xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 96230fab84e26a6435963032070e916a951a8b2e)
1 /*	$NetBSD: tcp_usrreq.c,v 1.149 2008/10/11 13:40:57 pooka 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 #include <sys/cdefs.h>
98 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.149 2008/10/11 13:40:57 pooka Exp $");
99 
100 #include "opt_inet.h"
101 #include "opt_ipsec.h"
102 #include "opt_tcp_debug.h"
103 #include "opt_mbuftrace.h"
104 #include "rnd.h"
105 
106 #include <sys/param.h>
107 #include <sys/systm.h>
108 #include <sys/kernel.h>
109 #include <sys/malloc.h>
110 #include <sys/mbuf.h>
111 #include <sys/socket.h>
112 #include <sys/socketvar.h>
113 #include <sys/protosw.h>
114 #include <sys/errno.h>
115 #include <sys/stat.h>
116 #include <sys/proc.h>
117 #include <sys/domain.h>
118 #include <sys/sysctl.h>
119 #include <sys/kauth.h>
120 #include <sys/uidinfo.h>
121 
122 #include <net/if.h>
123 #include <net/route.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/in_var.h>
128 #include <netinet/ip.h>
129 #include <netinet/in_pcb.h>
130 #include <netinet/ip_var.h>
131 #include <netinet/in_offload.h>
132 
133 #ifdef INET6
134 #ifndef INET
135 #include <netinet/in.h>
136 #endif
137 #include <netinet/ip6.h>
138 #include <netinet6/in6_pcb.h>
139 #include <netinet6/ip6_var.h>
140 #include <netinet6/scope6_var.h>
141 #endif
142 
143 #include <netinet/tcp.h>
144 #include <netinet/tcp_fsm.h>
145 #include <netinet/tcp_seq.h>
146 #include <netinet/tcp_timer.h>
147 #include <netinet/tcp_var.h>
148 #include <netinet/tcp_private.h>
149 #include <netinet/tcp_congctl.h>
150 #include <netinet/tcpip.h>
151 #include <netinet/tcp_debug.h>
152 
153 #include "opt_tcp_space.h"
154 
155 #ifdef IPSEC
156 #include <netinet6/ipsec.h>
157 #endif /*IPSEC*/
158 
159 /*
160  * TCP protocol interface to socket abstraction.
161  */
162 
163 /*
164  * Process a TCP user request for TCP tb.  If this is a send request
165  * then m is the mbuf chain of send data.  If this is a timer expiration
166  * (called from the software clock routine), then timertype tells which timer.
167  */
168 /*ARGSUSED*/
169 int
170 tcp_usrreq(struct socket *so, int req,
171     struct mbuf *m, struct mbuf *nam, struct mbuf *control, struct lwp *l)
172 {
173 	struct inpcb *inp;
174 #ifdef INET6
175 	struct in6pcb *in6p;
176 #endif
177 	struct tcpcb *tp = NULL;
178 	int s;
179 	int error = 0;
180 #ifdef TCP_DEBUG
181 	int ostate = 0;
182 #endif
183 	int family;	/* family of the socket */
184 
185 	family = so->so_proto->pr_domain->dom_family;
186 
187 	if (req == PRU_CONTROL) {
188 		switch (family) {
189 #ifdef INET
190 		case PF_INET:
191 			return (in_control(so, (long)m, (void *)nam,
192 			    (struct ifnet *)control, l));
193 #endif
194 #ifdef INET6
195 		case PF_INET6:
196 			return (in6_control(so, (long)m, (void *)nam,
197 			    (struct ifnet *)control, l));
198 #endif
199 		default:
200 			return EAFNOSUPPORT;
201 		}
202 	}
203 
204 	s = splsoftnet();
205 
206 	if (req == PRU_PURGEIF) {
207 		mutex_enter(softnet_lock);
208 		switch (family) {
209 #ifdef INET
210 		case PF_INET:
211 			in_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
212 			in_purgeif((struct ifnet *)control);
213 			in_pcbpurgeif(&tcbtable, (struct ifnet *)control);
214 			break;
215 #endif
216 #ifdef INET6
217 		case PF_INET6:
218 			in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
219 			in6_purgeif((struct ifnet *)control);
220 			in6_pcbpurgeif(&tcbtable, (struct ifnet *)control);
221 			break;
222 #endif
223 		default:
224 			mutex_exit(softnet_lock);
225 			splx(s);
226 			return (EAFNOSUPPORT);
227 		}
228 		mutex_exit(softnet_lock);
229 		splx(s);
230 		return (0);
231 	}
232 
233 	if (req == PRU_ATTACH)
234 		sosetlock(so);
235 
236 	switch (family) {
237 #ifdef INET
238 	case PF_INET:
239 		inp = sotoinpcb(so);
240 #ifdef INET6
241 		in6p = NULL;
242 #endif
243 		break;
244 #endif
245 #ifdef INET6
246 	case PF_INET6:
247 		inp = NULL;
248 		in6p = sotoin6pcb(so);
249 		break;
250 #endif
251 	default:
252 		splx(s);
253 		return EAFNOSUPPORT;
254 	}
255 
256 #ifdef DIAGNOSTIC
257 #ifdef INET6
258 	if (inp && in6p)
259 		panic("tcp_usrreq: both inp and in6p set to non-NULL");
260 #endif
261 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
262 		panic("tcp_usrreq: unexpected control mbuf");
263 #endif
264 	/*
265 	 * When a TCP is attached to a socket, then there will be
266 	 * a (struct inpcb) pointed at by the socket, and this
267 	 * structure will point at a subsidary (struct tcpcb).
268 	 */
269 #ifndef INET6
270 	if (inp == 0 && req != PRU_ATTACH)
271 #else
272 	if ((inp == 0 && in6p == 0) && req != PRU_ATTACH)
273 #endif
274 	{
275 		error = EINVAL;
276 		goto release;
277 	}
278 #ifdef INET
279 	if (inp) {
280 		tp = intotcpcb(inp);
281 		/* WHAT IF TP IS 0? */
282 #ifdef KPROF
283 		tcp_acounts[tp->t_state][req]++;
284 #endif
285 #ifdef TCP_DEBUG
286 		ostate = tp->t_state;
287 #endif
288 	}
289 #endif
290 #ifdef INET6
291 	if (in6p) {
292 		tp = in6totcpcb(in6p);
293 		/* WHAT IF TP IS 0? */
294 #ifdef KPROF
295 		tcp_acounts[tp->t_state][req]++;
296 #endif
297 #ifdef TCP_DEBUG
298 		ostate = tp->t_state;
299 #endif
300 	}
301 #endif
302 
303 	switch (req) {
304 
305 	/*
306 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
307 	 * and an internet control block.
308 	 */
309 	case PRU_ATTACH:
310 #ifndef INET6
311 		if (inp != 0)
312 #else
313 		if (inp != 0 || in6p != 0)
314 #endif
315 		{
316 			error = EISCONN;
317 			break;
318 		}
319 		error = tcp_attach(so);
320 		if (error)
321 			break;
322 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
323 			so->so_linger = TCP_LINGERTIME;
324 		tp = sototcpcb(so);
325 		break;
326 
327 	/*
328 	 * PRU_DETACH detaches the TCP protocol from the socket.
329 	 */
330 	case PRU_DETACH:
331 		tp = tcp_disconnect(tp);
332 		break;
333 
334 	/*
335 	 * Give the socket an address.
336 	 */
337 	case PRU_BIND:
338 		switch (family) {
339 #ifdef INET
340 		case PF_INET:
341 			error = in_pcbbind(inp, nam, l);
342 			break;
343 #endif
344 #ifdef INET6
345 		case PF_INET6:
346 			error = in6_pcbbind(in6p, nam, l);
347 			if (!error) {
348 				/* mapped addr case */
349 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
350 					tp->t_family = AF_INET;
351 				else
352 					tp->t_family = AF_INET6;
353 			}
354 			break;
355 #endif
356 		}
357 		break;
358 
359 	/*
360 	 * Prepare to accept connections.
361 	 */
362 	case PRU_LISTEN:
363 #ifdef INET
364 		if (inp && inp->inp_lport == 0) {
365 			error = in_pcbbind(inp, (struct mbuf *)0, l);
366 			if (error)
367 				break;
368 		}
369 #endif
370 #ifdef INET6
371 		if (in6p && in6p->in6p_lport == 0) {
372 			error = in6_pcbbind(in6p, (struct mbuf *)0,
373 			    (struct lwp *)0);
374 			if (error)
375 				break;
376 		}
377 #endif
378 		tp->t_state = TCPS_LISTEN;
379 		break;
380 
381 	/*
382 	 * Initiate connection to peer.
383 	 * Create a template for use in transmissions on this connection.
384 	 * Enter SYN_SENT state, and mark socket as connecting.
385 	 * Start keep-alive timer, and seed output sequence space.
386 	 * Send initial segment on connection.
387 	 */
388 	case PRU_CONNECT:
389 #ifdef INET
390 		if (inp) {
391 			if (inp->inp_lport == 0) {
392 				error = in_pcbbind(inp, (struct mbuf *)0, l);
393 				if (error)
394 					break;
395 			}
396 			error = in_pcbconnect(inp, nam, l);
397 		}
398 #endif
399 #ifdef INET6
400 		if (in6p) {
401 			if (in6p->in6p_lport == 0) {
402 				error = in6_pcbbind(in6p, (struct mbuf *)0,
403 				    (struct lwp *)0);
404 				if (error)
405 					break;
406 			}
407 			error = in6_pcbconnect(in6p, nam, l);
408 			if (!error) {
409 				/* mapped addr case */
410 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
411 					tp->t_family = AF_INET;
412 				else
413 					tp->t_family = AF_INET6;
414 			}
415 		}
416 #endif
417 		if (error)
418 			break;
419 		tp->t_template = tcp_template(tp);
420 		if (tp->t_template == 0) {
421 #ifdef INET
422 			if (inp)
423 				in_pcbdisconnect(inp);
424 #endif
425 #ifdef INET6
426 			if (in6p)
427 				in6_pcbdisconnect(in6p);
428 #endif
429 			error = ENOBUFS;
430 			break;
431 		}
432 		/*
433 		 * Compute window scaling to request.
434 		 * XXX: This should be moved to tcp_output().
435 		 */
436 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
437 		    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
438 			tp->request_r_scale++;
439 		soisconnecting(so);
440 		TCP_STATINC(TCP_STAT_CONNATTEMPT);
441 		tp->t_state = TCPS_SYN_SENT;
442 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
443 		tp->iss = tcp_new_iss(tp, 0);
444 		tcp_sendseqinit(tp);
445 		error = tcp_output(tp);
446 		break;
447 
448 	/*
449 	 * Create a TCP connection between two sockets.
450 	 */
451 	case PRU_CONNECT2:
452 		error = EOPNOTSUPP;
453 		break;
454 
455 	/*
456 	 * Initiate disconnect from peer.
457 	 * If connection never passed embryonic stage, just drop;
458 	 * else if don't need to let data drain, then can just drop anyways,
459 	 * else have to begin TCP shutdown process: mark socket disconnecting,
460 	 * drain unread data, state switch to reflect user close, and
461 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
462 	 * when peer sends FIN and acks ours.
463 	 *
464 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
465 	 */
466 	case PRU_DISCONNECT:
467 		tp = tcp_disconnect(tp);
468 		break;
469 
470 	/*
471 	 * Accept a connection.  Essentially all the work is
472 	 * done at higher levels; just return the address
473 	 * of the peer, storing through addr.
474 	 */
475 	case PRU_ACCEPT:
476 #ifdef INET
477 		if (inp)
478 			in_setpeeraddr(inp, nam);
479 #endif
480 #ifdef INET6
481 		if (in6p)
482 			in6_setpeeraddr(in6p, nam);
483 #endif
484 		break;
485 
486 	/*
487 	 * Mark the connection as being incapable of further output.
488 	 */
489 	case PRU_SHUTDOWN:
490 		socantsendmore(so);
491 		tp = tcp_usrclosed(tp);
492 		if (tp)
493 			error = tcp_output(tp);
494 		break;
495 
496 	/*
497 	 * After a receive, possibly send window update to peer.
498 	 */
499 	case PRU_RCVD:
500 		/*
501 		 * soreceive() calls this function when a user receives
502 		 * ancillary data on a listening socket. We don't call
503 		 * tcp_output in such a case, since there is no header
504 		 * template for a listening socket and hence the kernel
505 		 * will panic.
506 		 */
507 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
508 			(void) tcp_output(tp);
509 		break;
510 
511 	/*
512 	 * Do a send by putting data in output queue and updating urgent
513 	 * marker if URG set.  Possibly send more data.
514 	 */
515 	case PRU_SEND:
516 		if (control && control->m_len) {
517 			m_freem(control);
518 			m_freem(m);
519 			error = EINVAL;
520 			break;
521 		}
522 		sbappendstream(&so->so_snd, m);
523 		error = tcp_output(tp);
524 		break;
525 
526 	/*
527 	 * Abort the TCP.
528 	 */
529 	case PRU_ABORT:
530 		tp = tcp_drop(tp, ECONNABORTED);
531 		break;
532 
533 	case PRU_SENSE:
534 		/*
535 		 * stat: don't bother with a blocksize.
536 		 */
537 		splx(s);
538 		return (0);
539 
540 	case PRU_RCVOOB:
541 		if (control && control->m_len) {
542 			m_freem(control);
543 			m_freem(m);
544 			error = EINVAL;
545 			break;
546 		}
547 		if ((so->so_oobmark == 0 &&
548 		    (so->so_state & SS_RCVATMARK) == 0) ||
549 		    so->so_options & SO_OOBINLINE ||
550 		    tp->t_oobflags & TCPOOB_HADDATA) {
551 			error = EINVAL;
552 			break;
553 		}
554 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
555 			error = EWOULDBLOCK;
556 			break;
557 		}
558 		m->m_len = 1;
559 		*mtod(m, char *) = tp->t_iobc;
560 		if (((long)nam & MSG_PEEK) == 0)
561 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
562 		break;
563 
564 	case PRU_SENDOOB:
565 		if (sbspace(&so->so_snd) < -512) {
566 			m_freem(m);
567 			error = ENOBUFS;
568 			break;
569 		}
570 		/*
571 		 * According to RFC961 (Assigned Protocols),
572 		 * the urgent pointer points to the last octet
573 		 * of urgent data.  We continue, however,
574 		 * to consider it to indicate the first octet
575 		 * of data past the urgent section.
576 		 * Otherwise, snd_up should be one lower.
577 		 */
578 		sbappendstream(&so->so_snd, m);
579 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
580 		tp->t_force = 1;
581 		error = tcp_output(tp);
582 		tp->t_force = 0;
583 		break;
584 
585 	case PRU_SOCKADDR:
586 #ifdef INET
587 		if (inp)
588 			in_setsockaddr(inp, nam);
589 #endif
590 #ifdef INET6
591 		if (in6p)
592 			in6_setsockaddr(in6p, nam);
593 #endif
594 		break;
595 
596 	case PRU_PEERADDR:
597 #ifdef INET
598 		if (inp)
599 			in_setpeeraddr(inp, nam);
600 #endif
601 #ifdef INET6
602 		if (in6p)
603 			in6_setpeeraddr(in6p, nam);
604 #endif
605 		break;
606 
607 	default:
608 		panic("tcp_usrreq");
609 	}
610 #ifdef TCP_DEBUG
611 	if (tp && (so->so_options & SO_DEBUG))
612 		tcp_trace(TA_USER, ostate, tp, NULL, req);
613 #endif
614 
615 release:
616 	splx(s);
617 	return (error);
618 }
619 
620 static void
621 change_keepalive(struct socket *so, struct tcpcb *tp)
622 {
623 	tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl;
624 	TCP_TIMER_DISARM(tp, TCPT_KEEP);
625 	TCP_TIMER_DISARM(tp, TCPT_2MSL);
626 
627 	if (tp->t_state == TCPS_SYN_RECEIVED ||
628 	    tp->t_state == TCPS_SYN_SENT) {
629 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
630 	} else if (so->so_options & SO_KEEPALIVE &&
631 	    tp->t_state <= TCPS_CLOSE_WAIT) {
632 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl);
633 	} else {
634 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
635 	}
636 
637 	if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
638 		TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
639 }
640 
641 
642 int
643 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
644 {
645 	int error = 0, s;
646 	struct inpcb *inp;
647 #ifdef INET6
648 	struct in6pcb *in6p;
649 #endif
650 	struct tcpcb *tp;
651 	u_int ui;
652 	int family;	/* family of the socket */
653 	int level, optname, optval;
654 
655 	level = sopt->sopt_level;
656 	optname = sopt->sopt_name;
657 
658 	family = so->so_proto->pr_domain->dom_family;
659 
660 	s = splsoftnet();
661 	switch (family) {
662 #ifdef INET
663 	case PF_INET:
664 		inp = sotoinpcb(so);
665 #ifdef INET6
666 		in6p = NULL;
667 #endif
668 		break;
669 #endif
670 #ifdef INET6
671 	case PF_INET6:
672 		inp = NULL;
673 		in6p = sotoin6pcb(so);
674 		break;
675 #endif
676 	default:
677 		splx(s);
678 		panic("%s: af %d", __func__, family);
679 	}
680 #ifndef INET6
681 	if (inp == NULL)
682 #else
683 	if (inp == NULL && in6p == NULL)
684 #endif
685 	{
686 		splx(s);
687 		return (ECONNRESET);
688 	}
689 	if (level != IPPROTO_TCP) {
690 		switch (family) {
691 #ifdef INET
692 		case PF_INET:
693 			error = ip_ctloutput(op, so, sopt);
694 			break;
695 #endif
696 #ifdef INET6
697 		case PF_INET6:
698 			error = ip6_ctloutput(op, so, sopt);
699 			break;
700 #endif
701 		}
702 		splx(s);
703 		return (error);
704 	}
705 	if (inp)
706 		tp = intotcpcb(inp);
707 #ifdef INET6
708 	else if (in6p)
709 		tp = in6totcpcb(in6p);
710 #endif
711 	else
712 		tp = NULL;
713 
714 	switch (op) {
715 	case PRCO_SETOPT:
716 		switch (optname) {
717 #ifdef TCP_SIGNATURE
718 		case TCP_MD5SIG:
719 			error = sockopt_getint(sopt, &optval);
720 			if (error)
721 				break;
722 			if (optval > 0)
723 				tp->t_flags |= TF_SIGNATURE;
724 			else
725 				tp->t_flags &= ~TF_SIGNATURE;
726 			break;
727 #endif /* TCP_SIGNATURE */
728 
729 		case TCP_NODELAY:
730 			error = sockopt_getint(sopt, &optval);
731 			if (error)
732 				break;
733 			if (optval)
734 				tp->t_flags |= TF_NODELAY;
735 			else
736 				tp->t_flags &= ~TF_NODELAY;
737 			break;
738 
739 		case TCP_MAXSEG:
740 			error = sockopt_getint(sopt, &optval);
741 			if (error)
742 				break;
743 			if (optval > 0 && optval <= tp->t_peermss)
744 				tp->t_peermss = optval; /* limit on send size */
745 			else
746 				error = EINVAL;
747 			break;
748 #ifdef notyet
749 		case TCP_CONGCTL:
750 			/* XXX string overflow XXX */
751 			error = tcp_congctl_select(tp, sopt->sopt_data);
752 			break;
753 #endif
754 
755 		case TCP_KEEPIDLE:
756 			error = sockopt_get(sopt, &ui, sizeof(ui));
757 			if (error)
758 				break;
759 			if (ui > 0) {
760 				tp->t_keepidle = ui;
761 				change_keepalive(so, tp);
762 			} else
763 				error = EINVAL;
764 			break;
765 
766 		case TCP_KEEPINTVL:
767 			error = sockopt_get(sopt, &ui, sizeof(ui));
768 			if (error)
769 				break;
770 			if (ui > 0) {
771 				tp->t_keepintvl = ui;
772 				change_keepalive(so, tp);
773 			} else
774 				error = EINVAL;
775 			break;
776 
777 		case TCP_KEEPCNT:
778 			error = sockopt_get(sopt, &ui, sizeof(ui));
779 			if (error)
780 				break;
781 			if (ui > 0) {
782 				tp->t_keepcnt = ui;
783 				change_keepalive(so, tp);
784 			} else
785 				error = EINVAL;
786 			break;
787 
788 		case TCP_KEEPINIT:
789 			error = sockopt_get(sopt, &ui, sizeof(ui));
790 			if (error)
791 				break;
792 			if (ui > 0) {
793 				tp->t_keepinit = ui;
794 				change_keepalive(so, tp);
795 			} else
796 				error = EINVAL;
797 			break;
798 
799 		default:
800 			error = ENOPROTOOPT;
801 			break;
802 		}
803 		break;
804 
805 	case PRCO_GETOPT:
806 		switch (optname) {
807 #ifdef TCP_SIGNATURE
808 		case TCP_MD5SIG:
809 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
810 			error = sockopt_set(sopt, &optval, sizeof(optval));
811 			break;
812 #endif
813 		case TCP_NODELAY:
814 			optval = tp->t_flags & TF_NODELAY;
815 			error = sockopt_set(sopt, &optval, sizeof(optval));
816 			break;
817 		case TCP_MAXSEG:
818 			optval = tp->t_peermss;
819 			error = sockopt_set(sopt, &optval, sizeof(optval));
820 			break;
821 #ifdef notyet
822 		case TCP_CONGCTL:
823 			break;
824 #endif
825 		default:
826 			error = ENOPROTOOPT;
827 			break;
828 		}
829 		break;
830 	}
831 	splx(s);
832 	return (error);
833 }
834 
835 #ifndef TCP_SENDSPACE
836 #define	TCP_SENDSPACE	1024*32
837 #endif
838 int	tcp_sendspace = TCP_SENDSPACE;
839 #ifndef TCP_RECVSPACE
840 #define	TCP_RECVSPACE	1024*32
841 #endif
842 int	tcp_recvspace = TCP_RECVSPACE;
843 
844 /*
845  * Attach TCP protocol to socket, allocating
846  * internet protocol control block, tcp control block,
847  * bufer space, and entering LISTEN state if to accept connections.
848  */
849 int
850 tcp_attach(struct socket *so)
851 {
852 	struct tcpcb *tp;
853 	struct inpcb *inp;
854 #ifdef INET6
855 	struct in6pcb *in6p;
856 #endif
857 	int error;
858 	int family;	/* family of the socket */
859 
860 	family = so->so_proto->pr_domain->dom_family;
861 
862 #ifdef MBUFTRACE
863 	so->so_mowner = &tcp_sock_mowner;
864 	so->so_rcv.sb_mowner = &tcp_sock_rx_mowner;
865 	so->so_snd.sb_mowner = &tcp_sock_tx_mowner;
866 #endif
867 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
868 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
869 		if (error)
870 			return (error);
871 	}
872 
873 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
874 	so->so_snd.sb_flags |= SB_AUTOSIZE;
875 
876 	switch (family) {
877 #ifdef INET
878 	case PF_INET:
879 		error = in_pcballoc(so, &tcbtable);
880 		if (error)
881 			return (error);
882 		inp = sotoinpcb(so);
883 #ifdef INET6
884 		in6p = NULL;
885 #endif
886 		break;
887 #endif
888 #ifdef INET6
889 	case PF_INET6:
890 		error = in6_pcballoc(so, &tcbtable);
891 		if (error)
892 			return (error);
893 		inp = NULL;
894 		in6p = sotoin6pcb(so);
895 		break;
896 #endif
897 	default:
898 		return EAFNOSUPPORT;
899 	}
900 	if (inp)
901 		tp = tcp_newtcpcb(family, (void *)inp);
902 #ifdef INET6
903 	else if (in6p)
904 		tp = tcp_newtcpcb(family, (void *)in6p);
905 #endif
906 	else
907 		tp = NULL;
908 
909 	if (tp == 0) {
910 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
911 
912 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
913 #ifdef INET
914 		if (inp)
915 			in_pcbdetach(inp);
916 #endif
917 #ifdef INET6
918 		if (in6p)
919 			in6_pcbdetach(in6p);
920 #endif
921 		so->so_state |= nofd;
922 		return (ENOBUFS);
923 	}
924 	tp->t_state = TCPS_CLOSED;
925 	return (0);
926 }
927 
928 /*
929  * Initiate (or continue) disconnect.
930  * If embryonic state, just send reset (once).
931  * If in ``let data drain'' option and linger null, just drop.
932  * Otherwise (hard), mark socket disconnecting and drop
933  * current input data; switch states based on user close, and
934  * send segment to peer (with FIN).
935  */
936 struct tcpcb *
937 tcp_disconnect(struct tcpcb *tp)
938 {
939 	struct socket *so;
940 
941 	if (tp->t_inpcb)
942 		so = tp->t_inpcb->inp_socket;
943 #ifdef INET6
944 	else if (tp->t_in6pcb)
945 		so = tp->t_in6pcb->in6p_socket;
946 #endif
947 	else
948 		so = NULL;
949 
950 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
951 		tp = tcp_close(tp);
952 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
953 		tp = tcp_drop(tp, 0);
954 	else {
955 		soisdisconnecting(so);
956 		sbflush(&so->so_rcv);
957 		tp = tcp_usrclosed(tp);
958 		if (tp)
959 			(void) tcp_output(tp);
960 	}
961 	return (tp);
962 }
963 
964 /*
965  * User issued close, and wish to trail through shutdown states:
966  * if never received SYN, just forget it.  If got a SYN from peer,
967  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
968  * If already got a FIN from peer, then almost done; go to LAST_ACK
969  * state.  In all other cases, have already sent FIN to peer (e.g.
970  * after PRU_SHUTDOWN), and just have to play tedious game waiting
971  * for peer to send FIN or not respond to keep-alives, etc.
972  * We can let the user exit from the close as soon as the FIN is acked.
973  */
974 struct tcpcb *
975 tcp_usrclosed(struct tcpcb *tp)
976 {
977 
978 	switch (tp->t_state) {
979 
980 	case TCPS_CLOSED:
981 	case TCPS_LISTEN:
982 	case TCPS_SYN_SENT:
983 		tp->t_state = TCPS_CLOSED;
984 		tp = tcp_close(tp);
985 		break;
986 
987 	case TCPS_SYN_RECEIVED:
988 	case TCPS_ESTABLISHED:
989 		tp->t_state = TCPS_FIN_WAIT_1;
990 		break;
991 
992 	case TCPS_CLOSE_WAIT:
993 		tp->t_state = TCPS_LAST_ACK;
994 		break;
995 	}
996 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
997 		struct socket *so;
998 		if (tp->t_inpcb)
999 			so = tp->t_inpcb->inp_socket;
1000 #ifdef INET6
1001 		else if (tp->t_in6pcb)
1002 			so = tp->t_in6pcb->in6p_socket;
1003 #endif
1004 		else
1005 			so = NULL;
1006 		if (so)
1007 			soisdisconnected(so);
1008 		/*
1009 		 * If we are in FIN_WAIT_2, we arrived here because the
1010 		 * application did a shutdown of the send side.  Like the
1011 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
1012 		 * a full close, we start a timer to make sure sockets are
1013 		 * not left in FIN_WAIT_2 forever.
1014 		 */
1015 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
1016 			TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
1017 	}
1018 	return (tp);
1019 }
1020 
1021 /*
1022  * sysctl helper routine for net.inet.ip.mssdflt.  it can't be less
1023  * than 32.
1024  */
1025 static int
1026 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS)
1027 {
1028 	int error, mssdflt;
1029 	struct sysctlnode node;
1030 
1031 	mssdflt = tcp_mssdflt;
1032 	node = *rnode;
1033 	node.sysctl_data = &mssdflt;
1034 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1035 	if (error || newp == NULL)
1036 		return (error);
1037 
1038 	if (mssdflt < 32)
1039 		return (EINVAL);
1040 	tcp_mssdflt = mssdflt;
1041 
1042 	return (0);
1043 }
1044 
1045 /*
1046  * sysctl helper routine for setting port related values under
1047  * net.inet.ip and net.inet6.ip6.  does basic range checking and does
1048  * additional checks for each type.  this code has placed in
1049  * tcp_input.c since INET and INET6 both use the same tcp code.
1050  *
1051  * this helper is not static so that both inet and inet6 can use it.
1052  */
1053 int
1054 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS)
1055 {
1056 	int error, tmp;
1057 	int apmin, apmax;
1058 #ifndef IPNOPRIVPORTS
1059 	int lpmin, lpmax;
1060 #endif /* IPNOPRIVPORTS */
1061 	struct sysctlnode node;
1062 
1063 	if (namelen != 0)
1064 		return (EINVAL);
1065 
1066 	switch (name[-3]) {
1067 #ifdef INET
1068 	    case PF_INET:
1069 		apmin = anonportmin;
1070 		apmax = anonportmax;
1071 #ifndef IPNOPRIVPORTS
1072 		lpmin = lowportmin;
1073 		lpmax = lowportmax;
1074 #endif /* IPNOPRIVPORTS */
1075 		break;
1076 #endif /* INET */
1077 #ifdef INET6
1078 	    case PF_INET6:
1079 		apmin = ip6_anonportmin;
1080 		apmax = ip6_anonportmax;
1081 #ifndef IPNOPRIVPORTS
1082 		lpmin = ip6_lowportmin;
1083 		lpmax = ip6_lowportmax;
1084 #endif /* IPNOPRIVPORTS */
1085 		break;
1086 #endif /* INET6 */
1087 	    default:
1088 		return (EINVAL);
1089 	}
1090 
1091 	/*
1092 	 * insert temporary copy into node, perform lookup on
1093 	 * temporary, then restore pointer
1094 	 */
1095 	node = *rnode;
1096 	tmp = *(int*)rnode->sysctl_data;
1097 	node.sysctl_data = &tmp;
1098 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1099 	if (error || newp == NULL)
1100 		return (error);
1101 
1102 	/*
1103 	 * simple port range check
1104 	 */
1105 	if (tmp < 0 || tmp > 65535)
1106 		return (EINVAL);
1107 
1108 	/*
1109 	 * per-node range checks
1110 	 */
1111 	switch (rnode->sysctl_num) {
1112 	case IPCTL_ANONPORTMIN:
1113 		if (tmp >= apmax)
1114 			return (EINVAL);
1115 #ifndef IPNOPRIVPORTS
1116 		if (tmp < IPPORT_RESERVED)
1117                         return (EINVAL);
1118 #endif /* IPNOPRIVPORTS */
1119 		break;
1120 
1121 	case IPCTL_ANONPORTMAX:
1122                 if (apmin >= tmp)
1123 			return (EINVAL);
1124 #ifndef IPNOPRIVPORTS
1125 		if (tmp < IPPORT_RESERVED)
1126                         return (EINVAL);
1127 #endif /* IPNOPRIVPORTS */
1128 		break;
1129 
1130 #ifndef IPNOPRIVPORTS
1131 	case IPCTL_LOWPORTMIN:
1132 		if (tmp >= lpmax ||
1133 		    tmp > IPPORT_RESERVEDMAX ||
1134 		    tmp < IPPORT_RESERVEDMIN)
1135 			return (EINVAL);
1136 		break;
1137 
1138 	case IPCTL_LOWPORTMAX:
1139 		if (lpmin >= tmp ||
1140 		    tmp > IPPORT_RESERVEDMAX ||
1141 		    tmp < IPPORT_RESERVEDMIN)
1142 			return (EINVAL);
1143 		break;
1144 #endif /* IPNOPRIVPORTS */
1145 
1146 	default:
1147 		return (EINVAL);
1148 	}
1149 
1150 	*(int*)rnode->sysctl_data = tmp;
1151 
1152 	return (0);
1153 }
1154 
1155 /*
1156  * The superuser can drop any connection.  Normal users can only drop
1157  * their own connections.
1158  */
1159 static inline int
1160 check_sockuid(struct socket *sockp, kauth_cred_t cred)
1161 {
1162 	uid_t sockuid;
1163 
1164 	sockuid = sockp->so_uidinfo->ui_uid;
1165 	if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) == 0 ||
1166 	    sockuid == kauth_cred_getuid(cred) ||
1167 	    sockuid == kauth_cred_geteuid(cred))
1168 		return 0;
1169 	return EACCES;
1170 }
1171 
1172 static inline int
1173 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp)
1174 {
1175 	size_t sz;
1176 	int error;
1177 	uid_t uid;
1178 
1179 	uid = sockp->so_uidinfo->ui_uid;
1180 	if (oldp) {
1181 		sz = MIN(sizeof(uid), *oldlenp);
1182 		error = copyout(&uid, oldp, sz);
1183 		if (error)
1184 			return error;
1185 	}
1186 	*oldlenp = sizeof(uid);
1187 	return 0;
1188 }
1189 
1190 static inline int
1191 inet4_ident_core(struct in_addr raddr, u_int rport,
1192     struct in_addr laddr, u_int lport,
1193     void *oldp, size_t *oldlenp,
1194     struct lwp *l, int dodrop)
1195 {
1196 	struct inpcb *inp;
1197 	struct socket *sockp;
1198 
1199 	inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport);
1200 
1201 	if (inp == NULL || (sockp = inp->inp_socket) == NULL)
1202 		return ESRCH;
1203 
1204 	if (dodrop) {
1205 		struct tcpcb *tp;
1206 
1207 		if (inp == NULL || (tp = intotcpcb(inp)) == NULL ||
1208 		    (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0)
1209 			return ESRCH;
1210 
1211 		if (check_sockuid(inp->inp_socket, l->l_cred) != 0)
1212 			return EACCES;
1213 
1214 		(void)tcp_drop(tp, ECONNABORTED);
1215 		return 0;
1216 	}
1217 	else
1218 		return copyout_uid(sockp, oldp, oldlenp);
1219 }
1220 
1221 #ifdef INET6
1222 static inline int
1223 inet6_ident_core(struct in6_addr *raddr, u_int rport,
1224     struct in6_addr *laddr, u_int lport,
1225     void *oldp, size_t *oldlenp,
1226     struct lwp *l, int dodrop)
1227 {
1228 	struct in6pcb *in6p;
1229 	struct socket *sockp;
1230 
1231 	in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0);
1232 
1233 	if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL)
1234 		return ESRCH;
1235 
1236 	if (dodrop) {
1237 		struct tcpcb *tp;
1238 
1239 		if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL ||
1240 		    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0)
1241 			return ESRCH;
1242 
1243 		if (check_sockuid(in6p->in6p_socket, l->l_cred) != 0)
1244 			return EACCES;
1245 
1246 		(void)tcp_drop(tp, ECONNABORTED);
1247 		return 0;
1248 	}
1249 	else
1250 		return copyout_uid(sockp, oldp, oldlenp);
1251 }
1252 #endif
1253 
1254 /*
1255  * sysctl helper routine for the net.inet.tcp.drop and
1256  * net.inet6.tcp6.drop nodes.
1257  */
1258 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident
1259 
1260 /*
1261  * sysctl helper routine for the net.inet.tcp.ident and
1262  * net.inet6.tcp6.ident nodes.  contains backwards compat code for the
1263  * old way of looking up the ident information for ipv4 which involves
1264  * stuffing the port/addr pairs into the mib lookup.
1265  */
1266 static int
1267 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS)
1268 {
1269 #ifdef INET
1270 	struct sockaddr_in *si4[2];
1271 #endif /* INET */
1272 #ifdef INET6
1273 	struct sockaddr_in6 *si6[2];
1274 #endif /* INET6 */
1275 	struct sockaddr_storage sa[2];
1276 	int error, pf, dodrop;
1277 
1278 	dodrop = name[-1] == TCPCTL_DROP;
1279 	if (dodrop) {
1280 		if (oldp != NULL || *oldlenp != 0)
1281 			return EINVAL;
1282 		if (newp == NULL)
1283 			return EPERM;
1284 		if (newlen < sizeof(sa))
1285 			return ENOMEM;
1286 	}
1287 	if (namelen != 4 && namelen != 0)
1288 		return EINVAL;
1289 	if (name[-2] != IPPROTO_TCP)
1290 		return EINVAL;
1291 	pf = name[-3];
1292 
1293 	/* old style lookup, ipv4 only */
1294 	if (namelen == 4) {
1295 #ifdef INET
1296 		struct in_addr laddr, raddr;
1297 		u_int lport, rport;
1298 
1299 		if (pf != PF_INET)
1300 			return EPROTONOSUPPORT;
1301 		raddr.s_addr = (uint32_t)name[0];
1302 		rport = (u_int)name[1];
1303 		laddr.s_addr = (uint32_t)name[2];
1304 		lport = (u_int)name[3];
1305 
1306 		mutex_enter(softnet_lock);
1307 		error = inet4_ident_core(raddr, rport, laddr, lport,
1308 		    oldp, oldlenp, l, dodrop);
1309 		mutex_exit(softnet_lock);
1310 		return error;
1311 #else /* INET */
1312 		return EINVAL;
1313 #endif /* INET */
1314 	}
1315 
1316 	if (newp == NULL || newlen != sizeof(sa))
1317 		return EINVAL;
1318 	error = copyin(newp, &sa, newlen);
1319 	if (error)
1320 		return error;
1321 
1322 	/*
1323 	 * requested families must match
1324 	 */
1325 	if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family)
1326 		return EINVAL;
1327 
1328 	switch (pf) {
1329 #ifdef INET6
1330 	case PF_INET6:
1331 		si6[0] = (struct sockaddr_in6*)&sa[0];
1332 		si6[1] = (struct sockaddr_in6*)&sa[1];
1333 		if (si6[0]->sin6_len != sizeof(*si6[0]) ||
1334 		    si6[1]->sin6_len != sizeof(*si6[1]))
1335 			return EINVAL;
1336 
1337 		if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) &&
1338 		    !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) {
1339 			error = sa6_embedscope(si6[0], ip6_use_defzone);
1340 			if (error)
1341 				return error;
1342 			error = sa6_embedscope(si6[1], ip6_use_defzone);
1343 			if (error)
1344 				return error;
1345 
1346 			mutex_enter(softnet_lock);
1347 			error = inet6_ident_core(&si6[0]->sin6_addr,
1348 			    si6[0]->sin6_port, &si6[1]->sin6_addr,
1349 			    si6[1]->sin6_port, oldp, oldlenp, l, dodrop);
1350 			mutex_exit(softnet_lock);
1351 			return error;
1352 		}
1353 
1354 		if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) !=
1355 		    IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr))
1356 			return EINVAL;
1357 
1358 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]);
1359 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]);
1360 		/*FALLTHROUGH*/
1361 #endif /* INET6 */
1362 #ifdef INET
1363 	case PF_INET:
1364 		si4[0] = (struct sockaddr_in*)&sa[0];
1365 		si4[1] = (struct sockaddr_in*)&sa[1];
1366 		if (si4[0]->sin_len != sizeof(*si4[0]) ||
1367 		    si4[0]->sin_len != sizeof(*si4[1]))
1368 			return EINVAL;
1369 
1370 		mutex_enter(softnet_lock);
1371 		error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port,
1372 		    si4[1]->sin_addr, si4[1]->sin_port,
1373 		    oldp, oldlenp, l, dodrop);
1374 		mutex_exit(softnet_lock);
1375 		return error;
1376 #endif /* INET */
1377 	default:
1378 		return EPROTONOSUPPORT;
1379 	}
1380 }
1381 
1382 /*
1383  * sysctl helper for the inet and inet6 pcblists.  handles tcp/udp and
1384  * inet/inet6, as well as raw pcbs for each.  specifically not
1385  * declared static so that raw sockets and udp/udp6 can use it as
1386  * well.
1387  */
1388 int
1389 sysctl_inpcblist(SYSCTLFN_ARGS)
1390 {
1391 #ifdef INET
1392 	struct sockaddr_in *in;
1393 	const struct inpcb *inp;
1394 #endif
1395 #ifdef INET6
1396 	struct sockaddr_in6 *in6;
1397 	const struct in6pcb *in6p;
1398 #endif
1399 	/*
1400 	 * sysctl_data is const, but CIRCLEQ_FOREACH can't use a const
1401 	 * struct inpcbtable pointer, so we have to discard const.  :-/
1402 	 */
1403 	struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data);
1404 	const struct inpcb_hdr *inph;
1405 	struct tcpcb *tp;
1406 	struct kinfo_pcb pcb;
1407 	char *dp;
1408 	u_int op, arg;
1409 	size_t len, needed, elem_size, out_size;
1410 	int error, elem_count, pf, proto, pf2;
1411 
1412 	if (namelen != 4)
1413 		return (EINVAL);
1414 
1415 	if (oldp != NULL) {
1416 		    len = *oldlenp;
1417 		    elem_size = name[2];
1418 		    elem_count = name[3];
1419 		    if (elem_size != sizeof(pcb))
1420 			    return EINVAL;
1421 	} else {
1422 		    len = 0;
1423 		    elem_count = INT_MAX;
1424 		    elem_size = sizeof(pcb);
1425 	}
1426 	error = 0;
1427 	dp = oldp;
1428 	op = name[0];
1429 	arg = name[1];
1430 	out_size = elem_size;
1431 	needed = 0;
1432 
1433 	if (namelen == 1 && name[0] == CTL_QUERY)
1434 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1435 
1436 	if (name - oname != 4)
1437 		return (EINVAL);
1438 
1439 	pf = oname[1];
1440 	proto = oname[2];
1441 	pf2 = (oldp != NULL) ? pf : 0;
1442 
1443 	mutex_enter(softnet_lock);
1444 
1445 	CIRCLEQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) {
1446 #ifdef INET
1447 		inp = (const struct inpcb *)inph;
1448 #endif
1449 #ifdef INET6
1450 		in6p = (const struct in6pcb *)inph;
1451 #endif
1452 
1453 		if (inph->inph_af != pf)
1454 			continue;
1455 
1456 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1457 		    KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL,
1458 		    NULL) != 0)
1459 			continue;
1460 
1461 		memset(&pcb, 0, sizeof(pcb));
1462 
1463 		pcb.ki_family = pf;
1464 		pcb.ki_type = proto;
1465 
1466 		switch (pf2) {
1467 		case 0:
1468 			/* just probing for size */
1469 			break;
1470 #ifdef INET
1471 		case PF_INET:
1472 			pcb.ki_family = inp->inp_socket->so_proto->
1473 			    pr_domain->dom_family;
1474 			pcb.ki_type = inp->inp_socket->so_proto->
1475 			    pr_type;
1476 			pcb.ki_protocol = inp->inp_socket->so_proto->
1477 			    pr_protocol;
1478 			pcb.ki_pflags = inp->inp_flags;
1479 
1480 			pcb.ki_sostate = inp->inp_socket->so_state;
1481 			pcb.ki_prstate = inp->inp_state;
1482 			if (proto == IPPROTO_TCP) {
1483 				tp = intotcpcb(inp);
1484 				pcb.ki_tstate = tp->t_state;
1485 				pcb.ki_tflags = tp->t_flags;
1486 			}
1487 
1488 			pcb.ki_pcbaddr = PTRTOUINT64(inp);
1489 			pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb);
1490 			pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket);
1491 
1492 			pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc;
1493 			pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc;
1494 
1495 			in = satosin(&pcb.ki_src);
1496 			in->sin_len = sizeof(*in);
1497 			in->sin_family = pf;
1498 			in->sin_port = inp->inp_lport;
1499 			in->sin_addr = inp->inp_laddr;
1500 			if (pcb.ki_prstate >= INP_CONNECTED) {
1501 				in = satosin(&pcb.ki_dst);
1502 				in->sin_len = sizeof(*in);
1503 				in->sin_family = pf;
1504 				in->sin_port = inp->inp_fport;
1505 				in->sin_addr = inp->inp_faddr;
1506 			}
1507 			break;
1508 #endif
1509 #ifdef INET6
1510 		case PF_INET6:
1511 			pcb.ki_family = in6p->in6p_socket->so_proto->
1512 			    pr_domain->dom_family;
1513 			pcb.ki_type = in6p->in6p_socket->so_proto->pr_type;
1514 			pcb.ki_protocol = in6p->in6p_socket->so_proto->
1515 			    pr_protocol;
1516 			pcb.ki_pflags = in6p->in6p_flags;
1517 
1518 			pcb.ki_sostate = in6p->in6p_socket->so_state;
1519 			pcb.ki_prstate = in6p->in6p_state;
1520 			if (proto == IPPROTO_TCP) {
1521 				tp = in6totcpcb(in6p);
1522 				pcb.ki_tstate = tp->t_state;
1523 				pcb.ki_tflags = tp->t_flags;
1524 			}
1525 
1526 			pcb.ki_pcbaddr = PTRTOUINT64(in6p);
1527 			pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb);
1528 			pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket);
1529 
1530 			pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc;
1531 			pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc;
1532 
1533 			in6 = satosin6(&pcb.ki_src);
1534 			in6->sin6_len = sizeof(*in6);
1535 			in6->sin6_family = pf;
1536 			in6->sin6_port = in6p->in6p_lport;
1537 			in6->sin6_flowinfo = in6p->in6p_flowinfo;
1538 			in6->sin6_addr = in6p->in6p_laddr;
1539 			in6->sin6_scope_id = 0; /* XXX? */
1540 
1541 			if (pcb.ki_prstate >= IN6P_CONNECTED) {
1542 				in6 = satosin6(&pcb.ki_dst);
1543 				in6->sin6_len = sizeof(*in6);
1544 				in6->sin6_family = pf;
1545 				in6->sin6_port = in6p->in6p_fport;
1546 				in6->sin6_flowinfo = in6p->in6p_flowinfo;
1547 				in6->sin6_addr = in6p->in6p_faddr;
1548 				in6->sin6_scope_id = 0; /* XXX? */
1549 			}
1550 			break;
1551 #endif
1552 		}
1553 
1554 		if (len >= elem_size && elem_count > 0) {
1555 			error = copyout(&pcb, dp, out_size);
1556 			if (error)
1557 				return (error);
1558 			dp += elem_size;
1559 			len -= elem_size;
1560 		}
1561 		if (elem_count > 0) {
1562 			needed += elem_size;
1563 			if (elem_count != INT_MAX)
1564 				elem_count--;
1565 		}
1566 	}
1567 
1568 	*oldlenp = needed;
1569 	if (oldp == NULL)
1570 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
1571 
1572 	mutex_exit(softnet_lock);
1573 
1574 	return (error);
1575 }
1576 
1577 static int
1578 sysctl_tcp_congctl(SYSCTLFN_ARGS)
1579 {
1580 	struct sysctlnode node;
1581 	int error;
1582 	char newname[TCPCC_MAXLEN];
1583 
1584 	strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1);
1585 
1586 	node = *rnode;
1587 	node.sysctl_data = newname;
1588 	node.sysctl_size = sizeof(newname);
1589 
1590 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1591 
1592 	if (error ||
1593 	    newp == NULL ||
1594 	    strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0)
1595 		return error;
1596 
1597 	mutex_enter(softnet_lock);
1598 	error = tcp_congctl_select(NULL, newname);
1599 	mutex_exit(softnet_lock);
1600 
1601 	return error;
1602 }
1603 
1604 static int
1605 sysctl_tcp_keep(SYSCTLFN_ARGS)
1606 {
1607 	int error;
1608 	u_int tmp;
1609 	struct sysctlnode node;
1610 
1611 	node = *rnode;
1612 	tmp = *(u_int *)rnode->sysctl_data;
1613 	node.sysctl_data = &tmp;
1614 
1615 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1616 	if (error || newp == NULL)
1617 		return error;
1618 
1619 	mutex_enter(softnet_lock);
1620 
1621 	*(u_int *)rnode->sysctl_data = tmp;
1622 	tcp_tcpcb_template();	/* update the template */
1623 
1624 	mutex_exit(softnet_lock);
1625 	return 0;
1626 }
1627 
1628 static int
1629 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS)
1630 {
1631 
1632 	return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS));
1633 }
1634 
1635 /*
1636  * this (second stage) setup routine is a replacement for tcp_sysctl()
1637  * (which is currently used for ipv4 and ipv6)
1638  */
1639 static void
1640 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
1641 			   const char *tcpname)
1642 {
1643 	const struct sysctlnode *sack_node;
1644 	const struct sysctlnode *abc_node;
1645 	const struct sysctlnode *ecn_node;
1646 	const struct sysctlnode *congctl_node;
1647 #ifdef TCP_DEBUG
1648 	extern struct tcp_debug tcp_debug[TCP_NDEBUG];
1649 	extern int tcp_debx;
1650 #endif
1651 
1652 	sysctl_createv(clog, 0, NULL, NULL,
1653 		       CTLFLAG_PERMANENT,
1654 		       CTLTYPE_NODE, "net", NULL,
1655 		       NULL, 0, NULL, 0,
1656 		       CTL_NET, CTL_EOL);
1657 	sysctl_createv(clog, 0, NULL, NULL,
1658 		       CTLFLAG_PERMANENT,
1659 		       CTLTYPE_NODE, pfname, NULL,
1660 		       NULL, 0, NULL, 0,
1661 		       CTL_NET, pf, CTL_EOL);
1662 	sysctl_createv(clog, 0, NULL, NULL,
1663 		       CTLFLAG_PERMANENT,
1664 		       CTLTYPE_NODE, tcpname,
1665 		       SYSCTL_DESCR("TCP related settings"),
1666 		       NULL, 0, NULL, 0,
1667 		       CTL_NET, pf, IPPROTO_TCP, CTL_EOL);
1668 
1669 	sysctl_createv(clog, 0, NULL, NULL,
1670 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1671 		       CTLTYPE_INT, "rfc1323",
1672 		       SYSCTL_DESCR("Enable RFC1323 TCP extensions"),
1673 		       NULL, 0, &tcp_do_rfc1323, 0,
1674 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL);
1675 	sysctl_createv(clog, 0, NULL, NULL,
1676 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1677 		       CTLTYPE_INT, "sendspace",
1678 		       SYSCTL_DESCR("Default TCP send buffer size"),
1679 		       NULL, 0, &tcp_sendspace, 0,
1680 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL);
1681 	sysctl_createv(clog, 0, NULL, NULL,
1682 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1683 		       CTLTYPE_INT, "recvspace",
1684 		       SYSCTL_DESCR("Default TCP receive buffer size"),
1685 		       NULL, 0, &tcp_recvspace, 0,
1686 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL);
1687 	sysctl_createv(clog, 0, NULL, NULL,
1688 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1689 		       CTLTYPE_INT, "mssdflt",
1690 		       SYSCTL_DESCR("Default maximum segment size"),
1691 		       sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0,
1692 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL);
1693 	sysctl_createv(clog, 0, NULL, NULL,
1694 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1695 		       CTLTYPE_INT, "minmss",
1696 		       SYSCTL_DESCR("Lower limit for TCP maximum segment size"),
1697 		       NULL, 0, &tcp_minmss, 0,
1698 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1699 	sysctl_createv(clog, 0, NULL, NULL,
1700 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1701 		       CTLTYPE_INT, "syn_cache_limit",
1702 		       SYSCTL_DESCR("Maximum number of entries in the TCP "
1703 				    "compressed state engine"),
1704 		       NULL, 0, &tcp_syn_cache_limit, 0,
1705 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT,
1706 		       CTL_EOL);
1707 	sysctl_createv(clog, 0, NULL, NULL,
1708 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1709 		       CTLTYPE_INT, "syn_bucket_limit",
1710 		       SYSCTL_DESCR("Maximum number of entries per hash "
1711 				    "bucket in the TCP compressed state "
1712 				    "engine"),
1713 		       NULL, 0, &tcp_syn_bucket_limit, 0,
1714 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT,
1715 		       CTL_EOL);
1716 #if 0 /* obsoleted */
1717 	sysctl_createv(clog, 0, NULL, NULL,
1718 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1719 		       CTLTYPE_INT, "syn_cache_interval",
1720 		       SYSCTL_DESCR("TCP compressed state engine's timer interval"),
1721 		       NULL, 0, &tcp_syn_cache_interval, 0,
1722 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER,
1723 		       CTL_EOL);
1724 #endif
1725 	sysctl_createv(clog, 0, NULL, NULL,
1726 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1727 		       CTLTYPE_INT, "init_win",
1728 		       SYSCTL_DESCR("Initial TCP congestion window"),
1729 		       NULL, 0, &tcp_init_win, 0,
1730 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL);
1731 	sysctl_createv(clog, 0, NULL, NULL,
1732 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1733 		       CTLTYPE_INT, "mss_ifmtu",
1734 		       SYSCTL_DESCR("Use interface MTU for calculating MSS"),
1735 		       NULL, 0, &tcp_mss_ifmtu, 0,
1736 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL);
1737 	sysctl_createv(clog, 0, NULL, &sack_node,
1738 		       CTLFLAG_PERMANENT,
1739 		       CTLTYPE_NODE, "sack",
1740 		       SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
1741 		       NULL, 0, NULL, 0,
1742 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
1743 
1744 	/* Congctl subtree */
1745 	sysctl_createv(clog, 0, NULL, &congctl_node,
1746 		       CTLFLAG_PERMANENT,
1747 		       CTLTYPE_NODE, "congctl",
1748 		       SYSCTL_DESCR("TCP Congestion Control"),
1749 	    	       NULL, 0, NULL, 0,
1750 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1751 	sysctl_createv(clog, 0, &congctl_node, NULL,
1752 		       CTLFLAG_PERMANENT,
1753 		       CTLTYPE_STRING, "available",
1754 		       SYSCTL_DESCR("Available Congestion Control Mechanisms"),
1755 		       NULL, 0, &tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL);
1756 	sysctl_createv(clog, 0, &congctl_node, NULL,
1757 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1758 		       CTLTYPE_STRING, "selected",
1759 		       SYSCTL_DESCR("Selected Congestion Control Mechanism"),
1760 		       sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN,
1761 		       CTL_CREATE, CTL_EOL);
1762 
1763 	sysctl_createv(clog, 0, NULL, NULL,
1764 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1765 		       CTLTYPE_INT, "win_scale",
1766 		       SYSCTL_DESCR("Use RFC1323 window scale options"),
1767 		       NULL, 0, &tcp_do_win_scale, 0,
1768 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL);
1769 	sysctl_createv(clog, 0, NULL, NULL,
1770 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1771 		       CTLTYPE_INT, "timestamps",
1772 		       SYSCTL_DESCR("Use RFC1323 time stamp options"),
1773 		       NULL, 0, &tcp_do_timestamps, 0,
1774 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL);
1775 	sysctl_createv(clog, 0, NULL, NULL,
1776 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1777 		       CTLTYPE_INT, "compat_42",
1778 		       SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"),
1779 		       NULL, 0, &tcp_compat_42, 0,
1780 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL);
1781 	sysctl_createv(clog, 0, NULL, NULL,
1782 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1783 		       CTLTYPE_INT, "cwm",
1784 		       SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window "
1785 				    "Monitoring"),
1786 		       NULL, 0, &tcp_cwm, 0,
1787 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL);
1788 	sysctl_createv(clog, 0, NULL, NULL,
1789 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1790 		       CTLTYPE_INT, "cwm_burstsize",
1791 		       SYSCTL_DESCR("Congestion Window Monitoring allowed "
1792 				    "burst count in packets"),
1793 		       NULL, 0, &tcp_cwm_burstsize, 0,
1794 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE,
1795 		       CTL_EOL);
1796 	sysctl_createv(clog, 0, NULL, NULL,
1797 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1798 		       CTLTYPE_INT, "ack_on_push",
1799 		       SYSCTL_DESCR("Immediately return ACK when PSH is "
1800 				    "received"),
1801 		       NULL, 0, &tcp_ack_on_push, 0,
1802 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL);
1803 	sysctl_createv(clog, 0, NULL, NULL,
1804 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1805 		       CTLTYPE_INT, "keepidle",
1806 		       SYSCTL_DESCR("Allowed connection idle ticks before a "
1807 				    "keepalive probe is sent"),
1808 		       sysctl_tcp_keep, 0, &tcp_keepidle, 0,
1809 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL);
1810 	sysctl_createv(clog, 0, NULL, NULL,
1811 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1812 		       CTLTYPE_INT, "keepintvl",
1813 		       SYSCTL_DESCR("Ticks before next keepalive probe is sent"),
1814 		       sysctl_tcp_keep, 0, &tcp_keepintvl, 0,
1815 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL);
1816 	sysctl_createv(clog, 0, NULL, NULL,
1817 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1818 		       CTLTYPE_INT, "keepcnt",
1819 		       SYSCTL_DESCR("Number of keepalive probes to send"),
1820 		       sysctl_tcp_keep, 0, &tcp_keepcnt, 0,
1821 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL);
1822 	sysctl_createv(clog, 0, NULL, NULL,
1823 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
1824 		       CTLTYPE_INT, "slowhz",
1825 		       SYSCTL_DESCR("Keepalive ticks per second"),
1826 		       NULL, PR_SLOWHZ, NULL, 0,
1827 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL);
1828 	sysctl_createv(clog, 0, NULL, NULL,
1829 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1830 		       CTLTYPE_INT, "log_refused",
1831 		       SYSCTL_DESCR("Log refused TCP connections"),
1832 		       NULL, 0, &tcp_log_refused, 0,
1833 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL);
1834 #if 0 /* obsoleted */
1835 	sysctl_createv(clog, 0, NULL, NULL,
1836 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1837 		       CTLTYPE_INT, "rstratelimit", NULL,
1838 		       NULL, 0, &tcp_rst_ratelim, 0,
1839 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL);
1840 #endif
1841 	sysctl_createv(clog, 0, NULL, NULL,
1842 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1843 		       CTLTYPE_INT, "rstppslimit",
1844 		       SYSCTL_DESCR("Maximum number of RST packets to send "
1845 				    "per second"),
1846 		       NULL, 0, &tcp_rst_ppslim, 0,
1847 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL);
1848 	sysctl_createv(clog, 0, NULL, NULL,
1849 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1850 		       CTLTYPE_INT, "delack_ticks",
1851 		       SYSCTL_DESCR("Number of ticks to delay sending an ACK"),
1852 		       NULL, 0, &tcp_delack_ticks, 0,
1853 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL);
1854 	sysctl_createv(clog, 0, NULL, NULL,
1855 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1856 		       CTLTYPE_INT, "init_win_local",
1857 		       SYSCTL_DESCR("Initial TCP window size (in segments)"),
1858 		       NULL, 0, &tcp_init_win_local, 0,
1859 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL,
1860 		       CTL_EOL);
1861 	sysctl_createv(clog, 0, NULL, NULL,
1862 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1863 		       CTLTYPE_STRUCT, "ident",
1864 		       SYSCTL_DESCR("RFC1413 Identification Protocol lookups"),
1865 		       sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t),
1866 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL);
1867 	sysctl_createv(clog, 0, NULL, NULL,
1868 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1869 		       CTLTYPE_INT, "do_loopback_cksum",
1870 		       SYSCTL_DESCR("Perform TCP checksum on loopback"),
1871 		       NULL, 0, &tcp_do_loopback_cksum, 0,
1872 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM,
1873 		       CTL_EOL);
1874 	sysctl_createv(clog, 0, NULL, NULL,
1875 		       CTLFLAG_PERMANENT,
1876 		       CTLTYPE_STRUCT, "pcblist",
1877 		       SYSCTL_DESCR("TCP protocol control block list"),
1878 		       sysctl_inpcblist, 0, &tcbtable, 0,
1879 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
1880 		       CTL_EOL);
1881 	sysctl_createv(clog, 0, NULL, NULL,
1882 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1883 		       CTLTYPE_INT, "keepinit",
1884 		       SYSCTL_DESCR("Ticks before initial tcp connection times out"),
1885 		       sysctl_tcp_keep, 0, &tcp_keepinit, 0,
1886 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1887 
1888 	/* TCP socket buffers auto-sizing nodes */
1889 	sysctl_createv(clog, 0, NULL, NULL,
1890 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1891 		       CTLTYPE_INT, "recvbuf_auto",
1892 		       SYSCTL_DESCR("Enable automatic receive "
1893 		           "buffer sizing (experimental)"),
1894 		       NULL, 0, &tcp_do_autorcvbuf, 0,
1895 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1896 	sysctl_createv(clog, 0, NULL, NULL,
1897 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1898 		       CTLTYPE_INT, "recvbuf_inc",
1899 		       SYSCTL_DESCR("Incrementor step size of "
1900 		           "automatic receive buffer"),
1901 		       NULL, 0, &tcp_autorcvbuf_inc, 0,
1902 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1903 	sysctl_createv(clog, 0, NULL, NULL,
1904 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1905 		       CTLTYPE_INT, "recvbuf_max",
1906 		       SYSCTL_DESCR("Max size of automatic receive buffer"),
1907 		       NULL, 0, &tcp_autorcvbuf_max, 0,
1908 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1909 
1910 	sysctl_createv(clog, 0, NULL, NULL,
1911 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1912 		       CTLTYPE_INT, "sendbuf_auto",
1913 		       SYSCTL_DESCR("Enable automatic send "
1914 		           "buffer sizing (experimental)"),
1915 		       NULL, 0, &tcp_do_autosndbuf, 0,
1916 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1917 	sysctl_createv(clog, 0, NULL, NULL,
1918 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1919 		       CTLTYPE_INT, "sendbuf_inc",
1920 		       SYSCTL_DESCR("Incrementor step size of "
1921 		           "automatic send buffer"),
1922 		       NULL, 0, &tcp_autosndbuf_inc, 0,
1923 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1924 	sysctl_createv(clog, 0, NULL, NULL,
1925 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1926 		       CTLTYPE_INT, "sendbuf_max",
1927 		       SYSCTL_DESCR("Max size of automatic send buffer"),
1928 		       NULL, 0, &tcp_autosndbuf_max, 0,
1929 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1930 
1931 	/* ECN subtree */
1932 	sysctl_createv(clog, 0, NULL, &ecn_node,
1933 	    	       CTLFLAG_PERMANENT,
1934 		       CTLTYPE_NODE, "ecn",
1935 	    	       SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
1936 	    	       NULL, 0, NULL, 0,
1937 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1938 	sysctl_createv(clog, 0, &ecn_node, NULL,
1939 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1940 		       CTLTYPE_INT, "enable",
1941 		       SYSCTL_DESCR("Enable TCP Explicit Congestion "
1942 			   "Notification"),
1943 	    	       NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL);
1944 	sysctl_createv(clog, 0, &ecn_node, NULL,
1945 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1946 		       CTLTYPE_INT, "maxretries",
1947 		       SYSCTL_DESCR("Number of times to retry ECN setup "
1948 			       "before disabling ECN on the connection"),
1949 	    	       NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL);
1950 
1951 	/* SACK gets it's own little subtree. */
1952 	sysctl_createv(clog, 0, NULL, &sack_node,
1953 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1954 		       CTLTYPE_INT, "enable",
1955 		       SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"),
1956 		       NULL, 0, &tcp_do_sack, 0,
1957 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1958 	sysctl_createv(clog, 0, NULL, &sack_node,
1959 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1960 		       CTLTYPE_INT, "maxholes",
1961 		       SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"),
1962 		       NULL, 0, &tcp_sack_tp_maxholes, 0,
1963 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1964 	sysctl_createv(clog, 0, NULL, &sack_node,
1965 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1966 		       CTLTYPE_INT, "globalmaxholes",
1967 		       SYSCTL_DESCR("Global maximum number of TCP SACK holes"),
1968 		       NULL, 0, &tcp_sack_globalmaxholes, 0,
1969 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1970 	sysctl_createv(clog, 0, NULL, &sack_node,
1971 		       CTLFLAG_PERMANENT,
1972 		       CTLTYPE_INT, "globalholes",
1973 		       SYSCTL_DESCR("Global number of TCP SACK holes"),
1974 		       NULL, 0, &tcp_sack_globalholes, 0,
1975 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1976 
1977 	sysctl_createv(clog, 0, NULL, NULL,
1978 		       CTLFLAG_PERMANENT,
1979 		       CTLTYPE_STRUCT, "stats",
1980 		       SYSCTL_DESCR("TCP statistics"),
1981 		       sysctl_net_inet_tcp_stats, 0, NULL, 0,
1982 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS,
1983 		       CTL_EOL);
1984 #ifdef TCP_DEBUG
1985 	sysctl_createv(clog, 0, NULL, NULL,
1986 		       CTLFLAG_PERMANENT,
1987 		       CTLTYPE_STRUCT, "debug",
1988 		       SYSCTL_DESCR("TCP sockets debug information"),
1989 		       NULL, 0, &tcp_debug, sizeof(tcp_debug),
1990 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG,
1991 		       CTL_EOL);
1992 	sysctl_createv(clog, 0, NULL, NULL,
1993 		       CTLFLAG_PERMANENT,
1994 		       CTLTYPE_INT, "debx",
1995 		       SYSCTL_DESCR("Number of TCP debug sockets messages"),
1996 		       NULL, 0, &tcp_debx, sizeof(tcp_debx),
1997 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX,
1998 		       CTL_EOL);
1999 #endif
2000 	sysctl_createv(clog, 0, NULL, NULL,
2001 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2002 		       CTLTYPE_STRUCT, "drop",
2003 		       SYSCTL_DESCR("TCP drop connection"),
2004 		       sysctl_net_inet_tcp_drop, 0, NULL, 0,
2005 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL);
2006 #if NRND > 0
2007 	sysctl_createv(clog, 0, NULL, NULL,
2008 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2009 		       CTLTYPE_INT, "iss_hash",
2010 		       SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic "
2011 				    "hash computation"),
2012 		       NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948),
2013 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2014 		       CTL_EOL);
2015 #endif
2016 
2017 	/* ABC subtree */
2018 
2019 	sysctl_createv(clog, 0, NULL, &abc_node,
2020 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc",
2021 		       SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"),
2022 		       NULL, 0, NULL, 0,
2023 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2024 	sysctl_createv(clog, 0, &abc_node, NULL,
2025 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2026 		       CTLTYPE_INT, "enable",
2027 		       SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"),
2028 		       NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL);
2029 	sysctl_createv(clog, 0, &abc_node, NULL,
2030 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2031 		       CTLTYPE_INT, "aggressive",
2032 		       SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"),
2033 		       NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL);
2034 }
2035 
2036 /*
2037  * Sysctl for tcp variables.
2038  */
2039 #ifdef INET
2040 SYSCTL_SETUP(sysctl_net_inet_tcp_setup, "sysctl net.inet.tcp subtree setup")
2041 {
2042 
2043 	sysctl_net_inet_tcp_setup2(clog, PF_INET, "inet", "tcp");
2044 }
2045 #endif /* INET */
2046 
2047 #ifdef INET6
2048 SYSCTL_SETUP(sysctl_net_inet6_tcp6_setup, "sysctl net.inet6.tcp6 subtree setup")
2049 {
2050 
2051 	sysctl_net_inet_tcp_setup2(clog, PF_INET6, "inet6", "tcp6");
2052 }
2053 #endif /* INET6 */
2054