[Xorp-users] IPv6 & BGP
Atanu Ghosh
atanu@ICSI.Berkeley.EDU
Sat, 17 Dec 2005 10:27:51 -0800
>>>>> "Hasso" == Hasso Tepper <hasso@linux.ee> writes:
Hasso> Mark Handley wrote:
>> On 12/17/05, Kristian Larsson <kristian@juniks.net> wrote: > I'm
>> having problem setting up a v6 only BGP > connection. it would
>> seem it requires a next-hop > parameter which is v4. Since I
>> don't have v4 on > this machine at all it is difficult. > Please
>> advice
>>
>> The BGP spec requires a NEXTHOP attribute in all routes, and this
>> is specified as an IPv4 address. It needs this even if you're
>> running IPv6 only, but it shouldn't be used for the routing
>> calculation. I've not tried it (probably Atanu has) but you
>> should be able to specify 10.0.0.1 or something similar.
Hasso> Sorry, but this is nonsense.
Hasso> From draft-ietf-idr-rfc2858bis-07.txt (the same text is there
Hasso> in the RFC2858 as well):
Hasso> "An UPDATE message that carries no NLRI, other than the one
Hasso> encoded in the MP_REACH_NLRI attribute, SHOULD NOT carry the
Hasso> NEXT_HOP attribute. If such a message contains the NEXT_HOP
Hasso> attribute, the BGP speaker that receives the message SHOULD
Hasso> ignore this attribute."
Actually this is not a protocol issue but a configuration issue. In the
configuration we have an IPv4 nexthop marked as a mandatory item. In the
case where you want to use BGP for IPv6 only we have no way of disabling
the check (yet). If you set the ipv4-unicast and ipv4-multicast flags to
false and you have no local IPv4 addresses then just choose a legal IPv4
address. The IPv4 nexthop address should *not* be used it just needs to
satisify the configuration constraint. I have never tried an IPv6
peering on a machine that does not have any IPv4 addresses configured
but I would expect it to work.
Atanu.