xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 0920b4f20b78ab1ccd9f2312fbe10deaf000cbf3)
1 /*	$NetBSD: tcp_usrreq.c,v 1.136 2007/08/02 02:42:41 rmind 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  * 3. All advertising materials mentioning features or use of this software
53  *    must display the following acknowledgement:
54  *	This product includes software developed by the NetBSD
55  *	Foundation, Inc. and its contributors.
56  * 4. Neither the name of The NetBSD Foundation nor the names of its
57  *    contributors may be used to endorse or promote products derived
58  *    from this software without specific prior written permission.
59  *
60  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
61  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
62  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
63  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
64  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
65  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
66  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
67  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
68  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
69  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
70  * POSSIBILITY OF SUCH DAMAGE.
71  */
72 
73 /*
74  * Copyright (c) 1982, 1986, 1988, 1993, 1995
75  *	The Regents of the University of California.  All rights reserved.
76  *
77  * Redistribution and use in source and binary forms, with or without
78  * modification, are permitted provided that the following conditions
79  * are met:
80  * 1. Redistributions of source code must retain the above copyright
81  *    notice, this list of conditions and the following disclaimer.
82  * 2. Redistributions in binary form must reproduce the above copyright
83  *    notice, this list of conditions and the following disclaimer in the
84  *    documentation and/or other materials provided with the distribution.
85  * 3. Neither the name of the University nor the names of its contributors
86  *    may be used to endorse or promote products derived from this software
87  *    without specific prior written permission.
88  *
89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99  * SUCH DAMAGE.
100  *
101  *	@(#)tcp_usrreq.c	8.5 (Berkeley) 6/21/95
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.136 2007/08/02 02:42:41 rmind Exp $");
106 
107 #include "opt_inet.h"
108 #include "opt_ipsec.h"
109 #include "opt_tcp_debug.h"
110 #include "opt_mbuftrace.h"
111 #include "rnd.h"
112 
113 #include <sys/param.h>
114 #include <sys/systm.h>
115 #include <sys/kernel.h>
116 #include <sys/malloc.h>
117 #include <sys/mbuf.h>
118 #include <sys/socket.h>
119 #include <sys/socketvar.h>
120 #include <sys/protosw.h>
121 #include <sys/errno.h>
122 #include <sys/stat.h>
123 #include <sys/proc.h>
124 #include <sys/domain.h>
125 #include <sys/sysctl.h>
126 #include <sys/kauth.h>
127 
128 #include <net/if.h>
129 #include <net/route.h>
130 
131 #include <netinet/in.h>
132 #include <netinet/in_systm.h>
133 #include <netinet/in_var.h>
134 #include <netinet/ip.h>
135 #include <netinet/in_pcb.h>
136 #include <netinet/ip_var.h>
137 #include <netinet/in_offload.h>
138 
139 #ifdef INET6
140 #ifndef INET
141 #include <netinet/in.h>
142 #endif
143 #include <netinet/ip6.h>
144 #include <netinet6/in6_pcb.h>
145 #include <netinet6/ip6_var.h>
146 #include <netinet6/scope6_var.h>
147 #endif
148 
149 #include <netinet/tcp.h>
150 #include <netinet/tcp_fsm.h>
151 #include <netinet/tcp_seq.h>
152 #include <netinet/tcp_timer.h>
153 #include <netinet/tcp_var.h>
154 #include <netinet/tcp_congctl.h>
155 #include <netinet/tcpip.h>
156 #include <netinet/tcp_debug.h>
157 
158 #include "opt_tcp_space.h"
159 
160 #ifdef IPSEC
161 #include <netinet6/ipsec.h>
162 #endif /*IPSEC*/
163 
164 /*
165  * TCP protocol interface to socket abstraction.
166  */
167 
168 /*
169  * Process a TCP user request for TCP tb.  If this is a send request
170  * then m is the mbuf chain of send data.  If this is a timer expiration
171  * (called from the software clock routine), then timertype tells which timer.
172  */
173 /*ARGSUSED*/
174 int
175 tcp_usrreq(struct socket *so, int req,
176     struct mbuf *m, struct mbuf *nam, struct mbuf *control, struct lwp *l)
177 {
178 	struct inpcb *inp;
179 #ifdef INET6
180 	struct in6pcb *in6p;
181 #endif
182 	struct tcpcb *tp = NULL;
183 	int s;
184 	int error = 0;
185 #ifdef TCP_DEBUG
186 	int ostate = 0;
187 #endif
188 	int family;	/* family of the socket */
189 
190 	family = so->so_proto->pr_domain->dom_family;
191 
192 	if (req == PRU_CONTROL) {
193 		switch (family) {
194 #ifdef INET
195 		case PF_INET:
196 			return (in_control(so, (long)m, (void *)nam,
197 			    (struct ifnet *)control, l));
198 #endif
199 #ifdef INET6
200 		case PF_INET6:
201 			return (in6_control(so, (long)m, (void *)nam,
202 			    (struct ifnet *)control, l));
203 #endif
204 		default:
205 			return EAFNOSUPPORT;
206 		}
207 	}
208 
209 	s = splsoftnet();
210 
211 	if (req == PRU_PURGEIF) {
212 		switch (family) {
213 #ifdef INET
214 		case PF_INET:
215 			in_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
216 			in_purgeif((struct ifnet *)control);
217 			in_pcbpurgeif(&tcbtable, (struct ifnet *)control);
218 			break;
219 #endif
220 #ifdef INET6
221 		case PF_INET6:
222 			in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
223 			in6_purgeif((struct ifnet *)control);
224 			in6_pcbpurgeif(&tcbtable, (struct ifnet *)control);
225 			break;
226 #endif
227 		default:
228 			splx(s);
229 			return (EAFNOSUPPORT);
230 		}
231 		splx(s);
232 		return (0);
233 	}
234 
235 	switch (family) {
236 #ifdef INET
237 	case PF_INET:
238 		inp = sotoinpcb(so);
239 #ifdef INET6
240 		in6p = NULL;
241 #endif
242 		break;
243 #endif
244 #ifdef INET6
245 	case PF_INET6:
246 		inp = NULL;
247 		in6p = sotoin6pcb(so);
248 		break;
249 #endif
250 	default:
251 		splx(s);
252 		return EAFNOSUPPORT;
253 	}
254 
255 #ifdef DIAGNOSTIC
256 #ifdef INET6
257 	if (inp && in6p)
258 		panic("tcp_usrreq: both inp and in6p set to non-NULL");
259 #endif
260 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
261 		panic("tcp_usrreq: unexpected control mbuf");
262 #endif
263 	/*
264 	 * When a TCP is attached to a socket, then there will be
265 	 * a (struct inpcb) pointed at by the socket, and this
266 	 * structure will point at a subsidary (struct tcpcb).
267 	 */
268 #ifndef INET6
269 	if (inp == 0 && req != PRU_ATTACH)
270 #else
271 	if ((inp == 0 && in6p == 0) && req != PRU_ATTACH)
272 #endif
273 	{
274 		error = EINVAL;
275 		goto release;
276 	}
277 #ifdef INET
278 	if (inp) {
279 		tp = intotcpcb(inp);
280 		/* WHAT IF TP IS 0? */
281 #ifdef KPROF
282 		tcp_acounts[tp->t_state][req]++;
283 #endif
284 #ifdef TCP_DEBUG
285 		ostate = tp->t_state;
286 #endif
287 	}
288 #endif
289 #ifdef INET6
290 	if (in6p) {
291 		tp = in6totcpcb(in6p);
292 		/* WHAT IF TP IS 0? */
293 #ifdef KPROF
294 		tcp_acounts[tp->t_state][req]++;
295 #endif
296 #ifdef TCP_DEBUG
297 		ostate = tp->t_state;
298 #endif
299 	}
300 #endif
301 
302 	switch (req) {
303 
304 	/*
305 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
306 	 * and an internet control block.
307 	 */
308 	case PRU_ATTACH:
309 #ifndef INET6
310 		if (inp != 0)
311 #else
312 		if (inp != 0 || in6p != 0)
313 #endif
314 		{
315 			error = EISCONN;
316 			break;
317 		}
318 		error = tcp_attach(so);
319 		if (error)
320 			break;
321 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
322 			so->so_linger = TCP_LINGERTIME;
323 		tp = sototcpcb(so);
324 		break;
325 
326 	/*
327 	 * PRU_DETACH detaches the TCP protocol from the socket.
328 	 */
329 	case PRU_DETACH:
330 		tp = tcp_disconnect(tp);
331 		break;
332 
333 	/*
334 	 * Give the socket an address.
335 	 */
336 	case PRU_BIND:
337 		switch (family) {
338 #ifdef INET
339 		case PF_INET:
340 			error = in_pcbbind(inp, nam, l);
341 			break;
342 #endif
343 #ifdef INET6
344 		case PF_INET6:
345 			error = in6_pcbbind(in6p, nam, l);
346 			if (!error) {
347 				/* mapped addr case */
348 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
349 					tp->t_family = AF_INET;
350 				else
351 					tp->t_family = AF_INET6;
352 			}
353 			break;
354 #endif
355 		}
356 		break;
357 
358 	/*
359 	 * Prepare to accept connections.
360 	 */
361 	case PRU_LISTEN:
362 #ifdef INET
363 		if (inp && inp->inp_lport == 0) {
364 			error = in_pcbbind(inp, (struct mbuf *)0,
365 			    (struct lwp *)0);
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,
393 				    (struct lwp *)0);
394 				if (error)
395 					break;
396 			}
397 			error = in_pcbconnect(inp, nam, l);
398 		}
399 #endif
400 #ifdef INET6
401 		if (in6p) {
402 			if (in6p->in6p_lport == 0) {
403 				error = in6_pcbbind(in6p, (struct mbuf *)0,
404 				    (struct lwp *)0);
405 				if (error)
406 					break;
407 			}
408 			error = in6_pcbconnect(in6p, nam, l);
409 			if (!error) {
410 				/* mapped addr case */
411 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
412 					tp->t_family = AF_INET;
413 				else
414 					tp->t_family = AF_INET6;
415 			}
416 		}
417 #endif
418 		if (error)
419 			break;
420 		tp->t_template = tcp_template(tp);
421 		if (tp->t_template == 0) {
422 #ifdef INET
423 			if (inp)
424 				in_pcbdisconnect(inp);
425 #endif
426 #ifdef INET6
427 			if (in6p)
428 				in6_pcbdisconnect(in6p);
429 #endif
430 			error = ENOBUFS;
431 			break;
432 		}
433 		/*
434 		 * Compute window scaling to request:
435 		 * XXX: This should be moved to tcp_output(),
436 		 * and should be based on the actual MSS.
437 		 */
438 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
439 		    (0x1 << tp->request_r_scale) < tcp_minmss)
440 			tp->request_r_scale++;
441 		soisconnecting(so);
442 		tcpstat.tcps_connattempt++;
443 		tp->t_state = TCPS_SYN_SENT;
444 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
445 		tp->iss = tcp_new_iss(tp, 0);
446 		tcp_sendseqinit(tp);
447 		error = tcp_output(tp);
448 		break;
449 
450 	/*
451 	 * Create a TCP connection between two sockets.
452 	 */
453 	case PRU_CONNECT2:
454 		error = EOPNOTSUPP;
455 		break;
456 
457 	/*
458 	 * Initiate disconnect from peer.
459 	 * If connection never passed embryonic stage, just drop;
460 	 * else if don't need to let data drain, then can just drop anyways,
461 	 * else have to begin TCP shutdown process: mark socket disconnecting,
462 	 * drain unread data, state switch to reflect user close, and
463 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
464 	 * when peer sends FIN and acks ours.
465 	 *
466 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
467 	 */
468 	case PRU_DISCONNECT:
469 		tp = tcp_disconnect(tp);
470 		break;
471 
472 	/*
473 	 * Accept a connection.  Essentially all the work is
474 	 * done at higher levels; just return the address
475 	 * of the peer, storing through addr.
476 	 */
477 	case PRU_ACCEPT:
478 #ifdef INET
479 		if (inp)
480 			in_setpeeraddr(inp, nam);
481 #endif
482 #ifdef INET6
483 		if (in6p)
484 			in6_setpeeraddr(in6p, nam);
485 #endif
486 		break;
487 
488 	/*
489 	 * Mark the connection as being incapable of further output.
490 	 */
491 	case PRU_SHUTDOWN:
492 		socantsendmore(so);
493 		tp = tcp_usrclosed(tp);
494 		if (tp)
495 			error = tcp_output(tp);
496 		break;
497 
498 	/*
499 	 * After a receive, possibly send window update to peer.
500 	 */
501 	case PRU_RCVD:
502 		/*
503 		 * soreceive() calls this function when a user receives
504 		 * ancillary data on a listening socket. We don't call
505 		 * tcp_output in such a case, since there is no header
506 		 * template for a listening socket and hence the kernel
507 		 * will panic.
508 		 */
509 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
510 			(void) tcp_output(tp);
511 		break;
512 
513 	/*
514 	 * Do a send by putting data in output queue and updating urgent
515 	 * marker if URG set.  Possibly send more data.
516 	 */
517 	case PRU_SEND:
518 		if (control && control->m_len) {
519 			m_freem(control);
520 			m_freem(m);
521 			error = EINVAL;
522 			break;
523 		}
524 		sbappendstream(&so->so_snd, m);
525 		error = tcp_output(tp);
526 		break;
527 
528 	/*
529 	 * Abort the TCP.
530 	 */
531 	case PRU_ABORT:
532 		tp = tcp_drop(tp, ECONNABORTED);
533 		break;
534 
535 	case PRU_SENSE:
536 		/*
537 		 * stat: don't bother with a blocksize.
538 		 */
539 		splx(s);
540 		return (0);
541 
542 	case PRU_RCVOOB:
543 		if (control && control->m_len) {
544 			m_freem(control);
545 			m_freem(m);
546 			error = EINVAL;
547 			break;
548 		}
549 		if ((so->so_oobmark == 0 &&
550 		    (so->so_state & SS_RCVATMARK) == 0) ||
551 		    so->so_options & SO_OOBINLINE ||
552 		    tp->t_oobflags & TCPOOB_HADDATA) {
553 			error = EINVAL;
554 			break;
555 		}
556 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
557 			error = EWOULDBLOCK;
558 			break;
559 		}
560 		m->m_len = 1;
561 		*mtod(m, char *) = tp->t_iobc;
562 		if (((long)nam & MSG_PEEK) == 0)
563 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
564 		break;
565 
566 	case PRU_SENDOOB:
567 		if (sbspace(&so->so_snd) < -512) {
568 			m_freem(m);
569 			error = ENOBUFS;
570 			break;
571 		}
572 		/*
573 		 * According to RFC961 (Assigned Protocols),
574 		 * the urgent pointer points to the last octet
575 		 * of urgent data.  We continue, however,
576 		 * to consider it to indicate the first octet
577 		 * of data past the urgent section.
578 		 * Otherwise, snd_up should be one lower.
579 		 */
580 		sbappendstream(&so->so_snd, m);
581 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
582 		tp->t_force = 1;
583 		error = tcp_output(tp);
584 		tp->t_force = 0;
585 		break;
586 
587 	case PRU_SOCKADDR:
588 #ifdef INET
589 		if (inp)
590 			in_setsockaddr(inp, nam);
591 #endif
592 #ifdef INET6
593 		if (in6p)
594 			in6_setsockaddr(in6p, nam);
595 #endif
596 		break;
597 
598 	case PRU_PEERADDR:
599 #ifdef INET
600 		if (inp)
601 			in_setpeeraddr(inp, nam);
602 #endif
603 #ifdef INET6
604 		if (in6p)
605 			in6_setpeeraddr(in6p, nam);
606 #endif
607 		break;
608 
609 	default:
610 		panic("tcp_usrreq");
611 	}
612 #ifdef TCP_DEBUG
613 	if (tp && (so->so_options & SO_DEBUG))
614 		tcp_trace(TA_USER, ostate, tp, NULL, req);
615 #endif
616 
617 release:
618 	splx(s);
619 	return (error);
620 }
621 
622 static void
623 change_keepalive(struct socket *so, struct tcpcb *tp)
624 {
625 	tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl;
626 	TCP_TIMER_DISARM(tp, TCPT_KEEP);
627 	TCP_TIMER_DISARM(tp, TCPT_2MSL);
628 
629 	if (tp->t_state == TCPS_SYN_RECEIVED ||
630 	    tp->t_state == TCPS_SYN_SENT) {
631 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
632 	} else if (so->so_options & SO_KEEPALIVE &&
633 	    tp->t_state <= TCPS_CLOSE_WAIT) {
634 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl);
635 	} else {
636 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
637 	}
638 
639 	if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
640 		TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
641 }
642 
643 
644 int
645 tcp_ctloutput(int op, struct socket *so, int level, int optname,
646     struct mbuf **mp)
647 {
648 	int error = 0, s;
649 	struct inpcb *inp;
650 #ifdef INET6
651 	struct in6pcb *in6p;
652 #endif
653 	struct tcpcb *tp;
654 	struct mbuf *m;
655 	int i;
656 	u_int ui;
657 	int family;	/* family of the socket */
658 
659 	family = so->so_proto->pr_domain->dom_family;
660 
661 	s = splsoftnet();
662 	switch (family) {
663 #ifdef INET
664 	case PF_INET:
665 		inp = sotoinpcb(so);
666 #ifdef INET6
667 		in6p = NULL;
668 #endif
669 		break;
670 #endif
671 #ifdef INET6
672 	case PF_INET6:
673 		inp = NULL;
674 		in6p = sotoin6pcb(so);
675 		break;
676 #endif
677 	default:
678 		splx(s);
679 		panic("%s: af %d", __func__, family);
680 	}
681 #ifndef INET6
682 	if (inp == NULL)
683 #else
684 	if (inp == NULL && in6p == NULL)
685 #endif
686 	{
687 		splx(s);
688 		if (op == PRCO_SETOPT && *mp)
689 			(void) m_free(*mp);
690 		return (ECONNRESET);
691 	}
692 	if (level != IPPROTO_TCP) {
693 		switch (family) {
694 #ifdef INET
695 		case PF_INET:
696 			error = ip_ctloutput(op, so, level, optname, mp);
697 			break;
698 #endif
699 #ifdef INET6
700 		case PF_INET6:
701 			error = ip6_ctloutput(op, so, level, optname, mp);
702 			break;
703 #endif
704 		}
705 		splx(s);
706 		return (error);
707 	}
708 	if (inp)
709 		tp = intotcpcb(inp);
710 #ifdef INET6
711 	else if (in6p)
712 		tp = in6totcpcb(in6p);
713 #endif
714 	else
715 		tp = NULL;
716 
717 	switch (op) {
718 
719 	case PRCO_SETOPT:
720 		m = *mp;
721 		switch (optname) {
722 
723 #ifdef TCP_SIGNATURE
724 		case TCP_MD5SIG:
725 			if (m == NULL || m->m_len < sizeof (int))
726 				error = EINVAL;
727 			if (error)
728 				break;
729 			if (*mtod(m, int *) > 0)
730 				tp->t_flags |= TF_SIGNATURE;
731 			else
732 				tp->t_flags &= ~TF_SIGNATURE;
733 			break;
734 #endif /* TCP_SIGNATURE */
735 
736 		case TCP_NODELAY:
737 			if (m == NULL || m->m_len < sizeof (int))
738 				error = EINVAL;
739 			else if (*mtod(m, int *))
740 				tp->t_flags |= TF_NODELAY;
741 			else
742 				tp->t_flags &= ~TF_NODELAY;
743 			break;
744 
745 		case TCP_MAXSEG:
746 			if (m && m->m_len >= sizeof(int) &&
747 			    (i = *mtod(m, int *)) > 0 &&
748 			    i <= tp->t_peermss)
749 				tp->t_peermss = i;  /* limit on send size */
750 			else
751 				error = EINVAL;
752 			break;
753 #ifdef notyet
754 		case TCP_CONGCTL:
755 			if (m == NULL)
756 				error = EINVAL;
757 			error = tcp_congctl_select(tp, mtod(m, char *));
758 #endif
759 			break;
760 
761 		case TCP_KEEPIDLE:
762 			if (m && m->m_len >= sizeof(u_int) &&
763 			    (ui = *mtod(m, u_int *)) > 0) {
764 				tp->t_keepidle = ui;
765 				change_keepalive(so, tp);
766 			} else
767 				error = EINVAL;
768 			break;
769 
770 		case TCP_KEEPINTVL:
771 			if (m && m->m_len >= sizeof(u_int) &&
772 			    (ui = *mtod(m, u_int *)) > 0) {
773 				tp->t_keepintvl = ui;
774 				change_keepalive(so, tp);
775 			} else
776 				error = EINVAL;
777 			break;
778 
779 		case TCP_KEEPCNT:
780 			if (m && m->m_len >= sizeof(u_int) &&
781 			    (ui = *mtod(m, u_int *)) > 0) {
782 				tp->t_keepcnt = ui;
783 				change_keepalive(so, tp);
784 			} else
785 				error = EINVAL;
786 			break;
787 
788 		case TCP_KEEPINIT:
789 			if (m && m->m_len >= sizeof(u_int) &&
790 			    (ui = *mtod(m, u_int *)) > 0) {
791 				tp->t_keepinit = ui;
792 				change_keepalive(so, tp);
793 			} else
794 				error = EINVAL;
795 			break;
796 
797 		default:
798 			error = ENOPROTOOPT;
799 			break;
800 		}
801 		if (m)
802 			(void) m_free(m);
803 		break;
804 
805 	case PRCO_GETOPT:
806 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
807 		m->m_len = sizeof(int);
808 		MCLAIM(m, so->so_mowner);
809 
810 		switch (optname) {
811 #ifdef TCP_SIGNATURE
812 		case TCP_MD5SIG:
813 			*mtod(m, int *) = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
814 			break;
815 #endif
816 		case TCP_NODELAY:
817 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
818 			break;
819 		case TCP_MAXSEG:
820 			*mtod(m, int *) = tp->t_peermss;
821 			break;
822 #ifdef notyet
823 		case TCP_CONGCTL:
824 			break;
825 #endif
826 		default:
827 			error = ENOPROTOOPT;
828 			break;
829 		}
830 		break;
831 	}
832 	splx(s);
833 	return (error);
834 }
835 
836 #ifndef TCP_SENDSPACE
837 #define	TCP_SENDSPACE	1024*32
838 #endif
839 int	tcp_sendspace = TCP_SENDSPACE;
840 #ifndef TCP_RECVSPACE
841 #define	TCP_RECVSPACE	1024*32
842 #endif
843 int	tcp_recvspace = TCP_RECVSPACE;
844 
845 /*
846  * Attach TCP protocol to socket, allocating
847  * internet protocol control block, tcp control block,
848  * bufer space, and entering LISTEN state if to accept connections.
849  */
850 int
851 tcp_attach(struct socket *so)
852 {
853 	struct tcpcb *tp;
854 	struct inpcb *inp;
855 #ifdef INET6
856 	struct in6pcb *in6p;
857 #endif
858 	int error;
859 	int family;	/* family of the socket */
860 
861 	family = so->so_proto->pr_domain->dom_family;
862 
863 #ifdef MBUFTRACE
864 	so->so_mowner = &tcp_sock_mowner;
865 	so->so_rcv.sb_mowner = &tcp_sock_rx_mowner;
866 	so->so_snd.sb_mowner = &tcp_sock_tx_mowner;
867 #endif
868 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
869 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
870 		if (error)
871 			return (error);
872 	}
873 
874 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
875 	so->so_snd.sb_flags |= SB_AUTOSIZE;
876 
877 	switch (family) {
878 #ifdef INET
879 	case PF_INET:
880 		error = in_pcballoc(so, &tcbtable);
881 		if (error)
882 			return (error);
883 		inp = sotoinpcb(so);
884 #ifdef INET6
885 		in6p = NULL;
886 #endif
887 		break;
888 #endif
889 #ifdef INET6
890 	case PF_INET6:
891 		error = in6_pcballoc(so, &tcbtable);
892 		if (error)
893 			return (error);
894 		inp = NULL;
895 		in6p = sotoin6pcb(so);
896 		break;
897 #endif
898 	default:
899 		return EAFNOSUPPORT;
900 	}
901 	if (inp)
902 		tp = tcp_newtcpcb(family, (void *)inp);
903 #ifdef INET6
904 	else if (in6p)
905 		tp = tcp_newtcpcb(family, (void *)in6p);
906 #endif
907 	else
908 		tp = NULL;
909 
910 	if (tp == 0) {
911 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
912 
913 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
914 #ifdef INET
915 		if (inp)
916 			in_pcbdetach(inp);
917 #endif
918 #ifdef INET6
919 		if (in6p)
920 			in6_pcbdetach(in6p);
921 #endif
922 		so->so_state |= nofd;
923 		return (ENOBUFS);
924 	}
925 	tp->t_state = TCPS_CLOSED;
926 	return (0);
927 }
928 
929 /*
930  * Initiate (or continue) disconnect.
931  * If embryonic state, just send reset (once).
932  * If in ``let data drain'' option and linger null, just drop.
933  * Otherwise (hard), mark socket disconnecting and drop
934  * current input data; switch states based on user close, and
935  * send segment to peer (with FIN).
936  */
937 struct tcpcb *
938 tcp_disconnect(struct tcpcb *tp)
939 {
940 	struct socket *so;
941 
942 	if (tp->t_inpcb)
943 		so = tp->t_inpcb->inp_socket;
944 #ifdef INET6
945 	else if (tp->t_in6pcb)
946 		so = tp->t_in6pcb->in6p_socket;
947 #endif
948 	else
949 		so = NULL;
950 
951 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
952 		tp = tcp_close(tp);
953 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
954 		tp = tcp_drop(tp, 0);
955 	else {
956 		soisdisconnecting(so);
957 		sbflush(&so->so_rcv);
958 		tp = tcp_usrclosed(tp);
959 		if (tp)
960 			(void) tcp_output(tp);
961 	}
962 	return (tp);
963 }
964 
965 /*
966  * User issued close, and wish to trail through shutdown states:
967  * if never received SYN, just forget it.  If got a SYN from peer,
968  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
969  * If already got a FIN from peer, then almost done; go to LAST_ACK
970  * state.  In all other cases, have already sent FIN to peer (e.g.
971  * after PRU_SHUTDOWN), and just have to play tedious game waiting
972  * for peer to send FIN or not respond to keep-alives, etc.
973  * We can let the user exit from the close as soon as the FIN is acked.
974  */
975 struct tcpcb *
976 tcp_usrclosed(struct tcpcb *tp)
977 {
978 
979 	switch (tp->t_state) {
980 
981 	case TCPS_CLOSED:
982 	case TCPS_LISTEN:
983 	case TCPS_SYN_SENT:
984 		tp->t_state = TCPS_CLOSED;
985 		tp = tcp_close(tp);
986 		break;
987 
988 	case TCPS_SYN_RECEIVED:
989 	case TCPS_ESTABLISHED:
990 		tp->t_state = TCPS_FIN_WAIT_1;
991 		break;
992 
993 	case TCPS_CLOSE_WAIT:
994 		tp->t_state = TCPS_LAST_ACK;
995 		break;
996 	}
997 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
998 		struct socket *so;
999 		if (tp->t_inpcb)
1000 			so = tp->t_inpcb->inp_socket;
1001 #ifdef INET6
1002 		else if (tp->t_in6pcb)
1003 			so = tp->t_in6pcb->in6p_socket;
1004 #endif
1005 		else
1006 			so = NULL;
1007 		if (so)
1008 			soisdisconnected(so);
1009 		/*
1010 		 * If we are in FIN_WAIT_2, we arrived here because the
1011 		 * application did a shutdown of the send side.  Like the
1012 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
1013 		 * a full close, we start a timer to make sure sockets are
1014 		 * not left in FIN_WAIT_2 forever.
1015 		 */
1016 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0))
1017 			TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle);
1018 	}
1019 	return (tp);
1020 }
1021 
1022 /*
1023  * sysctl helper routine for net.inet.ip.mssdflt.  it can't be less
1024  * than 32.
1025  */
1026 static int
1027 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS)
1028 {
1029 	int error, mssdflt;
1030 	struct sysctlnode node;
1031 
1032 	mssdflt = tcp_mssdflt;
1033 	node = *rnode;
1034 	node.sysctl_data = &mssdflt;
1035 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1036 	if (error || newp == NULL)
1037 		return (error);
1038 
1039 	if (mssdflt < 32)
1040 		return (EINVAL);
1041 	tcp_mssdflt = mssdflt;
1042 
1043 	return (0);
1044 }
1045 
1046 /*
1047  * sysctl helper routine for setting port related values under
1048  * net.inet.ip and net.inet6.ip6.  does basic range checking and does
1049  * additional checks for each type.  this code has placed in
1050  * tcp_input.c since INET and INET6 both use the same tcp code.
1051  *
1052  * this helper is not static so that both inet and inet6 can use it.
1053  */
1054 int
1055 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS)
1056 {
1057 	int error, tmp;
1058 	int apmin, apmax;
1059 #ifndef IPNOPRIVPORTS
1060 	int lpmin, lpmax;
1061 #endif /* IPNOPRIVPORTS */
1062 	struct sysctlnode node;
1063 
1064 	if (namelen != 0)
1065 		return (EINVAL);
1066 
1067 	switch (name[-3]) {
1068 #ifdef INET
1069 	    case PF_INET:
1070 		apmin = anonportmin;
1071 		apmax = anonportmax;
1072 #ifndef IPNOPRIVPORTS
1073 		lpmin = lowportmin;
1074 		lpmax = lowportmax;
1075 #endif /* IPNOPRIVPORTS */
1076 		break;
1077 #endif /* INET */
1078 #ifdef INET6
1079 	    case PF_INET6:
1080 		apmin = ip6_anonportmin;
1081 		apmax = ip6_anonportmax;
1082 #ifndef IPNOPRIVPORTS
1083 		lpmin = ip6_lowportmin;
1084 		lpmax = ip6_lowportmax;
1085 #endif /* IPNOPRIVPORTS */
1086 		break;
1087 #endif /* INET6 */
1088 	    default:
1089 		return (EINVAL);
1090 	}
1091 
1092 	/*
1093 	 * insert temporary copy into node, perform lookup on
1094 	 * temporary, then restore pointer
1095 	 */
1096 	node = *rnode;
1097 	tmp = *(int*)rnode->sysctl_data;
1098 	node.sysctl_data = &tmp;
1099 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1100 	if (error || newp == NULL)
1101 		return (error);
1102 
1103 	/*
1104 	 * simple port range check
1105 	 */
1106 	if (tmp < 0 || tmp > 65535)
1107 		return (EINVAL);
1108 
1109 	/*
1110 	 * per-node range checks
1111 	 */
1112 	switch (rnode->sysctl_num) {
1113 	case IPCTL_ANONPORTMIN:
1114 		if (tmp >= apmax)
1115 			return (EINVAL);
1116 #ifndef IPNOPRIVPORTS
1117 		if (tmp < IPPORT_RESERVED)
1118                         return (EINVAL);
1119 #endif /* IPNOPRIVPORTS */
1120 		break;
1121 
1122 	case IPCTL_ANONPORTMAX:
1123                 if (apmin >= tmp)
1124 			return (EINVAL);
1125 #ifndef IPNOPRIVPORTS
1126 		if (tmp < IPPORT_RESERVED)
1127                         return (EINVAL);
1128 #endif /* IPNOPRIVPORTS */
1129 		break;
1130 
1131 #ifndef IPNOPRIVPORTS
1132 	case IPCTL_LOWPORTMIN:
1133 		if (tmp >= lpmax ||
1134 		    tmp > IPPORT_RESERVEDMAX ||
1135 		    tmp < IPPORT_RESERVEDMIN)
1136 			return (EINVAL);
1137 		break;
1138 
1139 	case IPCTL_LOWPORTMAX:
1140 		if (lpmin >= tmp ||
1141 		    tmp > IPPORT_RESERVEDMAX ||
1142 		    tmp < IPPORT_RESERVEDMIN)
1143 			return (EINVAL);
1144 		break;
1145 #endif /* IPNOPRIVPORTS */
1146 
1147 	default:
1148 		return (EINVAL);
1149 	}
1150 
1151 	*(int*)rnode->sysctl_data = tmp;
1152 
1153 	return (0);
1154 }
1155 
1156 /*
1157  * The superuser can drop any connection.  Normal users can only drop
1158  * their own connections.
1159  */
1160 static inline int
1161 check_sockuid(struct socket *sockp, kauth_cred_t cred)
1162 {
1163 	uid_t sockuid;
1164 
1165 	sockuid = sockp->so_uidinfo->ui_uid;
1166 	if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) == 0 ||
1167 	    sockuid == kauth_cred_getuid(cred) ||
1168 	    sockuid == kauth_cred_geteuid(cred))
1169 		return 0;
1170 	return EACCES;
1171 }
1172 
1173 static inline int
1174 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp)
1175 {
1176 	size_t sz;
1177 	int error;
1178 	uid_t uid;
1179 
1180 	uid = sockp->so_uidinfo->ui_uid;
1181 	if (oldp) {
1182 		sz = MIN(sizeof(uid), *oldlenp);
1183 		error = copyout(&uid, oldp, sz);
1184 		if (error)
1185 			return error;
1186 	}
1187 	*oldlenp = sizeof(uid);
1188 	return 0;
1189 }
1190 
1191 static inline int
1192 inet4_ident_core(struct in_addr raddr, u_int rport,
1193     struct in_addr laddr, u_int lport,
1194     void *oldp, size_t *oldlenp,
1195     struct lwp *l, int dodrop)
1196 {
1197 	struct inpcb *inp;
1198 	struct socket *sockp;
1199 
1200 	inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport);
1201 
1202 	if (inp == NULL || (sockp = inp->inp_socket) == NULL)
1203 		return ESRCH;
1204 
1205 	if (dodrop) {
1206 		struct tcpcb *tp;
1207 
1208 		if (inp == NULL || (tp = intotcpcb(inp)) == NULL ||
1209 		    (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0)
1210 			return ESRCH;
1211 
1212 		if (check_sockuid(inp->inp_socket, l->l_cred) != 0)
1213 			return EACCES;
1214 
1215 		(void)tcp_drop(tp, ECONNABORTED);
1216 		return 0;
1217 	}
1218 	else
1219 		return copyout_uid(sockp, oldp, oldlenp);
1220 }
1221 
1222 #ifdef INET6
1223 static inline int
1224 inet6_ident_core(struct in6_addr *raddr, u_int rport,
1225     struct in6_addr *laddr, u_int lport,
1226     void *oldp, size_t *oldlenp,
1227     struct lwp *l, int dodrop)
1228 {
1229 	struct in6pcb *in6p;
1230 	struct socket *sockp;
1231 
1232 	in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0);
1233 
1234 	if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL)
1235 		return ESRCH;
1236 
1237 	if (dodrop) {
1238 		struct tcpcb *tp;
1239 
1240 		if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL ||
1241 		    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0)
1242 			return ESRCH;
1243 
1244 		if (check_sockuid(in6p->in6p_socket, l->l_cred) != 0)
1245 			return EACCES;
1246 
1247 		(void)tcp_drop(tp, ECONNABORTED);
1248 		return 0;
1249 	}
1250 	else
1251 		return copyout_uid(sockp, oldp, oldlenp);
1252 }
1253 #endif
1254 
1255 /*
1256  * sysctl helper routine for the net.inet.tcp.drop and
1257  * net.inet6.tcp6.drop nodes.
1258  */
1259 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident
1260 
1261 /*
1262  * sysctl helper routine for the net.inet.tcp.ident and
1263  * net.inet6.tcp6.ident nodes.  contains backwards compat code for the
1264  * old way of looking up the ident information for ipv4 which involves
1265  * stuffing the port/addr pairs into the mib lookup.
1266  */
1267 static int
1268 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS)
1269 {
1270 #ifdef INET
1271 	struct sockaddr_in *si4[2];
1272 #endif /* INET */
1273 #ifdef INET6
1274 	struct sockaddr_in6 *si6[2];
1275 #endif /* INET6 */
1276 	struct sockaddr_storage sa[2];
1277 	int error, pf, dodrop, s;
1278 
1279 	dodrop = name[-1] == TCPCTL_DROP;
1280 	if (dodrop) {
1281 		if (oldp != NULL || *oldlenp != 0)
1282 			return EINVAL;
1283 		if (newp == NULL)
1284 			return EPERM;
1285 		if (newlen < sizeof(sa))
1286 			return ENOMEM;
1287 	}
1288 	if (namelen != 4 && namelen != 0)
1289 		return EINVAL;
1290 	if (name[-2] != IPPROTO_TCP)
1291 		return EINVAL;
1292 	pf = name[-3];
1293 
1294 	/* old style lookup, ipv4 only */
1295 	if (namelen == 4) {
1296 #ifdef INET
1297 		struct in_addr laddr, raddr;
1298 		u_int lport, rport;
1299 
1300 		if (pf != PF_INET)
1301 			return EPROTONOSUPPORT;
1302 		raddr.s_addr = (uint32_t)name[0];
1303 		rport = (u_int)name[1];
1304 		laddr.s_addr = (uint32_t)name[2];
1305 		lport = (u_int)name[3];
1306 
1307 		s = splsoftnet();
1308 		error = inet4_ident_core(raddr, rport, laddr, lport,
1309 		    oldp, oldlenp, l, dodrop);
1310 		splx(s);
1311 		return error;
1312 #else /* INET */
1313 		return EINVAL;
1314 #endif /* INET */
1315 	}
1316 
1317 	if (newp == NULL || newlen != sizeof(sa))
1318 		return EINVAL;
1319 	error = copyin(newp, &sa, newlen);
1320 	if (error)
1321 		return error;
1322 
1323 	/*
1324 	 * requested families must match
1325 	 */
1326 	if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family)
1327 		return EINVAL;
1328 
1329 	switch (pf) {
1330 #ifdef INET6
1331 	case PF_INET6:
1332 		si6[0] = (struct sockaddr_in6*)&sa[0];
1333 		si6[1] = (struct sockaddr_in6*)&sa[1];
1334 		if (si6[0]->sin6_len != sizeof(*si6[0]) ||
1335 		    si6[1]->sin6_len != sizeof(*si6[1]))
1336 			return EINVAL;
1337 
1338 		if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) &&
1339 		    !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) {
1340 			error = sa6_embedscope(si6[0], ip6_use_defzone);
1341 			if (error)
1342 				return error;
1343 			error = sa6_embedscope(si6[1], ip6_use_defzone);
1344 			if (error)
1345 				return error;
1346 
1347 			s = splsoftnet();
1348 			error = inet6_ident_core(&si6[0]->sin6_addr,
1349 			    si6[0]->sin6_port, &si6[1]->sin6_addr,
1350 			    si6[1]->sin6_port, oldp, oldlenp, l, dodrop);
1351 			splx(s);
1352 			return error;
1353 		}
1354 
1355 		if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) !=
1356 		    IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr))
1357 			return EINVAL;
1358 
1359 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]);
1360 		in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]);
1361 		/*FALLTHROUGH*/
1362 #endif /* INET6 */
1363 #ifdef INET
1364 	case PF_INET:
1365 		si4[0] = (struct sockaddr_in*)&sa[0];
1366 		si4[1] = (struct sockaddr_in*)&sa[1];
1367 		if (si4[0]->sin_len != sizeof(*si4[0]) ||
1368 		    si4[0]->sin_len != sizeof(*si4[1]))
1369 			return EINVAL;
1370 
1371 		s = splsoftnet();
1372 		error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port,
1373 		    si4[1]->sin_addr, si4[1]->sin_port,
1374 		    oldp, oldlenp, l, dodrop);
1375 		splx(s);
1376 		return error;
1377 #endif /* INET */
1378 	default:
1379 		return EPROTONOSUPPORT;
1380 	}
1381 }
1382 
1383 /*
1384  * sysctl helper for the inet and inet6 pcblists.  handles tcp/udp and
1385  * inet/inet6, as well as raw pcbs for each.  specifically not
1386  * declared static so that raw sockets and udp/udp6 can use it as
1387  * well.
1388  */
1389 int
1390 sysctl_inpcblist(SYSCTLFN_ARGS)
1391 {
1392 #ifdef INET
1393 	struct sockaddr_in *in;
1394 	const struct inpcb *inp;
1395 #endif
1396 #ifdef INET6
1397 	struct sockaddr_in6 *in6;
1398 	const struct in6pcb *in6p;
1399 #endif
1400 	/*
1401 	 * sysctl_data is const, but CIRCLEQ_FOREACH can't use a const
1402 	 * struct inpcbtable pointer, so we have to discard const.  :-/
1403 	 */
1404 	struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data);
1405 	const struct inpcb_hdr *inph;
1406 	struct tcpcb *tp;
1407 	struct kinfo_pcb pcb;
1408 	char *dp;
1409 	u_int op, arg;
1410 	size_t len, needed, elem_size, out_size;
1411 	int error, elem_count, pf, proto, pf2;
1412 
1413 	if (namelen != 4)
1414 		return (EINVAL);
1415 
1416 	if (oldp != NULL) {
1417 		    len = *oldlenp;
1418 		    elem_size = name[2];
1419 		    elem_count = name[3];
1420 		    if (elem_size != sizeof(pcb))
1421 			    return EINVAL;
1422 	} else {
1423 		    len = 0;
1424 		    elem_count = INT_MAX;
1425 		    elem_size = sizeof(pcb);
1426 	}
1427 	error = 0;
1428 	dp = oldp;
1429 	op = name[0];
1430 	arg = name[1];
1431 	out_size = elem_size;
1432 	needed = 0;
1433 
1434 	if (namelen == 1 && name[0] == CTL_QUERY)
1435 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1436 
1437 	if (name - oname != 4)
1438 		return (EINVAL);
1439 
1440 	pf = oname[1];
1441 	proto = oname[2];
1442 	pf2 = (oldp != NULL) ? pf : 0;
1443 
1444 	CIRCLEQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) {
1445 #ifdef INET
1446 		inp = (const struct inpcb *)inph;
1447 #endif
1448 #ifdef INET6
1449 		in6p = (const struct in6pcb *)inph;
1450 #endif
1451 
1452 		if (inph->inph_af != pf)
1453 			continue;
1454 
1455 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
1456 		    KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL,
1457 		    NULL) != 0)
1458 			continue;
1459 
1460 		memset(&pcb, 0, sizeof(pcb));
1461 
1462 		pcb.ki_family = pf;
1463 		pcb.ki_type = proto;
1464 
1465 		switch (pf2) {
1466 		case 0:
1467 			/* just probing for size */
1468 			break;
1469 #ifdef INET
1470 		case PF_INET:
1471 			pcb.ki_family = inp->inp_socket->so_proto->
1472 			    pr_domain->dom_family;
1473 			pcb.ki_type = inp->inp_socket->so_proto->
1474 			    pr_type;
1475 			pcb.ki_protocol = inp->inp_socket->so_proto->
1476 			    pr_protocol;
1477 			pcb.ki_pflags = inp->inp_flags;
1478 
1479 			pcb.ki_sostate = inp->inp_socket->so_state;
1480 			pcb.ki_prstate = inp->inp_state;
1481 			if (proto == IPPROTO_TCP) {
1482 				tp = intotcpcb(inp);
1483 				pcb.ki_tstate = tp->t_state;
1484 				pcb.ki_tflags = tp->t_flags;
1485 			}
1486 
1487 			pcb.ki_pcbaddr = PTRTOUINT64(inp);
1488 			pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb);
1489 			pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket);
1490 
1491 			pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc;
1492 			pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc;
1493 
1494 			in = satosin(&pcb.ki_src);
1495 			in->sin_len = sizeof(*in);
1496 			in->sin_family = pf;
1497 			in->sin_port = inp->inp_lport;
1498 			in->sin_addr = inp->inp_laddr;
1499 			if (pcb.ki_prstate >= INP_CONNECTED) {
1500 				in = satosin(&pcb.ki_dst);
1501 				in->sin_len = sizeof(*in);
1502 				in->sin_family = pf;
1503 				in->sin_port = inp->inp_fport;
1504 				in->sin_addr = inp->inp_faddr;
1505 			}
1506 			break;
1507 #endif
1508 #ifdef INET6
1509 		case PF_INET6:
1510 			pcb.ki_family = in6p->in6p_socket->so_proto->
1511 			    pr_domain->dom_family;
1512 			pcb.ki_type = in6p->in6p_socket->so_proto->pr_type;
1513 			pcb.ki_protocol = in6p->in6p_socket->so_proto->
1514 			    pr_protocol;
1515 			pcb.ki_pflags = in6p->in6p_flags;
1516 
1517 			pcb.ki_sostate = in6p->in6p_socket->so_state;
1518 			pcb.ki_prstate = in6p->in6p_state;
1519 			if (proto == IPPROTO_TCP) {
1520 				tp = in6totcpcb(in6p);
1521 				pcb.ki_tstate = tp->t_state;
1522 				pcb.ki_tflags = tp->t_flags;
1523 			}
1524 
1525 			pcb.ki_pcbaddr = PTRTOUINT64(in6p);
1526 			pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb);
1527 			pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket);
1528 
1529 			pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc;
1530 			pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc;
1531 
1532 			in6 = satosin6(&pcb.ki_src);
1533 			in6->sin6_len = sizeof(*in6);
1534 			in6->sin6_family = pf;
1535 			in6->sin6_port = in6p->in6p_lport;
1536 			in6->sin6_flowinfo = in6p->in6p_flowinfo;
1537 			in6->sin6_addr = in6p->in6p_laddr;
1538 			in6->sin6_scope_id = 0; /* XXX? */
1539 
1540 			if (pcb.ki_prstate >= IN6P_CONNECTED) {
1541 				in6 = satosin6(&pcb.ki_dst);
1542 				in6->sin6_len = sizeof(*in6);
1543 				in6->sin6_family = pf;
1544 				in6->sin6_port = in6p->in6p_fport;
1545 				in6->sin6_flowinfo = in6p->in6p_flowinfo;
1546 				in6->sin6_addr = in6p->in6p_faddr;
1547 				in6->sin6_scope_id = 0; /* XXX? */
1548 			}
1549 			break;
1550 #endif
1551 		}
1552 
1553 		if (len >= elem_size && elem_count > 0) {
1554 			error = copyout(&pcb, dp, out_size);
1555 			if (error)
1556 				return (error);
1557 			dp += elem_size;
1558 			len -= elem_size;
1559 		}
1560 		if (elem_count > 0) {
1561 			needed += elem_size;
1562 			if (elem_count != INT_MAX)
1563 				elem_count--;
1564 		}
1565 	}
1566 
1567 	*oldlenp = needed;
1568 	if (oldp == NULL)
1569 		*oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb);
1570 
1571 	return (error);
1572 }
1573 
1574 static int
1575 sysctl_tcp_congctl(SYSCTLFN_ARGS)
1576 {
1577 	struct sysctlnode node;
1578 	int error, r;
1579 	char newname[TCPCC_MAXLEN];
1580 
1581 	strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1);
1582 
1583 	node = *rnode;
1584 	node.sysctl_data = newname;
1585 	node.sysctl_size = sizeof(newname);
1586 
1587 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1588 
1589 	if (error ||
1590 	    newp == NULL ||
1591 	    strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0)
1592 		return error;
1593 
1594 	if ((r = tcp_congctl_select(NULL, newname)))
1595 		return r;
1596 
1597 	return error;
1598 }
1599 
1600 static int
1601 sysctl_tcp_keep(SYSCTLFN_ARGS)
1602 {
1603 	int error;
1604 	u_int tmp;
1605 	struct sysctlnode node;
1606 
1607 	node = *rnode;
1608 	tmp = *(u_int *)rnode->sysctl_data;
1609 	node.sysctl_data = &tmp;
1610 
1611 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1612 	if (error || newp == NULL)
1613 		return error;
1614 
1615 	*(u_int *)rnode->sysctl_data = tmp;
1616 	tcp_tcpcb_template();	/* update the template */
1617 	return 0;
1618 }
1619 
1620 
1621 /*
1622  * this (second stage) setup routine is a replacement for tcp_sysctl()
1623  * (which is currently used for ipv4 and ipv6)
1624  */
1625 static void
1626 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
1627 			   const char *tcpname)
1628 {
1629 	const struct sysctlnode *sack_node;
1630 	const struct sysctlnode *abc_node;
1631 	const struct sysctlnode *ecn_node;
1632 	const struct sysctlnode *congctl_node;
1633 #ifdef TCP_DEBUG
1634 	extern struct tcp_debug tcp_debug[TCP_NDEBUG];
1635 	extern int tcp_debx;
1636 #endif
1637 
1638 	sysctl_createv(clog, 0, NULL, NULL,
1639 		       CTLFLAG_PERMANENT,
1640 		       CTLTYPE_NODE, "net", NULL,
1641 		       NULL, 0, NULL, 0,
1642 		       CTL_NET, CTL_EOL);
1643 	sysctl_createv(clog, 0, NULL, NULL,
1644 		       CTLFLAG_PERMANENT,
1645 		       CTLTYPE_NODE, pfname, NULL,
1646 		       NULL, 0, NULL, 0,
1647 		       CTL_NET, pf, CTL_EOL);
1648 	sysctl_createv(clog, 0, NULL, NULL,
1649 		       CTLFLAG_PERMANENT,
1650 		       CTLTYPE_NODE, tcpname,
1651 		       SYSCTL_DESCR("TCP related settings"),
1652 		       NULL, 0, NULL, 0,
1653 		       CTL_NET, pf, IPPROTO_TCP, CTL_EOL);
1654 
1655 	sysctl_createv(clog, 0, NULL, NULL,
1656 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1657 		       CTLTYPE_INT, "rfc1323",
1658 		       SYSCTL_DESCR("Enable RFC1323 TCP extensions"),
1659 		       NULL, 0, &tcp_do_rfc1323, 0,
1660 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL);
1661 	sysctl_createv(clog, 0, NULL, NULL,
1662 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1663 		       CTLTYPE_INT, "sendspace",
1664 		       SYSCTL_DESCR("Default TCP send buffer size"),
1665 		       NULL, 0, &tcp_sendspace, 0,
1666 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL);
1667 	sysctl_createv(clog, 0, NULL, NULL,
1668 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1669 		       CTLTYPE_INT, "recvspace",
1670 		       SYSCTL_DESCR("Default TCP receive buffer size"),
1671 		       NULL, 0, &tcp_recvspace, 0,
1672 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL);
1673 	sysctl_createv(clog, 0, NULL, NULL,
1674 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1675 		       CTLTYPE_INT, "mssdflt",
1676 		       SYSCTL_DESCR("Default maximum segment size"),
1677 		       sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0,
1678 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL);
1679 	sysctl_createv(clog, 0, NULL, NULL,
1680 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1681 		       CTLTYPE_INT, "minmss",
1682 		       SYSCTL_DESCR("Lower limit for TCP maximum segment size"),
1683 		       NULL, 0, &tcp_minmss, 0,
1684 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1685 	sysctl_createv(clog, 0, NULL, NULL,
1686 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1687 		       CTLTYPE_INT, "syn_cache_limit",
1688 		       SYSCTL_DESCR("Maximum number of entries in the TCP "
1689 				    "compressed state engine"),
1690 		       NULL, 0, &tcp_syn_cache_limit, 0,
1691 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT,
1692 		       CTL_EOL);
1693 	sysctl_createv(clog, 0, NULL, NULL,
1694 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1695 		       CTLTYPE_INT, "syn_bucket_limit",
1696 		       SYSCTL_DESCR("Maximum number of entries per hash "
1697 				    "bucket in the TCP compressed state "
1698 				    "engine"),
1699 		       NULL, 0, &tcp_syn_bucket_limit, 0,
1700 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT,
1701 		       CTL_EOL);
1702 #if 0 /* obsoleted */
1703 	sysctl_createv(clog, 0, NULL, NULL,
1704 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1705 		       CTLTYPE_INT, "syn_cache_interval",
1706 		       SYSCTL_DESCR("TCP compressed state engine's timer interval"),
1707 		       NULL, 0, &tcp_syn_cache_interval, 0,
1708 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER,
1709 		       CTL_EOL);
1710 #endif
1711 	sysctl_createv(clog, 0, NULL, NULL,
1712 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1713 		       CTLTYPE_INT, "init_win",
1714 		       SYSCTL_DESCR("Initial TCP congestion window"),
1715 		       NULL, 0, &tcp_init_win, 0,
1716 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL);
1717 	sysctl_createv(clog, 0, NULL, NULL,
1718 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1719 		       CTLTYPE_INT, "mss_ifmtu",
1720 		       SYSCTL_DESCR("Use interface MTU for calculating MSS"),
1721 		       NULL, 0, &tcp_mss_ifmtu, 0,
1722 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL);
1723 	sysctl_createv(clog, 0, NULL, &sack_node,
1724 		       CTLFLAG_PERMANENT,
1725 		       CTLTYPE_NODE, "sack",
1726 		       SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
1727 		       NULL, 0, NULL, 0,
1728 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
1729 
1730 	/* Congctl subtree */
1731 	sysctl_createv(clog, 0, NULL, &congctl_node,
1732 		       CTLFLAG_PERMANENT,
1733 		       CTLTYPE_NODE, "congctl",
1734 		       SYSCTL_DESCR("TCP Congestion Control"),
1735 	    	       NULL, 0, NULL, 0,
1736 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1737 	sysctl_createv(clog, 0, &congctl_node, NULL,
1738 		       CTLFLAG_PERMANENT,
1739 		       CTLTYPE_STRING, "available",
1740 		       SYSCTL_DESCR("Available Congestion Control Mechanisms"),
1741 		       NULL, 0, &tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL);
1742 	sysctl_createv(clog, 0, &congctl_node, NULL,
1743 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1744 		       CTLTYPE_STRING, "selected",
1745 		       SYSCTL_DESCR("Selected Congestion Control Mechanism"),
1746 		       sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN,
1747 		       CTL_CREATE, CTL_EOL);
1748 
1749 	sysctl_createv(clog, 0, NULL, NULL,
1750 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1751 		       CTLTYPE_INT, "win_scale",
1752 		       SYSCTL_DESCR("Use RFC1323 window scale options"),
1753 		       NULL, 0, &tcp_do_win_scale, 0,
1754 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL);
1755 	sysctl_createv(clog, 0, NULL, NULL,
1756 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1757 		       CTLTYPE_INT, "timestamps",
1758 		       SYSCTL_DESCR("Use RFC1323 time stamp options"),
1759 		       NULL, 0, &tcp_do_timestamps, 0,
1760 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL);
1761 	sysctl_createv(clog, 0, NULL, NULL,
1762 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1763 		       CTLTYPE_INT, "compat_42",
1764 		       SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"),
1765 		       NULL, 0, &tcp_compat_42, 0,
1766 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL);
1767 	sysctl_createv(clog, 0, NULL, NULL,
1768 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1769 		       CTLTYPE_INT, "cwm",
1770 		       SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window "
1771 				    "Monitoring"),
1772 		       NULL, 0, &tcp_cwm, 0,
1773 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL);
1774 	sysctl_createv(clog, 0, NULL, NULL,
1775 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1776 		       CTLTYPE_INT, "cwm_burstsize",
1777 		       SYSCTL_DESCR("Congestion Window Monitoring allowed "
1778 				    "burst count in packets"),
1779 		       NULL, 0, &tcp_cwm_burstsize, 0,
1780 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE,
1781 		       CTL_EOL);
1782 	sysctl_createv(clog, 0, NULL, NULL,
1783 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1784 		       CTLTYPE_INT, "ack_on_push",
1785 		       SYSCTL_DESCR("Immediately return ACK when PSH is "
1786 				    "received"),
1787 		       NULL, 0, &tcp_ack_on_push, 0,
1788 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL);
1789 	sysctl_createv(clog, 0, NULL, NULL,
1790 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1791 		       CTLTYPE_INT, "keepidle",
1792 		       SYSCTL_DESCR("Allowed connection idle ticks before a "
1793 				    "keepalive probe is sent"),
1794 		       sysctl_tcp_keep, 0, &tcp_keepidle, 0,
1795 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL);
1796 	sysctl_createv(clog, 0, NULL, NULL,
1797 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1798 		       CTLTYPE_INT, "keepintvl",
1799 		       SYSCTL_DESCR("Ticks before next keepalive probe is sent"),
1800 		       sysctl_tcp_keep, 0, &tcp_keepintvl, 0,
1801 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL);
1802 	sysctl_createv(clog, 0, NULL, NULL,
1803 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1804 		       CTLTYPE_INT, "keepcnt",
1805 		       SYSCTL_DESCR("Number of keepalive probes to send"),
1806 		       sysctl_tcp_keep, 0, &tcp_keepcnt, 0,
1807 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL);
1808 	sysctl_createv(clog, 0, NULL, NULL,
1809 		       CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE,
1810 		       CTLTYPE_INT, "slowhz",
1811 		       SYSCTL_DESCR("Keepalive ticks per second"),
1812 		       NULL, PR_SLOWHZ, NULL, 0,
1813 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL);
1814 	sysctl_createv(clog, 0, NULL, NULL,
1815 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1816 		       CTLTYPE_INT, "log_refused",
1817 		       SYSCTL_DESCR("Log refused TCP connections"),
1818 		       NULL, 0, &tcp_log_refused, 0,
1819 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL);
1820 #if 0 /* obsoleted */
1821 	sysctl_createv(clog, 0, NULL, NULL,
1822 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1823 		       CTLTYPE_INT, "rstratelimit", NULL,
1824 		       NULL, 0, &tcp_rst_ratelim, 0,
1825 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL);
1826 #endif
1827 	sysctl_createv(clog, 0, NULL, NULL,
1828 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1829 		       CTLTYPE_INT, "rstppslimit",
1830 		       SYSCTL_DESCR("Maximum number of RST packets to send "
1831 				    "per second"),
1832 		       NULL, 0, &tcp_rst_ppslim, 0,
1833 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL);
1834 	sysctl_createv(clog, 0, NULL, NULL,
1835 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1836 		       CTLTYPE_INT, "delack_ticks",
1837 		       SYSCTL_DESCR("Number of ticks to delay sending an ACK"),
1838 		       NULL, 0, &tcp_delack_ticks, 0,
1839 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL);
1840 	sysctl_createv(clog, 0, NULL, NULL,
1841 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1842 		       CTLTYPE_INT, "init_win_local",
1843 		       SYSCTL_DESCR("Initial TCP window size (in segments)"),
1844 		       NULL, 0, &tcp_init_win_local, 0,
1845 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL,
1846 		       CTL_EOL);
1847 	sysctl_createv(clog, 0, NULL, NULL,
1848 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1849 		       CTLTYPE_STRUCT, "ident",
1850 		       SYSCTL_DESCR("RFC1413 Identification Protocol lookups"),
1851 		       sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t),
1852 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL);
1853 	sysctl_createv(clog, 0, NULL, NULL,
1854 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1855 		       CTLTYPE_INT, "do_loopback_cksum",
1856 		       SYSCTL_DESCR("Perform TCP checksum on loopback"),
1857 		       NULL, 0, &tcp_do_loopback_cksum, 0,
1858 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM,
1859 		       CTL_EOL);
1860 	sysctl_createv(clog, 0, NULL, NULL,
1861 		       CTLFLAG_PERMANENT,
1862 		       CTLTYPE_STRUCT, "pcblist",
1863 		       SYSCTL_DESCR("TCP protocol control block list"),
1864 		       sysctl_inpcblist, 0, &tcbtable, 0,
1865 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
1866 		       CTL_EOL);
1867 	sysctl_createv(clog, 0, NULL, NULL,
1868 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1869 		       CTLTYPE_INT, "keepinit",
1870 		       SYSCTL_DESCR("Ticks before initial tcp connection times out"),
1871 		       sysctl_tcp_keep, 0, &tcp_keepinit, 0,
1872 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1873 
1874 	/* TCP socket buffers auto-sizing nodes */
1875 	sysctl_createv(clog, 0, NULL, NULL,
1876 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1877 		       CTLTYPE_INT, "recvbuf_auto",
1878 		       SYSCTL_DESCR("Enable automatic receive "
1879 		           "buffer sizing (experimental)"),
1880 		       NULL, 0, &tcp_do_autorcvbuf, 0,
1881 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1882 	sysctl_createv(clog, 0, NULL, NULL,
1883 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1884 		       CTLTYPE_INT, "recvbuf_inc",
1885 		       SYSCTL_DESCR("Incrementor step size of "
1886 		           "automatic receive buffer"),
1887 		       NULL, 0, &tcp_autorcvbuf_inc, 0,
1888 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1889 	sysctl_createv(clog, 0, NULL, NULL,
1890 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1891 		       CTLTYPE_INT, "recvbuf_max",
1892 		       SYSCTL_DESCR("Max size of automatic receive buffer"),
1893 		       NULL, 0, &tcp_autorcvbuf_max, 0,
1894 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1895 
1896 	sysctl_createv(clog, 0, NULL, NULL,
1897 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1898 		       CTLTYPE_INT, "sendbuf_auto",
1899 		       SYSCTL_DESCR("Enable automatic send "
1900 		           "buffer sizing (experimental)"),
1901 		       NULL, 0, &tcp_do_autosndbuf, 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, "sendbuf_inc",
1906 		       SYSCTL_DESCR("Incrementor step size of "
1907 		           "automatic send buffer"),
1908 		       NULL, 0, &tcp_autosndbuf_inc, 0,
1909 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1910 	sysctl_createv(clog, 0, NULL, NULL,
1911 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1912 		       CTLTYPE_INT, "sendbuf_max",
1913 		       SYSCTL_DESCR("Max size of automatic send buffer"),
1914 		       NULL, 0, &tcp_autosndbuf_max, 0,
1915 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1916 
1917 	/* ECN subtree */
1918 	sysctl_createv(clog, 0, NULL, &ecn_node,
1919 	    	       CTLFLAG_PERMANENT,
1920 		       CTLTYPE_NODE, "ecn",
1921 	    	       SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
1922 	    	       NULL, 0, NULL, 0,
1923 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
1924 	sysctl_createv(clog, 0, &ecn_node, NULL,
1925 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1926 		       CTLTYPE_INT, "enable",
1927 		       SYSCTL_DESCR("Enable TCP Explicit Congestion "
1928 			   "Notification"),
1929 	    	       NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL);
1930 	sysctl_createv(clog, 0, &ecn_node, NULL,
1931 	    	       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1932 		       CTLTYPE_INT, "maxretries",
1933 		       SYSCTL_DESCR("Number of times to retry ECN setup "
1934 			       "before disabling ECN on the connection"),
1935 	    	       NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL);
1936 
1937 	/* SACK gets it's own little subtree. */
1938 	sysctl_createv(clog, 0, NULL, &sack_node,
1939 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1940 		       CTLTYPE_INT, "enable",
1941 		       SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"),
1942 		       NULL, 0, &tcp_do_sack, 0,
1943 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1944 	sysctl_createv(clog, 0, NULL, &sack_node,
1945 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1946 		       CTLTYPE_INT, "maxholes",
1947 		       SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"),
1948 		       NULL, 0, &tcp_sack_tp_maxholes, 0,
1949 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1950 	sysctl_createv(clog, 0, NULL, &sack_node,
1951 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1952 		       CTLTYPE_INT, "globalmaxholes",
1953 		       SYSCTL_DESCR("Global maximum number of TCP SACK holes"),
1954 		       NULL, 0, &tcp_sack_globalmaxholes, 0,
1955 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1956 	sysctl_createv(clog, 0, NULL, &sack_node,
1957 		       CTLFLAG_PERMANENT,
1958 		       CTLTYPE_INT, "globalholes",
1959 		       SYSCTL_DESCR("Global number of TCP SACK holes"),
1960 		       NULL, 0, &tcp_sack_globalholes, 0,
1961 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL);
1962 
1963 	sysctl_createv(clog, 0, NULL, NULL,
1964 		       CTLFLAG_PERMANENT,
1965 		       CTLTYPE_STRUCT, "stats",
1966 		       SYSCTL_DESCR("TCP statistics"),
1967 		       NULL, 0, &tcpstat, sizeof(tcpstat),
1968 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS,
1969 		       CTL_EOL);
1970 #ifdef TCP_DEBUG
1971 	sysctl_createv(clog, 0, NULL, NULL,
1972 		       CTLFLAG_PERMANENT,
1973 		       CTLTYPE_STRUCT, "debug",
1974 		       SYSCTL_DESCR("TCP sockets debug information"),
1975 		       NULL, 0, &tcp_debug, sizeof(tcp_debug),
1976 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG,
1977 		       CTL_EOL);
1978 	sysctl_createv(clog, 0, NULL, NULL,
1979 		       CTLFLAG_PERMANENT,
1980 		       CTLTYPE_INT, "debx",
1981 		       SYSCTL_DESCR("Number of TCP debug sockets messages"),
1982 		       NULL, 0, &tcp_debx, sizeof(tcp_debx),
1983 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX,
1984 		       CTL_EOL);
1985 #endif
1986 	sysctl_createv(clog, 0, NULL, NULL,
1987 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1988 		       CTLTYPE_STRUCT, "drop",
1989 		       SYSCTL_DESCR("TCP drop connection"),
1990 		       sysctl_net_inet_tcp_drop, 0, NULL, 0,
1991 		       CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL);
1992 #if NRND > 0
1993 	sysctl_createv(clog, 0, NULL, NULL,
1994 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1995 		       CTLTYPE_INT, "iss_hash",
1996 		       SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic "
1997 				    "hash computation"),
1998 		       NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948),
1999 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
2000 		       CTL_EOL);
2001 #endif
2002 
2003 	/* ABC subtree */
2004 
2005 	sysctl_createv(clog, 0, NULL, &abc_node,
2006 		       CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc",
2007 		       SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"),
2008 		       NULL, 0, NULL, 0,
2009 		       CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL);
2010 	sysctl_createv(clog, 0, &abc_node, NULL,
2011 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2012 		       CTLTYPE_INT, "enable",
2013 		       SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"),
2014 		       NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL);
2015 	sysctl_createv(clog, 0, &abc_node, NULL,
2016 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2017 		       CTLTYPE_INT, "aggressive",
2018 		       SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"),
2019 		       NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL);
2020 }
2021 
2022 /*
2023  * Sysctl for tcp variables.
2024  */
2025 #ifdef INET
2026 SYSCTL_SETUP(sysctl_net_inet_tcp_setup, "sysctl net.inet.tcp subtree setup")
2027 {
2028 
2029 	sysctl_net_inet_tcp_setup2(clog, PF_INET, "inet", "tcp");
2030 }
2031 #endif /* INET */
2032 
2033 #ifdef INET6
2034 SYSCTL_SETUP(sysctl_net_inet6_tcp6_setup, "sysctl net.inet6.tcp6 subtree setup")
2035 {
2036 
2037 	sysctl_net_inet_tcp_setup2(clog, PF_INET6, "inet6", "tcp6");
2038 }
2039 #endif /* INET6 */
2040