Why does self-ping go out of point-point/multipoint interfaces?

Unanswered Question
Nov 6th, 2007
User Badges:

Hi all

I'm trying to find a definitive answer on this one...

A lot of network engineers (myself included) when conducting loop testing on a /30 pe-ce wan connection (hdlc encapsulation) will ping from the pe the local interface of the pe, and given that the pe sees the loop and is up, up, the ping works a treat.


However, I'm wondering how this actually works...

Reason being, a show ip cef of the local interface ip, will show the "destination" as receive.

Now, if cef WAS being used, cef would realise that the packet was destined for a local interface on its own router.

So why would cef forward the icmp echo out of its own interface?

Is the locally generated ping only using process switching when sending out the icmp echo (with a sh ip ro <destination>

showing the destination is directly connected VIA the interface, so therefore sends icmp out of the interface).

Then when icmp packet is received (via the loop), cef running on the input interface checks and sees that the packet is indeed destined for itself?


I understand that hdlc is needed in order for the interface to be in an up, up (looped) state in order for the routing table to be populated

with the directly connected interface, and hence cef table is updated, but some clarification as to how the ping to loop actually works

would be greatly appreciated.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
paolo bevilacqua Tue, 11/06/2007 - 13:22
User Badges:
  • Super Gold, 25000 points or more
  • Hall of Fame,

    Founding Member

Hi Roy,


the issue of how pings to a local interface works, has also and 'historic' component. In most cases, where applicable to the interface HW, the implementation was for packets addressed to a local point-to-point interface , to be sent out, received on the other end, sent back, ditto for ping reply.

This was done especially because this way, ping success you would be always be positive indication that the circuit associated to the interface was actually working.

Admittedly, this approach has the drawback of making so that a misconfigured 'other end' would cause ping failure while in fact the local interface is up/up and perfectly functional.


I understand the reasoning you made about the dependencies with CEF, however the simpler case may be that for compatibility with the past, a decision was made not to change this behavior and likely it will never change.


Hope this helps, please rate post if it does!


Actions

This Discussion