How PHP works?

[ha_lh]

"PHP is used only for directly connected subnets or aggregate routes", is it true? Can you explain more clearly?

[Harold Ritter]

This is correct indeed.

The router directly connected to a given prefix or aggregating it advertises an implicit null label (value 3, per RFC3032) for that same prefix or for the aggregate route.

The penultimate hop router receives an implicit null label, which indicates taht it should pop the label before forwarding the packet to the next hop. If the label is at the bottom of the label stack, it is removed, the frame is forwarded and the next-hop receives an IP frame. If it is not, the label is removed, the frame is forwarded and the next-hop receives an MPLS frame.

Hope this helps,

[ha_lh]

But, why "used only for"?

Is that in all other cases, the L2 informations in LFIB or FIB are enough for properly forwarding packets? and no L3 lookup is necessary (no double look-up => no PHP needed)?

So, is this one of the reasons that we must enable CEF in all LSRs in MPLS domain? If it's not, why can't we use other Switching mechanism?

Thanks in advance,

[Harold Ritter]

"use only for" because it wouldn't make sense to pop the label before you get to the penultimate hop router.

CEF is a basic system requirement if you want to run MPLS. This is because the LFIB is built from the FIB (CEF) and the LIB (LDP).

Here's the life of a packet through an MPLS core.

Ingress LSR does a FIB lookup, does the label imposition and forward the packet.

P router does a TFIB lookup, does the label swapping and forward the packet.

Penultimate hop router (P router) does a TFIB lookup, does the label disposition and forward the packet.

Egress LSR does a FIB lookup and forward the packet (assuming you don't have another label in the packet for L3VPN or other)

Let me know if that answers your questions,

[ha_lh]

I got your idea with many thanks,