Return-path: Received: from nbd.name ([46.4.11.11]:55303 "EHLO nbd.name" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759792Ab1CDTeM (ORCPT ); Fri, 4 Mar 2011 14:34:12 -0500 Message-ID: <4D713EAD.1020803@openwrt.org> Date: Fri, 04 Mar 2011 20:34:05 +0100 From: Felix Fietkau MIME-Version: 1.0 To: Helmut Schaa CC: linville@tuxdriver.com, linux-wireless@vger.kernel.org, johannes@sipsolutions.net Subject: Re: [PATCH 2/2] mac80211: Shortcut minstrel_ht rate setup for non-MRR capable devices References: <1299241892-6719-1-git-send-email-helmut.schaa@googlemail.com> <1299241892-6719-2-git-send-email-helmut.schaa@googlemail.com> <4D711C04.8030309@openwrt.org> <201103041823.04001.helmut.schaa@googlemail.com> In-Reply-To: <201103041823.04001.helmut.schaa@googlemail.com> Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 2011-03-04 6:23 PM, Helmut Schaa wrote: > Am Freitag, 4. M?rz 2011 schrieb Felix Fietkau: >> On 2011-03-04 1:31 PM, Helmut Schaa wrote: >> > Devices without multi rate retry support won't be able to use all rates >> > as specified by mintrel_ht. Hence, we can simply skip setting up further >> > rates as the devices will only use the first one. >> > >> > Signed-off-by: Helmut Schaa >> > --- >> > >> > has_mrr is true if the device is able to make use of at least four rates. Not >> > sure if we might want to consider further special cases for devices with for >> > example a maximum of two rates. >> I think we should add a special case for handling a maximum of two >> rates. Right now this is being used by brcm80211, so applying this patch >> might make things worse there. > > John, please drop this patch then. > > Felix, currently minstrel_ht uses a maximum of 3 rates as far as I can see. > > In the sampling case it's > sample_rate -> highest_tp_rate -> highest_prob_rate > and in the default case it's > highest_tp_rate -> second_highest_tp_rate -> highest_prob_rate > > For devices with just one possible rate it would be either the sample_rate > or the highest_tp_rate. > > And for devices with two rates it might make sense to change the logic to > something like > sample_rate -> highest_prob_rate > and > highest_tp_rate -> highest_prob_rate > > Does that sound reasonable? Yes, sounds good. - Felix