xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 76c7fc5f6b13ed0b1508e6b313e88e59977ed78e)
1 /*	$NetBSD: tcp_usrreq.c,v 1.225 2019/08/06 15:48:18 riastradh 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 1.225 2019/08/06 15:48:18 riastradh 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)(hardclock_ticks -
260 					 (int)tp->t_rcvtime) * tick;
261 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
262 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
263 
264 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
265 	/* Linux API wants these in # of segments, apparently */
266 	ti->tcpi_snd_cwnd = tp->snd_cwnd / tp->t_segsz;
267 	ti->tcpi_snd_wnd = tp->snd_wnd / tp->t_segsz;
268 
269 	/*
270 	 * FreeBSD-specific extension fields for tcp_info.
271 	 */
272 	ti->tcpi_rcv_space = tp->rcv_wnd;
273 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
274 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
275 	ti->tcpi_snd_nxt = tp->snd_nxt;
276 	ti->tcpi_snd_mss = tp->t_segsz;
277 	ti->tcpi_rcv_mss = tp->t_segsz;
278 #ifdef TF_TOE
279 	if (tp->t_flags & TF_TOE)
280 		ti->tcpi_options |= TCPI_OPT_TOE;
281 #endif
282 	/* From the redundant department of redundancies... */
283 	ti->__tcpi_retransmits = ti->__tcpi_retrans =
284 		ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
285 
286 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
287 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
288 }
289 
290 int
291 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
292 {
293 	int error = 0, s;
294 	struct inpcb *inp;
295 #ifdef INET6
296 	struct in6pcb *in6p;
297 #endif
298 	struct tcpcb *tp;
299 	struct tcp_info ti;
300 	u_int ui;
301 	int family;	/* family of the socket */
302 	int level, optname, optval;
303 
304 	level = sopt->sopt_level;
305 	optname = sopt->sopt_name;
306 
307 	family = so->so_proto->pr_domain->dom_family;
308 
309 	s = splsoftnet();
310 	switch (family) {
311 	case PF_INET:
312 		inp = sotoinpcb(so);
313 #ifdef INET6
314 		in6p = NULL;
315 #endif
316 		break;
317 #ifdef INET6
318 	case PF_INET6:
319 		inp = NULL;
320 		in6p = sotoin6pcb(so);
321 		break;
322 #endif
323 	default:
324 		splx(s);
325 		panic("%s: af %d", __func__, family);
326 	}
327 #ifndef INET6
328 	if (inp == NULL)
329 #else
330 	if (inp == NULL && in6p == NULL)
331 #endif
332 	{
333 		splx(s);
334 		return ECONNRESET;
335 	}
336 	if (level != IPPROTO_TCP) {
337 		switch (family) {
338 		case PF_INET:
339 			error = ip_ctloutput(op, so, sopt);
340 			break;
341 #ifdef INET6
342 		case PF_INET6:
343 			error = ip6_ctloutput(op, so, sopt);
344 			break;
345 #endif
346 		}
347 		splx(s);
348 		return error;
349 	}
350 	if (inp)
351 		tp = intotcpcb(inp);
352 #ifdef INET6
353 	else if (in6p)
354 		tp = in6totcpcb(in6p);
355 #endif
356 	else
357 		tp = NULL;
358 
359 	switch (op) {
360 	case PRCO_SETOPT:
361 		switch (optname) {
362 #ifdef TCP_SIGNATURE
363 		case TCP_MD5SIG:
364 			error = sockopt_getint(sopt, &optval);
365 			if (error)
366 				break;
367 			if (optval > 0)
368 				tp->t_flags |= TF_SIGNATURE;
369 			else
370 				tp->t_flags &= ~TF_SIGNATURE;
371 			break;
372 #endif /* TCP_SIGNATURE */
373 
374 		case TCP_NODELAY:
375 			error = sockopt_getint(sopt, &optval);
376 			if (error)
377 				break;
378 			if (optval)
379 				tp->t_flags |= TF_NODELAY;
380 			else
381 				tp->t_flags &= ~TF_NODELAY;
382 			break;
383 
384 		case TCP_MAXSEG:
385 			error = sockopt_getint(sopt, &optval);
386 			if (error)
387 				break;
388 			if (optval > 0 && optval <= tp->t_peermss)
389 				tp->t_peermss = optval; /* limit on send size */
390 			else
391 				error = EINVAL;
392 			break;
393 #ifdef notyet
394 		case TCP_CONGCTL:
395 			/* XXX string overflow XXX */
396 			error = tcp_congctl_select(tp, sopt->sopt_data);
397 			break;
398 #endif
399 
400 		case TCP_KEEPIDLE:
401 			error = sockopt_get(sopt, &ui, sizeof(ui));
402 			if (error)
403 				break;
404 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
405 				tp->t_keepidle = ui;
406 				change_keepalive(so, tp);
407 			} else
408 				error = EINVAL;
409 			break;
410 
411 		case TCP_KEEPINTVL:
412 			error = sockopt_get(sopt, &ui, sizeof(ui));
413 			if (error)
414 				break;
415 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
416 				tp->t_keepintvl = ui;
417 				change_keepalive(so, tp);
418 			} else
419 				error = EINVAL;
420 			break;
421 
422 		case TCP_KEEPCNT:
423 			error = sockopt_get(sopt, &ui, sizeof(ui));
424 			if (error)
425 				break;
426 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
427 				tp->t_keepcnt = ui;
428 				change_keepalive(so, tp);
429 			} else
430 				error = EINVAL;
431 			break;
432 
433 		case TCP_KEEPINIT:
434 			error = sockopt_get(sopt, &ui, sizeof(ui));
435 			if (error)
436 				break;
437 			if (ui > 0 && ui <= TCP_TIMER_MAXTICKS) {
438 				tp->t_keepinit = ui;
439 				change_keepalive(so, tp);
440 			} else
441 				error = EINVAL;
442 			break;
443 
444 		default:
445 			error = ENOPROTOOPT;
446 			break;
447 		}
448 		break;
449 
450 	case PRCO_GETOPT:
451 		switch (optname) {
452 #ifdef TCP_SIGNATURE
453 		case TCP_MD5SIG:
454 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
455 			goto setval;
456 #endif
457 		case TCP_NODELAY:
458 			optval = tp->t_flags & TF_NODELAY;
459 			goto setval;
460 		case TCP_MAXSEG:
461 			optval = tp->t_peermss;
462 			goto setval;
463 		case TCP_INFO:
464 			tcp_fill_info(tp, &ti);
465 			error = sockopt_set(sopt, &ti, sizeof ti);
466 			break;
467 #ifdef notyet
468 		case TCP_CONGCTL:
469 			break;
470 #endif
471 		case TCP_KEEPIDLE:
472 			optval = tp->t_keepidle;
473 			goto setval;
474 		case TCP_KEEPINTVL:
475 			optval = tp->t_keepintvl;
476 			goto setval;
477 		case TCP_KEEPCNT:
478 			optval = tp->t_keepcnt;
479 			goto setval;
480 		case TCP_KEEPINIT:
481 			optval = tp->t_keepcnt;
482 setval:			error = sockopt_set(sopt, &optval, sizeof(optval));
483 			break;
484 		default:
485 			error = ENOPROTOOPT;
486 			break;
487 		}
488 		break;
489 	}
490 	splx(s);
491 	return error;
492 }
493 
494 #ifndef TCP_SENDSPACE
495 #define	TCP_SENDSPACE	1024*32
496 #endif
497 int	tcp_sendspace = TCP_SENDSPACE;
498 #ifndef TCP_RECVSPACE
499 #define	TCP_RECVSPACE	1024*32
500 #endif
501 int	tcp_recvspace = TCP_RECVSPACE;
502 
503 /*
504  * tcp_attach: attach TCP protocol to socket, allocating internet protocol
505  * control block, TCP control block, buffer space and entering LISTEN state
506  * if to accept connections.
507  */
508 static int
509 tcp_attach(struct socket *so, int proto)
510 {
511 	struct tcpcb *tp;
512 	struct inpcb *inp;
513 #ifdef INET6
514 	struct in6pcb *in6p;
515 #endif
516 	int s, error, family;
517 
518 	/* Assign the lock (must happen even if we will error out). */
519 	s = splsoftnet();
520 	sosetlock(so);
521 	KASSERT(solocked(so));
522 
523 	family = so->so_proto->pr_domain->dom_family;
524 	switch (family) {
525 	case PF_INET:
526 		inp = sotoinpcb(so);
527 #ifdef INET6
528 		in6p = NULL;
529 #endif
530 		break;
531 #ifdef INET6
532 	case PF_INET6:
533 		inp = NULL;
534 		in6p = sotoin6pcb(so);
535 		break;
536 #endif
537 	default:
538 		error = EAFNOSUPPORT;
539 		goto out;
540 	}
541 
542 	KASSERT(inp == NULL);
543 #ifdef INET6
544 	KASSERT(in6p == NULL);
545 #endif
546 
547 #ifdef MBUFTRACE
548 	so->so_mowner = &tcp_sock_mowner;
549 	so->so_rcv.sb_mowner = &tcp_sock_rx_mowner;
550 	so->so_snd.sb_mowner = &tcp_sock_tx_mowner;
551 #endif
552 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
553 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
554 		if (error)
555 			goto out;
556 	}
557 
558 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
559 	so->so_snd.sb_flags |= SB_AUTOSIZE;
560 
561 	switch (family) {
562 	case PF_INET:
563 		error = in_pcballoc(so, &tcbtable);
564 		if (error)
565 			goto out;
566 		inp = sotoinpcb(so);
567 #ifdef INET6
568 		in6p = NULL;
569 #endif
570 		break;
571 #ifdef INET6
572 	case PF_INET6:
573 		error = in6_pcballoc(so, &tcbtable);
574 		if (error)
575 			goto out;
576 		inp = NULL;
577 		in6p = sotoin6pcb(so);
578 		break;
579 #endif
580 	default:
581 		error = EAFNOSUPPORT;
582 		goto out;
583 	}
584 	if (inp)
585 		tp = tcp_newtcpcb(family, (void *)inp);
586 #ifdef INET6
587 	else if (in6p)
588 		tp = tcp_newtcpcb(family, (void *)in6p);
589 #endif
590 	else
591 		tp = NULL;
592 
593 	if (tp == NULL) {
594 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
595 
596 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
597 		if (inp)
598 			in_pcbdetach(inp);
599 #ifdef INET6
600 		if (in6p)
601 			in6_pcbdetach(in6p);
602 #endif
603 		so->so_state |= nofd;
604 		error = ENOBUFS;
605 		goto out;
606 	}
607 	tp->t_state = TCPS_CLOSED;
608 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
609 		so->so_linger = TCP_LINGERTIME;
610 	}
611 out:
612 	KASSERT(solocked(so));
613 	splx(s);
614 	return error;
615 }
616 
617 static void
618 tcp_detach(struct socket *so)
619 {
620 	struct inpcb *inp = NULL;
621 	struct in6pcb *in6p = NULL;
622 	struct tcpcb *tp = NULL;
623 	int s;
624 
625 	if (tcp_getpcb(so, &inp, &in6p, &tp) != 0)
626 		return;
627 
628 	s = splsoftnet();
629 	(void)tcp_disconnect1(tp);
630 	splx(s);
631 }
632 
633 static int
634 tcp_accept(struct socket *so, struct sockaddr *nam)
635 {
636 	struct inpcb *inp = NULL;
637 	struct in6pcb *in6p = NULL;
638 	struct tcpcb *tp = NULL;
639 	int ostate = 0;
640 	int error = 0;
641 	int s;
642 
643 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
644 		return error;
645 
646 	ostate = tcp_debug_capture(tp, PRU_ACCEPT);
647 
648 	/*
649 	 * Accept a connection.  Essentially all the work is
650 	 * done at higher levels; just return the address
651 	 * of the peer, storing through addr.
652 	 */
653 	s = splsoftnet();
654 	if (inp) {
655 		in_setpeeraddr(inp, (struct sockaddr_in *)nam);
656 	}
657 #ifdef INET6
658 	if (in6p) {
659 		in6_setpeeraddr(in6p, (struct sockaddr_in6 *)nam);
660 	}
661 #endif
662 	tcp_debug_trace(so, tp, ostate, PRU_ACCEPT);
663 	splx(s);
664 
665 	return 0;
666 }
667 
668 static int
669 tcp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
670 {
671 	struct inpcb *inp = NULL;
672 	struct in6pcb *in6p = NULL;
673 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
674 #ifdef INET6
675 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
676 #endif /* INET6 */
677 	struct tcpcb *tp = NULL;
678 	int s;
679 	int error = 0;
680 	int ostate = 0;
681 
682 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
683 		return error;
684 
685 	ostate = tcp_debug_capture(tp, PRU_BIND);
686 
687 	/*
688 	 * Give the socket an address.
689 	 */
690 	s = splsoftnet();
691 	switch (so->so_proto->pr_domain->dom_family) {
692 	case PF_INET:
693 		error = in_pcbbind(inp, sin, l);
694 		break;
695 #ifdef INET6
696 	case PF_INET6:
697 		error = in6_pcbbind(in6p, sin6, l);
698 		if (!error) {
699 			/* mapped addr case */
700 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
701 				tp->t_family = AF_INET;
702 			else
703 				tp->t_family = AF_INET6;
704 		}
705 		break;
706 #endif
707 	}
708 	tcp_debug_trace(so, tp, ostate, PRU_BIND);
709 	splx(s);
710 
711 	return error;
712 }
713 
714 static int
715 tcp_listen(struct socket *so, struct lwp *l)
716 {
717 	struct inpcb *inp = NULL;
718 	struct in6pcb *in6p = NULL;
719 	struct tcpcb *tp = NULL;
720 	int error = 0;
721 	int ostate = 0;
722 	int s;
723 
724 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
725 		return error;
726 
727 	ostate = tcp_debug_capture(tp, PRU_LISTEN);
728 
729 	/*
730 	 * Prepare to accept connections.
731 	 */
732 	s = splsoftnet();
733 	if (inp && inp->inp_lport == 0) {
734 		error = in_pcbbind(inp, NULL, l);
735 		if (error)
736 			goto release;
737 	}
738 #ifdef INET6
739 	if (in6p && in6p->in6p_lport == 0) {
740 		error = in6_pcbbind(in6p, NULL, l);
741 		if (error)
742 			goto release;
743 	}
744 #endif
745 	tp->t_state = TCPS_LISTEN;
746 
747 release:
748 	tcp_debug_trace(so, tp, ostate, PRU_LISTEN);
749 	splx(s);
750 
751 	return error;
752 }
753 
754 static int
755 tcp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
756 {
757 	struct inpcb *inp = NULL;
758 	struct in6pcb *in6p = NULL;
759 	struct tcpcb *tp = NULL;
760 	int s;
761 	int error = 0;
762 	int ostate = 0;
763 
764 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
765 		return error;
766 
767 	ostate = tcp_debug_capture(tp, PRU_CONNECT);
768 
769 	/*
770 	 * Initiate connection to peer.
771 	 * Create a template for use in transmissions on this connection.
772 	 * Enter SYN_SENT state, and mark socket as connecting.
773 	 * Start keep-alive timer, and seed output sequence space.
774 	 * Send initial segment on connection.
775 	 */
776 	s = splsoftnet();
777 
778 	if (inp) {
779 		if (inp->inp_lport == 0) {
780 			error = in_pcbbind(inp, NULL, l);
781 			if (error)
782 				goto release;
783 		}
784 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
785 	}
786 #ifdef INET6
787 	if (in6p) {
788 		if (in6p->in6p_lport == 0) {
789 			error = in6_pcbbind(in6p, NULL, l);
790 			if (error)
791 				goto release;
792 		}
793 		error = in6_pcbconnect(in6p, (struct sockaddr_in6 *)nam, l);
794 		if (!error) {
795 			/* mapped addr case */
796 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
797 				tp->t_family = AF_INET;
798 			else
799 				tp->t_family = AF_INET6;
800 		}
801 	}
802 #endif
803 	if (error)
804 		goto release;
805 	tp->t_template = tcp_template(tp);
806 	if (tp->t_template == 0) {
807 		if (inp)
808 			in_pcbdisconnect(inp);
809 #ifdef INET6
810 		if (in6p)
811 			in6_pcbdisconnect(in6p);
812 #endif
813 		error = ENOBUFS;
814 		goto release;
815 	}
816 	/*
817 	 * Compute window scaling to request.
818 	 * XXX: This should be moved to tcp_output().
819 	 */
820 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
821 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
822 		tp->request_r_scale++;
823 	soisconnecting(so);
824 	TCP_STATINC(TCP_STAT_CONNATTEMPT);
825 	tp->t_state = TCPS_SYN_SENT;
826 	TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
827 	tp->iss = tcp_new_iss(tp, 0);
828 	tcp_sendseqinit(tp);
829 	error = tcp_output(tp);
830 
831 release:
832 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT);
833 	splx(s);
834 
835 	return error;
836 }
837 
838 static int
839 tcp_connect2(struct socket *so, struct socket *so2)
840 {
841 	struct inpcb *inp = NULL;
842 	struct in6pcb *in6p = NULL;
843 	struct tcpcb *tp = NULL;
844 	int error = 0;
845 	int ostate = 0;
846 
847 	KASSERT(solocked(so));
848 
849 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
850 		return error;
851 
852 	ostate = tcp_debug_capture(tp, PRU_CONNECT2);
853 
854 	tcp_debug_trace(so, tp, ostate, PRU_CONNECT2);
855 
856 	return EOPNOTSUPP;
857 }
858 
859 static int
860 tcp_disconnect(struct socket *so)
861 {
862 	struct inpcb *inp = NULL;
863 	struct in6pcb *in6p = NULL;
864 	struct tcpcb *tp = NULL;
865 	int error = 0;
866 	int ostate = 0;
867 	int s;
868 
869 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
870 		return error;
871 
872 	ostate = tcp_debug_capture(tp, PRU_DISCONNECT);
873 
874 	/*
875 	 * Initiate disconnect from peer.
876 	 * If connection never passed embryonic stage, just drop;
877 	 * else if don't need to let data drain, then can just drop anyways,
878 	 * else have to begin TCP shutdown process: mark socket disconnecting,
879 	 * drain unread data, state switch to reflect user close, and
880 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
881 	 * when peer sends FIN and acks ours.
882 	 *
883 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
884 	 */
885 	s = splsoftnet();
886 	tp = tcp_disconnect1(tp);
887 	tcp_debug_trace(so, tp, ostate, PRU_DISCONNECT);
888 	splx(s);
889 
890 	return error;
891 }
892 
893 static int
894 tcp_shutdown(struct socket *so)
895 {
896 	struct inpcb *inp = NULL;
897 	struct in6pcb *in6p = NULL;
898 	struct tcpcb *tp = NULL;
899 	int error = 0;
900 	int ostate = 0;
901 	int s;
902 
903 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
904 		return error;
905 
906 	ostate = tcp_debug_capture(tp, PRU_SHUTDOWN);
907 	/*
908 	 * Mark the connection as being incapable of further output.
909 	 */
910 	s = splsoftnet();
911 	socantsendmore(so);
912 	tp = tcp_usrclosed(tp);
913 	if (tp)
914 		error = tcp_output(tp);
915 	tcp_debug_trace(so, tp, ostate, PRU_SHUTDOWN);
916 	splx(s);
917 
918 	return error;
919 }
920 
921 static int
922 tcp_abort(struct socket *so)
923 {
924 	struct inpcb *inp = NULL;
925 	struct in6pcb *in6p = NULL;
926 	struct tcpcb *tp = NULL;
927 	int error = 0;
928 	int ostate = 0;
929 	int s;
930 
931 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
932 		return error;
933 
934 	ostate = tcp_debug_capture(tp, PRU_ABORT);
935 
936 	/*
937 	 * Abort the TCP.
938 	 */
939 	s = splsoftnet();
940 	tp = tcp_drop(tp, ECONNABORTED);
941 	tcp_debug_trace(so, tp, ostate, PRU_ABORT);
942 	splx(s);
943 
944 	return error;
945 }
946 
947 static int
948 tcp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
949 {
950 	switch (so->so_proto->pr_domain->dom_family) {
951 	case PF_INET:
952 		return in_control(so, cmd, nam, ifp);
953 #ifdef INET6
954 	case PF_INET6:
955 		return in6_control(so, cmd, nam, ifp);
956 #endif
957 	default:
958 		return EAFNOSUPPORT;
959 	}
960 }
961 
962 static int
963 tcp_stat(struct socket *so, struct stat *ub)
964 {
965 	KASSERT(solocked(so));
966 
967 	/* stat: don't bother with a blocksize.  */
968 	return 0;
969 }
970 
971 static int
972 tcp_peeraddr(struct socket *so, struct sockaddr *nam)
973 {
974 	struct inpcb *inp = NULL;
975 	struct in6pcb *in6p = NULL;
976 	struct tcpcb *tp = NULL;
977 	int ostate = 0;
978 	int error = 0;
979 	int s;
980 
981 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
982 		return error;
983 
984 	ostate = tcp_debug_capture(tp, PRU_PEERADDR);
985 
986 	s = splsoftnet();
987 	if (inp) {
988 		in_setpeeraddr(inp, (struct sockaddr_in *)nam);
989 	}
990 #ifdef INET6
991 	if (in6p) {
992 		in6_setpeeraddr(in6p, (struct sockaddr_in6 *)nam);
993 	}
994 #endif
995 	tcp_debug_trace(so, tp, ostate, PRU_PEERADDR);
996 	splx(s);
997 
998 	return 0;
999 }
1000 
1001 static int
1002 tcp_sockaddr(struct socket *so, struct sockaddr *nam)
1003 {
1004 	struct inpcb *inp = NULL;
1005 	struct in6pcb *in6p = NULL;
1006 	struct tcpcb *tp = NULL;
1007 	int ostate = 0;
1008 	int error = 0;
1009 	int s;
1010 
1011 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1012 		return error;
1013 
1014 	ostate = tcp_debug_capture(tp, PRU_SOCKADDR);
1015 
1016 	s = splsoftnet();
1017 	if (inp) {
1018 		in_setsockaddr(inp, (struct sockaddr_in *)nam);
1019 	}
1020 #ifdef INET6
1021 	if (in6p) {
1022 		in6_setsockaddr(in6p, (struct sockaddr_in6 *)nam);
1023 	}
1024 #endif
1025 	tcp_debug_trace(so, tp, ostate, PRU_SOCKADDR);
1026 	splx(s);
1027 
1028 	return 0;
1029 }
1030 
1031 static int
1032 tcp_rcvd(struct socket *so, int flags, struct lwp *l)
1033 {
1034 	struct inpcb *inp = NULL;
1035 	struct in6pcb *in6p = NULL;
1036 	struct tcpcb *tp = NULL;
1037 	int ostate = 0;
1038 	int error = 0;
1039 	int s;
1040 
1041 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1042 		return error;
1043 
1044 	ostate = tcp_debug_capture(tp, PRU_RCVD);
1045 
1046 	/*
1047 	 * After a receive, possibly send window update to peer.
1048 	 *
1049 	 * soreceive() calls this function when a user receives
1050 	 * ancillary data on a listening socket. We don't call
1051 	 * tcp_output in such a case, since there is no header
1052 	 * template for a listening socket and hence the kernel
1053 	 * will panic.
1054 	 */
1055 	s = splsoftnet();
1056 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
1057 		(void) tcp_output(tp);
1058 	splx(s);
1059 
1060 	tcp_debug_trace(so, tp, ostate, PRU_RCVD);
1061 
1062 	return 0;
1063 }
1064 
1065 static int
1066 tcp_recvoob(struct socket *so, struct mbuf *m, int flags)
1067 {
1068 	struct inpcb *inp = NULL;
1069 	struct in6pcb *in6p = NULL;
1070 	struct tcpcb *tp = NULL;
1071 	int ostate = 0;
1072 	int error = 0;
1073 	int s;
1074 
1075 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1076 		return error;
1077 
1078 	ostate = tcp_debug_capture(tp, PRU_RCVOOB);
1079 
1080 	s = splsoftnet();
1081 	if ((so->so_oobmark == 0 &&
1082 	    (so->so_state & SS_RCVATMARK) == 0) ||
1083 	    so->so_options & SO_OOBINLINE ||
1084 	    tp->t_oobflags & TCPOOB_HADDATA) {
1085 		splx(s);
1086 		return EINVAL;
1087 	}
1088 
1089 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1090 		splx(s);
1091 		return EWOULDBLOCK;
1092 	}
1093 
1094 	m->m_len = 1;
1095 	*mtod(m, char *) = tp->t_iobc;
1096 	if ((flags & MSG_PEEK) == 0)
1097 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1098 
1099 	tcp_debug_trace(so, tp, ostate, PRU_RCVOOB);
1100 	splx(s);
1101 
1102 	return 0;
1103 }
1104 
1105 static int
1106 tcp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1107     struct mbuf *control, struct lwp *l)
1108 {
1109 	struct inpcb *inp = NULL;
1110 	struct in6pcb *in6p = NULL;
1111 	struct tcpcb *tp = NULL;
1112 	int ostate = 0;
1113 	int error = 0;
1114 	int s;
1115 
1116 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0)
1117 		return error;
1118 
1119 	ostate = tcp_debug_capture(tp, PRU_SEND);
1120 
1121 	/*
1122 	 * Do a send by putting data in output queue and updating urgent
1123 	 * marker if URG set.  Possibly send more data.
1124 	 */
1125 	s = splsoftnet();
1126 	if (control && control->m_len) {
1127 		m_freem(control);
1128 		m_freem(m);
1129 		tcp_debug_trace(so, tp, ostate, PRU_SEND);
1130 		splx(s);
1131 		return EINVAL;
1132 	}
1133 
1134 	sbappendstream(&so->so_snd, m);
1135 	error = tcp_output(tp);
1136 	tcp_debug_trace(so, tp, ostate, PRU_SEND);
1137 	splx(s);
1138 
1139 	return error;
1140 }
1141 
1142 static int
1143 tcp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1144 {
1145 	struct inpcb *inp = NULL;
1146 	struct in6pcb *in6p = NULL;
1147 	struct tcpcb *tp = NULL;
1148 	int ostate = 0;
1149 	int error = 0;
1150 	int s;
1151 
1152 	if ((error = tcp_getpcb(so, &inp, &in6p, &tp)) != 0) {
1153 		m_freem(m);
1154 		m_freem(control);
1155 		return error;
1156 	}
1157 
1158 	ostate = tcp_debug_capture(tp, PRU_SENDOOB);
1159 
1160 	s = splsoftnet();
1161 	if (sbspace_oob(&so->so_snd) == 0) {
1162 		m_freem(m);
1163 		m_freem(control);
1164 		splx(s);
1165 		return ENOBUFS;
1166 	}
1167 	/*
1168 	 * According to RFC961 (Assigned Protocols),
1169 	 * the urgent pointer points to the last octet
1170 	 * of urgent data.  We continue, however,
1171 	 * to consider it to indicate the first octet
1172 	 * of data past the urgent section.
1173 	 * Otherwise, snd_up should be one lower.
1174 	 */
1175 	sbappendstream(&so->so_snd, m);
1176 	tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1177 	tp->t_force = 1;
1178 	error = tcp_output(tp);
1179 	tp->t_force = 0;
1180 	tcp_debug_trace(so, tp, ostate, PRU_SENDOOB);
1181 	splx(s);
1182 	m_freem(control);
1183 
1184 	return error;
1185 }
1186 
1187 static int
1188 tcp_purgeif(struct socket *so, struct ifnet *ifp)
1189 {
1190 	int s;
1191 	int error = 0;
1192 
1193 	s = splsoftnet();
1194 
1195 	mutex_enter(softnet_lock);
1196 	switch (so->so_proto->pr_domain->dom_family) {
1197 	case PF_INET:
1198 		in_pcbpurgeif0(&tcbtable, ifp);
1199 #ifdef NET_MPSAFE
1200 		mutex_exit(softnet_lock);
1201 #endif
1202 		in_purgeif(ifp);
1203 #ifdef NET_MPSAFE
1204 		mutex_enter(softnet_lock);
1205 #endif
1206 		in_pcbpurgeif(&tcbtable, ifp);
1207 		break;
1208 #ifdef INET6
1209 	case PF_INET6:
1210 		in6_pcbpurgeif0(&tcbtable, ifp);
1211 #ifdef NET_MPSAFE
1212 		mutex_exit(softnet_lock);
1213 #endif
1214 		in6_purgeif(ifp);
1215 #ifdef NET_MPSAFE
1216 		mutex_enter(softnet_lock);
1217 #endif
1218 		in6_pcbpurgeif(&tcbtable, ifp);
1219 		break;
1220 #endif
1221 	default:
1222 		error = EAFNOSUPPORT;
1223 		break;
1224 	}
1225 	mutex_exit(softnet_lock);
1226 	splx(s);
1227 
1228 	return error;
1229 }
1230 
1231 /*
1232  * Initiate (or continue) disconnect.
1233  * If embryonic state, just send reset (once).
1234  * If in ``let data drain'' option and linger null, just drop.
1235  * Otherwise (hard), mark socket disconnecting and drop
1236  * current input data; switch states based on user close, and
1237  * send segment to peer (with FIN).
1238  */
1239 struct tcpcb *
1240 tcp_disconnect1(struct tcpcb *tp)
1241 {
1242 	struct socket *so;
1243 
1244 	if (tp->t_inpcb)
1245 		so = tp->t_inpcb->inp_socket;
1246 #ifdef INET6
1247 	else if (tp->t_in6pcb)
1248 		so = tp->t_in6pcb->in6p_socket;
1249 #endif
1250 	else
1251 		so = NULL;
1252 
1253 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
1254 		tp = tcp_close(tp);
1255 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1256 		tp = tcp_drop(tp, 0);
1257 	else {
1258 		soisdisconnecting(so);
1259 		sbflush(&so->so_rcv);
1260 		tp = tcp_usrclosed(tp);
1261 		if (tp)
1262 			(void) tcp_output(tp);
1263 	}
1264 	return tp;
1265 }
1266 
1267 /*
1268  * User issued close, and wish to trail through shutdown states:
1269  * if never received SYN, just forget it.  If got a SYN from peer,
1270  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1271  * If already got a FIN from peer, then almost done; go to LAST_ACK
1272  * state.  In all other cases, have already sent FIN to peer (e.g.
1273  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1274  * for peer to send FIN or not respond to keep-alives, etc.
1275  * We can let the user exit from the close as soon as the FIN is acked.
1276  */
1277 struct tcpcb *
1278 tcp_usrclosed(struct tcpcb *tp)
1279 {
1280 
1281 	switch (tp->t_state) {
1282 
1283 	case TCPS_CLOSED:
1284 	case TCPS_LISTEN:
1285 	case TCPS_SYN_SENT:
1286 		tp->t_state = TCPS_CLOSED;
1287 		tp = tcp_close(tp);
1288 		break;
1289 
1290 	case TCPS_SYN_RECEIVED:
1291 	case TCPS_ESTABLISHED:
1292 		tp->t_state = TCPS_FIN_WAIT_1;
1293 		break;
1294 
1295 	case TCPS_CLOSE_WAIT:
1296 		tp->t_state = TCPS_LAST_ACK;
1297 		break;
1298 	}
1299 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1300 		struct socket *so;
1301 		if (tp->t_inpcb)
1302 			so = tp->t_inpcb->inp_socket;
1303 #ifdef INET6
1304 		else if (tp->t_in6pcb)
1305 			so = tp->t_in6pcb->in6p_socket;
1306 #endif
1307 		else
1308 			so = NULL;
1309 		if (so)
1310 			soisdisconnected(so);
1311 		/*
1312 		 * If we are in FIN_WAIT_2, we arrived here because the
1313 		 * application did a shutdown of the send side.  Like the
1314 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
1315 		 * a full close, we start a timer to make sure sockets are
1316 		 * not left in FIN_WAIT_2 forever.
1317 		 */
1318 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
1319 			TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
1320 		else if (tp->t_state == TCPS_TIME_WAIT
1321 			 && ((tp->t_inpcb
1322 			      && (tcp4_vtw_enable & 1)
1323 			      && vtw_add(AF_INET, tp))
1324 			     ||
1325 			     (tp->t_in6pcb
1326 			      && (tcp6_vtw_enable & 1)
1327 			      && vtw_add(AF_INET6, tp)))) {
1328 			tp = 0;
1329 		}
1330 	}
1331 	return tp;
1332 }
1333 
1334 /*
1335  * sysctl helper routine for net.inet.ip.mssdflt.  it can't be less
1336  * than 32.
1337  */
1338 static int
1339 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS)
1340 {
1341 	int error, mssdflt;
1342 	struct sysctlnode node;
1343 
1344 	mssdflt = tcp_mssdflt;
1345 	node = *rnode;
1346 	node.sysctl_data = &mssdflt;
1347 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1348 	if (error || newp == NULL)
1349 		return error;
1350 
1351 	if (mssdflt < 32)
1352 		return EINVAL;
1353 	tcp_mssdflt = mssdflt;
1354 
1355 	mutex_enter(softnet_lock);
1356 	tcp_tcpcb_template();
1357 	mutex_exit(softnet_lock);
1358 
1359 	return 0;
1360 }
1361 
1362 /*
1363  * sysctl helper for TCP CB template update
1364  */
1365 static int
1366 sysctl_update_tcpcb_template(SYSCTLFN_ARGS)
1367 {
1368 	int t, error;
1369 	struct sysctlnode node;
1370 
1371 	/* follow procedures in sysctl(9) manpage */
1372 	t = *(int *)rnode->sysctl_data;
1373 	node = *rnode;
1374 	node.sysctl_data = &t;
1375 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1376 	if (error || newp == NULL)
1377 		return error;
1378 
1379 	if (t < 0)
1380 		return EINVAL;
1381 
1382 	*(int *)rnode->sysctl_data = t;
1383 
1384 	mutex_enter(softnet_lock);
1385 	tcp_tcpcb_template();
1386 	mutex_exit(softnet_lock);
1387 
1388 	return 0;
1389 }
1390 
1391 /*
1392  * sysctl helper routine for setting port related values under
1393  * net.inet.ip and net.inet6.ip6.  does basic range checking and does
1394  * additional checks for each type.  this code has placed in
1395  * tcp_input.c since INET and INET6 both use the same tcp code.
1396  *
1397  * this helper is not static so that both inet and inet6 can use it.
1398  */
1399 int
1400 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS)
1401 {
1402 	int error, tmp;
1403 	int apmin, apmax;
1404 #ifndef IPNOPRIVPORTS
1405 	int lpmin, lpmax;
1406 #endif /* IPNOPRIVPORTS */
1407 	struct sysctlnode node;
1408 
1409 	if (namelen != 0)
1410 		return EINVAL;
1411 
1412 	switch (name[-3]) {
1413 	    case PF_INET:
1414 		apmin = anonportmin;
1415 		apmax = anonportmax;
1416 #ifndef IPNOPRIVPORTS
1417 		lpmin = lowportmin;
1418 		lpmax = lowportmax;
1419 #endif /* IPNOPRIVPORTS */
1420 		break;
1421 #ifdef INET6
1422 	    case PF_INET6:
1423 		apmin = ip6_anonportmin;
1424 		apmax = ip6_anonportmax;
1425 #ifndef IPNOPRIVPORTS
1426 		lpmin = ip6_lowportmin;
1427 		lpmax = ip6_lowportmax;
1428 #endif /* IPNOPRIVPORTS */
1429 		break;
1430 #endif /* INET6 */
1431 	    default:
1432 		return EINVAL;
1433 	}
1434 
1435 	/*
1436 	 * insert temporary copy into node, perform lookup on
1437 	 * temporary, then restore pointer
1438 	 */
1439 	node = *rnode;
1440 	tmp = *(int*)rnode->sysctl_data;
1441 	node.sysctl_data = &tmp;
1442 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1443 	if (error || newp == NULL)
1444 		return error;
1445 
1446 	/*
1447 	 * simple port range check
1448 	 */
1449 	if (tmp < 0 || tmp > 65535)
1450 		return EINVAL;
1451 
1452 	/*
1453 	 * per-node range checks
1454 	 */
1455 	switch (rnode->sysctl_num) {
1456 	case IPCTL_ANONPORTMIN:
1457 	case IPV6CTL_ANONPORTMIN:
1458 		if (tmp >= apmax)
1459 			return EINVAL;
1460 #ifndef IPNOPRIVPORTS
1461 		if (tmp < IPPORT_RESERVED)
1462                         return EINVAL;
1463 #endif /* IPNOPRIVPORTS */
1464 		break;
1465 
1466 	case IPCTL_ANONPORTMAX:
1467 	case IPV6CTL_ANONPORTMAX:
1468                 if (apmin >= tmp)
1469 			return EINVAL;
1470 #ifndef IPNOPRIVPORTS
1471 		if (tmp < IPPORT_RESERVED)
1472                         return EINVAL;
1473 #endif /* IPNOPRIVPORTS */
1474 		break;
1475 
1476 #ifndef IPNOPRIVPORTS
1477 	case IPCTL_LOWPORTMIN:
1478 	case IPV6CTL_LOWPORTMIN:
1479 		if (tmp >= lpmax ||
1480 		    tmp > IPPORT_RESERVEDMAX ||
1481 		    tmp < IPPORT_RESERVEDMIN)
1482 			return EINVAL;
1483 		break;
1484 
1485 	case IPCTL_LOWPORTMAX:
1486 	case IPV6CTL_LOWPORTMAX:
1487 		if (lpmin >= tmp ||
1488 		    tmp > IPPORT_RESERVEDMAX ||
1489 		    tmp < IPPORT_RESERVEDMIN)
1490 			return EINVAL;
1491 		break;
1492 #endif /* IPNOPRIVPORTS */
1493 
1494 	default:
1495 		return EINVAL;
1496 	}
1497 
1498 	*(int*)rnode->sysctl_data = tmp;
1499 
1500 	return 0;
1501 }
1502 
1503 static inline int
1504 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp)
1505 {
1506 	if (oldp) {
1507 		size_t sz;
1508 		uid_t uid;
1509 		int error;
1510 
1511 		if (sockp->so_cred == NULL)
1512 			return EPERM;
1513 
1514 		uid = kauth_cred_geteuid(sockp->so_cred);
1515 		sz = MIN(sizeof(uid), *oldlenp);
1516 		if ((error = copyout(&uid, oldp, sz)) != 0)
1517 			return error;
1518 	}
1519 	*oldlenp = sizeof(uid_t);
1520 	return 0;
1521 }
1522 
1523 static inline int
1524 inet4_ident_core(struct in_addr raddr, u_int rport,
1525     struct in_addr laddr, u_int lport,
1526     void *oldp, size_t *oldlenp,
1527     struct lwp *l, int dodrop)
1528 {
1529 	struct inpcb *inp;
1530 	struct socket *sockp;
1531 
1532 	inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0);
1533 
1534 	if (inp == NULL || (sockp = inp->inp_socket) == NULL)
1535 		return ESRCH;
1536 
1537 	if (dodrop) {
1538 		struct tcpcb *tp;
1539 		int error;
1540 
1541 		if (inp == NULL || (tp = intotcpcb(inp)) == NULL ||
1542 		    (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0)
1543 			return ESRCH;
1544 
1545 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1546 		    KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL);
1547 		if (error)
1548 			return error;
1549 
1550 		(void)tcp_drop(tp, ECONNABORTED);
1551 		return 0;
1552 	}
1553 
1554 	return copyout_uid(sockp, oldp, oldlenp);
1555 }
1556 
1557 #ifdef INET6
1558 static inline int
1559 inet6_ident_core(struct in6_addr *raddr, u_int rport,
1560     struct in6_addr *laddr, u_int lport,
1561     void *oldp, size_t *oldlenp,
1562     struct lwp *l, int dodrop)
1563 {
1564 	struct in6pcb *in6p;
1565 	struct socket *sockp;
1566 
1567 	in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0);
1568 
1569 	if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL)
1570 		return ESRCH;
1571 
1572 	if (dodrop) {
1573 		struct tcpcb *tp;
1574 		int error;
1575 
1576 		if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL ||
1577 		    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0)
1578 			return ESRCH;
1579 
1580 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1581 		    KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL);
1582 		if (error)
1583 			return error;
1584 
1585 		(void)tcp_drop(tp, ECONNABORTED);
1586 		return 0;
1587 	}
1588 
1589 	return copyout_uid(sockp, oldp, oldlenp);
1590 }
1591 #endif
1592 
1593 /*
1594  * sysctl helper routine for the net.inet.tcp.drop and
1595  * net.inet6.tcp6.drop nodes.
1596  */
1597 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident
1598 
1599 /*
1600  * sysctl helper routine for the net.inet.tcp.ident and
1601  * net.inet6.tcp6.ident nodes.  contains backwards compat code for the
1602  * old way of looking up the ident information for ipv4 which involves
1603  * stuffing the port/addr pairs into the mib lookup.
1604  */
1605 static int
1606 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS)
1607 {
1608 	struct sockaddr_in *si4[2];
1609 #ifdef INET6
1610 	struct sockaddr_in6 *si6[2];
1611 #endif
1612 	struct sockaddr_storage sa[2];
1613 	int error, pf, dodrop;
1614 
1615 	dodrop = name[-1] == TCPCTL_DROP;
1616 	if (dodrop) {
1617 		if (oldp != NULL || *oldlenp != 0)
1618 			return EINVAL;
1619 		if (newp == NULL)
1620 			return EPERM;
1621 		if (newlen < sizeof(sa))
1622 			return ENOMEM;
1623 	}
1624 	if (namelen != 4 && namelen != 0)
1625 		return EINVAL;
1626 	if (name[-2] != IPPROTO_TCP)
1627 		return EINVAL;
1628 	pf = name[-3];
1629 
1630 	/* old style lookup, ipv4 only */
1631 	if (namelen == 4) {
1632 		struct in_addr laddr, raddr;
1633 		u_int lport, rport;
1634 
1635 		if (pf != PF_INET)
1636 			return EPROTONOSUPPORT;
1637 		raddr.s_addr = (uint32_t)name[0];
1638 		rport = (u_int)name[1];
1639 		laddr.s_addr = (uint32_t)name[2];
1640 		lport = (u_int)name[3];
1641 
1642 		mutex_enter(softnet_lock);
1643 		error = inet4_ident_core(raddr, rport, laddr, lport,
1644 		    oldp, oldlenp, l, dodrop);
1645 		mutex_exit(softnet_lock);
1646 		return error;
1647 	}
1648 
1649 	if (newp == NULL || newlen != sizeof(sa))
1650 		return EINVAL;
1651 	error = copyin(newp, &sa, newlen);
1652 	if (error)
1653 		return error;
1654 
1655 	/*
1656 	 * requested families must match
1657 	 */
1658 	if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family)
1659 		return EINVAL;
1660 
1661 	switch (pf) {
1662 #ifdef INET6
1663 	case PF_INET6:
1664 		si6[0] = (struct sockaddr_in6*)&sa[0];
1665 		si6[1] = (struct sockaddr_in6*)&sa[1];
1666 		if (si6[0]->sin6_len != sizeof(*si6[0]) ||
1667 		    si6[1]->sin6_len != sizeof(*si6[1]))
1668 			return EINVAL;
1669 
1670 		if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) &&
1671 		    !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) {
1672 			error = sa6_embedscope(si6[0], ip6_use_defzone);
1673 			if (error)
1674 				return error;
1675 			error = sa6_embedscope(si6[1], ip6_use_defzone);
1676 			if (error)
1677 				return error;
1678 
1679 			mutex_enter(softnet_lock);
1680 			error = inet6_ident_core(&si6[0]->sin6_addr,
1681 			    si6[0]->sin6_port, &si6[1]->sin6_addr,
1682 			    si6[1]->sin6_port, oldp, oldlenp, l, dodrop);
1683 			mutex_exit(softnet_lock);
1684 			return error;
1685 		}
1686 
1687 		if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) !=
1688 		    IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr))
1689 			return EINVAL;
1690 
1691 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]);
1692 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]);
1693 #endif /* INET6 */
1694 		/*FALLTHROUGH*/
1695 	case PF_INET:
1696 		si4[0] = (struct sockaddr_in*)&sa[0];
1697 		si4[1] = (struct sockaddr_in*)&sa[1];
1698 		if (si4[0]->sin_len != sizeof(*si4[0]) ||
1699 		    si4[0]->sin_len != sizeof(*si4[1]))
1700 			return EINVAL;
1701 
1702 		mutex_enter(softnet_lock);
1703 		error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port,
1704 		    si4[1]->sin_addr, si4[1]->sin_port,
1705 		    oldp, oldlenp, l, dodrop);
1706 		mutex_exit(softnet_lock);
1707 		return error;
1708 	default:
1709 		return EPROTONOSUPPORT;
1710 	}
1711 }
1712 
1713 /*
1714  * sysctl helper for the inet and inet6 pcblists.  handles tcp/udp and
1715  * inet/inet6, as well as raw pcbs for each.  specifically not
1716  * declared static so that raw sockets and udp/udp6 can use it as
1717  * well.
1718  */
1719 int
1720 sysctl_inpcblist(SYSCTLFN_ARGS)
1721 {
1722 	const bool allowaddr = get_expose_address(curproc);
1723 	struct sockaddr_in *in;
1724 	const struct inpcb *inp;
1725 #ifdef INET6
1726 	struct sockaddr_in6 *in6;
1727 	const struct in6pcb *in6p;
1728 #endif
1729 	struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data);
1730 	const struct inpcb_hdr *inph;
1731 	struct tcpcb *tp;
1732 	struct kinfo_pcb pcb;
1733 	char *dp;
1734 	size_t len, needed, elem_size, out_size;
1735 	int error, elem_count, pf, proto, pf2;
1736 
1737 	if (namelen != 4)
1738 		return EINVAL;
1739 
1740 	if (oldp != NULL) {
1741 		    len = *oldlenp;
1742 		    elem_size = name[2];
1743 		    elem_count = name[3];
1744 		    if (elem_size != sizeof(pcb))
1745 			    return EINVAL;
1746 	} else {
1747 		    len = 0;
1748 		    elem_count = INT_MAX;
1749 		    elem_size = sizeof(pcb);
1750 	}
1751 	error = 0;
1752 	dp = oldp;
1753 	out_size = elem_size;
1754 	needed = 0;
1755 
1756 	if (namelen == 1 && name[0] == CTL_QUERY)
1757 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1758 
1759 	if (name - oname != 4)
1760 		return EINVAL;
1761 
1762 	pf = oname[1];
1763 	proto = oname[2];
1764 	pf2 = (oldp != NULL) ? pf : 0;
1765 
1766 	mutex_enter(softnet_lock);
1767 
1768 	TAILQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) {
1769 		inp = (const struct inpcb *)inph;
1770 #ifdef INET6
1771 		in6p = (const struct in6pcb *)inph;
1772 #endif
1773 
1774 		if (inph->inph_af != pf)
1775 			continue;
1776 
1777 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1778 		    KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL,
1779 		    NULL) != 0)
1780 			continue;
1781 
1782 		memset(&pcb, 0, sizeof(pcb));
1783 
1784 		pcb.ki_family = pf;
1785 		pcb.ki_type = proto;
1786 
1787 		switch (pf2) {
1788 		case 0:
1789 			/* just probing for size */
1790 			break;
1791 		case PF_INET:
1792 			pcb.ki_family = inp->inp_socket->so_proto->
1793 			    pr_domain->dom_family;
1794 			pcb.ki_type = inp->inp_socket->so_proto->
1795 			    pr_type;
1796 			pcb.ki_protocol = inp->inp_socket->so_proto->
1797 			    pr_protocol;
1798 			pcb.ki_pflags = inp->inp_flags;
1799 
1800 			pcb.ki_sostate = inp->inp_socket->so_state;
1801 			pcb.ki_prstate = inp->inp_state;
1802 			if (proto == IPPROTO_TCP) {
1803 				tp = intotcpcb(inp);
1804 				pcb.ki_tstate = tp->t_state;
1805 				pcb.ki_tflags = tp->t_flags;
1806 			}
1807 
1808 			COND_SET_VALUE(pcb.ki_pcbaddr,
1809 			    PTRTOUINT64(inp), allowaddr);
1810 			COND_SET_VALUE(pcb.ki_ppcbaddr,
1811 			    PTRTOUINT64(inp->inp_ppcb), allowaddr);
1812 			COND_SET_VALUE(pcb.ki_sockaddr,
1813 			    PTRTOUINT64(inp->inp_socket), allowaddr);
1814 
1815 			pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc;
1816 			pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc;
1817 
1818 			in = satosin(&pcb.ki_src);
1819 			in->sin_len = sizeof(*in);
1820 			in->sin_family = pf;
1821 			in->sin_port = inp->inp_lport;
1822 			in->sin_addr = inp->inp_laddr;
1823 			if (pcb.ki_prstate >= INP_CONNECTED) {
1824 				in = satosin(&pcb.ki_dst);
1825 				in->sin_len = sizeof(*in);
1826 				in->sin_family = pf;
1827 				in->sin_port = inp->inp_fport;
1828 				in->sin_addr = inp->inp_faddr;
1829 			}
1830 			break;
1831 #ifdef INET6
1832 		case PF_INET6:
1833 			pcb.ki_family = in6p->in6p_socket->so_proto->
1834 			    pr_domain->dom_family;
1835 			pcb.ki_type = in6p->in6p_socket->so_proto->pr_type;
1836 			pcb.ki_protocol = in6p->in6p_socket->so_proto->
1837 			    pr_protocol;
1838 			pcb.ki_pflags = in6p->in6p_flags;
1839 
1840 			pcb.ki_sostate = in6p->in6p_socket->so_state;
1841 			pcb.ki_prstate = in6p->in6p_state;
1842 			if (proto == IPPROTO_TCP) {
1843 				tp = in6totcpcb(in6p);
1844 				pcb.ki_tstate = tp->t_state;
1845 				pcb.ki_tflags = tp->t_flags;
1846 			}
1847 
1848 			COND_SET_VALUE(pcb.ki_pcbaddr,
1849 			    PTRTOUINT64(in6p), allowaddr);
1850 			COND_SET_VALUE(pcb.ki_ppcbaddr,
1851 			    PTRTOUINT64(in6p->in6p_ppcb), allowaddr);
1852 			COND_SET_VALUE(pcb.ki_sockaddr,
1853 			    PTRTOUINT64(in6p->in6p_socket), allowaddr);
1854 
1855 			pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc;
1856 			pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc;
1857 
1858 			in6 = satosin6(&pcb.ki_src);
1859 			in6->sin6_len = sizeof(*in6);
1860 			in6->sin6_family = pf;
1861 			in6->sin6_port = in6p->in6p_lport;
1862 			in6->sin6_flowinfo = in6p->in6p_flowinfo;
1863 			in6->sin6_addr = in6p->in6p_laddr;
1864 			in6->sin6_scope_id = 0; /* XXX? */
1865 
1866 			if (pcb.ki_prstate >= IN6P_CONNECTED) {
1867 				in6 = satosin6(&pcb.ki_dst);
1868 				in6->sin6_len = sizeof(*in6);
1869 				in6->sin6_family = pf;
1870 				in6->sin6_port = in6p->in6p_fport;
1871 				in6->sin6_flowinfo = in6p->in6p_flowinfo;
1872 				in6->sin6_addr = in6p->in6p_faddr;
1873 				in6->sin6_scope_id = 0; /* XXX? */
1874 			}
1875 			break;
1876 #endif
1877 		}
1878 
1879 		if (len >= elem_size && elem_count > 0) {
1880 			error = copyout(&pcb, dp, out_size);
1881 			if (error) {
1882 				mutex_exit(softnet_lock);
1883 				return error;
1884 			}
1885 			dp += elem_size;
1886 			len -= elem_size;
1887 		}
1888 		needed += elem_size;
1889 		if (elem_count > 0 && elem_count != INT_MAX)
1890 			elem_count--;
1891 	}
1892 
1893 	*oldlenp = needed;
1894 	if (oldp == NULL)
1895 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
1896 
1897 	mutex_exit(softnet_lock);
1898 
1899 	return error;
1900 }
1901 
1902 static int
1903 sysctl_tcp_congctl(SYSCTLFN_ARGS)
1904 {
1905 	struct sysctlnode node;
1906 	int error;
1907 	char newname[TCPCC_MAXLEN];
1908 
1909 	strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1);
1910 
1911 	node = *rnode;
1912 	node.sysctl_data = newname;
1913 	node.sysctl_size = sizeof(newname);
1914 
1915 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1916 
1917 	if (error ||
1918 	    newp == NULL ||
1919 	    strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0)
1920 		return error;
1921 
1922 	mutex_enter(softnet_lock);
1923 	error = tcp_congctl_select(NULL, newname);
1924 	mutex_exit(softnet_lock);
1925 
1926 	return error;
1927 }
1928 
1929 static int
1930 sysctl_tcp_init_win(SYSCTLFN_ARGS)
1931 {
1932 	int error;
1933 	u_int iw;
1934 	struct sysctlnode node;
1935 
1936 	iw = *(u_int *)rnode->sysctl_data;
1937 	node = *rnode;
1938 	node.sysctl_data = &iw;
1939 	node.sysctl_size = sizeof(iw);
1940 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1941 	if (error || newp == NULL)
1942 		return error;
1943 
1944 	if (iw >= __arraycount(tcp_init_win_max))
1945 		return EINVAL;
1946 	*(u_int *)rnode->sysctl_data = iw;
1947 	return 0;
1948 }
1949 
1950 static int
1951 sysctl_tcp_keep(SYSCTLFN_ARGS)
1952 {
1953 	int error;
1954 	u_int tmp;
1955 	struct sysctlnode node;
1956 
1957 	node = *rnode;
1958 	tmp = *(u_int *)rnode->sysctl_data;
1959 	node.sysctl_data = &tmp;
1960 
1961 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1962 	if (error || newp == NULL)
1963 		return error;
1964 
1965 	if (!(tmp > 0 && tmp <= TCP_TIMER_MAXTICKS))
1966 		return EINVAL;
1967 
1968 	mutex_enter(softnet_lock);
1969 
1970 	*(u_int *)rnode->sysctl_data = tmp;
1971 	tcp_tcpcb_template();	/* update the template */
1972 
1973 	mutex_exit(softnet_lock);
1974 	return 0;
1975 }
1976 
1977 static int
1978 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS)
1979 {
1980 
1981 	return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS));
1982 }
1983 
1984 /*
1985  * this (second stage) setup routine is a replacement for tcp_sysctl()
1986  * (which is currently used for ipv4 and ipv6)
1987  */
1988 static void
1989 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
1990 			   const char *tcpname)
1991 {
1992 	const struct sysctlnode *sack_node;
1993 	const struct sysctlnode *abc_node;
1994 	const struct sysctlnode *ecn_node;
1995 	const struct sysctlnode *congctl_node;
1996 	const struct sysctlnode *mslt_node;
1997 	const struct sysctlnode *vtw_node;
1998 #ifdef TCP_DEBUG
1999 	extern struct tcp_debug tcp_debug[TCP_NDEBUG];
2000 	extern int tcp_debx;
2001 #endif
2002 
2003 	sysctl_createv(clog, 0, NULL, NULL,
2004 		       CTLFLAG_PERMANENT,
2005 		       CTLTYPE_NODE, pfname, NULL,
2006 		       NULL, 0, NULL, 0,
2007 		       CTL_NET, pf, CTL_EOL);
2008 	sysctl_createv(clog, 0, NULL, NULL,
2009 		       CTLFLAG_PERMANENT,
2010 		       CTLTYPE_NODE, tcpname,
2011 		       SYSCTL_DESCR("TCP related settings"),
2012 		       NULL, 0, NULL, 0,
2013 		       CTL_NET, pf, IPPROTO_TCP, CTL_EOL);
2014 
2015 	sysctl_createv(clog, 0, NULL, NULL,
2016 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2017 		       CTLTYPE_INT, "rfc1323",
2018 		       SYSCTL_DESCR("Enable RFC1323 TCP extensions"),
2019 		       sysctl_update_tcpcb_template, 0, &tcp_do_rfc1323, 0,
2020 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL);
2021 	sysctl_createv(clog, 0, NULL, NULL,
2022 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2023 		       CTLTYPE_INT, "sendspace",
2024 		       SYSCTL_DESCR("Default TCP send buffer size"),
2025 		       NULL, 0, &tcp_sendspace, 0,
2026 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL);
2027 	sysctl_createv(clog, 0, NULL, NULL,
2028 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2029 		       CTLTYPE_INT, "recvspace",
2030 		       SYSCTL_DESCR("Default TCP receive buffer size"),
2031 		       NULL, 0, &tcp_recvspace, 0,
2032 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL);
2033 	sysctl_createv(clog, 0, NULL, NULL,
2034 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2035 		       CTLTYPE_INT, "mssdflt",
2036 		       SYSCTL_DESCR("Default maximum segment size"),
2037 		       sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0,
2038 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL);
2039 	sysctl_createv(clog, 0, NULL, NULL,
2040 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2041 		       CTLTYPE_INT, "minmss",
2042 		       SYSCTL_DESCR("Lower limit for TCP maximum segment size"),
2043 		       NULL, 0, &tcp_minmss, 0,
2044 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2045 	sysctl_createv(clog, 0, NULL, NULL,
2046 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2047 		       CTLTYPE_INT, "msl",
2048 		       SYSCTL_DESCR("Maximum Segment Life"),
2049 		       NULL, 0, &tcp_msl, 0,
2050 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL);
2051 	sysctl_createv(clog, 0, NULL, NULL,
2052 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2053 		       CTLTYPE_INT, "syn_cache_limit",
2054 		       SYSCTL_DESCR("Maximum number of entries in the TCP "
2055 				    "compressed state engine"),
2056 		       NULL, 0, &tcp_syn_cache_limit, 0,
2057 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT,
2058 		       CTL_EOL);
2059 	sysctl_createv(clog, 0, NULL, NULL,
2060 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2061 		       CTLTYPE_INT, "syn_bucket_limit",
2062 		       SYSCTL_DESCR("Maximum number of entries per hash "
2063 				    "bucket in the TCP compressed state "
2064 				    "engine"),
2065 		       NULL, 0, &tcp_syn_bucket_limit, 0,
2066 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT,
2067 		       CTL_EOL);
2068 #if 0 /* obsoleted */
2069 	sysctl_createv(clog, 0, NULL, NULL,
2070 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2071 		       CTLTYPE_INT, "syn_cache_interval",
2072 		       SYSCTL_DESCR("TCP compressed state engine's timer interval"),
2073 		       NULL, 0, &tcp_syn_cache_interval, 0,
2074 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER,
2075 		       CTL_EOL);
2076 #endif
2077 	sysctl_createv(clog, 0, NULL, NULL,
2078 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2079 		       CTLTYPE_INT, "init_win",
2080 		       SYSCTL_DESCR("Initial TCP congestion window"),
2081 		       sysctl_tcp_init_win, 0, &tcp_init_win, 0,
2082 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL);
2083 	sysctl_createv(clog, 0, NULL, NULL,
2084 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2085 		       CTLTYPE_INT, "mss_ifmtu",
2086 		       SYSCTL_DESCR("Use interface MTU for calculating MSS"),
2087 		       NULL, 0, &tcp_mss_ifmtu, 0,
2088 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL);
2089 	sysctl_createv(clog, 0, NULL, &sack_node,
2090 		       CTLFLAG_PERMANENT,
2091 		       CTLTYPE_NODE, "sack",
2092 		       SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
2093 		       NULL, 0, NULL, 0,
2094 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
2095 
2096 	/* Congctl subtree */
2097 	sysctl_createv(clog, 0, NULL, &congctl_node,
2098 		       CTLFLAG_PERMANENT,
2099 		       CTLTYPE_NODE, "congctl",
2100 		       SYSCTL_DESCR("TCP Congestion Control"),
2101 	    	       NULL, 0, NULL, 0,
2102 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2103 	sysctl_createv(clog, 0, &congctl_node, NULL,
2104 		       CTLFLAG_PERMANENT,
2105 		       CTLTYPE_STRING, "available",
2106 		       SYSCTL_DESCR("Available Congestion Control Mechanisms"),
2107 		       NULL, 0, tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL);
2108 	sysctl_createv(clog, 0, &congctl_node, NULL,
2109 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2110 		       CTLTYPE_STRING, "selected",
2111 		       SYSCTL_DESCR("Selected Congestion Control Mechanism"),
2112 		       sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN,
2113 		       CTL_CREATE, CTL_EOL);
2114 
2115 	sysctl_createv(clog, 0, NULL, NULL,
2116 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2117 		       CTLTYPE_INT, "win_scale",
2118 		       SYSCTL_DESCR("Use RFC1323 window scale options"),
2119 		       sysctl_update_tcpcb_template, 0, &tcp_do_win_scale, 0,
2120 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL);
2121 	sysctl_createv(clog, 0, NULL, NULL,
2122 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2123 		       CTLTYPE_INT, "timestamps",
2124 		       SYSCTL_DESCR("Use RFC1323 time stamp options"),
2125 		       sysctl_update_tcpcb_template, 0, &tcp_do_timestamps, 0,
2126 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL);
2127 	sysctl_createv(clog, 0, NULL, NULL,
2128 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2129 		       CTLTYPE_INT, "cwm",
2130 		       SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window "
2131 				    "Monitoring"),
2132 		       NULL, 0, &tcp_cwm, 0,
2133 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL);
2134 	sysctl_createv(clog, 0, NULL, NULL,
2135 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2136 		       CTLTYPE_INT, "cwm_burstsize",
2137 		       SYSCTL_DESCR("Congestion Window Monitoring allowed "
2138 				    "burst count in packets"),
2139 		       NULL, 0, &tcp_cwm_burstsize, 0,
2140 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE,
2141 		       CTL_EOL);
2142 	sysctl_createv(clog, 0, NULL, NULL,
2143 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2144 		       CTLTYPE_INT, "ack_on_push",
2145 		       SYSCTL_DESCR("Immediately return ACK when PSH is "
2146 				    "received"),
2147 		       NULL, 0, &tcp_ack_on_push, 0,
2148 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL);
2149 	sysctl_createv(clog, 0, NULL, NULL,
2150 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2151 		       CTLTYPE_INT, "keepidle",
2152 		       SYSCTL_DESCR("Allowed connection idle ticks before a "
2153 				    "keepalive probe is sent"),
2154 		       sysctl_tcp_keep, 0, &tcp_keepidle, 0,
2155 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL);
2156 	sysctl_createv(clog, 0, NULL, NULL,
2157 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2158 		       CTLTYPE_INT, "keepintvl",
2159 		       SYSCTL_DESCR("Ticks before next keepalive probe is sent"),
2160 		       sysctl_tcp_keep, 0, &tcp_keepintvl, 0,
2161 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL);
2162 	sysctl_createv(clog, 0, NULL, NULL,
2163 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2164 		       CTLTYPE_INT, "keepcnt",
2165 		       SYSCTL_DESCR("Number of keepalive probes to send"),
2166 		       sysctl_tcp_keep, 0, &tcp_keepcnt, 0,
2167 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL);
2168 	sysctl_createv(clog, 0, NULL, NULL,
2169 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
2170 		       CTLTYPE_INT, "slowhz",
2171 		       SYSCTL_DESCR("Keepalive ticks per second"),
2172 		       NULL, PR_SLOWHZ, NULL, 0,
2173 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL);
2174 	sysctl_createv(clog, 0, NULL, NULL,
2175 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2176 		       CTLTYPE_INT, "log_refused",
2177 		       SYSCTL_DESCR("Log refused TCP connections"),
2178 		       NULL, 0, &tcp_log_refused, 0,
2179 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL);
2180 #if 0 /* obsoleted */
2181 	sysctl_createv(clog, 0, NULL, NULL,
2182 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2183 		       CTLTYPE_INT, "rstratelimit", NULL,
2184 		       NULL, 0, &tcp_rst_ratelim, 0,
2185 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL);
2186 #endif
2187 	sysctl_createv(clog, 0, NULL, NULL,
2188 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2189 		       CTLTYPE_INT, "rstppslimit",
2190 		       SYSCTL_DESCR("Maximum number of RST packets to send "
2191 				    "per second"),
2192 		       NULL, 0, &tcp_rst_ppslim, 0,
2193 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL);
2194 	sysctl_createv(clog, 0, NULL, NULL,
2195 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2196 		       CTLTYPE_INT, "delack_ticks",
2197 		       SYSCTL_DESCR("Number of ticks to delay sending an ACK"),
2198 		       NULL, 0, &tcp_delack_ticks, 0,
2199 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL);
2200 	sysctl_createv(clog, 0, NULL, NULL,
2201 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2202 		       CTLTYPE_INT, "init_win_local",
2203 		       SYSCTL_DESCR("Initial TCP window size (in segments)"),
2204 		       sysctl_tcp_init_win, 0, &tcp_init_win_local, 0,
2205 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL,
2206 		       CTL_EOL);
2207 	sysctl_createv(clog, 0, NULL, NULL,
2208 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2209 		       CTLTYPE_STRUCT, "ident",
2210 		       SYSCTL_DESCR("RFC1413 Identification Protocol lookups"),
2211 		       sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t),
2212 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL);
2213 	sysctl_createv(clog, 0, NULL, NULL,
2214 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2215 		       CTLTYPE_INT, "do_loopback_cksum",
2216 		       SYSCTL_DESCR("Perform TCP checksum on loopback"),
2217 		       NULL, 0, &tcp_do_loopback_cksum, 0,
2218 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM,
2219 		       CTL_EOL);
2220 	sysctl_createv(clog, 0, NULL, NULL,
2221 		       CTLFLAG_PERMANENT,
2222 		       CTLTYPE_STRUCT, "pcblist",
2223 		       SYSCTL_DESCR("TCP protocol control block list"),
2224 		       sysctl_inpcblist, 0, &tcbtable, 0,
2225 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2226 		       CTL_EOL);
2227 	sysctl_createv(clog, 0, NULL, NULL,
2228 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2229 		       CTLTYPE_INT, "keepinit",
2230 		       SYSCTL_DESCR("Ticks before initial tcp connection times out"),
2231 		       sysctl_tcp_keep, 0, &tcp_keepinit, 0,
2232 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2233 
2234 	/* TCP socket buffers auto-sizing nodes */
2235 	sysctl_createv(clog, 0, NULL, NULL,
2236 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2237 		       CTLTYPE_INT, "recvbuf_auto",
2238 		       SYSCTL_DESCR("Enable automatic receive "
2239 		           "buffer sizing (experimental)"),
2240 		       NULL, 0, &tcp_do_autorcvbuf, 0,
2241 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2242 	sysctl_createv(clog, 0, NULL, NULL,
2243 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2244 		       CTLTYPE_INT, "recvbuf_inc",
2245 		       SYSCTL_DESCR("Incrementor step size of "
2246 		           "automatic receive buffer"),
2247 		       NULL, 0, &tcp_autorcvbuf_inc, 0,
2248 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2249 	sysctl_createv(clog, 0, NULL, NULL,
2250 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2251 		       CTLTYPE_INT, "recvbuf_max",
2252 		       SYSCTL_DESCR("Max size of automatic receive buffer"),
2253 		       NULL, 0, &tcp_autorcvbuf_max, 0,
2254 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2255 
2256 	sysctl_createv(clog, 0, NULL, NULL,
2257 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2258 		       CTLTYPE_INT, "sendbuf_auto",
2259 		       SYSCTL_DESCR("Enable automatic send "
2260 		           "buffer sizing (experimental)"),
2261 		       NULL, 0, &tcp_do_autosndbuf, 0,
2262 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2263 	sysctl_createv(clog, 0, NULL, NULL,
2264 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2265 		       CTLTYPE_INT, "sendbuf_inc",
2266 		       SYSCTL_DESCR("Incrementor step size of "
2267 		           "automatic send buffer"),
2268 		       NULL, 0, &tcp_autosndbuf_inc, 0,
2269 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2270 	sysctl_createv(clog, 0, NULL, NULL,
2271 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2272 		       CTLTYPE_INT, "sendbuf_max",
2273 		       SYSCTL_DESCR("Max size of automatic send buffer"),
2274 		       NULL, 0, &tcp_autosndbuf_max, 0,
2275 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2276 
2277 	/* ECN subtree */
2278 	sysctl_createv(clog, 0, NULL, &ecn_node,
2279 	    	       CTLFLAG_PERMANENT,
2280 		       CTLTYPE_NODE, "ecn",
2281 	    	       SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
2282 	    	       NULL, 0, NULL, 0,
2283 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2284 	sysctl_createv(clog, 0, &ecn_node, NULL,
2285 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2286 		       CTLTYPE_INT, "enable",
2287 		       SYSCTL_DESCR("Enable TCP Explicit Congestion "
2288 			   "Notification"),
2289 	    	       NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL);
2290 	sysctl_createv(clog, 0, &ecn_node, NULL,
2291 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2292 		       CTLTYPE_INT, "maxretries",
2293 		       SYSCTL_DESCR("Number of times to retry ECN setup "
2294 			       "before disabling ECN on the connection"),
2295 	    	       NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL);
2296 
2297 	/* SACK gets its own little subtree. */
2298 	sysctl_createv(clog, 0, NULL, &sack_node,
2299 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2300 		       CTLTYPE_INT, "enable",
2301 		       SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"),
2302 		       NULL, 0, &tcp_do_sack, 0,
2303 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2304 	sysctl_createv(clog, 0, NULL, &sack_node,
2305 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2306 		       CTLTYPE_INT, "maxholes",
2307 		       SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"),
2308 		       NULL, 0, &tcp_sack_tp_maxholes, 0,
2309 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2310 	sysctl_createv(clog, 0, NULL, &sack_node,
2311 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2312 		       CTLTYPE_INT, "globalmaxholes",
2313 		       SYSCTL_DESCR("Global maximum number of TCP SACK holes"),
2314 		       NULL, 0, &tcp_sack_globalmaxholes, 0,
2315 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2316 	sysctl_createv(clog, 0, NULL, &sack_node,
2317 		       CTLFLAG_PERMANENT,
2318 		       CTLTYPE_INT, "globalholes",
2319 		       SYSCTL_DESCR("Global number of TCP SACK holes"),
2320 		       NULL, 0, &tcp_sack_globalholes, 0,
2321 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
2322 
2323 	sysctl_createv(clog, 0, NULL, NULL,
2324 		       CTLFLAG_PERMANENT,
2325 		       CTLTYPE_STRUCT, "stats",
2326 		       SYSCTL_DESCR("TCP statistics"),
2327 		       sysctl_net_inet_tcp_stats, 0, NULL, 0,
2328 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS,
2329 		       CTL_EOL);
2330         sysctl_createv(clog, 0, NULL, NULL,
2331                        CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2332                        CTLTYPE_INT, "local_by_rtt",
2333                        SYSCTL_DESCR("Use RTT estimator to decide which hosts "
2334 				    "are local"),
2335 		       NULL, 0, &tcp_rttlocal, 0,
2336 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2337 #ifdef TCP_DEBUG
2338 	sysctl_createv(clog, 0, NULL, NULL,
2339 		       CTLFLAG_PERMANENT,
2340 		       CTLTYPE_STRUCT, "debug",
2341 		       SYSCTL_DESCR("TCP sockets debug information"),
2342 		       NULL, 0, &tcp_debug, sizeof(tcp_debug),
2343 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG,
2344 		       CTL_EOL);
2345 	sysctl_createv(clog, 0, NULL, NULL,
2346 		       CTLFLAG_PERMANENT,
2347 		       CTLTYPE_INT, "debx",
2348 		       SYSCTL_DESCR("Number of TCP debug sockets messages"),
2349 		       NULL, 0, &tcp_debx, sizeof(tcp_debx),
2350 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX,
2351 		       CTL_EOL);
2352 #endif
2353 	sysctl_createv(clog, 0, NULL, NULL,
2354 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2355 		       CTLTYPE_STRUCT, "drop",
2356 		       SYSCTL_DESCR("TCP drop connection"),
2357 		       sysctl_net_inet_tcp_drop, 0, NULL, 0,
2358 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL);
2359 	sysctl_createv(clog, 0, NULL, NULL,
2360 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2361 		       CTLTYPE_INT, "iss_hash",
2362 		       SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic "
2363 				    "hash computation"),
2364 		       NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948),
2365 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2366 		       CTL_EOL);
2367 
2368 	/* ABC subtree */
2369 
2370 	sysctl_createv(clog, 0, NULL, &abc_node,
2371 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc",
2372 		       SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"),
2373 		       NULL, 0, NULL, 0,
2374 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2375 	sysctl_createv(clog, 0, &abc_node, NULL,
2376 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2377 		       CTLTYPE_INT, "enable",
2378 		       SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"),
2379 		       NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL);
2380 	sysctl_createv(clog, 0, &abc_node, NULL,
2381 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2382 		       CTLTYPE_INT, "aggressive",
2383 		       SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"),
2384 		       NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL);
2385 
2386 	/* MSL tuning subtree */
2387 
2388 	sysctl_createv(clog, 0, NULL, &mslt_node,
2389 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt",
2390 		       SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"),
2391 		       NULL, 0, NULL, 0,
2392 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2393 	sysctl_createv(clog, 0, &mslt_node, NULL,
2394 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2395 		       CTLTYPE_INT, "enable",
2396 		       SYSCTL_DESCR("Enable TIME_WAIT truncation"),
2397 		       NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL);
2398 	sysctl_createv(clog, 0, &mslt_node, NULL,
2399 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2400 		       CTLTYPE_INT, "loopback",
2401 		       SYSCTL_DESCR("MSL value to use for loopback connections"),
2402 		       NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL);
2403 	sysctl_createv(clog, 0, &mslt_node, NULL,
2404 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2405 		       CTLTYPE_INT, "local",
2406 		       SYSCTL_DESCR("MSL value to use for local connections"),
2407 		       NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL);
2408 	sysctl_createv(clog, 0, &mslt_node, NULL,
2409 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2410 		       CTLTYPE_INT, "remote",
2411 		       SYSCTL_DESCR("MSL value to use for remote connections"),
2412 		       NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL);
2413 	sysctl_createv(clog, 0, &mslt_node, NULL,
2414 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2415 		       CTLTYPE_INT, "remote_threshold",
2416 		       SYSCTL_DESCR("RTT estimate value to promote local to remote"),
2417 		       NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL);
2418 
2419 	/* vestigial TIME_WAIT tuning subtree */
2420 
2421 	sysctl_createv(clog, 0, NULL, &vtw_node,
2422 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw",
2423 		       SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"),
2424 		       NULL, 0, NULL, 0,
2425 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2426 	sysctl_createv(clog, 0, &vtw_node, NULL,
2427 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2428 		       CTLTYPE_INT, "enable",
2429 		       SYSCTL_DESCR("Enable Vestigial TIME_WAIT"),
2430 		       sysctl_tcp_vtw_enable, 0,
2431 	               (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable,
2432 		       0, CTL_CREATE, CTL_EOL);
2433 	sysctl_createv(clog, 0, &vtw_node, NULL,
2434 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY,
2435 		       CTLTYPE_INT, "entries",
2436 		       SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"),
2437 		       NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL);
2438 }
2439 
2440 void
2441 tcp_usrreq_init(void)
2442 {
2443 
2444 	sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp");
2445 #ifdef INET6
2446 	sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6");
2447 #endif
2448 }
2449 
2450 PR_WRAP_USRREQS(tcp)
2451 #define	tcp_attach	tcp_attach_wrapper
2452 #define	tcp_detach	tcp_detach_wrapper
2453 #define	tcp_accept	tcp_accept_wrapper
2454 #define	tcp_bind	tcp_bind_wrapper
2455 #define	tcp_listen	tcp_listen_wrapper
2456 #define	tcp_connect	tcp_connect_wrapper
2457 #define	tcp_connect2	tcp_connect2_wrapper
2458 #define	tcp_disconnect	tcp_disconnect_wrapper
2459 #define	tcp_shutdown	tcp_shutdown_wrapper
2460 #define	tcp_abort	tcp_abort_wrapper
2461 #define	tcp_ioctl	tcp_ioctl_wrapper
2462 #define	tcp_stat	tcp_stat_wrapper
2463 #define	tcp_peeraddr	tcp_peeraddr_wrapper
2464 #define	tcp_sockaddr	tcp_sockaddr_wrapper
2465 #define	tcp_rcvd	tcp_rcvd_wrapper
2466 #define	tcp_recvoob	tcp_recvoob_wrapper
2467 #define	tcp_send	tcp_send_wrapper
2468 #define	tcp_sendoob	tcp_sendoob_wrapper
2469 #define	tcp_purgeif	tcp_purgeif_wrapper
2470 
2471 const struct pr_usrreqs tcp_usrreqs = {
2472 	.pr_attach	= tcp_attach,
2473 	.pr_detach	= tcp_detach,
2474 	.pr_accept	= tcp_accept,
2475 	.pr_bind	= tcp_bind,
2476 	.pr_listen	= tcp_listen,
2477 	.pr_connect	= tcp_connect,
2478 	.pr_connect2	= tcp_connect2,
2479 	.pr_disconnect	= tcp_disconnect,
2480 	.pr_shutdown	= tcp_shutdown,
2481 	.pr_abort	= tcp_abort,
2482 	.pr_ioctl	= tcp_ioctl,
2483 	.pr_stat	= tcp_stat,
2484 	.pr_peeraddr	= tcp_peeraddr,
2485 	.pr_sockaddr	= tcp_sockaddr,
2486 	.pr_rcvd	= tcp_rcvd,
2487 	.pr_recvoob	= tcp_recvoob,
2488 	.pr_send	= tcp_send,
2489 	.pr_sendoob	= tcp_sendoob,
2490 	.pr_purgeif	= tcp_purgeif,
2491 };
2492