The MPLS WG Archive

[Jack Brennen <John.Brennen@marconi.com>]

> > But the problem has nothing to do with DoD mode.  It occurs in DU mode as well:
> >
> >   Router A                          Router B
> >   --------                          --------
> >
> >   Sends Label Mapping
> >    (Path Vector=<B,C,D,A>)
> >                                     Receives Label Mapping
> >   Router A gets a new next hop
> >    for the FEC
> >   Sends Label Mapping
> >    (Path Vector=<E,F,G,A>)
> >                                     Sends Label Release
> >                                      (loop detected)
> >   Receives Label Release
> >                                     Receives Label Mapping
> >
> > At the end of this scenario, Router B holds a label which he believes was
> > validly advertised by Router A.  Router A believes that Router B released
> > the label.
> 
> This is a smallish sort of window, which may be sorted by simply
> having Router A send a Label Withdraw when it receives packets
> that are labeled with the labels that Router B mistakenly thinks
> are valid.  I think that a robust LSR implementation will do
> that in any case.
> 

But Router A believes Router B to have released the label, so he may
reuse it for another purpose.  When Router B eventually puts traffic
onto the label (which may occur hours later), there is no reason to
believe that the label will still be unassigned.

And it is a smallish sort of window, but I'm not bringing this to the WG's
attention because of a theoretical concern.  This problem is being observed
in our lab under stress conditions.  It really happens in the real world.  :-(

Another way to look at this, which I hope *clearly* demonstrates the
intractable nature of this problem.  Consider these three protocol
conversations:

   Conversation 1                           Conversation 2
   --------------                           --------------

   Router A            Router B             Router A        Router B
   --------            --------             --------        --------
   Send LMap                                Send LMap
                       Recv LMap                            Recv LMap
                       Send LRel                            Send LRel
   NHOP Changes                             Recv LRel
   Send LMap                                NHOP Changes
   Recv LRel                                Send LMap
                       Recv LMap                            Recv LMap


   Conversation 3
   --------------

   Router A            Router B
   --------            --------
   Send LMap
                       Recv LMap
   NHOP Changes
   Send LMap
                       Recv LMap
                       Send LRel
   Recv LRel

These conversations are all legal according to the LDP spec,
and all are reasonably likely to happen.  Note that Router B can't
tell the difference between Conversations 1 & 2, and Router A can't
tell the difference between Conversations 1 & 3.

  Conversation 1:  A ends up with no label allocated, B has a label
  Conversation 2:  A and B both end up with a label allocated
  Conversation 3:  A and B both end up with no label allocated

Clearly, Conversation 1 leads to problems, but neither Router A nor
Router B can distinguish this from the valid Conversations 2 & 3.

This is a deficiency in the protocol, and we can brainstorm any number
of methods to work around the problem, but the problem is there,
and should be addressed.

   Jack