Cleanup downstream lan78xx

[nbuchwitz]

The downstream microchip,eee-enabled and microchip,tx-lpi-timer DT properties and the corresponding driver-side get/set_eee changes are superseded by Oleksij Rempel's proper phylink LPI integration that landed in 6.17 (673d455). That series adds mac_enable_tx_lpi/mac_disable_tx_lpi callbacks, sets eee_enabled_default = true and lpi_timer_default = 50, so EEE works out of the box without any downstream driver patches.

This reverts both downstream driver patches and converts the DT to use standard eee-broken-* PHY properties (same approach as the genet EEE cleanup in 54a9cd7), keeping the dtparam=eee=off` override working for users who need to disable EEE.

Also reverts the downstream SOF_TIMESTAMPING_TX_SOFTWARE patch which added a duplicate get_ts_info entry - the same fix was upstreamed independently in v6.0 (33e6b16), leaving a harmless but confusing duplicate in the ethtool_ops struct.

8633b35 ("lan78xx: Ack pending PHY ints when resetting") is also a candidate for revert, as the phylink refactoring reworked the IRQ ack path. @pelwell, thoughts?

[popcornmix]

@pelwell I'm in favour of removing unneeded downstream code. Are you happy this is okay?

[pelwell]

Aside from the spurious extra blank line, I'm happy with these. I am, however, nervous about the proposed reversion of 8633b35 - I only see one other read of LAN88XX_INT_STS, and that is conditional on (phydev->state == PHY_NOLINK && !phydev->autoneg && phydev->speed == 100). Convince me that there is no cause for concern.

[nbuchwitz]

Aside from the spurious extra blank line, I'm happy with these. I am, however, nervous about the proposed reversion of 8633b35 - I only see one other read of LAN88XX_INT_STS, and that is conditional on (phydev->state == PHY_NOLINK && !phydev->autoneg && phydev->speed == 100). Convince me that there is no cause for concern.

That's why I wanted to check beforehand. I will try to build a regression setup and test with and without. But this can be merged in a follow-up PR. The downshift patches [1] have also been merged, so we can revert the downstream equivalent in the future too.

[1] https://lore.kernel.org/all/[email protected]/