xref: /netbsd-src/sys/netinet/tcp_timer.c (revision aad9773e38ed2370a628a6416e098f9008fc10a7)
1 /*	$NetBSD: tcp_timer.c,v 1.88 2014/11/10 18:52:51 maxv Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1997, 1998, 2001, 2005 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation
38  * Facility, NASA Ames Research Center.
39  * This code is derived from software contributed to The NetBSD Foundation
40  * by Charles M. Hannum.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  *
51  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
52  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
53  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
54  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
55  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
56  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
57  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
58  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
59  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
60  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
61  * POSSIBILITY OF SUCH DAMAGE.
62  */
63 
64 /*
65  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
66  *	The Regents of the University of California.  All rights reserved.
67  *
68  * Redistribution and use in source and binary forms, with or without
69  * modification, are permitted provided that the following conditions
70  * are met:
71  * 1. Redistributions of source code must retain the above copyright
72  *    notice, this list of conditions and the following disclaimer.
73  * 2. Redistributions in binary form must reproduce the above copyright
74  *    notice, this list of conditions and the following disclaimer in the
75  *    documentation and/or other materials provided with the distribution.
76  * 3. Neither the name of the University nor the names of its contributors
77  *    may be used to endorse or promote products derived from this software
78  *    without specific prior written permission.
79  *
80  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
81  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
82  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
83  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
84  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
85  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
86  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
87  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
88  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
89  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
90  * SUCH DAMAGE.
91  *
92  *	@(#)tcp_timer.c	8.2 (Berkeley) 5/24/95
93  */
94 
95 #include <sys/cdefs.h>
96 __KERNEL_RCSID(0, "$NetBSD: tcp_timer.c,v 1.88 2014/11/10 18:52:51 maxv Exp $");
97 
98 #include "opt_inet.h"
99 #include "opt_tcp_debug.h"
100 
101 #include <sys/param.h>
102 #include <sys/systm.h>
103 #include <sys/mbuf.h>
104 #include <sys/socket.h>
105 #include <sys/socketvar.h>
106 #include <sys/protosw.h>
107 #include <sys/errno.h>
108 #include <sys/kernel.h>
109 
110 #include <net/if.h>
111 #include <net/route.h>
112 
113 #include <netinet/in.h>
114 #include <netinet/in_systm.h>
115 #include <netinet/ip.h>
116 #include <netinet/in_pcb.h>
117 #include <netinet/ip_var.h>
118 #include <netinet/ip_icmp.h>
119 
120 #ifdef INET6
121 #ifndef INET
122 #include <netinet/in.h>
123 #endif
124 #include <netinet/ip6.h>
125 #include <netinet6/in6_pcb.h>
126 #endif
127 
128 #include <netinet/tcp.h>
129 #include <netinet/tcp_fsm.h>
130 #include <netinet/tcp_seq.h>
131 #include <netinet/tcp_timer.h>
132 #include <netinet/tcp_var.h>
133 #include <netinet/tcp_private.h>
134 #include <netinet/tcp_congctl.h>
135 #include <netinet/tcpip.h>
136 #ifdef TCP_DEBUG
137 #include <netinet/tcp_debug.h>
138 #endif
139 
140 /*
141  * Various tunable timer parameters.  These are initialized in tcp_init(),
142  * unless they are patched.
143  */
144 u_int	tcp_keepinit = 0;
145 u_int	tcp_keepidle = 0;
146 u_int	tcp_keepintvl = 0;
147 u_int	tcp_keepcnt = 0;		/* max idle probes */
148 
149 int	tcp_maxpersistidle = 0;		/* max idle time in persist */
150 
151 /*
152  * Time to delay the ACK.  This is initialized in tcp_init(), unless
153  * its patched.
154  */
155 int	tcp_delack_ticks = 0;
156 
157 void	tcp_timer_rexmt(void *);
158 void	tcp_timer_persist(void *);
159 void	tcp_timer_keep(void *);
160 void	tcp_timer_2msl(void *);
161 
162 const tcp_timer_func_t tcp_timer_funcs[TCPT_NTIMERS] = {
163 	tcp_timer_rexmt,
164 	tcp_timer_persist,
165 	tcp_timer_keep,
166 	tcp_timer_2msl,
167 };
168 
169 /*
170  * Timer state initialization, called from tcp_init().
171  */
172 void
173 tcp_timer_init(void)
174 {
175 
176 	if (tcp_keepinit == 0)
177 		tcp_keepinit = TCPTV_KEEP_INIT;
178 
179 	if (tcp_keepidle == 0)
180 		tcp_keepidle = TCPTV_KEEP_IDLE;
181 
182 	if (tcp_keepintvl == 0)
183 		tcp_keepintvl = TCPTV_KEEPINTVL;
184 
185 	if (tcp_keepcnt == 0)
186 		tcp_keepcnt = TCPTV_KEEPCNT;
187 
188 	if (tcp_maxpersistidle == 0)
189 		tcp_maxpersistidle = TCPTV_KEEP_IDLE;
190 
191 	if (tcp_delack_ticks == 0)
192 		tcp_delack_ticks = TCP_DELACK_TICKS;
193 }
194 
195 /*
196  * Callout to process delayed ACKs for a TCPCB.
197  */
198 void
199 tcp_delack(void *arg)
200 {
201 	struct tcpcb *tp = arg;
202 
203 	/*
204 	 * If tcp_output() wasn't able to transmit the ACK
205 	 * for whatever reason, it will restart the delayed
206 	 * ACK callout.
207 	 */
208 
209 	mutex_enter(softnet_lock);
210 	if ((tp->t_flags & (TF_DEAD | TF_DELACK)) != TF_DELACK) {
211 		mutex_exit(softnet_lock);
212 		return;
213 	}
214 	if (!callout_expired(&tp->t_delack_ch)) {
215 		mutex_exit(softnet_lock);
216 		return;
217 	}
218 
219 	tp->t_flags |= TF_ACKNOW;
220 	KERNEL_LOCK(1, NULL);
221 	(void) tcp_output(tp);
222 	KERNEL_UNLOCK_ONE(NULL);
223 	mutex_exit(softnet_lock);
224 }
225 
226 /*
227  * Tcp protocol timeout routine called every 500 ms.
228  * Updates the timers in all active tcb's and
229  * causes finite state machine actions if timers expire.
230  */
231 void
232 tcp_slowtimo(void *arg)
233 {
234 
235 	mutex_enter(softnet_lock);
236 	tcp_iss_seq += TCP_ISSINCR;			/* increment iss */
237 	tcp_now++;					/* for timestamps */
238 	mutex_exit(softnet_lock);
239 
240 	callout_schedule(&tcp_slowtimo_ch, hz / PR_SLOWHZ);
241 }
242 
243 /*
244  * Cancel all timers for TCP tp.
245  */
246 void
247 tcp_canceltimers(struct tcpcb *tp)
248 {
249 	int i;
250 
251 	for (i = 0; i < TCPT_NTIMERS; i++)
252 		TCP_TIMER_DISARM(tp, i);
253 }
254 
255 const int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
256     { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
257 
258 const int	tcp_totbackoff = 511;	/* sum of tcp_backoff[] */
259 
260 /*
261  * TCP timer processing.
262  */
263 
264 void
265 tcp_timer_rexmt(void *arg)
266 {
267 	struct tcpcb *tp = arg;
268 	uint32_t rto;
269 #ifdef TCP_DEBUG
270 	struct socket *so = NULL;
271 	short ostate;
272 #endif
273 
274 	mutex_enter(softnet_lock);
275 	if ((tp->t_flags & TF_DEAD) != 0) {
276 		mutex_exit(softnet_lock);
277 		return;
278 	}
279 	if (!callout_expired(&tp->t_timer[TCPT_REXMT])) {
280 		mutex_exit(softnet_lock);
281 		return;
282 	}
283 
284 	KERNEL_LOCK(1, NULL);
285 	if ((tp->t_flags & TF_PMTUD_PEND) && tp->t_inpcb &&
286 	    SEQ_GEQ(tp->t_pmtud_th_seq, tp->snd_una) &&
287 	    SEQ_LT(tp->t_pmtud_th_seq, (int)(tp->snd_una + tp->t_ourmss))) {
288 		extern struct sockaddr_in icmpsrc;
289 		struct icmp icmp;
290 
291 		tp->t_flags &= ~TF_PMTUD_PEND;
292 
293 		/* XXX create fake icmp message with relevant entries */
294 		icmp.icmp_nextmtu = tp->t_pmtud_nextmtu;
295 		icmp.icmp_ip.ip_len = tp->t_pmtud_ip_len;
296 		icmp.icmp_ip.ip_hl = tp->t_pmtud_ip_hl;
297 		icmpsrc.sin_addr = tp->t_inpcb->inp_faddr;
298 		icmp_mtudisc(&icmp, icmpsrc.sin_addr);
299 
300 		/*
301 		 * Notify all connections to the same peer about
302 		 * new mss and trigger retransmit.
303 		 */
304 		in_pcbnotifyall(&tcbtable, icmpsrc.sin_addr, EMSGSIZE,
305 		    tcp_mtudisc);
306 		KERNEL_UNLOCK_ONE(NULL);
307 		mutex_exit(softnet_lock);
308  		return;
309  	}
310 #ifdef TCP_DEBUG
311 #ifdef INET
312 	if (tp->t_inpcb)
313 		so = tp->t_inpcb->inp_socket;
314 #endif
315 #ifdef INET6
316 	if (tp->t_in6pcb)
317 		so = tp->t_in6pcb->in6p_socket;
318 #endif
319 	ostate = tp->t_state;
320 #endif /* TCP_DEBUG */
321 
322 	/*
323 	 * Clear the SACK scoreboard, reset FACK estimate.
324 	 */
325 	tcp_free_sackholes(tp);
326 	tp->snd_fack = tp->snd_una;
327 
328 	/*
329 	 * Retransmission timer went off.  Message has not
330 	 * been acked within retransmit interval.  Back off
331 	 * to a longer retransmit interval and retransmit one segment.
332 	 */
333 
334 	if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
335 		tp->t_rxtshift = TCP_MAXRXTSHIFT;
336 		TCP_STATINC(TCP_STAT_TIMEOUTDROP);
337 		tp = tcp_drop(tp, tp->t_softerror ?
338 		    tp->t_softerror : ETIMEDOUT);
339 		goto out;
340 	}
341 	TCP_STATINC(TCP_STAT_REXMTTIMEO);
342 	rto = TCP_REXMTVAL(tp);
343 	if (rto < tp->t_rttmin)
344 		rto = tp->t_rttmin;
345 	TCPT_RANGESET(tp->t_rxtcur, rto * tcp_backoff[tp->t_rxtshift],
346 	    tp->t_rttmin, TCPTV_REXMTMAX);
347 	TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
348 
349 	/*
350 	 * If we are losing and we are trying path MTU discovery,
351 	 * try turning it off.  This will avoid black holes in
352 	 * the network which suppress or fail to send "packet
353 	 * too big" ICMP messages.  We should ideally do
354 	 * lots more sophisticated searching to find the right
355 	 * value here...
356 	 */
357 	if (tp->t_mtudisc && tp->t_rxtshift > TCP_MAXRXTSHIFT / 6) {
358 		TCP_STATINC(TCP_STAT_PMTUBLACKHOLE);
359 
360 #ifdef INET
361 		/* try turning PMTUD off */
362 		if (tp->t_inpcb)
363 			tp->t_mtudisc = 0;
364 #endif
365 #ifdef INET6
366 		/* try using IPv6 minimum MTU */
367 		if (tp->t_in6pcb)
368 			tp->t_mtudisc = 0;
369 #endif
370 
371 		/* XXX: more sophisticated Black hole recovery code? */
372 	}
373 
374 	/*
375 	 * If losing, let the lower level know and try for
376 	 * a better route.  Also, if we backed off this far,
377 	 * our srtt estimate is probably bogus.  Clobber it
378 	 * so we'll take the next rtt measurement as our srtt;
379 	 * move the current srtt into rttvar to keep the current
380 	 * retransmit times until then.
381 	 */
382 	if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
383 #ifdef INET
384 		if (tp->t_inpcb)
385 			in_losing(tp->t_inpcb);
386 #endif
387 #ifdef INET6
388 		if (tp->t_in6pcb)
389 			in6_losing(tp->t_in6pcb);
390 #endif
391 		/*
392 		 * This operation is not described in RFC2988.  The
393 		 * point is to keep srtt+4*rttvar constant, so we
394 		 * should shift right 2 bits to divide by 4, and then
395 		 * shift right one bit because the storage
396 		 * representation of rttvar is 1/16s vs 1/32s for
397 		 * srtt.
398 		 */
399 		tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
400 		tp->t_srtt = 0;
401 	}
402 	tp->snd_nxt = tp->snd_una;
403 	tp->snd_high = tp->snd_max;
404 	/*
405 	 * If timing a segment in this window, stop the timer.
406 	 */
407 	tp->t_rtttime = 0;
408 	/*
409 	 * Remember if we are retransmitting a SYN, because if
410 	 * we do, set the initial congestion window must be set
411 	 * to 1 segment.
412 	 */
413 	if (tp->t_state == TCPS_SYN_SENT)
414 		tp->t_flags |= TF_SYN_REXMT;
415 
416 	/*
417 	 * Adjust congestion control parameters.
418 	 */
419 	tp->t_congctl->slow_retransmit(tp);
420 
421 	(void) tcp_output(tp);
422 
423  out:
424 #ifdef TCP_DEBUG
425 	if (tp && so->so_options & SO_DEBUG)
426 		tcp_trace(TA_USER, ostate, tp, NULL,
427 		    PRU_SLOWTIMO | (TCPT_REXMT << 8));
428 #endif
429 	KERNEL_UNLOCK_ONE(NULL);
430 	mutex_exit(softnet_lock);
431 }
432 
433 void
434 tcp_timer_persist(void *arg)
435 {
436 	struct tcpcb *tp = arg;
437 	uint32_t rto;
438 #ifdef TCP_DEBUG
439 	struct socket *so = NULL;
440 	short ostate;
441 #endif
442 
443 	mutex_enter(softnet_lock);
444 	if ((tp->t_flags & TF_DEAD) != 0) {
445 		mutex_exit(softnet_lock);
446 		return;
447 	}
448 	if (!callout_expired(&tp->t_timer[TCPT_PERSIST])) {
449 		mutex_exit(softnet_lock);
450 		return;
451 	}
452 
453 	KERNEL_LOCK(1, NULL);
454 #ifdef TCP_DEBUG
455 #ifdef INET
456 	if (tp->t_inpcb)
457 		so = tp->t_inpcb->inp_socket;
458 #endif
459 #ifdef INET6
460 	if (tp->t_in6pcb)
461 		so = tp->t_in6pcb->in6p_socket;
462 #endif
463 
464 	ostate = tp->t_state;
465 #endif /* TCP_DEBUG */
466 
467 	/*
468 	 * Persistance timer into zero window.
469 	 * Force a byte to be output, if possible.
470 	 */
471 
472 	/*
473 	 * Hack: if the peer is dead/unreachable, we do not
474 	 * time out if the window is closed.  After a full
475 	 * backoff, drop the connection if the idle time
476 	 * (no responses to probes) reaches the maximum
477 	 * backoff that we would use if retransmitting.
478 	 */
479 	rto = TCP_REXMTVAL(tp);
480 	if (rto < tp->t_rttmin)
481 		rto = tp->t_rttmin;
482 	if (tp->t_rxtshift == TCP_MAXRXTSHIFT &&
483 	    ((tcp_now - tp->t_rcvtime) >= tcp_maxpersistidle ||
484 	    (tcp_now - tp->t_rcvtime) >= rto * tcp_totbackoff)) {
485 		TCP_STATINC(TCP_STAT_PERSISTDROPS);
486 		tp = tcp_drop(tp, ETIMEDOUT);
487 		goto out;
488 	}
489 	TCP_STATINC(TCP_STAT_PERSISTTIMEO);
490 	tcp_setpersist(tp);
491 	tp->t_force = 1;
492 	(void) tcp_output(tp);
493 	tp->t_force = 0;
494 
495  out:
496 #ifdef TCP_DEBUG
497 	if (tp && so->so_options & SO_DEBUG)
498 		tcp_trace(TA_USER, ostate, tp, NULL,
499 		    PRU_SLOWTIMO | (TCPT_PERSIST << 8));
500 #endif
501 	KERNEL_UNLOCK_ONE(NULL);
502 	mutex_exit(softnet_lock);
503 }
504 
505 void
506 tcp_timer_keep(void *arg)
507 {
508 	struct tcpcb *tp = arg;
509 	struct socket *so = NULL;	/* Quell compiler warning */
510 #ifdef TCP_DEBUG
511 	short ostate;
512 #endif
513 
514 	mutex_enter(softnet_lock);
515 	if ((tp->t_flags & TF_DEAD) != 0) {
516 		mutex_exit(softnet_lock);
517 		return;
518 	}
519 	if (!callout_expired(&tp->t_timer[TCPT_KEEP])) {
520 		mutex_exit(softnet_lock);
521 		return;
522 	}
523 
524 	KERNEL_LOCK(1, NULL);
525 
526 #ifdef TCP_DEBUG
527 	ostate = tp->t_state;
528 #endif /* TCP_DEBUG */
529 
530 	/*
531 	 * Keep-alive timer went off; send something
532 	 * or drop connection if idle for too long.
533 	 */
534 
535 	TCP_STATINC(TCP_STAT_KEEPTIMEO);
536 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
537 		goto dropit;
538 #ifdef INET
539 	if (tp->t_inpcb)
540 		so = tp->t_inpcb->inp_socket;
541 #endif
542 #ifdef INET6
543 	if (tp->t_in6pcb)
544 		so = tp->t_in6pcb->in6p_socket;
545 #endif
546 	KASSERT(so != NULL);
547 	if (so->so_options & SO_KEEPALIVE &&
548 	    tp->t_state <= TCPS_CLOSE_WAIT) {
549 	    	if ((tp->t_maxidle > 0) &&
550 		    ((tcp_now - tp->t_rcvtime) >=
551 		     tp->t_keepidle + tp->t_maxidle))
552 			goto dropit;
553 		/*
554 		 * Send a packet designed to force a response
555 		 * if the peer is up and reachable:
556 		 * either an ACK if the connection is still alive,
557 		 * or an RST if the peer has closed the connection
558 		 * due to timeout or reboot.
559 		 * Using sequence number tp->snd_una-1
560 		 * causes the transmitted zero-length segment
561 		 * to lie outside the receive window;
562 		 * by the protocol spec, this requires the
563 		 * correspondent TCP to respond.
564 		 */
565 		TCP_STATINC(TCP_STAT_KEEPPROBE);
566 		if (tcp_compat_42) {
567 			/*
568 			 * The keepalive packet must have nonzero
569 			 * length to get a 4.2 host to respond.
570 			 */
571 			(void)tcp_respond(tp, tp->t_template,
572 			    NULL, NULL, tp->rcv_nxt - 1,
573 			    tp->snd_una - 1, 0);
574 		} else {
575 			(void)tcp_respond(tp, tp->t_template,
576 			    NULL, NULL, tp->rcv_nxt,
577 			    tp->snd_una - 1, 0);
578 		}
579 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl);
580 	} else
581 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
582 
583 #ifdef TCP_DEBUG
584 	if (tp && so->so_options & SO_DEBUG)
585 		tcp_trace(TA_USER, ostate, tp, NULL,
586 		    PRU_SLOWTIMO | (TCPT_KEEP << 8));
587 #endif
588 	KERNEL_UNLOCK_ONE(NULL);
589 	mutex_exit(softnet_lock);
590 	return;
591 
592  dropit:
593 	TCP_STATINC(TCP_STAT_KEEPDROPS);
594 	(void) tcp_drop(tp, ETIMEDOUT);
595 	KERNEL_UNLOCK_ONE(NULL);
596 	mutex_exit(softnet_lock);
597 }
598 
599 void
600 tcp_timer_2msl(void *arg)
601 {
602 	struct tcpcb *tp = arg;
603 #ifdef TCP_DEBUG
604 	struct socket *so = NULL;
605 	short ostate;
606 #endif
607 
608 	mutex_enter(softnet_lock);
609 	if ((tp->t_flags & TF_DEAD) != 0) {
610 		mutex_exit(softnet_lock);
611 		return;
612 	}
613 	if (!callout_expired(&tp->t_timer[TCPT_2MSL])) {
614 		mutex_exit(softnet_lock);
615 		return;
616 	}
617 
618 	/*
619 	 * 2 MSL timeout went off, clear the SACK scoreboard, reset
620 	 * the FACK estimate.
621 	 */
622 	KERNEL_LOCK(1, NULL);
623 	tcp_free_sackholes(tp);
624 	tp->snd_fack = tp->snd_una;
625 
626 #ifdef TCP_DEBUG
627 #ifdef INET
628 	if (tp->t_inpcb)
629 		so = tp->t_inpcb->inp_socket;
630 #endif
631 #ifdef INET6
632 	if (tp->t_in6pcb)
633 		so = tp->t_in6pcb->in6p_socket;
634 #endif
635 
636 	ostate = tp->t_state;
637 #endif /* TCP_DEBUG */
638 
639 	/*
640 	 * 2 MSL timeout in shutdown went off.  If we're closed but
641 	 * still waiting for peer to close and connection has been idle
642 	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
643 	 * control block.  Otherwise, check again in a bit.
644 	 */
645 	if (tp->t_state != TCPS_TIME_WAIT &&
646 	    ((tp->t_maxidle == 0) ||
647 	    ((tcp_now - tp->t_rcvtime) <= tp->t_maxidle)))
648 	    TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_keepintvl);
649 	else
650 		tp = tcp_close(tp);
651 
652 #ifdef TCP_DEBUG
653 	if (tp && so->so_options & SO_DEBUG)
654 		tcp_trace(TA_USER, ostate, tp, NULL,
655 		    PRU_SLOWTIMO | (TCPT_2MSL << 8));
656 #endif
657 	mutex_exit(softnet_lock);
658 	KERNEL_UNLOCK_ONE(NULL);
659 }
660