From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hagen Paul Pfeifer Subject: Re: [PATCH V2] net/sched: =?UTF-8?Q?sch=5Fplug=20-=20Queue=20traf?= =?UTF-8?Q?fic=20until=20an=20explicit=20release=20command?= Date: Tue, 31 Jan 2012 16:25:06 +0100 Message-ID: <6790d1d5d8aa930222bbfc2ef6a9d48e@localhost> References: <1327823619-29274-1-git-send-email-rshriram@cs.ubc.ca> <1327934756.2243.3.camel@mojatatu> <313929ea4dca3c8fd6948797e03121a8@localhost> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: Jamal Hadi Salim , " " , Brendan Cully To: Shriram Rajagopalan Return-path: Received: from alternativer.internetendpunkt.de ([88.198.24.89]:49051 "EHLO geheimer.internetendpunkt.de" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1753506Ab2AaPZJ (ORCPT ); Tue, 31 Jan 2012 10:25:09 -0500 In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Mon, 30 Jan 2012 08:47:45 -0800, Shriram Rajagopalan wrote: > Could you elaborate a little on the packet-based-unplug ? > > I got your earlier comment on "indefinite unplug" until an explicit > plug is received. Is that what you mean by packet-based-unplug ? Sure, imagine a multihop MANET network. Sometimes we have high-priority crosstraffic in the next hop (router). Due to OLSR traffic information we know in advance that the next hop is not in the ability to forward our low priority packets. With this knowledge we can stop (unplug) local generated and forwarded traffic and if the next hop has enough free bandwidth we can restart (plug) sending already enqueued packets. So for us a really simple plug/unplug mechanism is superior. Maybe a head-drop FIFO based policy for forwarded traffic but I can provide a patch on top of your patch to implement a head-drop policy. Hagen