From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: [PATCH] net_sched: make traffic control network namespace aware Date: Sun, 11 Apr 2010 19:48:15 +0200 Message-ID: <4BC20B5F.60200@trash.net> References: <20100320014013.GC13239@boeing.com> <4BA7AF32.8020001@trash.net> <4BA7B13C.7020304@trash.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: Tom Goff Return-path: Received: from stinky.trash.net ([213.144.137.162]:37765 "EHLO stinky.trash.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752546Ab0DKRsb (ORCPT ); Sun, 11 Apr 2010 13:48:31 -0400 In-Reply-To: <4BA7B13C.7020304@trash.net> Sender: netdev-owner@vger.kernel.org List-ID: Patrick McHardy wrote: > Patrick McHardy wrote: >> Tom Goff wrote: >>> Mostly minor changes to add a net argument to various functions and >>> remove initial network namespace checks. >>> >>> Make /proc/net/psched per network namespace. >> Looks fine from a qdisc POV. One thing that appears to be missing >> though is teql master netdev registration in other than the initial >> namespace. > > Actually we could take this opportunity and add rtnl_link support > for teql device registration. I can look into this in a couple of > days. I tried to do that, but adding proper netns support is more complicated than I expected. sch_teql registers a qdisc for each master device using the name of the master. Currently qdisc registrations are global, so this doesn't work with network namespaces. We could of course make them per netns, but that would require duplicating all global registrations for each namespace. I'm not convinced that its worth doing this since its only teql that needs it and it doesn't seem to be very useful to use teql in a virtual environment.