Re: Raspberry Pi 4 ethernet issues with latest UEFI and 11.0_RC6

["David H. Gutteridge" <david%gutteridge.ca@localhost>]

On Tue, 28 Jul 2026 at 08:14:43 +0000, sc dying wrote:
> On Sun, Jul 26, 2026 at 4:47 PM John Klos <john%klos.com@localhost>
> wrote:
> >
> > Hi,
> 
> hello,
> 
> >
> > On both a Raspberry Pi 4 and a Pi 400, I've noticed that genet0
> ethernet
> > doesn't work a good fraction of the time after boot.
> >
> > This happened infrequently in the past, with older UEFI and older
> versions
> > of NetBSD 11. I've recently updated to the latest UEFI version 1.52
> and
> > NetBSD 11.0_RC6, and the problem seems to still happen on the Pi 4
> and now
> > it happens very often on the Pi 400.
> >
> > When this happens, genet0 is configured, and tcpdump shows packets
> coming
> > in, but nothing goes out.
> >
> > Occasionally, the network interface can communicate with certain
> hosts,
> > but not others, as if MAC filtering works selectively, or something
> like
> > that.
> 
> Have you seen some messages like this in dmesg?
> 
> > [     1.000004] genet0: Ethernet address f5:77:00:f2:3a:3b
>  :
> > [     1.000004] autoconfiguration error: The multicast bit is set in
> the MAC address. It's wrong.
> 
> If uefi firmware is v1.41 or later, RD4(UMAC_MAC[01]) return 0.
> genet_get_eaddr() tries to invent mac address, but it needs htobe().

Another way of putting it would be that until r. 1.16 of bcmgenet.c, the
MAC random generation code shared the ordering used in other drivers.
Then it was changed per "Don't mix endianess and MAC address layout."
But the initial generation of the maclo and machi values was left
unchanged, which can lead to randomly ending up with the multicast bit
set, it seems.
https://nxr.netbsd.org/diff/src/sys/dev/ic/bcmgenet.c?r2=%2Fsrc%2Fsys%2Fdev%2Fic%2Fbcmgenet.c%401.16&r1=%2Fsrc%2Fsys%2Fdev%2Fic%2Fbcmgenet.c%401.15

(Now, I have no familiarity at all with this driver or the firmware
differences, nor do I have an RPI 4, so I can't claim any understanding
of or judgment about why this was done. I did happen to notice it when
I was comparing drivers with MAC randomization code, since it's another
variant of how the "same" thing is expressed.)

> > Has anyone else experienced this yet? If someone who knows how to
> diagnose
> > would like serial console access to the Pi 4 to see what's up when
> this is
> > happening, please reach out.
> >
> > Thanks!
> > John