When the bridge fails to create a port group for a new MDB entry, the
entry is left behind with no ports and is never deleted, but it is still
counted towards mcast_hash_max. Since the introduction of
mcast_max_groups, a host behind a port that reached its maximum can use
this to exhaust the MDB and disable multicast snooping.
Patch #1 fixes the problem by deleting the entry upon failure.
Patch #2 adds a selftest.
Ido Schimmel (2):
bridge: mcast: Delete empty MDB entry when port group creation fails
selftests: forwarding: bridge_mdb_max: Add a test for rejected MDB
entries
net/bridge/br_mdb.c | 16 +-
net/bridge/br_multicast.c | 9 ++
net/bridge/br_private.h | 1 +
.../net/forwarding/bridge_mdb_max.sh | 138 ++++++++++++++++++
4 files changed, 161 insertions(+), 3 deletions(-)
--
2.56.0
