tcp_output.c

嵌入式操作系统ECOS的网络开发包

	 * to send into a small window), then must resend.
	 */
	if (len) {
		if (len == tp->t_maxseg)
			goto send;
		if ((idle || tp->t_flags & TF_NODELAY) &&
		    len + off >= so->so_snd.sb_cc)
			goto send;
		if (tp->t_force)
			goto send;
		if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0)
			goto send;
		if (SEQ_LT(tp->snd_nxt, tp->snd_max))
			goto send;
#ifdef TCP_SACK
		if (sack_rxmit)
			goto send;
#endif
	}

	/*
	 * Compare available window to amount of window
	 * known to peer (as advertised window less
	 * next expected input).  If the difference is at least two
	 * max size segments, or at least 50% of the maximum possible
	 * window, then want to send a window update to peer.
	 */
	if (win > 0) {
		/* 
		 * "adv" is the amount we can increase the window,
		 * taking into account that we are limited by
		 * TCP_MAXWIN << tp->rcv_scale.
		 */
		long adv = lmin(win, (long)TCP_MAXWIN << tp->rcv_scale) -
			(tp->rcv_adv - tp->rcv_nxt);

		if (adv >= (long) (2 * tp->t_maxseg))
			goto send;
		if (2 * adv >= (long) so->so_rcv.sb_hiwat)
			goto send;
	}

	/*
	 * Send if we owe peer an ACK.
	 */
	if (tp->t_flags & TF_ACKNOW)
		goto send;
	if (flags & (TH_SYN|TH_RST))
		goto send;
	if (SEQ_GT(tp->snd_up, tp->snd_una))
		goto send;
	/*
	 * If our state indicates that FIN should be sent
	 * and we have not yet done so, or we're retransmitting the FIN,
	 * then we need to send.
	 */
	if (flags & TH_FIN &&
	    ((tp->t_flags & TF_SENTFIN) == 0 || tp->snd_nxt == tp->snd_una))
		goto send;
#ifdef TCP_SACK
	/*
	 * In SACK, it is possible for tcp_output to fail to send a segment 
	 * after the retransmission timer has been turned off.  Make sure
	 * that the retransmission timer is set.
	 */
	if (SEQ_GT(tp->snd_max, tp->snd_una) &&
	    tp->t_timer[TCPT_REXMT] == 0 &&
	    tp->t_timer[TCPT_PERSIST] == 0) {
		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
		return (0);
	}
#endif /* TCP_SACK */

	/*
	 * TCP window updates are not reliable, rather a polling protocol
	 * using ``persist'' packets is used to insure receipt of window
	 * updates.  The three ``states'' for the output side are:
	 *	idle			not doing retransmits or persists
	 *	persisting		to move a small or zero window
	 *	(re)transmitting	and thereby not persisting
	 *
	 * tp->t_timer[TCPT_PERSIST]
	 *	is set when we are in persist state.
	 * tp->t_force
	 *	is set when we are called to send a persist packet.
	 * tp->t_timer[TCPT_REXMT]
	 *	is set when we are retransmitting
	 * The output side is idle when both timers are zero.
	 *
	 * If send window is too small, there is data to transmit, and no
	 * retransmit or persist is pending, then go to persist state.
	 * If nothing happens soon, send when timer expires:
	 * if window is nonzero, transmit what we can,
	 * otherwise force out a byte.
	 */
	if (so->so_snd.sb_cc && tp->t_timer[TCPT_REXMT] == 0 &&
	    tp->t_timer[TCPT_PERSIST] == 0) {
		tp->t_rxtshift = 0;
		tcp_setpersist(tp);
	}

	/*
	 * No reason to send a segment, just return.
	 */
	return (0);

send:
	/*
	 * Before ESTABLISHED, force sending of initial options
	 * unless TCP set not to do any options.
	 * NOTE: we assume that the IP/TCP header plus TCP options
	 * always fit in a single mbuf, leaving room for a maximum
	 * link header, i.e.
	 *	max_linkhdr + sizeof(network header) + sizeof(struct tcphdr +
	 * 		optlen <= MHLEN
	 */
	optlen = 0;

	switch (tp->pf) {
	case 0:	/*default to PF_INET*/
#ifdef INET
	case PF_INET:
		hdrlen = sizeof(struct ip) + sizeof(struct tcphdr);
		break;
#endif /* INET */
#ifdef INET6
	case PF_INET6:
		hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
		break;
#endif /* INET6 */
	default:
		return (EPFNOSUPPORT);
	}

	if (flags & TH_SYN) {
		tp->snd_nxt = tp->iss;
		if ((tp->t_flags & TF_NOOPT) == 0) {
			u_int16_t mss;

			opt[0] = TCPOPT_MAXSEG;
			opt[1] = 4;
			mss = htons((u_int16_t) tcp_mss(tp, 0));
			bcopy((caddr_t)&mss, (caddr_t)(opt + 2), sizeof(mss));
			optlen = 4;
#ifdef TCP_SACK
			/* 
			 * If this is the first SYN of connection (not a SYN 
			 * ACK), include SACK_PERMIT_HDR option.  If this is a 
			 * SYN ACK, include SACK_PERMIT_HDR option if peer has 
			 * already done so.
			 */
			if (!tp->sack_disable && ((flags & TH_ACK) == 0 ||
			    (tp->t_flags & TF_SACK_PERMIT))) {
				*((u_int32_t *) (opt + optlen)) =
				    htonl(TCPOPT_SACK_PERMIT_HDR);
				optlen += 4;
			}
#endif
	 
			if ((tp->t_flags & TF_REQ_SCALE) &&
			    ((flags & TH_ACK) == 0 ||
			    (tp->t_flags & TF_RCVD_SCALE))) {
				*((u_int32_t *) (opt + optlen)) = htonl(
					TCPOPT_NOP << 24 |
					TCPOPT_WINDOW << 16 |
					TCPOLEN_WINDOW << 8 |
					tp->request_r_scale);
				optlen += 4;
			}
		}
	}
 
	/*
	 * Send a timestamp and echo-reply if this is a SYN and our side 
	 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side
	 * and our peer have sent timestamps in our SYN's.
	 */
	if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
	     (flags & TH_RST) == 0 &&
	    ((flags & (TH_SYN|TH_ACK)) == TH_SYN ||
	     (tp->t_flags & TF_RCVD_TSTMP))) {
		u_int32_t *lp = (u_int32_t *)(opt + optlen);
 
		/* Form timestamp option as shown in appendix A of RFC 1323. */
		*lp++ = htonl(TCPOPT_TSTAMP_HDR);
		*lp++ = htonl(tcp_now);
		*lp   = htonl(tp->ts_recent);
		optlen += TCPOLEN_TSTAMP_APPA;
	}

#ifdef TCP_SIGNATURE
	if (tp->t_flags & TF_SIGNATURE) {
		u_int8_t *bp = (u_int8_t *)(opt + optlen);

		/* Send signature option */
		*(bp++) = TCPOPT_SIGNATURE;
		*(bp++) = TCPOLEN_SIGNATURE;
		sigoff = optlen + 2;

		{
			unsigned int i;

			for (i = 0; i < 16; i++)
				*(bp++) = 0;
		}

		optlen += TCPOLEN_SIGNATURE;

		/* Pad options list to the next 32 bit boundary and 
		 * terminate it.
		 */
		*bp++ = TCPOPT_NOP;
		*bp++ = TCPOPT_EOL;
		optlen += 2;
	}
#endif /* TCP_SIGNATURE */

#ifdef TCP_SACK
	/*
	 * Send SACKs if necessary.  This should be the last option processed.
	 * Only as many SACKs are sent as are permitted by the maximum options
	 * size.  No more than three SACKs are sent.
	 */
	if (!tp->sack_disable && tp->t_state == TCPS_ESTABLISHED &&
	    (tp->t_flags & (TF_SACK_PERMIT|TF_NOOPT)) == TF_SACK_PERMIT &&
	    tp->rcv_numsacks) {
		u_int32_t *lp = (u_int32_t *)(opt + optlen);
		u_int32_t *olp = lp++;
		int count = 0;  /* actual number of SACKs inserted */
		int maxsack = (MAX_TCPOPTLEN - (optlen + 4))/TCPOLEN_SACK;

		maxsack = min(maxsack, TCP_MAX_SACK);
		for (i = 0; (i < tp->rcv_numsacks && count < maxsack); i++) {
			struct sackblk sack = tp->sackblks[i];
			if (sack.start == 0 && sack.end == 0)
				continue;
			*lp++ = htonl(sack.start);
			*lp++ = htonl(sack.end);
			count++;
		}
		*olp = htonl(TCPOPT_SACK_HDR|(TCPOLEN_SACK*count+2));
		optlen += TCPOLEN_SACK*count + 4; /* including leading NOPs */
	}
#endif /* TCP_SACK */

#ifdef DIAGNOSTIC
	if (optlen > MAX_TCPOPTLEN)
		panic("tcp_output: options too long");
#endif /* DIAGNOSTIC */

	hdrlen += optlen;
 
	/*
	 * Adjust data length if insertion of options will
	 * bump the packet length beyond the t_maxopd length.
	 */
	if (len > tp->t_maxopd - optlen) {
		len = tp->t_maxopd - optlen;
		sendalot = 1;
		flags &= ~TH_FIN;
	 }

#ifdef DIAGNOSTIC
	if (max_linkhdr + hdrlen > MCLBYTES)
		panic("tcphdr too big");
#endif

	/*
	 * Grab a header mbuf, attaching a copy of data to
	 * be transmitted, and initialize the header from
	 * the template for sends on this connection.
	 */
	if (len) {
		if (tp->t_force && len == 1)
			tcpstat.tcps_sndprobe++;
		else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) {
			tcpstat.tcps_sndrexmitpack++;
			tcpstat.tcps_sndrexmitbyte += len;
		} else {
			tcpstat.tcps_sndpack++;
			tcpstat.tcps_sndbyte += len;
		}
#ifdef notyet
		if ((m = m_copypack(so->so_snd.sb_mb, off,
		    (int)len, max_linkhdr + hdrlen)) == 0) {
			error = ENOBUFS;
			goto out;
		}
		/*
		 * m_copypack left space for our hdr; use it.
		 */
		m->m_len += hdrlen;
		m->m_data -= hdrlen;
#else
		MGETHDR(m, M_DONTWAIT, MT_HEADER);
		if (m != NULL) {
			MCLGET(m, M_DONTWAIT);
			if ((m->m_flags & M_EXT) == 0) {
				m_freem(m);
				m = NULL;
			}
		}
		if (m == NULL) {
			error = ENOBUFS;
			goto out;
		}
		m->m_data += max_linkhdr;
		m->m_len = hdrlen;
		if (len <= MCLBYTES - hdrlen - max_linkhdr) {
			m_copydata(so->so_snd.sb_mb, off, (int) len,
			    mtod(m, caddr_t) + hdrlen);
			m->m_len += len;
		} else {
			m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
			if (m->m_next == 0) {
				(void) m_free(m);
				error = ENOBUFS;
				goto out;
			}
		}
#endif
		/*
		 * If we're sending everything we've got, set PUSH.
		 * (This will keep happy those implementations which only
		 * give data to the user when a buffer fills or
		 * a PUSH comes in.)
		 */
		if (off + len == so->so_snd.sb_cc)
			flags |= TH_PUSH;
	} else {
		if (tp->t_flags & TF_ACKNOW)
			tcpstat.tcps_sndacks++;
		else if (flags & (TH_SYN|TH_FIN|TH_RST))
			tcpstat.tcps_sndctrl++;
		else if (SEQ_GT(tp->snd_up, tp->snd_una))
			tcpstat.tcps_sndurg++;
		else
			tcpstat.tcps_sndwinup++;

		MGETHDR(m, M_DONTWAIT, MT_HEADER);
		if (m != NULL) {
			MCLGET(m, M_DONTWAIT);
			if ((m->m_flags & M_EXT) == 0) {
				m_freem(m);
				m = NULL;
			}
		}
		if (m == NULL) {
			error = ENOBUFS;
			goto out;
		}
		m->m_data += max_linkhdr;
		m->m_len = hdrlen;
	}
	m->m_pkthdr.rcvif = (struct ifnet *)0;

	if (!tp->t_template)
		panic("tcp_output");
#ifdef DIAGNOSTIC
	if (tp->t_template->m_len != hdrlen - optlen)
		panic("tcp_output: template len != hdrlen - optlen");
#endif /* DIAGNOSTIC */
	bcopy(mtod(tp->t_template, caddr_t), mtod(m, caddr_t),
		tp->t_template->m_len);
	th = (struct tcphdr *)(mtod(m, caddr_t) + tp->t_template->m_len -
		sizeof(struct tcphdr));

	/*
	 * Fill in fields, remembering maximum advertised
	 * window for use in delaying messages about window sizes.
	 * If resending a FIN, be sure not to use a new sequence number.
	 */
	if ((flags & TH_FIN) && (tp->t_flags & TF_SENTFIN) && 
	    (tp->snd_nxt == tp->snd_max))
		tp->snd_nxt--;
	/*