[RAM] Re: question about LISP v1

[Dino Farinacci <dino at cisco.com>]

> How does an initiator node informs the receiver about its multiple  
> RLOCs?

This has not changed from the -00 draft. the receiver can glean *a*  
locator from the packet it is decaping (the ETR that is). If the ETR  
wants the complete set of locators with their associated priorities  
and weights, it must either 1) send a packet returning with the outer  
DA equal to the inner DA (I call this a data probe) or 2) if there is  
a mapping service in place, do a lookup to get the mapping.

For the later, if LISP-NERD is used, the lookup is local the the ITR  
since the mappings are pushed. If LISP-CONS is used, a Map-Request is  
sent by the ITR to it's configured CAR over a TCP connection.

> I mean, in the described packet flow sequence, the initiator sends  
> a LISP data packet containing its EID and RLOC in the source  
> address fields of the inner and outer header, and this is how the  
> receiver ETR learns about the EID-to-RLOC mapping of the initiator.  
> However, this only conveys a single RLOC, what happens if the  
> initiator has multiple RLOCs (the most common in a multihomed site)  
> and wants to inform the receiver ETR about them?

Right, the ETR has some flexibility here. It can use the single  
locator to return packets while in parallel send a Map-Request to get  
the rest of them. When the ITR is sending packets this ETR, the  
Locoator Reachability bits are continually updated for each packet  
the ITR encaps to the ETR. So the ETR will know the reachability status.

There is a detail about which loc-reach-bit to associated with the  
outer SA but I can tell you about that offline.

> In the previous version of LISP, it was possible for the initiator  
> to send an unsolicited Map-Reply message containing the whole RLOC  
> set to the receiver.

No, it was never unsolicited, because the sender of the Map-Reply  
would not know where to send it.

> however, the current version includes the nonce, so Map-reply  
> messages can only be sent as replies to data packets or map request  
> messages. So, does this means that

The -01 draft uses both the fact that the Nonce much match as well as  
the ITR keeping a local cached entry for the EID in a "incomplete"  
state. Similar to how some ARP implementations are coded.

> the initiator cannot inform spontaneously of its locator set and  
> must wait for the receiver to ask for it? or there is a mean for  
> the initiator to send the locator set to the receiver even if the  
> receiver haven't asked for it?

The Map-Reply is never sent unsolicited.

Dino

_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram