From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261409AbTISIGk (ORCPT ); Fri, 19 Sep 2003 04:06:40 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261421AbTISIGk (ORCPT ); Fri, 19 Sep 2003 04:06:40 -0400 Received: from mail.videonetworks.com ([193.128.170.146]:51463 "EHLO natst.videonetworks.com") by vger.kernel.org with ESMTP id S261409AbTISIGj (ORCPT ); Fri, 19 Sep 2003 04:06:39 -0400 Message-ID: <3F6ABA4A.64DF0652@videonetworks.com> Date: Fri, 19 Sep 2003 09:11:54 +0100 From: Dominic Robinson X-Mailer: Mozilla 4.72 [en] (X11; U; Linux 2.2.14-5.0 i686) X-Accept-Language: en MIME-Version: 1.0 To: linux-kernel@vger.kernel.org Subject: 2.4.22 multicast address problem, pcap workaround Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Please cc to d.robinson@videonetworks.com My source machine sends multicast packets on all the following addresses: 230.1.1.1, 230.1.1.2 through to 230.1.1.64 My target machine (on local network, or with cross-over cable) joins all these multicast groups, and receives packets on all of these addresses EXCEPT 230.1.1.18 and 230.1.1.32. However, if I run tcpdump at the same time as my program, I find my program suddenly starts receiving on 230.1.1.[18,32] If I kill tcpdump, my program stops receiving. It's the "pcap_open_live()" function in tcpdump which causes this, and as a workaround I just call it direct in my receiver. Could this be a bug in the multicast kernel code? NOTES: (1) 230.1.1.1 to 230.1.64.1 and 230.1.1.1 to 230.64.1.1 has the same problem. (2) 230.1.1.1 to 230.1.1.63 works (3) 230.1.1.1 to 230.1.1.128 fails with slightly different behaviour (4) It makes no difference if I run a single program listening on all addresses, or 64 individual programs each listening on a single address. Thanks and regards, Dominic