From mboxrd@z Thu Jan 1 00:00:00 1970 From: Simon Kelley Subject: Please explain this ARP behaviour. Date: Wed, 26 Apr 2006 12:58:30 +0100 Message-ID: <444F6066.8020805@thekelleys.org.uk> Reply-To: simon@thekelleys.org.uk Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from cpc1-cmbg4-0-0-cust914.cmbg.cable.ntl.com ([81.98.247.147]:28427 "EHLO thekelleys.org.uk") by vger.kernel.org with ESMTP id S932405AbWDZL6b (ORCPT ); Wed, 26 Apr 2006 07:58:31 -0400 Received: from guest491.wtgc.org ([193.62.205.238]) by thekelleys.org.uk with asmtp (Exim 3.36 #1 (Debian)) id 1FYif8-00005L-00 for ; Wed, 26 Apr 2006 12:58:30 +0100 To: netdev@vger.kernel.org Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Two machines, many configuration differences, I need to find which one is causing this difference in behaviour: Machine 1: (eth1 is on the correct 192.168.3.x network) # ip neighbour add to 192.168.2.15 lladdr 00:0F:20:98:F8:86 nud \ reachable dev eth1 # cat /proc/net/arp IP address HW type Flags HW address Mask Device 192.168.0.4 0x1 0x2 00:00:F8:04:9B:69 * eth0 192.168.2.15 0x1 0x2 00:0F:20:98:F8:86 * eth1 192.168.2.148 0x1 0x2 00:10:A4:F2:CF:CB * eth1 # ip neighbour show 192.168.0.4 dev eth0 lladdr 00:00:f8:04:9b:69 nud reachable 192.168.2.15 dev eth1 lladdr 00:0f:20:98:f8:86 nud reachable 192.168.2.148 dev eth1 lladdr 00:10:a4:f2:cf:cb nud stale Note that hand-created entry appears in /proc and ip .. show Machine 2: (similarly correct network) # ip neighbour add to 192.62.204.15 lladdr 00:0F:20:98:F8:86\ nud reachable dev eth1 IP address HW type Flags HW address Mask Device 193.62.204.1 0x1 0x2 00:0F:20:98:F8:87 * eth1 # ip neighbour show 192.62.204.15 dev eth1 lladdr 00:0f:20:98:f8:86 nud reachable 193.62.204.1 dev eth1 lladdr 00:0f:20:98:f8:87 nud stale Here, hand-created interface is in ip .. show, but missing from /proc There is some reason to believe that the functional ARP behaves as /proc/net/arp (ie the hand-created entry is ignored) but I don't have absolute proof of that yet. Machine 1 (works) is 2.6.12.2 hand compiled Machine 2 (broken) is Ubunutu's 2.6.10-5-386, but this has possibly been seen in Debian 2.6.16.2 as well. Please could someone suggest where I look next? Cheers, Simon.