tcp_input.c

Linux内核源代码 为压缩文件 是<<Linux内核>>一书中的源代码

		/* Happy end! We did not retransmit anything
		 * or our original transmission succeeded.
		 */
		DBGUNDO(sk, tp, tp->ca_state == TCP_CA_Loss ? "loss" : "retrans");
		tcp_undo_cwr(tp, 1);
		if (tp->ca_state == TCP_CA_Loss)
			NET_INC_STATS_BH(TCPLossUndo);
		else
			NET_INC_STATS_BH(TCPFullUndo);
		tp->undo_marker = 0;
	}
	if (tp->snd_una == tp->high_seq && IsReno(tp)) {
		/* Hold old state until something *above* high_seq
		 * is ACKed. For Reno it is MUST to prevent false
		 * fast retransmits (RFC2582). SACK TCP is safe. */
		tcp_moderate_cwnd(tp);
		return 1;
	}
	tp->ca_state = TCP_CA_Open;
	return 0;
}

/* Try to undo cwnd reduction, because D-SACKs acked all retransmitted data */
static void tcp_try_undo_dsack(struct sock *sk, struct tcp_opt *tp)
{
	if (tp->undo_marker && !tp->undo_retrans) {
		DBGUNDO(sk, tp, "D-SACK");
		tcp_undo_cwr(tp, 1);
		tp->undo_marker = 0;
		NET_INC_STATS_BH(TCPDSACKUndo);
	}
}

/* Undo during fast recovery after partial ACK. */

static int tcp_try_undo_partial(struct sock *sk, struct tcp_opt *tp, int acked)
{
	/* Partial ACK arrived. Force Hoe's retransmit. */
	int failed = IsReno(tp) || tp->fackets_out>tp->reordering;

	if (tcp_may_undo(tp)) {
		/* Plain luck! Hole if filled with delayed
		 * packet, rather than with a retransmit.
		 */
		if (tp->retrans_out == 0)
			tp->retrans_stamp = 0;

		tcp_update_reordering(tp, tcp_fackets_out(tp)+acked, 1);

		DBGUNDO(sk, tp, "Hoe");
		tcp_undo_cwr(tp, 0);
		NET_INC_STATS_BH(TCPPartialUndo);

		/* So... Do not make Hoe's retransmit yet.
		 * If the first packet was delayed, the rest
		 * ones are most probably delayed as well.
		 */
		failed = 0;
	}
	return failed;
}

/* Undo during loss recovery after partial ACK. */
static int tcp_try_undo_loss(struct sock *sk, struct tcp_opt *tp)
{
	if (tcp_may_undo(tp)) {
		struct sk_buff *skb;
		for_retrans_queue(skb, sk, tp) {
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST;
		}
		DBGUNDO(sk, tp, "partial loss");
		tp->lost_out = 0;
		tp->left_out = tp->sacked_out;
		tcp_undo_cwr(tp, 1);
		NET_INC_STATS_BH(TCPLossUndo);
		tp->retransmits = 0;
		tp->undo_marker = 0;
		if (!IsReno(tp))
			tp->ca_state = TCP_CA_Open;
		return 1;
	}
	return 0;
}

static __inline__ void tcp_complete_cwr(struct tcp_opt *tp)
{
	tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh);
	tp->snd_cwnd_stamp = tcp_time_stamp;
}

static void tcp_try_to_open(struct sock *sk, struct tcp_opt *tp, int flag)
{
	tp->left_out = tp->sacked_out;

	if (tp->retrans_out == 0)
		tp->retrans_stamp = 0;

	if (flag&FLAG_ECE)
		tcp_enter_cwr(tp);

	if (tp->ca_state != TCP_CA_CWR) {
		int state = TCP_CA_Open;

		if (tp->left_out ||
		    tp->retrans_out ||
		    tp->undo_marker)
			state = TCP_CA_Disorder;

		if (tp->ca_state != state) {
			tp->ca_state = state;
			tp->high_seq = tp->snd_nxt;
		}
		tcp_moderate_cwnd(tp);
	} else {
		tcp_cwnd_down(tp);
	}
}

/* Process an event, which can update packets-in-flight not trivially.
 * Main goal of this function is to calculate new estimate for left_out,
 * taking into account both packets sitting in receiver's buffer and
 * packets lost by network.
 *
 * Besides that it does CWND reduction, when packet loss is detected
 * and changes state of machine.
 *
 * It does _not_ decide what to send, it is made in function
 * tcp_xmit_retransmit_queue().
 */
static void
tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una,
		      int prior_packets, int flag)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	int is_dupack = (tp->snd_una == prior_snd_una && !(flag&FLAG_NOT_DUP));

	/* Some technical things:
	 * 1. Reno does not count dupacks (sacked_out) automatically. */
	if (!tp->packets_out)
		tp->sacked_out = 0;
        /* 2. SACK counts snd_fack in packets inaccurately. */
	if (tp->sacked_out == 0)
		tp->fackets_out = 0;

        /* Now state machine starts.
	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
	if (flag&FLAG_ECE)
		tp->prior_ssthresh = 0;

	/* B. In all the states check for reneging SACKs. */
	if (tp->sacked_out && tcp_check_sack_reneging(sk, tp))
		return;

	/* C. Process data loss notification, provided it is valid. */
	if ((flag&FLAG_DATA_LOST) &&
	    before(tp->snd_una, tp->high_seq) &&
	    tp->ca_state != TCP_CA_Open &&
	    tp->fackets_out > tp->reordering) {
		tcp_mark_head_lost(sk, tp, tp->fackets_out-tp->reordering, tp->high_seq);
		NET_INC_STATS_BH(TCPLoss);
	}

	/* D. Synchronize left_out to current state. */
	tp->left_out = tp->sacked_out + tp->lost_out;

	/* E. Check state exit conditions. State can be terminated
	 *    when high_seq is ACKed. */
	if (tp->ca_state == TCP_CA_Open) {
		BUG_TRAP(tp->retrans_out == 0);
		tp->retrans_stamp = 0;
	} else if (!before(tp->snd_una, tp->high_seq)) {
		switch (tp->ca_state) {
		case TCP_CA_Loss:
			tp->retransmits = 0;
			if (tcp_try_undo_recovery(sk, tp))
				return;
			break;

		case TCP_CA_CWR:
			/* CWR is to be held something *above* high_seq
			 * 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;
			}
			break;

		case TCP_CA_Disorder:
			tcp_try_undo_dsack(sk, tp);
			tp->undo_marker = 0;
			tp->ca_state = TCP_CA_Open;
			break;

		case TCP_CA_Recovery:
			if (IsReno(tp))
				tcp_reset_reno_sack(tp);
			if (tcp_try_undo_recovery(sk, tp))
				return;
			tcp_complete_cwr(tp);
			break;
		}
	}

	/* F. Process state. */
	switch (tp->ca_state) {
	case TCP_CA_Recovery:
		if (prior_snd_una == tp->snd_una) {
			if (IsReno(tp) && is_dupack)
				tcp_add_reno_sack(tp);
		} else {
			int acked = prior_packets - tp->packets_out;
			if (IsReno(tp))
				tcp_remove_reno_sacks(sk, tp, acked);
			is_dupack = tcp_try_undo_partial(sk, tp, acked);
		}
		break;
	case TCP_CA_Loss:
		if (flag & FLAG_ACKED)
			tcp_reset_xmit_timer(sk, TCP_TIME_RETRANS, tp->rto);
		if (!tcp_try_undo_loss(sk, tp)) {
			tcp_moderate_cwnd(tp);
			tcp_xmit_retransmit_queue(sk);
			return;
		}
		if (tp->ca_state != TCP_CA_Open)
			return;
		/* Loss is undone; fall through to processing in Open state. */
	default:
		if (IsReno(tp)) {
			if (tp->snd_una != prior_snd_una)
				tcp_reset_reno_sack(tp);
			if (is_dupack)
				tcp_add_reno_sack(tp);
		}

		if (tp->ca_state == TCP_CA_Disorder)
			tcp_try_undo_dsack(sk, tp);

		if (!tcp_time_to_recover(sk, tp)) {
			tcp_try_to_open(sk, tp, flag);
			return;
		}

		/* Otherwise enter Recovery state */

		if (IsReno(tp))
			NET_INC_STATS_BH(TCPRenoRecovery);
		else
			NET_INC_STATS_BH(TCPSackRecovery);

		tp->high_seq = tp->snd_nxt;
		tp->prior_ssthresh = 0;
		tp->undo_marker = tp->snd_una;
		tp->undo_retrans = tp->retrans_out;

		if (tp->ca_state < TCP_CA_CWR) {
			if (!(flag&FLAG_ECE))
				tp->prior_ssthresh = tcp_current_ssthresh(tp);
			tp->snd_ssthresh = tcp_recalc_ssthresh(tp);
			TCP_ECN_queue_cwr(tp);
		}

		tp->snd_cwnd_cnt = 0;
		tp->ca_state = TCP_CA_Recovery;
	}

	if (is_dupack)
		tcp_update_scoreboard(sk, tp);
	tcp_cwnd_down(tp);
	tcp_xmit_retransmit_queue(sk);
}

/* Read draft-ietf-tcplw-high-performance before mucking
 * with this code. (Superceeds RFC1323)
 */
static void tcp_ack_saw_tstamp(struct tcp_opt *tp, int flag)
{
	__u32 seq_rtt;

	/* RTTM Rule: A TSecr value received in a segment is used to
	 * update the averaged RTT measurement only if the segment
	 * acknowledges some new data, i.e., only if it advances the
	 * left edge of the send window.
	 *
	 * See draft-ietf-tcplw-high-performance-00, section 3.3.
	 * 1998/04/10 Andrey V. Savochkin <[e-mail]>
	 */
	seq_rtt = tcp_time_stamp - tp->rcv_tsecr;
	tcp_rtt_estimator(tp, seq_rtt);
	tcp_set_rto(tp);
	if (tp->backoff) {
		if (!tp->retransmits || !(flag & FLAG_RETRANS_DATA_ACKED))
			tp->backoff = 0;
		else
			tp->rto <<= tp->backoff;
	}
	tcp_bound_rto(tp);
}

static void tcp_ack_no_tstamp(struct tcp_opt *tp, u32 seq_rtt, int flag)
{
	/* We don't have a timestamp. Can only use
	 * packets that are not retransmitted to determine
	 * rtt estimates. Also, we must not reset the
	 * backoff for rto until we get a non-retransmitted
	 * packet. This allows us to deal with a situation
	 * where the network delay has increased suddenly.
	 * I.e. Karn's algorithm. (SIGCOMM '87, p5.)
	 */

	if (flag & FLAG_RETRANS_DATA_ACKED)
		return;

	tcp_rtt_estimator(tp, seq_rtt);
	tcp_set_rto(tp);
	if (tp->backoff) {
		/* To relax it? We have valid sample as soon as we are
		 * here. Why not to clear backoff?
		 */
		if (!tp->retransmits)
			tp->backoff = 0;
		else
			tp->rto <<= tp->backoff;
	}
	tcp_bound_rto(tp);
}

static __inline__ void
tcp_ack_update_rtt(struct tcp_opt *tp, int flag, s32 seq_rtt)
{
	/* Note that peer MAY send zero echo. In this case it is ignored. (rfc1323) */
	if (tp->saw_tstamp && tp->rcv_tsecr)
		tcp_ack_saw_tstamp(tp, flag);
	else if (seq_rtt >= 0)
		tcp_ack_no_tstamp(tp, seq_rtt, flag);
}

/* This is Jacobson's slow start and congestion avoidance. 
 * SIGCOMM '88, p. 328.
 */
static __inline__ void tcp_cong_avoid(struct tcp_opt *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 >= tp->snd_cwnd) {
			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
				tp->snd_cwnd++;
			tp->snd_cwnd_cnt=0;
		} else
			tp->snd_cwnd_cnt++;
        }
}

/* Restart timer after forward progress on connection.
 * RFC2988 recommends (and BSD does) to restart timer to now+rto,
 * which is certainly wrong and effectively means that
 * rto includes one more _full_ rtt.
 *
 * For details see:
 * 	ftp://ftp.inr.ac.ru:/ip-routing/README.rto
 */

static __inline__ void tcp_ack_packets_out(struct sock *sk, struct tcp_opt *tp)
{
	if (tp->packets_out==0) {
		tcp_clear_xmit_timer(sk, TCP_TIME_RETRANS);
	} else {
		struct sk_buff *skb = skb_peek(&sk->write_queue);
		__u32 when = tp->rto + tp->rttvar - (tcp_time_stamp - TCP_SKB_CB(skb)->when);

		if ((__s32)when < (__s32)tp->rttvar)
			when = tp->rttvar;
		tcp_reset_xmit_timer(sk, TCP_TIME_RETRANS, when);
	}
}

/* Remove acknowledged frames from the retransmission queue. */
static int tcp_clean_rtx_queue(struct sock *sk)
{
	struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
	struct sk_buff *skb;
	__u32 now = tcp_time_stamp;
	int acked = 0;
	__s32 seq_rtt = -1;

	while((skb=skb_peek(&sk->write_queue)) && (skb != tp->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))
			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;
		}

		if (sacked) {
			if(sacked & TCPCB_RETRANS) {
				if(sacked & TCPCB_SACKED_RETRANS)
					tp->retrans_out--;
				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--;
			if(sacked & TCPCB_LOST)
				tp->lost_out--;
			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;
		if(tp->fackets_out)
			tp->fackets_out--;
		tp->packets_out--;
		__skb_unlink(skb, skb->list);
		tcp_free_skb(sk, skb);
	}

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