On Tue, Mar 01, 2005 at 03:51:34PM +0200, Jari Arkko wrote:
> On the other hand, if the problem is in bad drivers (as Bill points
> out), this may be a different issue. I guess part of the problem is
> that for plain old IPv4 usage multicast features in NICs and the
> drivers don't get tested well, so you see problems when you
> switch to IPv6 or other things that depend on multicast capabilities.
> Is there any way to determine how widespread this problem is?
I've seen this problem once or twice on FreeBSD, where the driver
author either didn't have enough information to program the multicast
filters or didn't complete the programming of the multicast filters
because it wasn't necessary for the normal opperation of IPv4. Mind
you, all such problems that I knew about have been fixed for some.
(The practical version of the promiscuous mode work-around for this
problem is to run "tcpdump -i blah" in the background on a problem
machine.)
I've seen a few other reasons that multicast ethernet frames have
been dropped. I've seen an IGMP snooping that only understands IPv4,
but turning off IGMP snooping fixed that. One type of VLAN tagging
on some old Cisco switches didn't seem to pass multicast traffic
properly (switching to 802.1q fixed that). I've heard reports of
802.11 APs that don't behave, but never seen one myself.
David.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
