via sysctl
  dev.fusb302.0.force_hard_reset=1 mid-contract. Driver
  returns to USBC_PD_SRC_STARTUP and renegotiates within
  300 ms.
• kldload fusb302; kldload rk3399_fusb302_helper; kldunload rk3399_fusb302_helper; kldunload fusb302 -- no leaks, no
  interrupts left enabled at the controller, no spurious GPIO
  edges observed under dtrace.

Size note

At ~4.5k lines this is on the larger side for a single
Differential. If reviewers prefer it split, the natural cut is:

• D4a: fusb302.c + fusb302_var.h +
  sys/modules/fusb302/Makefile (the platform-independent
  driver, ~4.4k lines).
• D4b: rk3399_fusb302_helper.c + its Makefile (Rockchip-
  specific extcon/typec-phy glue, ~140 lines).

I left it as one review because D4b is functionally meaningless
without D4a, and reviewing the helper in isolation is harder
than reviewing it as a small section of the same patch. Happy
to split on request.

Suggested reviewers

bsdimp, emaste, wulf7 (recent USB / iicbus committers per
git log); evadot, mhorne (sys/arm64/rockchip).