[PATCH net v3 2/2] net: ethernet: mtk_eth_soc: populate lpi_interfaces to fix EEE support

[Aleksei Sviridkin]

> For a 'net' patch, I think it makes more sense to restrict the lpi
> interface to the chip revisions with support confirmed.

v4 will gate lpi_interfaces on a per-SoC capability bit, so only chips
with a confirmed LPI exit get it and the rest keep the current
behaviour.

MT7981 is confirmed on my bench: with tx-lpi armed the PHY's MMD 3.1
shows Tx LPI asserted by the MAC, and thousands of LPI exits at 1G and
100M against a BCM5720 lost nothing. Details in the v4 cover.

Qingfang, which SoC did you validate the wake-up values on when adding
EEE? That would be the second entry.