Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next into for-davem

13 files changed, 295 insertions, 150 deletions

diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c
index 09d2e58..f9ae9b8 100644
--- a/net/mac80211/cfg.c
+++ b/net/mac80211/cfg.c
@@ -1035,6 +1035,7 @@ static int ieee80211_start_ap(struct wiphy *wiphy, struct net_device *dev,
 		return err;
 	}
 
+	ieee80211_recalc_dtim(local, sdata);
 	ieee80211_bss_info_change_notify(sdata, changed);
 
 	netif_carrier_on(dev);
@@ -3854,7 +3855,7 @@ static int ieee80211_set_qos_map(struct wiphy *wiphy,
 		new_qos_map = NULL;
 	}
 
-	old_qos_map = rtnl_dereference(sdata->qos_map);
+	old_qos_map = sdata_dereference(sdata->qos_map, sdata);
 	rcu_assign_pointer(sdata->qos_map, new_qos_map);
 	if (old_qos_map)
 		kfree_rcu(old_qos_map, rcu_head);
diff --git a/net/mac80211/debugfs_netdev.c b/net/mac80211/debugfs_netdev.c
index 04b5a14..ebf80f3 100644
--- a/net/mac80211/debugfs_netdev.c
+++ b/net/mac80211/debugfs_netdev.c
@@ -133,7 +133,15 @@ static ssize_t ieee80211_if_fmt_##name(					\
 			 jiffies_to_msecs(sdata->field));		\
 }
 
-#define __IEEE80211_IF_FILE(name, _write)				\
+#define _IEEE80211_IF_FILE_OPS(name, _read, _write)			\
+static const struct file_operations name##_ops = {			\
+	.read = (_read),						\
+	.write = (_write),						\
+	.open = simple_open,						\
+	.llseek = generic_file_llseek,					\
+}
+
+#define _IEEE80211_IF_FILE_R_FN(name)					\
 static ssize_t ieee80211_if_read_##name(struct file *file,		\
 					char __user *userbuf,		\
 					size_t count, loff_t *ppos)	\
@@ -141,28 +149,34 @@ static ssize_t ieee80211_if_read_##name(struct file *file,		\
 	return ieee80211_if_read(file->private_data,			\
 				 userbuf, count, ppos,			\
 				 ieee80211_if_fmt_##name);		\
-}									\
-static const struct file_operations name##_ops = {			\
-	.read = ieee80211_if_read_##name,				\
-	.write = (_write),						\
-	.open = simple_open,						\
-	.llseek = generic_file_llseek,					\
 }
 
-#define __IEEE80211_IF_FILE_W(name)					\
+#define _IEEE80211_IF_FILE_W_FN(name)					\
 static ssize_t ieee80211_if_write_##name(struct file *file,		\
 					 const char __user *userbuf,	\
 					 size_t count, loff_t *ppos)	\
 {									\
 	return ieee80211_if_write(file->private_data, userbuf, count,	\
 				  ppos, ieee80211_if_parse_##name);	\
-}									\
-__IEEE80211_IF_FILE(name, ieee80211_if_write_##name)
+}
+
+#define IEEE80211_IF_FILE_R(name)					\
+	_IEEE80211_IF_FILE_R_FN(name)					\
+	_IEEE80211_IF_FILE_OPS(name, ieee80211_if_read_##name, NULL)
+
+#define IEEE80211_IF_FILE_W(name)					\
+	_IEEE80211_IF_FILE_W_FN(name)					\
+	_IEEE80211_IF_FILE_OPS(name, NULL, ieee80211_if_write_##name)
 
+#define IEEE80211_IF_FILE_RW(name)					\
+	_IEEE80211_IF_FILE_R_FN(name)					\
+	_IEEE80211_IF_FILE_W_FN(name)					\
+	_IEEE80211_IF_FILE_OPS(name, ieee80211_if_read_##name,		\
+			       ieee80211_if_write_##name)
 
 #define IEEE80211_IF_FILE(name, field, format)				\
-		IEEE80211_IF_FMT_##format(name, field)			\
-		__IEEE80211_IF_FILE(name, NULL)
+	IEEE80211_IF_FMT_##format(name, field)				\
+	IEEE80211_IF_FILE_R(name)
 
 /* common attributes */
 IEEE80211_IF_FILE(drop_unencrypted, drop_unencrypted, DEC);
@@ -199,7 +213,7 @@ ieee80211_if_fmt_hw_queues(const struct ieee80211_sub_if_data *sdata,
 
 	return len;
 }
-__IEEE80211_IF_FILE(hw_queues, NULL);
+IEEE80211_IF_FILE_R(hw_queues);
 
 /* STA attributes */
 IEEE80211_IF_FILE(bssid, u.mgd.bssid, MAC);
@@ -275,14 +289,7 @@ static ssize_t ieee80211_if_parse_smps(struct ieee80211_sub_if_data *sdata,
 
 	return -EINVAL;
 }
-
-__IEEE80211_IF_FILE_W(smps);
-
-static ssize_t ieee80211_if_fmt_tkip_mic_test(
-	const struct ieee80211_sub_if_data *sdata, char *buf, int buflen)
-{
-	return -EOPNOTSUPP;
-}
+IEEE80211_IF_FILE_RW(smps);
 
 static ssize_t ieee80211_if_parse_tkip_mic_test(
 	struct ieee80211_sub_if_data *sdata, const char *buf, int buflen)
@@ -349,8 +356,7 @@ static ssize_t ieee80211_if_parse_tkip_mic_test(
 
 	return buflen;
 }
-
-__IEEE80211_IF_FILE_W(tkip_mic_test);
+IEEE80211_IF_FILE_W(tkip_mic_test);
 
 static ssize_t ieee80211_if_fmt_uapsd_queues(
 	const struct ieee80211_sub_if_data *sdata, char *buf, int buflen)
@@ -378,7 +384,7 @@ static ssize_t ieee80211_if_parse_uapsd_queues(
 
 	return buflen;
 }
-__IEEE80211_IF_FILE_W(uapsd_queues);
+IEEE80211_IF_FILE_RW(uapsd_queues);
 
 static ssize_t ieee80211_if_fmt_uapsd_max_sp_len(
 	const struct ieee80211_sub_if_data *sdata, char *buf, int buflen)
@@ -406,7 +412,7 @@ static ssize_t ieee80211_if_parse_uapsd_max_sp_len(
 
 	return buflen;
 }
-__IEEE80211_IF_FILE_W(uapsd_max_sp_len);
+IEEE80211_IF_FILE_RW(uapsd_max_sp_len);
 
 /* AP attributes */
 IEEE80211_IF_FILE(num_mcast_sta, u.ap.num_mcast_sta, ATOMIC);
@@ -419,7 +425,7 @@ static ssize_t ieee80211_if_fmt_num_buffered_multicast(
 	return scnprintf(buf, buflen, "%u\n",
 			 skb_queue_len(&sdata->u.ap.ps.bc_buf));
 }
-__IEEE80211_IF_FILE(num_buffered_multicast, NULL);
+IEEE80211_IF_FILE_R(num_buffered_multicast);
 
 /* IBSS attributes */
 static ssize_t ieee80211_if_fmt_tsf(
@@ -468,9 +474,10 @@ static ssize_t ieee80211_if_parse_tsf(
 		}
 	}
 
+	ieee80211_recalc_dtim(local, sdata);
 	return buflen;
 }
-__IEEE80211_IF_FILE_W(tsf);
+IEEE80211_IF_FILE_RW(tsf);
 
 
 /* WDS attributes */
diff --git a/net/mac80211/ht.c b/net/mac80211/ht.c
index 9a8be8f..fab7b91 100644
--- a/net/mac80211/ht.c
+++ b/net/mac80211/ht.c
@@ -479,10 +479,9 @@ void ieee80211_request_smps(struct ieee80211_vif *vif,
 			 vif->type != NL80211_IFTYPE_AP))
 		return;
 
-	if (WARN_ON(smps_mode == IEEE80211_SMPS_OFF))
-		smps_mode = IEEE80211_SMPS_AUTOMATIC;
-
 	if (vif->type == NL80211_IFTYPE_STATION) {
+		if (WARN_ON(smps_mode == IEEE80211_SMPS_OFF))
+			smps_mode = IEEE80211_SMPS_AUTOMATIC;
 		if (sdata->u.mgd.driver_smps_mode == smps_mode)
 			return;
 		sdata->u.mgd.driver_smps_mode = smps_mode;
diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h
index 953b9e2..3701930 100644
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -1800,6 +1800,8 @@ ieee80211_cs_get(struct ieee80211_local *local, u32 cipher,
 int ieee80211_cs_headroom(struct ieee80211_local *local,
 			  struct cfg80211_crypto_settings *crypto,
 			  enum nl80211_iftype iftype);
+void ieee80211_recalc_dtim(struct ieee80211_local *local,
+			   struct ieee80211_sub_if_data *sdata);
 
 #ifdef CONFIG_MAC80211_NOINLINE
 #define debug_noinline noinline
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index 2bd5b55..d767cfb 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -846,17 +846,6 @@ int ieee80211_register_hw(struct ieee80211_hw *hw)
 		/* TODO: consider VHT for RX chains, hopefully it's the same */
 	}
 
-	local->int_scan_req = kzalloc(sizeof(*local->int_scan_req) +
-				      sizeof(void *) * channels, GFP_KERNEL);
-	if (!local->int_scan_req)
-		return -ENOMEM;
-
-	for (band = 0; band < IEEE80211_NUM_BANDS; band++) {
-		if (!local->hw.wiphy->bands[band])
-			continue;
-		local->int_scan_req->rates[band] = (u32) -1;
-	}
-
 	/* if low-level driver supports AP, we also support VLAN */
 	if (local->hw.wiphy->interface_modes & BIT(NL80211_IFTYPE_AP)) {
 		hw->wiphy->interface_modes |= BIT(NL80211_IFTYPE_AP_VLAN);
@@ -880,6 +869,17 @@ int ieee80211_register_hw(struct ieee80211_hw *hw)
 				return -EINVAL;
 	}
 
+	local->int_scan_req = kzalloc(sizeof(*local->int_scan_req) +
+				      sizeof(void *) * channels, GFP_KERNEL);
+	if (!local->int_scan_req)
+		return -ENOMEM;
+
+	for (band = 0; band < IEEE80211_NUM_BANDS; band++) {
+		if (!local->hw.wiphy->bands[band])
+			continue;
+		local->int_scan_req->rates[band] = (u32) -1;
+	}
+
 #ifndef CONFIG_MAC80211_MESH
 	/* mesh depends on Kconfig, but drivers should set it if they want */
 	local->hw.wiphy->interface_modes &= ~BIT(NL80211_IFTYPE_MESH_POINT);
diff --git a/net/mac80211/mesh.c b/net/mac80211/mesh.c
index 5a74b24..5b919ca 100644
--- a/net/mac80211/mesh.c
+++ b/net/mac80211/mesh.c
@@ -807,6 +807,7 @@ int ieee80211_start_mesh(struct ieee80211_sub_if_data *sdata)
 		return -ENOMEM;
 	}
 
+	ieee80211_recalc_dtim(local, sdata);
 	ieee80211_bss_info_change_notify(sdata, changed);
 
 	netif_carrier_on(sdata->dev);
diff --git a/net/mac80211/mesh_plink.c b/net/mac80211/mesh_plink.c
index cf83217..e8f60aa 100644
--- a/net/mac80211/mesh_plink.c
+++ b/net/mac80211/mesh_plink.c
@@ -437,6 +437,7 @@ __mesh_sta_info_alloc(struct ieee80211_sub_if_data *sdata, u8 *hw_addr)
 	sta_info_pre_move_state(sta, IEEE80211_STA_AUTHORIZED);
 
 	set_sta_flag(sta, WLAN_STA_WME);
+	sta->sta.wme = true;
 
 	return sta;
 }
diff --git a/net/mac80211/rx.c b/net/mac80211/rx.c
index 5a2afe9..c24ca0d 100644
--- a/net/mac80211/rx.c
+++ b/net/mac80211/rx.c
@@ -3076,8 +3076,8 @@ void ieee80211_release_reorder_timeout(struct sta_info *sta, int tid)
 
 /* main receive path */
 
-static int prepare_for_handlers(struct ieee80211_rx_data *rx,
-				struct ieee80211_hdr *hdr)
+static bool prepare_for_handlers(struct ieee80211_rx_data *rx,
+				 struct ieee80211_hdr *hdr)
 {
 	struct ieee80211_sub_if_data *sdata = rx->sdata;
 	struct sk_buff *skb = rx->skb;
@@ -3088,29 +3088,29 @@ static int prepare_for_handlers(struct ieee80211_rx_data *rx,
 	switch (sdata->vif.type) {
 	case NL80211_IFTYPE_STATION:
 		if (!bssid && !sdata->u.mgd.use_4addr)
-			return 0;
+			return false;
 		if (!multicast &&
 		    !ether_addr_equal(sdata->vif.addr, hdr->addr1)) {
 			if (!(sdata->dev->flags & IFF_PROMISC) ||
 			    sdata->u.mgd.use_4addr)
-				return 0;
+				return false;
 			status->rx_flags &= ~IEEE80211_RX_RA_MATCH;
 		}
 		break;
 	case NL80211_IFTYPE_ADHOC:
 		if (!bssid)
-			return 0;
+			return false;
 		if (ether_addr_equal(sdata->vif.addr, hdr->addr2) ||
 		    ether_addr_equal(sdata->u.ibss.bssid, hdr->addr2))
-			return 0;
+			return false;
 		if (ieee80211_is_beacon(hdr->frame_control)) {
-			return 1;
+			return true;
 		} else if (!ieee80211_bssid_match(bssid, sdata->u.ibss.bssid)) {
-			return 0;
+			return false;
 		} else if (!multicast &&
 			   !ether_addr_equal(sdata->vif.addr, hdr->addr1)) {
 			if (!(sdata->dev->flags & IFF_PROMISC))
-				return 0;
+				return false;
 			status->rx_flags &= ~IEEE80211_RX_RA_MATCH;
 		} else if (!rx->sta) {
 			int rate_idx;
@@ -3126,7 +3126,7 @@ static int prepare_for_handlers(struct ieee80211_rx_data *rx,
 		if (!multicast &&
 		    !ether_addr_equal(sdata->vif.addr, hdr->addr1)) {
 			if (!(sdata->dev->flags & IFF_PROMISC))
-				return 0;
+				return false;
 
 			status->rx_flags &= ~IEEE80211_RX_RA_MATCH;
 		}
@@ -3135,7 +3135,7 @@ static int prepare_for_handlers(struct ieee80211_rx_data *rx,
 	case NL80211_IFTYPE_AP:
 		if (!bssid) {
 			if (!ether_addr_equal(sdata->vif.addr, hdr->addr1))
-				return 0;
+				return false;
 		} else if (!ieee80211_bssid_match(bssid, sdata->vif.addr)) {
 			/*
 			 * Accept public action frames even when the
@@ -3145,26 +3145,26 @@ static int prepare_for_handlers(struct ieee80211_rx_data *rx,
 			 */
 			if (!multicast &&
 			    !ether_addr_equal(sdata->vif.addr, hdr->addr1))
-				return 0;
+				return false;
 			if (ieee80211_is_public_action(hdr, skb->len))
-				return 1;
+				return true;
 			if (!ieee80211_is_beacon(hdr->frame_control))
-				return 0;
+				return false;
 			status->rx_flags &= ~IEEE80211_RX_RA_MATCH;
 		}
 		break;
 	case NL80211_IFTYPE_WDS:
 		if (bssid || !ieee80211_is_data(hdr->frame_control))
-			return 0;
+			return false;
 		if (!ether_addr_equal(sdata->u.wds.remote_addr, hdr->addr2))
-			return 0;
+			return false;
 		break;
 	case NL80211_IFTYPE_P2P_DEVICE:
 		if (!ieee80211_is_public_action(hdr, skb->len) &&
 		    !ieee80211_is_probe_req(hdr->frame_control) &&
 		    !ieee80211_is_probe_resp(hdr->frame_control) &&
 		    !ieee80211_is_beacon(hdr->frame_control))
-			return 0;
+			return false;
 		if (!ether_addr_equal(sdata->vif.addr, hdr->addr1) &&
 		    !multicast)
 			status->rx_flags &= ~IEEE80211_RX_RA_MATCH;
@@ -3175,7 +3175,7 @@ static int prepare_for_handlers(struct ieee80211_rx_data *rx,
 		break;
 	}
 
-	return 1;
+	return true;
 }
 
 /*
@@ -3191,13 +3191,11 @@ static bool ieee80211_prepare_and_rx_handle(struct ieee80211_rx_data *rx,
 	struct ieee80211_sub_if_data *sdata = rx->sdata;
 	struct ieee80211_rx_status *status = IEEE80211_SKB_RXCB(skb);
 	struct ieee80211_hdr *hdr = (void *)skb->data;
-	int prepares;
 
 	rx->skb = skb;
 	status->rx_flags |= IEEE80211_RX_RA_MATCH;
-	prepares = prepare_for_handlers(rx, hdr);
 
-	if (!prepares)
+	if (!prepare_for_handlers(rx, hdr))
 		return false;
 
 	if (!consume) {
diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
index 4576ba0..decd30c 100644
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -300,6 +300,35 @@ struct sta_info *sta_info_alloc(struct ieee80211_sub_if_data *sdata,
 	if (!sta)
 		return NULL;
 
+	rcu_read_lock();
+	tx_latency = rcu_dereference(local->tx_latency);
+	/* init stations Tx latency statistics && TID bins */
+	if (tx_latency) {
+		sta->tx_lat = kzalloc(IEEE80211_NUM_TIDS *
+				      sizeof(struct ieee80211_tx_latency_stat),
+				      GFP_ATOMIC);
+		if (!sta->tx_lat) {
+			rcu_read_unlock();
+			goto free;
+		}
+
+		if (tx_latency->n_ranges) {
+			for (i = 0; i < IEEE80211_NUM_TIDS; i++) {
+				/* size of bins is size of the ranges +1 */
+				sta->tx_lat[i].bin_count =
+					tx_latency->n_ranges + 1;
+				sta->tx_lat[i].bins =
+					kcalloc(sta->tx_lat[i].bin_count,
+						sizeof(u32), GFP_ATOMIC);
+				if (!sta->tx_lat[i].bins) {
+					rcu_read_unlock();
+					goto free;
+				}
+			}
+		}
+	}
+	rcu_read_unlock();
+
 	spin_lock_init(&sta->lock);
 	INIT_WORK(&sta->drv_unblock_wk, sta_unblock);
 	INIT_WORK(&sta->ampdu_mlme.work, ieee80211_ba_session_work);
@@ -324,10 +353,8 @@ struct sta_info *sta_info_alloc(struct ieee80211_sub_if_data *sdata,
 	for (i = 0; i < ARRAY_SIZE(sta->chain_signal_avg); i++)
 		ewma_init(&sta->chain_signal_avg[i], 1024, 8);
 
-	if (sta_prepare_rate_control(local, sta, gfp)) {
-		kfree(sta);
-		return NULL;
-	}
+	if (sta_prepare_rate_control(local, sta, gfp))
+		goto free;
 
 	for (i = 0; i < IEEE80211_NUM_TIDS; i++) {
 		/*
@@ -371,34 +398,17 @@ struct sta_info *sta_info_alloc(struct ieee80211_sub_if_data *sdata,
 		}
 	}
 
-	rcu_read_lock();
-
-	tx_latency = rcu_dereference(local->tx_latency);
-	/* init stations Tx latency statistics && TID bins */
-	if (tx_latency)
-		sta->tx_lat = kzalloc(IEEE80211_NUM_TIDS *
-				      sizeof(struct ieee80211_tx_latency_stat),
-				      GFP_ATOMIC);
-
-	/*
-	 * if Tx latency and bins are enabled and the previous allocation
-	 * succeeded
-	 */
-	if (tx_latency && tx_latency->n_ranges && sta->tx_lat)
-		for (i = 0; i < IEEE80211_NUM_TIDS; i++) {
-			/* size of bins is size of the ranges +1 */
-			sta->tx_lat[i].bin_count =
-				tx_latency->n_ranges + 1;
-			sta->tx_lat[i].bins  = kcalloc(sta->tx_lat[i].bin_count,
-						       sizeof(u32),
-						       GFP_ATOMIC);
-		}
-
-	rcu_read_unlock();
-
 	sta_dbg(sdata, "Allocated STA %pM\n", sta->sta.addr);
-
 	return sta;
+
+free:
+	if (sta->tx_lat) {
+		for (i = 0; i < IEEE80211_NUM_TIDS; i++)
+			kfree(sta->tx_lat[i].bins);
+		kfree(sta->tx_lat);
+	}
+	kfree(sta);
+	return NULL;
 }
 
 static int sta_info_insert_check(struct sta_info *sta)
@@ -1143,7 +1153,8 @@ void ieee80211_sta_ps_deliver_wakeup(struct sta_info *sta)
 
 static void ieee80211_send_null_response(struct ieee80211_sub_if_data *sdata,
 					 struct sta_info *sta, int tid,
-					 enum ieee80211_frame_release_type reason)
+					 enum ieee80211_frame_release_type reason,
+					 bool call_driver)
 {
 	struct ieee80211_local *local = sdata->local;
 	struct ieee80211_qos_hdr *nullfunc;
@@ -1201,7 +1212,9 @@ static void ieee80211_send_null_response(struct ieee80211_sub_if_data *sdata,
 		       IEEE80211_TX_STATUS_EOSP |
 		       IEEE80211_TX_CTL_REQ_TX_STATUS;
 
-	drv_allow_buffered_frames(local, sta, BIT(tid), 1, reason, false);
+	if (call_driver)
+		drv_allow_buffered_frames(local, sta, BIT(tid), 1,
+					  reason, false);
 
 	skb->dev = sdata->dev;
 
@@ -1217,6 +1230,17 @@ static void ieee80211_send_null_response(struct ieee80211_sub_if_data *sdata,
 	rcu_read_unlock();
 }
 
+static int find_highest_prio_tid(unsigned long tids)
+{
+	/* lower 3 TIDs aren't ordered perfectly */
+	if (tids & 0xF8)
+		return fls(tids) - 1;
+	/* TID 0 is BE just like TID 3 */
+	if (tids & BIT(0))
+		return 0;
+	return fls(tids) - 1;
+}
+
 static void
 ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 				  int n_frames, u8 ignored_acs,
@@ -1224,7 +1248,6 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 {
 	struct ieee80211_sub_if_data *sdata = sta->sdata;
 	struct ieee80211_local *local = sdata->local;
-	bool found = false;
 	bool more_data = false;
 	int ac;
 	unsigned long driver_release_tids = 0;
@@ -1235,9 +1258,7 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 
 	__skb_queue_head_init(&frames);
 
-	/*
-	 * Get response frame(s) and more data bit for it.
-	 */
+	/* Get response frame(s) and more data bit for the last one. */
 	for (ac = 0; ac < IEEE80211_NUM_ACS; ac++) {
 		unsigned long tids;
 
@@ -1246,43 +1267,48 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 
 		tids = ieee80211_tids_for_ac(ac);
 
-		if (!found) {
-			driver_release_tids = sta->driver_buffered_tids & tids;
-			if (driver_release_tids) {
-				found = true;
-			} else {
-				struct sk_buff *skb;
-
-				while (n_frames > 0) {
-					skb = skb_dequeue(&sta->tx_filtered[ac]);
-					if (!skb) {
-						skb = skb_dequeue(
-							&sta->ps_tx_buf[ac]);
-						if (skb)
-							local->total_ps_buffered--;
-					}
-					if (!skb)
-						break;
-					n_frames--;
-					found = true;
-					__skb_queue_tail(&frames, skb);
-				}
-			}
+		/* if we already have frames from software, then we can't also
+		 * release from hardware queues
+		 */
+		if (skb_queue_empty(&frames))
+			driver_release_tids |= sta->driver_buffered_tids & tids;
 
-			/*
-			 * If the driver has data on more than one TID then
+		if (driver_release_tids) {
+			/* If the driver has data on more than one TID then
 			 * certainly there's more data if we release just a
-			 * single frame now (from a single TID).
+			 * single frame now (from a single TID). This will
+			 * only happen for PS-Poll.
 			 */
 			if (reason == IEEE80211_FRAME_RELEASE_PSPOLL &&
 			    hweight16(driver_release_tids) > 1) {
 				more_data = true;
 				driver_release_tids =
-					BIT(ffs(driver_release_tids) - 1);
+					BIT(find_highest_prio_tid(
+						driver_release_tids));
 				break;
 			}
+		} else {
+			struct sk_buff *skb;
+
+			while (n_frames > 0) {
+				skb = skb_dequeue(&sta->tx_filtered[ac]);
+				if (!skb) {
+					skb = skb_dequeue(
+						&sta->ps_tx_buf[ac]);
+					if (skb)
+						local->total_ps_buffered--;
+				}
+				if (!skb)
+					break;
+				n_frames--;
+				__skb_queue_tail(&frames, skb);
+			}
 		}
 
+		/* If we have more frames buffered on this AC, then set the
+		 * more-data bit and abort the loop since we can't send more
+		 * data from other ACs before the buffered frames from this.
+		 */
 		if (!skb_queue_empty(&sta->tx_filtered[ac]) ||
 		    !skb_queue_empty(&sta->ps_tx_buf[ac])) {
 			more_data = true;
@@ -1290,7 +1316,7 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 		}
 	}
 
-	if (!found) {
+	if (skb_queue_empty(&frames) && !driver_release_tids) {
 		int tid;
 
 		/*
@@ -1311,15 +1337,13 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 		/* This will evaluate to 1, 3, 5 or 7. */
 		tid = 7 - ((ffs(~ignored_acs) - 1) << 1);
 
-		ieee80211_send_null_response(sdata, sta, tid, reason);
-		return;
-	}
-
-	if (!driver_release_tids) {
+		ieee80211_send_null_response(sdata, sta, tid, reason, true);
+	} else if (!driver_release_tids) {
 		struct sk_buff_head pending;
 		struct sk_buff *skb;
 		int num = 0;
 		u16 tids = 0;
+		bool need_null = false;
 
 		skb_queue_head_init(&pending);
 
@@ -1353,22 +1377,57 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 			    ieee80211_is_qos_nullfunc(hdr->frame_control))
 				qoshdr = ieee80211_get_qos_ctl(hdr);
 
-			/* end service period after last frame */
-			if (skb_queue_empty(&frames)) {
-				if (reason == IEEE80211_FRAME_RELEASE_UAPSD &&
-				    qoshdr)
-					*qoshdr |= IEEE80211_QOS_CTL_EOSP;
+			tids |= BIT(skb->priority);
+
+			__skb_queue_tail(&pending, skb);
+
+			/* end service period after last frame or add one */
+			if (!skb_queue_empty(&frames))
+				continue;
 
+			if (reason != IEEE80211_FRAME_RELEASE_UAPSD) {
+				/* for PS-Poll, there's only one frame */
 				info->flags |= IEEE80211_TX_STATUS_EOSP |
 					       IEEE80211_TX_CTL_REQ_TX_STATUS;
+				break;
 			}
 
-			if (qoshdr)
-				tids |= BIT(*qoshdr & IEEE80211_QOS_CTL_TID_MASK);
-			else
-				tids |= BIT(0);
+			/* For uAPSD, things are a bit more complicated. If the
+			 * last frame has a QoS header (i.e. is a QoS-data or
+			 * QoS-nulldata frame) then just set the EOSP bit there
+			 * and be done.
+			 * If the frame doesn't have a QoS header (which means
+			 * it should be a bufferable MMPDU) then we can't set
+			 * the EOSP bit in the QoS header; add a QoS-nulldata
+			 * frame to the list to send it after the MMPDU.
+			 *
+			 * Note that this code is only in the mac80211-release
+			 * code path, we assume that the driver will not buffer
+			 * anything but QoS-data frames, or if it does, will
+			 * create the QoS-nulldata frame by itself if needed.
+			 *
+			 * Cf. 802.11-2012 10.2.1.10 (c).
+			 */
+			if (qoshdr) {
+				*qoshdr |= IEEE80211_QOS_CTL_EOSP;
 
-			__skb_queue_tail(&pending, skb);
+				info->flags |= IEEE80211_TX_STATUS_EOSP |
+					       IEEE80211_TX_CTL_REQ_TX_STATUS;
+			} else {
+				/* The standard isn't completely clear on this
+				 * as it says the more-data bit should be set
+				 * if there are more BUs. The QoS-Null frame
+				 * we're about to send isn't buffered yet, we
+				 * only create it below, but let's pretend it
+				 * was buffered just in case some clients only
+				 * expect more-data=0 when eosp=1.
+				 */
+				hdr->frame_control |=
+					cpu_to_le16(IEEE80211_FCTL_MOREDATA);
+				need_null = true;
+				num++;
+			}
+			break;
 		}
 
 		drv_allow_buffered_frames(local, sta, tids, num,
@@ -1376,17 +1435,22 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 
 		ieee80211_add_pending_skbs(local, &pending);
 
+		if (need_null)
+			ieee80211_send_null_response(
+				sdata, sta, find_highest_prio_tid(tids),
+				reason, false);
+
 		sta_info_recalc_tim(sta);
 	} else {
 		/*
 		 * We need to release a frame that is buffered somewhere in the
 		 * driver ... it'll have to handle that.
-		 * Note that, as per the comment above, it'll also have to see
-		 * if there is more than just one frame on the specific TID that
-		 * we're releasing from, and it needs to set the more-data bit
-		 * accordingly if we tell it that there's no more data. If we do
-		 * tell it there's more data, then of course the more-data bit
-		 * needs to be set anyway.
+		 * Note that the driver also has to check the number of frames
+		 * on the TIDs we're releasing from - if there are more than
+		 * n_frames it has to set the more-data bit (if we didn't ask
+		 * it to set it anyway due to other buffered frames); if there
+		 * are fewer than n_frames it has to make sure to adjust that
+		 * to allow the service period to end properly.
 		 */
 		drv_release_buffered_frames(local, sta, driver_release_tids,
 					    n_frames, reason, more_data);
@@ -1394,9 +1458,9 @@ ieee80211_sta_ps_deliver_response(struct sta_info *sta,
 		/*
 		 * Note that we don't recalculate the TIM bit here as it would
 		 * most likely have no effect at all unless the driver told us
-		 * that the TID became empty before returning here from the
+		 * that the TID(s) became empty before returning here from the
 		 * release function.
-		 * Either way, however, when the driver tells us that the TID
+		 * Either way, however, when the driver tells us that the TID(s)
 		 * became empty we'll do the TIM recalculation.
 		 */
 	}
@@ -1485,6 +1549,8 @@ void ieee80211_sta_set_buffered(struct ieee80211_sta *pubsta,
 	if (WARN_ON(tid >= IEEE80211_NUM_TIDS))
 		return;
 
+	trace_api_sta_set_buffered(sta->local, pubsta, tid, buffered);
+
 	if (buffered)
 		set_bit(tid, &sta->driver_buffered_tids);
 	else
diff --git a/net/mac80211/trace.h b/net/mac80211/trace.h
index da93666..a0b0aea 100644
--- a/net/mac80211/trace.h
+++ b/net/mac80211/trace.h
@@ -1835,6 +1835,33 @@ TRACE_EVENT(api_eosp,
 	)
 );
 
+TRACE_EVENT(api_sta_set_buffered,
+	TP_PROTO(struct ieee80211_local *local,
+		 struct ieee80211_sta *sta,
+		 u8 tid, bool buffered),
+
+	TP_ARGS(local, sta, tid, buffered),
+
+	TP_STRUCT__entry(
+		LOCAL_ENTRY
+		STA_ENTRY
+		__field(u8, tid)
+		__field(bool, buffered)
+	),
+
+	TP_fast_assign(
+		LOCAL_ASSIGN;
+		STA_ASSIGN;
+		__entry->tid = tid;
+		__entry->buffered = buffered;
+	),
+
+	TP_printk(
+		LOCAL_PR_FMT STA_PR_FMT " tid:%d buffered:%d",
+		LOCAL_PR_ARG, STA_PR_ARG, __entry->tid, __entry->buffered
+	)
+);
+
 /*
  * Tracing for internal functions
  * (which may also be called in response to driver calls)
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index ef3555e..27c990b 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -490,6 +490,7 @@ ieee80211_tx_h_unicast_ps_buf(struct ieee80211_tx_data *tx)
 		info->control.jiffies = jiffies;
 		info->control.vif = &tx->sdata->vif;
 		info->flags |= IEEE80211_TX_INTFL_NEED_TXPROCESSING;
+		info->flags &= ~IEEE80211_TX_TEMPORARY_FLAGS;
 		skb_queue_tail(&sta->ps_tx_buf[ac], tx->skb);
 
 		if (!timer_pending(&local->sta_cleanup))
@@ -1076,6 +1077,7 @@ static bool ieee80211_tx_prep_agg(struct ieee80211_tx_data *tx,
 			queued = true;
 			info->control.vif = &tx->sdata->vif;
 			info->flags |= IEEE80211_TX_INTFL_NEED_TXPROCESSING;
+			info->flags &= ~IEEE80211_TX_TEMPORARY_FLAGS;
 			__skb_queue_tail(&tid_tx->pending, skb);
 			if (skb_queue_len(&tid_tx->pending) > STA_MAX_TX_BUFFER)
 				purge_skb = __skb_dequeue(&tid_tx->pending);
diff --git a/net/mac80211/util.c b/net/mac80211/util.c
index df00f19..676dc09 100644
--- a/net/mac80211/util.c
+++ b/net/mac80211/util.c
@@ -2734,3 +2734,44 @@ int ieee80211_parse_p2p_noa(const struct ieee80211_p2p_noa_attr *attr,
 	return ret;
 }
 EXPORT_SYMBOL(ieee80211_parse_p2p_noa);
+
+void ieee80211_recalc_dtim(struct ieee80211_local *local,
+			   struct ieee80211_sub_if_data *sdata)
+{
+	u64 tsf = drv_get_tsf(local, sdata);
+	u64 dtim_count = 0;
+	u16 beacon_int = sdata->vif.bss_conf.beacon_int * 1024;
+	u8 dtim_period = sdata->vif.bss_conf.dtim_period;
+	struct ps_data *ps;
+	u8 bcns_from_dtim;
+
+	if (tsf == -1ULL || !beacon_int || !dtim_period)
+		return;
+
+	if (sdata->vif.type == NL80211_IFTYPE_AP ||
+	    sdata->vif.type == NL80211_IFTYPE_AP_VLAN) {
+		if (!sdata->bss)
+			return;
+
+		ps = &sdata->bss->ps;
+	} else if (ieee80211_vif_is_mesh(&sdata->vif)) {
+		ps = &sdata->u.mesh.ps;
+	} else {
+		return;
+	}
+
+	/*
+	 * actually finds last dtim_count, mac80211 will update in
+	 * __beacon_add_tim().
+	 * dtim_count = dtim_period - (tsf / bcn_int) % dtim_period
+	 */
+	do_div(tsf, beacon_int);
+	bcns_from_dtim = do_div(tsf, dtim_period);
+	/* just had a DTIM */
+	if (!bcns_from_dtim)
+		dtim_count = 0;
+	else
+		dtim_count = dtim_period - bcns_from_dtim;
+
+	ps->dtim_count = dtim_count;
+}
diff --git a/net/mac80211/wpa.c b/net/mac80211/wpa.c
index 7313d37..21448d6 100644
--- a/net/mac80211/wpa.c
+++ b/net/mac80211/wpa.c
@@ -127,7 +127,7 @@ ieee80211_rx_h_michael_mic_verify(struct ieee80211_rx_data *rx)
 		 * APs with pairwise keys should never receive Michael MIC
 		 * errors for non-zero keyidx because these are reserved for
 		 * group keys and only the AP is sending real multicast
-		 * frames in the BSS. (
+		 * frames in the BSS.
 		 */
 		return RX_DROP_UNUSABLE;
 	}