ath9k: fix powersave frame filtering/buffering in AP mode

This patch fixes a long standing issue of pending packets in the queue being
sent (and retransmitted many times) to sleeping stations.
This was made worse by aggregation through driver-internal retransmitting
of A-MDPU subframes.
Previously the hardware tx filter was cleared unconditionally for every
single packet - with this patch it uses the IEEE80211_TX_CTL_CLEAR_PS_FILT
for unaggregated frames.
A sta_notify driver op is added to stop aggregation for stations when they
enter powersave mode. Subframes stay buffered inside the driver, to ensure
that the BlockAck window keeps a sane state.
Since the driver uses software aggregation, the clearing of the tx filter
needs to be handled by the driver instead of mac80211 for aggregated frames.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

drivers/net/wireless/ath/ath9k/ar9002_mac.c

@@ -290,7 +290,6 @@ static void ar9002_hw_set11n_txdesc(struct ath_hw *ah, void *ds,
 		| (flags & ATH9K_TXDESC_VMF ? AR_VirtMoreFrag : 0)
 		| SM(txPower, AR_XmitPower)
 		| (flags & ATH9K_TXDESC_VEOL ? AR_VEOL : 0)
-		| (flags & ATH9K_TXDESC_CLRDMASK ? AR_ClrDestMask : 0)
 		| (flags & ATH9K_TXDESC_INTREQ ? AR_TxIntrReq : 0)
 		| (keyIx != ATH9K_TXKEYIX_INVALID ? AR_DestIdxValid : 0);
 
@@ -311,6 +310,16 @@ static void ar9002_hw_set11n_txdesc(struct ath_hw *ah, void *ds,
 	}
 }
 
+static void ar9002_hw_set_clrdmask(struct ath_hw *ah, void *ds, bool val)
+{
+	struct ar5416_desc *ads = AR5416DESC(ds);
+
+	if (val)
+		ads->ds_ctl0 |= AR_ClrDestMask;
+	else
+		ads->ds_ctl0 &= ~AR_ClrDestMask;
+}
+
 static void ar9002_hw_set11n_ratescenario(struct ath_hw *ah, void *ds,
 					  void *lastds,
 					  u32 durUpdateEn, u32 rtsctsRate,
@@ -448,4 +457,5 @@ void ar9002_hw_attach_mac_ops(struct ath_hw *ah)
 	ops->set11n_aggr_last = ar9002_hw_set11n_aggr_last;
 	ops->clr11n_aggr = ar9002_hw_clr11n_aggr;
 	ops->set11n_burstduration = ar9002_hw_set11n_burstduration;
+	ops->set_clrdmask = ar9002_hw_set_clrdmask;
 }

drivers/net/wireless/ath/ath9k/ar9003_mac.c

@@ -329,7 +329,6 @@ static void ar9003_hw_set11n_txdesc(struct ath_hw *ah, void *ds,
 		| (flags & ATH9K_TXDESC_VMF ? AR_VirtMoreFrag : 0)
 		| SM(txpower, AR_XmitPower)
 		| (flags & ATH9K_TXDESC_VEOL ? AR_VEOL : 0)
-		| (flags & ATH9K_TXDESC_CLRDMASK ? AR_ClrDestMask : 0)
 		| (keyIx != ATH9K_TXKEYIX_INVALID ? AR_DestIdxValid : 0)
 		| (flags & ATH9K_TXDESC_LOWRXCHAIN ? AR_LowRxChain : 0);
 
@@ -350,6 +349,16 @@ static void ar9003_hw_set11n_txdesc(struct ath_hw *ah, void *ds,
 	ads->ctl22 = 0;
 }
 
+static void ar9003_hw_set_clrdmask(struct ath_hw *ah, void *ds, bool val)
+{
+	struct ar9003_txc *ads = (struct ar9003_txc *) ds;
+
+	if (val)
+		ads->ctl11 |= AR_ClrDestMask;
+	else
+		ads->ctl11 &= ~AR_ClrDestMask;
+}
+
 static void ar9003_hw_set11n_ratescenario(struct ath_hw *ah, void *ds,
 					  void *lastds,
 					  u32 durUpdateEn, u32 rtsctsRate,
@@ -510,6 +519,7 @@ void ar9003_hw_attach_mac_ops(struct ath_hw *hw)
 	ops->set11n_aggr_last = ar9003_hw_set11n_aggr_last;
 	ops->clr11n_aggr = ar9003_hw_clr11n_aggr;
 	ops->set11n_burstduration = ar9003_hw_set11n_burstduration;
+	ops->set_clrdmask = ar9003_hw_set_clrdmask;
 }
 
 void ath9k_hw_set_rx_bufsize(struct ath_hw *ah, u16 buf_size)

drivers/net/wireless/ath/ath9k/ath9k.h

@@ -200,6 +200,7 @@ struct ath_atx_ac {
 	int sched;
 	struct list_head list;
 	struct list_head tid_q;
+	bool clear_ps_filter;
 };
 
 struct ath_frame_info {
@@ -257,6 +258,8 @@ struct ath_node {
 	struct ath_atx_ac ac[WME_NUM_AC];
 	u16 maxampdu;
 	u8 mpdudensity;
+
+	bool sleeping;
 };
 
 #define AGGR_CLEANUP         BIT(1)
@@ -338,6 +341,9 @@ int ath_tx_aggr_start(struct ath_softc *sc, struct ieee80211_sta *sta,
 void ath_tx_aggr_stop(struct ath_softc *sc, struct ieee80211_sta *sta, u16 tid);
 void ath_tx_aggr_resume(struct ath_softc *sc, struct ieee80211_sta *sta, u16 tid);
 
+void ath_tx_aggr_wakeup(struct ath_softc *sc, struct ath_node *an);
+bool ath_tx_aggr_sleep(struct ath_softc *sc, struct ath_node *an);
+
 /********/
 /* VIFs */
 /********/

drivers/net/wireless/ath/ath9k/hw-ops.h

@@ -122,6 +122,11 @@ static inline void ath9k_hw_set11n_burstduration(struct ath_hw *ah, void *ds,
 	ath9k_hw_ops(ah)->set11n_burstduration(ah, ds, burstDuration);
 }
 
+static inline void ath9k_hw_set_clrdmask(struct ath_hw *ah, void *ds, bool val)
+{
+	ath9k_hw_ops(ah)->set_clrdmask(ah, ds, val);
+}
+
 /* Private hardware call ops */
 
 /* PHY ops */

drivers/net/wireless/ath/ath9k/hw.h

@@ -626,6 +626,7 @@ struct ath_hw_ops {
 	void (*clr11n_aggr)(struct ath_hw *ah, void *ds);
 	void (*set11n_burstduration)(struct ath_hw *ah, void *ds,
 				     u32 burstDuration);
+	void (*set_clrdmask)(struct ath_hw *ah, void *ds, bool val);
 };
 
 struct ath_nf_limits {

drivers/net/wireless/ath/ath9k/mac.h

@@ -239,7 +239,6 @@ struct ath_desc {
 	void *ds_vdata;
 } __packed __aligned(4);
 
-#define ATH9K_TXDESC_CLRDMASK		0x0001
 #define ATH9K_TXDESC_NOACK		0x0002
 #define ATH9K_TXDESC_RTSENA		0x0004
 #define ATH9K_TXDESC_CTSENA		0x0008

drivers/net/wireless/ath/ath9k/main.c

@@ -1749,6 +1749,27 @@ static int ath9k_sta_remove(struct ieee80211_hw *hw,
 	return 0;
 }
 
+static void ath9k_sta_notify(struct ieee80211_hw *hw,
+			 struct ieee80211_vif *vif,
+			 enum sta_notify_cmd cmd,
+			 struct ieee80211_sta *sta)
+{
+	struct ath_softc *sc = hw->priv;
+	struct ath_node *an = (struct ath_node *) sta->drv_priv;
+
+	switch (cmd) {
+	case STA_NOTIFY_SLEEP:
+		an->sleeping = true;
+		if (ath_tx_aggr_sleep(sc, an))
+			ieee80211_sta_set_tim(sta);
+		break;
+	case STA_NOTIFY_AWAKE:
+		an->sleeping = false;
+		ath_tx_aggr_wakeup(sc, an);
+		break;
+	}
+}
+
 static int ath9k_conf_tx(struct ieee80211_hw *hw, u16 queue,
 			 const struct ieee80211_tx_queue_params *params)
 {
@@ -2230,6 +2251,7 @@ struct ieee80211_ops ath9k_ops = {
 	.configure_filter   = ath9k_configure_filter,
 	.sta_add	    = ath9k_sta_add,
 	.sta_remove	    = ath9k_sta_remove,
+	.sta_notify         = ath9k_sta_notify,
 	.conf_tx 	    = ath9k_conf_tx,
 	.bss_info_changed   = ath9k_bss_info_changed,
 	.set_key            = ath9k_set_key,

drivers/net/wireless/ath/ath9k/xmit.c

@@ -357,6 +357,7 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 	struct ath_frame_info *fi;
 	int nframes;
 	u8 tidno;
+	bool clear_filter;
 
 	skb = bf->bf_mpdu;
 	hdr = (struct ieee80211_hdr *)skb->data;
@@ -441,22 +442,24 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 			/* transmit completion */
 			acked_cnt++;
 		} else {
-			if (!(tid->state & AGGR_CLEANUP) && retry) {
-				if (fi->retries < ATH_MAX_SW_RETRIES) {
-					ath_tx_set_retry(sc, txq, bf->bf_mpdu);
-					txpending = 1;
-				} else {
-					bf->bf_state.bf_type |= BUF_XRETRY;
-					txfail = 1;
-					sendbar = 1;
-					txfail_cnt++;
-				}
-			} else {
+			if ((tid->state & AGGR_CLEANUP) || !retry) {
 				/*
 				 * cleanup in progress, just fail
 				 * the un-acked sub-frames
 				 */
 				txfail = 1;
+			} else if (fi->retries < ATH_MAX_SW_RETRIES) {
+				if (!(ts->ts_status & ATH9K_TXERR_FILT) ||
+				    !an->sleeping)
+					ath_tx_set_retry(sc, txq, bf->bf_mpdu);
+
+				clear_filter = true;
+				txpending = 1;
+			} else {
+				bf->bf_state.bf_type |= BUF_XRETRY;
+				txfail = 1;
+				sendbar = 1;
+				txfail_cnt++;
 			}
 		}
 
@@ -496,6 +499,7 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 				!txfail, sendbar);
 		} else {
 			/* retry the un-acked ones */
+			ath9k_hw_set_clrdmask(sc->sc_ah, bf->bf_desc, false);
 			if (!(sc->sc_ah->caps.hw_caps & ATH9K_HW_CAP_EDMA)) {
 				if (bf->bf_next == NULL && bf_last->bf_stale) {
 					struct ath_buf *tbf;
@@ -546,7 +550,12 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 
 	/* prepend un-acked frames to the beginning of the pending frame queue */
 	if (!list_empty(&bf_pending)) {
+		if (an->sleeping)
+			ieee80211_sta_set_tim(sta);
+
 		spin_lock_bh(&txq->axq_lock);
+		if (clear_filter)
+			tid->ac->clear_ps_filter = true;
 		list_splice(&bf_pending, &tid->buf_q);
 		ath_tx_queue_tid(txq, tid);
 		spin_unlock_bh(&txq->axq_lock);
@@ -816,6 +825,11 @@ static void ath_tx_sched_aggr(struct ath_softc *sc, struct ath_txq *txq,
 		bf = list_first_entry(&bf_q, struct ath_buf, list);
 		bf->bf_lastbf = list_entry(bf_q.prev, struct ath_buf, list);
 
+		if (tid->ac->clear_ps_filter) {
+			tid->ac->clear_ps_filter = false;
+			ath9k_hw_set_clrdmask(sc->sc_ah, bf->bf_desc, true);
+		}
+
 		/* if only one frame, send as non-aggregate */
 		if (bf == bf->bf_lastbf) {
 			fi = get_frame_info(bf->bf_mpdu);
@@ -896,6 +910,67 @@ void ath_tx_aggr_stop(struct ath_softc *sc, struct ieee80211_sta *sta, u16 tid)
 	ath_tx_flush_tid(sc, txtid);
 }
 
+bool ath_tx_aggr_sleep(struct ath_softc *sc, struct ath_node *an)
+{
+	struct ath_atx_tid *tid;
+	struct ath_atx_ac *ac;
+	struct ath_txq *txq;
+	bool buffered = false;
+	int tidno;
+
+	for (tidno = 0, tid = &an->tid[tidno];
+	     tidno < WME_NUM_TID; tidno++, tid++) {
+
+		if (!tid->sched)
+			continue;
+
+		ac = tid->ac;
+		txq = ac->txq;
+
+		spin_lock_bh(&txq->axq_lock);
+
+		if (!list_empty(&tid->buf_q))
+			buffered = true;
+
+		tid->sched = false;
+		list_del(&tid->list);
+
+		if (ac->sched) {
+			ac->sched = false;
+			list_del(&ac->list);
+		}
+
+		spin_unlock_bh(&txq->axq_lock);
+	}
+
+	return buffered;
+}
+
+void ath_tx_aggr_wakeup(struct ath_softc *sc, struct ath_node *an)
+{
+	struct ath_atx_tid *tid;
+	struct ath_atx_ac *ac;
+	struct ath_txq *txq;
+	int tidno;
+
+	for (tidno = 0, tid = &an->tid[tidno];
+	     tidno < WME_NUM_TID; tidno++, tid++) {
+
+		ac = tid->ac;
+		txq = ac->txq;
+
+		spin_lock_bh(&txq->axq_lock);
+		ac->clear_ps_filter = true;
+
+		if (!list_empty(&tid->buf_q) && !tid->paused) {
+			ath_tx_queue_tid(txq, tid);
+			ath_txq_schedule(sc, txq);
+		}
+
+		spin_unlock_bh(&txq->axq_lock);
+	}
+}
+
 void ath_tx_aggr_resume(struct ath_softc *sc, struct ieee80211_sta *sta, u16 tid)
 {
 	struct ath_atx_tid *txtid;
@@ -1491,7 +1566,6 @@ static int setup_tx_flags(struct sk_buff *skb)
 	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
 	int flags = 0;
 
-	flags |= ATH9K_TXDESC_CLRDMASK; /* needed for crypto errors */
 	flags |= ATH9K_TXDESC_INTREQ;
 
 	if (tx_info->flags & IEEE80211_TX_CTL_NO_ACK)
@@ -1754,6 +1828,9 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
 		if (txctl->paprd)
 			bf->bf_state.bfs_paprd_timestamp = jiffies;
 
+		if (tx_info->flags & IEEE80211_TX_CTL_CLEAR_PS_FILT)
+			ath9k_hw_set_clrdmask(sc->sc_ah, bf->bf_desc, true);
+
 		ath_tx_send_normal(sc, txctl->txq, tid, &bf_head);
 	}