xref: /openbsd-src/sys/netinet/tcp_output.c (revision 64c77b9332c5892d33899e1580df890d2afb4ad0)
1 /*	$OpenBSD: tcp_output.c,v 1.74 2004/10/28 19:22:52 mcbride 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 <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/kernel.h>
78 
79 #include <net/route.h>
80 #include <net/if.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/in_systm.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 INET6
97 #include <netinet6/tcpipv6.h>
98 #include <netinet6/in6_var.h>
99 #endif /* INET6 */
100 
101 #ifdef notyet
102 extern struct mbuf *m_copypack();
103 #endif
104 
105 #ifdef TCP_SACK
106 extern int tcprexmtthresh;
107 #endif
108 
109 #ifdef TCP_SACK
110 #ifdef TCP_SACK_DEBUG
111 void tcp_print_holes(struct tcpcb *tp);
112 
113 void
114 tcp_print_holes(struct tcpcb *tp)
115 {
116 	struct sackhole *p = tp->snd_holes;
117 	if (p == 0)
118 		return;
119 	printf("Hole report: start--end dups rxmit\n");
120 	while (p) {
121 		printf("%x--%x d %d r %x\n", p->start, p->end, p->dups,
122 		    p->rxmit);
123 		p = p->next;
124 	}
125 	printf("\n");
126 }
127 #endif /* TCP_SACK_DEBUG */
128 
129 /*
130  * Returns pointer to a sackhole if there are any pending retransmissions;
131  * NULL otherwise.
132  */
133 struct sackhole *
134 tcp_sack_output(struct tcpcb *tp)
135 {
136 	struct sackhole *p;
137 
138 	if (!tp->sack_enable)
139 		return (NULL);
140 	p = tp->snd_holes;
141 	while (p) {
142 #ifndef TCP_FACK
143 		if (p->dups >= tcprexmtthresh && SEQ_LT(p->rxmit, p->end)) {
144 #else
145 		/* In FACK, if p->dups is less than tcprexmtthresh, but
146 		 * snd_fack advances more than tcprextmtthresh * tp->t_maxseg,
147 		 * tcp_input() will try fast retransmit. This forces output.
148 		 */
149 		if ((p->dups >= tcprexmtthresh ||
150 		     tp->t_dupacks == tcprexmtthresh) &&
151 		    SEQ_LT(p->rxmit, p->end)) {
152 #endif /* TCP_FACK */
153 			if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */
154 				p = p->next;
155 				continue;
156 			}
157 #ifdef TCP_SACK_DEBUG
158 			if (p)
159 				tcp_print_holes(tp);
160 #endif
161 			return (p);
162 		}
163         	p = p->next;
164 	}
165 	return (NULL);
166 }
167 
168 /*
169  * After a timeout, the SACK list may be rebuilt.  This SACK information
170  * should be used to avoid retransmitting SACKed data.  This function
171  * traverses the SACK list to see if snd_nxt should be moved forward.
172  */
173 
174 void
175 tcp_sack_adjust(struct tcpcb *tp)
176 {
177 	struct sackhole *cur = tp->snd_holes;
178 	if (cur == NULL)
179 		return; /* No holes */
180 	if (SEQ_GEQ(tp->snd_nxt, tp->rcv_lastsack))
181 		return; /* We're already beyond any SACKed blocks */
182 	/*
183 	 * Two cases for which we want to advance snd_nxt:
184 	 * i) snd_nxt lies between end of one hole and beginning of another
185 	 * ii) snd_nxt lies between end of last hole and rcv_lastsack
186 	 */
187 	while (cur->next) {
188 		if (SEQ_LT(tp->snd_nxt, cur->end))
189 			return;
190 		if (SEQ_GEQ(tp->snd_nxt, cur->next->start))
191 			cur = cur->next;
192 		else {
193 			tp->snd_nxt = cur->next->start;
194 			return;
195 		}
196 	}
197 	if (SEQ_LT(tp->snd_nxt, cur->end))
198 		return;
199 	tp->snd_nxt = tp->rcv_lastsack;
200 	return;
201 }
202 #endif /* TCP_SACK */
203 
204 /*
205  * Tcp output routine: figure out what should be sent and send it.
206  */
207 int
208 tcp_output(tp)
209 	struct tcpcb *tp;
210 {
211 	struct socket *so = tp->t_inpcb->inp_socket;
212 	long len, win, txmaxseg;
213 	int off, flags, error;
214 	struct mbuf *m;
215 	struct tcphdr *th;
216 	u_char opt[MAX_TCPOPTLEN];
217 	unsigned int optlen, hdrlen;
218 	int idle, sendalot = 0;
219 #ifdef TCP_SACK
220 	int i, sack_rxmit = 0;
221 	struct sackhole *p;
222 #endif
223 #if defined(TCP_SACK)
224 	int maxburst = TCP_MAXBURST;
225 #endif
226 #ifdef TCP_SIGNATURE
227 	unsigned int sigoff;
228 #endif /* TCP_SIGNATURE */
229 #ifdef TCP_ECN
230 	int needect;
231 #endif
232 
233 #if defined(TCP_SACK) && defined(TCP_SIGNATURE) && defined(DIAGNOSTIC)
234 	if (tp->sack_enable && (tp->t_flags & TF_SIGNATURE))
235 		return (EINVAL);
236 #endif /* defined(TCP_SACK) && defined(TCP_SIGNATURE) && defined(DIAGNOSTIC) */
237 
238 	/*
239 	 * Determine length of data that should be transmitted,
240 	 * and flags that will be used.
241 	 * If there is some data or critical controls (SYN, RST)
242 	 * to send, then transmit; otherwise, investigate further.
243 	 */
244 	idle = (tp->t_flags & TF_LASTIDLE) || (tp->snd_max == tp->snd_una);
245 	if (idle && (tcp_now - tp->t_rcvtime) >= tp->t_rxtcur)
246 		/*
247 		 * We have been idle for "a while" and no acks are
248 		 * expected to clock out any data we send --
249 		 * slow start to get ack "clock" running again.
250 		 */
251 		tp->snd_cwnd = 2 * tp->t_maxseg;
252 
253 	/* remember 'idle' for next invocation of tcp_output */
254 	if (idle && soissending(so)) {
255 		tp->t_flags |= TF_LASTIDLE;
256 		idle = 0;
257 	} else
258 		tp->t_flags &= ~TF_LASTIDLE;
259 
260 again:
261 #ifdef TCP_SACK
262 	/*
263 	 * If we've recently taken a timeout, snd_max will be greater than
264 	 * snd_nxt.  There may be SACK information that allows us to avoid
265 	 * resending already delivered data.  Adjust snd_nxt accordingly.
266 	 */
267 	if (tp->sack_enable && SEQ_LT(tp->snd_nxt, tp->snd_max))
268 		tcp_sack_adjust(tp);
269 #endif
270 	off = tp->snd_nxt - tp->snd_una;
271 #if defined(TCP_SACK) && defined(TCP_FACK)
272 	/* Normally, sendable data is limited by off < tp->snd_cwnd.
273 	 * But in FACK, sendable data is limited by snd_awnd < snd_cwnd,
274 	 * regardless of offset.
275 	 */
276 	if (tp->sack_enable && (tp->t_dupacks > tcprexmtthresh))
277 		win = tp->snd_wnd;
278 	else
279 #endif
280 	win = ulmin(tp->snd_wnd, tp->snd_cwnd);
281 
282 	flags = tcp_outflags[tp->t_state];
283 
284 #ifdef TCP_SACK
285 	/*
286 	 * Send any SACK-generated retransmissions.  If we're explicitly trying
287 	 * to send out new data (when sendalot is 1), bypass this function.
288 	 * If we retransmit in fast recovery mode, decrement snd_cwnd, since
289 	 * we're replacing a (future) new transmission with a retransmission
290 	 * now, and we previously incremented snd_cwnd in tcp_input().
291 	 */
292 	if (tp->sack_enable && !sendalot) {
293 		if (tp->t_dupacks >= tcprexmtthresh &&
294 		    (p = tcp_sack_output(tp))) {
295 			off = p->rxmit - tp->snd_una;
296 			sack_rxmit = 1;
297 #if 0
298 			/* Coalesce holes into a single retransmission */
299 #endif
300 			len = min(tp->t_maxseg, p->end - p->rxmit);
301 #ifndef TCP_FACK
302 			/* in FACK, hold snd_cwnd constant during recovery */
303 			if (SEQ_LT(tp->snd_una, tp->snd_last))
304 				tp->snd_cwnd -= tp->t_maxseg;
305 #endif
306     		}
307 	}
308 #endif /* TCP_SACK */
309 
310 	sendalot = 0;
311 	/*
312 	 * If in persist timeout with window of 0, send 1 byte.
313 	 * Otherwise, if window is small but nonzero
314 	 * and timer expired, we will send what we can
315 	 * and go to transmit state.
316 	 */
317 	if (tp->t_force) {
318 		if (win == 0) {
319 			/*
320 			 * If we still have some data to send, then
321 			 * clear the FIN bit.  Usually this would
322 			 * happen below when it realizes that we
323 			 * aren't sending all the data.  However,
324 			 * if we have exactly 1 byte of unset data,
325 			 * then it won't clear the FIN bit below,
326 			 * and if we are in persist state, we wind
327 			 * up sending the packet without recording
328 			 * that we sent the FIN bit.
329 			 *
330 			 * We can't just blindly clear the FIN bit,
331 			 * because if we don't have any more data
332 			 * to send then the probe will be the FIN
333 			 * itself.
334 			 */
335 			if (off < so->so_snd.sb_cc)
336 				flags &= ~TH_FIN;
337 			win = 1;
338 		} else {
339 			TCP_TIMER_DISARM(tp, TCPT_PERSIST);
340 			tp->t_rxtshift = 0;
341 		}
342 	}
343 
344 #ifdef TCP_SACK
345 	if (!sack_rxmit) {
346 #endif
347 	len = ulmin(so->so_snd.sb_cc, win) - off;
348 
349 #if defined(TCP_SACK) && defined(TCP_FACK)
350 	/*
351 	 * If we're in fast recovery (SEQ_GT(tp->snd_last, tp->snd_una)), and
352 	 * amount of outstanding data (snd_awnd) is >= snd_cwnd, then
353 	 * do not send data (like zero window conditions)
354 	 */
355 	if (tp->sack_enable && len && SEQ_GT(tp->snd_last, tp->snd_una) &&
356 	    (tp->snd_awnd >= tp->snd_cwnd))
357 		len = 0;
358 #endif /* TCP_FACK */
359 #ifdef TCP_SACK
360 	}
361 #endif
362 
363 	if (len < 0) {
364 		/*
365 		 * If FIN has been sent but not acked,
366 		 * but we haven't been called to retransmit,
367 		 * len will be -1.  Otherwise, window shrank
368 		 * after we sent into it.  If window shrank to 0,
369 		 * cancel pending retransmit, pull snd_nxt back
370 		 * to (closed) window, and set the persist timer
371 		 * if it isn't already going.  If the window didn't
372 		 * close completely, just wait for an ACK.
373 		 */
374 		len = 0;
375 		if (win == 0) {
376 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
377 			tp->t_rxtshift = 0;
378 			tp->snd_nxt = tp->snd_una;
379 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0)
380 				tcp_setpersist(tp);
381 		}
382 	}
383 
384         /*
385          * Never send more than half a buffer full.  This insures that we can
386          * always keep 2 packets on the wire, no matter what SO_SNDBUF is, and
387          * therefore acks will never be delayed unless we run out of data to
388          * transmit.
389          */
390 	txmaxseg = ulmin(so->so_snd.sb_hiwat / 2, tp->t_maxseg);
391 
392 	if (len > txmaxseg) {
393 		len = txmaxseg;
394 		sendalot = 1;
395 	}
396 	if (off + len < so->so_snd.sb_cc)
397 		flags &= ~TH_FIN;
398 
399 	win = sbspace(&so->so_rcv);
400 
401 	/*
402 	 * Sender silly window avoidance.  If connection is idle
403 	 * and can send all data, a maximum segment,
404 	 * at least a maximum default-size segment do it,
405 	 * or are forced, do it; otherwise don't bother.
406 	 * If peer's buffer is tiny, then send
407 	 * when window is at least half open.
408 	 * If retransmitting (possibly after persist timer forced us
409 	 * to send into a small window), then must resend.
410 	 */
411 	if (len) {
412 		if (len == txmaxseg)
413 			goto send;
414 		if ((idle || tp->t_flags & TF_NODELAY) &&
415 		    len + off >= so->so_snd.sb_cc && !soissending(so))
416 			goto send;
417 		if (tp->t_force)
418 			goto send;
419 		if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0)
420 			goto send;
421 		if (SEQ_LT(tp->snd_nxt, tp->snd_max))
422 			goto send;
423 #ifdef TCP_SACK
424 		if (sack_rxmit)
425 			goto send;
426 #endif
427 	}
428 
429 	/*
430 	 * Compare available window to amount of window
431 	 * known to peer (as advertised window less
432 	 * next expected input).  If the difference is at least two
433 	 * max size segments, or at least 50% of the maximum possible
434 	 * window, then want to send a window update to peer.
435 	 */
436 	if (win > 0) {
437 		/*
438 		 * "adv" is the amount we can increase the window,
439 		 * taking into account that we are limited by
440 		 * TCP_MAXWIN << tp->rcv_scale.
441 		 */
442 		long adv = lmin(win, (long)TCP_MAXWIN << tp->rcv_scale) -
443 			(tp->rcv_adv - tp->rcv_nxt);
444 
445 		if (adv >= (long) (2 * tp->t_maxseg))
446 			goto send;
447 		if (2 * adv >= (long) so->so_rcv.sb_hiwat)
448 			goto send;
449 	}
450 
451 	/*
452 	 * Send if we owe peer an ACK.
453 	 */
454 	if (tp->t_flags & TF_ACKNOW)
455 		goto send;
456 	if (flags & (TH_SYN|TH_RST))
457 		goto send;
458 	if (SEQ_GT(tp->snd_up, tp->snd_una))
459 		goto send;
460 	/*
461 	 * If our state indicates that FIN should be sent
462 	 * and we have not yet done so, or we're retransmitting the FIN,
463 	 * then we need to send.
464 	 */
465 	if (flags & TH_FIN &&
466 	    ((tp->t_flags & TF_SENTFIN) == 0 || tp->snd_nxt == tp->snd_una))
467 		goto send;
468 #ifdef TCP_SACK
469 	/*
470 	 * In SACK, it is possible for tcp_output to fail to send a segment
471 	 * after the retransmission timer has been turned off.  Make sure
472 	 * that the retransmission timer is set.
473 	 */
474 	if (SEQ_GT(tp->snd_max, tp->snd_una) &&
475 	    TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
476 	    TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
477 		TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
478 		return (0);
479 	}
480 #endif /* TCP_SACK */
481 
482 	/*
483 	 * TCP window updates are not reliable, rather a polling protocol
484 	 * using ``persist'' packets is used to insure receipt of window
485 	 * updates.  The three ``states'' for the output side are:
486 	 *	idle			not doing retransmits or persists
487 	 *	persisting		to move a small or zero window
488 	 *	(re)transmitting	and thereby not persisting
489 	 *
490 	 * tp->t_timer[TCPT_PERSIST]
491 	 *	is set when we are in persist state.
492 	 * tp->t_force
493 	 *	is set when we are called to send a persist packet.
494 	 * tp->t_timer[TCPT_REXMT]
495 	 *	is set when we are retransmitting
496 	 * The output side is idle when both timers are zero.
497 	 *
498 	 * If send window is too small, there is data to transmit, and no
499 	 * retransmit or persist is pending, then go to persist state.
500 	 * If nothing happens soon, send when timer expires:
501 	 * if window is nonzero, transmit what we can,
502 	 * otherwise force out a byte.
503 	 */
504 	if (so->so_snd.sb_cc && TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
505 	    TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
506 		tp->t_rxtshift = 0;
507 		tcp_setpersist(tp);
508 	}
509 
510 	/*
511 	 * No reason to send a segment, just return.
512 	 */
513 	return (0);
514 
515 send:
516 	/*
517 	 * Before ESTABLISHED, force sending of initial options
518 	 * unless TCP set not to do any options.
519 	 * NOTE: we assume that the IP/TCP header plus TCP options
520 	 * always fit in a single mbuf, leaving room for a maximum
521 	 * link header, i.e.
522 	 *	max_linkhdr + sizeof(network header) + sizeof(struct tcphdr +
523 	 * 		optlen <= MHLEN
524 	 */
525 	optlen = 0;
526 
527 	switch (tp->pf) {
528 	case 0:	/*default to PF_INET*/
529 #ifdef INET
530 	case PF_INET:
531 		hdrlen = sizeof(struct ip) + sizeof(struct tcphdr);
532 		break;
533 #endif /* INET */
534 #ifdef INET6
535 	case PF_INET6:
536 		hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
537 		break;
538 #endif /* INET6 */
539 	default:
540 		return (EPFNOSUPPORT);
541 	}
542 
543 	if (flags & TH_SYN) {
544 		tp->snd_nxt = tp->iss;
545 		if ((tp->t_flags & TF_NOOPT) == 0) {
546 			u_int16_t mss;
547 
548 			opt[0] = TCPOPT_MAXSEG;
549 			opt[1] = 4;
550 			mss = htons((u_int16_t) tcp_mss(tp, 0));
551 			bcopy((caddr_t)&mss, (caddr_t)(opt + 2), sizeof(mss));
552 			optlen = 4;
553 
554 			if (flags & TH_ACK)
555 				tcp_mss_update(tp);
556 #ifdef TCP_SACK
557 			/*
558 			 * If this is the first SYN of connection (not a SYN
559 			 * ACK), include SACK_PERMIT_HDR option.  If this is a
560 			 * SYN ACK, include SACK_PERMIT_HDR option if peer has
561 			 * already done so.
562 			 */
563 			if (tp->sack_enable && ((flags & TH_ACK) == 0 ||
564 			    (tp->t_flags & TF_SACK_PERMIT))) {
565 				*((u_int32_t *) (opt + optlen)) =
566 				    htonl(TCPOPT_SACK_PERMIT_HDR);
567 				optlen += 4;
568 			}
569 #endif
570 
571 			if ((tp->t_flags & TF_REQ_SCALE) &&
572 			    ((flags & TH_ACK) == 0 ||
573 			    (tp->t_flags & TF_RCVD_SCALE))) {
574 				*((u_int32_t *) (opt + optlen)) = htonl(
575 					TCPOPT_NOP << 24 |
576 					TCPOPT_WINDOW << 16 |
577 					TCPOLEN_WINDOW << 8 |
578 					tp->request_r_scale);
579 				optlen += 4;
580 			}
581 		}
582 	}
583 
584 	/*
585 	 * Send a timestamp and echo-reply if this is a SYN and our side
586 	 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side
587 	 * and our peer have sent timestamps in our SYN's.
588 	 */
589 	if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
590 	     (flags & TH_RST) == 0 &&
591 	    ((flags & (TH_SYN|TH_ACK)) == TH_SYN ||
592 	     (tp->t_flags & TF_RCVD_TSTMP))) {
593 		u_int32_t *lp = (u_int32_t *)(opt + optlen);
594 
595 		/* Form timestamp option as shown in appendix A of RFC 1323. */
596 		*lp++ = htonl(TCPOPT_TSTAMP_HDR);
597 		*lp++ = htonl(tcp_now + tp->ts_modulate);
598 		*lp   = htonl(tp->ts_recent);
599 		optlen += TCPOLEN_TSTAMP_APPA;
600 	}
601 
602 #ifdef TCP_SIGNATURE
603 	if (tp->t_flags & TF_SIGNATURE) {
604 		u_int8_t *bp = (u_int8_t *)(opt + optlen);
605 
606 		/* Send signature option */
607 		*(bp++) = TCPOPT_SIGNATURE;
608 		*(bp++) = TCPOLEN_SIGNATURE;
609 		sigoff = optlen + 2;
610 
611 		{
612 			unsigned int i;
613 
614 			for (i = 0; i < 16; i++)
615 				*(bp++) = 0;
616 		}
617 
618 
619 		/* Pad options list to the next 32 bit boundary and
620 		 * terminate it.
621 		 */
622 		*bp++ = TCPOPT_NOP;
623 		*bp++ = TCPOPT_EOL;
624 
625 		optlen += TCPOLEN_SIGLEN;
626 	}
627 #endif /* TCP_SIGNATURE */
628 
629 #ifdef TCP_SACK
630 	/*
631 	 * Send SACKs if necessary.  This should be the last option processed.
632 	 * Only as many SACKs are sent as are permitted by the maximum options
633 	 * size.  No more than three SACKs are sent.
634 	 */
635 	if (tp->sack_enable && tp->t_state == TCPS_ESTABLISHED &&
636 	    (tp->t_flags & (TF_SACK_PERMIT|TF_NOOPT)) == TF_SACK_PERMIT &&
637 	    tp->rcv_numsacks) {
638 		u_int32_t *lp = (u_int32_t *)(opt + optlen);
639 		u_int32_t *olp = lp++;
640 		int count = 0;  /* actual number of SACKs inserted */
641 		int maxsack = (MAX_TCPOPTLEN - (optlen + 4))/TCPOLEN_SACK;
642 
643 		maxsack = min(maxsack, TCP_MAX_SACK);
644 		for (i = 0; (i < tp->rcv_numsacks && count < maxsack); i++) {
645 			struct sackblk sack = tp->sackblks[i];
646 			if (sack.start == 0 && sack.end == 0)
647 				continue;
648 			*lp++ = htonl(sack.start);
649 			*lp++ = htonl(sack.end);
650 			count++;
651 		}
652 		*olp = htonl(TCPOPT_SACK_HDR|(TCPOLEN_SACK*count+2));
653 		optlen += TCPOLEN_SACK*count + 4; /* including leading NOPs */
654 	}
655 #endif /* TCP_SACK */
656 
657 #ifdef DIAGNOSTIC
658 	if (optlen > MAX_TCPOPTLEN)
659 		panic("tcp_output: options too long");
660 #endif /* DIAGNOSTIC */
661 
662 	hdrlen += optlen;
663 
664 	/*
665 	 * Adjust data length if insertion of options will
666 	 * bump the packet length beyond the t_maxopd length.
667 	 */
668 	if (len > tp->t_maxopd - optlen) {
669 		len = tp->t_maxopd - optlen;
670 		sendalot = 1;
671 		flags &= ~TH_FIN;
672 	 }
673 
674 #ifdef DIAGNOSTIC
675 	if (max_linkhdr + hdrlen > MCLBYTES)
676 		panic("tcphdr too big");
677 #endif
678 
679 	/*
680 	 * Grab a header mbuf, attaching a copy of data to
681 	 * be transmitted, and initialize the header from
682 	 * the template for sends on this connection.
683 	 */
684 	if (len) {
685 		if (tp->t_force && len == 1)
686 			tcpstat.tcps_sndprobe++;
687 		else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) {
688 			tcpstat.tcps_sndrexmitpack++;
689 			tcpstat.tcps_sndrexmitbyte += len;
690 		} else {
691 			tcpstat.tcps_sndpack++;
692 			tcpstat.tcps_sndbyte += len;
693 		}
694 #ifdef notyet
695 		if ((m = m_copypack(so->so_snd.sb_mb, off,
696 		    (int)len, max_linkhdr + hdrlen)) == 0) {
697 			error = ENOBUFS;
698 			goto out;
699 		}
700 		/*
701 		 * m_copypack left space for our hdr; use it.
702 		 */
703 		m->m_len += hdrlen;
704 		m->m_data -= hdrlen;
705 #else
706 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
707 		if (m != NULL && max_linkhdr + hdrlen > MHLEN) {
708 			MCLGET(m, M_DONTWAIT);
709 			if ((m->m_flags & M_EXT) == 0) {
710 				m_freem(m);
711 				m = NULL;
712 			}
713 		}
714 		if (m == NULL) {
715 			error = ENOBUFS;
716 			goto out;
717 		}
718 		m->m_data += max_linkhdr;
719 		m->m_len = hdrlen;
720 		if (len <= M_TRAILINGSPACE(m)) {
721 			m_copydata(so->so_snd.sb_mb, off, (int) len,
722 			    mtod(m, caddr_t) + hdrlen);
723 			m->m_len += len;
724 		} else {
725 			m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
726 			if (m->m_next == 0) {
727 				(void) m_free(m);
728 				error = ENOBUFS;
729 				goto out;
730 			}
731 		}
732 #endif
733 		/*
734 		 * If we're sending everything we've got, set PUSH.
735 		 * (This will keep happy those implementations which only
736 		 * give data to the user when a buffer fills or
737 		 * a PUSH comes in.)
738 		 */
739 		if (off + len == so->so_snd.sb_cc && !soissending(so))
740 			flags |= TH_PUSH;
741 	} else {
742 		if (tp->t_flags & TF_ACKNOW)
743 			tcpstat.tcps_sndacks++;
744 		else if (flags & (TH_SYN|TH_FIN|TH_RST))
745 			tcpstat.tcps_sndctrl++;
746 		else if (SEQ_GT(tp->snd_up, tp->snd_una))
747 			tcpstat.tcps_sndurg++;
748 		else
749 			tcpstat.tcps_sndwinup++;
750 
751 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
752 		if (m != NULL && max_linkhdr + hdrlen > MHLEN) {
753 			MCLGET(m, M_DONTWAIT);
754 			if ((m->m_flags & M_EXT) == 0) {
755 				m_freem(m);
756 				m = NULL;
757 			}
758 		}
759 		if (m == NULL) {
760 			error = ENOBUFS;
761 			goto out;
762 		}
763 		m->m_data += max_linkhdr;
764 		m->m_len = hdrlen;
765 	}
766 	m->m_pkthdr.rcvif = (struct ifnet *)0;
767 	m->m_pkthdr.len = hdrlen + len;
768 
769 	if (!tp->t_template)
770 		panic("tcp_output");
771 #ifdef DIAGNOSTIC
772 	if (tp->t_template->m_len != hdrlen - optlen)
773 		panic("tcp_output: template len != hdrlen - optlen");
774 #endif /* DIAGNOSTIC */
775 	bcopy(mtod(tp->t_template, caddr_t), mtod(m, caddr_t),
776 		tp->t_template->m_len);
777 	th = (struct tcphdr *)(mtod(m, caddr_t) + tp->t_template->m_len -
778 		sizeof(struct tcphdr));
779 
780 	/*
781 	 * Fill in fields, remembering maximum advertised
782 	 * window for use in delaying messages about window sizes.
783 	 * If resending a FIN, be sure not to use a new sequence number.
784 	 */
785 	if ((flags & TH_FIN) && (tp->t_flags & TF_SENTFIN) &&
786 	    (tp->snd_nxt == tp->snd_max))
787 		tp->snd_nxt--;
788 	/*
789 	 * If we are doing retransmissions, then snd_nxt will
790 	 * not reflect the first unsent octet.  For ACK only
791 	 * packets, we do not want the sequence number of the
792 	 * retransmitted packet, we want the sequence number
793 	 * of the next unsent octet.  So, if there is no data
794 	 * (and no SYN or FIN), use snd_max instead of snd_nxt
795 	 * when filling in ti_seq.  But if we are in persist
796 	 * state, snd_max might reflect one byte beyond the
797 	 * right edge of the window, so use snd_nxt in that
798 	 * case, since we know we aren't doing a retransmission.
799 	 * (retransmit and persist are mutually exclusive...)
800 	 */
801 	if (len || (flags & (TH_SYN|TH_FIN)) || TCP_TIMER_ISARMED(tp, TCPT_PERSIST))
802 		th->th_seq = htonl(tp->snd_nxt);
803 	else
804 		th->th_seq = htonl(tp->snd_max);
805 
806 #ifdef TCP_SACK
807 	if (sack_rxmit) {
808 		/*
809 		 * If sendalot was turned on (due to option stuffing), turn it
810 		 * off. Properly set th_seq field.  Advance the ret'x pointer
811 		 * by len.
812 		 */
813 		if (sendalot)
814 			sendalot = 0;
815 		th->th_seq = htonl(p->rxmit);
816 		p->rxmit += len;
817 #if defined(TCP_SACK) && defined(TCP_FACK)
818 		tp->retran_data += len;
819 #endif /* TCP_FACK */
820 	}
821 #endif /* TCP_SACK */
822 
823 	th->th_ack = htonl(tp->rcv_nxt);
824 	if (optlen) {
825 		bcopy((caddr_t)opt, (caddr_t)(th + 1), optlen);
826 		th->th_off = (sizeof (struct tcphdr) + optlen) >> 2;
827 	}
828 #ifdef TCP_ECN
829 	if (tcp_do_ecn) {
830 		/*
831 		 * if we have received congestion experienced segs,
832 		 * set ECE bit.
833 		 */
834 		if (tp->t_flags & TF_RCVD_CE) {
835 			flags |= TH_ECE;
836 			tcpstat.tcps_ecn_sndece++;
837 		}
838 		if (!(tp->t_flags & TF_DISABLE_ECN)) {
839 			/*
840 			 * if this is a SYN seg, set ECE and CWR.
841 			 * set only ECE for SYN-ACK if peer supports ECN.
842 			 */
843 			if ((flags & (TH_SYN|TH_ACK)) == TH_SYN)
844 				flags |= (TH_ECE|TH_CWR);
845 			else if ((tp->t_flags & TF_ECN_PERMIT) &&
846 				 (flags & (TH_SYN|TH_ACK)) == (TH_SYN|TH_ACK))
847 				flags |= TH_ECE;
848 		}
849 		/*
850 		 * if we have reduced the congestion window, notify
851 		 * the peer by setting CWR bit.
852 		 */
853 		if ((tp->t_flags & TF_ECN_PERMIT) &&
854 		    (tp->t_flags & TF_SEND_CWR)) {
855 			flags |= TH_CWR;
856 			tp->t_flags &= ~TF_SEND_CWR;
857 			tcpstat.tcps_ecn_sndcwr++;
858 		}
859 	}
860 #endif
861 	th->th_flags = flags;
862 
863 	/*
864 	 * Calculate receive window.  Don't shrink window,
865 	 * but avoid silly window syndrome.
866 	 */
867 	if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)tp->t_maxseg)
868 		win = 0;
869 	if (win > (long)TCP_MAXWIN << tp->rcv_scale)
870 		win = (long)TCP_MAXWIN << tp->rcv_scale;
871 	if (win < (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt))
872 		win = (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt);
873 	if (flags & TH_RST)
874 		win = 0;
875 	th->th_win = htons((u_int16_t) (win>>tp->rcv_scale));
876 	if (SEQ_GT(tp->snd_up, tp->snd_nxt)) {
877 		u_int32_t urp = tp->snd_up - tp->snd_nxt;
878 		if (urp > IP_MAXPACKET)
879 			urp = IP_MAXPACKET;
880 		th->th_urp = htons((u_int16_t)urp);
881 		th->th_flags |= TH_URG;
882 	} else
883 		/*
884 		 * If no urgent pointer to send, then we pull
885 		 * the urgent pointer to the left edge of the send window
886 		 * so that it doesn't drift into the send window on sequence
887 		 * number wraparound.
888 		 */
889 		tp->snd_up = tp->snd_una;		/* drag it along */
890 
891 #ifdef TCP_SIGNATURE
892 	if (tp->t_flags & TF_SIGNATURE) {
893 		int iphlen;
894 		union sockaddr_union src, dst;
895 		struct tdb *tdb;
896 
897 		bzero(&src, sizeof(union sockaddr_union));
898 		bzero(&dst, sizeof(union sockaddr_union));
899 
900 		switch (tp->pf) {
901 		case 0:	/*default to PF_INET*/
902 #ifdef INET
903 		case AF_INET:
904 			iphlen = sizeof(struct ip);
905 			src.sa.sa_len = sizeof(struct sockaddr_in);
906 			src.sa.sa_family = AF_INET;
907 			src.sin.sin_addr = mtod(m, struct ip *)->ip_src;
908 			dst.sa.sa_len = sizeof(struct sockaddr_in);
909 			dst.sa.sa_family = AF_INET;
910 			dst.sin.sin_addr = mtod(m, struct ip *)->ip_dst;
911 			break;
912 #endif /* INET */
913 #ifdef INET6
914 		case AF_INET6:
915 			iphlen = sizeof(struct ip6_hdr);
916 			src.sa.sa_len = sizeof(struct sockaddr_in6);
917 			src.sa.sa_family = AF_INET6;
918 			src.sin6.sin6_addr = mtod(m, struct ip6_hdr *)->ip6_src;
919 			dst.sa.sa_len = sizeof(struct sockaddr_in6);
920 			dst.sa.sa_family = AF_INET6;
921 			dst.sin6.sin6_addr = mtod(m, struct ip6_hdr *)->ip6_dst;
922 			break;
923 #endif /* INET6 */
924 		}
925 
926 		/* XXX gettdbbysrcdst() should really be called at spltdb(). */
927 		/* XXX this is splsoftnet(), currently they are the same. */
928 		tdb = gettdbbysrcdst(0, &src, &dst, IPPROTO_TCP);
929 		if (tdb == NULL)
930 			return (EPERM);
931 
932 		if (tcp_signature(tdb, tp->pf, m, th, iphlen, 0,
933 		    mtod(m, caddr_t) + hdrlen - optlen + sigoff) < 0)
934 			return (EINVAL);
935 	}
936 #endif /* TCP_SIGNATURE */
937 
938 	/*
939 	 * Put TCP length in extended header, and then
940 	 * checksum extended header and data.
941 	 */
942 	switch (tp->pf) {
943 	case 0:	/*default to PF_INET*/
944 #ifdef INET
945 	case AF_INET:
946 		/* Defer checksumming until later (ip_output() or hardware) */
947 		m->m_pkthdr.csum |= M_TCPV4_CSUM_OUT;
948 		if (len + optlen)
949 			th->th_sum = in_cksum_addword(th->th_sum,
950 			    htons((u_int16_t)(len + optlen)));
951 		break;
952 #endif /* INET */
953 #ifdef INET6
954 	case AF_INET6:
955 		th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr),
956 			hdrlen - sizeof(struct ip6_hdr) + len);
957 		break;
958 #endif /* INET6 */
959 	}
960 
961 	/*
962 	 * In transmit state, time the transmission and arrange for
963 	 * the retransmit.  In persist state, just set snd_max.
964 	 */
965 	if (tp->t_force == 0 || TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) {
966 		tcp_seq startseq = tp->snd_nxt;
967 
968 		/*
969 		 * Advance snd_nxt over sequence space of this segment.
970 		 */
971 		if (flags & (TH_SYN|TH_FIN)) {
972 			if (flags & TH_SYN)
973 				tp->snd_nxt++;
974 			if (flags & TH_FIN) {
975 				tp->snd_nxt++;
976 				tp->t_flags |= TF_SENTFIN;
977 			}
978 		}
979 #ifdef TCP_SACK
980 		if (tp->sack_enable) {
981 			if (sack_rxmit && (p->rxmit != tp->snd_nxt)) {
982 				goto timer;
983 			}
984 		}
985 #endif
986 		tp->snd_nxt += len;
987 		if (SEQ_GT(tp->snd_nxt, tp->snd_max)) {
988 			tp->snd_max = tp->snd_nxt;
989 			/*
990 			 * Time this transmission if not a retransmission and
991 			 * not currently timing anything.
992 			 */
993 			if (tp->t_rtttime == 0) {
994 				tp->t_rtttime = tcp_now;
995 				tp->t_rtseq = startseq;
996 				tcpstat.tcps_segstimed++;
997 			}
998 		}
999 
1000 		/*
1001 		 * Set retransmit timer if not currently set,
1002 		 * and not doing an ack or a keep-alive probe.
1003 		 * Initial value for retransmit timer is smoothed
1004 		 * round-trip time + 2 * round-trip time variance.
1005 		 * Initialize shift counter which is used for backoff
1006 		 * of retransmit time.
1007 		 */
1008 #ifdef TCP_SACK
1009  timer:
1010 		if (tp->sack_enable && sack_rxmit &&
1011 		    TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
1012 		    tp->snd_nxt != tp->snd_max) {
1013 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
1014 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) {
1015 				TCP_TIMER_DISARM(tp, TCPT_PERSIST);
1016 				tp->t_rxtshift = 0;
1017 			}
1018 		}
1019 #endif
1020 
1021 		if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 &&
1022 		    tp->snd_nxt != tp->snd_una) {
1023 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
1024 			if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) {
1025 				TCP_TIMER_DISARM(tp, TCPT_PERSIST);
1026 				tp->t_rxtshift = 0;
1027 			}
1028 		}
1029 	} else
1030 		if (SEQ_GT(tp->snd_nxt + len, tp->snd_max))
1031 			tp->snd_max = tp->snd_nxt + len;
1032 
1033 	/*
1034 	 * Trace.
1035 	 */
1036 	if (so->so_options & SO_DEBUG)
1037 		tcp_trace(TA_OUTPUT, tp->t_state, tp, mtod(m, caddr_t), 0,
1038 			len);
1039 
1040 	/*
1041 	 * Fill in IP length and desired time to live and
1042 	 * send to IP level.  There should be a better way
1043 	 * to handle ttl and tos; we could keep them in
1044 	 * the template, but need a way to checksum without them.
1045 	 */
1046 
1047 #ifdef TCP_ECN
1048 	/*
1049 	 * if peer is ECN capable, set the ECT bit in the IP header.
1050 	 * but don't set ECT for a pure ack, a retransmit or a window probe.
1051 	 */
1052 	needect = 0;
1053 	if (tcp_do_ecn && (tp->t_flags & TF_ECN_PERMIT)) {
1054 		if (len == 0 || SEQ_LT(tp->snd_nxt, tp->snd_max) ||
1055 		    (tp->t_force && len == 1)) {
1056 			/* don't set ECT */
1057 		} else {
1058 			needect = 1;
1059 			tcpstat.tcps_ecn_sndect++;
1060 		}
1061 	}
1062 #endif
1063 
1064 	switch (tp->pf) {
1065 	case 0:	/*default to PF_INET*/
1066 #ifdef INET
1067 	case AF_INET:
1068 		{
1069 			struct ip *ip;
1070 
1071 			ip = mtod(m, struct ip *);
1072 			ip->ip_len = htons(m->m_pkthdr.len);
1073 			ip->ip_ttl = tp->t_inpcb->inp_ip.ip_ttl;
1074 			ip->ip_tos = tp->t_inpcb->inp_ip.ip_tos;
1075 #ifdef TCP_ECN
1076 			if (needect)
1077 				ip->ip_tos |= IPTOS_ECN_ECT0;
1078 #endif
1079 		}
1080 		error = ip_output(m, tp->t_inpcb->inp_options,
1081 			&tp->t_inpcb->inp_route,
1082 			(ip_mtudisc ? IP_MTUDISC : 0) |
1083 				  (so->so_options & SO_DONTROUTE),
1084 			(void *)NULL, tp->t_inpcb);
1085 		break;
1086 #endif /* INET */
1087 #ifdef INET6
1088 	case AF_INET6:
1089 		{
1090 			struct ip6_hdr *ip6;
1091 
1092 			ip6 = mtod(m, struct ip6_hdr *);
1093 			ip6->ip6_plen = m->m_pkthdr.len -
1094 				sizeof(struct ip6_hdr);
1095 			ip6->ip6_nxt = IPPROTO_TCP;
1096 			ip6->ip6_hlim = in6_selecthlim(tp->t_inpcb, NULL);
1097 #ifdef TCP_ECN
1098 			if (needect)
1099 				ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20);
1100 #endif
1101 		}
1102 		error = ip6_output(m, tp->t_inpcb->inp_outputopts6,
1103 			  &tp->t_inpcb->inp_route6,
1104 			  (so->so_options & SO_DONTROUTE), NULL, NULL);
1105 		break;
1106 #endif /* INET6 */
1107 	}
1108 
1109 #if defined(TCP_SACK) && defined(TCP_FACK)
1110 	/* Update snd_awnd to reflect the new data that was sent.  */
1111 	tp->snd_awnd = tcp_seq_subtract(tp->snd_max, tp->snd_fack) +
1112 		tp->retran_data;
1113 #endif /* defined(TCP_SACK) && defined(TCP_FACK) */
1114 
1115 	if (error) {
1116 out:
1117 		if (error == ENOBUFS) {
1118 			tcp_quench(tp->t_inpcb, 0);
1119 			return (0);
1120 		}
1121 		if (error == EMSGSIZE) {
1122 			/*
1123 			 * ip_output() will have already fixed the route
1124 			 * for us.  tcp_mtudisc() will, as its last action,
1125 			 * initiate retransmission, so it is important to
1126 			 * not do so here.
1127 			 */
1128 			tcp_mtudisc(tp->t_inpcb, -1);
1129 			return (0);
1130 		}
1131 		if ((error == EHOSTUNREACH || error == ENETDOWN) &&
1132 		    TCPS_HAVERCVDSYN(tp->t_state)) {
1133 			tp->t_softerror = error;
1134 			return (0);
1135 		}
1136 
1137 		/* Restart the delayed ACK timer, if necessary. */
1138 		if (tp->t_flags & TF_DELACK)
1139 			TCP_RESTART_DELACK(tp);
1140 
1141 		return (error);
1142 	}
1143 	tcpstat.tcps_sndtotal++;
1144 	if (tp->t_flags & TF_DELACK)
1145 		tcpstat.tcps_delack++;
1146 
1147 	/*
1148 	 * Data sent (as far as we can tell).
1149 	 * If this advertises a larger window than any other segment,
1150 	 * then remember the size of the advertised window.
1151 	 * Any pending ACK has now been sent.
1152 	 */
1153 	if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv))
1154 		tp->rcv_adv = tp->rcv_nxt + win;
1155 	tp->last_ack_sent = tp->rcv_nxt;
1156 	tp->t_flags &= ~TF_ACKNOW;
1157 	TCP_CLEAR_DELACK(tp);
1158 #if defined(TCP_SACK)
1159 	if (sendalot && --maxburst)
1160 #else
1161 	if (sendalot)
1162 #endif
1163 		goto again;
1164 	return (0);
1165 }
1166 
1167 void
1168 tcp_setpersist(struct tcpcb *tp)
1169 {
1170 	int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1;
1171 	int nticks;
1172 
1173 	if (TCP_TIMER_ISARMED(tp, TCPT_REXMT))
1174 		panic("tcp_output REXMT");
1175 	/*
1176 	 * Start/restart persistance timer.
1177 	 */
1178 	if (t < tp->t_rttmin)
1179 		t = tp->t_rttmin;
1180 	TCPT_RANGESET(nticks, t * tcp_backoff[tp->t_rxtshift],
1181 	    TCPTV_PERSMIN, TCPTV_PERSMAX);
1182 	TCP_TIMER_ARM(tp, TCPT_PERSIST, nticks);
1183 	if (tp->t_rxtshift < TCP_MAXRXTSHIFT)
1184 		tp->t_rxtshift++;
1185 }
1186