Return-path: Received: from charlotte.tuxdriver.com ([70.61.120.58]:45844 "EHLO smtp.tuxdriver.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752936Ab1FGSQY (ORCPT ); Tue, 7 Jun 2011 14:16:24 -0400 Date: Tue, 7 Jun 2011 14:05:53 -0400 From: "John W. Linville" To: "Manoharan, Rajkumar" Cc: Johannes Berg , "linux-wireless@vger.kernel.org" Subject: Re: [PATCH 6/6] mac80211: stop queues before rate control updation Message-ID: <20110607180552.GF9320@tuxdriver.com> (sfid-20110607_201627_681498_BD86CFE6) References: <1305894135-14036-1-git-send-email-rmanoharan@atheros.com> <1305894135-14036-6-git-send-email-rmanoharan@atheros.com> <1306823717.9366.3.camel@jlt3.sipsolutions.net> <8F3AF1C9F856774F8C8D67AA7EDFEC88012CC51B@nasanexd02b.na.qualcomm.com> <20110606184447.GG2604@tuxdriver.com> <1307386634.3894.14.camel@jlt3.sipsolutions.net> <8F3AF1C9F856774F8C8D67AA7EDFEC88012CE709@nasanexd02d.na.qualcomm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <8F3AF1C9F856774F8C8D67AA7EDFEC88012CE709@nasanexd02d.na.qualcomm.com> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Tue, Jun 07, 2011 at 08:39:57AM +0000, Manoharan, Rajkumar wrote: > > > On Mon, 2011-06-06 at 14:44 -0400, John W. Linville wrote: > > > > > > During the channel type change, the pending tx frames in hw queues are dropped by hw config. > > > > But before updating rate control, the packets can be queued again with older HT rates. > > > > This contradicts with hw config mode and sometimes is causing baseband issues. This issue > > > > was observed only on flooding uplink traffic. To ensure that the frames are always xmitted with > > > > updated rates, the queues are stopped before hw config and waken up after rc updation. > >> > > > Johannes, do you find this explanation satisfactory (perhaps with > > > some new queue stop reason definition)? > > > > No, it doesn't address the fact that any packet that is pending will > > still be processed -- as I said before, the stop doesn't include a > > flush. > > Yes. I agree. Without flushing, still the packets can choose wrong rates. I missed that. > But I assumed that it would be better to stop queues before rc update. Thus we can avoid revisiting > the queued frames chosen with older (this case ht40) rate after rc changes. And fixing at mac80211 would help > to other rate controls too. OK, it seems like this isn't getting resolved quickly. I'm going to revert it for now, and hope for a more widely acceptable solution soon. John -- John W. Linville Someday the world will need a hero, and you linville@tuxdriver.com might be all we have. Be ready.