From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Jan Beulich" Subject: Re: [xen-4.0-testing test] 2045: regressions - FAIL Date: Fri, 17 Sep 2010 15:09:49 +0100 Message-ID: <4C9392CD02000078000176DC@vpn.id2.novell.com> References: <20100825154011.GO2804@reaktio.net> <19573.24591.154634.863643@mariner.uk.xensource.com> <1282806390.3469.22.camel@localhost.localdomain> <19581.4921.657543.786718@mariner.uk.xensource.com> <1283267148.12544.9437.camel@zakaz.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <1283267148.12544.9437.camel@zakaz.uk.xensource.com> Content-Disposition: inline List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Ian Campbell , Ian Jackson Cc: "xen-devel@lists.xensource.com" List-Id: xen-devel@lists.xenproject.org >>> On 31.08.10 at 17:05, Ian Campbell wrote: > On Tue, 2010-08-31 at 15:35 +0100, Ian Jackson wrote: >>=20 >> > Even with the fix in place the gratuitous ARP behaviour is disabled >> by >> > default so you need to enable the net.ipv4.conf..arp_notify >> sysctl >> > for any device you want to send the notifications. When I was >> testing I >> > did this by adding >> > net.ipv4.conf.default.arp_notify =3D 1 >> > to /etc/sysctl.conf and that seemed to do the trick. >>=20 >> I think this is a bug. I think the default should be for something to >> send this gratuitous arp and the most logical answer in the PV or >> PV-on-HVM case is the domU. >=20 > This is the upstream default, not something we control directly from > netfront etc. Wouldn't it seem possible/reasonable to force this on from netfront (via IN_DEV_CONF_SET()) at least until upstream possibly decides to not let the ARP_NOTIFY sysctl control the sending of an ARP explicitly requested through netif_notify_peers()? The perhaps undesirable effect of this (as well as setting the sysctl through /etc/sysctl.conf) is that it doesn't only control the ARP we want sent, but also one when bringing the interface up or changing its address, so I'd still favor moving the NETDEV_NOTIFY_PEERS case past the "if (IN_DEV_ARP_NOTIFY(in_dev))" in inetdev_event(). Have you got any feedback from Linux folks on such a potential change? Jan