[DjDeveloperr]

J714s (MacBook Pro 14" M5 Pro): describe the left-front USB-C port's (ATC1) USB4/Thunderbolt router and its PCIe tunnel, as the right port (ATC2) already is.

Before this, ATC1 had only its PD controller, USB 2/3 and DisplayPort alt mode. A USB4 device on it got a PD contract and the PHY switched to "USB3 PIPE on the USB4 tunnel", but with no router nothing trained the link. There was no tunnel, no USB 2/3 fallback and no device.

What changes

• New t6050-j714s-left-usb4.dtsi: acio1 (router, CPU mailbox, DART, NHI) and pcie-tunnel-1 (apciec1 and its DART), plus the ATC1 power domains. Addresses, interrupts, tunables and the host DROM come from the J714sAP ADT (/arm-io/acio1, acio-cpu1, dart-acio1, apciec1, dart-apciec1, atc-phy1).
  • The router, DART and tunnel tunables are identical to ATC2's in that ADT.
  • The PHY's CIO RX_TOP lane tunables are this port's own and differ from ATC2's.
  • The nodes are laid out like J716c's ATC1 (t6050-j716c-left-usb4.dtsi) and the right port's (t6050-max-atc2.dtsi).
• hpm1's connector gets port@3 → acio1, as hpm2's does.
• The router's LSTX pin (AOP GPIO 79) is exposed next to ATC2's (81).
• t6050-peripherals.dtsi: the inventory's disabled apciec1 DART is left out of J714s, as it already is for J716c, because the tunnel now defines it.

No driver changes. Related: #1579 adds disabled acio0/1/3 routers to shared files. It doesn't add the PCIe tunnels, and this PR doesn't depend on it.

Tested on J714s (one lab boot, 2026-11-13). Default kernel 7.3.0-m5-intg6 (integration/aurora eb6b241) with this DT on top of intg6's DT; the dtsi is byte-identical to this PR's. Attached was a Satechi USB4 NVMe SSD Pro Enclosure (ASMedia ASM2464PD) holding the owner's data, so the check was read-only by design.

• ATC1's PHY entered CIO mode, the router firmware came up, and both lanes trained and entered CL0. Thunderbolt sysfs reports device 1-1 "USB4 NVMe SSD Pro Enclosure" at 20 Gb/s × 2 lanes in both directions, authorized.
• The PCIe tunnel came up: cold init done and "Link up on /pcie-tunnel-1/pcie@1c60000000" at 7.90 s. Behind the root port are the ASM2464PD upstream and downstream ports and the NVMe controller (INNOGRIT IG5216) at 0002:03:00.0, linked at 8 GT/s x4 inside the enclosure.
• nvme1 ("NVME SSD 512GB", transport pcie) enumerated at 8.94 s with its partition table (6 partitions). macOS reports the same disk as an external PCI-Express NVMe disk on this port.
• The right port's dock (TB5, domain 0) came up as usual in the same boot. The ATC1 router is domain 1.
• The default DT is unchanged for every other T6050 board: all 11 T6050 DTBs were rebuilt, only J714s's differs, and dtc reports no new warnings. checkpatch: 0 errors (2 warnings: MAINTAINERS, and the From/Signed-off-by name).

Not tested

• Data transfer or throughput: the only device at hand is the owner's disk, which stayed read-only and unmounted.
• Hot-unplug/replug and suspend/resume on this port.
• DisplayPort over USB4 on this port: the router's DP-IN interrupts and the dcpext route to acio1 aren't described here, and no USB4 display was connected.

🤖 Generated with Claude Code