From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Hemminger Subject: Re: [RFC] ipv6: don't flush routes when setting loopback down Date: Sun, 23 Jan 2011 09:39:40 +1100 Message-ID: <20110123093940.4f6075a3@s6510> References: <20101209.122033.183046393.davem@davemloft.net> <20101216132812.2d7fd885@nehalam> <20101216.182656.226781473.davem@davemloft.net> <20110119191823.GC8442@midget.suse.cz> <20110119113817.0819ddf1@s6510> <20110119195632.GA27574@midget.suse.cz> <20110119120123.40974cbe@s6510> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: Jiri Bohac , David Miller , brian.haley@hp.com, netdev@vger.kernel.org, maheshkelkar@gmail.com, lorenzo@google.com, yoshfuji@linux-ipv6.org, stable@kernel.org To: ebiederm@xmission.com (Eric W. Biederman) Return-path: Received: from mail.vyatta.com ([76.74.103.46]:41116 "EHLO mail.vyatta.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750872Ab1AVWjs (ORCPT ); Sat, 22 Jan 2011 17:39:48 -0500 In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Sat, 22 Jan 2011 00:17:09 -0800 ebiederm@xmission.com (Eric W. Biederman) wrote: > Stephen Hemminger writes: > > > On Wed, 19 Jan 2011 20:56:32 +0100 > > Jiri Bohac wrote: > > > >> On Wed, Jan 19, 2011 at 11:38:17AM -0800, Stephen Hemminger wrote: > >> > Jiri Bohac wrote: > >> > > I have the feeling that Eric's patch is the safest solution we > >> > > have so far: > >> > Eric's patch has other regressions, see the discussion. > >> > >> What regression do you mean? I have read the whole discussion > >> thoroughly. You only say in one message that deleting ::1 would > >> propagate to routing daemons. And Eric correctly stated that > >> people couldn't hit this, because deleting ::1 would break > >> things on its own. > >> > >> Is there a real problem with Eric's fix? > >> > >> Thanks, > >> > > > > If address is assigned to loopback interface (other than ::1) then > > Eric's fix doesn't work. It is common to use an additional address > > on the lo device when doing routing protocols. > > Sigh. > > I just got back to looking through the rest of my failures in 2.6.37 and > despite it looking like it worked when i tested it, your patch doesn't > actually work on my real work load that has broken. > > At least your change that confirmed that the root problem is somewhere > in the routing. > > Eric The design problem behind all this is that sysctl disable_ipv6 as currently implemented is passive (just changes a variable). It needs to be implemented as a more active step that does the same thing as removing the interface from ipv6. I will look into it after LCA.