On Mon, Jul 01, 2013 at 11:56:50AM +0200, Baptiste Jonglez wrote: > On Sun, Jun 30, 2013 at 09:01:30PM +0200, Juliusz Chroboczek wrote: > > If you want to try your hand at fixing it, it's check_network in > > ahcpd.c. Note that the natural solution is racy (an interface could > > go down-up before ahcpd notices), and the right solution is quite > > a bit of work (it would require monitoring interfaces through netlink, > > yuck). > > While we're at it, I believe there are a number of ways to improve > ahcpd
Well, while we're dreaming of ahcpd as a really poweful tool, we could also implement RFC 4941. It probably wouldn't even be that hard if we implemented netlink monitoring first: install addresses with finite values for valid_lft and preferred_lft, let the kernel deal with the timeouts, and allocate new addresses when they expire (which you have to do anyway if the address disappears for any reason, such as a yoyo-interface). (Of course, this is slightly incorrect: you should ideally reuse the address if the interface goes down/up instead of choosing a new one, but it would mean duplicating the timeout mechanism in user-space which seems a bit silly.) -- Gabriel _______________________________________________ Babel-users mailing list [email protected] http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/babel-users
