mpls-linux-general — Mailing list for general discussion about MPLS for Linux

Re: [mpls-linux-general] Binding to FEC.

[James R. L. <jl...@mi...>]

On Thu, Jan 31, 2002 at 12:13:52PM +0200, Asviyan, Anatoly wrote:
> 	Hi
> 1)
> I can't make latest MPLS work:
> when I run this command
> > iptables -A OUTPUT -d 192.168.1.0/24 -j MPLS --set-key 0x2
> iptables complains about inkown chain/target/match or smth like that
> I checked that it indeed loads extension/libipt_MPLS and that 0x2 key does
> exist in mpls_out

I could be that the kernel is rejecting the target (MPLS).  Double check
your kenel config.  make sure you have:

Code maturity level options  --->
  [*] Prompt for development and/or incomplete code/drivers
Networking options  --->
  [*] Network packet filtering (replaces ipchains)
  [*] Multi Protocol Label Switching - MPLS
  IP: Netfilter Configuration  --->
    <*> IP tables support (required for filtering/masq/NAT)
    <*>   Packet mangling
    <*>     MPLS target support

NOTE!!! You need to turn on "Multi Protocol Label Switching" before
you'll see "MPLS target support".

Jim

> 
> my configurations is: patched iptables 1.2.4, patched kernel 2.4.17 and
> mpls-linux-1.126
> 
> Any ideas? Thanks
> 
> 2) 
> can somebody that has working mpls+iptables give exact instruction for
> mplsadm2 and iptables 
> how to connect 2 PCs (as in some famouse README :-) 
> Thanks
> 
> -- 
> Regards, Anatoly
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

Re: [mpls-linux-general] Label stack question.

[Radu D. <dr...@ee...>]

Apologize for the for the previous question.
The MPLS label stack encapsulation over ATM  DOES support a label stack of
level >2
So please ignore my previous email.

Thanks,
Radu



> Hi all,
>
> AFAIK the MPLS encapsulation over ATM (or FrameRelay) media suports a
max.
> level 2 stack.
>
> Then, what if in a hierarchical network an MPLS label with a label stack
of
> level m >2 enters to an ATM cloud?
>
> Thanks in advance,
> Radu
>
>
>
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

[mpls-linux-general] Label stack question.

[Radu D. <dr...@ee...>]

Hi all,

AFAIK the MPLS encapsulation over ATM (or FrameRelay) media suports a max.
level 2 stack.

Then, what if in a hierarchical network an MPLS label with a label stack of
level m >2 enters to an ATM cloud?

Thanks in advance,
Radu

[mpls-linux-general] problems using netfilter

[<due...@cs...>]

Hi all,

we have encountered some problems using
the mpls-iptables patch. We wanted to use
netfilter at the ingress node to bind labels
to specific ip traffic.
Here's what we did:
- first patched linux kernel 2.4.14 with mpls-linux-0.993
- second applied mpls-iptables-0.3 patch (.full patch)
- made menuconfig ("netfilter MPLS match support",
"Packet mangling/MPLS target support" + the "normal"
MPLS stuff)
- patched and compiled iptables 1.2.2

Excerpt from our configuration script:

...
mplsadm -v -A -O gen:33:eth1:ipv4:192.168.2.3

KEY3=`grep 'gen 33' /proc/net/mpls_out | cut -d' ' -f1`

iptables -v -I PREROUTING -t mangle -d 192.168.6.1 \
-s 192.168.5.2 -j MPLS --set-mpls 0x$KEY3


This all works so far (giving no error messages
from mplsadm or iptables), but the traffic seems
to be forwarded normally using the ip routing table,
although the packet counter of iptables shows that
the rule matched.

Has anybody experienced similar problems?
Which combination of the linux kernel/patches is known to be 
working best?


Thanks,

Uli & Daniel

RE: [mpls-linux-general] Binding to FEC.

[Asviyan, A. <ana...@te...>]

	Hi
1)
I can't make latest MPLS work:
when I run this command
> iptables -A OUTPUT -d 192.168.1.0/24 -j MPLS --set-key 0x2
iptables complains about inkown chain/target/match or smth like that
I checked that it indeed loads extension/libipt_MPLS and that 0x2 key does
exist in mpls_out

my configurations is: patched iptables 1.2.4, patched kernel 2.4.17 and
mpls-linux-1.126

Any ideas? Thanks

2) 
can somebody that has working mpls+iptables give exact instruction for
mplsadm2 and iptables 
how to connect 2 PCs (as in some famouse README :-) 
Thanks

-- 
Regards, Anatoly

Re: [mpls-linux-general] Binding to FEC.

[James R. L. <jl...@mi...>]

Sorry about that I gave you the wrong command line.

iptables -A OUTPUT -d 192.168.1.0/24 -j MPLS --set-key 0x2
          ^

Make sure you patched iptables, then do a full make install.
iptables must be run as root (it needs to load dynamic libraries which 
are places in your hosts files system via 'make install')

-A OUTPUT adds this rule to the OUTPUT netfilter.
--set-key <key> is what bind the packet going to the destination to
the outgoing label.  In the above example 0x2 is the key (found by looking
in /proc/net/mpls_output) for the outgoing label you want to use.

I hope this helps.

Jim

PS There is one other person who is using the diffserv instructions.
If you would like I could include him in the e-mails and he might be
able to help out.  Also, he has been finding lots of bugs, so make sure to
ask when things that 'should' work, aren't.

On Wed, Jan 30, 2002 at 12:39:28AM +0800, ncc49a wrote:
> Hi James,
> 
>     In a previous posting, you provided me with the following instruction to bind to an FEC.
> 
> >> iptables -O OUTPUT -d 192.168.1.0/24 -j MPLS --set-key 0x2.
> 
>     However, I was unable to execute this command. There were errors with the -O OUTPUT part.. May I know what does -O OUTPUT, and --set-key 0x2 do? How can a particular outgoing generic label of say value 16 be binded to the FEC 192.168.1.0/24?
> 
>     Thanks a lot for any help provided!
> 
> Paul

-- 
James R. Leu

[mpls-linux-general] Binding to FEC.

[ncc49a <nc...@si...>]

Hi James,

    In a previous posting, you provided me with the following =
instruction to bind to an FEC.

>> iptables -O OUTPUT -d 192.168.1.0/24 -j MPLS --set-key 0x2.

    However, I was unable to execute this command. There were errors =
with the -O OUTPUT part.. May I know what does -O OUTPUT, and --set-key =
0x2 do? How can a particular outgoing generic label of say value 16 be =
binded to the FEC 192.168.1.0/24?

    Thanks a lot for any help provided!

Paul

Re: [mpls-linux-general] Some Simple questions

[James R. L. <jl...@mi...>]

What version do you have?  Did you grab the latest version via CVS?

Jim

On Wed, Jan 30, 2002 at 07:18:17AM -0800, lingdong li wrote:
> 
> --- "James R. Leu" <jl...@mi...> wrote:
> > It means your routing tables don't have routes that
> > point at each other.
> > 
> > Jim
> Just as follow:
> 
>   dummy0 --------- eth0        eth0 -------- dummy0
>    <-----| LER 1 |------------------| LER B |---->-   
>      ---------                  --------
> First I set up the Devices.
> 
> On Ler 1:
> 	
> 	ifconfig eth0 10.0.0.1 netmask 255.255.255.0
> broadcast 10.0.0.255 up
> 	ifconfig dummy0 1.2.3.4
> 
> On Ler 2:
> 	
> 	ifconfig eth0 10.0.0.2 netmask 255.255.255.0
> broadcast 10.0.0.255 up
> 	ifconfig dummy0 2.3.4.5
> 
> Second I add routes to virtual networks.
> 
> On Ler 1:
> 	route add -net 78.1.1.0/24 gw 1.2.3.4
> 	route add -net 78.1.2.0/24 gw 1.2.3.4
> 
> 	route add -net 79.1.1.0/24 gw 10.0.0.2
> 	route add -net 79.1.2.0/24 gw 10.0.0.2
> 
> On Ler 2:
> 	route add -net 79.1.1.0/24 gw 2.3.4.5
> 	route add -net 79.1.2.0/24 gw 2.3.4.5
> 
> 	route add -net 78.1.1.0/24 gw 10.0.0.1
> 	route add -net 78.1.2.0/24 gw 10.0.0.1
> 
> Then I make shure the right labelspaces are
> associated:
> 
> on Ler 1 and Ler 2:
> 
> 	cd {your mpls dir}/utils
> 	./mplsadm -L eth0:0
> 
> Then start Ldp on both LERs:
> 
> On LER 1:
> 
> 	./ldp_linux
> 	add global 10.0.0.1
> 	set trace 0xFFFFFFFF
> 	add interface eth0
> 
> On LER 2:
> 
> 	./ldp_linux
> 	add global 10.0.0.2
> 	set trace 0xFFFFFFFF
> 	add interface eth0
> Now I can not find information from mpls_*,except a
> line in mpls_in.
> 
> And I find no MPLS packet have been
> transmitted.Whether without mpls_linux,the Ldp_linux
> only implement the label distribute?
> 
> And I find only "hello" and "Keepalive" packet.Did the
> LSP establish?
> 
> Li
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Great stuff seeking new owners in Yahoo! Auctions! 
> http://auctions.yahoo.com
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

Re: [mpls-linux-general] Some Simple questions

[lingdong li <lin...@ya...>]

--- "James R. Leu" <jl...@mi...> wrote:
> It means your routing tables don't have routes that
> point at each other.
> 
> Jim
Just as follow:

  dummy0 --------- eth0        eth0 -------- dummy0
   <-----| LER 1 |------------------| LER B |---->-   
     ---------                  --------
First I set up the Devices.

On Ler 1:
	
	ifconfig eth0 10.0.0.1 netmask 255.255.255.0
broadcast 10.0.0.255 up
	ifconfig dummy0 1.2.3.4

On Ler 2:
	
	ifconfig eth0 10.0.0.2 netmask 255.255.255.0
broadcast 10.0.0.255 up
	ifconfig dummy0 2.3.4.5

Second I add routes to virtual networks.

On Ler 1:
	route add -net 78.1.1.0/24 gw 1.2.3.4
	route add -net 78.1.2.0/24 gw 1.2.3.4

	route add -net 79.1.1.0/24 gw 10.0.0.2
	route add -net 79.1.2.0/24 gw 10.0.0.2

On Ler 2:
	route add -net 79.1.1.0/24 gw 2.3.4.5
	route add -net 79.1.2.0/24 gw 2.3.4.5

	route add -net 78.1.1.0/24 gw 10.0.0.1
	route add -net 78.1.2.0/24 gw 10.0.0.1

Then I make shure the right labelspaces are
associated:

on Ler 1 and Ler 2:

	cd {your mpls dir}/utils
	./mplsadm -L eth0:0

Then start Ldp on both LERs:

On LER 1:

	./ldp_linux
	add global 10.0.0.1
	set trace 0xFFFFFFFF
	add interface eth0

On LER 2:

	./ldp_linux
	add global 10.0.0.2
	set trace 0xFFFFFFFF
	add interface eth0
Now I can not find information from mpls_*,except a
line in mpls_in.

And I find no MPLS packet have been
transmitted.Whether without mpls_linux,the Ldp_linux
only implement the label distribute?

And I find only "hello" and "Keepalive" packet.Did the
LSP establish?

Li


__________________________________________________
Do You Yahoo!?
Great stuff seeking new owners in Yahoo! Auctions! 
http://auctions.yahoo.com

Re: [mpls-linux-general] Some Simple questions

[James R. L. <jl...@mi...>]

It means your routing tables don't have routes that point at each other.

Jim

On Wed, Jan 30, 2002 at 01:57:03AM -0800, lingdong li wrote:
> Hi,All,
>  when I run ldp_linux without run mplsadm firstly,I
> find there are no mpls packets transmitted between 2
> hosts.Is it true? 
> By ethereal,it shows "Hello", "Keeplive".Does ldp
> session establish?.
> In the mpls_in(/proc/net/),there is only 1 line
>     40004400 0/0/0 16 0 pop peek
> all other mpls_* are blank.That means no label
> bindings have been established between the 2 LSRs
> right? 
>   I guess I must have misunderstood something. I will 
> appreciate all help/advice provided. 
> 
> Best Regard 
> 
> Li
> 
> 
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Great stuff seeking new owners in Yahoo! Auctions! 
> http://auctions.yahoo.com
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

Re: [mpls-linux-general] A problem

[James R. L. <jl...@mi...>]

kernel.org is what I use.  But there are thousands of mirrors all over
the world.  Go to www.kernel.org with a web broser and look at the
mirror list.

Jim

On Wed, Jan 30, 2002 at 03:28:49AM -0800, farahmand zahra wrote:
> Hi,
> I have Redhat 7.0, but the example is implemented on a
> linux 2.4.x.
> I have problem to get linux 2.4.x,I tried to get
> kernel source from www.kernel.org but the source that
> I download is not complete.
> where can I get complete kernel scource?
> I know my english is poor.I am sorry.
> I appreciate your help.
> Sincerely yours.
> O.Jaberi.
>  
> 
> __________________________________________________
> Do You Yahoo!?
> Great stuff seeking new owners in Yahoo! Auctions! 
> http://auctions.yahoo.com
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

Re: [mpls-linux-general] (no subject)

[James R. L. <jl...@mi...>]

Ports below 1023 are can only be used by root.  Run ldp_linux as
root and you sjould have no problem.

Did you grab ldp-portable via CVS?  If not, you're wasting your time.

Jim


On Wed, Jan 30, 2002 at 08:12:19PM +0800, zjh...@so... wrote:
> hi,all
> Can the ldp_portable work without mpls-linux?
> I have tried this between to PCs,but not until I modified the port from 646 to 8000 ,each PC can't received the UDP hello message from other,the two PCs is connected  by ethernet .
> I tried to use the cmd:add peer 192.168.199.167(IP ADDRESS of the other PC),my PC send the HELLO 
> MSG to 192.168.199.167:646.But received the ICMP message:the port is unreachable.
> Can i must modified the kernal file : /etc/service ane /etc/xinetd.conf to solve the problem?
> If so ,how can i do this?
> all thanks!!
> 
> 
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

[mpls-linux-general] (no subject)

[<zjh...@so...>]

hi,all
Can the ldp_portable work without mpls-linux?
I have tried this between to PCs,but not until I modified the port from 646 to 8000 ,each PC can't received the UDP hello message from other,the two PCs is connected  by ethernet .
I tried to use the cmd:add peer 192.168.199.167(IP ADDRESS of the other PC),my PC send the HELLO 
MSG to 192.168.199.167:646.But received the ICMP message:the port is unreachable.
Can i must modified the kernal file : /etc/service ane /etc/xinetd.conf to solve the problem?
If so ,how can i do this?
all thanks!!

[mpls-linux-general] A problem

[farahmand z. <z_c...@ya...>]

Hi,
I have Redhat 7.0, but the example is implemented on a
linux 2.4.x.
I have problem to get linux 2.4.x,I tried to get
kernel source from www.kernel.org but the source that
I download is not complete.
where can I get complete kernel scource?
I know my english is poor.I am sorry.
I appreciate your help.
Sincerely yours.
O.Jaberi.
 

__________________________________________________
Do You Yahoo!?
Great stuff seeking new owners in Yahoo! Auctions! 
http://auctions.yahoo.com

[mpls-linux-general] about LDP_LINUX port with out MPLS_linux

[<zjh...@so...>]

hi,all
  I have installed LDP_PORTABLE on my pc.I have succeed on compiling the program.But
 when I run it between two PCs,the program on each PC can send the UDP HELLO message.
 But the message can't be received and so there is no session established!
What is the reason?
  I have tried to modified the TCP/UDP port used by LDP from 646 to 8000.then the session 
 have been established and everything is OK! 
I want to know can i make the program run OK under Linux with out MPLS_LINUX using th port 646?
Must I modified the *.conf in /etc/...? If so ,How?
 all thanks!!
 jianhui zhang 

[mpls-linux-general] Some Simple questions

[lingdong li <lin...@ya...>]

Hi,All,
 when I run ldp_linux without run mplsadm firstly,I
find there are no mpls packets transmitted between 2
hosts.Is it true? 
By ethereal,it shows "Hello", "Keeplive".Does ldp
session establish?.
In the mpls_in(/proc/net/),there is only 1 line
    40004400 0/0/0 16 0 pop peek
all other mpls_* are blank.That means no label
bindings have been established between the 2 LSRs
right? 
  I guess I must have misunderstood something. I will 
appreciate all help/advice provided. 

Best Regard 

Li




__________________________________________________
Do You Yahoo!?
Great stuff seeking new owners in Yahoo! Auctions! 
http://auctions.yahoo.com

[mpls-linux-general] BETA: mpls-linux 1.127

[James R. L. <jl...@mi...>]

A new beta is availble for download via CVS (tag mpls-linux_1_127).

Bug fixes
Locking
Better refence counting
Layer 2 Cross Connects (ethernet, port mode)

NOTE: There are a LOT of changes in this code.  None of these changes
are documented.  Look back in the archive to see the previous mpls-linux 1.1
annoucements form more info.  (if someone wants to write a CHANGES doc
from 1.0 to 1.1 that would be great!!)

The big change in this version is L2CCs.  To use:

Ingress LER:

mplsadm2 -A -O gen:16:eth1:ipv4:11.0.1.2
mplsadm2 -B -O key:0x2 -2 eth:eth2:port

Egress LER:

mplsadm2 -L eth1:0
mplsadm2 -A -I gen:16:0
mplsadm2 -B -I gen:16:0 -2 eth:eth3:port

The result is that all packet arriving on the ingress port (eth2)
get shipped across the LSP, and sent egress port (eth3) unchanged.

Do not enable any L3 protocols on the ingress interface.  Just do an 
'ifconfig eth2 up'.

What is this for?  Read draft-martini and you'll see where I'm going.

Jim

-- 
James R. Leu

Re: [mpls-linux-general] Integration with Zebra?

[James R. L. <jl...@mi...>]

No one has worked with the zebra porting layer for sometime.

So the error you are seeing are probably from code rot.

With in the next couple of weeks I will be shifting my attendiong away fr=
om
mpls-linux, and onto integrating MPLS services into zebra.  The first ste=
p
(which is about 50% complete) is to add static LSP commands to zebra.
After that is finished I will work on ldp-portable integration.

Jim


On Mon, Jan 28, 2002 at 11:10:10AM -0800, Sanjeev Chakravarty wrote:
> Hi Jim,
>=20
> I could compile ldp_zebra after making a number of changes:
>=20
> 1. inet_aton declaration in /usr/include/arpa/inet.h was causing a
> conflict(had to comment it)
> 2. In many zldp_* files, I had to change the following definition
> 	#define dprintf(args...) deprintf(dep_impl_fib, "FIB: " ## args)
> to 	#define dprintf(args...) deprintf(dep_impl_fib, args)
> though syntactically I couldn't see what's wrong. The error message was
> "nothing can be pasted after this token"
> 3. There are references to functions defined in mpls-linux/utils/netlin=
k.c
> and so I=20
>  copied the netlink.* to the directory containing zldp_* files and upda=
te
> the Makefile to include netlink.o
>=20
> Just wondering if it's just me and Sehla who are facing this problem or=
 we
> have missed something,
>=20
> thanks,
>=20
> Sanjeev
>=20
> -----Original Message-----
> From: sehla loussaief [mailto:seh...@ya...]
> Sent: Monday, January 28, 2002 3:11 AM
> To: Sanjeev Chakravarty
> Subject: RE: [mpls-linux-general] Integration with Zebra?
>=20
>=20
> Hi Sanjeev,
> yes i had some compilation errors, when i compile the
> ldp with 'ldp_zebra" the compilation failed but when i
> compile it whith "ldp_linux" it succed !!
>=20
> Sehla
>  --- Sanjeev Chakravarty <Sa...@co...>
> a =E9crit : > Hi Sehla,
> >=20
> > No, I haven't tried that. Plan to do that this
> > weekend. Are you getting
> > compilation problems?
> >=20
> > Sanjeev
> >=20
> > -----Original Message-----
> > From: sehla loussaief [mailto:seh...@ya...]
> > Sent: Friday, January 25, 2002 12:54 AM
> > To: Sanjeev Chakravarty
> > Subject: RE: [mpls-linux-general] Integration with
> > Zebra?
> >=20
> >=20
> > Hi Sanjeev,
> > Have you tested zebra with ldp-portable0.08?=20
> >=20
> > --- Sanjeev Chakravarty <Sa...@co...>
> > a
> > =E9crit : > Hi Sehla,
> > >=20
> > > Yeah it is. Check it out at www.zebra.org.
> > >=20
> > > Sanjeev
> > >=20
> > > -----Original Message-----
> > > From: sehla loussaief [mailto:seh...@ya...]
> > > Sent: Thursday, January 24, 2002 12:14 AM
> > > To: Sanjeev Chakravarty
> > > Subject: Re: [mpls-linux-general] Integration with
> > > Zebra?
> > >=20
> > >=20
> > > Hi,
> > > Is zebra an open source? if yes where can i get
> > it?
> > >=20
> > > thanks in advance.
> > >=20
> > >  --- Sanjeev Chakravarty
> > > <Sa...@co...>
> > > a =E9crit : >=20
> > > > Hi Folks,
> > > >=20
> > > > Since mpls-linux code is fairly robust, I was
> > > > wondering if there are any
> > > > thoughts of integrating it with other open
> > source
> > > > routing protocol suite
> > > > like Zebra. To me both the software are
> > > > complementary. Some of the pros and
> > > > cons:
> > > >=20
> > > > pros:
> > > > 1. mpls-linux will be more versatile
> > > > 2. creation of VPNs will now be possible
> > > > 3. use of BGP to exchange labels needed in case
> > of
> > > > VPNs would be possible
> > > > 4. RSVP-TE or other extensions of MPLS would be
> > > > easier to implement
> > > >=20
> > > > cons:
> > > > 1. kinda vendor specific (Cisco like)
> > > implementation
> > > > 2. some dependency may arise
> > > >=20
> > > > look forward to hearing your comments,
> > > >=20
> > > > Sanjeev
> > > >=20
> > > > _______________________________________________
> > > > mpls-linux-general mailing list
> > > > mpl...@li...
> > > >
> > >
> >
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general
> > >=20
> > >=20
> > >
> >
> ___________________________________________________________
> > > Do You Yahoo!? -- Une adresse @yahoo.fr gratuite
> > et
> > > en fran=E7ais !
> > > Yahoo! Mail : http://fr.mail.yahoo.fr=20
> >=20
> >
> ___________________________________________________________
> > Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et
> > en fran=E7ais !
> > Yahoo! Mail : http://fr.mail.yahoo.fr=20
>=20
> ___________________________________________________________
> Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en fran=E7ais !
> Yahoo! Mail : http://fr.mail.yahoo.fr

--=20
James R. Leu

Re: [mpls-linux-general] default entry in mpls forwarding table

[James R. L. <jl...@mi...>]

See my comments within:

On Mon, Jan 28, 2002 at 09:08:41PM +0530, Abhijit Gadgil wrote:
> anish  verma wrote :
> 
> >
> >hi jim,
> >i have some queries.. please help me
> >
> >1.  is it possible to forward packets to some default entry through
> >mpls-linux as it done in ip-network..
> 
> I can answer this query on Jim's behalf ;-)

Thank you.  You did a fine job of answering it as well :-)

> Binding labels to default entry is not a good thing. The reason is as
> follows.
> 
> 1. The labels bound to default entry will have to leave the MPLS domain
> through a 'default egress' router. As the fec, which is '0.0.0.0' in this
> case has to be uniquely interpreted throughout the domain.
> 
> eg. In following figure 'd is the default egress'
> 
> a-----b------c------d
>       |
>       |
>       e------f------g
> 
> Now assume that a label is bound to default entry in the routing table so
> a packet gets forwarded on LSP a-b-c-d . However it may be possible that
> the _best_ egress router in this case is router g. And d will indeed
> conclude so by inspecting the IP packet. (through BGP next hop. As the
> packet that will be routed through the default root is most likely to be
> external to the domain or your IGP is screwed up.) So this results in
> suboptimal routing. The TTL may expire in this prematurely.
> 
> >2. as i learn through ur code that mpls_input function recieve pkt from
> >DLC layer. if i am right then what is the significance of Push opcode in
> >mpls_input() function coz when a packet traverse up in network layer
> >headers r ripped off but in mpls_input (push) it is inserted. if i am
> >wrong please also explain the case when header is pushed in mpls_input()
> >happen.


It is for flexibility.  There is nothing saying you MUST do a push as part
mpls_input().  Infact to date, I haven't come up with a senerio that can
could use it.  But none-the-less I've left it there because there are
people in the world that are crazier then me :-)

Jim
-- 
James R. Leu

Re: [mpls-linux-general] default entry in mpls forwarding table

[Abhijit G. <gab...@ee...>]

anish  verma wrote :

>
>hi jim,
>i have some queries.. please help me
>
>1.  is it possible to forward packets to some default entry through
>mpls-linux as it done in ip-network..

I can answer this query on Jim's behalf ;-)

Binding labels to default entry is not a good thing. The reason is as
follows.

1. The labels bound to default entry will have to leave the MPLS domain
through a 'default egress' router. As the fec, which is '0.0.0.0' in this
case has to be uniquely interpreted throughout the domain.

eg. In following figure 'd is the default egress'

a-----b------c------d
      |
      |
      e------f------g

Now assume that a label is bound to default entry in the routing table so
a packet gets forwarded on LSP a-b-c-d . However it may be possible that
the _best_ egress router in this case is router g. And d will indeed
conclude so by inspecting the IP packet. (through BGP next hop. As the
packet that will be routed through the default root is most likely to be
external to the domain or your IGP is screwed up.) So this results in
suboptimal routing. The TTL may expire in this prematurely.

>2. as i learn through ur code that mpls_input function recieve pkt from
>DLC layer. if i am right then what is the significance of Push opcode in
>mpls_input() function coz when a packet traverse up in network layer
>headers r ripped off but in mpls_input (push) it is inserted. if i am
>wrong please also explain the case when header is pushed in mpls_input()
>happen.

no-idea

>
>thanx in advance
>bye
>anish

hth

-abhijit

[mpls-linux-general] default entry in mpls forwarding table

[anish v. <ani...@re...>]

hi jim,=0Ai have some queries.. please help me=0A=0A1.  is it possible to f=
orward packets to some default entry through mpls-linux as it done in ip-ne=
twork..=0A2. as i learn through ur code that mpls_input function recieve pk=
t from DLC layer. if i am right then what is the significance of Push opcod=
e in mpls_input() function coz when a packet traverse up in network layer h=
eaders r ripped off but in mpls_input (push) it is inserted. if i am wrong =
please also explain the case when header is pushed in mpls_input() happen.=
=0A=0Athanx in advance=0Abye=0Aanish=0A=0A =0A

Re: [mpls-linux-general] Mpls for Linux 1.126

[James R. L. <jl...@mi...>]

I thought I mentioned in the e-mail to you that the only way to bind
FECs to outging labels was via iptables.

If not .... You can can no longer use mplsadm to bind FEC(-f) to outlabels
you need to use iptables.  You will need to get the ip-tables source
and apply the ip-tables patch from mpls-linux.  Then you will be able to
use a command like this:

iptables -O OUTPUT -d 192.168.1.0/24 -j MPLS --set-key 0x2

Make sure your kernel has support for netfilter and the MPLS target!

Jim

On Mon, Jan 28, 2002 at 05:36:14PM +0800, ncc49a wrote:
> Hi,
> 
>   Downloaded the latest patch from the CVS. However, when I 
> tried to bind a label to an FEC, an error occurred that 
> the -f instruction is no longer supported.
> 
>   How then to bind to an FEC? Is the instruction set for 
> marking of EXP bits from DSCP same as the 1.108 version?

It should be.

> 
>   Thanks!
> 
> Paul
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

Re: [mpls-linux-general] Question about Data structure of Ldp-portable

[James R. L. <jl...@mi...>]

First thing, thank you for taking time to download and look at ldp-portable.
Second thing, you will want to get the latest version via CVS:

http://sourceforge.net/cvs/?group_id=15443

On Sun, Jan 27, 2002 at 04:09:38AM -0800, Cheng Zhengchao wrote:
> Hi,All,
>   In my reading the codes of ldp-portable-0.060,I was
> puzzled by some data structure.I need your help.
> 1. what does ldp_fs(FS) refer to?

Once you have the newest version, re-read ldp_struct.txt

> 2. what do the
> ldp_entity,ldp_adj,ldp_session,ldp_addr,ldp_attr
> mean?In RFC 3036,I can not find the defination and
> purpose of entity,attr.To me,these structures are too
> complex.

:-) Amazing how RFC speak can turn complicated isn't it.

> what is the relationship among them?I read the
> document "ldp_struct",but I still do know what is the

ldp_struct.txt is the best I have to give you to explain the relationships.
What more do you want?  It even tells you if they are 1:1 1:N or N:M!!!
From there you canm "use the source" to figure it out.

> purpose of them,do we need so many structure to
> implement ldp?

Remember that ldp-portable has to keep track of info which could be stored
in the kernel or other routing code.  It does this so that it can remain
OS/routing system independent.

Jim

>   Thanks for your reply.
> 
> Best Regard 
> 
> Cheng
>    
> 
> 
> __________________________________________________
> Do You Yahoo!?
> Great stuff seeking new owners in Yahoo! Auctions! 
> http://auctions.yahoo.com
> 
> _______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu

Re: [mpls-linux-general] LDP-Portable and other questions

[anish v. <ani...@re...>]

hi dani,=0A  ethereal will work with mpls-linux. and redhat 7.1.. after cap=
turing packet u will have IP fragmented.. n after cliking on it u will find=
 MPLS packet..=0Aanish=0A=0A=0AOn Mon, 28 Jan 2002 Daniel Gregorio Catalan =
wrote :=0A> Hi! I have installed linux-mpls 0.996 (mpls 1.0) over =0A> linu=
x kernel 2.4.13 (Red Hat 6.2 distribution) in =0A> pentium 133 with 32 Mb R=
AM it seems to work properly, =0A> but I have some questions:=0A> =0A> (Fir=
st of all, sorry for my english...)=0A> =0A> - Can someone tell me one easy=
 way to find out if I =0A> have installed correctly the mpls-linux?=0A> =0A=
> - How can I redirect the mplsadm debug messages to a =0A> file?=0A> =0A> =
- When I make the basic example that is in =0A> mpls-linux/utils/ (this is =
a simple LSP between two =0A> hosts) After executing all the commands to ma=
ke the LSP =0A> I put in both hosts ./mplsadm -d to set the debug =0A> opti=
on. Then I make a ping from one host to the other =0A> with the count optio=
n in order to send only one packet. =0A> In the sending-host it seems to wo=
rk properly because =0A> it ends with mpls_output: exit , however in the =
=0A> receiving-host it ends with the mpls_rcv: exit message =0A> but it doe=
sn't end executing the mplsadm, it seems to =0A> be waiting to another mess=
age. And when I make another =0A> ping, it receive it in the same way as th=
e first one =0A> and it continues waiting. It seems to be a no-end loop, =
=0A> the only way to exit from it is with 'control + c'. Is =0A> it correct=
? or Do I have installed or executed wrongly =0A> the mplsadm utility?=0A> =
    In receiving host side these are the messages:=0A> =0A>                =
 mpls_rcv: exit=0A>                 mpls_opcode_peek: enter=0A>            =
     mpls_opcode_peek: exit=0A>                 mpls_input: enter=0A>      =
           mpls_input:  labelspace=3D 0, label=3D 17, =0A> exp=3D 0, B.O.S.=
=3D 1, TTL=3D 64=0A>                 mpls_input: pop=0A>                 mp=
ls_opcode_pop: enter=0A>                 mpls_opcode_pop: exit=0A>         =
        mpls_input: mii_proto 8=0A>                 mpls_finish: enter=0A> =
                mpls_finish: exit=0A>                 mpls_input: sending t=
o IPv4=0A>                 skb_dump from eth0 with len 84 (284) =0A> headro=
om=3D 32 tailroom=3D 8=0A>                     (three lines with ASCII char=
s, I =0A> think they are the bytes that are in the packet)=0A>             =
    mpls_input: retval from ip_rcv 0=0A>                 mpls_rcv: exit=0A>=
 =0A> (And here is where it waits to receive another packet =0A> from the o=
ther host)=0A> =0A> - Now I have compiled the ldp-portable package, and I =
=0A> don't know how it exactly works.=0A> I have had some compiling problem=
s, the first time I =0A> try to build it, it failed and ended with a core_d=
ump. =0A> But then I made a clean, and then, when I compiled the =0A> packa=
ge it seems to be compiled correctly (only a few =0A> warning messages abou=
t unused variables).=0A> So, can someone tell me a basic list of instructio=
ns =0A> (in the ldp-portable prompt) to verify that =0A> ldp-portable works=
 properly? =0A> =0A> - It seems you have to add hosts and interfaces, but d=
o =0A> I have to add a LSP between the host with mplsadm =0A> before using =
ldp-portable?  or Does ldp-portable make =0A> the LSPs? =0A> =0A> - Is ther=
e some documentation about this package?=0A> =0A> - Now I'm trying to insta=
ll ethereal to take some =0A> statistics, but I don't know if it works with=
 =0A> mpls-linux. I'm tryng with ethereal 0.8.19, but I can't =0A> install =
it in Red Hat 6.2, so now I'm going to install =0A> Red Hat 7.1, but I don'=
t know if it will work with =0A> mpls-linux and ethereal. Can someone help =
me?=0A> =0A> Thanks in advance!=0A> =0A> Dani=0A =0A

[mpls-linux-general] LDP-Portable and other questions

[Daniel G. C. <da...@on...>]

Hi! I have installed linux-mpls 0.996 (mpls 1.0) over linux kernel =
2.4.13 (Red Hat 6.2 distribution) in pentium 133 with 32 Mb RAM it seems =
to work properly, but I have some questions:

(First of all, sorry for my english...)

- Can someone tell me one easy way to find out if I have installed =
correctly the mpls-linux?

- How can I redirect the mplsadm debug messages to a file?

- When I make the basic example that is in mpls-linux/utils/ (this is a =
simple LSP between two hosts) After executing all the commands to make =
the LSP I put in both hosts ./mplsadm -d to set the debug option. Then I =
make a ping from one host to the other with the count option in order to =
send only one packet. In the sending-host it seems to work properly =
because it ends with mpls_output: exit , however in the receiving-host =
it ends with the mpls_rcv: exit message but it doesn't end executing the =
mplsadm, it seems to be waiting to another message. And when I make =
another ping, it receive it in the same way as the first one and it =
continues waiting. It seems to be a no-end loop, the only way to exit =
from it is with 'control + c'. Is it correct? or Do I have installed or =
executed wrongly the mplsadm utility?
    In receiving host side these are the messages:

                mpls_rcv: exit
                mpls_opcode_peek: enter
                mpls_opcode_peek: exit
                mpls_input: enter
                mpls_input:  labelspace=3D 0, label=3D 17, exp=3D 0, =
B.O.S.=3D 1, TTL=3D 64
                mpls_input: pop
                mpls_opcode_pop: enter
                mpls_opcode_pop: exit
                mpls_input: mii_proto 8
                mpls_finish: enter
                mpls_finish: exit
                mpls_input: sending to IPv4
                skb_dump from eth0 with len 84 (284) headroom=3D 32 =
tailroom=3D 8
                    (three lines with ASCII chars, I think they are the =
bytes that are in the packet)
                mpls_input: retval from ip_rcv 0
                mpls_rcv: exit

(And here is where it waits to receive another packet from the other =
host)

- Now I have compiled the ldp-portable package, and I don't know how it =
exactly works.
I have had some compiling problems, the first time I try to build it, it =
failed and ended with a core_dump. But then I made a clean, and then, =
when I compiled the package it seems to be compiled correctly (only a =
few warning messages about unused variables).
So, can someone tell me a basic list of instructions (in the =
ldp-portable prompt) to verify that ldp-portable works properly?=20

- It seems you have to add hosts and interfaces, but do I have to add a =
LSP between the host with mplsadm before using ldp-portable?  or Does =
ldp-portable make the LSPs?=20

- Is there some documentation about this package?

- Now I'm trying to install ethereal to take some statistics, but I =
don't know if it works with mpls-linux. I'm tryng with ethereal 0.8.19, =
but I can't install it in Red Hat 6.2, so now I'm going to install Red =
Hat 7.1, but I don't know if it will work with mpls-linux and ethereal. =
Can someone help me?

Thanks in advance!

Dani