MikroTik RB22 changed behaviour since 2 patches

[Michał Kępień]

Hi!

> 9091c9f8cb ath79: mikrotik: compile SWCONFIG and AR8216_PHY as modules
> 12abb2413a ath79: mikrotik: also compile AG71XX_LEGACY as a module
> 
> When testing these changes on my RB922 boards, which have 2 ethernet
> ifaces, I noticed a change in behavior.
> Both interfaces got swapped without any notification:
> So anyone with these boards will suffer from broken iface config upon
> upgrade :-) (like i did)
> 
> Is this expected behavior?

I don't think so, just like breaking network connectivity altogether on
some MikroTik devices was not the expected behavior of 9091c9f8cb ;)

See https://github.com/openwrt/openwrt/issues/23739 for some context.

TL;DR:

  - 9091c9f8cb was necessary to support MikroTik RouterBOARD 960PGS,

  - 9091c9f8cb inadvertently broke switch initialization on boot for
    other devices, e.g. MikroTik RouterBOARD 951G-2HnD,

  - 12abb2413a was suggested as the cleanest way out.

What I am wondering about is whether both Ethernet interfaces on your
RB922 _work_ after 12abb2413a - note the PHY ID and, consequentially,
the driver bound to the 19000000.eth device:

> Before:
> 
> (...)
> [    1.091094] ag71xx-legacy 19000000.eth: connected to PHY at
> mdio.0:04 [uid=004dd074, driver=Qualcomm Atheros AR8031/AR8033]
> (...)

vs.

> after:
> 
> (...)
> [    7.194467] ag71xx-legacy 19000000.eth: connected to PHY at
> mdio.0:04 [uid=00000000, driver=Generic PHY]

Since the same MAC driver handles both of these interfaces, I sense that
the interface swap you're observing is not caused solely by the fact
that the driver is now built as a module, but rather that - for whatever
reason - that change messes up the probing process.

-- 
Best regards,
Michał Kępień