ath9k: fix dynack in IBSS mode

Currently, dynack was only tested upstream using AP/STA mode.
Testing it on IBSS, showed that late-ack detection was broken.

This is caused due to dynack using Association Request/Response
frames for late-ack detection, which IBSS does not use.
Also allowing Authentication frames here solves this.

A second issue also got fixed, which was also seen AP/STA mode:

When a station was added, the estimated value would be exponentially averaged
using 0 as a starting point.

This means that on larger distances, the ack timeout was still not high
enough before synchronizing would run out of late-ack's for estimation.

Fix this by using the initial estimated value as a baseline
and only start averaging in the following estimation rounds.

Test setup:
- 2x identical devices:  RB912UAG-5HPnD + 19dB sector
- IBSS
- 2x2 802.11an (ar9340), HT20, long GI
- RSSI's  -70 / -71
- Real distance: 23910 meter

Results (60s iperf runs):

Fixed coverage class 54 (up to 24300m):
* 21.5 Mbits/sec

Dynack:
* 28.9 Mbits/sec

Signed-off-by: Koen Vandeputte <[email protected]>

package/kernel/mac80211/patches/ath/553-ath9k-dynack-use-authentication-messages-for-late-ac.patch

@@ -0,0 +1,27 @@
+From: Lorenzo Bianconi <[email protected]>
+Date: Fri, 2 Nov 2018 21:49:55 +0100
+Subject: [PATCH] ath9k: dynack: use authentication messages for 'late' ack
+
+In order to properly support dynack in ad-hoc mode running
+wpa_supplicant, take into account authentication frames for
+'late ack' detection. This patch has been tested on devices
+mounted on offshore high-voltage stations connected through
+~24Km link
+
+Reported-by: Koen Vandeputte <[email protected]>
+Tested-by: Koen Vandeputte <[email protected]>
+Signed-off-by: Lorenzo Bianconi <[email protected]>
+---
+
+--- a/drivers/net/wireless/ath/ath9k/dynack.c
++++ b/drivers/net/wireless/ath/ath9k/dynack.c
+@@ -187,7 +187,8 @@ void ath_dynack_sample_tx_ts(struct ath_
+ 	/* late ACK */
+ 	if (ts->ts_status & ATH9K_TXERR_XRETRY) {
+ 		if (ieee80211_is_assoc_req(hdr->frame_control) ||
+-		    ieee80211_is_assoc_resp(hdr->frame_control)) {
++		    ieee80211_is_assoc_resp(hdr->frame_control) ||
++		    ieee80211_is_auth(hdr->frame_control)) {
+ 			ath_dbg(common, DYNACK, "late ack\n");
+ 			ath9k_hw_setslottime(ah, (LATEACK_TO - 3) / 2);
+ 			ath9k_hw_set_ack_timeout(ah, LATEACK_TO);

package/kernel/mac80211/patches/ath/554-ath9k-dynack-move-debug-log-after-buffer-increments.patch

@@ -0,0 +1,88 @@
+From: Lorenzo Bianconi <[email protected]>
+Date: Fri, 2 Nov 2018 21:49:56 +0100
+Subject: [PATCH] ath9k: dynack: move debug log after buffer increments
+
+Move debug log in ath_dynack_sample_tx_ts and ath_dynack_sample_ack_ts
+after timestamp buffer head/tail increments in order to make debugging
+more user friendly
+
+Tested-by: Koen Vandeputte <[email protected]>
+Signed-off-by: Lorenzo Bianconi <[email protected]>
+---
+
+--- a/drivers/net/wireless/ath/ath9k/dynack.c
++++ b/drivers/net/wireless/ath/ath9k/dynack.c
+@@ -171,11 +171,12 @@ static void ath_dynack_compute_to(struct
+ void ath_dynack_sample_tx_ts(struct ath_hw *ah, struct sk_buff *skb,
+ 			     struct ath_tx_status *ts)
+ {
+-	u8 ridx;
+ 	struct ieee80211_hdr *hdr;
+ 	struct ath_dynack *da = &ah->dynack;
+ 	struct ath_common *common = ath9k_hw_common(ah);
+ 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
++	u32 dur = ts->duration;
++	u8 ridx;
+ 
+ 	if ((info->flags & IEEE80211_TX_CTL_NO_ACK) || !da->enabled)
+ 		return;
+@@ -203,14 +204,13 @@ void ath_dynack_sample_tx_ts(struct ath_
+ 	ridx = ts->ts_rateindex;
+ 
+ 	da->st_rbf.ts[da->st_rbf.t_rb].tstamp = ts->ts_tstamp;
+-	da->st_rbf.ts[da->st_rbf.t_rb].dur = ts->duration;
+ 	ether_addr_copy(da->st_rbf.addr[da->st_rbf.t_rb].h_dest, hdr->addr1);
+ 	ether_addr_copy(da->st_rbf.addr[da->st_rbf.t_rb].h_src, hdr->addr2);
+ 
+ 	if (!(info->status.rates[ridx].flags & IEEE80211_TX_RC_MCS)) {
+-		u32 phy, sifs;
+ 		const struct ieee80211_rate *rate;
+ 		struct ieee80211_tx_rate *rates = info->status.rates;
++		u32 phy;
+ 
+ 		rate = &common->sbands[info->band].bitrates[rates[ridx].idx];
+ 		if (info->band == NL80211_BAND_2GHZ &&
+@@ -219,19 +219,18 @@ void ath_dynack_sample_tx_ts(struct ath_
+ 		else
+ 			phy = WLAN_RC_PHY_OFDM;
+ 
+-		sifs = ath_dynack_get_sifs(ah, phy);
+-		da->st_rbf.ts[da->st_rbf.t_rb].dur -= sifs;
++		dur -= ath_dynack_get_sifs(ah, phy);
+ 	}
+-
+-	ath_dbg(common, DYNACK, "{%pM} tx sample %u [dur %u][h %u-t %u]\n",
+-		hdr->addr1, da->st_rbf.ts[da->st_rbf.t_rb].tstamp,
+-		da->st_rbf.ts[da->st_rbf.t_rb].dur, da->st_rbf.h_rb,
+-		(da->st_rbf.t_rb + 1) % ATH_DYN_BUF);
++	da->st_rbf.ts[da->st_rbf.t_rb].dur = dur;
+ 
+ 	INCR(da->st_rbf.t_rb, ATH_DYN_BUF);
+ 	if (da->st_rbf.t_rb == da->st_rbf.h_rb)
+ 		INCR(da->st_rbf.h_rb, ATH_DYN_BUF);
+ 
++	ath_dbg(common, DYNACK, "{%pM} tx sample %u [dur %u][h %u-t %u]\n",
++		hdr->addr1, ts->ts_tstamp, dur, da->st_rbf.h_rb,
++		da->st_rbf.t_rb);
++
+ 	ath_dynack_compute_to(ah);
+ 
+ 	spin_unlock_bh(&da->qlock);
+@@ -258,14 +257,13 @@ void ath_dynack_sample_ack_ts(struct ath
+ 	spin_lock_bh(&da->qlock);
+ 	da->ack_rbf.tstamp[da->ack_rbf.t_rb] = ts;
+ 
+-	ath_dbg(common, DYNACK, "rx sample %u [h %u-t %u]\n",
+-		da->ack_rbf.tstamp[da->ack_rbf.t_rb],
+-		da->ack_rbf.h_rb, (da->ack_rbf.t_rb + 1) % ATH_DYN_BUF);
+-
+ 	INCR(da->ack_rbf.t_rb, ATH_DYN_BUF);
+ 	if (da->ack_rbf.t_rb == da->ack_rbf.h_rb)
+ 		INCR(da->ack_rbf.h_rb, ATH_DYN_BUF);
+ 
++	ath_dbg(common, DYNACK, "rx sample %u [h %u-t %u]\n",
++		ts, da->ack_rbf.h_rb, da->ack_rbf.t_rb);
++
+ 	ath_dynack_compute_to(ah);
+ 
+ 	spin_unlock_bh(&da->qlock);

package/kernel/mac80211/patches/ath/555-ath9k-dynack-check-da-enabled-first-in-sampling-rout.patch

@@ -0,0 +1,33 @@
+From: Lorenzo Bianconi <[email protected]>
+Date: Fri, 2 Nov 2018 21:49:57 +0100
+Subject: [PATCH] ath9k: dynack: check da->enabled first in sampling
+ routines
+
+Check da->enabled flag first in ath_dynack_sample_tx_ts and
+ath_dynack_sample_ack_ts routines in order to avoid useless
+processing
+
+Tested-by: Koen Vandeputte <[email protected]>
+Signed-off-by: Lorenzo Bianconi <[email protected]>
+---
+
+--- a/drivers/net/wireless/ath/ath9k/dynack.c
++++ b/drivers/net/wireless/ath/ath9k/dynack.c
+@@ -178,7 +178,7 @@ void ath_dynack_sample_tx_ts(struct ath_
+ 	u32 dur = ts->duration;
+ 	u8 ridx;
+ 
+-	if ((info->flags & IEEE80211_TX_CTL_NO_ACK) || !da->enabled)
++	if (!da->enabled || (info->flags & IEEE80211_TX_CTL_NO_ACK))
+ 		return;
+ 
+ 	spin_lock_bh(&da->qlock);
+@@ -251,7 +251,7 @@ void ath_dynack_sample_ack_ts(struct ath
+ 	struct ath_common *common = ath9k_hw_common(ah);
+ 	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
+ 
+-	if (!ath_dynack_bssidmask(ah, hdr->addr1) || !da->enabled)
++	if (!da->enabled || !ath_dynack_bssidmask(ah, hdr->addr1))
+ 		return;
+ 
+ 	spin_lock_bh(&da->qlock);

package/kernel/mac80211/patches/ath/556-ath9k-dynack-make-ewma-estimation-faster.patch

@@ -0,0 +1,139 @@
+From: Lorenzo Bianconi <[email protected]>
+Date: Fri, 2 Nov 2018 21:49:58 +0100
+Subject: [PATCH] ath9k: dynack: make ewma estimation faster
+
+In order to make propagation time estimation faster,
+use current sample as ewma output value during 'late ack'
+tracking
+
+Tested-by: Koen Vandeputte <[email protected]>
+Signed-off-by: Lorenzo Bianconi <[email protected]>
+---
+
+--- a/drivers/net/wireless/ath/ath9k/ath9k.h
++++ b/drivers/net/wireless/ath/ath9k/ath9k.h
+@@ -274,7 +274,7 @@ struct ath_node {
+ #endif
+ 	u8 key_idx[4];
+ 
+-	u32 ackto;
++	int ackto;
+ 	struct list_head list;
+ };
+ 
+--- a/drivers/net/wireless/ath/ath9k/dynack.c
++++ b/drivers/net/wireless/ath/ath9k/dynack.c
+@@ -29,9 +29,13 @@
+  * ath_dynack_ewma - EWMA (Exponentially Weighted Moving Average) calculation
+  *
+  */
+-static inline u32 ath_dynack_ewma(u32 old, u32 new)
++static inline int ath_dynack_ewma(int old, int new)
+ {
+-	return (new * (EWMA_DIV - EWMA_LEVEL) + old * EWMA_LEVEL) / EWMA_DIV;
++	if (old > 0)
++		return (new * (EWMA_DIV - EWMA_LEVEL) +
++			old * EWMA_LEVEL) / EWMA_DIV;
++	else
++		return new;
+ }
+ 
+ /**
+@@ -82,10 +86,10 @@ static inline bool ath_dynack_bssidmask(
+  */
+ static void ath_dynack_compute_ackto(struct ath_hw *ah)
+ {
+-	struct ath_node *an;
+-	u32 to = 0;
+-	struct ath_dynack *da = &ah->dynack;
+ 	struct ath_common *common = ath9k_hw_common(ah);
++	struct ath_dynack *da = &ah->dynack;
++	struct ath_node *an;
++	int to = 0;
+ 
+ 	list_for_each_entry(an, &da->nodes, list)
+ 		if (an->ackto > to)
+@@ -144,7 +148,8 @@ static void ath_dynack_compute_to(struct
+ 					an->ackto = ath_dynack_ewma(an->ackto,
+ 								    ackto);
+ 					ath_dbg(ath9k_hw_common(ah), DYNACK,
+-						"%pM to %u\n", dst, an->ackto);
++						"%pM to %d [%u]\n", dst,
++						an->ackto, ackto);
+ 					if (time_is_before_jiffies(da->lto)) {
+ 						ath_dynack_compute_ackto(ah);
+ 						da->lto = jiffies + COMPUTE_TO;
+@@ -166,10 +171,12 @@ static void ath_dynack_compute_to(struct
+  * @ah: ath hw
+  * @skb: socket buffer
+  * @ts: tx status info
++ * @sta: station pointer
+  *
+  */
+ void ath_dynack_sample_tx_ts(struct ath_hw *ah, struct sk_buff *skb,
+-			     struct ath_tx_status *ts)
++			     struct ath_tx_status *ts,
++			     struct ieee80211_sta *sta)
+ {
+ 	struct ieee80211_hdr *hdr;
+ 	struct ath_dynack *da = &ah->dynack;
+@@ -191,9 +198,16 @@ void ath_dynack_sample_tx_ts(struct ath_
+ 		    ieee80211_is_assoc_resp(hdr->frame_control) ||
+ 		    ieee80211_is_auth(hdr->frame_control)) {
+ 			ath_dbg(common, DYNACK, "late ack\n");
++
+ 			ath9k_hw_setslottime(ah, (LATEACK_TO - 3) / 2);
+ 			ath9k_hw_set_ack_timeout(ah, LATEACK_TO);
+ 			ath9k_hw_set_cts_timeout(ah, LATEACK_TO);
++			if (sta) {
++				struct ath_node *an;
++
++				an = (struct ath_node *)sta->drv_priv;
++				an->ackto = -1;
++			}
+ 			da->lto = jiffies + LATEACK_DELAY;
+ 		}
+ 
+--- a/drivers/net/wireless/ath/ath9k/dynack.h
++++ b/drivers/net/wireless/ath/ath9k/dynack.h
+@@ -86,7 +86,8 @@ void ath_dynack_node_deinit(struct ath_h
+ void ath_dynack_init(struct ath_hw *ah);
+ void ath_dynack_sample_ack_ts(struct ath_hw *ah, struct sk_buff *skb, u32 ts);
+ void ath_dynack_sample_tx_ts(struct ath_hw *ah, struct sk_buff *skb,
+-			     struct ath_tx_status *ts);
++			     struct ath_tx_status *ts,
++			     struct ieee80211_sta *sta);
+ #else
+ static inline void ath_dynack_init(struct ath_hw *ah) {}
+ static inline void ath_dynack_node_init(struct ath_hw *ah,
+@@ -97,7 +98,8 @@ static inline void ath_dynack_sample_ack
+ 					    struct sk_buff *skb, u32 ts) {}
+ static inline void ath_dynack_sample_tx_ts(struct ath_hw *ah,
+ 					   struct sk_buff *skb,
+-					   struct ath_tx_status *ts) {}
++					   struct ath_tx_status *ts,
++					   struct ieee80211_sta *sta) {}
+ #endif
+ 
+ #endif /* DYNACK_H */
+--- a/drivers/net/wireless/ath/ath9k/xmit.c
++++ b/drivers/net/wireless/ath/ath9k/xmit.c
+@@ -629,7 +629,7 @@ static void ath_tx_complete_aggr(struct
+ 				if (bf == bf->bf_lastbf)
+ 					ath_dynack_sample_tx_ts(sc->sc_ah,
+ 								bf->bf_mpdu,
+-								ts);
++								ts, sta);
+ 			}
+ 
+ 			ath_tx_complete_buf(sc, bf, txq, &bf_head, sta, ts,
+@@ -773,7 +773,8 @@ static void ath_tx_process_buffer(struct
+ 			memcpy(info->control.rates, bf->rates,
+ 			       sizeof(info->control.rates));
+ 			ath_tx_rc_status(sc, bf, ts, 1, txok ? 0 : 1, txok);
+-			ath_dynack_sample_tx_ts(sc->sc_ah, bf->bf_mpdu, ts);
++			ath_dynack_sample_tx_ts(sc->sc_ah, bf->bf_mpdu, ts,
++						sta);
+ 		}
+ 		ath_tx_complete_buf(sc, bf, txq, bf_head, sta, ts, txok);
+ 	} else

package/kernel/mac80211/patches/ath/557-ath9k-dynack-remove-experimental-tag.patch

@@ -0,0 +1,22 @@
+From: Lorenzo Bianconi <[email protected]>
+Date: Fri, 2 Nov 2018 21:49:59 +0100
+Subject: [PATCH] ath9k: dynack: remove 'experimental' tag
+
+Remove experimental tag from dynack Kconfig entry since it has
+been tested on outdoor 25Km links
+
+Tested-by: Koen Vandeputte <[email protected]>
+Signed-off-by: Lorenzo Bianconi <[email protected]>
+---
+
+--- a/drivers/net/wireless/ath/ath9k/Kconfig
++++ b/drivers/net/wireless/ath/ath9k/Kconfig
+@@ -132,7 +132,7 @@ config ATH9K_DFS_CERTIFIED
+ 	  except increase code size.
+ 
+ config ATH9K_DYNACK
+-	bool "Atheros ath9k ACK timeout estimation algorithm (EXPERIMENTAL)"
++	bool "Atheros ath9k ACK timeout estimation algorithm"
+ 	depends on ATH9K
+ 	default n
+ 	---help---