Re: bridge(4) learning

[Lloyd Parkes <lloyd%must-have-coffee.gen.nz@localhost>]

On 23/11/2013, at 10:08 am, Lloyd Parkes 
<lloyd%must-have-coffee.gen.nz@localhost> wrote:

> Hi all,
> I’ve started reworking the bridge(4) code a bit in order to do a better job 
> of handling multicast and broadcast packets. It’s not a big rework, it’s 
> mostly limited to bridge_input(). While looking at bridge_input(), I noticed 
> the following code:
> 
>       if (bif->bif_flags & IFBIF_LEARNING)
>               bridge_rtupdate(sc, eh->ether_shost, ifp, 0, IFBAF_DYNAMIC); 
> 
> It fairly obviously does the learning needed to actually do the bridging. 
> It’s only run when we successfully unicast a packet through the bridge. i.e. 
> we only learn source host addresses from packets when we already know where 
> the destination host is. 
> 
> Surely we should (or could) be learning source host address from all incoming 
> packets? 

And I’ve found the answer to my question. What matters is the state the bridge 
port is in, blocking vs. learning vs. listening vs. forwarding. This probably 
also explains the amount of repeated code in bridge_input().

Cheers,
Lloyd