xref: /openbsd-src/sys/netinet/tcp_output.c (revision cb39b41371628601fbe4c618205356d538b9d08a)
1 /*	$OpenBSD: tcp_output.c,v 1.109 2015/03/14 03:38:52 jsg Exp $	*/
2 /*	$NetBSD: tcp_output.c,v 1.16 1997/06/03 16:17:09 kml Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  * 	This product includes software developed by the University of
46  * 	California, Berkeley and its contributors.
47  * 	This product includes software developed at the Information
48  * 	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include "pf.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/kernel.h>
80 
81 #include <net/route.h>
82 
83 #include <netinet/in.h>
84 #include <netinet/ip.h>
85 #include <netinet/in_pcb.h>
86 #include <netinet/ip_var.h>
87 #include <netinet/tcp.h>
88 #define	TCPOUTFLAGS
89 #include <netinet/tcp_fsm.h>
90 #include <netinet/tcp_seq.h>
91 #include <netinet/tcp_timer.h>
92 #include <netinet/tcp_var.h>
93 #include <netinet/tcpip.h>
94 #include <netinet/tcp_debug.h>
95 
96 #ifdef notyet
97 extern struct mbuf *m_copypack();
98 #endif
99 
100 #ifdef TCP_SACK
101 extern int tcprexmtthresh;
102 #endif
103 
104 #ifdef TCP_SACK
105 #ifdef TCP_SACK_DEBUG
106 void tcp_print_holes(struct tcpcb *tp);
107 
108 void
109 tcp_print_holes(struct tcpcb *tp)
110 {
111 	struct sackhole *p = tp->snd_holes;
112 	if (p == 0)
113 		return;
114 	printf("Hole report: start--end dups rxmit\n");
115 	while (p) {
116 		printf("%x--%x d %d r %x\n", p->start, p->end, p->dups,
117 		    p->rxmit);
118 		p = p->next;
119 	}
120 	printf("\n");
121 }
122 #endif /* TCP_SACK_DEBUG */
123 
124 /*
125  * Returns pointer to a sackhole if there are any pending retransmissions;
126  * NULL otherwise.
127  */
128 struct sackhole *
129 tcp_sack_output(struct tcpcb *tp)
130 {
131 	struct sackhole *p;
132 
133 	if (!tp->sack_enable)
134 		return (NULL);
135 	p = tp->snd_holes;
136 	while (p) {
137 #ifndef TCP_FACK
138 		if (p->dups >= tcprexmtthresh && SEQ_LT(p->rxmit, p->end)) {
139 #else
140 		/* In FACK, if p->dups is less than tcprexmtthresh, but
141 		 * snd_fack advances more than tcprextmtthresh * tp->t_maxseg,
142 		 * tcp_input() will try fast retransmit. This forces output.
143 		 */
144 		if ((p->dups >= tcprexmtthresh ||
145 		     tp->t_dupacks == tcprexmtthresh) &&
146 		    SEQ_LT(p->rxmit, p->end)) {
147 #endif /* TCP_FACK */
148 			if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */
149 				p = p->next;
150 				continue;
151 			}
152 #ifdef TCP_SACK_DEBUG
153 			if (p)
154 				tcp_print_holes(tp);
155 #endif
156 			return (p);
157 		}
158         	p = p->next;
159 	}
160 	return (NULL);
161 }
162 
163 /*
164  * After a timeout, the SACK list may be rebuilt.  This SACK information
165  * should be used to avoid retransmitting SACKed data.  This function
166  * traverses the SACK list to see if snd_nxt should be moved forward.
167  */
168 
169 void
170 tcp_sack_adjust(struct tcpcb *tp)
171 {
172 	struct sackhole *cur = tp->snd_holes;
173 	if (cur == NULL)
174 		return; /* No holes */
175 	if (SEQ_GEQ(tp->snd_nxt, tp->rcv_lastsack))
176 		return; /* We're already beyond any SACKed blocks */
177 	/*
178 	 * Two cases for which we want to advance snd_nxt:
179 	 * i) snd_nxt lies between end of one hole and beginning of another
180 	 * ii) snd_nxt lies between end of last hole and rcv_lastsack
181 	 */
182 	while (cur->next) {
183 		if (SEQ_LT(tp->snd_nxt, cur->end))
184 			return;
185 		if (SEQ_GEQ(tp->snd_nxt, cur->next->start))
186 			cur = cur->next;
187 		else {
188 			tp->snd_nxt = cur->next->start;
189 			return;
190 		}
191 	}
192 	if (SEQ_LT(tp->snd_nxt, cur->end))
193 		return;
194 	tp->snd_nxt = tp->rcv_lastsack;
195 	return;
196 }
197 #endif /* TCP_SACK */
198 
199 /*
200  * Tcp output routine: figure out what should be sent and send it.
201  */
202 int
203 tcp_output(struct tcpcb *tp)
204 {
205 	struct socket *so = tp->t_inpcb->inp_socket;
206 	long len, win, txmaxseg;
207 	int off, flags, error;
208 	struct mbuf *m;
209 	struct tcphdr *th;
210 	u_int32_t optbuf[howmany(MAX_TCPOPTLEN, sizeof(u_int32_t))];
211 	u_char *opt = (u_char *)optbuf;
212 	unsigned int optlen, hdrlen, packetlen;
213 	int idle, sendalot = 0;
214 #ifdef TCP_SACK
215 	int i, sack_rxmit = 0;
216 	struct sackhole *p;
217 	int maxburst = TCP_MAXBURST;
218 #endif
219 #ifdef TCP_SIGNATURE
220 	unsigned int sigoff;
221 #endif /* TCP_SIGNATURE */
222 #ifdef TCP_ECN
223 	int needect;
224 #endif
225 
226 	if (tp->t_flags & TF_BLOCKOUTPUT) {
227 		tp->t_flags |= TF_NEEDOUTPUT;
228 		return (0);
229 	} else
230 		tp->t_flags &= ~TF_NEEDOUTPUT;
231 
232 #if defined(TCP_SACK) && defined(TCP_SIGNATURE) && defined(DIAGNOSTIC)
233 	if (tp->sack_enable && (tp->t_flags & TF_SIGNATURE))
234 		return (EINVAL);
235 #endif /* defined(TCP_SACK) && defined(TCP_SIGNATURE) && defined(DIAGNOSTIC) */
236 
237 	/*
238 	 * Determine length of data that should be transmitted,
239 	 * and flags that will be used.
240 	 * If there is some data or critical controls (SYN, RST)
241 	 * to send, then transmit; otherwise, investigate further.
242 	 */
243 	idle = (tp->t_flags & TF_LASTIDLE) || (tp->snd_max == tp->snd_una);
244 	if (idle && (tcp_now - tp->t_rcvtime) >= tp->t_rxtcur)
245 		/*
246 		 * We have been idle for "a while" and no acks are
247 		 * expected to clock out any data we send --
248 		 * slow start to get ack "clock" running again.
249 		 */
250 		tp->snd_cwnd = 2 * tp->t_maxseg;
251 
252 	/* remember 'idle' for next invocation of tcp_output */
253 	if (idle && soissending(so)) {
254 		tp->t_flags |= TF_LASTIDLE;
255 		idle = 0;
256 	} else
257 		tp->t_flags &= ~TF_LASTIDLE;
258 
259 again:
260 #ifdef TCP_SACK
261 	/*
262 	 * If we've recently taken a timeout, snd_max will be greater than
263 	 * snd_nxt.  There may be SACK information that allows us to avoid
264 	 * resending already delivered data.  Adjust snd_nxt accordingly.
265 	 */
266 	if (tp->sack_enable && SEQ_LT(tp->snd_nxt, tp->snd_max))
267 		tcp_sack_adjust(tp);
268 #endif
269 	off = tp->snd_nxt - tp->snd_una;
270 #if defined(TCP_SACK) && defined(TCP_FACK)
271 	/* Normally, sendable data is limited by off < tp->snd_cwnd.
272 	 * But in FACK, sendable data is limited by snd_awnd < snd_cwnd,
273 	 * regardless of offset.
274 	 */
275 	if (tp->sack_enable && (tp->t_dupacks > tcprexmtthresh))
276 		win = tp->snd_wnd;
277 	else
278 #endif
279 	win = ulmin(tp->snd_wnd, tp->snd_cwnd);
280 
281 	flags = tcp_outflags[tp->t_state];
282 
283 #ifdef TCP_SACK
284 	/*
285 	 * Send any SACK-generated retransmissions.  If we're explicitly trying
286 	 * to send out new data (when sendalot is 1), bypass this function.
287 	 * If we retransmit in fast recovery mode, decrement snd_cwnd, since
288 	 * we're replacing a (future) new transmission with a retransmission
289 	 * now, and we previously incremented snd_cwnd in tcp_input().
290 	 */
291 	if (tp->sack_enable && !sendalot) {
292 		if (tp->t_dupacks >= tcprexmtthresh &&
293 		    (p = tcp_sack_output(tp))) {
294 			off = p->rxmit - tp->snd_una;
295 			sack_rxmit = 1;
296 			/* Coalesce holes into a single retransmission */
297 			len = min(tp->t_maxseg, p->end - p->rxmit);
298 #ifndef TCP_FACK
299 			/* in FACK, hold snd_cwnd constant during recovery */
300 			if (SEQ_LT(tp->snd_una, tp->snd_last))
301 				tp->snd_cwnd -= tp->t_maxseg;
302 #endif
303     		}
304 	}
305 #endif /* TCP_SACK */
306 
307 	sendalot = 0;
308 	/*
309 	 * If in persist timeout with window of 0, send 1 byte.
310 	 * Otherwise, if window is small but nonzero
311 	 * and timer expired, we will send what we can
312 	 * and go to transmit state.
313 	 */
314 	if (tp->t_force) {
315 		if (win == 0) {
316 			/*
317 			 * If we still have some data to send, then
318 			 * clear the FIN bit.  Usually this would
319 			 * happen below when it realizes that we
320 			 * aren't sending all the data.  However,
321 			 * if we have exactly 1 byte of unset data,
322 			 * then it won't clear the FIN bit below,
323 			 * and if we are in persist state, we wind
324 			 * up sending the packet without recording
325 			 * that we sent the FIN bit.
326 			 *
327 			 * We can't just blindly clear the FIN bit,
328 			 * because if we don't have any more data
329 			 * to send then the probe will be the FIN
330 			 * itself.
331 			 */
332 			if (off < so->so_snd.sb_cc)
333 				flags &= ~TH_FIN;
334 			win = 1;
335 		} else {
336 			TCP_TIMER_DISARM(tp, TCPT_PERSIST);
337 			tp->t_rxtshift = 0;
338 		}
339 	}
340 
341 #ifdef TCP_SACK
342 	if (!sack_rxmit) {
343 #endif
344 	len = ulmin(so->so_snd.sb_cc, win) - off;
345 
346 #if defined(TCP_SACK) && defined(TCP_FACK)
347 	/*
348 	 * If we're in fast recovery (SEQ_GT(tp->snd_last, tp->snd_una)), and
349 	 * amount of outstanding data (snd_awnd) is >= snd_cwnd, then
350 	 * do not send data (like zero window conditions)
351 	 */
352 	if (tp->sack_enable && len && SEQ_GT(tp->snd_last, tp->snd_una) &&
353 	    (tp->snd_awnd >= tp->snd_cwnd))
354 		len = 0;
355 #endif /* TCP_FACK */
356 #ifdef TCP_SACK
357 	}
358 #endif
359 
360 	if (len < 0) {
361 		/*
362 		 * If FIN has been sent but not acked,
363 		 * but we haven't been called to retransmit,
364 		 * len will be -1.  Otherwise, window shrank
365 		 * after we sent into it.  If window shrank to 0,
366 		 * cancel pending retransmit, pull snd_nxt back
367 		 * to (closed) window, and set the persist timer
368 		 * if it isn't already going.  If the window didn't
369 		 * close completely, just wait for an ACK.
370 		 */
371 		len = 0;
372 		if (win == 0) {
373 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
374 			tp->t_rxtshift = 0;
375 			tp->snd_nxt = tp->snd_una;
376 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0)
377 				tcp_setpersist(tp);
378 		}
379 	}
380 
381         /*
382          * Never send more than half a buffer full.  This insures that we can
383          * always keep 2 packets on the wire, no matter what SO_SNDBUF is, and
384          * therefore acks will never be delayed unless we run out of data to
385          * transmit.
386          */
387 	txmaxseg = ulmin(so->so_snd.sb_hiwat / 2, tp->t_maxseg);
388 
389 	if (len > txmaxseg) {
390 		len = txmaxseg;
391 		sendalot = 1;
392 	}
393 	if (off + len < so->so_snd.sb_cc)
394 		flags &= ~TH_FIN;
395 
396 	win = sbspace(&so->so_rcv);
397 
398 	/*
399 	 * Sender silly window avoidance.  If connection is idle
400 	 * and can send all data, a maximum segment,
401 	 * at least a maximum default-size segment do it,
402 	 * or are forced, do it; otherwise don't bother.
403 	 * If peer's buffer is tiny, then send
404 	 * when window is at least half open.
405 	 * If retransmitting (possibly after persist timer forced us
406 	 * to send into a small window), then must resend.
407 	 */
408 	if (len) {
409 		if (len == txmaxseg)
410 			goto send;
411 		if ((idle || (tp->t_flags & TF_NODELAY)) &&
412 		    len + off >= so->so_snd.sb_cc && !soissending(so) &&
413 		    (tp->t_flags & TF_NOPUSH) == 0)
414 			goto send;
415 		if (tp->t_force)
416 			goto send;
417 		if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0)
418 			goto send;
419 		if (SEQ_LT(tp->snd_nxt, tp->snd_max))
420 			goto send;
421 #ifdef TCP_SACK
422 		if (sack_rxmit)
423 			goto send;
424 #endif
425 	}
426 
427 	/*
428 	 * Compare available window to amount of window
429 	 * known to peer (as advertised window less
430 	 * next expected input).  If the difference is at least two
431 	 * max size segments, or at least 50% of the maximum possible
432 	 * window, then want to send a window update to peer.
433 	 */
434 	if (win > 0) {
435 		/*
436 		 * "adv" is the amount we can increase the window,
437 		 * taking into account that we are limited by
438 		 * TCP_MAXWIN << tp->rcv_scale.
439 		 */
440 		long adv = lmin(win, (long)TCP_MAXWIN << tp->rcv_scale) -
441 			(tp->rcv_adv - tp->rcv_nxt);
442 
443 		if (adv >= (long) (2 * tp->t_maxseg))
444 			goto send;
445 		if (2 * adv >= (long) so->so_rcv.sb_hiwat)
446 			goto send;
447 	}
448 
449 	/*
450 	 * Send if we owe peer an ACK.
451 	 */
452 	if (tp->t_flags & TF_ACKNOW)
453 		goto send;
454 	if (flags & (TH_SYN|TH_RST))
455 		goto send;
456 	if (SEQ_GT(tp->snd_up, tp->snd_una))
457 		goto send;
458 	/*
459 	 * If our state indicates that FIN should be sent
460 	 * and we have not yet done so, or we're retransmitting the FIN,
461 	 * then we need to send.
462 	 */
463 	if (flags & TH_FIN &&
464 	    ((tp->t_flags & TF_SENTFIN) == 0 || tp->snd_nxt == tp->snd_una))
465 		goto send;
466 #ifdef TCP_SACK
467 	/*
468 	 * In SACK, it is possible for tcp_output to fail to send a segment
469 	 * after the retransmission timer has been turned off.  Make sure
470 	 * that the retransmission timer is set.
471 	 */
472 	if (SEQ_GT(tp->snd_max, tp->snd_una) &&
473 	    TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
474 	    TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
475 		TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
476 		return (0);
477 	}
478 #endif /* TCP_SACK */
479 
480 	/*
481 	 * TCP window updates are not reliable, rather a polling protocol
482 	 * using ``persist'' packets is used to insure receipt of window
483 	 * updates.  The three ``states'' for the output side are:
484 	 *	idle			not doing retransmits or persists
485 	 *	persisting		to move a small or zero window
486 	 *	(re)transmitting	and thereby not persisting
487 	 *
488 	 * tp->t_timer[TCPT_PERSIST]
489 	 *	is set when we are in persist state.
490 	 * tp->t_force
491 	 *	is set when we are called to send a persist packet.
492 	 * tp->t_timer[TCPT_REXMT]
493 	 *	is set when we are retransmitting
494 	 * The output side is idle when both timers are zero.
495 	 *
496 	 * If send window is too small, there is data to transmit, and no
497 	 * retransmit or persist is pending, then go to persist state.
498 	 * If nothing happens soon, send when timer expires:
499 	 * if window is nonzero, transmit what we can,
500 	 * otherwise force out a byte.
501 	 */
502 	if (so->so_snd.sb_cc && TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
503 	    TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
504 		tp->t_rxtshift = 0;
505 		tcp_setpersist(tp);
506 	}
507 
508 	/*
509 	 * No reason to send a segment, just return.
510 	 */
511 	return (0);
512 
513 send:
514 	/*
515 	 * Before ESTABLISHED, force sending of initial options
516 	 * unless TCP set not to do any options.
517 	 * NOTE: we assume that the IP/TCP header plus TCP options
518 	 * always fit in a single mbuf, leaving room for a maximum
519 	 * link header, i.e.
520 	 *	max_linkhdr + sizeof(network header) + sizeof(struct tcphdr +
521 	 * 		optlen <= MHLEN
522 	 */
523 	optlen = 0;
524 
525 	switch (tp->pf) {
526 	case 0:	/*default to PF_INET*/
527 	case PF_INET:
528 		hdrlen = sizeof(struct ip) + sizeof(struct tcphdr);
529 		break;
530 #ifdef INET6
531 	case PF_INET6:
532 		hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
533 		break;
534 #endif /* INET6 */
535 	default:
536 		return (EPFNOSUPPORT);
537 	}
538 
539 	if (flags & TH_SYN) {
540 		tp->snd_nxt = tp->iss;
541 		if ((tp->t_flags & TF_NOOPT) == 0) {
542 			u_int16_t mss;
543 
544 			opt[0] = TCPOPT_MAXSEG;
545 			opt[1] = 4;
546 			mss = htons((u_int16_t) tcp_mss(tp, 0));
547 			bcopy((caddr_t)&mss, (caddr_t)(opt + 2), sizeof(mss));
548 			optlen = 4;
549 
550 			if (flags & TH_ACK)
551 				tcp_mss_update(tp);
552 #ifdef TCP_SACK
553 			/*
554 			 * If this is the first SYN of connection (not a SYN
555 			 * ACK), include SACK_PERMIT_HDR option.  If this is a
556 			 * SYN ACK, include SACK_PERMIT_HDR option if peer has
557 			 * already done so.
558 			 */
559 			if (tp->sack_enable && ((flags & TH_ACK) == 0 ||
560 			    (tp->t_flags & TF_SACK_PERMIT))) {
561 				*((u_int32_t *) (opt + optlen)) =
562 				    htonl(TCPOPT_SACK_PERMIT_HDR);
563 				optlen += 4;
564 			}
565 #endif
566 
567 			if ((tp->t_flags & TF_REQ_SCALE) &&
568 			    ((flags & TH_ACK) == 0 ||
569 			    (tp->t_flags & TF_RCVD_SCALE))) {
570 				*((u_int32_t *) (opt + optlen)) = htonl(
571 					TCPOPT_NOP << 24 |
572 					TCPOPT_WINDOW << 16 |
573 					TCPOLEN_WINDOW << 8 |
574 					tp->request_r_scale);
575 				optlen += 4;
576 			}
577 		}
578 	}
579 
580 	/*
581 	 * Send a timestamp and echo-reply if this is a SYN and our side
582 	 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side
583 	 * and our peer have sent timestamps in our SYN's.
584 	 */
585 	if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
586 	     (flags & TH_RST) == 0 &&
587 	    ((flags & (TH_SYN|TH_ACK)) == TH_SYN ||
588 	     (tp->t_flags & TF_RCVD_TSTMP))) {
589 		u_int32_t *lp = (u_int32_t *)(opt + optlen);
590 
591 		/* Form timestamp option as shown in appendix A of RFC 1323. */
592 		*lp++ = htonl(TCPOPT_TSTAMP_HDR);
593 		*lp++ = htonl(tcp_now + tp->ts_modulate);
594 		*lp   = htonl(tp->ts_recent);
595 		optlen += TCPOLEN_TSTAMP_APPA;
596 
597 		/* Set receive buffer autosizing timestamp. */
598 		if (tp->rfbuf_ts == 0)
599 			tp->rfbuf_ts = tcp_now;
600 
601 	}
602 
603 #ifdef TCP_SIGNATURE
604 	if (tp->t_flags & TF_SIGNATURE) {
605 		u_int8_t *bp = (u_int8_t *)(opt + optlen);
606 
607 		/* Send signature option */
608 		*(bp++) = TCPOPT_SIGNATURE;
609 		*(bp++) = TCPOLEN_SIGNATURE;
610 		sigoff = optlen + 2;
611 
612 		{
613 			unsigned int i;
614 
615 			for (i = 0; i < 16; i++)
616 				*(bp++) = 0;
617 		}
618 
619 
620 		/* Pad options list to the next 32 bit boundary and
621 		 * terminate it.
622 		 */
623 		*bp++ = TCPOPT_NOP;
624 		*bp++ = TCPOPT_NOP;
625 
626 		optlen += TCPOLEN_SIGLEN;
627 	}
628 #endif /* TCP_SIGNATURE */
629 
630 #ifdef TCP_SACK
631 	/*
632 	 * Send SACKs if necessary.  This should be the last option processed.
633 	 * Only as many SACKs are sent as are permitted by the maximum options
634 	 * size.  No more than three SACKs are sent.
635 	 */
636 	if (tp->sack_enable && tp->t_state == TCPS_ESTABLISHED &&
637 	    (tp->t_flags & (TF_SACK_PERMIT|TF_NOOPT)) == TF_SACK_PERMIT &&
638 	    tp->rcv_numsacks) {
639 		u_int32_t *lp = (u_int32_t *)(opt + optlen);
640 		u_int32_t *olp = lp++;
641 		int count = 0;  /* actual number of SACKs inserted */
642 		int maxsack = (MAX_TCPOPTLEN - (optlen + 4))/TCPOLEN_SACK;
643 
644 		tcpstat.tcps_sack_snd_opts++;
645 		maxsack = min(maxsack, TCP_MAX_SACK);
646 		for (i = 0; (i < tp->rcv_numsacks && count < maxsack); i++) {
647 			struct sackblk sack = tp->sackblks[i];
648 			if (sack.start == 0 && sack.end == 0)
649 				continue;
650 			*lp++ = htonl(sack.start);
651 			*lp++ = htonl(sack.end);
652 			count++;
653 		}
654 		*olp = htonl(TCPOPT_SACK_HDR|(TCPOLEN_SACK*count+2));
655 		optlen += TCPOLEN_SACK*count + 4; /* including leading NOPs */
656 	}
657 #endif /* TCP_SACK */
658 
659 #ifdef DIAGNOSTIC
660 	if (optlen > MAX_TCPOPTLEN)
661 		panic("tcp_output: options too long");
662 #endif /* DIAGNOSTIC */
663 
664 	hdrlen += optlen;
665 
666 	/*
667 	 * Adjust data length if insertion of options will
668 	 * bump the packet length beyond the t_maxopd length.
669 	 */
670 	if (len > tp->t_maxopd - optlen) {
671 		len = tp->t_maxopd - optlen;
672 		sendalot = 1;
673 		flags &= ~TH_FIN;
674 	 }
675 
676 #ifdef DIAGNOSTIC
677 	if (max_linkhdr + hdrlen > MCLBYTES)
678 		panic("tcphdr too big");
679 #endif
680 
681 	/*
682 	 * Grab a header mbuf, attaching a copy of data to
683 	 * be transmitted, and initialize the header from
684 	 * the template for sends on this connection.
685 	 */
686 	if (len) {
687 		if (tp->t_force && len == 1)
688 			tcpstat.tcps_sndprobe++;
689 		else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) {
690 			tcpstat.tcps_sndrexmitpack++;
691 			tcpstat.tcps_sndrexmitbyte += len;
692 		} else {
693 			tcpstat.tcps_sndpack++;
694 			tcpstat.tcps_sndbyte += len;
695 		}
696 #ifdef notyet
697 		if ((m = m_copypack(so->so_snd.sb_mb, off,
698 		    (int)len, max_linkhdr + hdrlen)) == 0) {
699 			error = ENOBUFS;
700 			goto out;
701 		}
702 		/*
703 		 * m_copypack left space for our hdr; use it.
704 		 */
705 		m->m_len += hdrlen;
706 		m->m_data -= hdrlen;
707 #else
708 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
709 		if (m != NULL && max_linkhdr + hdrlen > MHLEN) {
710 			MCLGET(m, M_DONTWAIT);
711 			if ((m->m_flags & M_EXT) == 0) {
712 				m_freem(m);
713 				m = NULL;
714 			}
715 		}
716 		if (m == NULL) {
717 			error = ENOBUFS;
718 			goto out;
719 		}
720 		m->m_data += max_linkhdr;
721 		m->m_len = hdrlen;
722 		if (len <= M_TRAILINGSPACE(m)) {
723 			m_copydata(so->so_snd.sb_mb, off, (int) len,
724 			    mtod(m, caddr_t) + hdrlen);
725 			m->m_len += len;
726 		} else {
727 			m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
728 			if (m->m_next == 0) {
729 				(void) m_free(m);
730 				error = ENOBUFS;
731 				goto out;
732 			}
733 		}
734 #endif
735 		/*
736 		 * If we're sending everything we've got, set PUSH.
737 		 * (This will keep happy those implementations which only
738 		 * give data to the user when a buffer fills or
739 		 * a PUSH comes in.)
740 		 */
741 		if (off + len == so->so_snd.sb_cc && !soissending(so))
742 			flags |= TH_PUSH;
743 	} else {
744 		if (tp->t_flags & TF_ACKNOW)
745 			tcpstat.tcps_sndacks++;
746 		else if (flags & (TH_SYN|TH_FIN|TH_RST))
747 			tcpstat.tcps_sndctrl++;
748 		else if (SEQ_GT(tp->snd_up, tp->snd_una))
749 			tcpstat.tcps_sndurg++;
750 		else
751 			tcpstat.tcps_sndwinup++;
752 
753 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
754 		if (m != NULL && max_linkhdr + hdrlen > MHLEN) {
755 			MCLGET(m, M_DONTWAIT);
756 			if ((m->m_flags & M_EXT) == 0) {
757 				m_freem(m);
758 				m = NULL;
759 			}
760 		}
761 		if (m == NULL) {
762 			error = ENOBUFS;
763 			goto out;
764 		}
765 		m->m_data += max_linkhdr;
766 		m->m_len = hdrlen;
767 	}
768 	m->m_pkthdr.rcvif = (struct ifnet *)0;
769 	m->m_pkthdr.len = hdrlen + len;
770 
771 	if (!tp->t_template)
772 		panic("tcp_output");
773 #ifdef DIAGNOSTIC
774 	if (tp->t_template->m_len != hdrlen - optlen)
775 		panic("tcp_output: template len != hdrlen - optlen");
776 #endif /* DIAGNOSTIC */
777 	bcopy(mtod(tp->t_template, caddr_t), mtod(m, caddr_t),
778 		tp->t_template->m_len);
779 	th = (struct tcphdr *)(mtod(m, caddr_t) + tp->t_template->m_len -
780 		sizeof(struct tcphdr));
781 
782 	/*
783 	 * Fill in fields, remembering maximum advertised
784 	 * window for use in delaying messages about window sizes.
785 	 * If resending a FIN, be sure not to use a new sequence number.
786 	 */
787 	if ((flags & TH_FIN) && (tp->t_flags & TF_SENTFIN) &&
788 	    (tp->snd_nxt == tp->snd_max))
789 		tp->snd_nxt--;
790 	/*
791 	 * If we are doing retransmissions, then snd_nxt will
792 	 * not reflect the first unsent octet.  For ACK only
793 	 * packets, we do not want the sequence number of the
794 	 * retransmitted packet, we want the sequence number
795 	 * of the next unsent octet.  So, if there is no data
796 	 * (and no SYN or FIN), use snd_max instead of snd_nxt
797 	 * when filling in ti_seq.  But if we are in persist
798 	 * state, snd_max might reflect one byte beyond the
799 	 * right edge of the window, so use snd_nxt in that
800 	 * case, since we know we aren't doing a retransmission.
801 	 * (retransmit and persist are mutually exclusive...)
802 	 */
803 	if (len || (flags & (TH_SYN|TH_FIN)) || TCP_TIMER_ISARMED(tp, TCPT_PERSIST))
804 		th->th_seq = htonl(tp->snd_nxt);
805 	else
806 		th->th_seq = htonl(tp->snd_max);
807 
808 #ifdef TCP_SACK
809 	if (sack_rxmit) {
810 		/*
811 		 * If sendalot was turned on (due to option stuffing), turn it
812 		 * off. Properly set th_seq field.  Advance the ret'x pointer
813 		 * by len.
814 		 */
815 		if (sendalot)
816 			sendalot = 0;
817 		th->th_seq = htonl(p->rxmit);
818 		p->rxmit += len;
819 #if defined(TCP_SACK) && defined(TCP_FACK)
820 		tp->retran_data += len;
821 #endif /* TCP_FACK */
822 		tcpstat.tcps_sack_rexmits++;
823 		tcpstat.tcps_sack_rexmit_bytes += len;
824 	}
825 #endif /* TCP_SACK */
826 
827 	th->th_ack = htonl(tp->rcv_nxt);
828 	if (optlen) {
829 		bcopy((caddr_t)opt, (caddr_t)(th + 1), optlen);
830 		th->th_off = (sizeof (struct tcphdr) + optlen) >> 2;
831 	}
832 #ifdef TCP_ECN
833 	if (tcp_do_ecn) {
834 		/*
835 		 * if we have received congestion experienced segs,
836 		 * set ECE bit.
837 		 */
838 		if (tp->t_flags & TF_RCVD_CE) {
839 			flags |= TH_ECE;
840 			tcpstat.tcps_ecn_sndece++;
841 		}
842 		if (!(tp->t_flags & TF_DISABLE_ECN)) {
843 			/*
844 			 * if this is a SYN seg, set ECE and CWR.
845 			 * set only ECE for SYN-ACK if peer supports ECN.
846 			 */
847 			if ((flags & (TH_SYN|TH_ACK)) == TH_SYN)
848 				flags |= (TH_ECE|TH_CWR);
849 			else if ((tp->t_flags & TF_ECN_PERMIT) &&
850 				 (flags & (TH_SYN|TH_ACK)) == (TH_SYN|TH_ACK))
851 				flags |= TH_ECE;
852 		}
853 		/*
854 		 * if we have reduced the congestion window, notify
855 		 * the peer by setting CWR bit.
856 		 */
857 		if ((tp->t_flags & TF_ECN_PERMIT) &&
858 		    (tp->t_flags & TF_SEND_CWR)) {
859 			flags |= TH_CWR;
860 			tp->t_flags &= ~TF_SEND_CWR;
861 			tcpstat.tcps_ecn_sndcwr++;
862 		}
863 	}
864 #endif
865 	th->th_flags = flags;
866 
867 	/*
868 	 * Calculate receive window.  Don't shrink window,
869 	 * but avoid silly window syndrome.
870 	 */
871 	if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)tp->t_maxseg)
872 		win = 0;
873 	if (win > (long)TCP_MAXWIN << tp->rcv_scale)
874 		win = (long)TCP_MAXWIN << tp->rcv_scale;
875 	if (win < (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt))
876 		win = (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt);
877 	if (flags & TH_RST)
878 		win = 0;
879 	th->th_win = htons((u_int16_t) (win>>tp->rcv_scale));
880 	if (SEQ_GT(tp->snd_up, tp->snd_nxt)) {
881 		u_int32_t urp = tp->snd_up - tp->snd_nxt;
882 		if (urp > IP_MAXPACKET)
883 			urp = IP_MAXPACKET;
884 		th->th_urp = htons((u_int16_t)urp);
885 		th->th_flags |= TH_URG;
886 	} else
887 		/*
888 		 * If no urgent pointer to send, then we pull
889 		 * the urgent pointer to the left edge of the send window
890 		 * so that it doesn't drift into the send window on sequence
891 		 * number wraparound.
892 		 */
893 		tp->snd_up = tp->snd_una;		/* drag it along */
894 
895 #ifdef TCP_SIGNATURE
896 	if (tp->t_flags & TF_SIGNATURE) {
897 		int iphlen;
898 		union sockaddr_union src, dst;
899 		struct tdb *tdb;
900 
901 		bzero(&src, sizeof(union sockaddr_union));
902 		bzero(&dst, sizeof(union sockaddr_union));
903 
904 		switch (tp->pf) {
905 		case 0:	/*default to PF_INET*/
906 		case AF_INET:
907 			iphlen = sizeof(struct ip);
908 			src.sa.sa_len = sizeof(struct sockaddr_in);
909 			src.sa.sa_family = AF_INET;
910 			src.sin.sin_addr = mtod(m, struct ip *)->ip_src;
911 			dst.sa.sa_len = sizeof(struct sockaddr_in);
912 			dst.sa.sa_family = AF_INET;
913 			dst.sin.sin_addr = mtod(m, struct ip *)->ip_dst;
914 			break;
915 #ifdef INET6
916 		case AF_INET6:
917 			iphlen = sizeof(struct ip6_hdr);
918 			src.sa.sa_len = sizeof(struct sockaddr_in6);
919 			src.sa.sa_family = AF_INET6;
920 			src.sin6.sin6_addr = mtod(m, struct ip6_hdr *)->ip6_src;
921 			dst.sa.sa_len = sizeof(struct sockaddr_in6);
922 			dst.sa.sa_family = AF_INET6;
923 			dst.sin6.sin6_addr = mtod(m, struct ip6_hdr *)->ip6_dst;
924 			break;
925 #endif /* INET6 */
926 		}
927 
928 		tdb = gettdbbysrcdst(rtable_l2(tp->t_inpcb->inp_rtableid),
929 		    0, &src, &dst, IPPROTO_TCP);
930 		if (tdb == NULL)
931 			return (EPERM);
932 
933 		if (tcp_signature(tdb, tp->pf, m, th, iphlen, 0,
934 		    mtod(m, caddr_t) + hdrlen - optlen + sigoff) < 0)
935 			return (EINVAL);
936 	}
937 #endif /* TCP_SIGNATURE */
938 
939 	/* Defer checksumming until later (ip_output() or hardware) */
940 	m->m_pkthdr.csum_flags |= M_TCP_CSUM_OUT;
941 
942 	/*
943 	 * In transmit state, time the transmission and arrange for
944 	 * the retransmit.  In persist state, just set snd_max.
945 	 */
946 	if (tp->t_force == 0 || TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
947 		tcp_seq startseq = tp->snd_nxt;
948 
949 		/*
950 		 * Advance snd_nxt over sequence space of this segment.
951 		 */
952 		if (flags & (TH_SYN|TH_FIN)) {
953 			if (flags & TH_SYN)
954 				tp->snd_nxt++;
955 			if (flags & TH_FIN) {
956 				tp->snd_nxt++;
957 				tp->t_flags |= TF_SENTFIN;
958 			}
959 		}
960 #ifdef TCP_SACK
961 		if (tp->sack_enable) {
962 			if (sack_rxmit && (p->rxmit != tp->snd_nxt)) {
963 				goto timer;
964 			}
965 		}
966 #endif
967 		tp->snd_nxt += len;
968 		if (SEQ_GT(tp->snd_nxt, tp->snd_max)) {
969 			tp->snd_max = tp->snd_nxt;
970 			/*
971 			 * Time this transmission if not a retransmission and
972 			 * not currently timing anything.
973 			 */
974 			if (tp->t_rtttime == 0) {
975 				tp->t_rtttime = tcp_now;
976 				tp->t_rtseq = startseq;
977 				tcpstat.tcps_segstimed++;
978 			}
979 		}
980 
981 		/*
982 		 * Set retransmit timer if not currently set,
983 		 * and not doing an ack or a keep-alive probe.
984 		 * Initial value for retransmit timer is smoothed
985 		 * round-trip time + 2 * round-trip time variance.
986 		 * Initialize shift counter which is used for backoff
987 		 * of retransmit time.
988 		 */
989 #ifdef TCP_SACK
990  timer:
991 		if (tp->sack_enable && sack_rxmit &&
992 		    TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
993 		    tp->snd_nxt != tp->snd_max) {
994 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
995 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) {
996 				TCP_TIMER_DISARM(tp, TCPT_PERSIST);
997 				tp->t_rxtshift = 0;
998 			}
999 		}
1000 #endif
1001 
1002 		if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
1003 		    tp->snd_nxt != tp->snd_una) {
1004 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
1005 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) {
1006 				TCP_TIMER_DISARM(tp, TCPT_PERSIST);
1007 				tp->t_rxtshift = 0;
1008 			}
1009 		}
1010 	} else
1011 		if (SEQ_GT(tp->snd_nxt + len, tp->snd_max))
1012 			tp->snd_max = tp->snd_nxt + len;
1013 
1014 	tcp_update_sndspace(tp);
1015 
1016 	/*
1017 	 * Trace.
1018 	 */
1019 	if (so->so_options & SO_DEBUG)
1020 		tcp_trace(TA_OUTPUT, tp->t_state, tp, mtod(m, caddr_t), 0,
1021 			len);
1022 
1023 	/*
1024 	 * Fill in IP length and desired time to live and
1025 	 * send to IP level.  There should be a better way
1026 	 * to handle ttl and tos; we could keep them in
1027 	 * the template, but need a way to checksum without them.
1028 	 */
1029 
1030 #ifdef TCP_ECN
1031 	/*
1032 	 * if peer is ECN capable, set the ECT bit in the IP header.
1033 	 * but don't set ECT for a pure ack, a retransmit or a window probe.
1034 	 */
1035 	needect = 0;
1036 	if (tcp_do_ecn && (tp->t_flags & TF_ECN_PERMIT)) {
1037 		if (len == 0 || SEQ_LT(tp->snd_nxt, tp->snd_max) ||
1038 		    (tp->t_force && len == 1)) {
1039 			/* don't set ECT */
1040 		} else {
1041 			needect = 1;
1042 			tcpstat.tcps_ecn_sndect++;
1043 		}
1044 	}
1045 #endif
1046 
1047 	/* force routing table */
1048 	m->m_pkthdr.ph_rtableid = tp->t_inpcb->inp_rtableid;
1049 
1050 #if NPF > 0
1051 	m->m_pkthdr.pf.inp = tp->t_inpcb;
1052 #endif
1053 
1054 	switch (tp->pf) {
1055 	case 0:	/*default to PF_INET*/
1056 	case AF_INET:
1057 		{
1058 			struct ip *ip;
1059 
1060 			ip = mtod(m, struct ip *);
1061 			ip->ip_len = htons(m->m_pkthdr.len);
1062 			packetlen = m->m_pkthdr.len;
1063 			ip->ip_ttl = tp->t_inpcb->inp_ip.ip_ttl;
1064 			ip->ip_tos = tp->t_inpcb->inp_ip.ip_tos;
1065 #ifdef TCP_ECN
1066 			if (needect)
1067 				ip->ip_tos |= IPTOS_ECN_ECT0;
1068 #endif
1069 		}
1070 		error = ip_output(m, tp->t_inpcb->inp_options,
1071 			&tp->t_inpcb->inp_route,
1072 			(ip_mtudisc ? IP_MTUDISC : 0), NULL, tp->t_inpcb, 0);
1073 		break;
1074 #ifdef INET6
1075 	case AF_INET6:
1076 		{
1077 			struct ip6_hdr *ip6;
1078 
1079 			ip6 = mtod(m, struct ip6_hdr *);
1080 			ip6->ip6_plen = m->m_pkthdr.len -
1081 				sizeof(struct ip6_hdr);
1082 			packetlen = m->m_pkthdr.len;
1083 			ip6->ip6_nxt = IPPROTO_TCP;
1084 			ip6->ip6_hlim = in6_selecthlim(tp->t_inpcb, NULL);
1085 #ifdef TCP_ECN
1086 			if (needect)
1087 				ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20);
1088 #endif
1089 		}
1090 		error = ip6_output(m, tp->t_inpcb->inp_outputopts6,
1091 			  &tp->t_inpcb->inp_route6,
1092 			  0, NULL, NULL, tp->t_inpcb);
1093 		break;
1094 #endif /* INET6 */
1095 	}
1096 
1097 #if defined(TCP_SACK) && defined(TCP_FACK)
1098 	/* Update snd_awnd to reflect the new data that was sent.  */
1099 	tp->snd_awnd = tcp_seq_subtract(tp->snd_max, tp->snd_fack) +
1100 		tp->retran_data;
1101 #endif /* defined(TCP_SACK) && defined(TCP_FACK) */
1102 
1103 	if (error) {
1104 out:
1105 		if (error == ENOBUFS) {
1106 			/*
1107 			 * If the interface queue is full, or IP cannot
1108 			 * get an mbuf, trigger TCP slow start.
1109 			 */
1110 			tp->snd_cwnd = tp->t_maxseg;
1111 			return (0);
1112 		}
1113 		if (error == EMSGSIZE) {
1114 			/*
1115 			 * ip_output() will have already fixed the route
1116 			 * for us.  tcp_mtudisc() will, as its last action,
1117 			 * initiate retransmission, so it is important to
1118 			 * not do so here.
1119 			 */
1120 			tcp_mtudisc(tp->t_inpcb, -1);
1121 			return (0);
1122 		}
1123 		if (error == EACCES)	/* translate pf(4) error for userland */
1124 			error = EHOSTUNREACH;
1125 		if ((error == EHOSTUNREACH || error == ENETDOWN) &&
1126 		    TCPS_HAVERCVDSYN(tp->t_state)) {
1127 			tp->t_softerror = error;
1128 			return (0);
1129 		}
1130 
1131 		/* Restart the delayed ACK timer, if necessary. */
1132 		if (tp->t_flags & TF_DELACK)
1133 			TCP_RESTART_DELACK(tp);
1134 
1135 		return (error);
1136 	}
1137 
1138 	if (packetlen > tp->t_pmtud_mtu_sent)
1139 		tp->t_pmtud_mtu_sent = packetlen;
1140 
1141 	tcpstat.tcps_sndtotal++;
1142 	if (tp->t_flags & TF_DELACK)
1143 		tcpstat.tcps_delack++;
1144 
1145 	/*
1146 	 * Data sent (as far as we can tell).
1147 	 * If this advertises a larger window than any other segment,
1148 	 * then remember the size of the advertised window.
1149 	 * Any pending ACK has now been sent.
1150 	 */
1151 	if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv))
1152 		tp->rcv_adv = tp->rcv_nxt + win;
1153 	tp->last_ack_sent = tp->rcv_nxt;
1154 	tp->t_flags &= ~TF_ACKNOW;
1155 	TCP_CLEAR_DELACK(tp);
1156 #if defined(TCP_SACK)
1157 	if (sendalot && --maxburst)
1158 #else
1159 	if (sendalot)
1160 #endif
1161 		goto again;
1162 	return (0);
1163 }
1164 
1165 void
1166 tcp_setpersist(struct tcpcb *tp)
1167 {
1168 	int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + TCP_RTT_BASE_SHIFT);
1169 	int nticks;
1170 
1171 	if (TCP_TIMER_ISARMED(tp, TCPT_REXMT))
1172 		panic("tcp_output REXMT");
1173 	/*
1174 	 * Start/restart persistence timer.
1175 	 */
1176 	if (t < tp->t_rttmin)
1177 		t = tp->t_rttmin;
1178 	TCPT_RANGESET(nticks, t * tcp_backoff[tp->t_rxtshift],
1179 	    TCPTV_PERSMIN, TCPTV_PERSMAX);
1180 	TCP_TIMER_ARM(tp, TCPT_PERSIST, nticks);
1181 	if (tp->t_rxtshift < TCP_MAXRXTSHIFT)
1182 		tp->t_rxtshift++;
1183 }
1184