Skip to content

t6041-j575c: USB-C devices attached at boot stay at USB2 (SuperSpeed stuck in RxDetect) until replugged #1670

Closed

Description

@7robertodantas

Written with an AI assistant (Claude) on @7robertodantas's behalf; every hardware step was run and checked on the owner's J575c.

Board: J575c (Mac Studio 2025, M4 Max), linux-aurora 7.3.0, macOS 27.2.1 (26B131).

Summary. On the back Type-C ports, a USB3 device that is already attached when the Mac boots enumerates at USB2 (480M). The SuperSpeed port stays at Link=RxDetect although the T6040 PHY tunes the right lane and moves the PIPE over. Unplugging and plugging the same device in the same orientation gives SuperSpeed Plus Gen 2. #1546 described this as orientation-dependent; it is not: on a fresh plug both orientations train Gen 2. The earlier "lane 0 stays at 480M" case there was a boot-attached device too.

Reproduction (Crucial X10 Pro, USB 3.2 Gen 2, fourth back port):

Case PHY Result
Fresh plug USB3 PHY tuned: lane 1 SuperSpeed Plus Gen 2x1, uas
Fresh plug, rotated USB3 PHY tuned: lane 0 SuperSpeed Plus Gen 2x1, uas
Left attached across a reboot (3 boots) lane 0, USB3 PIPE on the PHY usb-storage 480M; SS port Link=RxDetect
Then replugged, same orientation lane 0 Gen 2 again
macOS 27.2.1, attached across a reboot - 10 Gb/s

Cause. The retimers (ADT i2c6/uatcrt0..3, one per back port) are set up by the port controller (SN201202x) when a partner attaches, not by the AP: macOS's only AP driver for them, AppleTypeCRetimer, has no mode or connection configuration. For a device attached before Linux, that setup happened before the host PHY ran and is never redone. macOS avoids it because its port controllers attach the devices again as macOS starts. The controller's status and data status read the same in the failing and working states; byte 7 of register 0x7e reads 0x04 on connections Linux inherits and 0x08 after a replug and under macOS.

Fix: #1671 renegotiates such a connection at probe through the port controller (the J716c's apple,reset-inherited-usb2-connection mechanism with the J575 key), on the second to fourth back ports.

Warnings for anyone touching these retimers:

  • Talking to them from the AP over I2C (even reading ID registers) dropped an attached SuperSpeed device to USB2 5-8 s later and blanked a DP adapter on that port, until a replug. Leave the atcrt nodes without a driver that talks to them.
  • i2c-pasemi's SMBus block-read emulation wedged this bus (Bus is still stuck) until reboot: it reads the length byte before it has arrived. Plain I2C transfers of the exact length work.
  • Registers above 0x21 NACK, and 0x22 clock-stretches into a controller timeout.

🤖 Generated with Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions