On Thu, Nov 26, 2020 at 05:49:10PM +0100, Florian Obser wrote: > The way rad(8) is normaly used is netstart(8) setup a bunch of > interfaces with IPv6 prefixes and one or more are listed in rad.conf. > Everything just works. > > But rad can also be configured to work on interfaces that are > dynamically added (tap, pair, vether...) It would then spam the log > with: > sendmsg on vether1: Can't assign requested address > as noted by tb. > > It's also not sensible to use such an interface, it can't possibly > work. So we are going to ignore interfaces without link local address.
This makes sense to me. It's much nicer to avoid EADDRNOTAVAIL in the first place than trying to handle it. It's also a big improvement in combination with the other diff we're testing and discussing, so I'm very much in favor of having it. The code reads fine fwiw. ok tb
