Re: [PATCH net v5 1/2] net: dsa: mt7530: populate lpi_interfaces to fix EEE support

[Aleksei Sviridkin]

On Sat, 5 Sep 2026 01:10:33 +0200, Aleksander Jan Bajkowski wrote:
> I'm a little confused by this patch. At the same time, there are patches
> floating around mailing list to disable EEE on the MT7530 PHY[1]. Does
> the EEE work, or is it broken?

Both, at different layers, and the two patches don't collide.

The MT7530 internal GE PHY really does have broken EEE, and Vladislav's
patch fixes that in the PHY driver where it belongs. Mine is on the MAC
side: since the phylink managed EEE conversion, lpi_interfaces was left
empty for every mt753x port, so phylink never saw the MAC as EEE capable
and ethtool answered "Not supported" everywhere, including on boards
whose PHY is fine.

They compose, and the PHY wins. phy_disable_eee() clears
eee_cfg.tx_lpi_enabled and fills eee_disabled_modes, phy.c:1027 builds
phydev->enable_tx_lpi from that flag, phylink copies it into
pl->phy_enable_tx_lpi at phylink.c:1997, and phylink_activate_lpi() is
gated on (pl->mac_supports_eee && pl->phy_enable_tx_lpi) at
phylink.c:1583. A PHY that disabled EEE keeps LPI off no matter what the
MAC declares.

So with both applied, boards with the broken MT7530 PHY stay EEE off
through the PHY driver, and boards with a working PHY get back the
ability to turn it on. Mine doesn't enable anything by itself, LPI stays
off until userspace asks.