-- change receive polling / forward a signal to this thread
send bugreports to linux kernel guys
-- In the opposite of IPv4, the IPv6 multicast stack of a current linux kernel doesn't have any group limits. This is very bad because the linux kernel will kill itself if it try to handle to much multicast groups.
-- We tried to testing IGMPv3 proxy with CDRouter, most of the test cases failed. The causes of failure are similar, the CDRouter can't receive expected IGMPv3 message. I listed the steps one of the failure cases here.
1. CDRouter sends a IGMPv3 report from LAN side, with a Group record ALLOW_NEW_SOURCES{x.x.x.x}
2. CDRouter waits IGMPv3 message on WAN side, it expects to receive the same Group record with ALLOW_NEW_SOURCES{x.x.x.x}
But our linux kernel always sends the Group record with CHANGE_TO_INCLUDE_MODE{x.x.x.x}, that doesn't meet the test criterion of CDRouter. So those cases failed.
I have contacted the CDRouter support team, they said the default filter mode should be INCLUDE mode, and they gave the proof:
The default filter-mode state is specified in Section 5.1 of the RFC: 5.1. Action on Change of Interface State
If no interface state existed for that multicast address before the change (i.e., the change consisted of creating a new per-interface record), or if no state exists after the change (i.e., the change consisted of
deleting a per-interface record), then the "non-existent" state is considered to have a filter mode of INCLUDE and an empty source list.
I debugged this issue in linux kernel, and found the default filter mode is EXCLUDE, it's strange when linux did so. So it always sends CHANGE_TO_INCLUDE_MODE{x.x.x.x} when a Group joining at first time.
If the upstream interface sends a TO_IN-record instead of a ALLOW-record at the startup, I think the biggest problem is that the querier of the corresponding network could be forced to send unnecessary
'Group Specific Queries' and/or 'Group and Source Specific Queries' depends on the current state of the querier. This does not lead to misbehavior but increases a little bit the overhead.
Email
-- How much will the implementation protect against killing SSM by
...
...
@@ -55,7 +72,7 @@ will it send out on the upstream interface ?
