Solved: Regarding ICMP "Packet Too Big" message in 6PE RFC 4798

ksriram29 · ‎11-25-2008

Dear All,

The penultimate para in Section 3, Page 6 of 6PE RFC 4798 states the following:

"Otherwise, routers in the IPv4 MPLS network have the option to generate an ICMP "Packet Too Big" message using mechanisms as described in Section 2.3.2 .... of [RFC3032]"

As per RFC3032, the routers in the IPv4 MPLS network can generate ICMP "Time Exceeded" message or "Destination Unreachable because fragmentation needed and DF set" message.

Can someone please explain, how a IPv4 MPLS router will generate an ICMP "Packet Too Big"? This requires that the router in IPv4 MPLS network be a dual stack router to understand the IPv6 header under the label stack. Is my understanding correct?

If the router is an intermediate LSR, how will it know the path to the IPv6 destination even if it is dual stack router?

Thanks in advance.

Cheers,

Sriram

Harold Ritter · ‎11-27-2008

Sriram,

The "Packet too big" ICMPv6 message will never exceed 1280 bytes (including the IPv6 header), which is the minimum MTU that all interfaces should support as per RFC2460, Section 5. So as long as all your MPLS core interfaces support an MTU of at least 1280+label overhead, you should be fine. This should be a non issue in todays MPLS cores.

Regards

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

View solution in original post

Harold Ritter · ‎11-26-2008

Sriram,

The IPv4 core router does need to have code to understand IPv6 and create the "Packet too big" message and so on. It does not need to be configured for IPv6 though (dual stack).

The forwarding will be done based on the incoming label stack for that IPv6 packet, which means that the ICMPv6 message will be constructed with the source of the original packet as the destination IPv6 address, the top label for the incoming IPv6 packet will be swapped, prepended to the ICMPv6 message and forwarded through the egress interface. The egress PE will received this ICMPv6 message and forwarding appropriately.

The same mechanism is used when performing traceroutes in an L3VPN context, as the core routers have no knowledge of the address space being used by the L3VPN customers and couldn't otherwise forward ICMP messages to the proper source.

Regards

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

ksriram29 · ‎11-26-2008

Thanks Harold. The router generates ICMP "Packet Too Big" message When the IPv6 data packet's length exceeds the MTU. There is a possibility of ICMP "Packet Too Big" message size exceeding the interface MTU (if the IPv6 payload itself is more than the MTU). In that case the source will not be able to receive the ICMP message itself, since the ICMP message itself cannot be transmitted over the egress interface.

Hence the RFC recommends engineering the IPv4 MPLS core's MTU on the core facing interface of the 6PE router.

Please correct me if I am wrong.

Cheers,

Sriram

Harold Ritter · ‎11-27-2008

Sriram,

The "Packet too big" ICMPv6 message will never exceed 1280 bytes (including the IPv6 header), which is the minimum MTU that all interfaces should support as per RFC2460, Section 5. So as long as all your MPLS core interfaces support an MTU of at least 1280+label overhead, you should be fine. This should be a non issue in todays MPLS cores.

Regards

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México