Members of bond's might be setting up incorrectly?

[donaldsharp]

FRR has code in zebra that is looking at the bond type and treating it as a bond_slave. I am not sure if this is correct but it would require some investigation to fully understand what the proper thing to do in zebra is. See code in zebra/if_netlink.c:

                if ((slave_kind != NULL) && strcmp(slave_kind, "bond") == 0)
                        netlink_determine_zebra_iftype("bond_slave", &zif_type);
                else
                        netlink_determine_zebra_iftype(kind, &zif_type);

At this point I am not sure what the correct thing to do here is nor am I sure that any changes I would make wouldn't break evpn code( or some implicit evpn code ). Let's add an issue to track this and figure out what the correct thing to do is.

From @idryzhov -> I understand that it's just a copy-paste from netlink_interface, but I think this change is wrong. Why should we replace the interface type with generic bond_slave when it is actually, for example, veth or gre? The fact that the interface is a bond slave is determined by zif_slave_type field, look at IS_ZEBRA_IF_BOND_SLAVE. Let's just drop this change.