Return-path: Received: from mail-bk0-f43.google.com ([209.85.214.43]:39303 "EHLO mail-bk0-f43.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752063Ab3GIOLV convert rfc822-to-8bit (ORCPT ); Tue, 9 Jul 2013 10:11:21 -0400 Received: by mail-bk0-f43.google.com with SMTP id jm2so2372693bkc.2 for ; Tue, 09 Jul 2013 07:11:20 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <20130709102751.GA21674@pandem0nium> References: <1373289250-12259-1-git-send-email-siwu@hrz.tu-chemnitz.de> <1373289250-12259-5-git-send-email-siwu@hrz.tu-chemnitz.de> <20130709102751.GA21674@pandem0nium> Date: Tue, 9 Jul 2013 16:11:19 +0200 Message-ID: (sfid-20130709_161124_410648_C58AF679) Subject: Re: [PATCHv3 4/5] mac80211: add channel switch command and beacon callbacks From: Michal Kazior To: Simon Wunderlich Cc: Johannes Berg , linux-wireless@vger.kernel.org, Mathias Kretschmer , Simon Wunderlich Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 9 July 2013 12:27, Simon Wunderlich wrote: > On Tue, Jul 09, 2013 at 09:17:41AM +0200, Michal Kazior wrote: >> I'm wondering if it's a huge hassle to support drivers that depend on >> channel context API? Perhaps local->open_count could be used instead >> of local->use_chantx and chanctx->refcount? > > Actually I'm not using any chanctx drivers, and as long as I can't test > I prefer to keep things disabled. :) Understandable :) I'm not able to test it now either. Hmm.. Just as an idea: mac80211_hwsim uses chanctx so it could be modified to support CSA to test this case. > local->open_count seems doable though, > it should be the same if there is only one channel context. I can change > that ... If you think it is safe I can enable support, nothing will happen > anyway as long as the driver don't set the "CSA supported" wiphy flag ... I looked at the code and I think local->open_count won't cut it. It will fail with AP_VLAN interfaces present (and this isn't a problem with your current approach). Monitor interfaces are counted in local->monitors, but AP_VLAN aren't counted at all. mac80211 doesn't really keep track of the number of interfaces it has reported to a driver nor the number of software interfaces (that aren't reported to the driver) unless I missed something. Another idea is to verify there's exactly 1 chanctx present and it's refcount is also 1. CSA + chanctx is just a nice to have. If you're short on resources it's fine the way it is I think. >> It seems there's nothing preventing from multiple calls to channel >> switch callback. Is this expected? Can this work if CSA is invoked >> while other CSA is still in progress? > > This should not happen as long as there is only one interface > doing a channel switch. If this is properly checked (need to fix > the point you mentioned above) that should be safe, I think? Yes it's okay as long as userspace doesn't do something stupid (i.e. call channel switch more than once quickly). Whether this is a concern is not for me to decide though. Pozdrawiam / Best regards, Michał Kazior.