I see two parts to this patch, of which I think one of them is actually being discussed:
a) If ONLINK is passed from a routing protocol: Pass this value to the kernel as ONLINK. b) If ONLINK is passed from a routing protocol: in nexthop_active_ipv4, bypass normal nexthop resolution and *only* check to see if the interface is up. I believe the discussion is centering around (a) where should we always pass in ONLINK to the kernel. I am not sure what that would mean, because the kernel does nothing with it(other than a pass through to another non Quagga routing protocol). So in theory I would be ok with that change... I also agree with the assertion that there is no need to spin a new NEXTHOP_IPV4_XXXX type when we can add it as a flag to NEXTHOP_IPV4_IFINDEX. I'll respin the patch(es) with these changes to see what people think about it. Thoughts? donald On Thu, Oct 1, 2015 at 12:39 PM, Joakim Tjernlund < [email protected]> wrote: > On Thu, 2015-10-01 at 18:31 +0200, Christian Franke wrote: > > On 10/01/2015 03:56 PM, Joakim Tjernlund wrote: > > > On Wed, 2015-09-30 at 21:46 -0400, Donald Sharp wrote: > > > > Joakim - > > > > > > > > Before I clean up the patch any further, I believe something like > this is being proposed, is this > > > > correct? > > > > > > Something like that although flags should be mandatory, even if is is > zero. > > > However, there was a discussion about having zebra autodetect ONLINK > instead? > > > > I am not sure how much of this made it to the archives, so let's just > > put it here: > > > > Imho, having a flag for NEXTHOP_IPV4_IFINDEX to specify the ONLINK > > semantics is better than adding a new nexthop type. > > > > I would also like to pose the following, possibly provocative, question, > > which might be an even stronger stand on ONLINK autodetection: Is this > > flag really needed? Couldn't Zebra just always set the rtnl onlink flag > > for nexthops that come in as NEXTHOP_IPV4_IFINDEX or > NEXTHOP_IPV6_IFINDEX? > > > > Is there any case where a routing protocol will send > > NEXTHOP_IPV4_IFINDEX or NEXTHOP_IPV6_IFINDEX to Zebra where not > > installing the route because of a failing onlink-check is the correct > > behavior? > > This is my thinking too, ospfd should never send a route it doesn't want. > Does any other protocol expect this? > Possibly one should add ONLINK for any XXX_IFINDEX route? > > Jocke >
_______________________________________________ Quagga-dev mailing list [email protected] https://lists.quagga.net/mailman/listinfo/quagga-dev
