[Bridge] Bridge responds on its IP address even when br0 is down.

[Joakim Tjernlund]

Stephen Hemminger <shemminger at linux-foundation.org> wrote on 2010/06/09 18:38:22:
>
> On Wed, 9 Jun 2010 18:30:17 +0200
> Joakim Tjernlund <joakim.tjernlund at transmode.se> wrote:
>
> > Stephen Hemminger <shemminger at linux-foundation.org> wrote on 2010/06/09 17:23:19:
> > >
> > > On Wed, 9 Jun 2010 16:46:01 +0200
> > > Joakim Tjernlund <joakim.tjernlund at transmode.se> wrote:
> > >
> > > >
> > > > System A:
> > > > eth0 has ip 192.168.1.16
> > > >
> > > > Then do in system A
> > > > brctl addbr br0
> > > > ifconfig br0 192.168.1.18 down
> > > >
> > > >
> > > > From system B I ping 192.168.1.18
> > > > and system A responds!
> > > >
> > > > Even though interface br0 is down I can still
> > > > access system A with 192.168.1.18, this feels
> > > > like a bridge bug?
> > > >
> > >
> > > No that is how Linux works. All addresses are property of the
> > > system not the interface.
> > >    http://en.wikipedia.org/wiki/Host_model
> >
> > I see, I know the about the host model but I was surprised it applied
> > to interfaces that is down. Does it not make sense to exclude the
> > IP address when the I/F is down?
>
> Responding on a all interfaces all the time does make sense on a multi-homed server
> (which was the original Linux use case), but it does on a router.
                                                  ^^^^^ - Should be "doesn't" I suppose?

> There is a configurable value in to control it in Vyatta to handle that.

hmm, this "configurable value" makes Linux host model strong or something in between?
You have a patch handy for me to look at?

> The patch got rejected when submitted upstream because of the developers worries
> about breaking existing applications.

Strange, how would a new knob, defaulting to off, break existing apps?

 Jocke