xref: /netbsd-src/sys/netinet/tcp_output.c (revision dc306354b0b29af51801a7632f1e95265a68cd81)
1 /*	$NetBSD: tcp_output.c,v 1.46 1998/12/16 00:33:14 thorpej Exp $	*/
2 
3 /*-
4  * Copyright (c) 1997, 1998 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation
9  * Facility, NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
42  *	The Regents of the University of California.  All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. All advertising materials mentioning features or use of this software
53  *    must display the following acknowledgement:
54  *	This product includes software developed by the University of
55  *	California, Berkeley and its contributors.
56  * 4. Neither the name of the University nor the names of its contributors
57  *    may be used to endorse or promote products derived from this software
58  *    without specific prior written permission.
59  *
60  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
61  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
62  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
63  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
64  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
65  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
66  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
67  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
68  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
69  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70  * SUCH DAMAGE.
71  *
72  *	@(#)tcp_output.c	8.4 (Berkeley) 5/24/95
73  */
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 
84 #include <net/if.h>
85 #include <net/route.h>
86 
87 #include <netinet/in.h>
88 #include <netinet/in_systm.h>
89 #include <netinet/ip.h>
90 #include <netinet/in_pcb.h>
91 #include <netinet/ip_var.h>
92 #include <netinet/tcp.h>
93 #define	TCPOUTFLAGS
94 #include <netinet/tcp_fsm.h>
95 #include <netinet/tcp_seq.h>
96 #include <netinet/tcp_timer.h>
97 #include <netinet/tcp_var.h>
98 #include <netinet/tcpip.h>
99 #include <netinet/tcp_debug.h>
100 
101 #ifdef notyet
102 extern struct mbuf *m_copypack();
103 #endif
104 
105 #define MAX_TCPOPTLEN	32	/* max # bytes that go in options */
106 
107 /*
108  * Knob to enable Congestion Window Monitoring, and control the
109  * the burst size it allows.  Default burst is 4 packets, per
110  * the Internet draft.
111  */
112 int	tcp_cwm = 0;
113 int	tcp_cwm_burstsize = 4;
114 
115 static __inline void tcp_segsize __P((struct tcpcb *, int *, int *));
116 static __inline void
117 tcp_segsize(tp, txsegsizep, rxsegsizep)
118 	struct tcpcb *tp;
119 	int *txsegsizep, *rxsegsizep;
120 {
121 	struct inpcb *inp = tp->t_inpcb;
122 	struct rtentry *rt;
123 	struct ifnet *ifp;
124 	int size;
125 
126 	if ((rt = in_pcbrtentry(inp)) == NULL) {
127 		size = tcp_mssdflt;
128 		goto out;
129 	}
130 
131 	ifp = rt->rt_ifp;
132 
133 	if (rt->rt_rmx.rmx_mtu != 0)
134 		size = rt->rt_rmx.rmx_mtu - sizeof(struct tcpiphdr);
135 	else if (ip_mtudisc || in_localaddr(inp->inp_faddr) ||
136 		 ifp->if_flags & IFF_LOOPBACK)
137 		size = ifp->if_mtu - sizeof(struct tcpiphdr);
138 	else
139 		size = tcp_mssdflt;
140 	size -= (tcp_optlen(tp) + ip_optlen(tp->t_inpcb));
141 
142  out:
143 	*txsegsizep = min(tp->t_peermss, size);
144 	*rxsegsizep = min(tp->t_ourmss, size);
145 
146 	if (*txsegsizep != tp->t_segsz) {
147 		/*
148 		 * If the new segment size is larger, we don't want to
149 		 * mess up the congestion window, but if it is smaller
150 		 * we'll have to reduce the congestion window to ensure
151 		 * that we don't get into trouble with initial windows
152 		 * and the rest.  In any case, if the segment size
153 		 * has changed, chances are the path has, too, and
154 		 * our congestion window will be different.
155 		 */
156 		if (*txsegsizep < tp->t_segsz) {
157 			tp->snd_cwnd = max((tp->snd_cwnd / tp->t_segsz)
158 					   * *txsegsizep, *txsegsizep);
159 			tp->snd_ssthresh = max((tp->snd_ssthresh / tp->t_segsz)
160 						* *txsegsizep, *txsegsizep);
161 		}
162 		tp->t_segsz = *txsegsizep;
163 	}
164 }
165 
166 /*
167  * Tcp output routine: figure out what should be sent and send it.
168  */
169 int
170 tcp_output(tp)
171 	register struct tcpcb *tp;
172 {
173 	register struct socket *so = tp->t_inpcb->inp_socket;
174 	register long len, win;
175 	int off, flags, error;
176 	register struct mbuf *m;
177 	register struct tcpiphdr *ti;
178 	u_char opt[MAX_TCPOPTLEN];
179 	unsigned optlen, hdrlen;
180 	int idle, sendalot, txsegsize, rxsegsize;
181 	int maxburst = TCP_MAXBURST;
182 
183 	tcp_segsize(tp, &txsegsize, &rxsegsize);
184 
185 	idle = (tp->snd_max == tp->snd_una);
186 
187 	/*
188 	 * Restart Window computation.  From draft-floyd-incr-init-win-03:
189 	 *
190 	 *	Optionally, a TCP MAY set the restart window to the
191 	 *	minimum of the value used for the initial window and
192 	 *	the current value of cwnd (in other words, using a
193 	 *	larger value for the restart window should never increase
194 	 *	the size of cwnd).
195 	 */
196 	if (tcp_cwm) {
197 		/*
198 		 * Hughes/Touch/Heidemann Congestion Window Monitoring.
199 		 * Count the number of packets currently pending
200 		 * acknowledgement, and limit our congestion window
201 		 * to a pre-determined allowed burst size plus that count.
202 		 * This prevents bursting once all pending packets have
203 		 * been acknowledged (i.e. transmission is idle).
204 		 *
205 		 * XXX Link this to Initial Window?
206 		 */
207 		tp->snd_cwnd = min(tp->snd_cwnd,
208 		    (tcp_cwm_burstsize * txsegsize) +
209 		    (tp->snd_nxt - tp->snd_una));
210 	} else {
211 		if (idle && tp->t_idle >= tp->t_rxtcur) {
212 			/*
213 			 * We have been idle for "a while" and no acks are
214 			 * expected to clock out any data we send --
215 			 * slow start to get ack "clock" running again.
216 			 */
217 			tp->snd_cwnd = min(tp->snd_cwnd,
218 			    TCP_INITIAL_WINDOW(tcp_init_win, txsegsize));
219 		}
220 	}
221 
222 again:
223 	/*
224 	 * Determine length of data that should be transmitted, and
225 	 * flags that should be used.  If there is some data or critical
226 	 * controls (SYN, RST) to send, then transmit; otherwise,
227 	 * investigate further.
228 	 */
229 	sendalot = 0;
230 	off = tp->snd_nxt - tp->snd_una;
231 	win = min(tp->snd_wnd, tp->snd_cwnd);
232 
233 	flags = tcp_outflags[tp->t_state];
234 	/*
235 	 * If in persist timeout with window of 0, send 1 byte.
236 	 * Otherwise, if window is small but nonzero
237 	 * and timer expired, we will send what we can
238 	 * and go to transmit state.
239 	 */
240 	if (tp->t_force) {
241 		if (win == 0) {
242 			/*
243 			 * If we still have some data to send, then
244 			 * clear the FIN bit.  Usually this would
245 			 * happen below when it realizes that we
246 			 * aren't sending all the data.  However,
247 			 * if we have exactly 1 byte of unset data,
248 			 * then it won't clear the FIN bit below,
249 			 * and if we are in persist state, we wind
250 			 * up sending the packet without recording
251 			 * that we sent the FIN bit.
252 			 *
253 			 * We can't just blindly clear the FIN bit,
254 			 * because if we don't have any more data
255 			 * to send then the probe will be the FIN
256 			 * itself.
257 			 */
258 			if (off < so->so_snd.sb_cc)
259 				flags &= ~TH_FIN;
260 			win = 1;
261 		} else {
262 			TCP_TIMER_DISARM(tp, TCPT_PERSIST);
263 			tp->t_rxtshift = 0;
264 		}
265 	}
266 
267 	if (win < so->so_snd.sb_cc) {
268 		len = win - off;
269 		flags &= ~TH_FIN;
270 	} else
271 		len = so->so_snd.sb_cc - off;
272 
273 	if (len < 0) {
274 		/*
275 		 * If FIN has been sent but not acked,
276 		 * but we haven't been called to retransmit,
277 		 * len will be -1.  Otherwise, window shrank
278 		 * after we sent into it.  If window shrank to 0,
279 		 * cancel pending retransmit, pull snd_nxt back
280 		 * to (closed) window, and set the persist timer
281 		 * if it isn't already going.  If the window didn't
282 		 * close completely, just wait for an ACK.
283 		 *
284 		 * If we have a pending FIN, either it has already been
285 		 * transmitted or it is outside the window, so drop it.
286 		 * If the FIN has been transmitted, but this is not a
287 		 * retransmission, then len must be -1.  Therefore we also
288 		 * prevent here the sending of `gratuitous FINs'.  This
289 		 * eliminates the need to check for that case below (e.g.
290 		 * to back up snd_nxt before the FIN so that the sequence
291 		 * number is correct).
292 		 */
293 		len = 0;
294 		flags &= ~TH_FIN;
295 		if (win == 0) {
296 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
297 			tp->t_rxtshift = 0;
298 			tp->snd_nxt = tp->snd_una;
299 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0)
300 				tcp_setpersist(tp);
301 		}
302 	}
303 	if (len > txsegsize) {
304 		len = txsegsize;
305 		flags &= ~TH_FIN;
306 		sendalot = 1;
307 	}
308 
309 	win = sbspace(&so->so_rcv);
310 
311 	/*
312 	 * Sender silly window avoidance.  If connection is idle
313 	 * and can send all data, a maximum segment,
314 	 * at least a maximum default-size segment do it,
315 	 * or are forced, do it; otherwise don't bother.
316 	 * If peer's buffer is tiny, then send
317 	 * when window is at least half open.
318 	 * If retransmitting (possibly after persist timer forced us
319 	 * to send into a small window), then must resend.
320 	 */
321 	if (len) {
322 		if (len == txsegsize)
323 			goto send;
324 		if ((so->so_state & SS_MORETOCOME) == 0 &&
325 		    ((idle || tp->t_flags & TF_NODELAY) &&
326 		     len + off >= so->so_snd.sb_cc))
327 			goto send;
328 		if (tp->t_force)
329 			goto send;
330 		if (len >= tp->max_sndwnd / 2)
331 			goto send;
332 		if (SEQ_LT(tp->snd_nxt, tp->snd_max))
333 			goto send;
334 	}
335 
336 	/*
337 	 * Compare available window to amount of window known to peer
338 	 * (as advertised window less next expected input).  If the
339 	 * difference is at least twice the size of the largest segment
340 	 * we expect to receive (i.e. two segments) or at least 50% of
341 	 * the maximum possible window, then want to send a window update
342 	 * to peer.
343 	 */
344 	if (win > 0) {
345 		/*
346 		 * "adv" is the amount we can increase the window,
347 		 * taking into account that we are limited by
348 		 * TCP_MAXWIN << tp->rcv_scale.
349 		 */
350 		long adv = min(win, (long)TCP_MAXWIN << tp->rcv_scale) -
351 			(tp->rcv_adv - tp->rcv_nxt);
352 
353 		if (adv >= (long) (2 * rxsegsize))
354 			goto send;
355 		if (2 * adv >= (long) so->so_rcv.sb_hiwat)
356 			goto send;
357 	}
358 
359 	/*
360 	 * Send if we owe peer an ACK.
361 	 */
362 	if (tp->t_flags & TF_ACKNOW)
363 		goto send;
364 	if (flags & (TH_SYN|TH_FIN|TH_RST))
365 		goto send;
366 	if (SEQ_GT(tp->snd_up, tp->snd_una))
367 		goto send;
368 
369 	/*
370 	 * TCP window updates are not reliable, rather a polling protocol
371 	 * using ``persist'' packets is used to insure receipt of window
372 	 * updates.  The three ``states'' for the output side are:
373 	 *	idle			not doing retransmits or persists
374 	 *	persisting		to move a small or zero window
375 	 *	(re)transmitting	and thereby not persisting
376 	 *
377 	 * tp->t_timer[TCPT_PERSIST]
378 	 *	is set when we are in persist state.
379 	 * tp->t_force
380 	 *	is set when we are called to send a persist packet.
381 	 * tp->t_timer[TCPT_REXMT]
382 	 *	is set when we are retransmitting
383 	 * The output side is idle when both timers are zero.
384 	 *
385 	 * If send window is too small, there is data to transmit, and no
386 	 * retransmit or persist is pending, then go to persist state.
387 	 * If nothing happens soon, send when timer expires:
388 	 * if window is nonzero, transmit what we can,
389 	 * otherwise force out a byte.
390 	 */
391 	if (so->so_snd.sb_cc && TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
392 	    TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
393 		tp->t_rxtshift = 0;
394 		tcp_setpersist(tp);
395 	}
396 
397 	/*
398 	 * No reason to send a segment, just return.
399 	 */
400 	return (0);
401 
402 send:
403 	/*
404 	 * Before ESTABLISHED, force sending of initial options
405 	 * unless TCP set not to do any options.
406 	 * NOTE: we assume that the IP/TCP header plus TCP options
407 	 * always fit in a single mbuf, leaving room for a maximum
408 	 * link header, i.e.
409 	 *	max_linkhdr + sizeof (struct tcpiphdr) + optlen <= MHLEN
410 	 */
411 	optlen = 0;
412 	hdrlen = sizeof (struct tcpiphdr);
413 	if (flags & TH_SYN) {
414 		struct rtentry *rt = in_pcbrtentry(tp->t_inpcb);
415 
416 		tp->snd_nxt = tp->iss;
417 		tp->t_ourmss = tcp_mss_to_advertise(rt != NULL ?
418 						    rt->rt_ifp : NULL);
419 		if ((tp->t_flags & TF_NOOPT) == 0) {
420 			opt[0] = TCPOPT_MAXSEG;
421 			opt[1] = 4;
422 			opt[2] = (tp->t_ourmss >> 8) & 0xff;
423 			opt[3] = tp->t_ourmss & 0xff;
424 			optlen = 4;
425 
426 			if ((tp->t_flags & TF_REQ_SCALE) &&
427 			    ((flags & TH_ACK) == 0 ||
428 			    (tp->t_flags & TF_RCVD_SCALE))) {
429 				*((u_int32_t *) (opt + optlen)) = htonl(
430 					TCPOPT_NOP << 24 |
431 					TCPOPT_WINDOW << 16 |
432 					TCPOLEN_WINDOW << 8 |
433 					tp->request_r_scale);
434 				optlen += 4;
435 			}
436 		}
437  	}
438 
439  	/*
440 	 * Send a timestamp and echo-reply if this is a SYN and our side
441 	 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side
442 	 * and our peer have sent timestamps in our SYN's.
443  	 */
444  	if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
445  	     (flags & TH_RST) == 0 &&
446  	    ((flags & (TH_SYN|TH_ACK)) == TH_SYN ||
447 	     (tp->t_flags & TF_RCVD_TSTMP))) {
448 		u_int32_t *lp = (u_int32_t *)(opt + optlen);
449 
450  		/* Form timestamp option as shown in appendix A of RFC 1323. */
451  		*lp++ = htonl(TCPOPT_TSTAMP_HDR);
452  		*lp++ = htonl(tcp_now);
453  		*lp   = htonl(tp->ts_recent);
454  		optlen += TCPOLEN_TSTAMP_APPA;
455  	}
456 
457  	hdrlen += optlen;
458 
459 #ifdef DIAGNOSTIC
460 	if (len > txsegsize)
461 		panic("tcp data to be sent is larger than segment");
462  	if (max_linkhdr + hdrlen > MHLEN)
463 		panic("tcphdr too big");
464 #endif
465 
466 	/*
467 	 * Grab a header mbuf, attaching a copy of data to
468 	 * be transmitted, and initialize the header from
469 	 * the template for sends on this connection.
470 	 */
471 	if (len) {
472 		if (tp->t_force && len == 1)
473 			tcpstat.tcps_sndprobe++;
474 		else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) {
475 			tcpstat.tcps_sndrexmitpack++;
476 			tcpstat.tcps_sndrexmitbyte += len;
477 		} else {
478 			tcpstat.tcps_sndpack++;
479 			tcpstat.tcps_sndbyte += len;
480 		}
481 #ifdef notyet
482 		if ((m = m_copypack(so->so_snd.sb_mb, off,
483 		    (int)len, max_linkhdr + hdrlen)) == 0) {
484 			error = ENOBUFS;
485 			goto out;
486 		}
487 		/*
488 		 * m_copypack left space for our hdr; use it.
489 		 */
490 		m->m_len += hdrlen;
491 		m->m_data -= hdrlen;
492 #else
493 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
494 		if (m == NULL) {
495 			error = ENOBUFS;
496 			goto out;
497 		}
498 		m->m_data += max_linkhdr;
499 		m->m_len = hdrlen;
500 		if (len <= MHLEN - hdrlen - max_linkhdr) {
501 			m_copydata(so->so_snd.sb_mb, off, (int) len,
502 			    mtod(m, caddr_t) + hdrlen);
503 			m->m_len += len;
504 		} else {
505 			m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
506 			if (m->m_next == 0) {
507 				m_freem(m);
508 				error = ENOBUFS;
509 				goto out;
510 			}
511 		}
512 #endif
513 		/*
514 		 * If we're sending everything we've got, set PUSH.
515 		 * (This will keep happy those implementations which only
516 		 * give data to the user when a buffer fills or
517 		 * a PUSH comes in.)
518 		 */
519 		if (off + len == so->so_snd.sb_cc)
520 			flags |= TH_PUSH;
521 	} else {
522 		if (tp->t_flags & TF_ACKNOW)
523 			tcpstat.tcps_sndacks++;
524 		else if (flags & (TH_SYN|TH_FIN|TH_RST))
525 			tcpstat.tcps_sndctrl++;
526 		else if (SEQ_GT(tp->snd_up, tp->snd_una))
527 			tcpstat.tcps_sndurg++;
528 		else
529 			tcpstat.tcps_sndwinup++;
530 
531 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
532 		if (m == NULL) {
533 			error = ENOBUFS;
534 			goto out;
535 		}
536 		m->m_data += max_linkhdr;
537 		m->m_len = hdrlen;
538 	}
539 	m->m_pkthdr.rcvif = (struct ifnet *)0;
540 	ti = mtod(m, struct tcpiphdr *);
541 	if (tp->t_template == 0)
542 		panic("tcp_output");
543 	bcopy((caddr_t)tp->t_template, (caddr_t)ti, sizeof (struct tcpiphdr));
544 
545 	/*
546 	 * If we are doing retransmissions, then snd_nxt will
547 	 * not reflect the first unsent octet.  For ACK only
548 	 * packets, we do not want the sequence number of the
549 	 * retransmitted packet, we want the sequence number
550 	 * of the next unsent octet.  So, if there is no data
551 	 * (and no SYN or FIN), use snd_max instead of snd_nxt
552 	 * when filling in ti_seq.  But if we are in persist
553 	 * state, snd_max might reflect one byte beyond the
554 	 * right edge of the window, so use snd_nxt in that
555 	 * case, since we know we aren't doing a retransmission.
556 	 * (retransmit and persist are mutually exclusive...)
557 	 */
558 	if (len || (flags & (TH_SYN|TH_FIN)) ||
559 	    TCP_TIMER_ISARMED(tp, TCPT_PERSIST))
560 		ti->ti_seq = htonl(tp->snd_nxt);
561 	else
562 		ti->ti_seq = htonl(tp->snd_max);
563 	ti->ti_ack = htonl(tp->rcv_nxt);
564 	if (optlen) {
565 		bcopy((caddr_t)opt, (caddr_t)(ti + 1), optlen);
566 		ti->ti_off = (sizeof (struct tcphdr) + optlen) >> 2;
567 	}
568 	ti->ti_flags = flags;
569 	/*
570 	 * Calculate receive window.  Don't shrink window,
571 	 * but avoid silly window syndrome.
572 	 */
573 	if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)rxsegsize)
574 		win = 0;
575 	if (win > (long)TCP_MAXWIN << tp->rcv_scale)
576 		win = (long)TCP_MAXWIN << tp->rcv_scale;
577 	if (win < (long)(tp->rcv_adv - tp->rcv_nxt))
578 		win = (long)(tp->rcv_adv - tp->rcv_nxt);
579 	ti->ti_win = htons((u_int16_t) (win>>tp->rcv_scale));
580 	if (SEQ_GT(tp->snd_up, tp->snd_nxt)) {
581 		u_int32_t urp = tp->snd_up - tp->snd_nxt;
582 		if (urp > IP_MAXPACKET)
583 			urp = IP_MAXPACKET;
584 		ti->ti_urp = htons((u_int16_t)urp);
585 		ti->ti_flags |= TH_URG;
586 	} else
587 		/*
588 		 * If no urgent pointer to send, then we pull
589 		 * the urgent pointer to the left edge of the send window
590 		 * so that it doesn't drift into the send window on sequence
591 		 * number wraparound.
592 		 */
593 		tp->snd_up = tp->snd_una;		/* drag it along */
594 
595 	/*
596 	 * Put TCP length in extended header, and then
597 	 * checksum extended header and data.
598 	 */
599 	if (len + optlen)
600 		ti->ti_len = htons((u_int16_t)(sizeof (struct tcphdr) +
601 		    optlen + len));
602 	ti->ti_sum = in_cksum(m, (int)(hdrlen + len));
603 
604 	/*
605 	 * In transmit state, time the transmission and arrange for
606 	 * the retransmit.  In persist state, just set snd_max.
607 	 */
608 	if (tp->t_force == 0 || TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
609 		tcp_seq startseq = tp->snd_nxt;
610 
611 		/*
612 		 * Advance snd_nxt over sequence space of this segment.
613 		 * There are no states in which we send both a SYN and a FIN,
614 		 * so we collapse the tests for these flags.
615 		 */
616 		if (flags & (TH_SYN|TH_FIN))
617 			tp->snd_nxt++;
618 		tp->snd_nxt += len;
619 		if (SEQ_GT(tp->snd_nxt, tp->snd_max)) {
620 			tp->snd_max = tp->snd_nxt;
621 			/*
622 			 * Time this transmission if not a retransmission and
623 			 * not currently timing anything.
624 			 */
625 			if (tp->t_rtt == 0) {
626 				tp->t_rtt = 1;
627 				tp->t_rtseq = startseq;
628 				tcpstat.tcps_segstimed++;
629 			}
630 		}
631 
632 		/*
633 		 * Set retransmit timer if not currently set,
634 		 * and not doing an ack or a keep-alive probe.
635 		 * Initial value for retransmit timer is smoothed
636 		 * round-trip time + 2 * round-trip time variance.
637 		 * Initialize shift counter which is used for backoff
638 		 * of retransmit time.
639 		 */
640 		if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
641 		    tp->snd_nxt != tp->snd_una) {
642 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
643 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) {
644 				TCP_TIMER_DISARM(tp, TCPT_PERSIST);
645 				tp->t_rxtshift = 0;
646 			}
647 		}
648 	} else
649 		if (SEQ_GT(tp->snd_nxt + len, tp->snd_max))
650 			tp->snd_max = tp->snd_nxt + len;
651 
652 	/*
653 	 * Trace.
654 	 */
655 	if (so->so_options & SO_DEBUG)
656 		tcp_trace(TA_OUTPUT, tp->t_state, tp, ti, 0);
657 
658 	/*
659 	 * Fill in IP length and desired time to live and
660 	 * send to IP level.  There should be a better way
661 	 * to handle ttl and tos; we could keep them in
662 	 * the template, but need a way to checksum without them.
663 	 */
664 	m->m_pkthdr.len = hdrlen + len;
665     {
666 	struct rtentry *rt;
667 
668 	((struct ip *)ti)->ip_len = m->m_pkthdr.len;
669 	((struct ip *)ti)->ip_ttl = tp->t_inpcb->inp_ip.ip_ttl;	/* XXX */
670 	((struct ip *)ti)->ip_tos = tp->t_inpcb->inp_ip.ip_tos;	/* XXX */
671 
672 	if (ip_mtudisc && (rt = in_pcbrtentry(tp->t_inpcb)) != 0 &&
673 	    (rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
674 		((struct ip *)ti)->ip_off |= IP_DF;
675 
676 #if BSD >= 43
677 	error = ip_output(m, tp->t_inpcb->inp_options, &tp->t_inpcb->inp_route,
678 	    so->so_options & SO_DONTROUTE, 0);
679 #else
680 	error = ip_output(m, (struct mbuf *)0, &tp->t_inpcb->inp_route,
681 	    so->so_options & SO_DONTROUTE);
682 #endif
683     }
684 	if (error) {
685 out:
686 		if (error == ENOBUFS) {
687 			tcp_quench(tp->t_inpcb, 0);
688 			return (0);
689 		}
690 		if ((error == EHOSTUNREACH || error == ENETDOWN)
691 		    && TCPS_HAVERCVDSYN(tp->t_state)) {
692 			tp->t_softerror = error;
693 			return (0);
694 		}
695 		return (error);
696 	}
697 	tcpstat.tcps_sndtotal++;
698 	if (tp->t_flags & TF_DELACK)
699 		tcpstat.tcps_delack++;
700 
701 	/*
702 	 * Data sent (as far as we can tell).
703 	 * If this advertises a larger window than any other segment,
704 	 * then remember the size of the advertised window.
705 	 * Any pending ACK has now been sent.
706 	 */
707 	if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv))
708 		tp->rcv_adv = tp->rcv_nxt + win;
709 	tp->last_ack_sent = tp->rcv_nxt;
710 	tp->t_flags &= ~TF_ACKNOW;
711 	TCP_CLEAR_DELACK(tp);
712 #ifdef DIAGNOSTIC
713 	if (maxburst < 0)
714 		printf("tcp_output: maxburst exceeded by %d\n", -maxburst);
715 #endif
716 	if (sendalot && (!tcp_do_newreno || --maxburst))
717 		goto again;
718 	return (0);
719 }
720 
721 void
722 tcp_setpersist(tp)
723 	register struct tcpcb *tp;
724 {
725 	register int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2);
726 	int nticks;
727 
728 	if (TCP_TIMER_ISARMED(tp, TCPT_REXMT))
729 		panic("tcp_output REXMT");
730 	/*
731 	 * Start/restart persistance timer.
732 	 */
733 	if (t < tp->t_rttmin)
734 		t = tp->t_rttmin;
735 	TCPT_RANGESET(nticks, t * tcp_backoff[tp->t_rxtshift],
736 	    TCPTV_PERSMIN, TCPTV_PERSMAX);
737 	TCP_TIMER_ARM(tp, TCPT_PERSIST, nticks);
738 	if (tp->t_rxtshift < TCP_MAXRXTSHIFT)
739 		tp->t_rxtshift++;
740 }
741