10 minutes
Using OSPF Unnumbered Interfaces with Infix Linux and FRR
When using OSPF over Ethernet point to point links, the work to make IP plans for the PtP connections can be tedious and limits flexibility as configurations must be coordinated and matched with each neighbor. The figure below is used as illustration.
Figure 1: OSPF network with point-to-point links
10.1.1.1 10.1.1.2 10.1.1.4 10.1.1.6
.----. .----. .----. .----.
| R1 +-------+ R2 +-------+ R4 +-------+ R6 |
'---+' '-+--' '--+-' '+---'
\ | | /
\ | | /
\ | | /
\ .-+--. .--+-. /
`---+ R3 +-------+ R5 +---´
'----' '----'
10.1.1.3 10.1.1.5
The setup disregards any user networks attached to the routers, and focus solely on establishing connectivity between routers. (Adding host subnets is described further below).
With OSPF, router connectivity would be easy to configure if the links can be treated as unnumbered point-to-point OSPF interfaces. With this approach, each router can be configured individually with minimum coordination with its neighbours. Ideally, R1 should have IP address 10.1.1.1 assigned to its loopback, advertise 10.1.1.1 via OSPF, and then connect one port (say eth0) towards R2 and another (eth1) towards R3.
In contrast, the custom way is to make an IP plan for each connection, e.g., 10.0.1.0/30 between R1 and R2, assign address 10.0.1.1/30 to eth0 on R1 and 10.0.1.2/30 to eth0 on R2. And do the same for subnet 10.0.1.4/30 between R1 and R3, etc. This work we wish to avoid.
With OSPF unnumbered interfaces we can skip coordinating and configuring IP addresses on Ethernet LANs used as point-to-point links between routers.
OSPF unnumbered interfaces in a nutshell
- Make an IP plan for your routers: In this example R1 is allocated the address 10.1.1.1, R2 address 10.1.1.2, etc.
- Configure the allocated address as /32 to loopback and to all Ethernet interfaces intended as unnumbered interfaces. On R1, we would configure 10.1.1.1/32 to lo, eth0 and eth1.
- In OSPF configuration there are two settings
- advertise the /32 address, e.g,
network 10.1.1.1/32 area 0 - set Ethernet interfaces as OSPF point-to-point networks
- advertise the /32 address, e.g,
- Plug the network together.
The routers should detect eachother, build routing tables for connectivity.
Minor test: Two nodes
We start out with a simple network, just two nodes
Figure 2: OSPF network with point-to-point link between two routers
10.1.1.1 10.1.1.2
.----. .----.
| R1 +-------+ R2 |
'----' '----'
The first step, the IP plan, is already shown in Figure 2.
Configuring IP address (and enabling IP forwarding)
Below we show how to set the address on R1, starting with the regular
Linux ip address command, then via Infix CLI
root@R1:~$ ip address add 10.1.1.1/32 dev lo
root@R1:~$ ip address add 10.1.1.1/32 dev eth0
root@R1:~$
The result can be viewed:
root@R1:~$ ip -brief address
lo UNKNOWN 127.0.0.1/8 10.1.1.1/32 ::1/128
eth0 UP 10.1.1.1/32 fe80::ee5:3bff:fead:0/64
eth1 DOWN
eth2 DOWN
eth3 DOWN
root@R1:~$
As the unit is intended to work as IPv4 router, ensure IP forwarding is enabled.
root@R1:~$ sysctl -w net.ipv4.ip_forward=1
net.ipv4.ip_forward = 1
root@R1:~$
Corresponding using Infix CLI
root@R1:~$ cli
root@R1:exec> configure
[edit]
root@R1:configure> edit interfaces
[edit interfaces]
root@R1:configure> set interface lo ipv4 address 10.1.1.1 prefix-length 32
[edit interfaces]
root@R1:configure> set interface eth0 ipv4 address 10.1.1.1 prefix-length 32
[edit interfaces]
root@R1:configure> set interface eth0 ipv4 forwarding true
[edit interfaces]
The resulting interface configuration can be listed
root@R1:configure> show
interface eth0 {
type ethernetCsmacd;
ipv4 {
forwarding true;
address 10.1.1.1 {
prefix-length 32;
}
}
}
interface eth1 {
type ethernetCsmacd;
}
interface eth2 {
type ethernetCsmacd;
}
interface eth3 {
type ethernetCsmacd;
}
interface lo {
type softwareLoopback;
ipv4 {
address 10.1.1.1 {
prefix-length 32;
}
address 127.0.0.1 {
prefix-length 8;
}
}
ipv6 {
address ::1 {
prefix-length 128;
}
}
}
[edit interfaces]
root@R1:configure>
OSPF configuration
Here we assume the routers use the FRR routing stack (zebra and ospfd daemons). OSPF support is not yet integrated into the Infix CLI, thus the FRR CLI is used directly. (See Infix Hybrid Mode for information on starting FRR in Infix.)
root@R1:~$ vtysh
Hello, this is FRRouting (version 8.4.2).
Copyright 1996-2005 Kunihiro Ishiguro, et al.
R1# configure
R1(config)# router ospf
R1(config-router)# network 10.1.1.1/32 area 0
R1(config-router)# interface eth0
R1(config-if)# ip ospf network point-to-point
R1(config-if)# end
R1# copy running-config startup-config
Building Configuration...
Integrated configuration saved to /etc/frr/frr.conf
[OK]
R1# show running-config
Building configuration...
Current configuration:
!
frr version 8.4.2
frr defaults traditional
hostname R1
no ip forwarding
no ipv6 forwarding
service integrated-vtysh-config
!
interface eth0
ip ospf network point-to-point
exit
!
router ospf
network 10.1.1.1/32 area 0
exit
!
end
R1#
Checking connectivity
Assuming the above configuration is done on R1 and R2 (with address 10.1.1.2), we can verify connectivity.
We can watch the IP routing table using the Linux ip route command
root@R1:~$ ip route
10.1.1.2 nhid 16 via 10.1.1.2 dev eth0 proto ospf metric 20 onlink
root@R1:~$
On R2, the corresponding would be
root@R2:~$ ip route
10.1.1.1 nhid 16 via 10.1.1.1 dev eth0 proto ospf metric 20 onlink
root@R2:~$
More details can be seen in FRR CLI
root@R1:~$ vtysh
Hello, this is FRRouting (version 8.4.2).
Copyright 1996-2005 Kunihiro Ishiguro, et al.
R1# show ip ospf route
============ OSPF network routing table ============
N 10.1.1.1/32 [0] area: 0.0.0.0
directly attached to lo
N 10.1.1.2/32 [10] area: 0.0.0.0
via 10.1.1.2, eth0
============ OSPF router routing table =============
============ OSPF external routing table ===========
R1# show ip ospf neighbor
Neighbor ID Pri State Up Time Dead Time Address Interface RXmtL RqstL DBsmL
10.1.1.2 1 Full/- 24.472s 36.561s 10.1.1.2 eth0:10.1.1.1 0 0 0
R1# show ip ospf interface
eth0 is up
ifindex 2, MTU 1500 bytes, BW 10000 Mbit <UP,BROADCAST,RUNNING,MULTICAST>
This interface is UNNUMBERED, Area 0.0.0.0
MTU mismatch detection: enabled
Router ID 10.1.1.1, Network Type POINTOPOINT, Cost: 10
Transmit Delay is 1 sec, State Point-To-Point, Priority 1
No backup designated router on this network
Multicast group memberships: OSPFAllRouters
Timer intervals configured, Hello 10s, Dead 40s, Wait 40s, Retransmit 5
Hello due in 4.351s
Neighbor Count is 1, Adjacent neighbor count is 1
lo is up
ifindex 1, MTU 65536 bytes, BW 0 Mbit <UP,LOOPBACK,RUNNING>
Internet Address 10.1.1.1/32, Broadcast 10.1.1.1, Area 0.0.0.0
MTU mismatch detection: enabled
Router ID 10.1.1.1, Network Type LOOPBACK, Cost: 10
Transmit Delay is 1 sec, State Loopback, Priority 1
No backup designated router on this network
Multicast group memberships: <None>
Timer intervals configured, Hello 10s, Dead 40s, Wait 40s, Retransmit 5
Hello due in inactive
Neighbor Count is 0, Adjacent neighbor count is 0
R1# show ip route
Codes: K - kernel route, C - connected, S - static, R - RIP,
O - OSPF, I - IS-IS, B - BGP, E - EIGRP, T - Table,
v - VNC, V - VNC-Direct, A - Babel, F - PBR,
f - OpenFabric,
> - selected route, * - FIB route, q - queued, r - rejected, b - backup
t - trapped, o - offload failure
O 10.1.1.1/32 [110/0] is directly connected, lo, weight 1, 00:03:23
C * 10.1.1.1/32 is directly connected, eth0, 00:05:05
C>* 10.1.1.1/32 is directly connected, lo, 00:05:05
O>* 10.1.1.2/32 [110/10] via 10.1.1.2, eth0 onlink, weight 1, 00:00:48
R1# ping 10.1.1.2
PING 10.1.1.2 (10.1.1.2): 56 data bytes
64 bytes from 10.1.1.2: seq=0 ttl=64 time=0.584 ms
64 bytes from 10.1.1.2: seq=1 ttl=64 time=0.690 ms
64 bytes from 10.1.1.2: seq=2 ttl=64 time=0.607 ms
^C
--- 10.1.1.2 ping statistics ---
3 packets transmitted, 3 packets received, 0% packet loss
round-trip min/avg/max = 0.584/0.627/0.690 ms
R1#
Troubleshooting
If OSPF fails to setup the routes, it is usually good to start
checking from bottom up. Is cable connected, are Ethernet interfaces
up, etc. To verify link connectivity, it can be hard to use IPv4 ping as we have not assigned any IPv4 address to eth0 than the /32
address. Instead we can use tcpdump to listen for LLDP packets.
root@R1:~$ tcpdump -lni eth0
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
12:34:45.853072 IP6 fe80::e31:9ff:fe4b:0 > ff02::2: ICMP6, router solicitation, length 16
12:34:47.129240 LLDP, length 211: R1
12:34:47.130563 LLDP, length 211: R2
^C
3 packets captured
3 packets received by filter
0 packets dropped by kernel
In this case, R1 could see both its outgoing LLDP message and R2’s incoming LLDP. A good sign of connectivity.
We can also use IPv6 ping to all-hosts address. Below R1 sends such a ping on interface eth0.
root@R1:~$ ping -6 ff02::1%eth0
PING ff02::1%eth0 (ff02::1%eth0): 56 data bytes
64 bytes from fe80::ee5:3bff:fead:0: seq=0 ttl=64 time=0.082 ms
64 bytes from fe80::e31:9ff:fe4b:0: seq=0 ttl=64 time=1.468 ms (DUP!)
64 bytes from fe80::ee5:3bff:fead:0: seq=1 ttl=64 time=0.103 ms
64 bytes from fe80::e31:9ff:fe4b:0: seq=1 ttl=64 time=1.239 ms (DUP!)
^C
--- ff02::1%eth0 ping statistics ---
2 packets transmitted, 2 packets received, 2 duplicates, 0% packet loss
round-trip min/avg/max = 0.082/0.723/1.468 ms
root@R1:~$
We see two responses, one from R1 itself and one from R2. This is also
a good sign of connectivity. (If available, the -L ping option can
be used to filter out local responses.)
Larger setup
To go from the simple setup with two nodes (Figure 2 to the larger in Figure 1, we should include the remaining interfaces on R1 and R2, and the do the corresponding configuration for R3-R5.
On R1, we need include eth1 for IP settings and OSPF.
Doing IP setting via Infix CLI is shown below (eth1 should get same config as eth0 got before).
root@R1:~$ cli
root@R1:exec> configure
[edit]
root@R1:configure> edit interfaces
[edit interfaces]
root@R1:configure> set interface eth1 ipv4 address 10.1.1.1 prefix-length 32
[edit interfaces]
root@R1:configure> set interface eth1 ipv4 forwarding true
[edit interfaces]
In OSPF, R1 needs to define eth1 as OSPF point-to-point link in FRR CLI.
R1# configure
R1(config)# interface eth1
R1(config-if)# ip ospf network point-to-point
Resulting OSPF configuration on R1
R1# show running-config
Building configuration...
Current configuration:
!
frr version 8.4.2
frr defaults traditional
hostname R1
no ip forwarding
no ipv6 forwarding
service integrated-vtysh-config
!
interface eth0
ip ospf network point-to-point
exit
!
interface eth1
ip ospf network point-to-point
exit
!
router ospf
network 10.1.1.1/32 area 0
exit
!
end
R1#
OSPF is able to establish routes to all routers. Here is the result at R1 (FRR CLI)
R1# show ip ospf route
============ OSPF network routing table ============
N 10.1.1.1/32 [0] area: 0.0.0.0
directly attached to lo
N 10.1.1.2/32 [10] area: 0.0.0.0
via 10.1.1.2, eth0
N 10.1.1.3/32 [10] area: 0.0.0.0
via 10.1.1.3, eth1
N 10.1.1.4/32 [20] area: 0.0.0.0
via 10.1.1.2, eth0
N 10.1.1.5/32 [20] area: 0.0.0.0
via 10.1.1.3, eth1
N 10.1.1.6/32 [30] area: 0.0.0.0
via 10.1.1.2, eth0
via 10.1.1.3, eth1
============ OSPF router routing table =============
============ OSPF external routing table ===========
r1#
Adding host subnets
In addition to the point-to-point links, the router likely has regular LAN subnets where hosts can be connected. In Figure 3, host subnets have been added to R1 and R6.
Figure 3: LAN broadcast networks added at R1 and R6
10.1.1.1 10.1.1.2 10.1.1.4 10.1.1.6
.----. .----. .----. .----.
.---. | R1 +-------+ R2 +-------+ R4 +-------+ R6 | .---.
|PC1| '+--+' '-+--' '--+-' '+--+' |PC2|
'-+-' |.1 \ | | / .1| '-+-'
.45| | \ | | / | |.26
---+----+-- \ | | / --+----+---
10.0.1.0/24 \ .-+--. .--+-. / 10.0.6.0/24
`---+ R3 +-------+ R5 +---´
'----' '----'
10.1.1.3 10.1.1.5
On R1, eth2 is configured with IP address 10.0.1.1/24.
root@R1:~$ cli
root@R1:exec> configure
[edit]
root@R1:configure> edit interfaces
[edit interfaces]
root@R1:configure> set interface eth2 ipv4 address 10.0.1.1 prefix-length 24
[edit interfaces]
root@R1:configure> set interface eth2 ipv4 forwarding true
[edit interfaces
and OSPF configuration in FRR CLI should be updated to advertise network 10.0.1.0/24. Note: The eth2 interface is kept as (regular) broadcast interface (not point-to-point).
R1# configure
R1(config)# router ospf
R1(config-router)# network 10.0.1.0/24 area 0
If corresponding setup is done on R6, OSPF would now establish the following routing table
R1# show ip ospf route
============ OSPF network routing table ============
N 10.0.1.0/24 [10] area: 0.0.0.0
directly attached to eth2
N 10.0.6.0/24 [40] area: 0.0.0.0
via 10.1.1.2, eth0
via 10.1.1.3, eth1
N 10.1.1.1/32 [0] area: 0.0.0.0
directly attached to lo
N 10.1.1.2/32 [10] area: 0.0.0.0
via 10.1.1.2, eth0
N 10.1.1.3/32 [10] area: 0.0.0.0
via 10.1.1.3, eth1
N 10.1.1.4/32 [20] area: 0.0.0.0
via 10.1.1.2, eth0
N 10.1.1.5/32 [20] area: 0.0.0.0
via 10.1.1.3, eth1
N 10.1.1.6/32 [30] area: 0.0.0.0
via 10.1.1.2, eth0
via 10.1.1.3, eth1
============ OSPF router routing table =============
============ OSPF external routing table ===========
r1#
Verify be letting PC1 ping PC2
PC1> ping 10.0.6.26
84 bytes from 10.0.6.26 icmp_seq=1 ttl=60 time=4.670 ms
84 bytes from 10.0.6.26 icmp_seq=2 ttl=60 time=4.054 ms
^C
PC1>
Done!