In message <87ek0alcg0.fsf@phi.internal.fnop.net>Rui Paulo writes
>jonathan@dsg.stanford.edu writes:
>
>> In message <87u096liq4.fsf@phi.internal.fnop.net>Rui Paulo writes
>>
>>>"Konstantin KABASSANOV" <Konstantin.Kabassanov@lip6.fr> writes:
>>>
>>>> If there is a packet with the same source and destination mac address
>>>> (typically a packet wrongly sent over an interface different from the
>>>> loopback), what could be the right behaviour of the kernel? Is the packet
>>>> treated twice (no way to make difference between incoming and outgoing
>>>> direction)?
>>>
>>>If I understad you correctly you are asking what should a system do if
>>>it recives a packet with the same src and dst MAC address ?
>>>Then I guess, it should be dropped.
>>
>> Rui,
>>
>> I can't see any reason why such packets should be dropped, given that
>> the local host decided to emit them?.
>
>I didn't understand that they were emitted by the local host. My
>mistake.

Oh, not necessarily your mistake at all.  It's more that I don't see
how a network stack or driver can distinguish the odd-but-legal case
of a self-addressed MAC frame, from case where a remote host sends a
packet to the local MAC but with the same MAC as a (spoofed) source
MAC address (which I confess I didn't really consider).

I deal regularly with multi-homed machines with mutilple interfaces,
all on the same Ethernet domain. Think about the implications of
trying to filter based on "is this source MAC address one of my local
MACs?", and consider multicast and broadcast traffic. That'll give you
some sense of where I'm coming from.

I concede that's a rather unusual viewpoint :-/.