Hi Ajith,

On Mon, Jun 23, 2014 at 07:58:21AM +0530, Ajith Adapa wrote:

Hi Luessing,

As per RFC 4541, snooping switches send query with source address as 0.0.0.0 since port in a L2 switch doesn't have ip-address configured.

Hm, I'm not quite sure whether this is true. Do you have a section/paragraph in RFC4541 which says that?

To me, it seems that 0.0.0.0/:: should only be used when the topology changes, on links becoming active (the RFC uses the example of STP). As a performance optimization but not as the source address for General Queries.

In section 2.1.1 1) b) says:

"The 0.0.0.0 address represents a special case where the switch is proxying IGMP Queries for faster network convergence, but is not itself the Querier."

In section 2.1.1 4) it says:

"If the switch is not the Querier, it should use the 'all-zeros' IP Source Address in these proxy queries (even though some hosts may elect to not process queries with a 0.0.0.0 IP Source Address)."

(which seems to imply that if a switch is the selected Querier, that it should use something other than the 'all-zeros' IP Source Address?)

Also, section 2.1.1 4) says:

"[...] even though some hosts may elect to not process queries with a 0.0.0.0 IP Source Address"

Which seems to indicate that 0.0.0.0 might be a bad choice to get reliable IGMP Reports and might only make sense for performance optimizations (at least for IPv4 - for IPv6 it doesn't make sense to me to send General Queries with '::' source address at all since '::' is ignored by hosts per RFC).

(*)

With current changes linux bridge will always select a querier who sends a query message with 0.0.0.0. Right ?

•   if (ntohl(saddr) <= ntohl(br->ip4_querier.addr.u.ip4))
  

•           goto update;
  

Yes. Which is what the querier election mechanism in the RFCs for IGMPv2/IGMPv3 (RFC2236/RFC3376) specifies, I think. (**)

Currently linux bridge sends a query message with source address as 0.0.0.0. Right ?

For IPv4, this is currently the default, yes. (there's a sysfs switch to change that, if desired) (***)

What should be the expected behaviour ?

Hm, good question.

Regarding (***): After rereading RFC4541 a few more times, I think (*) would be the right choice: The bridge code should probably be changed to always use an available, non-zero IPv4 source address for General Queries. If none is available it, doesn't make sense to me to become the selected Querier at all (should be changed for IPv6, too: Currently it refrains from sending Query messages then, but thinks a valid querier were in the network, enabling the multicast snooping of the bridge).

Regarding (**): Now that you're mentioning that, you're right, my change isn't quite straightforward in that regard, a few lines earlier there is:

•   if (!br->ip4_querier.addr.u.ip4)
  

•           goto update;
  

So it would select the all-zeros Querier if it receives such messages. But would happily switch to a non-all-zeros Querier as well if receiving such messages...

And now, this baffles me, section 2.1.1 4) again:

"When such proxy queries are received, they must not be included in the Querier election process."

If we were following that, a snooping switch and a true multicast router could potentially have an argument about which becomes the selected IGMP querier? (for IPv6/MLD this dilemma should not exist, I think)

I'm currently unsure, whether we should follow RFC2236(IGMPv2)+RFC3376(IGMPv3) or the informational (!) RFC4541.

Or am I overlooking something? What do others think?

Cheers, Linus