Re: multicast weirdness

[Otto Moerbeek <otto@drijf.net>]

On Wed, 2 Jul 2003, Chris Kuethe wrote:

> On Wed, 2 Jul 2003, Andy Wettstein wrote:
>
> > On Wed, Jul 02, 2003 at 06:36:39PM +0200, Otto Moerbeek wrote:
> > >
> > > On Wednesday, Jul 2, 2003, at 18:25 Europe/Amsterdam, Andy Wettstein
> > > wrote:
> > >
> > > >Hello
> > > >
> > > >I am using zebra 0.93a (the packaged version) and the RIP built in
> > > >to it on openBSD 3.3.  The problem is the RIP updates are not
> > > >properly multicasting.  But when I do a "tcpdump host 224.0.0.9" all
> > > >traffic is received and all the RIP updates work correctly.  When I
> > > >kill tcpdump everything stops working again.  I have a route to
> > > >224.0.0.9 through 127.0.0.1, but that doesn't seem to help.
> > > >
> > >
> > > What card? Looks like a multicast filter not working properly. Please
> > > provide full dmesg.
> > >
> >
> > This is on a Broadcom (bge) in a dell poweredge 2650.  I am using the
> > 3.3 stock kernel except for a patch to make vlans work better on the
> > em, and I turned multicast routing on.
>
> no joy with 3.3-rel, 3.3-stable or 3.3-current on my PE1650. (2 x em)
>
> if i use some random white-box pc with 2 fxp's it works great. if i
> switch my external (ospf-running) interface from my trusty fxp to an
> em card, multicast (and ospf) does not work. putting the gig card in
> promiscuous mode makes everything happy again, but that's inelegant.
>
> CK

Some time ago I fixed a very similar problem with a dc(4) card (PR 3263).
Even in chips of the same family, vendors introduce differences that break
multicast filtering. I suggest you file a bug report, including a full
dmesg and tcpdump of the relevant packets. Please use tcpdump -e to
include the ethernet headers.

There has been done quite some work on bge(4) lately in FreeBSD and
NetBSD, maybe a solution can be found there.

	-Otto

 > --
> Chris Kuethe, GCIA CISSP: Secure Systems Specialist - U of A CNS
>       office: 157 General Services Bldg.    +1.780.492.8135
>               chris.kuethe@[pyxis.cns.]ualberta.ca
>
>      GDB has a 'break' feature; why doesn't it have 'fix' too?