Hoi,

On 3/27/24 10:08, Benoit Chesneau wrote:
I was reading  the ospv3 spec and this link https://networklessons.com/ospf/ospfv3-for-ipv4-
 and  was wondering if such features is supported in bird 2. Can we 
announce loopbacks via OSPFv3 and remove the need to use OSPFv2 and ptp 
subnets ? 

I see it as a good opportunity to reduce the usage of the IPv4 addresses
 if we can just advertise loopbacks...
I tried this today on two machines directly connected as vpp0-2:e1 <-> vpp0-3:e0 (marked in green)

root@vpp0-2:~# ip -br a
lo               UNKNOWN        127.0.0.1/8 ::1/128
loop0            UNKNOWN        192.168.10.2/32 2001:678:d78:200::2/128 fe80::dcad:ff:fe00:0/64
e0               UP             192.168.10.2/32 2001:678:d78:200::2/128 fe80::5054:ff:fef0:1120/64
e1               UP             192.168.10.2/32 2001:678:d78:200::2/128 fe80::5054:ff:fef0:1121/64
e2               DOWN           
e3               DOWN           

root@vpp0-3:/etc/bird# ip -br a
lo               UNKNOWN        127.0.0.1/8 ::1/128
loop0            UNKNOWN        192.168.10.3/32 2001:678:d78:200::3/128 fe80::dcad:ff:fe00:0/64
e0               UP             192.168.10.3/32 2001:678:d78:200::3/128 fe80::5054:ff:fef0:1130/64
e1               UP             192.168.10.3/32 2001:678:d78:200::3/128 fe80::5054:ff:fef0:1131/64
e2               DOWN           
e3               DOWN           


They both took the following OSPF configuration with Bird 2.14 --

protocol ospf v3 ospf4 {
  ipv4 { export where (source = RTS_DEVICE || source = RTS_STATIC); import all; };
  area 0 {
    interface "loop0" { stub yes; };
    interface "e0" { type pointopoint; cost 5; bfd off; };
    interface "e1" { type pointopoint; cost 5; bfd off; };
    interface "e2" { type pointopoint; cost 5; bfd off; };
    interface "e3" { type pointopoint; cost 5; bfd off; };
  };
}

OSPFv3 adjacency did establish, but no routes were exchanged (also none with filter 'export all; import all;') --

root@vpp0-3:/etc/bird# birdc show os ne
BIRD 2.14 ready.
ospf4:
Router ID       Pri          State      DTime   Interface  Router IP
192.168.10.2      1     Full/PtP        30.036  e0         fe80::5054:ff:fef0:1121

root@vpp0-3:/etc/bird# birdc show route protocol ospf4
BIRD 2.14 ready.
Table master4:
192.168.10.3/32      unicast [ospf4 15:29:07.337] I (150/0) [192.168.10.3]
        dev loop0
        dev e0
        dev e1


It would be really useful to let OSPF v3 learn and distribute IPv4 routes with link local nexthops, like so:

pim@vpp0-2:~$ sudo ip ro add 192.168.10.3/32 via inet6 fe80::5054:ff:fef0:1130 dev e1
pim@vpp0-3:/etc/bird$ sudo ip ro add 192.168.10.2/32 via inet6 fe80::5054:ff:fef0:1121 dev e0
pim@vpp0-3:/etc/bird$ ping 192.168.10.2
PING 192.168.10.2 (192.168.10.2) 56(84) bytes of data.
64 bytes from 192.168.10.2: icmp_seq=1 ttl=64 time=3.90 ms
64 bytes from 192.168.10.2: icmp_seq=2 ttl=64 time=3.61 ms
 

However, I think this is not quite working. I've attached (a tiny tar of) config and logs from `vpp0-3` in case somebody wants to double check.
Can somebody confirm that the paragraph Benoit mentioned works for them?

OSPFv2 needs one IPv4 channel. OSPFv3 needs either one IPv6 channel, or one IPv4 channel (RFC 5838). Therefore, it is possible to use OSPFv3 for both IPv4 and Pv6 routing, but it is necessary to have two protocol instances anyway. If no channel is configured, appropriate channel is defined with default parameters.



groet,
Pim
-- 
Pim van Pelt
PBVP1-RIPE - https://ipng.ch/