conntrack thinks that ICMPv6 Echo reply to ICMPv6 Echo request sent to IPv6 multicast address is INVALID

Hi all,

1) I have two virtual machines with interfaces on the same link:
A) fe80::5054:ff:fe09:e0b9/64
B) fe80::5054:ff:fe80:d951/64

2) I set up IPv6 packet filter on A with ip6tables:

# ip6tables -F
# ip6tables -A INPUT -m conntrack --ctstate INVALID -j REJECT --reject-with icmp6-adm-prohibited
# ip6tables -A INPUT -p ipv6-icmp -j ACCEPT
# ip6tables -A INPUT -j REJECT --reject-with icmp6-adm-prohibited

3) ping6 B from A:

# ping6 -I eth2 fe80::5054:ff:fe80:d951
PING fe80::5054:ff:fe80:d951(fe80::5054:ff:fe80:d951) from
fe80::5054:ff:fe09:e0b9 eth2: 56 data bytes
64 bytes from fe80::5054:ff:fe80:d951: icmp_seq=1 ttl=64 time=0.265 ms
<OK>

4) ping6 'all nodes' from A:

# ping6 -I eth2 ff02::1
PING ff02::1(ff02::1) from fe80::5054:ff:fe09:e0b9 eth2: 56 data bytes
<no response, reply is rejected>

5) remove the first line from ip6tables

# ip6tables -D INPUT 1

6) ping6 'all nodes' from A:

# ping6 -I eth2 ff02::1
PING ff02::1(ff02::1) from fe80::5054:ff:fe09:e0b9 eth2: 56 data bytes
64 bytes from fe80::5054:ff:fe09:e0b9: icmp_seq=1 ttl=64 time=0.072 ms
64 bytes from fe80::5054:ff:fe80:d951: icmp_seq=1 ttl=64 time=0.318 ms (DUP!)
<OK>

I see this with kernels 2.6.32 and 3.3.0.
Is it a known bug or my misunderstanding ?

thanks

Jiri