From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from fg-out-1718.google.com ([72.14.220.152]:8722 "EHLO fg-out-1718.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751913AbYGNMWl (ORCPT ); Mon, 14 Jul 2008 08:22:41 -0400 Received: by fg-out-1718.google.com with SMTP id 19so2079222fgg.17 for ; Mon, 14 Jul 2008 05:22:40 -0700 (PDT) Date: Mon, 14 Jul 2008 12:27:33 +0000 From: Jarek Poplawski To: David Miller Cc: netdev@vger.kernel.org, vinay@linux.vnet.ibm.com, krkumar2@in.ibm.com, mchan@broadcom.com, Matheos.Worku@Sun.COM, linux-wireless@vger.kernel.org Subject: Re: [PATCH 11/39]: netdev: Move next_sched into struct netdev_queue. Message-ID: <20080714122733.GA4573@ff.dom.local> (sfid-20080714_142246_243442_51BDF33A) References: <20080703.000333.199735542.davem@davemloft.net> <20080714114913.GA4504@ff.dom.local> <20080714.050011.186100773.davem@davemloft.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <20080714.050011.186100773.davem@davemloft.net> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Mon, Jul 14, 2008 at 05:00:11AM -0700, David Miller wrote: > From: Jarek Poplawski > Date: Mon, 14 Jul 2008 11:49:13 +0000 > > > On 03-07-2008 09:03, David Miller wrote: > > > We schedule queues, not the device, for output queue > > > processing in BH. > > > > > ... > > > diff --git a/net/core/dev.c b/net/core/dev.c > > > index 8e38948..c5e3532 100644 > > > --- a/net/core/dev.c > > > +++ b/net/core/dev.c > > > @@ -1323,13 +1323,14 @@ static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev) > > > void __netif_schedule(struct net_device *dev) > > > { > > > if (!test_and_set_bit(__LINK_STATE_SCHED, &dev->state)) { > > > > Isn't this bit supposed to be set/tested per queue as well, or do I > > miss something? > > That shoule not happen until my later patches which switch that all > over. > > See patch 5/13 which I posted on Thursday, Jul 10. That's when the > transition from "global queue state + per-queue state" into "only > per-queue state" happens. > > Until that point, we have to keep the existing quasi-multi-queue > semantics in order to not break drivers etc. I've just wondered how this 1-13/13 step can work without this... Jarek P.