IPv6 Lo. for 6PE/6VPE


I was just wondering , while I'm planning my network to support 6PE/6VPE
why should i assign an IPv6 for Loopbacks?

Maybe it's needed for Point-Point links or external interfaces between my
peers, but anyone here know why i should assign IPv6 for all my Routers
inside my ISP if we will run PE/6VPE not dual stack.


Otherwise the intermediate P devices do not have an address to source
ICMPv6 "hop count exceeded" error replies => traceroute doesn't work

Best regards,

@Daniel +1


Per my understanding, it is not required to have ipv6 address in loopback intf on all P routers inorder to have 6PE work. If I remember it correctly, P router will use ::FFFF::<ipv4-addr> while originating ICMPv6 error message.


Right the ::FFFF::<ipv4-addr> sounds familiar
I guess there was also an option that the P router would just label switch
the packet towards the exit PE and the PE would than originate the ICMP back
to source
Or you can turn off TTL propagation across the core -so the ICMP could only
time out at the PEs


If it does, that's bad... You should never see IPv4 mapped addresses on the wire.
They should only be an internal representation of an IPv4 packet within the host.


Thanks all,

That's sound more logic to me, so we assign IPv6 for lo0 on every hop, so
it can understand ICMP control packet.

But who would use these packets? If I'm at my 6PE then i ping using Lo0
IPv4 address because we are not enabling IPV6 Routing/Dual stack so it can
carry IPv6 addresses across my backbone.

I mean to use IPv6 on loopbacks then i need to advertise them on some
routing protocol BGP/OPSF but actually we are not enabling these protocols
in the Core. An i correct in my logic?


... and some networks filter packets with source address in the mapped
range, so traceroute will be broken for 6PE intermediate P hops.

Best regards,

You mean like this? :wink:

  1. ???
  2. ldn-ipv6-b1.ipv6.telia.net 0.0% 3 1.0 1.2 1.0 1.4 0.2
  3. cogent-ic-125507-ldn-b5.c.telia.net 0.0% 2 40.6 40.4 40.2 40.6 0.3
  4. ::ffff: 0.0% 2 129.1 129.1 129.1 129.1 0.0
  5. ::ffff: 0.0% 2 120.2 120.0 119.8 120.2 0.3
  6. 2001:550::100 0.0% 2 120.2 120.3 120.2 120.5 0.2
  7. ::ffff: 0.0% 2 120.5 120.3 120.1 120.5 0.3
  8. ???
  9. cogentco.com 0.0% 2 119.9 119.9 119.9 119.9 0.0


How the heck is that supposed to work in an all-IPv6 network where you don't
*have* an ipv4-addr? Plus, as Owen noted, leaking those on the wire is
considered bad form.

Yes... That shouldn't happen.

Whoever is responsible for the routers at 154.54.{57.102,30.129,5.253} should fix their configurations.