Return-path: Received: from mail-bk0-f47.google.com ([209.85.214.47]:62741 "EHLO mail-bk0-f47.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751037AbaAOOLV convert rfc822-to-8bit (ORCPT ); Wed, 15 Jan 2014 09:11:21 -0500 Received: by mail-bk0-f47.google.com with SMTP id mx12so714869bkb.20 for ; Wed, 15 Jan 2014 06:11:19 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <1389792568.4338.12.camel@jlt4.sipsolutions.net> References: <1389787494-7361-1-git-send-email-michal.kazior@tieto.com> <1389787494-7361-7-git-send-email-michal.kazior@tieto.com> <1389791963.4338.4.camel@jlt4.sipsolutions.net> <1389792568.4338.12.camel@jlt4.sipsolutions.net> Date: Wed, 15 Jan 2014 15:11:19 +0100 Message-ID: (sfid-20140115_151124_192756_EC5BA030) Subject: Re: [RFC 6/9] mac80211: track CSA globally From: Michal Kazior To: Johannes Berg Cc: linux-wireless Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 15 January 2014 14:29, Johannes Berg wrote: > On Wed, 2014-01-15 at 14:28 +0100, Michal Kazior wrote: >> On 15 January 2014 14:19, Johannes Berg wrote: >> > On Wed, 2014-01-15 at 13:04 +0100, Michal Kazior wrote: >> >> For CSA to be safe it needs to be treated the same >> >> way as radar detection, scanning and remain on >> >> channel - all of those (including CSA) must be >> >> mutually exclusive. >> > >> > This I don't understand. Why couldn't you do a remain-on(some >> > other)-channel or scan while counting down the beacons? >> >> My concern is software offchannel (be it scan or roc) involves channel >> context switches. I wanted to avoid any channel context mangling while >> CSA is in progress. Does that make sense to you? > > No, I don't really get it. Why? Hmm. Perhaps I was a little overcautious. Offchannel stuff doesn't use channel contexts at all, right? It recalculates channel in hw_config() and doesn't touch channel contexts so it should be safe. Also, I just noticed the patch contains some code that doesn't belong here (vif_use_channel is protected with the csa_is_active() -- this should be in the final commit). And this is probably the only part that is necessary - to prevent anything new from binding to a channel context that is being used for CSA. Michał