It's a routing optimization for the network. Even if your dual-homed machine isn't a router, it's still on both segments. The alternative is that the net has to route your packet to the other segment. It's clearly more efficient to have the machine answer directly. On Wed, 2004-11-24 at 15:08 +0000, Illtud Daniel wrote: > Martin A. Brown wrote: > > > Greetings, > > Hi - thanks for your reply. Please bear with me, I'm not an > expert. > > > http://linux-ip.net/html/ether-arp.html#ether-arp-flux > > But *is* this an arp flux problem? I don't think it is. Arp flux > (which I see on a linux box routing here on our network) happens > on the *same* segment, according to this description. In my case, > the interfaces are on different segments, and they never exhibit > arp flux (ie the MAC-IP mapping is always constant), it's just that: > > > By default, an ARP reply will be generated for IP addresses available on a > > host, not just IP addresses on the interface through which the ARP query > > was received. > > ie. eth0 (193.a.b.c) receives an ARP query for 10.0.0.1 and returns > the MAC of eth1 (10.0.0.1) even though eth1 isn't on the same segment > (on nothing except a crossover cable). The ARP data is always > constant (193.a.b.c <-> eth0_mac, 10.0.0.1 <-> eth1_mac). > > This doesn't seem to me to be the same thing as arp flux. Am I wrong? > > What's the advantage of this behaviour being the default - seems to > me that it's information disclosure that shouldn't be happening. Why > should you tell another network segment what your other interface's > IP/MAC addresses are? > -- Lawrence MacIntyre 865.574.8696 macintyrelp@xxxxxxxx Oak Ridge National Laboratory High Performance Information Infrastructure Technology Group AKO: lawrence.macintyre@xxxxxxxxxxx SIPRNet: macintyrelp@xxxxxxxxxxxxxxxxx - : send the line "unsubscribe linux-net" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
