Return-path: Received: from mail-bw0-f227.google.com ([209.85.218.227]:44317 "EHLO mail-bw0-f227.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752201AbZLALoC (ORCPT ); Tue, 1 Dec 2009 06:44:02 -0500 Received: by bwz27 with SMTP id 27so3332971bwz.21 for ; Tue, 01 Dec 2009 03:44:08 -0800 (PST) To: Johannes Berg Cc: John Linville , linux-wireless@vger.kernel.org Subject: Re: [PATCH 1/2] cfg80211: add SM PS hooks References: <20091130153857.963079149@sipsolutions.net> <20091130154300.492667106@sipsolutions.net> <87pr6z6uud.fsf@purkki.valot.fi> <1259659771.32171.62.camel@johannes.local> From: Kalle Valo Date: Tue, 01 Dec 2009 13:44:06 +0200 In-Reply-To: <1259659771.32171.62.camel@johannes.local> (Johannes Berg's message of "Tue\, 01 Dec 2009 10\:29\:31 +0100") Message-ID: <871vje7wc9.fsf@purkki.valot.fi> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-wireless-owner@vger.kernel.org List-ID: Johannes Berg writes: > On Tue, 2009-12-01 at 09:01 +0200, Kalle Valo wrote: >> My concern here is that we will end up having, yet again, complicated >> user space interface for power save. The ideal situation would be that >> kernel would configure all this automatically and we would have a >> simple interface just to disable power save in cases where it doesn't >> work (broken APs etc). > > I'm not actually sure that there are APs broken wrt. SM PS, it seems > highly unlikely -- even if there are (like mac80211!!) the rate control > algorithm would recover the situation. Oh, that's good. > As such, I suppose it's somewhat a debugging API. On the other hand, > there are actual traffic throughput consequences of enabling SM PS, even > when in powersave. Powersave will be turned off automatically if there's > enough traffic, but SM PS won't. The default, automatic, means it just > follows powersave (dynamic SM PS if on, no SM PS if off), but I'm not > sure that is typically desirable. > > Of course, on the other hand, I don't see many users actually setting > it. An alternative could be to try to estimate traffic, but that gets > messy fairly quickly because the thresholds depend on the environment. I think at some point we need to start estimating the traffic somehow and having heuristics to dynamically change the power save settings. But it's a huge challenge, so it isn't coming anytime soon. > If you prefer to leave it out for now, I can probably move it to debugfs > until we have a more clearly defined use case, Great if you can convert it to debugfs for now. That way we have a bit more time to think about the power save interface. Thanks. > although if we decide that we never want it then obviously the > action frame handling can go away again. Yeah. -- Kalle Valo