Re: phy: qcom: qmp-combo: Glymur USB Gen2 fails on one port (chip rev 2.0)

[Burke Cates <burke.cates@gmail.com>]

On 27/10/2026 04:07, Konrad Dybcio wrote:
> Please check the attached patch.

Thanks, Konrad: your patch fixes C2. With it, the RTL8159 10G NIC
enumerates at 10000M on C2 in 8 of 8 tries (both cables, both
orientations, present at boot and hotplugged, including a hotplug
after a cold boot with C2 empty), with no regressions on C1 or DP.

I tested on the A16 (UX3607OA, SoC ID 662, revision 2.0, BIOS
UX3607OA.312), on v7.3 plus out-of-tree Glymur platform/board patches.
Your patch applied as-is, and I removed my local PHY changes so it
alone programs the USB tables.

  - RTL8159 on C2: 10000M (SuperSpeed Plus Gen 2x1) in 8 of 8 tries,
    with both C-C cables from my report, in both orientations, present
    at boot and hotplugged. No -71 or resets during SuperSpeed
    enumeration.
  - Cold boot with C2 empty (my original failure case), then hotplug:
    10000M in both orientations.
  - A VIA VL822 USB 3.2 Gen2 hub on C2 (a third cable) enumerates at
    10000M, and the NIC behind it does too.
  - C2 stays at Gen2 across s2idle (no disconnect or reset on resume).
  - C1 is still Gen2 in both orientations.
  - DP alt mode on C2 still trains 4 lanes at HBR (1080p60 monitor).
  - With the NIC present at boot on C2, iperf3 to a 10G host gave
    about 4.6 Gbit/s each way (TX 4.58, RX 4.71; 4.55 + 4.56 Gbit/s
    bidirectional over 60 s). The NIC's link partner is 5GbE.

Under sustained load I see about one burst of "r8152 ... Tx status
-71" per minute, each followed by an r8152 reset of the device.
Traffic continues and the link stays at Gen2.

Separately, the NIC only gets Ethernet carrier on a Type-C port when
it's present at boot or behind the hub. When hotplugged directly into
C1 or C2 it enumerates at Gen2 but never gets carrier, while its USB2
side errors out during the plug. That happens on both ports and
predates your patch, so I'm treating it as a separate issue and will
report it on its own once I've narrowed it down.

Tested-by: Burke Cates <burke.cates@gmail.com> # ASUS Zenbook A16 (UX3607OA)

AI coding assistants helped run these tests and draft this reply; the
results are from the hardware runs.

Thanks,
Burke