Re: [PATCH net-next v6 2/2] net: dsa: realtek: rtl8365mb: add HSGMII support for RTL8367S

[Stacho_P <kuncy7@gmail.com>]

Tested this series on a TP-Link Archer AX55 v1: RTL8367S with the
SerDes on external interface 1 running HSGMII (2500base-x fixed-link)
to the SoC (Qualcomm IPQ5018 stmmac gmac1). Base was OpenWrt's 6.12
kernel; the only adjustment needed was dropping the neg_mode parameter
from pcs_get_state() for the older phylink API there.

The trunk comes up at 2.5Gbps and passes traffic (all five front ports
are behind it, NAT routed through the CPU port plus wire-speed L2 on
the switch). Warm reboots and short power cycles bring it up clean
every time, no SerDes firmware involved.

One observation, quite possibly marginal silicon on my unit: after the
device has been powered off for several hours, the first boot brings
the link up (2.5Gbps/Full reported, phylink happy) but the data path
is heavily degraded - 60-70% packet loss, the surviving packets at
normal sub-ms RTT. Re-running the PCS sequence via admin down/up of
the CPU port re-rolls the dice (15% and 40% loss on two consecutive
attempts) but did not fully recover it; a soft reboot (full re-probe
including the chip reset) always restores a clean link. Short
power-offs (~a minute) do not reproduce this. For what it's worth, the
vendor firmware for this board keeps the DW8051 link-management loop
running permanently, which would effectively keep retrying out of
exactly such a state.

Happy to test follow-up patches if you want to poke at that; I don't
think it should block the series - the feature itself works as
advertised.

Tested-by: Stanisław Pal <kuncy7@gmail.com>