From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Rapoport Subject: Re: [PATCH net] net: vxlan: fix crash when interface is created with no group Date: Wed, 26 Mar 2014 19:50:18 +0200 Message-ID: <20140326175018.GA24065@zed> References: <20140326094754.GA19337@zed> <20140325.205324.1845346581147557473.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: David Miller , netdev , Or Gerlitz To: David Stevens Return-path: Received: from mail-ee0-f48.google.com ([74.125.83.48]:38771 "EHLO mail-ee0-f48.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751217AbaCZRuX (ORCPT ); Wed, 26 Mar 2014 13:50:23 -0400 Received: by mail-ee0-f48.google.com with SMTP id b57so1921677eek.21 for ; Wed, 26 Mar 2014 10:50:22 -0700 (PDT) Content-Disposition: inline In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Wed, Mar 26, 2014 at 08:47:19AM -0600, David Stevens wrote: > Mike Rapoport wrote on 03/26/2014 05:47:54 AM: > > > > > It fixes the instant crashes occurring when vxlan interface goes up, but > > I have other crashes with custom vxlan configuration, e.g. in the > > following scenario: > > > > Node A: > > $ ip link add vxlan1 type vxlan id 1 dev ethA > > $ ip addr add dev vxlan1 10.0.0.1/24 > > $ ip link set up dev vxlan1 > > $ bridge fdb append 00:00:00:00:00:00 dev vxlan1 dst > > I thought this MAC address would be an error, but I just noticed that you > submitted a patch adding interpretation of the all zeroes MAC address as > a default FDB. > > That patch, IMO, should've either not gone in, or should've replaced (entirely) > the existing default_dst. I think the problems you are seeing are due to having > two "defaults". > > Since the existing default_dst could be either a multicast group or a specific > destination, I don't understand the point of your patch: > afbd8bae9c798c5cdbe4439d3a50536b5438247c > > except possibly to support it as a list. In that case, at least that patch should've > removed default_dst and used the all-zeroes fdb entry instead, including group > joins and leaves when it is a multicast IP address. > > I certainly don't see the point of having both. The idea was to have a list of multiple destinations allow a weird form of multicast when infrastructure does not support it. I don't remember now why I didn't follow your advice about removing default_dst then (1). Maybe I had some thoughts that seemed smart and maybe it was just lazines. And now I definitely regret :) The problem is not only duplication of default destinations, but also the mis^Wuse of default_dst to distinguish whether IPv4 or IPv6 path should be taken because there is no explicit specification of the protocol if neither IFLA_VXLAN_GROUP or IFLA_VXLAN_LOCAL defined at vxlan_newlink. Moreover, if neither of these attributes is present, the custom configuration will always use v4 socket regardles of use of default_dst or fdb entry. > +-DLS > -- [1] http://thread.gmane.org/gmane.linux.network/270969/focus=271830 -- Sincerely yours, Mike.