patch-2.4.27 linux-2.4.27/net/ipv4/tcp_input.c

• Lines: 772
• Date: 2004-08-07 16:26:07.005443380 -0700
• Orig file: linux-2.4.26/net/ipv4/tcp_input.c
• Orig date: 2004-04-14 06:05:41.000000000 -0700

diff -urN linux-2.4.26/net/ipv4/tcp_input.c linux-2.4.27/net/ipv4/tcp_input.c
@@ -90,7 +90,23 @@
 int sysctl_tcp_max_orphans = NR_FILE;
 int sysctl_tcp_frto = 0;
 
+int sysctl_tcp_nometrics_save = 0;
+
 int sysctl_tcp_westwood = 0;
+int sysctl_tcp_vegas_cong_avoid = 0;
+
+int sysctl_tcp_moderate_rcvbuf = 0;
+
+/* Default values of the Vegas variables, in fixed-point representation
+ * with V_PARAM_SHIFT bits to the right of the binary point.
+ */
+#define V_PARAM_SHIFT 1
+int sysctl_tcp_vegas_alpha = 1<<V_PARAM_SHIFT;
+int sysctl_tcp_vegas_beta  = 3<<V_PARAM_SHIFT;
+int sysctl_tcp_vegas_gamma = 1<<V_PARAM_SHIFT;
+int sysctl_tcp_bic;
+int sysctl_tcp_bic_fast_convergence = 1;
+int sysctl_tcp_bic_low_window = 14;
 
 #define FLAG_DATA		0x01 /* Incoming frame contained data.		*/
 #define FLAG_WIN_UPDATE		0x02 /* Incoming ACK was a window update.	*/
@@ -296,6 +312,8 @@
 	if (!(sk->userlocks&SOCK_SNDBUF_LOCK))
 		tcp_fixup_sndbuf(sk);
 
+	tp->rcvq_space.space = tp->rcv_wnd;
+
 	maxwin = tcp_full_space(sk);
 
 	if (tp->window_clamp >= maxwin) {
@@ -315,6 +333,15 @@
 	tp->snd_cwnd_stamp = tcp_time_stamp;
 }
 
+static void init_bictcp(struct tcp_opt *tp)
+{
+	tp->bictcp.cnt = 0;
+
+	tp->bictcp.last_max_cwnd = 0;
+	tp->bictcp.last_cwnd = 0;
+	tp->bictcp.last_stamp = 0;
+}
+
 /* 5. Recalculate window clamp after socket hit its memory bounds. */
 static void tcp_clamp_window(struct sock *sk, struct tcp_opt *tp)
 {
@@ -352,6 +379,122 @@
 	}
 }
 
+/* Receiver "autotuning" code.
+ *
+ * The algorithm for RTT estimation w/o timestamps is based on
+ * Dynamic Right-Sizing (DRS) by Wu Feng and Mike Fisk of LANL.
+ * <http://www.lanl.gov/radiant/website/pubs/drs/lacsi2001.ps>
+ *
+ * More detail on this code can be found at
+ * <http://www.psc.edu/~jheffner/senior_thesis.ps>,
+ * though this reference is out of date.  A new paper
+ * is pending.
+ */
+static void tcp_rcv_rtt_update(struct tcp_opt *tp, u32 sample, int win_dep)
+{
+	u32 new_sample = tp->rcv_rtt_est.rtt;
+	long m = sample;
+
+	if (m == 0)
+		m = 1;
+
+	if (new_sample != 0) {
+		/* If we sample in larger samples in the non-timestamp
+		 * case, we could grossly overestimate the RTT especially
+		 * with chatty applications or bulk transfer apps which
+		 * are stalled on filesystem I/O.
+		 *
+		 * Also, since we are only going for a minimum in the
+		 * non-timestamp case, we do not smoothe things out
+		 * else with timestamps disabled convergance takes too
+		 * long.
+		 */
+		if (!win_dep) {
+			m -= (new_sample >> 3);
+			new_sample += m;
+		} else if (m < new_sample)
+			new_sample = m << 3;
+	} else {
+		/* No previous mesaure. */
+		new_sample = m << 3;
+	}
+
+	if (tp->rcv_rtt_est.rtt != new_sample)
+		tp->rcv_rtt_est.rtt = new_sample;
+}
+
+static inline void tcp_rcv_rtt_measure(struct tcp_opt *tp)
+{
+	if (tp->rcv_rtt_est.time == 0)
+		goto new_measure;
+	if (before(tp->rcv_nxt, tp->rcv_rtt_est.seq))
+		return;
+	tcp_rcv_rtt_update(tp,
+			   jiffies - tp->rcv_rtt_est.time,
+			   1);
+
+new_measure:
+	tp->rcv_rtt_est.seq = tp->rcv_nxt + tp->rcv_wnd;
+	tp->rcv_rtt_est.time = tcp_time_stamp;
+}
+
+static inline void tcp_rcv_rtt_measure_ts(struct tcp_opt *tp, struct sk_buff *skb)
+{
+	if (tp->rcv_tsecr &&
+	    (TCP_SKB_CB(skb)->end_seq -
+	     TCP_SKB_CB(skb)->seq >= tp->ack.rcv_mss))
+		tcp_rcv_rtt_update(tp, tcp_time_stamp - tp->rcv_tsecr, 0);
+}
+
+/*
+ * This function should be called every time data is copied to user space.
+ * It calculates the appropriate TCP receive buffer space.
+ */
+void tcp_rcv_space_adjust(struct sock *sk)
+{
+	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
+	int time;
+	int space;
+	
+	if (tp->rcvq_space.time == 0)
+		goto new_measure;
+	
+	time = tcp_time_stamp - tp->rcvq_space.time;
+	if (time < (tp->rcv_rtt_est.rtt >> 3) ||
+	    tp->rcv_rtt_est.rtt == 0)
+		return;
+	
+	space = 2 * (tp->copied_seq - tp->rcvq_space.seq);
+
+	space = max(tp->rcvq_space.space, space);
+
+	if (tp->rcvq_space.space != space) {
+		int rcvmem;
+
+		tp->rcvq_space.space = space;
+
+		if (sysctl_tcp_moderate_rcvbuf) {
+			/* Receive space grows, normalize in order to
+			 * take into account packet headers and sk_buff
+			 * structure overhead.
+			 */
+			space /= tp->advmss;
+			if (!space)
+				space = 1;
+			rcvmem = (tp->advmss + MAX_TCP_HEADER +
+				  16 + sizeof(struct sk_buff));
+			space *= rcvmem;
+			space = min(space, sysctl_tcp_rmem[2]);
+			if (space > sk->rcvbuf)
+				sk->rcvbuf = space;
+		}
+	}
+	
+new_measure:
+	tp->rcvq_space.seq = tp->copied_seq;
+	tp->rcvq_space.time = tcp_time_stamp;
+}
+
 /* There is something which you must keep in mind when you analyze the
  * behavior of the tp->ato delayed ack timeout interval.  When a
  * connection starts up, we want to ack as quickly as possible.  The
@@ -370,6 +513,8 @@
 
 	tcp_measure_rcv_mss(tp, skb);
 
+	tcp_rcv_rtt_measure(tp);
+
 	now = tcp_time_stamp;
 
 	if (!tp->ack.ato) {
@@ -404,6 +549,42 @@
 		tcp_grow_window(sk, tp, skb);
 }
 
+/* Set up a new TCP connection, depending on whether it should be
+ * using Vegas or not.
+ */    
+void tcp_vegas_init(struct tcp_opt *tp)
+{
+	if (sysctl_tcp_vegas_cong_avoid) {
+		tp->vegas.do_vegas = 1;
+		tp->vegas.baseRTT = 0x7fffffff;
+		tcp_vegas_enable(tp);
+	} else 
+		tcp_vegas_disable(tp);
+}
+
+/* Do RTT sampling needed for Vegas.
+ * Basically we:
+ *   o min-filter RTT samples from within an RTT to get the current
+ *     propagation delay + queuing delay (we are min-filtering to try to
+ *     avoid the effects of delayed ACKs)
+ *   o min-filter RTT samples from a much longer window (forever for now)
+ *     to find the propagation delay (baseRTT)
+ */
+static inline void vegas_rtt_calc(struct tcp_opt *tp, __u32 rtt)
+{
+	__u32 vrtt = rtt + 1; /* Never allow zero rtt or baseRTT */
+
+	/* Filter to find propagation delay: */
+	if (vrtt < tp->vegas.baseRTT) 
+		tp->vegas.baseRTT = vrtt;
+
+	/* Find the min RTT during the last RTT to find
+	 * the current prop. delay + queuing delay:
+	 */
+	tp->vegas.minRTT = min(tp->vegas.minRTT, vrtt);
+	tp->vegas.cntRTT++;
+}
+
 /* Called to compute a smoothed rtt estimate. The data fed to this
  * routine either comes from timestamps, or from segments that were
  * known _not_ to have been retransmitted [see Karn/Partridge
@@ -417,6 +598,9 @@
 {
 	long m = mrtt; /* RTT */
 
+	if (tcp_vegas_enabled(tp))
+		vegas_rtt_calc(tp, mrtt);
+
 	/*	The following amusing code comes from Jacobson's
 	 *	article in SIGCOMM '88.  Note that rtt and mdev
 	 *	are scaled versions of rtt and mean deviation.
@@ -519,6 +703,9 @@
 	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
 	struct dst_entry *dst = __sk_dst_get(sk);
 
+	if (sysctl_tcp_nometrics_save)
+		return;
+
 	dst_confirm(dst);
 
 	if (dst && (dst->flags&DST_HOST)) {
@@ -994,7 +1181,7 @@
 	}
 	tcp_sync_left_out(tp);
 
-	tp->ca_state = TCP_CA_Open;
+	tcp_set_ca_state(tp, TCP_CA_Open);
 	tp->frto_highmark = tp->snd_nxt;
 }
 
@@ -1040,9 +1227,11 @@
 
 	tp->reordering = min_t(unsigned int, tp->reordering,
 					     sysctl_tcp_reordering);
-	tp->ca_state = TCP_CA_Loss;
+	tcp_set_ca_state(tp, TCP_CA_Loss);
 	tp->high_seq = tp->frto_highmark;
 	TCP_ECN_queue_cwr(tp);
+
+	init_bictcp(tp);
 }
 
 void tcp_clear_retrans(struct tcp_opt *tp)
@@ -1105,7 +1294,7 @@
 	tcp_sync_left_out(tp);
 
 	tp->reordering = min_t(unsigned int, tp->reordering, sysctl_tcp_reordering);
-	tp->ca_state = TCP_CA_Loss;
+	tcp_set_ca_state(tp, TCP_CA_Loss);
 	tp->high_seq = tp->snd_nxt;
 	TCP_ECN_queue_cwr(tp);
 }
@@ -1480,7 +1669,7 @@
 		tcp_moderate_cwnd(tp);
 		return 1;
 	}
-	tp->ca_state = TCP_CA_Open;
+	tcp_set_ca_state(tp, TCP_CA_Open);
 	return 0;
 }
 
@@ -1540,7 +1729,7 @@
 		tp->retransmits = 0;
 		tp->undo_marker = 0;
 		if (!IsReno(tp))
-			tp->ca_state = TCP_CA_Open;
+			tcp_set_ca_state(tp, TCP_CA_Open);
 		return 1;
 	}
 	return 0;
@@ -1572,7 +1761,7 @@
 			state = TCP_CA_Disorder;
 
 		if (tp->ca_state != state) {
-			tp->ca_state = state;
+			tcp_set_ca_state(tp, state);
 			tp->high_seq = tp->snd_nxt;
 		}
 		tcp_moderate_cwnd(tp);
@@ -1647,7 +1836,7 @@
 			 * is ACKed for CWR bit to reach receiver. */
 			if (tp->snd_una != tp->high_seq) {
 				tcp_complete_cwr(tp);
-				tp->ca_state = TCP_CA_Open;
+				tcp_set_ca_state(tp, TCP_CA_Open);
 			}
 			break;
 
@@ -1658,7 +1847,7 @@
 			     * catching for all duplicate ACKs. */
 			    IsReno(tp) || tp->snd_una != tp->high_seq) {
 				tp->undo_marker = 0;
-				tp->ca_state = TCP_CA_Open;
+				tcp_set_ca_state(tp, TCP_CA_Open);
 			}
 			break;
 
@@ -1732,7 +1921,7 @@
 		}
 
 		tp->snd_cwnd_cnt = 0;
-		tp->ca_state = TCP_CA_Recovery;
+		tcp_set_ca_state(tp, TCP_CA_Recovery);
 	}
 
 	if (is_dupack || tcp_head_timedout(sk, tp))
@@ -1800,10 +1989,73 @@
 		tcp_ack_no_tstamp(tp, seq_rtt, flag);
 }
 
+/*
+ * Compute congestion window to use.
+ *
+ * This is from the implementation of BICTCP in
+ * Lison-Xu, Kahaled Harfoush, and Injog Rhee.
+ *  "Binary Increase Congestion Control for Fast, Long Distance
+ *  Networks" in InfoComm 2004
+ * Available from:
+ *  http://www.csc.ncsu.edu/faculty/rhee/export/bitcp.pdf
+ *
+ * Unless BIC is enabled and congestion window is large
+ * this behaves the same as the original Reno.
+ */
+static inline __u32 bictcp_cwnd(struct tcp_opt *tp)
+{
+	/* orignal Reno behaviour */
+	if (!sysctl_tcp_bic)
+		return tp->snd_cwnd;
+
+	if (tp->bictcp.last_cwnd == tp->snd_cwnd &&
+	   (s32)(tcp_time_stamp - tp->bictcp.last_stamp) <= (HZ>>5))
+		return tp->bictcp.cnt;
+
+	tp->bictcp.last_cwnd = tp->snd_cwnd;
+	tp->bictcp.last_stamp = tcp_time_stamp;
+      
+	/* start off normal */
+	if (tp->snd_cwnd <= sysctl_tcp_bic_low_window)
+		tp->bictcp.cnt = tp->snd_cwnd;
+
+	/* binary increase */
+	else if (tp->snd_cwnd < tp->bictcp.last_max_cwnd) {
+		__u32 	dist = (tp->bictcp.last_max_cwnd - tp->snd_cwnd)
+			/ BICTCP_B;
+
+		if (dist > BICTCP_MAX_INCREMENT)
+			/* linear increase */
+			tp->bictcp.cnt = tp->snd_cwnd / BICTCP_MAX_INCREMENT;
+		else if (dist <= 1U)
+			/* binary search increase */
+			tp->bictcp.cnt = tp->snd_cwnd * BICTCP_FUNC_OF_MIN_INCR
+				/ BICTCP_B;
+		else
+			/* binary search increase */
+			tp->bictcp.cnt = tp->snd_cwnd / dist;
+	} else {
+		/* slow start amd linear increase */
+		if (tp->snd_cwnd < tp->bictcp.last_max_cwnd + BICTCP_B)
+			/* slow start */
+			tp->bictcp.cnt = tp->snd_cwnd * BICTCP_FUNC_OF_MIN_INCR
+				/ BICTCP_B;
+		else if (tp->snd_cwnd < tp->bictcp.last_max_cwnd
+			 		+ BICTCP_MAX_INCREMENT*(BICTCP_B-1))
+			/* slow start */
+			tp->bictcp.cnt = tp->snd_cwnd * (BICTCP_B-1)
+				/ (tp->snd_cwnd-tp->bictcp.last_max_cwnd);
+		else
+			/* linear increase */
+			tp->bictcp.cnt = tp->snd_cwnd / BICTCP_MAX_INCREMENT;
+	}
+	return tp->bictcp.cnt;
+}
+
 /* This is Jacobson's slow start and congestion avoidance. 
  * SIGCOMM '88, p. 328.
  */
-static __inline__ void tcp_cong_avoid(struct tcp_opt *tp)
+static __inline__ void reno_cong_avoid(struct tcp_opt *tp)
 {
         if (tp->snd_cwnd <= tp->snd_ssthresh) {
                 /* In "safe" area, increase. */
@@ -1813,7 +2065,7 @@
                 /* In dangerous area, increase slowly.
 		 * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd
 		 */
-		if (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
+		if (tp->snd_cwnd_cnt >= bictcp_cwnd(tp)) {
 			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
 				tp->snd_cwnd++;
 			tp->snd_cwnd_cnt=0;
@@ -1823,6 +2075,236 @@
 	tp->snd_cwnd_stamp = tcp_time_stamp;
 }
 
+/* This is based on the congestion detection/avoidance scheme described in
+ *    Lawrence S. Brakmo and Larry L. Peterson.
+ *    "TCP Vegas: End to end congestion avoidance on a global internet."
+ *    IEEE Journal on Selected Areas in Communication, 13(8):1465--1480,
+ *    October 1995. Available from:
+ *	ftp://ftp.cs.arizona.edu/xkernel/Papers/jsac.ps
+ *
+ * See http://www.cs.arizona.edu/xkernel/ for their implementation.
+ * The main aspects that distinguish this implementation from the
+ * Arizona Vegas implementation are:
+ *   o We do not change the loss detection or recovery mechanisms of
+ *     Linux in any way. Linux already recovers from losses quite well,
+ *     using fine-grained timers, NewReno, and FACK.
+ *   o To avoid the performance penalty imposed by increasing cwnd
+ *     only every-other RTT during slow start, we increase during
+ *     every RTT during slow start, just like Reno.
+ *   o Largely to allow continuous cwnd growth during slow start,
+ *     we use the rate at which ACKs come back as the "actual"
+ *     rate, rather than the rate at which data is sent.
+ *   o To speed convergence to the right rate, we set the cwnd
+ *     to achieve the right ("actual") rate when we exit slow start.
+ *   o To filter out the noise caused by delayed ACKs, we use the
+ *     minimum RTT sample observed during the last RTT to calculate
+ *     the actual rate.
+ *   o When the sender re-starts from idle, it waits until it has
+ *     received ACKs for an entire flight of new data before making
+ *     a cwnd adjustment decision. The original Vegas implementation
+ *     assumed senders never went idle.
+ */
+static void vegas_cong_avoid(struct tcp_opt *tp, u32 ack, u32 seq_rtt)
+{
+	/* The key players are v_beg_snd_una and v_beg_snd_nxt.
+	 *
+	 * These are so named because they represent the approximate values
+	 * of snd_una and snd_nxt at the beginning of the current RTT. More
+	 * precisely, they represent the amount of data sent during the RTT.
+	 * At the end of the RTT, when we receive an ACK for v_beg_snd_nxt,
+	 * we will calculate that (v_beg_snd_nxt - v_beg_snd_una) outstanding
+	 * bytes of data have been ACKed during the course of the RTT, giving
+	 * an "actual" rate of:
+	 *
+	 *     (v_beg_snd_nxt - v_beg_snd_una) / (rtt duration)
+	 *
+	 * Unfortunately, v_beg_snd_una is not exactly equal to snd_una,
+	 * because delayed ACKs can cover more than one segment, so they
+	 * don't line up nicely with the boundaries of RTTs.
+	 *
+	 * Another unfortunate fact of life is that delayed ACKs delay the
+	 * advance of the left edge of our send window, so that the number
+	 * of bytes we send in an RTT is often less than our cwnd will allow.
+	 * So we keep track of our cwnd separately, in v_beg_snd_cwnd.
+	 */
+
+	if (after(ack, tp->vegas.beg_snd_nxt)) {
+		/* Do the Vegas once-per-RTT cwnd adjustment. */
+		u32 old_wnd, old_snd_cwnd;
+
+		
+		/* Here old_wnd is essentially the window of data that was
+		 * sent during the previous RTT, and has all
+		 * been acknowledged in the course of the RTT that ended
+		 * with the ACK we just received. Likewise, old_snd_cwnd
+		 * is the cwnd during the previous RTT.
+		 */
+		old_wnd = (tp->vegas.beg_snd_nxt - tp->vegas.beg_snd_una) /
+			tp->mss_cache;
+		old_snd_cwnd = tp->vegas.beg_snd_cwnd;
+
+		/* Save the extent of the current window so we can use this
+		 * at the end of the next RTT.
+		 */
+		tp->vegas.beg_snd_una  = tp->vegas.beg_snd_nxt;
+		tp->vegas.beg_snd_nxt  = tp->snd_nxt;
+		tp->vegas.beg_snd_cwnd = tp->snd_cwnd;
+
+		/* Take into account the current RTT sample too, to
+		 * decrease the impact of delayed acks. This double counts
+		 * this sample since we count it for the next window as well,
+		 * but that's not too awful, since we're taking the min,
+		 * rather than averaging.
+		 */
+		vegas_rtt_calc(tp, seq_rtt);
+
+		/* We do the Vegas calculations only if we got enough RTT
+		 * samples that we can be reasonably sure that we got
+		 * at least one RTT sample that wasn't from a delayed ACK.
+		 * If we only had 2 samples total,
+		 * then that means we're getting only 1 ACK per RTT, which
+		 * means they're almost certainly delayed ACKs.
+		 * If  we have 3 samples, we should be OK.
+		 */
+
+		if (tp->vegas.cntRTT <= 2) {
+			/* We don't have enough RTT samples to do the Vegas
+			 * calculation, so we'll behave like Reno.
+			 */
+			if (tp->snd_cwnd > tp->snd_ssthresh)
+				tp->snd_cwnd++;
+		} else {
+			u32 rtt, target_cwnd, diff;
+
+			/* We have enough RTT samples, so, using the Vegas
+			 * algorithm, we determine if we should increase or
+			 * decrease cwnd, and by how much.
+			 */
+
+			/* Pluck out the RTT we are using for the Vegas
+			 * calculations. This is the min RTT seen during the
+			 * last RTT. Taking the min filters out the effects
+			 * of delayed ACKs, at the cost of noticing congestion
+			 * a bit later.
+			 */
+			rtt = tp->vegas.minRTT;
+
+			/* Calculate the cwnd we should have, if we weren't
+			 * going too fast.
+			 *
+			 * This is:
+			 *     (actual rate in segments) * baseRTT
+			 * We keep it as a fixed point number with
+			 * V_PARAM_SHIFT bits to the right of the binary point.
+			 */
+			target_cwnd = ((old_wnd * tp->vegas.baseRTT)
+				       << V_PARAM_SHIFT) / rtt;
+
+			/* Calculate the difference between the window we had,
+			 * and the window we would like to have. This quantity
+			 * is the "Diff" from the Arizona Vegas papers.
+			 *
+			 * Again, this is a fixed point number with
+			 * V_PARAM_SHIFT bits to the right of the binary
+			 * point.
+			 */
+			diff = (old_wnd << V_PARAM_SHIFT) - target_cwnd;
+
+			if (tp->snd_cwnd < tp->snd_ssthresh) {
+				/* Slow start.  */
+				if (diff > sysctl_tcp_vegas_gamma) {
+					/* Going too fast. Time to slow down
+					 * and switch to congestion avoidance.
+					 */
+					tp->snd_ssthresh = 2;
+
+					/* Set cwnd to match the actual rate
+					 * exactly:
+					 *   cwnd = (actual rate) * baseRTT
+					 * Then we add 1 because the integer
+					 * truncation robs us of full link
+					 * utilization.
+					 */
+					tp->snd_cwnd = min(tp->snd_cwnd,
+							   (target_cwnd >>
+							    V_PARAM_SHIFT)+1);
+
+				}
+			} else {
+				/* Congestion avoidance. */
+				u32 next_snd_cwnd;
+
+				/* Figure out where we would like cwnd
+				 * to be.
+				 */
+				if (diff > sysctl_tcp_vegas_beta) {
+					/* The old window was too fast, so
+					 * we slow down.
+					 */
+					next_snd_cwnd = old_snd_cwnd - 1;
+				} else if (diff < sysctl_tcp_vegas_alpha) {
+					/* We don't have enough extra packets
+					 * in the network, so speed up.
+					 */
+					next_snd_cwnd = old_snd_cwnd + 1;
+				} else {
+					/* Sending just as fast as we
+					 * should be.
+					 */
+					next_snd_cwnd = old_snd_cwnd;
+				}
+
+				/* Adjust cwnd upward or downward, toward the
+				 * desired value.
+				 */
+				if (next_snd_cwnd > tp->snd_cwnd)
+					tp->snd_cwnd++;
+				else if (next_snd_cwnd < tp->snd_cwnd)
+					tp->snd_cwnd--;
+			}
+		}
+
+		/* Wipe the slate clean for the next RTT. */
+		tp->vegas.cntRTT = 0;
+		tp->vegas.minRTT = 0x7fffffff;
+	}
+
+	/* The following code is executed for every ack we receive,
+	 * except for conditions checked in should_advance_cwnd()
+	 * before the call to tcp_cong_avoid(). Mainly this means that
+	 * we only execute this code if the ack actually acked some
+	 * data.
+	 */
+
+	/* If we are in slow start, increase our cwnd in response to this ACK.
+	 * (If we are not in slow start then we are in congestion avoidance,
+	 * and adjust our congestion window only once per RTT. See the code
+	 * above.)
+	 */
+	if (tp->snd_cwnd <= tp->snd_ssthresh) 
+		tp->snd_cwnd++;
+
+	/* to keep cwnd from growing without bound */
+	tp->snd_cwnd = min_t(u32, tp->snd_cwnd, tp->snd_cwnd_clamp);
+
+	/* Make sure that we are never so timid as to reduce our cwnd below
+	 * 2 MSS.
+	 *
+	 * Going below 2 MSS would risk huge delayed ACKs from our receiver.
+	 */
+	tp->snd_cwnd = max(tp->snd_cwnd, 2U);
+
+	tp->snd_cwnd_stamp = tcp_time_stamp;
+}
+
+static inline void tcp_cong_avoid(struct tcp_opt *tp, u32 ack, u32 seq_rtt)
+{
+	if (tcp_vegas_enabled(tp))
+		vegas_cong_avoid(tp, ack, seq_rtt);
+	else
+		reno_cong_avoid(tp);
+}
+
 /* Restart timer after forward progress on connection.
  * RFC2988 recommends to restart timer to now+rto.
  */
@@ -1837,7 +2319,7 @@
 }
 
 /* Remove acknowledged frames from the retransmission queue. */
-static int tcp_clean_rtx_queue(struct sock *sk)
+static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p)
 {
 	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
 	struct sk_buff *skb;
@@ -1920,6 +2402,7 @@
 		}
 	}
 #endif
+	*seq_rtt_p = seq_rtt;
 	return acked;
 }
 
@@ -2273,6 +2756,7 @@
 	u32 ack_seq = TCP_SKB_CB(skb)->seq;
 	u32 ack = TCP_SKB_CB(skb)->ack_seq;
 	u32 prior_in_flight;
+	s32 seq_rtt;
 	int prior_packets;
 
 	/* If the ack is newer than sent or older than previous acks
@@ -2323,20 +2807,22 @@
 	prior_in_flight = tcp_packets_in_flight(tp);
 
 	/* See if we can take anything off of the retransmit queue. */
-	flag |= tcp_clean_rtx_queue(sk);
+	flag |= tcp_clean_rtx_queue(sk, &seq_rtt);
 
 	if (tp->frto_counter)
 		tcp_process_frto(sk, prior_snd_una);
 
 	if (tcp_ack_is_dubious(tp, flag)) {
 		/* Advanve CWND, if state allows this. */
-		if ((flag&FLAG_DATA_ACKED) && prior_in_flight >= tp->snd_cwnd &&
+		if ((flag&FLAG_DATA_ACKED) &&
+		    (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd) &&
 		    tcp_may_raise_cwnd(tp, flag))
-			tcp_cong_avoid(tp);
+			tcp_cong_avoid(tp, ack, seq_rtt);
 		tcp_fastretrans_alert(sk, prior_snd_una, prior_packets, flag);
 	} else {
-		if ((flag&FLAG_DATA_ACKED) && prior_in_flight >= tp->snd_cwnd)
-			tcp_cong_avoid(tp);
+		if ((flag & FLAG_DATA_ACKED) && 
+		    (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd))
+			tcp_cong_avoid(tp, ack, seq_rtt);
 	}
 
 	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag&FLAG_NOT_DUP))
@@ -2940,6 +3426,7 @@
 				tp->ucopy.len -= chunk;
 				tp->copied_seq += chunk;
 				eaten = (chunk == skb->len && !th->fin);
+				tcp_rcv_space_adjust(sk);
 			}
 			local_bh_disable();
 		}
@@ -3536,6 +4023,7 @@
 	if (!err) {
 		tp->ucopy.len -= chunk;
 		tp->copied_seq += chunk;
+		tcp_rcv_space_adjust(sk);
 	}
 
 	local_bh_disable();
@@ -3663,6 +4151,9 @@
 				    (sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED) &&
 				    tp->rcv_nxt == tp->rcv_wup)
 					tcp_store_ts_recent(tp);
+
+				tcp_rcv_rtt_measure_ts(tp, skb);
+
 				/* We know that such packets are checksummed
 				 * on entry.
 				 */
@@ -3694,6 +4185,8 @@
 					    tp->rcv_nxt == tp->rcv_wup)
 						tcp_store_ts_recent(tp);
 
+					tcp_rcv_rtt_measure_ts(tp, skb);
+
 					__skb_pull(skb, tcp_header_len);
 					tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
 					NET_INC_STATS_BH(TCPHPHitsToUser);
@@ -3713,6 +4206,8 @@
 				    tp->rcv_nxt == tp->rcv_wup)
 					tcp_store_ts_recent(tp);
 
+				tcp_rcv_rtt_measure_ts(tp, skb);
+
 				if ((int)skb->truesize > sk->forward_alloc)
 					goto step5;
 
@@ -3809,6 +4304,8 @@
 	if(th->ack)
 		tcp_ack(sk, skb, FLAG_SLOWPATH);
 
+	tcp_rcv_rtt_measure_ts(tp, skb);
+
 	/* Process urgent data. */
 	tcp_urg(sk, skb, th);
 
@@ -4079,6 +4576,7 @@
 				return 1;
 
 			tcp_init_westwood(sk);
+			init_bictcp(tp);
 
 			/* Now we have several options: In theory there is 
 			 * nothing else in the frame. KA9Q has an option to 
@@ -4102,6 +4600,7 @@
 
 	case TCP_SYN_SENT:
 		tcp_init_westwood(sk);
+		init_bictcp(tp);
 
 		queued = tcp_rcv_synsent_state_process(sk, skb, th, len);
 		if (queued >= 0)