xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision f89f6560d453f5e37386cc7938c072d2f528b9fa)
1 /*	$NetBSD: tcp_usrreq.c,v 1.205 2015/04/03 20:01:07 rtr 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 1.205 2015/04/03 20:01:07 rtr Exp $");
103 
104 #include "opt_inet.h"
105 #include "opt_tcp_debug.h"
106 #include "opt_mbuftrace.h"
107 
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/kernel.h>
111 #include <sys/mbuf.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #include <sys/protosw.h>
115 #include <sys/errno.h>
116 #include <sys/stat.h>
117 #include <sys/proc.h>
118 #include <sys/domain.h>
119 #include <sys/sysctl.h>
120 #include <sys/kauth.h>
121 #include <sys/kernel.h>
122 #include <sys/uidinfo.h>
123 
124 #include <net/if.h>
125 #include <net/route.h>
126 
127 #include <netinet/in.h>
128 #include <netinet/in_systm.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip.h>
131 #include <netinet/in_pcb.h>
132 #include <netinet/ip_var.h>
133 #include <netinet/in_offload.h>
134 
135 #ifdef INET6
136 #ifndef INET
137 #include <netinet/in.h>
138 #endif
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/tcpip.h>
153 #include <netinet/tcp_debug.h>
154 #include <netinet/tcp_vtw.h>
155 
156 #include "opt_tcp_space.h"
157 
158 static int
159 tcp_debug_capture(struct tcpcb *tp, int req)
160 {
161 #ifdef KPROF
162 	tcp_acounts[tp->t_state][req]++;
163 #endif
164 #ifdef TCP_DEBUG
165 	return tp->t_state;
166 #endif
167 	return 0;
168 }
169 
170 static inline void
171 tcp_debug_trace(struct socket *so, struct tcpcb *tp, int ostate, int req)
172 {
173 #ifdef TCP_DEBUG
174 	if (tp && (so->so_options & SO_DEBUG))
175 		tcp_trace(TA_USER, ostate, tp, NULL, req);
176 #endif
177 }
178 
179 static int
180 tcp_getpcb(struct socket *so, struct inpcb **inp,
181     struct in6pcb **in6p, struct tcpcb **tp)
182 {
183 
184 	KASSERT(solocked(so));
185 
186 	/*
187 	 * When a TCP is attached to a socket, then there will be
188 	 * a (struct inpcb) pointed at by the socket, and this
189 	 * structure will point at a subsidary (struct tcpcb).
190 	 */
191 	switch (so->so_proto->pr_domain->dom_family) {
192 #ifdef INET
193 	case PF_INET:
194 		*inp = sotoinpcb(so);
195 		if (*inp == NULL)
196 			return EINVAL;
197 		*tp = intotcpcb(*inp);
198 		break;
199 #endif
200 #ifdef INET6
201 	case PF_INET6:
202 		*in6p = sotoin6pcb(so);
203 		if (*in6p == NULL)
204 			return EINVAL;
205 		*tp = in6totcpcb(*in6p);
206 		break;
207 #endif
208 	default:
209 		return EAFNOSUPPORT;
210 	}
211 
212 	KASSERT(tp != NULL);
213 
214 	return 0;
215 }
216 
217 /*
218  * Process a TCP user request for TCP tb.  If this is a send request
219  * then m is the mbuf chain of send data.  If this is a timer expiration
220  * (called from the software clock routine), then timertype tells which timer.
221  */
222 static int
223 tcp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
224     struct mbuf *control, struct lwp *l)
225 {
226 	KASSERT(req != PRU_ATTACH);
227 	KASSERT(req != PRU_DETACH);
228 	KASSERT(req != PRU_ACCEPT);
229 	KASSERT(req != PRU_BIND);
230 	KASSERT(req != PRU_LISTEN);
231 	KASSERT(req != PRU_CONNECT);
232 	KASSERT(req != PRU_CONNECT2);
233 	KASSERT(req != PRU_DISCONNECT);
234 	KASSERT(req != PRU_SHUTDOWN);
235 	KASSERT(req != PRU_ABORT);
236 	KASSERT(req != PRU_CONTROL);
237 	KASSERT(req != PRU_SENSE);
238 	KASSERT(req != PRU_PEERADDR);
239 	KASSERT(req != PRU_SOCKADDR);
240 	KASSERT(req != PRU_RCVD);
241 	KASSERT(req != PRU_RCVOOB);
242 	KASSERT(req != PRU_SEND);
243 	KASSERT(req != PRU_SENDOOB);
244 	KASSERT(req != PRU_PURGEIF);
245 
246 	KASSERT(solocked(so));
247 
248 	panic("tcp_usrreq");
249 
250 	return 0;
251 }
252 
253 static void
254 change_keepalive(struct socket *so, struct tcpcb *tp)
255 {
256 	tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl;
257 	TCP_TIMER_DISARM(tp, TCPT_KEEP);
258 	TCP_TIMER_DISARM(tp, TCPT_2MSL);
259 
260 	if (tp->t_state == TCPS_SYN_RECEIVED ||
261 	    tp->t_state == TCPS_SYN_SENT) {
262 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
263 	} else if (so->so_options & SO_KEEPALIVE &&
264 	    tp->t_state <= TCPS_CLOSE_WAIT) {
265 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl);
266 	} else {
267 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
268 	}
269 
270 	if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
271 		TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
272 }
273 
274 /*
275  * Export TCP internal state information via a struct tcp_info, based on the
276  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
277  * (TCP state machine, etc).  We export all information using FreeBSD-native
278  * constants -- for example, the numeric values for tcpi_state will differ
279  * from Linux.
280  */
281 static void
282 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
283 {
284 
285 	bzero(ti, sizeof(*ti));
286 
287 	ti->tcpi_state = tp->t_state;
288 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
289 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
290 	if (tp->t_flags & TF_SACK_PERMIT)
291 		ti->tcpi_options |= TCPI_OPT_SACK;
292 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
293 		ti->tcpi_options |= TCPI_OPT_WSCALE;
294 		ti->tcpi_snd_wscale = tp->snd_scale;
295 		ti->tcpi_rcv_wscale = tp->rcv_scale;
296 	}
297 	if (tp->t_flags & TF_ECN_PERMIT) {
298 		ti->tcpi_options |= TCPI_OPT_ECN;
299 	}
300 
301 	ti->tcpi_rto = tp->t_rxtcur * tick;
302 	ti->tcpi_last_data_recv = (long)(hardclock_ticks -
303 					 (int)tp->t_rcvtime) * tick;
304 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
305 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
306 
307 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
308 	/* Linux API wants these in # of segments, apparently */
309 	ti->tcpi_snd_cwnd = tp->snd_cwnd / tp->t_segsz;
310 	ti->tcpi_snd_wnd = tp->snd_wnd / tp->t_segsz;
311 
312 	/*
313 	 * FreeBSD-specific extension fields for tcp_info.
314 	 */
315 	ti->tcpi_rcv_space = tp->rcv_wnd;
316 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
317 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
318 	ti->tcpi_snd_nxt = tp->snd_nxt;
319 	ti->tcpi_snd_mss = tp->t_segsz;
320 	ti->tcpi_rcv_mss = tp->t_segsz;
321 #ifdef TF_TOE
322 	if (tp->t_flags & TF_TOE)
323 		ti->tcpi_options |= TCPI_OPT_TOE;
324 #endif
325 	/* From the redundant department of redundancies... */
326 	ti->__tcpi_retransmits = ti->__tcpi_retrans =
327 		ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
328 
329 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
330 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
331 }
332 
333 int
334 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
335 {
336 	int error = 0, s;
337 	struct inpcb *inp;
338 #ifdef INET6
339 	struct in6pcb *in6p;
340 #endif
341 	struct tcpcb *tp;
342 	struct tcp_info ti;
343 	u_int ui;
344 	int family;	/* family of the socket */
345 	int level, optname, optval;
346 
347 	level = sopt->sopt_level;
348 	optname = sopt->sopt_name;
349 
350 	family = so->so_proto->pr_domain->dom_family;
351 
352 	s = splsoftnet();
353 	switch (family) {
354 #ifdef INET
355 	case PF_INET:
356 		inp = sotoinpcb(so);
357 #ifdef INET6
358 		in6p = NULL;
359 #endif
360 		break;
361 #endif
362 #ifdef INET6
363 	case PF_INET6:
364 		inp = NULL;
365 		in6p = sotoin6pcb(so);
366 		break;
367 #endif
368 	default:
369 		splx(s);
370 		panic("%s: af %d", __func__, family);
371 	}
372 #ifndef INET6
373 	if (inp == NULL)
374 #else
375 	if (inp == NULL && in6p == NULL)
376 #endif
377 	{
378 		splx(s);
379 		return (ECONNRESET);
380 	}
381 	if (level != IPPROTO_TCP) {
382 		switch (family) {
383 #ifdef INET
384 		case PF_INET:
385 			error = ip_ctloutput(op, so, sopt);
386 			break;
387 #endif
388 #ifdef INET6
389 		case PF_INET6:
390 			error = ip6_ctloutput(op, so, sopt);
391 			break;
392 #endif
393 		}
394 		splx(s);
395 		return (error);
396 	}
397 	if (inp)
398 		tp = intotcpcb(inp);
399 #ifdef INET6
400 	else if (in6p)
401 		tp = in6totcpcb(in6p);
402 #endif
403 	else
404 		tp = NULL;
405 
406 	switch (op) {
407 	case PRCO_SETOPT:
408 		switch (optname) {
409 #ifdef TCP_SIGNATURE
410 		case TCP_MD5SIG:
411 			error = sockopt_getint(sopt, &optval);
412 			if (error)
413 				break;
414 			if (optval > 0)
415 				tp->t_flags |= TF_SIGNATURE;
416 			else
417 				tp->t_flags &= ~TF_SIGNATURE;
418 			break;
419 #endif /* TCP_SIGNATURE */
420 
421 		case TCP_NODELAY:
422 			error = sockopt_getint(sopt, &optval);
423 			if (error)
424 				break;
425 			if (optval)
426 				tp->t_flags |= TF_NODELAY;
427 			else
428 				tp->t_flags &= ~TF_NODELAY;
429 			break;
430 
431 		case TCP_MAXSEG:
432 			error = sockopt_getint(sopt, &optval);
433 			if (error)
434 				break;
435 			if (optval > 0 && optval <= tp->t_peermss)
436 				tp->t_peermss = optval; /* limit on send size */
437 			else
438 				error = EINVAL;
439 			break;
440 #ifdef notyet
441 		case TCP_CONGCTL:
442 			/* XXX string overflow XXX */
443 			error = tcp_congctl_select(tp, sopt->sopt_data);
444 			break;
445 #endif
446 
447 		case TCP_KEEPIDLE:
448 			error = sockopt_get(sopt, &ui, sizeof(ui));
449 			if (error)
450 				break;
451 			if (ui > 0) {
452 				tp->t_keepidle = ui;
453 				change_keepalive(so, tp);
454 			} else
455 				error = EINVAL;
456 			break;
457 
458 		case TCP_KEEPINTVL:
459 			error = sockopt_get(sopt, &ui, sizeof(ui));
460 			if (error)
461 				break;
462 			if (ui > 0) {
463 				tp->t_keepintvl = ui;
464 				change_keepalive(so, tp);
465 			} else
466 				error = EINVAL;
467 			break;
468 
469 		case TCP_KEEPCNT:
470 			error = sockopt_get(sopt, &ui, sizeof(ui));
471 			if (error)
472 				break;
473 			if (ui > 0) {
474 				tp->t_keepcnt = ui;
475 				change_keepalive(so, tp);
476 			} else
477 				error = EINVAL;
478 			break;
479 
480 		case TCP_KEEPINIT:
481 			error = sockopt_get(sopt, &ui, sizeof(ui));
482 			if (error)
483 				break;
484 			if (ui > 0) {
485 				tp->t_keepinit = ui;
486 				change_keepalive(so, tp);
487 			} else
488 				error = EINVAL;
489 			break;
490 
491 		default:
492 			error = ENOPROTOOPT;
493 			break;
494 		}
495 		break;
496 
497 	case PRCO_GETOPT:
498 		switch (optname) {
499 #ifdef TCP_SIGNATURE
500 		case TCP_MD5SIG:
501 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
502 			error = sockopt_set(sopt, &optval, sizeof(optval));
503 			break;
504 #endif
505 		case TCP_NODELAY:
506 			optval = tp->t_flags & TF_NODELAY;
507 			error = sockopt_set(sopt, &optval, sizeof(optval));
508 			break;
509 		case TCP_MAXSEG:
510 			optval = tp->t_peermss;
511 			error = sockopt_set(sopt, &optval, sizeof(optval));
512 			break;
513 		case TCP_INFO:
514 			tcp_fill_info(tp, &ti);
515 			error = sockopt_set(sopt, &ti, sizeof ti);
516 			break;
517 #ifdef notyet
518 		case TCP_CONGCTL:
519 			break;
520 #endif
521 		default:
522 			error = ENOPROTOOPT;
523 			break;
524 		}
525 		break;
526 	}
527 	splx(s);
528 	return (error);
529 }
530 
531 #ifndef TCP_SENDSPACE
532 #define	TCP_SENDSPACE	1024*32
533 #endif
534 int	tcp_sendspace = TCP_SENDSPACE;
535 #ifndef TCP_RECVSPACE
536 #define	TCP_RECVSPACE	1024*32
537 #endif
538 int	tcp_recvspace = TCP_RECVSPACE;
539 
540 /*
541  * tcp_attach: attach TCP protocol to socket, allocating internet protocol
542  * control block, TCP control block, buffer space and entering LISTEN state
543  * if to accept connections.
544  */
545 static int
546 tcp_attach(struct socket *so, int proto)
547 {
548 	struct tcpcb *tp;
549 	struct inpcb *inp;
550 #ifdef INET6
551 	struct in6pcb *in6p;
552 #endif
553 	int s, error, family;
554 
555 	/* Assign the lock (must happen even if we will error out). */
556 	s = splsoftnet();
557 	sosetlock(so);
558 	KASSERT(solocked(so));
559 
560 	family = so->so_proto->pr_domain->dom_family;
561 	switch (family) {
562 #ifdef INET
563 	case PF_INET:
564 		inp = sotoinpcb(so);
565 #ifdef INET6
566 		in6p = NULL;
567 #endif
568 		break;
569 #endif
570 #ifdef INET6
571 	case PF_INET6:
572 		inp = NULL;
573 		in6p = sotoin6pcb(so);
574 		break;
575 #endif
576 	default:
577 		error = EAFNOSUPPORT;
578 		goto out;
579 	}
580 
581 	KASSERT(inp == NULL);
582 #ifdef INET6
583 	KASSERT(in6p == NULL);
584 #endif
585 
586 #ifdef MBUFTRACE
587 	so->so_mowner = &tcp_sock_mowner;
588 	so->so_rcv.sb_mowner = &tcp_sock_rx_mowner;
589 	so->so_snd.sb_mowner = &tcp_sock_tx_mowner;
590 #endif
591 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
592 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
593 		if (error)
594 			goto out;
595 	}
596 
597 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
598 	so->so_snd.sb_flags |= SB_AUTOSIZE;
599 
600 	switch (family) {
601 #ifdef INET
602 	case PF_INET:
603 		error = in_pcballoc(so, &tcbtable);
604 		if (error)
605 			goto out;
606 		inp = sotoinpcb(so);
607 #ifdef INET6
608 		in6p = NULL;
609 #endif
610 		break;
611 #endif
612 #ifdef INET6
613 	case PF_INET6:
614 		error = in6_pcballoc(so, &tcbtable);
615 		if (error)
616 			goto out;
617 		inp = NULL;
618 		in6p = sotoin6pcb(so);
619 		break;
620 #endif
621 	default:
622 		error = EAFNOSUPPORT;
623 		goto out;
624 	}
625 	if (inp)
626 		tp = tcp_newtcpcb(family, (void *)inp);
627 #ifdef INET6
628 	else if (in6p)
629 		tp = tcp_newtcpcb(family, (void *)in6p);
630 #endif
631 	else
632 		tp = NULL;
633 
634 	if (tp == NULL) {
635 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
636 
637 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
638 #ifdef INET
639 		if (inp)
640 			in_pcbdetach(inp);
641 #endif
642 #ifdef INET6
643 		if (in6p)
644 			in6_pcbdetach(in6p);
645 #endif
646 		so->so_state |= nofd;
647 		error = ENOBUFS;
648 		goto out;
649 	}
650 	tp->t_state = TCPS_CLOSED;
651 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
652 		so->so_linger = TCP_LINGERTIME;
653 	}
654 out:
655 	KASSERT(solocked(so));
656 	splx(s);
657 	return error;
658 }
659 
660 static void
661 tcp_detach(struct socket *so)
662 {
663 	struct inpcb *inp = NULL;
664 	struct in6pcb *in6p = NULL;
665 	struct tcpcb *tp = NULL;
666 	int s;
667 
668 	if (tcp_getpcb(so, &inp, &in6p, &tp) != 0)
669 		return;
670 
671 	s = splsoftnet();
672 	(void)tcp_disconnect1(tp);
673 	splx(s);
674 }
675 
676 static int
677 tcp_accept(struct socket *so, struct mbuf *nam)
678 {
679 	struct inpcb *inp = NULL;
680 	struct in6pcb *in6p = NULL;
681 	struct tcpcb *tp = NULL;
682 	int ostate = 0;
683 	int error = 0;
684 	int s;
685 
686 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
687 		return error;
688 
689 	ostate = tcp_debug_capture(tp, PRU_ACCEPT);
690 
691 	/*
692 	 * Accept a connection.  Essentially all the work is
693 	 * done at higher levels; just return the address
694 	 * of the peer, storing through addr.
695 	 */
696 	s = splsoftnet();
697 #ifdef INET
698 	if (inp) {
699 		in_setpeeraddr(inp, nam);
700 	}
701 #endif
702 #ifdef INET6
703 	if (in6p) {
704 		in6_setpeeraddr(in6p, nam);
705 	}
706 #endif
707 	tcp_debug_trace(so, tp, ostate, PRU_ACCEPT);
708 	splx(s);
709 
710 	return 0;
711 }
712 
713 static int
714 tcp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
715 {
716 	struct inpcb *inp = NULL;
717 	struct in6pcb *in6p = NULL;
718 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
719 #ifdef INET6
720 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
721 #endif /* INET6 */
722 	struct tcpcb *tp = NULL;
723 	int s;
724 	int error = 0;
725 	int ostate = 0;
726 
727 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
728 		return error;
729 
730 	ostate = tcp_debug_capture(tp, PRU_BIND);
731 
732 	/*
733 	 * Give the socket an address.
734 	 */
735 	s = splsoftnet();
736 	switch (so->so_proto->pr_domain->dom_family) {
737 #ifdef INET
738 	case PF_INET:
739 		error = in_pcbbind(inp, sin, l);
740 		break;
741 #endif
742 #ifdef INET6
743 	case PF_INET6:
744 		error = in6_pcbbind(in6p, sin6, l);
745 		if (!error) {
746 			/* mapped addr case */
747 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
748 				tp->t_family = AF_INET;
749 			else
750 				tp->t_family = AF_INET6;
751 		}
752 		break;
753 #endif
754 	}
755 	tcp_debug_trace(so, tp, ostate, PRU_BIND);
756 	splx(s);
757 
758 	return error;
759 }
760 
761 static int
762 tcp_listen(struct socket *so, struct lwp *l)
763 {
764 	struct inpcb *inp = NULL;
765 	struct in6pcb *in6p = NULL;
766 	struct tcpcb *tp = NULL;
767 	int error = 0;
768 	int ostate = 0;
769 	int s;
770 
771 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
772 		return error;
773 
774 	ostate = tcp_debug_capture(tp, PRU_LISTEN);
775 
776 	/*
777 	 * Prepare to accept connections.
778 	 */
779 	s = splsoftnet();
780 #ifdef INET
781 	if (inp && inp->inp_lport == 0) {
782 		error = in_pcbbind(inp, NULL, l);
783 		if (error)
784 			goto release;
785 	}
786 #endif
787 #ifdef INET6
788 	if (in6p && in6p->in6p_lport == 0) {
789 		error = in6_pcbbind(in6p, NULL, l);
790 		if (error)
791 			goto release;
792 	}
793 #endif
794 	tp->t_state = TCPS_LISTEN;
795 
796 release:
797 	tcp_debug_trace(so, tp, ostate, PRU_LISTEN);
798 	splx(s);
799 
800 	return error;
801 }
802 
803 static int
804 tcp_connect(struct socket *so, struct mbuf *nam, struct lwp *l)
805 {
806 	struct inpcb *inp = NULL;
807 	struct in6pcb *in6p = NULL;
808 	struct tcpcb *tp = NULL;
809 	int s;
810 	int error = 0;
811 	int ostate = 0;
812 
813 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
814 		return error;
815 
816 	ostate = tcp_debug_capture(tp, PRU_CONNECT);
817 
818 	/*
819 	 * Initiate connection to peer.
820 	 * Create a template for use in transmissions on this connection.
821 	 * Enter SYN_SENT state, and mark socket as connecting.
822 	 * Start keep-alive timer, and seed output sequence space.
823 	 * Send initial segment on connection.
824 	 */
825 	s = splsoftnet();
826 #ifdef INET
827 	if (inp) {
828 		if (inp->inp_lport == 0) {
829 			error = in_pcbbind(inp, NULL, l);
830 			if (error)
831 				goto release;
832 		}
833 		error = in_pcbconnect(inp, nam, l);
834 	}
835 #endif
836 #ifdef INET6
837 	if (in6p) {
838 		if (in6p->in6p_lport == 0) {
839 			error = in6_pcbbind(in6p, NULL, l);
840 			if (error)
841 				goto release;
842 		}
843 		error = in6_pcbconnect(in6p, nam, l);
844 		if (!error) {
845 			/* mapped addr case */
846 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
847 				tp->t_family = AF_INET;
848 			else
849 				tp->t_family = AF_INET6;
850 		}
851 	}
852 #endif
853 	if (error)
854 		goto release;
855 	tp->t_template = tcp_template(tp);
856 	if (tp->t_template == 0) {
857 #ifdef INET
858 		if (inp)
859 			in_pcbdisconnect(inp);
860 #endif
861 #ifdef INET6
862 		if (in6p)
863 			in6_pcbdisconnect(in6p);
864 #endif
865 		error = ENOBUFS;
866 		goto release;
867 	}
868 	/*
869 	 * Compute window scaling to request.
870 	 * XXX: This should be moved to tcp_output().
871 	 */
872 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
873 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
874 		tp->request_r_scale++;
875 	soisconnecting(so);
876 	TCP_STATINC(TCP_STAT_CONNATTEMPT);
877 	tp->t_state = TCPS_SYN_SENT;
878 	TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
879 	tp->iss = tcp_new_iss(tp, 0);
880 	tcp_sendseqinit(tp);
881 	error = tcp_output(tp);
882 
883 release:
884 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT);
885 	splx(s);
886 
887 	return error;
888 }
889 
890 static int
891 tcp_connect2(struct socket *so, struct socket *so2)
892 {
893 	struct inpcb *inp = NULL;
894 	struct in6pcb *in6p = NULL;
895 	struct tcpcb *tp = NULL;
896 	int error = 0;
897 	int ostate = 0;
898 
899 	KASSERT(solocked(so));
900 
901 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
902 		return error;
903 
904 	ostate = tcp_debug_capture(tp, PRU_CONNECT2);
905 
906 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT2);
907 
908 	return EOPNOTSUPP;
909 }
910 
911 static int
912 tcp_disconnect(struct socket *so)
913 {
914 	struct inpcb *inp = NULL;
915 	struct in6pcb *in6p = NULL;
916 	struct tcpcb *tp = NULL;
917 	int error = 0;
918 	int ostate = 0;
919 	int s;
920 
921 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
922 		return error;
923 
924 	ostate = tcp_debug_capture(tp, PRU_DISCONNECT);
925 
926 	/*
927 	 * Initiate disconnect from peer.
928 	 * If connection never passed embryonic stage, just drop;
929 	 * else if don't need to let data drain, then can just drop anyways,
930 	 * else have to begin TCP shutdown process: mark socket disconnecting,
931 	 * drain unread data, state switch to reflect user close, and
932 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
933 	 * when peer sends FIN and acks ours.
934 	 *
935 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
936 	 */
937 	s = splsoftnet();
938 	tp = tcp_disconnect1(tp);
939 	tcp_debug_trace(so, tp, ostate, PRU_DISCONNECT);
940 	splx(s);
941 
942 	return error;
943 }
944 
945 static int
946 tcp_shutdown(struct socket *so)
947 {
948 	struct inpcb *inp = NULL;
949 	struct in6pcb *in6p = NULL;
950 	struct tcpcb *tp = NULL;
951 	int error = 0;
952 	int ostate = 0;
953 	int s;
954 
955 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
956 		return error;
957 
958 	ostate = tcp_debug_capture(tp, PRU_SHUTDOWN);
959 	/*
960 	 * Mark the connection as being incapable of further output.
961 	 */
962 	s = splsoftnet();
963 	socantsendmore(so);
964 	tp = tcp_usrclosed(tp);
965 	if (tp)
966 		error = tcp_output(tp);
967 	tcp_debug_trace(so, tp, ostate, PRU_SHUTDOWN);
968 	splx(s);
969 
970 	return error;
971 }
972 
973 static int
974 tcp_abort(struct socket *so)
975 {
976 	struct inpcb *inp = NULL;
977 	struct in6pcb *in6p = NULL;
978 	struct tcpcb *tp = NULL;
979 	int error = 0;
980 	int ostate = 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_ABORT);
987 
988 	/*
989 	 * Abort the TCP.
990 	 */
991 	s = splsoftnet();
992 	tp = tcp_drop(tp, ECONNABORTED);
993 	tcp_debug_trace(so, tp, ostate, PRU_ABORT);
994 	splx(s);
995 
996 	return error;
997 }
998 
999 static int
1000 tcp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
1001 {
1002 	switch (so->so_proto->pr_domain->dom_family) {
1003 #ifdef INET
1004 	case PF_INET:
1005 		return in_control(so, cmd, nam, ifp);
1006 #endif
1007 #ifdef INET6
1008 	case PF_INET6:
1009 		return in6_control(so, cmd, nam, ifp);
1010 #endif
1011 	default:
1012 		return EAFNOSUPPORT;
1013 	}
1014 }
1015 
1016 static int
1017 tcp_stat(struct socket *so, struct stat *ub)
1018 {
1019 	KASSERT(solocked(so));
1020 
1021 	/* stat: don't bother with a blocksize.  */
1022 	return 0;
1023 }
1024 
1025 static int
1026 tcp_peeraddr(struct socket *so, struct mbuf *nam)
1027 {
1028 	struct inpcb *inp = NULL;
1029 	struct in6pcb *in6p = NULL;
1030 	struct tcpcb *tp = NULL;
1031 	int ostate = 0;
1032 	int error = 0;
1033 	int s;
1034 
1035 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1036 		return error;
1037 
1038 	ostate = tcp_debug_capture(tp, PRU_PEERADDR);
1039 
1040 	s = splsoftnet();
1041 #ifdef INET
1042 	if (inp)
1043 		in_setpeeraddr(inp, nam);
1044 #endif
1045 #ifdef INET6
1046 	if (in6p)
1047 		in6_setpeeraddr(in6p, nam);
1048 #endif
1049 	tcp_debug_trace(so, tp, ostate, PRU_PEERADDR);
1050 	splx(s);
1051 
1052 	return 0;
1053 }
1054 
1055 static int
1056 tcp_sockaddr(struct socket *so, struct mbuf *nam)
1057 {
1058 	struct inpcb *inp = NULL;
1059 	struct in6pcb *in6p = NULL;
1060 	struct tcpcb *tp = NULL;
1061 	int ostate = 0;
1062 	int error = 0;
1063 	int s;
1064 
1065 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1066 		return error;
1067 
1068 	ostate = tcp_debug_capture(tp, PRU_SOCKADDR);
1069 
1070 	s = splsoftnet();
1071 #ifdef INET
1072 	if (inp)
1073 		in_setsockaddr(inp, nam);
1074 #endif
1075 #ifdef INET6
1076 	if (in6p)
1077 		in6_setsockaddr(in6p, nam);
1078 #endif
1079 	tcp_debug_trace(so, tp, ostate, PRU_SOCKADDR);
1080 	splx(s);
1081 
1082 	return 0;
1083 }
1084 
1085 static int
1086 tcp_rcvd(struct socket *so, int flags, struct lwp *l)
1087 {
1088 	struct inpcb *inp = NULL;
1089 	struct in6pcb *in6p = NULL;
1090 	struct tcpcb *tp = NULL;
1091 	int ostate = 0;
1092 	int error = 0;
1093 	int s;
1094 
1095 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1096 		return error;
1097 
1098 	ostate = tcp_debug_capture(tp, PRU_RCVD);
1099 
1100 	/*
1101 	 * After a receive, possibly send window update to peer.
1102 	 *
1103 	 * soreceive() calls this function when a user receives
1104 	 * ancillary data on a listening socket. We don't call
1105 	 * tcp_output in such a case, since there is no header
1106 	 * template for a listening socket and hence the kernel
1107 	 * will panic.
1108 	 */
1109 	s = splsoftnet();
1110 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
1111 		(void) tcp_output(tp);
1112 	splx(s);
1113 
1114 	tcp_debug_trace(so, tp, ostate, PRU_RCVD);
1115 
1116 	return 0;
1117 }
1118 
1119 static int
1120 tcp_recvoob(struct socket *so, struct mbuf *m, int flags)
1121 {
1122 	struct inpcb *inp = NULL;
1123 	struct in6pcb *in6p = NULL;
1124 	struct tcpcb *tp = NULL;
1125 	int ostate = 0;
1126 	int error = 0;
1127 	int s;
1128 
1129 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1130 		return error;
1131 
1132 	ostate = tcp_debug_capture(tp, PRU_RCVOOB);
1133 
1134 	s = splsoftnet();
1135 	if ((so->so_oobmark == 0 &&
1136 	    (so->so_state & SS_RCVATMARK) == 0) ||
1137 	    so->so_options & SO_OOBINLINE ||
1138 	    tp->t_oobflags & TCPOOB_HADDATA) {
1139 		splx(s);
1140 		return EINVAL;
1141 	}
1142 
1143 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1144 		splx(s);
1145 		return EWOULDBLOCK;
1146 	}
1147 
1148 	m->m_len = 1;
1149 	*mtod(m, char *) = tp->t_iobc;
1150 	if ((flags & MSG_PEEK) == 0)
1151 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1152 
1153 	tcp_debug_trace(so, tp, ostate, PRU_RCVOOB);
1154 	splx(s);
1155 
1156 	return 0;
1157 }
1158 
1159 static int
1160 tcp_send(struct socket *so, struct mbuf *m, struct mbuf *nam,
1161     struct mbuf *control, struct lwp *l)
1162 {
1163 	struct inpcb *inp = NULL;
1164 	struct in6pcb *in6p = NULL;
1165 	struct tcpcb *tp = NULL;
1166 	int ostate = 0;
1167 	int error = 0;
1168 	int s;
1169 
1170 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1171 		return error;
1172 
1173 	ostate = tcp_debug_capture(tp, PRU_SEND);
1174 
1175 	/*
1176 	 * Do a send by putting data in output queue and updating urgent
1177 	 * marker if URG set.  Possibly send more data.
1178 	 */
1179 	s = splsoftnet();
1180 	if (control && control->m_len) {
1181 		m_freem(control);
1182 		m_freem(m);
1183 		tcp_debug_trace(so, tp, ostate, PRU_SEND);
1184 		splx(s);
1185 		return EINVAL;
1186 	}
1187 
1188 	sbappendstream(&so->so_snd, m);
1189 	error = tcp_output(tp);
1190 	tcp_debug_trace(so, tp, ostate, PRU_SEND);
1191 	splx(s);
1192 
1193 	return error;
1194 }
1195 
1196 static int
1197 tcp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1198 {
1199 	struct inpcb *inp = NULL;
1200 	struct in6pcb *in6p = NULL;
1201 	struct tcpcb *tp = NULL;
1202 	int ostate = 0;
1203 	int error = 0;
1204 	int s;
1205 
1206 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1207 		return error;
1208 
1209 	ostate = tcp_debug_capture(tp, PRU_SENDOOB);
1210 
1211 	s = splsoftnet();
1212 	if (sbspace(&so->so_snd) < -512) {
1213 		m_freem(m);
1214 		splx(s);
1215 		return ENOBUFS;
1216 	}
1217 	/*
1218 	 * According to RFC961 (Assigned Protocols),
1219 	 * the urgent pointer points to the last octet
1220 	 * of urgent data.  We continue, however,
1221 	 * to consider it to indicate the first octet
1222 	 * of data past the urgent section.
1223 	 * Otherwise, snd_up should be one lower.
1224 	 */
1225 	sbappendstream(&so->so_snd, m);
1226 	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1227 	tp->t_force = 1;
1228 	error = tcp_output(tp);
1229 	tp->t_force = 0;
1230 	tcp_debug_trace(so, tp, ostate, PRU_SENDOOB);
1231 	splx(s);
1232 
1233 	return error;
1234 }
1235 
1236 static int
1237 tcp_purgeif(struct socket *so, struct ifnet *ifp)
1238 {
1239 	int s;
1240 
1241 	s = splsoftnet();
1242 	mutex_enter(softnet_lock);
1243 	switch (so->so_proto->pr_domain->dom_family) {
1244 #ifdef INET
1245 	case PF_INET:
1246 		in_pcbpurgeif0(&tcbtable, ifp);
1247 		in_purgeif(ifp);
1248 		in_pcbpurgeif(&tcbtable, ifp);
1249 		break;
1250 #endif
1251 #ifdef INET6
1252 	case PF_INET6:
1253 		in6_pcbpurgeif0(&tcbtable, ifp);
1254 		in6_purgeif(ifp);
1255 		in6_pcbpurgeif(&tcbtable, ifp);
1256 		break;
1257 #endif
1258 	default:
1259 		mutex_exit(softnet_lock);
1260 		splx(s);
1261 		return EAFNOSUPPORT;
1262 	}
1263 	mutex_exit(softnet_lock);
1264 	splx(s);
1265 
1266 	return 0;
1267 }
1268 
1269 /*
1270  * Initiate (or continue) disconnect.
1271  * If embryonic state, just send reset (once).
1272  * If in ``let data drain'' option and linger null, just drop.
1273  * Otherwise (hard), mark socket disconnecting and drop
1274  * current input data; switch states based on user close, and
1275  * send segment to peer (with FIN).
1276  */
1277 struct tcpcb *
1278 tcp_disconnect1(struct tcpcb *tp)
1279 {
1280 	struct socket *so;
1281 
1282 	if (tp->t_inpcb)
1283 		so = tp->t_inpcb->inp_socket;
1284 #ifdef INET6
1285 	else if (tp->t_in6pcb)
1286 		so = tp->t_in6pcb->in6p_socket;
1287 #endif
1288 	else
1289 		so = NULL;
1290 
1291 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
1292 		tp = tcp_close(tp);
1293 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1294 		tp = tcp_drop(tp, 0);
1295 	else {
1296 		soisdisconnecting(so);
1297 		sbflush(&so->so_rcv);
1298 		tp = tcp_usrclosed(tp);
1299 		if (tp)
1300 			(void) tcp_output(tp);
1301 	}
1302 	return (tp);
1303 }
1304 
1305 /*
1306  * User issued close, and wish to trail through shutdown states:
1307  * if never received SYN, just forget it.  If got a SYN from peer,
1308  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1309  * If already got a FIN from peer, then almost done; go to LAST_ACK
1310  * state.  In all other cases, have already sent FIN to peer (e.g.
1311  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1312  * for peer to send FIN or not respond to keep-alives, etc.
1313  * We can let the user exit from the close as soon as the FIN is acked.
1314  */
1315 struct tcpcb *
1316 tcp_usrclosed(struct tcpcb *tp)
1317 {
1318 
1319 	switch (tp->t_state) {
1320 
1321 	case TCPS_CLOSED:
1322 	case TCPS_LISTEN:
1323 	case TCPS_SYN_SENT:
1324 		tp->t_state = TCPS_CLOSED;
1325 		tp = tcp_close(tp);
1326 		break;
1327 
1328 	case TCPS_SYN_RECEIVED:
1329 	case TCPS_ESTABLISHED:
1330 		tp->t_state = TCPS_FIN_WAIT_1;
1331 		break;
1332 
1333 	case TCPS_CLOSE_WAIT:
1334 		tp->t_state = TCPS_LAST_ACK;
1335 		break;
1336 	}
1337 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1338 		struct socket *so;
1339 		if (tp->t_inpcb)
1340 			so = tp->t_inpcb->inp_socket;
1341 #ifdef INET6
1342 		else if (tp->t_in6pcb)
1343 			so = tp->t_in6pcb->in6p_socket;
1344 #endif
1345 		else
1346 			so = NULL;
1347 		if (so)
1348 			soisdisconnected(so);
1349 		/*
1350 		 * If we are in FIN_WAIT_2, we arrived here because the
1351 		 * application did a shutdown of the send side.  Like the
1352 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
1353 		 * a full close, we start a timer to make sure sockets are
1354 		 * not left in FIN_WAIT_2 forever.
1355 		 */
1356 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
1357 			TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
1358 		else if (tp->t_state == TCPS_TIME_WAIT
1359 			 && ((tp->t_inpcb
1360 			      && (tcp4_vtw_enable & 1)
1361 			      && vtw_add(AF_INET, tp))
1362 			     ||
1363 			     (tp->t_in6pcb
1364 			      && (tcp6_vtw_enable & 1)
1365 			      && vtw_add(AF_INET6, tp)))) {
1366 			tp = 0;
1367 		}
1368 	}
1369 	return (tp);
1370 }
1371 
1372 /*
1373  * sysctl helper routine for net.inet.ip.mssdflt.  it can't be less
1374  * than 32.
1375  */
1376 static int
1377 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS)
1378 {
1379 	int error, mssdflt;
1380 	struct sysctlnode node;
1381 
1382 	mssdflt = tcp_mssdflt;
1383 	node = *rnode;
1384 	node.sysctl_data = &mssdflt;
1385 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1386 	if (error || newp == NULL)
1387 		return (error);
1388 
1389 	if (mssdflt < 32)
1390 		return (EINVAL);
1391 	tcp_mssdflt = mssdflt;
1392 
1393 	mutex_enter(softnet_lock);
1394 	tcp_tcpcb_template();
1395 	mutex_exit(softnet_lock);
1396 
1397 	return (0);
1398 }
1399 
1400 /*
1401  * sysctl helper for TCP CB template update
1402  */
1403 static int
1404 sysctl_update_tcpcb_template(SYSCTLFN_ARGS)
1405 {
1406 	int t, error;
1407 	struct sysctlnode node;
1408 
1409 	/* follow procedures in sysctl(9) manpage */
1410 	t = *(int *)rnode->sysctl_data;
1411 	node = *rnode;
1412 	node.sysctl_data = &t;
1413 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1414 	if (error || newp == NULL)
1415 		return error;
1416 
1417 	if (t < 0)
1418 		return EINVAL;
1419 
1420 	*(int *)rnode->sysctl_data = t;
1421 
1422 	mutex_enter(softnet_lock);
1423 	tcp_tcpcb_template();
1424 	mutex_exit(softnet_lock);
1425 
1426 	return 0;
1427 }
1428 
1429 /*
1430  * sysctl helper routine for setting port related values under
1431  * net.inet.ip and net.inet6.ip6.  does basic range checking and does
1432  * additional checks for each type.  this code has placed in
1433  * tcp_input.c since INET and INET6 both use the same tcp code.
1434  *
1435  * this helper is not static so that both inet and inet6 can use it.
1436  */
1437 int
1438 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS)
1439 {
1440 	int error, tmp;
1441 	int apmin, apmax;
1442 #ifndef IPNOPRIVPORTS
1443 	int lpmin, lpmax;
1444 #endif /* IPNOPRIVPORTS */
1445 	struct sysctlnode node;
1446 
1447 	if (namelen != 0)
1448 		return (EINVAL);
1449 
1450 	switch (name[-3]) {
1451 #ifdef INET
1452 	    case PF_INET:
1453 		apmin = anonportmin;
1454 		apmax = anonportmax;
1455 #ifndef IPNOPRIVPORTS
1456 		lpmin = lowportmin;
1457 		lpmax = lowportmax;
1458 #endif /* IPNOPRIVPORTS */
1459 		break;
1460 #endif /* INET */
1461 #ifdef INET6
1462 	    case PF_INET6:
1463 		apmin = ip6_anonportmin;
1464 		apmax = ip6_anonportmax;
1465 #ifndef IPNOPRIVPORTS
1466 		lpmin = ip6_lowportmin;
1467 		lpmax = ip6_lowportmax;
1468 #endif /* IPNOPRIVPORTS */
1469 		break;
1470 #endif /* INET6 */
1471 	    default:
1472 		return (EINVAL);
1473 	}
1474 
1475 	/*
1476 	 * insert temporary copy into node, perform lookup on
1477 	 * temporary, then restore pointer
1478 	 */
1479 	node = *rnode;
1480 	tmp = *(int*)rnode->sysctl_data;
1481 	node.sysctl_data = &tmp;
1482 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1483 	if (error || newp == NULL)
1484 		return (error);
1485 
1486 	/*
1487 	 * simple port range check
1488 	 */
1489 	if (tmp < 0 || tmp > 65535)
1490 		return (EINVAL);
1491 
1492 	/*
1493 	 * per-node range checks
1494 	 */
1495 	switch (rnode->sysctl_num) {
1496 	case IPCTL_ANONPORTMIN:
1497 	case IPV6CTL_ANONPORTMIN:
1498 		if (tmp >= apmax)
1499 			return (EINVAL);
1500 #ifndef IPNOPRIVPORTS
1501 		if (tmp < IPPORT_RESERVED)
1502                         return (EINVAL);
1503 #endif /* IPNOPRIVPORTS */
1504 		break;
1505 
1506 	case IPCTL_ANONPORTMAX:
1507 	case IPV6CTL_ANONPORTMAX:
1508                 if (apmin >= tmp)
1509 			return (EINVAL);
1510 #ifndef IPNOPRIVPORTS
1511 		if (tmp < IPPORT_RESERVED)
1512                         return (EINVAL);
1513 #endif /* IPNOPRIVPORTS */
1514 		break;
1515 
1516 #ifndef IPNOPRIVPORTS
1517 	case IPCTL_LOWPORTMIN:
1518 	case IPV6CTL_LOWPORTMIN:
1519 		if (tmp >= lpmax ||
1520 		    tmp > IPPORT_RESERVEDMAX ||
1521 		    tmp < IPPORT_RESERVEDMIN)
1522 			return (EINVAL);
1523 		break;
1524 
1525 	case IPCTL_LOWPORTMAX:
1526 	case IPV6CTL_LOWPORTMAX:
1527 		if (lpmin >= tmp ||
1528 		    tmp > IPPORT_RESERVEDMAX ||
1529 		    tmp < IPPORT_RESERVEDMIN)
1530 			return (EINVAL);
1531 		break;
1532 #endif /* IPNOPRIVPORTS */
1533 
1534 	default:
1535 		return (EINVAL);
1536 	}
1537 
1538 	*(int*)rnode->sysctl_data = tmp;
1539 
1540 	return (0);
1541 }
1542 
1543 static inline int
1544 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp)
1545 {
1546 	if (oldp) {
1547 		size_t sz;
1548 		uid_t uid;
1549 		int error;
1550 
1551 		if (sockp->so_cred == NULL)
1552 			return EPERM;
1553 
1554 		uid = kauth_cred_geteuid(sockp->so_cred);
1555 		sz = MIN(sizeof(uid), *oldlenp);
1556 		if ((error = copyout(&uid, oldp, sz)) != 0)
1557 			return error;
1558 	}
1559 	*oldlenp = sizeof(uid_t);
1560 	return 0;
1561 }
1562 
1563 static inline int
1564 inet4_ident_core(struct in_addr raddr, u_int rport,
1565     struct in_addr laddr, u_int lport,
1566     void *oldp, size_t *oldlenp,
1567     struct lwp *l, int dodrop)
1568 {
1569 	struct inpcb *inp;
1570 	struct socket *sockp;
1571 
1572 	inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0);
1573 
1574 	if (inp == NULL || (sockp = inp->inp_socket) == NULL)
1575 		return ESRCH;
1576 
1577 	if (dodrop) {
1578 		struct tcpcb *tp;
1579 		int error;
1580 
1581 		if (inp == NULL || (tp = intotcpcb(inp)) == NULL ||
1582 		    (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0)
1583 			return ESRCH;
1584 
1585 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1586 		    KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL);
1587 		if (error)
1588 			return (error);
1589 
1590 		(void)tcp_drop(tp, ECONNABORTED);
1591 		return 0;
1592 	}
1593 	else
1594 		return copyout_uid(sockp, oldp, oldlenp);
1595 }
1596 
1597 #ifdef INET6
1598 static inline int
1599 inet6_ident_core(struct in6_addr *raddr, u_int rport,
1600     struct in6_addr *laddr, u_int lport,
1601     void *oldp, size_t *oldlenp,
1602     struct lwp *l, int dodrop)
1603 {
1604 	struct in6pcb *in6p;
1605 	struct socket *sockp;
1606 
1607 	in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0);
1608 
1609 	if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL)
1610 		return ESRCH;
1611 
1612 	if (dodrop) {
1613 		struct tcpcb *tp;
1614 		int error;
1615 
1616 		if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL ||
1617 		    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0)
1618 			return ESRCH;
1619 
1620 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1621 		    KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL);
1622 		if (error)
1623 			return (error);
1624 
1625 		(void)tcp_drop(tp, ECONNABORTED);
1626 		return 0;
1627 	}
1628 	else
1629 		return copyout_uid(sockp, oldp, oldlenp);
1630 }
1631 #endif
1632 
1633 /*
1634  * sysctl helper routine for the net.inet.tcp.drop and
1635  * net.inet6.tcp6.drop nodes.
1636  */
1637 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident
1638 
1639 /*
1640  * sysctl helper routine for the net.inet.tcp.ident and
1641  * net.inet6.tcp6.ident nodes.  contains backwards compat code for the
1642  * old way of looking up the ident information for ipv4 which involves
1643  * stuffing the port/addr pairs into the mib lookup.
1644  */
1645 static int
1646 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS)
1647 {
1648 #ifdef INET
1649 	struct sockaddr_in *si4[2];
1650 #endif /* INET */
1651 #ifdef INET6
1652 	struct sockaddr_in6 *si6[2];
1653 #endif /* INET6 */
1654 	struct sockaddr_storage sa[2];
1655 	int error, pf, dodrop;
1656 
1657 	dodrop = name[-1] == TCPCTL_DROP;
1658 	if (dodrop) {
1659 		if (oldp != NULL || *oldlenp != 0)
1660 			return EINVAL;
1661 		if (newp == NULL)
1662 			return EPERM;
1663 		if (newlen < sizeof(sa))
1664 			return ENOMEM;
1665 	}
1666 	if (namelen != 4 && namelen != 0)
1667 		return EINVAL;
1668 	if (name[-2] != IPPROTO_TCP)
1669 		return EINVAL;
1670 	pf = name[-3];
1671 
1672 	/* old style lookup, ipv4 only */
1673 	if (namelen == 4) {
1674 #ifdef INET
1675 		struct in_addr laddr, raddr;
1676 		u_int lport, rport;
1677 
1678 		if (pf != PF_INET)
1679 			return EPROTONOSUPPORT;
1680 		raddr.s_addr = (uint32_t)name[0];
1681 		rport = (u_int)name[1];
1682 		laddr.s_addr = (uint32_t)name[2];
1683 		lport = (u_int)name[3];
1684 
1685 		mutex_enter(softnet_lock);
1686 		error = inet4_ident_core(raddr, rport, laddr, lport,
1687 		    oldp, oldlenp, l, dodrop);
1688 		mutex_exit(softnet_lock);
1689 		return error;
1690 #else /* INET */
1691 		return EINVAL;
1692 #endif /* INET */
1693 	}
1694 
1695 	if (newp == NULL || newlen != sizeof(sa))
1696 		return EINVAL;
1697 	error = copyin(newp, &sa, newlen);
1698 	if (error)
1699 		return error;
1700 
1701 	/*
1702 	 * requested families must match
1703 	 */
1704 	if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family)
1705 		return EINVAL;
1706 
1707 	switch (pf) {
1708 #ifdef INET6
1709 	case PF_INET6:
1710 		si6[0] = (struct sockaddr_in6*)&sa[0];
1711 		si6[1] = (struct sockaddr_in6*)&sa[1];
1712 		if (si6[0]->sin6_len != sizeof(*si6[0]) ||
1713 		    si6[1]->sin6_len != sizeof(*si6[1]))
1714 			return EINVAL;
1715 
1716 		if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) &&
1717 		    !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) {
1718 			error = sa6_embedscope(si6[0], ip6_use_defzone);
1719 			if (error)
1720 				return error;
1721 			error = sa6_embedscope(si6[1], ip6_use_defzone);
1722 			if (error)
1723 				return error;
1724 
1725 			mutex_enter(softnet_lock);
1726 			error = inet6_ident_core(&si6[0]->sin6_addr,
1727 			    si6[0]->sin6_port, &si6[1]->sin6_addr,
1728 			    si6[1]->sin6_port, oldp, oldlenp, l, dodrop);
1729 			mutex_exit(softnet_lock);
1730 			return error;
1731 		}
1732 
1733 		if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) !=
1734 		    IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr))
1735 			return EINVAL;
1736 
1737 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]);
1738 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]);
1739 		/*FALLTHROUGH*/
1740 #endif /* INET6 */
1741 #ifdef INET
1742 	case PF_INET:
1743 		si4[0] = (struct sockaddr_in*)&sa[0];
1744 		si4[1] = (struct sockaddr_in*)&sa[1];
1745 		if (si4[0]->sin_len != sizeof(*si4[0]) ||
1746 		    si4[0]->sin_len != sizeof(*si4[1]))
1747 			return EINVAL;
1748 
1749 		mutex_enter(softnet_lock);
1750 		error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port,
1751 		    si4[1]->sin_addr, si4[1]->sin_port,
1752 		    oldp, oldlenp, l, dodrop);
1753 		mutex_exit(softnet_lock);
1754 		return error;
1755 #endif /* INET */
1756 	default:
1757 		return EPROTONOSUPPORT;
1758 	}
1759 }
1760 
1761 /*
1762  * sysctl helper for the inet and inet6 pcblists.  handles tcp/udp and
1763  * inet/inet6, as well as raw pcbs for each.  specifically not
1764  * declared static so that raw sockets and udp/udp6 can use it as
1765  * well.
1766  */
1767 int
1768 sysctl_inpcblist(SYSCTLFN_ARGS)
1769 {
1770 #ifdef INET
1771 	struct sockaddr_in *in;
1772 	const struct inpcb *inp;
1773 #endif
1774 #ifdef INET6
1775 	struct sockaddr_in6 *in6;
1776 	const struct in6pcb *in6p;
1777 #endif
1778 	struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data);
1779 	const struct inpcb_hdr *inph;
1780 	struct tcpcb *tp;
1781 	struct kinfo_pcb pcb;
1782 	char *dp;
1783 	size_t len, needed, elem_size, out_size;
1784 	int error, elem_count, pf, proto, pf2;
1785 
1786 	if (namelen != 4)
1787 		return (EINVAL);
1788 
1789 	if (oldp != NULL) {
1790 		    len = *oldlenp;
1791 		    elem_size = name[2];
1792 		    elem_count = name[3];
1793 		    if (elem_size != sizeof(pcb))
1794 			    return EINVAL;
1795 	} else {
1796 		    len = 0;
1797 		    elem_count = INT_MAX;
1798 		    elem_size = sizeof(pcb);
1799 	}
1800 	error = 0;
1801 	dp = oldp;
1802 	out_size = elem_size;
1803 	needed = 0;
1804 
1805 	if (namelen == 1 && name[0] == CTL_QUERY)
1806 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1807 
1808 	if (name - oname != 4)
1809 		return (EINVAL);
1810 
1811 	pf = oname[1];
1812 	proto = oname[2];
1813 	pf2 = (oldp != NULL) ? pf : 0;
1814 
1815 	mutex_enter(softnet_lock);
1816 
1817 	TAILQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) {
1818 #ifdef INET
1819 		inp = (const struct inpcb *)inph;
1820 #endif
1821 #ifdef INET6
1822 		in6p = (const struct in6pcb *)inph;
1823 #endif
1824 
1825 		if (inph->inph_af != pf)
1826 			continue;
1827 
1828 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1829 		    KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL,
1830 		    NULL) != 0)
1831 			continue;
1832 
1833 		memset(&pcb, 0, sizeof(pcb));
1834 
1835 		pcb.ki_family = pf;
1836 		pcb.ki_type = proto;
1837 
1838 		switch (pf2) {
1839 		case 0:
1840 			/* just probing for size */
1841 			break;
1842 #ifdef INET
1843 		case PF_INET:
1844 			pcb.ki_family = inp->inp_socket->so_proto->
1845 			    pr_domain->dom_family;
1846 			pcb.ki_type = inp->inp_socket->so_proto->
1847 			    pr_type;
1848 			pcb.ki_protocol = inp->inp_socket->so_proto->
1849 			    pr_protocol;
1850 			pcb.ki_pflags = inp->inp_flags;
1851 
1852 			pcb.ki_sostate = inp->inp_socket->so_state;
1853 			pcb.ki_prstate = inp->inp_state;
1854 			if (proto == IPPROTO_TCP) {
1855 				tp = intotcpcb(inp);
1856 				pcb.ki_tstate = tp->t_state;
1857 				pcb.ki_tflags = tp->t_flags;
1858 			}
1859 
1860 			pcb.ki_pcbaddr = PTRTOUINT64(inp);
1861 			pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb);
1862 			pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket);
1863 
1864 			pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc;
1865 			pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc;
1866 
1867 			in = satosin(&pcb.ki_src);
1868 			in->sin_len = sizeof(*in);
1869 			in->sin_family = pf;
1870 			in->sin_port = inp->inp_lport;
1871 			in->sin_addr = inp->inp_laddr;
1872 			if (pcb.ki_prstate >= INP_CONNECTED) {
1873 				in = satosin(&pcb.ki_dst);
1874 				in->sin_len = sizeof(*in);
1875 				in->sin_family = pf;
1876 				in->sin_port = inp->inp_fport;
1877 				in->sin_addr = inp->inp_faddr;
1878 			}
1879 			break;
1880 #endif
1881 #ifdef INET6
1882 		case PF_INET6:
1883 			pcb.ki_family = in6p->in6p_socket->so_proto->
1884 			    pr_domain->dom_family;
1885 			pcb.ki_type = in6p->in6p_socket->so_proto->pr_type;
1886 			pcb.ki_protocol = in6p->in6p_socket->so_proto->
1887 			    pr_protocol;
1888 			pcb.ki_pflags = in6p->in6p_flags;
1889 
1890 			pcb.ki_sostate = in6p->in6p_socket->so_state;
1891 			pcb.ki_prstate = in6p->in6p_state;
1892 			if (proto == IPPROTO_TCP) {
1893 				tp = in6totcpcb(in6p);
1894 				pcb.ki_tstate = tp->t_state;
1895 				pcb.ki_tflags = tp->t_flags;
1896 			}
1897 
1898 			pcb.ki_pcbaddr = PTRTOUINT64(in6p);
1899 			pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb);
1900 			pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket);
1901 
1902 			pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc;
1903 			pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc;
1904 
1905 			in6 = satosin6(&pcb.ki_src);
1906 			in6->sin6_len = sizeof(*in6);
1907 			in6->sin6_family = pf;
1908 			in6->sin6_port = in6p->in6p_lport;
1909 			in6->sin6_flowinfo = in6p->in6p_flowinfo;
1910 			in6->sin6_addr = in6p->in6p_laddr;
1911 			in6->sin6_scope_id = 0; /* XXX? */
1912 
1913 			if (pcb.ki_prstate >= IN6P_CONNECTED) {
1914 				in6 = satosin6(&pcb.ki_dst);
1915 				in6->sin6_len = sizeof(*in6);
1916 				in6->sin6_family = pf;
1917 				in6->sin6_port = in6p->in6p_fport;
1918 				in6->sin6_flowinfo = in6p->in6p_flowinfo;
1919 				in6->sin6_addr = in6p->in6p_faddr;
1920 				in6->sin6_scope_id = 0; /* XXX? */
1921 			}
1922 			break;
1923 #endif
1924 		}
1925 
1926 		if (len >= elem_size && elem_count > 0) {
1927 			error = copyout(&pcb, dp, out_size);
1928 			if (error) {
1929 				mutex_exit(softnet_lock);
1930 				return (error);
1931 			}
1932 			dp += elem_size;
1933 			len -= elem_size;
1934 		}
1935 		needed += elem_size;
1936 		if (elem_count > 0 && elem_count != INT_MAX)
1937 			elem_count--;
1938 	}
1939 
1940 	*oldlenp = needed;
1941 	if (oldp == NULL)
1942 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
1943 
1944 	mutex_exit(softnet_lock);
1945 
1946 	return (error);
1947 }
1948 
1949 static int
1950 sysctl_tcp_congctl(SYSCTLFN_ARGS)
1951 {
1952 	struct sysctlnode node;
1953 	int error;
1954 	char newname[TCPCC_MAXLEN];
1955 
1956 	strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1);
1957 
1958 	node = *rnode;
1959 	node.sysctl_data = newname;
1960 	node.sysctl_size = sizeof(newname);
1961 
1962 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1963 
1964 	if (error ||
1965 	    newp == NULL ||
1966 	    strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0)
1967 		return error;
1968 
1969 	mutex_enter(softnet_lock);
1970 	error = tcp_congctl_select(NULL, newname);
1971 	mutex_exit(softnet_lock);
1972 
1973 	return error;
1974 }
1975 
1976 static int
1977 sysctl_tcp_init_win(SYSCTLFN_ARGS)
1978 {
1979 	int error;
1980 	u_int iw;
1981 	struct sysctlnode node;
1982 
1983 	iw = *(u_int *)rnode->sysctl_data;
1984 	node = *rnode;
1985 	node.sysctl_data = &iw;
1986 	node.sysctl_size = sizeof(iw);
1987 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1988 	if (error || newp == NULL)
1989 		return error;
1990 
1991 	if (iw >= __arraycount(tcp_init_win_max))
1992 		return EINVAL;
1993 	*(u_int *)rnode->sysctl_data = iw;
1994 	return 0;
1995 }
1996 
1997 static int
1998 sysctl_tcp_keep(SYSCTLFN_ARGS)
1999 {
2000 	int error;
2001 	u_int tmp;
2002 	struct sysctlnode node;
2003 
2004 	node = *rnode;
2005 	tmp = *(u_int *)rnode->sysctl_data;
2006 	node.sysctl_data = &tmp;
2007 
2008 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2009 	if (error || newp == NULL)
2010 		return error;
2011 
2012 	mutex_enter(softnet_lock);
2013 
2014 	*(u_int *)rnode->sysctl_data = tmp;
2015 	tcp_tcpcb_template();	/* update the template */
2016 
2017 	mutex_exit(softnet_lock);
2018 	return 0;
2019 }
2020 
2021 static int
2022 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS)
2023 {
2024 
2025 	return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS));
2026 }
2027 
2028 /*
2029  * this (second stage) setup routine is a replacement for tcp_sysctl()
2030  * (which is currently used for ipv4 and ipv6)
2031  */
2032 static void
2033 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
2034 			   const char *tcpname)
2035 {
2036 	const struct sysctlnode *sack_node;
2037 	const struct sysctlnode *abc_node;
2038 	const struct sysctlnode *ecn_node;
2039 	const struct sysctlnode *congctl_node;
2040 	const struct sysctlnode *mslt_node;
2041 	const struct sysctlnode *vtw_node;
2042 #ifdef TCP_DEBUG
2043 	extern struct tcp_debug tcp_debug[TCP_NDEBUG];
2044 	extern int tcp_debx;
2045 #endif
2046 
2047 	sysctl_createv(clog, 0, NULL, NULL,
2048 		       CTLFLAG_PERMANENT,
2049 		       CTLTYPE_NODE, pfname, NULL,
2050 		       NULL, 0, NULL, 0,
2051 		       CTL_NET, pf, CTL_EOL);
2052 	sysctl_createv(clog, 0, NULL, NULL,
2053 		       CTLFLAG_PERMANENT,
2054 		       CTLTYPE_NODE, tcpname,
2055 		       SYSCTL_DESCR("TCP related settings"),
2056 		       NULL, 0, NULL, 0,
2057 		       CTL_NET, pf, IPPROTO_TCP, CTL_EOL);
2058 
2059 	sysctl_createv(clog, 0, NULL, NULL,
2060 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2061 		       CTLTYPE_INT, "rfc1323",
2062 		       SYSCTL_DESCR("Enable RFC1323 TCP extensions"),
2063 		       sysctl_update_tcpcb_template, 0, &tcp_do_rfc1323, 0,
2064 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL);
2065 	sysctl_createv(clog, 0, NULL, NULL,
2066 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2067 		       CTLTYPE_INT, "sendspace",
2068 		       SYSCTL_DESCR("Default TCP send buffer size"),
2069 		       NULL, 0, &tcp_sendspace, 0,
2070 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL);
2071 	sysctl_createv(clog, 0, NULL, NULL,
2072 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2073 		       CTLTYPE_INT, "recvspace",
2074 		       SYSCTL_DESCR("Default TCP receive buffer size"),
2075 		       NULL, 0, &tcp_recvspace, 0,
2076 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL);
2077 	sysctl_createv(clog, 0, NULL, NULL,
2078 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2079 		       CTLTYPE_INT, "mssdflt",
2080 		       SYSCTL_DESCR("Default maximum segment size"),
2081 		       sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0,
2082 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL);
2083 	sysctl_createv(clog, 0, NULL, NULL,
2084 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2085 		       CTLTYPE_INT, "minmss",
2086 		       SYSCTL_DESCR("Lower limit for TCP maximum segment size"),
2087 		       NULL, 0, &tcp_minmss, 0,
2088 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2089 	sysctl_createv(clog, 0, NULL, NULL,
2090 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2091 		       CTLTYPE_INT, "msl",
2092 		       SYSCTL_DESCR("Maximum Segment Life"),
2093 		       NULL, 0, &tcp_msl, 0,
2094 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL);
2095 	sysctl_createv(clog, 0, NULL, NULL,
2096 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2097 		       CTLTYPE_INT, "syn_cache_limit",
2098 		       SYSCTL_DESCR("Maximum number of entries in the TCP "
2099 				    "compressed state engine"),
2100 		       NULL, 0, &tcp_syn_cache_limit, 0,
2101 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT,
2102 		       CTL_EOL);
2103 	sysctl_createv(clog, 0, NULL, NULL,
2104 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2105 		       CTLTYPE_INT, "syn_bucket_limit",
2106 		       SYSCTL_DESCR("Maximum number of entries per hash "
2107 				    "bucket in the TCP compressed state "
2108 				    "engine"),
2109 		       NULL, 0, &tcp_syn_bucket_limit, 0,
2110 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT,
2111 		       CTL_EOL);
2112 #if 0 /* obsoleted */
2113 	sysctl_createv(clog, 0, NULL, NULL,
2114 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2115 		       CTLTYPE_INT, "syn_cache_interval",
2116 		       SYSCTL_DESCR("TCP compressed state engine's timer interval"),
2117 		       NULL, 0, &tcp_syn_cache_interval, 0,
2118 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER,
2119 		       CTL_EOL);
2120 #endif
2121 	sysctl_createv(clog, 0, NULL, NULL,
2122 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2123 		       CTLTYPE_INT, "init_win",
2124 		       SYSCTL_DESCR("Initial TCP congestion window"),
2125 		       sysctl_tcp_init_win, 0, &tcp_init_win, 0,
2126 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL);
2127 	sysctl_createv(clog, 0, NULL, NULL,
2128 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2129 		       CTLTYPE_INT, "mss_ifmtu",
2130 		       SYSCTL_DESCR("Use interface MTU for calculating MSS"),
2131 		       NULL, 0, &tcp_mss_ifmtu, 0,
2132 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL);
2133 	sysctl_createv(clog, 0, NULL, &sack_node,
2134 		       CTLFLAG_PERMANENT,
2135 		       CTLTYPE_NODE, "sack",
2136 		       SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
2137 		       NULL, 0, NULL, 0,
2138 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
2139 
2140 	/* Congctl subtree */
2141 	sysctl_createv(clog, 0, NULL, &congctl_node,
2142 		       CTLFLAG_PERMANENT,
2143 		       CTLTYPE_NODE, "congctl",
2144 		       SYSCTL_DESCR("TCP Congestion Control"),
2145 	    	       NULL, 0, NULL, 0,
2146 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2147 	sysctl_createv(clog, 0, &congctl_node, NULL,
2148 		       CTLFLAG_PERMANENT,
2149 		       CTLTYPE_STRING, "available",
2150 		       SYSCTL_DESCR("Available Congestion Control Mechanisms"),
2151 		       NULL, 0, tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL);
2152 	sysctl_createv(clog, 0, &congctl_node, NULL,
2153 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2154 		       CTLTYPE_STRING, "selected",
2155 		       SYSCTL_DESCR("Selected Congestion Control Mechanism"),
2156 		       sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN,
2157 		       CTL_CREATE, CTL_EOL);
2158 
2159 	sysctl_createv(clog, 0, NULL, NULL,
2160 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2161 		       CTLTYPE_INT, "win_scale",
2162 		       SYSCTL_DESCR("Use RFC1323 window scale options"),
2163 		       sysctl_update_tcpcb_template, 0, &tcp_do_win_scale, 0,
2164 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL);
2165 	sysctl_createv(clog, 0, NULL, NULL,
2166 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2167 		       CTLTYPE_INT, "timestamps",
2168 		       SYSCTL_DESCR("Use RFC1323 time stamp options"),
2169 		       sysctl_update_tcpcb_template, 0, &tcp_do_timestamps, 0,
2170 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL);
2171 	sysctl_createv(clog, 0, NULL, NULL,
2172 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2173 		       CTLTYPE_INT, "compat_42",
2174 		       SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"),
2175 		       NULL, 0, &tcp_compat_42, 0,
2176 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL);
2177 	sysctl_createv(clog, 0, NULL, NULL,
2178 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2179 		       CTLTYPE_INT, "cwm",
2180 		       SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window "
2181 				    "Monitoring"),
2182 		       NULL, 0, &tcp_cwm, 0,
2183 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL);
2184 	sysctl_createv(clog, 0, NULL, NULL,
2185 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2186 		       CTLTYPE_INT, "cwm_burstsize",
2187 		       SYSCTL_DESCR("Congestion Window Monitoring allowed "
2188 				    "burst count in packets"),
2189 		       NULL, 0, &tcp_cwm_burstsize, 0,
2190 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE,
2191 		       CTL_EOL);
2192 	sysctl_createv(clog, 0, NULL, NULL,
2193 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2194 		       CTLTYPE_INT, "ack_on_push",
2195 		       SYSCTL_DESCR("Immediately return ACK when PSH is "
2196 				    "received"),
2197 		       NULL, 0, &tcp_ack_on_push, 0,
2198 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL);
2199 	sysctl_createv(clog, 0, NULL, NULL,
2200 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2201 		       CTLTYPE_INT, "keepidle",
2202 		       SYSCTL_DESCR("Allowed connection idle ticks before a "
2203 				    "keepalive probe is sent"),
2204 		       sysctl_tcp_keep, 0, &tcp_keepidle, 0,
2205 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL);
2206 	sysctl_createv(clog, 0, NULL, NULL,
2207 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2208 		       CTLTYPE_INT, "keepintvl",
2209 		       SYSCTL_DESCR("Ticks before next keepalive probe is sent"),
2210 		       sysctl_tcp_keep, 0, &tcp_keepintvl, 0,
2211 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL);
2212 	sysctl_createv(clog, 0, NULL, NULL,
2213 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2214 		       CTLTYPE_INT, "keepcnt",
2215 		       SYSCTL_DESCR("Number of keepalive probes to send"),
2216 		       sysctl_tcp_keep, 0, &tcp_keepcnt, 0,
2217 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL);
2218 	sysctl_createv(clog, 0, NULL, NULL,
2219 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
2220 		       CTLTYPE_INT, "slowhz",
2221 		       SYSCTL_DESCR("Keepalive ticks per second"),
2222 		       NULL, PR_SLOWHZ, NULL, 0,
2223 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL);
2224 	sysctl_createv(clog, 0, NULL, NULL,
2225 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2226 		       CTLTYPE_INT, "log_refused",
2227 		       SYSCTL_DESCR("Log refused TCP connections"),
2228 		       NULL, 0, &tcp_log_refused, 0,
2229 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL);
2230 #if 0 /* obsoleted */
2231 	sysctl_createv(clog, 0, NULL, NULL,
2232 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2233 		       CTLTYPE_INT, "rstratelimit", NULL,
2234 		       NULL, 0, &tcp_rst_ratelim, 0,
2235 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL);
2236 #endif
2237 	sysctl_createv(clog, 0, NULL, NULL,
2238 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2239 		       CTLTYPE_INT, "rstppslimit",
2240 		       SYSCTL_DESCR("Maximum number of RST packets to send "
2241 				    "per second"),
2242 		       NULL, 0, &tcp_rst_ppslim, 0,
2243 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL);
2244 	sysctl_createv(clog, 0, NULL, NULL,
2245 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2246 		       CTLTYPE_INT, "delack_ticks",
2247 		       SYSCTL_DESCR("Number of ticks to delay sending an ACK"),
2248 		       NULL, 0, &tcp_delack_ticks, 0,
2249 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL);
2250 	sysctl_createv(clog, 0, NULL, NULL,
2251 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2252 		       CTLTYPE_INT, "init_win_local",
2253 		       SYSCTL_DESCR("Initial TCP window size (in segments)"),
2254 		       sysctl_tcp_init_win, 0, &tcp_init_win_local, 0,
2255 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL,
2256 		       CTL_EOL);
2257 	sysctl_createv(clog, 0, NULL, NULL,
2258 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2259 		       CTLTYPE_STRUCT, "ident",
2260 		       SYSCTL_DESCR("RFC1413 Identification Protocol lookups"),
2261 		       sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t),
2262 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL);
2263 	sysctl_createv(clog, 0, NULL, NULL,
2264 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2265 		       CTLTYPE_INT, "do_loopback_cksum",
2266 		       SYSCTL_DESCR("Perform TCP checksum on loopback"),
2267 		       NULL, 0, &tcp_do_loopback_cksum, 0,
2268 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM,
2269 		       CTL_EOL);
2270 	sysctl_createv(clog, 0, NULL, NULL,
2271 		       CTLFLAG_PERMANENT,
2272 		       CTLTYPE_STRUCT, "pcblist",
2273 		       SYSCTL_DESCR("TCP protocol control block list"),
2274 		       sysctl_inpcblist, 0, &tcbtable, 0,
2275 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2276 		       CTL_EOL);
2277 	sysctl_createv(clog, 0, NULL, NULL,
2278 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2279 		       CTLTYPE_INT, "keepinit",
2280 		       SYSCTL_DESCR("Ticks before initial tcp connection times out"),
2281 		       sysctl_tcp_keep, 0, &tcp_keepinit, 0,
2282 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2283 
2284 	/* TCP socket buffers auto-sizing nodes */
2285 	sysctl_createv(clog, 0, NULL, NULL,
2286 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2287 		       CTLTYPE_INT, "recvbuf_auto",
2288 		       SYSCTL_DESCR("Enable automatic receive "
2289 		           "buffer sizing (experimental)"),
2290 		       NULL, 0, &tcp_do_autorcvbuf, 0,
2291 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2292 	sysctl_createv(clog, 0, NULL, NULL,
2293 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2294 		       CTLTYPE_INT, "recvbuf_inc",
2295 		       SYSCTL_DESCR("Incrementor step size of "
2296 		           "automatic receive buffer"),
2297 		       NULL, 0, &tcp_autorcvbuf_inc, 0,
2298 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2299 	sysctl_createv(clog, 0, NULL, NULL,
2300 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2301 		       CTLTYPE_INT, "recvbuf_max",
2302 		       SYSCTL_DESCR("Max size of automatic receive buffer"),
2303 		       NULL, 0, &tcp_autorcvbuf_max, 0,
2304 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2305 
2306 	sysctl_createv(clog, 0, NULL, NULL,
2307 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2308 		       CTLTYPE_INT, "sendbuf_auto",
2309 		       SYSCTL_DESCR("Enable automatic send "
2310 		           "buffer sizing (experimental)"),
2311 		       NULL, 0, &tcp_do_autosndbuf, 0,
2312 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2313 	sysctl_createv(clog, 0, NULL, NULL,
2314 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2315 		       CTLTYPE_INT, "sendbuf_inc",
2316 		       SYSCTL_DESCR("Incrementor step size of "
2317 		           "automatic send buffer"),
2318 		       NULL, 0, &tcp_autosndbuf_inc, 0,
2319 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2320 	sysctl_createv(clog, 0, NULL, NULL,
2321 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2322 		       CTLTYPE_INT, "sendbuf_max",
2323 		       SYSCTL_DESCR("Max size of automatic send buffer"),
2324 		       NULL, 0, &tcp_autosndbuf_max, 0,
2325 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2326 
2327 	/* ECN subtree */
2328 	sysctl_createv(clog, 0, NULL, &ecn_node,
2329 	    	       CTLFLAG_PERMANENT,
2330 		       CTLTYPE_NODE, "ecn",
2331 	    	       SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
2332 	    	       NULL, 0, NULL, 0,
2333 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2334 	sysctl_createv(clog, 0, &ecn_node, NULL,
2335 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2336 		       CTLTYPE_INT, "enable",
2337 		       SYSCTL_DESCR("Enable TCP Explicit Congestion "
2338 			   "Notification"),
2339 	    	       NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL);
2340 	sysctl_createv(clog, 0, &ecn_node, NULL,
2341 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2342 		       CTLTYPE_INT, "maxretries",
2343 		       SYSCTL_DESCR("Number of times to retry ECN setup "
2344 			       "before disabling ECN on the connection"),
2345 	    	       NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL);
2346 
2347 	/* SACK gets its own little subtree. */
2348 	sysctl_createv(clog, 0, NULL, &sack_node,
2349 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2350 		       CTLTYPE_INT, "enable",
2351 		       SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"),
2352 		       NULL, 0, &tcp_do_sack, 0,
2353 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2354 	sysctl_createv(clog, 0, NULL, &sack_node,
2355 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2356 		       CTLTYPE_INT, "maxholes",
2357 		       SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"),
2358 		       NULL, 0, &tcp_sack_tp_maxholes, 0,
2359 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2360 	sysctl_createv(clog, 0, NULL, &sack_node,
2361 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2362 		       CTLTYPE_INT, "globalmaxholes",
2363 		       SYSCTL_DESCR("Global maximum number of TCP SACK holes"),
2364 		       NULL, 0, &tcp_sack_globalmaxholes, 0,
2365 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2366 	sysctl_createv(clog, 0, NULL, &sack_node,
2367 		       CTLFLAG_PERMANENT,
2368 		       CTLTYPE_INT, "globalholes",
2369 		       SYSCTL_DESCR("Global number of TCP SACK holes"),
2370 		       NULL, 0, &tcp_sack_globalholes, 0,
2371 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2372 
2373 	sysctl_createv(clog, 0, NULL, NULL,
2374 		       CTLFLAG_PERMANENT,
2375 		       CTLTYPE_STRUCT, "stats",
2376 		       SYSCTL_DESCR("TCP statistics"),
2377 		       sysctl_net_inet_tcp_stats, 0, NULL, 0,
2378 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS,
2379 		       CTL_EOL);
2380         sysctl_createv(clog, 0, NULL, NULL,
2381                        CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2382                        CTLTYPE_INT, "local_by_rtt",
2383                        SYSCTL_DESCR("Use RTT estimator to decide which hosts "
2384 				    "are local"),
2385 		       NULL, 0, &tcp_rttlocal, 0,
2386 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2387 #ifdef TCP_DEBUG
2388 	sysctl_createv(clog, 0, NULL, NULL,
2389 		       CTLFLAG_PERMANENT,
2390 		       CTLTYPE_STRUCT, "debug",
2391 		       SYSCTL_DESCR("TCP sockets debug information"),
2392 		       NULL, 0, &tcp_debug, sizeof(tcp_debug),
2393 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG,
2394 		       CTL_EOL);
2395 	sysctl_createv(clog, 0, NULL, NULL,
2396 		       CTLFLAG_PERMANENT,
2397 		       CTLTYPE_INT, "debx",
2398 		       SYSCTL_DESCR("Number of TCP debug sockets messages"),
2399 		       NULL, 0, &tcp_debx, sizeof(tcp_debx),
2400 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX,
2401 		       CTL_EOL);
2402 #endif
2403 	sysctl_createv(clog, 0, NULL, NULL,
2404 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2405 		       CTLTYPE_STRUCT, "drop",
2406 		       SYSCTL_DESCR("TCP drop connection"),
2407 		       sysctl_net_inet_tcp_drop, 0, NULL, 0,
2408 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL);
2409 	sysctl_createv(clog, 0, NULL, NULL,
2410 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2411 		       CTLTYPE_INT, "iss_hash",
2412 		       SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic "
2413 				    "hash computation"),
2414 		       NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948),
2415 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2416 		       CTL_EOL);
2417 
2418 	/* ABC subtree */
2419 
2420 	sysctl_createv(clog, 0, NULL, &abc_node,
2421 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc",
2422 		       SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"),
2423 		       NULL, 0, NULL, 0,
2424 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2425 	sysctl_createv(clog, 0, &abc_node, NULL,
2426 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2427 		       CTLTYPE_INT, "enable",
2428 		       SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"),
2429 		       NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL);
2430 	sysctl_createv(clog, 0, &abc_node, NULL,
2431 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2432 		       CTLTYPE_INT, "aggressive",
2433 		       SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"),
2434 		       NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL);
2435 
2436 	/* MSL tuning subtree */
2437 
2438 	sysctl_createv(clog, 0, NULL, &mslt_node,
2439 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt",
2440 		       SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"),
2441 		       NULL, 0, NULL, 0,
2442 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2443 	sysctl_createv(clog, 0, &mslt_node, NULL,
2444 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2445 		       CTLTYPE_INT, "enable",
2446 		       SYSCTL_DESCR("Enable TIME_WAIT truncation"),
2447 		       NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL);
2448 	sysctl_createv(clog, 0, &mslt_node, NULL,
2449 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2450 		       CTLTYPE_INT, "loopback",
2451 		       SYSCTL_DESCR("MSL value to use for loopback connections"),
2452 		       NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL);
2453 	sysctl_createv(clog, 0, &mslt_node, NULL,
2454 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2455 		       CTLTYPE_INT, "local",
2456 		       SYSCTL_DESCR("MSL value to use for local connections"),
2457 		       NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL);
2458 	sysctl_createv(clog, 0, &mslt_node, NULL,
2459 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2460 		       CTLTYPE_INT, "remote",
2461 		       SYSCTL_DESCR("MSL value to use for remote connections"),
2462 		       NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL);
2463 	sysctl_createv(clog, 0, &mslt_node, NULL,
2464 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2465 		       CTLTYPE_INT, "remote_threshold",
2466 		       SYSCTL_DESCR("RTT estimate value to promote local to remote"),
2467 		       NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL);
2468 
2469 	/* vestigial TIME_WAIT tuning subtree */
2470 
2471 	sysctl_createv(clog, 0, NULL, &vtw_node,
2472 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw",
2473 		       SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"),
2474 		       NULL, 0, NULL, 0,
2475 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2476 	sysctl_createv(clog, 0, &vtw_node, NULL,
2477 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2478 		       CTLTYPE_INT, "enable",
2479 		       SYSCTL_DESCR("Enable Vestigial TIME_WAIT"),
2480 		       sysctl_tcp_vtw_enable, 0,
2481 	               (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable,
2482 		       0, CTL_CREATE, CTL_EOL);
2483 	sysctl_createv(clog, 0, &vtw_node, NULL,
2484 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY,
2485 		       CTLTYPE_INT, "entries",
2486 		       SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"),
2487 		       NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL);
2488 }
2489 
2490 void
2491 tcp_usrreq_init(void)
2492 {
2493 
2494 #ifdef INET
2495 	sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp");
2496 #endif
2497 #ifdef INET6
2498 	sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6");
2499 #endif
2500 }
2501 
2502 PR_WRAP_USRREQS(tcp)
2503 #define	tcp_attach	tcp_attach_wrapper
2504 #define	tcp_detach	tcp_detach_wrapper
2505 #define	tcp_accept	tcp_accept_wrapper
2506 #define	tcp_bind	tcp_bind_wrapper
2507 #define	tcp_listen	tcp_listen_wrapper
2508 #define	tcp_connect	tcp_connect_wrapper
2509 #define	tcp_connect2	tcp_connect2_wrapper
2510 #define	tcp_disconnect	tcp_disconnect_wrapper
2511 #define	tcp_shutdown	tcp_shutdown_wrapper
2512 #define	tcp_abort	tcp_abort_wrapper
2513 #define	tcp_ioctl	tcp_ioctl_wrapper
2514 #define	tcp_stat	tcp_stat_wrapper
2515 #define	tcp_peeraddr	tcp_peeraddr_wrapper
2516 #define	tcp_sockaddr	tcp_sockaddr_wrapper
2517 #define	tcp_rcvd	tcp_rcvd_wrapper
2518 #define	tcp_recvoob	tcp_recvoob_wrapper
2519 #define	tcp_send	tcp_send_wrapper
2520 #define	tcp_sendoob	tcp_sendoob_wrapper
2521 #define	tcp_purgeif	tcp_purgeif_wrapper
2522 #define	tcp_usrreq	tcp_usrreq_wrapper
2523 
2524 const struct pr_usrreqs tcp_usrreqs = {
2525 	.pr_attach	= tcp_attach,
2526 	.pr_detach	= tcp_detach,
2527 	.pr_accept	= tcp_accept,
2528 	.pr_bind	= tcp_bind,
2529 	.pr_listen	= tcp_listen,
2530 	.pr_connect	= tcp_connect,
2531 	.pr_connect2	= tcp_connect2,
2532 	.pr_disconnect	= tcp_disconnect,
2533 	.pr_shutdown	= tcp_shutdown,
2534 	.pr_abort	= tcp_abort,
2535 	.pr_ioctl	= tcp_ioctl,
2536 	.pr_stat	= tcp_stat,
2537 	.pr_peeraddr	= tcp_peeraddr,
2538 	.pr_sockaddr	= tcp_sockaddr,
2539 	.pr_rcvd	= tcp_rcvd,
2540 	.pr_recvoob	= tcp_recvoob,
2541 	.pr_send	= tcp_send,
2542 	.pr_sendoob	= tcp_sendoob,
2543 	.pr_purgeif	= tcp_purgeif,
2544 	.pr_generic	= tcp_usrreq,
2545 };
2546