* Stefan Sperling <s...@openbsd.org> [2014-07-15 11:06]: > On Sun, Jul 13, 2014 at 03:48:47PM +0200, Henning Brauer wrote: > > now that we have an uncontaminated, err, inet6-free system by default, > > IFXF_NOINET6 just doesn't make sense any more. > > fully go for no inet6 by default, get rid of the IFXF_NOINET6 guarded > > attachments etc. > > introduce IFAFATTACH and IFAFDETACH ioctls. note that they are NOT > > inet6 specific; the kernel only has code for inet6 so far and will > > properly return EAFNOSUPPORT for all others. > > > > there should be no user visible changes from this. > > I like this direction. > It's a lot better than the AF-specific "kill switch".
oh absolutely. and yes i plan to add "ifconfig <if> -inet" as a convenient way to get rid of all v4 addrs at once; inet doesn't really need that since it just doesn't have sth like the dreaded link-locals and the rtadv magic, but symmetry is good. and who knows whether we'll need it for inet7 anyway :) > However, since we're heading towards release, I believe we should wait > and do this flag removal + ioctl addition later. We've already got the > no-link-local-by-default change, and the new autoconf6 flag. > I don't see much harm in leaving it at that for this release cycle and > moving further in the next one. I'm looking forward to florian's in-kernel > rtsol as well, in the next cycle (it's not done yet anyway AFAIK). florian is done with that really, we're at the "last tiny nits" stage, polishing that could as well happen in-tree. I'm slightly undecided on whether this should make this release or not... it is actually kinda mechanic, i. e. we already know the relevant places since they are the ones that had the NONINET6 flag check before. I do think the risk is not big, so if we get some solid tests we should be golden. I for one haven't been able to make it misbehave no matter how hard I tried. kernel-rtsol should make release imo. > Perhaps we can also sort out the autoconf vs. ip6forward conundrum then, > and ship working IPv6 for SOHO routers in 5.7. yeah that limitation is bizarre - straight from the spec and not added by us, tho. > I'll be testing this regardless and will let you know if I run into issues. coolio, thx -- Henning Brauer, h...@bsws.de, henn...@openbsd.org BS Web Services GmbH, http://bsws.de, Full-Service ISP Secure Hosting, Mail and DNS. Virtual & Dedicated Servers, Root to Fully Managed Henning Brauer Consulting, http://henningbrauer.com/
