Return-path: Received: from mail-wi0-f173.google.com ([209.85.212.173]:52207 "EHLO mail-wi0-f173.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750895AbaK1Gy5 convert rfc822-to-8bit (ORCPT ); Fri, 28 Nov 2014 01:54:57 -0500 Received: by mail-wi0-f173.google.com with SMTP id r20so17598863wiv.6 for ; Thu, 27 Nov 2014 22:54:56 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <1417091446-16308-1-git-send-email-emmanuel.grumbach@intel.com> References: <1417091446-16308-1-git-send-email-emmanuel.grumbach@intel.com> Date: Fri, 28 Nov 2014 07:54:56 +0100 Message-ID: (sfid-20141128_075505_972449_FBA60A41) Subject: Re: [RFC] mac80211: take reserved vif into account when calculating the min_def From: Michal Kazior To: Emmanuel Grumbach Cc: linux-wireless Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 27 November 2014 at 13:30, Emmanuel Grumbach wrote: > When we want to calculate the minimal bandwidth needed for > a channel context, we need to take into account vifs that > have reserved the channel context. > I hit an issue with iwlwifi and channel switch as a client. > > We would allocate a virgin channel context and reserve it. > At that stage, the min_def was 20MHz. > Then we would use it after CSA, and start transmitting, but > the channel context was still 20MHz even if the GO was in > 40MHz. This made the firmware unhappy. > > Signed-off-by: Emmanuel Grumbach > --- > net/mac80211/chan.c | 6 +++++- > 1 file changed, 5 insertions(+), 1 deletion(-) > > diff --git a/net/mac80211/chan.c b/net/mac80211/chan.c > index 4c74e8d..769e0c5 100644 > --- a/net/mac80211/chan.c > +++ b/net/mac80211/chan.c > @@ -256,7 +256,8 @@ ieee80211_get_chanctx_max_required_bw(struct ieee80211_local *local, > if (!ieee80211_sdata_running(sdata)) > continue; > > - if (rcu_access_pointer(sdata->vif.chanctx_conf) != conf) > + if (rcu_access_pointer(sdata->vif.chanctx_conf) != conf && > + &sdata->reserved_chanctx->conf != conf) > continue; > > switch (vif->type) { > @@ -271,6 +272,7 @@ ieee80211_get_chanctx_max_required_bw(struct ieee80211_local *local, > case NL80211_IFTYPE_WDS: > case NL80211_IFTYPE_MESH_POINT: > width = vif->bss_conf.chandef.width; > + width = max(width, sdata->reserved_chandef.width); Not really sure why this is needed in this patch? > break; > case NL80211_IFTYPE_UNSPECIFIED: > case NUM_NL80211_IFTYPES: > @@ -899,6 +901,8 @@ int ieee80211_vif_reserve_chanctx(struct ieee80211_sub_if_data *sdata, > sdata->reserved_radar_required = radar_required; > sdata->reserved_ready = false; > > + ieee80211_recalc_chanctx_min_def(local, new_ctx); > + Hmm.. Wouldn't it make sense to recalc this in ieee80211_vif_unreserve_ chanctx() as well? Michał