Linux Ethernet Bridging - Oct 2021 - [Bridge] [PATCH net] net: bridge: mcast: use multicast_membership_interval for IGMPv3

On 15/10/2021 09:36, Hangbin Liu wrote:
> On Fri, Oct 15, 2021 at 10:52:00AM +0800, Hangbin Liu wrote:
>>> -	mod_timer(&p->timer, now +
br->multicast_membership_interval);
>>> +	if (igmpv2_mldv1)
>>> +		mod_timer(&p->timer, now +
br->multicast_membership_interval);
>>
>> Hi Nikolay,
>>
>> Our engineer found that the multicast_membership_interval will not work
with
>> IGMPv3. Is it intend as you said "IGMPv3/MLDv2 handling is not yet
>> implemented" ?
> 
> Ah, I saw in br_multicast_group_expired() it wait for mp->ports be freed
> before delete the mdb entry.
> 
> And in br_multicast_port_group_expired() it wait for src entry freed first.
> 
> But when mod group src timer, we use
> __grp_src_mod_timer(ent, now + br_multicast_gmi(brmctx));
> 
> instead of user configured multicast_membership_interval. I think we should
> fix it. WDYT?
> 
> Thanks
> Hangbin
> 
Hi,
I did that intentionally because of how RFC 3376 defines the GMI:
" 8.4. Group Membership Interval

   The Group Membership Interval is the amount of time that must pass
   before a multicast router decides there are no more members of a
   group or a particular source on a network.

   This value MUST be ((the Robustness Variable) times (the Query
   Interval)) plus (one Query Response Interval)."

So it is computed properly and the RFC says "MUST" be. The only thing
that is not
configurable is the QRV currently, if that is added then the GMI will be
automatically
correctly computed as per the RFC. I get the inconsistency though as we
don't compute all
timers as the RFC mandates, I'm inclined to revert the GMI computation as it
was before
and possibly fix how all timers are computed in the future as that may require
more work.
In addition this could've changed user-expected behaviour when switching to
v3 queries which
were supported before IGMPv3 was added, so yeah going forward looks like
it'd be better to
switch to the old GMI behaviour. I'll send the patch in a minute.

Thanks,
 Nik

From: Nikolay Aleksandrov <nikolay at nvidia.com>

When I added IGMPv3 support I decided to follow the RFC for computing
the GMI dynamically:
" 8.4. Group Membership Interval

   The Group Membership Interval is the amount of time that must pass
   before a multicast router decides there are no more members of a
   group or a particular source on a network.

   This value MUST be ((the Robustness Variable) times (the Query
   Interval)) plus (one Query Response Interval)."

But that actually is inconsistent with how the bridge used to compute it
for IGMPv2, where it was user-configurable that has a correct default value
but it is up to user-space to maintain it. This would make it consistent
with the other timer values which are also maintained correct by the user
instead of being dynamically computed. It also changes back to the previous
user-expected GMI behaviour for IGMPv3 queries which were supported before
IGMPv3 was added. Note that to properly compute it dynamically we would
need to add support for "Robustness Variable" which is currently
missing.

Reported-by: Hangbin Liu <liuhangbin at gmail.com>
Fixes: 0436862e417e ("net: bridge: mcast: support for IGMPv3/MLDv2
ALLOW_NEW_SOURCES report")
Signed-off-by: Nikolay Aleksandrov <nikolay at nvidia.com>
---
 net/bridge/br_private.h | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
index b4cef3a97f12..5c68052466a0 100644
--- a/net/bridge/br_private.h
+++ b/net/bridge/br_private.h
@@ -1125,9 +1125,7 @@ static inline unsigned long br_multicast_lmqt(const struct
net_bridge_mcast *brm
 
 static inline unsigned long br_multicast_gmi(const struct net_bridge_mcast
*brmctx)
 {
-	/* use the RFC default of 2 for QRV */
-	return 2 * brmctx->multicast_query_interval +
-	       brmctx->multicast_query_response_interval;
+	return brmctx->multicast_membership_interval;
 }
 
 static inline bool
-- 
2.31.1