Merge branch 'tcp_stretch_acks'

Neal Cardwell says:

====================
fix stretch ACK bugs in TCP CUBIC and Reno

This patch series fixes the TCP CUBIC and Reno congestion control
modules to properly handle stretch ACKs in their respective additive
increase modes, and in the transitions from slow start to additive
increase.

This finishes the project started by commit 9f9843a751d0a2057 ("tcp:
properly handle stretch acks in slow start"), which fixed behavior for
TCP congestion control when handling stretch ACKs in slow start mode.

Motivation: In the Jan 2015 netdev thread 'BW regression after "tcp:
refine TSO autosizing"', Eyal Perry documented a regression that Eric
Dumazet determined was caused by improper handling of TCP stretch
ACKs.

Background: LRO, GRO, delayed ACKs, and middleboxes can cause "stretch
ACKs" that cover more than the RFC-specified maximum of 2
packets. These stretch ACKs can cause serious performance shortfalls
in common congestion control algorithms, like Reno and CUBIC, which
were designed and tuned years ago with receiver hosts that were not
using LRO or GRO, and were instead ACKing every other packet.

Testing: at Google we have been using this approach for handling
stretch ACKs for CUBIC datacenter and Internet traffic for several
years, with good results.

v2:
 * fixed return type of tcp_slow_start() to be u32 instead of int
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

6 files changed, 42 insertions, 38 deletions

diff --git a/net/ipv4/tcp_bic.c b/net/ipv4/tcp_bic.c
index bb395d46a389..c037644eafb7 100644
--- a/net/ipv4/tcp_bic.c
+++ b/net/ipv4/tcp_bic.c
@@ -150,7 +150,7 @@ static void bictcp_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 		tcp_slow_start(tp, acked);
 	else {
 		bictcp_update(ca, tp->snd_cwnd);
-		tcp_cong_avoid_ai(tp, ca->cnt);
+		tcp_cong_avoid_ai(tp, ca->cnt, 1);
 	}
 }
 
diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c
index 27ead0dd16bc..8670e68e2ce6 100644
--- a/net/ipv4/tcp_cong.c
+++ b/net/ipv4/tcp_cong.c
@@ -291,26 +291,32 @@ int tcp_set_congestion_control(struct sock *sk, const char *name)
  * ABC caps N to 2. Slow start exits when cwnd grows over ssthresh and
  * returns the leftover acks to adjust cwnd in congestion avoidance mode.
  */
-void tcp_slow_start(struct tcp_sock *tp, u32 acked)
+u32 tcp_slow_start(struct tcp_sock *tp, u32 acked)
 {
 	u32 cwnd = tp->snd_cwnd + acked;
 
 	if (cwnd > tp->snd_ssthresh)
 		cwnd = tp->snd_ssthresh + 1;
+	acked -= cwnd - tp->snd_cwnd;
 	tp->snd_cwnd = min(cwnd, tp->snd_cwnd_clamp);
+
+	return acked;
 }
 EXPORT_SYMBOL_GPL(tcp_slow_start);
 
-/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd (or alternative w) */
-void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w)
+/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd (or alternative w),
+ * for every packet that was ACKed.
+ */
+void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w, u32 acked)
 {
+	tp->snd_cwnd_cnt += acked;
 	if (tp->snd_cwnd_cnt >= w) {
-		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
-			tp->snd_cwnd++;
-		tp->snd_cwnd_cnt = 0;
-	} else {
-		tp->snd_cwnd_cnt++;
+		u32 delta = tp->snd_cwnd_cnt / w;
+
+		tp->snd_cwnd_cnt -= delta * w;
+		tp->snd_cwnd += delta;
 	}
+	tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_cwnd_clamp);
 }
 EXPORT_SYMBOL_GPL(tcp_cong_avoid_ai);
 
@@ -329,11 +335,13 @@ void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 		return;
 
 	/* In "safe" area, increase. */
-	if (tp->snd_cwnd <= tp->snd_ssthresh)
-		tcp_slow_start(tp, acked);
+	if (tp->snd_cwnd <= tp->snd_ssthresh) {
+		acked = tcp_slow_start(tp, acked);
+		if (!acked)
+			return;
+	}
 	/* In dangerous area, increase slowly. */
-	else
-		tcp_cong_avoid_ai(tp, tp->snd_cwnd);
+	tcp_cong_avoid_ai(tp, tp->snd_cwnd, acked);
 }
 EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid);
 
diff --git a/net/ipv4/tcp_cubic.c b/net/ipv4/tcp_cubic.c
index 6b6002416a73..4b276d1ed980 100644
--- a/net/ipv4/tcp_cubic.c
+++ b/net/ipv4/tcp_cubic.c
@@ -93,9 +93,7 @@ struct bictcp {
 	u32	epoch_start;	/* beginning of an epoch */
 	u32	ack_cnt;	/* number of acks */
 	u32	tcp_cwnd;	/* estimated tcp cwnd */
-#define ACK_RATIO_SHIFT	4
-#define ACK_RATIO_LIMIT (32u << ACK_RATIO_SHIFT)
-	u16	delayed_ack;	/* estimate the ratio of Packets/ACKs << 4 */
+	u16	unused;
 	u8	sample_cnt;	/* number of samples to decide curr_rtt */
 	u8	found;		/* the exit point is found? */
 	u32	round_start;	/* beginning of each round */
@@ -114,7 +112,6 @@ static inline void bictcp_reset(struct bictcp *ca)
 	ca->bic_K = 0;
 	ca->delay_min = 0;
 	ca->epoch_start = 0;
-	ca->delayed_ack = 2 << ACK_RATIO_SHIFT;
 	ca->ack_cnt = 0;
 	ca->tcp_cwnd = 0;
 	ca->found = 0;
@@ -205,23 +202,30 @@ static u32 cubic_root(u64 a)
 /*
  * Compute congestion window to use.
  */
-static inline void bictcp_update(struct bictcp *ca, u32 cwnd)
+static inline void bictcp_update(struct bictcp *ca, u32 cwnd, u32 acked)
 {
 	u32 delta, bic_target, max_cnt;
 	u64 offs, t;
 
-	ca->ack_cnt++;	/* count the number of ACKs */
+	ca->ack_cnt += acked;	/* count the number of ACKed packets */
 
 	if (ca->last_cwnd == cwnd &&
 	    (s32)(tcp_time_stamp - ca->last_time) <= HZ / 32)
 		return;
 
+	/* The CUBIC function can update ca->cnt at most once per jiffy.
+	 * On all cwnd reduction events, ca->epoch_start is set to 0,
+	 * which will force a recalculation of ca->cnt.
+	 */
+	if (ca->epoch_start && tcp_time_stamp == ca->last_time)
+		goto tcp_friendliness;
+
 	ca->last_cwnd = cwnd;
 	ca->last_time = tcp_time_stamp;
 
 	if (ca->epoch_start == 0) {
 		ca->epoch_start = tcp_time_stamp;	/* record beginning */
-		ca->ack_cnt = 1;			/* start counting */
+		ca->ack_cnt = acked;			/* start counting */
 		ca->tcp_cwnd = cwnd;			/* syn with cubic */
 
 		if (ca->last_max_cwnd <= cwnd) {
@@ -283,6 +287,7 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd)
 	if (ca->last_max_cwnd == 0 && ca->cnt > 20)
 		ca->cnt = 20;	/* increase cwnd 5% per RTT */
 
+tcp_friendliness:
 	/* TCP Friendly */
 	if (tcp_friendliness) {
 		u32 scale = beta_scale;
@@ -301,7 +306,6 @@ static inline void bictcp_update(struct bictcp *ca, u32 cwnd)
 		}
 	}
 
-	ca->cnt = (ca->cnt << ACK_RATIO_SHIFT) / ca->delayed_ack;
 	if (ca->cnt == 0)			/* cannot be zero */
 		ca->cnt = 1;
 }
@@ -317,11 +321,12 @@ static void bictcp_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 	if (tp->snd_cwnd <= tp->snd_ssthresh) {
 		if (hystart && after(ack, ca->end_seq))
 			bictcp_hystart_reset(sk);
-		tcp_slow_start(tp, acked);
-	} else {
-		bictcp_update(ca, tp->snd_cwnd);
-		tcp_cong_avoid_ai(tp, ca->cnt);
+		acked = tcp_slow_start(tp, acked);
+		if (!acked)
+			return;
 	}
+	bictcp_update(ca, tp->snd_cwnd, acked);
+	tcp_cong_avoid_ai(tp, ca->cnt, acked);
 }
 
 static u32 bictcp_recalc_ssthresh(struct sock *sk)
@@ -411,20 +416,10 @@ static void hystart_update(struct sock *sk, u32 delay)
  */
 static void bictcp_acked(struct sock *sk, u32 cnt, s32 rtt_us)
 {
-	const struct inet_connection_sock *icsk = inet_csk(sk);
 	const struct tcp_sock *tp = tcp_sk(sk);
 	struct bictcp *ca = inet_csk_ca(sk);
 	u32 delay;
 
-	if (icsk->icsk_ca_state == TCP_CA_Open) {
-		u32 ratio = ca->delayed_ack;
-
-		ratio -= ca->delayed_ack >> ACK_RATIO_SHIFT;
-		ratio += cnt;
-
-		ca->delayed_ack = clamp(ratio, 1U, ACK_RATIO_LIMIT);
-	}
-
 	/* Some calls are for duplicates without timetamps */
 	if (rtt_us < 0)
 		return;
diff --git a/net/ipv4/tcp_scalable.c b/net/ipv4/tcp_scalable.c
index 6824afb65d93..333bcb2415ff 100644
--- a/net/ipv4/tcp_scalable.c
+++ b/net/ipv4/tcp_scalable.c
@@ -25,7 +25,8 @@ static void tcp_scalable_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 	if (tp->snd_cwnd <= tp->snd_ssthresh)
 		tcp_slow_start(tp, acked);
 	else
-		tcp_cong_avoid_ai(tp, min(tp->snd_cwnd, TCP_SCALABLE_AI_CNT));
+		tcp_cong_avoid_ai(tp, min(tp->snd_cwnd, TCP_SCALABLE_AI_CNT),
+				  1);
 }
 
 static u32 tcp_scalable_ssthresh(struct sock *sk)
diff --git a/net/ipv4/tcp_veno.c b/net/ipv4/tcp_veno.c
index a4d2d2d88dca..112151eeee45 100644
--- a/net/ipv4/tcp_veno.c
+++ b/net/ipv4/tcp_veno.c
@@ -159,7 +159,7 @@ static void tcp_veno_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 				/* In the "non-congestive state", increase cwnd
 				 *  every rtt.
 				 */
-				tcp_cong_avoid_ai(tp, tp->snd_cwnd);
+				tcp_cong_avoid_ai(tp, tp->snd_cwnd, 1);
 			} else {
 				/* In the "congestive state", increase cwnd
 				 * every other rtt.
diff --git a/net/ipv4/tcp_yeah.c b/net/ipv4/tcp_yeah.c
index cd7273218598..17d35662930d 100644
--- a/net/ipv4/tcp_yeah.c
+++ b/net/ipv4/tcp_yeah.c
@@ -92,7 +92,7 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, u32 acked)
 
 	} else {
 		/* Reno */
-		tcp_cong_avoid_ai(tp, tp->snd_cwnd);
+		tcp_cong_avoid_ai(tp, tp->snd_cwnd, 1);
 	}
 
 	/* The key players are v_vegas.beg_snd_una and v_beg_snd_nxt.