MIME-Version: 1.0
In-Reply-To: <1421408863.9214.7.camel@sipsolutions.net>
References: <1421404724-32326-1-git-send-email-janusz.dziedzic@tieto.com>
	<1421405364.9214.1.camel@sipsolutions.net>
	<CALhHN=payu27bPU+G5txsFkm6N=w_FL+LA8dVZAO96bwXXqiAg@mail.gmail.com>
	<1421408863.9214.7.camel@sipsolutions.net>
Date: Fri, 16 Jan 2015 13:18:57 +0100
Message-ID: <CALhHN=q=kpDnW1vj07k8Ar2bWYQGK_L-NcnxvnMttmDgN1FqFQ@mail.gmail.com> (sfid-20150116_131902_765944_E5F7C58D)
Subject: Re: [RFCv2 1/6] mac80211: ibss, fix chandef setup for HT40
From: Janusz Dziedzic <janusz.dziedzic@tieto.com>
To: Johannes Berg <johannes@sipsolutions.net>
Cc: linux-wireless@vger.kernel.org
Content-Type: text/plain; charset=UTF-8
Sender: linux-wireless-owner@vger.kernel.org

On 16 January 2015 at 12:47, Johannes Berg <johannes@sipsolutions.net> wrote:
> On Fri, 2015-01-16 at 12:24 +0100, Janusz Dziedzic wrote:
>> On 16 January 2015 at 11:49, Johannes Berg <johannes@sipsolutions.net> wrote:
>> > On Fri, 2015-01-16 at 11:38 +0100, Janusz Dziedzic wrote:
>> >> In some cases when used HT40+/HT40-, center_freq1
>> >> was configured incorrectly. Next this couse
>> >
>> > couse?
>> >
>> >> @@ -407,10 +407,15 @@ static void ieee80211_sta_join_ibss(struct ieee80211_sub_if_data *sdata,
>> >>       switch (sdata->u.ibss.chandef.width) {
>> >>       case NL80211_CHAN_WIDTH_20_NOHT:
>> >>       case NL80211_CHAN_WIDTH_20:
>> >> -     case NL80211_CHAN_WIDTH_40:
>> >>               chan_type = cfg80211_get_chandef_type(&sdata->u.ibss.chandef);
>> >>               cfg80211_chandef_create(&chandef, cbss->channel, chan_type);
>> >>               break;
>> >
>> > Sorry, I don't see the bug. "Sometimes"?
>> >
>> > cfg80211_get_chandef_type() will return HT40+/- correctly according to
>> > the chandef, and then you'll use it again to create the chandef - should be OK.
>> >
>> >> +       case NL80211_CHAN_WIDTH_40:
>> >> +               cfg80211_chandef_create(&chandef, cbss->channel,
>> >> +                                       NL80211_CHAN_WIDTH_20_NOHT);
>> >> +               chandef.width = sdata->u.ibss.chandef.width;
>> >> +               chandef.center_freq1 =
>> >> sdata->u.ibss.chandef.center_freq1;
>> >> +               break;
>> >
>> > Please don't mix chantype stuff and chandef stuff ... just create the
>> > chandef manually if you must. I really think the code above should have
>> > worked though, unless it's using the wrong inputs somehow.
>> >
>> This is scenario:
>>
>> 1) first ibss HT20 run on 5180
>> 2) second ibss configured HT40+ using 36 channel
>> 3) supplicant detect OBSS and decide to switch 40 and HT40-
>> 4) this code confgure chandef->center_freq1 = 5170
>> 5) reg_can_beacon() then fail and we fall back to NOHT_20
>
> Sounds more like we store and use invalid data?
>
I will check supplicant code then, and maybe skip obss scan when IBSS
already exist in scan results.

BR
Janusz