As a simple example, if a server has two Ethernet cards connected to the
same network configured as 10.0.0.1 and 10.0.0.2. From a client machine,
say 10.0.0.3, pinging both server addresses seems to work, but looking at
the ARP cache reveals that often one interface has responded to both ARP
requests. More interesting, if one interface is brought down the other will
continue to respond to ARP requests and any other network traffic for the
downed adapters IP address.
In a situation where multiple adapters are on the same network segment, the
Linux IPv4 stack behaves as if a many-to-many mapping of a pool of IP
addresses onto a pool of network adapters exists.
While this does provide some basic fail-over capabilities, I can't imagine
that this is expected or desired behavior. If this is the intended
behavior, I'd be interested in hearing why ARP was implemented this way. If
not, I may have time to try and change this and submit a patch.
(note: I'm having trouble sending to the list from my Intel account, please
ignore the fact that I'm sending from my mail.com account)
Chris Leech
christopher.leech@intel.com
Intel Network Communications Group
__________________________________________________
FREE Email for ALL! Sign up at http://www.mail.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/