On Mon, Jun 12, 2017 at 02:59:46PM +0100, Stuart Henderson wrote: > This seems the "least worst" option to me. (It's still racy, the > connection could come up and change the dest before the route is > added, so route addition would fail as we now require a "correct" > address on the p-p route addition, but that's not new). ok with me.
Yes, this race is a separate problem. I actually have ifstated(8) monitor pppoe interface state and the address of the peer, and it injects a default route once the peer's address changes to something other than 0.0.0.1.
