usbc: USB-PD R3.2 policy SM + TCPC abstraction

[b1nc0d3x]

Summary

Add a new in-tree framework for USB Power Delivery state-machine logic that is independent of any specific Type-C port controller chip. The framework lives at sys/dev/iicbus/usb/usbc/, builds as a kernel module (usbc.ko), and exposes four layers:

1. usbc_pd.h — base types, enums, and timing constants from USB PD R3.2: SOP* packet types, control/data message numbers, header bit layout, PDO/RDO field encoders, protocol timers (tSenderResponse, tPSTransition, ...), VDM SVID / object position constants for DisplayPort altmode.

2. usbc_pd_msg.{h,c} — hardware-independent build/parse for PD message headers and data objects. No I/O, no locking; callable from interrupt context:

   • usbc_pd_build_header() — pack header fields to a u16
   • usbc_pd_parse_header() — inverse, with bounds checking
   • usbc_pd_build_source_cap_pdo_fixed() — 5V fixed-supply PDO
   • usbc_pd_build_request_rdo() — fixed-supply request
   • usbc_pd_validate_pdo() — sanity-check before TX

3. usbc_pd_policy.{h,c} — the protocol/policy engine. Implements the source-port policy state machine from PD chapter 8 (Power Negotiation), the soft-reset and hard-reset paths, the VDM exchange for Discover Identity / Discover SVIDs / Enter Mode / DP_Status, and the timer-driven retry/abort logic. Caller-facing API:

   • usbc_pd_policy_attach(sc, tcpc_ops)
   • usbc_pd_policy_detach(sc)
   • usbc_pd_policy_rx(sc, msg) — chip-driver IRQ handler entry
   • usbc_pd_policy_role_change(sc, ...)
   • usbc_pd_policy_enter_dp_altmode(sc, port_pin_assign)

4. usbc_tcpc.h — the abstraction the chip driver implements. A struct of function pointers (transmit a PD msg, set CC state, set vbus, set role, etc.) that the policy SM invokes. This is what makes the framework hardware-independent — the FUSB302B driver in the companion (forthcoming) review supplies one implementation, and a future TUSB320 / TPS6598x / MT6360 driver supplies another without changing a line of policy code.

Why land this separately

• Dependency-free against the rest of the tree — no other in-tree driver consumes it yet.
• Reviewing the framework in isolation from any specific chip driver makes the abstraction boundary clearer.
• The companion fusb302 review (forthcoming) is materially smaller once it can refer to "the policy SM" and "the TCPC interface" instead of containing both itself.

Not in this review

• No chip driver. The fusb302(4) driver will be a separate, follow-on PR.
• No userspace policy hook — the SM is currently autonomous; the intent is that a future tcpmctl(8) or libusbc consumer can drive policy choices from userspace, but that is out of scope here.
• No PD R3.2 fast-role-swap or programmable-power-supply support. Source PDOs are limited to fixed-supply 5 V; extended-message handling is reserved for a follow-up.

Test plan

• Buildkernel + GENERIC + device usbc: builds clean, no external symbol references, kldload / kldunload work.
• Unit-level coverage: usbc_pd_build_header / parse_header round-trip 1000 random headers via a kgdb script; all field values preserved. PDO encoders verified by hand-decoding the resulting u32 against PD 6.4.1 Table 6-9.
• Integration: with the companion fusb302 driver (forthcoming) attached to a Pine64 RockPro64 v2.1 + XYM W156F1 USB-C portable monitor, the policy SM drives a full source contract (PD_REQUEST → ACCEPT → PS_RDY) and a DisplayPort altmode entry (Discover Identity → Discover SVIDs → Enter Mode → DP_Status) within ~1 s of plug insertion. Subsequent unplug / replug cycles re-run the entire flow without leaking state.
• Negative paths: forced PD soft-reset (sysctl) and hard-reset (peer TX fault) both return the SM to USBC_PD_SRC_STARTUP within tSenderResponseTimer + 50 ms.

Suggested reviewers

hps (USB stack), kib, manu, mhorne — based on git log of sys/dev/iicbus/usb/ and related arm64 work. Not @-mentioned to avoid spam; happy to ping on request.

Diff scope

File	Lines
sys/dev/iicbus/usb/usbc/usbc_pd.h	+179
sys/dev/iicbus/usb/usbc/usbc_pd_msg.c	+211
sys/dev/iicbus/usb/usbc/usbc_pd_msg.h	+174
sys/dev/iicbus/usb/usbc/usbc_pd_policy.c	+1992
sys/dev/iicbus/usb/usbc/usbc_pd_policy.h	+225
sys/dev/iicbus/usb/usbc/usbc_tcpc.h	+128
sys/modules/usbc/Makefile	+10
Total	+2919

All files net-new; no existing FreeBSD code modified. License is BSD-2-Clause throughout; SPDX-License-Identifier headers present on every file.

Notes

Third in a series landing the RockPro64 USB-C DisplayPort + audio path. Predecessors:

• rk_gpio: implement PIC masking methods and mask unhandled IRQs #2197 — rk_gpio: implement PIC masking methods and mask unhandled IRQs
• drm_fb_helper: re-probe on late EDID, don't clamp to fallback fb size #2198 — drm_fb_helper: re-probe on late EDID, don't clamp to fallback fb size

Successor patches (fusb302 chip driver, rk_typec_phy DP altmode, rk_cdn_dp framer, DP audio + hdmi codec stub) will follow as their own PRs, paced to avoid swamping reviewer attention.

Author: Kyle Crenshaw <B1nc0d3x@gmail.com>

[bms]

There's the slight problem of hps having passed away many years ago. I had the pleasure of meeting him in Warsaw over 2012.

[b1nc0d3x]

Thank you for the correction - I wasn’t aware of Hans Petter’s passing, and I’m genuinely sorry. I had no intention of being disrespectful and from here on I’ll cross-check before requesting specific reviewers…Thank You,Kyle CrenshawOn May 19, 2026, at 6:15 AM, Bruce M Simpson ***@***.***> wrote:bms left a comment (freebsd/freebsd-src#2211) There's the slight problem of hps having passed away many years ago. I had the pleasure of meeting him in Warsaw over 2012. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>