On 2010/08/26 15:25, Claudio Jeker wrote:
> On Wed, Aug 25, 2010 at 12:29:00AM +0100, Stuart Henderson wrote:
> > On 2010/08/24 23:14, Claudio Jeker wrote:
> > > carp(4) is normaly using IPv4 multicast for its announcements. In case
> > > there are no IPv4 addrs set IPv6 multicast will be used. But since the
> > > code forgot to embed the oh so freaking magic scope into the multicast
> > > address it failed to send the packet out.
> > > 
> > > I don't think anyone will OK this, so I will autocommit in about 2 days.
> > 
> > I must be missing something but even without your diff, I am seeing
> > outgoing carp packets locally and remotely when there are only v6
> > addresses on the carp(4)..
> > 
> > tcpdump output looks like this either way:
> > 
> > 00:23:24.709780 fe80::202:b3ff:fec8:dc4 > ff02::12: CARPv2-advertise 36: 
> > vhid=88 advbase=1 advskew=0 demote=0
> > 
> > fxp1: flags=8b43<UP,BROADCAST,RUNNING,PROMISC,ALLMULTI,SIMPLEX,MULTICAST> 
> > mtu 1500      
> >         lladdr 00:02:b3:c8:0d:c4
> >         priority: 0
> >         media: Ethernet autoselect (100baseTX full-duplex)
> >         status: active
> >         inet6 fe80::202:b3ff:fec8:dc4%fxp1 prefixlen 64 scopeid 0x5
> > 
> > carp88: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> >         lladdr 00:00:5e:00:01:58
> >         priority: 0
> >         carp: MASTER carpdev fxp1 vhid 88 advbase 1 advskew 0
> >         groups: carp
> >         status: master
> >         inet6 fe80::200:5eff:fe00:158%carp88 prefixlen 64 scopeid 0xe
> >         inet6 3001::88 prefixlen 64
> > 
> > This is on sparc64; I've also tried on amd64 without your diff and
> > see the same there.
> > 
> 
> Hmm. Something is funky because when I tried it did not work. Maybe
> it was may testing that managed to confuse ip6_output in a similar manner
> as it is confusing me.

One thing I ran into when testing (which I knew before, but forgot) was
that the autoconfigured link-local addresses are excluded from carp,
so unless you configure a real address you won't see any announcement
from carp.

>                        I still think that setting the scope on that link
> local multicast address is better then hoping that ifaof_ifpforaddr()
> returns a link local address.

Yes, I agree. ok.

Reply via email to