What macOS does to bring up USB4 on the T6034 #5

Activity

1. hungbenjamin402

   added

   wayfinder:researchwayfinder researchwayfinder research

   on Nov 9, 2026
2. hungbenjamin402

   added a parent issue on Nov 9, 2026

   • USB4 and Thunderbolt on the J516M (M3 Max) #1
3. hungbenjamin402

   marked this as blocking on Nov 9, 2026

   • Slice order and acceptance tests for the spec #8
4. hungbenjamin402 commented on Nov 9, 2026

   

   hungbenjamin402

   on Nov 9, 2026

   OwnerAuthor

   More actions

   Findings: https://github.com/hungbenjamin402/linux-aurora/blob/research/macos-usb4/research/macos-usb4.md

   • macOS runs the same NHI, MxWrap, T8122 PHY and T8122 PCIe-C kexts on the T6030 and the T6031/T6034. Every T6030/T6034 difference is in Linux's hw table and DT.
   • The code confirms the T6034 ACIO is an MxWrap IOP:
     • CPU_CTRL RUN: 0x10 at cpu+0x28.
     • LSTX: rc+0x0c = 3 (cable connected) or 2 (no cable).
     • Firmware ready: rc+0xa8 bits 30:24 == 1, polled every 5 ms.
     • Ten ADT tunable tables are applied only after that.
     • Linux's t6031 path does none of the first three (apple.c:154-160).
   • m1n1 swaps the hi_up_tx_desc and hi_up_rx_desc NHI tunable offsets on every chip except T8142 (kboot.c:1556-1558).
   • macOS waits up to 10 s for the CIO reconfigure to complete. Linux waits 100 ms.
   • The PCIe tunnel root port needs a T6031 tunnel DT, a T8122 MSI map and reset table, and apcie-common tunables. The doc lists 19 gaps, each with its file:line.
5. hungbenjamin402

   closed this as completedon Nov 9, 2026