Merge commit 'origin/master' into next

Conflicts:
	include/linux/kvm.h

1 files changed, 216 insertions, 138 deletions

diff --git a/drivers/net/wireless/ath/ath9k/xmit.c b/drivers/net/wireless/ath/ath9k/xmit.c
index 42551a48c8ac..564c6cb1c2b4 100644
--- a/drivers/net/wireless/ath/ath9k/xmit.c
+++ b/drivers/net/wireless/ath/ath9k/xmit.c
@@ -70,6 +70,29 @@ static int ath_tx_num_badfrms(struct ath_softc *sc, struct ath_buf *bf,
 static void ath_tx_rc_status(struct ath_buf *bf, struct ath_desc *ds,
 			     int nbad, int txok, bool update_rc);
 
+enum {
+	MCS_DEFAULT,
+	MCS_HT40,
+	MCS_HT40_SGI,
+};
+
+static int ath_max_4ms_framelen[3][16] = {
+	[MCS_DEFAULT] = {
+		3216,  6434,  9650,  12868, 19304, 25740,  28956,  32180,
+		6430,  12860, 19300, 25736, 38600, 51472,  57890,  64320,
+	},
+	[MCS_HT40] = {
+		6684,  13368, 20052, 26738, 40104, 53476,  60156,  66840,
+		13360, 26720, 40080, 53440, 80160, 106880, 120240, 133600,
+	},
+	[MCS_HT40_SGI] = {
+		/* TODO: Only MCS 7 and 15 updated, recalculate the rest */
+		6684,  13368, 20052, 26738, 40104, 53476,  60156,  74200,
+		13360, 26720, 40080, 53440, 80160, 106880, 120240, 148400,
+	}
+};
+
+
 /*********************/
 /* Aggregation logic */
 /*********************/
@@ -107,7 +130,7 @@ static void ath_tx_resume_tid(struct ath_softc *sc, struct ath_atx_tid *tid)
 {
 	struct ath_txq *txq = &sc->tx.txq[tid->ac->qnum];
 
-	ASSERT(tid->paused > 0);
+	BUG_ON(tid->paused <= 0);
 	spin_lock_bh(&txq->axq_lock);
 
 	tid->paused--;
@@ -131,7 +154,7 @@ static void ath_tx_flush_tid(struct ath_softc *sc, struct ath_atx_tid *tid)
 	struct list_head bf_head;
 	INIT_LIST_HEAD(&bf_head);
 
-	ASSERT(tid->paused > 0);
+	BUG_ON(tid->paused <= 0);
 	spin_lock_bh(&txq->axq_lock);
 
 	tid->paused--;
@@ -143,7 +166,7 @@ static void ath_tx_flush_tid(struct ath_softc *sc, struct ath_atx_tid *tid)
 
 	while (!list_empty(&tid->buf_q)) {
 		bf = list_first_entry(&tid->buf_q, struct ath_buf, list);
-		ASSERT(!bf_isretried(bf));
+		BUG_ON(bf_isretried(bf));
 		list_move_tail(&bf->list, &bf_head);
 		ath_tx_send_ht_normal(sc, txq, tid, &bf_head);
 	}
@@ -178,7 +201,7 @@ static void ath_tx_addto_baw(struct ath_softc *sc, struct ath_atx_tid *tid,
 	index  = ATH_BA_INDEX(tid->seq_start, bf->bf_seqno);
 	cindex = (tid->baw_head + index) & (ATH_TID_MAX_BUFS - 1);
 
-	ASSERT(tid->tx_buf[cindex] == NULL);
+	BUG_ON(tid->tx_buf[cindex] != NULL);
 	tid->tx_buf[cindex] = bf;
 
 	if (index >= ((tid->baw_tail - tid->baw_head) &
@@ -251,6 +274,7 @@ static struct ath_buf* ath_clone_txbuf(struct ath_softc *sc, struct ath_buf *bf)
 
 	ATH_TXBUF_RESET(tbf);
 
+	tbf->aphy = bf->aphy;
 	tbf->bf_mpdu = bf->bf_mpdu;
 	tbf->bf_buf_addr = bf->bf_buf_addr;
 	*(tbf->bf_desc) = *(bf->bf_desc);
@@ -267,7 +291,9 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 	struct ath_node *an = NULL;
 	struct sk_buff *skb;
 	struct ieee80211_sta *sta;
+	struct ieee80211_hw *hw;
 	struct ieee80211_hdr *hdr;
+	struct ieee80211_tx_info *tx_info;
 	struct ath_atx_tid *tid = NULL;
 	struct ath_buf *bf_next, *bf_last = bf->bf_lastbf;
 	struct ath_desc *ds = bf_last->bf_desc;
@@ -280,9 +306,13 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 	skb = bf->bf_mpdu;
 	hdr = (struct ieee80211_hdr *)skb->data;
 
+	tx_info = IEEE80211_SKB_CB(skb);
+	hw = bf->aphy->hw;
+
 	rcu_read_lock();
 
-	sta = ieee80211_find_sta(sc->hw, hdr->addr1);
+	/* XXX: use ieee80211_find_sta! */
+	sta = ieee80211_find_sta_by_hw(hw, hdr->addr1);
 	if (!sta) {
 		rcu_read_unlock();
 		return;
@@ -358,7 +388,7 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 			else
 				INIT_LIST_HEAD(&bf_head);
 		} else {
-			ASSERT(!list_empty(bf_q));
+			BUG_ON(list_empty(bf_q));
 			list_move_tail(&bf->list, &bf_head);
 		}
 
@@ -452,11 +482,9 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
 static u32 ath_lookup_rate(struct ath_softc *sc, struct ath_buf *bf,
 			   struct ath_atx_tid *tid)
 {
-	const struct ath_rate_table *rate_table = sc->cur_rate_table;
 	struct sk_buff *skb;
 	struct ieee80211_tx_info *tx_info;
 	struct ieee80211_tx_rate *rates;
-	struct ath_tx_info_priv *tx_info_priv;
 	u32 max_4ms_framelen, frmlen;
 	u16 aggr_limit, legacy = 0;
 	int i;
@@ -464,7 +492,6 @@ static u32 ath_lookup_rate(struct ath_softc *sc, struct ath_buf *bf,
 	skb = bf->bf_mpdu;
 	tx_info = IEEE80211_SKB_CB(skb);
 	rates = tx_info->control.rates;
-	tx_info_priv = (struct ath_tx_info_priv *)tx_info->rate_driver_data[0];
 
 	/*
 	 * Find the lowest frame length among the rate series that will have a
@@ -475,12 +502,20 @@ static u32 ath_lookup_rate(struct ath_softc *sc, struct ath_buf *bf,
 
 	for (i = 0; i < 4; i++) {
 		if (rates[i].count) {
-			if (!WLAN_RC_PHY_HT(rate_table->info[rates[i].idx].phy)) {
+			int modeidx;
+			if (!(rates[i].flags & IEEE80211_TX_RC_MCS)) {
 				legacy = 1;
 				break;
 			}
 
-			frmlen = rate_table->info[rates[i].idx].max_4ms_framelen;
+			if (rates[i].flags & IEEE80211_TX_RC_SHORT_GI)
+				modeidx = MCS_HT40_SGI;
+			else if (rates[i].flags & IEEE80211_TX_RC_40_MHZ_WIDTH)
+				modeidx = MCS_HT40;
+			else
+				modeidx = MCS_DEFAULT;
+
+			frmlen = ath_max_4ms_framelen[modeidx][rates[i].idx];
 			max_4ms_framelen = min(max_4ms_framelen, frmlen);
 		}
 	}
@@ -518,12 +553,11 @@ static u32 ath_lookup_rate(struct ath_softc *sc, struct ath_buf *bf,
 static int ath_compute_num_delims(struct ath_softc *sc, struct ath_atx_tid *tid,
 				  struct ath_buf *bf, u16 frmlen)
 {
-	const struct ath_rate_table *rt = sc->cur_rate_table;
 	struct sk_buff *skb = bf->bf_mpdu;
 	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
 	u32 nsymbits, nsymbols;
 	u16 minlen;
-	u8 rc, flags, rix;
+	u8 flags, rix;
 	int width, half_gi, ndelim, mindelim;
 
 	/* Select standard number of delimiters based on frame length alone */
@@ -553,7 +587,6 @@ static int ath_compute_num_delims(struct ath_softc *sc, struct ath_atx_tid *tid,
 
 	rix = tx_info->control.rates[0].idx;
 	flags = tx_info->control.rates[0].flags;
-	rc = rt->info[rix].ratecode;
 	width = (flags & IEEE80211_TX_RC_40_MHZ_WIDTH) ? 1 : 0;
 	half_gi = (flags & IEEE80211_TX_RC_SHORT_GI) ? 1 : 0;
 
@@ -565,7 +598,7 @@ static int ath_compute_num_delims(struct ath_softc *sc, struct ath_atx_tid *tid,
 	if (nsymbols == 0)
 		nsymbols = 1;
 
-	nsymbits = bits_per_symbol[HT_RC_2_MCS(rc)][width];
+	nsymbits = bits_per_symbol[rix][width];
 	minlen = (nsymbols * nsymbits) / BITS_PER_BYTE;
 
 	if (frmlen < minlen) {
@@ -694,7 +727,6 @@ static void ath_tx_sched_aggr(struct ath_softc *sc, struct ath_txq *txq,
 		/* anchor last desc of aggregate */
 		ath9k_hw_set11n_aggr_last(sc->sc_ah, bf->bf_lastbf->bf_desc);
 
-		txq->axq_aggr_depth++;
 		ath_tx_txqaddbuf(sc, txq, &bf_q);
 		TX_STAT_INC(txq->axq_qnum, a_aggr);
 
@@ -815,6 +847,7 @@ static void ath_txq_drain_pending_buffers(struct ath_softc *sc,
 struct ath_txq *ath_txq_setup(struct ath_softc *sc, int qtype, int subtype)
 {
 	struct ath_hw *ah = sc->sc_ah;
+	struct ath_common *common = ath9k_hw_common(ah);
 	struct ath9k_tx_queue_info qi;
 	int qnum;
 
@@ -854,9 +887,9 @@ struct ath_txq *ath_txq_setup(struct ath_softc *sc, int qtype, int subtype)
 		return NULL;
 	}
 	if (qnum >= ARRAY_SIZE(sc->tx.txq)) {
-		DPRINTF(sc, ATH_DBG_FATAL,
-			"qnum %u out of range, max %u!\n",
-			qnum, (unsigned int)ARRAY_SIZE(sc->tx.txq));
+		ath_print(common, ATH_DBG_FATAL,
+			  "qnum %u out of range, max %u!\n",
+			  qnum, (unsigned int)ARRAY_SIZE(sc->tx.txq));
 		ath9k_hw_releasetxqueue(ah, qnum);
 		return NULL;
 	}
@@ -869,8 +902,6 @@ struct ath_txq *ath_txq_setup(struct ath_softc *sc, int qtype, int subtype)
 		INIT_LIST_HEAD(&txq->axq_acq);
 		spin_lock_init(&txq->axq_lock);
 		txq->axq_depth = 0;
-		txq->axq_aggr_depth = 0;
-		txq->axq_linkbuf = NULL;
 		txq->axq_tx_inprogress = false;
 		sc->tx.txqsetup |= 1<<qnum;
 	}
@@ -884,9 +915,9 @@ int ath_tx_get_qnum(struct ath_softc *sc, int qtype, int haltype)
 	switch (qtype) {
 	case ATH9K_TX_QUEUE_DATA:
 		if (haltype >= ARRAY_SIZE(sc->tx.hwq_map)) {
-			DPRINTF(sc, ATH_DBG_FATAL,
-				"HAL AC %u out of range, max %zu!\n",
-				haltype, ARRAY_SIZE(sc->tx.hwq_map));
+			ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_FATAL,
+				  "HAL AC %u out of range, max %zu!\n",
+				  haltype, ARRAY_SIZE(sc->tx.hwq_map));
 			return -1;
 		}
 		qnum = sc->tx.hwq_map[haltype];
@@ -906,18 +937,19 @@ int ath_tx_get_qnum(struct ath_softc *sc, int qtype, int haltype)
 struct ath_txq *ath_test_get_txq(struct ath_softc *sc, struct sk_buff *skb)
 {
 	struct ath_txq *txq = NULL;
+	u16 skb_queue = skb_get_queue_mapping(skb);
 	int qnum;
 
-	qnum = ath_get_hal_qnum(skb_get_queue_mapping(skb), sc);
+	qnum = ath_get_hal_qnum(skb_queue, sc);
 	txq = &sc->tx.txq[qnum];
 
 	spin_lock_bh(&txq->axq_lock);
 
 	if (txq->axq_depth >= (ATH_TXBUF - 20)) {
-		DPRINTF(sc, ATH_DBG_XMIT,
-			"TX queue: %d is full, depth: %d\n",
-			qnum, txq->axq_depth);
-		ieee80211_stop_queue(sc->hw, skb_get_queue_mapping(skb));
+		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_XMIT,
+			  "TX queue: %d is full, depth: %d\n",
+			  qnum, txq->axq_depth);
+		ath_mac80211_stop_queue(sc, skb_queue);
 		txq->stopped = 1;
 		spin_unlock_bh(&txq->axq_lock);
 		return NULL;
@@ -945,7 +977,7 @@ int ath_txq_update(struct ath_softc *sc, int qnum,
 		return 0;
 	}
 
-	ASSERT(sc->tx.txq[qnum].axq_qnum == qnum);
+	BUG_ON(sc->tx.txq[qnum].axq_qnum != qnum);
 
 	ath9k_hw_get_txq_props(ah, qnum, &qi);
 	qi.tqi_aifs = qinfo->tqi_aifs;
@@ -955,8 +987,8 @@ int ath_txq_update(struct ath_softc *sc, int qnum,
 	qi.tqi_readyTime = qinfo->tqi_readyTime;
 
 	if (!ath9k_hw_set_txq_props(ah, qnum, &qi)) {
-		DPRINTF(sc, ATH_DBG_FATAL,
-			"Unable to update hardware queue %u!\n", qnum);
+		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_FATAL,
+			  "Unable to update hardware queue %u!\n", qnum);
 		error = -EIO;
 	} else {
 		ath9k_hw_resettxqueue(ah, qnum);
@@ -1004,7 +1036,6 @@ void ath_draintxq(struct ath_softc *sc, struct ath_txq *txq, bool retry_tx)
 
 		if (list_empty(&txq->axq_q)) {
 			txq->axq_link = NULL;
-			txq->axq_linkbuf = NULL;
 			spin_unlock_bh(&txq->axq_lock);
 			break;
 		}
@@ -1055,6 +1086,7 @@ void ath_draintxq(struct ath_softc *sc, struct ath_txq *txq, bool retry_tx)
 void ath_drain_all_txq(struct ath_softc *sc, bool retry_tx)
 {
 	struct ath_hw *ah = sc->sc_ah;
+	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
 	struct ath_txq *txq;
 	int i, npend = 0;
 
@@ -1076,14 +1108,15 @@ void ath_drain_all_txq(struct ath_softc *sc, bool retry_tx)
 	if (npend) {
 		int r;
 
-		DPRINTF(sc, ATH_DBG_XMIT, "Unable to stop TxDMA. Reset HAL!\n");
+		ath_print(common, ATH_DBG_XMIT,
+			  "Unable to stop TxDMA. Reset HAL!\n");
 
 		spin_lock_bh(&sc->sc_resetlock);
 		r = ath9k_hw_reset(ah, sc->sc_ah->curchan, true);
 		if (r)
-			DPRINTF(sc, ATH_DBG_FATAL,
-				"Unable to reset hardware; reset status %d\n",
-				r);
+			ath_print(common, ATH_DBG_FATAL,
+				  "Unable to reset hardware; reset status %d\n",
+				  r);
 		spin_unlock_bh(&sc->sc_resetlock);
 	}
 
@@ -1147,8 +1180,8 @@ int ath_tx_setup(struct ath_softc *sc, int haltype)
 	struct ath_txq *txq;
 
 	if (haltype >= ARRAY_SIZE(sc->tx.hwq_map)) {
-		DPRINTF(sc, ATH_DBG_FATAL,
-			"HAL AC %u out of range, max %zu!\n",
+		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_FATAL,
+			  "HAL AC %u out of range, max %zu!\n",
 			 haltype, ARRAY_SIZE(sc->tx.hwq_map));
 		return 0;
 	}
@@ -1172,6 +1205,7 @@ static void ath_tx_txqaddbuf(struct ath_softc *sc, struct ath_txq *txq,
 			     struct list_head *head)
 {
 	struct ath_hw *ah = sc->sc_ah;
+	struct ath_common *common = ath9k_hw_common(ah);
 	struct ath_buf *bf;
 
 	/*
@@ -1186,21 +1220,20 @@ static void ath_tx_txqaddbuf(struct ath_softc *sc, struct ath_txq *txq,
 
 	list_splice_tail_init(head, &txq->axq_q);
 	txq->axq_depth++;
-	txq->axq_linkbuf = list_entry(txq->axq_q.prev, struct ath_buf, list);
 
-	DPRINTF(sc, ATH_DBG_QUEUE,
-		"qnum: %d, txq depth: %d\n", txq->axq_qnum, txq->axq_depth);
+	ath_print(common, ATH_DBG_QUEUE,
+		  "qnum: %d, txq depth: %d\n", txq->axq_qnum, txq->axq_depth);
 
 	if (txq->axq_link == NULL) {
 		ath9k_hw_puttxbuf(ah, txq->axq_qnum, bf->bf_daddr);
-		DPRINTF(sc, ATH_DBG_XMIT,
-			"TXDP[%u] = %llx (%p)\n",
-			txq->axq_qnum, ito64(bf->bf_daddr), bf->bf_desc);
+		ath_print(common, ATH_DBG_XMIT,
+			  "TXDP[%u] = %llx (%p)\n",
+			  txq->axq_qnum, ito64(bf->bf_daddr), bf->bf_desc);
 	} else {
 		*txq->axq_link = bf->bf_daddr;
-		DPRINTF(sc, ATH_DBG_XMIT, "link[%u] (%p)=%llx (%p)\n",
-			txq->axq_qnum, txq->axq_link,
-			ito64(bf->bf_daddr), bf->bf_desc);
+		ath_print(common, ATH_DBG_XMIT, "link[%u] (%p)=%llx (%p)\n",
+			  txq->axq_qnum, txq->axq_link,
+			  ito64(bf->bf_daddr), bf->bf_desc);
 	}
 	txq->axq_link = &(bf->bf_lastbf->bf_desc->ds_link);
 	ath9k_hw_txstart(ah, txq->axq_qnum);
@@ -1420,22 +1453,14 @@ static int setup_tx_flags(struct ath_softc *sc, struct sk_buff *skb,
 static u32 ath_pkt_duration(struct ath_softc *sc, u8 rix, struct ath_buf *bf,
 			    int width, int half_gi, bool shortPreamble)
 {
-	const struct ath_rate_table *rate_table = sc->cur_rate_table;
 	u32 nbits, nsymbits, duration, nsymbols;
-	u8 rc;
 	int streams, pktlen;
 
 	pktlen = bf_isaggr(bf) ? bf->bf_al : bf->bf_frmlen;
-	rc = rate_table->info[rix].ratecode;
-
-	/* for legacy rates, use old function to compute packet duration */
-	if (!IS_HT_RATE(rc))
-		return ath9k_hw_computetxtime(sc->sc_ah, rate_table, pktlen,
-					      rix, shortPreamble);
 
 	/* find number of symbols: PLCP + data */
 	nbits = (pktlen << 3) + OFDM_PLCP_BITS;
-	nsymbits = bits_per_symbol[HT_RC_2_MCS(rc)][width];
+	nsymbits = bits_per_symbol[rix][width];
 	nsymbols = (nbits + nsymbits - 1) / nsymbits;
 
 	if (!half_gi)
@@ -1444,7 +1469,7 @@ static u32 ath_pkt_duration(struct ath_softc *sc, u8 rix, struct ath_buf *bf,
 		duration = SYMBOL_TIME_HALFGI(nsymbols);
 
 	/* addup duration for legacy/ht training and signal fields */
-	streams = HT_RC_2_STREAMS(rc);
+	streams = HT_RC_2_STREAMS(rix);
 	duration += L_STF + L_LTF + L_SIG + HT_SIG + HT_STF + HT_LTF(streams);
 
 	return duration;
@@ -1452,11 +1477,12 @@ static u32 ath_pkt_duration(struct ath_softc *sc, u8 rix, struct ath_buf *bf,
 
 static void ath_buf_set_rate(struct ath_softc *sc, struct ath_buf *bf)
 {
-	const struct ath_rate_table *rt = sc->cur_rate_table;
+	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
 	struct ath9k_11n_rate_series series[4];
 	struct sk_buff *skb;
 	struct ieee80211_tx_info *tx_info;
 	struct ieee80211_tx_rate *rates;
+	const struct ieee80211_rate *rate;
 	struct ieee80211_hdr *hdr;
 	int i, flags = 0;
 	u8 rix = 0, ctsrate = 0;
@@ -1475,11 +1501,10 @@ static void ath_buf_set_rate(struct ath_softc *sc, struct ath_buf *bf)
 	 * checking the BSS's global flag.
 	 * But for the rate series, IEEE80211_TX_RC_USE_SHORT_PREAMBLE is used.
 	 */
+	rate = ieee80211_get_rts_cts_rate(sc->hw, tx_info);
+	ctsrate = rate->hw_value;
 	if (sc->sc_flags & SC_OP_PREAMBLE_SHORT)
-		ctsrate = rt->info[tx_info->control.rts_cts_rate_idx].ratecode |
-			rt->info[tx_info->control.rts_cts_rate_idx].short_preamble;
-	else
-		ctsrate = rt->info[tx_info->control.rts_cts_rate_idx].ratecode;
+		ctsrate |= rate->hw_value_short;
 
 	/*
 	 * ATH9K_TXDESC_RTSENA and ATH9K_TXDESC_CTSENA are mutually exclusive.
@@ -1502,18 +1527,15 @@ static void ath_buf_set_rate(struct ath_softc *sc, struct ath_buf *bf)
 		flags &= ~(ATH9K_TXDESC_RTSENA);
 
 	for (i = 0; i < 4; i++) {
+		bool is_40, is_sgi, is_sp;
+		int phy;
+
 		if (!rates[i].count || (rates[i].idx < 0))
 			continue;
 
 		rix = rates[i].idx;
 		series[i].Tries = rates[i].count;
-		series[i].ChSel = sc->tx_chainmask;
-
-		if (rates[i].flags & IEEE80211_TX_RC_USE_SHORT_PREAMBLE)
-			series[i].Rate = rt->info[rix].ratecode |
-				rt->info[rix].short_preamble;
-		else
-			series[i].Rate = rt->info[rix].ratecode;
+		series[i].ChSel = common->tx_chainmask;
 
 		if (rates[i].flags & IEEE80211_TX_RC_USE_RTS_CTS)
 			series[i].RateFlags |= ATH9K_RATESERIES_RTS_CTS;
@@ -1522,10 +1544,36 @@ static void ath_buf_set_rate(struct ath_softc *sc, struct ath_buf *bf)
 		if (rates[i].flags & IEEE80211_TX_RC_SHORT_GI)
 			series[i].RateFlags |= ATH9K_RATESERIES_HALFGI;
 
-		series[i].PktDuration = ath_pkt_duration(sc, rix, bf,
-			 (rates[i].flags & IEEE80211_TX_RC_40_MHZ_WIDTH) != 0,
-			 (rates[i].flags & IEEE80211_TX_RC_SHORT_GI),
-			 (rates[i].flags & IEEE80211_TX_RC_USE_SHORT_PREAMBLE));
+		is_sgi = !!(rates[i].flags & IEEE80211_TX_RC_SHORT_GI);
+		is_40 = !!(rates[i].flags & IEEE80211_TX_RC_40_MHZ_WIDTH);
+		is_sp = !!(rates[i].flags & IEEE80211_TX_RC_USE_SHORT_PREAMBLE);
+
+		if (rates[i].flags & IEEE80211_TX_RC_MCS) {
+			/* MCS rates */
+			series[i].Rate = rix | 0x80;
+			series[i].PktDuration = ath_pkt_duration(sc, rix, bf,
+				 is_40, is_sgi, is_sp);
+			continue;
+		}
+
+		/* legcay rates */
+		if ((tx_info->band == IEEE80211_BAND_2GHZ) &&
+		    !(rate->flags & IEEE80211_RATE_ERP_G))
+			phy = WLAN_RC_PHY_CCK;
+		else
+			phy = WLAN_RC_PHY_OFDM;
+
+		rate = &sc->sbands[tx_info->band].bitrates[rates[i].idx];
+		series[i].Rate = rate->hw_value;
+		if (rate->hw_value_short) {
+			if (rates[i].flags & IEEE80211_TX_RC_USE_SHORT_PREAMBLE)
+				series[i].Rate |= rate->hw_value_short;
+		} else {
+			is_sp = false;
+		}
+
+		series[i].PktDuration = ath9k_hw_computetxtime(sc->sc_ah,
+			phy, rate->bitrate * 100, bf->bf_frmlen, rix, is_sp);
 	}
 
 	/* set dur_update_en for l-sig computation except for PS-Poll frames */
@@ -1546,24 +1594,36 @@ static int ath_tx_setup_buffer(struct ieee80211_hw *hw, struct ath_buf *bf,
 	struct ath_softc *sc = aphy->sc;
 	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
 	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
-	struct ath_tx_info_priv *tx_info_priv;
 	int hdrlen;
 	__le16 fc;
+	int padpos, padsize;
 
-	tx_info_priv = kzalloc(sizeof(*tx_info_priv), GFP_ATOMIC);
-	if (unlikely(!tx_info_priv))
-		return -ENOMEM;
-	tx_info->rate_driver_data[0] = tx_info_priv;
-	tx_info_priv->aphy = aphy;
-	tx_info_priv->frame_type = txctl->frame_type;
+	tx_info->pad[0] = 0;
+	switch (txctl->frame_type) {
+	case ATH9K_NOT_INTERNAL:
+		break;
+	case ATH9K_INT_PAUSE:
+		tx_info->pad[0] |= ATH_TX_INFO_FRAME_TYPE_PAUSE;
+		/* fall through */
+	case ATH9K_INT_UNPAUSE:
+		tx_info->pad[0] |= ATH_TX_INFO_FRAME_TYPE_INTERNAL;
+		break;
+	}
 	hdrlen = ieee80211_get_hdrlen_from_skb(skb);
 	fc = hdr->frame_control;
 
 	ATH_TXBUF_RESET(bf);
 
-	bf->bf_frmlen = skb->len + FCS_LEN - (hdrlen & 3);
+	bf->aphy = aphy;
+	bf->bf_frmlen = skb->len + FCS_LEN;
+	/* Remove the padding size from bf_frmlen, if any */
+	padpos = ath9k_cmn_padpos(hdr->frame_control);
+	padsize = padpos & 3;
+	if (padsize && skb->len>padpos+padsize) {
+		bf->bf_frmlen -= padsize;
+	}
 
-	if (conf_is_ht(&sc->hw->conf) && !is_pae(skb))
+	if (conf_is_ht(&hw->conf) && !is_pae(skb))
 		bf->bf_state.bf_type |= BUF_HT;
 
 	bf->bf_flags = setup_tx_flags(sc, skb, txctl->txq);
@@ -1585,13 +1645,20 @@ static int ath_tx_setup_buffer(struct ieee80211_hw *hw, struct ath_buf *bf,
 					   skb->len, DMA_TO_DEVICE);
 	if (unlikely(dma_mapping_error(sc->dev, bf->bf_dmacontext))) {
 		bf->bf_mpdu = NULL;
-		kfree(tx_info_priv);
-		tx_info->rate_driver_data[0] = NULL;
-		DPRINTF(sc, ATH_DBG_FATAL, "dma_mapping_error() on TX\n");
+		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_FATAL,
+			  "dma_mapping_error() on TX\n");
 		return -ENOMEM;
 	}
 
 	bf->bf_buf_addr = bf->bf_dmacontext;
+
+	/* tag if this is a nullfunc frame to enable PS when AP acks it */
+	if (ieee80211_is_nullfunc(fc) && ieee80211_has_pm(fc)) {
+		bf->bf_isnullfunc = true;
+		sc->sc_flags &= ~SC_OP_NULLFUNC_COMPLETED;
+	} else
+		bf->bf_isnullfunc = false;
+
 	return 0;
 }
 
@@ -1669,12 +1736,13 @@ int ath_tx_start(struct ieee80211_hw *hw, struct sk_buff *skb,
 {
 	struct ath_wiphy *aphy = hw->priv;
 	struct ath_softc *sc = aphy->sc;
+	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
 	struct ath_buf *bf;
 	int r;
 
 	bf = ath_tx_get_buffer(sc);
 	if (!bf) {
-		DPRINTF(sc, ATH_DBG_XMIT, "TX buffers are full\n");
+		ath_print(common, ATH_DBG_XMIT, "TX buffers are full\n");
 		return -1;
 	}
 
@@ -1682,7 +1750,7 @@ int ath_tx_start(struct ieee80211_hw *hw, struct sk_buff *skb,
 	if (unlikely(r)) {
 		struct ath_txq *txq = txctl->txq;
 
-		DPRINTF(sc, ATH_DBG_FATAL, "TX mem alloc failure\n");
+		ath_print(common, ATH_DBG_FATAL, "TX mem alloc failure\n");
 
 		/* upon ath_tx_processq() this TX queue will be resumed, we
 		 * guarantee this will happen by knowing beforehand that
@@ -1690,8 +1758,7 @@ int ath_tx_start(struct ieee80211_hw *hw, struct sk_buff *skb,
 		 * on the queue */
 		spin_lock_bh(&txq->axq_lock);
 		if (sc->tx.txq[txq->axq_qnum].axq_depth > 1) {
-			ieee80211_stop_queue(sc->hw,
-				skb_get_queue_mapping(skb));
+			ath_mac80211_stop_queue(sc, skb_get_queue_mapping(skb));
 			txq->stopped = 1;
 		}
 		spin_unlock_bh(&txq->axq_lock);
@@ -1712,6 +1779,7 @@ void ath_tx_cabq(struct ieee80211_hw *hw, struct sk_buff *skb)
 {
 	struct ath_wiphy *aphy = hw->priv;
 	struct ath_softc *sc = aphy->sc;
+	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
 	int hdrlen, padsize;
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
 	struct ath_tx_control txctl;
@@ -1736,7 +1804,8 @@ void ath_tx_cabq(struct ieee80211_hw *hw, struct sk_buff *skb)
 	if (hdrlen & 3) {
 		padsize = hdrlen % 4;
 		if (skb_headroom(skb) < padsize) {
-			DPRINTF(sc, ATH_DBG_XMIT, "TX CABQ padding failed\n");
+			ath_print(common, ATH_DBG_XMIT,
+				  "TX CABQ padding failed\n");
 			dev_kfree_skb_any(skb);
 			return;
 		}
@@ -1746,10 +1815,11 @@ void ath_tx_cabq(struct ieee80211_hw *hw, struct sk_buff *skb)
 
 	txctl.txq = sc->beacon.cabq;
 
-	DPRINTF(sc, ATH_DBG_XMIT, "transmitting CABQ packet, skb: %p\n", skb);
+	ath_print(common, ATH_DBG_XMIT,
+		  "transmitting CABQ packet, skb: %p\n", skb);
 
 	if (ath_tx_start(hw, skb, &txctl) != 0) {
-		DPRINTF(sc, ATH_DBG_XMIT, "CABQ TX failed\n");
+		ath_print(common, ATH_DBG_XMIT, "CABQ TX failed\n");
 		goto exit;
 	}
 
@@ -1763,26 +1833,17 @@ exit:
 /*****************/
 
 static void ath_tx_complete(struct ath_softc *sc, struct sk_buff *skb,
-			    int tx_flags)
+			    struct ath_wiphy *aphy, int tx_flags)
 {
 	struct ieee80211_hw *hw = sc->hw;
 	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
-	struct ath_tx_info_priv *tx_info_priv = ATH_TX_INFO_PRIV(tx_info);
+	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
 	int hdrlen, padsize;
-	int frame_type = ATH9K_NOT_INTERNAL;
 
-	DPRINTF(sc, ATH_DBG_XMIT, "TX complete: skb: %p\n", skb);
+	ath_print(common, ATH_DBG_XMIT, "TX complete: skb: %p\n", skb);
 
-	if (tx_info_priv) {
-		hw = tx_info_priv->aphy->hw;
-		frame_type = tx_info_priv->frame_type;
-	}
-
-	if (tx_info->flags & IEEE80211_TX_CTL_NO_ACK ||
-	    tx_info->flags & IEEE80211_TX_STAT_TX_FILTERED) {
-		kfree(tx_info_priv);
-		tx_info->rate_driver_data[0] = NULL;
-	}
+	if (aphy)
+		hw = aphy->hw;
 
 	if (tx_flags & ATH_TX_BAR)
 		tx_info->flags |= IEEE80211_TX_STAT_AMPDU_NO_BACK;
@@ -1805,18 +1866,19 @@ static void ath_tx_complete(struct ath_softc *sc, struct sk_buff *skb,
 
 	if (sc->sc_flags & SC_OP_WAIT_FOR_TX_ACK) {
 		sc->sc_flags &= ~SC_OP_WAIT_FOR_TX_ACK;
-		DPRINTF(sc, ATH_DBG_PS, "Going back to sleep after having "
-			"received TX status (0x%x)\n",
+		ath_print(common, ATH_DBG_PS,
+			  "Going back to sleep after having "
+			  "received TX status (0x%x)\n",
 			sc->sc_flags & (SC_OP_WAIT_FOR_BEACON |
 					SC_OP_WAIT_FOR_CAB |
 					SC_OP_WAIT_FOR_PSPOLL_DATA |
 					SC_OP_WAIT_FOR_TX_ACK));
 	}
 
-	if (frame_type == ATH9K_NOT_INTERNAL)
-		ieee80211_tx_status(hw, skb);
-	else
+	if (unlikely(tx_info->pad[0] & ATH_TX_INFO_FRAME_TYPE_INTERNAL))
 		ath9k_tx_status(hw, skb);
+	else
+		ieee80211_tx_status(hw, skb);
 }
 
 static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
@@ -1839,7 +1901,7 @@ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
 	}
 
 	dma_unmap_single(sc->dev, bf->bf_dmacontext, skb->len, DMA_TO_DEVICE);
-	ath_tx_complete(sc, skb, tx_flags);
+	ath_tx_complete(sc, skb, bf->aphy, tx_flags);
 	ath_debug_stat_tx(sc, txq, bf);
 
 	/*
@@ -1887,8 +1949,7 @@ static void ath_tx_rc_status(struct ath_buf *bf, struct ath_desc *ds,
 	struct sk_buff *skb = bf->bf_mpdu;
 	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
 	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
-	struct ath_tx_info_priv *tx_info_priv = ATH_TX_INFO_PRIV(tx_info);
-	struct ieee80211_hw *hw = tx_info_priv->aphy->hw;
+	struct ieee80211_hw *hw = bf->aphy->hw;
 	u8 i, tx_rateindex;
 
 	if (txok)
@@ -1897,22 +1958,29 @@ static void ath_tx_rc_status(struct ath_buf *bf, struct ath_desc *ds,
 	tx_rateindex = ds->ds_txstat.ts_rateindex;
 	WARN_ON(tx_rateindex >= hw->max_rates);
 
-	tx_info_priv->update_rc = update_rc;
+	if (update_rc)
+		tx_info->pad[0] |= ATH_TX_INFO_UPDATE_RC;
 	if (ds->ds_txstat.ts_status & ATH9K_TXERR_FILT)
 		tx_info->flags |= IEEE80211_TX_STAT_TX_FILTERED;
 
 	if ((ds->ds_txstat.ts_status & ATH9K_TXERR_FILT) == 0 &&
 	    (bf->bf_flags & ATH9K_TXDESC_NOACK) == 0 && update_rc) {
 		if (ieee80211_is_data(hdr->frame_control)) {
-			memcpy(&tx_info_priv->tx, &ds->ds_txstat,
-			       sizeof(tx_info_priv->tx));
-			tx_info_priv->n_frames = bf->bf_nframes;
-			tx_info_priv->n_bad_frames = nbad;
+			if (ds->ds_txstat.ts_flags &
+			    (ATH9K_TX_DATA_UNDERRUN | ATH9K_TX_DELIM_UNDERRUN))
+				tx_info->pad[0] |= ATH_TX_INFO_UNDERRUN;
+			if ((ds->ds_txstat.ts_status & ATH9K_TXERR_XRETRY) ||
+			    (ds->ds_txstat.ts_status & ATH9K_TXERR_FIFO))
+				tx_info->pad[0] |= ATH_TX_INFO_XRETRY;
+			tx_info->status.ampdu_len = bf->bf_nframes;
+			tx_info->status.ampdu_ack_len = bf->bf_nframes - nbad;
 		}
 	}
 
-	for (i = tx_rateindex + 1; i < hw->max_rates; i++)
+	for (i = tx_rateindex + 1; i < hw->max_rates; i++) {
 		tx_info->status.rates[i].count = 0;
+		tx_info->status.rates[i].idx = -1;
+	}
 
 	tx_info->status.rates[tx_rateindex].count = bf->bf_retries + 1;
 }
@@ -1926,7 +1994,7 @@ static void ath_wake_mac80211_queue(struct ath_softc *sc, struct ath_txq *txq)
 	    sc->tx.txq[txq->axq_qnum].axq_depth <= (ATH_TXBUF - 20)) {
 		qnum = ath_get_mac80211_qnum(txq->axq_qnum, sc);
 		if (qnum != -1) {
-			ieee80211_wake_queue(sc->hw, qnum);
+			ath_mac80211_start_queue(sc, qnum);
 			txq->stopped = 0;
 		}
 	}
@@ -1936,21 +2004,21 @@ static void ath_wake_mac80211_queue(struct ath_softc *sc, struct ath_txq *txq)
 static void ath_tx_processq(struct ath_softc *sc, struct ath_txq *txq)
 {
 	struct ath_hw *ah = sc->sc_ah;
+	struct ath_common *common = ath9k_hw_common(ah);
 	struct ath_buf *bf, *lastbf, *bf_held = NULL;
 	struct list_head bf_head;
 	struct ath_desc *ds;
 	int txok;
 	int status;
 
-	DPRINTF(sc, ATH_DBG_QUEUE, "tx queue %d (%x), link %p\n",
-		txq->axq_qnum, ath9k_hw_gettxbuf(sc->sc_ah, txq->axq_qnum),
-		txq->axq_link);
+	ath_print(common, ATH_DBG_QUEUE, "tx queue %d (%x), link %p\n",
+		  txq->axq_qnum, ath9k_hw_gettxbuf(sc->sc_ah, txq->axq_qnum),
+		  txq->axq_link);
 
 	for (;;) {
 		spin_lock_bh(&txq->axq_lock);
 		if (list_empty(&txq->axq_q)) {
 			txq->axq_link = NULL;
-			txq->axq_linkbuf = NULL;
 			spin_unlock_bh(&txq->axq_lock);
 			break;
 		}
@@ -1984,10 +2052,19 @@ static void ath_tx_processq(struct ath_softc *sc, struct ath_txq *txq)
 			spin_unlock_bh(&txq->axq_lock);
 			break;
 		}
-		if (bf->bf_desc == txq->axq_lastdsWithCTS)
-			txq->axq_lastdsWithCTS = NULL;
-		if (ds == txq->axq_gatingds)
-			txq->axq_gatingds = NULL;
+
+		/*
+		 * We now know the nullfunc frame has been ACKed so we
+		 * can disable RX.
+		 */
+		if (bf->bf_isnullfunc &&
+		    (ds->ds_txstat.ts_status & ATH9K_TX_ACKED)) {
+			if ((sc->sc_flags & SC_OP_PS_ENABLED)) {
+				sc->ps_enabled = true;
+				ath9k_hw_setrxabort(sc->sc_ah, 1);
+			} else
+				sc->sc_flags |= SC_OP_NULLFUNC_COMPLETED;
+		}
 
 		/*
 		 * Remove ath_buf's of the same transmit unit from txq,
@@ -2001,9 +2078,6 @@ static void ath_tx_processq(struct ath_softc *sc, struct ath_txq *txq)
 				&txq->axq_q, lastbf->list.prev);
 
 		txq->axq_depth--;
-		if (bf_isaggr(bf))
-			txq->axq_aggr_depth--;
-
 		txok = (ds->ds_txstat.ts_status == 0);
 		txq->axq_tx_inprogress = false;
 		spin_unlock_bh(&txq->axq_lock);
@@ -2064,8 +2138,11 @@ static void ath_tx_complete_poll_work(struct work_struct *work)
 		}
 
 	if (needreset) {
-		DPRINTF(sc, ATH_DBG_RESET, "tx hung, resetting the chip\n");
+		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_RESET,
+			  "tx hung, resetting the chip\n");
+		ath9k_ps_wakeup(sc);
 		ath_reset(sc, false);
+		ath9k_ps_restore(sc);
 	}
 
 	ieee80211_queue_delayed_work(sc->hw, &sc->tx_complete_work,
@@ -2093,6 +2170,7 @@ void ath_tx_tasklet(struct ath_softc *sc)
 
 int ath_tx_init(struct ath_softc *sc, int nbufs)
 {
+	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
 	int error = 0;
 
 	spin_lock_init(&sc->tx.txbuflock);
@@ -2100,16 +2178,16 @@ int ath_tx_init(struct ath_softc *sc, int nbufs)
 	error = ath_descdma_setup(sc, &sc->tx.txdma, &sc->tx.txbuf,
 				  "tx", nbufs, 1);
 	if (error != 0) {
-		DPRINTF(sc, ATH_DBG_FATAL,
-			"Failed to allocate tx descriptors: %d\n", error);
+		ath_print(common, ATH_DBG_FATAL,
+			  "Failed to allocate tx descriptors: %d\n", error);
 		goto err;
 	}
 
 	error = ath_descdma_setup(sc, &sc->beacon.bdma, &sc->beacon.bbuf,
 				  "beacon", ATH_BCBUF, 1);
 	if (error != 0) {
-		DPRINTF(sc, ATH_DBG_FATAL,
-			"Failed to allocate beacon descriptors: %d\n", error);
+		ath_print(common, ATH_DBG_FATAL,
+			  "Failed to allocate beacon descriptors: %d\n", error);
 		goto err;
 	}