When my MN moves to a subnet on which routers advertise multiple prefixes, things go wrong from time to time. Why?
when umip detects a link is up again, it performs prefix discovery by sending RS and waiting for RA. When a RA is received with multiple PIO (Prefix Information Options), it progressively configures addresses based on those prefixes. In parallel, when a new address is configured, another thread of the daemon receives a Netlink signal and checks if the associated interface has a higher preference value and if the address is suitable. If this is the case, umip make that address its new primary CoA. At that moment, if new addresses are configured on this interface, umip will not consider switch to them, i.e. the first address that gets configured is used. Let’s now rewind a little and makes things more practical on an example. A router on the new link returns a RA with following prefixes: 2002:db8:1::/64, 2001:db8::/64 and fdff:db8::/64. If an address based on the 6to4 prefix is configured first on the associated interface eth0, then, because it is suitable for joining the HA and no other
