Re: [mpls-linux-general] Some questions

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

In regards to the purpose of the MPLS tunnel interface, it is used as another
way to map traffic to a LSP.  You could create routes like:

    ip route add 10.0.0.0/8 dev mpls0

as opposed to

    ip route add 10.0.0.0/8 via 192.168.1.1 mpls 0x2

Same thing just a different technique for mapping.
It can be used for applications that cannot be modified
to bind routes directly to LSPs.

Also the "tunnel" paradigm is used by a couple of
popular router manufactures to represent a RSVP-TE LSP.

On Mon, Feb 18, 2008 at 03:12:29PM +0100, Javier Lorenzo wrote:
> James,
> Thank you for your answer.
> 
> 2008/2/16, James R. Leu <jl...@mi...>:
> > Javier,
> >
> > Thank you for trying MPLS Linux.  I've answered your questions in-line.
> > Please continue to send any questions you have to the mpls-linux-general
> > mailing list and I or others will do our best to answer your questions.
> >
> > On Fri, Feb 15, 2008 at 03:55:47PM +0100, Javier Lorenzo wrote:
> > > First of all, I'm working with mpls-1.950 on fc4 and I think these
> > > things aren't working well.
> > > I don't know if they work on newer releases.
> > >
> > > 1.
> > > I can add instructions to labelspace 0, but I can't to others:
> > > [root@localhost ~]# mpls ilm add label gen 100 labelspace 0
> > > instructions forward 0x2
> > > [root@localhost ~]# mpls ilm add label gen 100 labelspace 2
> > > instructions forward 0x2
> > > RTNETLINK answers: No such process
> >
> > I encourage you to upgrade to Fedora 7 or 8 and use the latest
> > packages available.
> >
> >     [root@mpls8-1 ~]# mpls ilm add label gen 100 labelspace 2
> >     [root@mpls8-1 ~]# mpls ilm add label gen 100 labelspace 0
> >     [root@mpls8-1 ~]# mpls ilm show
> >     ILM entry label gen 100 labelspace 0 proto ipv4
> >             pop peek  (0 bytes, 0 pkts)
> >     ILM entry label gen 100 labelspace 2 proto ipv4
> >             pop peek  (0 bytes, 0 pkts)
> >     [root@mpls8-1 ~]# uname -a
> >     Linux mpls8-1 2.6.23.15-137.fc8.mpls.1.960 #1 SMP Thu Feb 14 00:08:42 EST 2008 i686 i686 i386 GNU/Linux
> >     [root@mpls8-1 ~]#
> >
> 
> I don't mind if it don't work in this release, but I thougth it could
> help with debugging.
> If I do
> mpls ilm add label gen 200 labelspace 2
> it works well, but when I try to add some instructions:
> [root@localhost ~]# mpls ilm add label gen 200 labelspace 2
> instructions pop set-exp 0x3 forward 0x2
> RTNETLINK answers: No such process
> [root@localhost ~]# mpls ilm add label gen 200 labelspace 0
> instructions pop set-exp 0x3 forward 0x2
> [root@localhost ~]# mpls ilm add label gen 200 labelspace 2
> instructions pop set-exp 0x3 forward 0x2
> [root@localhost ~]# mpls ilm show
> ILM entry label gen 200 labelspace 2 proto ipv4
>         pop peek  (0 bytes, 0 pkts)
> ILM entry label gen 200 labelspace 0 proto ipv4
>         pop set-exp 3 forward key 0x00000002  (0 bytes, 0 pkts)
> 
> If I want to add the label to a labelspace per interface, and I want
> to do something (not the default pop), I get an error. If I add that
> label to the labelspace 0 before, I don't get the error, but the
> result is the default pop.
> 
> 
> > > 2.
> > > I have a problem when I try to add a label to a packet that already
> > > carries one (only one).
> > > In the LSR, I do:
> > > mpls nhlfe add key 0 instructions push gen 2000 nexthop eth0 ipv4 10.6.44.6
> > > mpls ilm add label gen 1000 instructions forward 0x2
> > >
> > > And in the ingress LER I do:
> > > mpls nhlfe change key 0x2 instructions push gen 500 push gen 1000
> > > nexthop eth0 ipv4 10.6.44.2
> > > or:
> > > mpls nhlfe change key 0x2 instructions push gen 1000 nexthop eth0 ipv4 10.6.44.2
> > >
> > > In the first case, there isn't any problems, but in the second, the S
> > > bit in the MPLS header
> > > with the label 2000 is set to 1. And it must be 0.
> > > I know that I can pop the label 1000 and push it again, solving the
> > > problem; but if I want
> > > to merge some LSP in that tunnel, I have to repeat the nhlfe line for each LSP.
> > > And I think it isn't the desired behaviour.
> >
> > push without pop is not supported and it against RFC 3031.  By doing
> > a push without pop your exposing nodes further along the LSP to labels
> > they have not assigned.
> >
> >        A ----> B -----> C
> >
> > In this picture the arrows represent the direction the packets are flowing.
> > Currently the only form of label allocation that is standardized is down
> > stream label allocation which mean the next hop peer assigns the label
> > on which a MPLS speaker can sent traffic.  In other words, B assigns labels
> > for which A can send traffic to B.  By doing a push without pop, you are
> > exposing some node beyond B to labels for which it did do assign.
> >
> > If you are trying to accomplish label stacking such as is found in MPLS BGP
> > VPNs or by-pass tunnels, then you will want to utilize the the 'forward'
> > instruction on a NHLFE to product layers of indirection.
> >
> > ILM (peek,pop,fwd) -> NHLFE (push,fwd) -> NHLFE (push,nexthop)
> >
> >
> 
> Ok. All right.
> 
> > > 3.
> > > I don't know how to use
> > > set-rx-if <NAME> forward <KEY>
> > > I tried this:
> > > [root@localhost ~]# mpls ilm add label gen 255 instructions set-rx-if
> > > eth0 forward 0x2
> > > RTNETLINK answers: Operation not permitted
> >
> > set-rx-if is used to set the interface you want to the kernel to believe
> > the packet arrived on.  It can be used in combination with the MPLS tunnel
> > interface to provide a bi-direction interface (MPLS tunnel interfaces are
> > uni-directional since they represent a uni-directional LSP).
> >
> > I believe set-rx-if must be the last instruction.
> >
> 
> I'm sorry. The command I wrote was right, but I forgot to set the
> interface into the labelspace. I've tried it with the tunnel interface
> and it works well. Thank you very much.
> 
> But I'm a bit confused with the tunnel interface. When or why is it useful?
> If I send traffic trought the mpls0 interface I have to do,
> mpls tunnel set dev mpls0 nhlfe KEY
> but I can forward traffic to that nhlfe key without the mpls0 interface.
> And when I use set-rx-if and the kernel thinks the traffic comes
> throught the mpls0 interface:
> if I'm going to do a MPLS forward I don't need the information;
> if I'm going to do a IP forward, I can do it with   ILM(pop,
> forward)-->NHLFE(nexthop)   or ILM(pop) and after an IP search.
> 
> > > Thank you.
> > > --
> > > Javier Lorenzo
> > >
> > > -------------------------------------------------------------------------
> > > This SF.net email is sponsored by: Microsoft
> > > Defy all challenges. Microsoft(R) Visual Studio 2008.
> > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> > > _______________________________________________
> > > mpls-linux-general mailing list
> > > mpl...@li...
> > > https://lists.sourceforge.net/lists/listinfo/mpls-linux-general
> >
> > --
> > James R. Leu
> > jl...@mi...
> >

-- 
James R. Leu
jl...@mi...