On Monday, 05/21/2012 at 07:41 EDT, David Boyes <dboyes@sine...>
> On 5/21/12 4:11 AM, "Ursula Braun" <ubraun@linu...> wrote:
> >we reported this problem in Novell bugzilla 617373. Solution has been
> >introduce a SEND_GRATUITOUS_ARP config option. Its default is "no".
> >Changing it to "yes" should trigger the sending of gratuitous ARPs. The
> >config option is offered starting with sysconfig-0.71.30-0.10.1.
> When would this NOT be the correct behavior? Seems like the default
> be yes, given that that's the way ARP has worked for decades.
You beat me to it. RFC 5227 (IPv4 address conflict detection, ACD) takes
Stevens' grat ARP (request, not reply - my bad) to the next level, and
establishes the standard for detection of duplicate IP address usage.
To quote from section RFC 5227 section 2.1:
Before beginning to use an IPv4 address (whether received from manual
configuration, DHCP, or some other means), a host implementing this
specification MUST test to see if the address is already in use, by
broadcasting ARP Probe packets. This also applies when a network
interface transitions from an inactive to an active state, when a
computer awakes from sleep, when a link-state change signals that an
Ethernet cable has been connected, when an 802.11 wireless interface
associates with a new base station, or when any other change in
connectivity occurs where a host becomes actively connected to a
Exceptions would be permitted for interfaces that, while physically
connected to the infrastructure, are not ready to receive packets for a
particular IP. I.e. backup interfaces.
If you're going to turn off grat ARPs (an incredibly dumb idea), you have
to turn off learning mode on the switches. Feh.
If there's going to be a setting to control grat ARPs, it needs to be a
flag on ifconfig and ioctl(), not some global server setting. IMHO, of
Senior Managing z/VM and Linux Consultant
IBM System Lab Services and Training
Alan_Altmark@us.i... IBM Endicott