From mboxrd@z Thu Jan 1 00:00:00 1970 From: Murali Karicheri Subject: IPv6 IGMP issue in v4.4.44 ?? Date: Tue, 21 Mar 2017 18:00:36 -0400 Message-ID: <58D1A284.7050802@ti.com> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit To: David Miller , "open list:TI NETCP ETHERNET DRIVER" Return-path: Received: from lelnx194.ext.ti.com ([198.47.27.80]:52459 "EHLO lelnx194.ext.ti.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757292AbdCUWAi (ORCPT ); Tue, 21 Mar 2017 18:00:38 -0400 Sender: netdev-owner@vger.kernel.org List-ID: Hello David, experts, I see an issue with IGMP for IPv6 when I test HSR redundancy network interface. As soon as I set up an HSR interface, I see some IGMP messages (destination mac address: 33 33 00 00 00 02 going over HSR interface to slave interfaces, at the egress where as for IPv6, I see similar messages going directly over the Ethernet interfaces that are attached to HSR master. It appears that the NETDEV_CHANGEUPPER is not handled properly and the mcast snoop sends the packets over the old interfaces at timer expiry. A dump of the message at the slave Ethernet interface looks like below. IPv4 [ 64.643842] 33 33 00 00 00 02 70 ff 76 1c 0f 8d 89 2f 10 3e fc [ 64.649910] 18 86 dd 60 00 00 00 00 10 3a ff fe 80 00 00 00 [ 64.655705] 00 00 00 72 ff 76 ff fe 1c 0f 8d ff 02 00 00 00 [ 64.661503] 00 00 00 00 00 00 00 00 00 00 02 85 00 8d dc You can see this is tagged with HSR. IPv6 [ 65.559130] 33 33 00 00 00 02 70 ff 76 1c 0f 8d 86 dd 60 00 00 [ 65.565205] 00 00 10 3a ff fe 80 00 00 00 00 00 00 72 ff 76 [ 65.571011] ff fe 1c 0f 8d ff 02 00 00 00 00 00 00 00 00 00 [ 65.576806] 00 00 00 00 02 85 00 8d dc 00 00 00 00 01 01 This is going directly to the slave Ethernet interface. When I put a WARN_ONCE, I found this is coming directly from mld_ifc_timer_expire() -> mld_sendpack() -> ip6_output() Do you think this is fixed in latest kernel at master? If so, could you point me to some commits. Thanks -- Murali Karicheri Linux Kernel, Keystone