xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 95d875fb90b1458e4f1de6950286ddcd6644bc61)
1 /*	$NetBSD: tcp_usrreq.c,v 1.43 1999/12/13 15:17:21 itojun 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 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  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1988, 1993, 1995
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. All advertising materials mentioning features or use of this software
82  *    must display the following acknowledgement:
83  *	This product includes software developed by the University of
84  *	California, Berkeley and its contributors.
85  * 4. 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 "opt_inet.h"
105 #include "opt_ipsec.h"
106 
107 #include <sys/param.h>
108 #include <sys/systm.h>
109 #include <sys/kernel.h>
110 #include <sys/malloc.h>
111 #include <sys/mbuf.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #include <sys/protosw.h>
115 #include <sys/errno.h>
116 #include <sys/stat.h>
117 #include <sys/proc.h>
118 #include <sys/ucred.h>
119 #include <sys/domain.h>
120 
121 #include <vm/vm.h>
122 #include <sys/sysctl.h>
123 
124 #include <net/if.h>
125 #include <net/route.h>
126 
127 #include <netinet/in.h>
128 #include <netinet/in_systm.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip.h>
131 #include <netinet/in_pcb.h>
132 #include <netinet/ip_var.h>
133 
134 #ifdef INET6
135 #ifndef INET
136 #include <netinet/in.h>
137 #endif
138 #include <netinet/ip6.h>
139 #include <netinet6/in6_pcb.h>
140 #include <netinet6/ip6_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/tcpip.h>
149 #include <netinet/tcp_debug.h>
150 
151 #include "opt_tcp_recvspace.h"
152 #include "opt_tcp_sendspace.h"
153 
154 #ifdef IPSEC
155 #include <netinet6/ipsec.h>
156 #endif /*IPSEC*/
157 
158 /*
159  * TCP protocol interface to socket abstraction.
160  */
161 extern	char *tcpstates[];
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(so, req, m, nam, control, p)
171 	struct socket *so;
172 	int req;
173 	struct mbuf *m, *nam, *control;
174 	struct proc *p;
175 {
176 	register struct inpcb *inp;
177 #ifdef INET6
178 	register struct in6pcb *in6p;
179 #endif
180 	register struct tcpcb *tp = NULL;
181 	int s;
182 	int error = 0;
183 	int ostate;
184 	int family;	/* family of the socket */
185 
186 	family = so->so_proto->pr_domain->dom_family;
187 
188 	if (req == PRU_CONTROL) {
189 		switch (family) {
190 		case PF_INET:
191 			return (in_control(so, (long)m, (caddr_t)nam,
192 			    (struct ifnet *)control, p));
193 #ifdef INET6
194 		case PF_INET6:
195 			return (in6_control(so, (long)m, (caddr_t)nam,
196 			    (struct ifnet *)control, p));
197 #endif
198 		default:
199 			return EAFNOSUPPORT;
200 		}
201 	}
202 
203 	s = splsoftnet();
204 	switch (family) {
205 	case PF_INET:
206 		inp = sotoinpcb(so);
207 #ifdef INET6
208 		in6p = NULL;
209 #endif
210 		break;
211 #ifdef INET6
212 	case PF_INET6:
213 		inp = NULL;
214 		in6p = sotoin6pcb(so);
215 		break;
216 #endif
217 	default:
218 		splx(s);
219 		return EAFNOSUPPORT;
220 	}
221 
222 #ifdef DIAGNOSTIC
223 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
224 		panic("tcp_usrreq: unexpected control mbuf");
225 #endif
226 	/*
227 	 * When a TCP is attached to a socket, then there will be
228 	 * a (struct inpcb) pointed at by the socket, and this
229 	 * structure will point at a subsidary (struct tcpcb).
230 	 */
231 #ifndef INET6
232 	if (inp == 0 && req != PRU_ATTACH)
233 #else
234 	if ((inp == 0 && in6p == 0) && req != PRU_ATTACH)
235 #endif
236 	{
237 		error = EINVAL;
238 		goto release;
239 	}
240 	if (inp) {
241 		tp = intotcpcb(inp);
242 		/* WHAT IF TP IS 0? */
243 #ifdef KPROF
244 		tcp_acounts[tp->t_state][req]++;
245 #endif
246 		ostate = tp->t_state;
247 	}
248 #ifdef INET6
249 	else if (in6p) {
250 		tp = in6totcpcb(in6p);
251 		/* WHAT IF TP IS 0? */
252 #ifdef KPROF
253 		tcp_acounts[tp->t_state][req]++;
254 #endif
255 		ostate = tp->t_state;
256 	}
257 #endif
258 	else
259 		ostate = 0;
260 
261 	switch (req) {
262 
263 	/*
264 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
265 	 * and an internet control block.
266 	 */
267 	case PRU_ATTACH:
268 #ifndef INET6
269 		if (inp != 0)
270 #else
271 		if (inp != 0 || in6p != 0)
272 #endif
273 		{
274 			error = EISCONN;
275 			break;
276 		}
277 		error = tcp_attach(so);
278 		if (error)
279 			break;
280 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
281 			so->so_linger = TCP_LINGERTIME;
282 		tp = sototcpcb(so);
283 		break;
284 
285 	/*
286 	 * PRU_DETACH detaches the TCP protocol from the socket.
287 	 */
288 	case PRU_DETACH:
289 		tp = tcp_disconnect(tp);
290 		break;
291 
292 	/*
293 	 * Give the socket an address.
294 	 */
295 	case PRU_BIND:
296 		switch (family) {
297 		case PF_INET:
298 			error = in_pcbbind(inp, nam, p);
299 			break;
300 #ifdef INET6
301 		case PF_INET6:
302 			error = in6_pcbbind(in6p, nam /*, p*/ );
303 			/* mapped addr case */
304 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
305 				tp->t_family = AF_INET;
306 			break;
307 #endif
308 		}
309 		break;
310 
311 	/*
312 	 * Prepare to accept connections.
313 	 */
314 	case PRU_LISTEN:
315 		if (inp && inp->inp_lport == 0) {
316 			error = in_pcbbind(inp, (struct mbuf *)0,
317 			    (struct proc *)0);
318 			if (error)
319 				break;
320 		}
321 #ifdef INET6
322 		else if (in6p && in6p->in6p_lport == 0) {
323 			error = in6_pcbbind(in6p, (struct mbuf *)0 /*,
324 			    (struct proc *)0 */ );
325 			if (error)
326 				break;
327 		}
328 #endif
329 		tp->t_state = TCPS_LISTEN;
330 		break;
331 
332 	/*
333 	 * Initiate connection to peer.
334 	 * Create a template for use in transmissions on this connection.
335 	 * Enter SYN_SENT state, and mark socket as connecting.
336 	 * Start keep-alive timer, and seed output sequence space.
337 	 * Send initial segment on connection.
338 	 */
339 	case PRU_CONNECT:
340 		if (inp) {
341 			if (inp->inp_lport == 0) {
342 				error = in_pcbbind(inp, (struct mbuf *)0,
343 				    (struct proc *)0);
344 				if (error)
345 					break;
346 			}
347 			error = in_pcbconnect(inp, nam);
348 		}
349 #ifdef INET6
350 		else if (in6p) {
351 			if (in6p->in6p_lport == 0) {
352 				error = in6_pcbbind(in6p, (struct mbuf *)0 /*,
353 				    (struct proc *)0 */ );
354 				if (error)
355 					break;
356 			}
357 			error = in6_pcbconnect(in6p, nam);
358 			/* mapped addr case */
359 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
360 				tp->t_family = AF_INET;
361 		}
362 #endif
363 		if (error)
364 			break;
365 		tp->t_template = tcp_template(tp);
366 		if (tp->t_template == 0) {
367 			if (inp)
368 				in_pcbdisconnect(inp);
369 #ifdef INET6
370 			else if (in6p)
371 				in6_pcbdisconnect(in6p);
372 #endif
373 			error = ENOBUFS;
374 			break;
375 		}
376 		/* Compute window scaling to request.  */
377 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
378 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
379 			tp->request_r_scale++;
380 		soisconnecting(so);
381 		tcpstat.tcps_connattempt++;
382 		tp->t_state = TCPS_SYN_SENT;
383 		TCP_TIMER_ARM(tp, TCPT_KEEP, TCPTV_KEEP_INIT);
384 		tp->iss = tcp_new_iss(tp, sizeof(struct tcpcb), 0);
385 		tcp_sendseqinit(tp);
386 		error = tcp_output(tp);
387 		break;
388 
389 	/*
390 	 * Create a TCP connection between two sockets.
391 	 */
392 	case PRU_CONNECT2:
393 		error = EOPNOTSUPP;
394 		break;
395 
396 	/*
397 	 * Initiate disconnect from peer.
398 	 * If connection never passed embryonic stage, just drop;
399 	 * else if don't need to let data drain, then can just drop anyways,
400 	 * else have to begin TCP shutdown process: mark socket disconnecting,
401 	 * drain unread data, state switch to reflect user close, and
402 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
403 	 * when peer sends FIN and acks ours.
404 	 *
405 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
406 	 */
407 	case PRU_DISCONNECT:
408 		tp = tcp_disconnect(tp);
409 		break;
410 
411 	/*
412 	 * Accept a connection.  Essentially all the work is
413 	 * done at higher levels; just return the address
414 	 * of the peer, storing through addr.
415 	 */
416 	case PRU_ACCEPT:
417 		if (inp)
418 			in_setpeeraddr(inp, nam);
419 #ifdef INET6
420 		else if (in6p)
421 			in6_setpeeraddr(in6p, nam);
422 #endif
423 		break;
424 
425 	/*
426 	 * Mark the connection as being incapable of further output.
427 	 */
428 	case PRU_SHUTDOWN:
429 		socantsendmore(so);
430 		tp = tcp_usrclosed(tp);
431 		if (tp)
432 			error = tcp_output(tp);
433 		break;
434 
435 	/*
436 	 * After a receive, possibly send window update to peer.
437 	 */
438 	case PRU_RCVD:
439 		(void) tcp_output(tp);
440 		break;
441 
442 	/*
443 	 * Do a send by putting data in output queue and updating urgent
444 	 * marker if URG set.  Possibly send more data.
445 	 */
446 	case PRU_SEND:
447 		if (control && control->m_len) {
448 			m_freem(control);
449 			m_freem(m);
450 			error = EINVAL;
451 			break;
452 		}
453 		sbappend(&so->so_snd, m);
454 		error = tcp_output(tp);
455 		break;
456 
457 	/*
458 	 * Abort the TCP.
459 	 */
460 	case PRU_ABORT:
461 		tp = tcp_drop(tp, ECONNABORTED);
462 		break;
463 
464 	case PRU_SENSE:
465 		/*
466 		 * stat: don't bother with a blocksize.
467 		 */
468 		splx(s);
469 		return (0);
470 
471 	case PRU_RCVOOB:
472 		if (control && control->m_len) {
473 			m_freem(control);
474 			m_freem(m);
475 			error = EINVAL;
476 			break;
477 		}
478 		if ((so->so_oobmark == 0 &&
479 		    (so->so_state & SS_RCVATMARK) == 0) ||
480 		    so->so_options & SO_OOBINLINE ||
481 		    tp->t_oobflags & TCPOOB_HADDATA) {
482 			error = EINVAL;
483 			break;
484 		}
485 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
486 			error = EWOULDBLOCK;
487 			break;
488 		}
489 		m->m_len = 1;
490 		*mtod(m, caddr_t) = tp->t_iobc;
491 		if (((long)nam & MSG_PEEK) == 0)
492 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
493 		break;
494 
495 	case PRU_SENDOOB:
496 		if (sbspace(&so->so_snd) < -512) {
497 			m_freem(m);
498 			error = ENOBUFS;
499 			break;
500 		}
501 		/*
502 		 * According to RFC961 (Assigned Protocols),
503 		 * the urgent pointer points to the last octet
504 		 * of urgent data.  We continue, however,
505 		 * to consider it to indicate the first octet
506 		 * of data past the urgent section.
507 		 * Otherwise, snd_up should be one lower.
508 		 */
509 		sbappend(&so->so_snd, m);
510 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
511 		tp->t_force = 1;
512 		error = tcp_output(tp);
513 		tp->t_force = 0;
514 		break;
515 
516 	case PRU_SOCKADDR:
517 		if (inp)
518 			in_setsockaddr(inp, nam);
519 #ifdef INET6
520 		else if (in6p)
521 			in6_setsockaddr(in6p, nam);
522 #endif
523 		break;
524 
525 	case PRU_PEERADDR:
526 		if (inp)
527 			in_setpeeraddr(inp, nam);
528 #ifdef INET6
529 		else if (in6p)
530 			in6_setpeeraddr(in6p, nam);
531 #endif
532 		break;
533 
534 	/*
535 	 * TCP slow timer went off; going through this
536 	 * routine for tracing's sake.
537 	 */
538 	case PRU_SLOWTIMO:
539 		tp = tcp_timers(tp, (long)nam);
540 		req |= (long)nam << 8;		/* for debug's sake */
541 		break;
542 
543 	default:
544 		panic("tcp_usrreq");
545 	}
546 	if (tp && (so->so_options & SO_DEBUG))
547 		tcp_trace(TA_USER, ostate, tp, NULL, req);
548 
549 release:
550 	splx(s);
551 	return (error);
552 }
553 
554 int
555 tcp_ctloutput(op, so, level, optname, mp)
556 	int op;
557 	struct socket *so;
558 	int level, optname;
559 	struct mbuf **mp;
560 {
561 	int error = 0, s;
562 	struct inpcb *inp;
563 #ifdef INET6
564 	register struct in6pcb *in6p;
565 #endif
566 	register struct tcpcb *tp;
567 	register struct mbuf *m;
568 	register int i;
569 	int family;	/* family of the socket */
570 
571 	family = so->so_proto->pr_domain->dom_family;
572 
573 	s = splsoftnet();
574 	switch (family) {
575 	case PF_INET:
576 		inp = sotoinpcb(so);
577 #ifdef INET6
578 		in6p = NULL;
579 #endif
580 		break;
581 #ifdef INET6
582 	case PF_INET6:
583 		inp = NULL;
584 		in6p = sotoin6pcb(so);
585 		break;
586 #endif
587 	default:
588 		splx(s);
589 		return EAFNOSUPPORT;
590 	}
591 #ifndef INET6
592 	if (inp == NULL)
593 #else
594 	if (inp == NULL && in6p == NULL)
595 #endif
596 	{
597 		splx(s);
598 		if (op == PRCO_SETOPT && *mp)
599 			(void) m_free(*mp);
600 		return (ECONNRESET);
601 	}
602 	if (level != IPPROTO_TCP) {
603 		switch (family) {
604 		case PF_INET:
605 			error = ip_ctloutput(op, so, level, optname, mp);
606 			break;
607 #ifdef INET6
608 		case PF_INET6:
609 			error = ip6_ctloutput(op, so, level, optname, mp);
610 			break;
611 #endif
612 		}
613 		splx(s);
614 		return (error);
615 	}
616 	if (inp)
617 		tp = intotcpcb(inp);
618 #ifdef INET6
619 	else if (in6p)
620 		tp = in6totcpcb(in6p);
621 #endif
622 	else
623 		tp = NULL;
624 
625 	switch (op) {
626 
627 	case PRCO_SETOPT:
628 		m = *mp;
629 		switch (optname) {
630 
631 		case TCP_NODELAY:
632 			if (m == NULL || m->m_len < sizeof (int))
633 				error = EINVAL;
634 			else if (*mtod(m, int *))
635 				tp->t_flags |= TF_NODELAY;
636 			else
637 				tp->t_flags &= ~TF_NODELAY;
638 			break;
639 
640 		case TCP_MAXSEG:
641 			if (m && (i = *mtod(m, int *)) > 0 &&
642 			    i <= tp->t_peermss)
643 				tp->t_peermss = i;  /* limit on send size */
644 			else
645 				error = EINVAL;
646 			break;
647 
648 		default:
649 			error = ENOPROTOOPT;
650 			break;
651 		}
652 		if (m)
653 			(void) m_free(m);
654 		break;
655 
656 	case PRCO_GETOPT:
657 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
658 		m->m_len = sizeof(int);
659 
660 		switch (optname) {
661 		case TCP_NODELAY:
662 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
663 			break;
664 		case TCP_MAXSEG:
665 			*mtod(m, int *) = tp->t_peermss;
666 			break;
667 		default:
668 			error = ENOPROTOOPT;
669 			break;
670 		}
671 		break;
672 	}
673 	splx(s);
674 	return (error);
675 }
676 
677 #ifndef TCP_SENDSPACE
678 #define	TCP_SENDSPACE	1024*16;
679 #endif
680 int	tcp_sendspace = TCP_SENDSPACE;
681 #ifndef TCP_RECVSPACE
682 #define	TCP_RECVSPACE	1024*16;
683 #endif
684 int	tcp_recvspace = TCP_RECVSPACE;
685 
686 /*
687  * Attach TCP protocol to socket, allocating
688  * internet protocol control block, tcp control block,
689  * bufer space, and entering LISTEN state if to accept connections.
690  */
691 int
692 tcp_attach(so)
693 	struct socket *so;
694 {
695 	register struct tcpcb *tp;
696 	struct inpcb *inp;
697 #ifdef INET6
698 	struct in6pcb *in6p;
699 #endif
700 	int error;
701 	int family;	/* family of the socket */
702 
703 	family = so->so_proto->pr_domain->dom_family;
704 
705 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
706 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
707 		if (error)
708 			return (error);
709 	}
710 	switch (family) {
711 	case PF_INET:
712 		error = in_pcballoc(so, &tcbtable);
713 		if (error)
714 			return (error);
715 		inp = sotoinpcb(so);
716 #ifdef INET6
717 		in6p = NULL;
718 #endif
719 		break;
720 #ifdef INET6
721 	case PF_INET6:
722 		error = in6_pcballoc(so, &tcb6);
723 		if (error)
724 			return (error);
725 		inp = NULL;
726 		in6p = sotoin6pcb(so);
727 		break;
728 #endif
729 	default:
730 		return EAFNOSUPPORT;
731 	}
732 #ifdef IPSEC
733 	if (inp) {
734 		error = ipsec_init_policy(&inp->inp_sp);
735 		if (error != 0) {
736 			in_pcbdetach(inp);
737 			return (error);
738 		}
739 	}
740 #ifdef INET6
741 	else if (in6p) {
742 		error = ipsec_init_policy(&in6p->in6p_sp);
743 		if (error != 0) {
744 			in6_pcbdetach(in6p);
745 			return (error);
746 		}
747 	}
748 #endif
749 #endif /*IPSEC*/
750 	if (inp)
751 		tp = tcp_newtcpcb(family, (void *)inp);
752 #ifdef INET6
753 	else if (in6p)
754 		tp = tcp_newtcpcb(family, (void *)in6p);
755 #endif
756 	else
757 		tp = NULL;
758 
759 	if (tp == 0) {
760 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
761 
762 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
763 		if (inp)
764 			in_pcbdetach(inp);
765 #ifdef INET6
766 		else if (in6p)
767 			in6_pcbdetach(in6p);
768 #endif
769 		so->so_state |= nofd;
770 		return (ENOBUFS);
771 	}
772 	tp->t_state = TCPS_CLOSED;
773 	return (0);
774 }
775 
776 /*
777  * Initiate (or continue) disconnect.
778  * If embryonic state, just send reset (once).
779  * If in ``let data drain'' option and linger null, just drop.
780  * Otherwise (hard), mark socket disconnecting and drop
781  * current input data; switch states based on user close, and
782  * send segment to peer (with FIN).
783  */
784 struct tcpcb *
785 tcp_disconnect(tp)
786 	register struct tcpcb *tp;
787 {
788 	struct socket *so;
789 
790 	if (tp->t_inpcb)
791 		so = tp->t_inpcb->inp_socket;
792 #ifdef INET6
793 	else if (tp->t_in6pcb)
794 		so = tp->t_in6pcb->in6p_socket;
795 #endif
796 	else
797 		so = NULL;
798 
799 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
800 		tp = tcp_close(tp);
801 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
802 		tp = tcp_drop(tp, 0);
803 	else {
804 		soisdisconnecting(so);
805 		sbflush(&so->so_rcv);
806 		tp = tcp_usrclosed(tp);
807 		if (tp)
808 			(void) tcp_output(tp);
809 	}
810 	return (tp);
811 }
812 
813 /*
814  * User issued close, and wish to trail through shutdown states:
815  * if never received SYN, just forget it.  If got a SYN from peer,
816  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
817  * If already got a FIN from peer, then almost done; go to LAST_ACK
818  * state.  In all other cases, have already sent FIN to peer (e.g.
819  * after PRU_SHUTDOWN), and just have to play tedious game waiting
820  * for peer to send FIN or not respond to keep-alives, etc.
821  * We can let the user exit from the close as soon as the FIN is acked.
822  */
823 struct tcpcb *
824 tcp_usrclosed(tp)
825 	register struct tcpcb *tp;
826 {
827 
828 	switch (tp->t_state) {
829 
830 	case TCPS_CLOSED:
831 	case TCPS_LISTEN:
832 	case TCPS_SYN_SENT:
833 		tp->t_state = TCPS_CLOSED;
834 		tp = tcp_close(tp);
835 		break;
836 
837 	case TCPS_SYN_RECEIVED:
838 	case TCPS_ESTABLISHED:
839 		tp->t_state = TCPS_FIN_WAIT_1;
840 		break;
841 
842 	case TCPS_CLOSE_WAIT:
843 		tp->t_state = TCPS_LAST_ACK;
844 		break;
845 	}
846 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
847 		struct socket *so;
848 		if (tp->t_inpcb)
849 			so = tp->t_inpcb->inp_socket;
850 #ifdef INET6
851 		else if (tp->t_in6pcb)
852 			so = tp->t_in6pcb->in6p_socket;
853 #endif
854 		else
855 			so = NULL;
856 		soisdisconnected(so);
857 		/*
858 		 * If we are in FIN_WAIT_2, we arrived here because the
859 		 * application did a shutdown of the send side.  Like the
860 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
861 		 * a full close, we start a timer to make sure sockets are
862 		 * not left in FIN_WAIT_2 forever.
863 		 */
864 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
865 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
866 	}
867 	return (tp);
868 }
869 
870 static struct {
871 	 unsigned int valid : 1;
872 	 unsigned int rdonly : 1;
873 	 int *var;
874 	 int val;
875 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
876 
877 /*
878  * Sysctl for tcp variables.
879  */
880 int
881 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
882 	int *name;
883 	u_int namelen;
884 	void *oldp;
885 	size_t *oldlenp;
886 	void *newp;
887 	size_t newlen;
888 {
889 	/* All sysctl names at this level are terminal. */
890 	if (namelen != 1)
891 		return (ENOTDIR);
892 
893 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
894 	    && tcp_ctlvars[name[0]].valid) {
895 		if (tcp_ctlvars[name[0]].rdonly)
896 			return (sysctl_rdint(oldp, oldlenp, newp,
897 			    tcp_ctlvars[name[0]].val));
898 		else
899 			return (sysctl_int(oldp, oldlenp, newp, newlen,
900 			    tcp_ctlvars[name[0]].var));
901 	}
902 
903 	return (ENOPROTOOPT);
904 }
905