From mboxrd@z Thu Jan 1 00:00:00 1970 From: Toshiaki Makita Subject: Re: [PATCH net-next 2/4] net: dev: don't set the same mac address for netdev Date: Thu, 05 Jun 2014 19:51:34 +0900 Message-ID: <53904BB6.7030803@lab.ntt.co.jp> References: <1401951028-9800-1-git-send-email-dingtianhong@huawei.com> <1401951028-9800-3-git-send-email-dingtianhong@huawei.com> <539033BC.3000703@lab.ntt.co.jp> <53903D5A.7050702@huawei.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: Ding Tianhong , kaber@trash.net, davem@davemloft.net, edumazet@google.com, vyasevic@redhat.com Return-path: Received: from tama50.ecl.ntt.co.jp ([129.60.39.147]:44180 "EHLO tama50.ecl.ntt.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751507AbaFEKv6 (ORCPT ); Thu, 5 Jun 2014 06:51:58 -0400 In-Reply-To: <53903D5A.7050702@huawei.com> Sender: netdev-owner@vger.kernel.org List-ID: (2014/06/05 18:50), Ding Tianhong wrote: > On 2014/6/5 17:09, Toshiaki Makita wrote: >> (2014/06/05 15:50), Ding Tianhong wrote: >>> Most of netdev just like bond, team, vlan will set the mac address >>> and propagate to the upperdev or lowerdev regardless the mac address >>> is same or not, I could not find that the same mac address could >>> make affect, so add equal check when set mac address. >>> >>> Signed-off-by: Ding Tianhong >>> --- >>> net/core/dev.c | 2 ++ >>> 1 file changed, 2 insertions(+) >>> >>> diff --git a/net/core/dev.c b/net/core/dev.c >>> index 5367bfb..4008a51 100644 >>> --- a/net/core/dev.c >>> +++ b/net/core/dev.c >>> @@ -5570,6 +5570,8 @@ int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) >>> return -EINVAL; >>> if (!netif_device_present(dev)) >>> return -ENODEV; >>> + if (ether_addr_equal_64bits(dev->dev_addr, sa->sa_data)) >>> + return 0; >>> err = ops->ndo_set_mac_address(dev, sa); >>> if (err) >>> return err; >>> >> >> Bridge uses addr_assign_type to check if bridge_id can be propageted by >> bridge ports. If user set mac address, and even if it is the same as >> current one, bridge uses the fact that the mac address is set by user. >> > > OK > >> Although I'm not aware of a driver that needs calling of >> ndo_set_mac_address() for the same mac address, this change looks a bit >> risky to me. >> (For example, old bridge code needed this call because it managed >> BR_SET_MAC_ADDR in bridge flags.) >> > Except the old bridge, I still don't think any other driver need to call ndo_set_mac_address() > for the same mac address, if the dev_set_mac_address() don't do anything for the same address, > I think some drivers should ignore the same mac address themselves just like bonding, macvlan, vlan and so on. Though I don't know why you think this is safe, looking over some drivers, br2684_mac_addr() seems to use a logic similar to old bridge's. Thanks, Toshiaki Makita