xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 82d56013d7b633d116a93943de88e08335357a7c)
1 /*	$NetBSD: tcp_usrreq.c,v 1.229 2021/03/08 18:17:27 christos 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, 2005, 2006 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  * This code is derived from software contributed to The NetBSD Foundation
42  * by Rui Paulo.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
54  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
56  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
57  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
58  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
59  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
60  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
61  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
62  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
63  * POSSIBILITY OF SUCH DAMAGE.
64  */
65 
66 /*
67  * Copyright (c) 1982, 1986, 1988, 1993, 1995
68  *	The Regents of the University of California.  All rights reserved.
69  *
70  * Redistribution and use in source and binary forms, with or without
71  * modification, are permitted provided that the following conditions
72  * are met:
73  * 1. Redistributions of source code must retain the above copyright
74  *    notice, this list of conditions and the following disclaimer.
75  * 2. Redistributions in binary form must reproduce the above copyright
76  *    notice, this list of conditions and the following disclaimer in the
77  *    documentation and/or other materials provided with the distribution.
78  * 3. Neither the name of the University nor the names of its contributors
79  *    may be used to endorse or promote products derived from this software
80  *    without specific prior written permission.
81  *
82  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
83  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
84  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
85  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
86  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
87  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
88  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
89  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
90  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
91  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
92  * SUCH DAMAGE.
93  *
94  *	@(#)tcp_usrreq.c	8.5 (Berkeley) 6/21/95
95  */
96 
97 /*
98  * TCP protocol interface to socket abstraction.
99  */
100 
101 #include <sys/cdefs.h>
102 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.229 2021/03/08 18:17:27 christos Exp $");
103 
104 #ifdef _KERNEL_OPT
105 #include "opt_inet.h"
106 #include "opt_tcp_debug.h"
107 #include "opt_mbuftrace.h"
108 #include "opt_tcp_space.h"
109 #include "opt_net_mpsafe.h"
110 #endif
111 
112 #include <sys/param.h>
113 #include <sys/systm.h>
114 #include <sys/kernel.h>
115 #include <sys/mbuf.h>
116 #include <sys/socket.h>
117 #include <sys/socketvar.h>
118 #include <sys/protosw.h>
119 #include <sys/errno.h>
120 #include <sys/stat.h>
121 #include <sys/proc.h>
122 #include <sys/domain.h>
123 #include <sys/sysctl.h>
124 #include <sys/kauth.h>
125 #include <sys/kernel.h>
126 #include <sys/uidinfo.h>
127 
128 #include <net/if.h>
129 
130 #include <netinet/in.h>
131 #include <netinet/in_systm.h>
132 #include <netinet/in_var.h>
133 #include <netinet/ip.h>
134 #include <netinet/in_pcb.h>
135 #include <netinet/ip_var.h>
136 #include <netinet/in_offload.h>
137 
138 #ifdef INET6
139 #include <netinet/ip6.h>
140 #include <netinet6/in6_pcb.h>
141 #include <netinet6/ip6_var.h>
142 #include <netinet6/scope6_var.h>
143 #endif
144 
145 #include <netinet/tcp.h>
146 #include <netinet/tcp_fsm.h>
147 #include <netinet/tcp_seq.h>
148 #include <netinet/tcp_timer.h>
149 #include <netinet/tcp_var.h>
150 #include <netinet/tcp_private.h>
151 #include <netinet/tcp_congctl.h>
152 #include <netinet/tcp_debug.h>
153 #include <netinet/tcp_vtw.h>
154 
155 static int
156 tcp_debug_capture(struct tcpcb *tp, int req)
157 {
158 #ifdef TCP_DEBUG
159 	return tp->t_state;
160 #endif
161 	return 0;
162 }
163 
164 static inline void
165 tcp_debug_trace(struct socket *so, struct tcpcb *tp, int ostate, int req)
166 {
167 #ifdef TCP_DEBUG
168 	if (tp && (so->so_options & SO_DEBUG))
169 		tcp_trace(TA_USER, ostate, tp, NULL, req);
170 #endif
171 }
172 
173 static int
174 tcp_getpcb(struct socket *so, struct inpcb **inp,
175     struct in6pcb **in6p, struct tcpcb **tp)
176 {
177 
178 	KASSERT(solocked(so));
179 
180 	/*
181 	 * When a TCP is attached to a socket, then there will be
182 	 * a (struct inpcb) pointed at by the socket, and this
183 	 * structure will point at a subsidary (struct tcpcb).
184 	 */
185 	switch (so->so_proto->pr_domain->dom_family) {
186 	case PF_INET:
187 		*inp = sotoinpcb(so);
188 		if (*inp == NULL)
189 			return EINVAL;
190 		*tp = intotcpcb(*inp);
191 		break;
192 #ifdef INET6
193 	case PF_INET6:
194 		*in6p = sotoin6pcb(so);
195 		if (*in6p == NULL)
196 			return EINVAL;
197 		*tp = in6totcpcb(*in6p);
198 		break;
199 #endif
200 	default:
201 		return EAFNOSUPPORT;
202 	}
203 
204 	KASSERT(tp != NULL);
205 
206 	return 0;
207 }
208 
209 static void
210 change_keepalive(struct socket *so, struct tcpcb *tp)
211 {
212 	tp->t_maxidle = tp->t_keepcnt * MIN(tp->t_keepintvl,
213 	    TCP_TIMER_MAXTICKS / tp->t_keepcnt);
214 	TCP_TIMER_DISARM(tp, TCPT_KEEP);
215 	TCP_TIMER_DISARM(tp, TCPT_2MSL);
216 
217 	if (tp->t_state == TCPS_SYN_RECEIVED ||
218 	    tp->t_state == TCPS_SYN_SENT) {
219 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
220 	} else if (so->so_options & SO_KEEPALIVE &&
221 	    tp->t_state <= TCPS_CLOSE_WAIT) {
222 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl);
223 	} else {
224 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
225 	}
226 
227 	if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
228 		TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
229 }
230 
231 /*
232  * Export TCP internal state information via a struct tcp_info, based on the
233  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
234  * (TCP state machine, etc).  We export all information using FreeBSD-native
235  * constants -- for example, the numeric values for tcpi_state will differ
236  * from Linux.
237  */
238 static void
239 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
240 {
241 
242 	bzero(ti, sizeof(*ti));
243 
244 	ti->tcpi_state = tp->t_state;
245 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
246 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
247 	if (tp->t_flags & TF_SACK_PERMIT)
248 		ti->tcpi_options |= TCPI_OPT_SACK;
249 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
250 		ti->tcpi_options |= TCPI_OPT_WSCALE;
251 		ti->tcpi_snd_wscale = tp->snd_scale;
252 		ti->tcpi_rcv_wscale = tp->rcv_scale;
253 	}
254 	if (tp->t_flags & TF_ECN_PERMIT) {
255 		ti->tcpi_options |= TCPI_OPT_ECN;
256 	}
257 
258 	ti->tcpi_rto = tp->t_rxtcur * tick;
259 	ti->tcpi_last_data_recv = (long)(getticks() -
260 					 (int)tp->t_rcvtime) * tick;
261 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick / PR_SLOWHZ)
262 	                   >> (TCP_RTT_SHIFT + 2);
263 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick / PR_SLOWHZ)
264 	                   >> (TCP_RTTVAR_SHIFT + 2);
265 
266 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
267 	/* Linux API wants these in # of segments, apparently */
268 	ti->tcpi_snd_cwnd = tp->snd_cwnd / tp->t_segsz;
269 	ti->tcpi_snd_wnd = tp->snd_wnd / tp->t_segsz;
270 
271 	/*
272 	 * FreeBSD-specific extension fields for tcp_info.
273 	 */
274 	ti->tcpi_rcv_space = tp->rcv_wnd;
275 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
276 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
277 	ti->tcpi_snd_nxt = tp->snd_nxt;
278 	ti->tcpi_snd_mss = tp->t_segsz;
279 	ti->tcpi_rcv_mss = tp->t_segsz;
280 #ifdef TF_TOE
281 	if (tp->t_flags & TF_TOE)
282 		ti->tcpi_options |= TCPI_OPT_TOE;
283 #endif
284 	/* From the redundant department of redundancies... */
285 	ti->__tcpi_retransmits = ti->__tcpi_retrans =
286 		ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
287 
288 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
289 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
290 }
291 
292 int
293 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
294 {
295 	int error = 0, s;
296 	struct inpcb *inp;
297 #ifdef INET6
298 	struct in6pcb *in6p;
299 #endif
300 	struct tcpcb *tp;
301 	struct tcp_info ti;
302 	u_int ui;
303 	int family;	/* family of the socket */
304 	int level, optname, optval;
305 
306 	level = sopt->sopt_level;
307 	optname = sopt->sopt_name;
308 
309 	family = so->so_proto->pr_domain->dom_family;
310 
311 	s = splsoftnet();
312 	switch (family) {
313 	case PF_INET:
314 		inp = sotoinpcb(so);
315 #ifdef INET6
316 		in6p = NULL;
317 #endif
318 		break;
319 #ifdef INET6
320 	case PF_INET6:
321 		inp = NULL;
322 		in6p = sotoin6pcb(so);
323 		break;
324 #endif
325 	default:
326 		splx(s);
327 		panic("%s: af %d", __func__, family);
328 	}
329 #ifndef INET6
330 	if (inp == NULL)
331 #else
332 	if (inp == NULL && in6p == NULL)
333 #endif
334 	{
335 		splx(s);
336 		return ECONNRESET;
337 	}
338 	if (level != IPPROTO_TCP) {
339 		switch (family) {
340 		case PF_INET:
341 			error = ip_ctloutput(op, so, sopt);
342 			break;
343 #ifdef INET6
344 		case PF_INET6:
345 			error = ip6_ctloutput(op, so, sopt);
346 			break;
347 #endif
348 		}
349 		splx(s);
350 		return error;
351 	}
352 	if (inp)
353 		tp = intotcpcb(inp);
354 #ifdef INET6
355 	else if (in6p)
356 		tp = in6totcpcb(in6p);
357 #endif
358 	else
359 		tp = NULL;
360 
361 	switch (op) {
362 	case PRCO_SETOPT:
363 		switch (optname) {
364 #ifdef TCP_SIGNATURE
365 		case TCP_MD5SIG:
366 			error = sockopt_getint(sopt, &optval);
367 			if (error)
368 				break;
369 			if (optval > 0)
370 				tp->t_flags |= TF_SIGNATURE;
371 			else
372 				tp->t_flags &= ~TF_SIGNATURE;
373 			break;
374 #endif /* TCP_SIGNATURE */
375 
376 		case TCP_NODELAY:
377 			error = sockopt_getint(sopt, &optval);
378 			if (error)
379 				break;
380 			if (optval)
381 				tp->t_flags |= TF_NODELAY;
382 			else
383 				tp->t_flags &= ~TF_NODELAY;
384 			break;
385 
386 		case TCP_MAXSEG:
387 			error = sockopt_getint(sopt, &optval);
388 			if (error)
389 				break;
390 			if (optval > 0 && optval <= tp->t_peermss)
391 				tp->t_peermss = optval; /* limit on send size */
392 			else
393 				error = EINVAL;
394 			break;
395 #ifdef notyet
396 		case TCP_CONGCTL:
397 			/* XXX string overflow XXX */
398 			error = tcp_congctl_select(tp, sopt->sopt_data);
399 			break;
400 #endif
401 
402 		case TCP_KEEPIDLE:
403 			error = sockopt_get(sopt, &ui, sizeof(ui));
404 			if (error)
405 				break;
406 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
407 				tp->t_keepidle = ui;
408 				change_keepalive(so, tp);
409 			} else
410 				error = EINVAL;
411 			break;
412 
413 		case TCP_KEEPINTVL:
414 			error = sockopt_get(sopt, &ui, sizeof(ui));
415 			if (error)
416 				break;
417 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
418 				tp->t_keepintvl = ui;
419 				change_keepalive(so, tp);
420 			} else
421 				error = EINVAL;
422 			break;
423 
424 		case TCP_KEEPCNT:
425 			error = sockopt_get(sopt, &ui, sizeof(ui));
426 			if (error)
427 				break;
428 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
429 				tp->t_keepcnt = ui;
430 				change_keepalive(so, tp);
431 			} else
432 				error = EINVAL;
433 			break;
434 
435 		case TCP_KEEPINIT:
436 			error = sockopt_get(sopt, &ui, sizeof(ui));
437 			if (error)
438 				break;
439 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
440 				tp->t_keepinit = ui;
441 				change_keepalive(so, tp);
442 			} else
443 				error = EINVAL;
444 			break;
445 
446 		default:
447 			error = ENOPROTOOPT;
448 			break;
449 		}
450 		break;
451 
452 	case PRCO_GETOPT:
453 		switch (optname) {
454 #ifdef TCP_SIGNATURE
455 		case TCP_MD5SIG:
456 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
457 			goto setval;
458 #endif
459 		case TCP_NODELAY:
460 			optval = tp->t_flags & TF_NODELAY;
461 			goto setval;
462 		case TCP_MAXSEG:
463 			optval = tp->t_peermss;
464 			goto setval;
465 		case TCP_INFO:
466 			tcp_fill_info(tp, &ti);
467 			error = sockopt_set(sopt, &ti, sizeof ti);
468 			break;
469 #ifdef notyet
470 		case TCP_CONGCTL:
471 			break;
472 #endif
473 		case TCP_KEEPIDLE:
474 			optval = tp->t_keepidle;
475 			goto setval;
476 		case TCP_KEEPINTVL:
477 			optval = tp->t_keepintvl;
478 			goto setval;
479 		case TCP_KEEPCNT:
480 			optval = tp->t_keepcnt;
481 			goto setval;
482 		case TCP_KEEPINIT:
483 			optval = tp->t_keepcnt;
484 setval:			error = sockopt_set(sopt, &optval, sizeof(optval));
485 			break;
486 		default:
487 			error = ENOPROTOOPT;
488 			break;
489 		}
490 		break;
491 	}
492 	splx(s);
493 	return error;
494 }
495 
496 #ifndef TCP_SENDSPACE
497 #define	TCP_SENDSPACE	1024*32
498 #endif
499 int	tcp_sendspace = TCP_SENDSPACE;
500 #ifndef TCP_RECVSPACE
501 #define	TCP_RECVSPACE	1024*32
502 #endif
503 int	tcp_recvspace = TCP_RECVSPACE;
504 
505 /*
506  * tcp_attach: attach TCP protocol to socket, allocating internet protocol
507  * control block, TCP control block, buffer space and entering LISTEN state
508  * if to accept connections.
509  */
510 static int
511 tcp_attach(struct socket *so, int proto)
512 {
513 	struct tcpcb *tp;
514 	struct inpcb *inp;
515 #ifdef INET6
516 	struct in6pcb *in6p;
517 #endif
518 	int s, error, family;
519 
520 	/* Assign the lock (must happen even if we will error out). */
521 	s = splsoftnet();
522 	sosetlock(so);
523 	KASSERT(solocked(so));
524 
525 	family = so->so_proto->pr_domain->dom_family;
526 	switch (family) {
527 	case PF_INET:
528 		inp = sotoinpcb(so);
529 #ifdef INET6
530 		in6p = NULL;
531 #endif
532 		break;
533 #ifdef INET6
534 	case PF_INET6:
535 		inp = NULL;
536 		in6p = sotoin6pcb(so);
537 		break;
538 #endif
539 	default:
540 		error = EAFNOSUPPORT;
541 		goto out;
542 	}
543 
544 	KASSERT(inp == NULL);
545 #ifdef INET6
546 	KASSERT(in6p == NULL);
547 #endif
548 
549 #ifdef MBUFTRACE
550 	so->so_mowner = &tcp_sock_mowner;
551 	so->so_rcv.sb_mowner = &tcp_sock_rx_mowner;
552 	so->so_snd.sb_mowner = &tcp_sock_tx_mowner;
553 #endif
554 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
555 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
556 		if (error)
557 			goto out;
558 	}
559 
560 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
561 	so->so_snd.sb_flags |= SB_AUTOSIZE;
562 
563 	switch (family) {
564 	case PF_INET:
565 		error = in_pcballoc(so, &tcbtable);
566 		if (error)
567 			goto out;
568 		inp = sotoinpcb(so);
569 #ifdef INET6
570 		in6p = NULL;
571 #endif
572 		break;
573 #ifdef INET6
574 	case PF_INET6:
575 		error = in6_pcballoc(so, &tcbtable);
576 		if (error)
577 			goto out;
578 		inp = NULL;
579 		in6p = sotoin6pcb(so);
580 		break;
581 #endif
582 	default:
583 		error = EAFNOSUPPORT;
584 		goto out;
585 	}
586 	if (inp)
587 		tp = tcp_newtcpcb(family, (void *)inp);
588 #ifdef INET6
589 	else if (in6p)
590 		tp = tcp_newtcpcb(family, (void *)in6p);
591 #endif
592 	else
593 		tp = NULL;
594 
595 	if (tp == NULL) {
596 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
597 
598 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
599 		if (inp)
600 			in_pcbdetach(inp);
601 #ifdef INET6
602 		if (in6p)
603 			in6_pcbdetach(in6p);
604 #endif
605 		so->so_state |= nofd;
606 		error = ENOBUFS;
607 		goto out;
608 	}
609 	tp->t_state = TCPS_CLOSED;
610 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
611 		so->so_linger = TCP_LINGERTIME;
612 	}
613 out:
614 	KASSERT(solocked(so));
615 	splx(s);
616 	return error;
617 }
618 
619 static void
620 tcp_detach(struct socket *so)
621 {
622 	struct inpcb *inp = NULL;
623 	struct in6pcb *in6p = NULL;
624 	struct tcpcb *tp = NULL;
625 	int s;
626 
627 	if (tcp_getpcb(so, &inp, &in6p, &tp) != 0)
628 		return;
629 
630 	s = splsoftnet();
631 	(void)tcp_disconnect1(tp);
632 	splx(s);
633 }
634 
635 static int
636 tcp_accept(struct socket *so, struct sockaddr *nam)
637 {
638 	struct inpcb *inp = NULL;
639 	struct in6pcb *in6p = NULL;
640 	struct tcpcb *tp = NULL;
641 	int ostate = 0;
642 	int error = 0;
643 	int s;
644 
645 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
646 		return error;
647 
648 	ostate = tcp_debug_capture(tp, PRU_ACCEPT);
649 
650 	/*
651 	 * Accept a connection.  Essentially all the work is
652 	 * done at higher levels; just return the address
653 	 * of the peer, storing through addr.
654 	 */
655 	s = splsoftnet();
656 	if (inp) {
657 		in_setpeeraddr(inp, (struct sockaddr_in *)nam);
658 	}
659 #ifdef INET6
660 	if (in6p) {
661 		in6_setpeeraddr(in6p, (struct sockaddr_in6 *)nam);
662 	}
663 #endif
664 	tcp_debug_trace(so, tp, ostate, PRU_ACCEPT);
665 	splx(s);
666 
667 	return 0;
668 }
669 
670 static int
671 tcp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
672 {
673 	struct inpcb *inp = NULL;
674 	struct in6pcb *in6p = NULL;
675 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
676 #ifdef INET6
677 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
678 #endif /* INET6 */
679 	struct tcpcb *tp = NULL;
680 	int s;
681 	int error = 0;
682 	int ostate = 0;
683 
684 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
685 		return error;
686 
687 	ostate = tcp_debug_capture(tp, PRU_BIND);
688 
689 	/*
690 	 * Give the socket an address.
691 	 */
692 	s = splsoftnet();
693 	switch (so->so_proto->pr_domain->dom_family) {
694 	case PF_INET:
695 		error = in_pcbbind(inp, sin, l);
696 		break;
697 #ifdef INET6
698 	case PF_INET6:
699 		error = in6_pcbbind(in6p, sin6, l);
700 		if (!error) {
701 			/* mapped addr case */
702 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
703 				tp->t_family = AF_INET;
704 			else
705 				tp->t_family = AF_INET6;
706 		}
707 		break;
708 #endif
709 	}
710 	tcp_debug_trace(so, tp, ostate, PRU_BIND);
711 	splx(s);
712 
713 	return error;
714 }
715 
716 static int
717 tcp_listen(struct socket *so, struct lwp *l)
718 {
719 	struct inpcb *inp = NULL;
720 	struct in6pcb *in6p = NULL;
721 	struct tcpcb *tp = NULL;
722 	int error = 0;
723 	int ostate = 0;
724 	int s;
725 
726 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
727 		return error;
728 
729 	ostate = tcp_debug_capture(tp, PRU_LISTEN);
730 
731 	/*
732 	 * Prepare to accept connections.
733 	 */
734 	s = splsoftnet();
735 	if (inp && inp->inp_lport == 0) {
736 		error = in_pcbbind(inp, NULL, l);
737 		if (error)
738 			goto release;
739 	}
740 #ifdef INET6
741 	if (in6p && in6p->in6p_lport == 0) {
742 		error = in6_pcbbind(in6p, NULL, l);
743 		if (error)
744 			goto release;
745 	}
746 #endif
747 	tp->t_state = TCPS_LISTEN;
748 
749 release:
750 	tcp_debug_trace(so, tp, ostate, PRU_LISTEN);
751 	splx(s);
752 
753 	return error;
754 }
755 
756 static int
757 tcp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
758 {
759 	struct inpcb *inp = NULL;
760 	struct in6pcb *in6p = NULL;
761 	struct tcpcb *tp = NULL;
762 	int s;
763 	int error = 0;
764 	int ostate = 0;
765 
766 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
767 		return error;
768 
769 	ostate = tcp_debug_capture(tp, PRU_CONNECT);
770 
771 	/*
772 	 * Initiate connection to peer.
773 	 * Create a template for use in transmissions on this connection.
774 	 * Enter SYN_SENT state, and mark socket as connecting.
775 	 * Start keep-alive timer, and seed output sequence space.
776 	 * Send initial segment on connection.
777 	 */
778 	s = splsoftnet();
779 
780 	if (inp) {
781 		if (inp->inp_lport == 0) {
782 			error = in_pcbbind(inp, NULL, l);
783 			if (error)
784 				goto release;
785 		}
786 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
787 	}
788 #ifdef INET6
789 	if (in6p) {
790 		if (in6p->in6p_lport == 0) {
791 			error = in6_pcbbind(in6p, NULL, l);
792 			if (error)
793 				goto release;
794 		}
795 		error = in6_pcbconnect(in6p, (struct sockaddr_in6 *)nam, l);
796 		if (!error) {
797 			/* mapped addr case */
798 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
799 				tp->t_family = AF_INET;
800 			else
801 				tp->t_family = AF_INET6;
802 		}
803 	}
804 #endif
805 	if (error)
806 		goto release;
807 	tp->t_template = tcp_template(tp);
808 	if (tp->t_template == 0) {
809 		if (inp)
810 			in_pcbdisconnect(inp);
811 #ifdef INET6
812 		if (in6p)
813 			in6_pcbdisconnect(in6p);
814 #endif
815 		error = ENOBUFS;
816 		goto release;
817 	}
818 	/*
819 	 * Compute window scaling to request.
820 	 * XXX: This should be moved to tcp_output().
821 	 */
822 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
823 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
824 		tp->request_r_scale++;
825 	soisconnecting(so);
826 	TCP_STATINC(TCP_STAT_CONNATTEMPT);
827 	tp->t_state = TCPS_SYN_SENT;
828 	TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
829 	tp->iss = tcp_new_iss(tp);
830 	tcp_sendseqinit(tp);
831 	error = tcp_output(tp);
832 
833 release:
834 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT);
835 	splx(s);
836 
837 	return error;
838 }
839 
840 static int
841 tcp_connect2(struct socket *so, struct socket *so2)
842 {
843 	struct inpcb *inp = NULL;
844 	struct in6pcb *in6p = NULL;
845 	struct tcpcb *tp = NULL;
846 	int error = 0;
847 	int ostate = 0;
848 
849 	KASSERT(solocked(so));
850 
851 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
852 		return error;
853 
854 	ostate = tcp_debug_capture(tp, PRU_CONNECT2);
855 
856 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT2);
857 
858 	return EOPNOTSUPP;
859 }
860 
861 static int
862 tcp_disconnect(struct socket *so)
863 {
864 	struct inpcb *inp = NULL;
865 	struct in6pcb *in6p = NULL;
866 	struct tcpcb *tp = NULL;
867 	int error = 0;
868 	int ostate = 0;
869 	int s;
870 
871 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
872 		return error;
873 
874 	ostate = tcp_debug_capture(tp, PRU_DISCONNECT);
875 
876 	/*
877 	 * Initiate disconnect from peer.
878 	 * If connection never passed embryonic stage, just drop;
879 	 * else if don't need to let data drain, then can just drop anyways,
880 	 * else have to begin TCP shutdown process: mark socket disconnecting,
881 	 * drain unread data, state switch to reflect user close, and
882 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
883 	 * when peer sends FIN and acks ours.
884 	 *
885 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
886 	 */
887 	s = splsoftnet();
888 	tp = tcp_disconnect1(tp);
889 	tcp_debug_trace(so, tp, ostate, PRU_DISCONNECT);
890 	splx(s);
891 
892 	return error;
893 }
894 
895 static int
896 tcp_shutdown(struct socket *so)
897 {
898 	struct inpcb *inp = NULL;
899 	struct in6pcb *in6p = NULL;
900 	struct tcpcb *tp = NULL;
901 	int error = 0;
902 	int ostate = 0;
903 	int s;
904 
905 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
906 		return error;
907 
908 	ostate = tcp_debug_capture(tp, PRU_SHUTDOWN);
909 	/*
910 	 * Mark the connection as being incapable of further output.
911 	 */
912 	s = splsoftnet();
913 	socantsendmore(so);
914 	tp = tcp_usrclosed(tp);
915 	if (tp)
916 		error = tcp_output(tp);
917 	tcp_debug_trace(so, tp, ostate, PRU_SHUTDOWN);
918 	splx(s);
919 
920 	return error;
921 }
922 
923 static int
924 tcp_abort(struct socket *so)
925 {
926 	struct inpcb *inp = NULL;
927 	struct in6pcb *in6p = NULL;
928 	struct tcpcb *tp = NULL;
929 	int error = 0;
930 	int ostate = 0;
931 	int s;
932 
933 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
934 		return error;
935 
936 	ostate = tcp_debug_capture(tp, PRU_ABORT);
937 
938 	/*
939 	 * Abort the TCP.
940 	 */
941 	s = splsoftnet();
942 	tp = tcp_drop(tp, ECONNABORTED);
943 	tcp_debug_trace(so, tp, ostate, PRU_ABORT);
944 	splx(s);
945 
946 	return error;
947 }
948 
949 static int
950 tcp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
951 {
952 	switch (so->so_proto->pr_domain->dom_family) {
953 	case PF_INET:
954 		return in_control(so, cmd, nam, ifp);
955 #ifdef INET6
956 	case PF_INET6:
957 		return in6_control(so, cmd, nam, ifp);
958 #endif
959 	default:
960 		return EAFNOSUPPORT;
961 	}
962 }
963 
964 static int
965 tcp_stat(struct socket *so, struct stat *ub)
966 {
967 	KASSERT(solocked(so));
968 
969 	/* stat: don't bother with a blocksize.  */
970 	return 0;
971 }
972 
973 static int
974 tcp_peeraddr(struct socket *so, struct sockaddr *nam)
975 {
976 	struct inpcb *inp = NULL;
977 	struct in6pcb *in6p = NULL;
978 	struct tcpcb *tp = NULL;
979 	int ostate = 0;
980 	int error = 0;
981 	int s;
982 
983 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
984 		return error;
985 
986 	ostate = tcp_debug_capture(tp, PRU_PEERADDR);
987 
988 	s = splsoftnet();
989 	if (inp) {
990 		in_setpeeraddr(inp, (struct sockaddr_in *)nam);
991 	}
992 #ifdef INET6
993 	if (in6p) {
994 		in6_setpeeraddr(in6p, (struct sockaddr_in6 *)nam);
995 	}
996 #endif
997 	tcp_debug_trace(so, tp, ostate, PRU_PEERADDR);
998 	splx(s);
999 
1000 	return 0;
1001 }
1002 
1003 static int
1004 tcp_sockaddr(struct socket *so, struct sockaddr *nam)
1005 {
1006 	struct inpcb *inp = NULL;
1007 	struct in6pcb *in6p = NULL;
1008 	struct tcpcb *tp = NULL;
1009 	int ostate = 0;
1010 	int error = 0;
1011 	int s;
1012 
1013 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1014 		return error;
1015 
1016 	ostate = tcp_debug_capture(tp, PRU_SOCKADDR);
1017 
1018 	s = splsoftnet();
1019 	if (inp) {
1020 		in_setsockaddr(inp, (struct sockaddr_in *)nam);
1021 	}
1022 #ifdef INET6
1023 	if (in6p) {
1024 		in6_setsockaddr(in6p, (struct sockaddr_in6 *)nam);
1025 	}
1026 #endif
1027 	tcp_debug_trace(so, tp, ostate, PRU_SOCKADDR);
1028 	splx(s);
1029 
1030 	return 0;
1031 }
1032 
1033 static int
1034 tcp_rcvd(struct socket *so, int flags, struct lwp *l)
1035 {
1036 	struct inpcb *inp = NULL;
1037 	struct in6pcb *in6p = NULL;
1038 	struct tcpcb *tp = NULL;
1039 	int ostate = 0;
1040 	int error = 0;
1041 	int s;
1042 
1043 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1044 		return error;
1045 
1046 	ostate = tcp_debug_capture(tp, PRU_RCVD);
1047 
1048 	/*
1049 	 * After a receive, possibly send window update to peer.
1050 	 *
1051 	 * soreceive() calls this function when a user receives
1052 	 * ancillary data on a listening socket. We don't call
1053 	 * tcp_output in such a case, since there is no header
1054 	 * template for a listening socket and hence the kernel
1055 	 * will panic.
1056 	 */
1057 	s = splsoftnet();
1058 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
1059 		(void) tcp_output(tp);
1060 	splx(s);
1061 
1062 	tcp_debug_trace(so, tp, ostate, PRU_RCVD);
1063 
1064 	return 0;
1065 }
1066 
1067 static int
1068 tcp_recvoob(struct socket *so, struct mbuf *m, int flags)
1069 {
1070 	struct inpcb *inp = NULL;
1071 	struct in6pcb *in6p = NULL;
1072 	struct tcpcb *tp = NULL;
1073 	int ostate = 0;
1074 	int error = 0;
1075 	int s;
1076 
1077 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1078 		return error;
1079 
1080 	ostate = tcp_debug_capture(tp, PRU_RCVOOB);
1081 
1082 	s = splsoftnet();
1083 	if ((so->so_oobmark == 0 &&
1084 	    (so->so_state & SS_RCVATMARK) == 0) ||
1085 	    so->so_options & SO_OOBINLINE ||
1086 	    tp->t_oobflags & TCPOOB_HADDATA) {
1087 		splx(s);
1088 		return EINVAL;
1089 	}
1090 
1091 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1092 		splx(s);
1093 		return EWOULDBLOCK;
1094 	}
1095 
1096 	m->m_len = 1;
1097 	*mtod(m, char *) = tp->t_iobc;
1098 	if ((flags & MSG_PEEK) == 0) {
1099 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1100 		so->so_state &= ~SS_POLLRDBAND;
1101 	}
1102 
1103 	tcp_debug_trace(so, tp, ostate, PRU_RCVOOB);
1104 	splx(s);
1105 
1106 	return 0;
1107 }
1108 
1109 static int
1110 tcp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1111     struct mbuf *control, struct lwp *l)
1112 {
1113 	struct inpcb *inp = NULL;
1114 	struct in6pcb *in6p = NULL;
1115 	struct tcpcb *tp = NULL;
1116 	int ostate = 0;
1117 	int error = 0;
1118 	int s;
1119 
1120 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1121 		return error;
1122 
1123 	ostate = tcp_debug_capture(tp, PRU_SEND);
1124 
1125 	/*
1126 	 * Do a send by putting data in output queue and updating urgent
1127 	 * marker if URG set.  Possibly send more data.
1128 	 */
1129 	s = splsoftnet();
1130 	if (control && control->m_len) {
1131 		m_freem(control);
1132 		m_freem(m);
1133 		tcp_debug_trace(so, tp, ostate, PRU_SEND);
1134 		splx(s);
1135 		return EINVAL;
1136 	}
1137 
1138 	sbappendstream(&so->so_snd, m);
1139 	error = tcp_output(tp);
1140 	tcp_debug_trace(so, tp, ostate, PRU_SEND);
1141 	splx(s);
1142 
1143 	return error;
1144 }
1145 
1146 static int
1147 tcp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1148 {
1149 	struct inpcb *inp = NULL;
1150 	struct in6pcb *in6p = NULL;
1151 	struct tcpcb *tp = NULL;
1152 	int ostate = 0;
1153 	int error = 0;
1154 	int s;
1155 
1156 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0) {
1157 		m_freem(m);
1158 		m_freem(control);
1159 		return error;
1160 	}
1161 
1162 	ostate = tcp_debug_capture(tp, PRU_SENDOOB);
1163 
1164 	s = splsoftnet();
1165 	if (sbspace_oob(&so->so_snd) == 0) {
1166 		m_freem(m);
1167 		m_freem(control);
1168 		splx(s);
1169 		return ENOBUFS;
1170 	}
1171 	/*
1172 	 * According to RFC961 (Assigned Protocols),
1173 	 * the urgent pointer points to the last octet
1174 	 * of urgent data.  We continue, however,
1175 	 * to consider it to indicate the first octet
1176 	 * of data past the urgent section.
1177 	 * Otherwise, snd_up should be one lower.
1178 	 */
1179 	sbappendstream(&so->so_snd, m);
1180 	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1181 	tp->t_force = 1;
1182 	error = tcp_output(tp);
1183 	tp->t_force = 0;
1184 	tcp_debug_trace(so, tp, ostate, PRU_SENDOOB);
1185 	splx(s);
1186 	m_freem(control);
1187 
1188 	return error;
1189 }
1190 
1191 static int
1192 tcp_purgeif(struct socket *so, struct ifnet *ifp)
1193 {
1194 	int s;
1195 	int error = 0;
1196 
1197 	s = splsoftnet();
1198 
1199 	mutex_enter(softnet_lock);
1200 	switch (so->so_proto->pr_domain->dom_family) {
1201 	case PF_INET:
1202 		in_pcbpurgeif0(&tcbtable, ifp);
1203 #ifdef NET_MPSAFE
1204 		mutex_exit(softnet_lock);
1205 #endif
1206 		in_purgeif(ifp);
1207 #ifdef NET_MPSAFE
1208 		mutex_enter(softnet_lock);
1209 #endif
1210 		in_pcbpurgeif(&tcbtable, ifp);
1211 		break;
1212 #ifdef INET6
1213 	case PF_INET6:
1214 		in6_pcbpurgeif0(&tcbtable, ifp);
1215 #ifdef NET_MPSAFE
1216 		mutex_exit(softnet_lock);
1217 #endif
1218 		in6_purgeif(ifp);
1219 #ifdef NET_MPSAFE
1220 		mutex_enter(softnet_lock);
1221 #endif
1222 		in6_pcbpurgeif(&tcbtable, ifp);
1223 		break;
1224 #endif
1225 	default:
1226 		error = EAFNOSUPPORT;
1227 		break;
1228 	}
1229 	mutex_exit(softnet_lock);
1230 	splx(s);
1231 
1232 	return error;
1233 }
1234 
1235 /*
1236  * Initiate (or continue) disconnect.
1237  * If embryonic state, just send reset (once).
1238  * If in ``let data drain'' option and linger null, just drop.
1239  * Otherwise (hard), mark socket disconnecting and drop
1240  * current input data; switch states based on user close, and
1241  * send segment to peer (with FIN).
1242  */
1243 struct tcpcb *
1244 tcp_disconnect1(struct tcpcb *tp)
1245 {
1246 	struct socket *so;
1247 
1248 	if (tp->t_inpcb)
1249 		so = tp->t_inpcb->inp_socket;
1250 #ifdef INET6
1251 	else if (tp->t_in6pcb)
1252 		so = tp->t_in6pcb->in6p_socket;
1253 #endif
1254 	else
1255 		so = NULL;
1256 
1257 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
1258 		tp = tcp_close(tp);
1259 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1260 		tp = tcp_drop(tp, 0);
1261 	else {
1262 		soisdisconnecting(so);
1263 		sbflush(&so->so_rcv);
1264 		tp = tcp_usrclosed(tp);
1265 		if (tp)
1266 			(void) tcp_output(tp);
1267 	}
1268 	return tp;
1269 }
1270 
1271 /*
1272  * User issued close, and wish to trail through shutdown states:
1273  * if never received SYN, just forget it.  If got a SYN from peer,
1274  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1275  * If already got a FIN from peer, then almost done; go to LAST_ACK
1276  * state.  In all other cases, have already sent FIN to peer (e.g.
1277  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1278  * for peer to send FIN or not respond to keep-alives, etc.
1279  * We can let the user exit from the close as soon as the FIN is acked.
1280  */
1281 struct tcpcb *
1282 tcp_usrclosed(struct tcpcb *tp)
1283 {
1284 
1285 	switch (tp->t_state) {
1286 
1287 	case TCPS_CLOSED:
1288 	case TCPS_LISTEN:
1289 	case TCPS_SYN_SENT:
1290 		tp->t_state = TCPS_CLOSED;
1291 		tp = tcp_close(tp);
1292 		break;
1293 
1294 	case TCPS_SYN_RECEIVED:
1295 	case TCPS_ESTABLISHED:
1296 		tp->t_state = TCPS_FIN_WAIT_1;
1297 		break;
1298 
1299 	case TCPS_CLOSE_WAIT:
1300 		tp->t_state = TCPS_LAST_ACK;
1301 		break;
1302 	}
1303 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1304 		struct socket *so;
1305 		if (tp->t_inpcb)
1306 			so = tp->t_inpcb->inp_socket;
1307 #ifdef INET6
1308 		else if (tp->t_in6pcb)
1309 			so = tp->t_in6pcb->in6p_socket;
1310 #endif
1311 		else
1312 			so = NULL;
1313 		if (so)
1314 			soisdisconnected(so);
1315 		/*
1316 		 * If we are in FIN_WAIT_2, we arrived here because the
1317 		 * application did a shutdown of the send side.  Like the
1318 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
1319 		 * a full close, we start a timer to make sure sockets are
1320 		 * not left in FIN_WAIT_2 forever.
1321 		 */
1322 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
1323 			TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
1324 		else if (tp->t_state == TCPS_TIME_WAIT
1325 			 && ((tp->t_inpcb
1326 			      && (tcp4_vtw_enable & 1)
1327 			      && vtw_add(AF_INET, tp))
1328 			     ||
1329 			     (tp->t_in6pcb
1330 			      && (tcp6_vtw_enable & 1)
1331 			      && vtw_add(AF_INET6, tp)))) {
1332 			tp = 0;
1333 		}
1334 	}
1335 	return tp;
1336 }
1337 
1338 /*
1339  * sysctl helper routine for net.inet.ip.mssdflt.  it can't be less
1340  * than 32.
1341  */
1342 static int
1343 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS)
1344 {
1345 	int error, mssdflt;
1346 	struct sysctlnode node;
1347 
1348 	mssdflt = tcp_mssdflt;
1349 	node = *rnode;
1350 	node.sysctl_data = &mssdflt;
1351 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1352 	if (error || newp == NULL)
1353 		return error;
1354 
1355 	if (mssdflt < 32)
1356 		return EINVAL;
1357 	tcp_mssdflt = mssdflt;
1358 
1359 	mutex_enter(softnet_lock);
1360 	tcp_tcpcb_template();
1361 	mutex_exit(softnet_lock);
1362 
1363 	return 0;
1364 }
1365 
1366 /*
1367  * sysctl helper for TCP CB template update
1368  */
1369 static int
1370 sysctl_update_tcpcb_template(SYSCTLFN_ARGS)
1371 {
1372 	int t, error;
1373 	struct sysctlnode node;
1374 
1375 	/* follow procedures in sysctl(9) manpage */
1376 	t = *(int *)rnode->sysctl_data;
1377 	node = *rnode;
1378 	node.sysctl_data = &t;
1379 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1380 	if (error || newp == NULL)
1381 		return error;
1382 
1383 	if (t < 0)
1384 		return EINVAL;
1385 
1386 	*(int *)rnode->sysctl_data = t;
1387 
1388 	mutex_enter(softnet_lock);
1389 	tcp_tcpcb_template();
1390 	mutex_exit(softnet_lock);
1391 
1392 	return 0;
1393 }
1394 
1395 /*
1396  * sysctl helper routine for setting port related values under
1397  * net.inet.ip and net.inet6.ip6.  does basic range checking and does
1398  * additional checks for each type.  this code has placed in
1399  * tcp_input.c since INET and INET6 both use the same tcp code.
1400  *
1401  * this helper is not static so that both inet and inet6 can use it.
1402  */
1403 int
1404 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS)
1405 {
1406 	int error, tmp;
1407 	int apmin, apmax;
1408 #ifndef IPNOPRIVPORTS
1409 	int lpmin, lpmax;
1410 #endif /* IPNOPRIVPORTS */
1411 	struct sysctlnode node;
1412 
1413 	if (namelen != 0)
1414 		return EINVAL;
1415 
1416 	switch (name[-3]) {
1417 	    case PF_INET:
1418 		apmin = anonportmin;
1419 		apmax = anonportmax;
1420 #ifndef IPNOPRIVPORTS
1421 		lpmin = lowportmin;
1422 		lpmax = lowportmax;
1423 #endif /* IPNOPRIVPORTS */
1424 		break;
1425 #ifdef INET6
1426 	    case PF_INET6:
1427 		apmin = ip6_anonportmin;
1428 		apmax = ip6_anonportmax;
1429 #ifndef IPNOPRIVPORTS
1430 		lpmin = ip6_lowportmin;
1431 		lpmax = ip6_lowportmax;
1432 #endif /* IPNOPRIVPORTS */
1433 		break;
1434 #endif /* INET6 */
1435 	    default:
1436 		return EINVAL;
1437 	}
1438 
1439 	/*
1440 	 * insert temporary copy into node, perform lookup on
1441 	 * temporary, then restore pointer
1442 	 */
1443 	node = *rnode;
1444 	tmp = *(int*)rnode->sysctl_data;
1445 	node.sysctl_data = &tmp;
1446 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1447 	if (error || newp == NULL)
1448 		return error;
1449 
1450 	/*
1451 	 * simple port range check
1452 	 */
1453 	if (tmp < 0 || tmp > 65535)
1454 		return EINVAL;
1455 
1456 	/*
1457 	 * per-node range checks
1458 	 */
1459 	switch (rnode->sysctl_num) {
1460 	case IPCTL_ANONPORTMIN:
1461 	case IPV6CTL_ANONPORTMIN:
1462 		if (tmp >= apmax)
1463 			return EINVAL;
1464 #ifndef IPNOPRIVPORTS
1465 		if (tmp < IPPORT_RESERVED)
1466                         return EINVAL;
1467 #endif /* IPNOPRIVPORTS */
1468 		break;
1469 
1470 	case IPCTL_ANONPORTMAX:
1471 	case IPV6CTL_ANONPORTMAX:
1472                 if (apmin >= tmp)
1473 			return EINVAL;
1474 #ifndef IPNOPRIVPORTS
1475 		if (tmp < IPPORT_RESERVED)
1476                         return EINVAL;
1477 #endif /* IPNOPRIVPORTS */
1478 		break;
1479 
1480 #ifndef IPNOPRIVPORTS
1481 	case IPCTL_LOWPORTMIN:
1482 	case IPV6CTL_LOWPORTMIN:
1483 		if (tmp >= lpmax ||
1484 		    tmp > IPPORT_RESERVEDMAX ||
1485 		    tmp < IPPORT_RESERVEDMIN)
1486 			return EINVAL;
1487 		break;
1488 
1489 	case IPCTL_LOWPORTMAX:
1490 	case IPV6CTL_LOWPORTMAX:
1491 		if (lpmin >= tmp ||
1492 		    tmp > IPPORT_RESERVEDMAX ||
1493 		    tmp < IPPORT_RESERVEDMIN)
1494 			return EINVAL;
1495 		break;
1496 #endif /* IPNOPRIVPORTS */
1497 
1498 	default:
1499 		return EINVAL;
1500 	}
1501 
1502 	*(int*)rnode->sysctl_data = tmp;
1503 
1504 	return 0;
1505 }
1506 
1507 static inline int
1508 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp)
1509 {
1510 	if (oldp) {
1511 		size_t sz;
1512 		uid_t uid;
1513 		int error;
1514 
1515 		if (sockp->so_cred == NULL)
1516 			return EPERM;
1517 
1518 		uid = kauth_cred_geteuid(sockp->so_cred);
1519 		sz = MIN(sizeof(uid), *oldlenp);
1520 		if ((error = copyout(&uid, oldp, sz)) != 0)
1521 			return error;
1522 	}
1523 	*oldlenp = sizeof(uid_t);
1524 	return 0;
1525 }
1526 
1527 static inline int
1528 inet4_ident_core(struct in_addr raddr, u_int rport,
1529     struct in_addr laddr, u_int lport,
1530     void *oldp, size_t *oldlenp,
1531     struct lwp *l, int dodrop)
1532 {
1533 	struct inpcb *inp;
1534 	struct socket *sockp;
1535 
1536 	inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0);
1537 
1538 	if (inp == NULL || (sockp = inp->inp_socket) == NULL)
1539 		return ESRCH;
1540 
1541 	if (dodrop) {
1542 		struct tcpcb *tp;
1543 		int error;
1544 
1545 		if (inp == NULL || (tp = intotcpcb(inp)) == NULL ||
1546 		    (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0)
1547 			return ESRCH;
1548 
1549 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1550 		    KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL);
1551 		if (error)
1552 			return error;
1553 
1554 		(void)tcp_drop(tp, ECONNABORTED);
1555 		return 0;
1556 	}
1557 
1558 	return copyout_uid(sockp, oldp, oldlenp);
1559 }
1560 
1561 #ifdef INET6
1562 static inline int
1563 inet6_ident_core(struct in6_addr *raddr, u_int rport,
1564     struct in6_addr *laddr, u_int lport,
1565     void *oldp, size_t *oldlenp,
1566     struct lwp *l, int dodrop)
1567 {
1568 	struct in6pcb *in6p;
1569 	struct socket *sockp;
1570 
1571 	in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0);
1572 
1573 	if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL)
1574 		return ESRCH;
1575 
1576 	if (dodrop) {
1577 		struct tcpcb *tp;
1578 		int error;
1579 
1580 		if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL ||
1581 		    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0)
1582 			return ESRCH;
1583 
1584 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1585 		    KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL);
1586 		if (error)
1587 			return error;
1588 
1589 		(void)tcp_drop(tp, ECONNABORTED);
1590 		return 0;
1591 	}
1592 
1593 	return copyout_uid(sockp, oldp, oldlenp);
1594 }
1595 #endif
1596 
1597 /*
1598  * sysctl helper routine for the net.inet.tcp.drop and
1599  * net.inet6.tcp6.drop nodes.
1600  */
1601 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident
1602 
1603 /*
1604  * sysctl helper routine for the net.inet.tcp.ident and
1605  * net.inet6.tcp6.ident nodes.  contains backwards compat code for the
1606  * old way of looking up the ident information for ipv4 which involves
1607  * stuffing the port/addr pairs into the mib lookup.
1608  */
1609 static int
1610 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS)
1611 {
1612 	struct sockaddr_in *si4[2];
1613 #ifdef INET6
1614 	struct sockaddr_in6 *si6[2];
1615 #endif
1616 	struct sockaddr_storage sa[2];
1617 	int error, pf, dodrop;
1618 
1619 	dodrop = name[-1] == TCPCTL_DROP;
1620 	if (dodrop) {
1621 		if (oldp != NULL || *oldlenp != 0)
1622 			return EINVAL;
1623 		if (newp == NULL)
1624 			return EPERM;
1625 		if (newlen < sizeof(sa))
1626 			return ENOMEM;
1627 	}
1628 	if (namelen != 4 && namelen != 0)
1629 		return EINVAL;
1630 	if (name[-2] != IPPROTO_TCP)
1631 		return EINVAL;
1632 	pf = name[-3];
1633 
1634 	/* old style lookup, ipv4 only */
1635 	if (namelen == 4) {
1636 		struct in_addr laddr, raddr;
1637 		u_int lport, rport;
1638 
1639 		if (pf != PF_INET)
1640 			return EPROTONOSUPPORT;
1641 		raddr.s_addr = (uint32_t)name[0];
1642 		rport = (u_int)name[1];
1643 		laddr.s_addr = (uint32_t)name[2];
1644 		lport = (u_int)name[3];
1645 
1646 		mutex_enter(softnet_lock);
1647 		error = inet4_ident_core(raddr, rport, laddr, lport,
1648 		    oldp, oldlenp, l, dodrop);
1649 		mutex_exit(softnet_lock);
1650 		return error;
1651 	}
1652 
1653 	if (newp == NULL || newlen != sizeof(sa))
1654 		return EINVAL;
1655 	error = copyin(newp, &sa, newlen);
1656 	if (error)
1657 		return error;
1658 
1659 	/*
1660 	 * requested families must match
1661 	 */
1662 	if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family)
1663 		return EINVAL;
1664 
1665 	switch (pf) {
1666 #ifdef INET6
1667 	case PF_INET6:
1668 		si6[0] = (struct sockaddr_in6*)&sa[0];
1669 		si6[1] = (struct sockaddr_in6*)&sa[1];
1670 		if (si6[0]->sin6_len != sizeof(*si6[0]) ||
1671 		    si6[1]->sin6_len != sizeof(*si6[1]))
1672 			return EINVAL;
1673 
1674 		if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) &&
1675 		    !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) {
1676 			error = sa6_embedscope(si6[0], ip6_use_defzone);
1677 			if (error)
1678 				return error;
1679 			error = sa6_embedscope(si6[1], ip6_use_defzone);
1680 			if (error)
1681 				return error;
1682 
1683 			mutex_enter(softnet_lock);
1684 			error = inet6_ident_core(&si6[0]->sin6_addr,
1685 			    si6[0]->sin6_port, &si6[1]->sin6_addr,
1686 			    si6[1]->sin6_port, oldp, oldlenp, l, dodrop);
1687 			mutex_exit(softnet_lock);
1688 			return error;
1689 		}
1690 
1691 		if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) !=
1692 		    IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr))
1693 			return EINVAL;
1694 
1695 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]);
1696 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]);
1697 #endif /* INET6 */
1698 		/*FALLTHROUGH*/
1699 	case PF_INET:
1700 		si4[0] = (struct sockaddr_in*)&sa[0];
1701 		si4[1] = (struct sockaddr_in*)&sa[1];
1702 		if (si4[0]->sin_len != sizeof(*si4[0]) ||
1703 		    si4[0]->sin_len != sizeof(*si4[1]))
1704 			return EINVAL;
1705 
1706 		mutex_enter(softnet_lock);
1707 		error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port,
1708 		    si4[1]->sin_addr, si4[1]->sin_port,
1709 		    oldp, oldlenp, l, dodrop);
1710 		mutex_exit(softnet_lock);
1711 		return error;
1712 	default:
1713 		return EPROTONOSUPPORT;
1714 	}
1715 }
1716 
1717 /*
1718  * sysctl helper for the inet and inet6 pcblists.  handles tcp/udp and
1719  * inet/inet6, as well as raw pcbs for each.  specifically not
1720  * declared static so that raw sockets and udp/udp6 can use it as
1721  * well.
1722  */
1723 int
1724 sysctl_inpcblist(SYSCTLFN_ARGS)
1725 {
1726 	const bool allowaddr = get_expose_address(curproc);
1727 	struct sockaddr_in *in;
1728 	const struct inpcb *inp;
1729 #ifdef INET6
1730 	struct sockaddr_in6 *in6;
1731 	const struct in6pcb *in6p;
1732 #endif
1733 	struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data);
1734 	const struct inpcb_hdr *inph;
1735 	struct tcpcb *tp;
1736 	struct kinfo_pcb pcb;
1737 	char *dp;
1738 	size_t len, needed, elem_size, out_size;
1739 	int error, elem_count, pf, proto, pf2;
1740 
1741 	if (namelen != 4)
1742 		return EINVAL;
1743 
1744 	if (oldp != NULL) {
1745 		    len = *oldlenp;
1746 		    elem_size = name[2];
1747 		    elem_count = name[3];
1748 		    if (elem_size != sizeof(pcb))
1749 			    return EINVAL;
1750 	} else {
1751 		    len = 0;
1752 		    elem_count = INT_MAX;
1753 		    elem_size = sizeof(pcb);
1754 	}
1755 	error = 0;
1756 	dp = oldp;
1757 	out_size = elem_size;
1758 	needed = 0;
1759 
1760 	if (namelen == 1 && name[0] == CTL_QUERY)
1761 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1762 
1763 	if (name - oname != 4)
1764 		return EINVAL;
1765 
1766 	pf = oname[1];
1767 	proto = oname[2];
1768 	pf2 = (oldp != NULL) ? pf : 0;
1769 
1770 	mutex_enter(softnet_lock);
1771 
1772 	TAILQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) {
1773 		inp = (const struct inpcb *)inph;
1774 #ifdef INET6
1775 		in6p = (const struct in6pcb *)inph;
1776 #endif
1777 
1778 		if (inph->inph_af != pf)
1779 			continue;
1780 
1781 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1782 		    KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL,
1783 		    NULL) != 0)
1784 			continue;
1785 
1786 		memset(&pcb, 0, sizeof(pcb));
1787 
1788 		pcb.ki_family = pf;
1789 		pcb.ki_type = proto;
1790 
1791 		switch (pf2) {
1792 		case 0:
1793 			/* just probing for size */
1794 			break;
1795 		case PF_INET:
1796 			pcb.ki_family = inp->inp_socket->so_proto->
1797 			    pr_domain->dom_family;
1798 			pcb.ki_type = inp->inp_socket->so_proto->
1799 			    pr_type;
1800 			pcb.ki_protocol = inp->inp_socket->so_proto->
1801 			    pr_protocol;
1802 			pcb.ki_pflags = inp->inp_flags;
1803 
1804 			pcb.ki_sostate = inp->inp_socket->so_state;
1805 			pcb.ki_prstate = inp->inp_state;
1806 			if (proto == IPPROTO_TCP) {
1807 				tp = intotcpcb(inp);
1808 				pcb.ki_tstate = tp->t_state;
1809 				pcb.ki_tflags = tp->t_flags;
1810 			}
1811 
1812 			COND_SET_VALUE(pcb.ki_pcbaddr,
1813 			    PTRTOUINT64(inp), allowaddr);
1814 			COND_SET_VALUE(pcb.ki_ppcbaddr,
1815 			    PTRTOUINT64(inp->inp_ppcb), allowaddr);
1816 			COND_SET_VALUE(pcb.ki_sockaddr,
1817 			    PTRTOUINT64(inp->inp_socket), allowaddr);
1818 
1819 			pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc;
1820 			pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc;
1821 
1822 			in = satosin(&pcb.ki_src);
1823 			in->sin_len = sizeof(*in);
1824 			in->sin_family = pf;
1825 			in->sin_port = inp->inp_lport;
1826 			in->sin_addr = inp->inp_laddr;
1827 			if (pcb.ki_prstate >= INP_CONNECTED) {
1828 				in = satosin(&pcb.ki_dst);
1829 				in->sin_len = sizeof(*in);
1830 				in->sin_family = pf;
1831 				in->sin_port = inp->inp_fport;
1832 				in->sin_addr = inp->inp_faddr;
1833 			}
1834 			break;
1835 #ifdef INET6
1836 		case PF_INET6:
1837 			pcb.ki_family = in6p->in6p_socket->so_proto->
1838 			    pr_domain->dom_family;
1839 			pcb.ki_type = in6p->in6p_socket->so_proto->pr_type;
1840 			pcb.ki_protocol = in6p->in6p_socket->so_proto->
1841 			    pr_protocol;
1842 			pcb.ki_pflags = in6p->in6p_flags;
1843 
1844 			pcb.ki_sostate = in6p->in6p_socket->so_state;
1845 			pcb.ki_prstate = in6p->in6p_state;
1846 			if (proto == IPPROTO_TCP) {
1847 				tp = in6totcpcb(in6p);
1848 				pcb.ki_tstate = tp->t_state;
1849 				pcb.ki_tflags = tp->t_flags;
1850 			}
1851 
1852 			COND_SET_VALUE(pcb.ki_pcbaddr,
1853 			    PTRTOUINT64(in6p), allowaddr);
1854 			COND_SET_VALUE(pcb.ki_ppcbaddr,
1855 			    PTRTOUINT64(in6p->in6p_ppcb), allowaddr);
1856 			COND_SET_VALUE(pcb.ki_sockaddr,
1857 			    PTRTOUINT64(in6p->in6p_socket), allowaddr);
1858 
1859 			pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc;
1860 			pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc;
1861 
1862 			in6 = satosin6(&pcb.ki_src);
1863 			in6->sin6_len = sizeof(*in6);
1864 			in6->sin6_family = pf;
1865 			in6->sin6_port = in6p->in6p_lport;
1866 			in6->sin6_flowinfo = in6p->in6p_flowinfo;
1867 			in6->sin6_addr = in6p->in6p_laddr;
1868 			in6->sin6_scope_id = 0; /* XXX? */
1869 
1870 			if (pcb.ki_prstate >= IN6P_CONNECTED) {
1871 				in6 = satosin6(&pcb.ki_dst);
1872 				in6->sin6_len = sizeof(*in6);
1873 				in6->sin6_family = pf;
1874 				in6->sin6_port = in6p->in6p_fport;
1875 				in6->sin6_flowinfo = in6p->in6p_flowinfo;
1876 				in6->sin6_addr = in6p->in6p_faddr;
1877 				in6->sin6_scope_id = 0; /* XXX? */
1878 			}
1879 			break;
1880 #endif
1881 		}
1882 
1883 		if (len >= elem_size && elem_count > 0) {
1884 			error = copyout(&pcb, dp, out_size);
1885 			if (error) {
1886 				mutex_exit(softnet_lock);
1887 				return error;
1888 			}
1889 			dp += elem_size;
1890 			len -= elem_size;
1891 		}
1892 		needed += elem_size;
1893 		if (elem_count > 0 && elem_count != INT_MAX)
1894 			elem_count--;
1895 	}
1896 
1897 	*oldlenp = needed;
1898 	if (oldp == NULL)
1899 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
1900 
1901 	mutex_exit(softnet_lock);
1902 
1903 	return error;
1904 }
1905 
1906 static int
1907 sysctl_tcp_congctl(SYSCTLFN_ARGS)
1908 {
1909 	struct sysctlnode node;
1910 	int error;
1911 	char newname[TCPCC_MAXLEN];
1912 
1913 	strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1);
1914 
1915 	node = *rnode;
1916 	node.sysctl_data = newname;
1917 	node.sysctl_size = sizeof(newname);
1918 
1919 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1920 
1921 	if (error ||
1922 	    newp == NULL ||
1923 	    strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0)
1924 		return error;
1925 
1926 	mutex_enter(softnet_lock);
1927 	error = tcp_congctl_select(NULL, newname);
1928 	mutex_exit(softnet_lock);
1929 
1930 	return error;
1931 }
1932 
1933 static int
1934 sysctl_tcp_init_win(SYSCTLFN_ARGS)
1935 {
1936 	int error;
1937 	u_int iw;
1938 	struct sysctlnode node;
1939 
1940 	iw = *(u_int *)rnode->sysctl_data;
1941 	node = *rnode;
1942 	node.sysctl_data = &iw;
1943 	node.sysctl_size = sizeof(iw);
1944 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1945 	if (error || newp == NULL)
1946 		return error;
1947 
1948 	if (iw >= __arraycount(tcp_init_win_max))
1949 		return EINVAL;
1950 	*(u_int *)rnode->sysctl_data = iw;
1951 	return 0;
1952 }
1953 
1954 static int
1955 sysctl_tcp_keep(SYSCTLFN_ARGS)
1956 {
1957 	int error;
1958 	u_int tmp;
1959 	struct sysctlnode node;
1960 
1961 	node = *rnode;
1962 	tmp = *(u_int *)rnode->sysctl_data;
1963 	node.sysctl_data = &tmp;
1964 
1965 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1966 	if (error || newp == NULL)
1967 		return error;
1968 
1969 	if (!(tmp > 0 && tmp <= TCP_TIMER_MAXTICKS))
1970 		return EINVAL;
1971 
1972 	mutex_enter(softnet_lock);
1973 
1974 	*(u_int *)rnode->sysctl_data = tmp;
1975 	tcp_tcpcb_template();	/* update the template */
1976 
1977 	mutex_exit(softnet_lock);
1978 	return 0;
1979 }
1980 
1981 static int
1982 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS)
1983 {
1984 
1985 	return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS));
1986 }
1987 
1988 /*
1989  * this (second stage) setup routine is a replacement for tcp_sysctl()
1990  * (which is currently used for ipv4 and ipv6)
1991  */
1992 static void
1993 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
1994 			   const char *tcpname)
1995 {
1996 	const struct sysctlnode *sack_node;
1997 	const struct sysctlnode *abc_node;
1998 	const struct sysctlnode *ecn_node;
1999 	const struct sysctlnode *congctl_node;
2000 	const struct sysctlnode *mslt_node;
2001 	const struct sysctlnode *vtw_node;
2002 #ifdef TCP_DEBUG
2003 	extern struct tcp_debug tcp_debug[TCP_NDEBUG];
2004 	extern int tcp_debx;
2005 #endif
2006 
2007 	sysctl_createv(clog, 0, NULL, NULL,
2008 		       CTLFLAG_PERMANENT,
2009 		       CTLTYPE_NODE, pfname, NULL,
2010 		       NULL, 0, NULL, 0,
2011 		       CTL_NET, pf, CTL_EOL);
2012 	sysctl_createv(clog, 0, NULL, NULL,
2013 		       CTLFLAG_PERMANENT,
2014 		       CTLTYPE_NODE, tcpname,
2015 		       SYSCTL_DESCR("TCP related settings"),
2016 		       NULL, 0, NULL, 0,
2017 		       CTL_NET, pf, IPPROTO_TCP, CTL_EOL);
2018 
2019 	sysctl_createv(clog, 0, NULL, NULL,
2020 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2021 		       CTLTYPE_INT, "rfc1323",
2022 		       SYSCTL_DESCR("Enable RFC1323 TCP extensions"),
2023 		       sysctl_update_tcpcb_template, 0, &tcp_do_rfc1323, 0,
2024 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL);
2025 	sysctl_createv(clog, 0, NULL, NULL,
2026 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2027 		       CTLTYPE_INT, "sendspace",
2028 		       SYSCTL_DESCR("Default TCP send buffer size"),
2029 		       NULL, 0, &tcp_sendspace, 0,
2030 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL);
2031 	sysctl_createv(clog, 0, NULL, NULL,
2032 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2033 		       CTLTYPE_INT, "recvspace",
2034 		       SYSCTL_DESCR("Default TCP receive buffer size"),
2035 		       NULL, 0, &tcp_recvspace, 0,
2036 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL);
2037 	sysctl_createv(clog, 0, NULL, NULL,
2038 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2039 		       CTLTYPE_INT, "mssdflt",
2040 		       SYSCTL_DESCR("Default maximum segment size"),
2041 		       sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0,
2042 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL);
2043 	sysctl_createv(clog, 0, NULL, NULL,
2044 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2045 		       CTLTYPE_INT, "minmss",
2046 		       SYSCTL_DESCR("Lower limit for TCP maximum segment size"),
2047 		       NULL, 0, &tcp_minmss, 0,
2048 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2049 	sysctl_createv(clog, 0, NULL, NULL,
2050 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2051 		       CTLTYPE_INT, "msl",
2052 		       SYSCTL_DESCR("Maximum Segment Life"),
2053 		       NULL, 0, &tcp_msl, 0,
2054 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL);
2055 	sysctl_createv(clog, 0, NULL, NULL,
2056 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2057 		       CTLTYPE_INT, "syn_cache_limit",
2058 		       SYSCTL_DESCR("Maximum number of entries in the TCP "
2059 				    "compressed state engine"),
2060 		       NULL, 0, &tcp_syn_cache_limit, 0,
2061 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT,
2062 		       CTL_EOL);
2063 	sysctl_createv(clog, 0, NULL, NULL,
2064 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2065 		       CTLTYPE_INT, "syn_bucket_limit",
2066 		       SYSCTL_DESCR("Maximum number of entries per hash "
2067 				    "bucket in the TCP compressed state "
2068 				    "engine"),
2069 		       NULL, 0, &tcp_syn_bucket_limit, 0,
2070 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT,
2071 		       CTL_EOL);
2072 #if 0 /* obsoleted */
2073 	sysctl_createv(clog, 0, NULL, NULL,
2074 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2075 		       CTLTYPE_INT, "syn_cache_interval",
2076 		       SYSCTL_DESCR("TCP compressed state engine's timer interval"),
2077 		       NULL, 0, &tcp_syn_cache_interval, 0,
2078 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER,
2079 		       CTL_EOL);
2080 #endif
2081 	sysctl_createv(clog, 0, NULL, NULL,
2082 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2083 		       CTLTYPE_INT, "init_win",
2084 		       SYSCTL_DESCR("Initial TCP congestion window"),
2085 		       sysctl_tcp_init_win, 0, &tcp_init_win, 0,
2086 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL);
2087 	sysctl_createv(clog, 0, NULL, NULL,
2088 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2089 		       CTLTYPE_INT, "mss_ifmtu",
2090 		       SYSCTL_DESCR("Use interface MTU for calculating MSS"),
2091 		       NULL, 0, &tcp_mss_ifmtu, 0,
2092 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL);
2093 	sysctl_createv(clog, 0, NULL, &sack_node,
2094 		       CTLFLAG_PERMANENT,
2095 		       CTLTYPE_NODE, "sack",
2096 		       SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
2097 		       NULL, 0, NULL, 0,
2098 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
2099 
2100 	/* Congctl subtree */
2101 	sysctl_createv(clog, 0, NULL, &congctl_node,
2102 		       CTLFLAG_PERMANENT,
2103 		       CTLTYPE_NODE, "congctl",
2104 		       SYSCTL_DESCR("TCP Congestion Control"),
2105 	    	       NULL, 0, NULL, 0,
2106 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2107 	sysctl_createv(clog, 0, &congctl_node, NULL,
2108 		       CTLFLAG_PERMANENT,
2109 		       CTLTYPE_STRING, "available",
2110 		       SYSCTL_DESCR("Available Congestion Control Mechanisms"),
2111 		       NULL, 0, tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL);
2112 	sysctl_createv(clog, 0, &congctl_node, NULL,
2113 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2114 		       CTLTYPE_STRING, "selected",
2115 		       SYSCTL_DESCR("Selected Congestion Control Mechanism"),
2116 		       sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN,
2117 		       CTL_CREATE, CTL_EOL);
2118 
2119 	sysctl_createv(clog, 0, NULL, NULL,
2120 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2121 		       CTLTYPE_INT, "win_scale",
2122 		       SYSCTL_DESCR("Use RFC1323 window scale options"),
2123 		       sysctl_update_tcpcb_template, 0, &tcp_do_win_scale, 0,
2124 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL);
2125 	sysctl_createv(clog, 0, NULL, NULL,
2126 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2127 		       CTLTYPE_INT, "timestamps",
2128 		       SYSCTL_DESCR("Use RFC1323 time stamp options"),
2129 		       sysctl_update_tcpcb_template, 0, &tcp_do_timestamps, 0,
2130 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL);
2131 	sysctl_createv(clog, 0, NULL, NULL,
2132 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2133 		       CTLTYPE_INT, "cwm",
2134 		       SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window "
2135 				    "Monitoring"),
2136 		       NULL, 0, &tcp_cwm, 0,
2137 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL);
2138 	sysctl_createv(clog, 0, NULL, NULL,
2139 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2140 		       CTLTYPE_INT, "cwm_burstsize",
2141 		       SYSCTL_DESCR("Congestion Window Monitoring allowed "
2142 				    "burst count in packets"),
2143 		       NULL, 0, &tcp_cwm_burstsize, 0,
2144 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE,
2145 		       CTL_EOL);
2146 	sysctl_createv(clog, 0, NULL, NULL,
2147 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2148 		       CTLTYPE_INT, "ack_on_push",
2149 		       SYSCTL_DESCR("Immediately return ACK when PSH is "
2150 				    "received"),
2151 		       NULL, 0, &tcp_ack_on_push, 0,
2152 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL);
2153 	sysctl_createv(clog, 0, NULL, NULL,
2154 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2155 		       CTLTYPE_INT, "keepidle",
2156 		       SYSCTL_DESCR("Allowed connection idle ticks before a "
2157 				    "keepalive probe is sent"),
2158 		       sysctl_tcp_keep, 0, &tcp_keepidle, 0,
2159 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL);
2160 	sysctl_createv(clog, 0, NULL, NULL,
2161 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2162 		       CTLTYPE_INT, "keepintvl",
2163 		       SYSCTL_DESCR("Ticks before next keepalive probe is sent"),
2164 		       sysctl_tcp_keep, 0, &tcp_keepintvl, 0,
2165 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL);
2166 	sysctl_createv(clog, 0, NULL, NULL,
2167 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2168 		       CTLTYPE_INT, "keepcnt",
2169 		       SYSCTL_DESCR("Number of keepalive probes to send"),
2170 		       sysctl_tcp_keep, 0, &tcp_keepcnt, 0,
2171 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL);
2172 	sysctl_createv(clog, 0, NULL, NULL,
2173 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
2174 		       CTLTYPE_INT, "slowhz",
2175 		       SYSCTL_DESCR("Keepalive ticks per second"),
2176 		       NULL, PR_SLOWHZ, NULL, 0,
2177 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL);
2178 	sysctl_createv(clog, 0, NULL, NULL,
2179 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2180 		       CTLTYPE_INT, "log_refused",
2181 		       SYSCTL_DESCR("Log refused TCP connections"),
2182 		       NULL, 0, &tcp_log_refused, 0,
2183 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL);
2184 #if 0 /* obsoleted */
2185 	sysctl_createv(clog, 0, NULL, NULL,
2186 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2187 		       CTLTYPE_INT, "rstratelimit", NULL,
2188 		       NULL, 0, &tcp_rst_ratelim, 0,
2189 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL);
2190 #endif
2191 	sysctl_createv(clog, 0, NULL, NULL,
2192 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2193 		       CTLTYPE_INT, "rstppslimit",
2194 		       SYSCTL_DESCR("Maximum number of RST packets to send "
2195 				    "per second"),
2196 		       NULL, 0, &tcp_rst_ppslim, 0,
2197 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL);
2198 	sysctl_createv(clog, 0, NULL, NULL,
2199 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2200 		       CTLTYPE_INT, "delack_ticks",
2201 		       SYSCTL_DESCR("Number of ticks to delay sending an ACK"),
2202 		       NULL, 0, &tcp_delack_ticks, 0,
2203 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL);
2204 	sysctl_createv(clog, 0, NULL, NULL,
2205 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2206 		       CTLTYPE_INT, "init_win_local",
2207 		       SYSCTL_DESCR("Initial TCP window size (in segments)"),
2208 		       sysctl_tcp_init_win, 0, &tcp_init_win_local, 0,
2209 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL,
2210 		       CTL_EOL);
2211 	sysctl_createv(clog, 0, NULL, NULL,
2212 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2213 		       CTLTYPE_STRUCT, "ident",
2214 		       SYSCTL_DESCR("RFC1413 Identification Protocol lookups"),
2215 		       sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t),
2216 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL);
2217 	sysctl_createv(clog, 0, NULL, NULL,
2218 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2219 		       CTLTYPE_INT, "do_loopback_cksum",
2220 		       SYSCTL_DESCR("Perform TCP checksum on loopback"),
2221 		       NULL, 0, &tcp_do_loopback_cksum, 0,
2222 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM,
2223 		       CTL_EOL);
2224 	sysctl_createv(clog, 0, NULL, NULL,
2225 		       CTLFLAG_PERMANENT,
2226 		       CTLTYPE_STRUCT, "pcblist",
2227 		       SYSCTL_DESCR("TCP protocol control block list"),
2228 		       sysctl_inpcblist, 0, &tcbtable, 0,
2229 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2230 		       CTL_EOL);
2231 	sysctl_createv(clog, 0, NULL, NULL,
2232 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2233 		       CTLTYPE_INT, "keepinit",
2234 		       SYSCTL_DESCR("Ticks before initial tcp connection times out"),
2235 		       sysctl_tcp_keep, 0, &tcp_keepinit, 0,
2236 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2237 
2238 	/* TCP socket buffers auto-sizing nodes */
2239 	sysctl_createv(clog, 0, NULL, NULL,
2240 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2241 		       CTLTYPE_INT, "recvbuf_auto",
2242 		       SYSCTL_DESCR("Enable automatic receive "
2243 		           "buffer sizing (experimental)"),
2244 		       NULL, 0, &tcp_do_autorcvbuf, 0,
2245 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2246 	sysctl_createv(clog, 0, NULL, NULL,
2247 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2248 		       CTLTYPE_INT, "recvbuf_inc",
2249 		       SYSCTL_DESCR("Incrementor step size of "
2250 		           "automatic receive buffer"),
2251 		       NULL, 0, &tcp_autorcvbuf_inc, 0,
2252 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2253 	sysctl_createv(clog, 0, NULL, NULL,
2254 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2255 		       CTLTYPE_INT, "recvbuf_max",
2256 		       SYSCTL_DESCR("Max size of automatic receive buffer"),
2257 		       NULL, 0, &tcp_autorcvbuf_max, 0,
2258 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2259 
2260 	sysctl_createv(clog, 0, NULL, NULL,
2261 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2262 		       CTLTYPE_INT, "sendbuf_auto",
2263 		       SYSCTL_DESCR("Enable automatic send "
2264 		           "buffer sizing (experimental)"),
2265 		       NULL, 0, &tcp_do_autosndbuf, 0,
2266 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2267 	sysctl_createv(clog, 0, NULL, NULL,
2268 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2269 		       CTLTYPE_INT, "sendbuf_inc",
2270 		       SYSCTL_DESCR("Incrementor step size of "
2271 		           "automatic send buffer"),
2272 		       NULL, 0, &tcp_autosndbuf_inc, 0,
2273 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2274 	sysctl_createv(clog, 0, NULL, NULL,
2275 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2276 		       CTLTYPE_INT, "sendbuf_max",
2277 		       SYSCTL_DESCR("Max size of automatic send buffer"),
2278 		       NULL, 0, &tcp_autosndbuf_max, 0,
2279 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2280 
2281 	/* ECN subtree */
2282 	sysctl_createv(clog, 0, NULL, &ecn_node,
2283 	    	       CTLFLAG_PERMANENT,
2284 		       CTLTYPE_NODE, "ecn",
2285 	    	       SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
2286 	    	       NULL, 0, NULL, 0,
2287 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2288 	sysctl_createv(clog, 0, &ecn_node, NULL,
2289 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2290 		       CTLTYPE_INT, "enable",
2291 		       SYSCTL_DESCR("Enable TCP Explicit Congestion "
2292 			   "Notification"),
2293 	    	       NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL);
2294 	sysctl_createv(clog, 0, &ecn_node, NULL,
2295 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2296 		       CTLTYPE_INT, "maxretries",
2297 		       SYSCTL_DESCR("Number of times to retry ECN setup "
2298 			       "before disabling ECN on the connection"),
2299 	    	       NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL);
2300 
2301 	/* SACK gets its own little subtree. */
2302 	sysctl_createv(clog, 0, NULL, &sack_node,
2303 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2304 		       CTLTYPE_INT, "enable",
2305 		       SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"),
2306 		       NULL, 0, &tcp_do_sack, 0,
2307 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2308 	sysctl_createv(clog, 0, NULL, &sack_node,
2309 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2310 		       CTLTYPE_INT, "maxholes",
2311 		       SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"),
2312 		       NULL, 0, &tcp_sack_tp_maxholes, 0,
2313 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2314 	sysctl_createv(clog, 0, NULL, &sack_node,
2315 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2316 		       CTLTYPE_INT, "globalmaxholes",
2317 		       SYSCTL_DESCR("Global maximum number of TCP SACK holes"),
2318 		       NULL, 0, &tcp_sack_globalmaxholes, 0,
2319 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2320 	sysctl_createv(clog, 0, NULL, &sack_node,
2321 		       CTLFLAG_PERMANENT,
2322 		       CTLTYPE_INT, "globalholes",
2323 		       SYSCTL_DESCR("Global number of TCP SACK holes"),
2324 		       NULL, 0, &tcp_sack_globalholes, 0,
2325 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2326 
2327 	sysctl_createv(clog, 0, NULL, NULL,
2328 		       CTLFLAG_PERMANENT,
2329 		       CTLTYPE_STRUCT, "stats",
2330 		       SYSCTL_DESCR("TCP statistics"),
2331 		       sysctl_net_inet_tcp_stats, 0, NULL, 0,
2332 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS,
2333 		       CTL_EOL);
2334         sysctl_createv(clog, 0, NULL, NULL,
2335                        CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2336                        CTLTYPE_INT, "local_by_rtt",
2337                        SYSCTL_DESCR("Use RTT estimator to decide which hosts "
2338 				    "are local"),
2339 		       NULL, 0, &tcp_rttlocal, 0,
2340 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2341 #ifdef TCP_DEBUG
2342 	sysctl_createv(clog, 0, NULL, NULL,
2343 		       CTLFLAG_PERMANENT,
2344 		       CTLTYPE_STRUCT, "debug",
2345 		       SYSCTL_DESCR("TCP sockets debug information"),
2346 		       NULL, 0, &tcp_debug, sizeof(tcp_debug),
2347 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG,
2348 		       CTL_EOL);
2349 	sysctl_createv(clog, 0, NULL, NULL,
2350 		       CTLFLAG_PERMANENT,
2351 		       CTLTYPE_INT, "debx",
2352 		       SYSCTL_DESCR("Number of TCP debug sockets messages"),
2353 		       NULL, 0, &tcp_debx, sizeof(tcp_debx),
2354 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX,
2355 		       CTL_EOL);
2356 #endif
2357 	sysctl_createv(clog, 0, NULL, NULL,
2358 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2359 		       CTLTYPE_STRUCT, "drop",
2360 		       SYSCTL_DESCR("TCP drop connection"),
2361 		       sysctl_net_inet_tcp_drop, 0, NULL, 0,
2362 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL);
2363 	sysctl_createv(clog, 0, NULL, NULL,
2364 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2365 		       CTLTYPE_INT, "iss_hash",
2366 		       SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic "
2367 				    "hash computation"),
2368 		       NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948),
2369 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2370 		       CTL_EOL);
2371 
2372 	/* ABC subtree */
2373 
2374 	sysctl_createv(clog, 0, NULL, &abc_node,
2375 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc",
2376 		       SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"),
2377 		       NULL, 0, NULL, 0,
2378 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2379 	sysctl_createv(clog, 0, &abc_node, NULL,
2380 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2381 		       CTLTYPE_INT, "enable",
2382 		       SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"),
2383 		       NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL);
2384 	sysctl_createv(clog, 0, &abc_node, NULL,
2385 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2386 		       CTLTYPE_INT, "aggressive",
2387 		       SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"),
2388 		       NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL);
2389 
2390 	/* MSL tuning subtree */
2391 
2392 	sysctl_createv(clog, 0, NULL, &mslt_node,
2393 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt",
2394 		       SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"),
2395 		       NULL, 0, NULL, 0,
2396 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2397 	sysctl_createv(clog, 0, &mslt_node, NULL,
2398 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2399 		       CTLTYPE_INT, "enable",
2400 		       SYSCTL_DESCR("Enable TIME_WAIT truncation"),
2401 		       NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL);
2402 	sysctl_createv(clog, 0, &mslt_node, NULL,
2403 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2404 		       CTLTYPE_INT, "loopback",
2405 		       SYSCTL_DESCR("MSL value to use for loopback connections"),
2406 		       NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL);
2407 	sysctl_createv(clog, 0, &mslt_node, NULL,
2408 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2409 		       CTLTYPE_INT, "local",
2410 		       SYSCTL_DESCR("MSL value to use for local connections"),
2411 		       NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL);
2412 	sysctl_createv(clog, 0, &mslt_node, NULL,
2413 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2414 		       CTLTYPE_INT, "remote",
2415 		       SYSCTL_DESCR("MSL value to use for remote connections"),
2416 		       NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL);
2417 	sysctl_createv(clog, 0, &mslt_node, NULL,
2418 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2419 		       CTLTYPE_INT, "remote_threshold",
2420 		       SYSCTL_DESCR("RTT estimate value to promote local to remote"),
2421 		       NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL);
2422 
2423 	/* vestigial TIME_WAIT tuning subtree */
2424 
2425 	sysctl_createv(clog, 0, NULL, &vtw_node,
2426 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw",
2427 		       SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"),
2428 		       NULL, 0, NULL, 0,
2429 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2430 	sysctl_createv(clog, 0, &vtw_node, NULL,
2431 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2432 		       CTLTYPE_INT, "enable",
2433 		       SYSCTL_DESCR("Enable Vestigial TIME_WAIT"),
2434 		       sysctl_tcp_vtw_enable, 0,
2435 	               (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable,
2436 		       0, CTL_CREATE, CTL_EOL);
2437 	sysctl_createv(clog, 0, &vtw_node, NULL,
2438 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY,
2439 		       CTLTYPE_INT, "entries",
2440 		       SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"),
2441 		       NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL);
2442 }
2443 
2444 void
2445 tcp_usrreq_init(void)
2446 {
2447 
2448 	sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp");
2449 #ifdef INET6
2450 	sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6");
2451 #endif
2452 }
2453 
2454 PR_WRAP_USRREQS(tcp)
2455 #define	tcp_attach	tcp_attach_wrapper
2456 #define	tcp_detach	tcp_detach_wrapper
2457 #define	tcp_accept	tcp_accept_wrapper
2458 #define	tcp_bind	tcp_bind_wrapper
2459 #define	tcp_listen	tcp_listen_wrapper
2460 #define	tcp_connect	tcp_connect_wrapper
2461 #define	tcp_connect2	tcp_connect2_wrapper
2462 #define	tcp_disconnect	tcp_disconnect_wrapper
2463 #define	tcp_shutdown	tcp_shutdown_wrapper
2464 #define	tcp_abort	tcp_abort_wrapper
2465 #define	tcp_ioctl	tcp_ioctl_wrapper
2466 #define	tcp_stat	tcp_stat_wrapper
2467 #define	tcp_peeraddr	tcp_peeraddr_wrapper
2468 #define	tcp_sockaddr	tcp_sockaddr_wrapper
2469 #define	tcp_rcvd	tcp_rcvd_wrapper
2470 #define	tcp_recvoob	tcp_recvoob_wrapper
2471 #define	tcp_send	tcp_send_wrapper
2472 #define	tcp_sendoob	tcp_sendoob_wrapper
2473 #define	tcp_purgeif	tcp_purgeif_wrapper
2474 
2475 const struct pr_usrreqs tcp_usrreqs = {
2476 	.pr_attach	= tcp_attach,
2477 	.pr_detach	= tcp_detach,
2478 	.pr_accept	= tcp_accept,
2479 	.pr_bind	= tcp_bind,
2480 	.pr_listen	= tcp_listen,
2481 	.pr_connect	= tcp_connect,
2482 	.pr_connect2	= tcp_connect2,
2483 	.pr_disconnect	= tcp_disconnect,
2484 	.pr_shutdown	= tcp_shutdown,
2485 	.pr_abort	= tcp_abort,
2486 	.pr_ioctl	= tcp_ioctl,
2487 	.pr_stat	= tcp_stat,
2488 	.pr_peeraddr	= tcp_peeraddr,
2489 	.pr_sockaddr	= tcp_sockaddr,
2490 	.pr_rcvd	= tcp_rcvd,
2491 	.pr_recvoob	= tcp_recvoob,
2492 	.pr_send	= tcp_send,
2493 	.pr_sendoob	= tcp_sendoob,
2494 	.pr_purgeif	= tcp_purgeif,
2495 };
2496