[funkzy]

Summary

The T6050 combo PHY's USB2 PHY and digital banks sit in the USB controller's power domain. Only a bound USB
controller holds it. On a port with no controller (on the M5 Ultra Mac Studio, J775D, with Linux on the hardware,
the rear Type-C ports drive DisplayPort with no USB controller bound), the first Type-C mode switch wrote an
unclocked bank and took an SError. A combo PHY node may now name that domain fourth, as usb, and the PHY then
holds it while it runs.

What it changes

• Binding: power-domains and power-domain-names take three or four entries; the fourth is usb. The fixed
  DisplayPort PHY (#phy-cells = <0>) keeps exactly three.
• Driver: probe accepts three domains, or four on a combo PHY, and attaches each one named. combo_start() takes
  the USB domain after the common one, before the USB2 sequence; combo_stop() drops it before the common one.
• A combo node that names three domains probes and runs exactly as before.

Why

Without it, a combo port with no USB controller cannot switch modes on bare metal: the SError takes the machine
down. Holding the domain in the PHY makes the port independent of whether a controller is bound.

Testing (J775D, M5 Ultra)

The kernel: our J775D series (integration/aurora's lineage, as on the Studio every day) with this PR's code and
that of the other T6050 PRs sent with it in place of the series' own forms, booted once on 2026-11-02 and restarted
through the SMC at the end. The same code is in the series kernel the machine boots by default.

• The combo PHY of the port the display is on (ATC2, phy@383000000) bound with its four domains: usb-aon, common,
  wrapper and usb. No USB controller is bound on that port.
• At hotplug the DCP took that port's DPTX PHY and the link came up: Hyprland ran on it, 682 swaps in the first
  minute. The mode switch that took the SError before ran without one.
• No panic, BUG, Oops, WARNING or SError line.
• Builds on phy: apple: Add T6050 DisplayPort and USB combo PHY #480: 0 errors, 0 warnings in the kernel. checkpatch --strict: clean.

Not tested

• A three-domain combo node (the guest's DT) on this code: J775D's DT names four. The three-domain path is phy: apple: Add T6050 DisplayPort and USB combo PHY #480's,
  unchanged apart from the count check.
• A combo port with a USB controller bound and the fourth domain named (both then hold it; runtime PM counts both).

Dependencies

#480 (t6050/display-phy). J775D's board DT names the fourth domain; it comes later with the J775D board. The DCP
side is the T6050 bare-metal DCP PR on #485.

For dj

Does the guest bind a USB controller on every combo port? If it does, the guest never needs the fourth domain. Is an
optional fourth domain the shape you want, or would you rather the guest DT name it too?

[funkzy]

Closing: this change now lives in the funkzy/j775d branch of aurora-silicon/linux-aurora (the J775D series as built and booted, 681521644671), which dj merges directly; no PR flow for it any more.