On Fri, Nov 28, 2014 at 06:34:33PM -0500, Geoff Steckel wrote:
> When an interface is given an IP6 address in anew rdomain,
> lo0 is named in various routes when that table is queried
> via netstat -r -f inet
> 
> Does the pseudo-interface lo0 actually "exist" in multiple
> routing tables simultaneously, or does the name 'lo0'
> signify an otherwise anonymous point to hang the extra
> routes defined when any address is configured in a rdomain?

lo(4) is a bit special since it will bounce back the traffic it got
unmodified. Because of this lo0 can be abused by a different rdomain.
But this is something we don't want and AFAIK mpi@ is on the way to
change the use of lo0 in IPv6 local addresses. There are still some
issues with IPv6 and rdomains (mainly with ND).

 
> -----
> river:5626$ sudo ifconfig re2 rdomain 3
> river:5627$ netstat -rn -f inet6 -T 3
> Routing tables
> 
> river:5629$ sudo ifconfig re2 inet6 2002:1234:4567::0
>                                   prefixlen 64 rdomain 3
> river:5630$ netstat -rn -f inet6 -T 3
> Routing tables
> 
> Internet6:
> Destination                        Gateway
>                        Flags   Refs      Use   Mtu  Prio Iface
> 2002:1234:4567::                   00:30:18:ab:ee:49
>                        HL         0        0     -     4 lo0
> 2002:1234:4567::/64                link#3
>                        C          0        0     -     4 re2
> fe80::%re2/64                      link#3
>                        C          0        0     -     4 re2
> fe80::230:18ff:feab:ee49%re2       00:30:18:ab:ee:49
>                        HL         0        0     -     4 lo0
> ff01::%re2/32                      link#3
>                        C          0        0     -     4 re2
> ff02::%re2/32                      link#3
>                        C          0        0     -     4 re2
> ------
> 
> In a more complex case, lo0 exists with a route but was never
> configured in this rdomain.
> 
> This is part of my implementation
> of an ip6 tunnel endpoint. The endpoint IP6 address must not
> be used as a source address from my net so that address is
> in a separate routing domain. pf transfers packets between
> the domains.
> 
> ------
>   netstat -rn -f inet6 -T 1
> Routing tables
> 
> Internet6:
> Destination                        Gateway
>            Flags   Refs      Use   Mtu  Prio Iface
> default                            2001:4830:xxxx:xxx::1
>            UGS        0      427     -     8 gif0
> ::1                                link#7
>            UHL        0        0     -     4 lo0
> 2001:4830:xxxx:xxx::1              2001:4830:xxxx:xxx::2
>            UH         1       61     -     4 gif0
> 2001:4830:xxxx:xxx::2              link#10
>            UHL        0        0     -     4 lo0
> fe80::%lo1/64                      fe80::1%lo1
>            U          0        0     -     4 lo1
> fe80::1%lo1                        link#7
>            UHL        0        0     -     4 lo0
> fe80::%gif0/64                     link#10
>            UC         0        0     -     4 gif0
> fe80::230:18ff:feab:ee47%gif0      link#10
>            UHL        0        0     -     4 lo0
> ff01::%lo1/32                      fe80::1%lo1
>            UC         0        0     -     4 lo1
> ff01::%gif0/32                     link#10
>            UC         0        0     -     4 gif0
> ff02::%lo1/32                      fe80::1%lo1
>            UC         0        0     -     4 lo1
> ff02::%gif0/32                     link#10
>            UC         0        0     -     4 gif0
> 
> If this is not the intended behavior, I can dig into
> the code - this would be in the ip6 route add code?
> 
> If it should be sent there  I'll copy this to tech@
> 
> thanks
> Geoff Steckel
> 

-- 
:wq Claudio

Reply via email to