Re: thunderbolt: AMD Phoenix USB4 host advertises Gen3 Support=0 in SB LINK_CONF with TB5 device, link stuck at Gen2

[James R Taylor <jrtaylor11@gmail.com>]

Hi Mika,

On Mon, Nov 09, 2026 at 04:50:07PM +0100, Mika Westerberg wrote:
> That's the host side retimer.

Thanks, that clears it up.

> Agree it looks like the AMD host only gets the link up Gen2 for some
> reason. It could be retimer/PD firmware issue. Anyways I'm adding the AMD
> experts if they have any ideas.

Thank you very much for looking at this and for adding them.

For Mario, Sanath and Basavaraj, a short summary of the setup and
what has been measured so far (the full logs and tbdump output are
earlier in this thread):

  - Host: Minisforum UM890 Pro (AMD Phoenix), BIOS 1.06
    (11/10/2025), AGESA PhoenixPI-FP8-FP7 1.2.0.0a, kernel 7.2.15.
  - Device: Minisforum DEG2 (TB5 eGPU dock, NVM 62.42), StarTech
    TBLT5MM50CM240W 0.5 m TB5 cable.
  - Host-side retimer 1-0:2.1, vendor 0x7fea device 0x1032.
  - Link comes up as TBT, 10.0 Gb/s x2. Host PORT_CS_18 reports
    Cable USB4 Version 0 and CG3=0; the DEG2 side reports Version 1
    and CG3=1.
  - Sideband LINK_CONFIGURATION (0x0c): host 03 d3 03 (Gen3 bit not
    set), DEG2 03 f3 1b (Gen3 bit set). Writing the Gen3 bit on the
    host side from Linux fails with -EIO.
  - Same result with thunderbolt.host_reset=0 and =1, and on a clean
    boot without host_reset/pcie_aspm=off.
  - The firmware exposes no UCSI device (no USBC000/PNP0CA0 in ACPI),
    so /sys/class/typec is absent and I can't see from Linux which
    mode the PD controller chose.

I'm happy to collect any other register reads or logs that would
help.

Thanks again,
James