Now that 2.6.x series in in a swing, being able to control, from the
userspace, when to send RA's and when to shut them off would be very
beneficial (2.4 too :).
In particular, consider a distribution which wants to allow disabling
autoconfig on one interface. When it's possible to do so, it's
already too late..
---------- Forwarded message ----------
Date: Mon, 27 Oct 2003 15:05:42 +0200 (EET)
From: Pekka Savola <pekkas@xxxxxxxxxx>
To: "YOSHIFUJI Hideaki / [iso-2022-jp] 吉藤英明"
<yoshfuji@xxxxxxxxxxxxxx>
Cc: netdev@xxxxxxxxxxx, sekiya@xxxxxxxxxx
Subject: Re: Disabling IPv6 accept_ra on just some interface
On Mon, 27 Oct 2003, YOSHIFUJI Hideaki / [iso-2022-jp] 吉藤英明 wrote:
> In article <Pine.LNX.4.44.0310231457110.3347-100000@xxxxxxxxxx> (at Thu, 23
> Oct 2003 15:22:47 +0300 (EEST)), Pekka Savola <pekkas@xxxxxxxxxx> says:
> > So, my thought (comments welcome) is:
> >
> > 1) when accept_ra changes from 0 -> 1, initiate the route
> > solicitation process, likewise as one would when the interface is
> > brought up.
> >
> > Makes sense?
> >
> > 2) (probably not a good idea, but some food for thought..) when accept_ra
> > changes from 1 -> 0, delete any autoconfigured routes or
> > prefixes. (could be ugly / dangerous..)
>
> Well, we'd propose to have another config "send_rs" or something like that
> because accept_ra is also effective against unsolicited RAs.
> It, "send_rs," tells kernel to start sending RS
> when the variable is changed 0 to 1 and/or
> when interface is going up.
I don't have any major objections to this model, I'm just worried that it
might make the configuration more complex (we already have accept_ra and
"autoconf" toggles which are confusing enough without documentation :-)
with little gain.
That is, is there any case when you'd want to accept an RA but *not* send
RS? I fail to see clear applicability for this, hence my proposal to
overload accept_ra :-)
> Assume the node has eth0 and eth1.
> Operation will be something like the following.
>
> If you want to listen RA and to send RS on some interfaces,
> sysctl -w net.ipv6.conf.default.accept_ra=0
> sysctl -w net.ipv6.conf.default.send_rs=0
> ifup -a
> sysctl -w net.ipv6.conf.eth0.accept_ra=1
> sysctl -w net.ipv6.conf.eth0.send_rs=1
>
> If you want to listen RA on all interfaces, but do not want to send RS on
> some of them,
> sysctl -w net.ipv6.conf.default.accept_ra=1
> sysctl -w net.ipv6.conf.default.send_rs=0
> ifup -a
> sysctl -w net.ipv6.cont.eth0.send_rs=1
>
>
--
Pekka Savola "You each name yourselves king, yet the
Netcore Oy kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings
|