Subject: Re: [mac80211] UAPSD: WLAN_STA_PS flag  cleared prematurely?
From: Johannes Berg <johannes@sipsolutions.net>
To: Marco Porsch <marco.porsch@etit.tu-chemnitz.de>
Cc: linux-wireless@vger.kernel.org
In-Reply-To: <4F634259.7040107@etit.tu-chemnitz.de>
References: <4F630757.6070901@etit.tu-chemnitz.de>
	 <1331901852.6753.2.camel@jlt3.sipsolutions.net>
	 <4F634259.7040107@etit.tu-chemnitz.de>
Content-Type: text/plain; charset="UTF-8"
Date: Fri, 16 Mar 2012 14:40:40 +0100
Message-ID: <1331905240.6753.14.camel@jlt3.sipsolutions.net> (sfid-20120316_144048_531360_661629AA)
Mime-Version: 1.0
Sender: linux-wireless-owner@vger.kernel.org

On Fri, 2012-03-16 at 14:38 +0100, Marco Porsch wrote:
> On 03/16/12 13:44, Johannes Berg wrote:
> > On Fri, 2012-03-16 at 10:26 +0100, Marco Porsch wrote:
> >> Hi all,
> >>
> >> in sta_info.c : ieee80211_sta_ps_deliver_response the
> >> IEEE80211_TX_STATUS_EOSP is set for all to-be-sent frames, not only for
> >> the last. But only the last buffered frame actually gets the EOSP flag.
> >>
> >>
> >> /* set EOSP for the frame */
> >> if (reason == IEEE80211_FRAME_RELEASE_UAPSD&&
> >>       qoshdr&&  skb_queue_empty(&frames))
> >> 	*qoshdr |= IEEE80211_QOS_CTL_EOSP;
> >>
> >> info->flags |= IEEE80211_TX_STATUS_EOSP |
> >>          IEEE80211_TX_CTL_REQ_TX_STATUS;
> >>
> >>
> >> Consequence is, that the WLAN_STA_SP flag gets cleared (multiple times)
> >> in ieee80211_tx_status before the last frame with EOSP has been sent.
> >> Is this correct?
> >
> > Looks like the bug is above, the EOSP/TX_STATUS should only be set for
> > the last frame?
> 
> I Agree. But what about the case, when the last frame is not a QoS 
> frame? Can this even happen - or is U-APSD only for QoS STA?

uAPSD can only be used by a QoS STA.

> Then we would have to manually append a QoS Null with the EOSP flag + 
> TX_STATUS?
> 
> So like this:
> 
> /* set EOSP for the _last_ frame or appended a QoS Null when needed */
> if (reason == IEEE80211_FRAME_RELEASE_UAPSD &&
>      skb_queue_empty(&frames)) {
> 	if (qoshdr) {
> 		*qoshdr |= IEEE80211_QOS_CTL_EOSP;
> 
> 		info->flags |= IEEE80211_TX_STATUS_EOSP |
> 		               IEEE80211_TX_CTL_REQ_TX_STATUS;
> 	} else {
> 		ieee80211_send_null_response(sdata, tid, reason);
> 	}
> }

No, I think it should be more like this:


/* set EOSP for the frame */

if (skb_queue_empty(&frames)) {
	if (reason == IEEE80211_FRAME_RELEASE_UAPSD && qoshdr)
	        *qoshdr |= IEEE80211_QOS_CTL_EOSP;

	info->flags |= IEEE80211_TX_STATUS_EOSP |
			IEEE80211_TX_CTL_REQ_TX_STATUS;
}

johannes