Great! Thanks for clearing that up for me. -Andre
----- Original Message ----- From: "Simon Horman" <ho...@verge.net.au> To: "Andre Khan" <andrek...@vmware.com> Cc: dev@openvswitch.org, "Isaku Yamahata" <yamah...@valinux.co.jp>, "Ravi K" <rke...@gmail.com>, "Ben Pfaff" <b...@nicira.com> Sent: Wednesday, January 23, 2013 11:00:11 PM Subject: Re: [ovs-dev] [PATCH 01/16] User-Space MPLS actions and matches On Wed, Jan 23, 2013 at 10:31:52PM -0800, Andre Khan wrote: > > Hi Simon, > > Regarding this part in your mail: > > >> push_mpls_lse() isn't documented as putting the new MPLS label before or > >> after the existing labels, but it appears to me that it puts the new > >> label after the existing labels. If I'm right about that, it's > >> surprising; one would ordinarily expect a new header to be an outermost > >> one, not an innermost one. > > >I believe that it pushes the new MPLS stack entry onto the MPLS stack > >before any existing entries. > > The newly pushed MPLS label is the "outermost" label (which is what I believe > Ben alluded to). > Typically this label is for the benefit of the MPLS Nexthop to which it has > local significance. > > I just started to scan this list, so forgive me if I lack the full context of > this thread and missed something... besides it is getting late here :) > > Thanks, > -Andre > > PS: I would suggest that that location of the label be signified with the > words "inner/outer" rather than "before/after" as it removes any ambiguity... > not to mention that is the MPLS speak :) Thanks. I believe that the behaviour of the code is consistent with what you describe. How about this updated comment for push_mpls_lse(). /* Push an new MPLS stack entry onto the MPLS stack * and adjust packet->l2 and packet->l2.5 accordingly. * The new entry will be the outermost entry on the stack. */ _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
