Re: Re: [mpls-linux-general] I need some help to read the codes related with MPLS patches

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

On Tue, Nov 20, 2001 at 03:03:17AM +0900, =C1=A4=B1=E2=BF=B5 wrote:
> Thanks for your comment, first.
> And thanks for you and other peoples that=20
> make the great job about the MPLS-Linux.

Thank you.

> Before I ask some questions, I must apology about
> my lack of English.  Actually I read the RFCs related=20
> with MPLS and LDP. But when I looked at the code, MPLS-Linux,=20
> I can't figure out the implementation of LIB tables such like
> ILM, NHLFE, FTN as descirbed in RFC 3031.
> I think that the problem I've been had is not just for
> my case. Ofcourse I read the codes.=20

ILM =3D incoming label info (mpls_in_info)
NHLFE =3D outgoing label info (mpls_out_info)
FTN =3D FIB entries -> outgoing label info (via proto_data)
LIB =3D all of the above :-)

> Anyway, I wonder the meaning of the res->proto_data[AUX_PROTO_DATA_MPLS=
]=20
> in route.c as I posted the prevoius mail. It is hard to

Heh heh.  I could give you the litteral meaning 'C' meaning :-)

> follow the code with this because the location of set the
> information of code is hard to find.

It's not hard to find it just uses lots of abstraction.

> Is there anybody explain me about that?=20

mpls_bind_out2fec() is what places MPLS info in a FIB entry.
rt_set_nexthop() which is called from ip_route_output_slow() and
ip_route_output_slow() grabs the MPLS info from the FIB entry and adds
it to a dst_entry.  Every SKB that goes through the Linux stack needs a
dst_entry to control where it goes.

>=20
>=20
>=20
> -------------------------------------------------------
> Ki Young Jung (jj...@ic...)
> Information and Communications University(ICU)
> Broadband Network Laboratory(http://bnlab.icu.ac.kr)
> 58-4 Hwaam-dong, Yusong-gu, Taejon, 305-732, KOREA
> Tel)+82-42-866-6182, 016-370-1751 Fax)+82-42-866-6110=20
> -------------------------------------------------------
>=20
> ----- Original Message -----=20
> From: James R. Leu=20
> To: &Aacute;=A2=B4=A1=BE&acirc;=A2=AF&micro;=20
> Cc: mpl...@li...=20
> Sent: Tuesday, November 20, 2001 1:21 AM
> Subject: Re: [mpls-linux-general] I need some help to read the codes re=
lated with MPLS patches
>=20
>=20
> You're already using the best document there is, the code.  Keep diggin=
g
> through it, you'll figure it out.
>=20
> If you run into a road block, and have specific questions feel free to
> e-mail that list with it.  I or someone else will answer.
>=20
> Jim
>=20
> On Mon, Nov 19, 2001 at 11:58:51PM +0900, &Aacute;=A2=B4=A1=BE&acirc;=A2=
=AF&micro; wrote:
> > I've looked at the code, and i think that the binding point of mpls_o=
utput() is in route.c
> >=20
> > After patching mpls codes, in ./net/ipv4/route.c
> >=20
> > void rt_set_nexthop(struct rtable *rt, struct fib_result *res, u32 it=
ag)
> > {
> >       struct fib_info *fi =3D res->fi;
> > =20
> > =A2=AE&brvbar;
> > =20
> > #ifdef CONFIG_MPLS
> >       /*  JLEU: we need to do this here because both
> >       * ip_route_output_slow() and ip_route_output_slow()
> >       * us this to finish the next hop calculations
> >       */
> >       if(res->proto_data[AUX_PROTO_DATA_MPLS]) {
> >           struct mpls_out_info *moi;
> >           int i;
> > =20
> >           MPLS_DEBUG(("rt_set_nexthop: adding label to dst\n"));
> >           MPLS_DEBUG(("rt_set_nexthop: before - mtu =3D %d\n",
> >             rt->u.dst.pmtu));
> >           rt->u.dst.output =3D mpls_output;    /* HERE */
> >           rt->u.dst.dst_proto_data[AUX_PROTO_DATA_MPLS] =3D=20
> >             res->proto_data[AUX_PROTO_DATA_MPLS];
> >           moi =3D res->proto_data[AUX_PROTO_DATA_MPLS];
> > mpls_again:
> >           for(i=3D0;i<moi->moi_instruction_length;i++) {
> >             switch(moi->moi_instruction[i].mi_opcode) {
> >                 case MPLS_OP_FWD:
> >                   moi =3D (struct mpls_out_info*)
> >                     moi->moi_instruction[i].mi_data;
> >                   goto mpls_again;
> >                   break;
> >                 case MPLS_OP_PUSH:
> >                   rt->u.dst.pmtu -=3D 4;
> >                   break;
> >             }
> >           }
> >           MPLS_DEBUG(("rt_set_nexthop: after - mtu =3D %d\n",
> >             rt->u.dst.pmtu));
> >       }
> > =20
> > =A2=AE&brvbar;
> > =20
> > }
> > =20
> > If res->proto_data[AUX_PROTO_DATA_MPLS] have a value (not zero)
> > then the output related function is binded with mpls_output()
> > so the function is invoked.
> > =20
> > But I still try to find work that describes the structures or methods=
 of MPLS in Linux .=20
> > Is there anyone who know this? I hope to learn from him or her.
> > =20
> > Especially, I want to know the semantics of proto_data[AUX_PROTO_DATA=
_MPLS].
> > =20
> > Thanks, in advance.
> > =20
> > B.E.
> > =20
> > Kiyoung Jung
> > -------------------------------------------------------
> > Ki Young Jung (jj...@ic...)
> > Information and Communications University(ICU)
> > Broadband Network Laboratory(http://bnlab.icu.ac.kr)
> > 58-4 Hwaam-dong, Yusong-gu, Taejon, 305-732, KOREA
> > Tel)+82-42-866-6182, 016-370-1751 Fax)+82-42-866-6110=20
> > -------------------------------------------------------
> > -------------------------------------------------
> > DreamWiz Free Mail @ http://www.dreamwiz.com/
> > DreamSearch Click the world!!! http://search.dreamwiz.com/
> >=20
>=20
> --=20
> James R. Leu
>=20
>=20
>=20
> -------------------------------------------------
> DreamWiz Free Mail @ http://www.dreamwiz.com/
> DreamSearch Click the world!!! http://search.dreamwiz.com/
>=20

--=20
James R. Leu