1 files changed, 55 insertions, 25 deletions

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 12fda8f27b08..12c26c9565b7 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -261,7 +261,8 @@ static void tcp_ecn_accept_cwr(struct sock *sk, const struct sk_buff *skb)
 		 * cwnd may be very low (even just 1 packet), so we should ACK
 		 * immediately.
 		 */
-		inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_NOW;
+		if (TCP_SKB_CB(skb)->seq != TCP_SKB_CB(skb)->end_seq)
+			inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_NOW;
 	}
 }
 
@@ -961,6 +962,15 @@ void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb)
 	}
 }
 
+/* Updates the delivered and delivered_ce counts */
+static void tcp_count_delivered(struct tcp_sock *tp, u32 delivered,
+				bool ece_ack)
+{
+	tp->delivered += delivered;
+	if (ece_ack)
+		tp->delivered_ce += delivered;
+}
+
 /* This procedure tags the retransmission queue when SACKs arrive.
  *
  * We have three tag bits: SACKED(S), RETRANS(R) and LOST(L).
@@ -1137,6 +1147,7 @@ struct tcp_sacktag_state {
 	struct rate_sample *rate;
 	int	flag;
 	unsigned int mss_now;
+	u32	sack_delivered;
 };
 
 /* Check if skb is fully within the SACK block. In presence of GSO skbs,
@@ -1257,7 +1268,8 @@ static u8 tcp_sacktag_one(struct sock *sk,
 		sacked |= TCPCB_SACKED_ACKED;
 		state->flag |= FLAG_DATA_SACKED;
 		tp->sacked_out += pcount;
-		tp->delivered += pcount;  /* Out-of-order packets delivered */
+		/* Out-of-order packets delivered */
+		state->sack_delivered += pcount;
 
 		/* Lost marker hint past SACKed? Tweak RFC3517 cnt */
 		if (tp->lost_skb_hint &&
@@ -1683,7 +1695,8 @@ tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb,
 					 num_sacks, prior_snd_una);
 	if (found_dup_sack) {
 		state->flag |= FLAG_DSACKING_ACK;
-		tp->delivered++; /* A spurious retransmission is delivered */
+		/* A spurious retransmission is delivered */
+		state->sack_delivered++;
 	}
 
 	/* Eliminate too old ACKs, but take into
@@ -1892,7 +1905,7 @@ static void tcp_check_reno_reordering(struct sock *sk, const int addend)
 
 /* Emulate SACKs for SACKless connection: account for a new dupack. */
 
-static void tcp_add_reno_sack(struct sock *sk, int num_dupack)
+static void tcp_add_reno_sack(struct sock *sk, int num_dupack, bool ece_ack)
 {
 	if (num_dupack) {
 		struct tcp_sock *tp = tcp_sk(sk);
@@ -1903,20 +1916,21 @@ static void tcp_add_reno_sack(struct sock *sk, int num_dupack)
 		tcp_check_reno_reordering(sk, 0);
 		delivered = tp->sacked_out - prior_sacked;
 		if (delivered > 0)
-			tp->delivered += delivered;
+			tcp_count_delivered(tp, delivered, ece_ack);
 		tcp_verify_left_out(tp);
 	}
 }
 
 /* Account for ACK, ACKing some data in Reno Recovery phase. */
 
-static void tcp_remove_reno_sacks(struct sock *sk, int acked)
+static void tcp_remove_reno_sacks(struct sock *sk, int acked, bool ece_ack)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 
 	if (acked > 0) {
 		/* One ACK acked hole. The rest eat duplicate ACKs. */
-		tp->delivered += max_t(int, acked - tp->sacked_out, 1);
+		tcp_count_delivered(tp, max_t(int, acked - tp->sacked_out, 1),
+				    ece_ack);
 		if (acked - 1 >= tp->sacked_out)
 			tp->sacked_out = 0;
 		else
@@ -2696,7 +2710,7 @@ static void tcp_process_loss(struct sock *sk, int flag, int num_dupack,
 		 * delivered. Lower inflight to clock out (re)tranmissions.
 		 */
 		if (after(tp->snd_nxt, tp->high_seq) && num_dupack)
-			tcp_add_reno_sack(sk, num_dupack);
+			tcp_add_reno_sack(sk, num_dupack, flag & FLAG_ECE);
 		else if (flag & FLAG_SND_UNA_ADVANCED)
 			tcp_reset_reno_sack(tp);
 	}
@@ -2778,6 +2792,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
 	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
 	int fast_rexmit = 0, flag = *ack_flag;
+	bool ece_ack = flag & FLAG_ECE;
 	bool do_lost = num_dupack || ((flag & FLAG_DATA_SACKED) &&
 				      tcp_force_fast_retransmit(sk));
 
@@ -2786,7 +2801,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
 
 	/* Now state machine starts.
 	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
-	if (flag & FLAG_ECE)
+	if (ece_ack)
 		tp->prior_ssthresh = 0;
 
 	/* B. In all the states check for reneging SACKs. */
@@ -2827,7 +2842,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
 	case TCP_CA_Recovery:
 		if (!(flag & FLAG_SND_UNA_ADVANCED)) {
 			if (tcp_is_reno(tp))
-				tcp_add_reno_sack(sk, num_dupack);
+				tcp_add_reno_sack(sk, num_dupack, ece_ack);
 		} else {
 			if (tcp_try_undo_partial(sk, prior_snd_una))
 				return;
@@ -2852,7 +2867,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
 		if (tcp_is_reno(tp)) {
 			if (flag & FLAG_SND_UNA_ADVANCED)
 				tcp_reset_reno_sack(tp);
-			tcp_add_reno_sack(sk, num_dupack);
+			tcp_add_reno_sack(sk, num_dupack, ece_ack);
 		}
 
 		if (icsk->icsk_ca_state <= TCP_CA_Disorder)
@@ -2876,7 +2891,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
 		}
 
 		/* Otherwise enter Recovery state */
-		tcp_enter_recovery(sk, (flag & FLAG_ECE));
+		tcp_enter_recovery(sk, ece_ack);
 		fast_rexmit = 1;
 	}
 
@@ -3052,7 +3067,7 @@ static void tcp_ack_tstamp(struct sock *sk, struct sk_buff *skb,
  */
 static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 			       u32 prior_snd_una,
-			       struct tcp_sacktag_state *sack)
+			       struct tcp_sacktag_state *sack, bool ece_ack)
 {
 	const struct inet_connection_sock *icsk = inet_csk(sk);
 	u64 first_ackt, last_ackt;
@@ -3077,8 +3092,6 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 		u8 sacked = scb->sacked;
 		u32 acked_pcount;
 
-		tcp_ack_tstamp(sk, skb, prior_snd_una);
-
 		/* Determine how many packets and what bytes were acked, tso and else */
 		if (after(scb->end_seq, tp->snd_una)) {
 			if (tcp_skb_pcount(skb) == 1 ||
@@ -3113,7 +3126,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 		if (sacked & TCPCB_SACKED_ACKED) {
 			tp->sacked_out -= acked_pcount;
 		} else if (tcp_is_sack(tp)) {
-			tp->delivered += acked_pcount;
+			tcp_count_delivered(tp, acked_pcount, ece_ack);
 			if (!tcp_skb_spurious_retrans(tp, skb))
 				tcp_rack_advance(tp, sacked, scb->end_seq,
 						 tcp_skb_timestamp_us(skb));
@@ -3142,6 +3155,8 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 		if (!fully_acked)
 			break;
 
+		tcp_ack_tstamp(sk, skb, prior_snd_una);
+
 		next = skb_rb_next(skb);
 		if (unlikely(skb == tp->retransmit_skb_hint))
 			tp->retransmit_skb_hint = NULL;
@@ -3157,8 +3172,11 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 	if (likely(between(tp->snd_up, prior_snd_una, tp->snd_una)))
 		tp->snd_up = tp->snd_una;
 
-	if (skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
-		flag |= FLAG_SACK_RENEGING;
+	if (skb) {
+		tcp_ack_tstamp(sk, skb, prior_snd_una);
+		if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
+			flag |= FLAG_SACK_RENEGING;
+	}
 
 	if (likely(first_ackt) && !(flag & FLAG_RETRANS_DATA_ACKED)) {
 		seq_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, first_ackt);
@@ -3190,7 +3208,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
 		}
 
 		if (tcp_is_reno(tp)) {
-			tcp_remove_reno_sacks(sk, pkts_acked);
+			tcp_remove_reno_sacks(sk, pkts_acked, ece_ack);
 
 			/* If any of the cumulatively ACKed segments was
 			 * retransmitted, non-SACK case cannot confirm that
@@ -3557,10 +3575,9 @@ static u32 tcp_newly_delivered(struct sock *sk, u32 prior_delivered, int flag)
 
 	delivered = tp->delivered - prior_delivered;
 	NET_ADD_STATS(net, LINUX_MIB_TCPDELIVERED, delivered);
-	if (flag & FLAG_ECE) {
-		tp->delivered_ce += delivered;
+	if (flag & FLAG_ECE)
 		NET_ADD_STATS(net, LINUX_MIB_TCPDELIVEREDCE, delivered);
-	}
+
 	return delivered;
 }
 
@@ -3584,6 +3601,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 
 	sack_state.first_sackt = 0;
 	sack_state.rate = &rs;
+	sack_state.sack_delivered = 0;
 
 	/* We very likely will need to access rtx queue. */
 	prefetch(sk->tcp_rtx_queue.rb_node);
@@ -3659,12 +3677,25 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 			ack_ev_flags |= CA_ACK_ECE;
 		}
 
+		if (sack_state.sack_delivered)
+			tcp_count_delivered(tp, sack_state.sack_delivered,
+					    flag & FLAG_ECE);
+
 		if (flag & FLAG_WIN_UPDATE)
 			ack_ev_flags |= CA_ACK_WIN_UPDATE;
 
 		tcp_in_ack_event(sk, ack_ev_flags);
 	}
 
+	/* This is a deviation from RFC3168 since it states that:
+	 * "When the TCP data sender is ready to set the CWR bit after reducing
+	 * the congestion window, it SHOULD set the CWR bit only on the first
+	 * new data packet that it transmits."
+	 * We accept CWR on pure ACKs to be more robust
+	 * with widely-deployed TCP implementations that do this.
+	 */
+	tcp_ecn_accept_cwr(sk, skb);
+
 	/* We passed data and got it acked, remove any soft error
 	 * log. Something worked...
 	 */
@@ -3675,7 +3706,8 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 		goto no_queue;
 
 	/* See if we can take anything off of the retransmit queue. */
-	flag |= tcp_clean_rtx_queue(sk, prior_fack, prior_snd_una, &sack_state);
+	flag |= tcp_clean_rtx_queue(sk, prior_fack, prior_snd_una, &sack_state,
+				    flag & FLAG_ECE);
 
 	tcp_rack_update_reo_wnd(sk, &rs);
 
@@ -4800,8 +4832,6 @@ static void tcp_data_queue(struct sock *sk, struct sk_buff *skb)
 	skb_dst_drop(skb);
 	__skb_pull(skb, tcp_hdr(skb)->doff * 4);
 
-	tcp_ecn_accept_cwr(sk, skb);
-
 	tp->rx_opt.dsack = 0;
 
 	/*  Queue data for delivery to the user.