Rockpi4B - ethernet duplicate MACs

[ca249]

I have three Rockpi-4B SBCs. Two of them have the same ethernet MAC address.
I ordered all three from Amazon Prime individually over a period of a week. All
three arrived with two day delivery. The MACs are not marked on the SBCs or in the
boxes in which they were received.

One of the SBCs with the offending MAC (46:af:95:af:26:7c) has ended up as one of
my feeders (108954). After getting the feeder up and running I started working with the
other two SBCs. Then the weirdness started. When I finally figured out that the impossible,
duplicate MACs had, in-fact happened, my first thought was to take a sledge hammer to
the second one. Sensibility then took over and I think I will return it to Amazon as defective
under their 30 return policy and get a free replacement.

I suspected dup MACs in troubleshooting early on but did not persue it because I thought
that dup MACs would never occur. Finally it became obvious that it had to be a dup MAC
problem. A google search indicated that it is more common than most people think.

Some symptoms:

When the dups were plugged in to different switches:
Hostnames changing unexpectedly.
Failure to ssh into one or both SBCs.
“Man in the middle” alerts from ssh.
Unable to assign static IPs.
When plugged in to the same switch:
Complete lost of connections to both SBCs.

Bottom line: Beware of duplicate ethernet MACs with Rockpi4Bs.

[mikkyo]

Did you try changing the MAC address?
https://en.wikibooks.org/wiki/Changing_Your_MAC_Address/Linux

[ca249]

@mikkyo

Did you try changing the MAC address?

No. I did not.

I thought MAC addresses were chiseled in stone.
I have already returned that rockpi for replacement.

I will be prepared for the next time.

Thank you for the info.

[obj]

FWIW there is a similar (but much less common) problem with Raspberry Pis (at least up to the Pi 3 - not sure about the Pi 4, as they seem to use a different scheme). The MAC address is a common prefix plus the low order bits from the CPU serial number; but there’s no guarantee that those low order bits are globally unique.

It’s not usually a problem because you’re unlikely to have two Pis with the same low-order bits on the same network. But it certainly caused problems for FA in the past when we used MAC address to uniquely identify feeders, just because of the scale (and the birthday paradox)