BGP FreeCore & 6PE

[Guillaume Gryszata]

Hi,

Before my IPv6 lab, I thought that 6PE had the same behavior than BGP FreeCore MPLS for IPv4.

I expected that the incoming IP packet on the ingress PE was labeled with just the IGP label. Forwarded to the egress PE and that the egress PE did a IPv6 lookup to forward the packet to the remote CE (exctly like BGP FreeCore MPLS).

It seems that in 6PE there is 2 labels (1 MP-BGP label and 1 IGP label). So I think in this case that there is IPv6 lookup on the egress PE (just forwarded with LFIB) ?

My question is why is there 2 different behavior ? Is it possible to do BGP FreeCore for IPv6 ? Is it really different that 6PE (or am I wrong)?

Thank you

[amit.bhagat]

Hi There,

You understanding is correct about 6PE solution using 2 label- one (bottom) label advertised by egress 6PE router for IPv6 prefixes and second (top) label for IGP within the MPLS core to reach the egress 6PE router. The ingress 6PE router forwards the traffic for remote IPv6 prefix over the LSP that is created for egress 6PE router.

RFC 4798 discusses 6PE solution in detail.

The fact that 6PE uses two labels has to do with LDP not available for IPv6. At present, LDP can only assign labels for IPv4 prefixes.

6PE cannot run without MP-BGP. MP-BGP uses MP_REACH_NLRI to advertise the label and the IPv6 prefix. The compatibility is notified during BGP session with AFI/SAFI = 2/4.

Hope this helps.

Regards,

Amit.

[Laurent Aubert]

Hi,

You can also think about it this way:

Your core is IPv4 aware only so with only one label in the stack, the PHP router will remove this label and your egress PE will receive an IPv6 packets on an interface which support IPv4 only.

HTH

Laurent.

[Guillaume Gryszata]

Thank you for your answer.

Yes IP lookup seems not possible if an IPv6 packet is received on the egress PE core interface.

I tried to configure 6PE for IPv4 (a kind of 4PE) = BGP Freecore with send-label in the purpose to suppress the IPv4 lookup on the egress PE router but it seems not work (no label for MP-BGP IPV4 prefix). Moreover there is not afi/safi 1/4 on my configuration (I think mandatory for this case). Is this configuration possible ?

Thank you

[Laurent Aubert]

It should work. Be sure all your iBGP neighbors negociated the MPLS capability (neighbor send-label) and if you have any outbound route-map, add the set mpls-label as well.

Here is an example: PE3 is receiving from an eBGP neighbor 3.1.1.1/32 and send it to PE4 with a label:

AS1-PE3#sh ip bgp labels

   Network          Next Hop      In label/Out label

   3.1.1.1/32       172.16.33.1     25/nolabel

AS1-PE3#sh mpls forwarding-table

Local  Outgoing      Prefix            Bytes Label   Outgoing   Next Hop   

Label  Label or VC   or Tunnel Id      Switched      interface             

...

25     No Label      3.1.1.1/32        0             Et1/0.15   172.16.33.1

Now we can check the cef entry for this prefix on PE4:

AS1-PE4#sh ip cef 3.1.1.1 internal

3.1.1.1/32, epoch 0, RIB[B], refcount 5, per-destination sharing

  sources: RIB

  feature space:

   LFD: 3.1.1.1/32 0 local labels

        contains path extension list

        label switch chain 0x03BBEA00

   IPRM: 0x00028000

  ifnums:

   Ethernet0/0(2): 192.168.2.13

  path 03BF3CC8, path list 03BE9E78, share 1/1, type recursive, for IPv4

    MPLS short path extensions: MOI flags = 0x4 label 25

  recursive via 10.1.1.3[IPv4:Default] label 25, fib 03C37A70, 1 terminal fib

    path 03BF3DB8, path list 03BE9F18, share 1/1, type attached nexthop, for IPv4

      MPLS short path extensions: MOI flags = 0x0 label 40

    nexthop 192.168.2.13 Ethernet0/0 label 40, adjacency IP adj out of Ethernet0/0, addr 192.168.2.13 044B7FD8

  output chain: label 25 label 40 TAG adj out of Ethernet0/0, addr 192.168.2.13 044B7E80

You can see the IGP and BGP tags added to the packet.

HTH

Laurent.

[Guillaume Gryszata]

Hi,

Thank you for your very interesting answer.

Guillaume