From mboxrd@z Thu Jan 1 00:00:00 1970 From: "David S. Miller" Subject: Re: Restrict local IP announcements in ARP requests Date: Wed, 11 Feb 2004 20:16:56 -0800 Sender: netdev-bounce@oss.sgi.com Message-ID: <20040211201656.4e4f708e.davem@redhat.com> References: <20040209140853.69ab8bea.davem@redhat.com> <20040209151053.1f63937a.davem@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: netdev@oss.sgi.com, linux-net@vger.kernel.org Return-path: To: Julian Anastasov In-Reply-To: Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Tue, 10 Feb 2004 03:00:24 +0200 (EET) Julian Anastasov wrote: > If you like it in this way you can apply now or after > some days or to delay it if we want to combine it with other > device flags or changes. I like the patch, but I want the documentation changed. If you'll recall, one of the big problems Alexey and myself had with the new ARP saddr selection behavior you propose is that it assumes that: if we do not know of a certain subnet being on a link than this subnet does not exist there. In some environments this assumption holds. But "some != all" as I often say to a friend of mine :-) No router has this knowledge completely, and in many environments such an assumption does not hold. I know you understand this, which is why your patch keeps the current behavior the default. But I want the docs for the sysctl to talk about this. Something simple like: + 1 - Try to avoid local addresses that are not in the target's + subnet for this interface as far as we know. In some environments + subnets may be active that this system is not aware of, thus this + setting might cause problems in such cases. Tweak it however you wish, just get the idea/warning across. I'd also prefer if you at least lightly define what "best" source address means in the context of arp_announce==2. Thanks.