Return-path: Received: from he.sipsolutions.net ([78.46.109.217]:59527 "EHLO sipsolutions.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755029Ab0JDSjb (ORCPT ); Mon, 4 Oct 2010 14:39:31 -0400 Subject: Re: [PATCH v2 3/3] mac80211: move to the home channel for disassociation when roaming From: Johannes Berg To: "Luis R. Rodriguez" Cc: Luis Rodriguez , "linville@tuxdriver.com" , "linux-wireless@vger.kernel.org" , "stable@kernel.org" , Jouni Malinen , Paul Stewart , Amod Bodas , Vasanth Thiagarajan In-Reply-To: <20101004172333.GK2105@tux> References: <1285965233-11097-1-git-send-email-lrodriguez@atheros.com> <1285965233-11097-4-git-send-email-lrodriguez@atheros.com> <1286198152.3620.35.camel@jlt3.sipsolutions.net> <20101004163829.GD2105@tux> <1286210516.3620.41.camel@jlt3.sipsolutions.net> <20101004172333.GK2105@tux> Content-Type: text/plain; charset="UTF-8" Date: Mon, 04 Oct 2010 20:39:29 +0200 Message-ID: <1286217569.3620.48.camel@jlt3.sipsolutions.net> Mime-Version: 1.0 Sender: linux-wireless-owner@vger.kernel.org List-ID: On Mon, 2010-10-04 at 10:23 -0700, Luis R. Rodriguez wrote: > On Mon, Oct 04, 2010 at 09:41:56AM -0700, Johannes Berg wrote: > > On Mon, 2010-10-04 at 09:38 -0700, Luis R. Rodriguez wrote: > > > On Mon, Oct 04, 2010 at 06:15:52AM -0700, Johannes Berg wrote: > > > > On Fri, 2010-10-01 at 16:33 -0400, Luis R. Rodriguez wrote: > > > > > > > > > @@ -2203,6 +2204,14 @@ int ieee80211_mgd_assoc(struct ieee80211_sub_if_data *sdata, > > > > > return -EALREADY; > > > > > } > > > > > > > > > > + /* > > > > > + * right before this was authentication, that was on > > > > > + * the same the wk->chan so we need to ensure we temporarily > > > > > + * go back to the operating channel to send the disassocation. > > > > > + */ > > > > > + local->tmp_channel = NULL; > > > > > + ieee80211_hw_config(local, 0); > > > > > + > > > > > > > > Yikes, no, you can't do this. You don't know what has been merged with > > > > the authentication work item, etc. > > > > > > What do you mean? Auth/Assoc will be handled separately and the > > > target channel will also be handled for each work item run, what > > > issues can arise from sending us back to the home channel towards > > > the end of this call here? > > > > Another work item, like a remain-on-channel, might have been started on > > the same channel ("merged"). > > But how do this break that other work from moving forward in its respective > channel? The work_work loop makes sure to check if we require a channel > change on the work item. But we already activated that other work item. johannes