From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Madore Subject: MTU blackhole discovery for IPv6 ? Date: Thu, 7 May 2009 23:19:26 +0200 Message-ID: <20090507211926.GA28662@regulus.madore.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii To: Linux NetDev Mailing-List Return-path: Received: from regulus.xn--kwg.net ([212.85.152.99]:58990 "EHLO regulus.xn--kwg.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751298AbZEGV0S (ORCPT ); Thu, 7 May 2009 17:26:18 -0400 Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: Hi, There's a sysctl called net.ipv4.tcp_mtu_probing: has there been any work done on the analogous key for IPv6? I'm encountering a number of asinine network providers who either discard ICMPs (whether v4 or v6...) or simply don't emit them. The IPv6 picture is even bleaker than IPv4 because so much traffic goes through tunnels like 6to4. An IPv6 MTU blackhole discovery implementation would truly save my day. I understand that RFC 4821 specifies the same algorithms for packetization layer path MTU discovery in IPv4 and IPv6. Did Linux implement this in IPv4 only because there is some difficulty in doing the same thing for IPv6, or simply because nobody was asking for it? Happy hacking, -- David A. Madore ( http://www.madore.org/~david/ )