Re: [v6ops] Q about IPv4-mapped IPv6 address & MPLS

[Brian E Carpenter <[e-mail]>]

On 01/05/2014 20:59, Gert Doering wrote:
> Hi,
> 
> On Thu, May 01, 2014 at 06:08:51AM +0000, Rajiv Asati (rajiva) wrote:
>> We need your guidance on handling v4-mapped v6 addresses (section 2.5.5.2
>> of [RFC4291]). 
>>
>> 1. Should/Would they appear in IPv6 routing table?
> 
> "Yes and no".  It depends on the context - if the packet comes in and has
> an IPv6 header, address lookup happens via the IPv6 routing table.  If 
> a packet comes in with an IPv4 header, address lookup happens via the
> IPv4 routing table.  
> 
> So, IPv4 should never "bleed over" to the IPv6 routing table, but if 
> you happen to receive an IPv6 packet with a destination IPv6 address
> in it's header of ::ffff:1.2.3.4, you'd either drop it right away,
> or use the IPv6 routing table to find the destination.
> 
> The router itself should never ever source packets with a v4-mapped
> destination address in the packet (see below).
> 
>> 2. Should an IPv6 packet with ::FFFF:127.0.0.0  be forwarded or dropped or
>> treated as a loopback packet, if ever received?
> 
> Dropped.  v4-mapped is an internal representation of an IPv4 address
> recevied on an ipv6 socket, and should never ever appear in packets on
> the wire.

Nobody seems to have mentioned RFC 4038, which makes this very clear.
These addresses are an artefact and have no place in the routing
system as such, and still less on the wire.

    Brian