Skip to content

aurora14: second left-hand Thunderbolt port never brings up a DP tunnel (DPRX negotiation failed, wedged until reboot) #374

Open

Description

@ollies0x

Summary

On linux-aurora 7.2.11.aurora14-1 (M1 Pro 14", Asahi), the two Thunderbolt ports on the
left-hand side behave differently:

  • cio@701ac0000 / nhi@701f00000 → USB4 connects, DP tunnel comes up, LG 40WP95C works.
  • cio@b01ac0000 / nhi@b01f00000 (DCP 28cc00000) → USB4 and PCIe come up, USB works,
    but the DP tunnel never establishes. It fails DPTX target validation, then DPRX
    negotiation, and stays wedged for the rest of the boot.

This is separate from #268, which I now understand to be a suspend/wake issue. Same box,
same cable, same monitor; only the port differs.

Hardware / versions

  • Apple MacBook Pro (14-inch, M1 Pro, 2021)
  • linux-aurora 7.2.11.aurora14-1, uname 7.2.11-14-1-ARCH
  • Monitor: LG 40WP95C, 5120x2160, direct with the Apple Thunderbolt cable, no dock
  • Lid closed (clamshell) and lid open both fail on this port

Port identification, for the record:

readlink -f /sys/class/typec/port*-partner
  port1 (usb-pd@38) -> cio@701ac0000     # works
  port2 (usb-pd@3f) -> cio@b01ac0000     # broken

Kernel log

First attempt after plugging into the b01 port (the tunnel is handed to dpin0 without a
Type-C PHY allocation):

03:24:59 apple-dcp 28cc00000.dcp: display routed to Thunderbolt DP tunnel dpin0
03:24:59 apple-dcp 28cc00000.dcp: dcp_dptx_connect_raw(port=0) die=0 atc=1 dfp_port=1 tunnel=1 typec=1 route=borrowed conn_type=20 connected=0
03:25:00 dcp_dptx_connect: failed to validate DPTX target 0:1: -110
03:25:01 dcp_dptx_connect: failed to validate DPTX target 0:1: -110
03:25:02 dcp_dptx_connect: failed to validate DPTX target 0:1: -110
03:25:04 dcp_dptx_connect: failed to validate DPTX target 0:1: -110
03:25:06 dcp_dptx_connect: failed to validate DPTX target 0:1: -110
03:25:08 dcp_dptx_connect: failed to validate DPTX target 0:1: -110
03:25:10 apple-dcp 28cc00000.dcp: Type-C DPTX reconnect failed after 5 retries: -110
03:25:11 0:5: DP IN resource unavailable: DPRX negotiation failed
03:25:12 dpin0: display teardown failed: -110
03:25:12 dpin0: DP tunnel down
03:25:12 apple-dcp 289c00000.dcp: display routed to Thunderbolt DP tunnel dpin1
03:25:12 apple-dcp 289c00000.dcp: no Type-C connector for the DP tunnel

Then it retries on the other adapter and gets a DPRX timeout storm:

03:25:23 0:6 <-> 1:13 (DP): DPRX read timeout          (repeating, ~40/s for 5 s)
03:25:23 0:6: DP IN resource unavailable: DPRX negotiation failed
03:25:23 no suitable DP IN adapter available, not tunneling

A later reconnect attempt does allocate a PHY and still fails the same way:

03:26:04 apple-dcp 28cc00000.dcp: allocated Type-C DPTX PHY 1
03:26:16 apple-dcp 28cc00000.dcp: dcp_hotplug() connected:0 valid_mode:0 nr_modes:0

28cc00000 logs nr_modes:0 here, while 289c00000 reports nr_modes:14 on the working
port — DCP never sees a mode list on this path.

Not a link or bandwidth problem

  • PCIe tunnel is up: Link up on /soc/usb4-pcie-tunnel-1/pcie@b30000000/pci@0,0, device
    authorized at 40 Gb/s (boltctl).
  • USB3 tunnel is up and stays up: the monitor's Ethernet adapter and a USB microphone on
    the monitor hub keep working through every attempt.
  • Bandwidth was free at allocation: available bandwidth for new DP tunnel 34650/34650 Mb/s.

Persistence

The wedge does not clear for the rest of the boot: unplug/replug, waiting, opening the
lid, and plugging into the working port (which recovers only because it is a different
NHI/domain) all leave this port at no suitable DP IN adapter available. A reboot is
required. card2-DP-1 stays disconnected.

Guess

28cc00000 is handed the tunnel before it has a Type-C PHY, and its DPTX target
validation then times out. On the working port the same sequence logs
allocated Type-C DPTX PHY 0 before display routed to Thunderbolt DP tunnel dpin0.
Whether the DCP↔ACIO PHY association is expected to be resolved per port is the part I
cannot tell from the logs.

Full log for the three cases (this port, the working port, and the clamshell case in #268):
https://gist.github.com/ollies0x/99581521c8eb362ddf9a5bd9e3304ef6

Happy to test candidate builds or capture with drivers/thunderbolt/* +p enabled.

Activity

  1. thetylerreiff commented on Nov 7, 2026

    @thetylerreiff

    Same behavior here on a second machine. In my case it goes through a dock instead of a direct connection.

    Setup

    • Apple MacBook Pro (14-inch, M1 Pro, 2021), Omarchy/Hyprland
    • linux-aurora 7.2.11.aurora14-1 (7.2.11-14-1-ARCH)
    • Anker Prime Docking Station (14-in-1, 8K, Thunderbolt 5) → ASUS PA32QCV (6K, TB4), daisy-chained off the dock's Thunderbolt downstream port

    Same port split as yours

    Port typec CIO / NHI DCP Result
    left, usb-pd@3f port2 cio@b01ac0000 / nhi@b01f00000 28cc00000 USB4, PCIe and USB through the dock all work. The DP tunnel never comes up.
    left, usb-pd@38 port1 cio@701ac0000 / nhi@701f00000 289c00000 Works: DP-1 at 6016x3384@60

    Both the dock and the monitor show as authorized in boltctl at 40 Gb/s on the broken port. Only the video fails.

    Log from the broken port (cold attach at boot, then a monitor replug about 3 minutes later):

    [   40.187015] thunderbolt-apple-nhi b01f00000.nhi: device links to tunneled native ports are missing!
    [   43.132649] thunderbolt 0-701: ASUS-Display PA32QCV
    [   43.778806] apple-dcp 28cc00000.dcp: display routed to Thunderbolt DP tunnel dpin0
    [   55.768035] thunderbolt-apple-nhi b01f00000.nhi: 0:5 <-> 701:14 (DP): not active, tearing down
    [   56.801507] thunderbolt-apple-acio b01ac0000.cio: dpin0: display teardown failed: -110
    [   56.803931] apple-dcp 289c00000.dcp: display routed to Thunderbolt DP tunnel dpin1
    [   56.804509] apple-dcp 289c00000.dcp: no Type-C connector for the DP tunnel
    [   67.839092] thunderbolt-apple-nhi b01f00000.nhi: 0:6 <-> 701:14 (DP): not active, tearing down
    ...
    [  213.017291] thunderbolt 0-701: ASUS-Display PA32QCV
    [  214.049714] thunderbolt-apple-acio b01ac0000.cio: dpin0: DP tunnel setup failed: -110
    [  225.066133] thunderbolt-apple-nhi b01f00000.nhi: 0:5 <-> 701:13 (DP): not active, tearing down
    [  225.100886] apple-dcp 289c00000.dcp: display routed to Thunderbolt DP tunnel dpin1
    [  225.101515] apple-dcp 289c00000.dcp: no Type-C connector for the DP tunnel
    [  237.099242] thunderbolt-apple-nhi b01f00000.nhi: 0:6 <-> 701:13 (DP): not active, tearing down
    

    After moving the dock to the usb-pd@38 port, with no reboot:

    [ 1094.078981] thunderbolt 0-701: ASUS-Display PA32QCV
    [ 1094.722348] apple-dcp 289c00000.dcp: display routed to Thunderbolt DP tunnel dpin0
    

    The display came up right away, which fits your finding that the other port is a separate NHI/domain.

    On the broken port, DCP 28cc00000 times out on dpin0. The tunnel then falls back to dpin1 on 289c00000, which has no Type-C connector. That second line may be the dpin1 device-tree gap that #373 addresses.

    I'm happy to capture a log with thunderbolt.dyndbg=+p thunderbolt_apple.dyndbg=+p phy_apple_atc.dyndbg=+p on the broken port if that would help.

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