[quagga-dev 11691] Re: [PATCH] zebra: Connected route addition shoudn't happen in MRIB

David Lamparter equinox at opensourcerouting.org
Thu Oct 30 05:10:00 GMT 2014


On Mon, Oct 27, 2014 at 05:01:51PM -0200, Everton Marques wrote:
> On Thu, Oct 23, 2014 at 2:26 PM, Balaji.G <balajig81 at gmail.com> wrote:
> > Signed-off-by: Balaji.G <balajig81 at gmail.com>
> > ---
> >  zebra/connected.c | 3 ---
> >  1 file changed, 3 deletions(-)
> >
> > diff --git a/zebra/connected.c b/zebra/connected.c
> > index c4f87f4..ef745a8 100644
> > --- a/zebra/connected.c
> > +++ b/zebra/connected.c
> > @@ -194,9 +194,6 @@ connected_up_ipv4 (struct interface *ifp, struct
> > connected *ifc)
> >    rib_add_ipv4 (ZEBRA_ROUTE_CONNECT, 0, &p, NULL, NULL, ifp->ifindex,
> >         RT_TABLE_MAIN, ifp->metric, 0, SAFI_UNICAST);
> >
> > -  rib_add_ipv4 (ZEBRA_ROUTE_CONNECT, 0, &p, NULL, NULL, ifp->ifindex,
> > -       RT_TABLE_MAIN, ifp->metric, 0, SAFI_MULTICAST);
> > -
> >    rib_update ();
> >  }
> >
> This change looks good to me.

This patch was discussed on IRC between Balaji, Everton and me;  the
outcome was that Balaji would resend it with a switch to control the
behaviour.

This is particularly relevant in cases where the multicast RIB contains
a default route, such that the fallback to unicast is not triggered.
Such a case would result in local networks not working for RPF checks
because connected routes are not in the MRIB.

(Of course, all IGPs have essentially the same problem, so we need
controlled RIB<>MRIB route exchange and/or MT support in OSPF & IS-IS.)


-David




More information about the Quagga-dev mailing list