Return-path: Received: from w1.fi ([128.177.27.249]:33228 "EHLO jmalinen.user.openhosting.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752461Ab1F2N4D (ORCPT ); Wed, 29 Jun 2011 09:56:03 -0400 Date: Wed, 29 Jun 2011 16:55:36 +0300 From: Jouni Malinen To: Michel Alexandre Salim Cc: "John W. Linville" , linux-wireless@vger.kernel.org, ath9k-devel@venema.h4ckr.net, "Luis R. Rodriguez" , Jouni Malinen , Vasanthakumar Thiagarajan , Senthil Balasubramanian Subject: Re: [PATCH] ath9k: add module option for disabling 11n functionality Message-ID: <20110629135536.GB3259@jm.kir.nu> (sfid-20110629_155607_609380_0CDB2CEE) References: <4E0A2F8A.8010102@fedoraproject.org> <20110628215757.GA14975@jm.kir.nu> <4E0AE8C1.9090809@fedoraproject.org> <20110629130910.GA2688@jm.kir.nu> <4E0B2BF8.6090300@fedoraproject.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <4E0B2BF8.6090300@fedoraproject.org> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Wed, Jun 29, 2011 at 03:43:20PM +0200, Michel Alexandre Salim wrote: > The argument would be that driver-specific implementations are > easier to get committed, and less intrusive than a general > framework, but yes, I would prefer a more general solution myself. Changes to the driver go through the same review process and if it makes this any simpler to understand, I'm NACKing the change to ath9k since that is not the correct place for this and module parameter is not the proper mechanism for this type of workaround. > Would this not require modifications to e.g. NetworkManager, ConnMan > etc. as well? While a module parameter is not an ideal workaround, > it's at least easy to use. If this goes in as a connection > parameter, it'd be nice to still have a way of manually adjusting it > through a command-line tool. Yes, doing this properly would indeed require changes somewhere in user space (at least in wpa_supplicant; potentially in other components, too). Were you assuming that users would manually unload and reload the driver module if they happen to hit some issues? Or write some modprobe config files to force 802.11n to be disabled for every connection? Neither sounds like a reasonable "fix" to me. > My first attempt to solve this was to use 'iwconfig wlan0 set rate > 54M' -- but this yields "Operation not supported". How about making > a request for a rate of 54M or below switch the device to 802.11g > mode (and optionally, a rate of 11M or below => 802.11b mode)? That > way we don't need to invent a new interface at all. How would I go > about adding support for rate setting requests to the driver? I would first move from WEXT to nl80211 in order to get any chance of getting the change accepted.. We have a somewhat similar mechanism in NL80211_CMD_SET_TX_BITRATE_MASK, but it does not address HT rates at all, so some additions would be needed. For example, a new attribute to the command could be used to set TX mask for MCS indexes and if none are allowed, HT could be disabled. Though, a cleaner mechanism would likely be to just provide an explicit attribute to disable 802.11n. If that is done outside the scope of a single connection, that would be available as a manual workaround even without modifications to wpa_supplicant/NM/ConnMan. -- Jouni Malinen PGP id EFC895FA