[Xorp-hackers] IGMP version used on an interface

[Sachin K]

Thank you for the reply.

I have a multicast router with multiple interfaces, each connected to
different network. Each network uses a different version of IGMP. I am
running Linux on the router.

If the user space multicast routing daemon had to send a group
specific query on all interfaces then how to know which interface is
using what version of IGMP?

Is there any ioctl or getsockopt call to know that on linux?

Thanks,
Sachin

On 3/13/07, Pavlin Radoslavov <pavlin at icir.org> wrote:
> > How XORP handles the following scenario?
> >
> >                 +---------------+
> > [A]-------|     |               |<---------->[S1]
> >            |-----|   MRouter|<---------->[S2]
> > [B]-------|     |               |<---------->[S3]
> >                 +---------------+
> > 1. A and B are on the same LAN and MRouter is the multicast router on that LAN.
> > 2. A joins (S1,G) and B joins (S2, G).
> > 3. A leaves G and sends a leave (*,G)
> >
> > How does MRouter handles this (*,G) LEAVE?
>
> The behavior is specified in the IGMPv3 spec (RFC 3376).
>
> Before Step 3, Mrouter is in INCLUDE (A+B) state.
> The IGMPv2 Leave (*,G) in Step 3 is equivalent to the IGMPv3
> TO_IN( {} ) message (see Section 7.3.2 in the IGMPv3 spec).
>
> Then according to Section 6.4.2 the new router state and actions
> are (note that I changed A->M and B->N in the original description
> to avoid confusion with your example):
>
> Router State   Report Rec'd New Router State        Actions
> ------------   ------------ ----------------        -------
> INCLUDE (M)    TO_IN (N)    INCLUDE (M+N)           (N)=GMI
>                                                    Send Q(G,M-N)
>
> In your example, M= A+B, while N is {} (i.e., empty). Therefore, the
> new router state is INCLUDE (A+B+{}) = INCLUDE(A+B).
> In addition, the router will send Group-and-Source Specific Query to
> G with source-list {A+B}.
> Future state of MRouter depends on the received Reports. See Section
> 6.4.2 in the IGMPv3 spec for complete description of the state
> transitions.
>
> Regards,
> Pavlin
>