In article <Pine.LNX.4.44.0206271522070.7655-100000@havoc.ittc.ku.edu> you wrote: >> That is wrong. ARP has nothing to do with it. Switches learn the MAC >> addresses by looking at the source MAC address coming in on the port, >> and nothing else. > That what I would have thought too. But my NetGear FS516 100Mbit switch > shows some characteristics that led me to believe that it "depends" on > catching ARPs. If you think about it, the first package of a new device on a hub port is usually the arp response. Therefoe your switch learns the address from the first arp reply, if your device is giving arp replys. > Now at the switch, I see something strange. The switch _broadcasts_ > these packets to _all_ connected ports, thus behaving like a hub. It does this as long as it does not know the port to the taret mac address. > But if I first send traffic between machines 1 and 2, and then try the > above experiment, it works. As soon as your target has responded the switch should know. Personally I think it is much likely the sweitch is only inspecting ARP packets. You could do a gratious arp announcement with arping on your systems and avoid that problem. This is also a good idea to do ip address duplicate detection that way. arping may be the tool you need to debug that problem further. Greetings Bernd - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
