xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 5aefcfdc06931dd97e76246d2fe0302f7b3fe094)
1 /*	$NetBSD: tcp_usrreq.c,v 1.58 2000/12/11 00:07:48 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 <uvm/uvm_extern.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 	struct inpcb *inp;
177 #ifdef INET6
178 	struct in6pcb *in6p;
179 #endif
180 	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 #ifdef INET
191 		case PF_INET:
192 			return (in_control(so, (long)m, (caddr_t)nam,
193 			    (struct ifnet *)control, p));
194 #endif
195 #ifdef INET6
196 		case PF_INET6:
197 			return (in6_control(so, (long)m, (caddr_t)nam,
198 			    (struct ifnet *)control, p));
199 #endif
200 		default:
201 			return EAFNOSUPPORT;
202 		}
203 	}
204 
205 	if (req == PRU_PURGEIF) {
206 		switch (family) {
207 #ifdef INET
208 		case PF_INET:
209 			in_purgeif((struct ifnet *)control);
210 			in_pcbpurgeif(&tcbtable, (struct ifnet *)control);
211 			break;
212 #endif
213 #ifdef INET6
214 		case PF_INET6:
215 			in6_purgeif((struct ifnet *)control);
216 			in6_pcbpurgeif(&tcb6, (struct ifnet *)control);
217 			break;
218 #endif
219 		default:
220 			return (EAFNOSUPPORT);
221 		}
222 		return (0);
223 	}
224 
225 	s = splsoftnet();
226 	switch (family) {
227 #ifdef INET
228 	case PF_INET:
229 		inp = sotoinpcb(so);
230 #ifdef INET6
231 		in6p = NULL;
232 #endif
233 		break;
234 #endif
235 #ifdef INET6
236 	case PF_INET6:
237 		inp = NULL;
238 		in6p = sotoin6pcb(so);
239 		break;
240 #endif
241 	default:
242 		splx(s);
243 		return EAFNOSUPPORT;
244 	}
245 
246 #ifdef DIAGNOSTIC
247 #ifdef INET6
248 	if (inp && in6p)
249 		panic("tcp_usrreq: both inp and in6p set to non-NULL");
250 #endif
251 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
252 		panic("tcp_usrreq: unexpected control mbuf");
253 #endif
254 	/*
255 	 * When a TCP is attached to a socket, then there will be
256 	 * a (struct inpcb) pointed at by the socket, and this
257 	 * structure will point at a subsidary (struct tcpcb).
258 	 */
259 #ifndef INET6
260 	if (inp == 0 && req != PRU_ATTACH)
261 #else
262 	if ((inp == 0 && in6p == 0) && req != PRU_ATTACH)
263 #endif
264 	{
265 		error = EINVAL;
266 		goto release;
267 	}
268 #ifdef INET
269 	if (inp) {
270 		tp = intotcpcb(inp);
271 		/* WHAT IF TP IS 0? */
272 #ifdef KPROF
273 		tcp_acounts[tp->t_state][req]++;
274 #endif
275 		ostate = tp->t_state;
276 	}
277 #endif
278 #ifdef INET6
279 	if (in6p) {
280 		tp = in6totcpcb(in6p);
281 		/* WHAT IF TP IS 0? */
282 #ifdef KPROF
283 		tcp_acounts[tp->t_state][req]++;
284 #endif
285 		ostate = tp->t_state;
286 	}
287 #endif
288 	else
289 		ostate = 0;
290 
291 	switch (req) {
292 
293 	/*
294 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
295 	 * and an internet control block.
296 	 */
297 	case PRU_ATTACH:
298 #ifndef INET6
299 		if (inp != 0)
300 #else
301 		if (inp != 0 || in6p != 0)
302 #endif
303 		{
304 			error = EISCONN;
305 			break;
306 		}
307 		error = tcp_attach(so);
308 		if (error)
309 			break;
310 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
311 			so->so_linger = TCP_LINGERTIME;
312 		tp = sototcpcb(so);
313 		break;
314 
315 	/*
316 	 * PRU_DETACH detaches the TCP protocol from the socket.
317 	 */
318 	case PRU_DETACH:
319 		tp = tcp_disconnect(tp);
320 		break;
321 
322 	/*
323 	 * Give the socket an address.
324 	 */
325 	case PRU_BIND:
326 		switch (family) {
327 #ifdef INET
328 		case PF_INET:
329 			error = in_pcbbind(inp, nam, p);
330 			break;
331 #endif
332 #ifdef INET6
333 		case PF_INET6:
334 			error = in6_pcbbind(in6p, nam, p);
335 			if (!error) {
336 				/* mapped addr case */
337 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
338 					tp->t_family = AF_INET;
339 				else
340 					tp->t_family = AF_INET6;
341 			}
342 			break;
343 #endif
344 		}
345 		break;
346 
347 	/*
348 	 * Prepare to accept connections.
349 	 */
350 	case PRU_LISTEN:
351 #ifdef INET
352 		if (inp && inp->inp_lport == 0) {
353 			error = in_pcbbind(inp, (struct mbuf *)0,
354 			    (struct proc *)0);
355 			if (error)
356 				break;
357 		}
358 #endif
359 #ifdef INET6
360 		if (in6p && in6p->in6p_lport == 0) {
361 			error = in6_pcbbind(in6p, (struct mbuf *)0,
362 			    (struct proc *)0);
363 			if (error)
364 				break;
365 		}
366 #endif
367 		tp->t_state = TCPS_LISTEN;
368 		break;
369 
370 	/*
371 	 * Initiate connection to peer.
372 	 * Create a template for use in transmissions on this connection.
373 	 * Enter SYN_SENT state, and mark socket as connecting.
374 	 * Start keep-alive timer, and seed output sequence space.
375 	 * Send initial segment on connection.
376 	 */
377 	case PRU_CONNECT:
378 #ifdef INET
379 		if (inp) {
380 			if (inp->inp_lport == 0) {
381 				error = in_pcbbind(inp, (struct mbuf *)0,
382 				    (struct proc *)0);
383 				if (error)
384 					break;
385 			}
386 			error = in_pcbconnect(inp, nam);
387 		}
388 #endif
389 #ifdef INET6
390 		if (in6p) {
391 			if (in6p->in6p_lport == 0) {
392 				error = in6_pcbbind(in6p, (struct mbuf *)0,
393 				    (struct proc *)0);
394 				if (error)
395 					break;
396 			}
397 			error = in6_pcbconnect(in6p, nam);
398 			if (!error) {
399 				/* mapped addr case */
400 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
401 					tp->t_family = AF_INET;
402 				else
403 					tp->t_family = AF_INET6;
404 			}
405 		}
406 #endif
407 		if (error)
408 			break;
409 		tp->t_template = tcp_template(tp);
410 		if (tp->t_template == 0) {
411 #ifdef INET
412 			if (inp)
413 				in_pcbdisconnect(inp);
414 #endif
415 #ifdef INET6
416 			if (in6p)
417 				in6_pcbdisconnect(in6p);
418 #endif
419 			error = ENOBUFS;
420 			break;
421 		}
422 		/* Compute window scaling to request.  */
423 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
424 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
425 			tp->request_r_scale++;
426 		soisconnecting(so);
427 		tcpstat.tcps_connattempt++;
428 		tp->t_state = TCPS_SYN_SENT;
429 		TCP_TIMER_ARM(tp, TCPT_KEEP, TCPTV_KEEP_INIT);
430 		tp->iss = tcp_new_iss(tp, sizeof(struct tcpcb), 0);
431 		tcp_sendseqinit(tp);
432 		error = tcp_output(tp);
433 		break;
434 
435 	/*
436 	 * Create a TCP connection between two sockets.
437 	 */
438 	case PRU_CONNECT2:
439 		error = EOPNOTSUPP;
440 		break;
441 
442 	/*
443 	 * Initiate disconnect from peer.
444 	 * If connection never passed embryonic stage, just drop;
445 	 * else if don't need to let data drain, then can just drop anyways,
446 	 * else have to begin TCP shutdown process: mark socket disconnecting,
447 	 * drain unread data, state switch to reflect user close, and
448 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
449 	 * when peer sends FIN and acks ours.
450 	 *
451 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
452 	 */
453 	case PRU_DISCONNECT:
454 		tp = tcp_disconnect(tp);
455 		break;
456 
457 	/*
458 	 * Accept a connection.  Essentially all the work is
459 	 * done at higher levels; just return the address
460 	 * of the peer, storing through addr.
461 	 */
462 	case PRU_ACCEPT:
463 #ifdef INET
464 		if (inp)
465 			in_setpeeraddr(inp, nam);
466 #endif
467 #ifdef INET6
468 		if (in6p)
469 			in6_setpeeraddr(in6p, nam);
470 #endif
471 		break;
472 
473 	/*
474 	 * Mark the connection as being incapable of further output.
475 	 */
476 	case PRU_SHUTDOWN:
477 		socantsendmore(so);
478 		tp = tcp_usrclosed(tp);
479 		if (tp)
480 			error = tcp_output(tp);
481 		break;
482 
483 	/*
484 	 * After a receive, possibly send window update to peer.
485 	 */
486 	case PRU_RCVD:
487 		(void) tcp_output(tp);
488 		break;
489 
490 	/*
491 	 * Do a send by putting data in output queue and updating urgent
492 	 * marker if URG set.  Possibly send more data.
493 	 */
494 	case PRU_SEND:
495 		if (control && control->m_len) {
496 			m_freem(control);
497 			m_freem(m);
498 			error = EINVAL;
499 			break;
500 		}
501 		sbappend(&so->so_snd, m);
502 		error = tcp_output(tp);
503 		break;
504 
505 	/*
506 	 * Abort the TCP.
507 	 */
508 	case PRU_ABORT:
509 		tp = tcp_drop(tp, ECONNABORTED);
510 		break;
511 
512 	case PRU_SENSE:
513 		/*
514 		 * stat: don't bother with a blocksize.
515 		 */
516 		splx(s);
517 		return (0);
518 
519 	case PRU_RCVOOB:
520 		if (control && control->m_len) {
521 			m_freem(control);
522 			m_freem(m);
523 			error = EINVAL;
524 			break;
525 		}
526 		if ((so->so_oobmark == 0 &&
527 		    (so->so_state & SS_RCVATMARK) == 0) ||
528 		    so->so_options & SO_OOBINLINE ||
529 		    tp->t_oobflags & TCPOOB_HADDATA) {
530 			error = EINVAL;
531 			break;
532 		}
533 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
534 			error = EWOULDBLOCK;
535 			break;
536 		}
537 		m->m_len = 1;
538 		*mtod(m, caddr_t) = tp->t_iobc;
539 		if (((long)nam & MSG_PEEK) == 0)
540 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
541 		break;
542 
543 	case PRU_SENDOOB:
544 		if (sbspace(&so->so_snd) < -512) {
545 			m_freem(m);
546 			error = ENOBUFS;
547 			break;
548 		}
549 		/*
550 		 * According to RFC961 (Assigned Protocols),
551 		 * the urgent pointer points to the last octet
552 		 * of urgent data.  We continue, however,
553 		 * to consider it to indicate the first octet
554 		 * of data past the urgent section.
555 		 * Otherwise, snd_up should be one lower.
556 		 */
557 		sbappend(&so->so_snd, m);
558 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
559 		tp->t_force = 1;
560 		error = tcp_output(tp);
561 		tp->t_force = 0;
562 		break;
563 
564 	case PRU_SOCKADDR:
565 #ifdef INET
566 		if (inp)
567 			in_setsockaddr(inp, nam);
568 #endif
569 #ifdef INET6
570 		if (in6p)
571 			in6_setsockaddr(in6p, nam);
572 #endif
573 		break;
574 
575 	case PRU_PEERADDR:
576 #ifdef INET
577 		if (inp)
578 			in_setpeeraddr(inp, nam);
579 #endif
580 #ifdef INET6
581 		if (in6p)
582 			in6_setpeeraddr(in6p, nam);
583 #endif
584 		break;
585 
586 	/*
587 	 * TCP slow timer went off; going through this
588 	 * routine for tracing's sake.
589 	 */
590 	case PRU_SLOWTIMO:
591 		tp = tcp_timers(tp, (long)nam);
592 		req |= (long)nam << 8;		/* for debug's sake */
593 		break;
594 
595 	default:
596 		panic("tcp_usrreq");
597 	}
598 	if (tp && (so->so_options & SO_DEBUG))
599 		tcp_trace(TA_USER, ostate, tp, NULL, req);
600 
601 release:
602 	splx(s);
603 	return (error);
604 }
605 
606 int
607 tcp_ctloutput(op, so, level, optname, mp)
608 	int op;
609 	struct socket *so;
610 	int level, optname;
611 	struct mbuf **mp;
612 {
613 	int error = 0, s;
614 	struct inpcb *inp;
615 #ifdef INET6
616 	struct in6pcb *in6p;
617 #endif
618 	struct tcpcb *tp;
619 	struct mbuf *m;
620 	int i;
621 	int family;	/* family of the socket */
622 
623 	family = so->so_proto->pr_domain->dom_family;
624 
625 	s = splsoftnet();
626 	switch (family) {
627 #ifdef INET
628 	case PF_INET:
629 		inp = sotoinpcb(so);
630 #ifdef INET6
631 		in6p = NULL;
632 #endif
633 		break;
634 #endif
635 #ifdef INET6
636 	case PF_INET6:
637 		inp = NULL;
638 		in6p = sotoin6pcb(so);
639 		break;
640 #endif
641 	default:
642 		splx(s);
643 		return EAFNOSUPPORT;
644 	}
645 #ifndef INET6
646 	if (inp == NULL)
647 #else
648 	if (inp == NULL && in6p == NULL)
649 #endif
650 	{
651 		splx(s);
652 		if (op == PRCO_SETOPT && *mp)
653 			(void) m_free(*mp);
654 		return (ECONNRESET);
655 	}
656 	if (level != IPPROTO_TCP) {
657 		switch (family) {
658 #ifdef INET
659 		case PF_INET:
660 			error = ip_ctloutput(op, so, level, optname, mp);
661 			break;
662 #endif
663 #ifdef INET6
664 		case PF_INET6:
665 			error = ip6_ctloutput(op, so, level, optname, mp);
666 			break;
667 #endif
668 		}
669 		splx(s);
670 		return (error);
671 	}
672 	if (inp)
673 		tp = intotcpcb(inp);
674 #ifdef INET6
675 	else if (in6p)
676 		tp = in6totcpcb(in6p);
677 #endif
678 	else
679 		tp = NULL;
680 
681 	switch (op) {
682 
683 	case PRCO_SETOPT:
684 		m = *mp;
685 		switch (optname) {
686 
687 		case TCP_NODELAY:
688 			if (m == NULL || m->m_len < sizeof (int))
689 				error = EINVAL;
690 			else if (*mtod(m, int *))
691 				tp->t_flags |= TF_NODELAY;
692 			else
693 				tp->t_flags &= ~TF_NODELAY;
694 			break;
695 
696 		case TCP_MAXSEG:
697 			if (m && (i = *mtod(m, int *)) > 0 &&
698 			    i <= tp->t_peermss)
699 				tp->t_peermss = i;  /* limit on send size */
700 			else
701 				error = EINVAL;
702 			break;
703 
704 		default:
705 			error = ENOPROTOOPT;
706 			break;
707 		}
708 		if (m)
709 			(void) m_free(m);
710 		break;
711 
712 	case PRCO_GETOPT:
713 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
714 		m->m_len = sizeof(int);
715 
716 		switch (optname) {
717 		case TCP_NODELAY:
718 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
719 			break;
720 		case TCP_MAXSEG:
721 			*mtod(m, int *) = tp->t_peermss;
722 			break;
723 		default:
724 			error = ENOPROTOOPT;
725 			break;
726 		}
727 		break;
728 	}
729 	splx(s);
730 	return (error);
731 }
732 
733 #ifndef TCP_SENDSPACE
734 #define	TCP_SENDSPACE	1024*16;
735 #endif
736 int	tcp_sendspace = TCP_SENDSPACE;
737 #ifndef TCP_RECVSPACE
738 #define	TCP_RECVSPACE	1024*16;
739 #endif
740 int	tcp_recvspace = TCP_RECVSPACE;
741 
742 /*
743  * Attach TCP protocol to socket, allocating
744  * internet protocol control block, tcp control block,
745  * bufer space, and entering LISTEN state if to accept connections.
746  */
747 int
748 tcp_attach(so)
749 	struct socket *so;
750 {
751 	struct tcpcb *tp;
752 	struct inpcb *inp;
753 #ifdef INET6
754 	struct in6pcb *in6p;
755 #endif
756 	int error;
757 	int family;	/* family of the socket */
758 
759 	family = so->so_proto->pr_domain->dom_family;
760 
761 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
762 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
763 		if (error)
764 			return (error);
765 	}
766 	switch (family) {
767 #ifdef INET
768 	case PF_INET:
769 		error = in_pcballoc(so, &tcbtable);
770 		if (error)
771 			return (error);
772 		inp = sotoinpcb(so);
773 #ifdef INET6
774 		in6p = NULL;
775 #endif
776 		break;
777 #endif
778 #ifdef INET6
779 	case PF_INET6:
780 		error = in6_pcballoc(so, &tcb6);
781 		if (error)
782 			return (error);
783 		inp = NULL;
784 		in6p = sotoin6pcb(so);
785 		break;
786 #endif
787 	default:
788 		return EAFNOSUPPORT;
789 	}
790 #ifdef IPSEC
791 #ifdef INET
792 	if (inp) {
793 		error = ipsec_init_policy(so, &inp->inp_sp);
794 		if (error != 0) {
795 			in_pcbdetach(inp);
796 			return (error);
797 		}
798 	}
799 #endif
800 #ifdef INET6
801 	if (in6p) {
802 		error = ipsec_init_policy(so, &in6p->in6p_sp);
803 		if (error != 0) {
804 			in6_pcbdetach(in6p);
805 			return (error);
806 		}
807 	}
808 #endif
809 #endif /*IPSEC*/
810 	if (inp)
811 		tp = tcp_newtcpcb(family, (void *)inp);
812 #ifdef INET6
813 	else if (in6p)
814 		tp = tcp_newtcpcb(family, (void *)in6p);
815 #endif
816 	else
817 		tp = NULL;
818 
819 	if (tp == 0) {
820 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
821 
822 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
823 #ifdef INET
824 		if (inp)
825 			in_pcbdetach(inp);
826 #endif
827 #ifdef INET6
828 		if (in6p)
829 			in6_pcbdetach(in6p);
830 #endif
831 		so->so_state |= nofd;
832 		return (ENOBUFS);
833 	}
834 	tp->t_state = TCPS_CLOSED;
835 	return (0);
836 }
837 
838 /*
839  * Initiate (or continue) disconnect.
840  * If embryonic state, just send reset (once).
841  * If in ``let data drain'' option and linger null, just drop.
842  * Otherwise (hard), mark socket disconnecting and drop
843  * current input data; switch states based on user close, and
844  * send segment to peer (with FIN).
845  */
846 struct tcpcb *
847 tcp_disconnect(tp)
848 	struct tcpcb *tp;
849 {
850 	struct socket *so;
851 
852 	if (tp->t_inpcb)
853 		so = tp->t_inpcb->inp_socket;
854 #ifdef INET6
855 	else if (tp->t_in6pcb)
856 		so = tp->t_in6pcb->in6p_socket;
857 #endif
858 	else
859 		so = NULL;
860 
861 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
862 		tp = tcp_close(tp);
863 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
864 		tp = tcp_drop(tp, 0);
865 	else {
866 		soisdisconnecting(so);
867 		sbflush(&so->so_rcv);
868 		tp = tcp_usrclosed(tp);
869 		if (tp)
870 			(void) tcp_output(tp);
871 	}
872 	return (tp);
873 }
874 
875 /*
876  * User issued close, and wish to trail through shutdown states:
877  * if never received SYN, just forget it.  If got a SYN from peer,
878  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
879  * If already got a FIN from peer, then almost done; go to LAST_ACK
880  * state.  In all other cases, have already sent FIN to peer (e.g.
881  * after PRU_SHUTDOWN), and just have to play tedious game waiting
882  * for peer to send FIN or not respond to keep-alives, etc.
883  * We can let the user exit from the close as soon as the FIN is acked.
884  */
885 struct tcpcb *
886 tcp_usrclosed(tp)
887 	struct tcpcb *tp;
888 {
889 
890 	switch (tp->t_state) {
891 
892 	case TCPS_CLOSED:
893 	case TCPS_LISTEN:
894 	case TCPS_SYN_SENT:
895 		tp->t_state = TCPS_CLOSED;
896 		tp = tcp_close(tp);
897 		break;
898 
899 	case TCPS_SYN_RECEIVED:
900 	case TCPS_ESTABLISHED:
901 		tp->t_state = TCPS_FIN_WAIT_1;
902 		break;
903 
904 	case TCPS_CLOSE_WAIT:
905 		tp->t_state = TCPS_LAST_ACK;
906 		break;
907 	}
908 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
909 		struct socket *so;
910 		if (tp->t_inpcb)
911 			so = tp->t_inpcb->inp_socket;
912 #ifdef INET6
913 		else if (tp->t_in6pcb)
914 			so = tp->t_in6pcb->in6p_socket;
915 #endif
916 		else
917 			so = NULL;
918 		soisdisconnected(so);
919 		/*
920 		 * If we are in FIN_WAIT_2, we arrived here because the
921 		 * application did a shutdown of the send side.  Like the
922 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
923 		 * a full close, we start a timer to make sure sockets are
924 		 * not left in FIN_WAIT_2 forever.
925 		 */
926 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
927 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
928 	}
929 	return (tp);
930 }
931 
932 static struct {
933 	 unsigned int valid : 1;
934 	 unsigned int rdonly : 1;
935 	 int *var;
936 	 int val;
937 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
938 
939 /*
940  * Sysctl for tcp variables.
941  */
942 int
943 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
944 	int *name;
945 	u_int namelen;
946 	void *oldp;
947 	size_t *oldlenp;
948 	void *newp;
949 	size_t newlen;
950 {
951 
952 	/* All sysctl names at this level are terminal. */
953 	if (namelen != 1)
954 		return (ENOTDIR);
955 
956 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
957 	    && tcp_ctlvars[name[0]].valid) {
958 		if (tcp_ctlvars[name[0]].rdonly)
959 			return (sysctl_rdint(oldp, oldlenp, newp,
960 			    tcp_ctlvars[name[0]].val));
961 		else
962 			return (sysctl_int(oldp, oldlenp, newp, newlen,
963 			    tcp_ctlvars[name[0]].var));
964 	}
965 
966 	return (ENOPROTOOPT);
967 }
968