From: Klemens Nanni <kn@openbsd.org>
Subject: Re: neuter tun(4)/tap(4) ioctls that change interface flags
To: David Gwynne <david@gwynne.id.au>, tech@openbsd.org
Date: Wed,  9 Oct 2024 11:29:53 +0000

09.10.2024 08:34, David Gwynne пишет:
> once upon a time there was just tun(4) and it handled both layer 3 (ipv4
> and ipv6) and ethernet. flipping interface type around at runtime was a
> recipe for disaster, so we've been progressively locking this down
> as time goes by splitting ethernet support out into the tap(4)
> interface. tun(4) is set up as a p2p interface with the right interface
> flags for p2p, and tap(4) is set up as an ethernet interface with the
> right interface flags for ethernet.
> 
> we locked down the ability for userland to reconfigure the interface
> type of tun/tap interfaces at runtime, but we still allow the flags to
> change. this diff removes that ability. it still allows the TUNSIFMODE
> and TUNSIFINFO ioctls, but you have to specify the interface flags
> appropriate for the interface type.
> 
> it also removes the ability to set IFF_UP using these ioctls.

Makes sense to me.

> 
> nothing in base uses them, so if there's going to be any fallout it's
> from ports. eg, openvpn does TUNGIFINFO, sets IFF_MULTICAST, and then
> TUNSIFINFO to apply it. because IFF_MULTICAST is set for both tun and
> tap, this is just a waste of cpu time more than anything else, but is
> otherwise handled by this diff. other software may be cleverer^Wdumber
> though.

If there's ports fallout, we'll fix it up.
Seems like now is a good time to clean up.

OK kn