From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from he.sipsolutions.net ([78.46.109.217]:54981 "EHLO sipsolutions.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753511Ab2KIVRO (ORCPT ); Fri, 9 Nov 2012 16:17:14 -0500 Message-ID: <1352495862.30006.1.camel@jlt4.sipsolutions.net> (sfid-20121109_221719_702580_AAD693CD) Subject: Re: [PATCH] nl/cfg80211: advertise OBSS scan offload capability From: Johannes Berg To: Bing Zhao Cc: "linux-wireless@vger.kernel.org" , Jouni Malinen , "John W. Linville" , Amitkumar Karwar Date: Fri, 09 Nov 2012 22:17:42 +0100 In-Reply-To: <477F20668A386D41ADCC57781B1F7043083BBD703C@SC-VEXCH1.marvell.com> References: <1352415240-5168-1-git-send-email-bzhao@marvell.com> <1352478572.9238.14.camel@jlt4.sipsolutions.net> <477F20668A386D41ADCC57781B1F7043083BBD703C@SC-VEXCH1.marvell.com> Content-Type: text/plain; charset="UTF-8" Mime-Version: 1.0 Sender: linux-wireless-owner@vger.kernel.org List-ID: On Fri, 2012-11-09 at 12:18 -0800, Bing Zhao wrote: > > > New flag is added to inform userspace that this driver doesn't > > > internally perform OBSS scan. It can use this feature available > > > in userspace application. > > > > Shouldn't that be a little more precisely worded? Most drivers already > > expect wpa_s to do this, but apparently the assumption was that without > > the flag, full-MAC drivers, i.e. drivers implementing the nl80211 > > connect API (rather than auth/assoc) do it themselves? > > That's right. > > > > > Or maybe mac80211 should just set the flag? > > Currently wpa_s checks WPA_DRIVER_FLAGS_SME flag for OBSS scan. > Since WPA_DRIVER_FLAGS_SME bit is set for mac80211, it's unnecessary to set this new flag. > > Or, are you suggesting that we should set the new flag in mac80211 > (and all drivers that need OBSS scan feature in wpa_s) and replace > WPA_DRIVER_FLAGS_SME check with the new flag? I don't think that can be done, for the simple reason that wpa_s should continue to work on older kernels. Thus, I guess that wpa_s is going to continue with the assumption that auth/assoc API implies also doing OBSS in wpa_s (which is a fair assumption), but then I think we should simply change the documentation of this new feature flag to indicate that it's interesting only in the case of a driver that has connect API. johannes