xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision b5cd2489b2952ee409ede7b61183dba7ab0359ad)
1 /*	$NetBSD: tcp_usrreq.c,v 1.66 2001/10/29 07:02:34 simonb 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 #include "opt_tcp_debug.h"
107 
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/kernel.h>
111 #include <sys/malloc.h>
112 #include <sys/mbuf.h>
113 #include <sys/socket.h>
114 #include <sys/socketvar.h>
115 #include <sys/protosw.h>
116 #include <sys/errno.h>
117 #include <sys/stat.h>
118 #include <sys/proc.h>
119 #include <sys/domain.h>
120 #include <sys/sysctl.h>
121 
122 #include <net/if.h>
123 #include <net/route.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/in_var.h>
128 #include <netinet/ip.h>
129 #include <netinet/in_pcb.h>
130 #include <netinet/ip_var.h>
131 
132 #ifdef INET6
133 #ifndef INET
134 #include <netinet/in.h>
135 #endif
136 #include <netinet/ip6.h>
137 #include <netinet6/in6_pcb.h>
138 #include <netinet6/ip6_var.h>
139 #endif
140 
141 #include <netinet/tcp.h>
142 #include <netinet/tcp_fsm.h>
143 #include <netinet/tcp_seq.h>
144 #include <netinet/tcp_timer.h>
145 #include <netinet/tcp_var.h>
146 #include <netinet/tcpip.h>
147 #include <netinet/tcp_debug.h>
148 
149 #include "opt_tcp_recvspace.h"
150 #include "opt_tcp_sendspace.h"
151 
152 #ifdef IPSEC
153 #include <netinet6/ipsec.h>
154 #endif /*IPSEC*/
155 
156 /*
157  * TCP protocol interface to socket abstraction.
158  */
159 extern	char *tcpstates[];
160 
161 /*
162  * Process a TCP user request for TCP tb.  If this is a send request
163  * then m is the mbuf chain of send data.  If this is a timer expiration
164  * (called from the software clock routine), then timertype tells which timer.
165  */
166 /*ARGSUSED*/
167 int
168 tcp_usrreq(so, req, m, nam, control, p)
169 	struct socket *so;
170 	int req;
171 	struct mbuf *m, *nam, *control;
172 	struct proc *p;
173 {
174 	struct inpcb *inp;
175 #ifdef INET6
176 	struct in6pcb *in6p;
177 #endif
178 	struct tcpcb *tp = NULL;
179 	int s;
180 	int error = 0;
181 	int ostate;
182 	int family;	/* family of the socket */
183 
184 	family = so->so_proto->pr_domain->dom_family;
185 
186 	if (req == PRU_CONTROL) {
187 		switch (family) {
188 #ifdef INET
189 		case PF_INET:
190 			return (in_control(so, (long)m, (caddr_t)nam,
191 			    (struct ifnet *)control, p));
192 #endif
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 	if (req == PRU_PURGEIF) {
204 		switch (family) {
205 #ifdef INET
206 		case PF_INET:
207 			in_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
208 			in_purgeif((struct ifnet *)control);
209 			in_pcbpurgeif(&tcbtable, (struct ifnet *)control);
210 			break;
211 #endif
212 #ifdef INET6
213 		case PF_INET6:
214 			in6_pcbpurgeif0(&tcb6, (struct ifnet *)control);
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, 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 		/*
488 		 * soreceive() calls this function when a user receives
489 		 * ancillary data on a listening socket. We don't call
490 		 * tcp_output in such a case, since there is no header
491 		 * template for a listening socket and hence the kernel
492 		 * will panic.
493 		 */
494 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
495 			(void) tcp_output(tp);
496 		break;
497 
498 	/*
499 	 * Do a send by putting data in output queue and updating urgent
500 	 * marker if URG set.  Possibly send more data.
501 	 */
502 	case PRU_SEND:
503 		if (control && control->m_len) {
504 			m_freem(control);
505 			m_freem(m);
506 			error = EINVAL;
507 			break;
508 		}
509 		sbappend(&so->so_snd, m);
510 		error = tcp_output(tp);
511 		break;
512 
513 	/*
514 	 * Abort the TCP.
515 	 */
516 	case PRU_ABORT:
517 		tp = tcp_drop(tp, ECONNABORTED);
518 		break;
519 
520 	case PRU_SENSE:
521 		/*
522 		 * stat: don't bother with a blocksize.
523 		 */
524 		splx(s);
525 		return (0);
526 
527 	case PRU_RCVOOB:
528 		if (control && control->m_len) {
529 			m_freem(control);
530 			m_freem(m);
531 			error = EINVAL;
532 			break;
533 		}
534 		if ((so->so_oobmark == 0 &&
535 		    (so->so_state & SS_RCVATMARK) == 0) ||
536 		    so->so_options & SO_OOBINLINE ||
537 		    tp->t_oobflags & TCPOOB_HADDATA) {
538 			error = EINVAL;
539 			break;
540 		}
541 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
542 			error = EWOULDBLOCK;
543 			break;
544 		}
545 		m->m_len = 1;
546 		*mtod(m, caddr_t) = tp->t_iobc;
547 		if (((long)nam & MSG_PEEK) == 0)
548 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
549 		break;
550 
551 	case PRU_SENDOOB:
552 		if (sbspace(&so->so_snd) < -512) {
553 			m_freem(m);
554 			error = ENOBUFS;
555 			break;
556 		}
557 		/*
558 		 * According to RFC961 (Assigned Protocols),
559 		 * the urgent pointer points to the last octet
560 		 * of urgent data.  We continue, however,
561 		 * to consider it to indicate the first octet
562 		 * of data past the urgent section.
563 		 * Otherwise, snd_up should be one lower.
564 		 */
565 		sbappend(&so->so_snd, m);
566 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
567 		tp->t_force = 1;
568 		error = tcp_output(tp);
569 		tp->t_force = 0;
570 		break;
571 
572 	case PRU_SOCKADDR:
573 #ifdef INET
574 		if (inp)
575 			in_setsockaddr(inp, nam);
576 #endif
577 #ifdef INET6
578 		if (in6p)
579 			in6_setsockaddr(in6p, nam);
580 #endif
581 		break;
582 
583 	case PRU_PEERADDR:
584 #ifdef INET
585 		if (inp)
586 			in_setpeeraddr(inp, nam);
587 #endif
588 #ifdef INET6
589 		if (in6p)
590 			in6_setpeeraddr(in6p, nam);
591 #endif
592 		break;
593 
594 	default:
595 		panic("tcp_usrreq");
596 	}
597 #ifdef TCP_DEBUG
598 	if (tp && (so->so_options & SO_DEBUG))
599 		tcp_trace(TA_USER, ostate, tp, NULL, req);
600 #endif
601 
602 release:
603 	splx(s);
604 	return (error);
605 }
606 
607 int
608 tcp_ctloutput(op, so, level, optname, mp)
609 	int op;
610 	struct socket *so;
611 	int level, optname;
612 	struct mbuf **mp;
613 {
614 	int error = 0, s;
615 	struct inpcb *inp;
616 #ifdef INET6
617 	struct in6pcb *in6p;
618 #endif
619 	struct tcpcb *tp;
620 	struct mbuf *m;
621 	int i;
622 	int family;	/* family of the socket */
623 
624 	family = so->so_proto->pr_domain->dom_family;
625 
626 	s = splsoftnet();
627 	switch (family) {
628 #ifdef INET
629 	case PF_INET:
630 		inp = sotoinpcb(so);
631 #ifdef INET6
632 		in6p = NULL;
633 #endif
634 		break;
635 #endif
636 #ifdef INET6
637 	case PF_INET6:
638 		inp = NULL;
639 		in6p = sotoin6pcb(so);
640 		break;
641 #endif
642 	default:
643 		splx(s);
644 		return EAFNOSUPPORT;
645 	}
646 #ifndef INET6
647 	if (inp == NULL)
648 #else
649 	if (inp == NULL && in6p == NULL)
650 #endif
651 	{
652 		splx(s);
653 		if (op == PRCO_SETOPT && *mp)
654 			(void) m_free(*mp);
655 		return (ECONNRESET);
656 	}
657 	if (level != IPPROTO_TCP) {
658 		switch (family) {
659 #ifdef INET
660 		case PF_INET:
661 			error = ip_ctloutput(op, so, level, optname, mp);
662 			break;
663 #endif
664 #ifdef INET6
665 		case PF_INET6:
666 			error = ip6_ctloutput(op, so, level, optname, mp);
667 			break;
668 #endif
669 		}
670 		splx(s);
671 		return (error);
672 	}
673 	if (inp)
674 		tp = intotcpcb(inp);
675 #ifdef INET6
676 	else if (in6p)
677 		tp = in6totcpcb(in6p);
678 #endif
679 	else
680 		tp = NULL;
681 
682 	switch (op) {
683 
684 	case PRCO_SETOPT:
685 		m = *mp;
686 		switch (optname) {
687 
688 		case TCP_NODELAY:
689 			if (m == NULL || m->m_len < sizeof (int))
690 				error = EINVAL;
691 			else if (*mtod(m, int *))
692 				tp->t_flags |= TF_NODELAY;
693 			else
694 				tp->t_flags &= ~TF_NODELAY;
695 			break;
696 
697 		case TCP_MAXSEG:
698 			if (m && (i = *mtod(m, int *)) > 0 &&
699 			    i <= tp->t_peermss)
700 				tp->t_peermss = i;  /* limit on send size */
701 			else
702 				error = EINVAL;
703 			break;
704 
705 		default:
706 			error = ENOPROTOOPT;
707 			break;
708 		}
709 		if (m)
710 			(void) m_free(m);
711 		break;
712 
713 	case PRCO_GETOPT:
714 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
715 		m->m_len = sizeof(int);
716 
717 		switch (optname) {
718 		case TCP_NODELAY:
719 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
720 			break;
721 		case TCP_MAXSEG:
722 			*mtod(m, int *) = tp->t_peermss;
723 			break;
724 		default:
725 			error = ENOPROTOOPT;
726 			break;
727 		}
728 		break;
729 	}
730 	splx(s);
731 	return (error);
732 }
733 
734 #ifndef TCP_SENDSPACE
735 #define	TCP_SENDSPACE	1024*16;
736 #endif
737 int	tcp_sendspace = TCP_SENDSPACE;
738 #ifndef TCP_RECVSPACE
739 #define	TCP_RECVSPACE	1024*16;
740 #endif
741 int	tcp_recvspace = TCP_RECVSPACE;
742 
743 /*
744  * Attach TCP protocol to socket, allocating
745  * internet protocol control block, tcp control block,
746  * bufer space, and entering LISTEN state if to accept connections.
747  */
748 int
749 tcp_attach(so)
750 	struct socket *so;
751 {
752 	struct tcpcb *tp;
753 	struct inpcb *inp;
754 #ifdef INET6
755 	struct in6pcb *in6p;
756 #endif
757 	int error;
758 	int family;	/* family of the socket */
759 
760 	family = so->so_proto->pr_domain->dom_family;
761 
762 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
763 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
764 		if (error)
765 			return (error);
766 	}
767 	switch (family) {
768 #ifdef INET
769 	case PF_INET:
770 		error = in_pcballoc(so, &tcbtable);
771 		if (error)
772 			return (error);
773 		inp = sotoinpcb(so);
774 #ifdef INET6
775 		in6p = NULL;
776 #endif
777 		break;
778 #endif
779 #ifdef INET6
780 	case PF_INET6:
781 		error = in6_pcballoc(so, &tcb6);
782 		if (error)
783 			return (error);
784 		inp = NULL;
785 		in6p = sotoin6pcb(so);
786 		break;
787 #endif
788 	default:
789 		return EAFNOSUPPORT;
790 	}
791 	if (inp)
792 		tp = tcp_newtcpcb(family, (void *)inp);
793 #ifdef INET6
794 	else if (in6p)
795 		tp = tcp_newtcpcb(family, (void *)in6p);
796 #endif
797 	else
798 		tp = NULL;
799 
800 	if (tp == 0) {
801 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
802 
803 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
804 #ifdef INET
805 		if (inp)
806 			in_pcbdetach(inp);
807 #endif
808 #ifdef INET6
809 		if (in6p)
810 			in6_pcbdetach(in6p);
811 #endif
812 		so->so_state |= nofd;
813 		return (ENOBUFS);
814 	}
815 	tp->t_state = TCPS_CLOSED;
816 	return (0);
817 }
818 
819 /*
820  * Initiate (or continue) disconnect.
821  * If embryonic state, just send reset (once).
822  * If in ``let data drain'' option and linger null, just drop.
823  * Otherwise (hard), mark socket disconnecting and drop
824  * current input data; switch states based on user close, and
825  * send segment to peer (with FIN).
826  */
827 struct tcpcb *
828 tcp_disconnect(tp)
829 	struct tcpcb *tp;
830 {
831 	struct socket *so;
832 
833 	if (tp->t_inpcb)
834 		so = tp->t_inpcb->inp_socket;
835 #ifdef INET6
836 	else if (tp->t_in6pcb)
837 		so = tp->t_in6pcb->in6p_socket;
838 #endif
839 	else
840 		so = NULL;
841 
842 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
843 		tp = tcp_close(tp);
844 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
845 		tp = tcp_drop(tp, 0);
846 	else {
847 		soisdisconnecting(so);
848 		sbflush(&so->so_rcv);
849 		tp = tcp_usrclosed(tp);
850 		if (tp)
851 			(void) tcp_output(tp);
852 	}
853 	return (tp);
854 }
855 
856 /*
857  * User issued close, and wish to trail through shutdown states:
858  * if never received SYN, just forget it.  If got a SYN from peer,
859  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
860  * If already got a FIN from peer, then almost done; go to LAST_ACK
861  * state.  In all other cases, have already sent FIN to peer (e.g.
862  * after PRU_SHUTDOWN), and just have to play tedious game waiting
863  * for peer to send FIN or not respond to keep-alives, etc.
864  * We can let the user exit from the close as soon as the FIN is acked.
865  */
866 struct tcpcb *
867 tcp_usrclosed(tp)
868 	struct tcpcb *tp;
869 {
870 
871 	switch (tp->t_state) {
872 
873 	case TCPS_CLOSED:
874 	case TCPS_LISTEN:
875 	case TCPS_SYN_SENT:
876 		tp->t_state = TCPS_CLOSED;
877 		tp = tcp_close(tp);
878 		break;
879 
880 	case TCPS_SYN_RECEIVED:
881 	case TCPS_ESTABLISHED:
882 		tp->t_state = TCPS_FIN_WAIT_1;
883 		break;
884 
885 	case TCPS_CLOSE_WAIT:
886 		tp->t_state = TCPS_LAST_ACK;
887 		break;
888 	}
889 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
890 		struct socket *so;
891 		if (tp->t_inpcb)
892 			so = tp->t_inpcb->inp_socket;
893 #ifdef INET6
894 		else if (tp->t_in6pcb)
895 			so = tp->t_in6pcb->in6p_socket;
896 #endif
897 		else
898 			so = NULL;
899 		soisdisconnected(so);
900 		/*
901 		 * If we are in FIN_WAIT_2, we arrived here because the
902 		 * application did a shutdown of the send side.  Like the
903 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
904 		 * a full close, we start a timer to make sure sockets are
905 		 * not left in FIN_WAIT_2 forever.
906 		 */
907 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
908 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
909 	}
910 	return (tp);
911 }
912 
913 static const struct {
914 	 unsigned int valid : 1;
915 	 unsigned int rdonly : 1;
916 	 int *var;
917 	 int val;
918 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
919 
920 /*
921  * Sysctl for tcp variables.
922  */
923 int
924 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
925 	int *name;
926 	u_int namelen;
927 	void *oldp;
928 	size_t *oldlenp;
929 	void *newp;
930 	size_t newlen;
931 {
932 
933 	/* All sysctl names at this level are terminal. */
934 	if (namelen != 1)
935 		return (ENOTDIR);
936 
937 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
938 	    && tcp_ctlvars[name[0]].valid) {
939 		if (tcp_ctlvars[name[0]].rdonly)
940 			return (sysctl_rdint(oldp, oldlenp, newp,
941 			    tcp_ctlvars[name[0]].val));
942 		else
943 			return (sysctl_int(oldp, oldlenp, newp, newlen,
944 			    tcp_ctlvars[name[0]].var));
945 	}
946 
947 	return (ENOPROTOOPT);
948 }
949