tcp_input.c

	 */

	if (flag & FLAG_RETRANS_DATA_ACKED)
		return;

	tcp_rtt_estimator(tp, seq_rtt);
	tcp_set_rto(tp);
	tp->backoff = 0;
	tcp_bound_rto(tp);
}

static inline void tcp_ack_update_rtt(struct tcp_sock *tp,
				      int flag, s32 seq_rtt)
{
	/* Note that peer MAY send zero echo. In this case it is ignored. (rfc1323) */
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
		tcp_ack_saw_tstamp(tp, flag);
	else if (seq_rtt >= 0)
		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_sock *tp)
{
	/* orignal Reno behaviour */
	if (!tcp_is_bic(tp))
		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 reno_cong_avoid(struct tcp_sock *tp)
{
        if (tp->snd_cwnd <= tp->snd_ssthresh) {
                /* In "safe" area, increase. */
		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
			tp->snd_cwnd++;
	} else {
                /* In dangerous area, increase slowly.
		 * In theory this is tp->snd_cwnd += 1 / 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;
		} else
			tp->snd_cwnd_cnt++;
        }
	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_sock *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_std;
		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_sock *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.
 */

static inline void tcp_ack_packets_out(struct sock *sk, struct tcp_sock *tp)
{
	if (!tp->packets_out) {
		tcp_clear_xmit_timer(sk, TCP_TIME_RETRANS);
	} else {
		tcp_reset_xmit_timer(sk, TCP_TIME_RETRANS, tp->rto);
	}
}

/* There is one downside to this scheme.  Although we keep the
 * ACK clock ticking, adjusting packet counters and advancing
 * congestion window, we do not liberate socket send buffer
 * space.
 *
 * Mucking with skb->truesize and sk->sk_wmem_alloc et al.
 * then making a write space wakeup callback is a possible
 * future enhancement.  WARNING: it is not trivial to make.
 */
static int tcp_tso_acked(struct sock *sk, struct sk_buff *skb,
			 __u32 now, __s32 *seq_rtt)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct tcp_skb_cb *scb = TCP_SKB_CB(skb); 
	__u32 seq = tp->snd_una;
	__u32 packets_acked;
	int acked = 0;

	/* If we get here, the whole TSO packet has not been
	 * acked.
	 */
	BUG_ON(!after(scb->end_seq, seq));

	packets_acked = tcp_skb_pcount(skb);
	if (tcp_trim_head(sk, skb, seq - scb->seq))
		return 0;
	packets_acked -= tcp_skb_pcount(skb);

	if (packets_acked) {
		__u8 sacked = scb->sacked;

		acked |= FLAG_DATA_ACKED;
		if (sacked) {
			if (sacked & TCPCB_RETRANS) {
				if (sacked & TCPCB_SACKED_RETRANS)
					tp->retrans_out -= packets_acked;
				acked |= FLAG_RETRANS_DATA_ACKED;
				*seq_rtt = -1;
			} else if (*seq_rtt < 0)
				*seq_rtt = now - scb->when;
			if (sacked & TCPCB_SACKED_ACKED)
				tp->sacked_out -= packets_acked;
			if (sacked & TCPCB_LOST)
				tp->lost_out -= packets_acked;
			if (sacked & TCPCB_URG) {
				if (tp->urg_mode &&
				    !before(seq, tp->snd_up))
					tp->urg_mode = 0;
			}
		} else if (*seq_rtt < 0)
			*seq_rtt = now - scb->when;

		if (tp->fackets_out) {
			__u32 dval = min(tp->fackets_out, packets_acked);
			tp->fackets_out -= dval;
		}
		tp->packets_out -= packets_acked;

		BUG_ON(tcp_skb_pcount(skb) == 0);
		BUG_ON(!before(scb->seq, scb->end_seq));
	}

	return acked;
}


/* Remove acknowledged frames from the retransmission queue. */
static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
	__u32 now = tcp_time_stamp;
	int acked = 0;
	__s32 seq_rtt = -1;

	while ((skb = skb_peek(&sk->sk_write_queue)) &&
	       skb != sk->sk_send_head) {
		struct tcp_skb_cb *scb = TCP_SKB_CB(skb); 
		__u8 sacked = scb->sacked;

		/* If our packet is before the ack sequence we can
		 * discard it as it's confirmed to have arrived at
		 * the other end.
		 */
		if (after(scb->end_seq, tp->snd_una)) {
			if (tcp_skb_pcount(skb) > 1)
				acked |= tcp_tso_acked(sk, skb,
						       now, &seq_rtt);
			break;
		}

		/* Initial outgoing SYN's get put onto the write_queue
		 * just like anything else we transmit.  It is not
		 * true data, and if we misinform our callers that
		 * this ACK acks real data, we will erroneously exit
		 * connection startup slow start one packet too
		 * quickly.  This is severely frowned upon behavior.
		 */
		if (!(scb->flags & TCPCB_FLAG_SYN)) {
			acked |= FLAG_DATA_ACKED;
		} else {
			acked |= FLAG_SYN_ACKED;
			tp->retrans_stamp = 0;
		}

		if (sacked) {
			if (sacked & TCPCB_RETRANS) {
				if(sacked & TCPCB_SACKED_RETRANS)
					tp->retrans_out -= tcp_skb_pcount(skb);
				acked |= FLAG_RETRANS_DATA_ACKED;
				seq_rtt = -1;
			} else if (seq_rtt < 0)
				seq_rtt = now - scb->when;
			if (sacked & TCPCB_SACKED_ACKED)
				tp->sacked_out -= tcp_skb_pcount(skb);
			if (sacked & TCPCB_LOST)
				tp->lost_out -= tcp_skb_pcount(skb);
			if (sacked & TCPCB_URG) {
				if (tp->urg_mode &&
				    !before(scb->end_seq, tp->snd_up))
					tp->urg_mode = 0;
			}
		} else if (seq_rtt < 0)
			seq_rtt = now - scb->when;
		tcp_dec_pcount_approx(&tp->fackets_out, skb);
		tcp_packets_out_dec(tp, skb);
		__skb_unlink(skb, skb->list);
		sk_stream_free_skb(sk, skb);
	}

	if (acked&FLAG_ACKED) {
		tcp_ack_update_rtt(tp, acked, seq_rtt);
		tcp_ack_packets_out(sk, tp);
	}

#if FASTRETRANS_DEBUG > 0
	BUG_TRAP((int)tp->sacked_out >= 0);
	BUG_TRAP((int)tp->lost_out >= 0);
	BUG_TRAP((int)tp->retrans_out >= 0);
	if (!tp->packets_out && tp->rx_opt.sack_ok) {
		if (tp->lost_out) {
			printk(KERN_DEBUG "Leak l=%u %d\n",
			       tp->lost_out, tp->ca_state);
			tp->lost_out = 0;
		}
		if (tp->sacked_out) {
			printk(KERN_DEBUG "Leak s=%u %d\n",
			       tp->sacked_out, tp->ca_state);
			tp->sacked_out = 0;
		}
		if (tp->retrans_out) {
			printk(KERN_DEBUG "Leak r=%u %d\n",
			       tp->retrans_out, tp->ca_state);
			tp->retrans_out = 0;
		}
	}
#endif
	*seq_rtt_p = seq_rtt;
	return acked;
}

static void tcp_ack_probe(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Was it a usable window open? */

	if (!after(TCP_SKB_CB(sk->sk_send_head)->end_seq,
		   tp->snd_una + tp->snd_wnd)) {
		tp->backoff = 0;
		tcp_clear_xmit_timer(sk, TCP_TIME_PROBE0);
		/* Socket must be waked up by subsequent tcp_data_snd_check().
		 * This function is not for random using!
		 */
	} else {
		tcp_reset_xmit_timer(sk, TCP_TIME_PROBE0,
				     min(tp->rto << tp->backoff, TCP_RTO_MAX));
	}
}

static inline int tcp_ack_is_dubious(struct tcp_sock *tp, int flag)
{
	return (!(flag & FLAG_NOT_DUP) || (flag & FLAG_CA_ALERT) ||
		tp->ca_state != TCP_CA_Open);
}

static inline int tcp_may_raise_cwnd(struct tcp_sock *tp, int flag)
{
	return (!(flag & FLAG_ECE) || tp->snd_cwnd < tp->snd_ssthresh) &&
		!((1<<tp->ca_state)&(TCPF_CA_Recovery|TCPF_CA_CWR));
}

/* Check that window update is acceptable.
 * The function assumes that snd_una<=ack<=snd_next.
 */
static inline int tcp_may_update_window(struct tcp_sock *tp, u32 ack,
					u32 ack_seq, u32 nwin)
{
	return (after(ack, tp->snd_una) ||
		after(ack_seq, tp->snd_wl1) ||
		(ack_seq == tp->snd_wl1 && nwin > tp->snd_wnd));
}

/* Update our send window.
 *
 * Window update algorithm, described in RFC793/RFC1122 (used in linux-2.2
 * and in FreeBSD. NetBSD's one is even worse.) is wrong.
 */
static int tcp_ack_update_window(struct sock *sk, struct tcp_sock *tp,
				 struct sk_buff *skb, u32 ack, u32 ack_seq)
{
	int flag = 0;
	u32 nwin = ntohs(skb->h.th->window);

	if (likely(!skb->h.th->syn))
		nwin <<= tp->rx_opt.snd_wscale;

	if (tcp_may_update_window(tp, ack, ack_seq, nwin)) {
		flag |= FLAG_WIN_UPDATE;
		tcp_update_wl(tp, ack, ack_seq);

		if (tp->snd_wnd != nwin) {
			tp->snd_wnd = nwin;

			/* Note, it is the only place, where
			 * fast path is recovered for sending TCP.
			 */
			tcp_fast_path_check(sk, tp);

			if (nwin > tp->max_window) {
				tp->max_window = nwin;
				tcp_sync_mss(sk, tp->pmtu_cookie);
			}
		}
	}

	tp->snd_una = ack;

	return flag;
}

static void tcp_process_frto(struct sock *sk, u32 prior_snd_una)
{
	struct tcp_sock *tp = tcp_sk(sk);
	
	tcp_sync_left_out(tp);
	
	if (tp->snd_una == prior_snd_una ||
	    !before(tp->snd_una, tp->frto_highmark)) {
		/* RTO was caused by loss, start retransmitting in
		 * go-back-N slow start
		 */
		tcp_enter_frto_loss(sk);
		return;
	}

	if (tp->frto_counter == 1) {
		/* First ACK after RTO advances the window: allow two new
		 * segments out.
		 */
		tp->snd_cwnd = tcp_packets_in_flight(tp) + 2;
	} else {
		/* Also the second ACK after RTO advances the window.
		 * The RTO was likely spurious. Reduce cwnd and continue
		 * in congestion avoidance
		 */
		tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh);
		tcp_moderate_cwnd(tp);
	}

	/* F-RTO affects on two new ACKs following RTO.
	 * At latest on third ACK the TCP behavor is back to normal.
	 */
	tp->frto_counter = (tp->frto_counter + 1) % 3;
}

/*
 * TCP Westwood+
 */

/*
 * @init_westwood
 * This function initializes fields used in TCP Westwood+. We can't
 * get no information about RTTmin at this time so we simply set it to
 * TCP_WESTWOOD_INIT_RTT. This value was chosen to be too conservative
 * since in this way we're sure it will be updated in a consistent
 * way as soon as possible. It will reasonably happen within the first
 * RTT period of the connection lifetime.
 */

static void init_westwood(struct sock *sk)
{
        struct tcp_sock *tp = tcp_sk(sk);

        tp->westwood.bw_ns_est = 0;
        tp->westwood.bw_est = 0;
        tp->westwood.accounted = 0;
        tp->westwood.cumul_ack = 0;
        tp->westwood.rtt_win_sx = tcp_time_stamp;
        tp->westwood.rtt = TCP_WESTWOOD_INIT_RTT;
        tp->westwood.rtt_min = TCP_WESTWOOD_INIT_RTT;
        tp->westwood.snd_una = tp->snd_una;
}

/*
 * @westwood_do_filter
 * Low-pass filter. Implemented using constant coeffients.
 */

static inline __u32 westwood_do_filter(__u32 a, __u32 b)
{
	return (((7 * a) + b) >> 3);
}

static void westwood_filter(struct sock *sk, __u32 delta)
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->westwood.bw_ns_est =
		westwood_do_filter(tp->westwood.bw_ns_est, 
				   tp->westwood.bk / delta);
	tp->westwood.bw_est =
		westwood_do_filter(tp->westwood.bw_est,
				   tp->westwood.bw_ns_est);
}

/* 
 * @westwood_update_rttmin
 * It is used to update RTTmin. In this case we MUST NOT use
 * WESTWOOD_RTT_MIN minimum bound since we could be on a LAN!
 */

static inline __u32 westwood_update_rttmin(const struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	__u32 rttmin = tp->westwood.rtt_min;

	if (tp->westwood.rtt != 0 &&
	    (tp->westwood.rtt < tp->westwood.rtt_min || !rttmin))
		rttmin = tp->westwood.rtt;

	return rttmin;
}

/*
 * @westwood_acked
 * Evaluate increases for dk. 
 */

static inline __u32 westwood_acked(const struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);

	return tp->snd_una - tp->westwood.snd_una;
}

/*
 * @westwood_new_window
 * It evaluates if we are receiving data inside the same RTT window as
 * when we started.
 * Return value:
 * It returns 0 if we are still evaluating samples in the same RTT
 * window, 1 if the sample has to be considered in the next window.
 */

static int westwood_new_window(const struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	__u32 left_bound;
	__u32 rtt;
	int ret = 0;

	left_bound = tp->westwood.rtt_win_sx;
	rtt = max(tp->westwood.rtt, (u32) TCP_WESTWOOD_RTT_MIN);

	/*
	 * A RTT-window has passed. Be careful since if RTT is less than
	 * 50ms we don't filter but we continue 'building the sample'.
	 * This minimum limit was choosen since an estimation on small
	 * time intervals is better to avoid...
	 * Obvioulsy on a LAN we reasonably will always have
	 * right_bound = left_bound + WESTWOOD_RTT_MIN
         */

	if ((left_bound + rtt) < tcp_time_stamp)
		ret = 1;

	return ret;
}

/*
 * @westwood_update_window
 * It updates RTT evaluation window if it is the right moment to do
 * it. If so it calls filter for evaluating bandwidth. 
 */

static void __westwood_update_window(struct sock *sk, __u32 now)
{
	struct tcp_sock *tp = tcp_sk(sk);
	__u32 delta = now - tp->westwood.rtt_win_sx;

        if (delta) {
		if (tp->westwood.rtt)
			westwood_filter(sk, delta);

		tp->westwood.bk = 0;
		tp->westwood.rtt_win_sx = tcp_time_stamp;
	}
}


static void westwood_update_window(struct sock *sk, __u32 now)
{
	if (westwood_new_window(sk)) 
		__westwood_update_window(sk, now);
}

/*
 * @__tcp_westwood_fast_bw
 * It is called when we are in fast path. In particular it is called when
 * header prediction is successfull. In such case infact update is
 * straight forward and doesn't need any particular care.
 */

static void __tcp_westwood_fast_bw(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);

	westwood_update_window(sk, tcp_time_stamp);

	tp->westwood.bk += westwood_acked(sk);
	tp->westwood.snd_una = tp->snd_una;
	tp->westwood.rtt_min = westwood_update_rttmin(sk);
}

static inline void tcp_westwood_fast_bw(struct sock *sk, struct sk_buff *skb)
{
        if (tcp_is_westwood(tcp_sk(sk)))
                __tcp_westwood_fast_bw(sk, skb);
}


/*
 * @westwood_dupack_update
 * It updates accounted and cumul_ack when receiving a dupack.
 */

static void westwood_dupack_update(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->westwood.accounted += tp->mss_cache_std;
	tp->westwood.cumul_ack = tp->mss_cache_std;
}

static inline int westwood_may_change_cumul(struct tcp_sock *tp)
{
	return (tp->westwood.cumul_ack > tp->mss_cache_std);
}

static inline void westwood_partial_update(struct tcp_sock *tp)
{
	tp->westwood.accounted -= tp->westwood.cumul_ack;
	tp->westwood.cumul_ack = tp->mss_cache_std;
}

static inline void westwood_complete_update(struct tcp_sock *tp)
{
	tp->westwood.cumul_ack -= tp->westwood.accounted;
	tp->westwood.accounted = 0;
}

/*
 * @westwood_acked_count
 * This function evaluates cumul_ack for evaluating dk in case of
 * delayed or partial acks.
 */

static inline __u32 westwood_acked_count(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->westwood.cumul_ack = westwood_acked(sk);

        /* If cumul_ack is 0 this is a dupack since it's not moving
         * tp->snd_una.
         */
        if (!(tp->westwood.cumul_ack))
                westwood_dupack_update(sk);

        if (westwood_may_change_cumul(tp)) {
		/* Partial or delayed ack */
		if (tp->westwood.accounted >= tp->westwood.cumul_ack)
			westwood_partial_update(tp);
		else
			westwood_complete_update(tp);
	}

	tp->westwood.snd_una = tp->snd_una;

	return tp->westwood.cumul_ack;
}


/*
 * @__tcp_westwood_slow_bw
 * It is called when something is going wrong..even if there could
 * be no problems! Infact a simple delayed packet may trigger a
 * dupack. But we need to be careful in such case.
 */

static void __tcp_westwood_slow_bw(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);

	westwood_update_window(sk, tcp_time_stamp);

	tp->westwood.bk += westwood_acked_count(sk);
	tp->westwood.rtt_min = westwood_update_rttmin(sk);
}

static inline void tcp_westwood_slow_bw(struct sock *sk, struct sk_buff *skb)
{
        if (tcp_is_westwood(tcp_sk(sk)))
                __tcp_westwood_slow_bw(sk, skb);
}

/* This routine deals with incoming acks, but not outgoing ones. */
static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag)
{
	struct tcp_sock *tp = tcp_sk(sk);
	u32 prior_snd_una = tp->snd_una;
	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
	 * then we can probably ignore it.
	 */
	if (after(ack, tp->snd_nxt))
		goto uninteresting_ack;

	if (before(ack, prior_snd_una))
		goto old_ack;

	if (!(flag&FLAG_SLOWPATH) && after(ack, prior_snd_una)) {
		/* Window is constant, pure forward advance.
		 * No more checks are required.
		 * Note, we use the fact that SND.UNA>=SND.WL2.
		 */
		tcp_update_wl(tp, ack, ack_seq);
		tp->snd_una = ack;
		tcp_westwood_fast_bw(sk, skb);
		flag |= FLAG_WIN_UPDATE;

		NET_INC_STATS_BH(LINUX_MIB_TCPHPACKS);
	} else {
		if (ack_seq != TCP_SKB_CB(skb)->end_seq)
			flag |= FLAG_DATA;
		else
			NET_INC_STATS_BH(LINUX_MIB_TCPPUREACKS);

		flag |= tcp_ack_update_window(sk, tp, skb, ack, ack_seq);

		if (TCP_SKB_CB(skb)->sacked)
			flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una);

		if (TCP_ECN_rcv_ecn_echo(tp, skb->h.th))
			flag |= FLAG_ECE;

		tcp_westwood_slow_bw(sk,skb);
	}

	/* We passed data and got it acked, remove any soft error
	 * log. Something worked...
	 */
	sk->sk_err_soft = 0;
	tp->rcv_tstamp = tcp_time_stamp;
	prior_packets = tp->packets_out;
	if (!prior_packets)
		goto no_queue;

	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, &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) &&
		    (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd) &&
		    tcp_may_raise_cwnd(tp, flag))
			tcp_cong_avoid(tp, ack, seq_rtt);
		tcp_fastretrans_alert(sk, prior_snd_una, prior_packets, flag);
	} else {
		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))
		dst_confirm(sk->sk_dst_cache);

	return 1;

no_queue:
	tp->probes_out = 0;

	/* If this ack opens up a zero window, clear backoff.  It was
	 * being used to time the probes, and is probably far higher than
	 * it needs to be for normal retransmission.
	 */
	if (sk->sk_send_head)
		tcp_ack_probe(sk);
	return 1;

old_ack:
	if (TCP_SKB_CB(skb)->sacked)
		tcp_sacktag_write_queue(sk, skb, prior_snd_una);

uninteresting_ack:
	SOCK_DEBUG(sk, "Ack %u out of %u:%u\n", ack, tp->snd_una, tp->snd_nxt);
	return 0;
}


/* Look for tcp options. Normally only called on SYN and SYNACK packets.
 * But, this can also be called on packets in the established flow when
 * the fast version below fails.
 */
void tcp_parse_options(struct sk_buff *skb, struct tcp_options_received *opt_rx, int estab)
{
	unsigned char *ptr;
	struct tcphdr *th = skb->h.th;
	int length=(th->doff*4)-sizeof(struct tcphdr);

	ptr = (unsigned char *)(th + 1);
	opt_rx->saw_tstamp = 0;

	while(length>0) {
	  	int opcode=*ptr++;
		int opsize;

		switch (opcode) {
			case TCPOPT_EOL: