mac80211: minstrel_ht: improve rate probing for devices with static fallback

On some devices that only support static rate fallback tables sending rate
control probing packets can be really expensive.
Probing lower rates can already hurt throughput quite a bit. What hurts even
more is the fact that on mt76x0/mt76x2, single probing packets can only be
forced by directing packets at a different internal hardware queue, which
causes some heavy reordering and extra latency.
The reordering issue is mainly problematic while pushing lots of packets to
a particular station. If there is little activity, the overhead of probing is
neglegible.

The static fallback behavior is designed to pretty much only handle rate
control algorithms that use only a very limited set of rates on which the
algorithm switches up/down based on packet error rate.

In order to better support that kind of hardware, this patch implements a
different approach to rate probing where it switches to a slightly higher rate,
waits for tx status feedback, then updates the stats and switches back to
the new max throughput rate. This only triggers above a packet rate of 100
per stats interval (~50ms).
For that kind of probing, the code has to reduce the set of probing rates
a lot more compared to single packet probing, so it uses only one packet
per MCS group which is either slightly faster, or as close as possible to
the max throughput rate.
This allows switching between similar rates with different numbers of
streams. The algorithm assumes that the hardware will work its way lower
within an MCS group in case of retransmissions, so that lower rates don't
have to be probed by the high packets per second rate probing code.

To further reduce the search space, it also does not probe rates with lower
channel bandwidth than the max throughput rate.

At the moment, these changes will only affect mt76x0/mt76x2.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20190820095449.45255-4-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

1 files changed, 212 insertions, 28 deletions

diff --git a/net/mac80211/rc80211_minstrel_ht.c b/net/mac80211/rc80211_minstrel_ht.c
index c5868a1de306..a01168514840 100644
--- a/net/mac80211/rc80211_minstrel_ht.c
+++ b/net/mac80211/rc80211_minstrel_ht.c
@@ -18,6 +18,8 @@
 #define AVG_AMPDU_SIZE	16
 #define AVG_PKT_SIZE	1200
 
+#define SAMPLE_SWITCH_THR	100
+
 /* Number of bits for an average sized packet */
 #define MCS_NBITS ((AVG_PKT_SIZE * AVG_AMPDU_SIZE) << 3)
 
@@ -58,6 +60,7 @@
 	[GROUP_IDX(_streams, _sgi, _ht40)] = {				\
 	.streams = _streams,						\
 	.shift = _s,							\
+	.bw = _ht40,							\
 	.flags =							\
 		IEEE80211_TX_RC_MCS |					\
 		(_sgi ? IEEE80211_TX_RC_SHORT_GI : 0) |			\
@@ -94,6 +97,7 @@
 	[VHT_GROUP_IDX(_streams, _sgi, _bw)] = {			\
 	.streams = _streams,						\
 	.shift = _s,							\
+	.bw = _bw,							\
 	.flags =							\
 		IEEE80211_TX_RC_VHT_MCS |				\
 		(_sgi ? IEEE80211_TX_RC_SHORT_GI : 0) |			\
@@ -526,6 +530,133 @@ minstrel_ht_prob_rate_reduce_streams(struct minstrel_ht_sta *mi)
 	}
 }
 
+static inline int
+minstrel_get_duration(int index)
+{
+	const struct mcs_group *group = &minstrel_mcs_groups[index / MCS_GROUP_RATES];
+	unsigned int duration = group->duration[index % MCS_GROUP_RATES];
+	return duration << group->shift;
+}
+
+static bool
+minstrel_ht_probe_group(struct minstrel_ht_sta *mi, const struct mcs_group *tp_group,
+						int tp_idx, const struct mcs_group *group)
+{
+	if (group->bw < tp_group->bw)
+		return false;
+
+	if (group->streams == tp_group->streams)
+		return true;
+
+	if (tp_idx < 4 && group->streams == tp_group->streams - 1)
+		return true;
+
+	return group->streams == tp_group->streams + 1;
+}
+
+static void
+minstrel_ht_find_probe_rates(struct minstrel_ht_sta *mi, u16 *rates, int *n_rates,
+			     bool faster_rate)
+{
+	const struct mcs_group *group, *tp_group;
+	int i, g, max_dur;
+	int tp_idx;
+
+	tp_group = &minstrel_mcs_groups[mi->max_tp_rate[0] / MCS_GROUP_RATES];
+	tp_idx = mi->max_tp_rate[0] % MCS_GROUP_RATES;
+
+	max_dur = minstrel_get_duration(mi->max_tp_rate[0]);
+	if (faster_rate)
+		max_dur -= max_dur / 16;
+
+	for (g = 0; g < MINSTREL_GROUPS_NB; g++) {
+		u16 supported = mi->supported[g];
+
+		if (!supported)
+			continue;
+
+		group = &minstrel_mcs_groups[g];
+		if (!minstrel_ht_probe_group(mi, tp_group, tp_idx, group))
+			continue;
+
+		for (i = 0; supported; supported >>= 1, i++) {
+			int idx;
+
+			if (!(supported & 1))
+				continue;
+
+			if ((group->duration[i] << group->shift) > max_dur)
+				continue;
+
+			idx = g * MCS_GROUP_RATES + i;
+			if (idx == mi->max_tp_rate[0])
+				continue;
+
+			rates[(*n_rates)++] = idx;
+			break;
+		}
+	}
+}
+
+static void
+minstrel_ht_rate_sample_switch(struct minstrel_priv *mp,
+			       struct minstrel_ht_sta *mi)
+{
+	struct minstrel_rate_stats *mrs;
+	u16 rates[MINSTREL_GROUPS_NB];
+	int n_rates = 0;
+	int probe_rate = 0;
+	bool faster_rate;
+	int i;
+	u8 random;
+
+	/*
+	 * Use rate switching instead of probing packets for devices with
+	 * little control over retry fallback behavior
+	 */
+	if (mp->hw->max_rates > 1)
+		return;
+
+	/*
+	 * If the current EWMA prob is >75%, look for a rate that's 6.25%
+	 * faster than the max tp rate.
+	 * If that fails, look again for a rate that is at least as fast
+	 */
+	mrs = minstrel_get_ratestats(mi, mi->max_tp_rate[0]);
+	faster_rate = mrs->prob_ewma > MINSTREL_FRAC(75, 100);
+	minstrel_ht_find_probe_rates(mi, rates, &n_rates, faster_rate);
+	if (!n_rates && faster_rate)
+		minstrel_ht_find_probe_rates(mi, rates, &n_rates, false);
+
+	/* If no suitable rate was found, try to pick the next one in the group */
+	if (!n_rates) {
+		int g_idx = mi->max_tp_rate[0] / MCS_GROUP_RATES;
+		u16 supported = mi->supported[g_idx];
+
+		supported >>= mi->max_tp_rate[0] % MCS_GROUP_RATES;
+		for (i = 0; supported; i++) {
+			if (!(supported & 1))
+				continue;
+
+			probe_rate = mi->max_tp_rate[0] + i;
+			goto out;
+		}
+
+		return;
+	}
+
+	i = 0;
+	if (n_rates > 1) {
+		random = prandom_u32();
+		i = random % n_rates;
+	}
+	probe_rate = rates[i];
+
+out:
+	mi->sample_rate = probe_rate;
+	mi->sample_mode = MINSTREL_SAMPLE_ACTIVE;
+}
+
 /*
  * Update rate statistics and select new primary rates
  *
@@ -536,7 +667,8 @@ minstrel_ht_prob_rate_reduce_streams(struct minstrel_ht_sta *mi)
  *    higher throughput rates, even if the probablity is a bit lower
  */
 static void
-minstrel_ht_update_stats(struct minstrel_priv *mp, struct minstrel_ht_sta *mi)
+minstrel_ht_update_stats(struct minstrel_priv *mp, struct minstrel_ht_sta *mi,
+			 bool sample)
 {
 	struct minstrel_mcs_group_data *mg;
 	struct minstrel_rate_stats *mrs;
@@ -544,6 +676,18 @@ minstrel_ht_update_stats(struct minstrel_priv *mp, struct minstrel_ht_sta *mi)
 	u16 tmp_mcs_tp_rate[MAX_THR_RATES], tmp_group_tp_rate[MAX_THR_RATES];
 	u16 tmp_cck_tp_rate[MAX_THR_RATES], index;
 
+	mi->sample_mode = MINSTREL_SAMPLE_IDLE;
+
+	if (sample) {
+		mi->total_packets_cur = mi->total_packets -
+					mi->total_packets_last;
+		mi->total_packets_last = mi->total_packets;
+	}
+	if (!mp->sample_switch)
+		sample = false;
+	if (mi->total_packets_cur < SAMPLE_SWITCH_THR && mp->sample_switch != 1)
+	    sample = false;
+
 	if (mi->ampdu_packets > 0) {
 		if (!ieee80211_hw_check(mp->hw, TX_STATUS_NO_AMPDU_LEN))
 			mi->avg_ampdu_len = minstrel_ewma(mi->avg_ampdu_len,
@@ -630,12 +774,16 @@ minstrel_ht_update_stats(struct minstrel_priv *mp, struct minstrel_ht_sta *mi)
 	/* try to sample all available rates during each interval */
 	mi->sample_count *= 8;
 
+	if (sample)
+		minstrel_ht_rate_sample_switch(mp, mi);
+
 #ifdef CONFIG_MAC80211_DEBUGFS
 	/* use fixed index if set */
 	if (mp->fixed_rate_idx != -1) {
 		for (i = 0; i < 4; i++)
 			mi->max_tp_rate[i] = mp->fixed_rate_idx;
 		mi->max_prob_rate = mp->fixed_rate_idx;
+		mi->sample_mode = MINSTREL_SAMPLE_IDLE;
 	}
 #endif
 
@@ -739,15 +887,17 @@ minstrel_ht_tx_status(void *priv, struct ieee80211_supported_band *sband,
 	struct minstrel_ht_sta_priv *msp = priv_sta;
 	struct minstrel_ht_sta *mi = &msp->ht;
 	struct ieee80211_tx_rate *ar = info->status.rates;
-	struct minstrel_rate_stats *rate, *rate2;
+	struct minstrel_rate_stats *rate, *rate2, *rate_sample = NULL;
 	struct minstrel_priv *mp = priv;
 	bool last, update = false;
+	bool sample_status = false;
 	int i;
 
 	if (!msp->is_ht)
 		return mac80211_minstrel.tx_status_ext(priv, sband,
 						       &msp->legacy, st);
 
+
 	/* This packet was aggregated but doesn't carry status info */
 	if ((info->flags & IEEE80211_TX_CTL_AMPDU) &&
 	    !(info->flags & IEEE80211_TX_STAT_AMPDU))
@@ -773,12 +923,17 @@ minstrel_ht_tx_status(void *priv, struct ieee80211_supported_band *sband,
 	if (info->flags & IEEE80211_TX_CTL_RATE_CTRL_PROBE)
 		mi->sample_packets += info->status.ampdu_len;
 
+	if (mi->sample_mode != MINSTREL_SAMPLE_IDLE)
+		rate_sample = minstrel_get_ratestats(mi, mi->sample_rate);
+
 	last = !minstrel_ht_txstat_valid(mp, &ar[0]);
 	for (i = 0; !last; i++) {
 		last = (i == IEEE80211_TX_MAX_RATES - 1) ||
 		       !minstrel_ht_txstat_valid(mp, &ar[i + 1]);
 
 		rate = minstrel_ht_get_stats(mp, mi, &ar[i]);
+		if (rate == rate_sample)
+			sample_status = true;
 
 		if (last)
 			rate->success += info->status.ampdu_ack_len;
@@ -786,44 +941,60 @@ minstrel_ht_tx_status(void *priv, struct ieee80211_supported_band *sband,
 		rate->attempts += ar[i].count * info->status.ampdu_len;
 	}
 
-	/*
-	 * check for sudden death of spatial multiplexing,
-	 * downgrade to a lower number of streams if necessary.
-	 */
-	rate = minstrel_get_ratestats(mi, mi->max_tp_rate[0]);
-	if (rate->attempts > 30 &&
-	    MINSTREL_FRAC(rate->success, rate->attempts) <
-	    MINSTREL_FRAC(20, 100)) {
-		minstrel_downgrade_rate(mi, &mi->max_tp_rate[0], true);
+	switch (mi->sample_mode) {
+	case MINSTREL_SAMPLE_IDLE:
+		break;
+
+	case MINSTREL_SAMPLE_ACTIVE:
+		if (!sample_status)
+			break;
+
+		mi->sample_mode = MINSTREL_SAMPLE_PENDING;
 		update = true;
-	}
+		break;
+
+	case MINSTREL_SAMPLE_PENDING:
+		if (sample_status)
+			break;
 
-	rate2 = minstrel_get_ratestats(mi, mi->max_tp_rate[1]);
-	if (rate2->attempts > 30 &&
-	    MINSTREL_FRAC(rate2->success, rate2->attempts) <
-	    MINSTREL_FRAC(20, 100)) {
-		minstrel_downgrade_rate(mi, &mi->max_tp_rate[1], false);
 		update = true;
+		minstrel_ht_update_stats(mp, mi, false);
+		break;
+	}
+
+
+	if (mp->hw->max_rates > 1) {
+		/*
+		 * check for sudden death of spatial multiplexing,
+		 * downgrade to a lower number of streams if necessary.
+		 */
+		rate = minstrel_get_ratestats(mi, mi->max_tp_rate[0]);
+		if (rate->attempts > 30 &&
+		    MINSTREL_FRAC(rate->success, rate->attempts) <
+		    MINSTREL_FRAC(20, 100)) {
+			minstrel_downgrade_rate(mi, &mi->max_tp_rate[0], true);
+			update = true;
+		}
+
+		rate2 = minstrel_get_ratestats(mi, mi->max_tp_rate[1]);
+		if (rate2->attempts > 30 &&
+		    MINSTREL_FRAC(rate2->success, rate2->attempts) <
+		    MINSTREL_FRAC(20, 100)) {
+			minstrel_downgrade_rate(mi, &mi->max_tp_rate[1], false);
+			update = true;
+		}
 	}
 
 	if (time_after(jiffies, mi->last_stats_update +
 				(mp->update_interval / 2 * HZ) / 1000)) {
 		update = true;
-		minstrel_ht_update_stats(mp, mi);
+		minstrel_ht_update_stats(mp, mi, true);
 	}
 
 	if (update)
 		minstrel_ht_update_rates(mp, mi);
 }
 
-static inline int
-minstrel_get_duration(int index)
-{
-	const struct mcs_group *group = &minstrel_mcs_groups[index / MCS_GROUP_RATES];
-	unsigned int duration = group->duration[index % MCS_GROUP_RATES];
-	return duration << group->shift;
-}
-
 static void
 minstrel_calc_retransmit(struct minstrel_priv *mp, struct minstrel_ht_sta *mi,
                          int index)
@@ -988,14 +1159,18 @@ static void
 minstrel_ht_update_rates(struct minstrel_priv *mp, struct minstrel_ht_sta *mi)
 {
 	struct ieee80211_sta_rates *rates;
+	u16 first_rate = mi->max_tp_rate[0];
 	int i = 0;
 
+	if (mi->sample_mode == MINSTREL_SAMPLE_ACTIVE)
+		first_rate = mi->sample_rate;
+
 	rates = kzalloc(sizeof(*rates), GFP_ATOMIC);
 	if (!rates)
 		return;
 
 	/* Start with max_tp_rate[0] */
-	minstrel_ht_set_rate(mp, mi, rates, i++, mi->max_tp_rate[0]);
+	minstrel_ht_set_rate(mp, mi, rates, i++, first_rate);
 
 	if (mp->hw->max_rates >= 3) {
 		/* At least 3 tx rates supported, use max_tp_rate[1] next */
@@ -1020,6 +1195,11 @@ minstrel_get_sample_rate(struct minstrel_priv *mp, struct minstrel_ht_sta *mi)
 	int tp_rate1, tp_rate2;
 	int sample_idx = 0;
 
+	if (mp->hw->max_rates == 1 && mp->sample_switch &&
+	    (mi->total_packets_cur >= SAMPLE_SWITCH_THR ||
+	     mp->sample_switch == 1))
+		return -1;
+
 	if (mi->sample_wait > 0) {
 		mi->sample_wait--;
 		return -1;
@@ -1341,7 +1521,7 @@ minstrel_ht_update_caps(void *priv, struct ieee80211_supported_band *sband,
 	mi->supported[MINSTREL_CCK_GROUP] |= mi->cck_supported_short << 4;
 
 	/* create an initial rate table with the lowest supported rates */
-	minstrel_ht_update_stats(mp, mi);
+	minstrel_ht_update_stats(mp, mi, true);
 	minstrel_ht_update_rates(mp, mi);
 
 	return;
@@ -1459,6 +1639,8 @@ minstrel_ht_alloc(struct ieee80211_hw *hw, struct dentry *debugfsdir)
 	if (!mp)
 		return NULL;
 
+	mp->sample_switch = -1;
+
 	/* contention window settings
 	 * Just an approximation. Using the per-queue values would complicate
 	 * the calculations and is probably unnecessary */
@@ -1490,6 +1672,8 @@ minstrel_ht_alloc(struct ieee80211_hw *hw, struct dentry *debugfsdir)
 	mp->fixed_rate_idx = (u32) -1;
 	debugfs_create_u32("fixed_rate_idx", S_IRUGO | S_IWUGO, debugfsdir,
 			   &mp->fixed_rate_idx);
+	debugfs_create_u32("sample_switch", S_IRUGO | S_IWUSR, debugfsdir,
+			   &mp->sample_switch);
 #endif
 
 	minstrel_ht_init_cck_rates(mp);