[Xorp-users] PIM-SM in xorp cvs

[Otto Solares]

On Tue, Aug 22, 2006 at 09:05:00AM -0700, Pavlin Radoslavov wrote:
> > FYI I fetch latest CVS and tried PIM-SM multicast routing and BGP
> > but BGP still don't respect the next-hop parameter so still using
> > quagga and trying xorp for PIM-SM:
> > 
> > What does this warning means?
> > 
> > [ 2006/08/22 07:15:48 WARNING xorp_pimsm4 PIM ] JoinDesired(*,G) = true: upstream neighbor for RP 200.0.204.169 for group 239.255.255.254: not found
> 
> Your XORP router wants to send (*,G) Join message toward the RP
> (200.0.204.169), but the next-hop router toward it is not PIM-SM
> capable (or simply there is no next-hop router).

There is next-hop router but PIM-SM is disabled.  See below why.

> Do you see this error all the time or only on startup? If it is only
> on startup, then it is probably a transient condition (until the
> MRIB information or the PIM-SM neighborhood information has reached
> PIM-SM process).

I see it all the time.

> Otherwise, to debug the problem, first you need to find the IP
> address of the next-hop (unicast) router toward 200.0.204.169.
> After that you need to double-check whether that router is indeed
> running PIM-SM. If it is running PIM-SM, then you need to find why
> the XORP router doesn't see the PIM-SM neighborhood info about the
> upstream.

My IP addres toward the RP (200.0.204.169) is 10.10.26.7, my (unicast)
next-hop router is 10.10.26.14 which have another NIC with
200.0.204.170 and it's next-hop router is my RP 200.0.204.169.

What is happening here is that my RP is in another net not directly
attached to my DR and my next-hop router is PIM-SM capable but is
not enabled because I don't want my next-hop router to be the RP,
because there is no MSDP implementation for Linux I need the other
router be the RP so he can exchange my multicast info with other
MSDP routers.  I think is a valid implementation.

> > When issuing in xorpsh 'show pim mrib' I get in xorpsh:
> > 
> > ERROR: Command "/usr/local/xorp/cli/tools/send_cli_processor_xrl": terminated with signal 11.
> 
> This means that the above process has terminated by SIGSEGV which is
> not good.
> Do you see this every time you execute "show pim mrib"?

Yes, every time.

> Also, do you see it when executing other "show pim" commands?

No, I try all others and everything works as expected, just the above
that trigger the errors.

> Finally, how many forwarding entries are in the kernel when you run
> "show pim mrib" and it fails.

Like 10,000 for IPv4 and 700 for IPv6 (IPv6 is not enabled).

> > And in daemons:
> > 
> > [ 2006/08/22 07:17:08  ERROR xorp_pimsm4:15829 LIBXORP +213 buffered_asyncio.cc io_event ] read error 104
> > [ 2006/08/22 07:17:08  ERROR xorp_pimsm4:15829 XRL +159 xrl_pf_stcp.cc read_event ] Read failed (error = 104)
> > [ 2006/08/22 07:17:08  ERROR xorp_pimsm4:15829 XRL +338 xrl_pf_stcp.cc die ] STCPRequestHandler died: read error
> 
> Those errors are probably related to the SIGSEGV termination of the
> above command.

Sorry for not being clear, what I am saying is that is part of the
problem but the errors are generated in the daemons, so clearly they are
related.

> > I have a lot of IP addresses in various NICs, the PIM register messages
> > comes from the RP connected interface but some registers comes from the
> > last IP address from the last interface, weird, and some messages comes
> > correctly from the RP connected interface.
> 
> Sorry, I don't understand the above, so can you provide some
> additional info.
> E.g., are you observing this in the XORP router that is also the RP?
> Also, are you saying that some of the PIM Register messages arrive
> with destination IP address of the advertised RP address, but other
> PIM Register messages come with a different destination address?
> Please clarify.

Sorry for no being clear, here is where my english has an end :)
anyway will try to explain as is a very weird condition:

As mentioned above my IP address in the interface connected towards
the RP is 10.10.26.7, in another NIC I have like 126 IP public
addresses but that NIC is completely disabled in XORP, what I am
seeing is this:

* some PIM-register comes from 10.10.26.7 with destination my RP
  (which is perfect).
* some PIM-register comes from 168.234.203.126 with destination my
  RP, the source is an IP in the NIC disabled in XORP, and from the
  126 IP addresses in that disabled (in XORP) NIC is using the
  last address not the first as would be the normal in any case.

I think this is very weird as all PIM-register always should come
from the primary IP in the NIC connected towards the RP (which by
the way is something you fix some time ago) and not come from
an IP address in another NIC specially a disabled one, this seems
to me like corruption in structures or something like that.

Thanks!

-otto